[Avcheck] Warning: bad AVP behaviour

Michael Tokarev mjt@tls.msk.ru
Mon, 20 Aug 2001 15:25:15 +0400


I performed some more tests against avpdaemon -- it looks
like it has more problems than I thought before...

When AvpDaemon's child process that lanched to handle a
request is terminated (killed), the same bad thing will
happen as with SIGSEGV case I described earlier, *or* it
will return "success".  I can't reproduce success case
easily (doing with strace always generates lockup instead
of success, but doing without strace shows 50/50 chance
to get either success or lockup -- seems like timing issues).
Previously I attempted to kill *parent* process (using
kill `cat /var/spool/avp/ctl/AvpPid'), now I trying to
use killall AvpDaemon (I still playing games with old
AvpDaemon).

It is still a question if the same bugs exists in never
kavdaemon -- when I will have a time, I'll check this too
(btw, anyone can do those checks using e.g. my "bugmail"
example file -- any volonter?).

For ones who uses avp, I suggest to file a bugreport
to kaspersky labs (I already tried this before, without
any luck, and I have no purchased version etc).  But
I'm quite shure that other bugs will be shown anyway --
overall quality of code is bad, and it need a strong
review and rewrite to be really stable.  I can't
trust it anymore, even with all the games I played
with it (chroot, safe permissions etc).

Sorry, folks -- c'est la vie...

Regards,
 Michael.