From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753588AbaE0TpL (ORCPT ); Tue, 27 May 2014 15:45:11 -0400 Received: from zene.cmpxchg.org ([85.214.230.12]:33391 "EHLO zene.cmpxchg.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752384AbaE0TpI (ORCPT ); Tue, 27 May 2014 15:45:08 -0400 Date: Tue, 27 May 2014 15:45:00 -0400 From: Johannes Weiner To: Michal Hocko Cc: linux-mm@kvack.org, Hugh Dickins , Tejun Heo , cgroups@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [patch 6/9] mm: memcontrol: remove ordering between pc->mem_cgroup and PageCgroupUsed Message-ID: <20140527194500.GC2878@cmpxchg.org> References: <1398889543-23671-1-git-send-email-hannes@cmpxchg.org> <1398889543-23671-7-git-send-email-hannes@cmpxchg.org> <20140523132043.GB22135@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20140523132043.GB22135@dhcp22.suse.cz> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, May 23, 2014 at 03:20:43PM +0200, Michal Hocko wrote: > On Wed 30-04-14 16:25:40, Johannes Weiner wrote: > > There is a write barrier between setting pc->mem_cgroup and > > PageCgroupUsed, which was added to allow LRU operations to lookup the > > memcg LRU list of a page without acquiring the page_cgroup lock. But > > ever since 38c5d72f3ebe ("memcg: simplify LRU handling by new rule"), > > pages are ensured to be off-LRU while charging, so nobody else is > > changing LRU state while pc->mem_cgroup is being written. > > This is quite confusing. Why do we have the lrucare path then? Some charge paths start with the page on the LRU, lrucare makes sure it's off during the charge. > The code is quite tricky so this deserves a more detailed explanation > IMO. > > There are only 3 paths which check both the flag and mem_cgroup ( > without page_cgroup_lock) get_mctgt_type* and mem_cgroup_page_lruvec AFAICS. > None of them have rmb so there was no guarantee about ordering anyway. Yeah, exactly. As per the changelog, this is a remnant of the way it used to work but it's no longer needed because of guaranteed off-LRU state. > > Signed-off-by: Johannes Weiner > > Anyway, the change is welcome > Acked-by: Michal Hocko Thanks!