[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 27 Jun 2012 09:56:20 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: bug or false alarm in BSDI format with clang
On Fri, Jun 22, 2012 at 07:36:03PM +0530, Dhiru Kholia wrote:
> >> DES_key[1] = key[1] & 0x7F;
>
> Can someone familiar with the format take a look? Testing DES itself
> which uses the same code (guessing?) passes with no problems.
DES_std_set_key() deliberately makes those reads potentially/sometimes
beyond the end of string, for performance reasons. Theoretically, this
could result in a crash if the string were at the very end of a page and
the next page were not mapped, but in practice what JtR passes to that
function is never located like that.
This dates back to 1990s when we did not yet have a bitslice DES
implementation and thus this non-bitslice DES code was
performance-critical. Perhaps I need to change this now.
Alexander
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.
