diff mbox series

[4/7] perf lock contention: Add data failure stat

Message ID 20230406210611.1622492-5-namhyung@kernel.org (mailing list archive)
State Not Applicable
Headers show
Series perf lock contention: Improve performance if map is full (v1) | expand

Checks

Context Check Description
netdev/tree_selection success Not a local patch

Commit Message

Namhyung Kim April 6, 2023, 9:06 p.m. UTC 
It's possible to fail to update the data when the lock_stat map is full.
We should check that case and show the number at the end.

  $ sudo ./perf lock con -ablv -E3 -- ./perf bench sched messaging
  ...
   contended   total wait     max wait     avg wait            address   symbol

        6157    208.48 ms     69.29 us     33.86 us   ffff934c001c1f00    (spinlock)
        4030     72.04 ms     61.84 us     17.88 us   ffff934c000415c0    (spinlock)
        3201     50.30 ms     47.73 us     15.71 us   ffff934c2eead850    (spinlock)

  === output for debug ===

  bad: 0, total: 13388
  bad rate: 0.00 %
  histogram of failure reasons
         task: 0
        stack: 0
         time: 0
         data: 0      <----- added

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/builtin-lock.c                      | 4 +++-
 tools/perf/util/bpf_lock_contention.c          | 1 +
 tools/perf/util/bpf_skel/lock_contention.bpf.c | 4 +++-
 tools/perf/util/lock-contention.h              | 1 +
 4 files changed, 8 insertions(+), 2 deletions(-)
diff mbox series

Patch

diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c
index 9b92c7a5aefb..01b318d6c80a 100644
--- a/tools/perf/builtin-lock.c
+++ b/tools/perf/builtin-lock.c
@@ -1626,7 +1626,7 @@  static void sort_contention_result(void)
 static void print_bpf_events(int total, struct lock_contention_fails *fails)
 {
 	/* Output for debug, this have to be removed */
-	int broken = fails->task + fails->stack + fails->time;
+	int broken = fails->task + fails->stack + fails->time + fails->data;
 
 	if (quiet || total == 0 || (broken == 0 && verbose <= 0))
 		return;
@@ -1640,7 +1640,9 @@  static void print_bpf_events(int total, struct lock_contention_fails *fails)
 	pr_info(" %10s: %d\n", "task", fails->task);
 	pr_info(" %10s: %d\n", "stack", fails->stack);
 	pr_info(" %10s: %d\n", "time", fails->time);
+	pr_info(" %10s: %d\n", "data", fails->data);
 }
+
 static void print_contention_result(struct lock_contention *con)
 {
 	struct lock_stat *st;
diff --git a/tools/perf/util/bpf_lock_contention.c b/tools/perf/util/bpf_lock_contention.c
index 8a5d0eb44189..0071058ac3d2 100644
--- a/tools/perf/util/bpf_lock_contention.c
+++ b/tools/perf/util/bpf_lock_contention.c
@@ -262,6 +262,7 @@  int lock_contention_read(struct lock_contention *con)
 	con->fails.task = skel->bss->task_fail;
 	con->fails.stack = skel->bss->stack_fail;
 	con->fails.time = skel->bss->time_fail;
+	con->fails.data = skel->bss->data_fail;
 
 	stack_trace = zalloc(stack_size);
 	if (stack_trace == NULL)
diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c
index f9d2d792ccc8..cb87c98e5340 100644
--- a/tools/perf/util/bpf_skel/lock_contention.bpf.c
+++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c
@@ -124,6 +124,7 @@  int aggr_mode;
 int task_fail;
 int stack_fail;
 int time_fail;
+int data_fail;
 
 static inline int can_record(u64 *ctx)
 {
@@ -380,7 +381,8 @@  int contention_end(u64 *ctx)
 		if (aggr_mode == LOCK_AGGR_ADDR)
 			first.flags |= check_lock_type(pelem->lock, pelem->flags);
 
-		bpf_map_update_elem(&lock_stat, &key, &first, BPF_NOEXIST);
+		if (bpf_map_update_elem(&lock_stat, &key, &first, BPF_NOEXIST) < 0)
+			__sync_fetch_and_add(&data_fail, 1);
 		bpf_map_delete_elem(&tstamp, &pid);
 		return 0;
 	}
diff --git a/tools/perf/util/lock-contention.h b/tools/perf/util/lock-contention.h
index 10c28302420c..3ed1cad370fc 100644
--- a/tools/perf/util/lock-contention.h
+++ b/tools/perf/util/lock-contention.h
@@ -126,6 +126,7 @@  struct lock_contention_fails {
 	int task;
 	int stack;
 	int time;
+	int data;
 };
 
 struct lock_contention {