Re: [PATCH] sched: tg_set_cfs_bandwidth() causes rq->lock deadlock

[Peter Zijlstra]

On Tue, May 20, 2014 at 06:21:15PM +0400, Roman Gushchin wrote:
> 20.05.2014, 17:15, "Peter Zijlstra" <peterz@xxxxxxxxxxxxx>:
> > Completely untested... does anybody have a 'useful' test case for this
> > bw stuff?
> 
> I can test it for stability.
> I'll start tomorrow, now I'm testing Ben's + my patches together.

Do you have an easy to reproduce setup and workload so I might be able
to test as well? Or are you simply dumping it in 'production' to see
what happens?

I mean, the trivial setup with a few cgroups and spinners isn't that
interesting because it'll just keep the timer ticking and never go idle.

I suppose I could try writing a sleep/spinner which randomly exceeds bw
constraints.

Attachment: pgpmigPkIDirX.pgp
Description: PGP signature