On Tue, Sep 04, 2012 at 12:01:19PM -0700, Tejun Heo wrote:

On Tue, Sep 04, 2012 at 11:55:40AM -0700, Tejun Heo wrote:

quoted

Actually, if the timer approach can reduce the frequency of rescuer
involvement, I think it could actually be better.

Ooh, it wouldn't.  It's kicking in only after alloc failure.  I don't
know.  I think conditioning it on alloc failure is cleaner and
converting all per-bio allocations to front-pad makes sense.  Using a
timer wouldn't make the mechanism any simpler, right?

Exactly