From mboxrd@z Thu Jan 1 00:00:00 1970 From: Nick Piggin Subject: Re: [patch 09/33] fs: dcache scale dentry refcount Date: Mon, 7 Sep 2009 13:35:49 +0200 Message-ID: <20090907113549.GA19521@wotan.suse.de> References: <20090904065142.114706411@nick.local0.net> <20090904065535.281876981@nick.local0.net> <7e0fb38c0909061101t6b4f337cvf691814c28bcf50e@mail.gmail.com> <20090907074431.GH1875@wotan.suse.de> <1252322481.2367.7.camel@dhcp231-106.rdu.redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Eric Paris , linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org To: Eric Paris Return-path: Content-Disposition: inline In-Reply-To: <1252322481.2367.7.camel@dhcp231-106.rdu.redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Mon, Sep 07, 2009 at 07:21:21AM -0400, Eric Paris wrote: > On Mon, 2009-09-07 at 09:44 +0200, Nick Piggin wrote: > > On Sun, Sep 06, 2009 at 02:01:05PM -0400, Eric Paris wrote: > > > On Fri, Sep 4, 2009 at 2:51 AM, wrote: > > > > > > And yet in this case we do check for dentry->d_parent == dentry. (my > > > unknowing self thinks we'd want to check in all places) > > > > I think you're right, I think we need checks there too... > > > > BTW Is there a way to test the fsnotify code? I thought inotify was > > supposed to be implemented with fsnotify, but I see quite a lot > > of duplicated (or very similar) code... We've also still got > > inotify calls in fs/ (inotify_umount_inodes). and CONFIG_INOTIFY and > > CONFIG_FSNOTIFY conditionals in there too. Would it be possible to > > move that out into fsnotify calls? (fsnotify_inode_init_once or > > whatever). > > > > (Sorry to hijack your good review comments :)) > > Actually inotify.c in linux-next is dead code which I plan to remove > in .32. In .31 inotify.c is used by the audit subsystem but inotify as > seen by userspace is implemented on top of fsnotify. Ah, fine. That explains why I had the check in fsnotify but not inotify. > I'll try to pull down the whole patch series and test for any problems. I guess I should make it available as a git tree or at least a patch rollup. People have spotted a few problems and I've found a few more in testing more filesystems, so maybe hold off testing and I'll get something newer out soon. It would be very appreciated though. Thanks, Nick