perf evlist: Fix per thread mmap setup

The PERF_EVENT_IOC_SET_OUTPUT ioctl was returning -EINVAL when using
--pid when monitoring multithreaded apps, as we can only share a ring
buffer for events on the same thread if not doing per cpu.

Fix it by using per thread ring buffers.

Tested with:

[root@felicio ~]# tuna -t 26131 -CP | nl
  1                      thread       ctxt_switches
  2    pid SCHED_ rtpri affinity voluntary nonvoluntary             cmd
  3 26131   OTHER     0      0,1  10814276      2397830 chromium-browse
  4  642    OTHER     0      0,1     14688            0 chromium-browse
  5  26148  OTHER     0      0,1    713602       115479 chromium-browse
  6  26149  OTHER     0      0,1    801958         2262 chromium-browse
  7  26150  OTHER     0      0,1   1271128          248 chromium-browse
  8  26151  OTHER     0      0,1         3            0 chromium-browse
  9  27049  OTHER     0      0,1     36796            9 chromium-browse
 10  618    OTHER     0      0,1     14711            0 chromium-browse
 11  661    OTHER     0      0,1     14593            0 chromium-browse
 12  29048  OTHER     0      0,1     28125            0 chromium-browse
 13  26143  OTHER     0      0,1   2202789          781 chromium-browse
[root@felicio ~]#

So 11 threads under pid 26131, then:

[root@felicio ~]# perf record -F 50000 --pid 26131

[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
  1 7fa4a2538000-7fa4a25b9000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  2 7fa4a25b9000-7fa4a263a000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  3 7fa4a263a000-7fa4a26bb000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  4 7fa4a26bb000-7fa4a273c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  5 7fa4a273c000-7fa4a27bd000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  6 7fa4a27bd000-7fa4a283e000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  7 7fa4a283e000-7fa4a28bf000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  8 7fa4a28bf000-7fa4a2940000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
  9 7fa4a2940000-7fa4a29c1000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
 10 7fa4a29c1000-7fa4a2a42000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
 11 7fa4a2a42000-7fa4a2ac3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#

11 mmaps, one per thread since we didn't specify any CPU list, so we need one
mmap per thread and:

[root@felicio ~]# perf record -F 50000 --pid 26131
^M
^C[ perf record: Woken up 79 times to write data ]
[ perf record: Captured and wrote 20.614 MB perf.data (~900639 samples) ]

[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
     1	 371310 26131
     2	  96516 26148
     3	  95694 26149
     4	  95203 26150
     5	   7291 26143
     6	     87 27049
     7	     76 661
     8	     60 29048
     9	     47 618
    10	     43 642
[root@felicio ~]#

Ok, one of the threads, 26151 was quiescent, so no samples there, but all the
others are there.

Then, if I specify one CPU:

[root@felicio ~]# perf record -F 50000 --pid 26131 --cpu 1
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.680 MB perf.data (~29730 samples) ]

[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
     1	   8444 26131
     2	   2584 26149
     3	   2518 26148
     4	   2324 26150
     5	    123 26143
     6	      9 661
     7	      9 29048
[root@felicio ~]#

This machine has two cores, so fewer threads appeared on the radar, and:

[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
 1 7f484b922000-7f484b9a3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#

Just one mmap, as now we can use just one per-cpu buffer instead of the
per-thread needed in the previous case.

For global profiling:

[root@felicio ~]# perf record -F 50000 -a
^C[ perf record: Woken up 26 times to write data ]
[ perf record: Captured and wrote 7.128 MB perf.data (~311412 samples) ]

[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
     1	7fb49b435000-7fb49b4b6000 rwxs 00000000 00:09 4064                       anon_inode:[perf_event]
     2	7fb49b4b6000-7fb49b537000 rwxs 00000000 00:09 4064                       anon_inode:[perf_event]
[root@felicio ~]#

It uses per-cpu buffers.

For just one thread:

[root@felicio ~]# perf record -F 50000 --tid 26148
^C[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.330 MB perf.data (~14426 samples) ]

[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
     1	   9969 26148
[root@felicio ~]#

[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
     1	7f286a51b000-7f286a59c000 rwxs 00000000 00:09 4064                       anon_inode:[perf_event]
[root@felicio ~]#

Tested-by: David Ahern <dsahern@gmail.com>
Tested-by: Lin Ming <ming.m.lin@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Link: http://lkml.kernel.org/r/20110426204401.GB1746@ghostprotocols.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

1 files changed, 106 insertions, 45 deletions

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 1884a7c7eb8..23eb22b05d2 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -166,11 +166,11 @@ struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id)
 	return NULL;
 }
 
-union perf_event *perf_evlist__read_on_cpu(struct perf_evlist *evlist, int cpu)
+union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 {
 	/* XXX Move this to perf.c, making it generally available */
 	unsigned int page_size = sysconf(_SC_PAGE_SIZE);
-	struct perf_mmap *md = &evlist->mmap[cpu];
+	struct perf_mmap *md = &evlist->mmap[idx];
 	unsigned int head = perf_mmap__read_head(md);
 	unsigned int old = md->prev;
 	unsigned char *data = md->base + page_size;
@@ -235,31 +235,37 @@ union perf_event *perf_evlist__read_on_cpu(struct perf_evlist *evlist, int cpu)
 
 void perf_evlist__munmap(struct perf_evlist *evlist)
 {
-	int cpu;
+	int i;
 
-	for (cpu = 0; cpu < evlist->cpus->nr; cpu++) {
-		if (evlist->mmap[cpu].base != NULL) {
-			munmap(evlist->mmap[cpu].base, evlist->mmap_len);
-			evlist->mmap[cpu].base = NULL;
+	for (i = 0; i < evlist->nr_mmaps; i++) {
+		if (evlist->mmap[i].base != NULL) {
+			munmap(evlist->mmap[i].base, evlist->mmap_len);
+			evlist->mmap[i].base = NULL;
 		}
 	}
+
+	free(evlist->mmap);
+	evlist->mmap = NULL;
 }
 
 int perf_evlist__alloc_mmap(struct perf_evlist *evlist)
 {
-	evlist->mmap = zalloc(evlist->cpus->nr * sizeof(struct perf_mmap));
+	evlist->nr_mmaps = evlist->cpus->nr;
+	if (evlist->cpus->map[0] == -1)
+		evlist->nr_mmaps = evlist->threads->nr;
+	evlist->mmap = zalloc(evlist->nr_mmaps * sizeof(struct perf_mmap));
 	return evlist->mmap != NULL ? 0 : -ENOMEM;
 }
 
 static int __perf_evlist__mmap(struct perf_evlist *evlist, struct perf_evsel *evsel,
-			       int cpu, int prot, int mask, int fd)
+			       int idx, int prot, int mask, int fd)
 {
-	evlist->mmap[cpu].prev = 0;
-	evlist->mmap[cpu].mask = mask;
-	evlist->mmap[cpu].base = mmap(NULL, evlist->mmap_len, prot,
+	evlist->mmap[idx].prev = 0;
+	evlist->mmap[idx].mask = mask;
+	evlist->mmap[idx].base = mmap(NULL, evlist->mmap_len, prot,
 				      MAP_SHARED, fd, 0);
-	if (evlist->mmap[cpu].base == MAP_FAILED) {
-		if (evlist->cpus->map[cpu] == -1 && evsel->attr.inherit)
+	if (evlist->mmap[idx].base == MAP_FAILED) {
+		if (evlist->cpus->map[idx] == -1 && evsel->attr.inherit)
 			ui__warning("Inherit is not allowed on per-task "
 				    "events using mmap.\n");
 		return -1;
@@ -269,6 +275,86 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, struct perf_evsel *ev
 	return 0;
 }
 
+static int perf_evlist__mmap_per_cpu(struct perf_evlist *evlist, int prot, int mask)
+{
+	struct perf_evsel *evsel;
+	int cpu, thread;
+
+	for (cpu = 0; cpu < evlist->cpus->nr; cpu++) {
+		int output = -1;
+
+		for (thread = 0; thread < evlist->threads->nr; thread++) {
+			list_for_each_entry(evsel, &evlist->entries, node) {
+				int fd = FD(evsel, cpu, thread);
+
+				if (output == -1) {
+					output = fd;
+					if (__perf_evlist__mmap(evlist, evsel, cpu,
+								prot, mask, output) < 0)
+						goto out_unmap;
+				} else {
+					if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT, output) != 0)
+						goto out_unmap;
+				}
+
+				if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
+				    perf_evlist__id_add_fd(evlist, evsel, cpu, thread, fd) < 0)
+					goto out_unmap;
+			}
+		}
+	}
+
+	return 0;
+
+out_unmap:
+	for (cpu = 0; cpu < evlist->cpus->nr; cpu++) {
+		if (evlist->mmap[cpu].base != NULL) {
+			munmap(evlist->mmap[cpu].base, evlist->mmap_len);
+			evlist->mmap[cpu].base = NULL;
+		}
+	}
+	return -1;
+}
+
+static int perf_evlist__mmap_per_thread(struct perf_evlist *evlist, int prot, int mask)
+{
+	struct perf_evsel *evsel;
+	int thread;
+
+	for (thread = 0; thread < evlist->threads->nr; thread++) {
+		int output = -1;
+
+		list_for_each_entry(evsel, &evlist->entries, node) {
+			int fd = FD(evsel, 0, thread);
+
+			if (output == -1) {
+				output = fd;
+				if (__perf_evlist__mmap(evlist, evsel, thread,
+							prot, mask, output) < 0)
+					goto out_unmap;
+			} else {
+				if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT, output) != 0)
+					goto out_unmap;
+			}
+
+			if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
+			    perf_evlist__id_add_fd(evlist, evsel, 0, thread, fd) < 0)
+				goto out_unmap;
+		}
+	}
+
+	return 0;
+
+out_unmap:
+	for (thread = 0; thread < evlist->threads->nr; thread++) {
+		if (evlist->mmap[thread].base != NULL) {
+			munmap(evlist->mmap[thread].base, evlist->mmap_len);
+			evlist->mmap[thread].base = NULL;
+		}
+	}
+	return -1;
+}
+
 /** perf_evlist__mmap - Create per cpu maps to receive events
  *
  * @evlist - list of events
@@ -287,11 +373,11 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, struct perf_evsel *ev
 int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite)
 {
 	unsigned int page_size = sysconf(_SC_PAGE_SIZE);
-	int mask = pages * page_size - 1, cpu;
-	struct perf_evsel *first_evsel, *evsel;
+	int mask = pages * page_size - 1;
+	struct perf_evsel *evsel;
 	const struct cpu_map *cpus = evlist->cpus;
 	const struct thread_map *threads = evlist->threads;
-	int thread, prot = PROT_READ | (overwrite ? 0 : PROT_WRITE);
+	int prot = PROT_READ | (overwrite ? 0 : PROT_WRITE);
 
 	if (evlist->mmap == NULL && perf_evlist__alloc_mmap(evlist) < 0)
 		return -ENOMEM;
@@ -301,43 +387,18 @@ int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite)
 
 	evlist->overwrite = overwrite;
 	evlist->mmap_len = (pages + 1) * page_size;
-	first_evsel = list_entry(evlist->entries.next, struct perf_evsel, node);
 
 	list_for_each_entry(evsel, &evlist->entries, node) {
 		if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
 		    evsel->sample_id == NULL &&
 		    perf_evsel__alloc_id(evsel, cpus->nr, threads->nr) < 0)
 			return -ENOMEM;
-
-		for (cpu = 0; cpu < cpus->nr; cpu++) {
-			for (thread = 0; thread < threads->nr; thread++) {
-				int fd = FD(evsel, cpu, thread);
-
-				if (evsel->idx || thread) {
-					if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT,
-						  FD(first_evsel, cpu, 0)) != 0)
-						goto out_unmap;
-				} else if (__perf_evlist__mmap(evlist, evsel, cpu,
-							       prot, mask, fd) < 0)
-					goto out_unmap;
-
-				if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
-				    perf_evlist__id_add_fd(evlist, evsel, cpu, thread, fd) < 0)
-					goto out_unmap;
-			}
-		}
 	}
 
-	return 0;
+	if (evlist->cpus->map[0] == -1)
+		return perf_evlist__mmap_per_thread(evlist, prot, mask);
 
-out_unmap:
-	for (cpu = 0; cpu < cpus->nr; cpu++) {
-		if (evlist->mmap[cpu].base != NULL) {
-			munmap(evlist->mmap[cpu].base, evlist->mmap_len);
-			evlist->mmap[cpu].base = NULL;
-		}
-	}
-	return -1;
+	return perf_evlist__mmap_per_cpu(evlist, prot, mask);
 }
 
 int perf_evlist__create_maps(struct perf_evlist *evlist, pid_t target_pid,