Prevent segfaults in SSE optimized Fletcher-4

In some cases, the compiler was not respecting the GNU aligned
attribute for stack variables in 35a76a0. This was resulting in
a segfault on CentOS 6.7 hosts using gcc 4.4.7-17.  This issue
was fixed in gcc 4.6.

To prevent this from occurring, use unaligned loads and stores
for all stack and global memory references in the SSE optimized
Fletcher-4 code.

Disable zimport testing against master where this flaw exists:

TEST_ZIMPORT_VERSIONS="installed"

Signed-off-by: Tyler J. Stachecki <stachecki.tyler@gmail.com>
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #4862
This commit is contained in:
Tyler J. Stachecki 2016-07-18 17:38:55 -04:00 committed by Brian Behlendorf
parent 1b87e0f532
commit 3d11ecbddd
1 changed files with 5 additions and 5 deletions

View File

@ -69,12 +69,12 @@ fletcher_4_sse2_fini(zio_cksum_t *zcp)
struct zfs_fletcher_sse_array a, b, c, d; struct zfs_fletcher_sse_array a, b, c, d;
uint64_t A, B, C, D; uint64_t A, B, C, D;
asm volatile("movdqa %%xmm0, %0":"=m" (a.v)); asm volatile("movdqu %%xmm0, %0":"=m" (a.v));
asm volatile("movdqa %%xmm1, %0":"=m" (b.v)); asm volatile("movdqu %%xmm1, %0":"=m" (b.v));
asm volatile("psllq $0x2, %xmm2"); asm volatile("psllq $0x2, %xmm2");
asm volatile("movdqa %%xmm2, %0":"=m" (c.v)); asm volatile("movdqu %%xmm2, %0":"=m" (c.v));
asm volatile("psllq $0x3, %xmm3"); asm volatile("psllq $0x3, %xmm3");
asm volatile("movdqa %%xmm3, %0":"=m" (d.v)); asm volatile("movdqu %%xmm3, %0":"=m" (d.v));
kfpu_end(); kfpu_end();
@ -168,7 +168,7 @@ fletcher_4_ssse3_byteswap(const void *buf, uint64_t size, zio_cksum_t *unused)
const uint64_t *ip = buf; const uint64_t *ip = buf;
const uint64_t *ipend = (uint64_t *)((uint8_t *)ip + size); const uint64_t *ipend = (uint64_t *)((uint8_t *)ip + size);
asm volatile("movdqa %0, %%xmm7"::"m" (mask)); asm volatile("movdqu %0, %%xmm7"::"m" (mask));
asm volatile("pxor %xmm4, %xmm4"); asm volatile("pxor %xmm4, %xmm4");
for (; ip < ipend; ip += 2) { for (; ip < ipend; ip += 2) {