Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 5 Jun 2011 21:53:57 -0500
From: "JFoug" <jfoug@....net>
To: <john-users@...ts.openwall.com>
Subject: Re: Issues with gen_md5 new linking method

The problem here, is in load salt (the part where I actually process OUT 
$$2, $$U, $$Fx, which is the salt2, the user id, and the field items).  The 
logic was being triggering on ANY $$ pair.  It would not have mattered if it 
was first, or last. A 2 byte $$ was enough to trigger the logic.   If the 
following value was not a 2, an U a Fx (with x from 0 to 9), then the 
program would exit.

I have made changes, so that the code will ONLY perform this logic if the $$ 
matches one of the patterns ($$2, $$U, $$Fx).   So, with the included patch 
(made against jumbo-5, so you will have to roll back the prior patch), it 
shoudl work through the problems you see.

NOTE that if you have any found hashes that have the salt of $$2 or $$U, 
then those simply will not work any more, sorry.  I 'may' be able to work 
around that issue also, but it may be a larger change, than this simple 
patch here.  However, hopefully, this will get you by until the next jumbo.

Jim.

----- Original Message ----- 
From: "Jean-Michel" <jtr@...izoku.org>
To: <john-users@...ts.openwall.com>
Sent: Sunday, June 05, 2011 8:51 PM
Subject: Re: [john-users] Issues with gen_md5 new linking method


> Well, I really have a silly database....
>
> Your patch nearly solved the problem. Except that I have entries with salt
> beginning with "$$" too !
> So the error still is "unknown salt string  $$7".
>
> Isn't it "easier" to add a loader flag, like "MGF_SALT_BASE16", that 
> simply
> tells the md5_gen loader to unhex the salt(s)/fields after parsing ?
> That was how the previous PHPS module was written (I mean with an 
> hexencoded
> salt). Maybe it will be also useful to further formats and it may
> uncomplexify your salts/fields parser.
>
> But if you want to try other things, feel free to send me patches and I'll
> try them against my database ASAP. It's quite uge so I can't easily spot
> problematic entries to give them to you.
>
>
> --
> Jean-Michel
>
>
> 2011/6/6 JFoug <jfoug@....net>
>
>> I have gotten your .pot file.
>>
>> The problem stems from having the first character of the salt be a '$'
>> char. There is some 'magic' logic within the salt function, since md5_gen
>> has to handle a totally 'unknown' salt condition. There can be a salt, a 
>> 2nd
>> salt, the user name, and up to 10 fields.   The salt starts with a $s. 
>> The
>> 2nd salt starts with $$2  the user id starts with $$U,  then the fields
>> start with $$Fx where x is 0 to 9
>>
>> So, what I 'believe' the code does, is if the first char of the salt is a
>> '$', then the salt function, grabs the entire salt (including the leading
>> $).  The logic behind this, was I was assuming that seeing the salt start
>> with $$ meant that there was a complex salt (which is broken apart 
>> later),
>> such as $$Uuser$$F2field2   So, the pointer that a 'normal' salt would 
>> copy
>> from is $Uuser$$F2field2, but the code is detecting the first $$ and 
>> backing
>> up and getting the whole thing, in preparation to later rip out the user 
>> ID,
>> and Field2 data.
>>
>> However, in this case, we have a simple salt, and the salt starts with a
>> '$' char.  Thus, the code is returning $$xy for the salt value, instead 
>> of
>> the proper $xy.
>>
>> I can make a simple fix to get this operating properly, and then put this
>> on my todo list.
>>
>> I have attached the patch here, since it is such a trivial patch.  Please
>> let the list know if this resolves the problems you have seen.  Also, if
>> there still are additional problems, then if you can get me a file 
>> (offline
>> again) that demonstrates this, I will continue getting it fixed.  I have 
>> not
>> placed this patch on the wiki, and will wait to hear if this corrects
>> Jean-Michel's problem before I place it there.
>>
>> Jim.
>>
>> ----- Original Message ----- From: "Jean-Michel" <jtr@...izoku.org>
>> To: <john-users@...ts.openwall.com>
>>
>> Sent: Sunday, June 05, 2011 5:25 PM
>> Subject: [john-users] Issues with gen_md5 new linking method
>>
>>
>>
>>  Hello,
>>>
>>> with the version 1.7.7 and the new jumbo patches, the method to link a
>>> format to a gen_md5() one has changed.
>>> For example, the old PHPS format now links to gen_md5(6) as a "thin"
>>> format.
>>> But with that, I can't load my database anymore because the salt used in
>>> this format as to be in "raw" format instead of base16.
>>> I have entries were the salt begins with "$" for example (or contains
>>> specials chars such as "\r"), and that totally messes the md5_gen 
>>> parser.
>>>
>>> Is it already on the TODO-list to add a flag to allow the salt to be
>>> hex-encoded ?
>>> I haven't seen a simple way to do so just by modifing the thin module...
>>>
>>> Thanks.
>>>
>>>
> 

Download attachment "john-1.7.7-jumbo-5-md5_gen_patch-2.diff" of type "application/octet-stream" (3370 bytes)

Powered by blists - more mailing lists

Your e-mail address:

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