When event group is enabled for forked task (i.e. no target task/cpu
was specified) all events were disabled and marked ->enable_on_exec.
However they wouldn't be counted at all since only group leader will
be enabled on exec actually.
In contrast to perf stat, perf record doesn't have a real problem
as it enables all the event before proceeding. But it needs to be
fixed anyway IMHO.
Signed-off-by: Namhyung Kim <[email protected]>
---
tools/perf/builtin-record.c | 2 +-
tools/perf/builtin-test.c | 2 +-
tools/perf/util/evlist.c | 5 +++--
tools/perf/util/evlist.h | 3 ++-
tools/perf/util/evsel.c | 5 +++--
tools/perf/util/evsel.h | 3 ++-
6 files changed, 12 insertions(+), 8 deletions(-)
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index fbf71edd31ba..a65b0a5a81ea 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -183,7 +183,7 @@ static void perf_record__open(struct perf_record *rec)
first = list_entry(evlist->entries.next, struct perf_evsel, node);
- perf_evlist__config_attrs(evlist, opts);
+ perf_evlist__config_attrs(evlist, opts, first);
list_for_each_entry(pos, &evlist->entries, node) {
struct perf_event_attr *attr = &pos->attr;
diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c
index 91a6eba000f2..b72b12aedd29 100644
--- a/tools/perf/builtin-test.c
+++ b/tools/perf/builtin-test.c
@@ -1083,7 +1083,7 @@ static int test__PERF_RECORD(void)
evsel->attr.sample_type |= PERF_SAMPLE_CPU;
evsel->attr.sample_type |= PERF_SAMPLE_TID;
evsel->attr.sample_type |= PERF_SAMPLE_TIME;
- perf_evlist__config_attrs(evlist, &opts);
+ perf_evlist__config_attrs(evlist, &opts, evsel);
err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask,
&cpu_mask_size);
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index aa0418034d82..7ab81906c1aa 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -49,7 +49,8 @@ struct perf_evlist *perf_evlist__new(struct cpu_map *cpus,
}
void perf_evlist__config_attrs(struct perf_evlist *evlist,
- struct perf_record_opts *opts)
+ struct perf_record_opts *opts,
+ struct perf_evsel *first)
{
struct perf_evsel *evsel;
@@ -57,7 +58,7 @@ void perf_evlist__config_attrs(struct perf_evlist *evlist,
opts->no_inherit = true;
list_for_each_entry(evsel, &evlist->entries, node) {
- perf_evsel__config(evsel, opts);
+ perf_evsel__config(evsel, opts, first);
if (evlist->nr_entries > 1)
evsel->attr.sample_type |= PERF_SAMPLE_ID;
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 7e99ac513893..7ebb6419e8d0 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -81,7 +81,8 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *self, int idx);
int perf_evlist__open(struct perf_evlist *evlist, bool group);
void perf_evlist__config_attrs(struct perf_evlist *evlist,
- struct perf_record_opts *opts);
+ struct perf_record_opts *opts,
+ struct perf_evsel *first);
int perf_evlist__prepare_workload(struct perf_evlist *evlist,
struct perf_record_opts *opts,
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index e7ddf27f240b..cb5c18b66d06 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -63,7 +63,8 @@ struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
return evsel;
}
-void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts)
+void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts,
+ struct perf_evsel *first)
{
struct perf_event_attr *attr = &evsel->attr;
int track = !evsel->idx; /* only the first counter needs these */
@@ -130,7 +131,7 @@ void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts)
attr->mmap = track;
attr->comm = track;
- if (no_target_maps(&opts->maps)) {
+ if (no_target_maps(&opts->maps) && (!opts->group || evsel == first)) {
attr->disabled = 1;
attr->enable_on_exec = 1;
}
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 326b8e4d5035..3158ca3d69a1 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -80,7 +80,8 @@ void perf_evsel__exit(struct perf_evsel *evsel);
void perf_evsel__delete(struct perf_evsel *evsel);
void perf_evsel__config(struct perf_evsel *evsel,
- struct perf_record_opts *opts);
+ struct perf_record_opts *opts,
+ struct perf_evsel *first);
int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads);
--
1.7.9
Em Mon, Feb 13, 2012 at 04:27:43PM +0900, Namhyung Kim escreveu:
> When event group is enabled for forked task (i.e. no target task/cpu
> was specified) all events were disabled and marked ->enable_on_exec.
> However they wouldn't be counted at all since only group leader will
> be enabled on exec actually.
>
> In contrast to perf stat, perf record doesn't have a real problem
> as it enables all the event before proceeding. But it needs to be
> fixed anyway IMHO.
>
> Signed-off-by: Namhyung Kim <[email protected]>
> ---
> tools/perf/builtin-record.c | 2 +-
> tools/perf/builtin-test.c | 2 +-
> tools/perf/util/evlist.c | 5 +++--
> tools/perf/util/evlist.h | 3 ++-
> tools/perf/util/evsel.c | 5 +++--
> tools/perf/util/evsel.h | 3 ++-
> 6 files changed, 12 insertions(+), 8 deletions(-)
>
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index fbf71edd31ba..a65b0a5a81ea 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -183,7 +183,7 @@ static void perf_record__open(struct perf_record *rec)
>
> first = list_entry(evlist->entries.next, struct perf_evsel, node);
>
> - perf_evlist__config_attrs(evlist, opts);
> + perf_evlist__config_attrs(evlist, opts, first);
No need to pass the first entry in the evlist, since we can find it at
perf_evlist__config_attrs using:
list_entry(evlist->entries.next, struct perf_evsel, node);
That way you reduce the patch size by not touching all the callers of
perf_evlist__config_attrs.
- Arnaldo
> list_for_each_entry(pos, &evlist->entries, node) {
> struct perf_event_attr *attr = &pos->attr;
> diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c
> index 91a6eba000f2..b72b12aedd29 100644
> --- a/tools/perf/builtin-test.c
> +++ b/tools/perf/builtin-test.c
> @@ -1083,7 +1083,7 @@ static int test__PERF_RECORD(void)
> evsel->attr.sample_type |= PERF_SAMPLE_CPU;
> evsel->attr.sample_type |= PERF_SAMPLE_TID;
> evsel->attr.sample_type |= PERF_SAMPLE_TIME;
> - perf_evlist__config_attrs(evlist, &opts);
> + perf_evlist__config_attrs(evlist, &opts, evsel);
>
> err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask,
> &cpu_mask_size);
> diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
> index aa0418034d82..7ab81906c1aa 100644
> --- a/tools/perf/util/evlist.c
> +++ b/tools/perf/util/evlist.c
> @@ -49,7 +49,8 @@ struct perf_evlist *perf_evlist__new(struct cpu_map *cpus,
> }
>
> void perf_evlist__config_attrs(struct perf_evlist *evlist,
> - struct perf_record_opts *opts)
> + struct perf_record_opts *opts,
> + struct perf_evsel *first)
> {
> struct perf_evsel *evsel;
>
> @@ -57,7 +58,7 @@ void perf_evlist__config_attrs(struct perf_evlist *evlist,
> opts->no_inherit = true;
>
> list_for_each_entry(evsel, &evlist->entries, node) {
> - perf_evsel__config(evsel, opts);
> + perf_evsel__config(evsel, opts, first);
>
> if (evlist->nr_entries > 1)
> evsel->attr.sample_type |= PERF_SAMPLE_ID;
> diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
> index 7e99ac513893..7ebb6419e8d0 100644
> --- a/tools/perf/util/evlist.h
> +++ b/tools/perf/util/evlist.h
> @@ -81,7 +81,8 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *self, int idx);
> int perf_evlist__open(struct perf_evlist *evlist, bool group);
>
> void perf_evlist__config_attrs(struct perf_evlist *evlist,
> - struct perf_record_opts *opts);
> + struct perf_record_opts *opts,
> + struct perf_evsel *first);
>
> int perf_evlist__prepare_workload(struct perf_evlist *evlist,
> struct perf_record_opts *opts,
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index e7ddf27f240b..cb5c18b66d06 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -63,7 +63,8 @@ struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
> return evsel;
> }
>
> -void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts)
> +void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts,
> + struct perf_evsel *first)
> {
> struct perf_event_attr *attr = &evsel->attr;
> int track = !evsel->idx; /* only the first counter needs these */
> @@ -130,7 +131,7 @@ void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts)
> attr->mmap = track;
> attr->comm = track;
>
> - if (no_target_maps(&opts->maps)) {
> + if (no_target_maps(&opts->maps) && (!opts->group || evsel == first)) {
> attr->disabled = 1;
> attr->enable_on_exec = 1;
> }
> diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
> index 326b8e4d5035..3158ca3d69a1 100644
> --- a/tools/perf/util/evsel.h
> +++ b/tools/perf/util/evsel.h
> @@ -80,7 +80,8 @@ void perf_evsel__exit(struct perf_evsel *evsel);
> void perf_evsel__delete(struct perf_evsel *evsel);
>
> void perf_evsel__config(struct perf_evsel *evsel,
> - struct perf_record_opts *opts);
> + struct perf_record_opts *opts,
> + struct perf_evsel *first);
>
> int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
> int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads);
> --
> 1.7.9