From: Andrea Parri <parri.andrea@gmail.com>
To: Christian Brauner <christian.brauner@ubuntu.com>
Cc: Will Deacon <will@kernel.org>,
linux-kernel@vger.kernel.org, bsingharora@gmail.com,
dvyukov@google.com, elver@google.com, stable@vger.kernel.org,
syzbot+c5d03165a1bd1dead0c1@syzkaller.appspotmail.com,
syzkaller-bugs@googlegroups.com
Subject: Re: [PATCH v6] taskstats: fix data-race
Date: Wed, 23 Oct 2019 14:16:03 +0200 [thread overview]
Message-ID: <20191023121603.GA16344@andrea.guest.corp.microsoft.com> (raw)
In-Reply-To: <20191021113327.22365-1-christian.brauner@ubuntu.com>
On Mon, Oct 21, 2019 at 01:33:27PM +0200, Christian Brauner wrote:
> When assiging and testing taskstats in taskstats_exit() there's a race
> when writing and reading sig->stats when a thread-group with more than
> one thread exits:
>
> cpu0:
> thread catches fatal signal and whole thread-group gets taken down
> do_exit()
> do_group_exit()
> taskstats_exit()
> taskstats_tgid_alloc()
> The tasks reads sig->stats without holding sighand lock.
>
> cpu1:
> task calls exit_group()
> do_exit()
> do_group_exit()
> taskstats_exit()
> taskstats_tgid_alloc()
> The task takes sighand lock and assigns new stats to sig->stats.
>
> The first approach used smp_load_acquire() and smp_store_release().
> However, after having discussed this it seems that the data dependency
> for kmem_cache_alloc() would be fixed by WRITE_ONCE().
> Furthermore, the smp_load_acquire() would only manage to order the stats
> check before the thread_group_empty() check. So it seems just using
> READ_ONCE() and WRITE_ONCE() will do the job and I wanted to bring this
> up for discussion at least.
Mmh, the RELEASE was intended to order the memory initialization in
kmem_cache_zalloc() with the later ->stats pointer assignment; AFAICT,
there is no data dependency between such memory accesses.
Correspondingly, the ACQUIRE was intended to order the ->stats pointer
load with later, _independent dereferences of the same pointer; the
latter are, e.g., in taskstats_exit() (but not thread_group_empty()).
Looking again, I see that fill_tgid_exit()'s dereferences of ->stats
are protected by ->siglock: maybe you meant to rely on such a critical
section pairing with the critical section in taskstats_tgid_alloc()?
That memcpy(-, tsk->signal->stats, -) at the end of taskstats_exit()
also bugs me: could these dereferences of ->stats happen concurrently
with other stores to the same memory locations?
Thanks,
Andrea
next prev parent reply other threads:[~2019-10-23 12:16 UTC|newest]
Thread overview: 62+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-05 4:26 KCSAN: data-race in taskstats_exit / taskstats_exit syzbot
2019-10-05 4:29 ` Dmitry Vyukov
2019-10-05 11:29 ` Christian Brauner
2019-10-05 11:28 ` [PATCH] taskstats: fix data-race Christian Brauner
2019-10-05 13:33 ` Marco Elver
2019-10-05 14:15 ` Christian Brauner
2019-10-05 14:34 ` Marco Elver
2019-10-06 10:00 ` Dmitry Vyukov
2019-10-06 10:59 ` Christian Brauner
2019-10-06 23:52 ` Christian Brauner
2019-10-07 7:31 ` Dmitry Vyukov
2019-10-07 9:29 ` Christian Brauner
2019-10-07 10:40 ` Andrea Parri
2019-10-07 10:50 ` Christian Brauner
2019-10-07 11:01 ` [PATCH v2] " Christian Brauner
2019-10-07 13:18 ` Andrea Parri
2019-10-07 13:28 ` Christian Brauner
2019-10-07 13:50 ` Dmitry Vyukov
2019-10-07 13:55 ` Christian Brauner
2019-10-07 14:08 ` Dmitry Vyukov
2019-10-07 14:10 ` Christian Brauner
2019-10-07 14:14 ` Andrea Parri
2019-10-07 14:18 ` Dmitry Vyukov
2019-10-08 14:20 ` Andrea Parri
2019-10-08 14:24 ` Christian Brauner
2019-10-08 15:26 ` Andrea Parri
2019-10-08 15:35 ` Christian Brauner
2019-10-08 15:44 ` Andrea Parri
2019-10-09 11:31 ` [PATCH] " Christian Brauner
2019-10-09 11:40 ` Christian Brauner
2019-10-09 11:48 ` [PATCH v5] " Christian Brauner
2019-10-09 12:08 ` Andrea Parri
2019-10-09 13:26 ` Christian Brauner
2019-10-21 11:33 ` [PATCH v6] " Christian Brauner
2019-10-21 12:19 ` Rasmus Villemoes
2019-10-21 13:04 ` Christian Brauner
2019-11-29 17:56 ` Will Deacon
2019-11-30 15:08 ` Christian Brauner
2019-10-23 12:16 ` Andrea Parri [this message]
2019-10-23 12:39 ` Dmitry Vyukov
2019-10-23 13:11 ` Christian Brauner
2019-10-23 13:20 ` Dmitry Vyukov
2019-10-24 11:31 ` Andrea Parri
2019-10-24 11:51 ` Dmitry Vyukov
2019-10-24 13:05 ` Andrea Parri
2019-10-24 13:13 ` Dmitry Vyukov
2019-10-24 13:21 ` Christian Brauner
2019-10-24 13:34 ` Dmitry Vyukov
2019-10-24 13:43 ` Andrea Parri
2019-10-24 13:58 ` Dmitry Vyukov
2019-10-24 14:40 ` Andrea Parri
2019-10-24 14:49 ` Dmitry Vyukov
2019-11-29 17:57 ` Will Deacon
2019-10-09 11:48 ` [PATCH] " Marco Elver
2019-10-09 11:53 ` Christian Brauner
2019-11-06 0:27 ` Balbir Singh
2019-11-06 0:09 ` KCSAN: data-race in taskstats_exit / taskstats_exit Balbir Singh
2019-11-06 10:23 ` Marco Elver
2019-11-07 10:39 ` Balbir Singh
2019-11-08 0:54 ` Balbir Singh
2019-11-08 8:55 ` Dmitry Vyukov
2019-11-09 3:42 ` Balbir Singh
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=20191023121603.GA16344@andrea.guest.corp.microsoft.com \
--to=parri.andrea@gmail.com \
--cc=bsingharora@gmail.com \
--cc=christian.brauner@ubuntu.com \
--cc=dvyukov@google.com \
--cc=elver@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=stable@vger.kernel.org \
--cc=syzbot+c5d03165a1bd1dead0c1@syzkaller.appspotmail.com \
--cc=syzkaller-bugs@googlegroups.com \
--cc=will@kernel.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 NNTP newsgroup(s).