help: semantic error: no match while resolving probe point module

[Frank Ch. Eigler]

Hi, Wendy -


s.wendy.cheng wrote:

> [...]
> I have a system (2.6.32-279.el6.x86_64) where
> /lib/modules/$kversion/kernel/drivers/infinibandel is obsoleted by
> /lib/modules/$kversion/updates/drivers/infiniband
>
> After debuginfo rpm was installed, system-tap seemed to be confused by
> the driver location. So I manually copied the
> /lib/modules/$kversion/updates/drivers/infiniband/ulp/ipoib/ib_ipoib.ko into
> /lib/modules/$kversion/kernel/drivers/infiniband/ulp/ipoib directory.

That sounds like a partial workaround for <http://sourceware.org/PR14454>.


> The following simple stap script worked correctly for a while:
>  [root@cst-kf5 wendy]# cat ipoib.stap
> probe module("*ib_ipoib*").function("*").call
> {
>         printf ("%d %s -> %s\n", tid(), thread_indent(1), probefunc())
> }
> probe module("*ib_ipoib*").function("*").return
> {
>         printf ("%d %s <- %s\n", tid(), thread_indent(-1), probefunc())
> }

(By the way, you could consider using the para-callgraph.stp script
from the distributed samples:)

# stap /usr/share/doc/systemtap*/examples/*/para-callgraph.stp \
       'module("*ib_ipoib*").function("*")'


> Then .. after some kmod coding/rebuilt work, kernel crashed. After
> reboot, system-tap no longer works:
> [root@cst-kf5 wendy]# stap ipoib.stap
> semantic error: no match while resolving probe point
> module("*ib_ipoib*").function("*").call
> [...]
> Any clue and help ?

Is it possible that the newly booted kernel was a different version
than the one before, and that copy of /lib/modules/`uname -r` didn't
have your new ip* module bits in it?  "stap --vp 02 ipoib.stap" should
give a little more clue about where stap is looking for files.


- FChE