Re: [PATCH v3 21/33] tracing: Add support for 'synthetic' events

[Namhyung Kim]

On Wed, Oct 04, 2017 at 02:50:18PM -0500, Tom Zanussi wrote:
> On Wed, 2017-10-04 at 14:08 -0400, Steven Rostedt wrote:
> > On Fri, 22 Sep 2017 15:00:01 -0500
> > Tom Zanussi <tom.zanussi@xxxxxxxxxxxxxxx> wrote:
> > 
> > 
> > 
> > > +/* This function releases synth_event_mutex */
> > > +static int unregister_synth_event(struct synth_event *event)
> > > +{
> > > +	struct trace_event_call *call = &event->call;
> > > +	int ret;
> > > +
> > > +	mutex_unlock(&synth_event_mutex);
> > > +	ret = trace_remove_event_call(call);
> > > +	mutex_lock(&synth_event_mutex);
> > > +
> > > +	return ret;
> > > +}
> > > +
> > > +static int remove_synth_event(struct synth_event *event)
> > > +{
> > > +	int ret = unregister_synth_event(event);
> > > +
> > > +	if (!ret)
> > > +		list_del(&event->list);
> > > +
> > > +	return ret;
> > > +}
> > > +
> > 
> > 
> > > +/* This function releases synth_event_mutex */
> > > +static int release_all_synth_events(void)
> > > +{
> > > +	struct synth_event *event, *e;
> > > +	int ret = 0;
> > > +
> > > +	mutex_lock(&synth_event_mutex);
> > > +
> > > +	list_for_each_entry(event, &synth_event_list, list) {
> > > +		if (event->ref) {
> > > +			ret = -EBUSY;
> > > +			goto out;
> > > +		}
> > > +	}
> > > +
> > > +	list_for_each_entry_safe(event, e, &synth_event_list, list) {
> > 
> > remove_synth_event() releases synth_event_mutex, which is racy, as more
> > than one instance can do the deletion.
> > 
> > Perhaps we should remove all the events off the synth_event_list under
> > the lock, release the lock, and then remove the trace events attached
> > to them?
> > 
> 
> Yeah, I think that makes sense, will change..

I suspect a similar race in creation too.  The create_synth_event()
checks existing events in the synth_event_list but it releases the
synth_event_mutex before adding a new event to the list.  So it's
possible to add two (or more) events have same name IMHO.

Thanks,
Namhyung