Return-Path: Received: from fieldses.org ([174.143.236.118]:60092 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752670Ab0IAVG0 (ORCPT ); Wed, 1 Sep 2010 17:06:26 -0400 Date: Wed, 1 Sep 2010 17:05:55 -0400 From: "J. Bruce Fields" To: Neil Brown Cc: Tim Gardner , linux-nfs@vger.kernel.org, "linux-kernel@vger.kernel.org" , Trond.Myklebust@netapp.com Subject: Re: nfsd deadlock, 2.6.36-rc3 Message-ID: <20100901210555.GB10507@fieldses.org> References: <4C7E73CB.7030603@canonical.com> <20100901165400.GB1201@fieldses.org> <20100902065551.079e297c@notabene> Content-Type: text/plain; charset=us-ascii In-Reply-To: <20100902065551.079e297c@notabene> Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Thu, Sep 02, 2010 at 06:55:51AM +1000, Neil Brown wrote: > On Wed, 1 Sep 2010 12:54:01 -0400 > "J. Bruce Fields" wrote: > > For that reason we just don't support loopback mounts--they're OK for > > light testing, but it would be difficult to make them completely robust > > under load. > > I wonder if we could use 'containers' to partition available memory between > 'nfsd threads' and 'everything else'?? Probably not worth the effort. cgroups, I don't know, I guess the essential thing would be to make sure that nfsd has the resources it needs to make forward progress, however slowly--even if it means, for example, only enough to keep a single thread processing requests. --b.