Re: select_task_rq_fair: WARNING: at kernel/lockdep.c match_held_lock

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Thu, 2011-08-04 at 18:53 +0300, Sergey Senozhatsky wrote:
> On (08/04/11 17:47), Peter Zijlstra wrote:
> > On Thu, 2011-08-04 at 18:37 +0300, Sergey Senozhatsky wrote:
> > > > > [  132.794685] WARNING: at kernel/lockdep.c:3117 match_held_lock+0xf6/0x12e()
> > 
> > Just to double check, that line is:
> > 
> >                 if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
> > 
> > in your kernel source?
> > 
> 
> Nope, that's `if (DEBUG_LOCKS_WARN_ON(!class))'
> 
> 3106 static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
> 3107 {                                                                                                                                                                                                                             
> 3108     if (hlock->instance == lock)
> 3109         return 1;
> 3110 
> 3111     if (hlock->references) {
> 3112         struct lock_class *class = lock->class_cache[0];
> 3113 
> 3114         if (!class)
> 3115             class = look_up_lock_class(lock, 0);
> 3116 
> 3117         if (DEBUG_LOCKS_WARN_ON(!class))
> 3118             return 0;
> 3119 
> 3120         if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
> 3121             return 0;
> 3122 
> 3123         if (hlock->class_idx == class - lock_classes + 1)
> 3124             return 1;
> 3125     }
> 3126 
> 3127     return 0;
> 3128 }
> 3129 

Ah, in that case my previous analysis was pointless and I shall need to
scratch my head some more.