Re: [PATCH 08/12] add trace events for each syscall entry/exit

[Frederic Weisbecker <fweisbec@gmail.com>]

On Tue, Aug 11, 2009 at 01:45:12PM +0200, Ingo Molnar wrote:
> 
> * Frederic Weisbecker <fweisbec@gmail.com> wrote:
> 
> > > +struct trace_event event_syscall_enter = {
> > > +	.trace			= print_syscall_enter,
> > > +	.type			= TRACE_SYSCALL_ENTER
> > > +};
> > > +
> > > +struct trace_event event_syscall_exit = {
> > > +	.trace			= print_syscall_exit,
> > > +	.type			= TRACE_SYSCALL_EXIT
> > > +};
> > > -- 
> > > 1.6.2.5
> > > 
> > 
> > Nice.
> > 
> > It's a bit too bad that enter and exit must be that separated 
> > whereas their callbacks are pretty the same.
> > 
> > But I guess if we want to nicely decouple both, we don't have the 
> > choice.
> 
> Yeah - and enter and exit are different, in terms of state.
> 
> One thing that would be nice in the future (as an add-on - this 
> patch-set looks useful already) is to allow the sampling of user 
> register state as well via these tracepoints. That way we'd have a 
> much faster (and completely transparent) implementation of strace in 
> essence, with unique features such as system-wide or per cpu 
> strace-ing.
> 
> 	Ingo


Indeed, a missing piece is the syscall record sampling with registers.
Actually, IMO the registers themselves are not the right piece to export
to perfcounter. It's too low-level.

What we need are the fetched arguments, because a lot of them are adresses
(even user adresses), pretty useless for perf tools.

We can already and easily implement that simple args, like we do for ftrace.
That's pretty trivial.