From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933094AbaHGQtx (ORCPT ); Thu, 7 Aug 2014 12:49:53 -0400 Received: from mail-qa0-f74.google.com ([209.85.216.74]:56389 "EHLO mail-qa0-f74.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932330AbaHGQtu (ORCPT ); Thu, 7 Aug 2014 12:49:50 -0400 References: <1407186897-21048-1-git-send-email-hannes@cmpxchg.org> <1407186897-21048-2-git-send-email-hannes@cmpxchg.org> <20140807130822.GB12730@dhcp22.suse.cz> <20140807153141.GD14734@cmpxchg.org> From: Greg Thelen To: Johannes Weiner Cc: Michal Hocko , Andrew Morton , Tejun Heo , linux-mm@kvack.org, cgroups@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [patch 1/4] mm: memcontrol: reduce reclaim invocations for higher order requests Date: Thu, 07 Aug 2014 09:10:43 -0700 In-reply-to: <20140807153141.GD14734@cmpxchg.org> Message-ID: MIME-Version: 1.0 Content-Type: text/plain Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Aug 07 2014, Johannes Weiner wrote: > On Thu, Aug 07, 2014 at 03:08:22PM +0200, Michal Hocko wrote: >> On Mon 04-08-14 17:14:54, Johannes Weiner wrote: >> > Instead of passing the request size to direct reclaim, memcg just >> > manually loops around reclaiming SWAP_CLUSTER_MAX pages until the >> > charge can succeed. That potentially wastes scan progress when huge >> > page allocations require multiple invocations, which always have to >> > restart from the default scan priority. >> > >> > Pass the request size as a reclaim target to direct reclaim and leave >> > it to that code to reach the goal. >> >> THP charge then will ask for 512 pages to be (direct) reclaimed. That >> is _a lot_ and I would expect long stalls to achieve this target. I >> would also expect quick priority drop down and potential over-reclaim >> for small and moderately sized memcgs (e.g. memcg with 1G worth of pages >> would need to drop down below DEF_PRIORITY-2 to have a chance to scan >> that many pages). All that done for a charge which can fallback to a >> single page charge. >> >> The current code is quite hostile to THP when we are close to the limit >> but solving this by introducing long stalls instead doesn't sound like a >> proper approach to me. > > THP latencies are actually the same when comparing high limit nr_pages > reclaim with the current hard limit SWAP_CLUSTER_MAX reclaim, although > system time is reduced with the high limit. > > High limit reclaim with SWAP_CLUSTER_MAX has better fault latency but > it doesn't actually contain the workload - with 1G high and a 4G load, > the consumption at the end of the run is 3.7G. > > So what I'm proposing works and is of equal quality from a THP POV. > This change is complicated enough when we stick to the facts, let's > not make up things based on gut feeling. I think that high order non THP page allocations also benefit from this. Such allocations don't have a small page fallback. This may be in flux, but linux-next shows me that: * mem_cgroup_reclaim() frees at least SWAP_CLUSTER_MAX (32) pages. * try_charge() calls mem_cgroup_reclaim() indefinitely for costly (3) or smaller orders assuming that something is reclaimed on each iteration. * try_charge() uses a loop of MEM_CGROUP_RECLAIM_RETRIES (5) for larger-than-costly orders. So for larger-than-costly allocations, try_charge() should be able to reclaim 160 (5*32) pages which satisfies an order:7 allocation. But for order:8+ allocations try_charge() and mem_cgroup_reclaim() are too eager to give up without something like this. So I think this patch is a step in the right direction. Coincidentally, we've been recently been experimenting with something like this. Though we didn't modify the interface between mem_cgroup_reclaim() and try_to_free_mem_cgroup_pages() - instead we looped within mem_cgroup_reclaim() until nr_pages of margin were found. But I have no objection the proposed plumbing of nr_pages all the way into try_to_free_mem_cgroup_pages().