Re: [PATCH] sched: Handle priority boosted tasks proper in setscheduler()

[Thomas Gleixner]

On Tue, 5 May 2015, Steven Rostedt wrote:
> On Tue, 5 May 2015 18:31:20 +0200 (CEST)
> Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> 
> > On Tue, 5 May 2015, Steven Rostedt wrote:
> > > I got pulled onto other things so I never finished it, but one thing
> > > that worried me about this fix is this:
> > > 
> > > 	T1 - FIFO policy (prio = 10)
> > > 	  lock(rtmutex);
> > > 
> > > 	T2 (prio = 20)
> > > 	  lock(rtmutex)
> > > 	    boost T1 (prio = 20)
> > > 
> > >         TI (prio = 20)
> > > 	  sys_sched_setscheduler(prio = 30)
> > > 	  TI (prio = 30)
> > > 
> > > 	T1 (prio = 30)
> > > 	  sys_sched_setscheduler(SCHED_OTHER)
> > > 	  new_effective_prio = 20, oldprio = 30
> > > 
> > > Before the code stopped at the rt_mutex_check_prio(), but now it
> > > continues. Will having the policy change cause problems here?
> > 
> > No, because it stays effective in the FIFO domain.
> > 
> 
> Ah, the policy passed in isn't used, so we are safe. But, but I also
> found that we still call __setscheduler(), which does:
> 
> 	p->prio = normal_prio();
> 
> Isn't that going to null out the boosting?

Crap. Yes, I missed that. So __setscheduler() assumes that there is no
boosting going on. So we need:

   p->prio = effective_prio(p);

there instead.

Thanks,

	tglx


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/