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          *
 114          * If the client reboots, this conn will need to be cleaned up.
 115          * rds_tcp_state_change() will do that cleanup
 116          */
 117         rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
 118         rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING);
 119         if (rs_tcp->t_sock) {
 120                 /* Need to resolve a duelling SYN between peers.
 121                  * We have an outstanding SYN to this peer, which may
 122                  * potentially have transitioned to the RDS_CONN_UP state,
 123                  * so we must quiesce any send threads before resetting
 124                  * c_transport_data.
 125                  */
 126                 wait_event(conn->c_waitq,
 127                            !test_bit(RDS_IN_XMIT, &conn->c_flags));
 128                 if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr)) {
 129                         struct sock *nsk = new_sock->sk;
 130
 131                         nsk->sk_user_data = NULL;
 132                         nsk->sk_prot->disconnect(nsk, 0);
 133                         tcp_done(nsk);
 134                         new_sock = NULL;
 135                         ret = 0;
 136                         goto out;
 137                 } else if (rs_tcp->t_sock) {
 138                         rds_tcp_restore_callbacks(rs_tcp->t_sock, rs_tcp);
 139                         conn->c_outgoing = 0;
 140                 }
 141         }
 142         rds_tcp_set_callbacks(new_sock, conn);
 143         rds_connect_complete(conn); /* marks RDS_CONN_UP */
 144         new_sock = NULL;
 145         ret = 0;
 146
 147 out:
 148         if (new_sock)
 149                 sock_release(new_sock);
 150         return ret;
 151 }
 152
 153 void rds_tcp_listen_data_ready(struct sock *sk)
 154 {
 155         void (*ready)(struct sock *sk);
 156
 157         rdsdebug("listen data ready sk %p\n", sk);
 158
 159         read_lock(&sk->sk_callback_lock);
 160         ready = sk->sk_user_data;
 161         if (!ready) { /* check for teardown race */
 162                 ready = sk->sk_data_ready;
 163                 goto out;
 164         }
 165
 166         /*
 167          * ->sk_data_ready is also called for a newly established child socket
 168          * before it has been accepted and the accepter has set up their
 169          * data_ready.. we only want to queue listen work for our listening
 170          * socket
 171          */
 172         if (sk->sk_state == TCP_LISTEN)
 173                 rds_tcp_accept_work(sk);
 174
 175 out:
 176         read_unlock(&sk->sk_callback_lock);
 177         ready(sk);
 178 }
 179
 180 struct socket *rds_tcp_listen_init(struct net *net)
 181 {
 182         struct sockaddr_in sin;
 183         struct socket *sock = NULL;
 184         int ret;
 185
 186         ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
 187         if (ret < 0)
 188                 goto out;
 189
 190         sock->sk->sk_reuse = SK_CAN_REUSE;
 191         rds_tcp_nonagle(sock);
 192
 193         write_lock_bh(&sock->sk->sk_callback_lock);
 194         sock->sk->sk_user_data = sock->sk->sk_data_ready;
 195         sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
 196         write_unlock_bh(&sock->sk->sk_callback_lock);
 197
 198         sin.sin_family = PF_INET;
 199         sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY);
 200         sin.sin_port = (__force u16)htons(RDS_TCP_PORT);
 201
 202         ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
 203         if (ret < 0)
 204                 goto out;
 205
 206         ret = sock->ops->listen(sock, 64);
 207         if (ret < 0)
 208                 goto out;
 209
 210         return sock;
 211 out:
 212         if (sock)
 213                 sock_release(sock);
 214         return NULL;
 215 }
 216
 217 void rds_tcp_listen_stop(struct socket *sock)
 218 {
 219         struct sock *sk;
 220
 221         if (!sock)
 222                 return;
 223
 224         sk = sock->sk;
 225
 226         /* serialize with and prevent further callbacks */
 227         lock_sock(sk);
 228         write_lock_bh(&sk->sk_callback_lock);
 229         if (sk->sk_user_data) {
 230                 sk->sk_data_ready = sk->sk_user_data;
 231                 sk->sk_user_data = NULL;
 232         }
 233         write_unlock_bh(&sk->sk_callback_lock);
 234         release_sock(sk);
 235
 236         /* wait for accepts to stop and close the socket */
 237         flush_workqueue(rds_wq);
 238         sock_release(sock);
 239 }