From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755763AbdGXMuc (ORCPT ); Mon, 24 Jul 2017 08:50:32 -0400 Received: from mx2.suse.de ([195.135.220.15]:34416 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1755716AbdGXMuY (ORCPT ); Mon, 24 Jul 2017 08:50:24 -0400 Date: Mon, 24 Jul 2017 14:50:15 +0200 From: Michal Hocko To: Vlastimil Babka Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Joonsoo Kim , Andrew Morton , Mel Gorman , Yang Shi , Laura Abbott , Vinayak Menon , zhong jiang Subject: Re: [PATCH 3/4] mm, page_owner: don't grab zone->lock for init_pages_in_zone() Message-ID: <20170724125015.GJ25221@dhcp22.suse.cz> References: <20170720134029.25268-1-vbabka@suse.cz> <20170720134029.25268-4-vbabka@suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170720134029.25268-4-vbabka@suse.cz> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu 20-07-17 15:40:28, Vlastimil Babka wrote: > init_pages_in_zone() is run under zone->lock, which means a long lock time and > disabled interrupts on large machines. This is currently not an issue since it > runs early in boot, but a later patch will change that. > However, like other pfn scanners, we don't actually need zone->lock even when > other cpus are running. The only potentially dangerous operation here is > reading bogus buddy page owner due to race, and we already know how to handle > that. The worse that can happen is that we skip some early allocated pages, > which should not affect the debugging power of page_owner noticeably. Makes sense to me > Signed-off-by: Vlastimil Babka Acked-by: Michal Hocko > --- > mm/page_owner.c | 16 ++++++++++------ > 1 file changed, 10 insertions(+), 6 deletions(-) > > diff --git a/mm/page_owner.c b/mm/page_owner.c > index 5aa21ca237d9..cf6568d1dc14 100644 > --- a/mm/page_owner.c > +++ b/mm/page_owner.c > @@ -567,11 +567,17 @@ static void init_pages_in_zone(pg_data_t *pgdat, struct zone *zone) > continue; > > /* > - * We are safe to check buddy flag and order, because > - * this is init stage and only single thread runs. > + * To avoid having to grab zone->lock, be a little > + * careful when reading buddy page order. The only > + * danger is that we skip too much and potentially miss > + * some early allocated pages, which is better than > + * heavy lock contention. > */ > if (PageBuddy(page)) { > - pfn += (1UL << page_order(page)) - 1; > + unsigned long order = page_order_unsafe(page); > + > + if (order > 0 && order < MAX_ORDER) > + pfn += (1UL << order) - 1; > continue; > } > > @@ -590,6 +596,7 @@ static void init_pages_in_zone(pg_data_t *pgdat, struct zone *zone) > __set_page_owner_init(page_ext, init_handle); > count++; > } > + cond_resched(); > } > > pr_info("Node %d, zone %8s: page owner found early allocated %lu pages\n", > @@ -600,15 +607,12 @@ static void init_zones_in_node(pg_data_t *pgdat) > { > struct zone *zone; > struct zone *node_zones = pgdat->node_zones; > - unsigned long flags; > > for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) { > if (!populated_zone(zone)) > continue; > > - spin_lock_irqsave(&zone->lock, flags); > init_pages_in_zone(pgdat, zone); > - spin_unlock_irqrestore(&zone->lock, flags); > } > } > > -- > 2.13.2 -- Michal Hocko SUSE Labs