From: Stefan Liebler <stli@linux.ibm.com>
To: Jakub Jelinek <jakub@redhat.com>
Cc: libc-alpha@sourceware.org, schwab@linux-m68k.org
Subject: Re: [PATCH] S390: Do not clobber r7 in clone [BZ #31402]
Date: Thu, 22 Feb 2024 15:04:30 +0100 [thread overview]
Message-ID: <e2951354-a5f6-4eee-a359-e52bfde0ffff@linux.ibm.com> (raw)
In-Reply-To: <ZdYlUYEx0J1J6YzR@tucnak>
On 21.02.24 17:31, Jakub Jelinek wrote:
> On Wed, Feb 21, 2024 at 05:13:07PM +0100, Stefan Liebler wrote:
>> Starting with commit e57d8fc97b90127de4ed3e3a9cdf663667580935
>> "S390: Always use svc 0"
>> clone clobbers the call-saved register r7 in error case:
>> function or stack is NULL.
>>
>> This patch restores the saved registers also in the error case.
>
> You could also just restore just %r7 and not both %r6 and %r7,
> because only %r7 has been modified. But no idea what is actually
> faster (and if equally fast what is smaller), especially when the
> saving has been done using stm/stmg of the pair.
Yes, you are right, restoring r7 is enough. Nevertheless I prefer that
the error-label is equal to the remaining code.
>
> Also, wonder if there shouldn't be a testcase covering it, just
> call clone (NULL, NULL, ...) and verify that it returned < 0/EINVAL
> and try to create high register preassure in the function across the
> call and make sure it fails without the patch and succeeds with it?
>
> Like below, though it will need to be adjusted for the glibc ways of doing
> testcases (instead of abort do what is usual for test failures, etc.),
> plus make sure it is tested only on arches which actually have clone (i.e.
> Linux).
I've extended the already existing testcase for all error cases and
added your suggested register clobber test. It fails without the fix and
passes with it.
Here is V2:
[PATCH v2] S390: Do not clobber r7 in clone [BZ #31402]
https://sourceware.org/pipermail/libc-alpha/2024-February/154911.html
>
> #define _GNU_SOURCE
> #include <stdlib.h>
> #include <sched.h>
> #include <signal.h>
> #include <errno.h>
>
> volatile unsigned v = 0xdeadbeef;
>
> int
> main ()
> {
> unsigned int a = v;
> unsigned int b = v;
> unsigned int c = v;
> unsigned int d = v;
> unsigned int e = v;
> unsigned int f = v;
> unsigned int g = v;
> unsigned int h = v;
> unsigned int i = v;
> unsigned int j = v;
> unsigned int k = v;
> unsigned int l = v;
> unsigned int m = v;
> unsigned int n = v;
> unsigned int o = v;
> if (clone (NULL, NULL, SIGCHLD, NULL) >= 0 || errno != EINVAL)
> abort ();
> if (a != v || b != v || c != v || d != v || e != v
> || f != v || g != v || h != v || i != v || j != v
> || k != v || l != v || m != v || n != v || o != v)
> abort ();
> }
>
> Jakub
>
prev parent reply other threads:[~2024-02-22 14:04 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-21 16:13 Stefan Liebler
2024-02-21 16:31 ` Jakub Jelinek
2024-02-22 14:04 ` Stefan Liebler [this message]
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=e2951354-a5f6-4eee-a359-e52bfde0ffff@linux.ibm.com \
--to=stli@linux.ibm.com \
--cc=jakub@redhat.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).