lkml.org 
[lkml]   [2009]   [Feb]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] tracing/markers: make markers select tracepoints

* Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> wrote:

> * Ingo Molnar (mingo@elte.hu) wrote:
> >
> > * Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> wrote:
> >
> > > > > so I wonder why don't we create linux-tracing new mailing
> > > > > list.
> >
> > Discussions related to upstream Linux tracing are generally
> > conducted on lkml.
> >
> > > > Yes, lets hide it from general view, sounds like a brilliant
> > > > plan.
> > >
> > > I guess Kosaki's point is that there may be a disconnect
> > > between the work that is done on the LTTng side and the ftrace
> > > side. [...]
> >
> > Could you please list the items we need to extend the tracing
> > framework in kernel/tracing/* with so that we can support the
> > aspects of LTTng you consider important and upstream-worthy?
> >
> > As i've mentioned it in the past, i'm definitely interested in
> > helping such an effort to further improve Linux tracing.
> >
> > Ingo
>
> Hi Ingo,
>
> I have not merged the "binary trace buffer to ascii" code done
> by Fujitsu yet because I still have to review it, but when
> it's ready, I will be ready to post everything that has been
> identified as wish list for a tracer v2. It includes a very
> generic buffering infrastructure which have been evolving for
> the past 4 years (and reviewed on LKML a couple of times in
> the past).
>
> So, either I start to merge this last feature, or I can start
> posting right away without the "binary buffer to ascii"
> module. The standard mechanism used in LTTng is binary-only,
> very high speed, and uses a userspace application to transform
> binary data into text.
>
> That will let us figure out what overlaps between ftrace and
> LTTng and find out how those two can fit together.
>
> Note that for the moment, LTTng sits in the /ltt directory of
> the kernel tree, so even if we have both tracers at some
> point, we should not step on each other's foot. That could let
> us room to figure out how to incrementally combine those
> tracers if that makes sense.

I think you got me wrong Mathieu. The way to combine those two
tracers is to do it _before_ it gets merged upstream. We
obviously dont want to parallel tracing frameworks in the
upstream kernel.

So ... please lets work on that, okay?

Ingo


\
 
 \ /
  Last update: 2009-02-23 09:55    [W:0.044 / U:1.240 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site