[PATCH] perf: fix stat attach bogus counts

[PATCH] perf: fix stat attach bogus counts

[Stephane Eranian]

	When perf stat -p pid is used, the events must be enabled
	immediately as there is no exec and thus no enable_on_exec.

	Signed-off-by: Stephane Eranian <eranian@google.com>

--
 builtin-stat.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 95db31c..3a91fad 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -159,8 +159,10 @@ static void create_perf_stat_counter(int counter, int pid)
 		}
 	} else {
 		attr->inherit	     = inherit;
-		attr->disabled	     = 1;
-		attr->enable_on_exec = 1;
+		if (target_pid == -1) {
+			attr->disabled	     = 1;
+			attr->enable_on_exec = 1;
+		}
 
 		fd[0][counter] = sys_perf_event_open(attr, pid, -1, -1, 0);
 		if (fd[0][counter] < 0 && verbose)

Re: [PATCH] perf: fix stat attach bogus counts

[Ingo Molnar]

* Stephane Eranian <eranian@google.com> wrote:

> 	When perf stat -p pid is used, the events must be enabled
> 	immediately as there is no exec and thus no enable_on_exec.
> 
> 	Signed-off-by: Stephane Eranian <eranian@google.com>
> 
> --
>  builtin-stat.c |    6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 95db31c..3a91fad 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -159,8 +159,10 @@ static void create_perf_stat_counter(int counter, int pid)
>  		}
>  	} else {
>  		attr->inherit	     = inherit;
> -		attr->disabled	     = 1;
> -		attr->enable_on_exec = 1;
> +		if (target_pid == -1) {
> +			attr->disabled	     = 1;
> +			attr->enable_on_exec = 1;
> +		}

Mind checking latest -tip, which has these commits:

d6d901c: perf events: Change perf parameter --pid to process-wide collection instead of thread-wid
46be604: perf record: Enable counters only when kernel is execing subcommand
6be2850: perf stat: Enable counters when collecting process-wide or system-wide data

as they fix/improve various --pid related behavioral aspects of perf. Does it 
work fine for you now?

	Ingo

Re: [PATCH] perf: fix stat attach bogus counts

[stephane eranian]

On Thu, Mar 18, 2010 at 6:36 PM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Stephane Eranian <eranian@google.com> wrote:
>
>>       When perf stat -p pid is used, the events must be enabled
>>       immediately as there is no exec and thus no enable_on_exec.
>>
>>       Signed-off-by: Stephane Eranian <eranian@google.com>
>>
>> --
>>  builtin-stat.c |    6 ++++--
>>  1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
>> index 95db31c..3a91fad 100644
>> --- a/tools/perf/builtin-stat.c
>> +++ b/tools/perf/builtin-stat.c
>> @@ -159,8 +159,10 @@ static void create_perf_stat_counter(int counter, int pid)
>>               }
>>       } else {
>>               attr->inherit        = inherit;
>> -             attr->disabled       = 1;
>> -             attr->enable_on_exec = 1;
>> +             if (target_pid == -1) {
>> +                     attr->disabled       = 1;
>> +                     attr->enable_on_exec = 1;
>> +             }
>
> Mind checking latest -tip, which has these commits:
>
> d6d901c: perf events: Change perf parameter --pid to process-wide collection instead of thread-wid
> 46be604: perf record: Enable counters only when kernel is execing subcommand
> 6be2850: perf stat: Enable counters when collecting process-wide or system-wide data
>
> as they fix/improve various --pid related behavioral aspects of perf. Does it
> work fine for you now?
>
They indeed included the same fix.
So that's fine now.