On Sat, Jun 4, 2016 at 1:49 AM, Marcelo Ricardo Leitner
<marcelo.leit...@gmail.com> wrote:
> On Fri, Jun 03, 2016 at 10:42:45PM +0800, Xin Long wrote:
>> Now sctp doesn't change socket state upon shutdown reception. It changes
>> just the assoc state, even though it's a TCP-style socket.
>>
>> For some cases, if we really need to check sk->sk_state, it's necessary to
>> fix this issue, at least when we use ss or netstat to dump, we can get a
>> more exact information.
>>
>> As an improvement, we will change sk->sk_state when we change asoc->state
>> to SHUTDOWN_RECEIVED, and also do it in sctp_shutdown to keep consistent with
>> sctp_close.
>>
>> Signed-off-by: Xin Long <lucien....@gmail.com>
>> ---
>> net/sctp/sm_sideeffect.c | 4 +++-
>> net/sctp/socket.c | 1 +
>> 2 files changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/sctp/sm_sideeffect.c b/net/sctp/sm_sideeffect.c
>> index aa37122..12d4519 100644
>> --- a/net/sctp/sm_sideeffect.c
>> +++ b/net/sctp/sm_sideeffect.c
>> @@ -806,8 +806,10 @@ static void sctp_cmd_new_state(sctp_cmd_seq_t *cmds,
>>
>> /* Set the RCV_SHUTDOWN flag when a SHUTDOWN is received. */
>> if (sctp_state(asoc, SHUTDOWN_RECEIVED) &&
>> - sctp_sstate(sk, ESTABLISHED))
>> + sctp_sstate(sk, ESTABLISHED)) {
>> + sk->sk_state = SCTP_SS_CLOSING;
>> sk->sk_shutdown |= RCV_SHUTDOWN;
>> + }
>> }
>>
>> if (sctp_state(asoc, COOKIE_WAIT)) {
>> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
>> index 67154b8..91a460d 100644
>> --- a/net/sctp/socket.c
>> +++ b/net/sctp/socket.c
>> @@ -4193,6 +4193,7 @@ static void sctp_shutdown(struct sock *sk, int how)
>> return;
>>
>> if (how & SEND_SHUTDOWN) {
>> + sk->sk_state = SCTP_SS_CLOSING;
>> ep = sctp_sk(sk)->ep;
>> if (!list_empty(&ep->asocs)) {
>> asoc = list_entry(ep->asocs.next,
>
> Oh but I think the removed chunk was important, it just needed fixing.
>
> A comment in there said:
> /* If the association on the newsk is already closed before
> * accept() is called, set RCV_SHUTDOWN flag.
> */
> Meaning that there is a way that it will create a socket in the state
> you're trying to avoid here, which now is not handled.
agreed, but it's another problem
there, we'd better free newsk and assoc after sctp_sock_migrate,
and let sctp_accept return err. I will think more about it and see if
I cook another patch for it.
>
> Marcelo
