public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Noah Goldstein <goldstein.w.n@gmail.com>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>,
	"Carlos O'Donell" <carlos@systemhalted.org>
Subject: Re: [PATCH v1] x86: Fix bug in strncmp-evex and strncmp-avx2 [BZ #28895]
Date: Tue, 15 Feb 2022 10:42:56 -0600	[thread overview]
Message-ID: <CAFUsyfJh34bpxYbuo6SNuj6L4Yjc+HceJB6BmHm7dH0c3f1Ljg@mail.gmail.com> (raw)
In-Reply-To: <CAMe9rOqpfMmkGcc+8ekkSg2_K3ywXFi3QTU8oifsXHTf7pGQcg@mail.gmail.com>

On Tue, Feb 15, 2022 at 10:32 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Tue, Feb 15, 2022 at 8:28 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >
> > Logic can read before the start of `s1` / `s2` if both `s1` and `s2`
> > are near the start of a page. To avoid having the result contimated by
> > these comparisons the `strcmp` variants would mask off these
> > comparisons. This was missing in the `strncmp` variants causing
> > the bug. This commit adds the masking to `strncmp` so that out of
> > range comparisons don't affect the result.
> >
> > test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass as
> > well a full xcheck on x86_64 linux.
> > ---
> >  string/test-strncmp.c                  | 24 ++++++++++++++++++++++++
> >  sysdeps/x86_64/multiarch/strcmp-avx2.S |  1 +
> >  sysdeps/x86_64/multiarch/strcmp-evex.S |  1 +
> >  3 files changed, 26 insertions(+)
> >
> > diff --git a/string/test-strncmp.c b/string/test-strncmp.c
> > index 831cb77893..501055de0c 100644
> > --- a/string/test-strncmp.c
> > +++ b/string/test-strncmp.c
> > @@ -423,6 +423,29 @@ check3 (void)
> >         }
> >  }
> >
> > +
>
> Remove the extra blank line.

Done in V2.
>
> > +static void
> > +check4 (void)
> > +{
> > +  /* To trigger bug 28895; We need 1) both s1 and s2 to be within 32 bytes of
> > +     the end of the page. 2) For there to be no mismatch/null byte before the
> > +     first page cross. 3) For length (`n`) to be large enough for one string to
> > +     cross the page. And 4) for there to be either mismatch/null bytes before
> > +     the start of the strings.  */
> > +
> > +  size_t size = 10;
> > +  CHAR *s1 = (CHAR *)(buf1 + (getpagesize () & 0xffa) / sizeof (CHAR));
> > +  CHAR *s2 = (CHAR *)(buf2 + (getpagesize () & 0xfed) / sizeof (CHAR));
> > +  int exp_result;
> > +
> > +  STRCPY (s1, L ("tst-tlsmod%"));
> > +  STRCPY (s2, L ("tst-tls-manydynamic73mod"));
> > +  exp_result = SIMPLE_STRNCMP (s1, s2, size);
> > +  FOR_EACH_IMPL (impl, 0)
> > +  check_result (impl, s1, s2, size, exp_result);
> > +}
> > +
> > +
>
> Remove the extra blank line.

Done in V2.
>
> >  static void
> >  check_overflow (void)
> >  {
> > @@ -546,6 +569,7 @@ test_main (void)
> >    check1 ();
> >    check2 ();
> >    check3 ();
> > +  check4 ();
> >
> >    printf ("%23s", "");
> >    FOR_EACH_IMPL (impl, 0)
> > diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> > index 99e5349be8..07a5a2c889 100644
> > --- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
> > +++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
> > @@ -661,6 +661,7 @@ L(ret8):
> >  # ifdef USE_AS_STRNCMP
> >         .p2align 4,, 10
> >  L(return_page_cross_end_check):
> > +       andl    %r10d, %ecx
> >         tzcntl  %ecx, %ecx
> >         leal    -VEC_SIZE(%rax, %rcx), %ecx
> >         cmpl    %ecx, %edx
> > diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
> > index 6f42e3155a..56d8c118e4 100644
> > --- a/sysdeps/x86_64/multiarch/strcmp-evex.S
> > +++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
> > @@ -689,6 +689,7 @@ L(ret8):
> >  # ifdef USE_AS_STRNCMP
> >         .p2align 4,, 10
> >  L(return_page_cross_end_check):
> > +       andl    %r10d, %ecx
> >         tzcntl  %ecx, %ecx
> >         leal    -VEC_SIZE(%rax, %rcx, SIZE_OF_CHAR), %ecx
> >  #  ifdef USE_AS_WCSCMP
> > --
> > 2.25.1
> >
>
> OK for master and 2.35 backport with the extra blank line removal.

Bug isn't present in 2.35. We had the foresight to wait until 2.36 to
push this.
>
> Thanks.
>
> --
> H.J.

  reply	other threads:[~2022-02-15 16:43 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-15 16:28 Noah Goldstein
2022-02-15 16:32 ` H.J. Lu
2022-02-15 16:42   ` Noah Goldstein [this message]
2022-02-15 16:44     ` H.J. Lu
2022-02-15 16:42 ` [PATCH v2] " Noah Goldstein
2022-02-15 16:45   ` H.J. Lu
2022-02-15 16:48 ` [PATCH v1] " Andreas Schwab
2022-02-15 16:49   ` Noah Goldstein
2022-02-15 17:04     ` Andreas Schwab
2022-02-15 17:24       ` Noah Goldstein
2022-02-15 17:23 ` [PATCH v3] " Noah Goldstein
2022-02-15 18:11   ` Andreas Schwab
2022-02-16  2:27 ` [PATCH v4] " Noah Goldstein
2022-02-16  2:28   ` H.J. Lu
2022-05-04  6:02     ` Sunil Pandey

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=CAFUsyfJh34bpxYbuo6SNuj6L4Yjc+HceJB6BmHm7dH0c3f1Ljg@mail.gmail.com \
    --to=goldstein.w.n@gmail.com \
    --cc=carlos@systemhalted.org \
    --cc=hjl.tools@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).