We get cc->order is -1 when user echos to /proc/sys/vm/compact_memory.
In this case, we should check that if we have enough pages for
the compaction in the zone.
If we do not check this, in our MX6Q board(arm), i ever observed
COMPACT_CLUSTER_MAX pages were compaction failed in per migrate_pages().
That's mean we can not alloc any pages by the free scanner in the zone.
This patch checks the watermark to avoid this problem.
Signed-off-by: Huang Shijie <b32955@xxxxxxxxxxxxx>
---
mm/compaction.c | 14 ++++++++++++--
1 files changed, 12 insertions(+), 2 deletions(-)
diff --git a/mm/compaction.c b/mm/compaction.c
index 899d956..0f12cc9 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -442,8 +442,13 @@ static int compact_finished(struct zone *zone,
* order == -1 is expected when compacting via
* /proc/sys/vm/compact_memory
*/
- if (cc->order == -1)
+ if (cc->order == -1) {
+ /* Check if we have enough pages now. */
+ watermark = low_wmark_pages(zone) + COMPACT_CLUSTER_MAX * 2;
+ if (!zone_watermark_ok(zone, 0, watermark, 0, 0))
+ return COMPACT_SKIPPED;
return COMPACT_CONTINUE;
+ }
/* Compaction run is not finished if the watermark is not met */
watermark = low_wmark_pages(zone);
@@ -482,8 +487,13 @@ unsigned long compaction_suitable(struct zone *zone, int order)
* order == -1 is expected when compacting via
* /proc/sys/vm/compact_memory
*/
- if (order == -1)
+ if (order == -1) {
+ /* Check if we have enough pages now. */
+ watermark = low_wmark_pages(zone) + COMPACT_CLUSTER_MAX * 2;
+ if (!zone_watermark_ok(zone, 0, watermark, 0, 0))
+ return COMPACT_SKIPPED;
return COMPACT_CONTINUE;
+ }
/*
* Watermarks for order-0 must be met for compaction. Note the 2UL.
--
1.7.3.2
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>