Re: [PATCH][RFC] tracing: Use class->reg() for all registering of events

[Srikar Dronamraju <srikar@linux.vnet.ibm.com>]

> On Tue, 2010-06-08 at 11:33 -0400, Steven Rostedt wrote:
> > I'm pushing this as an RFC first. This probably should be something that
> > makes it into 2.6.35.
> > 
> > Acks and perhaps a little testing from the perf and kprobe angle?
> 
> I'll have a look soon, but lets add Srikar to CC, he actually reported
> the problem :-)
> 
> > Steven Rostedt (1):
> >       tracing: Use class->reg() for all registering of evints
> > 

I tested the patch and it fixed the regression where
perf record -e probe:do_fork -aR sleep 1 would fail. 

Now with this patch, it records the events. However perf record
encounters a floating point exception .. (Peter said he was aware and had a
fix for the floating point exception problem)


However I still see another minor regression (atleast on the tip tree)
This regression was present even before this patch.
The first time I run a perf probe command, it fails, subsequent runs
pass. 

i.e 

411 [srikar@llm69 ]$  sudo perf probe do_fork 
kprobe_events file does not exist - please rebuild kernel with CONFIG_KPROBE_EVENT.
  Error: Failed to add events. (-1)
411 [srikar@llm69 ]$  sudo perf probe do_fork 
Add new event:
  probe:do_fork        (on do_fork)

You can now use it on all perf tools, such as:

        perf record -e probe:do_fork -aR sleep 1

411 [srikar@llm69 ]$ 


--
Thanks and Regards
Srikar