Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934549AbbEMPIy (ORCPT ); Wed, 13 May 2015 11:08:54 -0400 Received: from mail-wi0-f171.google.com ([209.85.212.171]:38340 "EHLO mail-wi0-f171.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934376AbbEMPIu (ORCPT ); Wed, 13 May 2015 11:08:50 -0400 Date: Wed, 13 May 2015 17:11:01 +0200 From: Miklos Szeredi To: Josh Boyer Cc: Miklos Szeredi , Vincent Batts , David Howells , linux-unionfs@vger.kernel.org, "Linux-Kernel@Vger. Kernel. Org" Subject: Re: Overalyfs regression in 4.0 Message-ID: <20150513151101.GA19583@tucsk.suse.de> References: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: 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 Content-Length: 2756 Lines: 86 On Wed, May 13, 2015 at 09:06:26AM -0400, Josh Boyer wrote: > Hi Miklos, > > Vincent reported[1] what appears to be a regression in Overlayfs with > 4.0. This was found in the upstream docker community[2] on Ubuntu > with 4.0.1 as well, so it is distro agnostic. The following sequence > of commands in the bug report seems to allow one to remove a non-empty > directory. > > Is this expected behavior now? I looked through the commits in 4.0 > and saw a few that might lead to a behavior change, but I am not > familiar enough with Overalyfs to know if this was intentional or not. > > josh > > [1] https://bugzilla.redhat.com/show_bug.cgi?id=1220915 > [2] https://github.com/docker/docker/issues/13108 Good report, thanks! Follwing patch should fix it. Thanks, Miklos --- Subject: ovl: don't remove non-empty opaque directory From: Miklos Szeredi When removing an opaque directory we can't just call rmdir() to check for emptyness, because the directory will need to be replaced with a whiteout. The replacement is done with RENAME_EXCHANGE, which doesn't check emptyness. Solution is just to check emptyness by reading the directory. In the future we could add a new rename flag to check for emptyness even for RENAME_EXCHANGE to optimize this case. Reported-by: Vincent Batts Signed-off-by: Miklos Szeredi Fixes: 263b4a0fee43 ("ovl: dont replace opaque dir") Cc: # v4.0+ --- fs/overlayfs/dir.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-) --- a/fs/overlayfs/dir.c +++ b/fs/overlayfs/dir.c @@ -506,11 +506,25 @@ static int ovl_remove_and_whiteout(struc struct dentry *opaquedir = NULL; int err; - if (is_dir && OVL_TYPE_MERGE_OR_LOWER(ovl_path_type(dentry))) { - opaquedir = ovl_check_empty_and_clear(dentry); - err = PTR_ERR(opaquedir); - if (IS_ERR(opaquedir)) - goto out; + if (is_dir) { + if (OVL_TYPE_MERGE_OR_LOWER(ovl_path_type(dentry))) { + opaquedir = ovl_check_empty_and_clear(dentry); + err = PTR_ERR(opaquedir); + if (IS_ERR(opaquedir)) + goto out; + } else { + LIST_HEAD(list); + + /* + * When removing an empty opaque directory, then it + * makes no sense to replace it with an exact replica of + * itself. But emptiness still needs to be checked. + */ + err = ovl_check_empty_dir(dentry, &list); + ovl_cache_free(&list); + if (err) + goto out; + } } err = ovl_lock_rename_workdir(workdir, upperdir); -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/