linux-kernel - Re: Possible lock-less list race in scheduler

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <933936114.242980.1425679623451.JavaMail.zimbra@efficios.com>
Date:	Fri, 6 Mar 2015 22:07:03 +0000 (UTC)
From:	Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
To:	Steven Rostedt <rostedt@...dmis.org>
Cc:	Linus Torvalds <torvalds@...ux-foundation.org>,
	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>,
	Huang Ying <ying.huang@...el.com>,
	Lai Jiangshan <laijs@...fujitsu.com>,
	Lai Jiangshan <eag0628@...il.com>,
	Peter Zijlstra <peterz@...radead.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Ingo Molnar <mingo@...nel.org>
Subject: Re: Possible lock-less list race in scheduler_ipi()

----- Original Message -----
> From: "Steven Rostedt" <rostedt@...dmis.org>
> To: "Mathieu Desnoyers" <mathieu.desnoyers@...icios.com>
> Cc: "Linus Torvalds" <torvalds@...ux-foundation.org>, "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>, "Huang Ying"
> <ying.huang@...el.com>, "Lai Jiangshan" <laijs@...fujitsu.com>, "Lai Jiangshan" <eag0628@...il.com>, "Peter
> Zijlstra" <peterz@...radead.org>, "LKML" <linux-kernel@...r.kernel.org>, "Ingo Molnar" <mingo@...nel.org>
> Sent: Friday, March 6, 2015 3:39:25 PM
> Subject: Re: Possible lock-less list race in scheduler_ipi()
> 
> On Fri, 6 Mar 2015 15:38:21 -0500
> Steven Rostedt <rostedt@...dmis.org> wrote:
> 
> > On Fri, 6 Mar 2015 19:39:44 +0000 (UTC)
> > Mathieu Desnoyers <mathieu.desnoyers@...icios.com> wrote:
> > 
> > ask concurrently with the llist iteration within sched_ttwu_pending().
> > > 
> > > AFAIU, ttwu_queue_remote() is called from ttwu_queue() without holding
> > > the rq lock. So I'm wondering what prevents corruption of the wake_list
> > > in this situation.
> > 
> > I guess if it is on the wake_list, then the task's state is already
> > RUNNING. Any other task can switch a task's state to RUNNING but only
> > the task itself can switch it back to something else. If the task is on
> > the wake_list, it's state is already RUNNING, but it has not run yet.
> > That means any other wakeup will jump to the "goto out" and skip over
> > the ttwu_queue() call.
> 
> If my assumption is indeed the case, then these types of subtleties
> really need comments in the code.

My understanding is that try_to_wake_up, by calling ttwu_queue(),
is responsible for enqueuing the task into the wake_list. Inspection
of try_to_wake_up() seems to show that the state of the task is set
to TASK_WAKING by try_to_wake_up.

Then when dequeuing the task from the llist, ttwu_do_wakeup sets the
task state to TASK_RUNNING.

Both TASK_WAKING and TASK_RUNNING mean that the try_to_wake_up check
for if (!(p->state & state)), which is typically done against TASK_NORMAL,
will skip the following ttwu_queue() for that task until it is set to
TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE again.

So there should not be any double-enqueue AFAIU.

Thanks,

Mathieu





-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/