From ff2d7db848f8db7cade39e55f78f86d77e0de01a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 25 Feb 2008 21:40:51 -0800 Subject: SUNRPC: Ensure that we read all available tcp data Don't stop until we run out of data, or we hit an error. Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'net/sunrpc/xprtsock.c') diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 30e7ac243a90..8bd3b0f73ac0 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1073,6 +1073,7 @@ static void xs_tcp_data_ready(struct sock *sk, int bytes) { struct rpc_xprt *xprt; read_descriptor_t rd_desc; + int read; dprintk("RPC: xs_tcp_data_ready...\n"); @@ -1084,8 +1085,10 @@ static void xs_tcp_data_ready(struct sock *sk, int bytes) /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ rd_desc.arg.data = xprt; - rd_desc.count = 65536; - tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); + do { + rd_desc.count = 65536; + read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); + } while (read > 0); out: read_unlock(&sk->sk_callback_lock); } -- cgit v1.2.3 From b6ddf64ffe9d59577a9176856bb6fe69a539f573 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 17 Apr 2008 18:52:19 -0400 Subject: SUNRPC: Fix up xprt_write_space() The rest of the networking layer uses SOCK_ASYNC_NOSPACE to signal whether or not we have someone waiting for buffer memory. Convert the SUNRPC layer to use the same idiom. Remove the unlikely()s in xs_udp_write_space and xs_tcp_write_space. In fact, the most common case will be that there is nobody waiting for buffer space. SOCK_NOSPACE is there to tell the TCP layer whether or not the cwnd was limited by the application window. Ensure that we follow the same idiom as the rest of the networking layer here too. Finally, ensure that we clear SOCK_ASYNC_NOSPACE once we wake up, so that write_space() doesn't keep waking things up on xprt->pending. Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 61 ++++++++++++++++++++++++++++++++++----------------- 1 file changed, 41 insertions(+), 20 deletions(-) (limited to 'net/sunrpc/xprtsock.c') diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 8bd3b0f73ac0..4c2462e2f2b2 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -516,6 +516,14 @@ out: return sent; } +static void xs_nospace_callback(struct rpc_task *task) +{ + struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt); + + transport->inet->sk_write_pending--; + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); +} + /** * xs_nospace - place task on wait queue if transmit was incomplete * @task: task to put to sleep @@ -531,20 +539,27 @@ static void xs_nospace(struct rpc_task *task) task->tk_pid, req->rq_slen - req->rq_bytes_sent, req->rq_slen); - if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { - /* Protect against races with write_space */ - spin_lock_bh(&xprt->transport_lock); - - /* Don't race with disconnect */ - if (!xprt_connected(xprt)) - task->tk_status = -ENOTCONN; - else if (test_bit(SOCK_NOSPACE, &transport->sock->flags)) - xprt_wait_for_buffer_space(task); + /* Protect against races with write_space */ + spin_lock_bh(&xprt->transport_lock); + + /* Don't race with disconnect */ + if (xprt_connected(xprt)) { + if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { + /* + * Notify TCP that we're limited by the application + * window size + */ + set_bit(SOCK_NOSPACE, &transport->sock->flags); + transport->inet->sk_write_pending++; + /* ...and wait for more buffer space */ + xprt_wait_for_buffer_space(task, xs_nospace_callback); + } + } else { + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); + task->tk_status = -ENOTCONN; + } - spin_unlock_bh(&xprt->transport_lock); - } else - /* Keep holding the socket if it is blocked */ - rpc_delay(task, HZ>>4); + spin_unlock_bh(&xprt->transport_lock); } /** @@ -588,19 +603,20 @@ static int xs_udp_send_request(struct rpc_task *task) } switch (status) { + case -EAGAIN: + xs_nospace(task); + break; case -ENETUNREACH: case -EPIPE: case -ECONNREFUSED: /* When the server has died, an ICMP port unreachable message * prompts ECONNREFUSED. */ - break; - case -EAGAIN: - xs_nospace(task); + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); break; default: + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); dprintk("RPC: sendmsg returned unrecognized error %d\n", -status); - break; } return status; @@ -695,12 +711,13 @@ static int xs_tcp_send_request(struct rpc_task *task) case -ENOTCONN: case -EPIPE: status = -ENOTCONN; + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); break; default: dprintk("RPC: sendmsg returned unrecognized error %d\n", -status); + clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); xs_tcp_shutdown(xprt); - break; } return status; @@ -1189,9 +1206,11 @@ static void xs_udp_write_space(struct sock *sk) if (unlikely(!(sock = sk->sk_socket))) goto out; + clear_bit(SOCK_NOSPACE, &sock->flags); + if (unlikely(!(xprt = xprt_from_sock(sk)))) goto out; - if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) goto out; xprt_write_space(xprt); @@ -1222,9 +1241,11 @@ static void xs_tcp_write_space(struct sock *sk) if (unlikely(!(sock = sk->sk_socket))) goto out; + clear_bit(SOCK_NOSPACE, &sock->flags); + if (unlikely(!(xprt = xprt_from_sock(sk)))) goto out; - if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) goto out; xprt_write_space(xprt); -- cgit v1.2.3 From 06b4b681ababc20596aa947595714710f557131d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Apr 2008 16:51:38 -0400 Subject: SUNRPC: remove XS_SENDMSG_RETRY The condition for exiting from the loop in xs_tcp_send_request() should be that we find we're not making progress (i.e. number of bytes sent is 0). Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) (limited to 'net/sunrpc/xprtsock.c') diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 4c2462e2f2b2..4a567a93e6ad 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -135,12 +135,6 @@ static ctl_table sunrpc_table[] = { #endif -/* - * How many times to try sending a request on a socket before waiting - * for the socket buffer to clear. - */ -#define XS_SENDMSG_RETRY (10U) - /* * Time out for an RPC UDP socket connect. UDP socket connects are * synchronous, but we set a timeout anyway in case of resource @@ -666,7 +660,6 @@ static int xs_tcp_send_request(struct rpc_task *task) struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); struct xdr_buf *xdr = &req->rq_snd_buf; int status; - unsigned int retry = 0; xs_encode_tcp_record_marker(&req->rq_snd_buf); @@ -697,9 +690,10 @@ static int xs_tcp_send_request(struct rpc_task *task) return 0; } + if (status != 0) + continue; status = -EAGAIN; - if (retry++ > XS_SENDMSG_RETRY) - break; + break; } switch (status) { -- cgit v1.2.3 From 7c1d71cf56feebfb5b98219b9d11dfc3a2feca62 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 17 Apr 2008 16:52:57 -0400 Subject: SUNRPC: Don't disconnect more than once if retransmitting NFSv4 requests NFSv4 requires us to ensure that we break the TCP connection before we're allowed to retransmit a request. However in the case where we're retransmitting several requests that have been sent on the same connection, we need to ensure that we don't interfere with the attempt to reconnect and/or break the connection again once it has been established. We therefore introduce a 'connection' cookie that is bumped every time a connection is broken. This allows requests to track if they need to force a disconnection. Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'net/sunrpc/xprtsock.c') diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 4a567a93e6ad..63d79e347c00 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1142,6 +1142,7 @@ static void xs_tcp_state_change(struct sock *sk) break; case TCP_FIN_WAIT1: /* The client initiated a shutdown of the socket */ + xprt->connect_cookie++; xprt->reestablish_timeout = 0; set_bit(XPRT_CLOSING, &xprt->state); smp_mb__before_clear_bit(); @@ -1154,6 +1155,7 @@ static void xs_tcp_state_change(struct sock *sk) set_bit(XPRT_CLOSING, &xprt->state); xprt_force_disconnect(xprt); case TCP_SYN_SENT: + xprt->connect_cookie++; case TCP_CLOSING: /* * If the server closed down the connection, make sure that -- cgit v1.2.3