Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 09 Aug 2012 15:15:06 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: Current -fixes GPU formats vs TS

On 2012-08-08 21:45, Lukas Odzioba wrote:
> If I will not send a patch for pwsafe within 24h we may want to remove
> it from tree and release without that format.

I just decreased its PLAINTEXT_LENGTH to 15 (like the opencl format) and
this works fine. However, it still intermittently fails self-test at
cmp_all(1) - something is not initialized correctly. When it passes
self-test, it passes TS too.

If I test xsha512-cuda right before it, it will fail:

magnum@...l:test [master]$ ./jtrts.pl xsha512-cuda pwsafe-cuda -stoponerror
-------------------------------------------------------------------------------
- JtR-TestSuite (jtrts). Version 1.12.11, July 20, 2012.  By, Jim
Fougeron & others
- Testing:  John the Ripper password cracker, ver: 1.7.9-jumbo-7RC
[linux-x86-64-gpu]
--------------------------------------------------------------------------------

John Jumbo build detected.

form=xsha512-cuda                 guesses: 1500 time: 0:00:00:03  [PASSED]
.pot CHK:xsha512-cuda             guesses: 1500 time: 0:00:00:03  [PASSED]

form=pwsafe-cuda                  guesses:    0 unk unk : Expected
count(s) (1485)(1375)  [!!!FAILED!!!]
Self test failed (cmp_all(1))

If I do not, it will work:

magnum@...l:test [master]$ ./jtrts.pl pwsafe-cuda -stoponerror
-------------------------------------------------------------------------------
- JtR-TestSuite (jtrts). Version 1.12.11, July 20, 2012.  By, Jim
Fougeron & others
- Testing:  John the Ripper password cracker, ver: 1.7.9-jumbo-7RC
[linux-x86-64-gpu]
--------------------------------------------------------------------------------

John Jumbo build detected.

form=pwsafe-cuda                  guesses: 1375 time: 0:00:13:34  [PASSED]

Also, if I just run a --test with xsha512-cuda right before pwsafe, it
will also fail:

magnum@...l:test [master]$ ../run/john -t -fo:xsha512-cuda
Benchmarking: Mac OS X 10.7+ salted SHA-512 [CUDA]... DONE
Many salts:     59986K c/s real, 59450K c/s virtual
Only one salt:  11838K c/s real, 11838K c/s virtual

magnum@...l:test [master]$ ../run/john -t -fo:pwsafe-cuda
Benchmarking: Password Safe SHA-256 [CUDA]... FAILED (cmp_all(1))

If I do not, it works fine:

magnum@...l:test [master]$ ../run/john -t -fo:pwsafe-cuda
Benchmarking: Password Safe SHA-256 [CUDA]... DONE
Raw:    106852 c/s real, 106852 c/s virtual


Lukas, if you can reproduce the above I bet you can nail it.

BTW I also moved the endian-swap out of crypt_all() and into get_salt()
for good measure (and same fix for the opencl format).

magnum

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ