Bugzilla Bug 156476

From:     wk@gnupg.org
        Subject:        GnuPG 1.4 and 2.0 buffer overflow
        Date:   November 27, 2006 12:13:02 PM EST
        To:       gnupg-announce@gnupg.org
        Cc:       bugtraq@securityfocus.com, lwn@lwn.net

            GnuPG 1.4 and 2.0 buffer overflow
           ==================================

Summary
=======

While fixing a bug reported by Hugh Warrington, a buffer overflow has
been identified in all released GnuPG versions.  The current versions
1.4.5 and 2.0.0 are affected.  A small patch is provided.

Please do not send private mail in response to this message.  The
mailing list gnupg-devel is the best place to discuss this problem
(please subscribe first so you don't need moderator approval [1]).


Impact
======

When running GnuPG interactively, special crafted messages may be used
to crash gpg or gpg2.  Running gpg in batch mode, as done by all
software using gpg as a backend (e.g. mailers), is not affected by
this bug.

Exploiting this overflow seems to be possible.

gpg-agent, gpgsm, gpgv or other tools from the GnuPG suite are not
affected.



Solution
========

Apply the following patch to GnuPG.  It should apply cleanly to
current versions (1.4.5 as well as 2.0.0) but might also work for
older versions. 

2006-11-27  Werner Koch  <wk@g10code.com>

        * openfile.c (ask_outfile_name): Fixed buffer overflow occurring
        if make_printable_string returns a longer string.  Fixes bug 728.

--- g10/openfile.c      (revision 4348)
+++ g10/openfile.c      (working copy)
@@ -144,8 +144,8 @@

     s = _("Enter new filename");

-    n = strlen(s) + namelen + 10;
     defname = name && namelen? make_printable_string( name, namelen, 0): NULL;
+    n = strlen(s) + (defname?strlen (defname):0) + 10;
     prompt = xmalloc(n);
     if( defname )
        sprintf(prompt, "%s [%s]: ", s, defname );



Background:
===========

The code in question has been introduced on July 1, 1999 and is a
pretty obvious bug.  make_printable_string is supposed to replace
possible dangerous characters from a prompt and returns a malloced
string.  Thus this string may be longer than the orginal one; the
buffer for the prompt has only be allocated at the size of the original
string - oops.  Note, that using snprintf would not have helped in
this case.  How I wish C-90 had introduced asprintf or at least it
would be available on more platforms.

The original bug report is at https://bugs.g10code.com/gnupg/issue728 .



===
[1] See http://lists.gnupg.org/mailman/listinfo/gnupg-devel .


-- 
Werner Koch                                      <wk@gnupg.org>
The GnuPG Experts                                http://g10code.com
Join the Fellowship and protect your Freedom!    http://www.fsfe.org

and if you read the rest of the announcements on the relevant mailing lists,
1.4.6/2.0.1 will be out in a day or two to resolve this.

doesnt seem like a security issue, unless you can get someone to type in lots
of (partially binary) characters.

Can anyone think of an attack using this bug?

where exactly do you have to type something? As I currently understand it, this
can be triggered by interactively runnung gpg on a malicious file (like the one
the reported attached to the upstream bug) - but I had no deep look here, since
the vendor came out with this (and well, I seriously hope that they know what
they are doing).

Ahh, I see what you mean, I hadnt read the bug.

Yes, this looks like it would be a problem then, ignore my last comment.

upstream has 2.0.1 out, but not 1.4.6 yet.

has not been fully wrangled so far... 
rating B2...

just to fill this comment field...

gnupg 1.4.6 is now available at http://www.gnupg.org/download/ 

Done.

Please don't close Security bugs.

This one is ready for GLSA.

GLSA 200612-03