Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 26 May 2011 00:28:13 +0200
From: magnum <>
Subject: Re: SSE bug still there in Jumbo-5-RC6++

On 2011-05-25 18:21, magnum wrote:
> Maybe the problem lies somewhere else and overwrites something in
> NSLDAP. I have traced the raw-MD4 segfault to be caused by
> raw-md5-unicode or MD5_go overwriting *out in raw_MD4's get_binary()
> before it is ever used. This makes it never get alloced... and it points
> to 0x67. Bang.
> If I disable raw-md5-unicode (by placing a "return 0" in valid()) the
> fault goes away. If I instead disable the "thatsworking" selftest (still
> in raw-md5-unicode) the fault goes away too. But that is just 12
> characters out of max 53 so I'm not sure what is wrong.

Just bumping up saved_plain size for rawMD5unicode_fmt.c with 1 (in 
MMX_COEF define block) seems to mitigate the problem but does not really 
fix the actual bug.

-static char saved_plain[BLOCK_LOOPS][2 * PLAINTEXT_LENGTH + 1];
+static char saved_plain[BLOCK_LOOPS][2 * PLAINTEXT_LENGTH + 2];

This format will be replaced with a thin as soon as md5-gen can handle 
it. Maybe this fix will do for now, it's better than not doing anything 
about it. Should be easy but I can't find it. I'm not a debugger wizard.


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.