From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266786AbUG1FSN (ORCPT ); Wed, 28 Jul 2004 01:18:13 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266781AbUG1FSN (ORCPT ); Wed, 28 Jul 2004 01:18:13 -0400 Received: from services.exanet.com ([212.143.73.102]:36107 "EHLO services.exanet.com") by vger.kernel.org with ESMTP id S266786AbUG1FSK (ORCPT ); Wed, 28 Jul 2004 01:18:10 -0400 Message-ID: <41073710.2020306@exanet.com> Date: Wed, 28 Jul 2004 08:18:08 +0300 From: Avi Kivity User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040510 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ulrich Weigand CC: linux-kernel@vger.kernel.org Subject: Re: [PATCH] Deadlock during heavy write activity to userspace NFS References: <200407280232.EAA14567@faui1m.informatik.uni-erlangen.de> In-Reply-To: <200407280232.EAA14567@faui1m.informatik.uni-erlangen.de> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 28 Jul 2004 05:18:09.0269 (UTC) FILETIME=[455F0A50:01C47462] Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Ulrich Weigand wrote: >Avi Kivity wrote: > > > >>In our case, all block I/O is done using >>unbuffered I/O, and all memory is preallocated, so we don't need kswapd >>at all, just that small bit of memory that syscalls consume. >> >> > >Does your userspace process need to send/receive network packets >in order to perform a write-out? > Yes. >If so, how can you make sure your >incoming packets aren't thrown away in out-of-memory situations? >(Outgoing packets can use PF_MEMALLOC memory I guess, but incoming >ones aren't associated to any process yet ...) > > > I did nothing to address this. So far it works well, even under heavy load. I guess a general solution needs to address this as well. The kernel NFS client (which kswapd depends on) has the same issue. Has anyone ever observed kswapd deadlock due to imcoming or outgoing NFS packets being discarded due to oom? Avi