john-users - FullMPI for 1.7.7

Openwall GNU/*/Linux 3.0 - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Wed, 04 May 2011 20:47:33 +0200
From: magnum <rawsmooth@...dband.net>
To: john-users@...ts.openwall.com
Subject: FullMPI for 1.7.7

john-1.7.7-jumbo-1-fullmpi-9.diff.gz is now uploaded to the wiki. It has 
a couple of improvements compared to 1.7.6-fullmpi8.

* -DHAVE_MPI, passing "the sha1sum test" when disabled.
* Do not show any ETA until progress is above 0.00%. NOTE that this 
should really be in non-MPI too but it's currently ifdef'ed.
* Incremental mode now has progress status and ETA (note the previous 
point about muting though), but not after resuming a job. It's useful 
for --inc=alnum or smaller keyspaces and/or very fast formats.
* Incremental mode: the empty string candidate is now only tested on one 
node.
* If one node runs out of hashes, it sends an abort to the other nodes 
(they would otherwise keep attacking already cracked hashes).
* If a node stops with error (eg. file locking problems because of 
flakey NFS) it will send an abort to all other nodes.
* Jobs could not be resumed if some nodes had finished. Now the .rec 
file does not get deleted until all nodes are done.
* Mitigated the interleaving of output lines when showing status for a 
running job (it doesn't happen if you print everything in one call).
* Higher preference of distributing rules instead of words.
* --mem=0 will force loading of wordfile to memory, regardless of size. 
This too could be in non-MPI as well IMO.
* Added some compilation hints to README.mpi.

I also did some tweaking so it applies together with the utf8 patch 
without rejecting.

The code has been stable for a long time now. If you get problems, they 
are most likely due to NFS problems (for multi-host setup) or using a 
buggy MPI version. I recommend using latest stable OpenMPI but mpich2 
and even mpich1 is supported.

enjoy!
magnum