openssh unix announce - Mar 2002 - OpenSSH Security Advisory (adv.channelalloc)

>         This bug can be exploited locally by an authenticated user
>         logging into a vulnerable OpenSSH server or by a malicious
>         SSH server attacking a vulnerable OpenSSH client.
OK, I must really be missing something.

Doesn't OpenSSH drop all privs long before either side gets the option to
open a corrupted channel?

If so, where's the route to sshd for a buffer overflow to exploit?  The
closest I can come up with is in a setuid ssh client being poked,
X-Forwarding style, by a corrupted server...in which case, that's another
reason why ssh shouldn't be setuid by default.

Incidentally, *someone* has actually seen a working attack, right?

--Dan

1. Systems affected:

	All versions of OpenSSH between 2.0 and 3.0.2 contain
	an off-by-one error in the channel code.

	OpenSSH 3.1 and later are not affected.

2. Impact:

        This bug can be exploited locally by an authenticated user
        logging into a vulnerable OpenSSH server or by a malicious
        SSH server attacking a vulnerable OpenSSH client.
	
3. Solution:

	Upgrade to OpenSSH 3.1 or apply the following patch.

4. Credits:

	This bug was discovered by Joost Pol <joost at pine.nl>


Appendix:

Index: channels.c
==================================================================RCS file:
/cvs/src/usr.bin/ssh/channels.c,v
retrieving revision 1.170
retrieving revision 1.171
diff -u -r1.170 -r1.171
--- channels.c	27 Feb 2002 21:23:13 -0000	1.170
+++ channels.c	4 Mar 2002 19:37:58 -0000	1.171
@@ -146,7 +146,7 @@
 {
 	Channel *c;
 
-	if (id < 0 || id > channels_alloc) {
+	if (id < 0 || id >= channels_alloc) {
 		log("channel_lookup: %d: bad id", id);
 		return NULL;
 	}