Skip to contentSkip to conversation

net/rose: Fix to not accept on connected socket

PATCH Superseded Hyunwoo Kim 2 participants 3 messages Jan 23, 2023 – Jan 25, 2023

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 net/rose: Fix to not accept on connected socket

If listen() and accept() are called on a rose socket that connect() is successful, accept() succeeds immediately. This is because rose_connect() queues the skb to sk->sk_receive_queue, and rose_accept() dequeues it.

This creates a child socket with the sk of the parent rose socket, which can cause confusion.

Fix rose_listen() to return -EINVAL if the socket has already been successfully connected, and add lock_sock to prevent this issue.

---
 net/rose/af_rose.c | 8 ++++++++
 1 file changed, 8 insertions(+)
net/rose/af_rose.c +8 -0
index 36fefc3957d7..ca2b17f32670 100644
@@ -488,6 +488,12 @@ static int rose_listen(struct socket *sock, int backlog)
 {
 	struct sock *sk = sock->sk;
 
+	lock_sock(sk);
+	if (sock->state != SS_UNCONNECTED) {
+		release_sock(sk);
+		return -EINVAL;
+	}
+
 	if (sk->sk_state != TCP_LISTEN) {
 		struct rose_sock *rose = rose_sk(sk);
 
@@ -497,8 +503,10 @@ static int rose_listen(struct socket *sock, int backlog)
 		memset(rose->dest_digis, 0, AX25_ADDR_LEN * ROSE_MAX_DIGIS);
 		sk->sk_max_ack_backlog = backlog;
 		sk->sk_state           = TCP_LISTEN;
+		release_sock(sk);
 		return 0;
 	}
+	release_sock(sk);
 
 	return -EOPNOTSUPP;
 }
Signature · 1 line
2.25.1
Reviewed-by on v2 1/1
#1 lore raw
From:   Hyunwoo Kim <[email protected]>
Date:   Mon, 23 Jan 2023 11:40:20 -0800
Quoted 5 lines

If listen() and accept() are called on a rose socket that connect() is successful, accept() succeeds immediately. This is because rose_connect() queues the skb to sk->sk_receive_queue, and rose_accept() dequeues it.

Quoted 8 lines

This creates a child socket with the sk of the parent rose socket, which can cause confusion.

Fix rose_listen() to return -EINVAL if the socket has already been successfully connected, and add lock_sock to prevent this issue.

Quoted 34 lines
---
 net/rose/af_rose.c | 8 ++++++++
 1 file changed, 8 insertions(+)
net/rose/af_rose.c +8 -0
index 36fefc3957d7..ca2b17f32670 100644
@@ -488,6 +488,12 @@ static int rose_listen(struct socket *sock, int backlog)
 {
 	struct sock *sk = sock->sk;
 
+	lock_sock(sk);
+	if (sock->state != SS_UNCONNECTED) {
+		release_sock(sk);
+		return -EINVAL;
+	}
+
 	if (sk->sk_state != TCP_LISTEN) {
 		struct rose_sock *rose = rose_sk(sk);
 
@@ -497,8 +503,10 @@ static int rose_listen(struct socket *sock, int backlog)
 		memset(rose->dest_digis, 0, AX25_ADDR_LEN * ROSE_MAX_DIGIS);
 		sk->sk_max_ack_backlog = backlog;
 		sk->sk_state           = TCP_LISTEN;
+		release_sock(sk);
 		return 0;
 	}
+	release_sock(sk);
 
 	return -EOPNOTSUPP;
 }
Signature · 1 line
2.25.1
#2 lore raw
From:   Kuniyuki Iwashima <[email protected]>
Date:   Tue, 24 Jan 2023 18:08:09 -0800
Quoted 6 lines
From:   Hyunwoo Kim <[email protected]>
Date:   Mon, 23 Jan 2023 11:40:20 -0800
Quoted 4 lines

If listen() and accept() are called on a rose socket that connect() is successful, accept() succeeds immediately. This is because rose_connect() queues the skb to sk->sk_receive_queue, and rose_accept() dequeues it.

Same comment for the netrom patch here.
https://lore.kernel.org/netdev/[email protected]/

The skb which the problematic accept() dequeues is created by sendmsg(), not connect(), right ?

Quoted 47 lines
Quoted 9 lines

This creates a child socket with the sk of the parent rose socket, which can cause confusion.

Fix rose_listen() to return -EINVAL if the socket has already been successfully connected, and add lock_sock to prevent this issue.

Quoted 34 lines
---
 net/rose/af_rose.c | 8 ++++++++
 1 file changed, 8 insertions(+)
net/rose/af_rose.c +8 -0
index 36fefc3957d7..ca2b17f32670 100644
@@ -488,6 +488,12 @@ static int rose_listen(struct socket *sock, int backlog)
 {
 	struct sock *sk = sock->sk;
 
+	lock_sock(sk);
+	if (sock->state != SS_UNCONNECTED) {
+		release_sock(sk);
+		return -EINVAL;
+	}
+
 	if (sk->sk_state != TCP_LISTEN) {
 		struct rose_sock *rose = rose_sk(sk);
 
@@ -497,8 +503,10 @@ static int rose_listen(struct socket *sock, int backlog)
 		memset(rose->dest_digis, 0, AX25_ADDR_LEN * ROSE_MAX_DIGIS);
 		sk->sk_max_ack_backlog = backlog;
 		sk->sk_state           = TCP_LISTEN;
+		release_sock(sk);
 		return 0;
 	}
+	release_sock(sk);
 
 	return -EOPNOTSUPP;
 }
Signature · 1 line
2.25.1