Re: [RFC][PATCH 1/4] perf: core, add group scheduling transactional APIs

[Lin Ming <ming.m.lin@intel.com>]

On Fri, 2010-04-23 at 01:31 +0800, Frederic Weisbecker wrote:
> On Thu, Apr 22, 2010 at 03:51:02PM +0800, Lin Ming wrote:
> > Add group scheduling transactional APIs to struct pmu.
> > These APIs will be implemented in arch code, based on Peter's idea as
> > below.
> > 
> > > the idea behind hw_perf_group_sched_in() is to not perform
> > > schedulability tests on each event in the group, but to add the group
> > as
> > > a whole and then perform one test.
> > >
> > > Of course, when that test fails, you'll have to roll-back the whole
> > > group again.
> > >
> > > So start_txn (or a better name) would simply toggle a flag in the pmu
> > > implementation that will make pmu::enable() not perform the
> > > schedulablilty test.
> > >
> > > Then commit_txn() will perform the schedulability test (so note the
> > > method has to have a !void return value, my mistake in the earlier
> > > email).
> > >
> > > This will allow us to use the regular
> > > kernel/perf_event.c::group_sched_in() and all the rollback code.
> > > Currently each hw_perf_group_sched_in() implementation duplicates all
> > > the rolllback code (with various bugs).
> > 
> > 
> > Reviewed-by: Stephane Eranian <eranian@google.com>
> > Reviewed-by: Peter Zijlstra <peterz@infradead.org>
> > Signed-off-by: Lin Ming <ming.m.lin@intel.com>
> > ---
> >  include/linux/perf_event.h |    8 +++++---
> >  kernel/perf_event.c        |   29 ++++++++++++++++-------------
> >  2 files changed, 21 insertions(+), 16 deletions(-)
> > 
> > diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
> > index ace31fb..b16cfba 100644
> > --- a/include/linux/perf_event.h
> > +++ b/include/linux/perf_event.h
> > @@ -532,6 +532,8 @@ struct hw_perf_event {
> >  
> >  struct perf_event;
> >  
> > +#define PERF_EVENT_TRAN_STARTED 1
> > +
> >  /**
> >   * struct pmu - generic performance monitoring unit
> >   */
> > @@ -542,6 +544,9 @@ struct pmu {
> >  	void (*stop)			(struct perf_event *event);
> >  	void (*read)			(struct perf_event *event);
> >  	void (*unthrottle)		(struct perf_event *event);
> > +	void (*start_txn)		(const struct pmu *pmu);
> > +	void (*stop_txn)		(const struct pmu *pmu);
> > +	int (*commit_txn)		(const struct pmu *pmu);
> 
> 
> Please add a few comments that briefly explain what these
> *_txn callbacks are supposed to mean.
> 
> Unless txn is an acronym that most kernel developers are used to.

How about below changes?

Thanks for review.

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index b16cfba..bba4c60 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -544,9 +544,21 @@ struct pmu {
 	void (*stop)			(struct perf_event *event);
 	void (*read)			(struct perf_event *event);
 	void (*unthrottle)		(struct perf_event *event);
-	void (*start_txn)		(const struct pmu *pmu);
-	void (*stop_txn)		(const struct pmu *pmu);
-	int (*commit_txn)		(const struct pmu *pmu);
+
+	/*
+	 * group events scheduling is treated as a transaction,
+	 * add group events as a whole and perform one schedulability test.
+	 * If test fails, roll back the whole group
+	 */
+
+	/* start group events transaction  */
+	void (*start_group_trans)	(const struct pmu *pmu);
+
+	/* stop group events transaction  */
+	void (*stop_group_trans)	(const struct pmu *pmu);
+
+	/* commit group events transaction */
+	int (*commit_group_trans)	(const struct pmu *pmu);
 };
 
 /**