Re: sched_rr_get_interval NULL pointer OOPS

On Fri, Jan 24, 2014 at 10:55:56PM +0200, Tommi Rantala wrote:
> Hello,
> 
> Trinity triggered the following bug in two separate qemu virtual
> machines after fuzzing v3.13-3995-g0dc3fd0 for a day or two. I have
> not been running Trinity in a while, so no idea if this is a
> regression or not.
> 
> If I'm reading this right, it's oopsing in kernel/sched/core.c:
> 
> SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
>     struct timespec __user *, interval)
> {
> ...
>     rq = task_rq_lock(p, &flags);
>     time_slice = p->sched_class->get_rr_interval(rq, p);   <==
>     task_rq_unlock(rq, p, &flags);
> ...


---
Subject: sched: make sched_class::get_rr_interval optional

Not all classes implement (or can implement) a useful get_rr_interval()
function, default to a 0 time-slice for them.

Reported-by: Tommi Rantala <[email protected]>
Signed-off-by: Peter Zijlstra <[email protected]>
---
Index: linux-2.6/kernel/sched/core.c
===================================================================
--- linux-2.6.orig/kernel/sched/core.c
+++ linux-2.6/kernel/sched/core.c
@@ -4324,7 +4324,9 @@ SYSCALL_DEFINE2(sched_rr_get_interval, p
 		goto out_unlock;
 
 	rq = task_rq_lock(p, &flags);
-	time_slice = p->sched_class->get_rr_interval(rq, p);
+	time_slice = 0;
+	if (p->sched_class->get_rr_interval)
+		time_slice = p->sched_class->get_rr_interval(rq, p);
 	task_rq_unlock(rq, p, &flags);
 
 	rcu_read_unlock();