Re: [PATCH 1/2] tracing/sysprof: add missing tracing_{start,stop}_record_cmdline()

[Frederic Weisbecker <fweisbec@gmail.com>]

On Tue, Feb 10, 2009 at 10:54:24AM -0200, Arnaldo Carvalho de Melo wrote:
> Em Tue, Feb 10, 2009 at 04:43:30AM +0100, Frederic Weisbecker escreveu:
> > Add the missing pair tracing_{start,stop}_record_cmdline() to record well
> > the cmdline associated with pid.
> > 
> > Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> 
> This is another thing that I think begs for a TRACE_INIT_CMDLINE in
> struct trace->flags so that just before calling ->init we call
> tracing_start_cmdline_record? no?


Right!
But the flags on struct tracer are more likely for private use, they have
a meaning only for the concerned tracer.

I guess it will be something for the global flags, when they will be
per-tracer.

 
> Another thing I noticed is that even we having a ->stop most of the time
> this semantic action is done thru ->reset :-\


Yes, perhaps the tracing of cmdline could be even performed on tracing_stop,
unless... perhaps it's not needed since the sched_switch tracer will be stopped too
anyway.

> 
> - Arnaldo
> 
> >  kernel/trace/trace_sysprof.c |    3 +++
> >  1 files changed, 3 insertions(+), 0 deletions(-)
> > 
> > diff --git a/kernel/trace/trace_sysprof.c b/kernel/trace/trace_sysprof.c
> > index 84ca9d8..9902c15 100644
> > --- a/kernel/trace/trace_sysprof.c
> > +++ b/kernel/trace/trace_sysprof.c
> > @@ -238,6 +238,8 @@ static int stack_trace_init(struct trace_array *tr)
> >  {
> >  	sysprof_trace = tr;
> >  
> > +	tracing_start_cmdline_record();
> > +
> >  	mutex_lock(&sample_timer_lock);
> >  	start_stack_timers();
> >  	tracer_enabled = 1;
> > @@ -247,6 +249,7 @@ static int stack_trace_init(struct trace_array *tr)
> >  
> >  static void stack_trace_reset(struct trace_array *tr)
> >  {
> > +	tracing_stop_cmdline_record();
> >  	stop_stack_trace(tr);
> >  }