From: Al Viro Subject: Re: kernel BUG when fsync'ing file in a overlayfs merged dir, located on btrfs Date: Thu, 24 Mar 2016 15:25:47 +0000 Message-ID: <20160324152547.GJ17997@ZenIV.linux.org.uk> References: <1443643065-16460-1-git-send-email-lebedev.ri@gmail.com> <563C171F.30702@jeffm.io> <20151106031845.GV22011@ZenIV.linux.org.uk> <563C26AE.1020403@jeffm.io> <20160324152043.GA22781@ZenIV.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20160324152043.GA22781@ZenIV.linux.org.uk> Sender: fstests-owner@vger.kernel.org To: Jeff Mahoney Cc: Roman Lebedev , David Howells , linux-btrfs@vger.kernel.org, linux-unionfs@vger.kernel.org, linux-fsdevel@vger.kernel.org, fstests@vger.kernel.org, Filipe Manana List-Id: linux-unionfs@vger.kernel.org On Thu, Mar 24, 2016 at 03:20:43PM +0000, Al Viro wrote: > > ocfs2_prepare_inode_for_write uses file->f_path.dentry for > > should_remove_suid (due to needing to do it early since cluster locking > > is unknown in setattr, according to the commit). Having > > should_remove_suid operate on an inode would solve that easily. > > Can't do - there are filesystems that _need_ dentry for ->setattr(). Grr... Sorry, misread what you'd written. should_remove_suid() ought to be switched to inode, indeed.