net/rds/tcp_listen.c

   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/gfp.h>
  35 #include <linux/in.h>
  36 #include <net/tcp.h>
  37
  38 #include "rds.h"
  39 #include "tcp.h"
  40
  41 int rds_tcp_keepalive(struct socket *sock)
  42 {
  43         /* values below based on xs_udp_default_timeout */
  44         int keepidle = 5; /* send a probe 'keepidle' secs after last data */
  45         int keepcnt = 5; /* number of unack'ed probes before declaring dead */
  46         int keepalive = 1;
  47         int ret = 0;
  48
  49         ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
  50                                 (char *)&keepalive, sizeof(keepalive));
  51         if (ret < 0)
  52                 goto bail;
  53
  54         ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
  55                                 (char *)&keepcnt, sizeof(keepcnt));
  56         if (ret < 0)
  57                 goto bail;
  58
  59         ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
  60                                 (char *)&keepidle, sizeof(keepidle));
  61         if (ret < 0)
  62                 goto bail;
  63
  64         /* KEEPINTVL is the interval between successive probes. We follow
  65          * the model in xs_tcp_finish_connecting() and re-use keepidle.
  66          */
  67         ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
  68                                 (char *)&keepidle, sizeof(keepidle));
  69 bail:
  70         return ret;
  71 }
  72
  73 int rds_tcp_accept_one(struct socket *sock)
  74 {
  75         struct socket *new_sock = NULL;
  76         struct rds_connection *conn;
  77         int ret;
  78         struct inet_sock *inet;
  79         struct rds_tcp_connection *rs_tcp;
  80
  81         ret = sock_create_kern(sock_net(sock->sk), sock->sk->sk_family,
  82                                sock->sk->sk_type, sock->sk->sk_protocol,
  83                                &new_sock);
  84         if (ret)
  85                 goto out;
  86
  87         new_sock->type = sock->type;
  88         new_sock->ops = sock->ops;
  89         ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
  90         if (ret < 0)
  91                 goto out;
  92
  93         ret = rds_tcp_keepalive(new_sock);
  94         if (ret < 0)
  95                 goto out;
  96
  97         rds_tcp_tune(new_sock);
  98
  99         inet = inet_sk(new_sock->sk);
 100
 101         rdsdebug("accepted tcp %pI4:%u -> %pI4:%u\n",
 102                  &inet->inet_saddr, ntohs(inet->inet_sport),
 103                  &inet->inet_daddr, ntohs(inet->inet_dport));
 104
 105         conn = rds_conn_create(sock_net(sock->sk),
 106                                inet->inet_saddr, inet->inet_daddr,
 107                                &rds_tcp_transport, GFP_KERNEL);
 108         if (IS_ERR(conn)) {
 109                 ret = PTR_ERR(conn);
 110                 goto out;
 111         }
 112         /* An incoming SYN request came in, and TCP just accepted it.
 113          * We always create a new conn for listen side of TCP, and do not
 114          * add it to the c_hash_list.
 115          *
 116          * If the client reboots, this conn will need to be cleaned up.
 117          * rds_tcp_state_change() will do that cleanup
 118          */
 119         rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
 120         WARN_ON(!rs_tcp || rs_tcp->t_sock);
 121
 122         /*
 123          * see the comment above rds_queue_delayed_reconnect()
 124          */
 125         if (!rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING)) {
 126                 if (rds_conn_state(conn) == RDS_CONN_UP)
 127                         rds_tcp_stats_inc(s_tcp_listen_closed_stale);
 128                 else
 129                         rds_tcp_stats_inc(s_tcp_connect_raced);
 130                 rds_conn_drop(conn);
 131                 ret = 0;
 132                 goto out;
 133         }
 134
 135         rds_tcp_set_callbacks(new_sock, conn);
 136         rds_connect_complete(conn);
 137         new_sock = NULL;
 138         ret = 0;
 139
 140 out:
 141         if (new_sock)
 142                 sock_release(new_sock);
 143         return ret;
 144 }
 145
 146 void rds_tcp_listen_data_ready(struct sock *sk)
 147 {
 148         void (*ready)(struct sock *sk);
 149
 150         rdsdebug("listen data ready sk %p\n", sk);
 151
 152         read_lock(&sk->sk_callback_lock);
 153         ready = sk->sk_user_data;
 154         if (!ready) { /* check for teardown race */
 155                 ready = sk->sk_data_ready;
 156                 goto out;
 157         }
 158
 159         /*
 160          * ->sk_data_ready is also called for a newly established child socket
 161          * before it has been accepted and the accepter has set up their
 162          * data_ready.. we only want to queue listen work for our listening
 163          * socket
 164          */
 165         if (sk->sk_state == TCP_LISTEN)
 166                 rds_tcp_accept_work(sk);
 167
 168 out:
 169         read_unlock(&sk->sk_callback_lock);
 170         ready(sk);
 171 }
 172
 173 struct socket *rds_tcp_listen_init(struct net *net)
 174 {
 175         struct sockaddr_in sin;
 176         struct socket *sock = NULL;
 177         int ret;
 178
 179         ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
 180         if (ret < 0)
 181                 goto out;
 182
 183         sock->sk->sk_reuse = SK_CAN_REUSE;
 184         rds_tcp_nonagle(sock);
 185
 186         write_lock_bh(&sock->sk->sk_callback_lock);
 187         sock->sk->sk_user_data = sock->sk->sk_data_ready;
 188         sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
 189         write_unlock_bh(&sock->sk->sk_callback_lock);
 190
 191         sin.sin_family = PF_INET;
 192         sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY);
 193         sin.sin_port = (__force u16)htons(RDS_TCP_PORT);
 194
 195         ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
 196         if (ret < 0)
 197                 goto out;
 198
 199         ret = sock->ops->listen(sock, 64);
 200         if (ret < 0)
 201                 goto out;
 202
 203         return sock;
 204 out:
 205         if (sock)
 206                 sock_release(sock);
 207         return NULL;
 208 }
 209
 210 void rds_tcp_listen_stop(struct socket *sock)
 211 {
 212         struct sock *sk;
 213
 214         if (!sock)
 215                 return;
 216
 217         sk = sock->sk;
 218
 219         /* serialize with and prevent further callbacks */
 220         lock_sock(sk);
 221         write_lock_bh(&sk->sk_callback_lock);
 222         if (sk->sk_user_data) {
 223                 sk->sk_data_ready = sk->sk_user_data;
 224                 sk->sk_user_data = NULL;
 225         }
 226         write_unlock_bh(&sk->sk_callback_lock);
 227         release_sock(sk);
 228
 229         /* wait for accepts to stop and close the socket */
 230         flush_workqueue(rds_wq);
 231         sock_release(sock);
 232 }