From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-ed1-f70.google.com (mail-ed1-f70.google.com [209.85.208.70]) by kanga.kvack.org (Postfix) with ESMTP id 53F7A6B000C for ; Wed, 31 Oct 2018 12:56:55 -0400 (EDT) Received: by mail-ed1-f70.google.com with SMTP id k17-v6so11125131edr.18 for ; Wed, 31 Oct 2018 09:56:55 -0700 (PDT) Received: from outbound-smtp12.blacknight.com (outbound-smtp12.blacknight.com. [46.22.139.17]) by mx.google.com with ESMTPS id y30-v6si4384830edb.128.2018.10.31.09.56.53 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 31 Oct 2018 09:56:54 -0700 (PDT) Received: from mail.blacknight.com (pemlinmail03.blacknight.ie [81.17.254.16]) by outbound-smtp12.blacknight.com (Postfix) with ESMTPS id B50581C1DC5 for ; Wed, 31 Oct 2018 16:56:53 +0000 (GMT) Date: Wed, 31 Oct 2018 16:56:52 +0000 From: Mel Gorman Subject: Re: [PATCH 3/5] mm: Reclaim small amounts of memory when an external fragmentation event occurs Message-ID: <20181031165651.GE23537@techsingularity.net> References: <20181031160645.7633-1-mgorman@techsingularity.net> <20181031160645.7633-4-mgorman@techsingularity.net> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20181031160645.7633-4-mgorman@techsingularity.net> Sender: owner-linux-mm@kvack.org List-ID: To: Linux-MM Cc: Andrew Morton , Vlastimil Babka , David Rientjes , Andrea Arcangeli , Zi Yan , LKML On Wed, Oct 31, 2018 at 04:06:43PM +0000, Mel Gorman wrote: > An external fragmentation event was previously described as > > When the page allocator fragments memory, it records the event using > the mm_page_alloc_extfrag event. If the fallback_order is smaller > than a pageblock order (order-9 on 64-bit x86) then it's considered > an event that will cause external fragmentation issues in the future. > This had a build error reported by the 0-day bot. It's trivially fixed with diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 77bcc35903e0..e36c279dfade 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -3317,8 +3317,8 @@ static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone) * probably too small. It only makes sense to spread allocations to avoid * fragmentation between the Normal and DMA32 zones. */ -static inline unsigned int alloc_flags_nofragment(struct zone *zone, - gfp_t gfp_mask) +static inline unsigned int +alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask) { if (zone_idx(zone) != ZONE_NORMAL) return 0; @@ -3340,7 +3340,8 @@ static inline unsigned int alloc_flags_nofragment(struct zone *zone, return ALLOC_NOFRAGMENT; } #else -static inline unsigned int alloc_flags_nofragment(struct zone *zone) +static inline unsigned int +alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask) { return 0; } -- Mel Gorman SUSE Labs