Realtime Linux: academia v. reality [LWN.net]

Realtime Linux: academia v. reality

Posted Jul 27, 2010 14:35 UTC (Tue) by mjthayer (guest, #39183)
In reply to: Realtime Linux: academia v. reality by nix
Parent article: Realtime Linux: academia v. reality

> systemtap will never ever ever go in until one day it suddenly does.

SystemTap or uprobes? For SystemTap itself, it looks to me as though Frank and friends have worked around requiring anything they can avoid in the kernel itself. And done it pretty well given the stones laid in the path of anyone doing kernel modules outside the official kernel tree.

to post comments

Realtime Linux: academia v. reality

Posted Jul 28, 2010 13:01 UTC (Wed) by nix (subscriber, #2304) [Link] (3 responses)

Yeah, but even now you have people like Christoph Hellwig turning up on e.g. the glibc list when probe point adding patches were proposed, saying 'no, uprobes will never exist in this form'. So uprobes may exist but some kernel hackers are trying as hard as possible to make sure they stay useless and out of *every* package's upstream.

Realtime Linux: academia v. reality

Posted Jul 28, 2010 16:08 UTC (Wed) by mjthayer (guest, #39183) [Link] (2 responses)

> Yeah, but even now you have people like Christoph Hellwig turning up on e.g. the glibc list when probe point adding patches were proposed, saying 'no, uprobes will never exist in this form'.

I thought I read about work to make SystemTap work with procfs. Any idea what is happening there (Frank?) Although something seems to be happening currently with uprobes, and Christoph Hellwig seemed to be actively (positively) involved in the discussion (see http://lkml.org/lkml/2010/7/27/121).

Realtime Linux: academia v. reality

Posted Jul 28, 2010 16:18 UTC (Wed) by fuhchee (guest, #40059) [Link] (1 responses)

> I thought I read about work to make SystemTap work with procfs.
> Any idea what is happening there (Frank?)

I'm not quite sure which effort you may be referring to.

> ... Christoph Hellwig seemed to be actively (positively) involved

Yes. It is unfortunate though that instructions of the form "Don't SYMBOL_EXPORT this facility [since you-know-who might use it]" are still the order of the day.

Realtime Linux: academia v. reality

Posted Jul 28, 2010 16:27 UTC (Wed) by mjthayer (guest, #39183) [Link]

>> I thought I read about work to make SystemTap work with procfs.
>> Any idea what is happening there (Frank?)
> I'm not quite sure which effort you may be referring to.
I can't seem to find any reference to it now. If you are not aware of it then I strongly assume I misunderstood something at some point.