| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <1305477500-20739-3-git-send-email-acme@infradead.org>
Date: Sun, 15 May 2011 13:38:20 -0300
From: Arnaldo Carvalho de Melo <acme@...radead.org>
To: Ingo Molnar <mingo@...e.hu>
Cc: linux-kernel@...r.kernel.org,
Arnaldo Carvalho de Melo <acme@...hat.com>,
Frederic Weisbecker <fweisbec@...il.com>,
Ingo Molnar <mingo@...e.hu>, Mike Galbraith <efault@....de>,
Paul Mackerras <paulus@...ba.org>,
Peter Zijlstra <peterz@...radead.org>,
Stephane Eranian <eranian@...gle.com>,
Tom Zanussi <tzanussi@...il.com>
Subject: [PATCH 2/2] perf evlist: Fix per thread mmap setup
From: Arnaldo Carvalho de Melo <acme@...hat.com>
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@...icio ~]# 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@...icio ~]#
So 11 threads under pid 26131, then:
[root@...icio ~]# perf record -F 50000 --pid 26131
[root@...icio ~]# 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@...icio ~]#
11 mmaps, one per thread since we didn't specify any CPU list, so we need one
mmap per thread and:
[root@...icio ~]# 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@...icio ~]# 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@...icio ~]#
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@...icio ~]# 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@...icio ~]# 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@...icio ~]#
This machine has two cores, so fewer threads appeared on the radar, and:
[root@...icio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f484b922000-7f484b9a3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@...icio ~]#
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@...icio ~]# 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@...icio ~]# 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@...icio ~]#
It uses per-cpu buffers.
For just one thread:
[root@...icio ~]# 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@...icio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 9969 26148
[root@...icio ~]#
[root@...icio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f286a51b000-7f286a59c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@...icio ~]#
Tested-by: David Ahern <dsahern@...il.com>
Tested-by: Lin Ming <ming.m.lin@...el.com>
Cc: Frederic Weisbecker <fweisbec@...il.com>
Cc: Ingo Molnar <mingo@...e.hu>
Cc: Mike Galbraith <efault@....de>
Cc: Paul Mackerras <paulus@...ba.org>
Cc: Peter Zijlstra <peterz@...radead.org>
Cc: Stephane Eranian <eranian@...gle.com>
Cc: Tom Zanussi <tzanussi@...il.com>
Link: http://lkml.kernel.org/r/20110426204401.GB1746@ghostprotocols.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@...hat.com>
---
tools/perf/builtin-record.c | 2 +-
tools/perf/builtin-test.c | 2 +-
tools/perf/builtin-top.c | 8 +-
tools/perf/util/evlist.c | 151 ++++++++++++++++++++++++++++++-------------
tools/perf/util/evlist.h | 3 +-
tools/perf/util/python.c | 2 +-
6 files changed, 115 insertions(+), 53 deletions(-)
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 4165382..0974f95 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -427,7 +427,7 @@ static void mmap_read_all(void)
{
int i;
- for (i = 0; i < evsel_list->cpus->nr; i++) {
+ for (i = 0; i < evsel_list->nr_mmaps; i++) {
if (evsel_list->mmap[i].base)
mmap_read(&evsel_list->mmap[i]);
}
diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c
index 11e3c84..2f9a337 100644
--- a/tools/perf/builtin-test.c
+++ b/tools/perf/builtin-test.c
@@ -549,7 +549,7 @@ static int test__basic_mmap(void)
++foo;
}
- while ((event = perf_evlist__read_on_cpu(evlist, 0)) != NULL) {
+ while ((event = perf_evlist__mmap_read(evlist, 0)) != NULL) {
struct perf_sample sample;
if (event->header.type != PERF_RECORD_SAMPLE) {
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 7e3d6e3..ebfc7cf 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -801,12 +801,12 @@ static void perf_event__process_sample(const union perf_event *event,
}
}
-static void perf_session__mmap_read_cpu(struct perf_session *self, int cpu)
+static void perf_session__mmap_read_idx(struct perf_session *self, int idx)
{
struct perf_sample sample;
union perf_event *event;
- while ((event = perf_evlist__read_on_cpu(top.evlist, cpu)) != NULL) {
+ while ((event = perf_evlist__mmap_read(top.evlist, idx)) != NULL) {
perf_session__parse_sample(self, event, &sample);
if (event->header.type == PERF_RECORD_SAMPLE)
@@ -820,8 +820,8 @@ static void perf_session__mmap_read(struct perf_session *self)
{
int i;
- for (i = 0; i < top.evlist->cpus->nr; i++)
- perf_session__mmap_read_cpu(self, i);
+ for (i = 0; i < top.evlist->nr_mmaps; i++)
+ perf_session__mmap_read_idx(self, i);
}
static void start_counters(struct perf_evlist *evlist)
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 1884a7c..23eb22b 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,
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 8b1cb7a..7109d7a 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -17,6 +17,7 @@ struct perf_evlist {
struct hlist_head heads[PERF_EVLIST__HLIST_SIZE];
int nr_entries;
int nr_fds;
+ int nr_mmaps;
int mmap_len;
bool overwrite;
union perf_event event_copy;
@@ -46,7 +47,7 @@ void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd);
struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id);
-union perf_event *perf_evlist__read_on_cpu(struct perf_evlist *self, int cpu);
+union perf_event *perf_evlist__mmap_read(struct perf_evlist *self, int idx);
int perf_evlist__alloc_mmap(struct perf_evlist *evlist);
int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite);
diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
index f5e3845..99c7226 100644
--- a/tools/perf/util/python.c
+++ b/tools/perf/util/python.c
@@ -680,7 +680,7 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
&cpu, &sample_id_all))
return NULL;
- event = perf_evlist__read_on_cpu(evlist, cpu);
+ event = perf_evlist__mmap_read(evlist, cpu);
if (event != NULL) {
struct perf_evsel *first;
PyObject *pyevent = pyrf_event__new(event);
--
1.6.2.5
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists