From: Carlos O'Donell <carlos@redhat.com>
To: "H.J. Lu" <hjl.tools@gmail.com>, Andreas Schwab <schwab@linux-m68k.org>
Cc: "H.J. Lu via Libc-alpha" <libc-alpha@sourceware.org>
Subject: Re: [PATCH v2] x86-64: Align child stack to 16 bytes [BZ #27902]
Date: Mon, 24 May 2021 10:19:26 -0400 [thread overview]
Message-ID: <9f26b2ac-d838-f106-211d-54c3ef829bd6@redhat.com> (raw)
In-Reply-To: <CAMe9rOopEdoNi2dDRMN-DaLC=SyKrJoFgxOZFOzz7oDUY18fHA@mail.gmail.com>
On 5/23/21 2:40 PM, H.J. Lu via Libc-alpha wrote:
> On Sun, May 23, 2021 at 10:57 AM Andreas Schwab <schwab@linux-m68k.org> wrote:
>>
>> On Mai 23 2021, H.J. Lu via Libc-alpha wrote:
>>
>>> + /* NB: Try to force misaligned child stack. */
>>> + char st[STACK_SIZE + 4] __attribute__ ((aligned (1)));
>>
>> If you want a misaligned pointer, you need to start with a known
>> alignment.
>>
>
> Fixed.
>
> Here is the v2 patch. OK for master?
Please post v3. See review below.
> From ae2dc0fe317b93d3ac1d62b2d191d45e9bee0cea Mon Sep 17 00:00:00 2001
> From: "H.J. Lu" <hjl.tools@gmail.com>
> Date: Sun, 23 May 2021 10:25:10 -0700
> Subject: [PATCH v2] x86-64: Align child stack to 16 bytes [BZ #27902]
>
> In the x86-64 clone wrapper, align child stack to 16 bytes per the
> x86-64 psABI.
> ---
> sysdeps/unix/sysv/linux/Makefile | 2 +-
> sysdeps/unix/sysv/linux/tst-misalign-clone.c | 99 ++++++++++++++++++++
> sysdeps/unix/sysv/linux/x86_64/clone.S | 3 +
> 3 files changed, 103 insertions(+), 1 deletion(-)
> create mode 100644 sysdeps/unix/sysv/linux/tst-misalign-clone.c
>
> diff --git a/sysdeps/unix/sysv/linux/Makefile b/sysdeps/unix/sysv/linux/Makefile
> index 70c3b3f8a3..d355b49033 100644
> --- a/sysdeps/unix/sysv/linux/Makefile
> +++ b/sysdeps/unix/sysv/linux/Makefile
> @@ -109,7 +109,7 @@ tests += tst-clone tst-clone2 tst-clone3 tst-fanotify tst-personality \
> tst-tgkill tst-sysvsem-linux tst-sysvmsg-linux tst-sysvshm-linux \
> tst-timerfd tst-ppoll \
> tst-clock_adjtime tst-adjtimex tst-ntp_adjtime tst-ntp_gettime \
> - tst-ntp_gettimex tst-sigtimedwait
> + tst-ntp_gettimex tst-sigtimedwait tst-misalign-clone
OK. Add two tests.
> # Test for the symbol version of fcntl that was replaced in glibc 2.28.
> ifeq ($(have-GLIBC_2.27)$(build-shared),yesyes)
> diff --git a/sysdeps/unix/sysv/linux/tst-misalign-clone.c b/sysdeps/unix/sysv/linux/tst-misalign-clone.c
> new file mode 100644
> index 0000000000..070a457d64
> --- /dev/null
> +++ b/sysdeps/unix/sysv/linux/tst-misalign-clone.c
> @@ -0,0 +1,99 @@
> +/* Copyright (C) 2021 Free Software Foundation, Inc.
Add one line test description.
> + This file is part of the GNU C Library.
> +
> + The GNU C Library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + The GNU C Library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with the GNU C Library; if not, see
> + <https://www.gnu.org/licenses/>. */
> +
> +#include <sched.h>
> +#include <stdbool.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <string.h>
> +#include <sys/wait.h>
> +#include <unistd.h>
> +#include <libc-pointer-arith.h>
> +#include <tst-stack-align.h>
> +#include <stackinfo.h>
> +#include <support/xunistd.h>
> +
> +static int
> +f (void *arg)
Please give this a real name e.g. check_func_align
> +{
> + bool ok = true;
> +
> + puts ("in f");
> +
> + if (TEST_STACK_ALIGN ())
> + ok = false;
> +
> + return ok ? 0 : 1;
> +}
OK. This is a fork-style clone for which everything should work including
calling puts() from libc.so. This answers Florian's question in his review.
> +
> +static int
> +do_test (void)
> +{
> + bool ok = true;
> +
> + puts ("in main");
s/in main/in do_test/g
> +
> + if (TEST_STACK_ALIGN ())
> + ok = false;
> +
> +#ifdef __ia64__
> +# define STACK_SIZE (256 * 1024)
> +#else
> +# define STACK_SIZE (128 * 1024)
> +#endif
> +
> + char st[STACK_SIZE + 1];
> + /* NB: Align child stack to 1 byte. */
> + char *stack = PTR_ALIGN_UP (&st[0], 2) + 1;
OK. Fixes Andreas' review.
> +
> +#ifdef __ia64__
> + extern int __clone2 (int (*__fn) (void *__arg), void *__child_stack_base,
> + size_t __child_stack_size, int __flags,
> + void *__arg, ...);
> + pid_t p = __clone2 (f, stack, STACK_SIZE, 0, 0);
> +#else
> +# if _STACK_GROWS_DOWN
> + pid_t p = clone (f, stack + STACK_SIZE, 0, 0);
> +# elif _STACK_GROWS_UP
> + pid_t p = clone (f, stack, 0, 0);
OK. Misaligned in both directions. Answers Florian's question.
The the _STACK_GROWS_UP part is only for hppa.
> +# else
> +# error "Define either _STACK_GROWS_DOWN or _STACK_GROWS_UP"
> +# endif
> +#endif
> + if (p == -1)
> + {
> + printf("clone failed: %m\n");
> + return 1;
> + }
Please use TEST_VERIFY* e.g.
/* Clone must not fail. */
TEST_VERIFY_EXIT (p != -1);
Or if you really want a specific message FAIL_EXIT.
e.g.
if (p == -1)
FAIL_EXIT1 ("clone failed: %m\n");
> +
> + int e;
> + xwaitpid (p, &e, __WCLONE);
> + if (!WIFEXITED (e))
> + {
> + if (WIFSIGNALED (e))
> + printf ("died from signal %s\n", strsignal (WTERMSIG (e)));
Add the extra information here if you want.
> + else
Remove the else.
> + puts ("did not terminate correctly");
> + return 1;
Then FAIL_EXIT1 ("Process did not termiante correctly");
> + }
> + if (WEXITSTATUS (e) != 0)
> + ok = false;
> +
> + return ok ? 0 : 1;
If (WEXITSTATUS (e) != 0)
FAIL_EXIT1 ("");
return 0;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/sysdeps/unix/sysv/linux/x86_64/clone.S b/sysdeps/unix/sysv/linux/x86_64/clone.S
> index 31ac12da0c..5f52ce7813 100644
> --- a/sysdeps/unix/sysv/linux/x86_64/clone.S
> +++ b/sysdeps/unix/sysv/linux/x86_64/clone.S
> @@ -57,6 +57,9 @@ ENTRY (__clone)
> testq %rsi,%rsi /* no NULL stack pointers */
> jz SYSCALL_ERROR_LABEL
>
> + /* Align stack to 16 bytes per the x86-64 psABI. */
> + andq $-16, %rsi
OK. Interesting to see this was missing, but maybe we never get this situation in production.
> +
> /* Insert the argument onto the new stack. */
> subq $16,%rsi
> movq %rcx,8(%rsi)
> --
> 2.31.1
>
--
Cheers,
Carlos.
next prev parent reply other threads:[~2021-05-24 14:19 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-23 17:29 [PATCH] x86-64: Align child stack to 16 bytes H.J. Lu
2021-05-23 17:57 ` Andreas Schwab
2021-05-23 18:40 ` [PATCH v2] x86-64: Align child stack to 16 bytes [BZ #27902] H.J. Lu
2021-05-24 12:07 ` Florian Weimer
2021-05-24 12:53 ` H.J. Lu
2021-05-24 14:19 ` Carlos O'Donell [this message]
2021-05-24 14:55 ` Noah Goldstein
2021-05-24 18:46 ` H.J. Lu
2021-05-24 18:46 ` 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=9f26b2ac-d838-f106-211d-54c3ef829bd6@redhat.com \
--to=carlos@redhat.com \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
--cc=schwab@linux-m68k.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).