OSDN Git Service

perf/x86/uncore: Fix event group support
authorKan Liang <kan.liang@linux.intel.com>
Fri, 25 Oct 2019 14:43:13 +0000 (07:43 -0700)
committerIngo Molnar <mingo@kernel.org>
Mon, 28 Oct 2019 10:02:01 +0000 (11:02 +0100)
The events in the same group don't start or stop simultaneously.
Here is the ftrace when enabling event group for uncore_iio_0:

  # perf stat -e "{uncore_iio_0/event=0x1/,uncore_iio_0/event=0xe/}"

            <idle>-0     [000] d.h.  8959.064832: read_msr: a41, value
  b2b0b030 //Read counter reg of IIO unit0 counter0
            <idle>-0     [000] d.h.  8959.064835: write_msr: a48, value
  400001 //Write Ctrl reg of IIO unit0 counter0 to enable
  counter0. <------ Although counter0 is enabled, Unit Ctrl is still
  freezed. Nothing will count. We are still good here.
            <idle>-0     [000] d.h.  8959.064836: read_msr: a40, value
  30100                   //Read Unit Ctrl reg of IIO unit0
            <idle>-0     [000] d.h.  8959.064838: write_msr: a40, value
  30000 //Write Unit Ctrl reg of IIO unit0 to enable all
  counters in the unit by clear Freeze bit  <------Unit0 is un-freezed.
  Counter0 has been enabled. Now it starts counting. But counter1 has not
  been enabled yet. The issue starts here.
            <idle>-0     [000] d.h.  8959.064846: read_msr: a42, value 0
//Read counter reg of IIO unit0 counter1
            <idle>-0     [000] d.h.  8959.064847: write_msr: a49, value
  40000e //Write Ctrl reg of IIO unit0 counter1 to enable
  counter1.   <------ Now, counter1 just starts to count. Counter0 has
  been running for a while.

Current code un-freezes the Unit Ctrl right after the first counter is
enabled. The subsequent group events always loses some counter values.

Implement pmu_enable and pmu_disable support for uncore, which can help
to batch hardware accesses.

No one uses uncore_enable_box and uncore_disable_box. Remove them.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: linux-drivers-review@eclists.intel.com
Cc: linux-perf@eclists.intel.com
Fixes: 087bfbb03269 ("perf/x86: Add generic Intel uncore PMU support")
Link: https://lkml.kernel.org/r/1572014593-31591-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/events/intel/uncore.c
arch/x86/events/intel/uncore.h

index 6fc2e06..86467f8 100644 (file)
@@ -502,10 +502,8 @@ void uncore_pmu_event_start(struct perf_event *event, int flags)
        local64_set(&event->hw.prev_count, uncore_read_counter(box, event));
        uncore_enable_event(box, event);
 
-       if (box->n_active == 1) {
-               uncore_enable_box(box);
+       if (box->n_active == 1)
                uncore_pmu_start_hrtimer(box);
-       }
 }
 
 void uncore_pmu_event_stop(struct perf_event *event, int flags)
@@ -529,10 +527,8 @@ void uncore_pmu_event_stop(struct perf_event *event, int flags)
                WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED);
                hwc->state |= PERF_HES_STOPPED;
 
-               if (box->n_active == 0) {
-                       uncore_disable_box(box);
+               if (box->n_active == 0)
                        uncore_pmu_cancel_hrtimer(box);
-               }
        }
 
        if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) {
@@ -778,6 +774,40 @@ static int uncore_pmu_event_init(struct perf_event *event)
        return ret;
 }
 
+static void uncore_pmu_enable(struct pmu *pmu)
+{
+       struct intel_uncore_pmu *uncore_pmu;
+       struct intel_uncore_box *box;
+
+       uncore_pmu = container_of(pmu, struct intel_uncore_pmu, pmu);
+       if (!uncore_pmu)
+               return;
+
+       box = uncore_pmu_to_box(uncore_pmu, smp_processor_id());
+       if (!box)
+               return;
+
+       if (uncore_pmu->type->ops->enable_box)
+               uncore_pmu->type->ops->enable_box(box);
+}
+
+static void uncore_pmu_disable(struct pmu *pmu)
+{
+       struct intel_uncore_pmu *uncore_pmu;
+       struct intel_uncore_box *box;
+
+       uncore_pmu = container_of(pmu, struct intel_uncore_pmu, pmu);
+       if (!uncore_pmu)
+               return;
+
+       box = uncore_pmu_to_box(uncore_pmu, smp_processor_id());
+       if (!box)
+               return;
+
+       if (uncore_pmu->type->ops->disable_box)
+               uncore_pmu->type->ops->disable_box(box);
+}
+
 static ssize_t uncore_get_attr_cpumask(struct device *dev,
                                struct device_attribute *attr, char *buf)
 {
@@ -803,6 +833,8 @@ static int uncore_pmu_register(struct intel_uncore_pmu *pmu)
                pmu->pmu = (struct pmu) {
                        .attr_groups    = pmu->type->attr_groups,
                        .task_ctx_nr    = perf_invalid_context,
+                       .pmu_enable     = uncore_pmu_enable,
+                       .pmu_disable    = uncore_pmu_disable,
                        .event_init     = uncore_pmu_event_init,
                        .add            = uncore_pmu_event_add,
                        .del            = uncore_pmu_event_del,
index f36f7be..bbfdaa7 100644 (file)
@@ -441,18 +441,6 @@ static inline int uncore_freerunning_hw_config(struct intel_uncore_box *box,
        return -EINVAL;
 }
 
-static inline void uncore_disable_box(struct intel_uncore_box *box)
-{
-       if (box->pmu->type->ops->disable_box)
-               box->pmu->type->ops->disable_box(box);
-}
-
-static inline void uncore_enable_box(struct intel_uncore_box *box)
-{
-       if (box->pmu->type->ops->enable_box)
-               box->pmu->type->ops->enable_box(box);
-}
-
 static inline void uncore_disable_event(struct intel_uncore_box *box,
                                struct perf_event *event)
 {