Return-Path: linux-nfs-owner@vger.kernel.org Received: from mx1.redhat.com ([209.132.183.28]:35824 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753326Ab3BPPLn (ORCPT ); Sat, 16 Feb 2013 10:11:43 -0500 Message-ID: <511FA1A6.2060903@RedHat.com> Date: Sat, 16 Feb 2013 10:11:34 -0500 From: Steve Dickson MIME-Version: 1.0 To: "Myklebust, Trond" CC: "J. Bruce Fields" , "linux-nfs@vger.kernel.org" Subject: Re: V4 unmount causes a GETATTR References: <511E5838.2020103@RedHat.com> <20130215182531.GB19923@fieldses.org> <511EAC8B.3050403@RedHat.com> <4FA345DA4F4AE44899BD2B03EEEC2FA91F3DA928@sacexcmbx05-prd.hq.netapp.com> In-Reply-To: <4FA345DA4F4AE44899BD2B03EEEC2FA91F3DA928@sacexcmbx05-prd.hq.netapp.com> Content-Type: text/plain; charset=windows-1252 Sender: linux-nfs-owner@vger.kernel.org List-ID: On 16/02/13 00:17, Myklebust, Trond wrote: >> -----Original Message----- >> From: linux-nfs-owner@vger.kernel.org [mailto:linux-nfs- >> owner@vger.kernel.org] On Behalf Of Steve Dickson >> Sent: Friday, February 15, 2013 4:46 PM >> To: J. Bruce Fields >> Cc: linux-nfs@vger.kernel.org >> Subject: Re: V4 unmount causes a GETATTR >> >> >> >> On 15/02/13 13:25, J. Bruce Fields wrote: >>> On Fri, Feb 15, 2013 at 10:46:00AM -0500, Steve Dickson wrote: >>>> Hello, >>>> >>>> I have not tracked down as to the exact reason, but it appears umount >>>> of v4 file system cause the directory to be revalidating causing the >>>> GETATTR. >>>> >>>> Is this revalidation really necessary on an unmount? >>> >>> I don't know, maybe not. But even if it's not, there are other things >>> (like cleaning up state) that the client will likely try to do on >>> unmount. In the presence of delegations it could have dirty data that >>> it's required to write back even if applications don't currently hold >>> any open files. >> Understood with dirty data... but what I as seeing was a v4 mount and then a >> reboot with no activity on the mount point... >> >>> >>> So I don't think we can promise umount will work without the network. >> Again with open files (aka activity on the mount point) I agree but with no >> activity or open files, I'm think that GETATTR is simply not necessary... > > You'd need to add a lookup intent specifically for umount. Otherwise the filesystem will not be able distinguish between path lookups for umount and other activity. > > Also note that "no activity or open files" is no guarantee that you won't be holding delegations or pNFS layouts to return; you may still see hangs. > I've also noticed force mounts this clean up does not happen (aka the GETATTR is not sent)... which is the price of doing forced mounts... steved.