On Tue, Jun 09, 2015 at 05:49:27PM +0530, Vineet Gupta wrote:
> @@ -99,8 +99,7 @@ static void arc_perf_event_update(struct perf_event *event,
> } while (local64_cmpxchg(&hwc->prev_count, prev_raw_count,
> new_raw_count) != prev_raw_count);
>
> - delta = (new_raw_count - prev_raw_count) &
> - ((1ULL << arc_pmu->counter_size) - 1ULL);
> + delta = (new_raw_count - prev_raw_count) & arc_pmu->max_period;
I don't know how your PMU works, but you seem to be assuming new_raw >
prev_raw, which implies its counting up.
Now, typically these things trigger when they reach 0, so when we're
counting up that would mean your values are negative.
In that case, where's the sign extension?
> @@ -182,6 +181,13 @@ static int arc_pmu_event_init(struct perf_event *event)
> struct hw_perf_event *hwc = &event->hw;
> int ret;
>
> + if (!is_sampling_event(event)) {
> + hwc->sample_period = arc_pmu->max_period;
> + hwc->last_period = hwc->sample_period;
> + local64_set(&hwc->period_left, hwc->sample_period);
> + } else
> + return -ENOENT;
-ENOENT is wrong for is_sampling_event().
Either the event is for this PMU, in which case you should return a
fatal error, or its not (-ENOENT is correct in that case).
--
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
