When migration scanner skips the rest of cc->order aligned block on isolation
failure, it avoids making migrations that cannot help the allocation at hand to
succeed, but the potential downside is not freeing pages in !MIGRATE_MOVABLE
pageblocks which could otherwise prevent allocations of same migratetype from
fallback to other pageblock types. Therefore let's restrict the skipping only
to MIGRATE_MOVABLE allocations, which in the async direct compaction mode only
scan MIGRATE_MOVABLE pageblocks.
Signed-off-by: Vlastimil Babka <vbabka@xxxxxxx>
---
mm/compaction.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/mm/compaction.c b/mm/compaction.c
index ce73badad464..95b8b5ae59c5 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -720,7 +720,8 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
if (compact_should_abort(cc))
return 0;
- if (cc->direct_compaction && (cc->mode == MIGRATE_ASYNC)) {
+ if (cc->direct_compaction && (cc->mode == MIGRATE_ASYNC) &&
+ cc->migratetype == MIGRATE_MOVABLE) {
skip_on_failure = true;
next_skip_pfn = block_end_pfn(low_pfn, cc->order);
}
--
2.15.1
--
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/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
