Re: [PATCH v12 3/3] tracing: Centralize preemptirq tracepoints and unify their usage

[Joel Fernandes]

On Mon, Aug 6, 2018 at 6:43 PM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> On Mon, 6 Aug 2018 17:43:19 -0700
> Joel Fernandes <joelaf@xxxxxxxxxx> wrote:
>
>> On Mon, Aug 6, 2018 at 12:50 PM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>> >
>> > With this patch applied, I'm constantly getting lockdep errors. Instead
>> > of doing a full revert of the patch, I did this, which makes all those
>> > errors go away. I may apply this for now, and we can revisit having
>> > lockdep use the tracepoint code. But since it's currently always
>> > enabled, I'm thinking of just leaving this as is. The macros are still
>> > clean from Joel's patch.
>> >
>> > Thoughts?
>>
>> I like your patch. Thanks a lot for doing this.. It keeps most of the
>> benefits of my patch while avoiding the issues with lockdep. I agree
>> we can look at the remaining lockdep issue after. There were several
>> lockdep issues with this patch that I fixed over the the months, but
>> there's still the one that Masami reported that I believe you're also
>> seeing. Once I'm back I'll work on figuring that one out.
>>
>> Could you pull in the fixes to the other issues I posted though? With
>> that we should be good.
>> https://lore.kernel.org/patchwork/patch/971104/
>> https://lore.kernel.org/patchwork/patch/971829/
>>
>
> I already had these applied when I created this patch ;-)
>
> Thanks, I'll add it.

Thanks, also one more thing I noticed in your patch,
lockdep_hardirqs_off needs to be called before all other probes but
you're calling it after. This is why I registered it with INT_MAX:

register_trace_prio_irq_disable(lockdep_hardirqs_off, NULL, INT_MAX);

Without it you may get annotation warnings. Thanks,

- Joel