On Wed, 23 Jan 2008, Frank Ch. Eigler wrote:
quoted text
> >
> > Also, the things the latency tracer and latecytop measure are quite
> > different. the latency tracer measures how long a non-schedulable
> > section is, latencytop measures how a task was scheduled away.
>
> The commonality I imagine is not in what sorts of reports they create,
> or where they complete the formatting for and end-user.  It is solely
> whether the kernel-side instrumentation sites can be represented by
> individual markers, so that these particular tools as well as future
> ones don't need to keep inserting custom code in similar or even the
> same places.

So basically just having the marker implementation there to let all
tracers access to irqs enabled/disabled.

One problem is that the hooks are alse in entry.S, so the tracing ends up
being architecture dependent. One solution is to have a "generic" hook for
trace_hardirqs_on/off (kind of like what I did here) and add a marker
inside that. The hooks in entry.S can call the generic function and have
the marker call the registered probes, be that lockdep, latencytap or
latency-trace.

Have the generic hook only be compiled in when one of the tracers or
lockdep is configured on and they select it.

-- Steve

--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/