net/smc: guarantee reset of write_blocked for heavy workload
authorUrsula Braun <ubraun@linux.vnet.ibm.com>
Mon, 10 Apr 2017 12:57:59 +0000 (14:57 +0200)
committerDavid S. Miller <davem@davemloft.net>
Wed, 12 Apr 2017 03:01:14 +0000 (23:01 -0400)
If peer indicates write_blocked, the cursor state of the received data
should be send to the peer immediately (in smc_tx_consumer_update()).
Afterwards the write_blocked indicator is cleared.

If there is no free slot for another write request, sending is postponed
to worker smc_tx_work, and the write_blocked indicator is not cleared.
Therefore another clearing check is needed in smc_tx_work().

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Reviewed-by: Thomas Richter <tmricht@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/smc/smc_tx.c

index 69a0013dd25cecbee0658168da577d15ed8913d8..21ec1832ab517d647ac8942ddbd5484492eb4c94 100644 (file)
@@ -431,9 +431,13 @@ static void smc_tx_work(struct work_struct *work)
                                                   struct smc_connection,
                                                   tx_work);
        struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
+       int rc;
 
        lock_sock(&smc->sk);
-       smc_tx_sndbuf_nonempty(conn);
+       rc = smc_tx_sndbuf_nonempty(conn);
+       if (!rc && conn->local_rx_ctrl.prod_flags.write_blocked &&
+           !atomic_read(&conn->bytes_to_rcv))
+               conn->local_rx_ctrl.prod_flags.write_blocked = 0;
        release_sock(&smc->sk);
 }