Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Tue, 24 Sep 2019 13:32:28 +0200
From: Solar Designer <>
Subject: Re: John doesn't detect my second graphics card

On Tue, Sep 24, 2019 at 08:22:59AM +0300, Xosted wrote:
> Thank you for your answer. If I remember correctly,  this is an incremental attack against a 7zip file. It has now been running for about 110 days.

I recommend that you leave this one running, and run other attacks on
your added GPU - various wordlists, etc.  There's little point in adding
more compute power to an incremental mode run that's been going for 110
days - the resulting increase in probability of success (until the point
where you'd give up) is tiny.  Speaking of better attacks to run, have
you already run through the RockYou list?  You should.

Also, if you recall anything about that password, you should provide
that information to JtR e.g. as a mask mode pattern.

> Actually, is there a way to "interrogate " a rec file so it reveals the exact parameters it was launched with? Maybe even getting the exact command line?

This is a text file - you can see the options used (and some that were
implied) in there, one option per line, along with other information.

> Sorry if the following doesn't make sense, I'm just thinking out loud:
> Do you think it would be possible or useful to have a resource pool architecture added to John? The advantage would of course be to solve cases like mine.
> Also, Correct me if I'm wrong but at the moment, I can use either of my graphics card or CPU but not both in an attack. By decoupling the attack and the resource on which it runs, maybe it would be possible to run attacks on heterogeneous pools? Port an attack from one system to another and have it continue in another pool rather than start again? Take advantage of temporarily available extra resources or survive the permanent disappearance of a resource?

Sure it does make sense.  It's just not the way we hacked in multi-GPU
support a few years back.  (It does work that way with our current use
of OpenMP - the number of CPU chips and/or cores can change between
interrupt/restore, and fewer or more cores would be used accordingly.)

Regarding your CPU, technically you can also use at as an OpenCL device
along with your GPUs, but that's far from optimal - both because of
shortcomings of our current architecture and because these devices can
best be used for different attacks.  On a CPU, you can more efficiently
run many quick attacks one after another.  On GPUs, you'd run heavier
attacks and keep them going for longer.

So I suggest that you run 3 separate attacks at a time, with different
session names.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.