On Thu, 13 Dec 2018 16:31:27 +0100
Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx> wrote:
> Currently, (E)TP transfers will be aborted, as soon application will
> call close() or exit, as the socket will be automatically closed by the
> kernel.
>
> Signed-off-by: Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx>
> ---
> net/can/j1939/socket.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/net/can/j1939/socket.c b/net/can/j1939/socket.c
> index d6332483350f..e003d3cae24c 100644
> --- a/net/can/j1939/socket.c
> +++ b/net/can/j1939/socket.c
> @@ -101,6 +101,13 @@ static inline void j1939_sock_pending_add(struct sock *sk)
> atomic_inc(&jsk->skb_pending);
> }
>
> +static int j1939_sock_pending_get(struct sock *sk)
> +{
> + struct j1939_sock *jsk = j1939_sk(sk);
> +
> + return atomic_read(&jsk->skb_pending);
> +}
> +
> void j1939_sock_pending_del(struct sock *sk)
> {
> struct j1939_sock *jsk = j1939_sk(sk);
> @@ -404,6 +411,9 @@ static int j1939_sk_release(struct socket *sock)
> if (jsk->state & J1939_SOCK_BOUND) {
> struct net_device *ndev;
>
> + wait_event_interruptible(jsk->waitq,
> + j1939_sock_pending_get(&jsk->sk) == 0);
> +
> spin_lock_bh(&j1939_socks_lock);
> list_del_init(&jsk->list);
> spin_unlock_bh(&j1939_socks_lock);
close() now blocks if data was successfully send on that socket.
This is probably due to a resource issue which is now brought to light.
Seems as if jsk->skb_pending is incremented on every j1939_sk_sendmsg()
call, but only decremented when an error occurs. So when a send()
succeeds, close() will block indefinitely.
Robin
