From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id F171DC77B76 for ; Fri, 21 Apr 2023 13:12:35 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231856AbjDUNMe (ORCPT ); Fri, 21 Apr 2023 09:12:34 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:41212 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229543AbjDUNMc (ORCPT ); Fri, 21 Apr 2023 09:12:32 -0400 Received: from outbound-smtp46.blacknight.com (outbound-smtp46.blacknight.com [46.22.136.58]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 29CC0139 for ; Fri, 21 Apr 2023 06:12:31 -0700 (PDT) Received: from mail.blacknight.com (pemlinmail01.blacknight.ie [81.17.254.10]) by outbound-smtp46.blacknight.com (Postfix) with ESMTPS id C3753FACCF for ; Fri, 21 Apr 2023 14:12:29 +0100 (IST) Received: (qmail 32508 invoked from network); 21 Apr 2023 13:12:29 -0000 Received: from unknown (HELO techsingularity.net) (mgorman@techsingularity.net@[84.203.21.103]) by 81.17.254.9 with ESMTPSA (AES256-SHA encrypted, authenticated); 21 Apr 2023 13:12:29 -0000 Date: Fri, 21 Apr 2023 14:12:27 +0100 From: Mel Gorman To: Johannes Weiner Cc: linux-mm@kvack.org, Kaiyang Zhao , Vlastimil Babka , David Rientjes , linux-kernel@vger.kernel.org, kernel-team@fb.com Subject: Re: [RFC PATCH 08/26] mm: page_alloc: claim blocks during compaction capturing Message-ID: <20230421131227.k2afmhb6kejdbhui@techsingularity.net> References: <20230418191313.268131-1-hannes@cmpxchg.org> <20230418191313.268131-9-hannes@cmpxchg.org> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20230418191313.268131-9-hannes@cmpxchg.org> Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Apr 18, 2023 at 03:12:55PM -0400, Johannes Weiner wrote: > When capturing a whole block, update the migratetype accordingly. For > example, a THP allocation might capture an unmovable block. If the THP > gets split and partially freed later, the remainder should group up > with movable allocations. > > Signed-off-by: Johannes Weiner > --- > mm/internal.h | 1 + > mm/page_alloc.c | 42 ++++++++++++++++++++++++------------------ > 2 files changed, 25 insertions(+), 18 deletions(-) > > diff --git a/mm/internal.h b/mm/internal.h > index 024affd4e4b5..39f65a463631 100644 > --- a/mm/internal.h > +++ b/mm/internal.h > @@ -432,6 +432,7 @@ struct compact_control { > */ > struct capture_control { > struct compact_control *cc; > + int migratetype; > struct page *page; > }; > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index 4d20513c83be..8e5996f8b4b4 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -615,6 +615,17 @@ void set_pageblock_migratetype(struct page *page, int migratetype) > page_to_pfn(page), MIGRATETYPE_MASK); > } > > +static void change_pageblock_range(struct page *pageblock_page, > + int start_order, int migratetype) > +{ > + int nr_pageblocks = 1 << (start_order - pageblock_order); > + > + while (nr_pageblocks--) { > + set_pageblock_migratetype(pageblock_page, migratetype); > + pageblock_page += pageblock_nr_pages; > + } > +} > + > #ifdef CONFIG_DEBUG_VM > static int page_outside_zone_boundaries(struct zone *zone, struct page *page) > { > @@ -962,14 +973,19 @@ compaction_capture(struct capture_control *capc, struct page *page, > is_migrate_isolate(migratetype)) > return false; > > - /* > - * Do not let lower order allocations pollute a movable pageblock. > - * This might let an unmovable request use a reclaimable pageblock > - * and vice-versa but no more than normal fallback logic which can > - * have trouble finding a high-order free page. > - */ > - if (order < pageblock_order && migratetype == MIGRATE_MOVABLE) > + if (order >= pageblock_order) { > + migratetype = capc->migratetype; > + change_pageblock_range(page, order, migratetype); > + } else if (migratetype == MIGRATE_MOVABLE) { > + /* > + * Do not let lower order allocations pollute a > + * movable pageblock. This might let an unmovable > + * request use a reclaimable pageblock and vice-versa > + * but no more than normal fallback logic which can > + * have trouble finding a high-order free page. > + */ > return false; > + } > For capturing pageblock order or larger, why not unconditionally make the block MOVABLE? Even if it's a zero page allocation, it would be nice to keep the pageblock for movable pages after the split as long as possible. -- Mel Gorman SUSE Labs