lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <001e01c3a554$32f29360$77bc5452@c0wboy>
Date: Fri, 7 Nov 2003 18:25:53 +0100
From: "c0wboy@...33" <c0wboy@...cali.it>
To: <bugtraq@...urityfocus.com>
Subject: terminatorX 3.8.1 local vulnerabilities



           -       0x333 OUTSIDERS SECURITY LABS       -
            -              www.0x333.org              -


~~~~~~~~~ contents ~~~~~~~~~

0x0 Info
0x1 Description
0x2 Stack Overflow in Switch (1)
0x3 Stack Overflow in Switch (2)
0x4 Stack Overflow in $LADSPA_PATH
0x5 Format String Bug in tx_note()
0x6 Solutions
0x7 Vendor Contact



[0x0 Info]

 author   : c0wboy
 mail     : c0wboy@...cali.it
 date     : 7 November 2003
 advisory : outsiders-terminatorX-001.txt
 vendor   : http://terminatorX.cx
 category : stack overflow, format bug
 platform : linux, unix


[0x1 Description]

>From terminatorX's offcial web page : "terminatorX is a realtime audio
synthesizer that allows
you to "scratch" on digitally sampled audio data (*.wav, *.au, *.ogg, *.mp3,
etc.)" . In Last
version (3.8.1) there are still lots of bugs, that could be used by user to
execute arbitrary
codes with root privileges.


[0x2 Stack Overflow in Switch (1)]

It is possible to overflow a static buffer, by passing a long string after
switch "-f"
(or "--file"). In function parse_args() (src/main.cc) we find the follow
piece of code:

  if ((strcmp(argv[i], "-f") == 0) || (strcmp(argv[i], "--file") == 0)) {
   ++i;
   globals.startup_set = argv[i];

The variable "globals.startup_set" will contain our (long) string. After
this the variable
will be passed to function load_tt_part() (src/tX_mastergui.cc) which is
called by main():

int main(int argc, char **argv)
{
...
 if (globals.startup_set) {
  while (gtk_events_pending()) gtk_main_iteration(); gdk_flush();
  tX_cursor::set_cursor(tX_cursor::WAIT_CURSOR);
  load_tt_part(globals.startup_set);
...

/* src/tX_mastergui.cc */
void load_tt_part(char * buffer)
{
 char idbuff[256];
 char wbuf[PATH_MAX];
...
 strcpy(globals.tables_filename, buffer);
...
   strcpy(idbuff, "Failed to access file: \"");
   strcat(idbuff, globals.tables_filename);
...

In load_tt_part() our string will be first copied in
"globals.tables_filename", that will
overflow the static buffer idbuff[256]. User could overwrite eip and
executes arbitrary
code.


[0x3 Stack Overflow in Switch (2)]

This case is similar to the first one. Here we can overflow a static buffer
by passing a
long string after option "-r" (or "--rc-file"). In function parse_args()
(src/main.cc) we


  if ((strcmp(argv[i], "-r") == 0) || (strcmp(argv[i], "--rc-file") == 0)) {
   ...
   globals.alternate_rc = argv[i];

In function get_rc_name() (src/tX_global.c) there is a unchecked strcpy()
that will overflow
a static buffer (passed to this function as argument) with the content of
"globals.alternate_rc"
variable:

void get_rc_name(char *buffer)
{
 strcpy(buffer,"");
 if (globals.alternate_rc) {
  strcpy(buffer, globals.alternate_rc);
...

This bug can be used to overwrite the eip, and execute arbitrary code.


[0x4 Stack Overflow in $LADSPA_PATH]

In file tX_ladspa.cc, is possible overflow a static buffer by setting a long
"$LADSPA_PATH"
environment variable. As we had before, there is an unchecked strcpy(), in
this case a long string
will overflow "ladspa_path[PATH_MAX]":

void LADSPA_Plugin :: init ()
{
 char *ladspa_path_ptr;
 char ladspa_path[PATH_MAX];
...

 /* Finding the LADSPA Path */
 ladspa_path_ptr=getenv("LADSPA_PATH");
...
 else strcpy(ladspa_path, ladspa_path_ptr);
...

By overflowing "ladspa_path[PATH_MAX]", eip will be overwritten, permitting
the execution of
arbitrary code.


[0x5 Format String Bug in tx_note()]

There is a format string bug in function tx_note(), that can be found in
src/tX_mastergui.cc. The
problem exists in function "gtk_message_dialog_new()". User can control the
contents of "message"
variable, and set it to special format string parameters (for example lots
of '%x'). Here's the
vulnerable code :

void tx_note(const char *message, bool isError)
{
 char buffer[4096]="terminatorX ";
 if (isError) {
  strcat(buffer, "note:\n\n");
 } else {
  strcat(buffer, "error:\n\n");
 }

 strcat(buffer, message);
 GtkWidget *dialog=gtk_message_dialog_new(GTK_WINDOW(main_window),
  GTK_DIALOG_DESTROY_WITH_PARENT,
  isError ? GTK_MESSAGE_ERROR : GTK_MESSAGE_INFO, GTK_BUTTONS_CLOSE,
message);
 gtk_dialog_run(GTK_DIALOG(dialog));                                      /*
|_____ fmt bug  */
 gtk_widget_destroy(dialog);
}

Is possible to exploit this bug, by passing a crafted string after witch
"-f" (or "--file"). TerminatorX
will be run, but it will display a warning-windows showing memory addresses.
There will be a segfault.

[c0wboy@...33 src]$ ./terminatorX -f %x%x%x%x%x%x%n
terminatorX Release 3.81 - Copyright (C) 1999-2003 by Alexander Knig
terminatorX comes with ABSOLUTELY NO WARRANTY - for details read the
license.
* tX_error: set_capabilities(): failed to set caps: Operation not permitted.
+ tX_warning: engine_thread_entry(): can't set SCHED_FIFO -> lacking
capabilities.
+ tX_warning: engine_thread_entry() - engine has no realtime priority
scheduling.
+ tX_warning: LADSPA_PATH not set. Trying
/usr/lib/ladspa:/usr/local/lib/ladspa
* tX_error: tX: Error: couldn't access directory "/usr/lib/ladspa".
+ tX_warning: Plugin "Stereo Amplifier" disabled. Not a 1-in/1-out plugin.
+ tX_warning: Plugin "White Noise Source" disabled. Not a 1-in/1-out plugin.
+ tX_warning: Plugin "Sine Oscillator (Freq:audio, Amp:audio)" disabled. Not
a 1-in/1-out plugin.
+ tX_warning: Plugin "Sine Oscillator (Freq:control, Amp:control)" disabled.
Not a 1-in/1-out plugin.
warning: failed to load external entity "%25x%25x%25x%25x%25x%25x%25n"
Segmentation fault
[c0wboy@...33 src]$

**NOTE** : There are lots of syntax-wrong gtk_message_dialog_new(), in
src/tX_mastergui.cc file.


[0x6 Solutions]

 0x1 (switch "-f"):

 /* src/tX_mastergui.cc */
 void load_tt_part(char * buffer)
 {
  char idbuff[256];
  char wbuf[PATH_MAX];
  xmlDocPtr doc;
 ...
 if (strlen(idbuff) + strlen(buffer) > 256-strlen("Failed to access file:
\"")-strlen("\""))
  return;
 else
  strcpy(globals.tables_filename, buffer);
  ...

 This will correctly check the "idbuff" lenght.


 0x2 (switch "-r"):

 void get_rc_name(char *buffer)
 {
  strcpy(buffer,"");
  if (globals.alternate_rc && (strlen(globals.alternate_rc) <
sizeof(buffer)-1) ) { /* bof fixed */
   strcpy(buffer, globals.alternate_rc);
 ...

 0x3 ($ env):

 The program already use a funtion to check the environment variables leghnt
(checkenv()).
 Just us it in main() also with $LADSPA_PATH.


 0x4 (fmt bug):

 void tx_note(const char *message, bool isError)
 {
  char buffer[4096]="terminatorX ";
  if (isError) {
   strcat(buffer, "note:\n\n");
  } else {
   strcat(buffer, "error:\n\n");
  }

  /* - */ strcat(buffer, message);
  /* + */ strncat(buffer, sizeof(message), message); /* i don't like this
strcat() ... */

  GtkWidget *dialog=gtk_message_dialog_new(GTK_WINDOW(main_window),
GTK_DIALOG_DESTROY_WITH_PARENT,
  isError ? GTK_MESSAGE_ERROR : GTK_MESSAGE_INFO, GTK_BUTTONS_CLOSE, "%s",
message); /* fmt bug fixed */
  ...


[0x7 Vendor Contact]

Vendor was informed about the bugs.




Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ