From: "H.J. Lu" <hjl.tools@gmail.com>
To: Noah Goldstein <goldstein.w.n@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>,
"Carlos O'Donell" <carlos@systemhalted.org>
Subject: Re: [PATCH v5] x86: Fallback {str|wcs}cmp RTM in the ncmp overflow case [BZ #28896]
Date: Thu, 17 Feb 2022 11:20:49 -0800 [thread overview]
Message-ID: <CAMe9rOo9mvpH+vPeZJTOvw8CmR1GuRY305GiB+1TqOV482mPdg@mail.gmail.com> (raw)
In-Reply-To: <20220217191524.2961663-1-goldstein.w.n@gmail.com>
On Thu, Feb 17, 2022 at 11:15 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> In the overflow fallback strncmp-avx2-rtm and wcsncmp-avx2-rtm would
> call strcmp-avx2 and wcsncmp-avx2 respectively. This would have
> not checks around vzeroupper and would trigger spurious
> aborts. This commit fixes that.
>
> test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass on
> AVX2 machines with and without RTM.
>
> Co-authored-by: H.J. Lu <hjl.tools@gmail.com>
> ---
> sysdeps/x86/Makefile | 2 +-
> sysdeps/x86/tst-strncmp-rtm.c | 17 ++++++++++++++++-
> sysdeps/x86_64/multiarch/strcmp-avx2.S | 8 ++------
> sysdeps/x86_64/multiarch/strncmp-avx2-rtm.S | 1 +
> sysdeps/x86_64/multiarch/strncmp-avx2.S | 1 +
> sysdeps/x86_64/multiarch/wcsncmp-avx2-rtm.S | 2 +-
> sysdeps/x86_64/multiarch/wcsncmp-avx2.S | 2 +-
> 7 files changed, 23 insertions(+), 10 deletions(-)
>
> diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile
> index 6cf708335c..d110f7b7f2 100644
> --- a/sysdeps/x86/Makefile
> +++ b/sysdeps/x86/Makefile
> @@ -109,7 +109,7 @@ CFLAGS-tst-memset-rtm.c += -mrtm
> CFLAGS-tst-strchr-rtm.c += -mrtm
> CFLAGS-tst-strcpy-rtm.c += -mrtm
> CFLAGS-tst-strlen-rtm.c += -mrtm
> -CFLAGS-tst-strncmp-rtm.c += -mrtm
> +CFLAGS-tst-strncmp-rtm.c += -mrtm -Wno-error
> CFLAGS-tst-strrchr-rtm.c += -mrtm
> endif
>
> diff --git a/sysdeps/x86/tst-strncmp-rtm.c b/sysdeps/x86/tst-strncmp-rtm.c
> index 09ed6fa0d6..9e20abaacc 100644
> --- a/sysdeps/x86/tst-strncmp-rtm.c
> +++ b/sysdeps/x86/tst-strncmp-rtm.c
> @@ -16,6 +16,7 @@
> License along with the GNU C Library; if not, see
> <https://www.gnu.org/licenses/>. */
>
> +#include <stdint.h>
> #include <tst-string-rtm.h>
>
> #define LOOP 3000
> @@ -45,8 +46,22 @@ function (void)
> return 1;
> }
>
> +__attribute__ ((noinline, noclone))
> +static int
> +function_overflow (void)
> +{
> + if (strncmp (string1, string2, SIZE_MAX) == 0)
> + return 0;
> + else
> + return 1;
> +}
> +
> static int
> do_test (void)
> {
> - return do_test_1 ("strncmp", LOOP, prepare, function);
> + int status = do_test_1 ("strncmp", LOOP, prepare, function);
> + if (status != EXIT_SUCCESS)
> + return status;
> + status = do_test_1 ("strncmp", LOOP, prepare, function_overflow);
> + return status;
> }
> diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> index 07a5a2c889..52ff5ad724 100644
> --- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
> +++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> @@ -193,10 +193,10 @@ L(ret_zero):
> .p2align 4,, 5
> L(one_or_less):
> jb L(ret_zero)
> -# ifdef USE_AS_WCSCMP
> /* 'nbe' covers the case where length is negative (large
> unsigned). */
> - jnbe __wcscmp_avx2
> + jnbe OVERFLOW_STRCMP
> +# ifdef USE_AS_WCSCMP
> movl (%rdi), %edx
> xorl %eax, %eax
> cmpl (%rsi), %edx
> @@ -205,10 +205,6 @@ L(one_or_less):
> negl %eax
> orl $1, %eax
> # else
> - /* 'nbe' covers the case where length is negative (large
> - unsigned). */
> -
> - jnbe __strcmp_avx2
> movzbl (%rdi), %eax
> movzbl (%rsi), %ecx
> subl %ecx, %eax
> diff --git a/sysdeps/x86_64/multiarch/strncmp-avx2-rtm.S b/sysdeps/x86_64/multiarch/strncmp-avx2-rtm.S
> index 37d1224bb9..68bad365ba 100644
> --- a/sysdeps/x86_64/multiarch/strncmp-avx2-rtm.S
> +++ b/sysdeps/x86_64/multiarch/strncmp-avx2-rtm.S
> @@ -1,3 +1,4 @@
> #define STRCMP __strncmp_avx2_rtm
> #define USE_AS_STRNCMP 1
> +#define OVERFLOW_STRCMP __strcmp_avx2_rtm
> #include "strcmp-avx2-rtm.S"
> diff --git a/sysdeps/x86_64/multiarch/strncmp-avx2.S b/sysdeps/x86_64/multiarch/strncmp-avx2.S
> index 1678bcc235..f138e9f1fd 100644
> --- a/sysdeps/x86_64/multiarch/strncmp-avx2.S
> +++ b/sysdeps/x86_64/multiarch/strncmp-avx2.S
> @@ -1,3 +1,4 @@
> #define STRCMP __strncmp_avx2
> #define USE_AS_STRNCMP 1
> +#define OVERFLOW_STRCMP __strcmp_avx2
> #include "strcmp-avx2.S"
> diff --git a/sysdeps/x86_64/multiarch/wcsncmp-avx2-rtm.S b/sysdeps/x86_64/multiarch/wcsncmp-avx2-rtm.S
> index 4e88c70cc6..f467582cbe 100644
> --- a/sysdeps/x86_64/multiarch/wcsncmp-avx2-rtm.S
> +++ b/sysdeps/x86_64/multiarch/wcsncmp-avx2-rtm.S
> @@ -1,5 +1,5 @@
> #define STRCMP __wcsncmp_avx2_rtm
> #define USE_AS_STRNCMP 1
> #define USE_AS_WCSCMP 1
> -
> +#define OVERFLOW_STRCMP __wcscmp_avx2_rtm
> #include "strcmp-avx2-rtm.S"
> diff --git a/sysdeps/x86_64/multiarch/wcsncmp-avx2.S b/sysdeps/x86_64/multiarch/wcsncmp-avx2.S
> index 4fa1de4d3f..e9ede522b8 100644
> --- a/sysdeps/x86_64/multiarch/wcsncmp-avx2.S
> +++ b/sysdeps/x86_64/multiarch/wcsncmp-avx2.S
> @@ -1,5 +1,5 @@
> #define STRCMP __wcsncmp_avx2
> #define USE_AS_STRNCMP 1
> #define USE_AS_WCSCMP 1
> -
> +#define OVERFLOW_STRCMP __wcscmp_avx2
> #include "strcmp-avx2.S"
> --
> 2.25.1
>
LGTM.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Thanks.
--
H.J.
next prev parent reply other threads:[~2022-02-17 19:21 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-02-15 16:27 [PATCH v1] " Noah Goldstein
2022-02-15 16:29 ` H.J. Lu
2022-02-15 16:51 ` Noah Goldstein
2022-02-15 16:59 ` H.J. Lu
2022-02-15 17:06 ` Noah Goldstein
2022-02-15 18:49 ` H.J. Lu
2022-02-16 8:10 ` Noah Goldstein
2022-02-15 16:49 ` [PATCH v2] " Noah Goldstein
2022-02-16 8:09 ` [PATCH v3] " Noah Goldstein
2022-02-16 13:17 ` H.J. Lu
2022-02-16 14:08 ` Carlos O'Donell
2022-02-16 14:26 ` H.J. Lu
2022-02-17 19:15 ` Noah Goldstein
2022-02-17 19:12 ` [PATCH v4] " Noah Goldstein
2022-02-17 19:15 ` [PATCH v5] " Noah Goldstein
2022-02-17 19:20 ` H.J. Lu [this message]
2022-05-25 19:51 ` Sunil Pandey
2022-05-25 19:56 ` H.J. Lu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAMe9rOo9mvpH+vPeZJTOvw8CmR1GuRY305GiB+1TqOV482mPdg@mail.gmail.com \
--to=hjl.tools@gmail.com \
--cc=carlos@systemhalted.org \
--cc=goldstein.w.n@gmail.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).