Add arena lock stats output.

include/jemalloc/internal/ctl_externs.h

@@ -1,6 +1,14 @@
 #ifndef JEMALLOC_INTERNAL_CTL_EXTERNS_H
 #define JEMALLOC_INTERNAL_CTL_EXTERNS_H
 
+/* Maximum ctl tree depth. */
+#define CTL_MAX_DEPTH	7
+
+#define NUM_ARENA_PROF_LOCKS 6
+#define NUM_LOCK_PROF_COUNTERS 7
+const char *arena_lock_names[NUM_ARENA_PROF_LOCKS];
+const char *lock_counter_names[NUM_LOCK_PROF_COUNTERS];
+
 int	ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
     void *newp, size_t newlen);
 int	ctl_nametomib(tsdn_t *tsdn, const char *name, size_t *mibp,

include/jemalloc/internal/mutex_inlines.h

@@ -29,9 +29,9 @@ malloc_mutex_trylock(malloc_mutex_t *mutex) {
 /* Aggregate lock prof data. */
 JEMALLOC_INLINE void
 malloc_lock_prof_merge(lock_prof_data_t *sum, lock_prof_data_t *data) {
-	nstime_add(&sum->tot_wait_time, &data->tot_wait_time);
-	if (nstime_compare(&data->max_wait_time, &sum->max_wait_time)) {
-		nstime_copy(&sum->max_wait_time, &data->max_wait_time);
+	sum->tot_wait_time += data->tot_wait_time;
+	if (data->max_wait_time > sum->max_wait_time) {
+		sum->max_wait_time = data->max_wait_time;
 	}
 	sum->n_wait_times += data->n_wait_times;
 	sum->n_spin_acquired += data->n_spin_acquired;

include/jemalloc/internal/mutex_structs.h

@@ -6,10 +6,10 @@ struct lock_prof_data_s {
 	 * Counters touched on the slow path, i.e. when there is lock
 	 * contention.  We update them once we have the lock.
 	 */
-	/* Total time spent waiting on this lock. */
-	nstime_t		tot_wait_time;
-	/* Max time spent on a single lock operation. */
-	nstime_t		max_wait_time;
+	/* Total time (in nano seconds) spent waiting on this lock. */
+	uint64_t		tot_wait_time;
+	/* Max time (in nano seconds) spent on a single lock operation. */
+	uint64_t		max_wait_time;
 	/* # of times have to wait for this lock (after spinning). */
 	uint64_t		n_wait_times;
 	/* # of times acquired the lock through local spinning. */

include/jemalloc/internal/mutex_types.h

@@ -28,8 +28,7 @@ typedef struct malloc_mutex_s malloc_mutex_t;
 #    define MALLOC_MUTEX_TRYLOCK(m) (pthread_mutex_trylock(&(m)->lock) != 0)
 #endif
 
-#define LOCK_PROF_DATA_INITIALIZER					\
-    {NSTIME_ZERO_INITIALIZER, NSTIME_ZERO_INITIALIZER, 0, 0, 0, 0, 0, NULL, 0}
+#define LOCK_PROF_DATA_INITIALIZER {0, 0, 0, 0, 0, 0, 0, NULL, 0}
 
 #ifdef _WIN32
 #  define MALLOC_MUTEX_INITIALIZER

include/jemalloc/internal/stats_structs.h

@@ -124,6 +124,13 @@ struct arena_stats_s {
 	/* Number of bytes cached in tcache associated with this arena. */
 	atomic_zu_t		tcache_bytes; /* Derived. */
 
+	lock_prof_data_t large_mtx_data;
+	lock_prof_data_t extent_freelist_mtx_data;
+	lock_prof_data_t extents_cached_mtx_data;
+	lock_prof_data_t extents_retained_mtx_data;
+	lock_prof_data_t decay_mtx_data;
+	lock_prof_data_t tcache_mtx_data;
+
 	/* One element for each large size class. */
 	malloc_large_stats_t	lstats[NSIZES - NBINS];
 };