aboutsummaryrefslogtreecommitdiffstats
path: root/mm/page_alloc.c
diff options
context:
space:
mode:
authorMel Gorman <mel@csn.ul.ie>2009-07-29 18:02:04 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-07-29 22:10:34 -0400
commite084b2d95e48b31aa45f9c49ffc6cdae8bdb21d4 (patch)
treeff15d36a3a1e49fdbd5080decb7ab00afdd60099 /mm/page_alloc.c
parent51fbb4bab6c8710eb897ab3fb06efbbc921f3a8d (diff)
page-allocator: preserve PFN ordering when __GFP_COLD is set
Fix a post-2.6.24 performace regression caused by 3dfa5721f12c3d5a441448086bee156887daa961 ("page-allocator: preserve PFN ordering when __GFP_COLD is set"). Narayanan reports "The regression is around 15%. There is no disk controller as our setup is based on Samsung OneNAND used as a memory mapped device on a OMAP2430 based board." The page allocator tries to preserve contiguous PFN ordering when returning pages such that repeated callers to the allocator have a strong chance of getting physically contiguous pages, particularly when external fragmentation is low. However, of the bulk of the allocations have __GFP_COLD set as they are due to aio_read() for example, then the PFNs are in reverse PFN order. This can cause performance degration when used with IO controllers that could have merged the requests. This patch attempts to preserve the contiguous ordering of PFNs for users of __GFP_COLD. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Reported-by: Narayananu Gopalakrishnan <narayanan.g@samsung.com> Tested-by: Narayanan Gopalakrishnan <narayanan.g@samsung.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'mm/page_alloc.c')
-rw-r--r--mm/page_alloc.c13
1 files changed, 9 insertions, 4 deletions
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index caa92689aac9..ae28c22a7fdb 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -882,7 +882,7 @@ retry_reserve:
882 */ 882 */
883static int rmqueue_bulk(struct zone *zone, unsigned int order, 883static int rmqueue_bulk(struct zone *zone, unsigned int order,
884 unsigned long count, struct list_head *list, 884 unsigned long count, struct list_head *list,
885 int migratetype) 885 int migratetype, int cold)
886{ 886{
887 int i; 887 int i;
888 888
@@ -901,7 +901,10 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
901 * merge IO requests if the physical pages are ordered 901 * merge IO requests if the physical pages are ordered
902 * properly. 902 * properly.
903 */ 903 */
904 list_add(&page->lru, list); 904 if (likely(cold == 0))
905 list_add(&page->lru, list);
906 else
907 list_add_tail(&page->lru, list);
905 set_page_private(page, migratetype); 908 set_page_private(page, migratetype);
906 list = &page->lru; 909 list = &page->lru;
907 } 910 }
@@ -1119,7 +1122,8 @@ again:
1119 local_irq_save(flags); 1122 local_irq_save(flags);
1120 if (!pcp->count) { 1123 if (!pcp->count) {
1121 pcp->count = rmqueue_bulk(zone, 0, 1124 pcp->count = rmqueue_bulk(zone, 0,
1122 pcp->batch, &pcp->list, migratetype); 1125 pcp->batch, &pcp->list,
1126 migratetype, cold);
1123 if (unlikely(!pcp->count)) 1127 if (unlikely(!pcp->count))
1124 goto failed; 1128 goto failed;
1125 } 1129 }
@@ -1138,7 +1142,8 @@ again:
1138 /* Allocate more to the pcp list if necessary */ 1142 /* Allocate more to the pcp list if necessary */
1139 if (unlikely(&page->lru == &pcp->list)) { 1143 if (unlikely(&page->lru == &pcp->list)) {
1140 pcp->count += rmqueue_bulk(zone, 0, 1144 pcp->count += rmqueue_bulk(zone, 0,
1141 pcp->batch, &pcp->list, migratetype); 1145 pcp->batch, &pcp->list,
1146 migratetype, cold);
1142 page = list_entry(pcp->list.next, struct page, lru); 1147 page = list_entry(pcp->list.next, struct page, lru);
1143 } 1148 }
1144 1149