perf tools: Improve thread comm resolution in perf sched

When we get sched traces that involve a task that was already created before opening the event, we won't have the comm event for it. So if we can't find the comm event for a given thread, we look at the traces that may contain these informations. Before: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | :5124:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | :6244:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | :6245:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | After: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | firefox:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | npviewer.bin:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | npviewer.bin:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1255012632-7882-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Improve thread comm resolution in perf sched
When we get sched traces that involve a task that was already created before opening the event, we won't have the comm event for it. So if we can't find the comm event for a given thread, we look at the traces that may contain these informations. Before: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | :5124:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | :6244:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | :6245:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | After: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | firefox:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | npviewer.bin:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | npviewer.bin:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1255012632-7882-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Frederic Weisbecker · Ingo Molnar
1 parent 016e92fbc9
Showing 3 changed files with 67 additions and 12 deletions Side-by-side Diff
tools/perf/builtin-sched.c
tools/perf/util/thread.c
tools/perf/util/thread.h
@@ -1034,6 +1034,36 @@
 	atoms->nb_atoms++;
 }
  
+static struct thread *
+threads__findnew_from_ctx(u32 pid, struct trace_switch_event *switch_event)
+{
+	struct thread *th;
+
+	th = threads__findnew_nocomm(pid, &threads, &last_match);
+	if (th->comm)
+		return th;
+
+	if (pid == switch_event->prev_pid)
+		thread__set_comm(th, switch_event->prev_comm);
+	else
+		thread__set_comm(th, switch_event->next_comm);
+	return th;
+}
+
+static struct thread *
+threads__findnew_from_wakeup(struct trace_wakeup_event *wakeup_event)
+{
+	struct thread *th;
+
+	th =  threads__findnew_nocomm(wakeup_event->pid, &threads, &last_match);
+	if (th->comm)
+		return th;
+
+	thread__set_comm(th, wakeup_event->comm);
+
+	return th;
+}
+
 static void
 latency_switch_event(struct trace_switch_event *switch_event,
 		     struct event *event __used,
@@ -1059,8 +1089,10 @@
 		die("hm, delta: %Ld < 0 ?\n", delta);
  
  
-	sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match);
-	sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match);
+	sched_out = threads__findnew_from_ctx(switch_event->prev_pid,
+					      switch_event);
+	sched_in = threads__findnew_from_ctx(switch_event->next_pid,
+					     switch_event);
  
 	out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
 	if (!out_events) {
@@ -1126,7 +1158,7 @@
 	if (!wakeup_event->success)
 		return;
  
-	wakee = threads__findnew(wakeup_event->pid, &threads, &last_match);
+	wakee = threads__findnew_from_wakeup(wakeup_event);
 	atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
 	if (!atoms) {
 		thread_atoms_insert(wakee);
@@ -1386,8 +1418,10 @@
 		die("hm, delta: %Ld < 0 ?\n", delta);
  
  
-	sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match);
-	sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match);
+	sched_out = threads__findnew_from_ctx(switch_event->prev_pid,
+					      switch_event);
+	sched_in = threads__findnew_from_ctx(switch_event->next_pid,
+					     switch_event);
  
 	curr_thread[this_cpu] = sched_in;
  
@@ -6,15 +6,17 @@
 #include "util.h"
 #include "debug.h"
  
-static struct thread *thread__new(pid_t pid)
+static struct thread *thread__new(pid_t pid, int set_comm)
 {
 	struct thread *self = calloc(1, sizeof(*self));
  
 	if (self != NULL) {
 		self->pid = pid;
-		self->comm = malloc(32);
-		if (self->comm)
-			snprintf(self->comm, 32, ":%d", self->pid);
+		if (set_comm) {
+			self->comm = malloc(32);
+			if (self->comm)
+				snprintf(self->comm, 32, ":%d", self->pid);
+		}
 		self->maps = RB_ROOT;
 		INIT_LIST_HEAD(&self->removed_maps);
 	}
@@ -50,8 +52,10 @@
 	return ret;
 }
  
-struct thread *
-threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match)
+static struct thread *
+__threads__findnew(pid_t pid, struct rb_root *threads,
+		   struct thread **last_match,
+		   int set_comm)
 {
 	struct rb_node **p = &threads->rb_node;
 	struct rb_node *parent = NULL;
@@ -80,7 +84,8 @@
 			p = &(*p)->rb_right;
 	}
  
-	th = thread__new(pid);
+	th = thread__new(pid, set_comm);
+
 	if (th != NULL) {
 		rb_link_node(&th->rb_node, parent, p);
 		rb_insert_color(&th->rb_node, threads);
@@ -88,6 +93,19 @@
 	}
  
 	return th;
+}
+
+struct thread *
+threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match)
+{
+	return __threads__findnew(pid, threads, last_match, 1);
+}
+
+struct thread *
+threads__findnew_nocomm(pid_t pid, struct rb_root *threads,
+			struct thread **last_match)
+{
+	return __threads__findnew(pid, threads, last_match, 0);
 }
  
 struct thread *
@@ -18,6 +18,9 @@
 struct thread *
 threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match);
 struct thread *
+threads__findnew_nocomm(pid_t pid, struct rb_root *threads,
+			struct thread **last_match);
+struct thread *
 register_idle_thread(struct rb_root *threads, struct thread **last_match);
 void thread__insert_map(struct thread *self, struct map *map);
 int thread__fork(struct thread *self, struct thread *parent);
...	...	@@ -1034,6 +1034,36 @@
1034	1034	atoms->nb_atoms++;
1035	1035	}
1036	1036
	1037	+static struct thread *
	1038	+threads__findnew_from_ctx(u32 pid, struct trace_switch_event *switch_event)
	1039	+{
	1040	+ struct thread *th;
	1041	+
	1042	+ th = threads__findnew_nocomm(pid, &threads, &last_match);
	1043	+ if (th->comm)
	1044	+ return th;
	1045	+
	1046	+ if (pid == switch_event->prev_pid)
	1047	+ thread__set_comm(th, switch_event->prev_comm);
	1048	+ else
	1049	+ thread__set_comm(th, switch_event->next_comm);
	1050	+ return th;
	1051	+}
	1052	+
	1053	+static struct thread *
	1054	+threads__findnew_from_wakeup(struct trace_wakeup_event *wakeup_event)
	1055	+{
	1056	+ struct thread *th;
	1057	+
	1058	+ th = threads__findnew_nocomm(wakeup_event->pid, &threads, &last_match);
	1059	+ if (th->comm)
	1060	+ return th;
	1061	+
	1062	+ thread__set_comm(th, wakeup_event->comm);
	1063	+
	1064	+ return th;
	1065	+}
	1066	+
1037	1067	static void
1038	1068	latency_switch_event(struct trace_switch_event *switch_event,
1039	1069	struct event *event __used,
...	...	@@ -1059,8 +1089,10 @@
1059	1089	die("hm, delta: %Ld < 0 ?\n", delta);
1060	1090
1061	1091
1062		- sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match);
1063		- sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match);
	1092	+ sched_out = threads__findnew_from_ctx(switch_event->prev_pid,
	1093	+ switch_event);
	1094	+ sched_in = threads__findnew_from_ctx(switch_event->next_pid,
	1095	+ switch_event);
1064	1096
1065	1097	out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
1066	1098	if (!out_events) {
...	...	@@ -1126,7 +1158,7 @@
1126	1158	if (!wakeup_event->success)
1127	1159	return;
1128	1160
1129		- wakee = threads__findnew(wakeup_event->pid, &threads, &last_match);
	1161	+ wakee = threads__findnew_from_wakeup(wakeup_event);
1130	1162	atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
1131	1163	if (!atoms) {
1132	1164	thread_atoms_insert(wakee);
...	...	@@ -1386,8 +1418,10 @@
1386	1418	die("hm, delta: %Ld < 0 ?\n", delta);
1387	1419
1388	1420
1389		- sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match);
1390		- sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match);
	1421	+ sched_out = threads__findnew_from_ctx(switch_event->prev_pid,
	1422	+ switch_event);
	1423	+ sched_in = threads__findnew_from_ctx(switch_event->next_pid,
	1424	+ switch_event);
1391	1425
1392	1426	curr_thread[this_cpu] = sched_in;
1393	1427
...	...	@@ -6,15 +6,17 @@
6	6	#include "util.h"
7	7	#include "debug.h"
8	8
9		-static struct thread *thread__new(pid_t pid)
	9	+static struct thread *thread__new(pid_t pid, int set_comm)
10	10	{
11	11	struct thread self = calloc(1, sizeof(self));
12	12
13	13	if (self != NULL) {
14	14	self->pid = pid;
15		- self->comm = malloc(32);
16		- if (self->comm)
17		- snprintf(self->comm, 32, ":%d", self->pid);
	15	+ if (set_comm) {
	16	+ self->comm = malloc(32);
	17	+ if (self->comm)
	18	+ snprintf(self->comm, 32, ":%d", self->pid);
	19	+ }
18	20	self->maps = RB_ROOT;
19	21	INIT_LIST_HEAD(&self->removed_maps);
20	22	}
...	...	@@ -50,8 +52,10 @@
50	52	return ret;
51	53	}
52	54
53		-struct thread *
54		-threads__findnew(pid_t pid, struct rb_root threads, struct thread *last_match)
	55	+static struct thread *
	56	+__threads__findnew(pid_t pid, struct rb_root *threads,
	57	+ struct thread **last_match,
	58	+ int set_comm)
55	59	{
56	60	struct rb_node **p = &threads->rb_node;
57	61	struct rb_node *parent = NULL;
...	...	@@ -80,7 +84,8 @@
80	84	p = &(*p)->rb_right;
81	85	}
82	86
83		- th = thread__new(pid);
	87	+ th = thread__new(pid, set_comm);
	88	+
84	89	if (th != NULL) {
85	90	rb_link_node(&th->rb_node, parent, p);
86	91	rb_insert_color(&th->rb_node, threads);
...	...	@@ -88,6 +93,19 @@
88	93	}
89	94
90	95	return th;
	96	+}
	97	+
	98	+struct thread *
	99	+threads__findnew(pid_t pid, struct rb_root threads, struct thread *last_match)
	100	+{
	101	+ return __threads__findnew(pid, threads, last_match, 1);
	102	+}
	103	+
	104	+struct thread *
	105	+threads__findnew_nocomm(pid_t pid, struct rb_root *threads,
	106	+ struct thread **last_match)
	107	+{
	108	+ return __threads__findnew(pid, threads, last_match, 0);
91	109	}
92	110
93	111	struct thread *
...	...	@@ -18,6 +18,9 @@
18	18	struct thread *
19	19	threads__findnew(pid_t pid, struct rb_root threads, struct thread *last_match);
20	20	struct thread *
	21	+threads__findnew_nocomm(pid_t pid, struct rb_root *threads,
	22	+ struct thread **last_match);
	23	+struct thread *
21	24	register_idle_thread(struct rb_root threads, struct thread *last_match);
22	25	void thread__insert_map(struct thread self, struct map map);
23	26	int thread__fork(struct thread self, struct thread parent);