|
 |
00db10 |
commit 9aaaab7c6e4176e61c59b0a63c6ba906d875dc0e
|
|
|
00db10 |
Author: Andreas Schwab <schwab@suse.de>
|
|
|
00db10 |
Date: Tue May 22 10:37:59 2018 +0200
|
|
|
00db10 |
|
|
|
00db10 |
Don't write beyond destination in __mempcpy_avx512_no_vzeroupper (bug 23196)
|
|
|
00db10 |
|
|
|
00db10 |
When compiled as mempcpy, the return value is the end of the destination
|
|
|
00db10 |
buffer, thus it cannot be used to refer to the start of it.
|
|
|
00db10 |
|
|
|
00db10 |
Conflicts:
|
|
|
00db10 |
sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
|
|
|
00db10 |
|
|
|
00db10 |
Change applied to sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
|
|
|
00db10 |
instead.
|
|
|
00db10 |
|
|
|
00db10 |
diff --git a/string/test-mempcpy.c b/string/test-mempcpy.c
|
|
|
00db10 |
index 1679e9d11f282c55..7ccce13e71bd10a9 100644
|
|
|
00db10 |
--- a/string/test-mempcpy.c
|
|
|
00db10 |
+++ b/string/test-mempcpy.c
|
|
|
00db10 |
@@ -18,6 +18,7 @@
|
|
|
00db10 |
<http://www.gnu.org/licenses/>. */
|
|
|
00db10 |
|
|
|
00db10 |
#define MEMCPY_RESULT(dst, len) (dst) + (len)
|
|
|
00db10 |
+#define MIN_PAGE_SIZE 131072
|
|
|
00db10 |
#define TEST_MAIN
|
|
|
00db10 |
#define TEST_NAME "mempcpy"
|
|
|
00db10 |
#include "test-string.h"
|
|
|
00db10 |
diff --git a/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
|
|
|
00db10 |
index 058e871847a77c45..b78710ea9359e129 100644
|
|
|
00db10 |
--- a/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
|
|
|
00db10 |
+++ b/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
|
|
|
00db10 |
@@ -333,6 +333,7 @@ L(preloop_large):
|
|
|
00db10 |
vmovups (%rsi), %zmm4
|
|
|
00db10 |
vmovups 0x40(%rsi), %zmm5
|
|
|
00db10 |
|
|
|
00db10 |
+ mov %rdi, %r11
|
|
|
00db10 |
/* Align destination for access with non-temporal stores in the loop. */
|
|
|
00db10 |
mov %rdi, %r8
|
|
|
00db10 |
and $-0x80, %rdi
|
|
|
00db10 |
@@ -363,8 +364,8 @@ L(gobble_256bytes_nt_loop):
|
|
|
00db10 |
cmp $256, %rdx
|
|
|
00db10 |
ja L(gobble_256bytes_nt_loop)
|
|
|
00db10 |
sfence
|
|
|
00db10 |
- vmovups %zmm4, (%rax)
|
|
|
00db10 |
- vmovups %zmm5, 0x40(%rax)
|
|
|
00db10 |
+ vmovups %zmm4, (%r11)
|
|
|
00db10 |
+ vmovups %zmm5, 0x40(%r11)
|
|
|
00db10 |
jmp L(check)
|
|
|
00db10 |
|
|
|
00db10 |
L(preloop_large_bkw):
|