kpriv/linux-hardened
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
linux-hardened/arch/mips/math-emu/cp1emu.c

2187 lines
53 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
* cp1emu.c: a MIPS coprocessor 1 (fpu) instruction emulator
*
* MIPS floating point support
* Copyright (C) 1994-2000 Algorithmics Ltd.
*
* Kevin D. Kissell, kevink@mips.com and Carsten Langgaard, carstenl@mips.com
* Copyright (C) 2000 MIPS Technologies, Inc.
*
* This program is free software; you can distribute it and/or modify it
* under the terms of the GNU General Public License (Version 2) as
* published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc.,
* 59 Temple Place - Suite 330, Boston MA 02111-1307, USA.
*
* A complete emulator for MIPS coprocessor 1 instructions. This is
* required for #float(switch) or #float(trap), where it catches all
* COP1 instructions via the "CoProcessor Unusable" exception.
*
* More surprisingly it is also required for #float(ieee), to help out
* the hardware fpu at the boundaries of the IEEE-754 representation
* (denormalised values, infinities, underflow, etc). It is made
* quite nasty because emulation of some non-COP1 instructions is
* required, e.g. in branch delay slots.
*
* Note if you know that you won't have an fpu, then you'll get much
* better performance by compiling with -msoft-float!
*/
#include <linux/sched.h>
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
#include <linux/module.h>
[MIPS] Add debugfs files to show fpuemu statistics Export contents of struct mips_fpu_emulator_stats via debugfs. There is no way to read these statistics for now but they (at least the "emulated" count) might be sometimes useful for performance tuning on FPU-less CPUs. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-07 16:21:49 +02:00
#include <linux/debugfs.h>
MIPS: add support for software performance events Software events are required as part of the measurable stuff by the Linux performance counter subsystem. Here is the list of events added by this patch: PERF_COUNT_SW_PAGE_FAULTS PERF_COUNT_SW_PAGE_FAULTS_MIN PERF_COUNT_SW_PAGE_FAULTS_MAJ PERF_COUNT_SW_ALIGNMENT_FAULTS PERF_COUNT_SW_EMULATION_FAULTS Signed-off-by: Deng-Cheng Zhu <dengcheng.zhu@gmail.com> To: linux-mips@linux-mips.org Cc: a.p.zijlstra@chello.nl Cc: paulus@samba.org Cc: mingo@elte.hu Cc: acme@redhat.com Cc: jamie.iles@picochip.com Acked-by: David Daney <ddaney@caviumnetworks.com> Reviewed-by: Matt Fleming <matt@console-pimps.org> Patchwork: https://patchwork.linux-mips.org/patch/1686/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-12 13:37:21 +02:00
#include <linux/perf_event.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#include <asm/inst.h>
#include <asm/bootinfo.h>
#include <asm/processor.h>
#include <asm/ptrace.h>
#include <asm/signal.h>
#include <asm/mipsregs.h>
#include <asm/fpu_emulator.h>
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
#include <asm/fpu.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#include <asm/uaccess.h>
#include <asm/branch.h>
#include "ieee754.h"
/* Strap kernel emulator for full MIPS IV emulation */
#ifdef __mips
#undef __mips
#endif
#define __mips 4
/* Function which emulates a floating point instruction. */
[MIPS] Unify mips_fpu_soft_struct and mips_fpu_hard_structs. The struct mips_fpu_soft_struct and mips_fpu_hard_struct are completely same now and the kernel fpu emulator assumes that. This patch unifies them to mips_fpu_struct and get rid of mips_fpu_union. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-05-15 18:26:03 +02:00
static int fpu_emu(struct pt_regs *, struct mips_fpu_struct *,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mips_instruction);
#if __mips >= 4 && __mips != 32
static int fpux_emu(struct pt_regs *,
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
struct mips_fpu_struct *, mips_instruction, void *__user *);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#endif
[MIPS] Unify mips_fpu_soft_struct and mips_fpu_hard_structs. The struct mips_fpu_soft_struct and mips_fpu_hard_struct are completely same now and the kernel fpu emulator assumes that. This patch unifies them to mips_fpu_struct and get rid of mips_fpu_union. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-05-15 18:26:03 +02:00
/* Further private data for which no space exists in mips_fpu_struct */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
#ifdef CONFIG_DEBUG_FS
DEFINE_PER_CPU(struct mips_fpu_emulator_stats, fpuemustats);
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* Control registers */
#define FPCREG_RID 0 /* $0 = revision id */
#define FPCREG_CSR 31 /* $31 = csr */
MIPS FPU emulator: allow Cause bits of FCSR to be writeable by ctc1 In the FPU emulator code of the MIPS, the Cause bits of the FCSR register are not currently writeable by the ctc1 instruction. In odd corner cases, this can cause problems. For example, a case existed where a divide-by-zero exception was generated by the FPU, and the signal handler attempted to restore the FPU registers to their state before the exception occurred. In this particular setup, writing the old value to the FCSR register would cause another divide-by-zero exception to occur immediately. The solution is to change the ctc1 instruction emulator code to allow the Cause bits of the FCSR register to be writeable. This is the behaviour of the hardware that the code is emulating. This problem was found by Shane McDonald, but the credit for the fix goes to Kevin Kissell. In Kevin's words: I submit that the bug is indeed in that ctc_op: case of the emulator. The Cause bits (17:12) are supposed to be writable by that instruction, but the CTC1 emulation won't let them be updated by the instruction. I think that actually if you just completely removed lines 387-388 [...] things would work a good deal better. At least, it would be a more accurate emulation of the architecturally defined FPU. If I wanted to be really, really pedantic (which I sometimes do), I'd also protect the reserved bits that aren't necessarily writable. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1205/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org> ---
2010-05-07 07:26:57 +02:00
/* Determine rounding mode from the RM bits of the FCSR */
#define modeindex(v) ((v) & FPU_CSR_RM)
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
/* microMIPS bitfields */
#define MM_POOL32A_MINOR_MASK 0x3f
#define MM_POOL32A_MINOR_SHIFT 0x6
#define MM_MIPS32_COND_FC 0x30
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* Convert Mips rounding mode (0..3) to IEEE library modes. */
static const unsigned char ieee_rm[4] = {
Fix preemption and SMP problems in the FP emulator code. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-04-28 15:39:10 +02:00
[FPU_CSR_RN] = IEEE754_RN,
[FPU_CSR_RZ] = IEEE754_RZ,
[FPU_CSR_RU] = IEEE754_RU,
[FPU_CSR_RD] = IEEE754_RD,
};
/* Convert IEEE library modes to Mips rounding mode (0..3). */
static const unsigned char mips_rm[4] = {
[IEEE754_RN] = FPU_CSR_RN,
[IEEE754_RZ] = FPU_CSR_RZ,
[IEEE754_RD] = FPU_CSR_RD,
[IEEE754_RU] = FPU_CSR_RU,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
#if __mips >= 4
/* convert condition code register number to csr bit */
static const unsigned int fpucondbit[8] = {
FPU_CSR_COND0,
FPU_CSR_COND1,
FPU_CSR_COND2,
FPU_CSR_COND3,
FPU_CSR_COND4,
FPU_CSR_COND5,
FPU_CSR_COND6,
FPU_CSR_COND7
};
#endif
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
/* (microMIPS) Convert 16-bit register encoding to 32-bit register encoding. */
static const unsigned int reg16to32map[8] = {16, 17, 2, 3, 4, 5, 6, 7};
/* (microMIPS) Convert certain microMIPS instructions to MIPS32 format. */
static const int sd_format[] = {16, 17, 0, 0, 0, 0, 0, 0};
static const int sdps_format[] = {16, 17, 22, 0, 0, 0, 0, 0};
static const int dwl_format[] = {17, 20, 21, 0, 0, 0, 0, 0};
static const int swl_format[] = {16, 20, 21, 0, 0, 0, 0, 0};
/*
* This functions translates a 32-bit microMIPS instruction
* into a 32-bit MIPS32 instruction. Returns 0 on success
* and SIGILL otherwise.
*/
static int microMIPS32_to_MIPS32(union mips_instruction *insn_ptr)
{
union mips_instruction insn = *insn_ptr;
union mips_instruction mips32_insn = insn;
int func, fmt, op;
switch (insn.mm_i_format.opcode) {
case mm_ldc132_op:
mips32_insn.mm_i_format.opcode = ldc1_op;
mips32_insn.mm_i_format.rt = insn.mm_i_format.rs;
mips32_insn.mm_i_format.rs = insn.mm_i_format.rt;
break;
case mm_lwc132_op:
mips32_insn.mm_i_format.opcode = lwc1_op;
mips32_insn.mm_i_format.rt = insn.mm_i_format.rs;
mips32_insn.mm_i_format.rs = insn.mm_i_format.rt;
break;
case mm_sdc132_op:
mips32_insn.mm_i_format.opcode = sdc1_op;
mips32_insn.mm_i_format.rt = insn.mm_i_format.rs;
mips32_insn.mm_i_format.rs = insn.mm_i_format.rt;
break;
case mm_swc132_op:
mips32_insn.mm_i_format.opcode = swc1_op;
mips32_insn.mm_i_format.rt = insn.mm_i_format.rs;
mips32_insn.mm_i_format.rs = insn.mm_i_format.rt;
break;
case mm_pool32i_op:
/* NOTE: offset is << by 1 if in microMIPS mode. */
if ((insn.mm_i_format.rt == mm_bc1f_op) ||
(insn.mm_i_format.rt == mm_bc1t_op)) {
mips32_insn.fb_format.opcode = cop1_op;
mips32_insn.fb_format.bc = bc_op;
mips32_insn.fb_format.flag =
(insn.mm_i_format.rt == mm_bc1t_op) ? 1 : 0;
} else
return SIGILL;
break;
case mm_pool32f_op:
switch (insn.mm_fp0_format.func) {
case mm_32f_01_op:
case mm_32f_11_op:
case mm_32f_02_op:
case mm_32f_12_op:
case mm_32f_41_op:
case mm_32f_51_op:
case mm_32f_42_op:
case mm_32f_52_op:
op = insn.mm_fp0_format.func;
if (op == mm_32f_01_op)
func = madd_s_op;
else if (op == mm_32f_11_op)
func = madd_d_op;
else if (op == mm_32f_02_op)
func = nmadd_s_op;
else if (op == mm_32f_12_op)
func = nmadd_d_op;
else if (op == mm_32f_41_op)
func = msub_s_op;
else if (op == mm_32f_51_op)
func = msub_d_op;
else if (op == mm_32f_42_op)
func = nmsub_s_op;
else
func = nmsub_d_op;
mips32_insn.fp6_format.opcode = cop1x_op;
mips32_insn.fp6_format.fr = insn.mm_fp6_format.fr;
mips32_insn.fp6_format.ft = insn.mm_fp6_format.ft;
mips32_insn.fp6_format.fs = insn.mm_fp6_format.fs;
mips32_insn.fp6_format.fd = insn.mm_fp6_format.fd;
mips32_insn.fp6_format.func = func;
break;
case mm_32f_10_op:
func = -1; /* Invalid */
op = insn.mm_fp5_format.op & 0x7;
if (op == mm_ldxc1_op)
func = ldxc1_op;
else if (op == mm_sdxc1_op)
func = sdxc1_op;
else if (op == mm_lwxc1_op)
func = lwxc1_op;
else if (op == mm_swxc1_op)
func = swxc1_op;
if (func != -1) {
mips32_insn.r_format.opcode = cop1x_op;
mips32_insn.r_format.rs =
insn.mm_fp5_format.base;
mips32_insn.r_format.rt =
insn.mm_fp5_format.index;
mips32_insn.r_format.rd = 0;
mips32_insn.r_format.re = insn.mm_fp5_format.fd;
mips32_insn.r_format.func = func;
} else
return SIGILL;
break;
case mm_32f_40_op:
op = -1; /* Invalid */
if (insn.mm_fp2_format.op == mm_fmovt_op)
op = 1;
else if (insn.mm_fp2_format.op == mm_fmovf_op)
op = 0;
if (op != -1) {
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp2_format.fmt];
mips32_insn.fp0_format.ft =
(insn.mm_fp2_format.cc<<2) + op;
mips32_insn.fp0_format.fs =
insn.mm_fp2_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp2_format.fd;
mips32_insn.fp0_format.func = fmovc_op;
} else
return SIGILL;
break;
case mm_32f_60_op:
func = -1; /* Invalid */
if (insn.mm_fp0_format.op == mm_fadd_op)
func = fadd_op;
else if (insn.mm_fp0_format.op == mm_fsub_op)
func = fsub_op;
else if (insn.mm_fp0_format.op == mm_fmul_op)
func = fmul_op;
else if (insn.mm_fp0_format.op == mm_fdiv_op)
func = fdiv_op;
if (func != -1) {
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp0_format.fmt];
mips32_insn.fp0_format.ft =
insn.mm_fp0_format.ft;
mips32_insn.fp0_format.fs =
insn.mm_fp0_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp0_format.fd;
mips32_insn.fp0_format.func = func;
} else
return SIGILL;
break;
case mm_32f_70_op:
func = -1; /* Invalid */
if (insn.mm_fp0_format.op == mm_fmovn_op)
func = fmovn_op;
else if (insn.mm_fp0_format.op == mm_fmovz_op)
func = fmovz_op;
if (func != -1) {
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp0_format.fmt];
mips32_insn.fp0_format.ft =
insn.mm_fp0_format.ft;
mips32_insn.fp0_format.fs =
insn.mm_fp0_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp0_format.fd;
mips32_insn.fp0_format.func = func;
} else
return SIGILL;
break;
case mm_32f_73_op: /* POOL32FXF */
switch (insn.mm_fp1_format.op) {
case mm_movf0_op:
case mm_movf1_op:
case mm_movt0_op:
case mm_movt1_op:
if ((insn.mm_fp1_format.op & 0x7f) ==
mm_movf0_op)
op = 0;
else
op = 1;
mips32_insn.r_format.opcode = spec_op;
mips32_insn.r_format.rs = insn.mm_fp4_format.fs;
mips32_insn.r_format.rt =
(insn.mm_fp4_format.cc << 2) + op;
mips32_insn.r_format.rd = insn.mm_fp4_format.rt;
mips32_insn.r_format.re = 0;
mips32_insn.r_format.func = movc_op;
break;
case mm_fcvtd0_op:
case mm_fcvtd1_op:
case mm_fcvts0_op:
case mm_fcvts1_op:
if ((insn.mm_fp1_format.op & 0x7f) ==
mm_fcvtd0_op) {
func = fcvtd_op;
fmt = swl_format[insn.mm_fp3_format.fmt];
} else {
func = fcvts_op;
fmt = dwl_format[insn.mm_fp3_format.fmt];
}
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt = fmt;
mips32_insn.fp0_format.ft = 0;
mips32_insn.fp0_format.fs =
insn.mm_fp3_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp3_format.rt;
mips32_insn.fp0_format.func = func;
break;
case mm_fmov0_op:
case mm_fmov1_op:
case mm_fabs0_op:
case mm_fabs1_op:
case mm_fneg0_op:
case mm_fneg1_op:
if ((insn.mm_fp1_format.op & 0x7f) ==
mm_fmov0_op)
func = fmov_op;
else if ((insn.mm_fp1_format.op & 0x7f) ==
mm_fabs0_op)
func = fabs_op;
else
func = fneg_op;
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp3_format.fmt];
mips32_insn.fp0_format.ft = 0;
mips32_insn.fp0_format.fs =
insn.mm_fp3_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp3_format.rt;
mips32_insn.fp0_format.func = func;
break;
case mm_ffloorl_op:
case mm_ffloorw_op:
case mm_fceill_op:
case mm_fceilw_op:
case mm_ftruncl_op:
case mm_ftruncw_op:
case mm_froundl_op:
case mm_froundw_op:
case mm_fcvtl_op:
case mm_fcvtw_op:
if (insn.mm_fp1_format.op == mm_ffloorl_op)
func = ffloorl_op;
else if (insn.mm_fp1_format.op == mm_ffloorw_op)
func = ffloor_op;
else if (insn.mm_fp1_format.op == mm_fceill_op)
func = fceill_op;
else if (insn.mm_fp1_format.op == mm_fceilw_op)
func = fceil_op;
else if (insn.mm_fp1_format.op == mm_ftruncl_op)
func = ftruncl_op;
else if (insn.mm_fp1_format.op == mm_ftruncw_op)
func = ftrunc_op;
else if (insn.mm_fp1_format.op == mm_froundl_op)
func = froundl_op;
else if (insn.mm_fp1_format.op == mm_froundw_op)
func = fround_op;
else if (insn.mm_fp1_format.op == mm_fcvtl_op)
func = fcvtl_op;
else
func = fcvtw_op;
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sd_format[insn.mm_fp1_format.fmt];
mips32_insn.fp0_format.ft = 0;
mips32_insn.fp0_format.fs =
insn.mm_fp1_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp1_format.rt;
mips32_insn.fp0_format.func = func;
break;
case mm_frsqrt_op:
case mm_fsqrt_op:
case mm_frecip_op:
if (insn.mm_fp1_format.op == mm_frsqrt_op)
func = frsqrt_op;
else if (insn.mm_fp1_format.op == mm_fsqrt_op)
func = fsqrt_op;
else
func = frecip_op;
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp1_format.fmt];
mips32_insn.fp0_format.ft = 0;
mips32_insn.fp0_format.fs =
insn.mm_fp1_format.fs;
mips32_insn.fp0_format.fd =
insn.mm_fp1_format.rt;
mips32_insn.fp0_format.func = func;
break;
case mm_mfc1_op:
case mm_mtc1_op:
case mm_cfc1_op:
case mm_ctc1_op:
MIPS: microMIPS: mfhc1 & mthc1 support for the FPU emulator This patch adds support for microMIPS encodings of the mfhc1 & mthc1 instructions introduced in release 2 of the mips32 & mips64 architectures, converting them to their mips32 equivalents for the FPU emulator. Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6110/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:29 +01:00
case mm_mfhc1_op:
case mm_mthc1_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (insn.mm_fp1_format.op == mm_mfc1_op)
op = mfc_op;
else if (insn.mm_fp1_format.op == mm_mtc1_op)
op = mtc_op;
else if (insn.mm_fp1_format.op == mm_cfc1_op)
op = cfc_op;
MIPS: microMIPS: mfhc1 & mthc1 support for the FPU emulator This patch adds support for microMIPS encodings of the mfhc1 & mthc1 instructions introduced in release 2 of the mips32 & mips64 architectures, converting them to their mips32 equivalents for the FPU emulator. Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6110/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:29 +01:00
else if (insn.mm_fp1_format.op == mm_ctc1_op)
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
op = ctc_op;
MIPS: microMIPS: mfhc1 & mthc1 support for the FPU emulator This patch adds support for microMIPS encodings of the mfhc1 & mthc1 instructions introduced in release 2 of the mips32 & mips64 architectures, converting them to their mips32 equivalents for the FPU emulator. Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6110/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:29 +01:00
else if (insn.mm_fp1_format.op == mm_mfhc1_op)
op = mfhc_op;
else
op = mthc_op;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
mips32_insn.fp1_format.opcode = cop1_op;
mips32_insn.fp1_format.op = op;
mips32_insn.fp1_format.rt =
insn.mm_fp1_format.rt;
mips32_insn.fp1_format.fs =
insn.mm_fp1_format.fs;
mips32_insn.fp1_format.fd = 0;
mips32_insn.fp1_format.func = 0;
break;
default:
return SIGILL;
}
break;
case mm_32f_74_op: /* c.cond.fmt */
mips32_insn.fp0_format.opcode = cop1_op;
mips32_insn.fp0_format.fmt =
sdps_format[insn.mm_fp4_format.fmt];
mips32_insn.fp0_format.ft = insn.mm_fp4_format.rt;
mips32_insn.fp0_format.fs = insn.mm_fp4_format.fs;
mips32_insn.fp0_format.fd = insn.mm_fp4_format.cc << 2;
mips32_insn.fp0_format.func =
insn.mm_fp4_format.cond | MM_MIPS32_COND_FC;
break;
default:
return SIGILL;
}
break;
default:
return SIGILL;
}
*insn_ptr = mips32_insn;
return 0;
}
int mm_isBranchInstr(struct pt_regs *regs, struct mm_decoded_insn dec_insn,
unsigned long *contpc)
{
union mips_instruction insn = (union mips_instruction)dec_insn.insn;
int bc_false = 0;
unsigned int fcr31;
unsigned int bit;
MIPS: Don't try to decode microMIPS branch instructions where they cannot exist. In mm_isBranchInstr() we can short circuit the entire function if !cpu_has_mmips. Signed-off-by: David Daney <david.daney@cavium.com> Acked-by: Steven J. Hill <Steven.Hill@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/5326/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-05-24 22:54:09 +02:00
if (!cpu_has_mmips)
return 0;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
switch (insn.mm_i_format.opcode) {
case mm_pool32a_op:
if ((insn.mm_i_format.simmediate & MM_POOL32A_MINOR_MASK) ==
mm_pool32axf_op) {
switch (insn.mm_i_format.simmediate >>
MM_POOL32A_MINOR_SHIFT) {
case mm_jalr_op:
case mm_jalrhb_op:
case mm_jalrs_op:
case mm_jalrshb_op:
if (insn.mm_i_format.rt != 0) /* Not mm_jr */
regs->regs[insn.mm_i_format.rt] =
regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
*contpc = regs->regs[insn.mm_i_format.rs];
return 1;
}
}
break;
case mm_pool32i_op:
switch (insn.mm_i_format.rt) {
case mm_bltzals_op:
case mm_bltzal_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
case mm_bltz_op:
if ((long)regs->regs[insn.mm_i_format.rs] < 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case mm_bgezals_op:
case mm_bgezal_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
case mm_bgez_op:
if ((long)regs->regs[insn.mm_i_format.rs] >= 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case mm_blez_op:
if ((long)regs->regs[insn.mm_i_format.rs] <= 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case mm_bgtz_op:
if ((long)regs->regs[insn.mm_i_format.rs] <= 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case mm_bc2f_op:
case mm_bc1f_op:
bc_false = 1;
/* Fall through */
case mm_bc2t_op:
case mm_bc1t_op:
preempt_disable();
if (is_fpu_owner())
asm volatile("cfc1\t%0,$31" : "=r" (fcr31));
else
fcr31 = current->thread.fpu.fcr31;
preempt_enable();
if (bc_false)
fcr31 = ~fcr31;
bit = (insn.mm_i_format.rs >> 2);
bit += (bit != 0);
bit += 23;
if (fcr31 & (1 << bit))
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
return 1;
}
break;
case mm_pool16c_op:
switch (insn.mm_i_format.rt) {
case mm_jalr16_op:
case mm_jalrs16_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
/* Fall through */
case mm_jr16_op:
*contpc = regs->regs[insn.mm_i_format.rs];
return 1;
}
break;
case mm_beqz16_op:
if ((long)regs->regs[reg16to32map[insn.mm_b1_format.rs]] == 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_b1_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
return 1;
case mm_bnez16_op:
if ((long)regs->regs[reg16to32map[insn.mm_b1_format.rs]] != 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_b1_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
return 1;
case mm_b16_op:
*contpc = regs->cp0_epc + dec_insn.pc_inc +
(insn.mm_b0_format.simmediate << 1);
return 1;
case mm_beq32_op:
if (regs->regs[insn.mm_i_format.rs] ==
regs->regs[insn.mm_i_format.rt])
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case mm_bne32_op:
if (regs->regs[insn.mm_i_format.rs] !=
regs->regs[insn.mm_i_format.rt])
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.mm_i_format.simmediate << 1);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
return 1;
case mm_jalx32_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
*contpc = regs->cp0_epc + dec_insn.pc_inc;
*contpc >>= 28;
*contpc <<= 28;
*contpc |= (insn.j_format.target << 2);
return 1;
case mm_jals32_op:
case mm_jal32_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc + dec_insn.next_pc_inc;
/* Fall through */
case mm_j32_op:
*contpc = regs->cp0_epc + dec_insn.pc_inc;
*contpc >>= 27;
*contpc <<= 27;
*contpc |= (insn.j_format.target << 1);
set_isa16_mode(*contpc);
return 1;
}
return 0;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
* Redundant with logic already in kernel/branch.c,
* embedded in compute_return_epc. At some point,
* a single subroutine should be used across both
* modules.
*/
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
static int isBranchInstr(struct pt_regs *regs, struct mm_decoded_insn dec_insn,
unsigned long *contpc)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
{
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
union mips_instruction insn = (union mips_instruction)dec_insn.insn;
unsigned int fcr31;
unsigned int bit = 0;
switch (insn.i_format.opcode) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case spec_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
switch (insn.r_format.func) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case jalr_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
regs->regs[insn.r_format.rd] =
regs->cp0_epc + dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case jr_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
*contpc = regs->regs[insn.r_format.rs];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return 1;
}
break;
case bcond_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
switch (insn.i_format.rt) {
case bltzal_op:
case bltzall_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case bltz_op:
case bltzl_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if ((long)regs->regs[insn.i_format.rs] < 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case bgezal_op:
case bgezall_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
case bgez_op:
case bgezl_op:
if ((long)regs->regs[insn.i_format.rs] >= 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return 1;
}
break;
case jalx_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
set_isa16_mode(bit);
case jal_op:
regs->regs[31] = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
/* Fall through */
case j_op:
*contpc = regs->cp0_epc + dec_insn.pc_inc;
*contpc >>= 28;
*contpc <<= 28;
*contpc |= (insn.j_format.target << 2);
/* Set microMIPS mode bit: XOR for jalx. */
*contpc ^= bit;
return 1;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case beq_op:
case beql_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (regs->regs[insn.i_format.rs] ==
regs->regs[insn.i_format.rt])
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case bne_op:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case bnel_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (regs->regs[insn.i_format.rs] !=
regs->regs[insn.i_format.rt])
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case blez_op:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case blezl_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if ((long)regs->regs[insn.i_format.rs] <= 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case bgtz_op:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case bgtzl_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if ((long)regs->regs[insn.i_format.rs] > 0)
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return 1;
MIPS: Handle OCTEON BBIT instructions in FPU emulator. The branch emulation needs to handle the OCTEON BBIT instructions, otherwise we get SIGILL instead of emulation. Signed-off-by: David Daney <david.daney@cavium.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/5726/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-08-19 21:10:34 +02:00
#ifdef CONFIG_CPU_CAVIUM_OCTEON
case lwc2_op: /* This is bbit0 on Octeon */
if ((regs->regs[insn.i_format.rs] & (1ull<<insn.i_format.rt)) == 0)
*contpc = regs->cp0_epc + 4 + (insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc + 8;
return 1;
case ldc2_op: /* This is bbit032 on Octeon */
if ((regs->regs[insn.i_format.rs] & (1ull<<(insn.i_format.rt + 32))) == 0)
*contpc = regs->cp0_epc + 4 + (insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc + 8;
return 1;
case swc2_op: /* This is bbit1 on Octeon */
if (regs->regs[insn.i_format.rs] & (1ull<<insn.i_format.rt))
*contpc = regs->cp0_epc + 4 + (insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc + 8;
return 1;
case sdc2_op: /* This is bbit132 on Octeon */
if (regs->regs[insn.i_format.rs] & (1ull<<(insn.i_format.rt + 32)))
*contpc = regs->cp0_epc + 4 + (insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc + 8;
return 1;
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case cop0_op:
case cop1_op:
case cop2_op:
case cop1x_op:
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (insn.i_format.rs == bc_op) {
preempt_disable();
if (is_fpu_owner())
asm volatile("cfc1\t%0,$31" : "=r" (fcr31));
else
fcr31 = current->thread.fpu.fcr31;
preempt_enable();
bit = (insn.i_format.rt >> 2);
bit += (bit != 0);
bit += 23;
switch (insn.i_format.rt & 3) {
case 0: /* bc1f */
case 2: /* bc1fl */
if (~fcr31 & (1 << bit))
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
case 1: /* bc1t */
case 3: /* bc1tl */
if (fcr31 & (1 << bit))
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
(insn.i_format.simmediate << 2);
else
*contpc = regs->cp0_epc +
dec_insn.pc_inc +
dec_insn.next_pc_inc;
return 1;
}
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
}
return 0;
}
/*
* In the Linux kernel, we support selection of FPR format on the
MIPS: Whitespace cleanup. Having received another series of whitespace patches I decided to do this once and for all rather than dealing with this kind of patches trickling in forever. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-01-22 12:59:30 +01:00
* basis of the Status.FR bit. If an FPU is not present, the FR bit
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
* is hardwired to zero, which would imply a 32-bit FPU even for
MIPS: Support for 64-bit FP with O32 binaries CPUs implementing MIPS32 R2 may include a 64-bit FPU, just as MIPS64 CPUs do. In order to preserve backwards compatibility a 64-bit FPU will act like a 32-bit FPU (by accessing doubles from the least significant 32 bits of an even-odd pair of FP registers) when the Status.FR bit is zero, again just like a mips64 CPU. The standard O32 ABI is defined expecting a 32-bit FPU, however recent toolchains support use of a 64-bit FPU from an O32 MIPS32 executable. When an ELF executable is built to use a 64-bit FPU a new flag (EF_MIPS_FP64) is set in the ELF header. With this patch the kernel will check the EF_MIPS_FP64 flag when executing an O32 binary, and set Status.FR accordingly. The addition of O32 64-bit FP support lessens the opportunity for optimisation in the FPU emulator, so a CONFIG_MIPS_O32_FP64_SUPPORT Kconfig option is introduced to allow this support to be disabled for those that don't require it. Inspired by an earlier patch by Leonid Yegoshin, but implemented more cleanly & correctly. Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Cc: Paul Burton <paul.burton@imgtec.com> Patchwork: https://patchwork.linux-mips.org/patch/6154/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-22 14:12:07 +01:00
* 64-bit CPUs so we rather look at TIF_32BIT_FPREGS.
MIPS: Fix for warning from FPU emulation code The default implementation of 'cpu_has_fpu' macro calls smp_processor_id() which causes this warning to be printed when preemption is enabled: [ 4.664000] Algorithmics/MIPS FPU Emulator v1.5 [ 4.676000] BUG: using smp_processor_id() in preemptible [00000000] code: ini [ 4.700000] caller is fpu_emulator_cop1Handler+0x434/0x27b8 This problem got introduced in November 2009 by af1d2af877ef6c36990671bc86a5b9c5bb50b1da (lmo) [MIPS: Fix emulation of 64-bit FPU on 64-bit CPUs.] rsp. da0bac33413b2888d3623dad3ad19ce76b688f07 (kernel.org) [MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs.] in 2.6.32. Fixed by rewriting cop1_64bit() to return a constant whenever possible but most importantly avoid the use pf cpu_has_fpu entirely. Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Reported-by: Jayachandran C <jchandra@broadcom.com> Initial-patch-by: Jayachandran C <jchandra@broadcom.com> Patchwork: https://patchwork.linux-mips.org/patch/4225/
2012-08-15 19:42:19 +02:00
* FPU emu is slow and bulky and optimizing this function offers fairly
* sizeable benefits so we try to be clever and make this function return
* a constant whenever possible, that is on 64-bit kernels without O32
MIPS: Support for 64-bit FP with O32 binaries CPUs implementing MIPS32 R2 may include a 64-bit FPU, just as MIPS64 CPUs do. In order to preserve backwards compatibility a 64-bit FPU will act like a 32-bit FPU (by accessing doubles from the least significant 32 bits of an even-odd pair of FP registers) when the Status.FR bit is zero, again just like a mips64 CPU. The standard O32 ABI is defined expecting a 32-bit FPU, however recent toolchains support use of a 64-bit FPU from an O32 MIPS32 executable. When an ELF executable is built to use a 64-bit FPU a new flag (EF_MIPS_FP64) is set in the ELF header. With this patch the kernel will check the EF_MIPS_FP64 flag when executing an O32 binary, and set Status.FR accordingly. The addition of O32 64-bit FP support lessens the opportunity for optimisation in the FPU emulator, so a CONFIG_MIPS_O32_FP64_SUPPORT Kconfig option is introduced to allow this support to be disabled for those that don't require it. Inspired by an earlier patch by Leonid Yegoshin, but implemented more cleanly & correctly. Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Cc: Paul Burton <paul.burton@imgtec.com> Patchwork: https://patchwork.linux-mips.org/patch/6154/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-22 14:12:07 +01:00
* compatibility enabled and on 32-bit without 64-bit FPU support.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
*/
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
static inline int cop1_64bit(struct pt_regs *xcp)
{
MIPS: Fix for warning from FPU emulation code The default implementation of 'cpu_has_fpu' macro calls smp_processor_id() which causes this warning to be printed when preemption is enabled: [ 4.664000] Algorithmics/MIPS FPU Emulator v1.5 [ 4.676000] BUG: using smp_processor_id() in preemptible [00000000] code: ini [ 4.700000] caller is fpu_emulator_cop1Handler+0x434/0x27b8 This problem got introduced in November 2009 by af1d2af877ef6c36990671bc86a5b9c5bb50b1da (lmo) [MIPS: Fix emulation of 64-bit FPU on 64-bit CPUs.] rsp. da0bac33413b2888d3623dad3ad19ce76b688f07 (kernel.org) [MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs.] in 2.6.32. Fixed by rewriting cop1_64bit() to return a constant whenever possible but most importantly avoid the use pf cpu_has_fpu entirely. Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Reported-by: Jayachandran C <jchandra@broadcom.com> Initial-patch-by: Jayachandran C <jchandra@broadcom.com> Patchwork: https://patchwork.linux-mips.org/patch/4225/
2012-08-15 19:42:19 +02:00
#if defined(CONFIG_64BIT) && !defined(CONFIG_MIPS32_O32)
return 1;
MIPS: Support for 64-bit FP with O32 binaries CPUs implementing MIPS32 R2 may include a 64-bit FPU, just as MIPS64 CPUs do. In order to preserve backwards compatibility a 64-bit FPU will act like a 32-bit FPU (by accessing doubles from the least significant 32 bits of an even-odd pair of FP registers) when the Status.FR bit is zero, again just like a mips64 CPU. The standard O32 ABI is defined expecting a 32-bit FPU, however recent toolchains support use of a 64-bit FPU from an O32 MIPS32 executable. When an ELF executable is built to use a 64-bit FPU a new flag (EF_MIPS_FP64) is set in the ELF header. With this patch the kernel will check the EF_MIPS_FP64 flag when executing an O32 binary, and set Status.FR accordingly. The addition of O32 64-bit FP support lessens the opportunity for optimisation in the FPU emulator, so a CONFIG_MIPS_O32_FP64_SUPPORT Kconfig option is introduced to allow this support to be disabled for those that don't require it. Inspired by an earlier patch by Leonid Yegoshin, but implemented more cleanly & correctly. Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Cc: Paul Burton <paul.burton@imgtec.com> Patchwork: https://patchwork.linux-mips.org/patch/6154/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-22 14:12:07 +01:00
#elif defined(CONFIG_32BIT) && !defined(CONFIG_MIPS_O32_FP64_SUPPORT)
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
return 0;
MIPS: Support for 64-bit FP with O32 binaries CPUs implementing MIPS32 R2 may include a 64-bit FPU, just as MIPS64 CPUs do. In order to preserve backwards compatibility a 64-bit FPU will act like a 32-bit FPU (by accessing doubles from the least significant 32 bits of an even-odd pair of FP registers) when the Status.FR bit is zero, again just like a mips64 CPU. The standard O32 ABI is defined expecting a 32-bit FPU, however recent toolchains support use of a 64-bit FPU from an O32 MIPS32 executable. When an ELF executable is built to use a 64-bit FPU a new flag (EF_MIPS_FP64) is set in the ELF header. With this patch the kernel will check the EF_MIPS_FP64 flag when executing an O32 binary, and set Status.FR accordingly. The addition of O32 64-bit FP support lessens the opportunity for optimisation in the FPU emulator, so a CONFIG_MIPS_O32_FP64_SUPPORT Kconfig option is introduced to allow this support to be disabled for those that don't require it. Inspired by an earlier patch by Leonid Yegoshin, but implemented more cleanly & correctly. Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Cc: Paul Burton <paul.burton@imgtec.com> Patchwork: https://patchwork.linux-mips.org/patch/6154/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-22 14:12:07 +01:00
#else
return !test_thread_flag(TIF_32BIT_FPREGS);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#endif
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
}
#define SIFROMREG(si, x) ((si) = cop1_64bit(xcp) || !(x & 1) ? \
(int)ctx->fpr[x] : (int)(ctx->fpr[x & ~1] >> 32))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
#define SITOREG(si, x) (ctx->fpr[x & ~(cop1_64bit(xcp) == 0)] = \
cop1_64bit(xcp) || !(x & 1) ? \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ctx->fpr[x & ~1] >> 32 << 32 | (u32)(si) : \
ctx->fpr[x & ~1] << 32 >> 32 | (u64)(si) << 32)
MIPS: math-emu: Add mfhc1 & mthc1 support. This patch adds support for the mfhc1 & mthc1 instructions to the FPU emulator. These instructions were introduced in release 2 of the MIPS32 & MIPS64 architectures and allow access to the most significant 32 bits of a 64-bit FP register. [ralf@linux-mips.org: Fix ifdef hell added by original patch.] Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6112/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:28 +01:00
#define SIFROMHREG(si, x) ((si) = (int)(ctx->fpr[x] >> 32))
#define SITOHREG(si, x) (ctx->fpr[x] = \
ctx->fpr[x] << 32 >> 32 | (u64)(si) << 32)
MIPS: Fix emulation of 64-bit FPU on FPU-less 64-bit CPUs. Running a 64-bit kernel on a 64-bit CPU without an FPU would cause the emulator to run in 32-bit mode. The c0_Status.FR bit is wired to zero on systems without an FPU, so using that bit to decide how the emulator behaves doesn't allow for proper emulation on 64-bit FPU-less processors. Instead, we need to select the emulator mode based on the user-space ABI. Since the thread flag TIF_32BIT_REGS is used to set c0_Status.FR, we can just use it to decide if the emulator should be in 32-bit or 64-bit mode. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-02 20:33:46 +01:00
#define DIFROMREG(di, x) ((di) = ctx->fpr[x & ~(cop1_64bit(xcp) == 0)])
#define DITOREG(di, x) (ctx->fpr[x & ~(cop1_64bit(xcp) == 0)] = (di))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[MIPS] checkfiles: Fix "need space after that ','" errors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
#define SPFROMREG(sp, x) SIFROMREG((sp).bits, x)
#define SPTOREG(sp, x) SITOREG((sp).bits, x)
#define DPFROMREG(dp, x) DIFROMREG((dp).bits, x)
#define DPTOREG(dp, x) DITOREG((dp).bits, x)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
* Emulate the single floating point instruction pointed at by EPC.
* Two instructions if the instruction is in a branch delay slot.
*/
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
static int cop1Emulate(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
struct mm_decoded_insn dec_insn, void *__user *fault_addr)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
{
mips_instruction ir;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
unsigned long contpc = xcp->cp0_epc + dec_insn.pc_inc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
unsigned int cond;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
int pc_inc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* XXX NEC Vr54xx bug workaround */
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (xcp->cp0_cause & CAUSEF_BD) {
if (dec_insn.micro_mips_mode) {
if (!mm_isBranchInstr(xcp, dec_insn, &contpc))
xcp->cp0_cause &= ~CAUSEF_BD;
} else {
if (!isBranchInstr(xcp, dec_insn, &contpc))
xcp->cp0_cause &= ~CAUSEF_BD;
}
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
if (xcp->cp0_cause & CAUSEF_BD) {
/*
* The instruction to be emulated is in a branch delay slot
MIPS: Whitespace cleanup. Having received another series of whitespace patches I decided to do this once and for all rather than dealing with this kind of patches trickling in forever. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-01-22 12:59:30 +01:00
* which means that we have to emulate the branch instruction
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
* BEFORE we do the cop1 instruction.
*
* This branch could be a COP1 branch, but in that case we
* would have had a trap for that instruction, and would not
* come through this route.
*
* Linux MIPS branch emulator operates on context, updating the
* cp0_epc.
*/
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
ir = dec_insn.next_insn; /* process delay slot instr */
pc_inc = dec_insn.next_pc_inc;
} else {
ir = dec_insn.insn; /* process current instr */
pc_inc = dec_insn.pc_inc;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
/*
* Since microMIPS FPU instructios are a subset of MIPS32 FPU
* instructions, we want to convert microMIPS FPU instructions
* into MIPS32 instructions so that we could reuse all of the
* FPU emulation code.
*
* NOTE: We cannot do this for branch instructions since they
* are not a subset. Example: Cannot emulate a 16-bit
* aligned target address with a MIPS32 instruction.
*/
if (dec_insn.micro_mips_mode) {
/*
* If next instruction is a 16-bit instruction, then it
* it cannot be a FPU instruction. This could happen
* since we can be called for non-FPU instructions.
*/
if ((pc_inc == 2) ||
(microMIPS32_to_MIPS32((union mips_instruction *)&ir)
== SIGILL))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGILL;
}
emul:
perf: Remove the nmi parameter from the swevent and overflow interface The nmi parameter indicated if we could do wakeups from the current context, if not, we would set some state and self-IPI and let the resulting interrupt do the wakeup. For the various event classes: - hardware: nmi=0; PMI is in fact an NMI or we run irq_work_run from the PMI-tail (ARM etc.) - tracepoint: nmi=0; since tracepoint could be from NMI context. - software: nmi=[0,1]; some, like the schedule thing cannot perform wakeups, and hence need 0. As one can see, there is very little nmi=1 usage, and the down-side of not using it is that on some platforms some software events can have a jiffy delay in wakeup (when arch_irq_work_raise isn't implemented). The up-side however is that we can remove the nmi parameter and save a bunch of conditionals in fast paths. Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Michael Cree <mcree@orcon.net.nz> Cc: Will Deacon <will.deacon@arm.com> Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com> Cc: Anton Blanchard <anton@samba.org> Cc: Eric B Munson <emunson@mgebm.net> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: David S. Miller <davem@davemloft.net> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Wessel <jason.wessel@windriver.com> Cc: Don Zickus <dzickus@redhat.com> Link: http://lkml.kernel.org/n/tip-agjev8eu666tvknpb3iaj0fg@git.kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
2011-06-27 14:41:57 +02:00
perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, xcp, 0);
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(emulated);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
switch (MIPSInst_OPCODE(ir)) {
case ldc1_op:{
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u64 __user *va = (u64 __user *) (xcp->regs[MIPSInst_RS(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_SIMM(ir));
u64 val;
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(loads);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_READ, va, sizeof(u64))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__get_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DITOREG(val, MIPSInst_RT(ir));
break;
}
case sdc1_op:{
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u64 __user *va = (u64 __user *) (xcp->regs[MIPSInst_RS(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_SIMM(ir));
u64 val;
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(stores);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DIFROMREG(val, MIPSInst_RT(ir));
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_WRITE, va, sizeof(u64))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__put_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
}
case lwc1_op:{
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u32 __user *va = (u32 __user *) (xcp->regs[MIPSInst_RS(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_SIMM(ir));
u32 val;
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(loads);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_READ, va, sizeof(u32))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__get_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
SITOREG(val, MIPSInst_RT(ir));
break;
}
case swc1_op:{
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u32 __user *va = (u32 __user *) (xcp->regs[MIPSInst_RS(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_SIMM(ir));
u32 val;
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(stores);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
SIFROMREG(val, MIPSInst_RT(ir));
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_WRITE, va, sizeof(u32))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__put_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
}
case cop1_op:
switch (MIPSInst_RS(ir)) {
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case dmfc_op:
/* copregister fs -> gpr[rt] */
if (MIPSInst_RT(ir) != 0) {
DIFROMREG(xcp->regs[MIPSInst_RT(ir)],
MIPSInst_RD(ir));
}
break;
case dmtc_op:
/* copregister fs <- rt */
DITOREG(xcp->regs[MIPSInst_RT(ir)], MIPSInst_RD(ir));
break;
#endif
MIPS: math-emu: Add mfhc1 & mthc1 support. This patch adds support for the mfhc1 & mthc1 instructions to the FPU emulator. These instructions were introduced in release 2 of the MIPS32 & MIPS64 architectures and allow access to the most significant 32 bits of a 64-bit FP register. [ralf@linux-mips.org: Fix ifdef hell added by original patch.] Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6112/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:28 +01:00
case mfhc_op:
if (!cpu_has_mips_r2)
goto sigill;
/* copregister rd -> gpr[rt] */
if (MIPSInst_RT(ir) != 0) {
SIFROMHREG(xcp->regs[MIPSInst_RT(ir)],
MIPSInst_RD(ir));
}
break;
case mthc_op:
if (!cpu_has_mips_r2)
goto sigill;
/* copregister rd <- gpr[rt] */
SITOHREG(xcp->regs[MIPSInst_RT(ir)], MIPSInst_RD(ir));
break;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case mfc_op:
/* copregister rd -> gpr[rt] */
if (MIPSInst_RT(ir) != 0) {
SIFROMREG(xcp->regs[MIPSInst_RT(ir)],
MIPSInst_RD(ir));
}
break;
case mtc_op:
/* copregister rd <- rt */
SITOREG(xcp->regs[MIPSInst_RT(ir)], MIPSInst_RD(ir));
break;
case cfc_op:{
/* cop control register rd -> gpr[rt] */
u32 value;
if (MIPSInst_RD(ir) == FPCREG_CSR) {
value = ctx->fcr31;
MIPS: Coding style cleanups of access of FCSR rounding mode bits Replaces references to the magic number 0x3 with constants and macros indicating the real purpose of those bits. They are the rounding mode bits of the FCSR register. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: linux-mips@linux-mips.org To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1206/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-05-07 08:02:09 +02:00
value = (value & ~FPU_CSR_RM) |
mips_rm[modeindex(value)];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#ifdef CSRTRACE
printk("%p gpr[%d]<-csr=%08x\n",
Gross macro abuse. Get rid of gpreg_t, vaddr_t, REG_TO_VA and VA_TO_REG. Who ever wrote this apparently did enjoy the C Puzzle Book. ISBN 0201604612, a little old but still fun reading for the next blackout ;) Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-02-28 18:55:57 +01:00
(void *) (xcp->cp0_epc),
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_RT(ir), value);
#endif
}
else if (MIPSInst_RD(ir) == FPCREG_RID)
value = 0;
else
value = 0;
if (MIPSInst_RT(ir))
xcp->regs[MIPSInst_RT(ir)] = value;
break;
}
case ctc_op:{
/* copregister rd <- rt */
u32 value;
if (MIPSInst_RT(ir) == 0)
value = 0;
else
value = xcp->regs[MIPSInst_RT(ir)];
/* we only have one writable control reg
*/
if (MIPSInst_RD(ir) == FPCREG_CSR) {
#ifdef CSRTRACE
printk("%p gpr[%d]->csr=%08x\n",
Gross macro abuse. Get rid of gpreg_t, vaddr_t, REG_TO_VA and VA_TO_REG. Who ever wrote this apparently did enjoy the C Puzzle Book. ISBN 0201604612, a little old but still fun reading for the next blackout ;) Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-02-28 18:55:57 +01:00
(void *) (xcp->cp0_epc),
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
MIPSInst_RT(ir), value);
#endif
MIPS FPU emulator: allow Cause bits of FCSR to be writeable by ctc1 In the FPU emulator code of the MIPS, the Cause bits of the FCSR register are not currently writeable by the ctc1 instruction. In odd corner cases, this can cause problems. For example, a case existed where a divide-by-zero exception was generated by the FPU, and the signal handler attempted to restore the FPU registers to their state before the exception occurred. In this particular setup, writing the old value to the FCSR register would cause another divide-by-zero exception to occur immediately. The solution is to change the ctc1 instruction emulator code to allow the Cause bits of the FCSR register to be writeable. This is the behaviour of the hardware that the code is emulating. This problem was found by Shane McDonald, but the credit for the fix goes to Kevin Kissell. In Kevin's words: I submit that the bug is indeed in that ctc_op: case of the emulator. The Cause bits (17:12) are supposed to be writable by that instruction, but the CTC1 emulation won't let them be updated by the instruction. I think that actually if you just completely removed lines 387-388 [...] things would work a good deal better. At least, it would be a more accurate emulation of the architecturally defined FPU. If I wanted to be really, really pedantic (which I sometimes do), I'd also protect the reserved bits that aren't necessarily writable. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1205/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org> ---
2010-05-07 07:26:57 +02:00
/*
* Don't write reserved bits,
* and convert to ieee library modes
*/
ctx->fcr31 = (value &
~(FPU_CSR_RSVD | FPU_CSR_RM)) |
ieee_rm[modeindex(value)];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
}
if ((ctx->fcr31 >> 5) & ctx->fcr31 & FPU_CSR_ALL_E) {
return SIGFPE;
}
break;
}
case bc_op:{
int likely = 0;
if (xcp->cp0_cause & CAUSEF_BD)
return SIGILL;
#if __mips >= 4
cond = ctx->fcr31 & fpucondbit[MIPSInst_RT(ir) >> 2];
#else
cond = ctx->fcr31 & FPU_CSR_COND;
#endif
switch (MIPSInst_RT(ir) & 3) {
case bcfl_op:
likely = 1;
case bcf_op:
cond = !cond;
break;
case bctl_op:
likely = 1;
case bct_op:
break;
default:
/* thats an illegal instruction */
return SIGILL;
}
xcp->cp0_cause |= CAUSEF_BD;
if (cond) {
/* branch taken: emulate dslot
* instruction
*/
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
xcp->cp0_epc += dec_insn.pc_inc;
contpc = MIPSInst_SIMM(ir);
ir = dec_insn.next_insn;
if (dec_insn.micro_mips_mode) {
contpc = (xcp->cp0_epc + (contpc << 1));
/* If 16-bit instruction, not FPU. */
if ((dec_insn.next_pc_inc == 2) ||
(microMIPS32_to_MIPS32((union mips_instruction *)&ir) == SIGILL)) {
/*
* Since this instruction will
* be put on the stack with
* 32-bit words, get around
* this problem by putting a
* NOP16 as the second one.
*/
if (dec_insn.next_pc_inc == 2)
ir = (ir & (~0xffff)) | MM_NOP16;
/*
* Single step the non-CP1
* instruction in the dslot.
*/
return mips_dsemul(xcp, ir, contpc);
}
} else
contpc = (xcp->cp0_epc + (contpc << 2));
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
switch (MIPSInst_OPCODE(ir)) {
case lwc1_op:
case swc1_op:
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if (__mips >= 2 || defined(__mips64))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case ldc1_op:
case sdc1_op:
#endif
case cop1_op:
#if __mips >= 4 && __mips != 32
case cop1x_op:
#endif
/* its one of ours */
goto emul;
#if __mips >= 4
case spec_op:
if (MIPSInst_FUNC(ir) == movc_op)
goto emul;
break;
#endif
}
/*
* Single step the non-cp1
* instruction in the dslot
*/
[MIPS] math-emu minor cleanup Declaring emulpc and contpc as "unsigned long" can get rid of some casts. This also get rid of some sparse warnings. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-13 16:02:29 +02:00
return mips_dsemul(xcp, ir, contpc);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
}
else {
/* branch not taken */
if (likely) {
/*
* branch likely nullifies
* dslot if not taken
*/
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
xcp->cp0_epc += dec_insn.pc_inc;
contpc += dec_insn.pc_inc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
* else continue & execute
* dslot as normal insn
*/
}
}
break;
}
default:
if (!(MIPSInst_RS(ir) & 0x10))
return SIGILL;
{
int sig;
/* a real fpu computation instruction */
if ((sig = fpu_emu(xcp, ctx, ir)))
return sig;
}
}
break;
#if __mips >= 4 && __mips != 32
case cop1x_op:{
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
int sig = fpux_emu(xcp, ctx, ir, fault_addr);
if (sig)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return sig;
break;
}
#endif
#if __mips >= 4
case spec_op:
if (MIPSInst_FUNC(ir) != movc_op)
return SIGILL;
cond = fpucondbit[MIPSInst_RT(ir) >> 2];
if (((ctx->fcr31 & cond) != 0) == ((MIPSInst_RT(ir) & 1) != 0))
xcp->regs[MIPSInst_RD(ir)] =
xcp->regs[MIPSInst_RS(ir)];
break;
#endif
default:
MIPS: math-emu: Add mfhc1 & mthc1 support. This patch adds support for the mfhc1 & mthc1 instructions to the FPU emulator. These instructions were introduced in release 2 of the MIPS32 & MIPS64 architectures and allow access to the most significant 32 bits of a 64-bit FP register. [ralf@linux-mips.org: Fix ifdef hell added by original patch.] Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com> Signed-off-by: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/6112/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-11-07 13:48:28 +01:00
sigill:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGILL;
}
/* we did it !! */
[MIPS] math-emu minor cleanup Declaring emulpc and contpc as "unsigned long" can get rid of some casts. This also get rid of some sparse warnings. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-13 16:02:29 +02:00
xcp->cp0_epc = contpc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
xcp->cp0_cause &= ~CAUSEF_BD;
Gross macro abuse. Get rid of gpreg_t, vaddr_t, REG_TO_VA and VA_TO_REG. Who ever wrote this apparently did enjoy the C Puzzle Book. ISBN 0201604612, a little old but still fun reading for the next blackout ;) Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-02-28 18:55:57 +01:00
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return 0;
}
/*
* Conversion table from MIPS compare ops 48-63
* cond = ieee754dp_cmp(x,y,IEEE754_UN,sig);
*/
static const unsigned char cmptab[8] = {
0, /* cmp_0 (sig) cmp_sf */
IEEE754_CUN, /* cmp_un (sig) cmp_ngle */
IEEE754_CEQ, /* cmp_eq (sig) cmp_seq */
IEEE754_CEQ | IEEE754_CUN, /* cmp_ueq (sig) cmp_ngl */
IEEE754_CLT, /* cmp_olt (sig) cmp_lt */
IEEE754_CLT | IEEE754_CUN, /* cmp_ult (sig) cmp_nge */
IEEE754_CLT | IEEE754_CEQ, /* cmp_ole (sig) cmp_le */
IEEE754_CLT | IEEE754_CEQ | IEEE754_CUN, /* cmp_ule (sig) cmp_ngt */
};
#if __mips >= 4 && __mips != 32
/*
* Additional MIPS4 instructions
*/
#define DEF3OP(name, p, f1, f2, f3) \
[MIPS] Fix "no space between function name and open parenthesis" warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
static ieee754##p fpemu_##p##_##name(ieee754##p r, ieee754##p s, \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ieee754##p t) \
{ \
Fix preemption and SMP problems in the FP emulator code. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-04-28 15:39:10 +02:00
struct _ieee754_csr ieee754_csr_save; \
[MIPS] Fix "no space between function name and open parenthesis" warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
s = f1(s, t); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ieee754_csr_save = ieee754_csr; \
[MIPS] Fix "no space between function name and open parenthesis" warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
s = f2(s, r); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ieee754_csr_save.cx |= ieee754_csr.cx; \
ieee754_csr_save.sx |= ieee754_csr.sx; \
[MIPS] Fix "no space between function name and open parenthesis" warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
s = f3(s); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ieee754_csr.cx |= ieee754_csr_save.cx; \
ieee754_csr.sx |= ieee754_csr_save.sx; \
return s; \
}
static ieee754dp fpemu_dp_recip(ieee754dp d)
{
return ieee754dp_div(ieee754dp_one(0), d);
}
static ieee754dp fpemu_dp_rsqrt(ieee754dp d)
{
return ieee754dp_div(ieee754dp_one(0), ieee754dp_sqrt(d));
}
static ieee754sp fpemu_sp_recip(ieee754sp s)
{
return ieee754sp_div(ieee754sp_one(0), s);
}
static ieee754sp fpemu_sp_rsqrt(ieee754sp s)
{
return ieee754sp_div(ieee754sp_one(0), ieee754sp_sqrt(s));
}
[MIPS] checkfiles: Fix "need space after that ','" errors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
DEF3OP(madd, sp, ieee754sp_mul, ieee754sp_add, );
DEF3OP(msub, sp, ieee754sp_mul, ieee754sp_sub, );
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DEF3OP(nmadd, sp, ieee754sp_mul, ieee754sp_add, ieee754sp_neg);
DEF3OP(nmsub, sp, ieee754sp_mul, ieee754sp_sub, ieee754sp_neg);
[MIPS] checkfiles: Fix "need space after that ','" errors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-10-12 00:46:15 +02:00
DEF3OP(madd, dp, ieee754dp_mul, ieee754dp_add, );
DEF3OP(msub, dp, ieee754dp_mul, ieee754dp_sub, );
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DEF3OP(nmadd, dp, ieee754dp_mul, ieee754dp_add, ieee754dp_neg);
DEF3OP(nmsub, dp, ieee754dp_mul, ieee754dp_sub, ieee754dp_neg);
[MIPS] Unify mips_fpu_soft_struct and mips_fpu_hard_structs. The struct mips_fpu_soft_struct and mips_fpu_hard_struct are completely same now and the kernel fpu emulator assumes that. This patch unifies them to mips_fpu_struct and get rid of mips_fpu_union. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-05-15 18:26:03 +02:00
static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
mips_instruction ir, void *__user *fault_addr)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
{
unsigned rcsr = 0; /* resulting csr */
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(cp1xops);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
switch (MIPSInst_FMA_FFMT(ir)) {
case s_fmt:{ /* 0 */
ieee754sp(*handler) (ieee754sp, ieee754sp, ieee754sp);
ieee754sp fd, fr, fs, ft;
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u32 __user *va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
u32 val;
switch (MIPSInst_FUNC(ir)) {
case lwxc1_op:
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
va = (void __user *) (xcp->regs[MIPSInst_FR(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
xcp->regs[MIPSInst_FT(ir)]);
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(loads);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_READ, va, sizeof(u32))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__get_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
SITOREG(val, MIPSInst_FD(ir));
break;
case swxc1_op:
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
va = (void __user *) (xcp->regs[MIPSInst_FR(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
xcp->regs[MIPSInst_FT(ir)]);
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(stores);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
SIFROMREG(val, MIPSInst_FS(ir));
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_WRITE, va, sizeof(u32))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (put_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
case madd_s_op:
handler = fpemu_sp_madd;
goto scoptop;
case msub_s_op:
handler = fpemu_sp_msub;
goto scoptop;
case nmadd_s_op:
handler = fpemu_sp_nmadd;
goto scoptop;
case nmsub_s_op:
handler = fpemu_sp_nmsub;
goto scoptop;
scoptop:
SPFROMREG(fr, MIPSInst_FR(ir));
SPFROMREG(fs, MIPSInst_FS(ir));
SPFROMREG(ft, MIPSInst_FT(ir));
fd = (*handler) (fr, fs, ft);
SPTOREG(fd, MIPSInst_FD(ir));
copcsr:
if (ieee754_cxtest(IEEE754_INEXACT))
rcsr |= FPU_CSR_INE_X | FPU_CSR_INE_S;
if (ieee754_cxtest(IEEE754_UNDERFLOW))
rcsr |= FPU_CSR_UDF_X | FPU_CSR_UDF_S;
if (ieee754_cxtest(IEEE754_OVERFLOW))
rcsr |= FPU_CSR_OVF_X | FPU_CSR_OVF_S;
if (ieee754_cxtest(IEEE754_INVALID_OPERATION))
rcsr |= FPU_CSR_INV_X | FPU_CSR_INV_S;
ctx->fcr31 = (ctx->fcr31 & ~FPU_CSR_ALL_X) | rcsr;
if ((ctx->fcr31 >> 5) & ctx->fcr31 & FPU_CSR_ALL_E) {
/*printk ("SIGFPE: fpu csr = %08x\n",
ctx->fcr31); */
return SIGFPE;
}
break;
default:
return SIGILL;
}
break;
}
case d_fmt:{ /* 1 */
ieee754dp(*handler) (ieee754dp, ieee754dp, ieee754dp);
ieee754dp fd, fr, fs, ft;
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
u64 __user *va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
u64 val;
switch (MIPSInst_FUNC(ir)) {
case ldxc1_op:
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
va = (void __user *) (xcp->regs[MIPSInst_FR(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
xcp->regs[MIPSInst_FT(ir)]);
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(loads);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_READ, va, sizeof(u64))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__get_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DITOREG(val, MIPSInst_FD(ir));
break;
case sdxc1_op:
Fix all the get_user / put_user related sparse warnings. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 14:58:21 +02:00
va = (void __user *) (xcp->regs[MIPSInst_FR(ir)] +
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
xcp->regs[MIPSInst_FT(ir)]);
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(stores);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
DIFROMREG(val, MIPSInst_FS(ir));
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (!access_ok(VERIFY_WRITE, va, sizeof(u64))) {
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(errors);
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
*fault_addr = va;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
return SIGBUS;
}
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
if (__put_user(val, va)) {
MIPS_FPU_EMU_INC_STATS(errors);
*fault_addr = va;
return SIGSEGV;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
case madd_d_op:
handler = fpemu_dp_madd;
goto dcoptop;
case msub_d_op:
handler = fpemu_dp_msub;
goto dcoptop;
case nmadd_d_op:
handler = fpemu_dp_nmadd;
goto dcoptop;
case nmsub_d_op:
handler = fpemu_dp_nmsub;
goto dcoptop;
dcoptop:
DPFROMREG(fr, MIPSInst_FR(ir));
DPFROMREG(fs, MIPSInst_FS(ir));
DPFROMREG(ft, MIPSInst_FT(ir));
fd = (*handler) (fr, fs, ft);
DPTOREG(fd, MIPSInst_FD(ir));
goto copcsr;
default:
return SIGILL;
}
break;
}
case 0x7: /* 7 */
if (MIPSInst_FUNC(ir) != pfetch_op) {
return SIGILL;
}
/* ignore prefx operation */
break;
default:
return SIGILL;
}
return 0;
}
#endif
/*
* Emulate a single COP1 arithmetic instruction.
*/
[MIPS] Unify mips_fpu_soft_struct and mips_fpu_hard_structs. The struct mips_fpu_soft_struct and mips_fpu_hard_struct are completely same now and the kernel fpu emulator assumes that. This patch unifies them to mips_fpu_struct and get rid of mips_fpu_union. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-05-15 18:26:03 +02:00
static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mips_instruction ir)
{
int rfmt; /* resulting format */
unsigned rcsr = 0; /* resulting csr */
unsigned cond;
union {
ieee754dp d;
ieee754sp s;
int w;
[PATCH] mips: fix build warnings This patch has fixed the following warnings. arch/mips/kernel/genex.S:250:5: warning: "CONFIG_64BIT" is not defined arch/mips/math-emu/cp1emu.c:1128:5: warning: "__mips64" is not defined arch/mips/math-emu/cp1emu.c:1206:5: warning: "__mips64" is not defined arch/mips/math-emu/cp1emu.c:1270:5: warning: "__mips64" is not defined arch/mips/math-emu/cp1emu.c:323:5: warning: "__mips64" is not defined arch/mips/math-emu/cp1emu.c:808:5: warning: "__mips64" is not defined arch/mips/math-emu/cp1emu.c:953:5: warning: "__mips64" is not defined arch/mips/mm/tlbex.c:519:5: warning: "CONFIG_64BIT" is not defined include/asm/reg.h:73:5: warning: "CONFIG_64BIT" is not defined Signed-off-by: Yoichi Yuasa <yuasa@hh.iij4u.or.jp> Cc: Ralf Baechle <ralf@linux-mips.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-04 00:56:22 +02:00
#ifdef __mips64
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
s64 l;
#endif
} rv; /* resulting value */
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
MIPS_FPU_EMU_INC_STATS(cp1ops);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
switch (rfmt = (MIPSInst_FFMT(ir) & 0xf)) {
case s_fmt:{ /* 0 */
union {
ieee754sp(*b) (ieee754sp, ieee754sp);
ieee754sp(*u) (ieee754sp);
} handler;
switch (MIPSInst_FUNC(ir)) {
/* binary ops */
case fadd_op:
handler.b = ieee754sp_add;
goto scopbop;
case fsub_op:
handler.b = ieee754sp_sub;
goto scopbop;
case fmul_op:
handler.b = ieee754sp_mul;
goto scopbop;
case fdiv_op:
handler.b = ieee754sp_div;
goto scopbop;
/* unary ops */
GCC bulletproofing: __mips64 is only defined when compiling for 64-bit processors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-09-15 10:52:34 +02:00
#if __mips >= 2 || defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fsqrt_op:
handler.u = ieee754sp_sqrt;
goto scopuop;
#endif
#if __mips >= 4 && __mips != 32
case frsqrt_op:
handler.u = fpemu_sp_rsqrt;
goto scopuop;
case frecip_op:
handler.u = fpemu_sp_recip;
goto scopuop;
#endif
#if __mips >= 4
case fmovc_op:
cond = fpucondbit[MIPSInst_FT(ir) >> 2];
if (((ctx->fcr31 & cond) != 0) !=
((MIPSInst_FT(ir) & 1) != 0))
return 0;
SPFROMREG(rv.s, MIPSInst_FS(ir));
break;
case fmovz_op:
if (xcp->regs[MIPSInst_FT(ir)] != 0)
return 0;
SPFROMREG(rv.s, MIPSInst_FS(ir));
break;
case fmovn_op:
if (xcp->regs[MIPSInst_FT(ir)] == 0)
return 0;
SPFROMREG(rv.s, MIPSInst_FS(ir));
break;
#endif
case fabs_op:
handler.u = ieee754sp_abs;
goto scopuop;
case fneg_op:
handler.u = ieee754sp_neg;
goto scopuop;
case fmov_op:
/* an easy one */
SPFROMREG(rv.s, MIPSInst_FS(ir));
goto copcsr;
/* binary op on handler */
scopbop:
{
ieee754sp fs, ft;
SPFROMREG(fs, MIPSInst_FS(ir));
SPFROMREG(ft, MIPSInst_FT(ir));
rv.s = (*handler.b) (fs, ft);
goto copcsr;
}
scopuop:
{
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
rv.s = (*handler.u) (fs);
goto copcsr;
}
copcsr:
if (ieee754_cxtest(IEEE754_INEXACT))
rcsr |= FPU_CSR_INE_X | FPU_CSR_INE_S;
if (ieee754_cxtest(IEEE754_UNDERFLOW))
rcsr |= FPU_CSR_UDF_X | FPU_CSR_UDF_S;
if (ieee754_cxtest(IEEE754_OVERFLOW))
rcsr |= FPU_CSR_OVF_X | FPU_CSR_OVF_S;
if (ieee754_cxtest(IEEE754_ZERO_DIVIDE))
rcsr |= FPU_CSR_DIV_X | FPU_CSR_DIV_S;
if (ieee754_cxtest(IEEE754_INVALID_OPERATION))
rcsr |= FPU_CSR_INV_X | FPU_CSR_INV_S;
break;
/* unary conv ops */
case fcvts_op:
return SIGILL; /* not defined */
case fcvtd_op:{
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
rv.d = ieee754dp_fsp(fs);
rfmt = d_fmt;
goto copcsr;
}
case fcvtw_op:{
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
rv.w = ieee754sp_tint(fs);
rfmt = w_fmt;
goto copcsr;
}
GCC bulletproofing: __mips64 is only defined when compiling for 64-bit processors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-09-15 10:52:34 +02:00
#if __mips >= 2 || defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fround_op:
case ftrunc_op:
case fceil_op:
case ffloor_op:{
unsigned int oldrm = ieee754_csr.rm;
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
MIPS: Coding style cleanups of access of FCSR rounding mode bits Replaces references to the magic number 0x3 with constants and macros indicating the real purpose of those bits. They are the rounding mode bits of the FCSR register. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: linux-mips@linux-mips.org To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1206/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-05-07 08:02:09 +02:00
ieee754_csr.rm = ieee_rm[modeindex(MIPSInst_FUNC(ir))];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rv.w = ieee754sp_tint(fs);
ieee754_csr.rm = oldrm;
rfmt = w_fmt;
goto copcsr;
}
#endif /* __mips >= 2 */
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fcvtl_op:{
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
rv.l = ieee754sp_tlong(fs);
rfmt = l_fmt;
goto copcsr;
}
case froundl_op:
case ftruncl_op:
case fceill_op:
case ffloorl_op:{
unsigned int oldrm = ieee754_csr.rm;
ieee754sp fs;
SPFROMREG(fs, MIPSInst_FS(ir));
MIPS: Coding style cleanups of access of FCSR rounding mode bits Replaces references to the magic number 0x3 with constants and macros indicating the real purpose of those bits. They are the rounding mode bits of the FCSR register. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: linux-mips@linux-mips.org To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1206/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-05-07 08:02:09 +02:00
ieee754_csr.rm = ieee_rm[modeindex(MIPSInst_FUNC(ir))];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rv.l = ieee754sp_tlong(fs);
ieee754_csr.rm = oldrm;
rfmt = l_fmt;
goto copcsr;
}
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#endif /* defined(__mips64) */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
default:
if (MIPSInst_FUNC(ir) >= fcmp_op) {
unsigned cmpop = MIPSInst_FUNC(ir) - fcmp_op;
ieee754sp fs, ft;
SPFROMREG(fs, MIPSInst_FS(ir));
SPFROMREG(ft, MIPSInst_FT(ir));
rv.w = ieee754sp_cmp(fs, ft,
cmptab[cmpop & 0x7], cmpop & 0x8);
rfmt = -1;
if ((cmpop & 0x8) && ieee754_cxtest
(IEEE754_INVALID_OPERATION))
rcsr = FPU_CSR_INV_X | FPU_CSR_INV_S;
else
goto copcsr;
}
else {
return SIGILL;
}
break;
}
break;
}
case d_fmt:{
union {
ieee754dp(*b) (ieee754dp, ieee754dp);
ieee754dp(*u) (ieee754dp);
} handler;
switch (MIPSInst_FUNC(ir)) {
/* binary ops */
case fadd_op:
handler.b = ieee754dp_add;
goto dcopbop;
case fsub_op:
handler.b = ieee754dp_sub;
goto dcopbop;
case fmul_op:
handler.b = ieee754dp_mul;
goto dcopbop;
case fdiv_op:
handler.b = ieee754dp_div;
goto dcopbop;
/* unary ops */
GCC bulletproofing: __mips64 is only defined when compiling for 64-bit processors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-09-15 10:52:34 +02:00
#if __mips >= 2 || defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fsqrt_op:
handler.u = ieee754dp_sqrt;
goto dcopuop;
#endif
#if __mips >= 4 && __mips != 32
case frsqrt_op:
handler.u = fpemu_dp_rsqrt;
goto dcopuop;
case frecip_op:
handler.u = fpemu_dp_recip;
goto dcopuop;
#endif
#if __mips >= 4
case fmovc_op:
cond = fpucondbit[MIPSInst_FT(ir) >> 2];
if (((ctx->fcr31 & cond) != 0) !=
((MIPSInst_FT(ir) & 1) != 0))
return 0;
DPFROMREG(rv.d, MIPSInst_FS(ir));
break;
case fmovz_op:
if (xcp->regs[MIPSInst_FT(ir)] != 0)
return 0;
DPFROMREG(rv.d, MIPSInst_FS(ir));
break;
case fmovn_op:
if (xcp->regs[MIPSInst_FT(ir)] == 0)
return 0;
DPFROMREG(rv.d, MIPSInst_FS(ir));
break;
#endif
case fabs_op:
handler.u = ieee754dp_abs;
goto dcopuop;
case fneg_op:
handler.u = ieee754dp_neg;
goto dcopuop;
case fmov_op:
/* an easy one */
DPFROMREG(rv.d, MIPSInst_FS(ir));
goto copcsr;
/* binary op on handler */
dcopbop:{
ieee754dp fs, ft;
DPFROMREG(fs, MIPSInst_FS(ir));
DPFROMREG(ft, MIPSInst_FT(ir));
rv.d = (*handler.b) (fs, ft);
goto copcsr;
}
dcopuop:{
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
rv.d = (*handler.u) (fs);
goto copcsr;
}
/* unary conv ops */
case fcvts_op:{
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
rv.s = ieee754sp_fdp(fs);
rfmt = s_fmt;
goto copcsr;
}
case fcvtd_op:
return SIGILL; /* not defined */
case fcvtw_op:{
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
rv.w = ieee754dp_tint(fs); /* wrong */
rfmt = w_fmt;
goto copcsr;
}
GCC bulletproofing: __mips64 is only defined when compiling for 64-bit processors. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-09-15 10:52:34 +02:00
#if __mips >= 2 || defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fround_op:
case ftrunc_op:
case fceil_op:
case ffloor_op:{
unsigned int oldrm = ieee754_csr.rm;
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
MIPS: Coding style cleanups of access of FCSR rounding mode bits Replaces references to the magic number 0x3 with constants and macros indicating the real purpose of those bits. They are the rounding mode bits of the FCSR register. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: linux-mips@linux-mips.org To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1206/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-05-07 08:02:09 +02:00
ieee754_csr.rm = ieee_rm[modeindex(MIPSInst_FUNC(ir))];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rv.w = ieee754dp_tint(fs);
ieee754_csr.rm = oldrm;
rfmt = w_fmt;
goto copcsr;
}
#endif
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case fcvtl_op:{
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
rv.l = ieee754dp_tlong(fs);
rfmt = l_fmt;
goto copcsr;
}
case froundl_op:
case ftruncl_op:
case fceill_op:
case ffloorl_op:{
unsigned int oldrm = ieee754_csr.rm;
ieee754dp fs;
DPFROMREG(fs, MIPSInst_FS(ir));
MIPS: Coding style cleanups of access of FCSR rounding mode bits Replaces references to the magic number 0x3 with constants and macros indicating the real purpose of those bits. They are the rounding mode bits of the FCSR register. Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com> To: anemo@mba.ocn.ne.jp To: kevink@paralogos.com To: linux-mips@linux-mips.org To: sshtylyov@mvista.com Patchwork: http://patchwork.linux-mips.org/patch/1206/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-05-07 08:02:09 +02:00
ieee754_csr.rm = ieee_rm[modeindex(MIPSInst_FUNC(ir))];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rv.l = ieee754dp_tlong(fs);
ieee754_csr.rm = oldrm;
rfmt = l_fmt;
goto copcsr;
}
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#endif /* __mips >= 3 */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
default:
if (MIPSInst_FUNC(ir) >= fcmp_op) {
unsigned cmpop = MIPSInst_FUNC(ir) - fcmp_op;
ieee754dp fs, ft;
DPFROMREG(fs, MIPSInst_FS(ir));
DPFROMREG(ft, MIPSInst_FT(ir));
rv.w = ieee754dp_cmp(fs, ft,
cmptab[cmpop & 0x7], cmpop & 0x8);
rfmt = -1;
if ((cmpop & 0x8)
&&
ieee754_cxtest
(IEEE754_INVALID_OPERATION))
rcsr = FPU_CSR_INV_X | FPU_CSR_INV_S;
else
goto copcsr;
}
else {
return SIGILL;
}
break;
}
break;
}
case w_fmt:{
ieee754sp fs;
switch (MIPSInst_FUNC(ir)) {
case fcvts_op:
/* convert word to single precision real */
SPFROMREG(fs, MIPSInst_FS(ir));
rv.s = ieee754sp_fint(fs.bits);
rfmt = s_fmt;
goto copcsr;
case fcvtd_op:
/* convert word to double precision real */
SPFROMREG(fs, MIPSInst_FS(ir));
rv.d = ieee754dp_fint(fs.bits);
rfmt = d_fmt;
goto copcsr;
default:
return SIGILL;
}
break;
}
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case l_fmt:{
switch (MIPSInst_FUNC(ir)) {
case fcvts_op:
/* convert long to single precision real */
rv.s = ieee754sp_flong(ctx->fpr[MIPSInst_FS(ir)]);
rfmt = s_fmt;
goto copcsr;
case fcvtd_op:
/* convert long to double precision real */
rv.d = ieee754dp_flong(ctx->fpr[MIPSInst_FS(ir)]);
rfmt = d_fmt;
goto copcsr;
default:
return SIGILL;
}
break;
}
#endif
default:
return SIGILL;
}
/*
* Update the fpu CSR register for this operation.
* If an exception is required, generate a tidy SIGFPE exception,
* without updating the result register.
* Note: cause exception bits do not accumulate, they are rewritten
* for each op; only the flag/sticky bits accumulate.
*/
ctx->fcr31 = (ctx->fcr31 & ~FPU_CSR_ALL_X) | rcsr;
if ((ctx->fcr31 >> 5) & ctx->fcr31 & FPU_CSR_ALL_E) {
/*printk ("SIGFPE: fpu csr = %08x\n",ctx->fcr31); */
return SIGFPE;
}
/*
* Now we can safely write the result back to the register file.
*/
switch (rfmt) {
case -1:{
#if __mips >= 4
cond = fpucondbit[MIPSInst_FD(ir) >> 2];
#else
cond = FPU_CSR_COND;
#endif
if (rv.w)
ctx->fcr31 |= cond;
else
ctx->fcr31 &= ~cond;
break;
}
case d_fmt:
DPTOREG(rv.d, MIPSInst_FD(ir));
break;
case s_fmt:
SPTOREG(rv.s, MIPSInst_FD(ir));
break;
case w_fmt:
SITOREG(rv.w, MIPSInst_FD(ir));
break;
Get rid of SINGLE_ONLY_FPU. Linux does not support half FPU other than by emulation of a full FPU. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-10-23 16:05:47 +02:00
#if defined(__mips64)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
case l_fmt:
DITOREG(rv.l, MIPSInst_FD(ir));
break;
#endif
default:
return SIGILL;
}
return 0;
}
[MIPS] Make sure cpu_has_fpu is used only in atomic context Make sure cpu_has_fpu (which uses smp_processor_id()) is used only in atomic context. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-10-08 17:10:01 +02:00
int fpu_emulator_cop1Handler(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
int has_fpu, void *__user *fault_addr)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
{
Gross macro abuse. Get rid of gpreg_t, vaddr_t, REG_TO_VA and VA_TO_REG. Who ever wrote this apparently did enjoy the C Puzzle Book. ISBN 0201604612, a little old but still fun reading for the next blackout ;) Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-02-28 18:55:57 +01:00
unsigned long oldepc, prevepc;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
struct mm_decoded_insn dec_insn;
u16 instr[4];
u16 *instr_ptr;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int sig = 0;
oldepc = xcp->cp0_epc;
do {
prevepc = xcp->cp0_epc;
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if (get_isa16_mode(prevepc) && cpu_has_mmips) {
/*
* Get next 2 microMIPS instructions and convert them
* into 32-bit instructions.
*/
if ((get_user(instr[0], (u16 __user *)msk_isa16_mode(xcp->cp0_epc))) ||
(get_user(instr[1], (u16 __user *)msk_isa16_mode(xcp->cp0_epc + 2))) ||
(get_user(instr[2], (u16 __user *)msk_isa16_mode(xcp->cp0_epc + 4))) ||
(get_user(instr[3], (u16 __user *)msk_isa16_mode(xcp->cp0_epc + 6)))) {
MIPS_FPU_EMU_INC_STATS(errors);
return SIGBUS;
}
instr_ptr = instr;
/* Get first instruction. */
if (mm_insn_16bit(*instr_ptr)) {
/* Duplicate the half-word. */
dec_insn.insn = (*instr_ptr << 16) |
(*instr_ptr);
/* 16-bit instruction. */
dec_insn.pc_inc = 2;
instr_ptr += 1;
} else {
dec_insn.insn = (*instr_ptr << 16) |
*(instr_ptr+1);
/* 32-bit instruction. */
dec_insn.pc_inc = 4;
instr_ptr += 2;
}
/* Get second instruction. */
if (mm_insn_16bit(*instr_ptr)) {
/* Duplicate the half-word. */
dec_insn.next_insn = (*instr_ptr << 16) |
(*instr_ptr);
/* 16-bit instruction. */
dec_insn.next_pc_inc = 2;
} else {
dec_insn.next_insn = (*instr_ptr << 16) |
*(instr_ptr+1);
/* 32-bit instruction. */
dec_insn.next_pc_inc = 4;
}
dec_insn.micro_mips_mode = 1;
} else {
if ((get_user(dec_insn.insn,
(mips_instruction __user *) xcp->cp0_epc)) ||
(get_user(dec_insn.next_insn,
(mips_instruction __user *)(xcp->cp0_epc+4)))) {
MIPS_FPU_EMU_INC_STATS(errors);
return SIGBUS;
}
dec_insn.pc_inc = 4;
dec_insn.next_pc_inc = 4;
dec_insn.micro_mips_mode = 0;
MIPS: Send proper signal and siginfo on FP emulator faults. We were unconditionally sending SIGBUS with an empty siginfo on FP emulator faults. This differs from what happens when real floating point hardware would get a fault. For most faults we need to send SIGSEGV with the faulting address filled in in the struct siginfo. Reported-by: Camm Maguire <camm@maguirefamily.org> Signed-off-by: David Daney <ddaney@caviumnetworks.com> To: linux-mips@linux-mips.org Cc: Camm Maguire <camm@maguirefamily.org> Patchwork: https://patchwork.linux-mips.org/patch/1727/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2010-10-22 01:32:26 +02:00
}
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
if ((dec_insn.insn == 0) ||
((dec_insn.pc_inc == 2) &&
((dec_insn.insn & 0xffff) == MM_NOP16)))
xcp->cp0_epc += dec_insn.pc_inc; /* Skip NOPs */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
else {
Fix preemption and SMP problems in the FP emulator code. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-04-28 15:39:10 +02:00
/*
* The 'ieee754_csr' is an alias of
MIPS: Whitespace cleanup. Having received another series of whitespace patches I decided to do this once and for all rather than dealing with this kind of patches trickling in forever. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2013-01-22 12:59:30 +01:00
* ctx->fcr31. No need to copy ctx->fcr31 to
* ieee754_csr. But ieee754_csr.rm is ieee
Fix preemption and SMP problems in the FP emulator code. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-04-28 15:39:10 +02:00
* library modes. (not mips rounding mode)
*/
/* convert to ieee library modes */
ieee754_csr.rm = ieee_rm[ieee754_csr.rm];
MIPS: microMIPS: Floating point support. Add logic needed to do floating point emulation in microMIPS mode. Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Signed-off-by: Steven J. Hill <Steven. Hill@imgtec.com>
2013-03-25 18:09:02 +01:00
sig = cop1Emulate(xcp, ctx, dec_insn, fault_addr);
Fix preemption and SMP problems in the FP emulator code. Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2005-04-28 15:39:10 +02:00
/* revert to mips rounding mode */
ieee754_csr.rm = mips_rm[ieee754_csr.rm];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
}
[MIPS] Make sure cpu_has_fpu is used only in atomic context Make sure cpu_has_fpu (which uses smp_processor_id()) is used only in atomic context. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-10-08 17:10:01 +02:00
if (has_fpu)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
break;
if (sig)
break;
cond_resched();
} while (xcp->cp0_epc > prevepc);
/* SIGILL indicates a non-fpu instruction */
if (sig == SIGILL && xcp->cp0_epc != oldepc)
/* but if epc has advanced, then ignore it */
sig = 0;
return sig;
}
[MIPS] Add debugfs files to show fpuemu statistics Export contents of struct mips_fpu_emulator_stats via debugfs. There is no way to read these statistics for now but they (at least the "emulated" count) might be sometimes useful for performance tuning on FPU-less CPUs. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-07 16:21:49 +02:00
#ifdef CONFIG_DEBUG_FS
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
static int fpuemu_stat_get(void *data, u64 *val)
{
int cpu;
unsigned long sum = 0;
for_each_online_cpu(cpu) {
struct mips_fpu_emulator_stats *ps;
local_t *pv;
ps = &per_cpu(fpuemustats, cpu);
pv = (void *)ps + (unsigned long)data;
sum += local_read(pv);
}
*val = sum;
return 0;
}
DEFINE_SIMPLE_ATTRIBUTE(fops_fpuemu_stat, fpuemu_stat_get, NULL, "%llu\n");
[MIPS] Add debugfs files to show fpuemu statistics Export contents of struct mips_fpu_emulator_stats via debugfs. There is no way to read these statistics for now but they (at least the "emulated" count) might be sometimes useful for performance tuning on FPU-less CPUs. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-07 16:21:49 +02:00
extern struct dentry *mips_debugfs_dir;
static int __init debugfs_fpuemu(void)
{
struct dentry *d, *dir;
if (!mips_debugfs_dir)
return -ENODEV;
dir = debugfs_create_dir("fpuemustats", mips_debugfs_dir);
MIPS: Fix debugfs_create_*'s error checking method for arch/mips/math-emu/ debugfs_create_*() returns NULL on error. Make its caller debugfs_fpuemu return -ENODEV on error. Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2008-10-17 13:12:30 +02:00
if (!dir)
return -ENOMEM;
MIPS: Collect FPU emulator statistics per-CPU. On SMP systems, the collection of statistics can cause cache line bouncing in the lines associated with the counters. Also there are races incrementing the counters on multiple CPUs. To fix both problems, we collect the statistics in per-CPU variables, and add them up in the debugfs read operation. As a test I ran the LTP float_bessel test on a 12 CPU Octeon system. Without CONFIG_DEBUG_FS : 2602 seconds. With CONFIG_DEBUG_FS: 2640 seconds. With non-cpu-local atomic statistics: 14569 seconds. Signed-off-by: David Daney <ddaney@caviumnetworks.com> Cc: linux-mips@linux-mips.org Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2009-11-05 20:34:26 +01:00
#define FPU_STAT_CREATE(M) \
do { \
d = debugfs_create_file(#M , S_IRUGO, dir, \
(void *)offsetof(struct mips_fpu_emulator_stats, M), \
&fops_fpuemu_stat); \
if (!d) \
return -ENOMEM; \
} while (0)
FPU_STAT_CREATE(emulated);
FPU_STAT_CREATE(loads);
FPU_STAT_CREATE(stores);
FPU_STAT_CREATE(cp1ops);
FPU_STAT_CREATE(cp1xops);
FPU_STAT_CREATE(errors);
[MIPS] Add debugfs files to show fpuemu statistics Export contents of struct mips_fpu_emulator_stats via debugfs. There is no way to read these statistics for now but they (at least the "emulated" count) might be sometimes useful for performance tuning on FPU-less CPUs. Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2007-07-07 16:21:49 +02:00
return 0;
}
__initcall(debugfs_fpuemu);
#endif
Reference in a new issue Copy permalink