From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sebastian Andrzej Siewior Subject: Re: [PATCH] fs/f2fs: add cond_resched() to sync_dirty_dir_inodes() Date: Tue, 03 Mar 2015 09:22:37 +0100 Message-ID: <54F56F4D.7000202@linutronix.de> References: <20150220201035.GA6005@linutronix.de> <20150223233353.GA35245@jaegeuk-mac02> <20150227121314.GA21939@linutronix.de> <20150303011303.GC6541@lcm-devel.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from sog-mx-2.v43.ch3.sourceforge.com ([172.29.43.192] helo=mx.sourceforge.net) by sfs-ml-3.v29.ch3.sourceforge.com with esmtp (Exim 4.76) (envelope-from ) id 1YSi6J-00062B-Mj for linux-f2fs-devel@lists.sourceforge.net; Tue, 03 Mar 2015 08:22:47 +0000 Received: from www.linutronix.de ([62.245.132.108] helo=Galois.linutronix.de) by sog-mx-2.v43.ch3.sourceforge.com with esmtps (TLSv1:AES128-SHA:128) (Exim 4.76) id 1YSi6H-0006Lq-RE for linux-f2fs-devel@lists.sourceforge.net; Tue, 03 Mar 2015 08:22:47 +0000 In-Reply-To: <20150303011303.GC6541@lcm-devel.org> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: linux-f2fs-devel-bounces@lists.sourceforge.net To: Changman Lee Cc: Jaegeuk Kim , linux-f2fs-devel@lists.sourceforge.net On 03/03/2015 02:13 AM, Changman Lee wrote: > Hi Sebastian, Hi Jaegeuk, > After this patch, your test is all done without any CPU stall, Right? correct. > IMHO, context should be switched without cond_resched() after consumed > own time quota. So, it just reduces system latency due to yielding. Correct but without preemption enabled there is no time quota. I still have the >5sec printk but since the task schedules away I don't have a "problem" anymore. > I thought another way to discard pages of inode to be evicted in merged bio > instead of submit. If so, evict() doesn't need to wait for writeback. You are lucky that the workqueue code is aware of "that". That means it will schedule another worker if the current task schedules away because it waits for something. Otherwise you would stall other worker. If you could solve it in another way, that would be better than just polling for it. > Just my curiousity out of this problem. > > Thanks, Sebastian ------------------------------------------------------------------------------ Dive into the World of Parallel Programming The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/