Re: Question about prio_changed_dl()

[luca abeni <luca.abeni@unitn.it>]

On Thu, 25 Feb 2016 15:52:02 +0100
Peter Zijlstra <peterz@infradead.org> wrote:

> On Thu, Feb 25, 2016 at 03:25:58PM +0100, luca abeni wrote:
> 
> > > > (BTW, it seems to me that switched_to_dl() is never invoked, for
> > > > some reason...)
> > > 
> > > Hmm, it should be invoked if you do sched_setattr() to get
> > > SCHED_DEADLINE.
> > 
> > Sorry, that was me being confused...
> > It is prio_changed_dl() that is not invoked when the deadline
> > parameters are changed (I am testing a change to fix this - it actually
> > is included in the "Move the remaining __dl_{sub,add}_ac() calls from
> > core.c to deadline.c" patch I posted on Monday; I am going to extract
> > it in a separate patch).
> 
> Ah ok. So the idea was that the || dl_task() part would ensure
> ->prio_changed() would always be called.
> 
> I'll await your patch.

>From 5a0bde5332bc1cc5701b2d9799f6ec197c1572cc Mon Sep 17 00:00:00 2001
From: Luca Abeni <luca.abeni@unitn.it>
Date: Fri, 26 Feb 2016 09:56:24 +0100
Subject: sched/core: fix __sched_setscheduler() to properly invoke prio_changed_dl()

Currently, prio_changed_dl() is not called when __sched_setscheduler()
changes the parameters of a SCHED_DEADLINE task. This happens because
when changing parameters deadline tasks do not change their priority,
so new_effective_prio == oldprio.
The issue is fixed by explicitly checking if the task is a deadline task.

Signed-off-by: Luca Abeni <luca.abeni@unitn.it>
---
 kernel/sched/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 9503d59..5646bde 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4079,6 +4079,8 @@ change:
 		new_effective_prio = rt_mutex_get_effective_prio(p, newprio);
 		if (new_effective_prio == oldprio) {
 			__setscheduler_params(p, attr);
+			if (p->sched_class == &dl_sched_class)
+				p->sched_class->prio_changed(rq, p, oldprio);
 			task_rq_unlock(rq, p, &flags);
 			return 0;
 		}
-- 
2.5.0