From mboxrd@z Thu Jan 1 00:00:00 1970 From: Al Viro Subject: Re: [git pull] vfs pile, part 2 Date: Fri, 28 May 2010 20:47:53 +0100 Message-ID: <20100528194753.GL31073@ZenIV.linux.org.uk> References: <20100528155102.GK31073@ZenIV.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org To: Sage Weil Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Fri, May 28, 2010 at 12:27:51PM -0700, Sage Weil wrote: > Hi Al, > > There are two patches > > 1f24668c vfs: make real_lookup do dentry revalidation with i_mutex held > 982ddae4 vfs: clean up do_lookup() > > that were in your for-next a few weeks back but aren't anymore (they're > still in your autofs branch). I just want to make sure those don't get > lost if they're not going to make it upstream this time around? I think > they've been in vfs-2.6.git limbo for two release cycles now. My apologies, I thought you'd been Cc'd on the discussions with Ian concerning that stuff, but... The bottom line: "autofs4: always use lookup for lookup" is needed for them and it had to be reverted since it'd been FUBAR. Details: we really, *really* can not do things in ->d_revalidate() and count on ->lookup() to come and finish them. Think for a minute and you'll see why - between these two calls we allocate a dentry to pass to ->lookup(). What happens if that allocation fails? And without that patch your stuff breaks autofs4. So it's waiting for the ->d_revalidate() interface redesign. Hopefully in the coming cycle...