Messages in this thread | | | Date | Tue, 18 Jun 2013 16:46:35 +0200 | From | Oleg Nesterov <> | Subject | Re: [PATCH 0/3] tracing: more list_empty(perf_events) checks |
| |
On 06/17, Steven Rostedt wrote: > > On Mon, 2013-06-17 at 22:18 +0200, Oleg Nesterov wrote: > > On 06/17, Oleg Nesterov wrote: > > > > > > DECLARE_EVENT_CLASS()->perf_trace_##call() is not trivial because > > > of __perf_task() > > > > Perhaps we can do something like below? > > Did this actually compile for you?
Why did you ask?
Perhaps you are trying to say that this patch needs more work...
Just because it can't be compiled? Pedant.
> > @@ -659,13 +665,12 @@ perf_trace_##call(void *__data, proto) \ > > int __data_size; \ > > int rctx; \ > > \ > > - perf_fetch_caller_regs(&__regs); \ > > - \ > > __data_size = ftrace_get_offsets_##call(&__data_offsets, args); \ > > OK, so here the task gets assigned the val, and so does count. > > This may not be a bad approach, but instead of having TP_perf_arg() in > events/sched.h, keep the TP_perf_task() and TP_perf_count(), and have > whatever is put there assigned.
Or this, yes.
OK. Let me try to make something working. At least, something I believe should work, I will mostly rely on your review anyway.
Oleg.
| |