From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752701AbdJDNSr (ORCPT ); Wed, 4 Oct 2017 09:18:47 -0400 Received: from mail.univention.de ([82.198.197.8]:17468 "EHLO mail.univention.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752191AbdJDNSn (ORCPT ); Wed, 4 Oct 2017 09:18:43 -0400 Subject: Re: OOPS: linux-4.9.33/fs/pnode.c:propagate_one() To: "Eric W. Biederman" Cc: linux-fsdevel@vger.kernel.org, "linux-kernel@vger.kernel.org" , Felix Botner , Dirk Wiesenthal References: <8760c2u5bh.fsf@xmission.com> <87lgksb2k5.fsf@xmission.com> From: Philipp Hahn Organization: Univention GmbH Message-ID: <98a9dcb0-1040-d783-d263-70ebe63a85de@univention.de> Date: Wed, 4 Oct 2017 15:18:02 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.3.0 MIME-Version: 1.0 In-Reply-To: <87lgksb2k5.fsf@xmission.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-GB Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hello Eric, Am 03.10.2017 um 18:48 schrieb Eric W. Biederman: > ebiederm@xmission.com (Eric W. Biederman) writes: > >> Philipp Hahn writes: >> >>> We will try 4.9.52 next and see, if we can still reproduce it. > > Can you still reproduce this on 4.9.52? My college Dirk is able to reproduce it only with .33, but not with .52 (yet). > I am not seeing anything obvious that would result in last_source > becomming NULL there so knowing I am not lookning for a phantom would be very > helpful. Looks like .52 closed the race windows sufficiently enough, so the bug does not get triggered there. The bug is triggered when using "docker cp" to copy a file into the container. It communicates with the long-running 'dockerd', which rebuilds the overlay FS itself to access the file. The failing syscall is: > mount("/fw/data_folder", "/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/fwxserver/Data Folder", 0xc82135e5c8, MS_BIND|MS_REC, NULL I just now noticed that before the mount() docker did this: > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/media", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/tmp", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/etc", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/log", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/var/lib/univention-appcenter/apps/filewave/data", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/var/lib/univention-appcenter/apps/filewave/conf", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/certs", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/fwcld", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/fwxserver/Data Folder", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/apache/conf", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/fwxserver/DB", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/apache/logs", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/private/var/log", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/apache/passwords", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/sys/fs/cgroup", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/ipa", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/etc/apt/apt.conf.d/80proxy", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/usr/local/filewave/postgresql/conf", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/etc/resolv.conf", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/etc/hostname", MNT_DETACH) = 0 > umount("/var/lib/docker/overlay/038e54bc5db6a180409aa0e5b0f84087b93ac7ffe4eaa3c8908aea79d43a45d3/merged/etc/hosts", MNT_DETACH) = 0 Is MNT_DETACH a problem, as it is described to do a "lazy" umount? Doing another > strace -ttt -b execve -f -ff -s 65536 -o /var/tmp/docker.strace -p $(