[v2] Fix x86_64 memchr for large input sizes
Commit Message
Changes from previous version
- Simplify uintptr_t generation and check.
If noone opposes it I will commit it shortly.
--
Current optimized memchr for x86_64 does for input arguments pointers
module 64 in range of [49,63] if there is no searchr char in the rest
of 64-byte block a pointer addition which might overflow:
* sysdeps/x86_64/memchr.S
77 .p2align 4
78 L(unaligned_no_match):
79 add %rcx, %rdx
Add (uintptr_t)s % 16 to n in %rdx.
80 sub $16, %rdx
81 jbe L(return_null)
This patch fixes by adding a saturated math that sets a maximum pointer
value if it overflows (UINTPTR_MAX).
Checked on x86_64-linux-gnu and powerpc64-linux-gnu.
[BZ# 19387]
* sysdeps/x86_64/memchr.S (memchr): Avoid overflow in pointer
addition.
* string/test-memchr.c (do_test): Remove alignment limitation.
(test_main): Add test that trigger BZ# 19387.
---
string/test-memchr.c | 9 ++++-----
sysdeps/x86_64/memchr.S | 6 ++++++
3 files changed, 18 insertions(+), 5 deletions(-)
Comments
FWIW, I'm seeing string/test-memchr failing on i686; I suppose at least
one of the i686 versions must have a similar issue.
@@ -76,7 +76,6 @@ do_test (size_t align, size_t pos, size_t len, size_t n, int seek_char)
size_t i;
CHAR *result;
- align &= 7;
if ((align + len) * sizeof (CHAR) >= page_size)
return;
@@ -194,12 +193,12 @@ test_main (void)
do_test (i, 64, 256, SIZE_MAX, 0);
}
- for (i = 1; i < 16; ++i)
+ for (i = 1; i < 64; ++i)
{
- for (j = 1; j < 16; j++)
+ for (j = 1; j < 64; j++)
{
- do_test (0, 16 - j, 16, SIZE_MAX, 23);
- do_test (i, 16 - j, 16, SIZE_MAX, 23);
+ do_test (0, 64 - j, 64, SIZE_MAX, 23);
+ do_test (i, 64 - j, 64, SIZE_MAX, 23);
}
}
@@ -76,7 +76,13 @@ L(crosscache):
.p2align 4
L(unaligned_no_match):
+ /* Calculate the last acceptable address and check for possible
+ addition overflow by using satured math:
+ rdx = rcx + rdx
+ rdx |= -(rdx < x) */
add %rcx, %rdx
+ sbb %rax, %rax
+ or %rax, %rdx
sub $16, %rdx
jbe L(return_null)
add $16, %rdi