Re: Null pointer crash at find_idlest_group on db845c w/ linus/master

[Vincent Guittot]

Le Wednesday 04 Dec 2019 à 09:42:17 (+0000), Qais Yousef a écrit :
> On 12/04/19 09:06, Vincent Guittot wrote:
> > Hi John,
> > 
> > On Tue, 3 Dec 2019 at 20:15, John Stultz <john.stultz@xxxxxxxxxx> wrote:
> > >
> > > With today's linus/master on db845c running android, I'm able to
> > > fairly easily reproduce the following crash. I've not had a chance to
> > > bisect it yet, but I'm suspecting its connected to Vincent's recent
> > > rework.
> > 
> > Does the crash happen randomly or after a specific action ?
> > I have a db845 so I can try to reproduce it locally.
> 
> Isn't there a chance we use local_sgs without initializing it in that function?

Normally not because the cpu belongs to its sched_domain

Now, we test that a group has at least one allowed CPU for the task so we
could skip the local group with the correct/wrong p->cpus_ptr

The path is used for fork/exec ibut also for wakeup path for b.L when the task doesn't fit in the CPUs

So we can probably imagine a scenario where we change task affinity while
sleeping. If the wakeup happens on a CPU that belongs to the group that is not
allowed, we can imagine that we skip the local_group

John,

Could you try the fix below ?

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 08a233e..bcd216d 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8417,6 +8417,10 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p,
        if (!idlest)
                return NULL;
 
+       /* The local group has been skipped because of cpu affinity */
+       if (!local)
+               return idlest;
+
        /*
         * If the local group is idler than the selected idlest group
         * don't try and push the task.


>
> AFAICS we define local_sgs on the stack but not always could be populated with
> the right values. I can't see tmp_sgs being used in the function too. Could
> this cause the/a problem?
> 
>  8377         struct sg_lb_stats local_sgs, tmp_sgs;
> .
> .
> .
>  8399                 if (local_group) {
>  8400                         sgs = &local_sgs;
>  8401                         local = group;
>  8402                 } else {
>  8403                         sgs = &tmp_sgs;
>  8404                 }
>  8405
>  8406                 update_sg_wakeup_stats(sd, group, sgs, p);
> 
> Cheers
> 
> --
> Qais Youef