Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 15 Sep 2013 03:27:19 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: ZedBoard / Parallella: bcrypt

Hi Katja,

On Sat, Sep 14, 2013 at 02:46:41PM +0200, Katja Malvoni wrote:
> FPGA bcrypt implementation is integrated in JtR, speed is 13.9 c/s. Code:
> git clone https://github.com/kmalvoni/JohnTheRipper -b master
> 
> I added bitstream file in repo as well, I plan to update wiki page next
> week with detailed instructions for bitstream generation (I will change
> files in repo accordingly).

This is great news, and yes please do update the wiki.

I've just tested your currently committed code, and here are my
observations:

1. After loading of your committed bcrypt.bit.bin, I am getting your
reported speed above from --test, and this implies that the self-test
passes.

2. However, when I tried to actually crack passwords, nothing got
cracked, even though john proceeds through the wordlist at this same
speed.  (I ran it as root, after a "sudo bash", so that the status
display on keypress worked for me.)  Obviously, you need to get this to
work.

3. bcrypt.bit.bin doesn't include(?) Epiphany support, right?  At least
my attempt to use format bcrypt-parallella failed (the john process was
stuck at 100% host CPU, presumably waiting for Epiphany), until I loaded
the Parallella project provided bitstream back.  Can the Epiphany
support be re-added without too much cost in FPGA resource utilization?
I guess the 1200+ c/s that we're getting from Epiphany may exceed
whatever we'll get from that portion of FPGA resources (or maybe not -
we'll see for real later).

4. In BF_std.c, can you separate your added code into a new source file
and invoke it via function calls?  At the same time, make it clear what
license applies to your added code.

5. Can you easily split bcrypt.v into several source files, separating
your hand-written code from Xilinx's code and from auto-generated code?

6. Once you get the current code working for real (cracking passwords),
please proceed with implementation and usage of multiple bcrypt cores,
and with optimizations.

Thanks,

Alexander

Powered by blists - more mailing lists

Your e-mail address:

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