Date:   Wed, 26 Jul 2017 12:50:04 +0200
From:   Michal Hocko <mhocko@...nel.org>
To:     Luiz Capitulino <lcapitulino@...hat.com>,
        Mike Kravetz <mike.kravetz@...cle.com>
Cc:     linux-mm@...ck.org, LKML <linux-kernel@...r.kernel.org>
Subject: gigantic hugepages vs. movable zones

Hi,
I've just noticed that alloc_gigantic_page ignores movability of the
gigantic page and it uses any existing zone. Considering that
hugepage_migration_supported only supports 2MB and pgd level hugepages
then 1GB pages are not migratable and as such allocating them from a
movable zone will break the basic expectation of this zone. Standard
hugetlb allocations try to avoid that by using htlb_alloc_mask and I
believe we should do the same for gigantic pages as well.

I suspect this behavior is not intentional. What do you think about the
following untested patch?
---
>From 542d32c1eca7dcf38afca1a91bca4a472f6e8651 Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@...e.com>
Date: Wed, 26 Jul 2017 12:43:43 +0200
Subject: [PATCH] mm, hugetlb: do not allocate non-migrateable gigantic pages
 from movable zones

alloc_gigantic_page doesn't consider movability of the gigantic hugetlb
when scanning eligible ranges for the allocation. As 1GB hugetlb pages
are not movable currently this can break the movable zone assumption
that all allocations are migrateable and as such break memory hotplug.

Reorganize the code and use the standard zonelist allocations scheme
that we use for standard hugetbl pages. htlb_alloc_mask will ensure that
only migratable hugetlb pages will ever see a movable zone.

Fixes: 944d9fec8d7a ("hugetlb: add support for gigantic page allocation at runtime")
Signed-off-by: Michal Hocko <mhocko@...e.com>
---
 mm/hugetlb.c | 35 ++++++++++++++++++++---------------
 1 file changed, 20 insertions(+), 15 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index bc48ee783dd9..60530bb3d228 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1066,11 +1066,11 @@ static void free_gigantic_page(struct page *page, unsigned int order)
 }
 
 static int __alloc_gigantic_page(unsigned long start_pfn,
-				unsigned long nr_pages)
+				unsigned long nr_pages, gfp_t gfp_mask)
 {
 	unsigned long end_pfn = start_pfn + nr_pages;
 	return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE,
-				  GFP_KERNEL);
+				  gfp_mask);
 }
 
 static bool pfn_range_valid_gigantic(struct zone *z,
@@ -1108,19 +1108,24 @@ static bool zone_spans_last_pfn(const struct zone *zone,
 	return zone_spans_pfn(zone, last_pfn);
 }
 
-static struct page *alloc_gigantic_page(int nid, unsigned int order)
+static struct page *alloc_gigantic_page(int nid, struct hstate *h)
 {
+	unsigned int order = huge_page_order(h);
 	unsigned long nr_pages = 1 << order;
 	unsigned long ret, pfn, flags;
-	struct zone *z;
+	struct zonelist *zonelist;
+	struct zone *zone;
+	struct zoneref *z;
+	gfp_t gfp_mask;
 
-	z = NODE_DATA(nid)->node_zones;
-	for (; z - NODE_DATA(nid)->node_zones < MAX_NR_ZONES; z++) {
-		spin_lock_irqsave(&z->lock, flags);
+	gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
+	zonelist = node_zonelist(nid, gfp_mask);
+	for_each_zone_zonelist_nodemask(zone, z, zonelist, gfp_zone(gfp_mask), NULL) {
+		spin_lock_irqsave(&zone->lock, flags);
 
-		pfn = ALIGN(z->zone_start_pfn, nr_pages);
-		while (zone_spans_last_pfn(z, pfn, nr_pages)) {
-			if (pfn_range_valid_gigantic(z, pfn, nr_pages)) {
+		pfn = ALIGN(zone->zone_start_pfn, nr_pages);
+		while (zone_spans_last_pfn(zone, pfn, nr_pages)) {
+			if (pfn_range_valid_gigantic(zone, pfn, nr_pages)) {
 				/*
 				 * We release the zone lock here because
 				 * alloc_contig_range() will also lock the zone
@@ -1128,16 +1133,16 @@ static struct page *alloc_gigantic_page(int nid, unsigned int order)
 				 * spinning on this lock, it may win the race
 				 * and cause alloc_contig_range() to fail...
 				 */
-				spin_unlock_irqrestore(&z->lock, flags);
-				ret = __alloc_gigantic_page(pfn, nr_pages);
+				spin_unlock_irqrestore(&zone->lock, flags);
+				ret = __alloc_gigantic_page(pfn, nr_pages, gfp_mask);
 				if (!ret)
 					return pfn_to_page(pfn);
-				spin_lock_irqsave(&z->lock, flags);
+				spin_lock_irqsave(&zone->lock, flags);
 			}
 			pfn += nr_pages;
 		}
 
-		spin_unlock_irqrestore(&z->lock, flags);
+		spin_unlock_irqrestore(&zone->lock, flags);
 	}
 
 	return NULL;
@@ -1150,7 +1155,7 @@ static struct page *alloc_fresh_gigantic_page_node(struct hstate *h, int nid)
 {
 	struct page *page;
 
-	page = alloc_gigantic_page(nid, huge_page_order(h));
+	page = alloc_gigantic_page(nid, h);
 	if (page) {
 		prep_compound_gigantic_page(page, huge_page_order(h));
 		prep_new_huge_page(h, page, nid);
-- 
2.13.2

-- 
Michal Hocko
SUSE Labs

Powered by blists - more mailing lists