potential buffer overflow in xinetd-2.1.8.9pre11-1 (fwd)

[Miquel Mart?n L?pez <miquel@casal.upc.es>]

Hi there!
What do you think of this? is 2.1.8.8 vulnerable? I checked the source code
for my potato xinetd and it doesn't  look like it is...
Any opinions?
Miquel Martín

----- Forwarded message from zen-parse@gmx.net -----

Mailing-List: contact bugtraq-help@securityfocus.com; run by ezmlm
List-Id: <bugtraq.list-id.securityfocus.com>
List-Post: <mailto:bugtraq@securityfocus.com>
List-Help: <mailto:bugtraq-help@securityfocus.com>
List-Unsubscribe: <mailto:bugtraq-unsubscribe@securityfocus.com>
List-Subscribe: <mailto:bugtraq-subscribe@securityfocus.com>
Date: Sat, 9 Jun 2001 02:10:41 +1200 (NZST)
From: <zen-parse@gmx.net>
X-X-Sender:  <zen-parse@clarity.local>
To: <bugtraq@securityfocus.com>
Subject: potential buffer overflow in xinetd-2.1.8.9pre11-1
Resent-To: Administrador del Sistema <admin@casal.upc.es>
Resent-To: Miquel Martín López <miquel@casal.upc.es>

***************************************************************
 Potential remote overflow fixed in patched version of xinetd
***************************************************************

***************************************************************
 Problem:   potential buffer overflow in xinetd-2.1.8.9pre11-1
***************************************************************

***************************************************************
    Fix:  update to xinetd-2.1.8.9pre15-2 (for redhat ppl)
***************************************************************

Please note:  This does not seem to be exploitable in a default
setup RH 7.0 machine. However There may be other distributions/
configutations that it is used in where it is explotable.


***************************************************************
                   Details of the problem
***************************************************************

svc_logprint (in xinetd/log.c) has a slight bug which may allow remote
root access.


...
        len = strx_nprint( buf, bufsize, "%s: %s ", line_id, SVC_ID( sp
) ) ;
        va_start( ap, fmt ) ;
*->    cc = strx_nprintv( &buf[ len ], bufsize, fmt, ap ) ;
        va_end( ap ) ;
...
(bufsize=sizeof(buf) ==  LOGBUF_SIZE = 1024)

If an argument to the marked line is longer than (bufsize-len) then it
will overflow the string.

The ident feature allows returning 1024 bytes of information, and that
information, less the
source,dest: componant and the \r\n s passed to svc_logprint() as an
argument.
1024,21:USERID:AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA...AA\r\n
such that the string totals 1024 characters for example.

If a malicious root user was to connect, he could set his own source
port to something like 1,
which would gain him another 3-4 characters.
1,21:USERID:AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA...AAAAA\r\n

The string is then truncated at the \r

(xinetd/ident.c)
...
       svc_logprint( SERVER_CONNSERVICE( serp ), USERID_ENTRY, "%s", p )
;
...

p would then be a string 1010 characters long.
if strlen(line_id)+strlen(SVC_ID( sp ) )>14 then we have a buffer
overflow.

With the ftp service I was only able to get a 1022 byte buffer written
but with other services
with longer names that use authentication, this could be a serious
problem.

The server is still running as root while this happens.


--zen-parse


----- End forwarded message -----

----- End forwarded message -----