mptcp: tag sequence_seq with socket state

Paolo Abeni suggested to avoid re-syncing new subflows because
they inherit options from listener. In case options were set on
listener but are not set on mptcp-socket there is no need to
do any synchronisation for new subflows.

This change sets sockopt_seq of new mptcp sockets to the seq of
the mptcp listener sock.

Subflow sequence is set to the embedded tcp listener sk.
Add a comment explaing why sk_state is involved in sockopt_seq
generation.

Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 1399d30..5cba909 100644 (file)
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -733,18 +733,23 @@ wake:
 static bool mptcp_do_flush_join_list(struct mptcp_sock *msk)
 {
        struct mptcp_subflow_context *subflow;
+       bool ret = false;
 
        if (likely(list_empty(&msk->join_list)))
                return false;
 
        spin_lock_bh(&msk->join_list_lock);
-       list_for_each_entry(subflow, &msk->join_list, node)
-               mptcp_propagate_sndbuf((struct sock *)msk, mptcp_subflow_tcp_sock(subflow));
+       list_for_each_entry(subflow, &msk->join_list, node) {
+               u32 sseq = READ_ONCE(subflow->setsockopt_seq);
 
+               mptcp_propagate_sndbuf((struct sock *)msk, mptcp_subflow_tcp_sock(subflow));
+               if (READ_ONCE(msk->setsockopt_seq) != sseq)
+                       ret = true;
+       }
        list_splice_tail_init(&msk->join_list, &msk->conn_list);
        spin_unlock_bh(&msk->join_list_lock);
 
-       return true;
+       return ret;
 }
 
 void __mptcp_flush_join_list(struct mptcp_sock *msk)
@@ -2718,6 +2723,7 @@ struct sock *mptcp_sk_clone(const struct sock *sk,
        msk->snd_nxt = msk->write_seq;
        msk->snd_una = msk->write_seq;
        msk->wnd_end = msk->snd_nxt + req->rsk_rcv_wnd;
+       msk->setsockopt_seq = mptcp_sk(sk)->setsockopt_seq;
 
        if (mp_opt->mp_capable) {
                msk->can_ack = true;

diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index 0186aad..df269c2 100644 (file)
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -256,6 +256,8 @@ struct mptcp_sock {
                u64     time;   /* start time of measurement window */
                u64     rtt_us; /* last maximum rtt of subflows */
        } rcvq_space;
+
+       u32 setsockopt_seq;
 };
 
 #define mptcp_lock_sock(___sk, cb) do {                                        \
@@ -414,6 +416,8 @@ struct mptcp_subflow_context {
        long    delegated_status;
        struct  list_head delegated_node;   /* link into delegated_action, protected by local BH */
 
+       u32 setsockopt_seq;
+
        struct  sock *tcp_sock;     /* tcp sk backpointer */
        struct  sock *conn;         /* parent mptcp_sock */
        const   struct inet_connection_sock_af_ops *icsk_af_ops;

diff --git a/net/mptcp/sockopt.c b/net/mptcp/sockopt.c
index 4fdc0ad..27b4954 100644 (file)
--- a/net/mptcp/sockopt.c
+++ b/net/mptcp/sockopt.c
@@ -24,6 +24,27 @@ static struct sock *__mptcp_tcp_fallback(struct mptcp_sock *msk)
        return msk->first;
 }
 
+static u32 sockopt_seq_reset(const struct sock *sk)
+{
+       sock_owned_by_me(sk);
+
+       /* Highbits contain state.  Allows to distinguish sockopt_seq
+        * of listener and established:
+        * s0 = new_listener()
+        * sockopt(s0) - seq is 1
+        * s1 = accept(s0) - s1 inherits seq 1 if listener sk (s0)
+        * sockopt(s0) - seq increments to 2 on s0
+        * sockopt(s1) // seq increments to 2 on s1 (different option)
+        * new ssk completes join, inherits options from s0 // seq 2
+        * Needs sync from mptcp join logic, but ssk->seq == msk->seq
+        *
+        * Set High order bits to sk_state so ssk->seq == msk->seq test
+        * will fail.
+        */
+
+       return (u32)sk->sk_state << 24u;
+}
+
 static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
                                       sockptr_t optval, unsigned int optlen)
 {
@@ -350,22 +371,44 @@ int mptcp_getsockopt(struct sock *sk, int level, int optname,
        return -EOPNOTSUPP;
 }
 
+static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
+{
+}
+
 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
 {
+       struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
+
        msk_owned_by_me(msk);
+
+       if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
+               __mptcp_sockopt_sync(msk, ssk);
+
+               subflow->setsockopt_seq = msk->setsockopt_seq;
+       }
 }
 
 void mptcp_sockopt_sync_all(struct mptcp_sock *msk)
 {
        struct mptcp_subflow_context *subflow;
+       struct sock *sk = (struct sock *)msk;
+       u32 seq;
 
-       msk_owned_by_me(msk);
+       seq = sockopt_seq_reset(sk);
 
        mptcp_for_each_subflow(msk, subflow) {
                struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
+               u32 sseq = READ_ONCE(subflow->setsockopt_seq);
 
-               mptcp_sockopt_sync(msk, ssk);
+               if (sseq != msk->setsockopt_seq) {
+                       __mptcp_sockopt_sync(msk, ssk);
+                       WRITE_ONCE(subflow->setsockopt_seq, seq);
+               } else if (sseq != seq) {
+                       WRITE_ONCE(subflow->setsockopt_seq, seq);
+               }
 
                cond_resched();
        }
+
+       msk->setsockopt_seq = seq;
 }

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 350c51c..c3da845 100644 (file)
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -679,6 +679,9 @@ create_child:
                        goto out;
                }
 
+               /* ssk inherits options of listener sk */
+               ctx->setsockopt_seq = listener->setsockopt_seq;
+
                if (ctx->mp_capable) {
                        /* this can't race with mptcp_close(), as the msk is
                         * not yet exposted to user-space
@@ -694,6 +697,7 @@ create_child:
                         * created mptcp socket
                         */
                        new_msk->sk_destruct = mptcp_sock_destruct;
+                       mptcp_sk(new_msk)->setsockopt_seq = ctx->setsockopt_seq;
                        mptcp_pm_new_connection(mptcp_sk(new_msk), child, 1);
                        mptcp_token_accept(subflow_req, mptcp_sk(new_msk));
                        ctx->conn = new_msk;