Re: ye olde task_ctx_sched_out trace.

[Adrian Hunter <adrian.hunter@intel.com>]

On 05/22/2014 10:04 AM, Peter Zijlstra wrote:
> On Thu, May 22, 2014 at 09:52:46AM +0300, Adrian Hunter wrote:
>>> diff --git a/fs/exec.c b/fs/exec.c
>>> index 476f3ebf437e..8d51d7ce3dcf 100644
>>> --- a/fs/exec.c
>>> +++ b/fs/exec.c
>>> @@ -1111,6 +1111,7 @@ void setup_new_exec(struct linux_binprm * bprm)
>>>  		set_dumpable(current->mm, suid_dumpable);
>>>  
>>>  	set_task_comm(current, kbasename(bprm->filename));
>>> +	perf_event_exec();
>>
>> Shouldn't that be the other way around i.e.
>>
>> +	perf_event_exec();
>> 	set_task_comm(current, kbasename(bprm->filename));
> 
> I suppose so indeed.
> 
>> Also what about flagging the comm event that corresponds to an exec e.g.
> 
> I think it was a mistake to conflate the two concepts, and separating
> them into different functions makes things clearer.

My patch was not related to that.  It was to get effectively an "exec"
event, by piggybacking the comm event.