Date: Fri, 29 Jun 2012 23:43:50 +0200 From: Frank Dittrich <frank_dittrich@...mail.com> To: john-dev@...ts.openwall.com Subject: Re: For some dynamic formats on linux-x86-mmx build cracking depends on password candidate sequence On 06/29/2012 11:32 PM, Frank Dittrich wrote: > But > $ grep -n "^swordfish$" pw.dic > 5762:swordfish > $ grep -n "^asdfasfga__7$" pw.dic > 5122:asdfasfga__7 > $ grep -n "^characters__5$" pw.dic > 5250:characters__5 > > > 5762 = 45 * 128 + 2 > 5122 = 40 * 128 + 2 > 5250 = 41 * 128 + 2 > > So, for the clang build not just the first candidate password of a new > 128 passwords block is affected, but the first two passwords. Sometimes I give up too early: $ grep -n "^PIII__4$" pw.dic 5378:PIII__4 $ grep -n "^pentium__4$" pw.dic 5379:pentium__4 5379 = 42 * 128 + 3 So, looks like (at least) the first 4 candidate passwords of a new MAX_KEYS_PER_CRYPT block can be affected for the clang build. (I can try to shuffle passwords around, to see if they are always affected. Or I could just check at which lines those passwords cracked in the first run were located. But that will be sometime next week, unless someone found the bug by that time.) And, of course, the workaround code change posted for dynamic_2 earlier in this thread works as well. But that costs 2.5 - 3 % performance for my 32 bit linux-x86 builds. Frank
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.