[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160615175512.GV30921@twins.programming.kicks-ass.net>
Date: Wed, 15 Jun 2016 19:55:12 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: mingo@...nel.org, tglx@...utronix.de, juri.lelli@....com,
xlpang@...hat.com, linux-kernel@...r.kernel.org,
mathieu.desnoyers@...icios.com, jdesfossez@...icios.com,
bristot@...hat.com, Ingo Molnar <mingo@...hat.com>
Subject: Re: [RFC][PATCH 3/8] sched/deadline/rtmutex: Dont miss the
dl_runtime/dl_period update
On Wed, Jun 15, 2016 at 12:30:07PM -0400, Steven Rostedt wrote:
> > +++ b/kernel/locking/rtmutex.c
> > @@ -488,7 +488,7 @@ static int rt_mutex_adjust_prio_chain(st
> > * enabled we continue, but stop the requeueing in the chain
> > * walk.
> > */
> > - if (waiter->prio == task->prio) {
> > + if (waiter->prio == task->prio && !dl_task(task)) {
>
> Isn't task the owner of the lock?
No, task is blocked on something.
> What happens if the waiter is a
> deadline task?
So the test here is a shortcut to terminate the Pi chain adjust, it says
that if the waiter and task have the same priority, we're done. Further
adjustments will not make a difference.
The problem is that for deadline tasks, prio is a useless number, so
even if they match (all deadline tasks have prio -1) they might still
not actually match.
After the last patch I suppose we could do something like:
waiter->prio == task->prio && waiter->deadline == task->dl.deadline
In any case, any condition that compares just two 'prio' values is per
definition broken if DL tasks are involved.
Powered by blists - more mailing lists