Bug 214729

Hi Maintainers,
I recently reviewed the uninitialized value use bug in Linux kernel:

https://syzkaller.appspot.com/bug?id=739ce0bc6e4097668cbf94c862f3b643b364d589. 
and patch:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b42b3a2744b3e8f427de79896720c72823af91ad

The vulnerable function is:
int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
		      int __user *usockaddr_len)
{
	struct socket *sock;
	struct sockaddr_storage address; // allocation
	int err, fput_needed;

	sock = sockfd_lookup_light(fd, &err, &fput_needed);
	if (!sock)
		goto out;

	err = security_socket_getsockname(sock);
	if (err)
		goto out_put;

	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); // initialization
	if (err < 0)
		goto out_put;
        /* "err" is actually length in this case */
	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); //use

out_put:
	fput_light(sock->file, fput_needed);
out:
	return err;
}

static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
			     void __user *uaddr, int __user *ulen)
{
	int err;
	int len;

	BUG_ON(klen > sizeof(struct sockaddr_storage));
	err = get_user(len, ulen);
	if (err)
		return err;
	if (len > klen)
		len = klen;
	if (len < 0)
		return -EINVAL;
	if (len) {
		if (audit_sockaddr(klen, kaddr))
			return -ENOMEM;
		if (copy_to_user(uaddr, kaddr, len)) // use 
			return -EFAULT;
	}
	/*
	 *      "fromlen shall refer to the value before truncation.."
	 *                      1003.1g
	 */
	return __put_user(klen, ulen);
}


the variable address is allocated in __sys_getsockname, and then is initialized in sock->ops->getname(sock, (struct sockaddr *)& address, 0); After that, address is passed to move_addr_to_user() and finally it passed to copy_to_user(), leading to uninitialized value use. 

Main reason for this bug: initialization in sock->ops->getname(sock, (struct sockaddr *)&address, 0) is partially initialized.  It only initializes the fields in the struct but not the holes between the fields. As a result, since uaddr will be passed to copy_to_user(), and the holes inside the uaddr struct  contain uninitialized data inherited from the kernel stack, it may cause information leakage from kernel space to user space

Here is the initialization function:

static int isotp_getname(struct socket *sock, struct sockaddr *uaddr, int peer)
{
	struct sockaddr_can *addr = (struct sockaddr_can *)uaddr;
	struct sock *sk = sock->sk;
	struct isotp_sock *so = isotp_sk(sk);

	if (peer)
		return -EOPNOTSUPP;

	+memset(addr, 0, ISOTP_MIN_NAMELEN);//patch
	addr->can_family = AF_CAN;
	addr->can_ifindex = so->ifindex;
	addr->can_addr.tp.rx_id = so->rxid;
	addr->can_addr.tp.tx_id = so->txid;

	return ISOTP_MIN_NAMELEN;
}

The patch: memset() initializes the whole struct sockaddr_can, including the holes within the struct sockaddr_can.
 
Sockaddr_can is declared here:
struct sockaddr_can {
	__kernel_sa_family_t can_family;
	int         can_ifindex;
	union {
		/* transport protocol class address information (e.g. ISOTP) */
		struct { canid_t rx_id, tx_id; } tp;

		/* J1939 address information */
		struct {
			/* 8 byte name when using dynamic addressing */
			__u64 name;

			/* pgn:
			 * 8 bit: PS in PDU2 case, else 0
			 * 8 bit: PF
			 * 1 bit: DP
			 * 1 bit: reserved
			 */
			__u32 pgn;

			/* 1 byte address */
			__u8 addr;
		} j1939;

		/* reserved for future CAN protocols address information */
	} can_addr;
};
There are a few holes inside the struct, but it doesn’t explicitly set to 0 in  isotp_getname()

At the same time, I realized  sock->ops->getname(sock, (struct sockaddr *)&address, 0) is an indirect call. Thus it may go to different functions at the run time. If one of these functions doesn't initialize the holes within the struct, it may cause an information leak from kernel space to userspace. 


My tools find similar cloned bugs
The same bug happen in /net/bluetooth/hci_sock.c

s
static int hci_sock_getname(struct socket *sock, struct sockaddr *addr,
			    int peer)
{
	struct sockaddr_hci *haddr = (struct sockaddr_hci *)addr;
	struct sock *sk = sock->sk;
	struct hci_dev *hdev;
	int err = 0;

	BT_DBG("sock %p sk %p", sock, sk);

	if (peer)
		return -EOPNOTSUPP;

	lock_sock(sk);

	hdev = hci_hdev_from_sock(sk);
	if (IS_ERR(hdev)) {
		err = PTR_ERR(hdev);
		goto done;
	}

	haddr->hci_family = AF_BLUETOOTH;
	haddr->hci_dev    = hdev->id;
	haddr->hci_channel= hci_pi(sk)->channel;
	err = sizeof(*haddr);

done:
	release_sock(sk);
	return err;
}
sockaddr_hci is declared here:

struct sockaddr_hci {
	sa_family_t    hci_family;
	unsigned short hci_dev;
	unsigned short hci_channel;
};

We can see there is a hole between family and dev. Thus, we have to explicitly set the hole to zero. Otherwise, the address will be passed to copy_to_user and cause information leakage.

Suggested patch:
memset(maddr, 0, sizeof(sockaddr_hci));


Thank you for the review. I appreciate your time. 

Andrew Bao