Re: [PATCH v2 2/4] ftrace: introduce workqueue_handler_exit tracepoint and rename workqueue_execution to workqueue_handler_entry

[Oleg Nesterov <oleg@redhat.com>]

I am very sorry for delay. I am going to study these patches later.

Just one note,

> On Mon, Apr 13, 2009 at 02:53:01PM +0900, KOSAKI Motohiro wrote:
>
> > Subject: [PATCH] ftrace: introduce workqueue_handler_exit tracepoint and rename workqueue_execution to workqueue_handler_entry
> >
> > Entry/exit handler pair is useful common tracepoint technique. it can mesure handler consumption time.
...
> >  static void run_workqueue(struct cpu_workqueue_struct *cwq)
> >  {
> > @@ -282,7 +283,6 @@ static void run_workqueue(struct cpu_wor
> >  		 */
> >  		struct lockdep_map lockdep_map = work->lockdep_map;
> >  #endif
> > -		trace_workqueue_execution(cwq->thread, work);
> >  		cwq->current_work = work;
> >  		list_del_init(cwq->worklist.next);
> >  		spin_unlock_irq(&cwq->lock);
> > @@ -291,7 +291,9 @@ static void run_workqueue(struct cpu_wor
> >  		work_clear_pending(work);
> >  		lock_map_acquire(&cwq->wq->lockdep_map);
> >  		lock_map_acquire(&lockdep_map);
> > +		trace_workqueue_handler_entry(cwq->thread, work);
> >  		f(work);
> > +		trace_workqueue_handler_exit(cwq->thread, work);

This doesn't look right. We must not use "work" after f(work). work->func()
can kfree its work.

That is why we copy lockdep_map beforehand. Perhaps ftrace should do
something similar.

Oleg.