Re: [RFC PATCH 6/6] perf: Disable IRQs in address filter sync path

[Mathieu Poirier]

On 26 September 2016 at 10:18, Alexander Shishkin
<alexander.shishkin@xxxxxxxxxxxxxxx> wrote:
> Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx> writes:
>
>> If PMU callbacks are executed in hardirq context, the address filter
>> sync code needs to disable interrupts when taking its spinlock to be
>> consistent with the rest of its users. This may happen if the PMU is
>> used in AUX sampling.
>
> Hi Mathieu,
>
> I've been meaning to CC you on this series and forgot. My concern was
> that on PMUs that run PMIs in hardirq context this patch should be
> required already now. Is this the case for you?

At this time cross triggers haven't been implemented and as such, PMIs
aren't an issue.  On the flip side I can see the value of your code
when we get around to do the implementation.

Mathieu

>
>>
>> Signed-off-by: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
>> ---
>>  kernel/events/core.c | 5 +++--
>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/events/core.c b/kernel/events/core.c
>> index f6582df1c9..047c495c94 100644
>> --- a/kernel/events/core.c
>> +++ b/kernel/events/core.c
>> @@ -2568,16 +2568,17 @@ static int perf_event_stop(struct perf_event *event, int restart)
>>  void perf_event_addr_filters_sync(struct perf_event *event)
>>  {
>>       struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
>> +     unsigned long flags;
>>
>>       if (!has_addr_filter(event))
>>               return;
>>
>> -     raw_spin_lock(&ifh->lock);
>> +     raw_spin_lock_irqsave(&ifh->lock, flags);
>>       if (event->addr_filters_gen != event->hw.addr_filters_gen) {
>>               event->pmu->addr_filters_sync(event);
>>               event->hw.addr_filters_gen = event->addr_filters_gen;
>>       }
>> -     raw_spin_unlock(&ifh->lock);
>> +     raw_spin_unlock_irqrestore(&ifh->lock, flags);
>>  }
>>  EXPORT_SYMBOL_GPL(perf_event_addr_filters_sync);
>>
>> --
>> 2.9.3