On Tue, Jun 21, 2016 at 09:37:09PM +0200, Peter Zijlstra wrote:

Hurm.. So I've applied it, just to get this issue sorted, but I'm not
entirely sure I like it.

I think I prefer ego's version because that makes it harder to get stuff
to run on !active,online cpus. I think we really want to be careful what
gets to run during that state.

The original patch just did set_cpus_allowed one more time late enough
so that the target kthread (in most cases) doesn't have to go through
fallback rq selection afterwards.  I don't know what the long term
solution is but CPU_ONLINE callbacks should be able to bind kthreads
to the new CPU one way or the other.

Thanks.

-- 
tejun