From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sebastian Andrzej Siewior Subject: Re: Hung task for proc_cgroup_show Date: Mon, 21 Dec 2015 19:11:53 +0100 Message-ID: <20151221181153.GB6714@linutronix.de> References: <20150714150016.GC21820@linutronix.de> <20151211161450.GA6720@linutronix.de> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Cc: Linux RT Users To: Christoph Mathys Return-path: Received: from www.linutronix.de ([62.245.132.108]:50497 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750914AbbLUSLz (ORCPT ); Mon, 21 Dec 2015 13:11:55 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-rt-users-owner@vger.kernel.org List-ID: * Christoph Mathys | 2015-12-21 09:58:11 [+0100]: >On Fri, Dec 11, 2015 at 5:14 PM, Sebastian Andrzej Siewior > wrote: >> This means css_release_work_fn() or css_killed_work_fn(). Can you issue >> a task dump which says what the tasks are doing? > >Here is another of those lockups, this time with 3.18.24-rt22. I took >the liberty to remove everything not related to cgroups from the logs. > >Seeing the memory control group inside one of the call stacks I tried >disabling this cgroup. I did not see the lockup again, but two system >freezes instead. > So based on this output: SysRq : Show Locks Held Showing all locks held in the system: 1 lock held by kswork/69: #0: (cgroup_mutex){+.+.+.}, at: [] css_release_work_fn+0x2f/0xe0 4 locks held by cgmanager/349: #0: (sb_writers#9){.+.+.+}, at: [] mnt_want_write+0x24/0x50 #1: (&type->i_mutex_dir_key#3/1){+.+.+.}, at: [] do_rmdir+0x15c/0x220 #2: (&type->i_mutex_dir_key#3){+.+.+.}, at: [] vfs_rmdir+0x57/0x150 #3: (cgroup_mutex){+.+.+.}, at: [] cgroup_kn_lock_live+0x50/0x180 3 locks held by kworker/2:2/3642: #0: ("cgroup_destroy"){.+.+.+}, at: [] process_one_work+0x171/0x5b0 #1: ((&css->destroy_work)){+.+.+.}, at: [] process_one_work+0x171/0x5b0 #2: (cgroup_mutex){+.+.+.}, at: [] css_killed_work_fn+0x1f/0x170 and the stack trace below it looks like each one of those three tasks block on the mutex but none of them owns it. So the question is if any of those tree waiters are marked as such and if so why is there no wake up in progress. Thank you for the steps how to reproduce it. I will try it next year. Sebastian