Return-Path: linux-nfs-owner@vger.kernel.org Received: from mx2.netapp.com ([216.240.18.37]:49433 "EHLO mx2.netapp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934915Ab2JXO45 convert rfc822-to-8bit (ORCPT ); Wed, 24 Oct 2012 10:56:57 -0400 From: "Myklebust, Trond" To: Chris Perl CC: "linux-nfs@vger.kernel.org" Subject: RE: RPC Race Condition Date: Wed, 24 Oct 2012 14:56:55 +0000 Message-ID: <4FA345DA4F4AE44899BD2B03EEEC2FA90928F920@SACEXCMBX04-PRD.hq.netapp.com> References: <20121022180339.GC24763@nyc-qws-132.nyc.delacy.com> <4FA345DA4F4AE44899BD2B03EEEC2FA909289403@SACEXCMBX04-PRD.hq.netapp.com> <20121022202611.GA27191@nyc-qws-132.nyc.delacy.com> <4FA345DA4F4AE44899BD2B03EEEC2FA90928AB39@SACEXCMBX04-PRD.hq.netapp.com> <20121023134551.GB27191@nyc-qws-132.nyc.delacy.com> <4FA345DA4F4AE44899BD2B03EEEC2FA90928C74A@SACEXCMBX04-PRD.hq.netapp.com> <20121023200855.GC27191@nyc-qws-132.nyc.delacy.com> <4FA345DA4F4AE44899BD2B03EEEC2FA90928DCDD@SACEXCMBX04-PRD.hq.netapp.com> <20121024124047.GD27191@nyc-qws-132.nyc.delacy.com> <4FA345DA4F4AE44899BD2B03EEEC2FA90928F571@SACEXCMBX04-PRD.hq.netapp.com> <20121024142333.GE27191@nyc-qws-132.nyc.delacy.com> In-Reply-To: <20121024142333.GE27191@nyc-qws-132.nyc.delacy.com> Content-Type: text/plain; charset="Windows-1252" MIME-Version: 1.0 Sender: linux-nfs-owner@vger.kernel.org List-ID: > -----Original Message----- > From: Chris Perl [mailto:chris.perl@gmail.com] > Sent: Wednesday, October 24, 2012 10:24 AM > To: Myklebust, Trond > Cc: linux-nfs@vger.kernel.org > Subject: Re: RPC Race Condition > > On Wed, Oct 24, 2012 at 01:04:16PM +0000, Myklebust, Trond wrote: > > Cool! Would you mind if I add "Tested-by" tags? > > Wouldn't mind at all. > > I've also run several of the filebench [1] workload personalities with this > patched kernel for ten minutes a piece (i.e. the `fileserver' and `networkfs' > personalities) and experienced no problems. During the tests I also added > the blackhole route to the server several times (along with restarting the NFS > server) and it recovered each time. I did this for filebench running > SCHED_OTHER and SCHED_RR with a priority of 50. > > > Sure. I'll let you know as soon as I push them out later today... > > Cool, I'll keep an eye out. OK, I've now posted the patches in the "bugfixes" branch of git://git.linux-nfs.org/projects/trondmy/linux-nfs.git By the way, you said that you were using a set of systemtap scripts to debug these races? If you'd be OK sharing those scripts, I'd love to see them. Debugging this kind of race is always difficult... Cheers Trond