Re: Plain kprobe genearation/bisect script - latest changes

[David Smith <dsmith@redhat.com>]

Mahesh Jagannath Salgaonkar wrote:
> Mahesh Jagannath Salgaonkar wrote:
>> Hi All,
>>
>> I have implemented shell scripts that helps me to narrow down a faulty
>> probe that causes problem/crash, while debugging systemtap issues. As
>> per the discussion in weekly systemtap meeting dated 12th March 2009,
>> I am attaching these scripts (kp_bisect.tgz) that may help us in
>> identifying probes that can be a potential candidate for blacklist.
>> These scripts will help user to extract probes (probe function name)
>> from stap generated C module and test plain kprobes/kretprobes
>> allowing user to run bisect-like functionality to narrow down to a
>> single faulty probe. It does handle ".call" and ".return" probes.
>>
>> Can someone can take look at these scripts and see if these can be
>> used as a base for automating the process of testing kprobes for
>> black/white list generation? The file "README.txt" describes the steps
>> to use these scripts.
>>
>> Thanks,
>> -Mahesh.
> 
> Please find the modified scripts attached. Modified script to isolate
> good probes/bad probes and show all bad probes at the end.

Mahesh,

I finally found some time to look at this.  The nice thing about your
scripts is that they are very simple and look like they would do the job.

The one thing that the scripts I worked on does differently is that it
ends up with 3 lists instead of 2 lists:

- a "bad" list - functions that cause crashes
- a "good" list - functions that were registered correctly and the
kprobe got hit with no problems
- an "unknown" list - functions that were registered correctly, but the
function wasn't actually called (so we don't really know if a kprobe can
be safely used with that function or not)

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)