From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753299AbbCMXTA (ORCPT ); Fri, 13 Mar 2015 19:19:00 -0400 Received: from mail-ie0-f172.google.com ([209.85.223.172]:35481 "EHLO mail-ie0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751038AbbCMXS6 (ORCPT ); Fri, 13 Mar 2015 19:18:58 -0400 Date: Fri, 13 Mar 2015 16:18:55 -0700 (PDT) From: David Rientjes X-X-Sender: rientjes@chino.kir.corp.google.com To: Eric B Munson cc: Rik van Riel , Andrew Morton , Vlastimil Babka , Thomas Gleixner , Christoph Lameter , Peter Zijlstra , Mel Gorman , Michal Hocko , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH V5] Allow compaction of unevictable pages In-Reply-To: <20150313190915.GA12589@akamai.com> Message-ID: References: <1426267597-25811-1-git-send-email-emunson@akamai.com> <550332CE.7040404@redhat.com> <20150313190915.GA12589@akamai.com> User-Agent: Alpine 2.10 (DEB 1266 2009-07-14) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 13 Mar 2015, Eric B Munson wrote: > > > --- a/mm/compaction.c > > > +++ b/mm/compaction.c > > > @@ -1046,6 +1046,8 @@ typedef enum { > > > ISOLATE_SUCCESS, /* Pages isolated, migrate */ > > > } isolate_migrate_t; > > > > > > +int sysctl_compact_unevictable; > > > + > > > /* > > > * Isolate all pages that can be migrated from the first suitable block, > > > * starting at the block pointed to by the migrate scanner pfn within > > > > I suspect that the use cases where users absolutely do not want > > unevictable pages migrated are special cases, and it may make > > sense to enable sysctl_compact_unevictable by default. > > Given that sysctl_compact_unevictable=0 is the way the kernel behaves > now and the push back against always enabling compaction on unevictable > pages, I left the default to be the behavior as it is today. I agree > that this is likely the minority case, but I'd really like Peter Z or > someone else from real time to say that they are okay with the default > changing. > It would be really disappointing to not enable this by default for !rt kernels. We haven't migrated mlocked pages in the past by way of memory compaction because it can theoretically result in consistent minor page faults, but I haven't yet heard a !rt objection to enabling this. If the rt patchset is going to carry a patch to disable this, then the question arises: why not just carry an ISOLATE_UNEVICTABLE patch instead? I think you've done the due diligence required to allow this to be disabled at any time in a very easy way from userspace by the new tunable. I think it should be enabled and I'd be very surprised to hear any other objection about it other than it's different from the status quo.