linux-hardened/include/asm-sparc/checksum.h
Bob Breuer 51bcf09291 [SPARC32]: Fix over-optimization by GCC near ip_fast_csum.
In some cases such as:
	iph->check = 0;
	iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
GCC may optimize out the previous store.

Observed as a failure of NFS over udp (bad checksums on ip fragments)
when compiled with GCC 3.4.2.

Signed-off-by: Bob Breuer <breuerr@mc.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2007-02-01 20:24:35 -08:00

242 lines
6.7 KiB
C

/* $Id: checksum.h,v 1.33 2002/02/01 22:01:05 davem Exp $ */
#ifndef __SPARC_CHECKSUM_H
#define __SPARC_CHECKSUM_H
/* checksum.h: IP/UDP/TCP checksum routines on the Sparc.
*
* Copyright(C) 1995 Linus Torvalds
* Copyright(C) 1995 Miguel de Icaza
* Copyright(C) 1996 David S. Miller
* Copyright(C) 1996 Eddie C. Dost
* Copyright(C) 1997 Jakub Jelinek
*
* derived from:
* Alpha checksum c-code
* ix86 inline assembly
* RFC1071 Computing the Internet Checksum
*/
#include <linux/in6.h>
#include <asm/uaccess.h>
/* computes the checksum of a memory block at buff, length len,
* and adds in "sum" (32-bit)
*
* returns a 32-bit number suitable for feeding into itself
* or csum_tcpudp_magic
*
* this function must be called with even lengths, except
* for the last fragment, which may be odd
*
* it's best to have buff aligned on a 32-bit boundary
*/
extern __wsum csum_partial(const void *buff, int len, __wsum sum);
/* the same as csum_partial, but copies from fs:src while it
* checksums
*
* here even more important to align src and dst on a 32-bit (or even
* better 64-bit) boundary
*/
extern unsigned int __csum_partial_copy_sparc_generic (const unsigned char *, unsigned char *);
static inline __wsum
csum_partial_copy_nocheck(const void *src, void *dst, int len, __wsum sum)
{
register unsigned int ret asm("o0") = (unsigned int)src;
register char *d asm("o1") = dst;
register int l asm("g1") = len;
__asm__ __volatile__ (
"call __csum_partial_copy_sparc_generic\n\t"
" mov %6, %%g7\n"
: "=&r" (ret), "=&r" (d), "=&r" (l)
: "0" (ret), "1" (d), "2" (l), "r" (sum)
: "o2", "o3", "o4", "o5", "o7",
"g2", "g3", "g4", "g5", "g7",
"memory", "cc");
return (__force __wsum)ret;
}
static inline __wsum
csum_partial_copy_from_user(const void __user *src, void *dst, int len,
__wsum sum, int *err)
{
register unsigned long ret asm("o0") = (unsigned long)src;
register char *d asm("o1") = dst;
register int l asm("g1") = len;
register __wsum s asm("g7") = sum;
__asm__ __volatile__ (
".section __ex_table,#alloc\n\t"
".align 4\n\t"
".word 1f,2\n\t"
".previous\n"
"1:\n\t"
"call __csum_partial_copy_sparc_generic\n\t"
" st %8, [%%sp + 64]\n"
: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
: "o2", "o3", "o4", "o5", "o7", "g2", "g3", "g4", "g5",
"cc", "memory");
return (__force __wsum)ret;
}
static inline __wsum
csum_partial_copy_to_user(const void *src, void __user *dst, int len,
__wsum sum, int *err)
{
if (!access_ok (VERIFY_WRITE, dst, len)) {
*err = -EFAULT;
return sum;
} else {
register unsigned long ret asm("o0") = (unsigned long)src;
register char __user *d asm("o1") = dst;
register int l asm("g1") = len;
register __wsum s asm("g7") = sum;
__asm__ __volatile__ (
".section __ex_table,#alloc\n\t"
".align 4\n\t"
".word 1f,1\n\t"
".previous\n"
"1:\n\t"
"call __csum_partial_copy_sparc_generic\n\t"
" st %8, [%%sp + 64]\n"
: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
: "o2", "o3", "o4", "o5", "o7",
"g2", "g3", "g4", "g5",
"cc", "memory");
return (__force __wsum)ret;
}
}
#define HAVE_CSUM_COPY_USER
#define csum_and_copy_to_user csum_partial_copy_to_user
/* ihl is always 5 or greater, almost always is 5, and iph is word aligned
* the majority of the time.
*/
static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
{
__sum16 sum;
/* Note: We must read %2 before we touch %0 for the first time,
* because GCC can legitimately use the same register for
* both operands.
*/
__asm__ __volatile__("sub\t%2, 4, %%g4\n\t"
"ld\t[%1 + 0x00], %0\n\t"
"ld\t[%1 + 0x04], %%g2\n\t"
"ld\t[%1 + 0x08], %%g3\n\t"
"addcc\t%%g2, %0, %0\n\t"
"addxcc\t%%g3, %0, %0\n\t"
"ld\t[%1 + 0x0c], %%g2\n\t"
"ld\t[%1 + 0x10], %%g3\n\t"
"addxcc\t%%g2, %0, %0\n\t"
"addx\t%0, %%g0, %0\n"
"1:\taddcc\t%%g3, %0, %0\n\t"
"add\t%1, 4, %1\n\t"
"addxcc\t%0, %%g0, %0\n\t"
"subcc\t%%g4, 1, %%g4\n\t"
"be,a\t2f\n\t"
"sll\t%0, 16, %%g2\n\t"
"b\t1b\n\t"
"ld\t[%1 + 0x10], %%g3\n"
"2:\taddcc\t%0, %%g2, %%g2\n\t"
"srl\t%%g2, 16, %0\n\t"
"addx\t%0, %%g0, %0\n\t"
"xnor\t%%g0, %0, %0"
: "=r" (sum), "=&r" (iph)
: "r" (ihl), "1" (iph)
: "g2", "g3", "g4", "cc", "memory");
return sum;
}
/* Fold a partial checksum without adding pseudo headers. */
static inline __sum16 csum_fold(__wsum sum)
{
unsigned int tmp;
__asm__ __volatile__("addcc\t%0, %1, %1\n\t"
"srl\t%1, 16, %1\n\t"
"addx\t%1, %%g0, %1\n\t"
"xnor\t%%g0, %1, %0"
: "=&r" (sum), "=r" (tmp)
: "0" (sum), "1" ((__force u32)sum<<16)
: "cc");
return (__force __sum16)sum;
}
static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr,
unsigned short len,
unsigned short proto,
__wsum sum)
{
__asm__ __volatile__("addcc\t%1, %0, %0\n\t"
"addxcc\t%2, %0, %0\n\t"
"addxcc\t%3, %0, %0\n\t"
"addx\t%0, %%g0, %0\n\t"
: "=r" (sum), "=r" (saddr)
: "r" (daddr), "r" (proto + len), "0" (sum),
"1" (saddr)
: "cc");
return sum;
}
/*
* computes the checksum of the TCP/UDP pseudo-header
* returns a 16-bit checksum, already complemented
*/
static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
unsigned short len,
unsigned short proto,
__wsum sum)
{
return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum));
}
#define _HAVE_ARCH_IPV6_CSUM
static inline __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
const struct in6_addr *daddr,
__u32 len, unsigned short proto,
__wsum sum)
{
__asm__ __volatile__ (
"addcc %3, %4, %%g4\n\t"
"addxcc %5, %%g4, %%g4\n\t"
"ld [%2 + 0x0c], %%g2\n\t"
"ld [%2 + 0x08], %%g3\n\t"
"addxcc %%g2, %%g4, %%g4\n\t"
"ld [%2 + 0x04], %%g2\n\t"
"addxcc %%g3, %%g4, %%g4\n\t"
"ld [%2 + 0x00], %%g3\n\t"
"addxcc %%g2, %%g4, %%g4\n\t"
"ld [%1 + 0x0c], %%g2\n\t"
"addxcc %%g3, %%g4, %%g4\n\t"
"ld [%1 + 0x08], %%g3\n\t"
"addxcc %%g2, %%g4, %%g4\n\t"
"ld [%1 + 0x04], %%g2\n\t"
"addxcc %%g3, %%g4, %%g4\n\t"
"ld [%1 + 0x00], %%g3\n\t"
"addxcc %%g2, %%g4, %%g4\n\t"
"addxcc %%g3, %%g4, %0\n\t"
"addx 0, %0, %0\n"
: "=&r" (sum)
: "r" (saddr), "r" (daddr),
"r"(htonl(len)), "r"(htonl(proto)), "r"(sum)
: "g2", "g3", "g4", "cc");
return csum_fold(sum);
}
/* this routine is used for miscellaneous IP-like checksums, mainly in icmp.c */
static inline __sum16 ip_compute_csum(const void *buff, int len)
{
return csum_fold(csum_partial(buff, len, 0));
}
#endif /* !(__SPARC_CHECKSUM_H) */