Re: [RFC][PATCH v2 19/31] timers: net: Use del_timer_shutdown() before freeing timer

[Eric Dumazet]

On Thu, Nov 3, 2022 at 2:51 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Sun, 30 Oct 2022 18:22:03 +0100
> Paolo Abeni <pabeni@xxxxxxxxxx> wrote:
>
> > On the positive side, I think converting the sk_stop_timer in
> > inet_csk_clear_xmit_timers() should be safe and should cover the issue
> > reported by Guenter
>
> Would something like this be OK?
>
> [ Note, talking with Thomas Gleixner, we agreed that we are changing the
>   name to: time_shutdown_sync() and timer_shutdown() (no wait version).
>   I'll be posting new patches soon. ]
>
> -- Steve
>
> diff --git a/include/net/sock.h b/include/net/sock.h
> index 22f8bab583dd..0ef58697d4e5 100644
> --- a/include/net/sock.h
> +++ b/include/net/sock.h
> @@ -2439,6 +2439,8 @@ void sk_stop_timer(struct sock *sk, struct timer_list *timer);
>
>  void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer);
>
> +void sk_shutdown_timer(struct sock *sk, struct timer_list *timer);
> +
>  int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
>                         struct sk_buff *skb, unsigned int flags,
>                         void (*destructor)(struct sock *sk,
> diff --git a/net/core/sock.c b/net/core/sock.c
> index a3ba0358c77c..82124862b594 100644
> --- a/net/core/sock.c
> +++ b/net/core/sock.c
> @@ -3357,6 +3357,13 @@ void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer)
>  }
>  EXPORT_SYMBOL(sk_stop_timer_sync);
>
> +void sk_shutdown_timer(struct sock *sk, struct timer_list* timer)
> +{
> +       if (timer_shutdown(timer))
> +               __sock_put(sk);
> +}
> +EXPORT_SYMBOL(sk_shutdown_timer);
> +
>  void sock_init_data(struct socket *sock, struct sock *sk)
>  {
>         sk_init_common(sk);
> diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
> index 5e70228c5ae9..71f398f51958 100644
> --- a/net/ipv4/inet_connection_sock.c
> +++ b/net/ipv4/inet_connection_sock.c
> @@ -722,15 +722,15 @@ void inet_csk_clear_xmit_timers(struct sock *sk)
>
>         icsk->icsk_pending = icsk->icsk_ack.pending = 0;
>
> -       sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
> -       sk_stop_timer(sk, &icsk->icsk_delack_timer);
> -       sk_stop_timer(sk, &sk->sk_timer);
> +       sk_shutdown_timer(sk, &icsk->icsk_retransmit_timer);
> +       sk_shutdown_timer(sk, &icsk->icsk_delack_timer);
> +       sk_shutdown_timer(sk, &sk->sk_timer);
>  }
>  EXPORT_SYMBOL(inet_csk_clear_xmit_timers);

 inet_csk_clear_xmit_timers() can be called multiple times during TCP
socket lifetime.

(See tcp_disconnect(), which can be followed by another connect() ... and loop)

Maybe add a second parameter, or add a new
inet_csk_shutdown_xmit_timers() only called from tcp_v4_destroy_sock() ?

>
>  void inet_csk_delete_keepalive_timer(struct sock *sk)
>  {
> -       sk_stop_timer(sk, &sk->sk_timer);
> +       sk_shutdown_timer(sk, &sk->sk_timer);

SO_KEEPALIVE can be called multiple times in a TCP socket lifetime,
on/off/on/off/...

I suggest leaving sk_stop_timer() here.

Eventually  inet_csk_clear_xmit_timers( sk, destroy=true) (or
inet_csk_shutdown_xmit_timers(())
   will  be called before the socket is destroyed.