Re: [PATCH 2/2] perf: Disable the event on a truncated AUX record

From: Alexander Shishkin
Date: Wed May 11 2016 - 05:41:34 EST


Peter Zijlstra <peterz@xxxxxxxxxxxxx> writes:

> On Tue, May 10, 2016 at 04:18:33PM +0300, Alexander Shishkin wrote:
>> When the PMU driver reports a truncated AUX record, it effectively means
>> that there is no more usable room in the event's AUX buffer (even though
>> there may still be some room, so that perf_aux_output_begin() doesn't take
>> action). At this point the consumer still has to be woken up and the event
>> has to be disabled, otherwise the event will just keep spinning between
>> perf_aux_output_begin() and perf_aux_output_end() until its context gets
>> unscheduled.
>>
>> Again, for cpu-wide events this means never, so once in this condition,
>> they will be forever losing data.
>>
>> Fix this by disabling the event and waking up the consumer in case of a
>> truncated AUX record.
>
>> + if (wakeup) {
>> + if (truncated)
>> + handle->event->pending_disable = 1;
>> + perf_output_wakeup(handle);
>> + }
>
> Does the userspace tool know how to deal with this and re-enable it?

Yes, it knows to ioctl(EVENT_ENABLE). Also, we're already doing the
disabling for the no-room-in-buffer condition in
perf_aux_output_begin().

Regards,
--
Alex