Re: Some newbie questions

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

avi wrote:

> [...]
> >>(Have you tried a stap script that merely traps all the same the
> >>function calls, and has empty probe handlers?)

> >I can try it.
> 
> It was actually pretty bad with empty handlers:
> 
> #
>     66.25%  scylla [kernel.kallsyms]   [k] > _raw_spin_lock
>                      |
>                      ---_raw_spin_lock
>                         |
>                         |--49.95%-- 0x62ab
>                         |          syscall_trace_leave
> [...]
>                         |--49.46%-- 0x619b
>                         |          syscall_trace_enter_phase2
> [...]
> 
> I don't have any system call probes.  Just two empty static probes, and 
> a timer.profile handler.

It would sure be nice if perf gave you an explanation of those 0x62ab
bits.  Maybe use dwarf unwind based callgraph collection?  The stap
runtime does sometimes create probes for internal purposes, such as
trapping exec/mmap operations (so probes on new programs can be
activated).  It is conceivable that this is the source of the spinlock
activity, but such a high rate doesn't make sense.


> Will timer.profile work with dyninst?

Profile timers, not yet, but normal periodic ones like timer.hz(997) yes.
Unfortunately backtracing is also not yet implemented there.


- FChE