From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754433Ab2DIMbE (ORCPT ); Mon, 9 Apr 2012 08:31:04 -0400 Received: from mail-bk0-f46.google.com ([209.85.214.46]:50308 "EHLO mail-bk0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752571Ab2DIMbC (ORCPT ); Mon, 9 Apr 2012 08:31:02 -0400 Date: Mon, 9 Apr 2012 16:29:50 +0400 From: Anton Vorontsov To: Pekka Enberg Cc: Leonid Moiseichuk , John Stultz , linux-mm@kvack.org, linux-kernel@vger.kernel.org, linaro-kernel@lists.linaro.org Subject: Re: [PATCH 1/3] vmevent: Should not grab mutex in the atomic context Message-ID: <20120409122950.GA21833@lizard> References: <20120408233550.GA3791@panacea> <20120408233802.GA4839@panacea> <1333960831.3943.4.camel@jaguar> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <1333960831.3943.4.camel@jaguar> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Apr 09, 2012 at 11:40:31AM +0300, Pekka Enberg wrote: > On Mon, 2012-04-09 at 03:38 +0400, Anton Vorontsov wrote: > > vmevent grabs a mutex in the atomic context, and so this pops up: > > > > BUG: sleeping function called from invalid context at kernel/mutex.c:271 > > in_atomic(): 1, irqs_disabled(): 0, pid: 0, name: swapper/0 [...] > > This patch fixes the issue by removing the mutex and making the logic > > lock-free. > > > > Signed-off-by: Anton Vorontsov > > What guarantees that there's only one thread writing to struct > vmevent_attr::value in vmevent_sample() now that the mutex is gone? Well, it is called from the timer function, which has the same guaranties as an interrupt handler: it can have only one execution thread (unlike bare softirq handler), so we don't need to worry about racing w/ ourselves? If you're concerned about several instances of timers accessing the same vmevent_watch, I don't really see how it is possible, as we allocate vmevent_watch together w/ the timer instance in vmevent_fd(), so there is always one timer per vmevent_watch. Thanks, -- Anton Vorontsov Email: cbouatmailru@gmail.com