Re: A 2.4.[57] kernel crypto problem
On Sun, Jan 06, 2002 at 04:10:12AM -0700, Stefan Srdic wrote:
> On January 6, 2002 02:00 pm, Pavel Minev Penev wrote:
> >
> > Hello.
> >
> > I had a peculiar experience with a password (forgot it). It is the
> > password for an AE
> > S-encrypted partition on my HDD. I am using the loop
> > device and the international kernel patch. I wrote a brute-forcer
> > (didn't find docs and stole a lot of code from mount and losetup).
> >
> > The problem is that when I leave the brute-forcer working for a longer
> > time (about 5 seconds on my 750 MHz Duron) when I press Ctrl-C (the
> > brute-forcer catches the signal and tries to save the session) the
> > kernel seems to deadlock. It is most probably the loop device driver.
> > There are about 3304 proceses with sequential PIDs and names of
> > "[loop7 <defunct>]", and are all zombies. If I try to start an app. as
> > root I get something like "fork: resource temporarily unavailable", or
> > "INIT: cannot fork; retrying...". Also, I can't `losetup /dev/loop7`,
> > since it pauses, issuing nothing. As a normal user I seem to be able to
> > work as usual. After a short period of time `init` seems to start
> > functioning and switch the run-level. The brute-forcer works as follows:
> > 1. Generate billions of passwords.
> > For each of them:
> > 1. Setup a loop device.
> > 2. Read the block after the 1024-th byte and check it
> > for Ext2/Ext3's magic ID.
> > If the ID matches:
> > 1. Print the password.
> > 3. Deconfigure the loop device.
> > The brute-forcer works fine for short periods of time.
> >
> > I've tried this on kernels 2.4.15 and 2.4.17, it's all the same
> > (although the 2.4.17 Changelog says about a number of bug-fixes in the
> > loop-back driver).
> >
> > I was wonder
> > ing if you could tell me about any known or unknown problems
> > with the kernel crypto, or help me realise my stupidity.
> >
> > Comment: Sorry if this is too off-topic, I could post it to the Linux
> > kernel mailing list if you prefer.
>
> There's a warning in announce.txt in the testing directory that you might not
> have seen:
>
> *WARNING* this is meant for the brave ones (read beta-testers ;), which
> want to do some tests, and hopefully report back any problems they
> encounter!
>
> Perhaps just using the standalone loop-aes module would be better in your
> case:
>
> http://loop-aes.sourceforge.net/
>
> It's worth a try.
Yes. This is great, it has a fabulous performance. If I could only find
some docs on the API I would be very thankful (the kernel module's
sources are huge to read).
Thanks,
--
Pav
Reply to: