Re: [PATCH v4 03/15] perf tools script: Filter COMM/FORK/.. events by CPU

From: Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com>
To: Jiri Olsa <jolsa@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>,
	Andi Kleen <andi@firstfloor.org>,
	jolsa@kernel.org, namhyung@kernel.org,
	linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
	Andi Kleen <ak@linux.intel.com>
Subject: Re: [PATCH v4 03/15] perf tools script: Filter COMM/FORK/.. events by CPU
Date: Fri, 8 Mar 2019 10:39:01 -0300	[thread overview]
Message-ID: <20190308133901.GB26260@kernel.org> (raw)
In-Reply-To: <20190307110231.GB8979@krava>

Em Thu, Mar 07, 2019 at 12:02:31PM +0100, Jiri Olsa escreveu:
> On Wed, Mar 06, 2019 at 04:23:40PM +0200, Adrian Hunter wrote:
> > On 5/03/19 4:47 PM, Andi Kleen wrote:
> > > From: Andi Kleen <ak@linux.intel.com>
> > > 
> > > The --cpu option only filtered samples. Filter other perf events,
> > > such as COMM, FORK, SWITCH by the CPU too.
> > 
> > Because tasks can migrate from cpu to cpu, we probably need to process most
> > of the events anyway, even if they are not printed.
> 
> agreed, I wonder we could just make the perf_event__fprintf conditional

Humm, probably just do the filtering on PERF_RECORD_SAMPLE is enough?
I.e. having the other PERF_RECORD_{COMM,MMAP,} etc is required in face
of migration.

- Arnaldo

> jirka
> 
> > 
> > > 
> > > Reported-by: Jiri Olsa <jolsa@kernel.org>
> > > Signed-off-by: Andi Kleen <ak@linux.intel.com>
> > > ---
> > >  tools/perf/builtin-script.c | 24 ++++++++++++++++++++++++
> > >  1 file changed, 24 insertions(+)
> > > 
> > > diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
> > > index fbc440bdf880..3813f60d1dc0 100644
> > > --- a/tools/perf/builtin-script.c
> > > +++ b/tools/perf/builtin-script.c
> > > @@ -2038,6 +2038,9 @@ static int process_comm_event(struct perf_tool *tool,
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  	int ret = -1;
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	thread = machine__findnew_thread(machine, event->comm.pid, event->comm.tid);
> > >  	if (thread == NULL) {
> > >  		pr_debug("problem processing COMM event, skipping it.\n");
> > > @@ -2073,6 +2076,9 @@ static int process_namespaces_event(struct perf_tool *tool,
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  	int ret = -1;
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	thread = machine__findnew_thread(machine, event->namespaces.pid,
> > >  					 event->namespaces.tid);
> > >  	if (thread == NULL) {
> > > @@ -2108,6 +2114,9 @@ static int process_fork_event(struct perf_tool *tool,
> > >  	struct perf_session *session = script->session;
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	if (perf_event__process_fork(tool, event, sample, machine) < 0)
> > >  		return -1;
> > >  
> > > @@ -2141,6 +2150,9 @@ static int process_exit_event(struct perf_tool *tool,
> > >  	struct perf_session *session = script->session;
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	thread = machine__findnew_thread(machine, event->fork.pid, event->fork.tid);
> > >  	if (thread == NULL) {
> > >  		pr_debug("problem processing EXIT event, skipping it.\n");
> > > @@ -2174,6 +2186,9 @@ static int process_mmap_event(struct perf_tool *tool,
> > >  	struct perf_session *session = script->session;
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	if (perf_event__process_mmap(tool, event, sample, machine) < 0)
> > >  		return -1;
> > >  
> > > @@ -2206,6 +2221,9 @@ static int process_mmap2_event(struct perf_tool *tool,
> > >  	struct perf_session *session = script->session;
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	if (perf_event__process_mmap2(tool, event, sample, machine) < 0)
> > >  		return -1;
> > >  
> > > @@ -2238,6 +2256,9 @@ static int process_switch_event(struct perf_tool *tool,
> > >  	struct perf_session *session = script->session;
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	if (perf_event__process_switch(tool, event, sample, machine) < 0)
> > >  		return -1;
> > >  
> > > @@ -2266,6 +2287,9 @@ process_lost_event(struct perf_tool *tool,
> > >  	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
> > >  	struct thread *thread;
> > >  
> > > +	if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
> > > +		return 0;
> > > +
> > >  	thread = machine__findnew_thread(machine, sample->pid,
> > >  					 sample->tid);
> > >  	if (thread == NULL)
> > > 
> > 

-- 

- Arnaldo