From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752622AbcD1ObU (ORCPT ); Thu, 28 Apr 2016 10:31:20 -0400 Received: from mail.kernel.org ([198.145.29.136]:58459 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751411AbcD1ObT (ORCPT ); Thu, 28 Apr 2016 10:31:19 -0400 Subject: Re: [f2fs-dev] [PATCH RESEND 2/2] f2fs: disable preemption when waiting on all pages writeback To: Peter Zijlstra , Chao Yu References: <5720C19C.3000503@kernel.org> <20160427180801.GA33099@jaegeuk.gateway> <0fda915e-2945-9c2b-0e5a-dff3b4e37a92@huawei.com> <20160428140314.GD3448@twins.programming.kicks-ass.net> Cc: Jaegeuk Kim , linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net From: Chao Yu Message-ID: <57221E9F.30002@kernel.org> Date: Thu, 28 Apr 2016 22:30:55 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101 Thunderbird/38.7.2 MIME-Version: 1.0 In-Reply-To: <20160428140314.GD3448@twins.programming.kicks-ass.net> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 2016/4/28 22:03, Peter Zijlstra wrote: > On Thu, Apr 28, 2016 at 07:51:04PM +0800, Chao Yu wrote: >>> On Wed, Apr 27, 2016 at 09:41:48PM +0800, Chao Yu wrote: >>>> From: Chao Yu >>>> >>>> The following condition can happen in a preemptible kernel, it may cause >>>> checkpointer hunging. >>>> >>>> CPU0: CPU1: >>>> - write_checkpoint >>>> - do_checkpoint >>>> - wait_on_all_pages_writeback >>>> - f2fs_write_end_io >>>> - wake_up >>>> this is last writebacked page, but >>>> no sleeper in sbi->cp_wait wait >>>> queue, wake_up is not been called. >>>> - prepare_to_wait(TASK_UNINTERRUPTIBLE) >>>> Here, current task can been preempted, >>>> but there will be no waker since last >>>> write_end_io has bypassed wake_up. So >>>> current task will sleep forever. > > But here, you should be verifying if you really should go sleep; as the > code did; it tests for !get_pages(, F2FS_WRITEBACK), and if you've just > completed that very last one, this will break out. You mean after being preempted with TASK_UNINTERRUPTIBLE status, that task still has chance to be scheduled to check '!get_pages(, F2FS_WRITEBACK)', is that right? > >>>> - io_schedule_timeout >>>