linux-hardened/mm/mmap.c

3773 lines
100 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* mm/mmap.c
*
* Written by obz.
*
* Address space accounting code <alan@lxorguk.ukuu.org.uk>
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
mm/mmap: check for RLIMIT_AS before unmapping Fix a corner case for MAP_FIXED when requested mapping length is larger than rlimit for virtual memory. In such case any overlapping mappings are unmapped before we check for the limit and return ENOMEM. The check is moved before the loop that unmaps overlapping parts of existing mappings. When we are about to hit the limit (currently mapped pages + len > limit) we scan for overlapping pages and check again accounting for them. This fixes situation when userspace program expects that the previous mappings are preserved after the mmap() syscall has returned with error. (POSIX clearly states that successfull mapping shall replace any previous mappings.) This corner case was found and can be tested with LTP testcase: testcases/open_posix_testsuite/conformance/interfaces/mmap/24-2.c In this case the mmap, which is clearly over current limit, unmaps dynamic libraries and the testcase segfaults right after returning into userspace. I've also looked at the second instance of the unmapping loop in the do_brk(). The do_brk() is called from brk() syscall and from vm_brk(). The brk() syscall checks for overlapping mappings and bails out when there are any (so it can't be triggered from the brk syscall). The vm_brk() is called only from binmft handlers so it shouldn't be triggered unless binmft handler created overlapping mappings. Signed-off-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:33 +02:00
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/backing-dev.h>
#include <linux/mm.h>
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
#include <linux/vmacache.h>
#include <linux/shm.h>
#include <linux/mman.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/syscalls.h>
#include <linux/capability.h>
#include <linux/init.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/personality.h>
#include <linux/security.h>
#include <linux/hugetlb.h>
#include <linux/shmem_fs.h>
#include <linux/profile.h>
#include <linux/export.h>
#include <linux/mount.h>
#include <linux/mempolicy.h>
#include <linux/rmap.h>
mmu-notifiers: core With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages. There are secondary MMUs (with secondary sptes and secondary tlbs) too. sptes in the kvm case are shadow pagetables, but when I say spte in mmu-notifier context, I mean "secondary pte". In GRU case there's no actual secondary pte and there's only a secondary tlb because the GRU secondary MMU has no knowledge about sptes and every secondary tlb miss event in the MMU always generates a page fault that has to be resolved by the CPU (this is not the case of KVM where the a secondary tlb miss will walk sptes in hardware and it will refill the secondary tlb transparently to software if the corresponding spte is present). The same way zap_page_range has to invalidate the pte before freeing the page, the spte (and secondary tlb) must also be invalidated before any page is freed and reused. Currently we take a page_count pin on every page mapped by sptes, but that means the pages can't be swapped whenever they're mapped by any spte because they're part of the guest working set. Furthermore a spte unmap event can immediately lead to a page to be freed when the pin is released (so requiring the same complex and relatively slow tlb_gather smp safe logic we have in zap_page_range and that can be avoided completely if the spte unmap event doesn't require an unpin of the page previously mapped in the secondary MMU). The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know when the VM is swapping or freeing or doing anything on the primary MMU so that the secondary MMU code can drop sptes before the pages are freed, avoiding all page pinning and allowing 100% reliable swapping of guest physical address space. Furthermore it avoids the code that teardown the mappings of the secondary MMU, to implement a logic like tlb_gather in zap_page_range that would require many IPI to flush other cpu tlbs, for each fixed number of spte unmapped. To make an example: if what happens on the primary MMU is a protection downgrade (from writeable to wrprotect) the secondary MMU mappings will be invalidated, and the next secondary-mmu-page-fault will call get_user_pages and trigger a do_wp_page through get_user_pages if it called get_user_pages with write=1, and it'll re-establishing an updated spte or secondary-tlb-mapping on the copied page. Or it will setup a readonly spte or readonly tlb mapping if it's a guest-read, if it calls get_user_pages with write=0. This is just an example. This allows to map any page pointed by any pte (and in turn visible in the primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an full MMU with both sptes and secondary-tlb like the shadow-pagetable layer with kvm), or a remote DMA in software like XPMEM (hence needing of schedule in XPMEM code to send the invalidate to the remote node, while no need to schedule in kvm/gru as it's an immediate event like invalidating primary-mmu pte). At least for KVM without this patch it's impossible to swap guests reliably. And having this feature and removing the page pin allows several other optimizations that simplify life considerably. Dependencies: 1) mm_take_all_locks() to register the mmu notifier when the whole VM isn't doing anything with "mm". This allows mmu notifier users to keep track if the VM is in the middle of the invalidate_range_begin/end critical section with an atomic counter incraese in range_begin and decreased in range_end. No secondary MMU page fault is allowed to map any spte or secondary tlb reference, while the VM is in the middle of range_begin/end as any page returned by get_user_pages in that critical section could later immediately be freed without any further ->invalidate_page notification (invalidate_range_begin/end works on ranges and ->invalidate_page isn't called immediately before freeing the page). To stop all page freeing and pagetable overwrites the mmap_sem must be taken in write mode and all other anon_vma/i_mmap locks must be taken too. 2) It'd be a waste to add branches in the VM if nobody could possibly run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of mmu notifiers, but this already allows to compile a KVM external module against a kernel with mmu notifiers enabled and from the next pull from kvm.git we'll start using them. And GRU/XPMEM will also be able to continue the development by enabling KVM=m in their config, until they submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n). This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM are all =n. The mmu_notifier_register call can fail because mm_take_all_locks may be interrupted by a signal and return -EINTR. Because mmu_notifier_reigster is used when a driver startup, a failure can be gracefully handled. Here an example of the change applied to kvm to register the mmu notifiers. Usually when a driver startups other allocations are required anyway and -ENOMEM failure paths exists already. struct kvm *kvm_arch_create_vm(void) { struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL); + int err; if (!kvm) return ERR_PTR(-ENOMEM); INIT_LIST_HEAD(&kvm->arch.active_mmu_pages); + kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops; + err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm); + if (err) { + kfree(kvm); + return ERR_PTR(err); + } + return kvm; } mmu_notifier_unregister returns void and it's reliable. The patch also adds a few needed but missing includes that would prevent kernel to compile after these changes on non-x86 archs (x86 didn't need them by luck). [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: fix mm/filemap_xip.c build] [akpm@linux-foundation.org: fix mm/mmu_notifier.c build] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:29 +02:00
#include <linux/mmu_notifier.h>
#include <linux/mmdebug.h>
perf: Do the big rename: Performance Counters -> Performance Events Bye-bye Performance Counters, welcome Performance Events! In the past few months the perfcounters subsystem has grown out its initial role of counting hardware events, and has become (and is becoming) a much broader generic event enumeration, reporting, logging, monitoring, analysis facility. Naming its core object 'perf_counter' and naming the subsystem 'perfcounters' has become more and more of a misnomer. With pending code like hw-breakpoints support the 'counter' name is less and less appropriate. All in one, we've decided to rename the subsystem to 'performance events' and to propagate this rename through all fields, variables and API names. (in an ABI compatible fashion) The word 'event' is also a bit shorter than 'counter' - which makes it slightly more convenient to write/handle as well. Thanks goes to Stephane Eranian who first observed this misnomer and suggested a rename. User-space tooling and ABI compatibility is not affected - this patch should be function-invariant. (Also, defconfigs were not touched to keep the size down.) This patch has been generated via the following script: FILES=$(find * -type f | grep -vE 'oprofile|[^K]config') sed -i \ -e 's/PERF_EVENT_/PERF_RECORD_/g' \ -e 's/PERF_COUNTER/PERF_EVENT/g' \ -e 's/perf_counter/perf_event/g' \ -e 's/nb_counters/nb_events/g' \ -e 's/swcounter/swevent/g' \ -e 's/tpcounter_event/tp_event/g' \ $FILES for N in $(find . -name perf_counter.[ch]); do M=$(echo $N | sed 's/perf_counter/perf_event/g') mv $N $M done FILES=$(find . -name perf_event.*) sed -i \ -e 's/COUNTER_MASK/REG_MASK/g' \ -e 's/COUNTER/EVENT/g' \ -e 's/\<event\>/event_id/g' \ -e 's/counter/event/g' \ -e 's/Counter/Event/g' \ $FILES ... to keep it as correct as possible. This script can also be used by anyone who has pending perfcounters patches - it converts a Linux kernel tree over to the new naming. We tried to time this change to the point in time where the amount of pending patches is the smallest: the end of the merge window. Namespace clashes were fixed up in a preparatory patch - and some stylistic fallout will be fixed up in a subsequent patch. ( NOTE: 'counters' are still the proper terminology when we deal with hardware registers - and these sed scripts are a bit over-eager in renaming them. I've undone some of that, but in case there's something left where 'counter' would be better than 'event' we can undo that on an individual basis instead of touching an otherwise nicely automated patch. ) Suggested-by: Stephane Eranian <eranian@google.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Paul Mackerras <paulus@samba.org> Reviewed-by: Arjan van de Ven <arjan@linux.intel.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Howells <dhowells@redhat.com> Cc: Kyle McMartin <kyle@mcmartin.ca> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: <linux-arch@vger.kernel.org> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 12:02:48 +02:00
#include <linux/perf_event.h>
#include <linux/audit.h>
#include <linux/khugepaged.h>
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
#include <linux/uprobes.h>
#include <linux/rbtree_augmented.h>
#include <linux/notifier.h>
#include <linux/memory.h>
#include <linux/printk.h>
#include <linux/userfaultfd_k.h>
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
#include <linux/moduleparam.h>
mm/core, x86/mm/pkeys: Add execute-only protection keys support Protection keys provide new page-based protection in hardware. But, they have an interesting attribute: they only affect data accesses and never affect instruction fetches. That means that if we set up some memory which is set as "access-disabled" via protection keys, we can still execute from it. This patch uses protection keys to set up mappings to do just that. If a user calls: mmap(..., PROT_EXEC); or mprotect(ptr, sz, PROT_EXEC); (note PROT_EXEC-only without PROT_READ/WRITE), the kernel will notice this, and set a special protection key on the memory. It also sets the appropriate bits in the Protection Keys User Rights (PKRU) register so that the memory becomes unreadable and unwritable. I haven't found any userspace that does this today. With this facility in place, we expect userspace to move to use it eventually. Userspace _could_ start doing this today. Any PROT_EXEC calls get converted to PROT_READ inside the kernel, and would transparently be upgraded to "true" PROT_EXEC with this code. IOW, userspace never has to do any PROT_EXEC runtime detection. This feature provides enhanced protection against leaking executable memory contents. This helps thwart attacks which are attempting to find ROP gadgets on the fly. But, the security provided by this approach is not comprehensive. The PKRU register which controls access permissions is a normal user register writable from unprivileged userspace. An attacker who can execute the 'wrpkru' instruction can easily disable the protection provided by this feature. The protection key that is used for execute-only support is permanently dedicated at compile time. This is fine for now because there is currently no API to set a protection key other than this one. Despite there being a constant PKRU value across the entire system, we do not set it unless this feature is in use in a process. That is to preserve the PKRU XSAVE 'init state', which can lead to faster context switches. PKRU *is* a user register and the kernel is modifying it. That means that code doing: pkru = rdpkru() pkru |= 0x100; mmap(..., PROT_EXEC); wrpkru(pkru); could lose the bits in PKRU that enforce execute-only permissions. To avoid this, we suggest avoiding ever calling mmap() or mprotect() when the PKRU value is expected to be unstable. Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Borislav Petkov <bp@suse.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Chen Gang <gang.chen.5i5j@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave@sr71.net> Cc: David Hildenbrand <dahi@linux.vnet.ibm.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Kees Cook <keescook@chromium.org> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Piotr Kwapulinski <kwapulinski.piotr@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: Vladimir Murzin <vladimir.murzin@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: keescook@google.com Cc: linux-kernel@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160212210240.CB4BB5CA@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-12 22:02:40 +01:00
#include <linux/pkeys.h>
mm: oom: let oom_reap_task and exit_mmap run concurrently This is purely required because exit_aio() may block and exit_mmap() may never start, if the oom_reap_task cannot start running on a mm with mm_users == 0. At the same time if the OOM reaper doesn't wait at all for the memory of the current OOM candidate to be freed by exit_mmap->unmap_vmas, it would generate a spurious OOM kill. If it wasn't because of the exit_aio or similar blocking functions in the last mmput, it would be enough to change the oom_reap_task() in the case it finds mm_users == 0, to wait for a timeout or to wait for __mmput to set MMF_OOM_SKIP itself, but it's not just exit_mmap the problem here so the concurrency of exit_mmap and oom_reap_task is apparently warranted. It's a non standard runtime, exit_mmap() runs without mmap_sem, and oom_reap_task runs with the mmap_sem for reading as usual (kind of MADV_DONTNEED). The race between the two is solved with a combination of tsk_is_oom_victim() (serialized by task_lock) and MMF_OOM_SKIP (serialized by a dummy down_write/up_write cycle on the same lines of the ksm_exit method). If the oom_reap_task() may be running concurrently during exit_mmap, exit_mmap will wait it to finish in down_write (before taking down mm structures that would make the oom_reap_task fail with use after free). If exit_mmap comes first, oom_reap_task() will skip the mm if MMF_OOM_SKIP is already set and in turn all memory is already freed and furthermore the mm data structures may already have been taken down by free_pgtables. [aarcange@redhat.com: incremental one liner] Link: http://lkml.kernel.org/r/20170726164319.GC29716@redhat.com [rientjes@google.com: remove unused mmput_async] Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1708141733130.50317@chino.kir.corp.google.com [aarcange@redhat.com: microoptimization] Link: http://lkml.kernel.org/r/20170817171240.GB5066@redhat.com Link: http://lkml.kernel.org/r/20170726162912.GA29716@redhat.com Fixes: 26db62f179d1 ("oom: keep mm of the killed task available") Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Signed-off-by: David Rientjes <rientjes@google.com> Reported-by: David Rientjes <rientjes@google.com> Tested-by: David Rientjes <rientjes@google.com> Reviewed-by: Michal Hocko <mhocko@suse.com> Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: "Kirill A. Shutemov" <kirill@shutemov.name> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-07 01:25:00 +02:00
#include <linux/oom.h>
coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017: https://lkml.kernel.org/r/alpine.LSU.2.11.1707191716030.2055@eggly.anvils "Not strictly relevant here, but a related note: I was very surprised to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct" In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently. Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code. Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix. For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped. Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case. In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit. Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen. Link: http://lkml.kernel.org/r/20190325224949.11068-1-aarcange@redhat.com Fixes: 86039bd3b4e6 ("userfaultfd: add new syscall to provide memory externalization") Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Jann Horn <jannh@google.com> Suggested-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Peter Xu <peterx@redhat.com> Reviewed-by: Mike Rapoport <rppt@linux.ibm.com> Reviewed-by: Oleg Nesterov <oleg@redhat.com> Reviewed-by: Jann Horn <jannh@google.com> Acked-by: Jason Gunthorpe <jgg@mellanox.com> Acked-by: Michal Hocko <mhocko@suse.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-04-19 02:50:52 +02:00
#include <linux/sched/mm.h>
#include <linux/uaccess.h>
#include <asm/cacheflush.h>
#include <asm/tlb.h>
#include <asm/mmu_context.h>
#include "internal.h"
#ifndef arch_mmap_check
#define arch_mmap_check(addr, len, flags) (0)
#endif
mm: mmap: add new /proc tunable for mmap_base ASLR Address Space Layout Randomization (ASLR) provides a barrier to exploitation of user-space processes in the presence of security vulnerabilities by making it more difficult to find desired code/data which could help an attack. This is done by adding a random offset to the location of regions in the process address space, with a greater range of potential offset values corresponding to better protection/a larger search-space for brute force, but also to greater potential for fragmentation. The offset added to the mmap_base address, which provides the basis for the majority of the mappings for a process, is set once on process exec in arch_pick_mmap_layout() and is done via hard-coded per-arch values, which reflect, hopefully, the best compromise for all systems. The trade-off between increased entropy in the offset value generation and the corresponding increased variability in address space fragmentation is not absolute, however, and some platforms may tolerate higher amounts of entropy. This patch introduces both new Kconfig values and a sysctl interface which may be used to change the amount of entropy used for offset generation on a system. The direct motivation for this change was in response to the libstagefright vulnerabilities that affected Android, specifically to information provided by Google's project zero at: http://googleprojectzero.blogspot.com/2015/09/stagefrightened.html The attack presented therein, by Google's project zero, specifically targeted the limited randomness used to generate the offset added to the mmap_base address in order to craft a brute-force-based attack. Concretely, the attack was against the mediaserver process, which was limited to respawning every 5 seconds, on an arm device. The hard-coded 8 bits used resulted in an average expected success rate of defeating the mmap ASLR after just over 10 minutes (128 tries at 5 seconds a piece). With this patch, and an accompanying increase in the entropy value to 16 bits, the same attack would take an average expected time of over 45 hours (32768 tries), which makes it both less feasible and more likely to be noticed. The introduced Kconfig and sysctl options are limited by per-arch minimum and maximum values, the minimum of which was chosen to match the current hard-coded value and the maximum of which was chosen so as to give the greatest flexibility without generating an invalid mmap_base address, generally a 3-4 bits less than the number of bits in the user-space accessible virtual address space. When decided whether or not to change the default value, a system developer should consider that mmap_base address could be placed anywhere up to 2^(value) bits away from the non-randomized location, which would introduce variable-sized areas above and below the mmap_base address such that the maximum vm_area_struct size may be reduced, preventing very large allocations. This patch (of 4): ASLR only uses as few as 8 bits to generate the random offset for the mmap base address on 32 bit architectures. This value was chosen to prevent a poorly chosen value from dividing the address space in such a way as to prevent large allocations. This may not be an issue on all platforms. Allow the specification of a minimum number of bits so that platforms desiring greater ASLR protection may determine where to place the trade-off. Signed-off-by: Daniel Cashman <dcashman@google.com> Cc: Russell King <linux@arm.linux.org.uk> Acked-by: Kees Cook <keescook@chromium.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Don Zickus <dzickus@redhat.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Heinrich Schuchardt <xypron.glpk@gmx.de> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: David Rientjes <rientjes@google.com> Cc: Mark Salyzyn <salyzyn@android.com> Cc: Jeff Vander Stoep <jeffv@google.com> Cc: Nick Kralevich <nnk@google.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Borislav Petkov <bp@suse.de> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:19:53 +01:00
#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
const int mmap_rnd_bits_min = CONFIG_ARCH_MMAP_RND_BITS_MIN;
const int mmap_rnd_bits_max = CONFIG_ARCH_MMAP_RND_BITS_MAX;
int mmap_rnd_bits __read_mostly = CONFIG_ARCH_MMAP_RND_BITS;
#endif
#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
const int mmap_rnd_compat_bits_min = CONFIG_ARCH_MMAP_RND_COMPAT_BITS_MIN;
const int mmap_rnd_compat_bits_max = CONFIG_ARCH_MMAP_RND_COMPAT_BITS_MAX;
int mmap_rnd_compat_bits __read_mostly = CONFIG_ARCH_MMAP_RND_COMPAT_BITS;
#endif
static bool ignore_rlimit_data;
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
core_param(ignore_rlimit_data, ignore_rlimit_data, bool, 0644);
mm: mmap: add new /proc tunable for mmap_base ASLR Address Space Layout Randomization (ASLR) provides a barrier to exploitation of user-space processes in the presence of security vulnerabilities by making it more difficult to find desired code/data which could help an attack. This is done by adding a random offset to the location of regions in the process address space, with a greater range of potential offset values corresponding to better protection/a larger search-space for brute force, but also to greater potential for fragmentation. The offset added to the mmap_base address, which provides the basis for the majority of the mappings for a process, is set once on process exec in arch_pick_mmap_layout() and is done via hard-coded per-arch values, which reflect, hopefully, the best compromise for all systems. The trade-off between increased entropy in the offset value generation and the corresponding increased variability in address space fragmentation is not absolute, however, and some platforms may tolerate higher amounts of entropy. This patch introduces both new Kconfig values and a sysctl interface which may be used to change the amount of entropy used for offset generation on a system. The direct motivation for this change was in response to the libstagefright vulnerabilities that affected Android, specifically to information provided by Google's project zero at: http://googleprojectzero.blogspot.com/2015/09/stagefrightened.html The attack presented therein, by Google's project zero, specifically targeted the limited randomness used to generate the offset added to the mmap_base address in order to craft a brute-force-based attack. Concretely, the attack was against the mediaserver process, which was limited to respawning every 5 seconds, on an arm device. The hard-coded 8 bits used resulted in an average expected success rate of defeating the mmap ASLR after just over 10 minutes (128 tries at 5 seconds a piece). With this patch, and an accompanying increase in the entropy value to 16 bits, the same attack would take an average expected time of over 45 hours (32768 tries), which makes it both less feasible and more likely to be noticed. The introduced Kconfig and sysctl options are limited by per-arch minimum and maximum values, the minimum of which was chosen to match the current hard-coded value and the maximum of which was chosen so as to give the greatest flexibility without generating an invalid mmap_base address, generally a 3-4 bits less than the number of bits in the user-space accessible virtual address space. When decided whether or not to change the default value, a system developer should consider that mmap_base address could be placed anywhere up to 2^(value) bits away from the non-randomized location, which would introduce variable-sized areas above and below the mmap_base address such that the maximum vm_area_struct size may be reduced, preventing very large allocations. This patch (of 4): ASLR only uses as few as 8 bits to generate the random offset for the mmap base address on 32 bit architectures. This value was chosen to prevent a poorly chosen value from dividing the address space in such a way as to prevent large allocations. This may not be an issue on all platforms. Allow the specification of a minimum number of bits so that platforms desiring greater ASLR protection may determine where to place the trade-off. Signed-off-by: Daniel Cashman <dcashman@google.com> Cc: Russell King <linux@arm.linux.org.uk> Acked-by: Kees Cook <keescook@chromium.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Don Zickus <dzickus@redhat.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Heinrich Schuchardt <xypron.glpk@gmx.de> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: David Rientjes <rientjes@google.com> Cc: Mark Salyzyn <salyzyn@android.com> Cc: Jeff Vander Stoep <jeffv@google.com> Cc: Nick Kralevich <nnk@google.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Borislav Petkov <bp@suse.de> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:19:53 +01:00
[PATCH] freepgt: free_pgtables use vma list Recent woes with some arches needing their own pgd_addr_end macro; and 4-level clear_page_range regression since 2.6.10's clear_page_tables; and its long-standing well-known inefficiency in searching throughout the higher-level page tables for those few entries to clear and free: all can be blamed on ignoring the list of vmas when we free page tables. Replace exit_mmap's clear_page_range of the total user address space by free_pgtables operating on the mm's vma list; unmap_region use it in the same way, giving floor and ceiling beyond which it may not free tables. This brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled, in which case latency fixes spoil unmap_vmas throughput). Beware: the do_mmap_pgoff driver failure case must now use unmap_region instead of zap_page_range, since a page table might have been allocated, and can only be freed while it is touched by some vma. Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted from the clear_page_range levels. (Most of free_pgtables' old code was actually for a non-existent case, prev not properly set up, dating from before hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we might want to add latency lockdrops later; but no attempt to do so yet, going by vma should itself reduce latency. But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful examination: put that off until a later patch of the series. What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma? And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that we need to do more than is done here - every PMD_SIZE ever occupied will be flushed, do we really have to flush every PGDIR_SIZE ever partially occupied? A shame to complicate it unnecessarily. Special thanks to David Miller for time spent repairing my ceilings. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 22:29:15 +02:00
static void unmap_region(struct mm_struct *mm,
struct vm_area_struct *vma, struct vm_area_struct *prev,
unsigned long start, unsigned long end);
/* description of effects of mapping type and prot in current implementation.
* this is due to the limited x86 page protection hardware. The expected
* behavior is in parens:
*
* map_type prot
* PROT_NONE PROT_READ PROT_WRITE PROT_EXEC
* MAP_SHARED r: (no) no r: (yes) yes r: (no) yes r: (no) yes
* w: (no) no w: (no) no w: (yes) yes w: (no) no
* x: (no) no x: (no) yes x: (no) yes x: (yes) yes
*
* MAP_PRIVATE r: (no) no r: (yes) yes r: (no) yes r: (no) yes
* w: (no) no w: (no) no w: (copy) copy w: (no) no
* x: (no) no x: (no) yes x: (no) yes x: (yes) yes
*
* On arm64, PROT_EXEC has the following behaviour for both MAP_SHARED and
* MAP_PRIVATE:
* r: (no) no
* w: (no) no
* x: (yes) yes
*/
pgprot_t protection_map[16] __ro_after_init = {
__P000, __P001, __P010, __P011, __P100, __P101, __P110, __P111,
__S000, __S001, __S010, __S011, __S100, __S101, __S110, __S111
};
x86/pti: Filter at vma->vm_page_prot population commit ce9962bf7e22bb3891655c349faff618922d4a73 0day reported warnings at boot on 32-bit systems without NX support: attempted to set unsupported pgprot: 8000000000000025 bits: 8000000000000000 supported: 7fffffffffffffff WARNING: CPU: 0 PID: 1 at arch/x86/include/asm/pgtable.h:540 handle_mm_fault+0xfc1/0xfe0: check_pgprot at arch/x86/include/asm/pgtable.h:535 (inlined by) pfn_pte at arch/x86/include/asm/pgtable.h:549 (inlined by) do_anonymous_page at mm/memory.c:3169 (inlined by) handle_pte_fault at mm/memory.c:3961 (inlined by) __handle_mm_fault at mm/memory.c:4087 (inlined by) handle_mm_fault at mm/memory.c:4124 The problem is that due to the recent commit which removed auto-massaging of page protections, filtering page permissions at PTE creation time is not longer done, so vma->vm_page_prot is passed unfiltered to PTE creation. Filter the page protections before they are installed in vma->vm_page_prot. Fixes: fb43d6cb91 ("x86/mm: Do not auto-massage page protections") Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@google.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Hugh Dickins <hughd@google.com> Cc: linux-mm@kvack.org Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Nadav Amit <namit@vmware.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Arjan van de Ven <arjan@linux.intel.com> Link: https://lkml.kernel.org/r/20180420222028.99D72858@viggo.jf.intel.com
2018-04-21 00:20:28 +02:00
#ifndef CONFIG_ARCH_HAS_FILTER_PGPROT
static inline pgprot_t arch_filter_pgprot(pgprot_t prot)
{
return prot;
}
#endif
pgprot_t vm_get_page_prot(unsigned long vm_flags)
{
x86/pti: Filter at vma->vm_page_prot population commit ce9962bf7e22bb3891655c349faff618922d4a73 0day reported warnings at boot on 32-bit systems without NX support: attempted to set unsupported pgprot: 8000000000000025 bits: 8000000000000000 supported: 7fffffffffffffff WARNING: CPU: 0 PID: 1 at arch/x86/include/asm/pgtable.h:540 handle_mm_fault+0xfc1/0xfe0: check_pgprot at arch/x86/include/asm/pgtable.h:535 (inlined by) pfn_pte at arch/x86/include/asm/pgtable.h:549 (inlined by) do_anonymous_page at mm/memory.c:3169 (inlined by) handle_pte_fault at mm/memory.c:3961 (inlined by) __handle_mm_fault at mm/memory.c:4087 (inlined by) handle_mm_fault at mm/memory.c:4124 The problem is that due to the recent commit which removed auto-massaging of page protections, filtering page permissions at PTE creation time is not longer done, so vma->vm_page_prot is passed unfiltered to PTE creation. Filter the page protections before they are installed in vma->vm_page_prot. Fixes: fb43d6cb91 ("x86/mm: Do not auto-massage page protections") Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@google.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Hugh Dickins <hughd@google.com> Cc: linux-mm@kvack.org Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Nadav Amit <namit@vmware.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Arjan van de Ven <arjan@linux.intel.com> Link: https://lkml.kernel.org/r/20180420222028.99D72858@viggo.jf.intel.com
2018-04-21 00:20:28 +02:00
pgprot_t ret = __pgprot(pgprot_val(protection_map[vm_flags &
(VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]) |
pgprot_val(arch_vm_get_page_prot(vm_flags)));
x86/pti: Filter at vma->vm_page_prot population commit ce9962bf7e22bb3891655c349faff618922d4a73 0day reported warnings at boot on 32-bit systems without NX support: attempted to set unsupported pgprot: 8000000000000025 bits: 8000000000000000 supported: 7fffffffffffffff WARNING: CPU: 0 PID: 1 at arch/x86/include/asm/pgtable.h:540 handle_mm_fault+0xfc1/0xfe0: check_pgprot at arch/x86/include/asm/pgtable.h:535 (inlined by) pfn_pte at arch/x86/include/asm/pgtable.h:549 (inlined by) do_anonymous_page at mm/memory.c:3169 (inlined by) handle_pte_fault at mm/memory.c:3961 (inlined by) __handle_mm_fault at mm/memory.c:4087 (inlined by) handle_mm_fault at mm/memory.c:4124 The problem is that due to the recent commit which removed auto-massaging of page protections, filtering page permissions at PTE creation time is not longer done, so vma->vm_page_prot is passed unfiltered to PTE creation. Filter the page protections before they are installed in vma->vm_page_prot. Fixes: fb43d6cb91 ("x86/mm: Do not auto-massage page protections") Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@google.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Hugh Dickins <hughd@google.com> Cc: linux-mm@kvack.org Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Nadav Amit <namit@vmware.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Arjan van de Ven <arjan@linux.intel.com> Link: https://lkml.kernel.org/r/20180420222028.99D72858@viggo.jf.intel.com
2018-04-21 00:20:28 +02:00
return arch_filter_pgprot(ret);
}
EXPORT_SYMBOL(vm_get_page_prot);
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
static pgprot_t vm_pgprot_modify(pgprot_t oldprot, unsigned long vm_flags)
{
return pgprot_modify(oldprot, vm_get_page_prot(vm_flags));
}
/* Update vma->vm_page_prot to reflect vma->vm_flags. */
void vma_set_page_prot(struct vm_area_struct *vma)
{
unsigned long vm_flags = vma->vm_flags;
pgprot_t vm_page_prot;
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
vm_page_prot = vm_pgprot_modify(vma->vm_page_prot, vm_flags);
if (vma_wants_writenotify(vma, vm_page_prot)) {
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
vm_flags &= ~VM_SHARED;
vm_page_prot = vm_pgprot_modify(vm_page_prot, vm_flags);
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
}
/* remove_protection_ptes reads vma->vm_page_prot without mmap_sem */
WRITE_ONCE(vma->vm_page_prot, vm_page_prot);
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
}
/*
* Requires inode->i_mapping->i_mmap_rwsem
*/
static void __remove_shared_vm_struct(struct vm_area_struct *vma,
struct file *file, struct address_space *mapping)
{
if (vma->vm_flags & VM_DENYWRITE)
atomic_inc(&file_inode(file)->i_writecount);
if (vma->vm_flags & VM_SHARED)
mapping_unmap_writable(mapping);
flush_dcache_mmap_lock(mapping);
vma_interval_tree_remove(vma, &mapping->i_mmap);
flush_dcache_mmap_unlock(mapping);
}
/*
* Unlink a file-based vm structure from its interval tree, to hide
* vma from rmap and vmtruncate before freeing its page tables.
*/
void unlink_file_vma(struct vm_area_struct *vma)
{
struct file *file = vma->vm_file;
if (file) {
struct address_space *mapping = file->f_mapping;
i_mmap_lock_write(mapping);
__remove_shared_vm_struct(vma, file, mapping);
i_mmap_unlock_write(mapping);
}
}
/*
* Close a vm structure and free it, returning the next.
*/
static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
{
struct vm_area_struct *next = vma->vm_next;
might_sleep();
if (vma->vm_ops && vma->vm_ops->close)
vma->vm_ops->close(vma);
mm: kill vma flag VM_EXECUTABLE and mm->num_exe_file_vmas Currently the kernel sets mm->exe_file during sys_execve() and then tracks number of vmas with VM_EXECUTABLE flag in mm->num_exe_file_vmas, as soon as this counter drops to zero kernel resets mm->exe_file to NULL. Plus it resets mm->exe_file at last mmput() when mm->mm_users drops to zero. VMA with VM_EXECUTABLE flag appears after mapping file with flag MAP_EXECUTABLE, such vmas can appears only at sys_execve() or after vma splitting, because sys_mmap ignores this flag. Usually binfmt module sets mm->exe_file and mmaps executable vmas with this file, they hold mm->exe_file while task is running. comment from v2.6.25-6245-g925d1c4 ("procfs task exe symlink"), where all this stuff was introduced: > The kernel implements readlink of /proc/pid/exe by getting the file from > the first executable VMA. Then the path to the file is reconstructed and > reported as the result. > > Because of the VMA walk the code is slightly different on nommu systems. > This patch avoids separate /proc/pid/exe code on nommu systems. Instead of > walking the VMAs to find the first executable file-backed VMA we store a > reference to the exec'd file in the mm_struct. > > That reference would prevent the filesystem holding the executable file > from being unmounted even after unmapping the VMAs. So we track the number > of VM_EXECUTABLE VMAs and drop the new reference when the last one is > unmapped. This avoids pinning the mounted filesystem. exe_file's vma accounting is hooked into every file mmap/unmmap and vma split/merge just to fix some hypothetical pinning fs from umounting by mm, which already unmapped all its executable files, but still alive. Seems like currently nobody depends on this behaviour. We can try to remove this logic and keep mm->exe_file until final mmput(). mm->exe_file is still protected with mm->mmap_sem, because we want to change it via new sys_prctl(PR_SET_MM_EXE_FILE). Also via this syscall task can change its mm->exe_file and unpin mountpoint explicitly. Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Carsten Otte <cotte@de.ibm.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Eric Paris <eparis@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hugh Dickins <hughd@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Kentaro Takeda <takedakn@nttdata.co.jp> Cc: Matt Helsley <matthltc@us.ibm.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Robert Richter <robert.richter@amd.com> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Venkatesh Pallipadi <venki@google.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:28:54 +02:00
if (vma->vm_file)
fput(vma->vm_file);
mpol_put(vma_policy(vma));
vm_area_free(vma);
return next;
}
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
static int do_brk_flags(unsigned long addr, unsigned long request, unsigned long flags,
struct list_head *uf);
SYSCALL_DEFINE1(brk, unsigned long, brk)
{
unsigned long retval;
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
unsigned long newbrk, oldbrk, origbrk;
struct mm_struct *mm = current->mm;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
struct vm_area_struct *next;
unsigned long min_brk;
bool populate;
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
bool downgraded = false;
LIST_HEAD(uf);
mm: make mmap_sem for write waits killable for mm syscalls This is a follow up work for oom_reaper [1]. As the async OOM killing depends on oom_sem for read we would really appreciate if a holder for write didn't stood in the way. This patchset is changing many of down_write calls to be killable to help those cases when the writer is blocked and waiting for readers to release the lock and so help __oom_reap_task to process the oom victim. Most of the patches are really trivial because the lock is help from a shallow syscall paths where we can return EINTR trivially and allow the current task to die (note that EINTR will never get to the userspace as the task has fatal signal pending). Others seem to be easy as well as the callers are already handling fatal errors and bail and return to userspace which should be sufficient to handle the failure gracefully. I am not familiar with all those code paths so a deeper review is really appreciated. As this work is touching more areas which are not directly connected I have tried to keep the CC list as small as possible and people who I believed would be familiar are CCed only to the specific patches (all should have received the cover though). This patchset is based on linux-next and it depends on down_write_killable for rw_semaphores which got merged into tip locking/rwsem branch and it is merged into this next tree. I guess it would be easiest to route these patches via mmotm because of the dependency on the tip tree but if respective maintainers prefer other way I have no objections. I haven't covered all the mmap_write(mm->mmap_sem) instances here $ git grep "down_write(.*\<mmap_sem\>)" next/master | wc -l 98 $ git grep "down_write(.*\<mmap_sem\>)" | wc -l 62 I have tried to cover those which should be relatively easy to review in this series because this alone should be a nice improvement. Other places can be changed on top. [0] http://lkml.kernel.org/r/1456752417-9626-1-git-send-email-mhocko@kernel.org [1] http://lkml.kernel.org/r/1452094975-551-1-git-send-email-mhocko@kernel.org [2] http://lkml.kernel.org/r/1456750705-7141-1-git-send-email-mhocko@kernel.org This patch (of 18): This is the first step in making mmap_sem write waiters killable. It focuses on the trivial ones which are taking the lock early after entering the syscall and they are not changing state before. Therefore it is very easy to change them to use down_write_killable and immediately return with -EINTR. This will allow the waiter to pass away without blocking the mmap_sem which might be required to make a forward progress. E.g. the oom reaper will need the lock for reading to dismantle the OOM victim address space. The only tricky function in this patch is vm_mmap_pgoff which has many call sites via vm_mmap. To reduce the risk keep vm_mmap with the original non-killable semantic for now. vm_munmap callers do not bother checking the return value so open code it into the munmap syscall path for now for simplicity. Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Mel Gorman <mgorman@suse.de> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Hugh Dickins <hughd@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: David Rientjes <rientjes@google.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-24 01:25:27 +02:00
if (down_write_killable(&mm->mmap_sem))
return -EINTR;
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
origbrk = mm->brk;
#ifdef CONFIG_COMPAT_BRK
/*
* CONFIG_COMPAT_BRK can still be overridden by setting
* randomize_va_space to 2, which will still cause mm->start_brk
* to be arbitrarily shifted
*/
if (current->brk_randomized)
min_brk = mm->start_brk;
else
min_brk = mm->end_data;
#else
min_brk = mm->start_brk;
#endif
if (brk < min_brk)
goto out;
/*
* Check against rlimit here. If this check is done later after the test
* of oldbrk with newbrk then it can escape the test and let the data
* segment grow beyond its set limit the in case where the limit is
* not page aligned -Ram Gupta
*/
if (check_data_rlimit(rlimit(RLIMIT_DATA), brk, mm->start_brk,
mm->end_data, mm->start_data))
goto out;
newbrk = PAGE_ALIGN(brk);
oldbrk = PAGE_ALIGN(mm->brk);
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
if (oldbrk == newbrk) {
mm->brk = brk;
goto success;
}
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
/*
* Always allow shrinking brk.
* __do_munmap() may downgrade mmap_sem to read.
*/
if (brk <= mm->brk) {
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
int ret;
/*
* mm->brk must to be protected by write mmap_sem so update it
* before downgrading mmap_sem. When __do_munmap() fails,
* mm->brk will be restored from origbrk.
*/
mm->brk = brk;
ret = __do_munmap(mm, newbrk, oldbrk-newbrk, &uf, true);
if (ret < 0) {
mm->brk = origbrk;
goto out;
} else if (ret == 1) {
downgraded = true;
}
goto success;
}
/* Check against existing mmap mappings. */
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
next = find_vma(mm, oldbrk);
if (next && newbrk + PAGE_SIZE > vm_start_gap(next))
goto out;
/* Ok, looks good - let it rip. */
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
if (do_brk_flags(oldbrk, newbrk-oldbrk, 0, &uf) < 0)
goto out;
mm->brk = brk;
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
success:
populate = newbrk > oldbrk && (mm->def_flags & VM_LOCKED) != 0;
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
if (downgraded)
up_read(&mm->mmap_sem);
else
up_write(&mm->mmap_sem);
userfaultfd_unmap_complete(mm, &uf);
if (populate)
mm_populate(oldbrk, newbrk - oldbrk);
return brk;
out:
mm: brk: downgrade mmap_sem to read when shrinking brk might be used to shrink memory mapping too other than munmap(). So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The brk() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case. But, it may set mm->brk after __do_munmap(), which needs hold write mmap_sem. However, a simple trick can workaround this by setting mm->brk before __do_munmap(). Then restore the original value if __do_munmap() fails. With this trick, it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-1-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:54 +02:00
retval = origbrk;
up_write(&mm->mmap_sem);
return retval;
}
static inline unsigned long vma_compute_gap(struct vm_area_struct *vma)
{
unsigned long gap, prev_end;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
/*
* Note: in the rare case of a VM_GROWSDOWN above a VM_GROWSUP, we
* allow two stack_guard_gaps between them here, and when choosing
* an unmapped area; whereas when expanding we only require one.
* That's a little inconsistent, but keeps the code here simpler.
*/
gap = vm_start_gap(vma);
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
if (vma->vm_prev) {
prev_end = vm_end_gap(vma->vm_prev);
if (gap > prev_end)
gap -= prev_end;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
else
gap = 0;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
}
return gap;
}
#ifdef CONFIG_DEBUG_VM_RB
static unsigned long vma_compute_subtree_gap(struct vm_area_struct *vma)
{
unsigned long max = vma_compute_gap(vma), subtree_gap;
if (vma->vm_rb.rb_left) {
subtree_gap = rb_entry(vma->vm_rb.rb_left,
struct vm_area_struct, vm_rb)->rb_subtree_gap;
if (subtree_gap > max)
max = subtree_gap;
}
if (vma->vm_rb.rb_right) {
subtree_gap = rb_entry(vma->vm_rb.rb_right,
struct vm_area_struct, vm_rb)->rb_subtree_gap;
if (subtree_gap > max)
max = subtree_gap;
}
return max;
}
static int browse_rb(struct mm_struct *mm)
{
struct rb_root *root = &mm->mm_rb;
int i = 0, j, bug = 0;
struct rb_node *nd, *pn = NULL;
unsigned long prev = 0, pend = 0;
for (nd = rb_first(root); nd; nd = rb_next(nd)) {
struct vm_area_struct *vma;
vma = rb_entry(nd, struct vm_area_struct, vm_rb);
if (vma->vm_start < prev) {
pr_emerg("vm_start %lx < prev %lx\n",
vma->vm_start, prev);
bug = 1;
}
if (vma->vm_start < pend) {
pr_emerg("vm_start %lx < pend %lx\n",
vma->vm_start, pend);
bug = 1;
}
if (vma->vm_start > vma->vm_end) {
pr_emerg("vm_start %lx > vm_end %lx\n",
vma->vm_start, vma->vm_end);
bug = 1;
}
spin_lock(&mm->page_table_lock);
if (vma->rb_subtree_gap != vma_compute_subtree_gap(vma)) {
pr_emerg("free gap %lx, correct %lx\n",
vma->rb_subtree_gap,
vma_compute_subtree_gap(vma));
bug = 1;
}
spin_unlock(&mm->page_table_lock);
i++;
pn = nd;
prev = vma->vm_start;
pend = vma->vm_end;
}
j = 0;
for (nd = pn; nd; nd = rb_prev(nd))
j++;
if (i != j) {
pr_emerg("backwards %d, forwards %d\n", j, i);
bug = 1;
}
return bug ? -1 : i;
}
static void validate_mm_rb(struct rb_root *root, struct vm_area_struct *ignore)
{
struct rb_node *nd;
for (nd = rb_first(root); nd; nd = rb_next(nd)) {
struct vm_area_struct *vma;
vma = rb_entry(nd, struct vm_area_struct, vm_rb);
VM_BUG_ON_VMA(vma != ignore &&
vma->rb_subtree_gap != vma_compute_subtree_gap(vma),
vma);
}
}
static void validate_mm(struct mm_struct *mm)
{
int bug = 0;
int i = 0;
unsigned long highest_address = 0;
struct vm_area_struct *vma = mm->mmap;
while (vma) {
struct anon_vma *anon_vma = vma->anon_vma;
struct anon_vma_chain *avc;
if (anon_vma) {
anon_vma_lock_read(anon_vma);
list_for_each_entry(avc, &vma->anon_vma_chain, same_vma)
anon_vma_interval_tree_verify(avc);
anon_vma_unlock_read(anon_vma);
}
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
highest_address = vm_end_gap(vma);
vma = vma->vm_next;
i++;
}
if (i != mm->map_count) {
pr_emerg("map_count %d vm_next %d\n", mm->map_count, i);
bug = 1;
}
if (highest_address != mm->highest_vm_end) {
pr_emerg("mm->highest_vm_end %lx, found %lx\n",
mm->highest_vm_end, highest_address);
bug = 1;
}
i = browse_rb(mm);
if (i != mm->map_count) {
if (i != -1)
pr_emerg("map_count %d rb %d\n", mm->map_count, i);
bug = 1;
}
VM_BUG_ON_MM(bug, mm);
}
#else
#define validate_mm_rb(root, ignore) do { } while (0)
#define validate_mm(mm) do { } while (0)
#endif
RB_DECLARE_CALLBACKS_MAX(static, vma_gap_callbacks,
struct vm_area_struct, vm_rb,
unsigned long, rb_subtree_gap, vma_compute_gap)
/*
* Update augmented rbtree rb_subtree_gap values after vma->vm_start or
* vma->vm_prev->vm_end values changed, without modifying the vma's position
* in the rbtree.
*/
static void vma_gap_update(struct vm_area_struct *vma)
{
/*
* As it turns out, RB_DECLARE_CALLBACKS_MAX() already created
* a callback function that does exactly what we want.
*/
vma_gap_callbacks_propagate(&vma->vm_rb, NULL);
}
static inline void vma_rb_insert(struct vm_area_struct *vma,
struct rb_root *root)
{
/* All rb_subtree_gap values must be consistent prior to insertion */
validate_mm_rb(root, NULL);
rb_insert_augmented(&vma->vm_rb, root, &vma_gap_callbacks);
}
static void __vma_rb_erase(struct vm_area_struct *vma, struct rb_root *root)
{
/*
* Note rb_erase_augmented is a fairly large inline function,
* so make sure we instantiate it only once with our desired
* augmented rbtree callbacks.
*/
rb_erase_augmented(&vma->vm_rb, root, &vma_gap_callbacks);
}
static __always_inline void vma_rb_erase_ignore(struct vm_area_struct *vma,
struct rb_root *root,
struct vm_area_struct *ignore)
{
/*
* All rb_subtree_gap values must be consistent prior to erase,
* with the possible exception of the "next" vma being erased if
* next->vm_start was reduced.
*/
validate_mm_rb(root, ignore);
__vma_rb_erase(vma, root);
}
static __always_inline void vma_rb_erase(struct vm_area_struct *vma,
struct rb_root *root)
{
/*
* All rb_subtree_gap values must be consistent prior to erase,
* with the possible exception of the vma being erased.
*/
validate_mm_rb(root, vma);
__vma_rb_erase(vma, root);
}
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
/*
* vma has some anon_vma assigned, and is already inserted on that
* anon_vma's interval trees.
*
* Before updating the vma's vm_start / vm_end / vm_pgoff fields, the
* vma must be removed from the anon_vma's interval trees using
* anon_vma_interval_tree_pre_update_vma().
*
* After the update, the vma will be reinserted using
* anon_vma_interval_tree_post_update_vma().
*
* The entire update must be protected by exclusive mmap_sem and by
* the root anon_vma's mutex.
*/
static inline void
anon_vma_interval_tree_pre_update_vma(struct vm_area_struct *vma)
{
struct anon_vma_chain *avc;
list_for_each_entry(avc, &vma->anon_vma_chain, same_vma)
anon_vma_interval_tree_remove(avc, &avc->anon_vma->rb_root);
}
static inline void
anon_vma_interval_tree_post_update_vma(struct vm_area_struct *vma)
{
struct anon_vma_chain *avc;
list_for_each_entry(avc, &vma->anon_vma_chain, same_vma)
anon_vma_interval_tree_insert(avc, &avc->anon_vma->rb_root);
}
static int find_vma_links(struct mm_struct *mm, unsigned long addr,
unsigned long end, struct vm_area_struct **pprev,
struct rb_node ***rb_link, struct rb_node **rb_parent)
{
struct rb_node **__rb_link, *__rb_parent, *rb_prev;
__rb_link = &mm->mm_rb.rb_node;
rb_prev = __rb_parent = NULL;
while (*__rb_link) {
struct vm_area_struct *vma_tmp;
__rb_parent = *__rb_link;
vma_tmp = rb_entry(__rb_parent, struct vm_area_struct, vm_rb);
if (vma_tmp->vm_end > addr) {
/* Fail if an existing vma overlaps the area */
if (vma_tmp->vm_start < end)
return -ENOMEM;
__rb_link = &__rb_parent->rb_left;
} else {
rb_prev = __rb_parent;
__rb_link = &__rb_parent->rb_right;
}
}
*pprev = NULL;
if (rb_prev)
*pprev = rb_entry(rb_prev, struct vm_area_struct, vm_rb);
*rb_link = __rb_link;
*rb_parent = __rb_parent;
return 0;
}
mm/mmap: check for RLIMIT_AS before unmapping Fix a corner case for MAP_FIXED when requested mapping length is larger than rlimit for virtual memory. In such case any overlapping mappings are unmapped before we check for the limit and return ENOMEM. The check is moved before the loop that unmaps overlapping parts of existing mappings. When we are about to hit the limit (currently mapped pages + len > limit) we scan for overlapping pages and check again accounting for them. This fixes situation when userspace program expects that the previous mappings are preserved after the mmap() syscall has returned with error. (POSIX clearly states that successfull mapping shall replace any previous mappings.) This corner case was found and can be tested with LTP testcase: testcases/open_posix_testsuite/conformance/interfaces/mmap/24-2.c In this case the mmap, which is clearly over current limit, unmaps dynamic libraries and the testcase segfaults right after returning into userspace. I've also looked at the second instance of the unmapping loop in the do_brk(). The do_brk() is called from brk() syscall and from vm_brk(). The brk() syscall checks for overlapping mappings and bails out when there are any (so it can't be triggered from the brk syscall). The vm_brk() is called only from binmft handlers so it shouldn't be triggered unless binmft handler created overlapping mappings. Signed-off-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:33 +02:00
static unsigned long count_vma_pages_range(struct mm_struct *mm,
unsigned long addr, unsigned long end)
{
unsigned long nr_pages = 0;
struct vm_area_struct *vma;
/* Find first overlaping mapping */
vma = find_vma_intersection(mm, addr, end);
if (!vma)
return 0;
nr_pages = (min(end, vma->vm_end) -
max(addr, vma->vm_start)) >> PAGE_SHIFT;
/* Iterate over the rest of the overlaps */
for (vma = vma->vm_next; vma; vma = vma->vm_next) {
unsigned long overlap_len;
if (vma->vm_start > end)
break;
overlap_len = min(end, vma->vm_end) - vma->vm_start;
nr_pages += overlap_len >> PAGE_SHIFT;
}
return nr_pages;
}
void __vma_link_rb(struct mm_struct *mm, struct vm_area_struct *vma,
struct rb_node **rb_link, struct rb_node *rb_parent)
{
/* Update tracking information for the gap following the new vma. */
if (vma->vm_next)
vma_gap_update(vma->vm_next);
else
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
mm->highest_vm_end = vm_end_gap(vma);
/*
* vma->vm_prev wasn't known when we followed the rbtree to find the
* correct insertion point for that vma. As a result, we could not
* update the vma vm_rb parents rb_subtree_gap values on the way down.
* So, we first insert the vma with a zero rb_subtree_gap value
* (to be consistent with what we did on the way down), and then
* immediately update the gap to the correct value. Finally we
* rebalance the rbtree after all augmented values have been set.
*/
rb_link_node(&vma->vm_rb, rb_parent, rb_link);
vma->rb_subtree_gap = 0;
vma_gap_update(vma);
vma_rb_insert(vma, &mm->mm_rb);
}
static void __vma_link_file(struct vm_area_struct *vma)
{
struct file *file;
file = vma->vm_file;
if (file) {
struct address_space *mapping = file->f_mapping;
if (vma->vm_flags & VM_DENYWRITE)
atomic_dec(&file_inode(file)->i_writecount);
if (vma->vm_flags & VM_SHARED)
atomic_inc(&mapping->i_mmap_writable);
flush_dcache_mmap_lock(mapping);
vma_interval_tree_insert(vma, &mapping->i_mmap);
flush_dcache_mmap_unlock(mapping);
}
}
static void
__vma_link(struct mm_struct *mm, struct vm_area_struct *vma,
struct vm_area_struct *prev, struct rb_node **rb_link,
struct rb_node *rb_parent)
{
__vma_link_list(mm, vma, prev, rb_parent);
__vma_link_rb(mm, vma, rb_link, rb_parent);
}
static void vma_link(struct mm_struct *mm, struct vm_area_struct *vma,
struct vm_area_struct *prev, struct rb_node **rb_link,
struct rb_node *rb_parent)
{
struct address_space *mapping = NULL;
if (vma->vm_file) {
mapping = vma->vm_file->f_mapping;
i_mmap_lock_write(mapping);
}
__vma_link(mm, vma, prev, rb_link, rb_parent);
__vma_link_file(vma);
if (mapping)
i_mmap_unlock_write(mapping);
mm->map_count++;
validate_mm(mm);
}
/*
* Helper for vma_adjust() in the split_vma insert case: insert a vma into the
* mm's list and rbtree. It has already been inserted into the interval tree.
*/
static void __insert_vm_struct(struct mm_struct *mm, struct vm_area_struct *vma)
{
struct vm_area_struct *prev;
struct rb_node **rb_link, *rb_parent;
if (find_vma_links(mm, vma->vm_start, vma->vm_end,
&prev, &rb_link, &rb_parent))
BUG();
__vma_link(mm, vma, prev, rb_link, rb_parent);
mm->map_count++;
}
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
static __always_inline void __vma_unlink_common(struct mm_struct *mm,
struct vm_area_struct *vma,
struct vm_area_struct *prev,
bool has_prev,
struct vm_area_struct *ignore)
{
struct vm_area_struct *next;
vma_rb_erase_ignore(vma, &mm->mm_rb, ignore);
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
next = vma->vm_next;
if (has_prev)
prev->vm_next = next;
else {
prev = vma->vm_prev;
if (prev)
prev->vm_next = next;
else
mm->mmap = next;
}
if (next)
next->vm_prev = prev;
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
/* Kill the cache */
vmacache_invalidate(mm);
}
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
static inline void __vma_unlink_prev(struct mm_struct *mm,
struct vm_area_struct *vma,
struct vm_area_struct *prev)
{
__vma_unlink_common(mm, vma, prev, true, vma);
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
}
/*
* We cannot adjust vm_start, vm_end, vm_pgoff fields of a vma that
* is already present in an i_mmap tree without adjusting the tree.
* The following helper function should be used when such adjustments
* are necessary. The "insert" vma (if any) is to be inserted
* before we drop the necessary locks.
*/
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
struct vm_area_struct *expand)
{
struct mm_struct *mm = vma->vm_mm;
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
struct vm_area_struct *next = vma->vm_next, *orig_vma = vma;
struct address_space *mapping = NULL;
struct rb_root_cached *root = NULL;
struct anon_vma *anon_vma = NULL;
struct file *file = vma->vm_file;
bool start_changed = false, end_changed = false;
long adjust_next = 0;
int remove_next = 0;
if (next && !insert) {
struct vm_area_struct *exporter = NULL, *importer = NULL;
if (end >= next->vm_end) {
/*
* vma expands, overlapping all the next, and
* perhaps the one after too (mprotect case 6).
* The only other cases that gets here are
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
* case 1, case 7 and case 8.
*/
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
if (next == expand) {
/*
* The only case where we don't expand "vma"
* and we expand "next" instead is case 8.
*/
VM_WARN_ON(end != next->vm_end);
/*
* remove_next == 3 means we're
* removing "vma" and that to do so we
* swapped "vma" and "next".
*/
remove_next = 3;
VM_WARN_ON(file != next->vm_file);
swap(vma, next);
} else {
VM_WARN_ON(expand != vma);
/*
* case 1, 6, 7, remove_next == 2 is case 6,
* remove_next == 1 is case 1 or 7.
*/
remove_next = 1 + (end > next->vm_end);
VM_WARN_ON(remove_next == 2 &&
end != next->vm_next->vm_end);
VM_WARN_ON(remove_next == 1 &&
end != next->vm_end);
/* trim end to next, for case 6 first pass */
end = next->vm_end;
}
exporter = next;
importer = vma;
/*
* If next doesn't have anon_vma, import from vma after
* next, if the vma overlaps with it.
*/
if (remove_next == 2 && !next->anon_vma)
exporter = next->vm_next;
} else if (end > next->vm_start) {
/*
* vma expands, overlapping part of the next:
* mprotect case 5 shifting the boundary up.
*/
adjust_next = (end - next->vm_start) >> PAGE_SHIFT;
exporter = next;
importer = vma;
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
VM_WARN_ON(expand != importer);
} else if (end < vma->vm_end) {
/*
* vma shrinks, and !insert tells it's not
* split_vma inserting another: so it must be
* mprotect case 4 shifting the boundary down.
*/
adjust_next = -((vma->vm_end - end) >> PAGE_SHIFT);
exporter = vma;
importer = next;
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
VM_WARN_ON(expand != importer);
}
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
/*
* Easily overlooked: when mprotect shifts the boundary,
* make sure the expanding vma has anon_vma set if the
* shrinking vma had, to cover any anon pages imported.
*/
if (exporter && exporter->anon_vma && !importer->anon_vma) {
int error;
importer->anon_vma = exporter->anon_vma;
error = anon_vma_clone(importer, exporter);
mm: fix anon_vma->degree underflow in anon_vma endless growing prevention I have constantly stumbled upon "kernel BUG at mm/rmap.c:399!" after upgrading to 3.19 and had no luck with 4.0-rc1 neither. So, after looking into new logic introduced by commit 7a3ef208e662 ("mm: prevent endless growth of anon_vma hierarchy"), I found chances are that unlink_anon_vmas() is called without incrementing dst->anon_vma->degree in anon_vma_clone() due to allocation failure. If dst->anon_vma is not NULL in error path, its degree will be incorrectly decremented in unlink_anon_vmas() and eventually underflow when exiting as a result of another call to unlink_anon_vmas(). That's how "kernel BUG at mm/rmap.c:399!" is triggered for me. This patch fixes the underflow by dropping dst->anon_vma when allocation fails. It's safe to do so regardless of original value of dst->anon_vma because dst->anon_vma doesn't have valid meaning if anon_vma_clone() fails. Besides, callers don't care dst->anon_vma in such case neither. Also suggested by Michal Hocko, we can clean up vma_adjust() a bit as anon_vma_clone() now does the work. [akpm@linux-foundation.org: tweak comment] Fixes: 7a3ef208e662 ("mm: prevent endless growth of anon_vma hierarchy") Signed-off-by: Leon Yu <chianglungyu@gmail.com> Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Reviewed-by: Michal Hocko <mhocko@suse.cz> Acked-by: Rik van Riel <riel@redhat.com> Acked-by: David Rientjes <rientjes@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-03-25 23:55:11 +01:00
if (error)
return error;
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
}
}
again:
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
vma_adjust_trans_huge(orig_vma, start, end, adjust_next);
if (file) {
mapping = file->f_mapping;
root = &mapping->i_mmap;
uprobe_munmap(vma, vma->vm_start, vma->vm_end);
uprobes/core: Optimize probe hits with the help of a counter Maintain a per-mm counter: number of uprobes that are inserted on this process address space. This counter can be used at probe hit time to determine if we need a lookup in the uprobes rbtree. Everytime a probe gets inserted successfully, the probe count is incremented and everytime a probe gets removed, the probe count is decremented. The new uprobe_munmap hook ensures the count is correct on a unmap or remap of a region. We expect that once a uprobe_munmap() is called, the vma goes away. So uprobe_unregister() finding a probe to unregister would either mean unmap event hasnt occurred yet or a mmap event on the same executable file occured after a unmap event. Additionally, uprobe_mmap hook now also gets called: a. on every executable vma that is COWed at fork. b. a vma of interest is newly mapped; breakpoint insertion also happens at the required address. On process creation, make sure the probes count in the child is set correctly. Special cases that are taken care include: a. mremap b. VM_DONTCOPY vmas on fork() c. insertion/removal races in the parent during fork(). Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Jim Keniston <jkenisto@linux.vnet.ibm.com> Cc: Linux-mm <linux-mm@kvack.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Anton Arapov <anton@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20120330182646.10018.85805.sendpatchset@srdronam.in.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2012-03-30 20:26:46 +02:00
if (adjust_next)
uprobe_munmap(next, next->vm_start, next->vm_end);
uprobes/core: Optimize probe hits with the help of a counter Maintain a per-mm counter: number of uprobes that are inserted on this process address space. This counter can be used at probe hit time to determine if we need a lookup in the uprobes rbtree. Everytime a probe gets inserted successfully, the probe count is incremented and everytime a probe gets removed, the probe count is decremented. The new uprobe_munmap hook ensures the count is correct on a unmap or remap of a region. We expect that once a uprobe_munmap() is called, the vma goes away. So uprobe_unregister() finding a probe to unregister would either mean unmap event hasnt occurred yet or a mmap event on the same executable file occured after a unmap event. Additionally, uprobe_mmap hook now also gets called: a. on every executable vma that is COWed at fork. b. a vma of interest is newly mapped; breakpoint insertion also happens at the required address. On process creation, make sure the probes count in the child is set correctly. Special cases that are taken care include: a. mremap b. VM_DONTCOPY vmas on fork() c. insertion/removal races in the parent during fork(). Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Jim Keniston <jkenisto@linux.vnet.ibm.com> Cc: Linux-mm <linux-mm@kvack.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Anton Arapov <anton@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20120330182646.10018.85805.sendpatchset@srdronam.in.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2012-03-30 20:26:46 +02:00
i_mmap_lock_write(mapping);
if (insert) {
/*
* Put into interval tree now, so instantiated pages
* are visible to arm/parisc __flush_dcache_page
* throughout; but we cannot insert into address
* space until vma start or end is updated.
*/
__vma_link_file(insert);
}
}
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma = vma->anon_vma;
if (!anon_vma && adjust_next)
anon_vma = next->anon_vma;
if (anon_vma) {
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
VM_WARN_ON(adjust_next && next->anon_vma &&
anon_vma != next->anon_vma);
mm/rmap, migration: Make rmap_walk_anon() and try_to_unmap_anon() more scalable rmap_walk_anon() and try_to_unmap_anon() appears to be too careful about locking the anon vma: while it needs protection against anon vma list modifications, it does not need exclusive access to the list itself. Transforming this exclusive lock to a read-locked rwsem removes a global lock from the hot path of page-migration intense threaded workloads which can cause pathological performance like this: 96.43% process 0 [kernel.kallsyms] [k] perf_trace_sched_switch | --- perf_trace_sched_switch __schedule schedule schedule_preempt_disabled __mutex_lock_common.isra.6 __mutex_lock_slowpath mutex_lock | |--50.61%-- rmap_walk | move_to_new_page | migrate_pages | migrate_misplaced_page | __do_numa_page.isra.69 | handle_pte_fault | handle_mm_fault | __do_page_fault | do_page_fault | page_fault | __memset_sse2 | | | --100.00%-- worker_thread | | | --100.00%-- start_thread | --49.39%-- page_lock_anon_vma try_to_unmap_anon try_to_unmap migrate_pages migrate_misplaced_page __do_numa_page.isra.69 handle_pte_fault handle_mm_fault __do_page_fault do_page_fault page_fault __memset_sse2 | --100.00%-- worker_thread start_thread With this change applied the profile is now nicely flat and there's no anon-vma related scheduling/blocking. Rename anon_vma_[un]lock() => anon_vma_[un]lock_write(), to make it clearer that it's an exclusive write-lock in that case - suggested by Rik van Riel. Suggested-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Turner <pjt@google.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Christoph Lameter <cl@linux.com> Cc: Rik van Riel <riel@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Mel Gorman <mgorman@suse.de>
2012-12-02 20:56:50 +01:00
anon_vma_lock_write(anon_vma);
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma_interval_tree_pre_update_vma(vma);
if (adjust_next)
anon_vma_interval_tree_pre_update_vma(next);
}
if (root) {
flush_dcache_mmap_lock(mapping);
vma_interval_tree_remove(vma, root);
if (adjust_next)
vma_interval_tree_remove(next, root);
}
if (start != vma->vm_start) {
vma->vm_start = start;
start_changed = true;
}
if (end != vma->vm_end) {
vma->vm_end = end;
end_changed = true;
}
vma->vm_pgoff = pgoff;
if (adjust_next) {
next->vm_start += adjust_next << PAGE_SHIFT;
next->vm_pgoff += adjust_next;
}
if (root) {
if (adjust_next)
vma_interval_tree_insert(next, root);
vma_interval_tree_insert(vma, root);
flush_dcache_mmap_unlock(mapping);
}
if (remove_next) {
/*
* vma_merge has merged next into vma, and needs
* us to remove next before dropping the locks.
*/
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
if (remove_next != 3)
__vma_unlink_prev(mm, next, vma);
else
/*
* vma is not before next if they've been
* swapped.
*
* pre-swap() next->vm_start was reduced so
* tell validate_mm_rb to ignore pre-swap()
* "next" (which is stored in post-swap()
* "vma").
*/
__vma_unlink_common(mm, next, NULL, false, vma);
if (file)
__remove_shared_vm_struct(next, file, mapping);
} else if (insert) {
/*
* split_vma has split insert from vma, and needs
* us to insert it before dropping the locks
* (it may either follow vma or precede it).
*/
__insert_vm_struct(mm, insert);
} else {
if (start_changed)
vma_gap_update(vma);
if (end_changed) {
if (!next)
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
mm->highest_vm_end = vm_end_gap(vma);
else if (!adjust_next)
vma_gap_update(next);
}
}
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
if (anon_vma) {
anon_vma_interval_tree_post_update_vma(vma);
if (adjust_next)
anon_vma_interval_tree_post_update_vma(next);
anon_vma_unlock_write(anon_vma);
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
}
if (mapping)
i_mmap_unlock_write(mapping);
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
if (root) {
uprobe_mmap(vma);
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
if (adjust_next)
uprobe_mmap(next);
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
}
if (remove_next) {
if (file) {
uprobes/core: Decrement uprobe count before the pages are unmapped Uprobes has a callback (uprobe_munmap()) in the unmap path to maintain the uprobes count. In the exit path this callback gets called in unlink_file_vma(). However by the time unlink_file_vma() is called, the pages would have been unmapped (in unmap_vmas()) and the task->rss_stat counts accounted (in zap_pte_range()). If the exiting process has probepoints, uprobe_munmap() checks if the breakpoint instruction was around before decrementing the probe count. This results in a file backed page being reread by uprobe_munmap() and hence it does not find the breakpoint. This patch fixes this problem by moving the callback to unmap_single_vma(). Since unmap_single_vma() may not unmap the complete vma, add start and end parameters to uprobe_munmap(). This bug became apparent courtesy of commit c3f0327f8e9d ("mm: add rss counters consistency check"). Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Jim Keniston <jkenisto@linux.vnet.ibm.com> Cc: Linux-mm <linux-mm@kvack.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Anton Arapov <anton@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20120411103527.23245.9835.sendpatchset@srdronam.in.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2012-04-11 12:35:27 +02:00
uprobe_munmap(next, next->vm_start, next->vm_end);
fput(file);
}
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
if (next->anon_vma)
anon_vma_merge(vma, next);
mm->map_count--;
mpol_put(vma_policy(next));
vm_area_free(next);
/*
* In mprotect's case 6 (see comments on vma_merge),
* we must remove another next too. It would clutter
* up the code too much to do both in one go.
*/
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
if (remove_next != 3) {
/*
* If "next" was removed and vma->vm_end was
* expanded (up) over it, in turn
* "next->vm_prev->vm_end" changed and the
* "vma->vm_next" gap must be updated.
*/
next = vma->vm_next;
} else {
/*
* For the scope of the comment "next" and
* "vma" considered pre-swap(): if "vma" was
* removed, next->vm_start was expanded (down)
* over it and the "next" gap must be updated.
* Because of the swap() the post-swap() "vma"
* actually points to pre-swap() "next"
* (post-swap() "next" as opposed is now a
* dangling pointer).
*/
next = vma;
}
if (remove_next == 2) {
remove_next = 1;
end = next->vm_end;
goto again;
}
else if (next)
vma_gap_update(next);
else {
/*
* If remove_next == 2 we obviously can't
* reach this path.
*
* If remove_next == 3 we can't reach this
* path because pre-swap() next is always not
* NULL. pre-swap() "next" is not being
* removed and its next->vm_end is not altered
* (and furthermore "end" already matches
* next->vm_end in remove_next == 3).
*
* We reach this only in the remove_next == 1
* case if the "next" vma that was removed was
* the highest vma of the mm. However in such
* case next->vm_end == "end" and the extended
* "vma" has vma->vm_end == next->vm_end so
* mm->highest_vm_end doesn't need any update
* in remove_next == 1 case.
*/
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
VM_WARN_ON(mm->highest_vm_end != vm_end_gap(vma));
}
}
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
if (insert && file)
uprobe_mmap(insert);
validate_mm(mm);
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
return 0;
}
/*
* If the vma has a ->close operation then the driver probably needs to release
* per-vma resources, so we don't attempt to merge those.
*/
static inline int is_mergeable_vma(struct vm_area_struct *vma,
struct file *file, unsigned long vm_flags,
struct vm_userfaultfd_ctx vm_userfaultfd_ctx)
{
mm: ignore VM_SOFTDIRTY on VMA merging The VM_SOFTDIRTY bit affects vma merge routine: if two VMAs has all bits in vm_flags matched except dirty bit the kernel can't longer merge them and this forces the kernel to generate new VMAs instead. It finally may lead to the situation when userspace application reaches vm.max_map_count limit and get crashed in worse case | (gimp:11768): GLib-ERROR **: gmem.c:110: failed to allocate 4096 bytes | | (file-tiff-load:12038): LibGimpBase-WARNING **: file-tiff-load: gimp_wire_read(): error | xinit: connection to X server lost | | waiting for X server to shut down | /usr/lib64/gimp/2.0/plug-ins/file-tiff-load terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup https://bugzilla.kernel.org/show_bug.cgi?id=67651 https://bugzilla.gnome.org/show_bug.cgi?id=719619#c0 Initial problem came from missed VM_SOFTDIRTY in do_brk() routine but even if we would set up VM_SOFTDIRTY here, there is still a way to prevent VMAs from merging: one can call | echo 4 > /proc/$PID/clear_refs and clear all VM_SOFTDIRTY over all VMAs presented in memory map, then new do_brk() will try to extend old VMA and finds that dirty bit doesn't match thus new VMA will be generated. As discussed with Pavel, the right approach should be to ignore VM_SOFTDIRTY bit when we're trying to merge VMAs and if merge successed we mark extended VMA with dirty bit where needed. Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Reported-by: Bastian Hougaard <gnome@rvzt.net> Reported-by: Mel Gorman <mgorman@suse.de> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:42 +01:00
/*
* VM_SOFTDIRTY should not prevent from VMA merging, if we
* match the flags but dirty bit -- the caller should mark
* merged VMA as dirty. If dirty bit won't be excluded from
* comparison, we increase pressure on the memory system forcing
mm: ignore VM_SOFTDIRTY on VMA merging The VM_SOFTDIRTY bit affects vma merge routine: if two VMAs has all bits in vm_flags matched except dirty bit the kernel can't longer merge them and this forces the kernel to generate new VMAs instead. It finally may lead to the situation when userspace application reaches vm.max_map_count limit and get crashed in worse case | (gimp:11768): GLib-ERROR **: gmem.c:110: failed to allocate 4096 bytes | | (file-tiff-load:12038): LibGimpBase-WARNING **: file-tiff-load: gimp_wire_read(): error | xinit: connection to X server lost | | waiting for X server to shut down | /usr/lib64/gimp/2.0/plug-ins/file-tiff-load terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup https://bugzilla.kernel.org/show_bug.cgi?id=67651 https://bugzilla.gnome.org/show_bug.cgi?id=719619#c0 Initial problem came from missed VM_SOFTDIRTY in do_brk() routine but even if we would set up VM_SOFTDIRTY here, there is still a way to prevent VMAs from merging: one can call | echo 4 > /proc/$PID/clear_refs and clear all VM_SOFTDIRTY over all VMAs presented in memory map, then new do_brk() will try to extend old VMA and finds that dirty bit doesn't match thus new VMA will be generated. As discussed with Pavel, the right approach should be to ignore VM_SOFTDIRTY bit when we're trying to merge VMAs and if merge successed we mark extended VMA with dirty bit where needed. Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Reported-by: Bastian Hougaard <gnome@rvzt.net> Reported-by: Mel Gorman <mgorman@suse.de> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:42 +01:00
* the kernel to generate new VMAs when old one could be
* extended instead.
*/
if ((vma->vm_flags ^ vm_flags) & ~VM_SOFTDIRTY)
return 0;
if (vma->vm_file != file)
return 0;
if (vma->vm_ops && vma->vm_ops->close)
return 0;
if (!is_mergeable_vm_userfaultfd_ctx(vma, vm_userfaultfd_ctx))
return 0;
return 1;
}
static inline int is_mergeable_anon_vma(struct anon_vma *anon_vma1,
struct anon_vma *anon_vma2,
struct vm_area_struct *vma)
{
/*
* The list_is_singular() test is to avoid merging VMA cloned from
* parents. This can improve scalability caused by anon_vma lock.
*/
if ((!anon_vma1 || !anon_vma2) && (!vma ||
list_is_singular(&vma->anon_vma_chain)))
return 1;
return anon_vma1 == anon_vma2;
}
/*
* Return true if we can merge this (vm_flags,anon_vma,file,vm_pgoff)
* in front of (at a lower virtual address and file offset than) the vma.
*
* We cannot merge two vmas if they have differently assigned (non-NULL)
* anon_vmas, nor if same anon_vma is assigned but offsets incompatible.
*
* We don't check here for the merged mmap wrapping around the end of pagecache
* indices (16TB on ia32) because do_mmap_pgoff() does not permit mmap's which
* wrap, nor mmaps which cover the final page at index -1UL.
*/
static int
can_vma_merge_before(struct vm_area_struct *vma, unsigned long vm_flags,
struct anon_vma *anon_vma, struct file *file,
pgoff_t vm_pgoff,
struct vm_userfaultfd_ctx vm_userfaultfd_ctx)
{
if (is_mergeable_vma(vma, file, vm_flags, vm_userfaultfd_ctx) &&
is_mergeable_anon_vma(anon_vma, vma->anon_vma, vma)) {
if (vma->vm_pgoff == vm_pgoff)
return 1;
}
return 0;
}
/*
* Return true if we can merge this (vm_flags,anon_vma,file,vm_pgoff)
* beyond (at a higher virtual address and file offset than) the vma.
*
* We cannot merge two vmas if they have differently assigned (non-NULL)
* anon_vmas, nor if same anon_vma is assigned but offsets incompatible.
*/
static int
can_vma_merge_after(struct vm_area_struct *vma, unsigned long vm_flags,
struct anon_vma *anon_vma, struct file *file,
pgoff_t vm_pgoff,
struct vm_userfaultfd_ctx vm_userfaultfd_ctx)
{
if (is_mergeable_vma(vma, file, vm_flags, vm_userfaultfd_ctx) &&
is_mergeable_anon_vma(anon_vma, vma->anon_vma, vma)) {
pgoff_t vm_pglen;
vm_pglen = vma_pages(vma);
if (vma->vm_pgoff + vm_pglen == vm_pgoff)
return 1;
}
return 0;
}
/*
* Given a mapping request (addr,end,vm_flags,file,pgoff), figure out
* whether that can be merged with its predecessor or its successor.
* Or both (it neatly fills a hole).
*
* In most cases - when called for mmap, brk or mremap - [addr,end) is
* certain not to be mapped by the time vma_merge is called; but when
* called for mprotect, it is certain to be already mapped (either at
* an offset within prev, or at the start of next), and the flags of
* this area are about to be changed to vm_flags - and the no-change
* case has already been eliminated.
*
* The following mprotect cases have to be considered, where AAAA is
* the area passed down from mprotect_fixup, never extending beyond one
* vma, PPPPPP is the prev vma specified, and NNNNNN the next vma after:
*
* AAAA AAAA AAAA AAAA
* PPPPPPNNNNNN PPPPPPNNNNNN PPPPPPNNNNNN PPPPNNNNXXXX
* cannot merge might become might become might become
* PPNNNNNNNNNN PPPPPPPPPPNN PPPPPPPPPPPP 6 or
* mmap, brk or case 4 below case 5 below PPPPPPPPXXXX 7 or
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
* mremap move: PPPPXXXXXXXX 8
* AAAA
* PPPP NNNN PPPPPPPPPPPP PPPPPPPPNNNN PPPPNNNNNNNN
* might become case 1 below case 2 below case 3 below
*
* It is important for case 8 that the vma NNNN overlapping the
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
* region AAAA is never going to extended over XXXX. Instead XXXX must
* be extended in region AAAA and NNNN must be removed. This way in
* all cases where vma_merge succeeds, the moment vma_adjust drops the
* rmap_locks, the properties of the merged vma will be already
* correct for the whole merged range. Some of those properties like
* vm_page_prot/vm_flags may be accessed by rmap_walks and they must
* be correct for the whole merged range immediately after the
* rmap_locks are released. Otherwise if XXXX would be removed and
* NNNN would be extended over the XXXX range, remove_migration_ptes
* or other rmap walkers (if working on addresses beyond the "end"
* parameter) may establish ptes with the wrong permissions of NNNN
* instead of the right permissions of XXXX.
*/
struct vm_area_struct *vma_merge(struct mm_struct *mm,
struct vm_area_struct *prev, unsigned long addr,
unsigned long end, unsigned long vm_flags,
struct anon_vma *anon_vma, struct file *file,
pgoff_t pgoff, struct mempolicy *policy,
struct vm_userfaultfd_ctx vm_userfaultfd_ctx)
{
pgoff_t pglen = (end - addr) >> PAGE_SHIFT;
struct vm_area_struct *area, *next;
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
int err;
/*
* We later require that vma->vm_flags == vm_flags,
* so this tests vma->vm_flags & VM_SPECIAL, too.
*/
if (vm_flags & VM_SPECIAL)
return NULL;
if (prev)
next = prev->vm_next;
else
next = mm->mmap;
area = next;
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
if (area && area->vm_end == end) /* cases 6, 7, 8 */
next = next->vm_next;
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
/* verify some invariant that must be enforced by the caller */
VM_WARN_ON(prev && addr <= prev->vm_start);
VM_WARN_ON(area && end > area->vm_end);
VM_WARN_ON(addr >= end);
/*
* Can it merge with the predecessor?
*/
if (prev && prev->vm_end == addr &&
mpol_equal(vma_policy(prev), policy) &&
can_vma_merge_after(prev, vm_flags,
anon_vma, file, pgoff,
vm_userfaultfd_ctx)) {
/*
* OK, it can. Can we now merge in the successor as well?
*/
if (next && end == next->vm_start &&
mpol_equal(policy, vma_policy(next)) &&
can_vma_merge_before(next, vm_flags,
anon_vma, file,
pgoff+pglen,
vm_userfaultfd_ctx) &&
is_mergeable_anon_vma(prev->anon_vma,
next->anon_vma, NULL)) {
/* cases 1, 6 */
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
err = __vma_adjust(prev, prev->vm_start,
next->vm_end, prev->vm_pgoff, NULL,
prev);
} else /* cases 2, 5, 7 */
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
err = __vma_adjust(prev, prev->vm_start,
end, prev->vm_pgoff, NULL, prev);
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
if (err)
return NULL;
khugepaged_enter_vma_merge(prev, vm_flags);
return prev;
}
/*
* Can this new request be merged in front of next?
*/
if (next && end == next->vm_start &&
mpol_equal(policy, vma_policy(next)) &&
can_vma_merge_before(next, vm_flags,
anon_vma, file, pgoff+pglen,
vm_userfaultfd_ctx)) {
if (prev && addr < prev->vm_end) /* case 4 */
mm: vma_merge: fix vm_page_prot SMP race condition against rmap_walk The rmap_walk can access vm_page_prot (and potentially vm_flags in the pte/pmd manipulations). So it's not safe to wait the caller to update the vm_page_prot/vm_flags after vma_merge returned potentially removing the "next" vma and extending the "current" vma over the next->vm_start,vm_end range, but still with the "current" vma vm_page_prot, after releasing the rmap locks. The vm_page_prot/vm_flags must be transferred from the "next" vma to the current vma while vma_merge still holds the rmap locks. The side effect of this race condition is pte corruption during migrate as remove_migration_ptes when run on a address of the "next" vma that got removed, used the vm_page_prot of the current vma. migrate mprotect ------------ ------------- migrating in "next" vma vma_merge() # removes "next" vma and # extends "current" vma # current vma is not with # vm_page_prot updated remove_migration_ptes read vm_page_prot of current "vma" establish pte with wrong permissions vm_set_page_prot(vma) # too late! change_protection in the old vma range only, next range is not updated This caused segmentation faults and potentially memory corruption in heavy mprotect loads with some light page migration caused by compaction in the background. Hugh Dickins pointed out the comment about the Odd case 8 in vma_merge which confirms the case 8 is only buggy one where the race can trigger, in all other vma_merge cases the above cannot happen. This fix removes the oddness factor from case 8 and it converts it from: AAAA PPPPNNNNXXXX -> PPPPNNNNNNNN to: AAAA PPPPNNNNXXXX -> PPPPXXXXXXXX XXXX has the right vma properties for the whole merged vma returned by vma_adjust, so it solves the problem fully. It has the added benefits that the callers could stop updating vma properties when vma_merge succeeds however the callers are not updated by this patch (there are bits like VM_SOFTDIRTY that still need special care for the whole range, as the vma merging ignores them, but as long as they're not processed by rmap walks and instead they're accessed with the mmap_sem at least for reading, they are fine not to be updated within vma_adjust before releasing the rmap_locks). Link: http://lkml.kernel.org/r/1474309513-20313-1-git-send-email-aarcange@redhat.com Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Aditya Mandaleeka <adityam@microsoft.com> Cc: Rik van Riel <riel@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Vorlicek <janvorli@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-08 02:01:28 +02:00
err = __vma_adjust(prev, prev->vm_start,
addr, prev->vm_pgoff, NULL, next);
else { /* cases 3, 8 */
err = __vma_adjust(area, addr, next->vm_end,
next->vm_pgoff - pglen, NULL, next);
/*
* In case 3 area is already equal to next and
* this is a noop, but in case 8 "area" has
* been removed and next was expanded over it.
*/
area = next;
}
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
if (err)
return NULL;
khugepaged_enter_vma_merge(area, vm_flags);
return area;
}
return NULL;
}
/*
* Rough compatbility check to quickly see if it's even worth looking
* at sharing an anon_vma.
*
* They need to have the same vm_file, and the flags can only differ
* in things that mprotect may change.
*
* NOTE! The fact that we share an anon_vma doesn't _have_ to mean that
* we can merge the two vma's. For example, we refuse to merge a vma if
* there is a vm_ops->close() function, because that indicates that the
* driver is doing some kind of reference counting. But that doesn't
* really matter for the anon_vma sharing case.
*/
static int anon_vma_compatible(struct vm_area_struct *a, struct vm_area_struct *b)
{
return a->vm_end == b->vm_start &&
mpol_equal(vma_policy(a), vma_policy(b)) &&
a->vm_file == b->vm_file &&
mm: ignore VM_SOFTDIRTY on VMA merging The VM_SOFTDIRTY bit affects vma merge routine: if two VMAs has all bits in vm_flags matched except dirty bit the kernel can't longer merge them and this forces the kernel to generate new VMAs instead. It finally may lead to the situation when userspace application reaches vm.max_map_count limit and get crashed in worse case | (gimp:11768): GLib-ERROR **: gmem.c:110: failed to allocate 4096 bytes | | (file-tiff-load:12038): LibGimpBase-WARNING **: file-tiff-load: gimp_wire_read(): error | xinit: connection to X server lost | | waiting for X server to shut down | /usr/lib64/gimp/2.0/plug-ins/file-tiff-load terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup | /usr/lib64/gimp/2.0/plug-ins/script-fu terminated: Hangup https://bugzilla.kernel.org/show_bug.cgi?id=67651 https://bugzilla.gnome.org/show_bug.cgi?id=719619#c0 Initial problem came from missed VM_SOFTDIRTY in do_brk() routine but even if we would set up VM_SOFTDIRTY here, there is still a way to prevent VMAs from merging: one can call | echo 4 > /proc/$PID/clear_refs and clear all VM_SOFTDIRTY over all VMAs presented in memory map, then new do_brk() will try to extend old VMA and finds that dirty bit doesn't match thus new VMA will be generated. As discussed with Pavel, the right approach should be to ignore VM_SOFTDIRTY bit when we're trying to merge VMAs and if merge successed we mark extended VMA with dirty bit where needed. Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Reported-by: Bastian Hougaard <gnome@rvzt.net> Reported-by: Mel Gorman <mgorman@suse.de> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:42 +01:00
!((a->vm_flags ^ b->vm_flags) & ~(VM_READ|VM_WRITE|VM_EXEC|VM_SOFTDIRTY)) &&
b->vm_pgoff == a->vm_pgoff + ((b->vm_start - a->vm_start) >> PAGE_SHIFT);
}
/*
* Do some basic sanity checking to see if we can re-use the anon_vma
* from 'old'. The 'a'/'b' vma's are in VM order - one of them will be
* the same as 'old', the other will be the new one that is trying
* to share the anon_vma.
*
* NOTE! This runs with mm_sem held for reading, so it is possible that
* the anon_vma of 'old' is concurrently in the process of being set up
* by another page fault trying to merge _that_. But that's ok: if it
* is being set up, that automatically means that it will be a singleton
* acceptable for merging, so we can do all of this optimistically. But
* we do that READ_ONCE() to make sure that we never re-load the pointer.
*
* IOW: that the "list_is_singular()" test on the anon_vma_chain only
* matters for the 'stable anon_vma' case (ie the thing we want to avoid
* is to return an anon_vma that is "complex" due to having gone through
* a fork).
*
* We also make sure that the two vma's are compatible (adjacent,
* and with the same memory policies). That's all stable, even with just
* a read lock on the mm_sem.
*/
static struct anon_vma *reusable_anon_vma(struct vm_area_struct *old, struct vm_area_struct *a, struct vm_area_struct *b)
{
if (anon_vma_compatible(a, b)) {
struct anon_vma *anon_vma = READ_ONCE(old->anon_vma);
if (anon_vma && list_is_singular(&old->anon_vma_chain))
return anon_vma;
}
return NULL;
}
/*
* find_mergeable_anon_vma is used by anon_vma_prepare, to check
* neighbouring vmas for a suitable anon_vma, before it goes off
* to allocate a new anon_vma. It checks because a repetitive
* sequence of mprotects and faults may otherwise lead to distinct
* anon_vmas being allocated, preventing vma merge in subsequent
* mprotect.
*/
struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *vma)
{
struct anon_vma *anon_vma;
struct vm_area_struct *near;
near = vma->vm_next;
if (!near)
goto try_prev;
anon_vma = reusable_anon_vma(near, vma, near);
if (anon_vma)
return anon_vma;
try_prev:
near = vma->vm_prev;
if (!near)
goto none;
anon_vma = reusable_anon_vma(near, near, vma);
if (anon_vma)
return anon_vma;
none:
/*
* There's no absolute need to look only at touching neighbours:
* we could search further afield for "compatible" anon_vmas.
* But it would probably just be a waste of time searching,
* or lead to too many vmas hanging off the same anon_vma.
* We're trying to allow mprotect remerging later on,
* not trying to minimize memory used for anon_vmas.
*/
return NULL;
}
/*
* If a hint addr is less than mmap_min_addr change hint to be as
* low as possible but still greater than mmap_min_addr
*/
static inline unsigned long round_hint_to_min(unsigned long hint)
{
hint &= PAGE_MASK;
if (((void *)hint != NULL) &&
(hint < mmap_min_addr))
return PAGE_ALIGN(mmap_min_addr);
return hint;
}
static inline int mlock_future_check(struct mm_struct *mm,
unsigned long flags,
unsigned long len)
{
unsigned long locked, lock_limit;
/* mlock MCL_FUTURE? */
if (flags & VM_LOCKED) {
locked = len >> PAGE_SHIFT;
locked += mm->locked_vm;
lock_limit = rlimit(RLIMIT_MEMLOCK);
lock_limit >>= PAGE_SHIFT;
if (locked > lock_limit && !capable(CAP_IPC_LOCK))
return -EAGAIN;
}
return 0;
}
mmap: introduce sane default mmap limits The internal VM "mmap()" interfaces are based on the mmap target doing everything using page indexes rather than byte offsets, because traditionally (ie 32-bit) we had the situation that the byte offset didn't fit in a register. So while the mmap virtual address was limited by the word size of the architecture, the backing store was not. So we're basically passing "pgoff" around as a page index, in order to be able to describe backing store locations that are much bigger than the word size (think files larger than 4GB etc). But while this all makes a ton of sense conceptually, we've been dogged by various drivers that don't really understand this, and internally work with byte offsets, and then try to work with the page index by turning it into a byte offset with "pgoff << PAGE_SHIFT". Which obviously can overflow. Adding the size of the mapping to it to get the byte offset of the end of the backing store just exacerbates the problem, and if you then use this overflow-prone value to check various limits of your device driver mmap capability, you're just setting yourself up for problems. The correct thing for drivers to do is to do their limit math in page indices, the way the interface is designed. Because the generic mmap code _does_ test that the index doesn't overflow, since that's what the mmap code really cares about. HOWEVER. Finding and fixing various random drivers is a sisyphean task, so let's just see if we can just make the core mmap() code do the limiting for us. Realistically, the only "big" backing stores we need to care about are regular files and block devices, both of which are known to do this properly, and which have nice well-defined limits for how much data they can access. So let's special-case just those two known cases, and then limit other random mmap users to a backing store that still fits in "unsigned long". Realistically, that's not much of a limit at all on 64-bit, and on 32-bit architectures the only worry might be the GPU drivers, which can have big physical address spaces. To make it possible for drivers like that to say that they are 64-bit clean, this patch does repurpose the "FMODE_UNSIGNED_OFFSET" bit in the file flags to allow drivers to mark their file descriptors as safe in the full 64-bit mmap address space. [ The timing for doing this is less than optimal, and this should really go in a merge window. But realistically, this needs wide testing more than it needs anything else, and being main-line is the only way to do that. So the earlier the better, even if it's outside the proper development cycle - Linus ] Cc: Kees Cook <keescook@chromium.org> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Willy Tarreau <w@1wt.eu> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-11 18:52:01 +02:00
static inline u64 file_mmap_size_max(struct file *file, struct inode *inode)
{
if (S_ISREG(inode->i_mode))
return MAX_LFS_FILESIZE;
mmap: introduce sane default mmap limits The internal VM "mmap()" interfaces are based on the mmap target doing everything using page indexes rather than byte offsets, because traditionally (ie 32-bit) we had the situation that the byte offset didn't fit in a register. So while the mmap virtual address was limited by the word size of the architecture, the backing store was not. So we're basically passing "pgoff" around as a page index, in order to be able to describe backing store locations that are much bigger than the word size (think files larger than 4GB etc). But while this all makes a ton of sense conceptually, we've been dogged by various drivers that don't really understand this, and internally work with byte offsets, and then try to work with the page index by turning it into a byte offset with "pgoff << PAGE_SHIFT". Which obviously can overflow. Adding the size of the mapping to it to get the byte offset of the end of the backing store just exacerbates the problem, and if you then use this overflow-prone value to check various limits of your device driver mmap capability, you're just setting yourself up for problems. The correct thing for drivers to do is to do their limit math in page indices, the way the interface is designed. Because the generic mmap code _does_ test that the index doesn't overflow, since that's what the mmap code really cares about. HOWEVER. Finding and fixing various random drivers is a sisyphean task, so let's just see if we can just make the core mmap() code do the limiting for us. Realistically, the only "big" backing stores we need to care about are regular files and block devices, both of which are known to do this properly, and which have nice well-defined limits for how much data they can access. So let's special-case just those two known cases, and then limit other random mmap users to a backing store that still fits in "unsigned long". Realistically, that's not much of a limit at all on 64-bit, and on 32-bit architectures the only worry might be the GPU drivers, which can have big physical address spaces. To make it possible for drivers like that to say that they are 64-bit clean, this patch does repurpose the "FMODE_UNSIGNED_OFFSET" bit in the file flags to allow drivers to mark their file descriptors as safe in the full 64-bit mmap address space. [ The timing for doing this is less than optimal, and this should really go in a merge window. But realistically, this needs wide testing more than it needs anything else, and being main-line is the only way to do that. So the earlier the better, even if it's outside the proper development cycle - Linus ] Cc: Kees Cook <keescook@chromium.org> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Willy Tarreau <w@1wt.eu> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-11 18:52:01 +02:00
if (S_ISBLK(inode->i_mode))
return MAX_LFS_FILESIZE;
mm: mmap: increase sockets maximum memory size pgoff for 32bits The AF_XDP sockets umem mapping interface uses XDP_UMEM_PGOFF_FILL_RING and XDP_UMEM_PGOFF_COMPLETION_RING offsets. These offsets are established already and are part of the configuration interface. But for 32-bit systems, using AF_XDP socket configuration, these values are too large to pass the maximum allowed file size verification. The offsets can be tuned off, but instead of changing the existing interface, let's extend the max allowed file size for sockets. No one has been using this until this patch with 32 bits as without this fix af_xdp sockets can't be used at all, so it unblocks af_xdp socket usage for 32bit systems. All list of mmap cbs for sockets was verified for side effects and all of them contain dummy cb - sock_no_mmap() at this moment, except the following: xsk_mmap() - it's what this fix is needed for. tcp_mmap() - doesn't have obvious issues with pgoff - no any references on it. packet_mmap() - return -EINVAL if it's even set. Link: http://lkml.kernel.org/r/20190812124326.32146-1-ivan.khoronzhuk@linaro.org Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Björn Töpel <bjorn.topel@intel.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Magnus Karlsson <magnus.karlsson@intel.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: David Miller <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-09-24 00:39:28 +02:00
if (S_ISSOCK(inode->i_mode))
return MAX_LFS_FILESIZE;
mmap: introduce sane default mmap limits The internal VM "mmap()" interfaces are based on the mmap target doing everything using page indexes rather than byte offsets, because traditionally (ie 32-bit) we had the situation that the byte offset didn't fit in a register. So while the mmap virtual address was limited by the word size of the architecture, the backing store was not. So we're basically passing "pgoff" around as a page index, in order to be able to describe backing store locations that are much bigger than the word size (think files larger than 4GB etc). But while this all makes a ton of sense conceptually, we've been dogged by various drivers that don't really understand this, and internally work with byte offsets, and then try to work with the page index by turning it into a byte offset with "pgoff << PAGE_SHIFT". Which obviously can overflow. Adding the size of the mapping to it to get the byte offset of the end of the backing store just exacerbates the problem, and if you then use this overflow-prone value to check various limits of your device driver mmap capability, you're just setting yourself up for problems. The correct thing for drivers to do is to do their limit math in page indices, the way the interface is designed. Because the generic mmap code _does_ test that the index doesn't overflow, since that's what the mmap code really cares about. HOWEVER. Finding and fixing various random drivers is a sisyphean task, so let's just see if we can just make the core mmap() code do the limiting for us. Realistically, the only "big" backing stores we need to care about are regular files and block devices, both of which are known to do this properly, and which have nice well-defined limits for how much data they can access. So let's special-case just those two known cases, and then limit other random mmap users to a backing store that still fits in "unsigned long". Realistically, that's not much of a limit at all on 64-bit, and on 32-bit architectures the only worry might be the GPU drivers, which can have big physical address spaces. To make it possible for drivers like that to say that they are 64-bit clean, this patch does repurpose the "FMODE_UNSIGNED_OFFSET" bit in the file flags to allow drivers to mark their file descriptors as safe in the full 64-bit mmap address space. [ The timing for doing this is less than optimal, and this should really go in a merge window. But realistically, this needs wide testing more than it needs anything else, and being main-line is the only way to do that. So the earlier the better, even if it's outside the proper development cycle - Linus ] Cc: Kees Cook <keescook@chromium.org> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Willy Tarreau <w@1wt.eu> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-11 18:52:01 +02:00
/* Special "we do even unsigned file positions" case */
if (file->f_mode & FMODE_UNSIGNED_OFFSET)
return 0;
/* Yes, random drivers might want more. But I'm tired of buggy drivers */
return ULONG_MAX;
}
static inline bool file_mmap_ok(struct file *file, struct inode *inode,
unsigned long pgoff, unsigned long len)
{
u64 maxsize = file_mmap_size_max(file, inode);
if (maxsize && len > maxsize)
return false;
maxsize -= len;
if (pgoff > maxsize >> PAGE_SHIFT)
return false;
return true;
}
/*
* The caller must hold down_write(&current->mm->mmap_sem).
*/
unsigned long do_mmap(struct file *file, unsigned long addr,
unsigned long len, unsigned long prot,
unsigned long flags, vm_flags_t vm_flags,
unsigned long pgoff, unsigned long *populate,
struct list_head *uf)
{
struct mm_struct *mm = current->mm;
mm/core, x86/mm/pkeys: Add execute-only protection keys support Protection keys provide new page-based protection in hardware. But, they have an interesting attribute: they only affect data accesses and never affect instruction fetches. That means that if we set up some memory which is set as "access-disabled" via protection keys, we can still execute from it. This patch uses protection keys to set up mappings to do just that. If a user calls: mmap(..., PROT_EXEC); or mprotect(ptr, sz, PROT_EXEC); (note PROT_EXEC-only without PROT_READ/WRITE), the kernel will notice this, and set a special protection key on the memory. It also sets the appropriate bits in the Protection Keys User Rights (PKRU) register so that the memory becomes unreadable and unwritable. I haven't found any userspace that does this today. With this facility in place, we expect userspace to move to use it eventually. Userspace _could_ start doing this today. Any PROT_EXEC calls get converted to PROT_READ inside the kernel, and would transparently be upgraded to "true" PROT_EXEC with this code. IOW, userspace never has to do any PROT_EXEC runtime detection. This feature provides enhanced protection against leaking executable memory contents. This helps thwart attacks which are attempting to find ROP gadgets on the fly. But, the security provided by this approach is not comprehensive. The PKRU register which controls access permissions is a normal user register writable from unprivileged userspace. An attacker who can execute the 'wrpkru' instruction can easily disable the protection provided by this feature. The protection key that is used for execute-only support is permanently dedicated at compile time. This is fine for now because there is currently no API to set a protection key other than this one. Despite there being a constant PKRU value across the entire system, we do not set it unless this feature is in use in a process. That is to preserve the PKRU XSAVE 'init state', which can lead to faster context switches. PKRU *is* a user register and the kernel is modifying it. That means that code doing: pkru = rdpkru() pkru |= 0x100; mmap(..., PROT_EXEC); wrpkru(pkru); could lose the bits in PKRU that enforce execute-only permissions. To avoid this, we suggest avoiding ever calling mmap() or mprotect() when the PKRU value is expected to be unstable. Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Borislav Petkov <bp@suse.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Chen Gang <gang.chen.5i5j@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave@sr71.net> Cc: David Hildenbrand <dahi@linux.vnet.ibm.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Kees Cook <keescook@chromium.org> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Piotr Kwapulinski <kwapulinski.piotr@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: Vladimir Murzin <vladimir.murzin@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: keescook@google.com Cc: linux-kernel@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160212210240.CB4BB5CA@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-12 22:02:40 +01:00
int pkey = 0;
*populate = 0;
if (!len)
return -EINVAL;
/*
* Does the application expect PROT_READ to imply PROT_EXEC?
*
* (the exception is when the underlying filesystem is noexec
* mounted, in which case we dont add PROT_EXEC.)
*/
if ((prot & PROT_READ) && (current->personality & READ_IMPLIES_EXEC))
vfs: Commit to never having exectuables on proc and sysfs. Today proc and sysfs do not contain any executable files. Several applications today mount proc or sysfs without noexec and nosuid and then depend on there being no exectuables files on proc or sysfs. Having any executable files show on proc or sysfs would cause a user space visible regression, and most likely security problems. Therefore commit to never allowing executables on proc and sysfs by adding a new flag to mark them as filesystems without executables and enforce that flag. Test the flag where MNT_NOEXEC is tested today, so that the only user visible effect will be that exectuables will be treated as if the execute bit is cleared. The filesystems proc and sysfs do not currently incoporate any executable files so this does not result in any user visible effects. This makes it unnecessary to vet changes to proc and sysfs tightly for adding exectuable files or changes to chattr that would modify existing files, as no matter what the individual file say they will not be treated as exectuable files by the vfs. Not having to vet changes to closely is important as without this we are only one proc_create call (or another goof up in the implementation of notify_change) from having problematic executables on proc. Those mistakes are all too easy to make and would create a situation where there are security issues or the assumptions of some program having to be broken (and cause userspace regressions). Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2015-06-29 21:42:03 +02:00
if (!(file && path_noexec(&file->f_path)))
prot |= PROT_EXEC;
mm: introduce MAP_FIXED_NOREPLACE Patch series "mm: introduce MAP_FIXED_NOREPLACE", v2. This has started as a follow up discussion [3][4] resulting in the runtime failure caused by hardening patch [5] which removes MAP_FIXED from the elf loader because MAP_FIXED is inherently dangerous as it might silently clobber an existing underlying mapping (e.g. stack). The reason for the failure is that some architectures enforce an alignment for the given address hint without MAP_FIXED used (e.g. for shared or file backed mappings). One way around this would be excluding those archs which do alignment tricks from the hardening [6]. The patch is really trivial but it has been objected, rightfully so, that this screams for a more generic solution. We basically want a non-destructive MAP_FIXED. The first patch introduced MAP_FIXED_NOREPLACE which enforces the given address but unlike MAP_FIXED it fails with EEXIST if the given range conflicts with an existing one. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. Except we won't export expose the new semantic to the userspace at all. It seems there are users who would like to have something like that. Jemalloc has been mentioned by Michael Ellerman [7] Florian Weimer has mentioned the following: : glibc ld.so currently maps DSOs without hints. This means that the kernel : will map right next to each other, and the offsets between them a completely : predictable. We would like to change that and supply a random address in a : window of the address space. If there is a conflict, we do not want the : kernel to pick a non-random address. Instead, we would try again with a : random address. John Hubbard has mentioned CUDA example : a) Searches /proc/<pid>/maps for a "suitable" region of available : VA space. "Suitable" generally means it has to have a base address : within a certain limited range (a particular device model might : have odd limitations, for example), it has to be large enough, and : alignment has to be large enough (again, various devices may have : constraints that lead us to do this). : : This is of course subject to races with other threads in the process. : : Let's say it finds a region starting at va. : : b) Next it does: : p = mmap(va, ...) : : *without* setting MAP_FIXED, of course (so va is just a hint), to : attempt to safely reserve that region. If p != va, then in most cases, : this is a failure (almost certainly due to another thread getting a : mapping from that region before we did), and so this layer now has to : call munmap(), before returning a "failure: retry" to upper layers. : : IMPROVEMENT: --> if instead, we could call this: : : p = mmap(va, ... MAP_FIXED_NOREPLACE ...) : : , then we could skip the munmap() call upon failure. This : is a small thing, but it is useful here. (Thanks to Piotr : Jaroszynski and Mark Hairgrove for helping me get that detail : exactly right, btw.) : : c) After that, CUDA suballocates from p, via: : : q = mmap(sub_region_start, ... MAP_FIXED ...) : : Interestingly enough, "freeing" is also done via MAP_FIXED, and : setting PROT_NONE to the subregion. Anyway, I just included (c) for : general interest. Atomic address range probing in the multithreaded programs in general sounds like an interesting thing to me. The second patch simply replaces MAP_FIXED use in elf loader by MAP_FIXED_NOREPLACE. I believe other places which rely on MAP_FIXED should follow. Actually real MAP_FIXED usages should be docummented properly and they should be more of an exception. [1] http://lkml.kernel.org/r/20171116101900.13621-1-mhocko@kernel.org [2] http://lkml.kernel.org/r/20171129144219.22867-1-mhocko@kernel.org [3] http://lkml.kernel.org/r/20171107162217.382cd754@canb.auug.org.au [4] http://lkml.kernel.org/r/1510048229.12079.7.camel@abdul.in.ibm.com [5] http://lkml.kernel.org/r/20171023082608.6167-1-mhocko@kernel.org [6] http://lkml.kernel.org/r/20171113094203.aofz2e7kueitk55y@dhcp22.suse.cz [7] http://lkml.kernel.org/r/87efp1w7vy.fsf@concordia.ellerman.id.au This patch (of 2): MAP_FIXED is used quite often to enforce mapping at the particular range. The main problem of this flag is, however, that it is inherently dangerous because it unmaps existing mappings covered by the requested range. This can cause silent memory corruptions. Some of them even with serious security implications. While the current semantic might be really desiderable in many cases there are others which would want to enforce the given range but rather see a failure than a silent memory corruption on a clashing range. Please note that there is no guarantee that a given range is obeyed by the mmap even when it is free - e.g. arch specific code is allowed to apply an alignment. Introduce a new MAP_FIXED_NOREPLACE flag for mmap to achieve this behavior. It has the same semantic as MAP_FIXED wrt. the given address request with a single exception that it fails with EEXIST if the requested address is already covered by an existing mapping. We still do rely on get_unmaped_area to handle all the arch specific MAP_FIXED treatment and check for a conflicting vma after it returns. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt. flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. [mpe@ellerman.id.au: fix whitespace] [fail on clashing range with EEXIST as per Florian Weimer] [set MAP_FIXED before round_hint_to_min as per Khalid Aziz] Link: http://lkml.kernel.org/r/20171213092550.2774-2-mhocko@kernel.org Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com> Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: Khalid Aziz <khalid.aziz@oracle.com> Cc: Russell King - ARM Linux <linux@armlinux.org.uk> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Florian Weimer <fweimer@redhat.com> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Abdul Haleem <abdhalee@linux.vnet.ibm.com> Cc: Joel Stanley <joel@jms.id.au> Cc: Kees Cook <keescook@chromium.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Jason Evans <jasone@google.com> Cc: David Goldblatt <davidtgoldblatt@gmail.com> Cc: Edward Tomasz Napierała <trasz@FreeBSD.org> Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-11 01:35:57 +02:00
/* force arch specific MAP_FIXED handling in get_unmapped_area */
if (flags & MAP_FIXED_NOREPLACE)
flags |= MAP_FIXED;
if (!(flags & MAP_FIXED))
addr = round_hint_to_min(addr);
/* Careful about overflows.. */
len = PAGE_ALIGN(len);
if (!len)
return -ENOMEM;
/* offset overflow? */
if ((pgoff + (len >> PAGE_SHIFT)) < pgoff)
return -EOVERFLOW;
/* Too many mappings? */
if (mm->map_count > sysctl_max_map_count)
return -ENOMEM;
/* Obtain the address to map to. we verify (or select) it and ensure
* that it represents a valid section of the address space.
*/
addr = get_unmapped_area(file, addr, len, pgoff, flags);
if (offset_in_page(addr))
return addr;
mm: introduce MAP_FIXED_NOREPLACE Patch series "mm: introduce MAP_FIXED_NOREPLACE", v2. This has started as a follow up discussion [3][4] resulting in the runtime failure caused by hardening patch [5] which removes MAP_FIXED from the elf loader because MAP_FIXED is inherently dangerous as it might silently clobber an existing underlying mapping (e.g. stack). The reason for the failure is that some architectures enforce an alignment for the given address hint without MAP_FIXED used (e.g. for shared or file backed mappings). One way around this would be excluding those archs which do alignment tricks from the hardening [6]. The patch is really trivial but it has been objected, rightfully so, that this screams for a more generic solution. We basically want a non-destructive MAP_FIXED. The first patch introduced MAP_FIXED_NOREPLACE which enforces the given address but unlike MAP_FIXED it fails with EEXIST if the given range conflicts with an existing one. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. Except we won't export expose the new semantic to the userspace at all. It seems there are users who would like to have something like that. Jemalloc has been mentioned by Michael Ellerman [7] Florian Weimer has mentioned the following: : glibc ld.so currently maps DSOs without hints. This means that the kernel : will map right next to each other, and the offsets between them a completely : predictable. We would like to change that and supply a random address in a : window of the address space. If there is a conflict, we do not want the : kernel to pick a non-random address. Instead, we would try again with a : random address. John Hubbard has mentioned CUDA example : a) Searches /proc/<pid>/maps for a "suitable" region of available : VA space. "Suitable" generally means it has to have a base address : within a certain limited range (a particular device model might : have odd limitations, for example), it has to be large enough, and : alignment has to be large enough (again, various devices may have : constraints that lead us to do this). : : This is of course subject to races with other threads in the process. : : Let's say it finds a region starting at va. : : b) Next it does: : p = mmap(va, ...) : : *without* setting MAP_FIXED, of course (so va is just a hint), to : attempt to safely reserve that region. If p != va, then in most cases, : this is a failure (almost certainly due to another thread getting a : mapping from that region before we did), and so this layer now has to : call munmap(), before returning a "failure: retry" to upper layers. : : IMPROVEMENT: --> if instead, we could call this: : : p = mmap(va, ... MAP_FIXED_NOREPLACE ...) : : , then we could skip the munmap() call upon failure. This : is a small thing, but it is useful here. (Thanks to Piotr : Jaroszynski and Mark Hairgrove for helping me get that detail : exactly right, btw.) : : c) After that, CUDA suballocates from p, via: : : q = mmap(sub_region_start, ... MAP_FIXED ...) : : Interestingly enough, "freeing" is also done via MAP_FIXED, and : setting PROT_NONE to the subregion. Anyway, I just included (c) for : general interest. Atomic address range probing in the multithreaded programs in general sounds like an interesting thing to me. The second patch simply replaces MAP_FIXED use in elf loader by MAP_FIXED_NOREPLACE. I believe other places which rely on MAP_FIXED should follow. Actually real MAP_FIXED usages should be docummented properly and they should be more of an exception. [1] http://lkml.kernel.org/r/20171116101900.13621-1-mhocko@kernel.org [2] http://lkml.kernel.org/r/20171129144219.22867-1-mhocko@kernel.org [3] http://lkml.kernel.org/r/20171107162217.382cd754@canb.auug.org.au [4] http://lkml.kernel.org/r/1510048229.12079.7.camel@abdul.in.ibm.com [5] http://lkml.kernel.org/r/20171023082608.6167-1-mhocko@kernel.org [6] http://lkml.kernel.org/r/20171113094203.aofz2e7kueitk55y@dhcp22.suse.cz [7] http://lkml.kernel.org/r/87efp1w7vy.fsf@concordia.ellerman.id.au This patch (of 2): MAP_FIXED is used quite often to enforce mapping at the particular range. The main problem of this flag is, however, that it is inherently dangerous because it unmaps existing mappings covered by the requested range. This can cause silent memory corruptions. Some of them even with serious security implications. While the current semantic might be really desiderable in many cases there are others which would want to enforce the given range but rather see a failure than a silent memory corruption on a clashing range. Please note that there is no guarantee that a given range is obeyed by the mmap even when it is free - e.g. arch specific code is allowed to apply an alignment. Introduce a new MAP_FIXED_NOREPLACE flag for mmap to achieve this behavior. It has the same semantic as MAP_FIXED wrt. the given address request with a single exception that it fails with EEXIST if the requested address is already covered by an existing mapping. We still do rely on get_unmaped_area to handle all the arch specific MAP_FIXED treatment and check for a conflicting vma after it returns. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt. flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. [mpe@ellerman.id.au: fix whitespace] [fail on clashing range with EEXIST as per Florian Weimer] [set MAP_FIXED before round_hint_to_min as per Khalid Aziz] Link: http://lkml.kernel.org/r/20171213092550.2774-2-mhocko@kernel.org Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com> Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: Khalid Aziz <khalid.aziz@oracle.com> Cc: Russell King - ARM Linux <linux@armlinux.org.uk> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Florian Weimer <fweimer@redhat.com> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Abdul Haleem <abdhalee@linux.vnet.ibm.com> Cc: Joel Stanley <joel@jms.id.au> Cc: Kees Cook <keescook@chromium.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Jason Evans <jasone@google.com> Cc: David Goldblatt <davidtgoldblatt@gmail.com> Cc: Edward Tomasz Napierała <trasz@FreeBSD.org> Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-11 01:35:57 +02:00
if (flags & MAP_FIXED_NOREPLACE) {
struct vm_area_struct *vma = find_vma(mm, addr);
mm/mmap.c: don't clobber partially overlapping VMA with MAP_FIXED_NOREPLACE Daniel Micay reports that attempting to use MAP_FIXED_NOREPLACE in an application causes that application to randomly crash. The existing check for handling MAP_FIXED_NOREPLACE looks up the first VMA that either overlaps or follows the requested region, and then bails out if that VMA overlaps *the start* of the requested region. It does not bail out if the VMA only overlaps another part of the requested region. Fix it by checking that the found VMA only starts at or after the end of the requested region, in which case there is no overlap. Test case: user@debian:~$ cat mmap_fixed_simple.c #include <sys/mman.h> #include <errno.h> #include <stdio.h> #include <stdlib.h> #include <unistd.h> #ifndef MAP_FIXED_NOREPLACE #define MAP_FIXED_NOREPLACE 0x100000 #endif int main(void) { char *p; errno = 0; p = mmap((void*)0x10001000, 0x4000, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED_NOREPLACE, -1, 0); printf("p1=%p err=%m\n", p); errno = 0; p = mmap((void*)0x10000000, 0x2000, PROT_READ, MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED_NOREPLACE, -1, 0); printf("p2=%p err=%m\n", p); char cmd[100]; sprintf(cmd, "cat /proc/%d/maps", getpid()); system(cmd); return 0; } user@debian:~$ gcc -o mmap_fixed_simple mmap_fixed_simple.c user@debian:~$ ./mmap_fixed_simple p1=0x10001000 err=Success p2=0x10000000 err=Success 10000000-10002000 r--p 00000000 00:00 0 10002000-10005000 ---p 00000000 00:00 0 564a9a06f000-564a9a070000 r-xp 00000000 fe:01 264004 /home/user/mmap_fixed_simple 564a9a26f000-564a9a270000 r--p 00000000 fe:01 264004 /home/user/mmap_fixed_simple 564a9a270000-564a9a271000 rw-p 00001000 fe:01 264004 /home/user/mmap_fixed_simple 564a9a54a000-564a9a56b000 rw-p 00000000 00:00 0 [heap] 7f8eba447000-7f8eba5dc000 r-xp 00000000 fe:01 405885 /lib/x86_64-linux-gnu/libc-2.24.so 7f8eba5dc000-7f8eba7dc000 ---p 00195000 fe:01 405885 /lib/x86_64-linux-gnu/libc-2.24.so 7f8eba7dc000-7f8eba7e0000 r--p 00195000 fe:01 405885 /lib/x86_64-linux-gnu/libc-2.24.so 7f8eba7e0000-7f8eba7e2000 rw-p 00199000 fe:01 405885 /lib/x86_64-linux-gnu/libc-2.24.so 7f8eba7e2000-7f8eba7e6000 rw-p 00000000 00:00 0 7f8eba7e6000-7f8eba809000 r-xp 00000000 fe:01 405876 /lib/x86_64-linux-gnu/ld-2.24.so 7f8eba9e9000-7f8eba9eb000 rw-p 00000000 00:00 0 7f8ebaa06000-7f8ebaa09000 rw-p 00000000 00:00 0 7f8ebaa09000-7f8ebaa0a000 r--p 00023000 fe:01 405876 /lib/x86_64-linux-gnu/ld-2.24.so 7f8ebaa0a000-7f8ebaa0b000 rw-p 00024000 fe:01 405876 /lib/x86_64-linux-gnu/ld-2.24.so 7f8ebaa0b000-7f8ebaa0c000 rw-p 00000000 00:00 0 7ffcc99fa000-7ffcc9a1b000 rw-p 00000000 00:00 0 [stack] 7ffcc9b44000-7ffcc9b47000 r--p 00000000 00:00 0 [vvar] 7ffcc9b47000-7ffcc9b49000 r-xp 00000000 00:00 0 [vdso] ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall] user@debian:~$ uname -a Linux debian 4.19.0-rc6+ #181 SMP Wed Oct 3 23:43:42 CEST 2018 x86_64 GNU/Linux user@debian:~$ As you can see, the first page of the mapping at 0x10001000 was clobbered. Link: http://lkml.kernel.org/r/20181010152736.99475-1-jannh@google.com Fixes: a4ff8e8620d3 ("mm: introduce MAP_FIXED_NOREPLACE") Signed-off-by: Jann Horn <jannh@google.com> Reported-by: Daniel Micay <danielmicay@gmail.com> Acked-by: Michal Hocko <mhocko@suse.com> Acked-by: John Hubbard <jhubbard@nvidia.com> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-10-13 06:34:32 +02:00
if (vma && vma->vm_start < addr + len)
mm: introduce MAP_FIXED_NOREPLACE Patch series "mm: introduce MAP_FIXED_NOREPLACE", v2. This has started as a follow up discussion [3][4] resulting in the runtime failure caused by hardening patch [5] which removes MAP_FIXED from the elf loader because MAP_FIXED is inherently dangerous as it might silently clobber an existing underlying mapping (e.g. stack). The reason for the failure is that some architectures enforce an alignment for the given address hint without MAP_FIXED used (e.g. for shared or file backed mappings). One way around this would be excluding those archs which do alignment tricks from the hardening [6]. The patch is really trivial but it has been objected, rightfully so, that this screams for a more generic solution. We basically want a non-destructive MAP_FIXED. The first patch introduced MAP_FIXED_NOREPLACE which enforces the given address but unlike MAP_FIXED it fails with EEXIST if the given range conflicts with an existing one. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. Except we won't export expose the new semantic to the userspace at all. It seems there are users who would like to have something like that. Jemalloc has been mentioned by Michael Ellerman [7] Florian Weimer has mentioned the following: : glibc ld.so currently maps DSOs without hints. This means that the kernel : will map right next to each other, and the offsets between them a completely : predictable. We would like to change that and supply a random address in a : window of the address space. If there is a conflict, we do not want the : kernel to pick a non-random address. Instead, we would try again with a : random address. John Hubbard has mentioned CUDA example : a) Searches /proc/<pid>/maps for a "suitable" region of available : VA space. "Suitable" generally means it has to have a base address : within a certain limited range (a particular device model might : have odd limitations, for example), it has to be large enough, and : alignment has to be large enough (again, various devices may have : constraints that lead us to do this). : : This is of course subject to races with other threads in the process. : : Let's say it finds a region starting at va. : : b) Next it does: : p = mmap(va, ...) : : *without* setting MAP_FIXED, of course (so va is just a hint), to : attempt to safely reserve that region. If p != va, then in most cases, : this is a failure (almost certainly due to another thread getting a : mapping from that region before we did), and so this layer now has to : call munmap(), before returning a "failure: retry" to upper layers. : : IMPROVEMENT: --> if instead, we could call this: : : p = mmap(va, ... MAP_FIXED_NOREPLACE ...) : : , then we could skip the munmap() call upon failure. This : is a small thing, but it is useful here. (Thanks to Piotr : Jaroszynski and Mark Hairgrove for helping me get that detail : exactly right, btw.) : : c) After that, CUDA suballocates from p, via: : : q = mmap(sub_region_start, ... MAP_FIXED ...) : : Interestingly enough, "freeing" is also done via MAP_FIXED, and : setting PROT_NONE to the subregion. Anyway, I just included (c) for : general interest. Atomic address range probing in the multithreaded programs in general sounds like an interesting thing to me. The second patch simply replaces MAP_FIXED use in elf loader by MAP_FIXED_NOREPLACE. I believe other places which rely on MAP_FIXED should follow. Actually real MAP_FIXED usages should be docummented properly and they should be more of an exception. [1] http://lkml.kernel.org/r/20171116101900.13621-1-mhocko@kernel.org [2] http://lkml.kernel.org/r/20171129144219.22867-1-mhocko@kernel.org [3] http://lkml.kernel.org/r/20171107162217.382cd754@canb.auug.org.au [4] http://lkml.kernel.org/r/1510048229.12079.7.camel@abdul.in.ibm.com [5] http://lkml.kernel.org/r/20171023082608.6167-1-mhocko@kernel.org [6] http://lkml.kernel.org/r/20171113094203.aofz2e7kueitk55y@dhcp22.suse.cz [7] http://lkml.kernel.org/r/87efp1w7vy.fsf@concordia.ellerman.id.au This patch (of 2): MAP_FIXED is used quite often to enforce mapping at the particular range. The main problem of this flag is, however, that it is inherently dangerous because it unmaps existing mappings covered by the requested range. This can cause silent memory corruptions. Some of them even with serious security implications. While the current semantic might be really desiderable in many cases there are others which would want to enforce the given range but rather see a failure than a silent memory corruption on a clashing range. Please note that there is no guarantee that a given range is obeyed by the mmap even when it is free - e.g. arch specific code is allowed to apply an alignment. Introduce a new MAP_FIXED_NOREPLACE flag for mmap to achieve this behavior. It has the same semantic as MAP_FIXED wrt. the given address request with a single exception that it fails with EEXIST if the requested address is already covered by an existing mapping. We still do rely on get_unmaped_area to handle all the arch specific MAP_FIXED treatment and check for a conflicting vma after it returns. The flag is introduced as a completely new one rather than a MAP_FIXED extension because of the backward compatibility. We really want a never-clobber semantic even on older kernels which do not recognize the flag. Unfortunately mmap sucks wrt. flags evaluation because we do not EINVAL on unknown flags. On those kernels we would simply use the traditional hint based semantic so the caller can still get a different address (which sucks) but at least not silently corrupt an existing mapping. I do not see a good way around that. [mpe@ellerman.id.au: fix whitespace] [fail on clashing range with EEXIST as per Florian Weimer] [set MAP_FIXED before round_hint_to_min as per Khalid Aziz] Link: http://lkml.kernel.org/r/20171213092550.2774-2-mhocko@kernel.org Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com> Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: Khalid Aziz <khalid.aziz@oracle.com> Cc: Russell King - ARM Linux <linux@armlinux.org.uk> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Florian Weimer <fweimer@redhat.com> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Abdul Haleem <abdhalee@linux.vnet.ibm.com> Cc: Joel Stanley <joel@jms.id.au> Cc: Kees Cook <keescook@chromium.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Jason Evans <jasone@google.com> Cc: David Goldblatt <davidtgoldblatt@gmail.com> Cc: Edward Tomasz Napierała <trasz@FreeBSD.org> Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-11 01:35:57 +02:00
return -EEXIST;
}
mm/core, x86/mm/pkeys: Add execute-only protection keys support Protection keys provide new page-based protection in hardware. But, they have an interesting attribute: they only affect data accesses and never affect instruction fetches. That means that if we set up some memory which is set as "access-disabled" via protection keys, we can still execute from it. This patch uses protection keys to set up mappings to do just that. If a user calls: mmap(..., PROT_EXEC); or mprotect(ptr, sz, PROT_EXEC); (note PROT_EXEC-only without PROT_READ/WRITE), the kernel will notice this, and set a special protection key on the memory. It also sets the appropriate bits in the Protection Keys User Rights (PKRU) register so that the memory becomes unreadable and unwritable. I haven't found any userspace that does this today. With this facility in place, we expect userspace to move to use it eventually. Userspace _could_ start doing this today. Any PROT_EXEC calls get converted to PROT_READ inside the kernel, and would transparently be upgraded to "true" PROT_EXEC with this code. IOW, userspace never has to do any PROT_EXEC runtime detection. This feature provides enhanced protection against leaking executable memory contents. This helps thwart attacks which are attempting to find ROP gadgets on the fly. But, the security provided by this approach is not comprehensive. The PKRU register which controls access permissions is a normal user register writable from unprivileged userspace. An attacker who can execute the 'wrpkru' instruction can easily disable the protection provided by this feature. The protection key that is used for execute-only support is permanently dedicated at compile time. This is fine for now because there is currently no API to set a protection key other than this one. Despite there being a constant PKRU value across the entire system, we do not set it unless this feature is in use in a process. That is to preserve the PKRU XSAVE 'init state', which can lead to faster context switches. PKRU *is* a user register and the kernel is modifying it. That means that code doing: pkru = rdpkru() pkru |= 0x100; mmap(..., PROT_EXEC); wrpkru(pkru); could lose the bits in PKRU that enforce execute-only permissions. To avoid this, we suggest avoiding ever calling mmap() or mprotect() when the PKRU value is expected to be unstable. Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Borislav Petkov <bp@suse.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Chen Gang <gang.chen.5i5j@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave@sr71.net> Cc: David Hildenbrand <dahi@linux.vnet.ibm.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Kees Cook <keescook@chromium.org> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Piotr Kwapulinski <kwapulinski.piotr@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: Vladimir Murzin <vladimir.murzin@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: keescook@google.com Cc: linux-kernel@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160212210240.CB4BB5CA@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-12 22:02:40 +01:00
if (prot == PROT_EXEC) {
pkey = execute_only_pkey(mm);
if (pkey < 0)
pkey = 0;
}
/* Do simple checking here so the lower-level routines won't have
* to. we assume access permissions have been handled by the open
* of the memory object, so we don't do any here.
*/
mm/core, x86/mm/pkeys: Add execute-only protection keys support Protection keys provide new page-based protection in hardware. But, they have an interesting attribute: they only affect data accesses and never affect instruction fetches. That means that if we set up some memory which is set as "access-disabled" via protection keys, we can still execute from it. This patch uses protection keys to set up mappings to do just that. If a user calls: mmap(..., PROT_EXEC); or mprotect(ptr, sz, PROT_EXEC); (note PROT_EXEC-only without PROT_READ/WRITE), the kernel will notice this, and set a special protection key on the memory. It also sets the appropriate bits in the Protection Keys User Rights (PKRU) register so that the memory becomes unreadable and unwritable. I haven't found any userspace that does this today. With this facility in place, we expect userspace to move to use it eventually. Userspace _could_ start doing this today. Any PROT_EXEC calls get converted to PROT_READ inside the kernel, and would transparently be upgraded to "true" PROT_EXEC with this code. IOW, userspace never has to do any PROT_EXEC runtime detection. This feature provides enhanced protection against leaking executable memory contents. This helps thwart attacks which are attempting to find ROP gadgets on the fly. But, the security provided by this approach is not comprehensive. The PKRU register which controls access permissions is a normal user register writable from unprivileged userspace. An attacker who can execute the 'wrpkru' instruction can easily disable the protection provided by this feature. The protection key that is used for execute-only support is permanently dedicated at compile time. This is fine for now because there is currently no API to set a protection key other than this one. Despite there being a constant PKRU value across the entire system, we do not set it unless this feature is in use in a process. That is to preserve the PKRU XSAVE 'init state', which can lead to faster context switches. PKRU *is* a user register and the kernel is modifying it. That means that code doing: pkru = rdpkru() pkru |= 0x100; mmap(..., PROT_EXEC); wrpkru(pkru); could lose the bits in PKRU that enforce execute-only permissions. To avoid this, we suggest avoiding ever calling mmap() or mprotect() when the PKRU value is expected to be unstable. Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Borislav Petkov <bp@suse.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Chen Gang <gang.chen.5i5j@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave@sr71.net> Cc: David Hildenbrand <dahi@linux.vnet.ibm.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Kees Cook <keescook@chromium.org> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Piotr Kwapulinski <kwapulinski.piotr@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: Vladimir Murzin <vladimir.murzin@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: keescook@google.com Cc: linux-kernel@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160212210240.CB4BB5CA@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-12 22:02:40 +01:00
vm_flags |= calc_vm_prot_bits(prot, pkey) | calc_vm_flag_bits(flags) |
mm->def_flags | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC;
if (flags & MAP_LOCKED)
if (!can_do_mlock())
return -EPERM;
if (mlock_future_check(mm, vm_flags, len))
return -EAGAIN;
if (file) {
struct inode *inode = file_inode(file);
mm: introduce MAP_SHARED_VALIDATE, a mechanism to safely define new mmap flags The mmap(2) syscall suffers from the ABI anti-pattern of not validating unknown flags. However, proposals like MAP_SYNC need a mechanism to define new behavior that is known to fail on older kernels without the support. Define a new MAP_SHARED_VALIDATE flag pattern that is guaranteed to fail on all legacy mmap implementations. It is worth noting that the original proposal was for a standalone MAP_VALIDATE flag. However, when that could not be supported by all archs Linus observed: I see why you *think* you want a bitmap. You think you want a bitmap because you want to make MAP_VALIDATE be part of MAP_SYNC etc, so that people can do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_SYNC, fd, 0); and "know" that MAP_SYNC actually takes. And I'm saying that whole wish is bogus. You're fundamentally depending on special semantics, just make it explicit. It's already not portable, so don't try to make it so. Rename that MAP_VALIDATE as MAP_SHARED_VALIDATE, make it have a value of 0x3, and make people do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED_VALIDATE | MAP_SYNC, fd, 0); and then the kernel side is easier too (none of that random garbage playing games with looking at the "MAP_VALIDATE bit", but just another case statement in that map type thing. Boom. Done. Similar to ->fallocate() we also want the ability to validate the support for new flags on a per ->mmap() 'struct file_operations' instance basis. Towards that end arrange for flags to be generically validated against a mmap_supported_flags exported by 'struct file_operations'. By default all existing flags are implicitly supported, but new flags require MAP_SHARED_VALIDATE and per-instance-opt-in. Cc: Jan Kara <jack@suse.cz> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Suggested-by: Christoph Hellwig <hch@lst.de> Suggested-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-11-01 16:36:30 +01:00
unsigned long flags_mask;
mmap: introduce sane default mmap limits The internal VM "mmap()" interfaces are based on the mmap target doing everything using page indexes rather than byte offsets, because traditionally (ie 32-bit) we had the situation that the byte offset didn't fit in a register. So while the mmap virtual address was limited by the word size of the architecture, the backing store was not. So we're basically passing "pgoff" around as a page index, in order to be able to describe backing store locations that are much bigger than the word size (think files larger than 4GB etc). But while this all makes a ton of sense conceptually, we've been dogged by various drivers that don't really understand this, and internally work with byte offsets, and then try to work with the page index by turning it into a byte offset with "pgoff << PAGE_SHIFT". Which obviously can overflow. Adding the size of the mapping to it to get the byte offset of the end of the backing store just exacerbates the problem, and if you then use this overflow-prone value to check various limits of your device driver mmap capability, you're just setting yourself up for problems. The correct thing for drivers to do is to do their limit math in page indices, the way the interface is designed. Because the generic mmap code _does_ test that the index doesn't overflow, since that's what the mmap code really cares about. HOWEVER. Finding and fixing various random drivers is a sisyphean task, so let's just see if we can just make the core mmap() code do the limiting for us. Realistically, the only "big" backing stores we need to care about are regular files and block devices, both of which are known to do this properly, and which have nice well-defined limits for how much data they can access. So let's special-case just those two known cases, and then limit other random mmap users to a backing store that still fits in "unsigned long". Realistically, that's not much of a limit at all on 64-bit, and on 32-bit architectures the only worry might be the GPU drivers, which can have big physical address spaces. To make it possible for drivers like that to say that they are 64-bit clean, this patch does repurpose the "FMODE_UNSIGNED_OFFSET" bit in the file flags to allow drivers to mark their file descriptors as safe in the full 64-bit mmap address space. [ The timing for doing this is less than optimal, and this should really go in a merge window. But realistically, this needs wide testing more than it needs anything else, and being main-line is the only way to do that. So the earlier the better, even if it's outside the proper development cycle - Linus ] Cc: Kees Cook <keescook@chromium.org> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Willy Tarreau <w@1wt.eu> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-11 18:52:01 +02:00
if (!file_mmap_ok(file, inode, pgoff, len))
return -EOVERFLOW;
mm: introduce MAP_SHARED_VALIDATE, a mechanism to safely define new mmap flags The mmap(2) syscall suffers from the ABI anti-pattern of not validating unknown flags. However, proposals like MAP_SYNC need a mechanism to define new behavior that is known to fail on older kernels without the support. Define a new MAP_SHARED_VALIDATE flag pattern that is guaranteed to fail on all legacy mmap implementations. It is worth noting that the original proposal was for a standalone MAP_VALIDATE flag. However, when that could not be supported by all archs Linus observed: I see why you *think* you want a bitmap. You think you want a bitmap because you want to make MAP_VALIDATE be part of MAP_SYNC etc, so that people can do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_SYNC, fd, 0); and "know" that MAP_SYNC actually takes. And I'm saying that whole wish is bogus. You're fundamentally depending on special semantics, just make it explicit. It's already not portable, so don't try to make it so. Rename that MAP_VALIDATE as MAP_SHARED_VALIDATE, make it have a value of 0x3, and make people do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED_VALIDATE | MAP_SYNC, fd, 0); and then the kernel side is easier too (none of that random garbage playing games with looking at the "MAP_VALIDATE bit", but just another case statement in that map type thing. Boom. Done. Similar to ->fallocate() we also want the ability to validate the support for new flags on a per ->mmap() 'struct file_operations' instance basis. Towards that end arrange for flags to be generically validated against a mmap_supported_flags exported by 'struct file_operations'. By default all existing flags are implicitly supported, but new flags require MAP_SHARED_VALIDATE and per-instance-opt-in. Cc: Jan Kara <jack@suse.cz> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Suggested-by: Christoph Hellwig <hch@lst.de> Suggested-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-11-01 16:36:30 +01:00
flags_mask = LEGACY_MAP_MASK | file->f_op->mmap_supported_flags;
switch (flags & MAP_TYPE) {
case MAP_SHARED:
mm: introduce MAP_SHARED_VALIDATE, a mechanism to safely define new mmap flags The mmap(2) syscall suffers from the ABI anti-pattern of not validating unknown flags. However, proposals like MAP_SYNC need a mechanism to define new behavior that is known to fail on older kernels without the support. Define a new MAP_SHARED_VALIDATE flag pattern that is guaranteed to fail on all legacy mmap implementations. It is worth noting that the original proposal was for a standalone MAP_VALIDATE flag. However, when that could not be supported by all archs Linus observed: I see why you *think* you want a bitmap. You think you want a bitmap because you want to make MAP_VALIDATE be part of MAP_SYNC etc, so that people can do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_SYNC, fd, 0); and "know" that MAP_SYNC actually takes. And I'm saying that whole wish is bogus. You're fundamentally depending on special semantics, just make it explicit. It's already not portable, so don't try to make it so. Rename that MAP_VALIDATE as MAP_SHARED_VALIDATE, make it have a value of 0x3, and make people do ret = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED_VALIDATE | MAP_SYNC, fd, 0); and then the kernel side is easier too (none of that random garbage playing games with looking at the "MAP_VALIDATE bit", but just another case statement in that map type thing. Boom. Done. Similar to ->fallocate() we also want the ability to validate the support for new flags on a per ->mmap() 'struct file_operations' instance basis. Towards that end arrange for flags to be generically validated against a mmap_supported_flags exported by 'struct file_operations'. By default all existing flags are implicitly supported, but new flags require MAP_SHARED_VALIDATE and per-instance-opt-in. Cc: Jan Kara <jack@suse.cz> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Suggested-by: Christoph Hellwig <hch@lst.de> Suggested-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-11-01 16:36:30 +01:00
/*
* Force use of MAP_SHARED_VALIDATE with non-legacy
* flags. E.g. MAP_SYNC is dangerous to use with
* MAP_SHARED as you don't know which consistency model
* you will get. We silently ignore unsupported flags
* with MAP_SHARED to preserve backward compatibility.
*/
flags &= LEGACY_MAP_MASK;
/* fall through */
case MAP_SHARED_VALIDATE:
if (flags & ~flags_mask)
return -EOPNOTSUPP;
if (prot & PROT_WRITE) {
if (!(file->f_mode & FMODE_WRITE))
return -EACCES;
if (IS_SWAPFILE(file->f_mapping->host))
return -ETXTBSY;
}
/*
* Make sure we don't allow writing to an append-only
* file..
*/
if (IS_APPEND(inode) && (file->f_mode & FMODE_WRITE))
return -EACCES;
/*
* Make sure there are no mandatory locks on the file.
*/
if (locks_verify_locked(file))
return -EAGAIN;
vm_flags |= VM_SHARED | VM_MAYSHARE;
if (!(file->f_mode & FMODE_WRITE))
vm_flags &= ~(VM_MAYWRITE | VM_SHARED);
/* fall through */
case MAP_PRIVATE:
if (!(file->f_mode & FMODE_READ))
return -EACCES;
vfs: Commit to never having exectuables on proc and sysfs. Today proc and sysfs do not contain any executable files. Several applications today mount proc or sysfs without noexec and nosuid and then depend on there being no exectuables files on proc or sysfs. Having any executable files show on proc or sysfs would cause a user space visible regression, and most likely security problems. Therefore commit to never allowing executables on proc and sysfs by adding a new flag to mark them as filesystems without executables and enforce that flag. Test the flag where MNT_NOEXEC is tested today, so that the only user visible effect will be that exectuables will be treated as if the execute bit is cleared. The filesystems proc and sysfs do not currently incoporate any executable files so this does not result in any user visible effects. This makes it unnecessary to vet changes to proc and sysfs tightly for adding exectuable files or changes to chattr that would modify existing files, as no matter what the individual file say they will not be treated as exectuable files by the vfs. Not having to vet changes to closely is important as without this we are only one proc_create call (or another goof up in the implementation of notify_change) from having problematic executables on proc. Those mistakes are all too easy to make and would create a situation where there are security issues or the assumptions of some program having to be broken (and cause userspace regressions). Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2015-06-29 21:42:03 +02:00
if (path_noexec(&file->f_path)) {
if (vm_flags & VM_EXEC)
return -EPERM;
vm_flags &= ~VM_MAYEXEC;
}
if (!file->f_op->mmap)
return -ENODEV;
if (vm_flags & (VM_GROWSDOWN|VM_GROWSUP))
return -EINVAL;
break;
default:
return -EINVAL;
}
} else {
switch (flags & MAP_TYPE) {
case MAP_SHARED:
if (vm_flags & (VM_GROWSDOWN|VM_GROWSUP))
return -EINVAL;
/*
* Ignore pgoff.
*/
pgoff = 0;
vm_flags |= VM_SHARED | VM_MAYSHARE;
break;
case MAP_PRIVATE:
/*
* Set pgoff according to addr for anon_vma.
*/
pgoff = addr >> PAGE_SHIFT;
break;
default:
return -EINVAL;
}
}
/*
* Set 'VM_NORESERVE' if we should not account for the
* memory use of this mapping.
*/
if (flags & MAP_NORESERVE) {
/* We honor MAP_NORESERVE if allowed to overcommit */
if (sysctl_overcommit_memory != OVERCOMMIT_NEVER)
vm_flags |= VM_NORESERVE;
/* hugetlb applies strict overcommit unless MAP_NORESERVE */
if (file && is_file_hugepages(file))
vm_flags |= VM_NORESERVE;
}
addr = mmap_region(file, addr, len, vm_flags, pgoff, uf);
if (!IS_ERR_VALUE(addr) &&
((vm_flags & VM_LOCKED) ||
(flags & (MAP_POPULATE | MAP_NONBLOCK)) == MAP_POPULATE))
*populate = len;
return addr;
}
unsigned long ksys_mmap_pgoff(unsigned long addr, unsigned long len,
unsigned long prot, unsigned long flags,
unsigned long fd, unsigned long pgoff)
{
struct file *file = NULL;
unsigned long retval;
if (!(flags & MAP_ANONYMOUS)) {
audit_mmap_fd(fd, flags);
file = fget(fd);
if (!file)
return -EBADF;
if (is_file_hugepages(file))
len = ALIGN(len, huge_page_size(hstate_file(file)));
retval = -EINVAL;
if (unlikely(flags & MAP_HUGETLB && !is_file_hugepages(file)))
goto out_fput;
} else if (flags & MAP_HUGETLB) {
struct user_struct *user = NULL;
struct hstate *hs;
hs = hstate_sizelog((flags >> MAP_HUGE_SHIFT) & MAP_HUGE_MASK);
if (!hs)
return -EINVAL;
len = ALIGN(len, huge_page_size(hs));
/*
* VM_NORESERVE is used because the reservations will be
* taken when vm_ops->mmap() is called
* A dummy user value is used because we are not locking
* memory so no accounting is necessary
*/
file = hugetlb_file_setup(HUGETLB_ANON_FILE, len,
mm: support more pagesizes for MAP_HUGETLB/SHM_HUGETLB There was some desire in large applications using MAP_HUGETLB or SHM_HUGETLB to use 1GB huge pages on some mappings, and stay with 2MB on others. This is useful together with NUMA policy: use 2MB interleaving on some mappings, but 1GB on local mappings. This patch extends the IPC/SHM syscall interfaces slightly to allow specifying the page size. It borrows some upper bits in the existing flag arguments and allows encoding the log of the desired page size in addition to the *_HUGETLB flag. When 0 is specified the default size is used, this makes the change fully compatible. Extending the internal hugetlb code to handle this is straight forward. Instead of a single mount it just keeps an array of them and selects the right mount based on the specified page size. When no page size is specified it uses the mount of the default page size. The change is not visible in /proc/mounts because internal mounts don't appear there. It also has very little overhead: the additional mounts just consume a super block, but not more memory when not used. I also exported the new flags to the user headers (they were previously under __KERNEL__). Right now only symbols for x86 and some other architecture for 1GB and 2MB are defined. The interface should already work for all other architectures though. Only architectures that define multiple hugetlb sizes actually need it (that is currently x86, tile, powerpc). However tile and powerpc have user configurable hugetlb sizes, so it's not easy to add defines. A program on those architectures would need to query sysfs and use the appropiate log2. [akpm@linux-foundation.org: cleanups] [rientjes@google.com: fix build] [akpm@linux-foundation.org: checkpatch fixes] Signed-off-by: Andi Kleen <ak@linux.intel.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Acked-by: Rik van Riel <riel@redhat.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Hillf Danton <dhillf@gmail.com> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-12-12 01:01:34 +01:00
VM_NORESERVE,
&user, HUGETLB_ANONHUGE_INODE,
(flags >> MAP_HUGE_SHIFT) & MAP_HUGE_MASK);
if (IS_ERR(file))
return PTR_ERR(file);
}
flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
retval = vm_mmap_pgoff(file, addr, len, prot, flags, pgoff);
out_fput:
if (file)
fput(file);
return retval;
}
SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len,
unsigned long, prot, unsigned long, flags,
unsigned long, fd, unsigned long, pgoff)
{
return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff);
}
#ifdef __ARCH_WANT_SYS_OLD_MMAP
struct mmap_arg_struct {
unsigned long addr;
unsigned long len;
unsigned long prot;
unsigned long flags;
unsigned long fd;
unsigned long offset;
};
SYSCALL_DEFINE1(old_mmap, struct mmap_arg_struct __user *, arg)
{
struct mmap_arg_struct a;
if (copy_from_user(&a, arg, sizeof(a)))
return -EFAULT;
if (offset_in_page(a.offset))
return -EINVAL;
return ksys_mmap_pgoff(a.addr, a.len, a.prot, a.flags, a.fd,
a.offset >> PAGE_SHIFT);
}
#endif /* __ARCH_WANT_SYS_OLD_MMAP */
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
/*
* Some shared mappings will want the pages marked read-only
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
* to track write events. If so, we'll downgrade vm_page_prot
* to the private version (using protection_map[] without the
* VM_SHARED bit).
*/
int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot)
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
{
vm_flags_t vm_flags = vma->vm_flags;
const struct vm_operations_struct *vm_ops = vma->vm_ops;
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
/* If it was private or non-writable, the write bit is already clear */
if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
return 0;
/* The backer wishes to know when pages are first written to? */
if (vm_ops && (vm_ops->page_mkwrite || vm_ops->pfn_mkwrite))
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
return 1;
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
/* The open routine did something to the protections that pgprot_modify
* won't preserve? */
if (pgprot_val(vm_page_prot) !=
pgprot_val(vm_pgprot_modify(vm_page_prot, vm_flags)))
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
return 0;
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
/* Do we need to track softdirty? */
if (IS_ENABLED(CONFIG_MEM_SOFT_DIRTY) && !(vm_flags & VM_SOFTDIRTY))
return 1;
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
/* Specialty mapping? */
mm: kill vma flag VM_INSERTPAGE Merge VM_INSERTPAGE into VM_MIXEDMAP. VM_MIXEDMAP VMA can mix pure-pfn ptes, special ptes and normal ptes. Now copy_page_range() always copies VM_MIXEDMAP VMA on fork like VM_PFNMAP. If driver populates whole VMA at mmap() it probably not expects page-faults. This patch removes special check from vma_wants_writenotify() which disables pages write tracking for VMA populated via vm_instert_page(). BDI below mapped file should not use dirty-accounting, moreover do_wp_page() can handle this. vm_insert_page() still marks vma after first usage. Usually it is called from f_op->mmap() handler under mm->mmap_sem write-lock, so it able to change vma->vm_flags. Caller must set VM_MIXEDMAP at mmap time if it wants to call this function from other places, for example from page-fault handler. Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Carsten Otte <cotte@de.ibm.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Eric Paris <eparis@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hugh Dickins <hughd@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Kentaro Takeda <takedakn@nttdata.co.jp> Cc: Matt Helsley <matthltc@us.ibm.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Robert Richter <robert.richter@amd.com> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Venkatesh Pallipadi <venki@google.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:28:40 +02:00
if (vm_flags & VM_PFNMAP)
Remove fs.h from mm.h Remove fs.h from mm.h. For this, 1) Uninline vma_wants_writenotify(). It's pretty huge anyway. 2) Add back fs.h or less bloated headers (err.h) to files that need it. As result, on x86_64 allyesconfig, fs.h dependencies cut down from 3929 files rebuilt down to 3444 (-12.3%). Cross-compile tested without regressions on my two usual configs and (sigh): alpha arm-mx1ads mips-bigsur powerpc-ebony alpha-allnoconfig arm-neponset mips-capcella powerpc-g5 alpha-defconfig arm-netwinder mips-cobalt powerpc-holly alpha-up arm-netx mips-db1000 powerpc-iseries arm arm-ns9xxx mips-db1100 powerpc-linkstation arm-assabet arm-omap_h2_1610 mips-db1200 powerpc-lite5200 arm-at91rm9200dk arm-onearm mips-db1500 powerpc-maple arm-at91rm9200ek arm-picotux200 mips-db1550 powerpc-mpc7448_hpc2 arm-at91sam9260ek arm-pleb mips-ddb5477 powerpc-mpc8272_ads arm-at91sam9261ek arm-pnx4008 mips-decstation powerpc-mpc8313_rdb arm-at91sam9263ek arm-pxa255-idp mips-e55 powerpc-mpc832x_mds arm-at91sam9rlek arm-realview mips-emma2rh powerpc-mpc832x_rdb arm-ateb9200 arm-realview-smp mips-excite powerpc-mpc834x_itx arm-badge4 arm-rpc mips-fulong powerpc-mpc834x_itxgp arm-carmeva arm-s3c2410 mips-ip22 powerpc-mpc834x_mds arm-cerfcube arm-shannon mips-ip27 powerpc-mpc836x_mds arm-clps7500 arm-shark mips-ip32 powerpc-mpc8540_ads arm-collie arm-simpad mips-jazz powerpc-mpc8544_ds arm-corgi arm-spitz mips-jmr3927 powerpc-mpc8560_ads arm-csb337 arm-trizeps4 mips-malta powerpc-mpc8568mds arm-csb637 arm-versatile mips-mipssim powerpc-mpc85xx_cds arm-ebsa110 i386 mips-mpc30x powerpc-mpc8641_hpcn arm-edb7211 i386-allnoconfig mips-msp71xx powerpc-mpc866_ads arm-em_x270 i386-defconfig mips-ocelot powerpc-mpc885_ads arm-ep93xx i386-up mips-pb1100 powerpc-pasemi arm-footbridge ia64 mips-pb1500 powerpc-pmac32 arm-fortunet ia64-allnoconfig mips-pb1550 powerpc-ppc64 arm-h3600 ia64-bigsur mips-pnx8550-jbs powerpc-prpmc2800 arm-h7201 ia64-defconfig mips-pnx8550-stb810 powerpc-ps3 arm-h7202 ia64-gensparse mips-qemu powerpc-pseries arm-hackkit ia64-sim mips-rbhma4200 powerpc-up arm-integrator ia64-sn2 mips-rbhma4500 s390 arm-iop13xx ia64-tiger mips-rm200 s390-allnoconfig arm-iop32x ia64-up mips-sb1250-swarm s390-defconfig arm-iop33x ia64-zx1 mips-sead s390-up arm-ixp2000 m68k mips-tb0219 sparc arm-ixp23xx m68k-amiga mips-tb0226 sparc-allnoconfig arm-ixp4xx m68k-apollo mips-tb0287 sparc-defconfig arm-jornada720 m68k-atari mips-workpad sparc-up arm-kafa m68k-bvme6000 mips-wrppmc sparc64 arm-kb9202 m68k-hp300 mips-yosemite sparc64-allnoconfig arm-ks8695 m68k-mac parisc sparc64-defconfig arm-lart m68k-mvme147 parisc-allnoconfig sparc64-up arm-lpd270 m68k-mvme16x parisc-defconfig um-x86_64 arm-lpd7a400 m68k-q40 parisc-up x86_64 arm-lpd7a404 m68k-sun3 powerpc x86_64-allnoconfig arm-lubbock m68k-sun3x powerpc-cell x86_64-defconfig arm-lusl7200 mips powerpc-celleb x86_64-up arm-mainstone mips-atlas powerpc-chrp32 Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-30 00:36:13 +02:00
return 0;
/* Can the mapping track the dirty pages? */
return vma->vm_file && vma->vm_file->f_mapping &&
mapping_cap_account_dirty(vma->vm_file->f_mapping);
}
/*
* We account for memory if it's a private writeable mapping,
Do not account for the address space used by hugetlbfs using VM_ACCOUNT When overcommit is disabled, the core VM accounts for pages used by anonymous shared, private mappings and special mappings. It keeps track of VMAs that should be accounted for with VM_ACCOUNT and VMAs that never had a reserve with VM_NORESERVE. Overcommit for hugetlbfs is much riskier than overcommit for base pages due to contiguity requirements. It avoids overcommiting on both shared and private mappings using reservation counters that are checked and updated during mmap(). This ensures (within limits) that hugepages exist in the future when faults occurs or it is too easy to applications to be SIGKILLed. As hugetlbfs makes its own reservations of a different unit to the base page size, VM_ACCOUNT should never be set. Even if the units were correct, we would double account for the usage in the core VM and hugetlbfs. VM_NORESERVE may be set because an application can request no reserves be made for hugetlbfs at the risk of getting killed later. With commit fc8744adc870a8d4366908221508bb113d8b72ee, VM_NORESERVE and VM_ACCOUNT are getting unconditionally set for hugetlbfs-backed mappings. This breaks the accounting for both the core VM and hugetlbfs, can trigger an OOM storm when hugepage pools are too small lockups and corrupted counters otherwise are used. This patch brings hugetlbfs more in line with how the core VM treats VM_NORESERVE but prevents VM_ACCOUNT being set. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-10 15:02:27 +01:00
* not hugepages and VM_NORESERVE wasn't set.
*/
static inline int accountable_mapping(struct file *file, vm_flags_t vm_flags)
{
Do not account for the address space used by hugetlbfs using VM_ACCOUNT When overcommit is disabled, the core VM accounts for pages used by anonymous shared, private mappings and special mappings. It keeps track of VMAs that should be accounted for with VM_ACCOUNT and VMAs that never had a reserve with VM_NORESERVE. Overcommit for hugetlbfs is much riskier than overcommit for base pages due to contiguity requirements. It avoids overcommiting on both shared and private mappings using reservation counters that are checked and updated during mmap(). This ensures (within limits) that hugepages exist in the future when faults occurs or it is too easy to applications to be SIGKILLed. As hugetlbfs makes its own reservations of a different unit to the base page size, VM_ACCOUNT should never be set. Even if the units were correct, we would double account for the usage in the core VM and hugetlbfs. VM_NORESERVE may be set because an application can request no reserves be made for hugetlbfs at the risk of getting killed later. With commit fc8744adc870a8d4366908221508bb113d8b72ee, VM_NORESERVE and VM_ACCOUNT are getting unconditionally set for hugetlbfs-backed mappings. This breaks the accounting for both the core VM and hugetlbfs, can trigger an OOM storm when hugepage pools are too small lockups and corrupted counters otherwise are used. This patch brings hugetlbfs more in line with how the core VM treats VM_NORESERVE but prevents VM_ACCOUNT being set. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-10 15:02:27 +01:00
/*
* hugetlb has its own accounting separate from the core VM
* VM_HUGETLB may not be set yet so we cannot check for that flag.
*/
if (file && is_file_hugepages(file))
return 0;
return (vm_flags & (VM_NORESERVE | VM_SHARED | VM_WRITE)) == VM_WRITE;
}
unsigned long mmap_region(struct file *file, unsigned long addr,
unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
struct list_head *uf)
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma, *prev;
int error;
struct rb_node **rb_link, *rb_parent;
unsigned long charged = 0;
mm/mmap: check for RLIMIT_AS before unmapping Fix a corner case for MAP_FIXED when requested mapping length is larger than rlimit for virtual memory. In such case any overlapping mappings are unmapped before we check for the limit and return ENOMEM. The check is moved before the loop that unmaps overlapping parts of existing mappings. When we are about to hit the limit (currently mapped pages + len > limit) we scan for overlapping pages and check again accounting for them. This fixes situation when userspace program expects that the previous mappings are preserved after the mmap() syscall has returned with error. (POSIX clearly states that successfull mapping shall replace any previous mappings.) This corner case was found and can be tested with LTP testcase: testcases/open_posix_testsuite/conformance/interfaces/mmap/24-2.c In this case the mmap, which is clearly over current limit, unmaps dynamic libraries and the testcase segfaults right after returning into userspace. I've also looked at the second instance of the unmapping loop in the do_brk(). The do_brk() is called from brk() syscall and from vm_brk(). The brk() syscall checks for overlapping mappings and bails out when there are any (so it can't be triggered from the brk syscall). The vm_brk() is called only from binmft handlers so it shouldn't be triggered unless binmft handler created overlapping mappings. Signed-off-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:33 +02:00
/* Check against address space limit. */
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
if (!may_expand_vm(mm, vm_flags, len >> PAGE_SHIFT)) {
mm/mmap: check for RLIMIT_AS before unmapping Fix a corner case for MAP_FIXED when requested mapping length is larger than rlimit for virtual memory. In such case any overlapping mappings are unmapped before we check for the limit and return ENOMEM. The check is moved before the loop that unmaps overlapping parts of existing mappings. When we are about to hit the limit (currently mapped pages + len > limit) we scan for overlapping pages and check again accounting for them. This fixes situation when userspace program expects that the previous mappings are preserved after the mmap() syscall has returned with error. (POSIX clearly states that successfull mapping shall replace any previous mappings.) This corner case was found and can be tested with LTP testcase: testcases/open_posix_testsuite/conformance/interfaces/mmap/24-2.c In this case the mmap, which is clearly over current limit, unmaps dynamic libraries and the testcase segfaults right after returning into userspace. I've also looked at the second instance of the unmapping loop in the do_brk(). The do_brk() is called from brk() syscall and from vm_brk(). The brk() syscall checks for overlapping mappings and bails out when there are any (so it can't be triggered from the brk syscall). The vm_brk() is called only from binmft handlers so it shouldn't be triggered unless binmft handler created overlapping mappings. Signed-off-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:33 +02:00
unsigned long nr_pages;
/*
* MAP_FIXED may remove pages of mappings that intersects with
* requested mapping. Account for the pages it would unmap.
*/
nr_pages = count_vma_pages_range(mm, addr, addr + len);
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
if (!may_expand_vm(mm, vm_flags,
(len >> PAGE_SHIFT) - nr_pages))
mm/mmap: check for RLIMIT_AS before unmapping Fix a corner case for MAP_FIXED when requested mapping length is larger than rlimit for virtual memory. In such case any overlapping mappings are unmapped before we check for the limit and return ENOMEM. The check is moved before the loop that unmaps overlapping parts of existing mappings. When we are about to hit the limit (currently mapped pages + len > limit) we scan for overlapping pages and check again accounting for them. This fixes situation when userspace program expects that the previous mappings are preserved after the mmap() syscall has returned with error. (POSIX clearly states that successfull mapping shall replace any previous mappings.) This corner case was found and can be tested with LTP testcase: testcases/open_posix_testsuite/conformance/interfaces/mmap/24-2.c In this case the mmap, which is clearly over current limit, unmaps dynamic libraries and the testcase segfaults right after returning into userspace. I've also looked at the second instance of the unmapping loop in the do_brk(). The do_brk() is called from brk() syscall and from vm_brk(). The brk() syscall checks for overlapping mappings and bails out when there are any (so it can't be triggered from the brk syscall). The vm_brk() is called only from binmft handlers so it shouldn't be triggered unless binmft handler created overlapping mappings. Signed-off-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:33 +02:00
return -ENOMEM;
}
/* Clear old maps */
while (find_vma_links(mm, addr, addr + len, &prev, &rb_link,
&rb_parent)) {
if (do_munmap(mm, addr, len, uf))
return -ENOMEM;
}
/*
* Private writable mapping: check memory availability
*/
Do not account for the address space used by hugetlbfs using VM_ACCOUNT When overcommit is disabled, the core VM accounts for pages used by anonymous shared, private mappings and special mappings. It keeps track of VMAs that should be accounted for with VM_ACCOUNT and VMAs that never had a reserve with VM_NORESERVE. Overcommit for hugetlbfs is much riskier than overcommit for base pages due to contiguity requirements. It avoids overcommiting on both shared and private mappings using reservation counters that are checked and updated during mmap(). This ensures (within limits) that hugepages exist in the future when faults occurs or it is too easy to applications to be SIGKILLed. As hugetlbfs makes its own reservations of a different unit to the base page size, VM_ACCOUNT should never be set. Even if the units were correct, we would double account for the usage in the core VM and hugetlbfs. VM_NORESERVE may be set because an application can request no reserves be made for hugetlbfs at the risk of getting killed later. With commit fc8744adc870a8d4366908221508bb113d8b72ee, VM_NORESERVE and VM_ACCOUNT are getting unconditionally set for hugetlbfs-backed mappings. This breaks the accounting for both the core VM and hugetlbfs, can trigger an OOM storm when hugepage pools are too small lockups and corrupted counters otherwise are used. This patch brings hugetlbfs more in line with how the core VM treats VM_NORESERVE but prevents VM_ACCOUNT being set. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-10 15:02:27 +01:00
if (accountable_mapping(file, vm_flags)) {
charged = len >> PAGE_SHIFT;
if (security_vm_enough_memory_mm(mm, charged))
return -ENOMEM;
vm_flags |= VM_ACCOUNT;
}
/*
* Can we just expand an old mapping?
*/
vma = vma_merge(mm, prev, addr, addr + len, vm_flags,
NULL, file, pgoff, NULL, NULL_VM_UFFD_CTX);
if (vma)
goto out;
/*
* Determine the object being mapped and call the appropriate
* specific mapper. the address has already been validated, but
* not unmapped, but the maps are removed from the list.
*/
vma = vm_area_alloc(mm);
if (!vma) {
error = -ENOMEM;
goto unacct_error;
}
vma->vm_start = addr;
vma->vm_end = addr + len;
vma->vm_flags = vm_flags;
vma->vm_page_prot = vm_get_page_prot(vm_flags);
vma->vm_pgoff = pgoff;
if (file) {
if (vm_flags & VM_DENYWRITE) {
error = deny_write_access(file);
if (error)
goto free_vma;
}
if (vm_flags & VM_SHARED) {
error = mapping_map_writable(file->f_mapping);
if (error)
goto allow_write_and_free_vma;
}
/* ->mmap() can change vma->vm_file, but must guarantee that
* vma_link() below can deny write-access if VM_DENYWRITE is set
* and map writably if VM_SHARED is set. This usually means the
* new file must not have been exposed to user-space, yet.
*/
vma->vm_file = get_file(file);
error = call_mmap(file, vma);
if (error)
goto unmap_and_free_vma;
/* Can addr have changed??
*
* Answer: Yes, several device drivers can do it in their
* f_op->mmap method. -DaveM
* Bug: If addr is changed, prev, rb_link, rb_parent should
* be updated for vma_link()
*/
WARN_ON_ONCE(addr != vma->vm_start);
addr = vma->vm_start;
vm_flags = vma->vm_flags;
} else if (vm_flags & VM_SHARED) {
error = shmem_zero_setup(vma);
if (error)
goto free_vma;
mm: fix vma_is_anonymous() false-positives vma_is_anonymous() relies on ->vm_ops being NULL to detect anonymous VMA. This is unreliable as ->mmap may not set ->vm_ops. False-positive vma_is_anonymous() may lead to crashes: next ffff8801ce5e7040 prev ffff8801d20eca50 mm ffff88019c1e13c0 prot 27 anon_vma ffff88019680cdd8 vm_ops 0000000000000000 pgoff 0 file ffff8801b2ec2d00 private_data 0000000000000000 flags: 0xff(read|write|exec|shared|mayread|maywrite|mayexec|mayshare) ------------[ cut here ]------------ kernel BUG at mm/memory.c:1422! invalid opcode: 0000 [#1] SMP KASAN CPU: 0 PID: 18486 Comm: syz-executor3 Not tainted 4.18.0-rc3+ #136 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:zap_pmd_range mm/memory.c:1421 [inline] RIP: 0010:zap_pud_range mm/memory.c:1466 [inline] RIP: 0010:zap_p4d_range mm/memory.c:1487 [inline] RIP: 0010:unmap_page_range+0x1c18/0x2220 mm/memory.c:1508 Call Trace: unmap_single_vma+0x1a0/0x310 mm/memory.c:1553 zap_page_range_single+0x3cc/0x580 mm/memory.c:1644 unmap_mapping_range_vma mm/memory.c:2792 [inline] unmap_mapping_range_tree mm/memory.c:2813 [inline] unmap_mapping_pages+0x3a7/0x5b0 mm/memory.c:2845 unmap_mapping_range+0x48/0x60 mm/memory.c:2880 truncate_pagecache+0x54/0x90 mm/truncate.c:800 truncate_setsize+0x70/0xb0 mm/truncate.c:826 simple_setattr+0xe9/0x110 fs/libfs.c:409 notify_change+0xf13/0x10f0 fs/attr.c:335 do_truncate+0x1ac/0x2b0 fs/open.c:63 do_sys_ftruncate+0x492/0x560 fs/open.c:205 __do_sys_ftruncate fs/open.c:215 [inline] __se_sys_ftruncate fs/open.c:213 [inline] __x64_sys_ftruncate+0x59/0x80 fs/open.c:213 do_syscall_64+0x1b9/0x820 arch/x86/entry/common.c:290 entry_SYSCALL_64_after_hwframe+0x49/0xbe Reproducer: #include <stdio.h> #include <stddef.h> #include <stdint.h> #include <stdlib.h> #include <string.h> #include <sys/types.h> #include <sys/stat.h> #include <sys/ioctl.h> #include <sys/mman.h> #include <unistd.h> #include <fcntl.h> #define KCOV_INIT_TRACE _IOR('c', 1, unsigned long) #define KCOV_ENABLE _IO('c', 100) #define KCOV_DISABLE _IO('c', 101) #define COVER_SIZE (1024<<10) #define KCOV_TRACE_PC 0 #define KCOV_TRACE_CMP 1 int main(int argc, char **argv) { int fd; unsigned long *cover; system("mount -t debugfs none /sys/kernel/debug"); fd = open("/sys/kernel/debug/kcov", O_RDWR); ioctl(fd, KCOV_INIT_TRACE, COVER_SIZE); cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long), PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0); munmap(cover, COVER_SIZE * sizeof(unsigned long)); cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long), PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0); memset(cover, 0, COVER_SIZE * sizeof(unsigned long)); ftruncate(fd, 3UL << 20); return 0; } This can be fixed by assigning anonymous VMAs own vm_ops and not relying on it being NULL. If ->mmap() failed to set ->vm_ops, mmap_region() will set it to dummy_vm_ops. This way we will have non-NULL ->vm_ops for all VMAs. Link: http://lkml.kernel.org/r/20180724121139.62570-4-kirill.shutemov@linux.intel.com Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Reported-by: syzbot+3f84280d52be9b7083cc@syzkaller.appspotmail.com Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-27 01:37:35 +02:00
} else {
vma_set_anonymous(vma);
}
vma_link(mm, vma, prev, rb_link, rb_parent);
/* Once vma denies write, undo our temporary denial count */
if (file) {
if (vm_flags & VM_SHARED)
mapping_unmap_writable(file->f_mapping);
if (vm_flags & VM_DENYWRITE)
allow_write_access(file);
}
file = vma->vm_file;
out:
perf: Do the big rename: Performance Counters -> Performance Events Bye-bye Performance Counters, welcome Performance Events! In the past few months the perfcounters subsystem has grown out its initial role of counting hardware events, and has become (and is becoming) a much broader generic event enumeration, reporting, logging, monitoring, analysis facility. Naming its core object 'perf_counter' and naming the subsystem 'perfcounters' has become more and more of a misnomer. With pending code like hw-breakpoints support the 'counter' name is less and less appropriate. All in one, we've decided to rename the subsystem to 'performance events' and to propagate this rename through all fields, variables and API names. (in an ABI compatible fashion) The word 'event' is also a bit shorter than 'counter' - which makes it slightly more convenient to write/handle as well. Thanks goes to Stephane Eranian who first observed this misnomer and suggested a rename. User-space tooling and ABI compatibility is not affected - this patch should be function-invariant. (Also, defconfigs were not touched to keep the size down.) This patch has been generated via the following script: FILES=$(find * -type f | grep -vE 'oprofile|[^K]config') sed -i \ -e 's/PERF_EVENT_/PERF_RECORD_/g' \ -e 's/PERF_COUNTER/PERF_EVENT/g' \ -e 's/perf_counter/perf_event/g' \ -e 's/nb_counters/nb_events/g' \ -e 's/swcounter/swevent/g' \ -e 's/tpcounter_event/tp_event/g' \ $FILES for N in $(find . -name perf_counter.[ch]); do M=$(echo $N | sed 's/perf_counter/perf_event/g') mv $N $M done FILES=$(find . -name perf_event.*) sed -i \ -e 's/COUNTER_MASK/REG_MASK/g' \ -e 's/COUNTER/EVENT/g' \ -e 's/\<event\>/event_id/g' \ -e 's/counter/event/g' \ -e 's/Counter/Event/g' \ $FILES ... to keep it as correct as possible. This script can also be used by anyone who has pending perfcounters patches - it converts a Linux kernel tree over to the new naming. We tried to time this change to the point in time where the amount of pending patches is the smallest: the end of the merge window. Namespace clashes were fixed up in a preparatory patch - and some stylistic fallout will be fixed up in a subsequent patch. ( NOTE: 'counters' are still the proper terminology when we deal with hardware registers - and these sed scripts are a bit over-eager in renaming them. I've undone some of that, but in case there's something left where 'counter' would be better than 'event' we can undo that on an individual basis instead of touching an otherwise nicely automated patch. ) Suggested-by: Stephane Eranian <eranian@google.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Paul Mackerras <paulus@samba.org> Reviewed-by: Arjan van de Ven <arjan@linux.intel.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Howells <dhowells@redhat.com> Cc: Kyle McMartin <kyle@mcmartin.ca> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: <linux-arch@vger.kernel.org> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 12:02:48 +02:00
perf_event_mmap(vma);
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
vm_stat_account(mm, vm_flags, len >> PAGE_SHIFT);
if (vm_flags & VM_LOCKED) {
dax: remove VM_MIXEDMAP for fsdax and device dax This patch is reworked from an earlier patch that Dan has posted: https://patchwork.kernel.org/patch/10131727/ VM_MIXEDMAP is used by dax to direct mm paths like vm_normal_page() that the memory page it is dealing with is not typical memory from the linear map. The get_user_pages_fast() path, since it does not resolve the vma, is already using {pte,pmd}_devmap() as a stand-in for VM_MIXEDMAP, so we use that as a VM_MIXEDMAP replacement in some locations. In the cases where there is no pte to consult we fallback to using vma_is_dax() to detect the VM_MIXEDMAP special case. Now that we have explicit driver pfn_t-flag opt-in/opt-out for get_user_pages() support for DAX we can stop setting VM_MIXEDMAP. This also means we no longer need to worry about safely manipulating vm_flags in a future where we support dynamically changing the dax mode of a file. DAX should also now be supported with madvise_behavior(), vma_merge(), and copy_page_range(). This patch has been tested against ndctl unit test. It has also been tested against xfstests commit: 625515d using fake pmem created by memmap and no additional issues have been observed. Link: http://lkml.kernel.org/r/152847720311.55924.16999195879201817653.stgit@djiang5-desk3.ch.intel.com Signed-off-by: Dave Jiang <dave.jiang@intel.com> Acked-by: Dan Williams <dan.j.williams@intel.com> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-18 00:43:40 +02:00
if ((vm_flags & VM_SPECIAL) || vma_is_dax(vma) ||
is_vm_hugetlb_page(vma) ||
vma == get_gate_vma(current->mm))
mm: introduce VM_LOCKONFAULT The cost of faulting in all memory to be locked can be very high when working with large mappings. If only portions of the mapping will be used this can incur a high penalty for locking. For the example of a large file, this is the usage pattern for a large statical language model (probably applies to other statical or graphical models as well). For the security example, any application transacting in data that cannot be swapped out (credit card data, medical records, etc). This patch introduces the ability to request that pages are not pre-faulted, but are placed on the unevictable LRU when they are finally faulted in. The VM_LOCKONFAULT flag will be used together with VM_LOCKED and has no effect when set without VM_LOCKED. Setting the VM_LOCKONFAULT flag for a VMA will cause pages faulted into that VMA to be added to the unevictable LRU when they are faulted or if they are already present, but will not cause any missing pages to be faulted in. Exposing this new lock state means that we cannot overload the meaning of the FOLL_POPULATE flag any longer. Prior to this patch it was used to mean that the VMA for a fault was locked. This means we need the new FOLL_MLOCK flag to communicate the locked state of a VMA. FOLL_POPULATE will now only control if the VMA should be populated and in the case of VM_LOCKONFAULT, it will not be set. Signed-off-by: Eric B Munson <emunson@akamai.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Michal Hocko <mhocko@suse.cz> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Shuah Khan <shuahkh@osg.samsung.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-06 03:51:36 +01:00
vma->vm_flags &= VM_LOCKED_CLEAR_MASK;
dax: remove VM_MIXEDMAP for fsdax and device dax This patch is reworked from an earlier patch that Dan has posted: https://patchwork.kernel.org/patch/10131727/ VM_MIXEDMAP is used by dax to direct mm paths like vm_normal_page() that the memory page it is dealing with is not typical memory from the linear map. The get_user_pages_fast() path, since it does not resolve the vma, is already using {pte,pmd}_devmap() as a stand-in for VM_MIXEDMAP, so we use that as a VM_MIXEDMAP replacement in some locations. In the cases where there is no pte to consult we fallback to using vma_is_dax() to detect the VM_MIXEDMAP special case. Now that we have explicit driver pfn_t-flag opt-in/opt-out for get_user_pages() support for DAX we can stop setting VM_MIXEDMAP. This also means we no longer need to worry about safely manipulating vm_flags in a future where we support dynamically changing the dax mode of a file. DAX should also now be supported with madvise_behavior(), vma_merge(), and copy_page_range(). This patch has been tested against ndctl unit test. It has also been tested against xfstests commit: 625515d using fake pmem created by memmap and no additional issues have been observed. Link: http://lkml.kernel.org/r/152847720311.55924.16999195879201817653.stgit@djiang5-desk3.ch.intel.com Signed-off-by: Dave Jiang <dave.jiang@intel.com> Acked-by: Dan Williams <dan.j.williams@intel.com> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-18 00:43:40 +02:00
else
mm->locked_vm += (len >> PAGE_SHIFT);
}
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
uprobes: Fix mmap_region()'s mm->mm_rb corruption if uprobe_mmap() fails This patch fixes: https://bugzilla.redhat.com/show_bug.cgi?id=843640 If mmap_region()->uprobe_mmap() fails, unmap_and_free_vma path does unmap_region() but does not remove the soon-to-be-freed vma from rb tree. Actually there are more problems but this is how William noticed this bug. Perhaps we could do do_munmap() + return in this case, but in fact it is simply wrong to abort if uprobe_mmap() fails. Until at least we move the !UPROBE_COPY_INSN code from install_breakpoint() to uprobe_register(). For example, uprobe_mmap()->install_breakpoint() can fail if the probed insn is not supported (remember, uprobe_register() succeeds if nobody mmaps inode/offset), mmap() should not fail in this case. dup_mmap()->uprobe_mmap() is wrong too by the same reason, fork() can race with uprobe_register() and fail for no reason if it wins the race and does install_breakpoint() first. And, if nothing else, both mmap_region() and dup_mmap() return success if uprobe_mmap() fails. Change them to ignore the error code from uprobe_mmap(). Reported-and-tested-by: William Cohen <wcohen@redhat.com> Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: <stable@vger.kernel.org> # v3.5 Cc: Anton Arapov <anton@redhat.com> Cc: William Cohen <wcohen@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: http://lkml.kernel.org/r/20120819171042.GB26957@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2012-08-19 19:10:42 +02:00
if (file)
uprobe_mmap(vma);
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
/*
* New (or expanded) vma always get soft dirty status.
* Otherwise user-space soft-dirty page tracker won't
* be able to distinguish situation when vma area unmapped,
* then new mapped in-place (which must be aimed as
* a completely new data area).
*/
vma->vm_flags |= VM_SOFTDIRTY;
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared For VMAs that don't want write notifications, PTEs created for read faults have their write bit set. If the read fault happens after VM_SOFTDIRTY is cleared, then the PTE's softdirty bit will remain clear after subsequent writes. Here's a simple code snippet to demonstrate the bug: char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_SHARED, -1, 0); system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */ assert(*m == '\0'); /* new PTE allows write access */ assert(!soft_dirty(x)); *m = 'x'; /* should dirty the page */ assert(soft_dirty(x)); /* fails */ With this patch, write notifications are enabled when VM_SOFTDIRTY is cleared. Furthermore, to avoid unnecessary faults, write notifications are disabled when VM_SOFTDIRTY is set. As a side effect of enabling and disabling write notifications with care, this patch fixes a bug in mprotect where vm_page_prot bits set by drivers were zapped on mprotect. An analogous bug was fixed in mmap by commit c9d0bf241451 ("mm: uncached vma support with writenotify"). Signed-off-by: Peter Feiner <pfeiner@google.com> Reported-by: Peter Feiner <pfeiner@google.com> Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Jamie Liu <jamieliu@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-14 00:55:46 +02:00
vma_set_page_prot(vma);
return addr;
unmap_and_free_vma:
vma->vm_file = NULL;
fput(file);
/* Undo any partial mapping done by a device driver. */
[PATCH] freepgt: free_pgtables use vma list Recent woes with some arches needing their own pgd_addr_end macro; and 4-level clear_page_range regression since 2.6.10's clear_page_tables; and its long-standing well-known inefficiency in searching throughout the higher-level page tables for those few entries to clear and free: all can be blamed on ignoring the list of vmas when we free page tables. Replace exit_mmap's clear_page_range of the total user address space by free_pgtables operating on the mm's vma list; unmap_region use it in the same way, giving floor and ceiling beyond which it may not free tables. This brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled, in which case latency fixes spoil unmap_vmas throughput). Beware: the do_mmap_pgoff driver failure case must now use unmap_region instead of zap_page_range, since a page table might have been allocated, and can only be freed while it is touched by some vma. Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted from the clear_page_range levels. (Most of free_pgtables' old code was actually for a non-existent case, prev not properly set up, dating from before hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we might want to add latency lockdrops later; but no attempt to do so yet, going by vma should itself reduce latency. But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful examination: put that off until a later patch of the series. What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma? And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that we need to do more than is done here - every PMD_SIZE ever occupied will be flushed, do we really have to flush every PGDIR_SIZE ever partially occupied? A shame to complicate it unnecessarily. Special thanks to David Miller for time spent repairing my ceilings. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 22:29:15 +02:00
unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
charged = 0;
if (vm_flags & VM_SHARED)
mapping_unmap_writable(file->f_mapping);
allow_write_and_free_vma:
if (vm_flags & VM_DENYWRITE)
allow_write_access(file);
free_vma:
vm_area_free(vma);
unacct_error:
if (charged)
vm_unacct_memory(charged);
return error;
}
unsigned long unmapped_area(struct vm_unmapped_area_info *info)
{
/*
* We implement the search by looking for an rbtree node that
* immediately follows a suitable gap. That is,
* - gap_start = vma->vm_prev->vm_end <= info->high_limit - length;
* - gap_end = vma->vm_start >= info->low_limit + length;
* - gap_end - gap_start >= length
*/
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
unsigned long length, low_limit, high_limit, gap_start, gap_end;
/* Adjust search length to account for worst case alignment overhead */
length = info->length + info->align_mask;
if (length < info->length)
return -ENOMEM;
/* Adjust search limits by the desired length */
if (info->high_limit < length)
return -ENOMEM;
high_limit = info->high_limit - length;
if (info->low_limit > high_limit)
return -ENOMEM;
low_limit = info->low_limit + length;
/* Check if rbtree root looks promising */
if (RB_EMPTY_ROOT(&mm->mm_rb))
goto check_highest;
vma = rb_entry(mm->mm_rb.rb_node, struct vm_area_struct, vm_rb);
if (vma->rb_subtree_gap < length)
goto check_highest;
while (true) {
/* Visit left subtree if it looks promising */
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
gap_end = vm_start_gap(vma);
if (gap_end >= low_limit && vma->vm_rb.rb_left) {
struct vm_area_struct *left =
rb_entry(vma->vm_rb.rb_left,
struct vm_area_struct, vm_rb);
if (left->rb_subtree_gap >= length) {
vma = left;
continue;
}
}
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
gap_start = vma->vm_prev ? vm_end_gap(vma->vm_prev) : 0;
check_current:
/* Check if current node has a suitable gap */
if (gap_start > high_limit)
return -ENOMEM;
if (gap_end >= low_limit &&
gap_end > gap_start && gap_end - gap_start >= length)
goto found;
/* Visit right subtree if it looks promising */
if (vma->vm_rb.rb_right) {
struct vm_area_struct *right =
rb_entry(vma->vm_rb.rb_right,
struct vm_area_struct, vm_rb);
if (right->rb_subtree_gap >= length) {
vma = right;
continue;
}
}
/* Go back up the rbtree to find next candidate node */
while (true) {
struct rb_node *prev = &vma->vm_rb;
if (!rb_parent(prev))
goto check_highest;
vma = rb_entry(rb_parent(prev),
struct vm_area_struct, vm_rb);
if (prev == vma->vm_rb.rb_left) {
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
gap_start = vm_end_gap(vma->vm_prev);
gap_end = vm_start_gap(vma);
goto check_current;
}
}
}
check_highest:
/* Check highest gap, which does not precede any rbtree node */
gap_start = mm->highest_vm_end;
gap_end = ULONG_MAX; /* Only for VM_BUG_ON below */
if (gap_start > high_limit)
return -ENOMEM;
found:
/* We found a suitable gap. Clip it with the original low_limit. */
if (gap_start < info->low_limit)
gap_start = info->low_limit;
/* Adjust gap address to the desired alignment */
gap_start += (info->align_offset - gap_start) & info->align_mask;
VM_BUG_ON(gap_start + info->length > info->high_limit);
VM_BUG_ON(gap_start + info->length > gap_end);
return gap_start;
}
unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info)
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
unsigned long length, low_limit, high_limit, gap_start, gap_end;
/* Adjust search length to account for worst case alignment overhead */
length = info->length + info->align_mask;
if (length < info->length)
return -ENOMEM;
/*
* Adjust search limits by the desired length.
* See implementation comment at top of unmapped_area().
*/
gap_end = info->high_limit;
if (gap_end < length)
return -ENOMEM;
high_limit = gap_end - length;
if (info->low_limit > high_limit)
return -ENOMEM;
low_limit = info->low_limit + length;
/* Check highest gap, which does not precede any rbtree node */
gap_start = mm->highest_vm_end;
if (gap_start <= high_limit)
goto found_highest;
/* Check if rbtree root looks promising */
if (RB_EMPTY_ROOT(&mm->mm_rb))
return -ENOMEM;
vma = rb_entry(mm->mm_rb.rb_node, struct vm_area_struct, vm_rb);
if (vma->rb_subtree_gap < length)
return -ENOMEM;
while (true) {
/* Visit right subtree if it looks promising */
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
gap_start = vma->vm_prev ? vm_end_gap(vma->vm_prev) : 0;
if (gap_start <= high_limit && vma->vm_rb.rb_right) {
struct vm_area_struct *right =
rb_entry(vma->vm_rb.rb_right,
struct vm_area_struct, vm_rb);
if (right->rb_subtree_gap >= length) {
vma = right;
continue;
}
}
check_current:
/* Check if current node has a suitable gap */
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
gap_end = vm_start_gap(vma);
if (gap_end < low_limit)
return -ENOMEM;
if (gap_start <= high_limit &&
gap_end > gap_start && gap_end - gap_start >= length)
goto found;
/* Visit left subtree if it looks promising */
if (vma->vm_rb.rb_left) {
struct vm_area_struct *left =
rb_entry(vma->vm_rb.rb_left,
struct vm_area_struct, vm_rb);
if (left->rb_subtree_gap >= length) {
vma = left;
continue;
}
}
/* Go back up the rbtree to find next candidate node */
while (true) {
struct rb_node *prev = &vma->vm_rb;
if (!rb_parent(prev))
return -ENOMEM;
vma = rb_entry(rb_parent(prev),
struct vm_area_struct, vm_rb);
if (prev == vma->vm_rb.rb_right) {
gap_start = vma->vm_prev ?
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
vm_end_gap(vma->vm_prev) : 0;
goto check_current;
}
}
}
found:
/* We found a suitable gap. Clip it with the original high_limit. */
if (gap_end > info->high_limit)
gap_end = info->high_limit;
found_highest:
/* Compute highest gap address at the desired alignment */
gap_end -= info->length;
gap_end -= (gap_end - info->align_offset) & info->align_mask;
VM_BUG_ON(gap_end < info->low_limit);
VM_BUG_ON(gap_end < gap_start);
return gap_end;
}
#ifndef arch_get_mmap_end
#define arch_get_mmap_end(addr) (TASK_SIZE)
#endif
#ifndef arch_get_mmap_base
#define arch_get_mmap_base(addr, base) (base)
#endif
/* Get an address range which is currently unmapped.
* For shmat() with addr=0.
*
* Ugly calling convention alert:
* Return value with the low bits set means error value,
* ie
* if (ret & ~PAGE_MASK)
* error = ret;
*
* This function "knows" that -ENOMEM has the bits set.
*/
#ifndef HAVE_ARCH_UNMAPPED_AREA
unsigned long
arch_get_unmapped_area(struct file *filp, unsigned long addr,
unsigned long len, unsigned long pgoff, unsigned long flags)
{
struct mm_struct *mm = current->mm;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
struct vm_area_struct *vma, *prev;
struct vm_unmapped_area_info info;
const unsigned long mmap_end = arch_get_mmap_end(addr);
if (len > mmap_end - mmap_min_addr)
return -ENOMEM;
if (flags & MAP_FIXED)
return addr;
if (addr) {
addr = PAGE_ALIGN(addr);
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
vma = find_vma_prev(mm, addr, &prev);
if (mmap_end - len >= addr && addr >= mmap_min_addr &&
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
(!vma || addr + len <= vm_start_gap(vma)) &&
(!prev || addr >= vm_end_gap(prev)))
return addr;
}
info.flags = 0;
info.length = len;
info.low_limit = mm->mmap_base;
info.high_limit = mmap_end;
info.align_mask = 0;
return vm_unmapped_area(&info);
}
#endif
/*
* This mmap-allocator allocates new areas top-down from below the
* stack's low limit (the base):
*/
#ifndef HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
unsigned long
arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
unsigned long len, unsigned long pgoff,
unsigned long flags)
{
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
struct vm_area_struct *vma, *prev;
struct mm_struct *mm = current->mm;
struct vm_unmapped_area_info info;
const unsigned long mmap_end = arch_get_mmap_end(addr);
/* requested length too big for entire address space */
if (len > mmap_end - mmap_min_addr)
return -ENOMEM;
if (flags & MAP_FIXED)
return addr;
/* requesting a specific address */
if (addr) {
addr = PAGE_ALIGN(addr);
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
vma = find_vma_prev(mm, addr, &prev);
if (mmap_end - len >= addr && addr >= mmap_min_addr &&
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
(!vma || addr + len <= vm_start_gap(vma)) &&
(!prev || addr >= vm_end_gap(prev)))
return addr;
}
info.flags = VM_UNMAPPED_AREA_TOPDOWN;
info.length = len;
mm: ensure get_unmapped_area() returns higher address than mmap_min_addr This patch fixes the problem that get_unmapped_area() can return illegal address and result in failing mmap(2) etc. In case that the address higher than PAGE_SIZE is set to /proc/sys/vm/mmap_min_addr, the address lower than mmap_min_addr can be returned by get_unmapped_area(), even if you do not pass any virtual address hint (i.e. the second argument). This is because the current get_unmapped_area() code does not take into account mmap_min_addr. This leads to two actual problems as follows: 1. mmap(2) can fail with EPERM on the process without CAP_SYS_RAWIO, although any illegal parameter is not passed. 2. The bottom-up search path after the top-down search might not work in arch_get_unmapped_area_topdown(). Note: The first and third chunk of my patch, which changes "len" check, are for more precise check using mmap_min_addr, and not for solving the above problem. [How to reproduce] --- test.c ------------------------------------------------- #include <stdio.h> #include <unistd.h> #include <sys/mman.h> #include <sys/errno.h> int main(int argc, char *argv[]) { void *ret = NULL, *last_map; size_t pagesize = sysconf(_SC_PAGESIZE); do { last_map = ret; ret = mmap(0, pagesize, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); // printf("ret=%p\n", ret); } while (ret != MAP_FAILED); if (errno != ENOMEM) { printf("ERR: unexpected errno: %d (last map=%p)\n", errno, last_map); } return 0; } --------------------------------------------------------------- $ gcc -m32 -o test test.c $ sudo sysctl -w vm.mmap_min_addr=65536 vm.mmap_min_addr = 65536 $ ./test (run as non-priviledge user) ERR: unexpected errno: 1 (last map=0x10000) Signed-off-by: Akira Takeuchi <takeuchi.akr@jp.panasonic.com> Signed-off-by: Kiyoshi Owada <owada.kiyoshi@jp.panasonic.com> Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-11-13 00:08:21 +01:00
info.low_limit = max(PAGE_SIZE, mmap_min_addr);
info.high_limit = arch_get_mmap_base(addr, mm->mmap_base);
info.align_mask = 0;
addr = vm_unmapped_area(&info);
/*
* A failed mmap() very likely causes application failure,
* so fall back to the bottom-up function here. This scenario
* can happen with large stack limits and large mmap()
* allocations.
*/
if (offset_in_page(addr)) {
VM_BUG_ON(addr != -ENOMEM);
info.flags = 0;
info.low_limit = TASK_UNMAPPED_BASE;
info.high_limit = mmap_end;
addr = vm_unmapped_area(&info);
}
return addr;
}
#endif
unsigned long
get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
unsigned long pgoff, unsigned long flags)
{
unsigned long (*get_area)(struct file *, unsigned long,
unsigned long, unsigned long, unsigned long);
unsigned long error = arch_mmap_check(addr, len, flags);
if (error)
return error;
/* Careful about overflows.. */
if (len > TASK_SIZE)
return -ENOMEM;
get_area = current->mm->get_unmapped_area;
if (file) {
if (file->f_op->get_unmapped_area)
get_area = file->f_op->get_unmapped_area;
} else if (flags & MAP_SHARED) {
/*
* mmap_region() will call shmem_zero_setup() to create a file,
* so use shmem's get_unmapped_area in case it can be huge.
* do_mmap_pgoff() will clear pgoff, so match alignment.
*/
pgoff = 0;
get_area = shmem_get_unmapped_area;
}
addr = get_area(file, addr, len, pgoff, flags);
if (IS_ERR_VALUE(addr))
return addr;
if (addr > TASK_SIZE - len)
return -ENOMEM;
if (offset_in_page(addr))
return -EINVAL;
error = security_mmap_addr(addr);
return error ? error : addr;
}
EXPORT_SYMBOL(get_unmapped_area);
/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
{
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
struct rb_node *rb_node;
struct vm_area_struct *vma;
/* Check the cache first. */
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
vma = vmacache_find(mm, addr);
if (likely(vma))
return vma;
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
rb_node = mm->mm_rb.rb_node;
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
while (rb_node) {
struct vm_area_struct *tmp;
tmp = rb_entry(rb_node, struct vm_area_struct, vm_rb);
if (tmp->vm_end > addr) {
vma = tmp;
if (tmp->vm_start <= addr)
break;
rb_node = rb_node->rb_left;
} else
rb_node = rb_node->rb_right;
}
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
if (vma)
vmacache_update(addr, vma);
return vma;
}
EXPORT_SYMBOL(find_vma);
/*
* Same as find_vma, but also return a pointer to the previous VMA in *pprev.
*/
struct vm_area_struct *
find_vma_prev(struct mm_struct *mm, unsigned long addr,
struct vm_area_struct **pprev)
{
struct vm_area_struct *vma;
vma = find_vma(mm, addr);
if (vma) {
*pprev = vma->vm_prev;
} else {
struct rb_node *rb_node = rb_last(&mm->mm_rb);
*pprev = rb_node ? rb_entry(rb_node, struct vm_area_struct, vm_rb) : NULL;
}
return vma;
}
/*
* Verify that the stack growth is acceptable and
* update accounting. This is shared with both the
* grow-up and grow-down cases.
*/
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
static int acct_stack_growth(struct vm_area_struct *vma,
unsigned long size, unsigned long grow)
{
struct mm_struct *mm = vma->vm_mm;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
unsigned long new_start;
/* address space limit tests */
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
if (!may_expand_vm(mm, vma->vm_flags, grow))
return -ENOMEM;
/* Stack limit test */
if (size > rlimit(RLIMIT_STACK))
return -ENOMEM;
/* mlock limit tests */
if (vma->vm_flags & VM_LOCKED) {
unsigned long locked;
unsigned long limit;
locked = mm->locked_vm + grow;
limit = rlimit(RLIMIT_MEMLOCK);
limit >>= PAGE_SHIFT;
if (locked > limit && !capable(CAP_IPC_LOCK))
return -ENOMEM;
}
/* Check to ensure the stack will not grow into a hugetlb-only region */
new_start = (vma->vm_flags & VM_GROWSUP) ? vma->vm_start :
vma->vm_end - size;
if (is_hugepage_only_range(vma->vm_mm, new_start, size))
return -EFAULT;
/*
* Overcommit.. This must be the final test, as it will
* update security statistics.
*/
if (security_vm_enough_memory_mm(mm, grow))
return -ENOMEM;
return 0;
}
#if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
/*
* PA-RISC uses this for its stack; IA64 for its Register Backing Store.
* vma is the last one with address > vma->vm_end. Have to extend vma.
*/
int expand_upwards(struct vm_area_struct *vma, unsigned long address)
{
struct mm_struct *mm = vma->vm_mm;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
struct vm_area_struct *next;
unsigned long gap_addr;
int error = 0;
if (!(vma->vm_flags & VM_GROWSUP))
return -EFAULT;
/* Guard against exceeding limits of the address space. */
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
address &= PAGE_MASK;
if (address >= (TASK_SIZE & PAGE_MASK))
return -ENOMEM;
address += PAGE_SIZE;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
/* Enforce stack_guard_gap */
gap_addr = address + stack_guard_gap;
/* Guard against overflow */
if (gap_addr < address || gap_addr > TASK_SIZE)
gap_addr = TASK_SIZE;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
next = vma->vm_next;
if (next && next->vm_start < gap_addr &&
(next->vm_flags & (VM_WRITE|VM_READ|VM_EXEC))) {
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
if (!(next->vm_flags & VM_GROWSUP))
return -ENOMEM;
/* Check that both stack segments have the same anon_vma? */
}
/* We must make sure the anon_vma is allocated. */
if (unlikely(anon_vma_prepare(vma)))
return -ENOMEM;
/*
* vma->vm_start/vm_end cannot change under us because the caller
* is required to hold the mmap_sem in read mode. We need the
* anon_vma lock to serialize against concurrent expand_stacks.
*/
anon_vma_lock_write(vma->anon_vma);
/* Somebody else might have raced and expanded it already */
if (address > vma->vm_end) {
unsigned long size, grow;
size = address - vma->vm_start;
grow = (address - vma->vm_end) >> PAGE_SHIFT;
error = -ENOMEM;
if (vma->vm_pgoff + (size >> PAGE_SHIFT) >= vma->vm_pgoff) {
error = acct_stack_growth(vma, size, grow);
if (!error) {
/*
* vma_gap_update() doesn't support concurrent
* updates, but we only hold a shared mmap_sem
* lock here, so we need to protect against
* concurrent vma expansions.
* anon_vma_lock_write() doesn't help here, as
* we don't guarantee that all growable vmas
* in a mm share the same root anon vma.
* So, we reuse mm->page_table_lock to guard
* against concurrent vma expansions.
*/
spin_lock(&mm->page_table_lock);
if (vma->vm_flags & VM_LOCKED)
mm->locked_vm += grow;
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
vm_stat_account(mm, vma->vm_flags, grow);
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma_interval_tree_pre_update_vma(vma);
vma->vm_end = address;
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma_interval_tree_post_update_vma(vma);
if (vma->vm_next)
vma_gap_update(vma->vm_next);
else
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
mm->highest_vm_end = vm_end_gap(vma);
spin_unlock(&mm->page_table_lock);
perf_event_mmap(vma);
}
}
}
anon_vma_unlock_write(vma->anon_vma);
khugepaged_enter_vma_merge(vma, vma->vm_flags);
validate_mm(mm);
return error;
}
#endif /* CONFIG_STACK_GROWSUP || CONFIG_IA64 */
/*
* vma is the first one with address < vma->vm_start. Have to extend vma.
*/
int expand_downwards(struct vm_area_struct *vma,
unsigned long address)
{
struct mm_struct *mm = vma->vm_mm;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
struct vm_area_struct *prev;
int error = 0;
address &= PAGE_MASK;
if (address < mmap_min_addr)
return -EPERM;
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
/* Enforce stack_guard_gap */
prev = vma->vm_prev;
/* Check that both stack segments have the same anon_vma? */
if (prev && !(prev->vm_flags & VM_GROWSDOWN) &&
(prev->vm_flags & (VM_WRITE|VM_READ|VM_EXEC))) {
if (address - prev->vm_end < stack_guard_gap)
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
return -ENOMEM;
}
/* We must make sure the anon_vma is allocated. */
if (unlikely(anon_vma_prepare(vma)))
return -ENOMEM;
/*
* vma->vm_start/vm_end cannot change under us because the caller
* is required to hold the mmap_sem in read mode. We need the
* anon_vma lock to serialize against concurrent expand_stacks.
*/
anon_vma_lock_write(vma->anon_vma);
/* Somebody else might have raced and expanded it already */
if (address < vma->vm_start) {
unsigned long size, grow;
size = vma->vm_end - address;
grow = (vma->vm_start - address) >> PAGE_SHIFT;
error = -ENOMEM;
if (grow <= vma->vm_pgoff) {
error = acct_stack_growth(vma, size, grow);
if (!error) {
/*
* vma_gap_update() doesn't support concurrent
* updates, but we only hold a shared mmap_sem
* lock here, so we need to protect against
* concurrent vma expansions.
* anon_vma_lock_write() doesn't help here, as
* we don't guarantee that all growable vmas
* in a mm share the same root anon vma.
* So, we reuse mm->page_table_lock to guard
* against concurrent vma expansions.
*/
spin_lock(&mm->page_table_lock);
if (vma->vm_flags & VM_LOCKED)
mm->locked_vm += grow;
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
vm_stat_account(mm, vma->vm_flags, grow);
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma_interval_tree_pre_update_vma(vma);
vma->vm_start = address;
vma->vm_pgoff -= grow;
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
anon_vma_interval_tree_post_update_vma(vma);
vma_gap_update(vma);
spin_unlock(&mm->page_table_lock);
perf_event_mmap(vma);
}
}
}
anon_vma_unlock_write(vma->anon_vma);
khugepaged_enter_vma_merge(vma, vma->vm_flags);
validate_mm(mm);
return error;
}
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
/* enforced gap between the expanding stack and other mappings. */
unsigned long stack_guard_gap = 256UL<<PAGE_SHIFT;
static int __init cmdline_parse_stack_guard_gap(char *p)
{
unsigned long val;
char *endptr;
val = simple_strtoul(p, &endptr, 10);
if (!*endptr)
stack_guard_gap = val << PAGE_SHIFT;
return 0;
}
__setup("stack_guard_gap=", cmdline_parse_stack_guard_gap);
#ifdef CONFIG_STACK_GROWSUP
int expand_stack(struct vm_area_struct *vma, unsigned long address)
{
return expand_upwards(vma, address);
}
struct vm_area_struct *
find_extend_vma(struct mm_struct *mm, unsigned long addr)
{
struct vm_area_struct *vma, *prev;
addr &= PAGE_MASK;
vma = find_vma_prev(mm, addr, &prev);
if (vma && (vma->vm_start <= addr))
return vma;
coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017: https://lkml.kernel.org/r/alpine.LSU.2.11.1707191716030.2055@eggly.anvils "Not strictly relevant here, but a related note: I was very surprised to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct" In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently. Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code. Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix. For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped. Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case. In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit. Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen. Link: http://lkml.kernel.org/r/20190325224949.11068-1-aarcange@redhat.com Fixes: 86039bd3b4e6 ("userfaultfd: add new syscall to provide memory externalization") Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Jann Horn <jannh@google.com> Suggested-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Peter Xu <peterx@redhat.com> Reviewed-by: Mike Rapoport <rppt@linux.ibm.com> Reviewed-by: Oleg Nesterov <oleg@redhat.com> Reviewed-by: Jann Horn <jannh@google.com> Acked-by: Jason Gunthorpe <jgg@mellanox.com> Acked-by: Michal Hocko <mhocko@suse.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-04-19 02:50:52 +02:00
/* don't alter vm_end if the coredump is running */
if (!prev || !mmget_still_valid(mm) || expand_stack(prev, addr))
return NULL;
if (prev->vm_flags & VM_LOCKED)
populate_vma_page_range(prev, addr, prev->vm_end, NULL);
return prev;
}
#else
int expand_stack(struct vm_area_struct *vma, unsigned long address)
{
return expand_downwards(vma, address);
}
struct vm_area_struct *
find_extend_vma(struct mm_struct *mm, unsigned long addr)
{
struct vm_area_struct *vma;
unsigned long start;
addr &= PAGE_MASK;
vma = find_vma(mm, addr);
if (!vma)
return NULL;
if (vma->vm_start <= addr)
return vma;
if (!(vma->vm_flags & VM_GROWSDOWN))
return NULL;
coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017: https://lkml.kernel.org/r/alpine.LSU.2.11.1707191716030.2055@eggly.anvils "Not strictly relevant here, but a related note: I was very surprised to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct" In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently. Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code. Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix. For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped. Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case. In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit. Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen. Link: http://lkml.kernel.org/r/20190325224949.11068-1-aarcange@redhat.com Fixes: 86039bd3b4e6 ("userfaultfd: add new syscall to provide memory externalization") Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Jann Horn <jannh@google.com> Suggested-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Peter Xu <peterx@redhat.com> Reviewed-by: Mike Rapoport <rppt@linux.ibm.com> Reviewed-by: Oleg Nesterov <oleg@redhat.com> Reviewed-by: Jann Horn <jannh@google.com> Acked-by: Jason Gunthorpe <jgg@mellanox.com> Acked-by: Michal Hocko <mhocko@suse.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-04-19 02:50:52 +02:00
/* don't alter vm_start if the coredump is running */
if (!mmget_still_valid(mm))
return NULL;
start = vma->vm_start;
if (expand_stack(vma, addr))
return NULL;
if (vma->vm_flags & VM_LOCKED)
populate_vma_page_range(vma, addr, start, NULL);
return vma;
}
#endif
EXPORT_SYMBOL_GPL(find_extend_vma);
/*
* Ok - we have the memory areas we should free on the vma list,
* so release them, and do the vma updates.
*
* Called with the mm semaphore held.
*/
static void remove_vma_list(struct mm_struct *mm, struct vm_area_struct *vma)
{
unsigned long nr_accounted = 0;
[PATCH] mm: update_hiwaters just in time update_mem_hiwater has attracted various criticisms, in particular from those concerned with mm scalability. Originally it was called whenever rss or total_vm got raised. Then many of those callsites were replaced by a timer tick call from account_system_time. Now Frank van Maarseveen reports that to be found inadequate. How about this? Works for Frank. Replace update_mem_hiwater, a poor combination of two unrelated ops, by macros update_hiwater_rss and update_hiwater_vm. Don't attempt to keep mm->hiwater_rss up to date at timer tick, nor every time we raise rss (usually by 1): those are hot paths. Do the opposite, update only when about to lower rss (usually by many), or just before final accounting in do_exit. Handle mm->hiwater_vm in the same way, though it's much less of an issue. Demand that whoever collects these hiwater statistics do the work of taking the maximum with rss or total_vm. And there has been no collector of these hiwater statistics in the tree. The new convention needs an example, so match Frank's usage by adding a VmPeak line above VmSize to /proc/<pid>/status, and also a VmHWM line above VmRSS (High-Water-Mark or High-Water-Memory). There was a particular anomaly during mremap move, that hiwater_vm might be captured too high. A fleeting such anomaly remains, but it's quickly corrected now, whereas before it would stick. What locking? None: if the app is racy then these statistics will be racy, it's not worth any overhead to make them exact. But whenever it suits, hiwater_vm is updated under exclusive mmap_sem, and hiwater_rss under page_table_lock (for now) or with preemption disabled (later on): without going to any trouble, minimize the time between reading current values and updating, to minimize those occasions when a racing thread bumps a count up and back down in between. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 02:16:18 +01:00
/* Update high watermark before we lower total_vm */
update_hiwater_vm(mm);
do {
long nrpages = vma_pages(vma);
if (vma->vm_flags & VM_ACCOUNT)
nr_accounted += nrpages;
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
vm_stat_account(mm, vma->vm_flags, -nrpages);
vma = remove_vma(vma);
} while (vma);
vm_unacct_memory(nr_accounted);
validate_mm(mm);
}
/*
* Get rid of page table information in the indicated region.
*
* Called with the mm semaphore held.
*/
static void unmap_region(struct mm_struct *mm,
[PATCH] freepgt: free_pgtables use vma list Recent woes with some arches needing their own pgd_addr_end macro; and 4-level clear_page_range regression since 2.6.10's clear_page_tables; and its long-standing well-known inefficiency in searching throughout the higher-level page tables for those few entries to clear and free: all can be blamed on ignoring the list of vmas when we free page tables. Replace exit_mmap's clear_page_range of the total user address space by free_pgtables operating on the mm's vma list; unmap_region use it in the same way, giving floor and ceiling beyond which it may not free tables. This brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled, in which case latency fixes spoil unmap_vmas throughput). Beware: the do_mmap_pgoff driver failure case must now use unmap_region instead of zap_page_range, since a page table might have been allocated, and can only be freed while it is touched by some vma. Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted from the clear_page_range levels. (Most of free_pgtables' old code was actually for a non-existent case, prev not properly set up, dating from before hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we might want to add latency lockdrops later; but no attempt to do so yet, going by vma should itself reduce latency. But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful examination: put that off until a later patch of the series. What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma? And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that we need to do more than is done here - every PMD_SIZE ever occupied will be flushed, do we really have to flush every PGDIR_SIZE ever partially occupied? A shame to complicate it unnecessarily. Special thanks to David Miller for time spent repairing my ceilings. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 22:29:15 +02:00
struct vm_area_struct *vma, struct vm_area_struct *prev,
unsigned long start, unsigned long end)
{
struct vm_area_struct *next = prev ? prev->vm_next : mm->mmap;
mm: mmu_gather rework Rework the existing mmu_gather infrastructure. The direct purpose of these patches was to allow preemptible mmu_gather, but even without that I think these patches provide an improvement to the status quo. The first 9 patches rework the mmu_gather infrastructure. For review purpose I've split them into generic and per-arch patches with the last of those a generic cleanup. The next patch provides generic RCU page-table freeing, and the followup is a patch converting s390 to use this. I've also got 4 patches from DaveM lined up (not included in this series) that uses this to implement gup_fast() for sparc64. Then there is one patch that extends the generic mmu_gather batching. After that follow the mm preemptibility patches, these make part of the mm a lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes which together with the mmu_gather rework makes mmu_gather preemptible as well. Making i_mmap_lock a mutex also enables a clean-up of the truncate code. This also allows for preemptible mmu_notifiers, something that XPMEM I think wants. Furthermore, it removes the new and universially detested unmap_mutex. This patch: Remove the first obstacle towards a fully preemptible mmu_gather. The current scheme assumes mmu_gather is always done with preemption disabled and uses per-cpu storage for the page batches. Change this to try and allocate a page for batching and in case of failure, use a small on-stack array to make some progress. Preemptible mmu_gather is desired in general and usable once i_mmap_lock becomes a mutex. Doing it before the mutex conversion saves us from having to rework the code by moving the mmu_gather bits inside the pte_lock. Also avoid flushing the tlb batches from under the pte lock, this is useful even without the i_mmap_lock conversion as it significantly reduces pte lock hold times. [akpm@linux-foundation.org: fix comment tpyo] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Miller <davem@davemloft.net> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Jeff Dike <jdike@addtoit.com> Cc: Richard Weinberger <richard@nod.at> Cc: Tony Luck <tony.luck@intel.com> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Hugh Dickins <hughd@google.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Namhyung Kim <namhyung@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-05-25 02:11:45 +02:00
struct mmu_gather tlb;
lru_add_drain();
Fix TLB gather virtual address range invalidation corner cases Ben Tebulin reported: "Since v3.7.2 on two independent machines a very specific Git repository fails in 9/10 cases on git-fsck due to an SHA1/memory failures. This only occurs on a very specific repository and can be reproduced stably on two independent laptops. Git mailing list ran out of ideas and for me this looks like some very exotic kernel issue" and bisected the failure to the backport of commit 53a59fc67f97 ("mm: limit mmu_gather batching to fix soft lockups on !CONFIG_PREEMPT"). That commit itself is not actually buggy, but what it does is to make it much more likely to hit the partial TLB invalidation case, since it introduces a new case in tlb_next_batch() that previously only ever happened when running out of memory. The real bug is that the TLB gather virtual memory range setup is subtly buggered. It was introduced in commit 597e1c3580b7 ("mm/mmu_gather: enable tlb flush range in generic mmu_gather"), and the range handling was already fixed at least once in commit e6c495a96ce0 ("mm: fix the TLB range flushed when __tlb_remove_page() runs out of slots"), but that fix was not complete. The problem with the TLB gather virtual address range is that it isn't set up by the initial tlb_gather_mmu() initialization (which didn't get the TLB range information), but it is set up ad-hoc later by the functions that actually flush the TLB. And so any such case that forgot to update the TLB range entries would potentially miss TLB invalidates. Rather than try to figure out exactly which particular ad-hoc range setup was missing (I personally suspect it's the hugetlb case in zap_huge_pmd(), which didn't have the same logic as zap_pte_range() did), this patch just gets rid of the problem at the source: make the TLB range information available to tlb_gather_mmu(), and initialize it when initializing all the other tlb gather fields. This makes the patch larger, but conceptually much simpler. And the end result is much more understandable; even if you want to play games with partial ranges when invalidating the TLB contents in chunks, now the range information is always there, and anybody who doesn't want to bother with it won't introduce subtle bugs. Ben verified that this fixes his problem. Reported-bisected-and-tested-by: Ben Tebulin <tebulin@googlemail.com> Build-testing-by: Stephen Rothwell <sfr@canb.auug.org.au> Build-testing-by: Richard Weinberger <richard.weinberger@gmail.com> Reviewed-by: Michal Hocko <mhocko@suse.cz> Acked-by: Peter Zijlstra <peterz@infradead.org> Cc: stable@vger.kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-08-15 20:42:25 +02:00
tlb_gather_mmu(&tlb, mm, start, end);
[PATCH] mm: update_hiwaters just in time update_mem_hiwater has attracted various criticisms, in particular from those concerned with mm scalability. Originally it was called whenever rss or total_vm got raised. Then many of those callsites were replaced by a timer tick call from account_system_time. Now Frank van Maarseveen reports that to be found inadequate. How about this? Works for Frank. Replace update_mem_hiwater, a poor combination of two unrelated ops, by macros update_hiwater_rss and update_hiwater_vm. Don't attempt to keep mm->hiwater_rss up to date at timer tick, nor every time we raise rss (usually by 1): those are hot paths. Do the opposite, update only when about to lower rss (usually by many), or just before final accounting in do_exit. Handle mm->hiwater_vm in the same way, though it's much less of an issue. Demand that whoever collects these hiwater statistics do the work of taking the maximum with rss or total_vm. And there has been no collector of these hiwater statistics in the tree. The new convention needs an example, so match Frank's usage by adding a VmPeak line above VmSize to /proc/<pid>/status, and also a VmHWM line above VmRSS (High-Water-Mark or High-Water-Memory). There was a particular anomaly during mremap move, that hiwater_vm might be captured too high. A fleeting such anomaly remains, but it's quickly corrected now, whereas before it would stick. What locking? None: if the app is racy then these statistics will be racy, it's not worth any overhead to make them exact. But whenever it suits, hiwater_vm is updated under exclusive mmap_sem, and hiwater_rss under page_table_lock (for now) or with preemption disabled (later on): without going to any trouble, minimize the time between reading current values and updating, to minimize those occasions when a racing thread bumps a count up and back down in between. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 02:16:18 +01:00
update_hiwater_rss(mm);
unmap_vmas(&tlb, vma, start, end);
mm: mmu_gather rework Rework the existing mmu_gather infrastructure. The direct purpose of these patches was to allow preemptible mmu_gather, but even without that I think these patches provide an improvement to the status quo. The first 9 patches rework the mmu_gather infrastructure. For review purpose I've split them into generic and per-arch patches with the last of those a generic cleanup. The next patch provides generic RCU page-table freeing, and the followup is a patch converting s390 to use this. I've also got 4 patches from DaveM lined up (not included in this series) that uses this to implement gup_fast() for sparc64. Then there is one patch that extends the generic mmu_gather batching. After that follow the mm preemptibility patches, these make part of the mm a lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes which together with the mmu_gather rework makes mmu_gather preemptible as well. Making i_mmap_lock a mutex also enables a clean-up of the truncate code. This also allows for preemptible mmu_notifiers, something that XPMEM I think wants. Furthermore, it removes the new and universially detested unmap_mutex. This patch: Remove the first obstacle towards a fully preemptible mmu_gather. The current scheme assumes mmu_gather is always done with preemption disabled and uses per-cpu storage for the page batches. Change this to try and allocate a page for batching and in case of failure, use a small on-stack array to make some progress. Preemptible mmu_gather is desired in general and usable once i_mmap_lock becomes a mutex. Doing it before the mutex conversion saves us from having to rework the code by moving the mmu_gather bits inside the pte_lock. Also avoid flushing the tlb batches from under the pte lock, this is useful even without the i_mmap_lock conversion as it significantly reduces pte lock hold times. [akpm@linux-foundation.org: fix comment tpyo] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Miller <davem@davemloft.net> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Jeff Dike <jdike@addtoit.com> Cc: Richard Weinberger <richard@nod.at> Cc: Tony Luck <tony.luck@intel.com> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Hugh Dickins <hughd@google.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Namhyung Kim <namhyung@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-05-25 02:11:45 +02:00
free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
next ? next->vm_start : USER_PGTABLES_CEILING);
mm: mmu_gather rework Rework the existing mmu_gather infrastructure. The direct purpose of these patches was to allow preemptible mmu_gather, but even without that I think these patches provide an improvement to the status quo. The first 9 patches rework the mmu_gather infrastructure. For review purpose I've split them into generic and per-arch patches with the last of those a generic cleanup. The next patch provides generic RCU page-table freeing, and the followup is a patch converting s390 to use this. I've also got 4 patches from DaveM lined up (not included in this series) that uses this to implement gup_fast() for sparc64. Then there is one patch that extends the generic mmu_gather batching. After that follow the mm preemptibility patches, these make part of the mm a lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes which together with the mmu_gather rework makes mmu_gather preemptible as well. Making i_mmap_lock a mutex also enables a clean-up of the truncate code. This also allows for preemptible mmu_notifiers, something that XPMEM I think wants. Furthermore, it removes the new and universially detested unmap_mutex. This patch: Remove the first obstacle towards a fully preemptible mmu_gather. The current scheme assumes mmu_gather is always done with preemption disabled and uses per-cpu storage for the page batches. Change this to try and allocate a page for batching and in case of failure, use a small on-stack array to make some progress. Preemptible mmu_gather is desired in general and usable once i_mmap_lock becomes a mutex. Doing it before the mutex conversion saves us from having to rework the code by moving the mmu_gather bits inside the pte_lock. Also avoid flushing the tlb batches from under the pte lock, this is useful even without the i_mmap_lock conversion as it significantly reduces pte lock hold times. [akpm@linux-foundation.org: fix comment tpyo] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Miller <davem@davemloft.net> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Jeff Dike <jdike@addtoit.com> Cc: Richard Weinberger <richard@nod.at> Cc: Tony Luck <tony.luck@intel.com> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Hugh Dickins <hughd@google.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Namhyung Kim <namhyung@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-05-25 02:11:45 +02:00
tlb_finish_mmu(&tlb, start, end);
}
/*
* Create a list of vma's touched by the unmap, removing them from the mm's
* vma list as we go..
*/
static void
detach_vmas_to_be_unmapped(struct mm_struct *mm, struct vm_area_struct *vma,
struct vm_area_struct *prev, unsigned long end)
{
struct vm_area_struct **insertion_point;
struct vm_area_struct *tail_vma = NULL;
insertion_point = (prev ? &prev->vm_next : &mm->mmap);
vma->vm_prev = NULL;
do {
vma_rb_erase(vma, &mm->mm_rb);
mm->map_count--;
tail_vma = vma;
vma = vma->vm_next;
} while (vma && vma->vm_start < end);
*insertion_point = vma;
if (vma) {
vma->vm_prev = prev;
vma_gap_update(vma);
} else
mm: larger stack guard gap, between vmas Stack guard page is a useful feature to reduce a risk of stack smashing into a different mapping. We have been using a single page gap which is sufficient to prevent having stack adjacent to a different mapping. But this seems to be insufficient in the light of the stack usage in userspace. E.g. glibc uses as large as 64kB alloca() in many commonly used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX] which is 256kB or stack strings with MAX_ARG_STRLEN. This will become especially dangerous for suid binaries and the default no limit for the stack size limit because those applications can be tricked to consume a large portion of the stack and a single glibc call could jump over the guard page. These attacks are not theoretical, unfortunatelly. Make those attacks less probable by increasing the stack guard gap to 1MB (on systems with 4k pages; but make it depend on the page size because systems with larger base pages might cap stack allocations in the PAGE_SIZE units) which should cover larger alloca() and VLA stack allocations. It is obviously not a full fix because the problem is somehow inherent, but it should reduce attack space a lot. One could argue that the gap size should be configurable from userspace, but that can be done later when somebody finds that the new 1MB is wrong for some special case applications. For now, add a kernel command line option (stack_guard_gap) to specify the stack gap size (in page units). Implementation wise, first delete all the old code for stack guard page: because although we could get away with accounting one extra page in a stack vma, accounting a larger gap can break userspace - case in point, a program run with "ulimit -S -v 20000" failed when the 1MB gap was counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK and strict non-overcommit mode. Instead of keeping gap inside the stack vma, maintain the stack guard gap as a gap between vmas: using vm_start_gap() in place of vm_start (or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few places which need to respect the gap - mainly arch_get_unmapped_area(), and and the vma tree's subtree_gap support for that. Original-patch-by: Oleg Nesterov <oleg@redhat.com> Original-patch-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Tested-by: Helge Deller <deller@gmx.de> # parisc Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-19 13:03:24 +02:00
mm->highest_vm_end = prev ? vm_end_gap(prev) : 0;
tail_vma->vm_next = NULL;
mm: per-thread vma caching This patch is a continuation of efforts trying to optimize find_vma(), avoiding potentially expensive rbtree walks to locate a vma upon faults. The original approach (https://lkml.org/lkml/2013/11/1/410), where the largest vma was also cached, ended up being too specific and random, thus further comparison with other approaches were needed. There are two things to consider when dealing with this, the cache hit rate and the latency of find_vma(). Improving the hit-rate does not necessarily translate in finding the vma any faster, as the overhead of any fancy caching schemes can be too high to consider. We currently cache the last used vma for the whole address space, which provides a nice optimization, reducing the total cycles in find_vma() by up to 250%, for workloads with good locality. On the other hand, this simple scheme is pretty much useless for workloads with poor locality. Analyzing ebizzy runs shows that, no matter how many threads are running, the mmap_cache hit rate is less than 2%, and in many situations below 1%. The proposed approach is to replace this scheme with a small per-thread cache, maximizing hit rates at a very low maintenance cost. Invalidations are performed by simply bumping up a 32-bit sequence number. The only expensive operation is in the rare case of a seq number overflow, where all caches that share the same address space are flushed. Upon a miss, the proposed replacement policy is based on the page number that contains the virtual address in question. Concretely, the following results are seen on an 80 core, 8 socket x86-64 box: 1) System bootup: Most programs are single threaded, so the per-thread scheme does improve ~50% hit rate by just adding a few more slots to the cache. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 50.61% | 19.90 | | patched | 73.45% | 13.58 | +----------------+----------+------------------+ 2) Kernel build: This one is already pretty good with the current approach as we're dealing with good locality. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 75.28% | 11.03 | | patched | 88.09% | 9.31 | +----------------+----------+------------------+ 3) Oracle 11g Data Mining (4k pages): Similar to the kernel build workload. +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 70.66% | 17.14 | | patched | 91.15% | 12.57 | +----------------+----------+------------------+ 4) Ebizzy: There's a fair amount of variation from run to run, but this approach always shows nearly perfect hit rates, while baseline is just about non-existent. The amounts of cycles can fluctuate between anywhere from ~60 to ~116 for the baseline scheme, but this approach reduces it considerably. For instance, with 80 threads: +----------------+----------+------------------+ | caching scheme | hit-rate | cycles (billion) | +----------------+----------+------------------+ | baseline | 1.06% | 91.54 | | patched | 99.97% | 14.18 | +----------------+----------+------------------+ [akpm@linux-foundation.org: fix nommu build, per Davidlohr] [akpm@linux-foundation.org: document vmacache_valid() logic] [akpm@linux-foundation.org: attempt to untangle header files] [akpm@linux-foundation.org: add vmacache_find() BUG_ON] [hughd@google.com: add vmacache_valid_mm() (from Oleg)] [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: adjust and enhance comments] Signed-off-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Michel Lespinasse <walken@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Tested-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-08 00:37:25 +02:00
/* Kill the cache */
vmacache_invalidate(mm);
}
/*
* __split_vma() bypasses sysctl_max_map_count checking. We use this where it
* has already been checked or doesn't make sense to fail.
*/
int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
unsigned long addr, int new_below)
{
struct vm_area_struct *new;
int err;
if (vma->vm_ops && vma->vm_ops->split) {
err = vma->vm_ops->split(vma, addr);
if (err)
return err;
}
new = vm_area_dup(vma);
if (!new)
return -ENOMEM;
if (new_below)
new->vm_end = addr;
else {
new->vm_start = addr;
new->vm_pgoff += ((addr - vma->vm_start) >> PAGE_SHIFT);
}
err = vma_dup_policy(vma, new);
if (err)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
goto out_free_vma;
err = anon_vma_clone(new, vma);
if (err)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
goto out_free_mpol;
mm: kill vma flag VM_EXECUTABLE and mm->num_exe_file_vmas Currently the kernel sets mm->exe_file during sys_execve() and then tracks number of vmas with VM_EXECUTABLE flag in mm->num_exe_file_vmas, as soon as this counter drops to zero kernel resets mm->exe_file to NULL. Plus it resets mm->exe_file at last mmput() when mm->mm_users drops to zero. VMA with VM_EXECUTABLE flag appears after mapping file with flag MAP_EXECUTABLE, such vmas can appears only at sys_execve() or after vma splitting, because sys_mmap ignores this flag. Usually binfmt module sets mm->exe_file and mmaps executable vmas with this file, they hold mm->exe_file while task is running. comment from v2.6.25-6245-g925d1c4 ("procfs task exe symlink"), where all this stuff was introduced: > The kernel implements readlink of /proc/pid/exe by getting the file from > the first executable VMA. Then the path to the file is reconstructed and > reported as the result. > > Because of the VMA walk the code is slightly different on nommu systems. > This patch avoids separate /proc/pid/exe code on nommu systems. Instead of > walking the VMAs to find the first executable file-backed VMA we store a > reference to the exec'd file in the mm_struct. > > That reference would prevent the filesystem holding the executable file > from being unmounted even after unmapping the VMAs. So we track the number > of VM_EXECUTABLE VMAs and drop the new reference when the last one is > unmapped. This avoids pinning the mounted filesystem. exe_file's vma accounting is hooked into every file mmap/unmmap and vma split/merge just to fix some hypothetical pinning fs from umounting by mm, which already unmapped all its executable files, but still alive. Seems like currently nobody depends on this behaviour. We can try to remove this logic and keep mm->exe_file until final mmput(). mm->exe_file is still protected with mm->mmap_sem, because we want to change it via new sys_prctl(PR_SET_MM_EXE_FILE). Also via this syscall task can change its mm->exe_file and unpin mountpoint explicitly. Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Carsten Otte <cotte@de.ibm.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Eric Paris <eparis@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hugh Dickins <hughd@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Kentaro Takeda <takedakn@nttdata.co.jp> Cc: Matt Helsley <matthltc@us.ibm.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Robert Richter <robert.richter@amd.com> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Venkatesh Pallipadi <venki@google.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:28:54 +02:00
if (new->vm_file)
get_file(new->vm_file);
if (new->vm_ops && new->vm_ops->open)
new->vm_ops->open(new);
if (new_below)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
err = vma_adjust(vma, addr, vma->vm_end, vma->vm_pgoff +
((addr - new->vm_start) >> PAGE_SHIFT), new);
else
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
err = vma_adjust(vma, vma->vm_start, addr, vma->vm_pgoff, new);
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
/* Success. */
if (!err)
return 0;
/* Clean everything up if vma_adjust failed. */
if (new->vm_ops && new->vm_ops->close)
new->vm_ops->close(new);
mm: kill vma flag VM_EXECUTABLE and mm->num_exe_file_vmas Currently the kernel sets mm->exe_file during sys_execve() and then tracks number of vmas with VM_EXECUTABLE flag in mm->num_exe_file_vmas, as soon as this counter drops to zero kernel resets mm->exe_file to NULL. Plus it resets mm->exe_file at last mmput() when mm->mm_users drops to zero. VMA with VM_EXECUTABLE flag appears after mapping file with flag MAP_EXECUTABLE, such vmas can appears only at sys_execve() or after vma splitting, because sys_mmap ignores this flag. Usually binfmt module sets mm->exe_file and mmaps executable vmas with this file, they hold mm->exe_file while task is running. comment from v2.6.25-6245-g925d1c4 ("procfs task exe symlink"), where all this stuff was introduced: > The kernel implements readlink of /proc/pid/exe by getting the file from > the first executable VMA. Then the path to the file is reconstructed and > reported as the result. > > Because of the VMA walk the code is slightly different on nommu systems. > This patch avoids separate /proc/pid/exe code on nommu systems. Instead of > walking the VMAs to find the first executable file-backed VMA we store a > reference to the exec'd file in the mm_struct. > > That reference would prevent the filesystem holding the executable file > from being unmounted even after unmapping the VMAs. So we track the number > of VM_EXECUTABLE VMAs and drop the new reference when the last one is > unmapped. This avoids pinning the mounted filesystem. exe_file's vma accounting is hooked into every file mmap/unmmap and vma split/merge just to fix some hypothetical pinning fs from umounting by mm, which already unmapped all its executable files, but still alive. Seems like currently nobody depends on this behaviour. We can try to remove this logic and keep mm->exe_file until final mmput(). mm->exe_file is still protected with mm->mmap_sem, because we want to change it via new sys_prctl(PR_SET_MM_EXE_FILE). Also via this syscall task can change its mm->exe_file and unpin mountpoint explicitly. Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Carsten Otte <cotte@de.ibm.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Eric Paris <eparis@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hugh Dickins <hughd@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Kentaro Takeda <takedakn@nttdata.co.jp> Cc: Matt Helsley <matthltc@us.ibm.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Robert Richter <robert.richter@amd.com> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Venkatesh Pallipadi <venki@google.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:28:54 +02:00
if (new->vm_file)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
fput(new->vm_file);
unlink_anon_vmas(new);
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
out_free_mpol:
mpol_put(vma_policy(new));
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
out_free_vma:
vm_area_free(new);
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
return err;
}
mmap: don't return ENOMEM when mapcount is temporarily exceeded in munmap() On ia64, the following test program exit abnormally, because glibc thread library called abort(). ======================================================== (gdb) bt #0 0xa000000000010620 in __kernel_syscall_via_break () #1 0x20000000003208e0 in raise () from /lib/libc.so.6.1 #2 0x2000000000324090 in abort () from /lib/libc.so.6.1 #3 0x200000000027c3e0 in __deallocate_stack () from /lib/libpthread.so.0 #4 0x200000000027f7c0 in start_thread () from /lib/libpthread.so.0 #5 0x200000000047ef60 in __clone2 () from /lib/libc.so.6.1 ======================================================== The fact is, glibc call munmap() when thread exitng time for freeing stack, and it assume munlock() never fail. However, munmap() often make vma splitting and it with many mapcount make -ENOMEM. Oh well, that's crazy, because stack unmapping never increase mapcount. The maxcount exceeding is only temporary. internal temporary exceeding shouldn't make ENOMEM. This patch does it. test_max_mapcount.c ================================================================== #include<stdio.h> #include<stdlib.h> #include<string.h> #include<pthread.h> #include<errno.h> #include<unistd.h> #define THREAD_NUM 30000 #define MAL_SIZE (8*1024*1024) void *wait_thread(void *args) { void *addr; addr = malloc(MAL_SIZE); sleep(10); return NULL; } void *wait_thread2(void *args) { sleep(60); return NULL; } int main(int argc, char *argv[]) { int i; pthread_t thread[THREAD_NUM], th; int ret, count = 0; pthread_attr_t attr; ret = pthread_attr_init(&attr); if(ret) { perror("pthread_attr_init"); } ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED); if(ret) { perror("pthread_attr_setdetachstate"); } for (i = 0; i < THREAD_NUM; i++) { ret = pthread_create(&th, &attr, wait_thread, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; ret = pthread_create(&thread[i], &attr, wait_thread2, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; } sleep(3600); return 0; } ================================================================== [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-15 02:57:56 +01:00
/*
* Split a vma into two pieces at address 'addr', a new vma is allocated
* either for the first part or the tail.
*/
int split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
unsigned long addr, int new_below)
{
if (mm->map_count >= sysctl_max_map_count)
return -ENOMEM;
return __split_vma(mm, vma, addr, new_below);
}
/* Munmap is split into 2 main parts -- this part which finds
* what needs doing, and the areas themselves, which do the
* work. This now handles partial unmappings.
* Jeremy Fitzhardinge <jeremy@goop.org>
*/
mm: mremap: downgrade mmap_sem to read when shrinking Other than munmap, mremap might be used to shrink memory mapping too. So, it may hold write mmap_sem for long time when shrinking large mapping, as what commit ("mm: mmap: zap pages with read mmap_sem in munmap") described. The mremap() will not manipulate vmas anymore after __do_munmap() call for the mapping shrink use case, so it is safe to downgrade to read mmap_sem. So, the same optimization, which downgrades mmap_sem to read for zapping pages, is also feasible and reasonable to this case. The period of holding exclusive mmap_sem for shrinking large mapping would be reduced significantly with this optimization. MREMAP_FIXED and MREMAP_MAYMOVE are more complicated to adopt this optimization since they need manipulate vmas after do_munmap(), downgrading mmap_sem may create race window. Simple mapping shrink is the low hanging fruit, and it may cover the most cases of unmap with munmap together. [akpm@linux-foundation.org: tweak comment] [yang.shi@linux.alibaba.com: fix unsigned compare against 0 issue] Link: http://lkml.kernel.org/r/1538687672-17795-2-git-send-email-yang.shi@linux.alibaba.com Link: http://lkml.kernel.org/r/1538067582-60038-1-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Colin Ian King <colin.king@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:08:50 +02:00
int __do_munmap(struct mm_struct *mm, unsigned long start, size_t len,
struct list_head *uf, bool downgrade)
{
unsigned long end;
struct vm_area_struct *vma, *prev, *last;
if ((offset_in_page(start)) || start > TASK_SIZE || len > TASK_SIZE-start)
return -EINVAL;
len = PAGE_ALIGN(len);
x86/mpx, mm/core: Fix recursive munmap() corruption This is a bit of a mess, to put it mildly. But, it's a bug that only seems to have showed up in 4.20 but wasn't noticed until now, because nobody uses MPX. MPX has the arch_unmap() hook inside of munmap() because MPX uses bounds tables that protect other areas of memory. When memory is unmapped, there is also a need to unmap the MPX bounds tables. Barring this, unused bounds tables can eat 80% of the address space. But, the recursive do_munmap() that gets called vi arch_unmap() wreaks havoc with __do_munmap()'s state. It can result in freeing populated page tables, accessing bogus VMA state, double-freed VMAs and more. See the "long story" further below for the gory details. To fix this, call arch_unmap() before __do_unmap() has a chance to do anything meaningful. Also, remove the 'vma' argument and force the MPX code to do its own, independent VMA lookup. == UML / unicore32 impact == Remove unused 'vma' argument to arch_unmap(). No functional change. I compile tested this on UML but not unicore32. == powerpc impact == powerpc uses arch_unmap() well to watch for munmap() on the VDSO and zeroes out 'current->mm->context.vdso_base'. Moving arch_unmap() makes this happen earlier in __do_munmap(). But, 'vdso_base' seems to only be used in perf and in the signal delivery that happens near the return to userspace. I can not find any likely impact to powerpc, other than the zeroing happening a little earlier. powerpc does not use the 'vma' argument and is unaffected by its removal. I compile-tested a 64-bit powerpc defconfig. == x86 impact == For the common success case this is functionally identical to what was there before. For the munmap() failure case, it's possible that some MPX tables will be zapped for memory that continues to be in use. But, this is an extraordinarily unlikely scenario and the harm would be that MPX provides no protection since the bounds table got reset (zeroed). I can't imagine anyone doing this: ptr = mmap(); // use ptr ret = munmap(ptr); if (ret) // oh, there was an error, I'll // keep using ptr. Because if you're doing munmap(), you are *done* with the memory. There's probably no good data in there _anyway_. This passes the original reproducer from Richard Biener as well as the existing mpx selftests/. The long story: munmap() has a couple of pieces: 1. Find the affected VMA(s) 2. Split the start/end one(s) if neceesary 3. Pull the VMAs out of the rbtree 4. Actually zap the memory via unmap_region(), including freeing page tables (or queueing them to be freed). 5. Fix up some of the accounting (like fput()) and actually free the VMA itself. This specific ordering was actually introduced by: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap") during the 4.20 merge window. The previous __do_munmap() code was actually safe because the only thing after arch_unmap() was remove_vma_list(). arch_unmap() could not see 'vma' in the rbtree because it was detached, so it is not even capable of doing operations unsafe for remove_vma_list()'s use of 'vma'. Richard Biener reported a test that shows this in dmesg: [1216548.787498] BUG: Bad rss-counter state mm:0000000017ce560b idx:1 val:551 [1216548.787500] BUG: non-zero pgtables_bytes on freeing mm: 24576 What triggered this was the recursive do_munmap() called via arch_unmap(). It was freeing page tables that has not been properly zapped. But, the problem was bigger than this. For one, arch_unmap() can free VMAs. But, the calling __do_munmap() has variables that *point* to VMAs and obviously can't handle them just getting freed while the pointer is still in use. I tried a couple of things here. First, I tried to fix the page table freeing problem in isolation, but I then found the VMA issue. I also tried having the MPX code return a flag if it modified the rbtree which would force __do_munmap() to re-walk to restart. That spiralled out of control in complexity pretty fast. Just moving arch_unmap() and accepting that the bonkers failure case might eat some bounds tables seems like the simplest viable fix. This was also reported in the following kernel bugzilla entry: https://bugzilla.kernel.org/show_bug.cgi?id=203123 There are some reports that this commit triggered this bug: dd2283f2605 ("mm: mmap: zap pages with read mmap_sem in munmap") While that commit certainly made the issues easier to hit, I believe the fundamental issue has been with us as long as MPX itself, thus the Fixes: tag below is for one of the original MPX commits. [ mingo: Minor edits to the changelog and the patch. ] Reported-by: Richard Biener <rguenther@suse.de> Reported-by: H.J. Lu <hjl.tools@gmail.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Anton Ivanov <anton.ivanov@cambridgegreys.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Guan Xuetao <gxt@pku.edu.cn> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jeff Dike <jdike@addtoit.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Richard Weinberger <richard@nod.at> Cc: Rik van Riel <riel@surriel.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: linux-arch@vger.kernel.org Cc: linux-mm@kvack.org Cc: linux-um@lists.infradead.org Cc: linuxppc-dev@lists.ozlabs.org Cc: stable@vger.kernel.org Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap") Link: http://lkml.kernel.org/r/20190419194747.5E1AD6DC@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-04-19 21:47:47 +02:00
end = start + len;
if (len == 0)
return -EINVAL;
x86/mpx, mm/core: Fix recursive munmap() corruption This is a bit of a mess, to put it mildly. But, it's a bug that only seems to have showed up in 4.20 but wasn't noticed until now, because nobody uses MPX. MPX has the arch_unmap() hook inside of munmap() because MPX uses bounds tables that protect other areas of memory. When memory is unmapped, there is also a need to unmap the MPX bounds tables. Barring this, unused bounds tables can eat 80% of the address space. But, the recursive do_munmap() that gets called vi arch_unmap() wreaks havoc with __do_munmap()'s state. It can result in freeing populated page tables, accessing bogus VMA state, double-freed VMAs and more. See the "long story" further below for the gory details. To fix this, call arch_unmap() before __do_unmap() has a chance to do anything meaningful. Also, remove the 'vma' argument and force the MPX code to do its own, independent VMA lookup. == UML / unicore32 impact == Remove unused 'vma' argument to arch_unmap(). No functional change. I compile tested this on UML but not unicore32. == powerpc impact == powerpc uses arch_unmap() well to watch for munmap() on the VDSO and zeroes out 'current->mm->context.vdso_base'. Moving arch_unmap() makes this happen earlier in __do_munmap(). But, 'vdso_base' seems to only be used in perf and in the signal delivery that happens near the return to userspace. I can not find any likely impact to powerpc, other than the zeroing happening a little earlier. powerpc does not use the 'vma' argument and is unaffected by its removal. I compile-tested a 64-bit powerpc defconfig. == x86 impact == For the common success case this is functionally identical to what was there before. For the munmap() failure case, it's possible that some MPX tables will be zapped for memory that continues to be in use. But, this is an extraordinarily unlikely scenario and the harm would be that MPX provides no protection since the bounds table got reset (zeroed). I can't imagine anyone doing this: ptr = mmap(); // use ptr ret = munmap(ptr); if (ret) // oh, there was an error, I'll // keep using ptr. Because if you're doing munmap(), you are *done* with the memory. There's probably no good data in there _anyway_. This passes the original reproducer from Richard Biener as well as the existing mpx selftests/. The long story: munmap() has a couple of pieces: 1. Find the affected VMA(s) 2. Split the start/end one(s) if neceesary 3. Pull the VMAs out of the rbtree 4. Actually zap the memory via unmap_region(), including freeing page tables (or queueing them to be freed). 5. Fix up some of the accounting (like fput()) and actually free the VMA itself. This specific ordering was actually introduced by: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap") during the 4.20 merge window. The previous __do_munmap() code was actually safe because the only thing after arch_unmap() was remove_vma_list(). arch_unmap() could not see 'vma' in the rbtree because it was detached, so it is not even capable of doing operations unsafe for remove_vma_list()'s use of 'vma'. Richard Biener reported a test that shows this in dmesg: [1216548.787498] BUG: Bad rss-counter state mm:0000000017ce560b idx:1 val:551 [1216548.787500] BUG: non-zero pgtables_bytes on freeing mm: 24576 What triggered this was the recursive do_munmap() called via arch_unmap(). It was freeing page tables that has not been properly zapped. But, the problem was bigger than this. For one, arch_unmap() can free VMAs. But, the calling __do_munmap() has variables that *point* to VMAs and obviously can't handle them just getting freed while the pointer is still in use. I tried a couple of things here. First, I tried to fix the page table freeing problem in isolation, but I then found the VMA issue. I also tried having the MPX code return a flag if it modified the rbtree which would force __do_munmap() to re-walk to restart. That spiralled out of control in complexity pretty fast. Just moving arch_unmap() and accepting that the bonkers failure case might eat some bounds tables seems like the simplest viable fix. This was also reported in the following kernel bugzilla entry: https://bugzilla.kernel.org/show_bug.cgi?id=203123 There are some reports that this commit triggered this bug: dd2283f2605 ("mm: mmap: zap pages with read mmap_sem in munmap") While that commit certainly made the issues easier to hit, I believe the fundamental issue has been with us as long as MPX itself, thus the Fixes: tag below is for one of the original MPX commits. [ mingo: Minor edits to the changelog and the patch. ] Reported-by: Richard Biener <rguenther@suse.de> Reported-by: H.J. Lu <hjl.tools@gmail.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Reviewed-by Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Anton Ivanov <anton.ivanov@cambridgegreys.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Guan Xuetao <gxt@pku.edu.cn> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jeff Dike <jdike@addtoit.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Richard Weinberger <richard@nod.at> Cc: Rik van Riel <riel@surriel.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: linux-arch@vger.kernel.org Cc: linux-mm@kvack.org Cc: linux-um@lists.infradead.org Cc: linuxppc-dev@lists.ozlabs.org Cc: stable@vger.kernel.org Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap") Link: http://lkml.kernel.org/r/20190419194747.5E1AD6DC@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-04-19 21:47:47 +02:00
/*
* arch_unmap() might do unmaps itself. It must be called
* and finish any rbtree manipulation before this code
* runs and also starts to manipulate the rbtree.
*/
arch_unmap(mm, start, end);
/* Find the first overlapping VMA */
vma = find_vma(mm, start);
if (!vma)
return 0;
prev = vma->vm_prev;
/* we have start < vma->vm_end */
/* if it doesn't overlap, we have nothing.. */
if (vma->vm_start >= end)
return 0;
/*
* If we need to split any vma, do it now to save pain later.
*
* Note: mremap's move_vma VM_ACCOUNT handling assumes a partially
* unmapped vm_area_struct will remain in use: so lower split_vma
* places tmp vma above, and higher split_vma places tmp vma below.
*/
if (start > vma->vm_start) {
mmap: don't return ENOMEM when mapcount is temporarily exceeded in munmap() On ia64, the following test program exit abnormally, because glibc thread library called abort(). ======================================================== (gdb) bt #0 0xa000000000010620 in __kernel_syscall_via_break () #1 0x20000000003208e0 in raise () from /lib/libc.so.6.1 #2 0x2000000000324090 in abort () from /lib/libc.so.6.1 #3 0x200000000027c3e0 in __deallocate_stack () from /lib/libpthread.so.0 #4 0x200000000027f7c0 in start_thread () from /lib/libpthread.so.0 #5 0x200000000047ef60 in __clone2 () from /lib/libc.so.6.1 ======================================================== The fact is, glibc call munmap() when thread exitng time for freeing stack, and it assume munlock() never fail. However, munmap() often make vma splitting and it with many mapcount make -ENOMEM. Oh well, that's crazy, because stack unmapping never increase mapcount. The maxcount exceeding is only temporary. internal temporary exceeding shouldn't make ENOMEM. This patch does it. test_max_mapcount.c ================================================================== #include<stdio.h> #include<stdlib.h> #include<string.h> #include<pthread.h> #include<errno.h> #include<unistd.h> #define THREAD_NUM 30000 #define MAL_SIZE (8*1024*1024) void *wait_thread(void *args) { void *addr; addr = malloc(MAL_SIZE); sleep(10); return NULL; } void *wait_thread2(void *args) { sleep(60); return NULL; } int main(int argc, char *argv[]) { int i; pthread_t thread[THREAD_NUM], th; int ret, count = 0; pthread_attr_t attr; ret = pthread_attr_init(&attr); if(ret) { perror("pthread_attr_init"); } ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED); if(ret) { perror("pthread_attr_setdetachstate"); } for (i = 0; i < THREAD_NUM; i++) { ret = pthread_create(&th, &attr, wait_thread, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; ret = pthread_create(&thread[i], &attr, wait_thread2, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; } sleep(3600); return 0; } ================================================================== [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-15 02:57:56 +01:00
int error;
/*
* Make sure that map_count on return from munmap() will
* not exceed its limit; but let map_count go just above
* its limit temporarily, to help free resources as expected.
*/
if (end < vma->vm_end && mm->map_count >= sysctl_max_map_count)
return -ENOMEM;
error = __split_vma(mm, vma, start, 0);
if (error)
return error;
prev = vma;
}
/* Does it split the last one? */
last = find_vma(mm, end);
if (last && end > last->vm_start) {
mmap: don't return ENOMEM when mapcount is temporarily exceeded in munmap() On ia64, the following test program exit abnormally, because glibc thread library called abort(). ======================================================== (gdb) bt #0 0xa000000000010620 in __kernel_syscall_via_break () #1 0x20000000003208e0 in raise () from /lib/libc.so.6.1 #2 0x2000000000324090 in abort () from /lib/libc.so.6.1 #3 0x200000000027c3e0 in __deallocate_stack () from /lib/libpthread.so.0 #4 0x200000000027f7c0 in start_thread () from /lib/libpthread.so.0 #5 0x200000000047ef60 in __clone2 () from /lib/libc.so.6.1 ======================================================== The fact is, glibc call munmap() when thread exitng time for freeing stack, and it assume munlock() never fail. However, munmap() often make vma splitting and it with many mapcount make -ENOMEM. Oh well, that's crazy, because stack unmapping never increase mapcount. The maxcount exceeding is only temporary. internal temporary exceeding shouldn't make ENOMEM. This patch does it. test_max_mapcount.c ================================================================== #include<stdio.h> #include<stdlib.h> #include<string.h> #include<pthread.h> #include<errno.h> #include<unistd.h> #define THREAD_NUM 30000 #define MAL_SIZE (8*1024*1024) void *wait_thread(void *args) { void *addr; addr = malloc(MAL_SIZE); sleep(10); return NULL; } void *wait_thread2(void *args) { sleep(60); return NULL; } int main(int argc, char *argv[]) { int i; pthread_t thread[THREAD_NUM], th; int ret, count = 0; pthread_attr_t attr; ret = pthread_attr_init(&attr); if(ret) { perror("pthread_attr_init"); } ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED); if(ret) { perror("pthread_attr_setdetachstate"); } for (i = 0; i < THREAD_NUM; i++) { ret = pthread_create(&th, &attr, wait_thread, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; ret = pthread_create(&thread[i], &attr, wait_thread2, NULL); if(ret) { fprintf(stderr, "[%d] ", count); perror("pthread_create"); } else { printf("[%d] create OK.\n", count); } count++; } sleep(3600); return 0; } ================================================================== [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-15 02:57:56 +01:00
int error = __split_vma(mm, last, end, 1);
if (error)
return error;
}
vma = prev ? prev->vm_next : mm->mmap;
if (unlikely(uf)) {
/*
* If userfaultfd_unmap_prep returns an error the vmas
* will remain splitted, but userland will get a
* highly unexpected error anyway. This is no
* different than the case where the first of the two
* __split_vma fails, but we don't undo the first
* split, despite we could. This is unlikely enough
* failure that it's not worth optimizing it for.
*/
int error = userfaultfd_unmap_prep(vma, start, end, uf);
if (error)
return error;
}
/*
* unlock any mlock()ed ranges before detaching vmas
*/
if (mm->locked_vm) {
struct vm_area_struct *tmp = vma;
while (tmp && tmp->vm_start < end) {
if (tmp->vm_flags & VM_LOCKED) {
mm->locked_vm -= vma_pages(tmp);
munlock_vma_pages_all(tmp);
}
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
tmp = tmp->vm_next;
}
}
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
/* Detach vmas from rbtree */
detach_vmas_to_be_unmapped(mm, vma, prev, end);
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
if (downgrade)
downgrade_write(&mm->mmap_sem);
unmap_region(mm, vma, prev, start, end);
/* Fix up all other VM information */
remove_vma_list(mm, vma);
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
return downgrade ? 1 : 0;
}
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
int do_munmap(struct mm_struct *mm, unsigned long start, size_t len,
struct list_head *uf)
{
return __do_munmap(mm, start, len, uf, false);
}
static int __vm_munmap(unsigned long start, size_t len, bool downgrade)
{
int ret;
struct mm_struct *mm = current->mm;
LIST_HEAD(uf);
if (down_write_killable(&mm->mmap_sem))
return -EINTR;
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
ret = __do_munmap(mm, start, len, &uf, downgrade);
/*
* Returning 1 indicates mmap_sem is downgraded.
* But 1 is not legal return value of vm_munmap() and munmap(), reset
* it to 0 before return.
*/
if (ret == 1) {
up_read(&mm->mmap_sem);
ret = 0;
} else
up_write(&mm->mmap_sem);
userfaultfd_unmap_complete(mm, &uf);
return ret;
}
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
int vm_munmap(unsigned long start, size_t len)
{
return __vm_munmap(start, len, false);
}
EXPORT_SYMBOL(vm_munmap);
SYSCALL_DEFINE2(munmap, unsigned long, addr, size_t, len)
{
profile_munmap(addr);
mm: mmap: zap pages with read mmap_sem in munmap Patch series "mm: zap pages with read mmap_sem in munmap for large mapping", v11. Background: Recently, when we ran some vm scalability tests on machines with large memory, we ran into a couple of mmap_sem scalability issues when unmapping large memory space, please refer to https://lkml.org/lkml/2017/12/14/733 and https://lkml.org/lkml/2018/2/20/576. History: Then akpm suggested to unmap large mapping section by section and drop mmap_sem at a time to mitigate it (see https://lkml.org/lkml/2018/3/6/784). V1 patch series was submitted to the mailing list per Andrew's suggestion (see https://lkml.org/lkml/2018/3/20/786). Then I received a lot great feedback and suggestions. Then this topic was discussed on LSFMM summit 2018. In the summit, Michal Hocko suggested (also in the v1 patches review) to try "two phases" approach. Zapping pages with read mmap_sem, then doing via cleanup with write mmap_sem (for discussion detail, see https://lwn.net/Articles/753269/) Approach: Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, we can't call MADV_DONTNEED directly, since there are two major drawbacks: * The unexpected state from PF if it wins the race in the middle of munmap. It may return zero page, instead of the content or SIGSEGV. * Can't handle VM_LOCKED | VM_HUGETLB | VM_PFNMAP and uprobe mappings, which is a showstopper from akpm But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by falling back to regular do_munmap() with exclusive mmap_sem held in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. So, uprobe unmap will not be handled by the regular path. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. This patch (of 3): When running some mmap/munmap scalability tests with large memory (i.e. > 300GB), the below hung task issue may happen occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [<ffffffff817154d0>] ? __schedule+0x250/0x730 [<ffffffff817159e6>] schedule+0x36/0x80 [<ffffffff81718560>] rwsem_down_read_failed+0xf0/0x150 [<ffffffff81390a28>] call_rwsem_down_read_failed+0x18/0x30 [<ffffffff81717db0>] down_read+0x20/0x40 [<ffffffff812b9439>] proc_pid_cmdline_read+0xd9/0x4e0 [<ffffffff81253c95>] ? do_filp_open+0xa5/0x100 [<ffffffff81241d87>] __vfs_read+0x37/0x150 [<ffffffff812f824b>] ? security_file_permission+0x9b/0xc0 [<ffffffff81242266>] vfs_read+0x96/0x130 [<ffffffff812437b5>] SyS_read+0x55/0xc0 [<ffffffff8171a6da>] entry_SYSCALL_64_fastpath+0x1a/0xc5 It is because munmap holds mmap_sem exclusively from very beginning to all the way down to the end, and doesn't release it in the middle. When unmapping large mapping, it may take long time (take ~18 seconds to unmap 320GB mapping with every single page mapped on an idle machine). Zapping pages is the most time consuming part, according to the suggestion from Michal Hocko [1], zapping pages can be done with holding read mmap_sem, like what MADV_DONTNEED does. Then re-acquire write mmap_sem to cleanup vmas. But, some part may need write mmap_sem, for example, vma splitting. So, the design is as follows: acquire write mmap_sem lookup vmas (find and split vmas) deal with special mappings detach vmas downgrade_write zap pages free page tables release mmap_sem The vm events with read mmap_sem may come in during page zapping, but since vmas have been detached before, they, i.e. page fault, gup, etc, will not be able to find valid vma, then just return SIGSEGV or -EFAULT as expected. If the vma has VM_HUGETLB | VM_PFNMAP, they are considered as special mappings. They will be handled by without downgrading mmap_sem in this patch since they may update vm flags. But, with the "detach vmas first" approach, the vmas have been detached when vm flags are updated, so it sounds safe to update vm flags with read mmap_sem for this specific case. So, VM_HUGETLB and VM_PFNMAP will be handled by using the optimized path in the following separate patches for bisectable sake. Unmapping uprobe areas may need update mm flags (MMF_RECALC_UPROBES). However it is fine to have false-positive MMF_RECALC_UPROBES according to uprobes developer. With the "detach vmas first" approach we don't have to re-acquire mmap_sem again to clean up vmas to avoid race window which might get the address space changed since downgrade_write() doesn't release the lock to lead regression, which simply downgrades to read lock. And, since the lock acquire/release cost is managed to the minimum and almost as same as before, the optimization could be extended to any size of mapping without incurring significant penalty to small mappings. For the time being, just do this in munmap syscall path. Other vm_munmap() or do_munmap() call sites (i.e mmap, mremap, etc) remain intact due to some implementation difficulties since they acquire write mmap_sem from very beginning and hold it until the end, do_munmap() might be called in the middle. But, the optimized do_munmap would like to be called without mmap_sem held so that we can do the optimization. So, if we want to do the similar optimization for mmap/mremap path, I'm afraid we would have to redesign them. mremap might be called on very large area depending on the usecases, the optimization to it will be considered in the future. With the patches, exclusive mmap_sem hold time when munmap a 80GB address space on a machine with 32 cores of E5-2680 @ 2.70GHz dropped to us level from second. munmap_test-15002 [008] 594.380138: funcgraph_entry: | __vm_munmap() { munmap_test-15002 [008] 594.380146: funcgraph_entry: !2485684 us | unmap_region(); munmap_test-15002 [008] 596.865836: funcgraph_exit: !2485692 us | } Here the execution time of unmap_region() is used to evaluate the time of holding read mmap_sem, then the remaining time is used with holding exclusive lock. [1] https://lwn.net/Articles/753269/ Link: http://lkml.kernel.org/r/1537376621-51150-2-git-send-email-yang.shi@linux.alibaba.com Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>Suggested-by: Michal Hocko <mhocko@kernel.org> Suggested-by: Kirill A. Shutemov <kirill@shutemov.name> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Matthew Wilcox <willy@infradead.org> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-27 00:07:11 +02:00
return __vm_munmap(addr, len, true);
}
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
/*
* Emulation of deprecated remap_file_pages() syscall.
*/
SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
unsigned long, prot, unsigned long, pgoff, unsigned long, flags)
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
unsigned long populate = 0;
unsigned long ret = -EINVAL;
struct file *file;
pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/vm/remap_file_pages.rst.\n",
current->comm, current->pid);
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
if (prot)
return ret;
start = start & PAGE_MASK;
size = size & PAGE_MASK;
if (start + size <= start)
return ret;
/* Does pgoff wrap? */
if (pgoff + (size >> PAGE_SHIFT) < pgoff)
return ret;
mm: make mmap_sem for write waits killable for mm syscalls This is a follow up work for oom_reaper [1]. As the async OOM killing depends on oom_sem for read we would really appreciate if a holder for write didn't stood in the way. This patchset is changing many of down_write calls to be killable to help those cases when the writer is blocked and waiting for readers to release the lock and so help __oom_reap_task to process the oom victim. Most of the patches are really trivial because the lock is help from a shallow syscall paths where we can return EINTR trivially and allow the current task to die (note that EINTR will never get to the userspace as the task has fatal signal pending). Others seem to be easy as well as the callers are already handling fatal errors and bail and return to userspace which should be sufficient to handle the failure gracefully. I am not familiar with all those code paths so a deeper review is really appreciated. As this work is touching more areas which are not directly connected I have tried to keep the CC list as small as possible and people who I believed would be familiar are CCed only to the specific patches (all should have received the cover though). This patchset is based on linux-next and it depends on down_write_killable for rw_semaphores which got merged into tip locking/rwsem branch and it is merged into this next tree. I guess it would be easiest to route these patches via mmotm because of the dependency on the tip tree but if respective maintainers prefer other way I have no objections. I haven't covered all the mmap_write(mm->mmap_sem) instances here $ git grep "down_write(.*\<mmap_sem\>)" next/master | wc -l 98 $ git grep "down_write(.*\<mmap_sem\>)" | wc -l 62 I have tried to cover those which should be relatively easy to review in this series because this alone should be a nice improvement. Other places can be changed on top. [0] http://lkml.kernel.org/r/1456752417-9626-1-git-send-email-mhocko@kernel.org [1] http://lkml.kernel.org/r/1452094975-551-1-git-send-email-mhocko@kernel.org [2] http://lkml.kernel.org/r/1456750705-7141-1-git-send-email-mhocko@kernel.org This patch (of 18): This is the first step in making mmap_sem write waiters killable. It focuses on the trivial ones which are taking the lock early after entering the syscall and they are not changing state before. Therefore it is very easy to change them to use down_write_killable and immediately return with -EINTR. This will allow the waiter to pass away without blocking the mmap_sem which might be required to make a forward progress. E.g. the oom reaper will need the lock for reading to dismantle the OOM victim address space. The only tricky function in this patch is vm_mmap_pgoff which has many call sites via vm_mmap. To reduce the risk keep vm_mmap with the original non-killable semantic for now. vm_munmap callers do not bother checking the return value so open code it into the munmap syscall path for now for simplicity. Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Mel Gorman <mgorman@suse.de> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Hugh Dickins <hughd@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: David Rientjes <rientjes@google.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-24 01:25:27 +02:00
if (down_write_killable(&mm->mmap_sem))
return -EINTR;
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
vma = find_vma(mm, start);
if (!vma || !(vma->vm_flags & VM_SHARED))
goto out;
2016-02-17 22:11:15 +01:00
if (start < vma->vm_start)
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
goto out;
2016-02-17 22:11:15 +01:00
if (start + size > vma->vm_end) {
struct vm_area_struct *next;
for (next = vma->vm_next; next; next = next->vm_next) {
/* hole between vmas ? */
if (next->vm_start != next->vm_prev->vm_end)
goto out;
if (next->vm_file != vma->vm_file)
goto out;
if (next->vm_flags != vma->vm_flags)
goto out;
if (start + size <= next->vm_end)
break;
}
if (!next)
goto out;
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
}
prot |= vma->vm_flags & VM_READ ? PROT_READ : 0;
prot |= vma->vm_flags & VM_WRITE ? PROT_WRITE : 0;
prot |= vma->vm_flags & VM_EXEC ? PROT_EXEC : 0;
flags &= MAP_NONBLOCK;
flags |= MAP_SHARED | MAP_FIXED | MAP_POPULATE;
if (vma->vm_flags & VM_LOCKED) {
2016-02-17 22:11:15 +01:00
struct vm_area_struct *tmp;
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
flags |= MAP_LOCKED;
2016-02-17 22:11:15 +01:00
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
/* drop PG_Mlocked flag for over-mapped range */
2016-02-17 22:11:15 +01:00
for (tmp = vma; tmp->vm_start >= start + size;
tmp = tmp->vm_next) {
/*
* Split pmd and munlock page on the border
* of the range.
*/
vma_adjust_trans_huge(tmp, start, start + size, 0);
2016-02-17 22:11:15 +01:00
munlock_vma_pages_range(tmp,
max(tmp->vm_start, start),
min(tmp->vm_end, start + size));
}
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
}
file = get_file(vma->vm_file);
ret = do_mmap_pgoff(vma->vm_file, start, size,
prot, flags, pgoff, &populate, NULL);
mm: replace remap_file_pages() syscall with emulation remap_file_pages(2) was invented to be able efficiently map parts of huge file into limited 32-bit virtual address space such as in database workloads. Nonlinear mappings are pain to support and it seems there's no legitimate use-cases nowadays since 64-bit systems are widely available. Let's drop it and get rid of all these special-cased code. The patch replaces the syscall with emulation which creates new VMA on each remap_file_pages(), unless they it can be merged with an adjacent one. I didn't find *any* real code that uses remap_file_pages(2) to test emulation impact on. I've checked Debian code search and source of all packages in ALT Linux. No real users: libc wrappers, mentions in strace, gdb, valgrind and this kind of stuff. There are few basic tests in LTP for the syscall. They work just fine with emulation. To test performance impact, I've written small test case which demonstrate pretty much worst case scenario: map 4G shmfs file, write to begin of every page pgoff of the page, remap pages in reverse order, read every page. The test creates 1 million of VMAs if emulation is in use, so I had to set vm.max_map_count to 1100000 to avoid -ENOMEM. Before: 23.3 ( +- 4.31% ) seconds After: 43.9 ( +- 0.85% ) seconds Slowdown: 1.88x I believe we can live with that. Test case: #define _GNU_SOURCE #include <assert.h> #include <stdlib.h> #include <stdio.h> #include <sys/mman.h> #define MB (1024UL * 1024) #define SIZE (4096 * MB) int main(int argc, char **argv) { unsigned long *p; long i, pass; for (pass = 0; pass < 10; pass++) { p = mmap(NULL, SIZE, PROT_READ|PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { perror("mmap"); return -1; } for (i = 0; i < SIZE / 4096; i++) p[i * 4096 / sizeof(*p)] = i; for (i = 0; i < SIZE / 4096; i++) { if (remap_file_pages(p + i * 4096 / sizeof(*p), 4096, 0, (SIZE - 4096 * (i + 1)) >> 12, 0)) { perror("remap_file_pages"); return -1; } } for (i = SIZE / 4096 - 1; i >= 0; i--) assert(p[i * 4096 / sizeof(*p)] == SIZE / 4096 - i - 1); munmap(p, SIZE); } return 0; } [akpm@linux-foundation.org: fix spello] [sasha.levin@oracle.com: initialize populate before usage] [sasha.levin@oracle.com: grab file ref to prevent race while mmaping] Signed-off-by: "Kirill A. Shutemov" <kirill@shutemov.name> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Dave Jones <davej@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Armin Rigo <arigo@tunes.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-10 23:09:46 +01:00
fput(file);
out:
up_write(&mm->mmap_sem);
if (populate)
mm_populate(ret, populate);
if (!IS_ERR_VALUE(ret))
ret = 0;
return ret;
}
/*
* this is really a simplified "do_mmap". it only handles
* anonymous maps. eventually we may be able to do some
* brk-specific accounting here.
*/
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
static int do_brk_flags(unsigned long addr, unsigned long len, unsigned long flags, struct list_head *uf)
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma, *prev;
struct rb_node **rb_link, *rb_parent;
pgoff_t pgoff = addr >> PAGE_SHIFT;
int error;
powerpc: do not make the entire heap executable On 32-bit powerpc the ELF PLT sections of binaries (built with --bss-plt, or with a toolchain which defaults to it) look like this: [17] .sbss NOBITS 0002aff8 01aff8 000014 00 WA 0 0 4 [18] .plt NOBITS 0002b00c 01aff8 000084 00 WAX 0 0 4 [19] .bss NOBITS 0002b090 01aff8 0000a4 00 WA 0 0 4 Which results in an ELF load header: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x019c70 0x00029c70 0x00029c70 0x01388 0x014c4 RWE 0x10000 This is all correct, the load region containing the PLT is marked as executable. Note that the PLT starts at 0002b00c but the file mapping ends at 0002aff8, so the PLT falls in the 0 fill section described by the load header, and after a page boundary. Unfortunately the generic ELF loader ignores the X bit in the load headers when it creates the 0 filled non-file backed mappings. It assumes all of these mappings are RW BSS sections, which is not the case for PPC. gcc/ld has an option (--secure-plt) to not do this, this is said to incur a small performance penalty. Currently, to support 32-bit binaries with PLT in BSS kernel maps *entire brk area* with executable rights for all binaries, even --secure-plt ones. Stop doing that. Teach the ELF loader to check the X bit in the relevant load header and create 0 filled anonymous mappings that are executable if the load header requests that. Test program showing the difference in /proc/$PID/maps: int main() { char buf[16*1024]; char *p = malloc(123); /* make "[heap]" mapping appear */ int fd = open("/proc/self/maps", O_RDONLY); int len = read(fd, buf, sizeof(buf)); write(1, buf, len); printf("%p\n", p); return 0; } Compiled using: gcc -mbss-plt -m32 -Os test.c -otest Unpatched ppc64 kernel: 00100000-00120000 r-xp 00000000 00:00 0 [vdso] 0fe10000-0ffd0000 r-xp 00000000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffd0000-0ffe0000 r--p 001b0000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffe0000-0fff0000 rw-p 001c0000 fd:00 67898094 /usr/lib/libc-2.17.so 10000000-10010000 r-xp 00000000 fd:00 100674505 /home/user/test 10010000-10020000 r--p 00000000 fd:00 100674505 /home/user/test 10020000-10030000 rw-p 00010000 fd:00 100674505 /home/user/test 10690000-106c0000 rwxp 00000000 00:00 0 [heap] f7f70000-f7fa0000 r-xp 00000000 fd:00 67898089 /usr/lib/ld-2.17.so f7fa0000-f7fb0000 r--p 00020000 fd:00 67898089 /usr/lib/ld-2.17.so f7fb0000-f7fc0000 rw-p 00030000 fd:00 67898089 /usr/lib/ld-2.17.so ffa90000-ffac0000 rw-p 00000000 00:00 0 [stack] 0x10690008 Patched ppc64 kernel: 00100000-00120000 r-xp 00000000 00:00 0 [vdso] 0fe10000-0ffd0000 r-xp 00000000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffd0000-0ffe0000 r--p 001b0000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffe0000-0fff0000 rw-p 001c0000 fd:00 67898094 /usr/lib/libc-2.17.so 10000000-10010000 r-xp 00000000 fd:00 100674505 /home/user/test 10010000-10020000 r--p 00000000 fd:00 100674505 /home/user/test 10020000-10030000 rw-p 00010000 fd:00 100674505 /home/user/test 10180000-101b0000 rw-p 00000000 00:00 0 [heap] ^^^^ this has changed f7c60000-f7c90000 r-xp 00000000 fd:00 67898089 /usr/lib/ld-2.17.so f7c90000-f7ca0000 r--p 00020000 fd:00 67898089 /usr/lib/ld-2.17.so f7ca0000-f7cb0000 rw-p 00030000 fd:00 67898089 /usr/lib/ld-2.17.so ff860000-ff890000 rw-p 00000000 00:00 0 [stack] 0x10180008 The patch was originally posted in 2012 by Jason Gunthorpe and apparently ignored: https://lkml.org/lkml/2012/9/30/138 Lightly run-tested. Link: http://lkml.kernel.org/r/20161215131950.23054-1-dvlasenk@redhat.com Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Tested-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Florian Weimer <fweimer@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-02-23 00:45:16 +01:00
/* Until we need other flags, refuse anything except VM_EXEC. */
if ((flags & (~VM_EXEC)) != 0)
return -EINVAL;
flags |= VM_DATA_DEFAULT_FLAGS | VM_ACCOUNT | mm->def_flags;
error = get_unmapped_area(NULL, addr, len, 0, MAP_FIXED);
if (offset_in_page(error))
return error;
error = mlock_future_check(mm, mm->def_flags, len);
if (error)
return error;
/*
* Clear old maps. this also does some error checking for us
*/
while (find_vma_links(mm, addr, addr + len, &prev, &rb_link,
&rb_parent)) {
if (do_munmap(mm, addr, len, uf))
return -ENOMEM;
}
/* Check against address space limits *after* clearing old maps... */
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
if (!may_expand_vm(mm, flags, len >> PAGE_SHIFT))
return -ENOMEM;
if (mm->map_count > sysctl_max_map_count)
return -ENOMEM;
if (security_vm_enough_memory_mm(mm, len >> PAGE_SHIFT))
return -ENOMEM;
/* Can we just expand an old private anonymous mapping? */
vma = vma_merge(mm, prev, addr, addr + len, flags,
NULL, NULL, pgoff, NULL, NULL_VM_UFFD_CTX);
if (vma)
goto out;
/*
* create a vma struct for an anonymous mapping
*/
vma = vm_area_alloc(mm);
if (!vma) {
vm_unacct_memory(len >> PAGE_SHIFT);
return -ENOMEM;
}
mm: fix vma_is_anonymous() false-positives vma_is_anonymous() relies on ->vm_ops being NULL to detect anonymous VMA. This is unreliable as ->mmap may not set ->vm_ops. False-positive vma_is_anonymous() may lead to crashes: next ffff8801ce5e7040 prev ffff8801d20eca50 mm ffff88019c1e13c0 prot 27 anon_vma ffff88019680cdd8 vm_ops 0000000000000000 pgoff 0 file ffff8801b2ec2d00 private_data 0000000000000000 flags: 0xff(read|write|exec|shared|mayread|maywrite|mayexec|mayshare) ------------[ cut here ]------------ kernel BUG at mm/memory.c:1422! invalid opcode: 0000 [#1] SMP KASAN CPU: 0 PID: 18486 Comm: syz-executor3 Not tainted 4.18.0-rc3+ #136 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:zap_pmd_range mm/memory.c:1421 [inline] RIP: 0010:zap_pud_range mm/memory.c:1466 [inline] RIP: 0010:zap_p4d_range mm/memory.c:1487 [inline] RIP: 0010:unmap_page_range+0x1c18/0x2220 mm/memory.c:1508 Call Trace: unmap_single_vma+0x1a0/0x310 mm/memory.c:1553 zap_page_range_single+0x3cc/0x580 mm/memory.c:1644 unmap_mapping_range_vma mm/memory.c:2792 [inline] unmap_mapping_range_tree mm/memory.c:2813 [inline] unmap_mapping_pages+0x3a7/0x5b0 mm/memory.c:2845 unmap_mapping_range+0x48/0x60 mm/memory.c:2880 truncate_pagecache+0x54/0x90 mm/truncate.c:800 truncate_setsize+0x70/0xb0 mm/truncate.c:826 simple_setattr+0xe9/0x110 fs/libfs.c:409 notify_change+0xf13/0x10f0 fs/attr.c:335 do_truncate+0x1ac/0x2b0 fs/open.c:63 do_sys_ftruncate+0x492/0x560 fs/open.c:205 __do_sys_ftruncate fs/open.c:215 [inline] __se_sys_ftruncate fs/open.c:213 [inline] __x64_sys_ftruncate+0x59/0x80 fs/open.c:213 do_syscall_64+0x1b9/0x820 arch/x86/entry/common.c:290 entry_SYSCALL_64_after_hwframe+0x49/0xbe Reproducer: #include <stdio.h> #include <stddef.h> #include <stdint.h> #include <stdlib.h> #include <string.h> #include <sys/types.h> #include <sys/stat.h> #include <sys/ioctl.h> #include <sys/mman.h> #include <unistd.h> #include <fcntl.h> #define KCOV_INIT_TRACE _IOR('c', 1, unsigned long) #define KCOV_ENABLE _IO('c', 100) #define KCOV_DISABLE _IO('c', 101) #define COVER_SIZE (1024<<10) #define KCOV_TRACE_PC 0 #define KCOV_TRACE_CMP 1 int main(int argc, char **argv) { int fd; unsigned long *cover; system("mount -t debugfs none /sys/kernel/debug"); fd = open("/sys/kernel/debug/kcov", O_RDWR); ioctl(fd, KCOV_INIT_TRACE, COVER_SIZE); cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long), PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0); munmap(cover, COVER_SIZE * sizeof(unsigned long)); cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long), PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0); memset(cover, 0, COVER_SIZE * sizeof(unsigned long)); ftruncate(fd, 3UL << 20); return 0; } This can be fixed by assigning anonymous VMAs own vm_ops and not relying on it being NULL. If ->mmap() failed to set ->vm_ops, mmap_region() will set it to dummy_vm_ops. This way we will have non-NULL ->vm_ops for all VMAs. Link: http://lkml.kernel.org/r/20180724121139.62570-4-kirill.shutemov@linux.intel.com Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Reported-by: syzbot+3f84280d52be9b7083cc@syzkaller.appspotmail.com Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-27 01:37:35 +02:00
vma_set_anonymous(vma);
vma->vm_start = addr;
vma->vm_end = addr + len;
vma->vm_pgoff = pgoff;
vma->vm_flags = flags;
vma->vm_page_prot = vm_get_page_prot(flags);
vma_link(mm, vma, prev, rb_link, rb_parent);
out:
perf_event_mmap(vma);
mm->total_vm += len >> PAGE_SHIFT;
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
mm->data_vm += len >> PAGE_SHIFT;
if (flags & VM_LOCKED)
mm->locked_vm += (len >> PAGE_SHIFT);
vma->vm_flags |= VM_SOFTDIRTY;
return 0;
}
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
int vm_brk_flags(unsigned long addr, unsigned long request, unsigned long flags)
{
struct mm_struct *mm = current->mm;
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
unsigned long len;
int ret;
bool populate;
LIST_HEAD(uf);
mm: do not bug_on on incorrect length in __mm_populate() syzbot has noticed that a specially crafted library can easily hit VM_BUG_ON in __mm_populate kernel BUG at mm/gup.c:1242! invalid opcode: 0000 [#1] SMP CPU: 2 PID: 9667 Comm: a.out Not tainted 4.18.0-rc3 #644 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 RIP: 0010:__mm_populate+0x1e2/0x1f0 Code: 55 d0 65 48 33 14 25 28 00 00 00 89 d8 75 21 48 83 c4 20 5b 41 5c 41 5d 41 5e 41 5f 5d c3 e8 75 18 f1 ff 0f 0b e8 6e 18 f1 ff <0f> 0b 31 db eb c9 e8 93 06 e0 ff 0f 1f 00 55 48 89 e5 53 48 89 fb Call Trace: vm_brk_flags+0xc3/0x100 vm_brk+0x1f/0x30 load_elf_library+0x281/0x2e0 __ia32_sys_uselib+0x170/0x1e0 do_fast_syscall_32+0xca/0x420 entry_SYSENTER_compat+0x70/0x7f The reason is that the length of the new brk is not page aligned when we try to populate the it. There is no reason to bug on that though. do_brk_flags already aligns the length properly so the mapping is expanded as it should. All we need is to tell mm_populate about it. Besides that there is absolutely no reason to to bug_on in the first place. The worst thing that could happen is that the last page wouldn't get populated and that is far from putting system into an inconsistent state. Fix the issue by moving the length sanitization code from do_brk_flags up to vm_brk_flags. The only other caller of do_brk_flags is brk syscall entry and it makes sure to provide the proper length so t here is no need for sanitation and so we can use do_brk_flags without it. Also remove the bogus BUG_ONs. [osalvador@techadventures.net: fix up vm_brk_flags s@request@len@] Link: http://lkml.kernel.org/r/20180706090217.GI32658@dhcp22.suse.cz Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: syzbot <syzbot+5dcb560fe12aa5091c06@syzkaller.appspotmail.com> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: Zi Yan <zi.yan@cs.rutgers.edu> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-07-14 01:59:20 +02:00
len = PAGE_ALIGN(request);
if (len < request)
return -ENOMEM;
if (!len)
return 0;
if (down_write_killable(&mm->mmap_sem))
return -EINTR;
ret = do_brk_flags(addr, len, flags, &uf);
populate = ((mm->def_flags & VM_LOCKED) != 0);
up_write(&mm->mmap_sem);
userfaultfd_unmap_complete(mm, &uf);
if (populate && !ret)
mm_populate(addr, len);
return ret;
}
powerpc: do not make the entire heap executable On 32-bit powerpc the ELF PLT sections of binaries (built with --bss-plt, or with a toolchain which defaults to it) look like this: [17] .sbss NOBITS 0002aff8 01aff8 000014 00 WA 0 0 4 [18] .plt NOBITS 0002b00c 01aff8 000084 00 WAX 0 0 4 [19] .bss NOBITS 0002b090 01aff8 0000a4 00 WA 0 0 4 Which results in an ELF load header: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x019c70 0x00029c70 0x00029c70 0x01388 0x014c4 RWE 0x10000 This is all correct, the load region containing the PLT is marked as executable. Note that the PLT starts at 0002b00c but the file mapping ends at 0002aff8, so the PLT falls in the 0 fill section described by the load header, and after a page boundary. Unfortunately the generic ELF loader ignores the X bit in the load headers when it creates the 0 filled non-file backed mappings. It assumes all of these mappings are RW BSS sections, which is not the case for PPC. gcc/ld has an option (--secure-plt) to not do this, this is said to incur a small performance penalty. Currently, to support 32-bit binaries with PLT in BSS kernel maps *entire brk area* with executable rights for all binaries, even --secure-plt ones. Stop doing that. Teach the ELF loader to check the X bit in the relevant load header and create 0 filled anonymous mappings that are executable if the load header requests that. Test program showing the difference in /proc/$PID/maps: int main() { char buf[16*1024]; char *p = malloc(123); /* make "[heap]" mapping appear */ int fd = open("/proc/self/maps", O_RDONLY); int len = read(fd, buf, sizeof(buf)); write(1, buf, len); printf("%p\n", p); return 0; } Compiled using: gcc -mbss-plt -m32 -Os test.c -otest Unpatched ppc64 kernel: 00100000-00120000 r-xp 00000000 00:00 0 [vdso] 0fe10000-0ffd0000 r-xp 00000000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffd0000-0ffe0000 r--p 001b0000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffe0000-0fff0000 rw-p 001c0000 fd:00 67898094 /usr/lib/libc-2.17.so 10000000-10010000 r-xp 00000000 fd:00 100674505 /home/user/test 10010000-10020000 r--p 00000000 fd:00 100674505 /home/user/test 10020000-10030000 rw-p 00010000 fd:00 100674505 /home/user/test 10690000-106c0000 rwxp 00000000 00:00 0 [heap] f7f70000-f7fa0000 r-xp 00000000 fd:00 67898089 /usr/lib/ld-2.17.so f7fa0000-f7fb0000 r--p 00020000 fd:00 67898089 /usr/lib/ld-2.17.so f7fb0000-f7fc0000 rw-p 00030000 fd:00 67898089 /usr/lib/ld-2.17.so ffa90000-ffac0000 rw-p 00000000 00:00 0 [stack] 0x10690008 Patched ppc64 kernel: 00100000-00120000 r-xp 00000000 00:00 0 [vdso] 0fe10000-0ffd0000 r-xp 00000000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffd0000-0ffe0000 r--p 001b0000 fd:00 67898094 /usr/lib/libc-2.17.so 0ffe0000-0fff0000 rw-p 001c0000 fd:00 67898094 /usr/lib/libc-2.17.so 10000000-10010000 r-xp 00000000 fd:00 100674505 /home/user/test 10010000-10020000 r--p 00000000 fd:00 100674505 /home/user/test 10020000-10030000 rw-p 00010000 fd:00 100674505 /home/user/test 10180000-101b0000 rw-p 00000000 00:00 0 [heap] ^^^^ this has changed f7c60000-f7c90000 r-xp 00000000 fd:00 67898089 /usr/lib/ld-2.17.so f7c90000-f7ca0000 r--p 00020000 fd:00 67898089 /usr/lib/ld-2.17.so f7ca0000-f7cb0000 rw-p 00030000 fd:00 67898089 /usr/lib/ld-2.17.so ff860000-ff890000 rw-p 00000000 00:00 0 [stack] 0x10180008 The patch was originally posted in 2012 by Jason Gunthorpe and apparently ignored: https://lkml.org/lkml/2012/9/30/138 Lightly run-tested. Link: http://lkml.kernel.org/r/20161215131950.23054-1-dvlasenk@redhat.com Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Tested-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Florian Weimer <fweimer@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-02-23 00:45:16 +01:00
EXPORT_SYMBOL(vm_brk_flags);
int vm_brk(unsigned long addr, unsigned long len)
{
return vm_brk_flags(addr, len, 0);
}
EXPORT_SYMBOL(vm_brk);
/* Release all mmaps. */
void exit_mmap(struct mm_struct *mm)
{
mm: mmu_gather rework Rework the existing mmu_gather infrastructure. The direct purpose of these patches was to allow preemptible mmu_gather, but even without that I think these patches provide an improvement to the status quo. The first 9 patches rework the mmu_gather infrastructure. For review purpose I've split them into generic and per-arch patches with the last of those a generic cleanup. The next patch provides generic RCU page-table freeing, and the followup is a patch converting s390 to use this. I've also got 4 patches from DaveM lined up (not included in this series) that uses this to implement gup_fast() for sparc64. Then there is one patch that extends the generic mmu_gather batching. After that follow the mm preemptibility patches, these make part of the mm a lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes which together with the mmu_gather rework makes mmu_gather preemptible as well. Making i_mmap_lock a mutex also enables a clean-up of the truncate code. This also allows for preemptible mmu_notifiers, something that XPMEM I think wants. Furthermore, it removes the new and universially detested unmap_mutex. This patch: Remove the first obstacle towards a fully preemptible mmu_gather. The current scheme assumes mmu_gather is always done with preemption disabled and uses per-cpu storage for the page batches. Change this to try and allocate a page for batching and in case of failure, use a small on-stack array to make some progress. Preemptible mmu_gather is desired in general and usable once i_mmap_lock becomes a mutex. Doing it before the mutex conversion saves us from having to rework the code by moving the mmu_gather bits inside the pte_lock. Also avoid flushing the tlb batches from under the pte lock, this is useful even without the i_mmap_lock conversion as it significantly reduces pte lock hold times. [akpm@linux-foundation.org: fix comment tpyo] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Miller <davem@davemloft.net> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Jeff Dike <jdike@addtoit.com> Cc: Richard Weinberger <richard@nod.at> Cc: Tony Luck <tony.luck@intel.com> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Hugh Dickins <hughd@google.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Namhyung Kim <namhyung@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-05-25 02:11:45 +02:00
struct mmu_gather tlb;
struct vm_area_struct *vma;
unsigned long nr_accounted = 0;
/* mm's last user has gone, and its about to be pulled down */
mmu-notifiers: core With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages. There are secondary MMUs (with secondary sptes and secondary tlbs) too. sptes in the kvm case are shadow pagetables, but when I say spte in mmu-notifier context, I mean "secondary pte". In GRU case there's no actual secondary pte and there's only a secondary tlb because the GRU secondary MMU has no knowledge about sptes and every secondary tlb miss event in the MMU always generates a page fault that has to be resolved by the CPU (this is not the case of KVM where the a secondary tlb miss will walk sptes in hardware and it will refill the secondary tlb transparently to software if the corresponding spte is present). The same way zap_page_range has to invalidate the pte before freeing the page, the spte (and secondary tlb) must also be invalidated before any page is freed and reused. Currently we take a page_count pin on every page mapped by sptes, but that means the pages can't be swapped whenever they're mapped by any spte because they're part of the guest working set. Furthermore a spte unmap event can immediately lead to a page to be freed when the pin is released (so requiring the same complex and relatively slow tlb_gather smp safe logic we have in zap_page_range and that can be avoided completely if the spte unmap event doesn't require an unpin of the page previously mapped in the secondary MMU). The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know when the VM is swapping or freeing or doing anything on the primary MMU so that the secondary MMU code can drop sptes before the pages are freed, avoiding all page pinning and allowing 100% reliable swapping of guest physical address space. Furthermore it avoids the code that teardown the mappings of the secondary MMU, to implement a logic like tlb_gather in zap_page_range that would require many IPI to flush other cpu tlbs, for each fixed number of spte unmapped. To make an example: if what happens on the primary MMU is a protection downgrade (from writeable to wrprotect) the secondary MMU mappings will be invalidated, and the next secondary-mmu-page-fault will call get_user_pages and trigger a do_wp_page through get_user_pages if it called get_user_pages with write=1, and it'll re-establishing an updated spte or secondary-tlb-mapping on the copied page. Or it will setup a readonly spte or readonly tlb mapping if it's a guest-read, if it calls get_user_pages with write=0. This is just an example. This allows to map any page pointed by any pte (and in turn visible in the primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an full MMU with both sptes and secondary-tlb like the shadow-pagetable layer with kvm), or a remote DMA in software like XPMEM (hence needing of schedule in XPMEM code to send the invalidate to the remote node, while no need to schedule in kvm/gru as it's an immediate event like invalidating primary-mmu pte). At least for KVM without this patch it's impossible to swap guests reliably. And having this feature and removing the page pin allows several other optimizations that simplify life considerably. Dependencies: 1) mm_take_all_locks() to register the mmu notifier when the whole VM isn't doing anything with "mm". This allows mmu notifier users to keep track if the VM is in the middle of the invalidate_range_begin/end critical section with an atomic counter incraese in range_begin and decreased in range_end. No secondary MMU page fault is allowed to map any spte or secondary tlb reference, while the VM is in the middle of range_begin/end as any page returned by get_user_pages in that critical section could later immediately be freed without any further ->invalidate_page notification (invalidate_range_begin/end works on ranges and ->invalidate_page isn't called immediately before freeing the page). To stop all page freeing and pagetable overwrites the mmap_sem must be taken in write mode and all other anon_vma/i_mmap locks must be taken too. 2) It'd be a waste to add branches in the VM if nobody could possibly run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of mmu notifiers, but this already allows to compile a KVM external module against a kernel with mmu notifiers enabled and from the next pull from kvm.git we'll start using them. And GRU/XPMEM will also be able to continue the development by enabling KVM=m in their config, until they submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n). This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM are all =n. The mmu_notifier_register call can fail because mm_take_all_locks may be interrupted by a signal and return -EINTR. Because mmu_notifier_reigster is used when a driver startup, a failure can be gracefully handled. Here an example of the change applied to kvm to register the mmu notifiers. Usually when a driver startups other allocations are required anyway and -ENOMEM failure paths exists already. struct kvm *kvm_arch_create_vm(void) { struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL); + int err; if (!kvm) return ERR_PTR(-ENOMEM); INIT_LIST_HEAD(&kvm->arch.active_mmu_pages); + kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops; + err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm); + if (err) { + kfree(kvm); + return ERR_PTR(err); + } + return kvm; } mmu_notifier_unregister returns void and it's reliable. The patch also adds a few needed but missing includes that would prevent kernel to compile after these changes on non-x86 archs (x86 didn't need them by luck). [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: fix mm/filemap_xip.c build] [akpm@linux-foundation.org: fix mm/mmu_notifier.c build] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:29 +02:00
mmu_notifier_release(mm);
mm, oom: fix concurrent munlock and oom reaper unmap, v3 Since exit_mmap() is done without the protection of mm->mmap_sem, it is possible for the oom reaper to concurrently operate on an mm until MMF_OOM_SKIP is set. This allows munlock_vma_pages_all() to concurrently run while the oom reaper is operating on a vma. Since munlock_vma_pages_range() depends on clearing VM_LOCKED from vm_flags before actually doing the munlock to determine if any other vmas are locking the same memory, the check for VM_LOCKED in the oom reaper is racy. This is especially noticeable on architectures such as powerpc where clearing a huge pmd requires serialize_against_pte_lookup(). If the pmd is zapped by the oom reaper during follow_page_mask() after the check for pmd_none() is bypassed, this ends up deferencing a NULL ptl or a kernel oops. Fix this by manually freeing all possible memory from the mm before doing the munlock and then setting MMF_OOM_SKIP. The oom reaper can not run on the mm anymore so the munlock is safe to do in exit_mmap(). It also matches the logic that the oom reaper currently uses for determining when to set MMF_OOM_SKIP itself, so there's no new risk of excessive oom killing. This issue fixes CVE-2018-1000200. Link: http://lkml.kernel.org/r/alpine.DEB.2.21.1804241526320.238665@chino.kir.corp.google.com Fixes: 212925802454 ("mm: oom: let oom_reap_task and exit_mmap run concurrently") Signed-off-by: David Rientjes <rientjes@google.com> Suggested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: <stable@vger.kernel.org> [4.14+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-12 01:02:04 +02:00
if (unlikely(mm_is_oom_victim(mm))) {
/*
* Manually reap the mm to free as much memory as possible.
* Then, as the oom reaper does, set MMF_OOM_SKIP to disregard
* this mm from further consideration. Taking mm->mmap_sem for
* write after setting MMF_OOM_SKIP will guarantee that the oom
* reaper will not run on this mm again after mmap_sem is
* dropped.
*
* Nothing can be holding mm->mmap_sem here and the above call
* to mmu_notifier_release(mm) ensures mmu notifier callbacks in
* __oom_reap_task_mm() will not block.
*
* This needs to be done before calling munlock_vma_pages_all(),
* which clears VM_LOCKED, otherwise the oom reaper cannot
* reliably test it.
*/
mm, oom: distinguish blockable mode for mmu notifiers There are several blockable mmu notifiers which might sleep in mmu_notifier_invalidate_range_start and that is a problem for the oom_reaper because it needs to guarantee a forward progress so it cannot depend on any sleepable locks. Currently we simply back off and mark an oom victim with blockable mmu notifiers as done after a short sleep. That can result in selecting a new oom victim prematurely because the previous one still hasn't torn its memory down yet. We can do much better though. Even if mmu notifiers use sleepable locks there is no reason to automatically assume those locks are held. Moreover majority of notifiers only care about a portion of the address space and there is absolutely zero reason to fail when we are unmapping an unrelated range. Many notifiers do really block and wait for HW which is harder to handle and we have to bail out though. This patch handles the low hanging fruit. __mmu_notifier_invalidate_range_start gets a blockable flag and callbacks are not allowed to sleep if the flag is set to false. This is achieved by using trylock instead of the sleepable lock for most callbacks and continue as long as we do not block down the call chain. I think we can improve that even further because there is a common pattern to do a range lookup first and then do something about that. The first part can be done without a sleeping lock in most cases AFAICS. The oom_reaper end then simply retries if there is at least one notifier which couldn't make any progress in !blockable mode. A retry loop is already implemented to wait for the mmap_sem and this is basically the same thing. The simplest way for driver developers to test this code path is to wrap userspace code which uses these notifiers into a memcg and set the hard limit to hit the oom. This can be done e.g. after the test faults in all the mmu notifier managed memory and set the hard limit to something really small. Then we are looking for a proper process tear down. [akpm@linux-foundation.org: coding style fixes] [akpm@linux-foundation.org: minor code simplification] Link: http://lkml.kernel.org/r/20180716115058.5559-1-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Acked-by: Christian König <christian.koenig@amd.com> # AMD notifiers Acked-by: Leon Romanovsky <leonro@mellanox.com> # mlx and umem_odp Reported-by: David Rientjes <rientjes@google.com> Cc: "David (ChunMing) Zhou" <David1.Zhou@amd.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: David Airlie <airlied@linux.ie> Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Doug Ledford <dledford@redhat.com> Cc: Jason Gunthorpe <jgg@ziepe.ca> Cc: Mike Marciniszyn <mike.marciniszyn@intel.com> Cc: Dennis Dalessandro <dennis.dalessandro@intel.com> Cc: Sudeep Dutt <sudeep.dutt@intel.com> Cc: Ashutosh Dixit <ashutosh.dixit@intel.com> Cc: Dimitri Sivanich <sivanich@sgi.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: Juergen Gross <jgross@suse.com> Cc: "Jérôme Glisse" <jglisse@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Felix Kuehling <felix.kuehling@amd.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 06:52:33 +02:00
(void)__oom_reap_task_mm(mm);
mm, oom: fix concurrent munlock and oom reaper unmap, v3 Since exit_mmap() is done without the protection of mm->mmap_sem, it is possible for the oom reaper to concurrently operate on an mm until MMF_OOM_SKIP is set. This allows munlock_vma_pages_all() to concurrently run while the oom reaper is operating on a vma. Since munlock_vma_pages_range() depends on clearing VM_LOCKED from vm_flags before actually doing the munlock to determine if any other vmas are locking the same memory, the check for VM_LOCKED in the oom reaper is racy. This is especially noticeable on architectures such as powerpc where clearing a huge pmd requires serialize_against_pte_lookup(). If the pmd is zapped by the oom reaper during follow_page_mask() after the check for pmd_none() is bypassed, this ends up deferencing a NULL ptl or a kernel oops. Fix this by manually freeing all possible memory from the mm before doing the munlock and then setting MMF_OOM_SKIP. The oom reaper can not run on the mm anymore so the munlock is safe to do in exit_mmap(). It also matches the logic that the oom reaper currently uses for determining when to set MMF_OOM_SKIP itself, so there's no new risk of excessive oom killing. This issue fixes CVE-2018-1000200. Link: http://lkml.kernel.org/r/alpine.DEB.2.21.1804241526320.238665@chino.kir.corp.google.com Fixes: 212925802454 ("mm: oom: let oom_reap_task and exit_mmap run concurrently") Signed-off-by: David Rientjes <rientjes@google.com> Suggested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: <stable@vger.kernel.org> [4.14+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-05-12 01:02:04 +02:00
set_bit(MMF_OOM_SKIP, &mm->flags);
down_write(&mm->mmap_sem);
up_write(&mm->mmap_sem);
}
if (mm->locked_vm) {
vma = mm->mmap;
while (vma) {
if (vma->vm_flags & VM_LOCKED)
munlock_vma_pages_all(vma);
vma = vma->vm_next;
}
}
mm: rearrange exit_mmap() to unlock before arch_exit_mmap Christophe Saout reported [in precursor to: http://marc.info/?l=linux-kernel&m=123209902707347&w=4]: > Note that I also some a different issue with CONFIG_UNEVICTABLE_LRU. > Seems like Xen tears down current->mm early on process termination, so > that __get_user_pages in exit_mmap causes nasty messages when the > process had any mlocked pages. (in fact, it somehow manages to get into > the swapping code and produces a null pointer dereference trying to get > a swap token) Jeremy explained: Yes. In the normal case under Xen, an in-use pagetable is "pinned", meaning that it is RO to the kernel, and all updates must go via hypercall (or writes are trapped and emulated, which is much the same thing). An unpinned pagetable is not currently in use by any process, and can be directly accessed as normal RW pages. As an optimisation at process exit time, we unpin the pagetable as early as possible (switching the process to init_mm), so that all the normal pagetable teardown can happen with direct memory accesses. This happens in exit_mmap() -> arch_exit_mmap(). The munlocking happens a few lines below. The obvious thing to do would be to move arch_exit_mmap() to below the munlock code, but I think we'd want to call it even if mm->mmap is NULL, just to be on the safe side. Thus, this patch: exit_mmap() needs to unlock any locked vmas before calling arch_exit_mmap, as the latter may switch the current mm to init_mm, which would cause the former to fail. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Christophe Saout <christophe@saout.de> Cc: Keir Fraser <keir.fraser@eu.citrix.com> Cc: Christophe Saout <christophe@saout.de> Cc: Alex Williamson <alex.williamson@hp.com> Cc: <stable@kernel.org> [2.6.28.x] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-11 22:04:41 +01:00
arch_exit_mmap(mm);
vma = mm->mmap;
mm: rearrange exit_mmap() to unlock before arch_exit_mmap Christophe Saout reported [in precursor to: http://marc.info/?l=linux-kernel&m=123209902707347&w=4]: > Note that I also some a different issue with CONFIG_UNEVICTABLE_LRU. > Seems like Xen tears down current->mm early on process termination, so > that __get_user_pages in exit_mmap causes nasty messages when the > process had any mlocked pages. (in fact, it somehow manages to get into > the swapping code and produces a null pointer dereference trying to get > a swap token) Jeremy explained: Yes. In the normal case under Xen, an in-use pagetable is "pinned", meaning that it is RO to the kernel, and all updates must go via hypercall (or writes are trapped and emulated, which is much the same thing). An unpinned pagetable is not currently in use by any process, and can be directly accessed as normal RW pages. As an optimisation at process exit time, we unpin the pagetable as early as possible (switching the process to init_mm), so that all the normal pagetable teardown can happen with direct memory accesses. This happens in exit_mmap() -> arch_exit_mmap(). The munlocking happens a few lines below. The obvious thing to do would be to move arch_exit_mmap() to below the munlock code, but I think we'd want to call it even if mm->mmap is NULL, just to be on the safe side. Thus, this patch: exit_mmap() needs to unlock any locked vmas before calling arch_exit_mmap, as the latter may switch the current mm to init_mm, which would cause the former to fail. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Christophe Saout <christophe@saout.de> Cc: Keir Fraser <keir.fraser@eu.citrix.com> Cc: Christophe Saout <christophe@saout.de> Cc: Alex Williamson <alex.williamson@hp.com> Cc: <stable@kernel.org> [2.6.28.x] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-11 22:04:41 +01:00
if (!vma) /* Can happen if dup_mmap() received an OOM */
return;
lru_add_drain();
flush_cache_mm(mm);
Fix TLB gather virtual address range invalidation corner cases Ben Tebulin reported: "Since v3.7.2 on two independent machines a very specific Git repository fails in 9/10 cases on git-fsck due to an SHA1/memory failures. This only occurs on a very specific repository and can be reproduced stably on two independent laptops. Git mailing list ran out of ideas and for me this looks like some very exotic kernel issue" and bisected the failure to the backport of commit 53a59fc67f97 ("mm: limit mmu_gather batching to fix soft lockups on !CONFIG_PREEMPT"). That commit itself is not actually buggy, but what it does is to make it much more likely to hit the partial TLB invalidation case, since it introduces a new case in tlb_next_batch() that previously only ever happened when running out of memory. The real bug is that the TLB gather virtual memory range setup is subtly buggered. It was introduced in commit 597e1c3580b7 ("mm/mmu_gather: enable tlb flush range in generic mmu_gather"), and the range handling was already fixed at least once in commit e6c495a96ce0 ("mm: fix the TLB range flushed when __tlb_remove_page() runs out of slots"), but that fix was not complete. The problem with the TLB gather virtual address range is that it isn't set up by the initial tlb_gather_mmu() initialization (which didn't get the TLB range information), but it is set up ad-hoc later by the functions that actually flush the TLB. And so any such case that forgot to update the TLB range entries would potentially miss TLB invalidates. Rather than try to figure out exactly which particular ad-hoc range setup was missing (I personally suspect it's the hugetlb case in zap_huge_pmd(), which didn't have the same logic as zap_pte_range() did), this patch just gets rid of the problem at the source: make the TLB range information available to tlb_gather_mmu(), and initialize it when initializing all the other tlb gather fields. This makes the patch larger, but conceptually much simpler. And the end result is much more understandable; even if you want to play games with partial ranges when invalidating the TLB contents in chunks, now the range information is always there, and anybody who doesn't want to bother with it won't introduce subtle bugs. Ben verified that this fixes his problem. Reported-bisected-and-tested-by: Ben Tebulin <tebulin@googlemail.com> Build-testing-by: Stephen Rothwell <sfr@canb.auug.org.au> Build-testing-by: Richard Weinberger <richard.weinberger@gmail.com> Reviewed-by: Michal Hocko <mhocko@suse.cz> Acked-by: Peter Zijlstra <peterz@infradead.org> Cc: stable@vger.kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-08-15 20:42:25 +02:00
tlb_gather_mmu(&tlb, mm, 0, -1);
/* update_hiwater_rss(mm) here? but nobody should be looking */
[PATCH] freepgt: free_pgtables use vma list Recent woes with some arches needing their own pgd_addr_end macro; and 4-level clear_page_range regression since 2.6.10's clear_page_tables; and its long-standing well-known inefficiency in searching throughout the higher-level page tables for those few entries to clear and free: all can be blamed on ignoring the list of vmas when we free page tables. Replace exit_mmap's clear_page_range of the total user address space by free_pgtables operating on the mm's vma list; unmap_region use it in the same way, giving floor and ceiling beyond which it may not free tables. This brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled, in which case latency fixes spoil unmap_vmas throughput). Beware: the do_mmap_pgoff driver failure case must now use unmap_region instead of zap_page_range, since a page table might have been allocated, and can only be freed while it is touched by some vma. Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted from the clear_page_range levels. (Most of free_pgtables' old code was actually for a non-existent case, prev not properly set up, dating from before hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we might want to add latency lockdrops later; but no attempt to do so yet, going by vma should itself reduce latency. But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful examination: put that off until a later patch of the series. What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma? And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that we need to do more than is done here - every PMD_SIZE ever occupied will be flushed, do we really have to flush every PGDIR_SIZE ever partially occupied? A shame to complicate it unnecessarily. Special thanks to David Miller for time spent repairing my ceilings. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 22:29:15 +02:00
/* Use -1 here to ensure all VMAs in the mm are unmapped */
unmap_vmas(&tlb, vma, 0, -1);
free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, USER_PGTABLES_CEILING);
tlb_finish_mmu(&tlb, 0, -1);
/*
[PATCH] mm: unlink vma before pagetables In most places the descent from pgd to pud to pmd to pte holds mmap_sem (exclusively or not), which ensures that free_pgtables cannot be freeing page tables from any level at the same time. But truncation and reverse mapping descend without mmap_sem. No problem: just make sure that a vma is unlinked from its prio_tree (or nonlinear list) and from its anon_vma list, after zapping the vma, but before freeing its page tables. Then neither vmtruncate nor rmap can reach that vma whose page tables are now volatile (nor do they need to reach it, since all its page entries have been zapped by this stage). The i_mmap_lock and anon_vma->lock already serialize this correctly; but the locking hierarchy is such that we cannot take them while holding page_table_lock. Well, we're trying to push that down anyway. So in this patch, move anon_vma_unlink and unlink_file_vma into free_pgtables, at the same time as moving page_table_lock around calls to unmap_vmas. tlb_gather_mmu and tlb_finish_mmu then fall outside the page_table_lock, but we made them preempt_disable and preempt_enable earlier; and a long source audit of all the architectures has shown no problem with removing page_table_lock from them. free_pgtables doesn't need page_table_lock for itself, nor for what it calls; tlb->mm->nr_ptes is usually protected by page_table_lock, but partly by non-exclusive mmap_sem - here it's decremented with exclusive mmap_sem, or mm_users 0. update_hiwater_rss and vm_unacct_memory don't need page_table_lock either. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 02:16:29 +01:00
* Walk the list again, actually closing and freeing it,
* with preemption enabled, without holding any MM locks.
*/
while (vma) {
if (vma->vm_flags & VM_ACCOUNT)
nr_accounted += vma_pages(vma);
vma = remove_vma(vma);
}
vm_unacct_memory(nr_accounted);
}
/* Insert vm structure into process list sorted by address
* and into the inode's i_mmap tree. If vm_file is non-NULL
* then i_mmap_rwsem is taken here.
*/
int insert_vm_struct(struct mm_struct *mm, struct vm_area_struct *vma)
{
struct vm_area_struct *prev;
struct rb_node **rb_link, *rb_parent;
if (find_vma_links(mm, vma->vm_start, vma->vm_end,
&prev, &rb_link, &rb_parent))
return -ENOMEM;
if ((vma->vm_flags & VM_ACCOUNT) &&
security_vm_enough_memory_mm(mm, vma_pages(vma)))
return -ENOMEM;
/*
* The vm_pgoff of a purely anonymous vma should be irrelevant
* until its first write fault, when page's anon_vma and index
* are set. But now set the vm_pgoff it will almost certainly
* end up with (unless mremap moves it elsewhere before that
* first wfault), so /proc/pid/maps tells a consistent story.
*
* By setting it to reflect the virtual start address of the
* vma, merges and splits can happen in a seamless way, just
* using the existing file pgoff checks and manipulations.
* Similarly in do_mmap_pgoff and in do_brk.
*/
mmap: fix the usage of ->vm_pgoff in special_mapping paths Test-case: #include <stdio.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/mman.h> #include <assert.h> void *find_vdso_vaddr(void) { FILE *perl; char buf[32] = {}; perl = popen("perl -e 'open STDIN,qq|/proc/@{[getppid]}/maps|;" "/^(.*?)-.*vdso/ && print hex $1 while <>'", "r"); fread(buf, sizeof(buf), 1, perl); fclose(perl); return (void *)atol(buf); } #define PAGE_SIZE 4096 int main(void) { void *vdso = find_vdso_vaddr(); assert(vdso); // of course they should differ, and they do so far printf("vdso pages differ: %d\n", !!memcmp(vdso, vdso + PAGE_SIZE, PAGE_SIZE)); // split into 2 vma's assert(mprotect(vdso, PAGE_SIZE, PROT_READ) == 0); // force another fault on the next check assert(madvise(vdso, 2 * PAGE_SIZE, MADV_DONTNEED) == 0); // now they no longer differ, the 2nd vm_pgoff is wrong printf("vdso pages differ: %d\n", !!memcmp(vdso, vdso + PAGE_SIZE, PAGE_SIZE)); return 0; } Output: vdso pages differ: 1 vdso pages differ: 0 This is because split_vma() correctly updates ->vm_pgoff, but the logic in insert_vm_struct() and special_mapping_fault() is absolutely broken, so the fault at vdso + PAGE_SIZE return the 1st page. The same happens if you simply unmap the 1st page. special_mapping_fault() does: pgoff = vmf->pgoff - vma->vm_pgoff; and this is _only_ correct if vma->vm_start mmaps the first page from ->vm_private_data array. vdso or any other user of install_special_mapping() is not anonymous, it has the "backing storage" even if it is just the array of pages. So we actually need to make vm_pgoff work as an offset in this array. Note: this also allows to fix another problem: currently gdb can't access "[vvar]" memory because in this case special_mapping_fault() doesn't work. Now that we can use ->vm_pgoff we can implement ->access() and fix this. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Hugh Dickins <hughd@google.com> Cc: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 23:58:31 +02:00
if (vma_is_anonymous(vma)) {
BUG_ON(vma->anon_vma);
vma->vm_pgoff = vma->vm_start >> PAGE_SHIFT;
}
uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints Add uprobes support to the core kernel, with x86 support. This commit adds the kernel facilities, the actual uprobes user-space ABI and perf probe support comes in later commits. General design: Uprobes are maintained in an rb-tree indexed by inode and offset (the offset here is from the start of the mapping). For a unique (inode, offset) tuple, there can be at most one uprobe in the rb-tree. Since the (inode, offset) tuple identifies a unique uprobe, more than one user may be interested in the same uprobe. This provides the ability to connect multiple 'consumers' to the same uprobe. Each consumer defines a handler and a filter (optional). The 'handler' is run every time the uprobe is hit, if it matches the 'filter' criteria. The first consumer of a uprobe causes the breakpoint to be inserted at the specified address and subsequent consumers are appended to this list. On subsequent probes, the consumer gets appended to the existing list of consumers. The breakpoint is removed when the last consumer unregisters. For all other unregisterations, the consumer is removed from the list of consumers. Given a inode, we get a list of the mms that have mapped the inode. Do the actual registration if mm maps the page where a probe needs to be inserted/removed. We use a temporary list to walk through the vmas that map the inode. - The number of maps that map the inode, is not known before we walk the rmap and keeps changing. - extending vm_area_struct wasn't recommended, it's a size-critical data structure. - There can be more than one maps of the inode in the same mm. We add callbacks to the mmap methods to keep an eye on text vmas that are of interest to uprobes. When a vma of interest is mapped, we insert the breakpoint at the right address. Uprobe works by replacing the instruction at the address defined by (inode, offset) with the arch specific breakpoint instruction. We save a copy of the original instruction at the uprobed address. This is needed for: a. executing the instruction out-of-line (xol). b. instruction analysis for any subsequent fixups. c. restoring the instruction back when the uprobe is unregistered. We insert or delete a breakpoint instruction, and this breakpoint instruction is assumed to be the smallest instruction available on the platform. For fixed size instruction platforms this is trivially true, for variable size instruction platforms the breakpoint instruction is typically the smallest (often a single byte). Writing the instruction is done by COWing the page and changing the instruction during the copy, this even though most platforms allow atomic writes of the breakpoint instruction. This also mirrors the behaviour of a ptrace() memory write to a PRIVATE file map. The core worker is derived from KSM's replace_page() logic. In essence, similar to KSM: a. allocate a new page and copy over contents of the page that has the uprobed vaddr b. modify the copy and insert the breakpoint at the required address c. switch the original page with the copy containing the breakpoint d. flush page tables. replace_page() is being replicated here because of some minor changes in the type of pages and also because Hugh Dickins had plans to improve replace_page() for KSM specific work. Instruction analysis on x86 is based on instruction decoder and determines if an instruction can be probed and determines the necessary fixups after singlestep. Instruction analysis is done at probe insertion time so that we avoid having to repeat the same analysis every time a probe is hit. A lot of code here is due to the improvement/suggestions/inputs from Peter Zijlstra. Changelog: (v10): - Add code to clear REX.B prefix as suggested by Denys Vlasenko and Masami Hiramatsu. (v9): - Use insn_offset_modrm as suggested by Masami Hiramatsu. (v7): Handle comments from Peter Zijlstra: - Dont take reference to inode. (expect inode to uprobe_register to be sane). - Use PTR_ERR to set the return value. - No need to take reference to inode. - use PTR_ERR to return error value. - register and uprobe_unregister share code. (v5): - Modified del_consumer as per comments from Peter. - Drop reference to inode before dropping reference to uprobe. - Use i_size_read(inode) instead of inode->i_size. - Ensure uprobe->consumers is NULL, before __uprobe_unregister() is called. - Includes errno.h as recommended by Stephen Rothwell to fix a build issue on sparc defconfig - Remove restrictions while unregistering. - Earlier code leaked inode references under some conditions while registering/unregistering. - Continue the vma-rmap walk even if the intermediate vma doesnt meet the requirements. - Validate the vma found by find_vma before inserting/removing the breakpoint - Call del_consumer under mutex_lock. - Use hash locks. - Handle mremap. - Introduce find_least_offset_node() instead of close match logic in find_uprobe - Uprobes no more depends on MM_OWNER; No reference to task_structs while inserting/removing a probe. - Uses read_mapping_page instead of grab_cache_page so that the pages have valid content. - pass NULL to get_user_pages for the task parameter. - call SetPageUptodate on the new page allocated in write_opcode. - fix leaking a reference to the new page under certain conditions. - Include Instruction Decoder if Uprobes gets defined. - Remove const attributes for instruction prefix arrays. - Uses mm_context to know if the application is 32 bit. Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Also-written-by: Jim Keniston <jkenisto@us.ibm.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Roland McGrath <roland@hack.frob.com> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Anton Arapov <anton@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Denys Vlasenko <vda.linux@googlemail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linux-mm <linux-mm@kvack.org> Link: http://lkml.kernel.org/r/20120209092642.GE16600@linux.vnet.ibm.com [ Made various small edits to the commit log ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
2012-02-09 10:26:42 +01:00
vma_link(mm, vma, prev, rb_link, rb_parent);
return 0;
}
/*
* Copy the vma structure to a new location in the same mm,
* prior to moving page table entries, to effect an mremap move.
*/
struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
unsigned long addr, unsigned long len, pgoff_t pgoff,
bool *need_rmap_locks)
{
struct vm_area_struct *vma = *vmap;
unsigned long vma_start = vma->vm_start;
struct mm_struct *mm = vma->vm_mm;
struct vm_area_struct *new_vma, *prev;
struct rb_node **rb_link, *rb_parent;
mremap: enforce rmap src/dst vma ordering in case of vma_merge() succeeding in copy_vma() migrate was doing an rmap_walk with speculative lock-less access on pagetables. That could lead it to not serializing properly against mremap PT locks. But a second problem remains in the order of vmas in the same_anon_vma list used by the rmap_walk. If vma_merge succeeds in copy_vma, the src vma could be placed after the dst vma in the same_anon_vma list. That could still lead to migrate missing some pte. This patch adds an anon_vma_moveto_tail() function to force the dst vma at the end of the list before mremap starts to solve the problem. If the mremap is very large and there are a lots of parents or childs sharing the anon_vma root lock, this should still scale better than taking the anon_vma root lock around every pte copy practically for the whole duration of mremap. Update: Hugh noticed special care is needed in the error path where move_page_tables goes in the reverse direction, a second anon_vma_moveto_tail() call is needed in the error path. This program exercises the anon_vma_moveto_tail: === int main() { static struct timeval oldstamp, newstamp; long diffsec; char *p, *p2, *p3, *p4; if (posix_memalign((void **)&p, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p2, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p3, 2*1024*1024, SIZE)) perror("memalign"), exit(1); memset(p, 0xff, SIZE); printf("%p\n", p); memset(p2, 0xff, SIZE); memset(p3, 0x77, 4096); if (memcmp(p, p2, SIZE)) printf("error\n"); p4 = mremap(p+SIZE/2, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p3); if (p4 != p3) perror("mremap"), exit(1); p4 = mremap(p4, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p+SIZE/2); if (p4 != p+SIZE/2) perror("mremap"), exit(1); if (memcmp(p, p2, SIZE)) printf("error\n"); printf("ok\n"); return 0; } === $ perf probe -a anon_vma_moveto_tail Add new event: probe:anon_vma_moveto_tail (on anon_vma_moveto_tail) You can now use it on all perf tools, such as: perf record -e probe:anon_vma_moveto_tail -aR sleep 1 $ perf record -e probe:anon_vma_moveto_tail -aR ./anon_vma_moveto_tail 0x7f2ca2800000 ok [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.043 MB perf.data (~1860 samples) ] $ perf report --stdio 100.00% anon_vma_moveto [kernel.kallsyms] [k] anon_vma_moveto_tail Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Nai Xia <nai.xia@gmail.com> Acked-by: Mel Gorman <mgorman@suse.de> Cc: Hugh Dickins <hughd@google.com> Cc: Pawel Sikora <pluto@agmk.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-11 00:08:05 +01:00
bool faulted_in_anon_vma = true;
/*
* If anonymous vma has not yet been faulted, update new pgoff
* to match new location, to increase its chance of merging.
*/
mremap: fix the wrong !vma->vm_file check in copy_vma() Test-case: #define _GNU_SOURCE #include <stdio.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/mman.h> #include <assert.h> void *find_vdso_vaddr(void) { FILE *perl; char buf[32] = {}; perl = popen("perl -e 'open STDIN,qq|/proc/@{[getppid]}/maps|;" "/^(.*?)-.*vdso/ && print hex $1 while <>'", "r"); fread(buf, sizeof(buf), 1, perl); fclose(perl); return (void *)atol(buf); } #define PAGE_SIZE 4096 void *get_unmapped_area(void) { void *p = mmap(0, PAGE_SIZE, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS, -1,0); assert(p != MAP_FAILED); munmap(p, PAGE_SIZE); return p; } char save[2][PAGE_SIZE]; int main(void) { void *vdso = find_vdso_vaddr(); void *page[2]; assert(vdso); memcpy(save, vdso, sizeof (save)); // force another fault on the next check assert(madvise(vdso, 2 * PAGE_SIZE, MADV_DONTNEED) == 0); page[0] = mremap(vdso, PAGE_SIZE, PAGE_SIZE, MREMAP_FIXED | MREMAP_MAYMOVE, get_unmapped_area()); page[1] = mremap(vdso + PAGE_SIZE, PAGE_SIZE, PAGE_SIZE, MREMAP_FIXED | MREMAP_MAYMOVE, get_unmapped_area()); assert(page[0] != MAP_FAILED && page[1] != MAP_FAILED); printf("match: %d %d\n", !memcmp(save[0], page[0], PAGE_SIZE), !memcmp(save[1], page[1], PAGE_SIZE)); return 0; } fails without this patch. Before the previous commit it gets the wrong page, now it segfaults (which is imho better). This is because copy_vma() wrongly assumes that if vma->vm_file == NULL is irrelevant until the first fault which will use do_anonymous_page(). This is obviously wrong for the special mapping. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Hugh Dickins <hughd@google.com> Cc: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 23:58:34 +02:00
if (unlikely(vma_is_anonymous(vma) && !vma->anon_vma)) {
pgoff = addr >> PAGE_SHIFT;
mremap: enforce rmap src/dst vma ordering in case of vma_merge() succeeding in copy_vma() migrate was doing an rmap_walk with speculative lock-less access on pagetables. That could lead it to not serializing properly against mremap PT locks. But a second problem remains in the order of vmas in the same_anon_vma list used by the rmap_walk. If vma_merge succeeds in copy_vma, the src vma could be placed after the dst vma in the same_anon_vma list. That could still lead to migrate missing some pte. This patch adds an anon_vma_moveto_tail() function to force the dst vma at the end of the list before mremap starts to solve the problem. If the mremap is very large and there are a lots of parents or childs sharing the anon_vma root lock, this should still scale better than taking the anon_vma root lock around every pte copy practically for the whole duration of mremap. Update: Hugh noticed special care is needed in the error path where move_page_tables goes in the reverse direction, a second anon_vma_moveto_tail() call is needed in the error path. This program exercises the anon_vma_moveto_tail: === int main() { static struct timeval oldstamp, newstamp; long diffsec; char *p, *p2, *p3, *p4; if (posix_memalign((void **)&p, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p2, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p3, 2*1024*1024, SIZE)) perror("memalign"), exit(1); memset(p, 0xff, SIZE); printf("%p\n", p); memset(p2, 0xff, SIZE); memset(p3, 0x77, 4096); if (memcmp(p, p2, SIZE)) printf("error\n"); p4 = mremap(p+SIZE/2, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p3); if (p4 != p3) perror("mremap"), exit(1); p4 = mremap(p4, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p+SIZE/2); if (p4 != p+SIZE/2) perror("mremap"), exit(1); if (memcmp(p, p2, SIZE)) printf("error\n"); printf("ok\n"); return 0; } === $ perf probe -a anon_vma_moveto_tail Add new event: probe:anon_vma_moveto_tail (on anon_vma_moveto_tail) You can now use it on all perf tools, such as: perf record -e probe:anon_vma_moveto_tail -aR sleep 1 $ perf record -e probe:anon_vma_moveto_tail -aR ./anon_vma_moveto_tail 0x7f2ca2800000 ok [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.043 MB perf.data (~1860 samples) ] $ perf report --stdio 100.00% anon_vma_moveto [kernel.kallsyms] [k] anon_vma_moveto_tail Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Nai Xia <nai.xia@gmail.com> Acked-by: Mel Gorman <mgorman@suse.de> Cc: Hugh Dickins <hughd@google.com> Cc: Pawel Sikora <pluto@agmk.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-11 00:08:05 +01:00
faulted_in_anon_vma = false;
}
if (find_vma_links(mm, addr, addr + len, &prev, &rb_link, &rb_parent))
return NULL; /* should never get here */
new_vma = vma_merge(mm, prev, addr, addr + len, vma->vm_flags,
vma->anon_vma, vma->vm_file, pgoff, vma_policy(vma),
vma->vm_userfaultfd_ctx);
if (new_vma) {
/*
* Source vma may have been merged into new_vma
*/
mremap: enforce rmap src/dst vma ordering in case of vma_merge() succeeding in copy_vma() migrate was doing an rmap_walk with speculative lock-less access on pagetables. That could lead it to not serializing properly against mremap PT locks. But a second problem remains in the order of vmas in the same_anon_vma list used by the rmap_walk. If vma_merge succeeds in copy_vma, the src vma could be placed after the dst vma in the same_anon_vma list. That could still lead to migrate missing some pte. This patch adds an anon_vma_moveto_tail() function to force the dst vma at the end of the list before mremap starts to solve the problem. If the mremap is very large and there are a lots of parents or childs sharing the anon_vma root lock, this should still scale better than taking the anon_vma root lock around every pte copy practically for the whole duration of mremap. Update: Hugh noticed special care is needed in the error path where move_page_tables goes in the reverse direction, a second anon_vma_moveto_tail() call is needed in the error path. This program exercises the anon_vma_moveto_tail: === int main() { static struct timeval oldstamp, newstamp; long diffsec; char *p, *p2, *p3, *p4; if (posix_memalign((void **)&p, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p2, 2*1024*1024, SIZE)) perror("memalign"), exit(1); if (posix_memalign((void **)&p3, 2*1024*1024, SIZE)) perror("memalign"), exit(1); memset(p, 0xff, SIZE); printf("%p\n", p); memset(p2, 0xff, SIZE); memset(p3, 0x77, 4096); if (memcmp(p, p2, SIZE)) printf("error\n"); p4 = mremap(p+SIZE/2, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p3); if (p4 != p3) perror("mremap"), exit(1); p4 = mremap(p4, SIZE/2, SIZE/2, MREMAP_FIXED|MREMAP_MAYMOVE, p+SIZE/2); if (p4 != p+SIZE/2) perror("mremap"), exit(1); if (memcmp(p, p2, SIZE)) printf("error\n"); printf("ok\n"); return 0; } === $ perf probe -a anon_vma_moveto_tail Add new event: probe:anon_vma_moveto_tail (on anon_vma_moveto_tail) You can now use it on all perf tools, such as: perf record -e probe:anon_vma_moveto_tail -aR sleep 1 $ perf record -e probe:anon_vma_moveto_tail -aR ./anon_vma_moveto_tail 0x7f2ca2800000 ok [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.043 MB perf.data (~1860 samples) ] $ perf report --stdio 100.00% anon_vma_moveto [kernel.kallsyms] [k] anon_vma_moveto_tail Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Nai Xia <nai.xia@gmail.com> Acked-by: Mel Gorman <mgorman@suse.de> Cc: Hugh Dickins <hughd@google.com> Cc: Pawel Sikora <pluto@agmk.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-11 00:08:05 +01:00
if (unlikely(vma_start >= new_vma->vm_start &&
vma_start < new_vma->vm_end)) {
/*
* The only way we can get a vma_merge with
* self during an mremap is if the vma hasn't
* been faulted in yet and we were allowed to
* reset the dst vma->vm_pgoff to the
* destination address of the mremap to allow
* the merge to happen. mremap must change the
* vm_pgoff linearity between src and dst vmas
* (in turn preventing a vma_merge) to be
* safe. It is only safe to keep the vm_pgoff
* linear if there are no pages mapped yet.
*/
VM_BUG_ON_VMA(faulted_in_anon_vma, new_vma);
*vmap = vma = new_vma;
}
*need_rmap_locks = (new_vma->vm_pgoff <= vma->vm_pgoff);
} else {
new_vma = vm_area_dup(vma);
if (!new_vma)
goto out;
new_vma->vm_start = addr;
new_vma->vm_end = addr + len;
new_vma->vm_pgoff = pgoff;
if (vma_dup_policy(vma, new_vma))
goto out_free_vma;
if (anon_vma_clone(new_vma, vma))
goto out_free_mempol;
if (new_vma->vm_file)
get_file(new_vma->vm_file);
if (new_vma->vm_ops && new_vma->vm_ops->open)
new_vma->vm_ops->open(new_vma);
vma_link(mm, new_vma, prev, rb_link, rb_parent);
*need_rmap_locks = false;
}
return new_vma;
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
out_free_mempol:
mpol_put(vma_policy(new_vma));
out_free_vma:
vm_area_free(new_vma);
out:
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
return NULL;
}
/*
* Return true if the calling process may expand its vm space by the passed
* number of pages
*/
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
bool may_expand_vm(struct mm_struct *mm, vm_flags_t flags, unsigned long npages)
{
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
if (mm->total_vm + npages > rlimit(RLIMIT_AS) >> PAGE_SHIFT)
return false;
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
if (is_data_mapping(flags) &&
mm->data_vm + npages > rlimit(RLIMIT_DATA) >> PAGE_SHIFT) {
/* Workaround for Valgrind */
if (rlimit(RLIMIT_DATA) == 0 &&
mm->data_vm + npages <= rlimit_max(RLIMIT_DATA) >> PAGE_SHIFT)
return true;
pr_warn_once("%s (%d): VmData %lu exceed data ulimit %lu. Update limits%s.\n",
current->comm, current->pid,
(mm->data_vm + npages) << PAGE_SHIFT,
rlimit(RLIMIT_DATA),
ignore_rlimit_data ? "" : " or use boot option ignore_rlimit_data");
if (!ignore_rlimit_data)
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
return false;
}
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
return true;
}
void vm_stat_account(struct mm_struct *mm, vm_flags_t flags, long npages)
{
mm->total_vm += npages;
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
if (is_exec_mapping(flags))
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
mm->exec_vm += npages;
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
else if (is_stack_mapping(flags))
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
mm->stack_vm += npages;
mm: warn about VmData over RLIMIT_DATA This patch provides a way of working around a slight regression introduced by commit 84638335900f ("mm: rework virtual memory accounting"). Before that commit RLIMIT_DATA have control only over size of the brk region. But that change have caused problems with all existing versions of valgrind, because it set RLIMIT_DATA to zero. This patch fixes rlimit check (limit actually in bytes, not pages) and by default turns it into warning which prints at first VmData misuse: "mmap: top (795): VmData 516096 exceed data ulimit 512000. Will be forbidden soon." Behavior is controlled by boot param ignore_rlimit_data=y/n and by sysfs /sys/module/kernel/parameters/ignore_rlimit_data. For now it set to "y". [akpm@linux-foundation.org: tweak kernel-parameters.txt text[ Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Link: http://lkml.kernel.org/r/20151228211015.GL2194@uranus Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Cyrill Gorcunov <gorcunov@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Vegard Nossum <vegard.nossum@oracle.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Kees Cook <keescook@google.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-02-03 01:57:43 +01:00
else if (is_data_mapping(flags))
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
mm->data_vm += npages;
}
static vm_fault_t special_mapping_fault(struct vm_fault *vmf);
/*
* Having a close hook prevents vma merging regardless of flags.
*/
static void special_mapping_close(struct vm_area_struct *vma)
{
}
static const char *special_mapping_name(struct vm_area_struct *vma)
{
return ((struct vm_special_mapping *)vma->vm_private_data)->name;
}
x86/vdso: Add mremap hook to vm_special_mapping Add possibility for 32-bit user-space applications to move the vDSO mapping. Previously, when a user-space app called mremap() for the vDSO address, in the syscall return path it would land on the previous address of the vDSOpage, resulting in segmentation violation. Now it lands fine and returns to userspace with a remapped vDSO. This will also fix the context.vdso pointer for 64-bit, which does not affect the user of vDSO after mremap() currently, but this may change in the future. As suggested by Andy, return -EINVAL for mremap() that would split the vDSO image: that operation cannot possibly result in a working system so reject it. Renamed and moved the text_mapping structure declaration inside map_vdso(), as it used only there and now it complements the vvar_mapping variable. There is still a problem for remapping the vDSO in glibc applications: the linker relocates addresses for syscalls on the vDSO page, so you need to relink with the new addresses. Without that the next syscall through glibc may fail: Program received signal SIGSEGV, Segmentation fault. #0 0xf7fd9b80 in __kernel_vsyscall () #1 0xf7ec8238 in _exit () from /usr/lib32/libc.so.6 Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com> Acked-by: Andy Lutomirski <luto@kernel.org> Cc: 0x7f454c46@gmail.com Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160628113539.13606-2-dsafonov@virtuozzo.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-06-28 13:35:38 +02:00
static int special_mapping_mremap(struct vm_area_struct *new_vma)
{
struct vm_special_mapping *sm = new_vma->vm_private_data;
if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
return -EFAULT;
x86/vdso: Add mremap hook to vm_special_mapping Add possibility for 32-bit user-space applications to move the vDSO mapping. Previously, when a user-space app called mremap() for the vDSO address, in the syscall return path it would land on the previous address of the vDSOpage, resulting in segmentation violation. Now it lands fine and returns to userspace with a remapped vDSO. This will also fix the context.vdso pointer for 64-bit, which does not affect the user of vDSO after mremap() currently, but this may change in the future. As suggested by Andy, return -EINVAL for mremap() that would split the vDSO image: that operation cannot possibly result in a working system so reject it. Renamed and moved the text_mapping structure declaration inside map_vdso(), as it used only there and now it complements the vvar_mapping variable. There is still a problem for remapping the vDSO in glibc applications: the linker relocates addresses for syscalls on the vDSO page, so you need to relink with the new addresses. Without that the next syscall through glibc may fail: Program received signal SIGSEGV, Segmentation fault. #0 0xf7fd9b80 in __kernel_vsyscall () #1 0xf7ec8238 in _exit () from /usr/lib32/libc.so.6 Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com> Acked-by: Andy Lutomirski <luto@kernel.org> Cc: 0x7f454c46@gmail.com Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160628113539.13606-2-dsafonov@virtuozzo.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-06-28 13:35:38 +02:00
if (sm->mremap)
return sm->mremap(sm, new_vma);
x86/vdso: Add mremap hook to vm_special_mapping Add possibility for 32-bit user-space applications to move the vDSO mapping. Previously, when a user-space app called mremap() for the vDSO address, in the syscall return path it would land on the previous address of the vDSOpage, resulting in segmentation violation. Now it lands fine and returns to userspace with a remapped vDSO. This will also fix the context.vdso pointer for 64-bit, which does not affect the user of vDSO after mremap() currently, but this may change in the future. As suggested by Andy, return -EINVAL for mremap() that would split the vDSO image: that operation cannot possibly result in a working system so reject it. Renamed and moved the text_mapping structure declaration inside map_vdso(), as it used only there and now it complements the vvar_mapping variable. There is still a problem for remapping the vDSO in glibc applications: the linker relocates addresses for syscalls on the vDSO page, so you need to relink with the new addresses. Without that the next syscall through glibc may fail: Program received signal SIGSEGV, Segmentation fault. #0 0xf7fd9b80 in __kernel_vsyscall () #1 0xf7ec8238 in _exit () from /usr/lib32/libc.so.6 Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com> Acked-by: Andy Lutomirski <luto@kernel.org> Cc: 0x7f454c46@gmail.com Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160628113539.13606-2-dsafonov@virtuozzo.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-06-28 13:35:38 +02:00
return 0;
}
static const struct vm_operations_struct special_mapping_vmops = {
.close = special_mapping_close,
.fault = special_mapping_fault,
x86/vdso: Add mremap hook to vm_special_mapping Add possibility for 32-bit user-space applications to move the vDSO mapping. Previously, when a user-space app called mremap() for the vDSO address, in the syscall return path it would land on the previous address of the vDSOpage, resulting in segmentation violation. Now it lands fine and returns to userspace with a remapped vDSO. This will also fix the context.vdso pointer for 64-bit, which does not affect the user of vDSO after mremap() currently, but this may change in the future. As suggested by Andy, return -EINVAL for mremap() that would split the vDSO image: that operation cannot possibly result in a working system so reject it. Renamed and moved the text_mapping structure declaration inside map_vdso(), as it used only there and now it complements the vvar_mapping variable. There is still a problem for remapping the vDSO in glibc applications: the linker relocates addresses for syscalls on the vDSO page, so you need to relink with the new addresses. Without that the next syscall through glibc may fail: Program received signal SIGSEGV, Segmentation fault. #0 0xf7fd9b80 in __kernel_vsyscall () #1 0xf7ec8238 in _exit () from /usr/lib32/libc.so.6 Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com> Acked-by: Andy Lutomirski <luto@kernel.org> Cc: 0x7f454c46@gmail.com Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/20160628113539.13606-2-dsafonov@virtuozzo.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-06-28 13:35:38 +02:00
.mremap = special_mapping_mremap,
.name = special_mapping_name,
};
static const struct vm_operations_struct legacy_special_mapping_vmops = {
.close = special_mapping_close,
.fault = special_mapping_fault,
};
static vm_fault_t special_mapping_fault(struct vm_fault *vmf)
{
struct vm_area_struct *vma = vmf->vma;
pgoff_t pgoff;
struct page **pages;
if (vma->vm_ops == &legacy_special_mapping_vmops) {
pages = vma->vm_private_data;
} else {
struct vm_special_mapping *sm = vma->vm_private_data;
if (sm->fault)
return sm->fault(sm, vmf->vma, vmf);
pages = sm->pages;
}
mmap: fix the usage of ->vm_pgoff in special_mapping paths Test-case: #include <stdio.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/mman.h> #include <assert.h> void *find_vdso_vaddr(void) { FILE *perl; char buf[32] = {}; perl = popen("perl -e 'open STDIN,qq|/proc/@{[getppid]}/maps|;" "/^(.*?)-.*vdso/ && print hex $1 while <>'", "r"); fread(buf, sizeof(buf), 1, perl); fclose(perl); return (void *)atol(buf); } #define PAGE_SIZE 4096 int main(void) { void *vdso = find_vdso_vaddr(); assert(vdso); // of course they should differ, and they do so far printf("vdso pages differ: %d\n", !!memcmp(vdso, vdso + PAGE_SIZE, PAGE_SIZE)); // split into 2 vma's assert(mprotect(vdso, PAGE_SIZE, PROT_READ) == 0); // force another fault on the next check assert(madvise(vdso, 2 * PAGE_SIZE, MADV_DONTNEED) == 0); // now they no longer differ, the 2nd vm_pgoff is wrong printf("vdso pages differ: %d\n", !!memcmp(vdso, vdso + PAGE_SIZE, PAGE_SIZE)); return 0; } Output: vdso pages differ: 1 vdso pages differ: 0 This is because split_vma() correctly updates ->vm_pgoff, but the logic in insert_vm_struct() and special_mapping_fault() is absolutely broken, so the fault at vdso + PAGE_SIZE return the 1st page. The same happens if you simply unmap the 1st page. special_mapping_fault() does: pgoff = vmf->pgoff - vma->vm_pgoff; and this is _only_ correct if vma->vm_start mmaps the first page from ->vm_private_data array. vdso or any other user of install_special_mapping() is not anonymous, it has the "backing storage" even if it is just the array of pages. So we actually need to make vm_pgoff work as an offset in this array. Note: this also allows to fix another problem: currently gdb can't access "[vvar]" memory because in this case special_mapping_fault() doesn't work. Now that we can use ->vm_pgoff we can implement ->access() and fix this. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Hugh Dickins <hughd@google.com> Cc: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 23:58:31 +02:00
for (pgoff = vmf->pgoff; pgoff && *pages; ++pages)
pgoff--;
if (*pages) {
struct page *page = *pages;
get_page(page);
vmf->page = page;
return 0;
}
return VM_FAULT_SIGBUS;
}
static struct vm_area_struct *__install_special_mapping(
struct mm_struct *mm,
unsigned long addr, unsigned long len,
unsigned long vm_flags, void *priv,
const struct vm_operations_struct *ops)
{
int ret;
struct vm_area_struct *vma;
vma = vm_area_alloc(mm);
if (unlikely(vma == NULL))
return ERR_PTR(-ENOMEM);
vma->vm_start = addr;
vma->vm_end = addr + len;
vma->vm_flags = vm_flags | mm->def_flags | VM_DONTEXPAND | VM_SOFTDIRTY;
vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
vma->vm_ops = ops;
vma->vm_private_data = priv;
ret = insert_vm_struct(mm, vma);
if (ret)
goto out;
mm: rework virtual memory accounting When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations. Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. But in this patch we go further, and the changes are bundled together as: * keep vma counting if CONFIG_PROC_FS=n, will be used for limits * replace mm->shared_vm with better defined mm->data_vm * account anonymous executable areas as executable * account file-backed growsdown/up areas as stack * drop struct file* argument from vm_stat_account * enforce RLIMIT_DATA for size of data areas This way code looks cleaner: now code/stack/data classification depends only on vm_flags state: VM_EXEC & ~VM_WRITE -> code (VmExe + VmLib in proc) VM_GROWSUP | VM_GROWSDOWN -> stack (VmStk) VM_WRITE & ~VM_SHARED & !stack -> data (VmData) The rest (VmSize - VmData - VmStk - VmExe - VmLib) could be called "shared", but that might be strange beast like readonly-private or VM_IO area. - RLIMIT_AS limits whole address space "VmSize" - RLIMIT_STACK limits stack "VmStk" (but each vma individually) - RLIMIT_DATA now limits "VmData" Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Vegard Nossum <vegard.nossum@oracle.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Kees Cook <keescook@google.com> Cc: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 00:22:07 +01:00
vm_stat_account(mm, vma->vm_flags, len >> PAGE_SHIFT);
perf: Do the big rename: Performance Counters -> Performance Events Bye-bye Performance Counters, welcome Performance Events! In the past few months the perfcounters subsystem has grown out its initial role of counting hardware events, and has become (and is becoming) a much broader generic event enumeration, reporting, logging, monitoring, analysis facility. Naming its core object 'perf_counter' and naming the subsystem 'perfcounters' has become more and more of a misnomer. With pending code like hw-breakpoints support the 'counter' name is less and less appropriate. All in one, we've decided to rename the subsystem to 'performance events' and to propagate this rename through all fields, variables and API names. (in an ABI compatible fashion) The word 'event' is also a bit shorter than 'counter' - which makes it slightly more convenient to write/handle as well. Thanks goes to Stephane Eranian who first observed this misnomer and suggested a rename. User-space tooling and ABI compatibility is not affected - this patch should be function-invariant. (Also, defconfigs were not touched to keep the size down.) This patch has been generated via the following script: FILES=$(find * -type f | grep -vE 'oprofile|[^K]config') sed -i \ -e 's/PERF_EVENT_/PERF_RECORD_/g' \ -e 's/PERF_COUNTER/PERF_EVENT/g' \ -e 's/perf_counter/perf_event/g' \ -e 's/nb_counters/nb_events/g' \ -e 's/swcounter/swevent/g' \ -e 's/tpcounter_event/tp_event/g' \ $FILES for N in $(find . -name perf_counter.[ch]); do M=$(echo $N | sed 's/perf_counter/perf_event/g') mv $N $M done FILES=$(find . -name perf_event.*) sed -i \ -e 's/COUNTER_MASK/REG_MASK/g' \ -e 's/COUNTER/EVENT/g' \ -e 's/\<event\>/event_id/g' \ -e 's/counter/event/g' \ -e 's/Counter/Event/g' \ $FILES ... to keep it as correct as possible. This script can also be used by anyone who has pending perfcounters patches - it converts a Linux kernel tree over to the new naming. We tried to time this change to the point in time where the amount of pending patches is the smallest: the end of the merge window. Namespace clashes were fixed up in a preparatory patch - and some stylistic fallout will be fixed up in a subsequent patch. ( NOTE: 'counters' are still the proper terminology when we deal with hardware registers - and these sed scripts are a bit over-eager in renaming them. I've undone some of that, but in case there's something left where 'counter' would be better than 'event' we can undo that on an individual basis instead of touching an otherwise nicely automated patch. ) Suggested-by: Stephane Eranian <eranian@google.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Paul Mackerras <paulus@samba.org> Reviewed-by: Arjan van de Ven <arjan@linux.intel.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Howells <dhowells@redhat.com> Cc: Kyle McMartin <kyle@mcmartin.ca> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: <linux-arch@vger.kernel.org> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 12:02:48 +02:00
perf_event_mmap(vma);
return vma;
out:
vm_area_free(vma);
return ERR_PTR(ret);
}
bool vma_is_special_mapping(const struct vm_area_struct *vma,
const struct vm_special_mapping *sm)
{
return vma->vm_private_data == sm &&
(vma->vm_ops == &special_mapping_vmops ||
vma->vm_ops == &legacy_special_mapping_vmops);
}
/*
* Called with mm->mmap_sem held for writing.
* Insert a new vma covering the given region, with the given flags.
* Its pages are supplied by the given array of struct page *.
* The array can be shorter than len >> PAGE_SHIFT if it's null-terminated.
* The region past the last page supplied will always produce SIGBUS.
* The array pointer and the pages it points to are assumed to stay alive
* for as long as this mapping might exist.
*/
struct vm_area_struct *_install_special_mapping(
struct mm_struct *mm,
unsigned long addr, unsigned long len,
unsigned long vm_flags, const struct vm_special_mapping *spec)
{
return __install_special_mapping(mm, addr, len, vm_flags, (void *)spec,
&special_mapping_vmops);
}
int install_special_mapping(struct mm_struct *mm,
unsigned long addr, unsigned long len,
unsigned long vm_flags, struct page **pages)
{
struct vm_area_struct *vma = __install_special_mapping(
mm, addr, len, vm_flags, (void *)pages,
&legacy_special_mapping_vmops);
return PTR_ERR_OR_ZERO(vma);
}
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
static DEFINE_MUTEX(mm_all_locks_mutex);
static void vm_lock_anon_vma(struct mm_struct *mm, struct anon_vma *anon_vma)
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
{
if (!test_bit(0, (unsigned long *) &anon_vma->root->rb_root.rb_root.rb_node)) {
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
/*
* The LSB of head.next can't change from under us
* because we hold the mm_all_locks_mutex.
*/
mm: mmap: annotate vm_lock_anon_vma locking properly for lockdep Commit 5a505085f043 ("mm/rmap: Convert the struct anon_vma::mutex to an rwsem") turned anon_vma mutex to rwsem. However, the properly annotated nested locking in mm_take_all_locks() has been converted from mutex_lock_nest_lock(&anon_vma->root->mutex, &mm->mmap_sem); to down_write(&anon_vma->root->rwsem); which is incomplete, and causes the false positive report from lockdep below. Annotate the fact that mmap_sem is used as an outter lock to serialize taking of all the anon_vma rwsems at once no matter the order, using the down_write_nest_lock() primitive. This patch fixes this lockdep report: ============================================= [ INFO: possible recursive locking detected ] 3.8.0-rc2-00036-g5f73896 #171 Not tainted --------------------------------------------- qemu-kvm/2315 is trying to acquire lock: (&anon_vma->rwsem){+.+...}, at: mm_take_all_locks+0x149/0x1b0 but task is already holding lock: (&anon_vma->rwsem){+.+...}, at: mm_take_all_locks+0x149/0x1b0 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(&anon_vma->rwsem); lock(&anon_vma->rwsem); *** DEADLOCK *** May be due to missing lock nesting notation 4 locks held by qemu-kvm/2315: #0: (&mm->mmap_sem){++++++}, at: do_mmu_notifier_register+0xfc/0x170 #1: (mm_all_locks_mutex){+.+...}, at: mm_take_all_locks+0x36/0x1b0 #2: (&mapping->i_mmap_mutex){+.+...}, at: mm_take_all_locks+0xc9/0x1b0 #3: (&anon_vma->rwsem){+.+...}, at: mm_take_all_locks+0x149/0x1b0 stack backtrace: Pid: 2315, comm: qemu-kvm Not tainted 3.8.0-rc2-00036-g5f73896 #171 Call Trace: print_deadlock_bug+0xf2/0x100 validate_chain+0x4f6/0x720 __lock_acquire+0x359/0x580 lock_acquire+0x121/0x190 down_write+0x3f/0x70 mm_take_all_locks+0x149/0x1b0 do_mmu_notifier_register+0x68/0x170 mmu_notifier_register+0xe/0x10 kvm_create_vm+0x22b/0x330 [kvm] kvm_dev_ioctl+0xf8/0x1a0 [kvm] do_vfs_ioctl+0x9d/0x350 sys_ioctl+0x91/0xb0 system_call_fastpath+0x16/0x1b Signed-off-by: Jiri Kosina <jkosina@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mel Gorman <mel@csn.ul.ie> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-01-11 23:31:59 +01:00
down_write_nest_lock(&anon_vma->root->rwsem, &mm->mmap_sem);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
/*
* We can safely modify head.next after taking the
* anon_vma->root->rwsem. If some other vma in this mm shares
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
* the same anon_vma we won't take it again.
*
* No need of atomic instructions here, head.next
* can't change from under us thanks to the
* anon_vma->root->rwsem.
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
*/
if (__test_and_set_bit(0, (unsigned long *)
&anon_vma->root->rb_root.rb_root.rb_node))
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
BUG();
}
}
static void vm_lock_mapping(struct mm_struct *mm, struct address_space *mapping)
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
{
if (!test_bit(AS_MM_ALL_LOCKS, &mapping->flags)) {
/*
* AS_MM_ALL_LOCKS can't change from under us because
* we hold the mm_all_locks_mutex.
*
* Operations on ->flags have to be atomic because
* even if AS_MM_ALL_LOCKS is stable thanks to the
* mm_all_locks_mutex, there may be other cpus
* changing other bitflags in parallel to us.
*/
if (test_and_set_bit(AS_MM_ALL_LOCKS, &mapping->flags))
BUG();
down_write_nest_lock(&mapping->i_mmap_rwsem, &mm->mmap_sem);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
}
}
/*
* This operation locks against the VM for all pte/vma/mm related
* operations that could ever happen on a certain mm. This includes
* vmtruncate, try_to_unmap, and all page faults.
*
* The caller must take the mmap_sem in write mode before calling
* mm_take_all_locks(). The caller isn't allowed to release the
* mmap_sem until mm_drop_all_locks() returns.
*
* mmap_sem in write mode is required in order to block all operations
* that could modify pagetables and free pages without need of
* altering the vma layout. It's also needed in write mode to avoid new
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
* anon_vmas to be associated with existing vmas.
*
* A single task can't take more than one mm_take_all_locks() in a row
* or it would deadlock.
*
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
* The LSB in anon_vma->rb_root.rb_node and the AS_MM_ALL_LOCKS bitflag in
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
* mapping->flags avoid to take the same lock twice, if more than one
* vma in this mm is backed by the same anon_vma or address_space.
*
* We take locks in following order, accordingly to comment at beginning
* of mm/rmap.c:
* - all hugetlbfs_i_mmap_rwsem_key locks (aka mapping->i_mmap_rwsem for
* hugetlb mapping);
* - all i_mmap_rwsem locks;
* - all anon_vma->rwseml
*
* We can take all locks within these types randomly because the VM code
* doesn't nest them and we protected from parallel mm_take_all_locks() by
* mm_all_locks_mutex.
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
*
* mm_take_all_locks() and mm_drop_all_locks are expensive operations
* that may have to take thousand of locks.
*
* mm_take_all_locks() can fail if it's interrupted by signals.
*/
int mm_take_all_locks(struct mm_struct *mm)
{
struct vm_area_struct *vma;
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
struct anon_vma_chain *avc;
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
BUG_ON(down_read_trylock(&mm->mmap_sem));
mutex_lock(&mm_all_locks_mutex);
for (vma = mm->mmap; vma; vma = vma->vm_next) {
if (signal_pending(current))
goto out_unlock;
if (vma->vm_file && vma->vm_file->f_mapping &&
is_vm_hugetlb_page(vma))
vm_lock_mapping(mm, vma->vm_file->f_mapping);
}
for (vma = mm->mmap; vma; vma = vma->vm_next) {
if (signal_pending(current))
goto out_unlock;
if (vma->vm_file && vma->vm_file->f_mapping &&
!is_vm_hugetlb_page(vma))
vm_lock_mapping(mm, vma->vm_file->f_mapping);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
}
mm: fix mm_take_all_locks() locking order Lockdep spotted: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.27-rc1 #270 ------------------------------------------------------- qemu-kvm/2033 is trying to acquire lock: (&inode->i_data.i_mmap_lock){----}, at: [<ffffffff802996cc>] mm_take_all_locks+0xc2/0xea but task is already holding lock: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&anon_vma->lock){----}: [<ffffffff8025cd37>] __lock_acquire+0x11be/0x14d2 [<ffffffff8025d0a9>] lock_acquire+0x5e/0x7a [<ffffffff804c655b>] _spin_lock+0x3b/0x47 [<ffffffff8029a2ef>] vma_adjust+0x200/0x444 [<ffffffff8029a662>] split_vma+0x12f/0x146 [<ffffffff8029bc60>] mprotect_fixup+0x13c/0x536 [<ffffffff8029c203>] sys_mprotect+0x1a9/0x21e [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff -> #0 (&inode->i_data.i_mmap_lock){----}: [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2 [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219 [<ffffffff8025d515>] lock_release+0x127/0x14a [<ffffffff804c6403>] _spin_unlock+0x1e/0x50 [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0 [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112 [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10 [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm] [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78 [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274 [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78 [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff other info that might help us debug this: 5 locks held by qemu-kvm/2033: #0: (&mm->mmap_sem){----}, at: [<ffffffff802a95d0>] do_mmu_notifier_register+0x55/0x112 #1: (mm_all_locks_mutex){--..}, at: [<ffffffff8029963e>] mm_take_all_locks+0x34/0xea #2: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea #3: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea #4: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea stack backtrace: Pid: 2033, comm: qemu-kvm Not tainted 2.6.27-rc1 #270 Call Trace: [<ffffffff8025b7c7>] print_circular_bug_tail+0xb8/0xc3 [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2 [<ffffffff80259bb1>] ? add_lock_to_list+0x7e/0xad [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219 [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea [<ffffffff8025b202>] ? trace_hardirqs_on_caller+0x4d/0x115 [<ffffffff802995d9>] ? mm_drop_all_locks+0x7f/0xb0 [<ffffffff8025d515>] lock_release+0x127/0x14a [<ffffffff804c6403>] _spin_unlock+0x1e/0x50 [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0 [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112 [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10 [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm] [<ffffffff8033f9f2>] ? file_has_perm+0x83/0x8e [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78 [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274 [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78 [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b Which the locking hierarchy in mm/rmap.c confirms as valid. Fix this by first taking all the mapping->i_mmap_lock instances and then take all anon_vma->lock instances. Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-08-11 09:30:25 +02:00
for (vma = mm->mmap; vma; vma = vma->vm_next) {
if (signal_pending(current))
goto out_unlock;
if (vma->anon_vma)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
list_for_each_entry(avc, &vma->anon_vma_chain, same_vma)
vm_lock_anon_vma(mm, avc->anon_vma);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
}
mm: fix mm_take_all_locks() locking order Lockdep spotted: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.27-rc1 #270 ------------------------------------------------------- qemu-kvm/2033 is trying to acquire lock: (&inode->i_data.i_mmap_lock){----}, at: [<ffffffff802996cc>] mm_take_all_locks+0xc2/0xea but task is already holding lock: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&anon_vma->lock){----}: [<ffffffff8025cd37>] __lock_acquire+0x11be/0x14d2 [<ffffffff8025d0a9>] lock_acquire+0x5e/0x7a [<ffffffff804c655b>] _spin_lock+0x3b/0x47 [<ffffffff8029a2ef>] vma_adjust+0x200/0x444 [<ffffffff8029a662>] split_vma+0x12f/0x146 [<ffffffff8029bc60>] mprotect_fixup+0x13c/0x536 [<ffffffff8029c203>] sys_mprotect+0x1a9/0x21e [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff -> #0 (&inode->i_data.i_mmap_lock){----}: [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2 [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219 [<ffffffff8025d515>] lock_release+0x127/0x14a [<ffffffff804c6403>] _spin_unlock+0x1e/0x50 [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0 [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112 [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10 [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm] [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78 [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274 [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78 [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff other info that might help us debug this: 5 locks held by qemu-kvm/2033: #0: (&mm->mmap_sem){----}, at: [<ffffffff802a95d0>] do_mmu_notifier_register+0x55/0x112 #1: (mm_all_locks_mutex){--..}, at: [<ffffffff8029963e>] mm_take_all_locks+0x34/0xea #2: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea #3: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea #4: (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea stack backtrace: Pid: 2033, comm: qemu-kvm Not tainted 2.6.27-rc1 #270 Call Trace: [<ffffffff8025b7c7>] print_circular_bug_tail+0xb8/0xc3 [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2 [<ffffffff80259bb1>] ? add_lock_to_list+0x7e/0xad [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219 [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea [<ffffffff8025b202>] ? trace_hardirqs_on_caller+0x4d/0x115 [<ffffffff802995d9>] ? mm_drop_all_locks+0x7f/0xb0 [<ffffffff8025d515>] lock_release+0x127/0x14a [<ffffffff804c6403>] _spin_unlock+0x1e/0x50 [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0 [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112 [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10 [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm] [<ffffffff8033f9f2>] ? file_has_perm+0x83/0x8e [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78 [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274 [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78 [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b Which the locking hierarchy in mm/rmap.c confirms as valid. Fix this by first taking all the mapping->i_mmap_lock instances and then take all anon_vma->lock instances. Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-08-11 09:30:25 +02:00
return 0;
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
out_unlock:
mm_drop_all_locks(mm);
return -EINTR;
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
}
static void vm_unlock_anon_vma(struct anon_vma *anon_vma)
{
if (test_bit(0, (unsigned long *) &anon_vma->root->rb_root.rb_root.rb_node)) {
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
/*
* The LSB of head.next can't change to 0 from under
* us because we hold the mm_all_locks_mutex.
*
* We must however clear the bitflag before unlocking
mm anon rmap: replace same_anon_vma linked list with an interval tree. When a large VMA (anon or private file mapping) is first touched, which will populate its anon_vma field, and then split into many regions through the use of mprotect(), the original anon_vma ends up linking all of the vmas on a linked list. This can cause rmap to become inefficient, as we have to walk potentially thousands of irrelevent vmas before finding the one a given anon page might fall into. By replacing the same_anon_vma linked list with an interval tree (where each avc's interval is determined by its vma's start and last pgoffs), we can make rmap efficient for this use case again. While the change is large, all of its pieces are fairly simple. Most places that were walking the same_anon_vma list were looking for a known pgoff, so they can just use the anon_vma_interval_tree_foreach() interval tree iterator instead. The exception here is ksm, where the page's index is not known. It would probably be possible to rework ksm so that the index would be known, but for now I have decided to keep things simple and just walk the entirety of the interval tree there. When updating vma's that already have an anon_vma assigned, we must take care to re-index the corresponding avc's on their interval tree. This is done through the use of anon_vma_interval_tree_pre_update_vma() and anon_vma_interval_tree_post_update_vma(), which remove the avc's from their interval tree before the update and re-insert them after the update. The anon_vma stays locked during the update, so there is no chance that rmap would miss the vmas that are being updated. Signed-off-by: Michel Lespinasse <walken@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Daniel Santos <daniel.santos@pobox.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-09 01:31:39 +02:00
* the vma so the users using the anon_vma->rb_root will
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
* never see our bitflag.
*
* No need of atomic instructions here, head.next
* can't change from under us until we release the
* anon_vma->root->rwsem.
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
*/
if (!__test_and_clear_bit(0, (unsigned long *)
&anon_vma->root->rb_root.rb_root.rb_node))
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
BUG();
anon_vma_unlock_write(anon_vma);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
}
}
static void vm_unlock_mapping(struct address_space *mapping)
{
if (test_bit(AS_MM_ALL_LOCKS, &mapping->flags)) {
/*
* AS_MM_ALL_LOCKS can't change to 0 from under us
* because we hold the mm_all_locks_mutex.
*/
i_mmap_unlock_write(mapping);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
if (!test_and_clear_bit(AS_MM_ALL_LOCKS,
&mapping->flags))
BUG();
}
}
/*
* The mmap_sem cannot be released by the caller until
* mm_drop_all_locks() returns.
*/
void mm_drop_all_locks(struct mm_struct *mm)
{
struct vm_area_struct *vma;
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
struct anon_vma_chain *avc;
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
BUG_ON(down_read_trylock(&mm->mmap_sem));
BUG_ON(!mutex_is_locked(&mm_all_locks_mutex));
for (vma = mm->mmap; vma; vma = vma->vm_next) {
if (vma->anon_vma)
mm: change anon_vma linking to fix multi-process server scalability issue The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel <riel@redhat.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 22:42:07 +01:00
list_for_each_entry(avc, &vma->anon_vma_chain, same_vma)
vm_unlock_anon_vma(avc->anon_vma);
mmu-notifiers: add mm_take_all_locks() operation mm_take_all_locks holds off reclaim from an entire mm_struct. This allows mmu notifiers to register into the mm at any time with the guarantee that no mmu operation is in progress on the mm. This operation locks against the VM for all pte/vma/mm related operations that could ever happen on a certain mm. This includes vmtruncate, try_to_unmap, and all page faults. The caller must take the mmap_sem in write mode before calling mm_take_all_locks(). The caller isn't allowed to release the mmap_sem until mm_drop_all_locks() returns. mmap_sem in write mode is required in order to block all operations that could modify pagetables and free pages without need of altering the vma layout (for example populate_range() with nonlinear vmas). It's also needed in write mode to avoid new anon_vmas to be associated with existing vmas. A single task can't take more than one mm_take_all_locks() in a row or it would deadlock. mm_take_all_locks() and mm_drop_all_locks are expensive operations that may have to take thousand of locks. mm_take_all_locks() can fail if it's interrupted by signals. When mmu_notifier_register returns, we must be sure that the driver is notified if some task is in the middle of a vmtruncate for the 'mm' where the mmu notifier was registered (mmu_notifier_invalidate_range_start/end is run around the vmtruncation but mmu_notifier_register can run after mmu_notifier_invalidate_range_start and before mmu_notifier_invalidate_range_end). Same problem for rmap paths. And we've to remove page pinning to avoid replicating the tlb_gather logic inside KVM (and GRU doesn't work well with page pinning regardless of needing tlb_gather), so without mm_take_all_locks when vmtruncate frees the page, kvm would have no way to notice that it mapped into sptes a page that is going into the freelist without a chance of any further mmu_notifier notification. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Andrea Arcangeli <andrea@qumranet.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Jack Steiner <steiner@sgi.com> Cc: Robin Holt <holt@sgi.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Kanoj Sarcar <kanojsarcar@yahoo.com> Cc: Roland Dreier <rdreier@cisco.com> Cc: Steve Wise <swise@opengridcomputing.com> Cc: Avi Kivity <avi@qumranet.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Chris Wright <chrisw@redhat.com> Cc: Marcelo Tosatti <marcelo@kvack.org> Cc: Eric Dumazet <dada1@cosmosbay.com> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Izik Eidus <izike@qumranet.com> Cc: Anthony Liguori <aliguori@us.ibm.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:26 +02:00
if (vma->vm_file && vma->vm_file->f_mapping)
vm_unlock_mapping(vma->vm_file->f_mapping);
}
mutex_unlock(&mm_all_locks_mutex);
}
NOMMU: Make VMAs per MM as for MMU-mode linux Make VMAs per mm_struct as for MMU-mode linux. This solves two problems: (1) In SYSV SHM where nattch for a segment does not reflect the number of shmat's (and forks) done. (2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an exec'ing process when VM_EXECUTABLE is specified, regardless of the fact that a VMA might be shared and already have its vm_mm assigned to another process or a dead process. A new struct (vm_region) is introduced to track a mapped region and to remember the circumstances under which it may be shared and the vm_list_struct structure is discarded as it's no longer required. This patch makes the following additional changes: (1) Regions are now allocated with alloc_pages() rather than kmalloc() and with no recourse to __GFP_COMP, so the pages are not composite. Instead, each page has a reference on it held by the region. Anything else that is interested in such a page will have to get a reference on it to retain it. When the pages are released due to unmapping, each page is passed to put_page() and will be freed when the page usage count reaches zero. (2) Excess pages are trimmed after an allocation as the allocation must be made as a power-of-2 quantity of pages. (3) VMAs are added to the parent MM's R/B tree and mmap lists. As an MM may end up with overlapping VMAs within the tree, the VMA struct address is appended to the sort key. (4) Non-anonymous VMAs are now added to the backing inode's prio list. (5) Holes may be punched in anonymous VMAs with munmap(), releasing parts of the backing region. The VMA and region structs will be split if necessary. (6) sys_shmdt() only releases one attachment to a SYSV IPC shared memory segment instead of all the attachments at that addresss. Multiple shmat()'s return the same address under NOMMU-mode instead of different virtual addresses as under MMU-mode. (7) Core dumping for ELF-FDPIC requires fewer exceptions for NOMMU-mode. (8) /proc/maps is now the global list of mapped regions, and may list bits that aren't actually mapped anywhere. (9) /proc/meminfo gains a line (tagged "MmapCopy") that indicates the amount of RAM currently allocated by mmap to hold mappable regions that can't be mapped directly. These are copies of the backing device or file if not anonymous. These changes make NOMMU mode more similar to MMU mode. The downside is that NOMMU mode requires some extra memory to track things over NOMMU without this patch (VMAs are no longer shared, and there are now region structs). Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Mike Frysinger <vapier.adi@gmail.com> Acked-by: Paul Mundt <lethal@linux-sh.org>
2009-01-08 13:04:47 +01:00
/*
* initialise the percpu counter for VM
NOMMU: Make VMAs per MM as for MMU-mode linux Make VMAs per mm_struct as for MMU-mode linux. This solves two problems: (1) In SYSV SHM where nattch for a segment does not reflect the number of shmat's (and forks) done. (2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an exec'ing process when VM_EXECUTABLE is specified, regardless of the fact that a VMA might be shared and already have its vm_mm assigned to another process or a dead process. A new struct (vm_region) is introduced to track a mapped region and to remember the circumstances under which it may be shared and the vm_list_struct structure is discarded as it's no longer required. This patch makes the following additional changes: (1) Regions are now allocated with alloc_pages() rather than kmalloc() and with no recourse to __GFP_COMP, so the pages are not composite. Instead, each page has a reference on it held by the region. Anything else that is interested in such a page will have to get a reference on it to retain it. When the pages are released due to unmapping, each page is passed to put_page() and will be freed when the page usage count reaches zero. (2) Excess pages are trimmed after an allocation as the allocation must be made as a power-of-2 quantity of pages. (3) VMAs are added to the parent MM's R/B tree and mmap lists. As an MM may end up with overlapping VMAs within the tree, the VMA struct address is appended to the sort key. (4) Non-anonymous VMAs are now added to the backing inode's prio list. (5) Holes may be punched in anonymous VMAs with munmap(), releasing parts of the backing region. The VMA and region structs will be split if necessary. (6) sys_shmdt() only releases one attachment to a SYSV IPC shared memory segment instead of all the attachments at that addresss. Multiple shmat()'s return the same address under NOMMU-mode instead of different virtual addresses as under MMU-mode. (7) Core dumping for ELF-FDPIC requires fewer exceptions for NOMMU-mode. (8) /proc/maps is now the global list of mapped regions, and may list bits that aren't actually mapped anywhere. (9) /proc/meminfo gains a line (tagged "MmapCopy") that indicates the amount of RAM currently allocated by mmap to hold mappable regions that can't be mapped directly. These are copies of the backing device or file if not anonymous. These changes make NOMMU mode more similar to MMU mode. The downside is that NOMMU mode requires some extra memory to track things over NOMMU without this patch (VMAs are no longer shared, and there are now region structs). Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Mike Frysinger <vapier.adi@gmail.com> Acked-by: Paul Mundt <lethal@linux-sh.org>
2009-01-08 13:04:47 +01:00
*/
void __init mmap_init(void)
{
mm: fix Committed_AS underflow on large NR_CPUS environment The Committed_AS field can underflow in certain situations: > # while true; do cat /proc/meminfo | grep _AS; sleep 1; done | uniq -c > 1 Committed_AS: 18446744073709323392 kB > 11 Committed_AS: 18446744073709455488 kB > 6 Committed_AS: 35136 kB > 5 Committed_AS: 18446744073709454400 kB > 7 Committed_AS: 35904 kB > 3 Committed_AS: 18446744073709453248 kB > 2 Committed_AS: 34752 kB > 9 Committed_AS: 18446744073709453248 kB > 8 Committed_AS: 34752 kB > 3 Committed_AS: 18446744073709320960 kB > 7 Committed_AS: 18446744073709454080 kB > 3 Committed_AS: 18446744073709320960 kB > 5 Committed_AS: 18446744073709454080 kB > 6 Committed_AS: 18446744073709320960 kB Because NR_CPUS can be greater than 1000 and meminfo_proc_show() does not check for underflow. But NR_CPUS proportional isn't good calculation. In general, possibility of lock contention is proportional to the number of online cpus, not theorical maximum cpus (NR_CPUS). The current kernel has generic percpu-counter stuff. using it is right way. it makes code simplify and percpu_counter_read_positive() don't make underflow issue. Reported-by: Dave Hansen <dave@linux.vnet.ibm.com> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Eric B Munson <ebmunson@us.ibm.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: <stable@kernel.org> [All kernel versions] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-05-01 00:08:51 +02:00
int ret;
ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
mm: fix Committed_AS underflow on large NR_CPUS environment The Committed_AS field can underflow in certain situations: > # while true; do cat /proc/meminfo | grep _AS; sleep 1; done | uniq -c > 1 Committed_AS: 18446744073709323392 kB > 11 Committed_AS: 18446744073709455488 kB > 6 Committed_AS: 35136 kB > 5 Committed_AS: 18446744073709454400 kB > 7 Committed_AS: 35904 kB > 3 Committed_AS: 18446744073709453248 kB > 2 Committed_AS: 34752 kB > 9 Committed_AS: 18446744073709453248 kB > 8 Committed_AS: 34752 kB > 3 Committed_AS: 18446744073709320960 kB > 7 Committed_AS: 18446744073709454080 kB > 3 Committed_AS: 18446744073709320960 kB > 5 Committed_AS: 18446744073709454080 kB > 6 Committed_AS: 18446744073709320960 kB Because NR_CPUS can be greater than 1000 and meminfo_proc_show() does not check for underflow. But NR_CPUS proportional isn't good calculation. In general, possibility of lock contention is proportional to the number of online cpus, not theorical maximum cpus (NR_CPUS). The current kernel has generic percpu-counter stuff. using it is right way. it makes code simplify and percpu_counter_read_positive() don't make underflow issue. Reported-by: Dave Hansen <dave@linux.vnet.ibm.com> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Eric B Munson <ebmunson@us.ibm.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: <stable@kernel.org> [All kernel versions] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-05-01 00:08:51 +02:00
VM_BUG_ON(ret);
NOMMU: Make VMAs per MM as for MMU-mode linux Make VMAs per mm_struct as for MMU-mode linux. This solves two problems: (1) In SYSV SHM where nattch for a segment does not reflect the number of shmat's (and forks) done. (2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an exec'ing process when VM_EXECUTABLE is specified, regardless of the fact that a VMA might be shared and already have its vm_mm assigned to another process or a dead process. A new struct (vm_region) is introduced to track a mapped region and to remember the circumstances under which it may be shared and the vm_list_struct structure is discarded as it's no longer required. This patch makes the following additional changes: (1) Regions are now allocated with alloc_pages() rather than kmalloc() and with no recourse to __GFP_COMP, so the pages are not composite. Instead, each page has a reference on it held by the region. Anything else that is interested in such a page will have to get a reference on it to retain it. When the pages are released due to unmapping, each page is passed to put_page() and will be freed when the page usage count reaches zero. (2) Excess pages are trimmed after an allocation as the allocation must be made as a power-of-2 quantity of pages. (3) VMAs are added to the parent MM's R/B tree and mmap lists. As an MM may end up with overlapping VMAs within the tree, the VMA struct address is appended to the sort key. (4) Non-anonymous VMAs are now added to the backing inode's prio list. (5) Holes may be punched in anonymous VMAs with munmap(), releasing parts of the backing region. The VMA and region structs will be split if necessary. (6) sys_shmdt() only releases one attachment to a SYSV IPC shared memory segment instead of all the attachments at that addresss. Multiple shmat()'s return the same address under NOMMU-mode instead of different virtual addresses as under MMU-mode. (7) Core dumping for ELF-FDPIC requires fewer exceptions for NOMMU-mode. (8) /proc/maps is now the global list of mapped regions, and may list bits that aren't actually mapped anywhere. (9) /proc/meminfo gains a line (tagged "MmapCopy") that indicates the amount of RAM currently allocated by mmap to hold mappable regions that can't be mapped directly. These are copies of the backing device or file if not anonymous. These changes make NOMMU mode more similar to MMU mode. The downside is that NOMMU mode requires some extra memory to track things over NOMMU without this patch (VMAs are no longer shared, and there are now region structs). Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Mike Frysinger <vapier.adi@gmail.com> Acked-by: Paul Mundt <lethal@linux-sh.org>
2009-01-08 13:04:47 +01:00
}
mm: limit growth of 3% hardcoded other user reserve Add user_reserve_kbytes knob. Limit the growth of the memory reserved for other user processes to min(3% current process size, user_reserve_pages). Only about 8MB is necessary to enable recovery in the default mode, and only a few hundred MB are required even when overcommit is disabled. user_reserve_pages defaults to min(3% free pages, 128MB) I arrived at 128MB by taking the max VSZ of sshd, login, bash, and top ... then adding the RSS of each. This only affects OVERCOMMIT_NEVER mode. Background 1. user reserve __vm_enough_memory reserves a hardcoded 3% of the current process size for other applications when overcommit is disabled. This was done so that a user could recover if they launched a memory hogging process. Without the reserve, a user would easily run into a message such as: bash: fork: Cannot allocate memory 2. admin reserve Additionally, a hardcoded 3% of free memory is reserved for root in both overcommit 'guess' and 'never' modes. This was intended to prevent a scenario where root-cant-log-in and perform recovery operations. Note that this reserve shrinks, and doesn't guarantee a useful reserve. Motivation The two hardcoded memory reserves should be updated to account for current memory sizes. Also, the admin reserve would be more useful if it didn't shrink too much. When the current code was originally written, 1GB was considered "enterprise". Now the 3% reserve can grow to multiple GB on large memory systems, and it only needs to be a few hundred MB at most to enable a user or admin to recover a system with an unwanted memory hogging process. I've found that reducing these reserves is especially beneficial for a specific type of application load: * single application system * one or few processes (e.g. one per core) * allocating all available memory * not initializing every page immediately * long running I've run scientific clusters with this sort of load. A long running job sometimes failed many hours (weeks of CPU time) into a calculation. They weren't initializing all of their memory immediately, and they weren't using calloc, so I put systems into overcommit 'never' mode. These clusters run diskless and have no swap. However, with the current reserves, a user wishing to allocate as much memory as possible to one process may be prevented from using, for example, almost 2GB out of 32GB. The effect is less, but still significant when a user starts a job with one process per core. I have repeatedly seen a set of processes requesting the same amount of memory fail because one of them could not allocate the amount of memory a user would expect to be able to allocate. For example, Message Passing Interfce (MPI) processes, one per core. And it is similar for other parallel programming frameworks. Changing this reserve code will make the overcommit never mode more useful by allowing applications to allocate nearly all of the available memory. Also, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Risks * "bash: fork: Cannot allocate memory" The downside of the first patch-- which creates a tunable user reserve that is only used in overcommit 'never' mode--is that an admin can set it so low that a user may not be able to kill their process, even if they already have a shell prompt. Of course, a user can get in the same predicament with the current 3% reserve--they just have to launch processes until 3% becomes negligible. * root-cant-log-in problem The second patch, adding the tunable rootuser_reserve_pages, allows the admin to shoot themselves in the foot by setting it too small. They can easily get the system into a state where root-can't-log-in. However, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Alternatives * Memory cgroups provide a more flexible way to limit application memory. Not everyone wants to set up cgroups or deal with their overhead. * We could create a fourth overcommit mode which provides smaller reserves. The size of useful reserves may be drastically different depending on the whether the system is embedded or enterprise. * Force users to initialize all of their memory or use calloc. Some users don't want/expect the system to overcommit when they malloc. Overcommit 'never' mode is for this scenario, and it should work well. The new user and admin reserve tunables are simple to use, with low overhead compared to cgroups. The patches preserve current behavior where 3% of memory is less than 128MB, except that the admin reserve doesn't shrink to an unusable size under pressure. The code allows admins to tune for embedded and enterprise usage. FAQ * How is the root-cant-login problem addressed? What happens if admin_reserve_pages is set to 0? Root is free to shoot themselves in the foot by setting admin_reserve_kbytes too low. On x86_64, the minimum useful reserve is: 8MB for overcommit 'guess' 128MB for overcommit 'never' admin_reserve_pages defaults to min(3% free memory, 8MB) So, anyone switching to 'never' mode needs to adjust admin_reserve_pages. * How do you calculate a minimum useful reserve? A user or the admin needs enough memory to login and perform recovery operations, which includes, at a minimum: sshd or login + bash (or some other shell) + top (or ps, kill, etc.) For overcommit 'guess', we can sum resident set sizes (RSS) because we only need enough memory to handle what the recovery programs will typically use. On x86_64 this is about 8MB. For overcommit 'never', we can take the max of their virtual sizes (VSZ) and add the sum of their RSS. We use VSZ instead of RSS because mode forces us to ensure we can fulfill all of the requested memory allocations-- even if the programs only use a fraction of what they ask for. On x86_64 this is about 128MB. When swap is enabled, reserves are useful even when they are as small as 10MB, regardless of overcommit mode. When both swap and overcommit are disabled, then the admin should tune the reserves higher to be absolutley safe. Over 230MB each was safest in my testing. * What happens if user_reserve_pages is set to 0? Note, this only affects overcomitt 'never' mode. Then a user will be able to allocate all available memory minus admin_reserve_kbytes. However, they will easily see a message such as: "bash: fork: Cannot allocate memory" And they won't be able to recover/kill their application. The admin should be able to recover the system if admin_reserve_kbytes is set appropriately. * What's the difference between overcommit 'guess' and 'never'? "Guess" allows an allocation if there are enough free + reclaimable pages. It has a hardcoded 3% of free pages reserved for root. "Never" allows an allocation if there is enough swap + a configurable percentage (default is 50) of physical RAM. It has a hardcoded 3% of free pages reserved for root, like "Guess" mode. It also has a hardcoded 3% of the current process size reserved for additional applications. * Why is overcommit 'guess' not suitable even when an app eventually writes to every page? It takes free pages, file pages, available swap pages, reclaimable slab pages into consideration. In other words, these are all pages available, then why isn't overcommit suitable? Because it only looks at the present state of the system. It does not take into account the memory that other applications have malloced, but haven't initialized yet. It overcommits the system. Test Summary There was little change in behavior in the default overcommit 'guess' mode with swap enabled before and after the patch. This was expected. Systems run most predictably (i.e. no oom kills) in overcommit 'never' mode with swap enabled. This also allowed the most memory to be allocated to a user application. Overcommit 'guess' mode without swap is a bad idea. It is easy to crash the system. None of the other tested combinations crashed. This matches my experience on the Roadrunner supercomputer. Without the tunable user reserve, a system in overcommit 'never' mode and without swap does not allow the admin to recover, although the admin can. With the new tunable reserves, a system in overcommit 'never' mode and without swap can be configured to: 1. maximize user-allocatable memory, running close to the edge of recoverability 2. maximize recoverability, sacrificing allocatable memory to ensure that a user cannot take down a system Test Description Fedora 18 VM - 4 x86_64 cores, 5725MB RAM, 4GB Swap System is booted into multiuser console mode, with unnecessary services turned off. Caches were dropped before each test. Hogs are user memtester processes that attempt to allocate all free memory as reported by /proc/meminfo In overcommit 'never' mode, memory_ratio=100 Test Results 3.9.0-rc1-mm1 Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5432/5432 no yes yes guess yes 4 5444/5444 1 yes yes guess no 1 5302/5449 no yes yes guess no 4 - crash no no never yes 1 5460/5460 1 yes yes never yes 4 5460/5460 1 yes yes never no 1 5218/5432 no no yes never no 4 5203/5448 no no yes 3.9.0-rc1-mm1-tunablereserves User and Admin Recovery show their respective reserves, if applicable. Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5419/5419 no - yes 8MB yes guess yes 4 5436/5436 1 - yes 8MB yes guess no 1 5440/5440 * - yes 8MB yes guess no 4 - crash - no 8MB no * process would successfully mlock, then the oom killer would pick it never yes 1 5446/5446 no 10MB yes 20MB yes never yes 4 5456/5456 no 10MB yes 20MB yes never no 1 5387/5429 no 128MB no 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5359/5448 no 10MB no 10MB barely never no 1 5323/5428 no 0MB no 10MB barely never no 1 5332/5428 no 0MB no 50MB yes never no 1 5293/5429 no 0MB no 90MB yes never no 1 5001/5427 no 230MB yes 338MB yes never no 4* 4998/5424 no 230MB yes 338MB yes * more memtesters were launched, able to allocate approximately another 100MB Future Work - Test larger memory systems. - Test an embedded image. - Test other architectures. - Time malloc microbenchmarks. - Would it be useful to be able to set overcommit policy for each memory cgroup? - Some lines are slightly above 80 chars. Perhaps define a macro to convert between pages and kb? Other places in the kernel do this. [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: make init_user_reserve() static] Signed-off-by: Andrew Shewmaker <agshew@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:10 +02:00
/*
* Initialise sysctl_user_reserve_kbytes.
*
* This is intended to prevent a user from starting a single memory hogging
* process, such that they cannot recover (kill the hog) in OVERCOMMIT_NEVER
* mode.
*
* The default value is min(3% of free memory, 128MB)
* 128MB is enough to recover with sshd/login, bash, and top/kill.
*/
static int init_user_reserve(void)
mm: limit growth of 3% hardcoded other user reserve Add user_reserve_kbytes knob. Limit the growth of the memory reserved for other user processes to min(3% current process size, user_reserve_pages). Only about 8MB is necessary to enable recovery in the default mode, and only a few hundred MB are required even when overcommit is disabled. user_reserve_pages defaults to min(3% free pages, 128MB) I arrived at 128MB by taking the max VSZ of sshd, login, bash, and top ... then adding the RSS of each. This only affects OVERCOMMIT_NEVER mode. Background 1. user reserve __vm_enough_memory reserves a hardcoded 3% of the current process size for other applications when overcommit is disabled. This was done so that a user could recover if they launched a memory hogging process. Without the reserve, a user would easily run into a message such as: bash: fork: Cannot allocate memory 2. admin reserve Additionally, a hardcoded 3% of free memory is reserved for root in both overcommit 'guess' and 'never' modes. This was intended to prevent a scenario where root-cant-log-in and perform recovery operations. Note that this reserve shrinks, and doesn't guarantee a useful reserve. Motivation The two hardcoded memory reserves should be updated to account for current memory sizes. Also, the admin reserve would be more useful if it didn't shrink too much. When the current code was originally written, 1GB was considered "enterprise". Now the 3% reserve can grow to multiple GB on large memory systems, and it only needs to be a few hundred MB at most to enable a user or admin to recover a system with an unwanted memory hogging process. I've found that reducing these reserves is especially beneficial for a specific type of application load: * single application system * one or few processes (e.g. one per core) * allocating all available memory * not initializing every page immediately * long running I've run scientific clusters with this sort of load. A long running job sometimes failed many hours (weeks of CPU time) into a calculation. They weren't initializing all of their memory immediately, and they weren't using calloc, so I put systems into overcommit 'never' mode. These clusters run diskless and have no swap. However, with the current reserves, a user wishing to allocate as much memory as possible to one process may be prevented from using, for example, almost 2GB out of 32GB. The effect is less, but still significant when a user starts a job with one process per core. I have repeatedly seen a set of processes requesting the same amount of memory fail because one of them could not allocate the amount of memory a user would expect to be able to allocate. For example, Message Passing Interfce (MPI) processes, one per core. And it is similar for other parallel programming frameworks. Changing this reserve code will make the overcommit never mode more useful by allowing applications to allocate nearly all of the available memory. Also, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Risks * "bash: fork: Cannot allocate memory" The downside of the first patch-- which creates a tunable user reserve that is only used in overcommit 'never' mode--is that an admin can set it so low that a user may not be able to kill their process, even if they already have a shell prompt. Of course, a user can get in the same predicament with the current 3% reserve--they just have to launch processes until 3% becomes negligible. * root-cant-log-in problem The second patch, adding the tunable rootuser_reserve_pages, allows the admin to shoot themselves in the foot by setting it too small. They can easily get the system into a state where root-can't-log-in. However, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Alternatives * Memory cgroups provide a more flexible way to limit application memory. Not everyone wants to set up cgroups or deal with their overhead. * We could create a fourth overcommit mode which provides smaller reserves. The size of useful reserves may be drastically different depending on the whether the system is embedded or enterprise. * Force users to initialize all of their memory or use calloc. Some users don't want/expect the system to overcommit when they malloc. Overcommit 'never' mode is for this scenario, and it should work well. The new user and admin reserve tunables are simple to use, with low overhead compared to cgroups. The patches preserve current behavior where 3% of memory is less than 128MB, except that the admin reserve doesn't shrink to an unusable size under pressure. The code allows admins to tune for embedded and enterprise usage. FAQ * How is the root-cant-login problem addressed? What happens if admin_reserve_pages is set to 0? Root is free to shoot themselves in the foot by setting admin_reserve_kbytes too low. On x86_64, the minimum useful reserve is: 8MB for overcommit 'guess' 128MB for overcommit 'never' admin_reserve_pages defaults to min(3% free memory, 8MB) So, anyone switching to 'never' mode needs to adjust admin_reserve_pages. * How do you calculate a minimum useful reserve? A user or the admin needs enough memory to login and perform recovery operations, which includes, at a minimum: sshd or login + bash (or some other shell) + top (or ps, kill, etc.) For overcommit 'guess', we can sum resident set sizes (RSS) because we only need enough memory to handle what the recovery programs will typically use. On x86_64 this is about 8MB. For overcommit 'never', we can take the max of their virtual sizes (VSZ) and add the sum of their RSS. We use VSZ instead of RSS because mode forces us to ensure we can fulfill all of the requested memory allocations-- even if the programs only use a fraction of what they ask for. On x86_64 this is about 128MB. When swap is enabled, reserves are useful even when they are as small as 10MB, regardless of overcommit mode. When both swap and overcommit are disabled, then the admin should tune the reserves higher to be absolutley safe. Over 230MB each was safest in my testing. * What happens if user_reserve_pages is set to 0? Note, this only affects overcomitt 'never' mode. Then a user will be able to allocate all available memory minus admin_reserve_kbytes. However, they will easily see a message such as: "bash: fork: Cannot allocate memory" And they won't be able to recover/kill their application. The admin should be able to recover the system if admin_reserve_kbytes is set appropriately. * What's the difference between overcommit 'guess' and 'never'? "Guess" allows an allocation if there are enough free + reclaimable pages. It has a hardcoded 3% of free pages reserved for root. "Never" allows an allocation if there is enough swap + a configurable percentage (default is 50) of physical RAM. It has a hardcoded 3% of free pages reserved for root, like "Guess" mode. It also has a hardcoded 3% of the current process size reserved for additional applications. * Why is overcommit 'guess' not suitable even when an app eventually writes to every page? It takes free pages, file pages, available swap pages, reclaimable slab pages into consideration. In other words, these are all pages available, then why isn't overcommit suitable? Because it only looks at the present state of the system. It does not take into account the memory that other applications have malloced, but haven't initialized yet. It overcommits the system. Test Summary There was little change in behavior in the default overcommit 'guess' mode with swap enabled before and after the patch. This was expected. Systems run most predictably (i.e. no oom kills) in overcommit 'never' mode with swap enabled. This also allowed the most memory to be allocated to a user application. Overcommit 'guess' mode without swap is a bad idea. It is easy to crash the system. None of the other tested combinations crashed. This matches my experience on the Roadrunner supercomputer. Without the tunable user reserve, a system in overcommit 'never' mode and without swap does not allow the admin to recover, although the admin can. With the new tunable reserves, a system in overcommit 'never' mode and without swap can be configured to: 1. maximize user-allocatable memory, running close to the edge of recoverability 2. maximize recoverability, sacrificing allocatable memory to ensure that a user cannot take down a system Test Description Fedora 18 VM - 4 x86_64 cores, 5725MB RAM, 4GB Swap System is booted into multiuser console mode, with unnecessary services turned off. Caches were dropped before each test. Hogs are user memtester processes that attempt to allocate all free memory as reported by /proc/meminfo In overcommit 'never' mode, memory_ratio=100 Test Results 3.9.0-rc1-mm1 Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5432/5432 no yes yes guess yes 4 5444/5444 1 yes yes guess no 1 5302/5449 no yes yes guess no 4 - crash no no never yes 1 5460/5460 1 yes yes never yes 4 5460/5460 1 yes yes never no 1 5218/5432 no no yes never no 4 5203/5448 no no yes 3.9.0-rc1-mm1-tunablereserves User and Admin Recovery show their respective reserves, if applicable. Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5419/5419 no - yes 8MB yes guess yes 4 5436/5436 1 - yes 8MB yes guess no 1 5440/5440 * - yes 8MB yes guess no 4 - crash - no 8MB no * process would successfully mlock, then the oom killer would pick it never yes 1 5446/5446 no 10MB yes 20MB yes never yes 4 5456/5456 no 10MB yes 20MB yes never no 1 5387/5429 no 128MB no 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5359/5448 no 10MB no 10MB barely never no 1 5323/5428 no 0MB no 10MB barely never no 1 5332/5428 no 0MB no 50MB yes never no 1 5293/5429 no 0MB no 90MB yes never no 1 5001/5427 no 230MB yes 338MB yes never no 4* 4998/5424 no 230MB yes 338MB yes * more memtesters were launched, able to allocate approximately another 100MB Future Work - Test larger memory systems. - Test an embedded image. - Test other architectures. - Time malloc microbenchmarks. - Would it be useful to be able to set overcommit policy for each memory cgroup? - Some lines are slightly above 80 chars. Perhaps define a macro to convert between pages and kb? Other places in the kernel do this. [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: make init_user_reserve() static] Signed-off-by: Andrew Shewmaker <agshew@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:10 +02:00
{
unsigned long free_kbytes;
free_kbytes = global_zone_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10);
mm: limit growth of 3% hardcoded other user reserve Add user_reserve_kbytes knob. Limit the growth of the memory reserved for other user processes to min(3% current process size, user_reserve_pages). Only about 8MB is necessary to enable recovery in the default mode, and only a few hundred MB are required even when overcommit is disabled. user_reserve_pages defaults to min(3% free pages, 128MB) I arrived at 128MB by taking the max VSZ of sshd, login, bash, and top ... then adding the RSS of each. This only affects OVERCOMMIT_NEVER mode. Background 1. user reserve __vm_enough_memory reserves a hardcoded 3% of the current process size for other applications when overcommit is disabled. This was done so that a user could recover if they launched a memory hogging process. Without the reserve, a user would easily run into a message such as: bash: fork: Cannot allocate memory 2. admin reserve Additionally, a hardcoded 3% of free memory is reserved for root in both overcommit 'guess' and 'never' modes. This was intended to prevent a scenario where root-cant-log-in and perform recovery operations. Note that this reserve shrinks, and doesn't guarantee a useful reserve. Motivation The two hardcoded memory reserves should be updated to account for current memory sizes. Also, the admin reserve would be more useful if it didn't shrink too much. When the current code was originally written, 1GB was considered "enterprise". Now the 3% reserve can grow to multiple GB on large memory systems, and it only needs to be a few hundred MB at most to enable a user or admin to recover a system with an unwanted memory hogging process. I've found that reducing these reserves is especially beneficial for a specific type of application load: * single application system * one or few processes (e.g. one per core) * allocating all available memory * not initializing every page immediately * long running I've run scientific clusters with this sort of load. A long running job sometimes failed many hours (weeks of CPU time) into a calculation. They weren't initializing all of their memory immediately, and they weren't using calloc, so I put systems into overcommit 'never' mode. These clusters run diskless and have no swap. However, with the current reserves, a user wishing to allocate as much memory as possible to one process may be prevented from using, for example, almost 2GB out of 32GB. The effect is less, but still significant when a user starts a job with one process per core. I have repeatedly seen a set of processes requesting the same amount of memory fail because one of them could not allocate the amount of memory a user would expect to be able to allocate. For example, Message Passing Interfce (MPI) processes, one per core. And it is similar for other parallel programming frameworks. Changing this reserve code will make the overcommit never mode more useful by allowing applications to allocate nearly all of the available memory. Also, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Risks * "bash: fork: Cannot allocate memory" The downside of the first patch-- which creates a tunable user reserve that is only used in overcommit 'never' mode--is that an admin can set it so low that a user may not be able to kill their process, even if they already have a shell prompt. Of course, a user can get in the same predicament with the current 3% reserve--they just have to launch processes until 3% becomes negligible. * root-cant-log-in problem The second patch, adding the tunable rootuser_reserve_pages, allows the admin to shoot themselves in the foot by setting it too small. They can easily get the system into a state where root-can't-log-in. However, the new admin_reserve_kbytes will be safer than the current behavior since the hardcoded 3% of available memory reserve can shrink to something useless in the case where applications have grabbed all available memory. Alternatives * Memory cgroups provide a more flexible way to limit application memory. Not everyone wants to set up cgroups or deal with their overhead. * We could create a fourth overcommit mode which provides smaller reserves. The size of useful reserves may be drastically different depending on the whether the system is embedded or enterprise. * Force users to initialize all of their memory or use calloc. Some users don't want/expect the system to overcommit when they malloc. Overcommit 'never' mode is for this scenario, and it should work well. The new user and admin reserve tunables are simple to use, with low overhead compared to cgroups. The patches preserve current behavior where 3% of memory is less than 128MB, except that the admin reserve doesn't shrink to an unusable size under pressure. The code allows admins to tune for embedded and enterprise usage. FAQ * How is the root-cant-login problem addressed? What happens if admin_reserve_pages is set to 0? Root is free to shoot themselves in the foot by setting admin_reserve_kbytes too low. On x86_64, the minimum useful reserve is: 8MB for overcommit 'guess' 128MB for overcommit 'never' admin_reserve_pages defaults to min(3% free memory, 8MB) So, anyone switching to 'never' mode needs to adjust admin_reserve_pages. * How do you calculate a minimum useful reserve? A user or the admin needs enough memory to login and perform recovery operations, which includes, at a minimum: sshd or login + bash (or some other shell) + top (or ps, kill, etc.) For overcommit 'guess', we can sum resident set sizes (RSS) because we only need enough memory to handle what the recovery programs will typically use. On x86_64 this is about 8MB. For overcommit 'never', we can take the max of their virtual sizes (VSZ) and add the sum of their RSS. We use VSZ instead of RSS because mode forces us to ensure we can fulfill all of the requested memory allocations-- even if the programs only use a fraction of what they ask for. On x86_64 this is about 128MB. When swap is enabled, reserves are useful even when they are as small as 10MB, regardless of overcommit mode. When both swap and overcommit are disabled, then the admin should tune the reserves higher to be absolutley safe. Over 230MB each was safest in my testing. * What happens if user_reserve_pages is set to 0? Note, this only affects overcomitt 'never' mode. Then a user will be able to allocate all available memory minus admin_reserve_kbytes. However, they will easily see a message such as: "bash: fork: Cannot allocate memory" And they won't be able to recover/kill their application. The admin should be able to recover the system if admin_reserve_kbytes is set appropriately. * What's the difference between overcommit 'guess' and 'never'? "Guess" allows an allocation if there are enough free + reclaimable pages. It has a hardcoded 3% of free pages reserved for root. "Never" allows an allocation if there is enough swap + a configurable percentage (default is 50) of physical RAM. It has a hardcoded 3% of free pages reserved for root, like "Guess" mode. It also has a hardcoded 3% of the current process size reserved for additional applications. * Why is overcommit 'guess' not suitable even when an app eventually writes to every page? It takes free pages, file pages, available swap pages, reclaimable slab pages into consideration. In other words, these are all pages available, then why isn't overcommit suitable? Because it only looks at the present state of the system. It does not take into account the memory that other applications have malloced, but haven't initialized yet. It overcommits the system. Test Summary There was little change in behavior in the default overcommit 'guess' mode with swap enabled before and after the patch. This was expected. Systems run most predictably (i.e. no oom kills) in overcommit 'never' mode with swap enabled. This also allowed the most memory to be allocated to a user application. Overcommit 'guess' mode without swap is a bad idea. It is easy to crash the system. None of the other tested combinations crashed. This matches my experience on the Roadrunner supercomputer. Without the tunable user reserve, a system in overcommit 'never' mode and without swap does not allow the admin to recover, although the admin can. With the new tunable reserves, a system in overcommit 'never' mode and without swap can be configured to: 1. maximize user-allocatable memory, running close to the edge of recoverability 2. maximize recoverability, sacrificing allocatable memory to ensure that a user cannot take down a system Test Description Fedora 18 VM - 4 x86_64 cores, 5725MB RAM, 4GB Swap System is booted into multiuser console mode, with unnecessary services turned off. Caches were dropped before each test. Hogs are user memtester processes that attempt to allocate all free memory as reported by /proc/meminfo In overcommit 'never' mode, memory_ratio=100 Test Results 3.9.0-rc1-mm1 Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5432/5432 no yes yes guess yes 4 5444/5444 1 yes yes guess no 1 5302/5449 no yes yes guess no 4 - crash no no never yes 1 5460/5460 1 yes yes never yes 4 5460/5460 1 yes yes never no 1 5218/5432 no no yes never no 4 5203/5448 no no yes 3.9.0-rc1-mm1-tunablereserves User and Admin Recovery show their respective reserves, if applicable. Overcommit | Swap | Hogs | MB Got/Wanted | OOMs | User Recovery | Admin Recovery ---------- ---- ---- ------------- ---- ------------- -------------- guess yes 1 5419/5419 no - yes 8MB yes guess yes 4 5436/5436 1 - yes 8MB yes guess no 1 5440/5440 * - yes 8MB yes guess no 4 - crash - no 8MB no * process would successfully mlock, then the oom killer would pick it never yes 1 5446/5446 no 10MB yes 20MB yes never yes 4 5456/5456 no 10MB yes 20MB yes never no 1 5387/5429 no 128MB no 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5323/5428 no 226MB barely 8MB barely never no 1 5359/5448 no 10MB no 10MB barely never no 1 5323/5428 no 0MB no 10MB barely never no 1 5332/5428 no 0MB no 50MB yes never no 1 5293/5429 no 0MB no 90MB yes never no 1 5001/5427 no 230MB yes 338MB yes never no 4* 4998/5424 no 230MB yes 338MB yes * more memtesters were launched, able to allocate approximately another 100MB Future Work - Test larger memory systems. - Test an embedded image. - Test other architectures. - Time malloc microbenchmarks. - Would it be useful to be able to set overcommit policy for each memory cgroup? - Some lines are slightly above 80 chars. Perhaps define a macro to convert between pages and kb? Other places in the kernel do this. [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: make init_user_reserve() static] Signed-off-by: Andrew Shewmaker <agshew@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 00:08:10 +02:00
sysctl_user_reserve_kbytes = min(free_kbytes / 32, 1UL << 17);
return 0;
}
mm: audit/fix non-modular users of module_init in core code Code that is obj-y (always built-in) or dependent on a bool Kconfig (built-in or absent) can never be modular. So using module_init as an alias for __initcall can be somewhat misleading. Fix these up now, so that we can relocate module_init from init.h into module.h in the future. If we don't do this, we'd have to add module.h to obviously non-modular code, and that would be a worse thing. The audit targets the following module_init users for change: mm/ksm.c bool KSM mm/mmap.c bool MMU mm/huge_memory.c bool TRANSPARENT_HUGEPAGE mm/mmu_notifier.c bool MMU_NOTIFIER Note that direct use of __initcall is discouraged, vs. one of the priority categorized subgroups. As __initcall gets mapped onto device_initcall, our use of subsys_initcall (which makes sense for these files) will thus change this registration from level 6-device to level 4-subsys (i.e. slightly earlier). However no observable impact of that difference has been observed during testing. One might think that core_initcall (l2) or postcore_initcall (l3) would be more appropriate for anything in mm/ but if we look at some actual init functions themselves, we see things like: mm/huge_memory.c --> hugepage_init --> hugepage_init_sysfs mm/mmap.c --> init_user_reserve --> sysctl_user_reserve_kbytes mm/ksm.c --> ksm_init --> sysfs_create_group and hence the choice of subsys_initcall (l4) seems reasonable, and at the same time minimizes the risk of changing the priority too drastically all at once. We can adjust further in the future. Also, several instances of missing ";" at EOL are fixed. Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:30 +01:00
subsys_initcall(init_user_reserve);
/*
* Initialise sysctl_admin_reserve_kbytes.
*
* The purpose of sysctl_admin_reserve_kbytes is to allow the sys admin
* to log in and kill a memory hogging process.
*
* Systems with more than 256MB will reserve 8MB, enough to recover
* with sshd, bash, and top in OVERCOMMIT_GUESS. Smaller systems will
* only reserve 3% of free pages by default.
*/
static int init_admin_reserve(void)
{
unsigned long free_kbytes;
free_kbytes = global_zone_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10);
sysctl_admin_reserve_kbytes = min(free_kbytes / 32, 1UL << 13);
return 0;
}
mm: audit/fix non-modular users of module_init in core code Code that is obj-y (always built-in) or dependent on a bool Kconfig (built-in or absent) can never be modular. So using module_init as an alias for __initcall can be somewhat misleading. Fix these up now, so that we can relocate module_init from init.h into module.h in the future. If we don't do this, we'd have to add module.h to obviously non-modular code, and that would be a worse thing. The audit targets the following module_init users for change: mm/ksm.c bool KSM mm/mmap.c bool MMU mm/huge_memory.c bool TRANSPARENT_HUGEPAGE mm/mmu_notifier.c bool MMU_NOTIFIER Note that direct use of __initcall is discouraged, vs. one of the priority categorized subgroups. As __initcall gets mapped onto device_initcall, our use of subsys_initcall (which makes sense for these files) will thus change this registration from level 6-device to level 4-subsys (i.e. slightly earlier). However no observable impact of that difference has been observed during testing. One might think that core_initcall (l2) or postcore_initcall (l3) would be more appropriate for anything in mm/ but if we look at some actual init functions themselves, we see things like: mm/huge_memory.c --> hugepage_init --> hugepage_init_sysfs mm/mmap.c --> init_user_reserve --> sysctl_user_reserve_kbytes mm/ksm.c --> ksm_init --> sysfs_create_group and hence the choice of subsys_initcall (l4) seems reasonable, and at the same time minimizes the risk of changing the priority too drastically all at once. We can adjust further in the future. Also, several instances of missing ";" at EOL are fixed. Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:30 +01:00
subsys_initcall(init_admin_reserve);
/*
* Reinititalise user and admin reserves if memory is added or removed.
*
* The default user reserve max is 128MB, and the default max for the
* admin reserve is 8MB. These are usually, but not always, enough to
* enable recovery from a memory hogging process using login/sshd, a shell,
* and tools like top. It may make sense to increase or even disable the
* reserve depending on the existence of swap or variations in the recovery
* tools. So, the admin may have changed them.
*
* If memory is added and the reserves have been eliminated or increased above
* the default max, then we'll trust the admin.
*
* If memory is removed and there isn't enough free memory, then we
* need to reset the reserves.
*
* Otherwise keep the reserve set by the admin.
*/
static int reserve_mem_notifier(struct notifier_block *nb,
unsigned long action, void *data)
{
unsigned long tmp, free_kbytes;
switch (action) {
case MEM_ONLINE:
/* Default max is 128MB. Leave alone if modified by operator. */
tmp = sysctl_user_reserve_kbytes;
if (0 < tmp && tmp < (1UL << 17))
init_user_reserve();
/* Default max is 8MB. Leave alone if modified by operator. */
tmp = sysctl_admin_reserve_kbytes;
if (0 < tmp && tmp < (1UL << 13))
init_admin_reserve();
break;
case MEM_OFFLINE:
free_kbytes = global_zone_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10);
if (sysctl_user_reserve_kbytes > free_kbytes) {
init_user_reserve();
pr_info("vm.user_reserve_kbytes reset to %lu\n",
sysctl_user_reserve_kbytes);
}
if (sysctl_admin_reserve_kbytes > free_kbytes) {
init_admin_reserve();
pr_info("vm.admin_reserve_kbytes reset to %lu\n",
sysctl_admin_reserve_kbytes);
}
break;
default:
break;
}
return NOTIFY_OK;
}
static struct notifier_block reserve_mem_nb = {
.notifier_call = reserve_mem_notifier,
};
static int __meminit init_reserve_notifier(void)
{
if (register_hotmemory_notifier(&reserve_mem_nb))
pr_err("Failed registering memory add/remove notifier for admin reserve\n");
return 0;
}
mm: audit/fix non-modular users of module_init in core code Code that is obj-y (always built-in) or dependent on a bool Kconfig (built-in or absent) can never be modular. So using module_init as an alias for __initcall can be somewhat misleading. Fix these up now, so that we can relocate module_init from init.h into module.h in the future. If we don't do this, we'd have to add module.h to obviously non-modular code, and that would be a worse thing. The audit targets the following module_init users for change: mm/ksm.c bool KSM mm/mmap.c bool MMU mm/huge_memory.c bool TRANSPARENT_HUGEPAGE mm/mmu_notifier.c bool MMU_NOTIFIER Note that direct use of __initcall is discouraged, vs. one of the priority categorized subgroups. As __initcall gets mapped onto device_initcall, our use of subsys_initcall (which makes sense for these files) will thus change this registration from level 6-device to level 4-subsys (i.e. slightly earlier). However no observable impact of that difference has been observed during testing. One might think that core_initcall (l2) or postcore_initcall (l3) would be more appropriate for anything in mm/ but if we look at some actual init functions themselves, we see things like: mm/huge_memory.c --> hugepage_init --> hugepage_init_sysfs mm/mmap.c --> init_user_reserve --> sysctl_user_reserve_kbytes mm/ksm.c --> ksm_init --> sysfs_create_group and hence the choice of subsys_initcall (l4) seems reasonable, and at the same time minimizes the risk of changing the priority too drastically all at once. We can adjust further in the future. Also, several instances of missing ";" at EOL are fixed. Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 00:53:30 +01:00
subsys_initcall(init_reserve_notifier);