Re: [PATCH v2 12/31] x86: Fix fallback for wcsncmp_avx2 in strcmp-avx2.S [BZ #28896]

["H.J. Lu" <hjl.tools@gmail.com>]

On Thu, Mar 24, 2022 at 1:50 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> Overflow case for __wcsncmp_avx2_rtm should be __wcscmp_avx2_rtm not
> __wcscmp_avx2.
>
> commit ddf0992cf57a93200e0c782e2a94d0733a5a0b87
> Author: Noah Goldstein <goldstein.w.n@gmail.com>
> Date:   Sun Jan 9 16:02:21 2022 -0600
>
>     x86: Fix __wcsncmp_avx2 in strcmp-avx2.S [BZ# 28755]
>
> Set the wrong fallback function for `__wcsncmp_avx2_rtm`. It was set
> to fallback on to `__wcscmp_avx2` instead of `__wcscmp_avx2_rtm` which
> can cause spurious aborts.
>
> This change will need to be backported.
>
> All string/memory tests pass.
> ---
>  sysdeps/x86/tst-strncmp-rtm.c          | 15 +++++++++++++++
>  sysdeps/x86_64/multiarch/strcmp-avx2.S |  2 +-
>  2 files changed, 16 insertions(+), 1 deletion(-)
>
> diff --git a/sysdeps/x86/tst-strncmp-rtm.c b/sysdeps/x86/tst-strncmp-rtm.c
> index 300bc8c281..a3b14e72ff 100644
> --- a/sysdeps/x86/tst-strncmp-rtm.c
> +++ b/sysdeps/x86/tst-strncmp-rtm.c
> @@ -70,6 +70,16 @@ function_overflow (void)
>      return 1;
>  }
>
> +__attribute__ ((noinline, noclone))
> +static int
> +function_overflow2 (void)
> +{
> +  if (STRNCMP (string1, string2, SIZE_MAX >> 4) == 0)
> +    return 0;
> +  else
> +    return 1;
> +}
> +
>  static int
>  do_test (void)
>  {
> @@ -77,5 +87,10 @@ do_test (void)
>    if (status != EXIT_SUCCESS)
>      return status;
>    status = do_test_1 (TEST_NAME, LOOP, prepare, function_overflow);
> +  if (status != EXIT_SUCCESS)
> +    return status;
> +  status = do_test_1 (TEST_NAME, LOOP, prepare, function_overflow2);
> +  if (status != EXIT_SUCCESS)
> +    return status;
>    return status;
>  }
> diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> index 52ff5ad724..86a86b68e3 100644
> --- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
> +++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> @@ -122,7 +122,7 @@ ENTRY(STRCMP)
>            are cases where length is large enough that it can never be a
>            bound on valid memory so just use wcscmp.  */
>         shrq    $56, %rcx
> -       jnz     __wcscmp_avx2
> +       jnz     OVERFLOW_STRCMP
>
>         leaq    (, %rdx, 4), %rdx
>  #  endif
> --
> 2.25.1
>

LGTM.   Verified on AVX2 machine with RTM.  Without the fix,
the new testcase failed.

Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

Thanks.

-- 
H.J.