1 /* 2 * Copyright (c) 2006 Oracle. All rights reserved. 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 18 * - Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials 21 * provided with the distribution. 22 * 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 30 * SOFTWARE. 31 * 32 */ 33 #include <linux/kernel.h> 34 #include <linux/in.h> 35 #include <net/tcp.h> 36 37 #include "rds.h" 38 #include "tcp.h" 39 40 void rds_tcp_state_change(struct sock *sk) 41 { 42 void (*state_change)(struct sock *sk); 43 struct rds_conn_path *cp; 44 struct rds_tcp_connection *tc; 45 46 read_lock_bh(&sk->sk_callback_lock); 47 cp = sk->sk_user_data; 48 if (!cp) { 49 state_change = sk->sk_state_change; 50 goto out; 51 } 52 tc = cp->cp_transport_data; 53 state_change = tc->t_orig_state_change; 54 55 rdsdebug("sock %p state_change to %d\n", tc->t_sock, sk->sk_state); 56 57 switch (sk->sk_state) { 58 /* ignore connecting sockets as they make progress */ 59 case TCP_SYN_SENT: 60 case TCP_SYN_RECV: 61 break; 62 case TCP_ESTABLISHED: 63 rds_connect_path_complete(cp, RDS_CONN_CONNECTING); 64 break; 65 case TCP_CLOSE_WAIT: 66 case TCP_CLOSE: 67 rds_conn_path_drop(cp); 68 default: 69 break; 70 } 71 out: 72 read_unlock_bh(&sk->sk_callback_lock); 73 state_change(sk); 74 } 75 76 int rds_tcp_conn_path_connect(struct rds_conn_path *cp) 77 { 78 struct socket *sock = NULL; 79 struct sockaddr_in src, dest; 80 int ret; 81 struct rds_connection *conn = cp->cp_conn; 82 struct rds_tcp_connection *tc = cp->cp_transport_data; 83 84 /* for multipath rds,we only trigger the connection after 85 * the handshake probe has determined the number of paths. 86 */ 87 if (cp->cp_index > 0 && cp->cp_conn->c_npaths < 2) 88 return -EAGAIN; 89 90 mutex_lock(&tc->t_conn_path_lock); 91 92 if (rds_conn_path_up(cp)) { 93 mutex_unlock(&tc->t_conn_path_lock); 94 return 0; 95 } 96 ret = sock_create_kern(rds_conn_net(conn), PF_INET, 97 SOCK_STREAM, IPPROTO_TCP, &sock); 98 if (ret < 0) 99 goto out; 100 101 rds_tcp_tune(sock); 102 103 src.sin_family = AF_INET; 104 src.sin_addr.s_addr = (__force u32)conn->c_laddr; 105 src.sin_port = (__force u16)htons(0); 106 107 ret = sock->ops->bind(sock, (struct sockaddr *)&src, sizeof(src)); 108 if (ret) { 109 rdsdebug("bind failed with %d at address %pI4\n", 110 ret, &conn->c_laddr); 111 goto out; 112 } 113 114 dest.sin_family = AF_INET; 115 dest.sin_addr.s_addr = (__force u32)conn->c_faddr; 116 dest.sin_port = (__force u16)htons(RDS_TCP_PORT); 117 118 /* 119 * once we call connect() we can start getting callbacks and they 120 * own the socket 121 */ 122 rds_tcp_set_callbacks(sock, cp); 123 ret = sock->ops->connect(sock, (struct sockaddr *)&dest, sizeof(dest), 124 O_NONBLOCK); 125 126 cp->cp_outgoing = 1; 127 rdsdebug("connect to address %pI4 returned %d\n", &conn->c_faddr, ret); 128 if (ret == -EINPROGRESS) 129 ret = 0; 130 if (ret == 0) { 131 rds_tcp_keepalive(sock); 132 sock = NULL; 133 } else { 134 rds_tcp_restore_callbacks(sock, cp->cp_transport_data); 135 } 136 137 out: 138 mutex_unlock(&tc->t_conn_path_lock); 139 if (sock) 140 sock_release(sock); 141 return ret; 142 } 143 144 /* 145 * Before killing the tcp socket this needs to serialize with callbacks. The 146 * caller has already grabbed the sending sem so we're serialized with other 147 * senders. 148 * 149 * TCP calls the callbacks with the sock lock so we hold it while we reset the 150 * callbacks to those set by TCP. Our callbacks won't execute again once we 151 * hold the sock lock. 152 */ 153 void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp) 154 { 155 struct rds_tcp_connection *tc = cp->cp_transport_data; 156 struct socket *sock = tc->t_sock; 157 158 rdsdebug("shutting down conn %p tc %p sock %p\n", 159 cp->cp_conn, tc, sock); 160 161 if (sock) { 162 sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN); 163 lock_sock(sock->sk); 164 rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */ 165 166 release_sock(sock->sk); 167 sock_release(sock); 168 } 169 170 if (tc->t_tinc) { 171 rds_inc_put(&tc->t_tinc->ti_inc); 172 tc->t_tinc = NULL; 173 } 174 tc->t_tinc_hdr_rem = sizeof(struct rds_header); 175 tc->t_tinc_data_rem = 0; 176 } 177