[PATCH] perf/smmuv3: Don't trample existing events with global filter

Subsystems: arm pmu profiling and debugging, the rest

STALE1879d LANDED

Landed in mainline as 4c1daba15c20 on 2021-06-11.

2 messages, 2 authors, 2021-06-11 · open the first message on its own page

[PATCH] perf/smmuv3: Don't trample existing events with global filter

From: Robin Murphy <robin.murphy@arm.com>
Date: 2021-06-08 11:55:24

With global filtering, we only allow an event to be scheduled if its
filter settings exactly match those of any existing events, therefore
it is pointless to reapply the filter in that case. Much worse, though,
is that in doing that we trample the event type of counter 0 if it's
already active, and never touch the appropriate PMEVTYPERn so the new
event is likely not counting the right thing either. Don't do that.

CC: stable@vger.kernel.org
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---
 drivers/perf/arm_smmuv3_pmu.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
index ff6fab4bae30..863d9f702aa1 100644
--- a/drivers/perf/arm_smmuv3_pmu.c
+++ b/drivers/perf/arm_smmuv3_pmu.c
@@ -277,7 +277,7 @@ static int smmu_pmu_apply_event_filter(struct smmu_pmu *smmu_pmu,
 				       struct perf_event *event, int idx)
 {
 	u32 span, sid;
-	unsigned int num_ctrs = smmu_pmu->num_counters;
+	unsigned int cur_idx, num_ctrs = smmu_pmu->num_counters;
 	bool filter_en = !!get_filter_enable(event);
 
 	span = filter_en ? get_filter_span(event) :
@@ -285,17 +285,19 @@ static int smmu_pmu_apply_event_filter(struct smmu_pmu *smmu_pmu,
 	sid = filter_en ? get_filter_stream_id(event) :
 			   SMMU_PMCG_DEFAULT_FILTER_SID;
 
-	/* Support individual filter settings */
-	if (!smmu_pmu->global_filter) {
+	cur_idx = find_first_bit(smmu_pmu->used_counters, num_ctrs);
+	/*
+	 * Per-counter filtering, or scheduling the first globally-filtered
+	 * event into an empty PMU so idx == 0 and it works out equivalent.
+	 */
+	if (!smmu_pmu->global_filter || cur_idx == num_ctrs) {
 		smmu_pmu_set_event_filter(event, idx, span, sid);
 		return 0;
 	}
 
-	/* Requested settings same as current global settings*/
-	idx = find_first_bit(smmu_pmu->used_counters, num_ctrs);
-	if (idx == num_ctrs ||
-	    smmu_pmu_check_global_filter(smmu_pmu->events[idx], event)) {
-		smmu_pmu_set_event_filter(event, 0, span, sid);
+	/* Otherwise, must match whatever's currently scheduled */
+	if (smmu_pmu_check_global_filter(smmu_pmu->events[cur_idx], event)) {
+		smmu_pmu_set_evtyper(smmu_pmu, idx, get_event(event));
 		return 0;
 	}
 
-- 
2.25.1


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] perf/smmuv3: Don't trample existing events with global filter

From: Will Deacon <will@kernel.org>
Date: 2021-06-11 16:15:53

On Tue, 8 Jun 2021 12:55:12 +0100, Robin Murphy wrote:
With global filtering, we only allow an event to be scheduled if its
filter settings exactly match those of any existing events, therefore
it is pointless to reapply the filter in that case. Much worse, though,
is that in doing that we trample the event type of counter 0 if it's
already active, and never touch the appropriate PMEVTYPERn so the new
event is likely not counting the right thing either. Don't do that.
Applied to will (for-next/perf), thanks!

[1/1] perf/smmuv3: Don't trample existing events with global filter
      https://git.kernel.org/will/c/4c1daba15c20

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help