Re: [RFC][Patch 0/2]kprobe: kprobe-booster against 2.6.14-mm1 for i386

[Masami Hiramatsu <hiramatu at sdl dot hitachi dot co dot jp>]

Hi, Frank

Frank Ch. Eigler wrote:
>>>Can you describe what kinds of tests you ran on this exciting
>>>optimization, and their results?
>>
>>I made a test program attached to this mail.
>>I checked correctness of kprobe-booster (take2) by using
>>this program. [...]
>
> OK.  Would you be able to try a larger test, such as running any
> existing kprobes or especially systemtap pass-5 tests on a kernel with
> this modification?

Yes, I tested.
For kprobes, I ran a small test script on the
boosted kernel.
The script disassembles sys_symlink function,
inserts kprobes in each instructions and
runs ln commands which executes a success
path and two failure paths.

for i in `./disym.sh -a sys_symlink | grep ^c | cut -f1 -d\: `;do
(echo $i | grep ^c > /dev/null ) || continue;
sudo /sbin/insmod ./noop_kprobe.ko addr=0x$i; echo install $i
(echo OK; ln -s /dev/null /tmp; rm /tmp/null;
 echo eperm; ln -s /dev/null /usr;
 echo enoent; ln -s /dev/null /hoge/huga/ )
sudo /sbin/rmmod noop_kprobe
done

For SystemTap, I ran my original tapsets which I
demonstrated in Beaverton.

I confirmed both tests worked correctly.

I will try to run the test suite which can be got
from sourceware.org by CVS.

Best Regards.
-- 
Masami HIRAMATSU
2nd Research Dept.
Hitachi, Ltd., Systems Development Laboratory
E-mail: hiramatu@sdl.hitachi.co.jp