Re: [PATCH 08/44] perf session: Add hooks to allow transparent decoding of AUX area tracing data

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Tue, Apr 21, 2015 at 11:41:54AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Thu, Apr 09, 2015 at 06:53:48PM +0300, Adrian Hunter escreveu:
> > Hook into session processing so that AUX area decoding can
> > synthesize events transparently to the tools.
> > 
> > The advantages of transparent decoding are that tools can be used
> > directly with perf.data files containing AUX area tracing data,
> > which is easier for the user and more efficient than having a
> > separate decoding tool.
> > 
> > This will work as follows:
> > 1. Tools will feed auxtrace events to the decoder using
> > perf_tool->auxtrace() (support for that still to come).
> > 2. The decoder can process side-band events as needed due
> > to the auxtrace->process_event() hook.
> > 3. The decoder can deliver synthesized events into the
> > event stream using perf_session__deliver_synth_event().
> 
> So, here this clashes with work I'm doing to untie ordered_events from
> perf_session, undoes it even, trying to decipher if this is really
> needed...

Sorry, gut reaction, it only touches perf_session code, it can be left
for later to figure out if that auxtrace fits better in 'struct
perf_evlist' or even 'struct machine', as perf_session is about a
'perf.data' file, i.e. post processing.

But I think that processing intel PT events _as they are generated_ (in
fact as fast as possible, as there is buffering, etc), i.e. a la 'perf
trace' or 'perf top' will be difficult/impossible, given its volume...
But perhaps with some filtering/narrowing... anyway, will leave this for
later, applying this patch.

- Arnaldo
 
> - Arnaldo
>  
> > Note the expectation is that decoding will work on data that is
> > time-ordered with respect to the per-cpu or per-thread contexts
> > that were recorded.
>  
> > Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
> > ---
> >  tools/perf/util/auxtrace.h | 55 ++++++++++++++++++++++++++++++++++++++++++++++
> >  tools/perf/util/session.c  | 54 ++++++++++++++++++++++++++++++++++++---------
> >  tools/perf/util/session.h  |  3 +++
> >  3 files changed, 102 insertions(+), 10 deletions(-)
> > 
> > diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
> > index 7ab4850..199fc27 100644
> > --- a/tools/perf/util/auxtrace.h
> > +++ b/tools/perf/util/auxtrace.h
> > @@ -23,6 +23,7 @@
> >  #include <linux/types.h>
> >  
> >  #include "../perf.h"
> > +#include "session.h"
> >  
> >  union perf_event;
> >  struct perf_session;
> > @@ -32,6 +33,24 @@ struct record_opts;
> >  struct auxtrace_info_event;
> >  
> >  /**
> > + * struct auxtrace - session callbacks to allow AUX area data decoding.
> > + * @process_event: lets the decoder see all session events
> > + * @flush_events: process any remaining data
> > + * @free_events: free resources associated with event processing
> > + * @free: free resources associated with the session
> > + */
> > +struct auxtrace {
> > +	int (*process_event)(struct perf_session *session,
> > +			     union perf_event *event,
> > +			     struct perf_sample *sample,
> > +			     struct perf_tool *tool);
> > +	int (*flush_events)(struct perf_session *session,
> > +			    struct perf_tool *tool);
> > +	void (*free_events)(struct perf_session *session);
> > +	void (*free)(struct perf_session *session);
> > +};
> > +
> > +/**
> >   * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
> >   * @base: address of mapped area
> >   * @userpg: pointer to buffer's perf_event_mmap_page
> > @@ -168,4 +187,40 @@ int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
> >  					 struct perf_session *session,
> >  					 perf_event__handler_t process);
> >  
> > +static inline int auxtrace__process_event(struct perf_session *session,
> > +					  union perf_event *event,
> > +					  struct perf_sample *sample,
> > +					  struct perf_tool *tool)
> > +{
> > +	if (!session->auxtrace)
> > +		return 0;
> > +
> > +	return session->auxtrace->process_event(session, event, sample, tool);
> > +}
> > +
> > +static inline int auxtrace__flush_events(struct perf_session *session,
> > +					 struct perf_tool *tool)
> > +{
> > +	if (!session->auxtrace)
> > +		return 0;
> > +
> > +	return session->auxtrace->flush_events(session, tool);
> > +}
> > +
> > +static inline void auxtrace__free_events(struct perf_session *session)
> > +{
> > +	if (!session->auxtrace)
> > +		return;
> > +
> > +	return session->auxtrace->free_events(session);
> > +}
> > +
> > +static inline void auxtrace__free(struct perf_session *session)
> > +{
> > +	if (!session->auxtrace)
> > +		return;
> > +
> > +	return session->auxtrace->free(session);
> > +}
> > +
> >  #endif
> > diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
> > index 18e4034..cf9d678 100644
> > --- a/tools/perf/util/session.c
> > +++ b/tools/perf/util/session.c
> > @@ -15,12 +15,13 @@
> >  #include "cpumap.h"
> >  #include "perf_regs.h"
> >  #include "asm/bug.h"
> > +#include "auxtrace.h"
> >  
> > -static int machines__deliver_event(struct machines *machines,
> > -				   struct perf_evlist *evlist,
> > -				   union perf_event *event,
> > -				   struct perf_sample *sample,
> > -				   struct perf_tool *tool, u64 file_offset);
> > +static int perf_session__deliver_event(struct perf_session *session,
> > +				       union perf_event *event,
> > +				       struct perf_sample *sample,
> > +				       struct perf_tool *tool,
> > +				       u64 file_offset);
> >  
> >  static int perf_session__open(struct perf_session *session)
> >  {
> > @@ -105,8 +106,8 @@ static int ordered_events__deliver_event(struct ordered_events *oe,
> >  		return ret;
> >  	}
> >  
> > -	return machines__deliver_event(&session->machines, session->evlist, event->event,
> > -				       &sample, session->tool, event->file_offset);
> > +	return perf_session__deliver_event(session, event->event, &sample,
> > +					   session->tool, event->file_offset);
> >  }
> >  
> >  struct perf_session *perf_session__new(struct perf_data_file *file,
> > @@ -185,6 +186,7 @@ static void perf_session_env__delete(struct perf_session_env *env)
> >  
> >  void perf_session__delete(struct perf_session *session)
> >  {
> > +	auxtrace__free(session);
> >  	perf_session__destroy_kernel_maps(session);
> >  	perf_session__delete_threads(session);
> >  	perf_session_env__delete(&session->header.env);
> > @@ -1030,6 +1032,24 @@ static int machines__deliver_event(struct machines *machines,
> >  	}
> >  }
> >  
> > +static int perf_session__deliver_event(struct perf_session *session,
> > +				       union perf_event *event,
> > +				       struct perf_sample *sample,
> > +				       struct perf_tool *tool,
> > +				       u64 file_offset)
> > +{
> > +	int ret;
> > +
> > +	ret = auxtrace__process_event(session, event, sample, tool);
> > +	if (ret < 0)
> > +		return ret;
> > +	if (ret > 0)
> > +		return 0;
> > +
> > +	return machines__deliver_event(&session->machines, session->evlist,
> > +				       event, sample, tool, file_offset);
> > +}
> > +
> >  static s64 perf_session__process_user_event(struct perf_session *session,
> >  					    union perf_event *event,
> >  					    u64 file_offset)
> > @@ -1190,8 +1210,8 @@ static s64 perf_session__process_event(struct perf_session *session,
> >  			return ret;
> >  	}
> >  
> > -	return machines__deliver_event(&session->machines, evlist, event,
> > -				       &sample, tool, file_offset);
> > +	return perf_session__deliver_event(session, event, &sample, tool,
> > +					   file_offset);
> >  }
> >  
> >  void perf_event_header__bswap(struct perf_event_header *hdr)
> > @@ -1350,10 +1370,14 @@ more:
> >  done:
> >  	/* do the final flush for ordered samples */
> >  	err = ordered_events__flush(oe, OE_FLUSH__FINAL);
> > +	if (err)
> > +		goto out_err;
> > +	err = auxtrace__flush_events(session, tool);
> >  out_err:
> >  	free(buf);
> >  	perf_session__warn_about_errors(session);
> >  	ordered_events__free(&session->ordered_events);
> > +	auxtrace__free_events(session);
> >  	return err;
> >  }
> >  
> > @@ -1496,10 +1520,14 @@ more:
> >  out:
> >  	/* do the final flush for ordered samples */
> >  	err = ordered_events__flush(oe, OE_FLUSH__FINAL);
> > +	if (err)
> > +		goto out_err;
> > +	err = auxtrace__flush_events(session, tool);
> >  out_err:
> >  	ui_progress__finish();
> >  	perf_session__warn_about_errors(session);
> >  	ordered_events__free(&session->ordered_events);
> > +	auxtrace__free_events(session);
> >  	session->one_mmap = false;
> >  	return err;
> >  }
> > @@ -1582,7 +1610,13 @@ size_t perf_session__fprintf_dsos_buildid(struct perf_session *session, FILE *fp
> >  
> >  size_t perf_session__fprintf_nr_events(struct perf_session *session, FILE *fp)
> >  {
> > -	size_t ret = fprintf(fp, "Aggregated stats:\n");
> > +	size_t ret;
> > +	const char *msg = "";
> > +
> > +	if (perf_header__has_feat(&session->header, HEADER_AUXTRACE))
> > +		msg = " (excludes AUX area (e.g. instruction trace) decoded / synthesized events)";
> > +
> > +	ret = fprintf(fp, "Aggregated stats:%s\n", msg);
> >  
> >  	ret += events_stats__fprintf(&session->evlist->stats, fp);
> >  	return ret;
> > diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
> > index d5fa7b791..8a69d3b 100644
> > --- a/tools/perf/util/session.h
> > +++ b/tools/perf/util/session.h
> > @@ -15,10 +15,13 @@
> >  struct ip_callchain;
> >  struct thread;
> >  
> > +struct auxtrace;
> > +
> >  struct perf_session {
> >  	struct perf_header	header;
> >  	struct machines		machines;
> >  	struct perf_evlist	*evlist;
> > +	struct auxtrace		*auxtrace;
> >  	struct trace_event	tevent;
> >  	bool			repipe;
> >  	bool			one_mmap;
> > -- 
> > 1.9.1