Bad behaviour when unintentionally mixing ipv4 and ipv6 addresses

Hi,

By chance Reinhard Max spotted an interesting flaw in Linux bind(2)...

If you create a IPv4 socket and then incorrectly bind(2) to a IPv6 address (which you got from getaddrinfo(3) or similar), the socket will be bound to INADDR_ANY.

The reason is that the kernel just takes the sockaddr_in6 struct and evaluates it as a sockaddr_in struct, with the port being OK, but the IPv4 sin_addr overlaying the IPv6 sin6_flowinfo field.

As the sin6_flowinfo field is usually 0, your service can end up listening to the world.

A testprogram that you can strace is attached, run netstat -apn |grep 12345 afterwards to see it binds 0.0.0.0:12345.

Perhaps add a check like the one below? (untested)
Or use if (addr->sin_family == AF_INET6) to just catch the IPv6 case?

Ciao, Marcus

Subject: [PATCH] net/ipv4: Check for mistakenly passed in non-IPv4 address

Hi,

Check against mistakenly passing in IPv6 addresses (which would result in an INADDR_ANY bind) or similar incompatible sockaddrs.

Ciao, Marcus

---
 net/ipv4/af_inet.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

index cc14631..9c19260 100644
@@ -465,6 +465,9 @@ int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 	if (addr_len < sizeof(struct sockaddr_in))
 		goto out;
 
+	if (addr->sin_family != AF_INET)
+		goto out;
+
 	chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 
 	/* Not specified by any standard per-se, however it breaks too

1.7.4.1

#include <stdio.h>
#include <stdlib.h>
#include <netinet/in.h>
#include <unistd.h>
#include <string.h>

int main(int argc, char **argv) {
	int			fd,fd2;
	struct sockaddr_in6	saddrin,saddrin2;
	size_t			len;

	fd = socket(AF_INET,SOCK_STREAM,0);
	if (fd == -1) {
		perror("socket");
		exit(1);
	}
	memset(&saddrin,0,sizeof(saddrin));

	/*memcpy(&saddrin.sin6_addr, &in6addr_loopback, sizeof(saddrin.sin6_addr));*/
	memset(&saddrin.sin6_addr, 0x42, sizeof(saddrin.sin6_addr));

	saddrin.sin6_family = AF_INET6;
	saddrin.sin6_port = htons(12345);
	if (-1 == bind(fd, (struct sockaddr*)&saddrin, sizeof(saddrin))) {
		perror("bind");
		exit(1);
	}
	if (-1 == listen(fd, 5)) {
		perror("listen");
		exit(1);
	}
	len  = sizeof(saddrin2);
	fd2 = accept (fd, (struct sockaddr*)&saddrin2, &len);
	if (fd2 == -1) perror("accept");
	close(fd);
	return 0;
}

From: Marcus Meissner <[email protected]>
Date: Wed, 25 May 2011 17:59:18 +0200

I think a corresponding check should also go into inet6_bind() in net/ipv6/af_inet6.c .

cu
Reinhard

Good idea,

Same check as for IPv4, also do for IPv6.

(If you passed in a IPv4 sockaddr_in here, the sizeof check
in the line before would have triggered already though.)

Ciao, Marcus

---
 net/ipv6/af_inet6.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

index b7919f9..d450a2f 100644
@@ -272,6 +272,10 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 
 	if (addr_len < SIN6_LEN_RFC2133)
 		return -EINVAL;
+
+	if (addr->sin6_family != AF_INET6)
+		return -EINVAL;
+
 	addr_type = ipv6_addr_type(&addr->sin6_addr);
 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 		return -EINVAL;

1.7.4.1

From: Marcus Meissner <[email protected]>
Date: Mon, 6 Jun 2011 18:00:07 +0200

Applied, thanks.