The following commit has been merged into the perf/core branch of tip:
Commit-ID: 66d258c5b048840991de49697264af75f5b09def
Gitweb: https://git.kernel.org/tip/66d258c5b048840991de49697264af75f5b09def
Author: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
AuthorDate: Thu, 17 Oct 2019 20:31:03 +02:00
Committer: Ingo Molnar <mingo@xxxxxxxxxx>
CommitterDate: Mon, 28 Oct 2019 12:51:02 +01:00
perf/core: Optimize perf_init_event()
Andi reported that he was hitting the linear search in
perf_init_event() a lot. Make more agressive use of the IDR lookup to
avoid hitting the linear search.
With exception of PERF_TYPE_SOFTWARE (which relies on a hideous hack),
we can put everything in the IDR. On top of that, we can alias
TYPE_HARDWARE and TYPE_HW_CACHE to TYPE_RAW on the lookup side.
This greatly reduces the chances of hitting the linear search.
Reported-by: Andi Kleen <andi@xxxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Cc: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
Cc: Kan <kan.liang@xxxxxxxxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Mark Rutland <mark.rutland@xxxxxxx>
Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Stephane Eranian <eranian@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Vince Weaver <vincent.weaver@xxxxxxxxx>
Signed-off-by: Ingo Molnar <mingo@xxxxxxxxxx>
---
kernel/events/core.c | 41 ++++++++++++++++++++++++++++++-----------
1 file changed, 30 insertions(+), 11 deletions(-)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index ea70ca6..4d67c5d 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -10080,7 +10080,7 @@ static struct lock_class_key cpuctx_lock;
int perf_pmu_register(struct pmu *pmu, const char *name, int type)
{
- int cpu, ret;
+ int cpu, ret, max = PERF_TYPE_MAX;
mutex_lock(&pmus_lock);
ret = -ENOMEM;
@@ -10093,12 +10093,17 @@ int perf_pmu_register(struct pmu *pmu, const char *name, int type)
goto skip_type;
pmu->name = name;
- if (type < 0) {
- type = idr_alloc(&pmu_idr, pmu, PERF_TYPE_MAX, 0, GFP_KERNEL);
- if (type < 0) {
- ret = type;
+ if (type != PERF_TYPE_SOFTWARE) {
+ if (type >= 0)
+ max = type;
+
+ ret = idr_alloc(&pmu_idr, pmu, max, 0, GFP_KERNEL);
+ if (ret < 0)
goto free_pdc;
- }
+
+ WARN_ON(type >= 0 && ret != type);
+
+ type = ret;
}
pmu->type = type;
@@ -10188,7 +10193,7 @@ free_dev:
put_device(pmu->dev);
free_idr:
- if (pmu->type >= PERF_TYPE_MAX)
+ if (pmu->type != PERF_TYPE_SOFTWARE)
idr_remove(&pmu_idr, pmu->type);
free_pdc:
@@ -10210,7 +10215,7 @@ void perf_pmu_unregister(struct pmu *pmu)
synchronize_rcu();
free_percpu(pmu->pmu_disable_count);
- if (pmu->type >= PERF_TYPE_MAX)
+ if (pmu->type != PERF_TYPE_SOFTWARE)
idr_remove(&pmu_idr, pmu->type);
if (pmu_bus_running) {
if (pmu->nr_addr_filters)
@@ -10280,9 +10285,8 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)
static struct pmu *perf_init_event(struct perf_event *event)
{
+ int idx, type, ret;
struct pmu *pmu;
- int idx;
- int ret;
idx = srcu_read_lock(&pmus_srcu);
@@ -10295,12 +10299,27 @@ static struct pmu *perf_init_event(struct perf_event *event)
}
rcu_read_lock();
- pmu = idr_find(&pmu_idr, event->attr.type);
+ /*
+ * PERF_TYPE_HARDWARE and PERF_TYPE_HW_CACHE
+ * are often aliases for PERF_TYPE_RAW.
+ */
+ type = event->attr.type;
+ if (type == PERF_TYPE_HARDWARE || type == PERF_TYPE_HW_CACHE)
+ type = PERF_TYPE_RAW;
+
+again:
+ pmu = idr_find(&pmu_idr, type);
rcu_read_unlock();
if (pmu) {
ret = perf_try_init_event(pmu, event);
+ if (ret == -ENOENT && event->attr.type != type) {
+ type = event->attr.type;
+ goto again;
+ }
+
if (ret)
pmu = ERR_PTR(ret);
+
goto unlock;
}
|