[ltt-dev] LTTng roadmap (updated)

[ltt-dev] LTTng roadmap (updated)

[Mathieu Desnoyers]

Well well, let's see what we have left to do before releasing LTTng to
LKML :

- Add LTTng "streaming" feature (buffer switch at periodical interval)
  - Should make the default number of subbuffers larger for channels
    which implement this kind of periodical subbuffer switch.
- Make sure lttd only send the data (without padding) to disk to make
  sure we don't end up with large empty traces in in streaming mode.
- Modify LTTV so it supports variable-sized buffers (create an index).
- Create ltt-ascii.ko
  - Note : the last lttng version implements the open()/close() refcount
    that will be used by an in-kernel interface for ltt-ascii.
- Patchset cleanup, patch folding
  Mathieu Desnoyers <mathieu.desnoyers at polymtl.ca>

Mathieu

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

[ltt-dev] LTTng roadmap (updated)

[Jason Baron]

On Mon, Dec 15, 2008 at 06:59:12PM -0500, Mathieu Desnoyers wrote:
> Well well, let's see what we have left to do before releasing LTTng to
> LKML :
> 

hi,

I find these roadmap updates quite helpful, thanks.

I was just wondering what the merge plans were for 2.6.29 merge window?
Specifically, I'd (and I think others as well) are interested in the tracepoints 
themselves, as found in include/trace/*.h and their respective source file 
locations. In addition to lttng, Systemtap, and even simple ftrace
widgets would make use of these.

thanks,

-Jason

[ltt-dev] LTTng roadmap (updated)

[Mathieu Desnoyers]

* Jason Baron (jbaron at redhat.com) wrote:
> On Mon, Dec 15, 2008 at 06:59:12PM -0500, Mathieu Desnoyers wrote:
> > Well well, let's see what we have left to do before releasing LTTng to
> > LKML :
> > 
> 
> hi,
> 
> I find these roadmap updates quite helpful, thanks.
> 
> I was just wondering what the merge plans were for 2.6.29 merge window?
> Specifically, I'd (and I think others as well) are interested in the tracepoints 
> themselves, as found in include/trace/*.h and their respective source file 
> locations. In addition to lttng, Systemtap, and even simple ftrace
> widgets would make use of these.
> 

Hi Jason,

Linus has been very clear at OLS on that he does not want to merge
instrumentation if there is no tool that does *something* useful with
it.

AFAIK, Systemtap does not support getting tracepoints fields, am I
correct ?

We already merged the sched.h tracepoints because we have simple ftrace
widgets.

LTTng consists of many probe modules which expand the tracepoint
parameters into meaningful event data. They can also be the location for
probe-specific filtering (see the ext4 probe I just did for an example).
Those tracepoints have no other known users and therefore I plan to
submit the LTTng probing/buffering/trace control infrastructure before I
submit the tracepoints.

Best regards,

Mathieu


> thanks,
> 
> -Jason
> 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68