Some protocols using TCP encapsulation (e.g., espintcp, openvpn) deliver userspace-bound packets through a custom skb queue rather than the standard sk_receive_queue. Introduce datagram_poll_queue that accepts an explicit receive queue, and convert datagram_poll into a wrapper around datagram_poll_queue. This allows protocols with custom skb queues to reuse the core polling logic without relying on sk_receive_queue. Cc: Sabrina Dubroca Cc: Antonio Quartulli Signed-off-by: Ralf Lici --- include/linux/skbuff.h | 3 +++ net/core/datagram.c | 44 ++++++++++++++++++++++++++++++++---------- 2 files changed, 37 insertions(+), 10 deletions(-) diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index fb3fec9affaa..a7cc3d1f4fd1 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -4204,6 +4204,9 @@ struct sk_buff *__skb_recv_datagram(struct sock *sk, struct sk_buff_head *sk_queue, unsigned int flags, int *off, int *err); struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err); +__poll_t datagram_poll_queue(struct file *file, struct socket *sock, + struct poll_table_struct *wait, + struct sk_buff_head *rcv_queue); __poll_t datagram_poll(struct file *file, struct socket *sock, struct poll_table_struct *wait); int skb_copy_datagram_iter(const struct sk_buff *from, int offset, diff --git a/net/core/datagram.c b/net/core/datagram.c index cb4b9ef2e4e3..c285c6465923 100644 --- a/net/core/datagram.c +++ b/net/core/datagram.c @@ -920,21 +920,22 @@ int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, EXPORT_SYMBOL(skb_copy_and_csum_datagram_msg); /** - * datagram_poll - generic datagram poll + * datagram_poll_queue - same as datagram_poll, but on a specific receive + * queue * @file: file struct * @sock: socket * @wait: poll table + * @rcv_queue: receive queue to poll * - * Datagram poll: Again totally generic. This also handles - * sequenced packet sockets providing the socket receive queue - * is only ever holding data ready to receive. + * Performs polling on the given receive queue, handling shutdown, error, + * and connection state. This is useful for protocols that deliver + * userspace-bound packets through a custom queue instead of + * sk->sk_receive_queue. * - * Note: when you *don't* use this routine for this protocol, - * and you use a different write policy from sock_writeable() - * then please supply your own write_space callback. + * Return: poll bitmask indicating the socket's current state */ -__poll_t datagram_poll(struct file *file, struct socket *sock, - poll_table *wait) +__poll_t datagram_poll_queue(struct file *file, struct socket *sock, + poll_table *wait, struct sk_buff_head *rcv_queue) { struct sock *sk = sock->sk; __poll_t mask; @@ -956,7 +957,7 @@ __poll_t datagram_poll(struct file *file, struct socket *sock, mask |= EPOLLHUP; /* readable? */ - if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) + if (!skb_queue_empty_lockless(rcv_queue)) mask |= EPOLLIN | EPOLLRDNORM; /* Connection-based need to check for termination and startup */ @@ -978,4 +979,27 @@ __poll_t datagram_poll(struct file *file, struct socket *sock, return mask; } +EXPORT_SYMBOL(datagram_poll_queue); + +/** + * datagram_poll - generic datagram poll + * @file: file struct + * @sock: socket + * @wait: poll table + * + * Datagram poll: Again totally generic. This also handles + * sequenced packet sockets providing the socket receive queue + * is only ever holding data ready to receive. + * + * Note: when you *don't* use this routine for this protocol, + * and you use a different write policy from sock_writeable() + * then please supply your own write_space callback. + * + * Return: poll bitmask indicating the socket's current state + */ +__poll_t datagram_poll(struct file *file, struct socket *sock, poll_table *wait) +{ + return datagram_poll_queue(file, sock, wait, + &sock->sk->sk_receive_queue); +} EXPORT_SYMBOL(datagram_poll); -- 2.51.0 espintcp uses a custom queue (ike_queue) to deliver packets to userspace. The polling logic relies on datagram_poll, which checks sk_receive_queue, which can lead to false readiness signals when that queue contains non-userspace packets. Switch espintcp_poll to use datagram_poll_queue with ike_queue, ensuring poll only signals readiness when userspace data is actually available. Fixes: e27cca96cd68 ("xfrm: add espintcp (RFC 8229)") Signed-off-by: Ralf Lici --- net/xfrm/espintcp.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/net/xfrm/espintcp.c b/net/xfrm/espintcp.c index fc7a603b04f1..bf744ac9d5a7 100644 --- a/net/xfrm/espintcp.c +++ b/net/xfrm/espintcp.c @@ -555,14 +555,10 @@ static void espintcp_close(struct sock *sk, long timeout) static __poll_t espintcp_poll(struct file *file, struct socket *sock, poll_table *wait) { - __poll_t mask = datagram_poll(file, sock, wait); struct sock *sk = sock->sk; struct espintcp_ctx *ctx = espintcp_getctx(sk); - if (!skb_queue_empty(&ctx->ike_queue)) - mask |= EPOLLIN | EPOLLRDNORM; - - return mask; + return datagram_poll_queue(file, sock, wait, &ctx->ike_queue); } static void build_protos(struct proto *espintcp_prot, -- 2.51.0 openvpn TCP encapsulation uses a custom queue to deliver packets to userspace. Currently it relies on datagram_poll, which checks sk_receive_queue, leading to false readiness signals when that queue contains non-userspace packets. Switch ovpn_tcp_poll to use datagram_poll_queue with the peer's user_queue, ensuring poll only signals readiness when userspace data is actually available. Also refactor ovpn_tcp_poll in order to enforce the assumption we can make on the lifetime of ovpn_sock and peer. Fixes: 11851cbd60ea ("ovpn: implement TCP transport") Signed-off-by: Antonio Quartulli Signed-off-by: Ralf Lici --- drivers/net/ovpn/tcp.c | 26 ++++++++++++++++++++++---- 1 file changed, 22 insertions(+), 4 deletions(-) diff --git a/drivers/net/ovpn/tcp.c b/drivers/net/ovpn/tcp.c index 289f62c5d2c7..0d7f30360d87 100644 --- a/drivers/net/ovpn/tcp.c +++ b/drivers/net/ovpn/tcp.c @@ -560,16 +560,34 @@ static void ovpn_tcp_close(struct sock *sk, long timeout) static __poll_t ovpn_tcp_poll(struct file *file, struct socket *sock, poll_table *wait) { - __poll_t mask = datagram_poll(file, sock, wait); + struct sk_buff_head *queue = &sock->sk->sk_receive_queue; struct ovpn_socket *ovpn_sock; + struct ovpn_peer *peer = NULL; + __poll_t mask; rcu_read_lock(); ovpn_sock = rcu_dereference_sk_user_data(sock->sk); - if (ovpn_sock && ovpn_sock->peer && - !skb_queue_empty(&ovpn_sock->peer->tcp.user_queue)) - mask |= EPOLLIN | EPOLLRDNORM; + /* if we landed in this callback, we expect to have a + * meaningful state. The ovpn_socket lifecycle would + * prevent it otherwise. + */ + if (WARN(!ovpn_sock || !ovpn_sock->peer, + "ovpn: null state in ovpn_tcp_poll!")) { + rcu_read_unlock(); + return 0; + } + + if (ovpn_peer_hold(ovpn_sock->peer)) { + peer = ovpn_sock->peer; + queue = &peer->tcp.user_queue; + } rcu_read_unlock(); + mask = datagram_poll_queue(file, sock, wait, queue); + + if (peer) + ovpn_peer_put(peer); + return mask; } -- 2.51.0