This is the mail archive of the
mailing list for the systemtap project.
Re: Issues with probing multiple modules in script
- From: fche at redhat dot com (Frank Ch. Eigler)
- To: "Klein, Matt" <mklein at amazon dot com>
- Cc: "systemtap at sourceware dot org" <systemtap at sourceware dot org>
- Date: Wed, 28 Dec 2011 12:08:32 -0500
- Subject: Re: Issues with probing multiple modules in script
- References: <FCDCC4AE6C28B44B90993D9894683575B35B07D5@EX-SEA3-A.ant.amazon.com>
> SystemTap translator/driver (version 1.3/0.152 non-git sources)
That's a little old, but probably not a factor.
> If I try to probe a single module, everything works as
> expected. However if I probe multiple modules, it seems like probes
> in the first module found works fine, but any probes on subsequent
> modules don't actually work. I also see the following error for all
> modules other than the first:
> WARNING: missing unwind/symbol data for module ...
This message indicates only that operations such as symname() /
backtrace() might not be able to cover those other modules. (This
information is extracted during pass 3, in translate.cxx
dump_unwindsyms().) There may be a bug here, but I couldn't reproduce
the problem here. Could you share the "stap --vp 009 -p4 ..." output?
The probes' basic activity should be otherwise unaffected, meaning
that you should be able to place probes whereever.
> I have looked through the -vvvvv output, and it looks to me like all
> of the module probe points are being identified correctly, but then
> I see the WARNING message and it doesn't look like the probes ever
> actually get invoked.
Why do you think they're not getting invoked? You can try "stap -t
..." to get hit counts/times for individual probe points. You could
try "stap -DDEBUG_SYMBOLS ..." to see some of the runtime
> Is there anything obvious I am doing wrong (like probing multiple
> modules from the same script doesn't work)? [...]
It should all work smoothly. We just need more information to figure
out what's going wrong.