Re: [PATCH] sched: Do not release current rq lock on non contended double_lock_balance()

From: Steven Rostedt
Date: Wed Jun 15 2016 - 12:14:01 EST


On Wed, 15 Jun 2016 13:14:53 +0200
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> OK, so should not the whole HAVE_RT_PUSH_IPI thing have avoided that
> loop entirely? And therefore made the point moot?

I believe there was another issue that we had in our tests. But I don't
have the trace available with me. I'll rerun the tests when I get back
home and have some more concrete examples for you.

>
> In any case, can't we add another cpupri for pushable tasks and use that
> to find the highest priority task to pull and avoid the loop thus?

I thought about this too, but I was a bit concerned about complexities
this would add. But I can look into it. Currently I'm in NYC for
personal reasons and will take a look at this when I get back.

-- Steve