af_unix: fix races in sk_peer_pid and sk_peer_cred accesses

Jann Horn reported that SO_PEERCRED and SO_PEERGROUPS implementations
are racy, as af_unix can concurrently change sk_peer_pid and sk_peer_cred.

In order to fix this issue, this patch adds a new spinlock that needs
to be used whenever these fields are read or written.

Jann also pointed out that l2cap_sock_get_peer_pid_cb() is currently
reading sk->sk_peer_pid which makes no sense, as this field
is only possibly set by AF_UNIX sockets.
We will have to clean this in a separate patch.
This could be done by reverting b48596d "Bluetooth: L2CAP: Add get_peer_pid callback"
or implementing what was truly expected.

Fixes: 109f6e3 ("af_unix: Allow SO_PEERCRED to work across namespaces.")
Signed-off-by: Eric Dumazet <[email protected]>
Reported-by: Jann Horn <[email protected]>
Cc: Eric W. Biederman <[email protected]>
Cc: Luiz Augusto von Dentz <[email protected]>
Cc: Marcel Holtmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

include/net/sock.h

@@ -488,8 +488,10 @@ struct sock {
 	u8			sk_prefer_busy_poll;
 	u16			sk_busy_poll_budget;
 #endif
+	spinlock_t		sk_peer_lock;
 	struct pid		*sk_peer_pid;
 	const struct cred	*sk_peer_cred;
+
 	long			sk_rcvtimeo;
 	ktime_t			sk_stamp;
 #if BITS_PER_LONG==32

net/core/sock.c

@@ -1376,6 +1376,16 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 }
 EXPORT_SYMBOL(sock_setsockopt);
 
+static const struct cred *sk_get_peer_cred(struct sock *sk)
+{
+	const struct cred *cred;
+
+	spin_lock(&sk->sk_peer_lock);
+	cred = get_cred(sk->sk_peer_cred);
+	spin_unlock(&sk->sk_peer_lock);
+
+	return cred;
+}
 
 static void cred_to_ucred(struct pid *pid, const struct cred *cred,
 			  struct ucred *ucred)
@@ -1552,28 +1562,35 @@ int sock_getsockopt(struct socket *sock, int level, int optname,
 		struct ucred peercred;
 		if (len > sizeof(peercred))
 			len = sizeof(peercred);
+
+		spin_lock(&sk->sk_peer_lock);
 		cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
+		spin_unlock(&sk->sk_peer_lock);
+
 		if (copy_to_user(optval, &peercred, len))
 			return -EFAULT;
 		goto lenout;
 	}
 
 	case SO_PEERGROUPS:
 	{
+		const struct cred *cred;
 		int ret, n;
 
-		if (!sk->sk_peer_cred)
+		cred = sk_get_peer_cred(sk);
+		if (!cred)
 			return -ENODATA;
 
-		n = sk->sk_peer_cred->group_info->ngroups;
+		n = cred->group_info->ngroups;
 		if (len < n * sizeof(gid_t)) {
 			len = n * sizeof(gid_t);
+			put_cred(cred);
 			return put_user(len, optlen) ? -EFAULT : -ERANGE;
 		}
 		len = n * sizeof(gid_t);
 
-		ret = groups_to_user((gid_t __user *)optval,
-				     sk->sk_peer_cred->group_info);
+		ret = groups_to_user((gid_t __user *)optval, cred->group_info);
+		put_cred(cred);
 		if (ret)
 			return ret;
 		goto lenout;
@@ -1935,9 +1952,10 @@ static void __sk_destruct(struct rcu_head *head)
 		sk->sk_frag.page = NULL;
 	}
 
-	if (sk->sk_peer_cred)
-		put_cred(sk->sk_peer_cred);
+	/* We do not need to acquire sk->sk_peer_lock, we are the last user. */
+	put_cred(sk->sk_peer_cred);
 	put_pid(sk->sk_peer_pid);
+
 	if (likely(sk->sk_net_refcnt))
 		put_net(sock_net(sk));
 	sk_prot_free(sk->sk_prot_creator, sk);
@@ -3145,6 +3163,8 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 
 	sk->sk_peer_pid 	=	NULL;
 	sk->sk_peer_cred	=	NULL;
+	spin_lock_init(&sk->sk_peer_lock);
+
 	sk->sk_write_pending	=	0;
 	sk->sk_rcvlowat		=	1;
 	sk->sk_rcvtimeo		=	MAX_SCHEDULE_TIMEOUT;

net/unix/af_unix.c

@@ -608,20 +608,42 @@ static void unix_release_sock(struct sock *sk, int embrion)
 
 static void init_peercred(struct sock *sk)
 {
-	put_pid(sk->sk_peer_pid);
-	if (sk->sk_peer_cred)
-		put_cred(sk->sk_peer_cred);
+	const struct cred *old_cred;
+	struct pid *old_pid;
+
+	spin_lock(&sk->sk_peer_lock);
+	old_pid = sk->sk_peer_pid;
+	old_cred = sk->sk_peer_cred;
 	sk->sk_peer_pid  = get_pid(task_tgid(current));
 	sk->sk_peer_cred = get_current_cred();
+	spin_unlock(&sk->sk_peer_lock);
+
+	put_pid(old_pid);
+	put_cred(old_cred);
 }
 
 static void copy_peercred(struct sock *sk, struct sock *peersk)
 {
-	put_pid(sk->sk_peer_pid);
-	if (sk->sk_peer_cred)
-		put_cred(sk->sk_peer_cred);
+	const struct cred *old_cred;
+	struct pid *old_pid;
+
+	if (sk < peersk) {
+		spin_lock(&sk->sk_peer_lock);
+		spin_lock_nested(&peersk->sk_peer_lock, SINGLE_DEPTH_NESTING);
+	} else {
+		spin_lock(&peersk->sk_peer_lock);
+		spin_lock_nested(&sk->sk_peer_lock, SINGLE_DEPTH_NESTING);
+	}
+	old_pid = sk->sk_peer_pid;
+	old_cred = sk->sk_peer_cred;
 	sk->sk_peer_pid  = get_pid(peersk->sk_peer_pid);
 	sk->sk_peer_cred = get_cred(peersk->sk_peer_cred);
+
+	spin_unlock(&sk->sk_peer_lock);
+	spin_unlock(&peersk->sk_peer_lock);
+
+	put_pid(old_pid);
+	put_cred(old_cred);
 }
 
 static int unix_listen(struct socket *sock, int backlog)