[PATCH 5/9] xfs: reduce the number of AIL push wakeups

[Christoph Hellwig]

On Fri, Dec 17, 2010 at 08:50:50AM +1100, Dave Chinner wrote:
> What tests?

I couldn't really see a pattern yet - it's different ones and different
amount of warnings during each run.

> So this is saying is that a 20ms uninterruptible sleep lasting for more
> than 120s? Doesn't that imply some kind of scheduler starvation, not
> an actual XFS problem?

It does for sure look odd to me.  I'll do a sub-commit bisect between
the two hunk in the patch tomorrow to drill a bit deeper into what's
going on.