On Tue, Mar 01, 2011 at 09:57:59PM -0800, Andrew Morton wrote: > On Wed, 2 Mar 2011 06:52:21 +0100 Andrea Arcangeli <aarcange@xxxxxxxxxx> wrote: > > > These are the other two patches that are needed for both workloads to > > be better than before. > > > > mm-compaction-minimise-the-time-irqs-are-disabled-while-isolating-pages-for-migration > > mm-compaction-minimise-the-time-irqs-are-disabled-while-isolating-free-pages > > I have those queued for 2.6.39 - they didn't seem terribly critical and > no mention of NMI watchdog timeouts was made. > > Guys, this stuff matters :( Should both go into 2.6.38? If so, why? Ok: before commit 5a03b051ed87e72b959f32a86054e1142ac4cf55 it was unnoticeable problem (the above two patches are fixing longstanding bugs). But the combination of kswapd running compaction in a loop after commit 5a03b051ed87e72b959f32a86054e1142ac4cf55, and compaction keeping irqs disabled for too long (longstanding bug, but unnoticed before 5a03b051ed87e72b959f32a86054e1142ac4cf55), shows both problems. If you have a single problem you don't notice so much the other one. But kswapd calling compaction in a loop, and compaction keeping irqs disabled for too long are separate problems that shows each other. I think we want the above two patches and the patch Mel sent with Message-ID: <20110302142542.GE14162@xxxxxxxxx> in 2.6.38. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx 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>