mm, compaction: use correct watermark when checking compaction success
authorVlastimil Babka <vbabka@suse.cz>
Fri, 7 Oct 2016 23:57:50 +0000 (16:57 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 8 Oct 2016 01:46:27 +0000 (18:46 -0700)
The __compact_finished() function uses low watermark in a check that has
to pass if the direct compaction is to finish and allocation should
succeed.  This is too pessimistic, as the allocation will typically use
min watermark.  It may happen that during compaction, we drop below the
low watermark (due to parallel activity), but still form the target
high-order page.  By checking against low watermark, we might needlessly
continue compaction.

Similarly, __compaction_suitable() uses low watermark in a check whether
allocation can succeed without compaction.  Again, this is unnecessarily
pessimistic.

After this patch, these check will use direct compactor's alloc_flags to
determine the watermark, which is effectively the min watermark.

Link: http://lkml.kernel.org/r/20160810091226.6709-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Lorenzo Stoakes <lstoakes@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/compaction.c

index 21040304f4d20238a0c2595a9f6b470cb3dbc7ed..e2618ac062a660f97390f0fd43aa7b2a09754bfa 100644 (file)
@@ -1316,7 +1316,7 @@ static enum compact_result __compact_finished(struct zone *zone, struct compact_
                return COMPACT_CONTINUE;
 
        /* Compaction run is not finished if the watermark is not met */
-       watermark = low_wmark_pages(zone);
+       watermark = zone->watermark[cc->alloc_flags & ALLOC_WMARK_MASK];
 
        if (!zone_watermark_ok(zone, cc->order, watermark, cc->classzone_idx,
                                                        cc->alloc_flags))
@@ -1381,7 +1381,7 @@ static enum compact_result __compaction_suitable(struct zone *zone, int order,
        if (is_via_compact_memory(order))
                return COMPACT_CONTINUE;
 
-       watermark = low_wmark_pages(zone);
+       watermark = zone->watermark[alloc_flags & ALLOC_WMARK_MASK];
        /*
         * If watermarks for high-order allocation are already met, there
         * should be no need for compaction at all.
@@ -1395,7 +1395,7 @@ static enum compact_result __compaction_suitable(struct zone *zone, int order,
         * This is because during migration, copies of pages need to be
         * allocated and for a short time, the footprint is higher
         */
-       watermark += (2UL << order);
+       watermark = low_wmark_pages(zone) + (2UL << order);
        if (!__zone_watermark_ok(zone, 0, watermark, classzone_idx,
                                 alloc_flags, wmark_target))
                return COMPACT_SKIPPED;