mm, compaction: extend pageblock_skip_persistent() to all compound pages
authorVlastimil Babka <vbabka@suse.cz>
Fri, 17 Nov 2017 23:26:34 +0000 (15:26 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 18 Nov 2017 00:10:00 +0000 (16:10 -0800)
commitb527cfe5bc23208cf9a346879501333cec638aba
tree3302cb8445a9ae8edd1ac187ffef8dc8f84f9781
parent21dc7e023611fbcf8e38f255731bcf3cc38e7638
mm, compaction: extend pageblock_skip_persistent() to all compound pages

pageblock_skip_persistent() checks for HugeTLB pages of pageblock order.
When clearing pageblock skip bits for compaction, the bits are not
cleared for such pageblocks, because they cannot contain base pages
suitable for migration, nor free pages to use as migration targets.

This optimization can be simply extended to all compound pages of order
equal or larger than pageblock order, because migrating such pages (if
they support it) cannot help sub-pageblock fragmentation.  This includes
THP's and also gigantic HugeTLB pages, which the current implementation
doesn't persistently skip due to a strict pageblock_order equality check
and not recognizing tail pages.

While THP pages are generally less "persistent" than HugeTLB, we can
still expect that if a THP exists at the point of
__reset_isolation_suitable(), it will exist also during the subsequent
compaction run.  The time difference here could be actually smaller than
between a compaction run that sets a (non-persistent) skip bit on a THP,
and the next compaction run that observes it.

Link: http://lkml.kernel.org/r/20171102121706.21504-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/compaction.c