From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S966660AbbBDXqV (ORCPT <rfc822;w@1wt.eu>);
	Wed, 4 Feb 2015 18:46:21 -0500
Received: from mail.linuxfoundation.org ([140.211.169.12]:48726 "EHLO
	mail.linuxfoundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S966070AbbBDXqU (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 4 Feb 2015 18:46:20 -0500
Date: Wed, 4 Feb 2015 15:46:18 -0800
From: Andrew Morton <akpm@linux-foundation.org>
To: Sasha Levin <sasha.levin@oracle.com>
Cc: Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Andrey Ryabinin <a.ryabinin@samsung.com>,
        Dave Jones <davej@codemonkey.org.uk>,
        LKML <linux-kernel@vger.kernel.org>
Subject: Re: sched: memory corruption on completing completions
Message-Id: <20150204154618.36511661ab558c0a9e047a77@linux-foundation.org>
In-Reply-To: <54D2AA16.6030706@oracle.com>
References: <54D2AA16.6030706@oracle.com>
X-Mailer: Sylpheed 3.4.1 (GTK+ 2.24.23; x86_64-pc-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, 04 Feb 2015 18:24:06 -0500 Sasha Levin <sasha.levin@oracle.com> wrote:

> Hi all,
> 
> I was fuzzing with trinity on a -next kernel with the KASan patchset, and
> got what initially appeared to be a rather odd trace:
> 
> ...
>
> 
> I now have a theory for why it happens:
> 
> Thread A				Thread B
> ----------------------------------------------------------
> 
> [Enter function]
> DECLARE_COMPLETION_ONSTACK(x)
> wait_for_completion(x)
> 					complete(x)
> 					[In complete(x):]
> 				        spin_lock_irqsave(&x->wait.lock, flags);
> 				        x->done++;
> 				        __wake_up_locked(&x->wait, TASK_NORMAL, 1);
> [Done waiting, wakes up]
> [Exit function]
> 				        spin_unlock_irqrestore(&x->wait.lock, flags);
> 
> 
> 
> So the spin_unlock_irqrestore() at the end of complete() would proceed to corruption
> the stack of thread A.

But wait_for_completion() takes ->wait.lock as well, which should
provide the needed synchronization (__wait_for_common,
do_wait_for_common).  I'm not seeing a hole in the logic, but it looks
like there might be one.