From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH RFC 0/5] IO-less balance_dirty_pages() v2 (simple approach) Date: Thu, 17 Mar 2011 12:43:54 -0400 Message-ID: <20110317164354.GA26093@infradead.org> References: <1299623475-5512-1-git-send-email-jack@suse.cz> <20110317155139.GA16195@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Christoph Hellwig , Jan Kara , linux-fsdevel@vger.kernel.org, linux-mm@kvack.org, Wu Fengguang , Peter Zijlstra , Andrew Morton To: Curt Wohlgemuth Return-path: Received: from bombadil.infradead.org ([18.85.46.34]:44353 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751572Ab1CQQn6 (ORCPT ); Thu, 17 Mar 2011 12:43:58 -0400 Content-Disposition: inline In-Reply-To: Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Thu, Mar 17, 2011 at 09:24:28AM -0700, Curt Wohlgemuth wrote: > Which is indeed part of the patchset I referred to above ("[RFC] > [PATCH 0/6] Provide cgroup isolation for buffered writes", > https://lkml.org/lkml/2011/3/8/332 ). So what about letting us fix normal writeback first and then later look into cgroups properly. And to do it properly we'll need to implement something similar to the I/O less balance dirty pages - be that targeted writeback from the flusher thread including proper tagging of pages, or be that writeback from balance_dirty_pages in a why that we keep multiple processes from writing at the same time. Although I'd prefer something that keeps the CG case as close as possible to the normal code, right now we already have a huge mess with memcg and it's own handrolled version of direct reclaim which is an even worse stack hog than the already overly painfull "normal" direct reclaim.