Re: perf: multiplexing broken on amd fam16h?

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Jul 22, 2015 at 05:06:24PM +0200, Peter Zijlstra wrote:
> On Wed, Jul 22, 2015 at 11:02:21AM -0400, Vince Weaver wrote:
> > On Wed, 22 Jul 2015, Vince Weaver wrote:
> > > 
> > > OK, so I have to admit I'm a bit weak on ftrace.  What type of command 
> > > line are you interested in seeing?  The ones I've tried with trace-cmd
> > > have  activated  your trace_printks but there's 500MB of other data mixed 
> > > in so maybe I am doing something wrong.
> > 
> > maybe the "nop" plugin was what I wanted.  I've posted the trace with that 
> > here:
> > 
> > http://web.eece.maine.edu/~vweaver/junk/jaguar.mpx.out.gz
> 
> OK, that looks about right, lemme to stare at that.

Does this help?

---
 kernel/events/core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index d3dae3419b99..10d076b2572c 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -1868,8 +1868,6 @@ event_sched_in(struct perf_event *event,
 
 	perf_pmu_disable(event->pmu);
 
-	event->tstamp_running += tstamp - event->tstamp_stopped;
-
 	perf_set_shadow_time(event, ctx, tstamp);
 
 	perf_log_itrace_start(event);
@@ -1881,6 +1879,8 @@ event_sched_in(struct perf_event *event,
 		goto out;
 	}
 
+	event->tstamp_running += tstamp - event->tstamp_stopped;
+
 	if (!is_software_event(event))
 		cpuctx->active_oncpu++;
 	if (!ctx->nr_active++)