--------
Overview
--------
There seems to be a remotely exploitable vulnerability in Sendmail up to
and including the latest version, 8.12.9. The problem lies in prescan()
function, but is not related to previous issues with this code.
The primary attack vector is an indirect invocation via parseaddr(),
although other routes are possible. Heap or stack structures, depending
on the calling location, can be overwritten due to the ability to go
past end of the input buffer in strtok()-alike routines.
This is an early release, thanks to my sheer stupidity.
--------------
Attack details
--------------
Local exploitation on little endian Linux is confirmed to be trivial
via recipient.c and sendtolist(), with a pointer overwrite leading to a
neat case of free() on user-supplied data, i.e.:
eip = 0x40178ae2
edx = 0x41414141
esi = 0x61616161
SEGV in chunk_free (ar_ptr=0x4022a160, p=0x81337e0) at malloc.c:3242
0x40178ae2 <chunk_free+486>: mov %esi,0xc(%edx)
0x40178ae5 <chunk_free+489>: mov %edx,0x8(%esi)
Remote attack is believed to be possible.
----------------
Workaround / fix
----------------
Vendor was notified, and released an early patch attached below.
There are no known workarounds.
Index: parseaddr.c
===================================================================
RCS file: /cvs/src/gnu/usr.sbin/sendmail/sendmail/parseaddr.c,v
retrieving revision 1.16
diff -u -r1.16 parseaddr.c
--- parseaddr.c 29 Mar 2003 19:44:01 -0000 1.16
+++ parseaddr.c 16 Sep 2003 17:37:26 -0000
@@ -700,7 +700,11 @@
addr[MAXNAME] = '\0';
returnnull:
if (delimptr != NULL)
+ {
+ if (p > addr)
+ p--;
*delimptr = p;
+ }
CurEnv->e_to = saveto;
return NULL;
}
--
------------------------- bash$ :(){ :|:&};: --
Michal Zalewski * [
http://lcamtuf.coredump.cx]
Did you know that clones never use mirrors?
--------------------------- 2003-09-16 21:18 --
