Re: [PATCH 2/2] tracing/workqueue: Rename workqueue_execute to worklet_entry and add worklet_exit

[KOSAKI Motohiro]

Hi

> In Kosaki's patch, we had the work struct address displayed too.
> Your version is supposed to be sufficient because we know a workqueue
> serializes its works. Then we know that an exit event will always follow
> and match the previous entry event from the same workqueue thread.
> The workqueue pid then provides a sufficient key for that.
> 
> That said, we should worry about possible lost events from
> perf in some circumstances. And userspace profiling needs something
> to ensure the accuracy about this entry/exit pair.
> 
> We could have:
> 	entry work1
> 	exit work 1 <--- lost event
> 	entry work2 <--- lost event
> 	exit work2
> 
> And then the pair would be misinterpreted.
> 
> (Although we could have even other misinterpretation
> with other kind of scenarios, even if we have this work
> address. But that's still more safety).
> 
> So I'd prefer to keep Kosaki's idea about these keys.
> But I prefer your event naming.
> 
> May be I can unearth Kosaki's patch, change it with your naming
> and add your Signed-off-by?
> 
> Kosaki, no problem about it?

Sure.
Thank you paid attention my patch :-)





--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/