From: "H.J. Lu" <hjl.tools@gmail.com>
To: Noah Goldstein <goldstein.w.n@gmail.com>
Cc: libc-alpha@sourceware.org, carlos@systemhalted.org
Subject: Re: [PATCH v1] x86: Fix wcsnlen-avx2 page cross length comparison [BZ #29591]
Date: Wed, 21 Sep 2022 15:01:25 -0700 [thread overview]
Message-ID: <CAMe9rOobcrtq0dSy5pPqDNahBjJVd7LVs-iT6FcQw5VO7mRi=g@mail.gmail.com> (raw)
In-Reply-To: <20220921005804.7131-1-goldstein.w.n@gmail.com>
On Tue, Sep 20, 2022 at 5:58 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> Previous implementation was adjusting length (rsi) to match
> bytes (eax), but since there is no bound to length this can cause
> overflow.
>
> Fix is to just convert the byte-count (eax) to length by dividing by
> sizeof (wchar_t) before the comparison.
>
> Full check passes on x86-64 and build succeeds w/ and w/o multiarch.
> ---
> string/test-strnlen.c | 70 +++++++++++++++-----------
> sysdeps/x86_64/multiarch/strlen-avx2.S | 7 +--
> 2 files changed, 43 insertions(+), 34 deletions(-)
>
> diff --git a/string/test-strnlen.c b/string/test-strnlen.c
> index 4a9375112a..5cbaf4b734 100644
> --- a/string/test-strnlen.c
> +++ b/string/test-strnlen.c
> @@ -73,7 +73,7 @@ do_test (size_t align, size_t len, size_t maxlen, int max_char)
> {
> size_t i;
>
> - align &= 63;
> + align &= (getpagesize () / sizeof (CHAR) - 1);
> if ((align + len) * sizeof (CHAR) >= page_size)
> return;
>
> @@ -90,38 +90,50 @@ do_test (size_t align, size_t len, size_t maxlen, int max_char)
> static void
> do_overflow_tests (void)
> {
> - size_t i, j, len;
> + size_t i, j, al_idx, repeats, len;
> const size_t one = 1;
> uintptr_t buf_addr = (uintptr_t) buf1;
> + const size_t alignments[] = { 0, 1, 7, 9, 31, 33, 63, 65, 95, 97, 127, 129 };
>
> - for (i = 0; i < 750; ++i)
> + for (al_idx = 0; al_idx < sizeof (alignments) / sizeof (alignments[0]);
> + al_idx++)
> {
> - do_test (1, i, SIZE_MAX, BIG_CHAR);
> -
> - do_test (0, i, SIZE_MAX - i, BIG_CHAR);
> - do_test (0, i, i - buf_addr, BIG_CHAR);
> - do_test (0, i, -buf_addr - i, BIG_CHAR);
> - do_test (0, i, SIZE_MAX - buf_addr - i, BIG_CHAR);
> - do_test (0, i, SIZE_MAX - buf_addr + i, BIG_CHAR);
> -
> - len = 0;
> - for (j = 8 * sizeof(size_t) - 1; j ; --j)
> - {
> - len |= one << j;
> - do_test (0, i, len - i, BIG_CHAR);
> - do_test (0, i, len + i, BIG_CHAR);
> - do_test (0, i, len - buf_addr - i, BIG_CHAR);
> - do_test (0, i, len - buf_addr + i, BIG_CHAR);
> -
> - do_test (0, i, ~len - i, BIG_CHAR);
> - do_test (0, i, ~len + i, BIG_CHAR);
> - do_test (0, i, ~len - buf_addr - i, BIG_CHAR);
> - do_test (0, i, ~len - buf_addr + i, BIG_CHAR);
> -
> - do_test (0, i, -buf_addr, BIG_CHAR);
> - do_test (0, i, j - buf_addr, BIG_CHAR);
> - do_test (0, i, -buf_addr - j, BIG_CHAR);
> - }
> + for (repeats = 0; repeats < 2; ++repeats)
> + {
> + size_t align = repeats ? (getpagesize () - alignments[al_idx])
> + : alignments[al_idx];
> + align /= sizeof (CHAR);
> + for (i = 0; i < 750; ++i)
> + {
> + do_test (align, i, SIZE_MAX, BIG_CHAR);
> +
> + do_test (align, i, SIZE_MAX - i, BIG_CHAR);
> + do_test (align, i, i - buf_addr, BIG_CHAR);
> + do_test (align, i, -buf_addr - i, BIG_CHAR);
> + do_test (align, i, SIZE_MAX - buf_addr - i, BIG_CHAR);
> + do_test (align, i, SIZE_MAX - buf_addr + i, BIG_CHAR);
> +
> + len = 0;
> + for (j = 8 * sizeof (size_t) - 1; j; --j)
> + {
> + len |= one << j;
> + do_test (align, i, len, BIG_CHAR);
> + do_test (align, i, len - i, BIG_CHAR);
> + do_test (align, i, len + i, BIG_CHAR);
> + do_test (align, i, len - buf_addr - i, BIG_CHAR);
> + do_test (align, i, len - buf_addr + i, BIG_CHAR);
> +
> + do_test (align, i, ~len - i, BIG_CHAR);
> + do_test (align, i, ~len + i, BIG_CHAR);
> + do_test (align, i, ~len - buf_addr - i, BIG_CHAR);
> + do_test (align, i, ~len - buf_addr + i, BIG_CHAR);
> +
> + do_test (align, i, -buf_addr, BIG_CHAR);
> + do_test (align, i, j - buf_addr, BIG_CHAR);
> + do_test (align, i, -buf_addr - j, BIG_CHAR);
> + }
> + }
> + }
> }
> }
>
> diff --git a/sysdeps/x86_64/multiarch/strlen-avx2.S b/sysdeps/x86_64/multiarch/strlen-avx2.S
> index 0593fb303b..b9b58ef599 100644
> --- a/sysdeps/x86_64/multiarch/strlen-avx2.S
> +++ b/sysdeps/x86_64/multiarch/strlen-avx2.S
> @@ -544,14 +544,11 @@ L(return_vzeroupper):
> L(cross_page_less_vec):
> tzcntl %eax, %eax
> # ifdef USE_AS_WCSLEN
> - /* NB: Multiply length by 4 to get byte count. */
> - sall $2, %esi
> + /* NB: Divide by 4 to convert from byte-count to length. */
> + shrl $2, %eax
> # endif
> cmpq %rax, %rsi
> cmovb %esi, %eax
> -# ifdef USE_AS_WCSLEN
> - shrl $2, %eax
> -# endif
> VZEROUPPER_RETURN
> # endif
>
> --
> 2.34.1
>
LGTM.
Thanks.
--
H.J.
next prev parent reply other threads:[~2022-09-21 22:02 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-21 0:58 Noah Goldstein
2022-09-21 22:01 ` H.J. Lu [this message]
2022-11-23 22:20 ` Sunil Pandey
2022-11-24 0:22 ` H.J. Lu
2022-11-24 3:04 ` Sunil Pandey
2022-11-24 3:12 ` 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='CAMe9rOobcrtq0dSy5pPqDNahBjJVd7LVs-iT6FcQw5VO7mRi=g@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).