Andy Grover | 7004108 | 2009-08-21 12:28:31 +0000 | [diff] [blame] | 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 | /* |
| 41 | * cheesy, but simple.. |
| 42 | */ |
| 43 | static void rds_tcp_accept_worker(struct work_struct *work); |
| 44 | static DECLARE_WORK(rds_tcp_listen_work, rds_tcp_accept_worker); |
| 45 | static struct socket *rds_tcp_listen_sock; |
| 46 | |
| 47 | static int rds_tcp_accept_one(struct socket *sock) |
| 48 | { |
| 49 | struct socket *new_sock = NULL; |
| 50 | struct rds_connection *conn; |
| 51 | int ret; |
| 52 | struct inet_sock *inet; |
| 53 | |
| 54 | ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type, |
| 55 | sock->sk->sk_protocol, &new_sock); |
| 56 | if (ret) |
| 57 | goto out; |
| 58 | |
| 59 | new_sock->type = sock->type; |
| 60 | new_sock->ops = sock->ops; |
| 61 | ret = sock->ops->accept(sock, new_sock, O_NONBLOCK); |
| 62 | if (ret < 0) |
| 63 | goto out; |
| 64 | |
| 65 | rds_tcp_tune(new_sock); |
| 66 | |
| 67 | inet = inet_sk(new_sock->sk); |
| 68 | |
| 69 | rdsdebug("accepted tcp %u.%u.%u.%u:%u -> %u.%u.%u.%u:%u\n", |
| 70 | NIPQUAD(inet->saddr), ntohs(inet->sport), |
| 71 | NIPQUAD(inet->daddr), ntohs(inet->dport)); |
| 72 | |
| 73 | conn = rds_conn_create(inet->saddr, inet->daddr, &rds_tcp_transport, |
| 74 | GFP_KERNEL); |
| 75 | if (IS_ERR(conn)) { |
| 76 | ret = PTR_ERR(conn); |
| 77 | goto out; |
| 78 | } |
| 79 | |
| 80 | /* |
| 81 | * see the comment above rds_queue_delayed_reconnect() |
| 82 | */ |
| 83 | if (!rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING)) { |
| 84 | if (rds_conn_state(conn) == RDS_CONN_UP) |
| 85 | rds_tcp_stats_inc(s_tcp_listen_closed_stale); |
| 86 | else |
| 87 | rds_tcp_stats_inc(s_tcp_connect_raced); |
| 88 | rds_conn_drop(conn); |
| 89 | ret = 0; |
| 90 | goto out; |
| 91 | } |
| 92 | |
| 93 | rds_tcp_set_callbacks(new_sock, conn); |
| 94 | rds_connect_complete(conn); |
| 95 | new_sock = NULL; |
| 96 | ret = 0; |
| 97 | |
| 98 | out: |
| 99 | if (new_sock) |
| 100 | sock_release(new_sock); |
| 101 | return ret; |
| 102 | } |
| 103 | |
| 104 | static void rds_tcp_accept_worker(struct work_struct *work) |
| 105 | { |
| 106 | while (rds_tcp_accept_one(rds_tcp_listen_sock) == 0) |
| 107 | cond_resched(); |
| 108 | } |
| 109 | |
| 110 | void rds_tcp_listen_data_ready(struct sock *sk, int bytes) |
| 111 | { |
| 112 | void (*ready)(struct sock *sk, int bytes); |
| 113 | |
| 114 | rdsdebug("listen data ready sk %p\n", sk); |
| 115 | |
| 116 | read_lock(&sk->sk_callback_lock); |
| 117 | ready = sk->sk_user_data; |
| 118 | if (ready == NULL) { /* check for teardown race */ |
| 119 | ready = sk->sk_data_ready; |
| 120 | goto out; |
| 121 | } |
| 122 | |
| 123 | /* |
| 124 | * ->sk_data_ready is also called for a newly established child socket |
| 125 | * before it has been accepted and the accepter has set up their |
| 126 | * data_ready.. we only want to queue listen work for our listening |
| 127 | * socket |
| 128 | */ |
| 129 | if (sk->sk_state == TCP_LISTEN) |
| 130 | queue_work(rds_wq, &rds_tcp_listen_work); |
| 131 | |
| 132 | out: |
| 133 | read_unlock(&sk->sk_callback_lock); |
| 134 | ready(sk, bytes); |
| 135 | } |
| 136 | |
| 137 | int __init rds_tcp_listen_init(void) |
| 138 | { |
| 139 | struct sockaddr_in sin; |
| 140 | struct socket *sock = NULL; |
| 141 | int ret; |
| 142 | |
| 143 | ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock); |
| 144 | if (ret < 0) |
| 145 | goto out; |
| 146 | |
| 147 | sock->sk->sk_reuse = 1; |
| 148 | rds_tcp_nonagle(sock); |
| 149 | |
| 150 | write_lock_bh(&sock->sk->sk_callback_lock); |
| 151 | sock->sk->sk_user_data = sock->sk->sk_data_ready; |
| 152 | sock->sk->sk_data_ready = rds_tcp_listen_data_ready; |
| 153 | write_unlock_bh(&sock->sk->sk_callback_lock); |
| 154 | |
| 155 | sin.sin_family = PF_INET, |
| 156 | sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY); |
| 157 | sin.sin_port = (__force u16)htons(RDS_TCP_PORT); |
| 158 | |
| 159 | ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); |
| 160 | if (ret < 0) |
| 161 | goto out; |
| 162 | |
| 163 | ret = sock->ops->listen(sock, 64); |
| 164 | if (ret < 0) |
| 165 | goto out; |
| 166 | |
| 167 | rds_tcp_listen_sock = sock; |
| 168 | sock = NULL; |
| 169 | out: |
| 170 | if (sock) |
| 171 | sock_release(sock); |
| 172 | return ret; |
| 173 | } |
| 174 | |
| 175 | void rds_tcp_listen_stop(void) |
| 176 | { |
| 177 | struct socket *sock = rds_tcp_listen_sock; |
| 178 | struct sock *sk; |
| 179 | |
| 180 | if (sock == NULL) |
| 181 | return; |
| 182 | |
| 183 | sk = sock->sk; |
| 184 | |
| 185 | /* serialize with and prevent further callbacks */ |
| 186 | lock_sock(sk); |
| 187 | write_lock_bh(&sk->sk_callback_lock); |
| 188 | if (sk->sk_user_data) { |
| 189 | sk->sk_data_ready = sk->sk_user_data; |
| 190 | sk->sk_user_data = NULL; |
| 191 | } |
| 192 | write_unlock_bh(&sk->sk_callback_lock); |
| 193 | release_sock(sk); |
| 194 | |
| 195 | /* wait for accepts to stop and close the socket */ |
| 196 | flush_workqueue(rds_wq); |
| 197 | sock_release(sock); |
| 198 | rds_tcp_listen_sock = NULL; |
| 199 | } |