mm/vmscan.c: avoid throttling reclaim for loop-back nfsd threads
authorNeilBrown <neilb@suse.de>
Wed, 4 Jun 2014 23:07:42 +0000 (16:07 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 4 Jun 2014 23:54:01 +0000 (16:54 -0700)
commit399ba0b95670c70aaaa3f4f1623ea9e76c391681
tree6db21b9d0b97a6015654f6806eb4101c61ea949d
parent11de9927f9dd3cb0a0f18064fa4b6976fc37e79c
mm/vmscan.c: avoid throttling reclaim for loop-back nfsd threads

When a loopback NFS mount is active and the backing device for the NFS
mount becomes congested, that can impose throttling delays on the nfsd
threads.

These delays significantly reduce throughput and so the NFS mount remains
congested.

This results in a livelock and the reduced throughput persists.

This livelock has been found in testing with the 'wait_iff_congested'
call, and could possibly be caused by the 'congestion_wait' call.

This livelock is similar to the deadlock which justified the introduction
of PF_LESS_THROTTLE, and the same flag can be used to remove this
livelock.

To minimise the impact of the change, we still throttle nfsd when the
filesystem it is writing to is congested, but not when some separate
filesystem (e.g.  the NFS filesystem) is congested.

Signed-off-by: NeilBrown <neilb@suse.de>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/vmscan.c