Re: [PATCH 03/11] tracing: Inject lock_class_init events on registration

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2010-02-05 at 09:13 -0500, Steven Rostedt wrote:
> On Wed, 2010-02-03 at 10:14 +0100, Frederic Weisbecker wrote:
> 
> >  static inline void lockdep_off(void)
> > diff --git a/include/trace/events/lock.h b/include/trace/events/lock.h
> > index 90af03c..6999f16 100644
> > --- a/include/trace/events/lock.h
> > +++ b/include/trace/events/lock.h
> > @@ -9,7 +9,7 @@
> >  
> >  #ifdef CONFIG_LOCKDEP
> >  
> > -TRACE_EVENT(lock_class_init,
> > +TRACE_EVENT_INJECT(lock_class_init,
> >  
> >  	TP_PROTO(struct lock_class *class),
> >  
> > @@ -30,7 +30,14 @@ TRACE_EVENT(lock_class_init,
> >  		__entry->class_id = (void *)class->name;
> >  	),
> >  
> > -	TP_printk("%p %s", __entry->class_id, __get_str(class_name))
> > +	TP_printk("%p %s", __entry->class_id, __get_str(class_name)),
> > +
> > +	/*
> > +	 * On activation, we want to send an event for each lock
> > +	 * classes that have been recorded by lockdep, so that we
> > +	 * catch up with the existing class:name mappings.
> > +	 */
> > +	lock_class_init_inject_events
> >  );
> >  
> 
> I also suggested this type of registering:
> 
> register_event_callback("event", "command", command_func, 
> 		command_enable_func,
> 		command_disable_func);
> 
> Where command_func, command_enable_func and command_disable_func can all
> be a function pointer or NULL.
> 
> command_func gets called every time the event is hit while enabled.
> 
> command_enable_func gets called when the event is enabled
> 
> command_disable_func gets called when the event is disabled.

Well, that will solve johill's polling thing, but not work nicely with
the proposed collection iteration stuff.