Commit | Line | Data |
---|---|---|
70041088 AG |
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> | |
5a0e3ad6 | 34 | #include <linux/gfp.h> |
70041088 AG |
35 | #include <linux/in.h> |
36 | #include <net/tcp.h> | |
37 | ||
38 | #include "rds.h" | |
39 | #include "tcp.h" | |
40 | ||
467fa153 | 41 | int rds_tcp_keepalive(struct socket *sock) |
f711a6ae SV |
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 | ||
467fa153 | 73 | int rds_tcp_accept_one(struct socket *sock) |
70041088 AG |
74 | { |
75 | struct socket *new_sock = NULL; | |
76 | struct rds_connection *conn; | |
77 | int ret; | |
78 | struct inet_sock *inet; | |
bd7c5f98 SV |
79 | struct rds_tcp_connection *rs_tcp = NULL; |
80 | int conn_state; | |
81 | struct sock *nsk; | |
70041088 | 82 | |
37e14f4f SV |
83 | if (!sock) /* module unload or netns delete in progress */ |
84 | return -ENETUNREACH; | |
85 | ||
d5a8ac28 SV |
86 | ret = sock_create_kern(sock_net(sock->sk), sock->sk->sk_family, |
87 | sock->sk->sk_type, sock->sk->sk_protocol, | |
88 | &new_sock); | |
70041088 AG |
89 | if (ret) |
90 | goto out; | |
91 | ||
92 | new_sock->type = sock->type; | |
93 | new_sock->ops = sock->ops; | |
94 | ret = sock->ops->accept(sock, new_sock, O_NONBLOCK); | |
95 | if (ret < 0) | |
96 | goto out; | |
97 | ||
f711a6ae SV |
98 | ret = rds_tcp_keepalive(new_sock); |
99 | if (ret < 0) | |
100 | goto out; | |
101 | ||
70041088 AG |
102 | rds_tcp_tune(new_sock); |
103 | ||
104 | inet = inet_sk(new_sock->sk); | |
105 | ||
6884b348 JP |
106 | rdsdebug("accepted tcp %pI4:%u -> %pI4:%u\n", |
107 | &inet->inet_saddr, ntohs(inet->inet_sport), | |
108 | &inet->inet_daddr, ntohs(inet->inet_dport)); | |
70041088 | 109 | |
d5a8ac28 SV |
110 | conn = rds_conn_create(sock_net(sock->sk), |
111 | inet->inet_saddr, inet->inet_daddr, | |
c720c7e8 | 112 | &rds_tcp_transport, GFP_KERNEL); |
70041088 AG |
113 | if (IS_ERR(conn)) { |
114 | ret = PTR_ERR(conn); | |
115 | goto out; | |
116 | } | |
f711a6ae | 117 | /* An incoming SYN request came in, and TCP just accepted it. |
f711a6ae SV |
118 | * |
119 | * If the client reboots, this conn will need to be cleaned up. | |
120 | * rds_tcp_state_change() will do that cleanup | |
121 | */ | |
122 | rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data; | |
3b20fc38 | 123 | rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING); |
bd7c5f98 SV |
124 | mutex_lock(&rs_tcp->t_conn_lock); |
125 | conn_state = rds_conn_state(conn); | |
126 | if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_UP) | |
127 | goto rst_nsk; | |
eb192840 SV |
128 | if (rs_tcp->t_sock) { |
129 | /* Need to resolve a duelling SYN between peers. | |
130 | * We have an outstanding SYN to this peer, which may | |
131 | * potentially have transitioned to the RDS_CONN_UP state, | |
132 | * so we must quiesce any send threads before resetting | |
133 | * c_transport_data. | |
134 | */ | |
c948bb5c SV |
135 | if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr) || |
136 | !conn->c_outgoing) { | |
bd7c5f98 | 137 | goto rst_nsk; |
c948bb5c SV |
138 | } else { |
139 | atomic_set(&conn->c_state, RDS_CONN_CONNECTING); | |
140 | wait_event(conn->c_waitq, | |
141 | !test_bit(RDS_IN_XMIT, &conn->c_flags)); | |
eb192840 SV |
142 | rds_tcp_restore_callbacks(rs_tcp->t_sock, rs_tcp); |
143 | conn->c_outgoing = 0; | |
144 | } | |
145 | } | |
70041088 | 146 | rds_tcp_set_callbacks(new_sock, conn); |
eb192840 | 147 | rds_connect_complete(conn); /* marks RDS_CONN_UP */ |
70041088 AG |
148 | new_sock = NULL; |
149 | ret = 0; | |
bd7c5f98 SV |
150 | goto out; |
151 | rst_nsk: | |
152 | /* reset the newly returned accept sock and bail */ | |
153 | nsk = new_sock->sk; | |
154 | rds_tcp_stats_inc(s_tcp_listen_closed_stale); | |
155 | nsk->sk_user_data = NULL; | |
156 | nsk->sk_prot->disconnect(nsk, 0); | |
157 | tcp_done(nsk); | |
158 | new_sock = NULL; | |
159 | ret = 0; | |
70041088 | 160 | out: |
bd7c5f98 SV |
161 | if (rs_tcp) |
162 | mutex_unlock(&rs_tcp->t_conn_lock); | |
70041088 AG |
163 | if (new_sock) |
164 | sock_release(new_sock); | |
165 | return ret; | |
166 | } | |
167 | ||
676d2369 | 168 | void rds_tcp_listen_data_ready(struct sock *sk) |
70041088 | 169 | { |
676d2369 | 170 | void (*ready)(struct sock *sk); |
70041088 AG |
171 | |
172 | rdsdebug("listen data ready sk %p\n", sk); | |
173 | ||
38036629 | 174 | read_lock_bh(&sk->sk_callback_lock); |
70041088 | 175 | ready = sk->sk_user_data; |
8690bfa1 | 176 | if (!ready) { /* check for teardown race */ |
70041088 AG |
177 | ready = sk->sk_data_ready; |
178 | goto out; | |
179 | } | |
180 | ||
181 | /* | |
182 | * ->sk_data_ready is also called for a newly established child socket | |
183 | * before it has been accepted and the accepter has set up their | |
184 | * data_ready.. we only want to queue listen work for our listening | |
185 | * socket | |
186 | */ | |
187 | if (sk->sk_state == TCP_LISTEN) | |
467fa153 | 188 | rds_tcp_accept_work(sk); |
70041088 AG |
189 | |
190 | out: | |
38036629 | 191 | read_unlock_bh(&sk->sk_callback_lock); |
676d2369 | 192 | ready(sk); |
70041088 AG |
193 | } |
194 | ||
467fa153 | 195 | struct socket *rds_tcp_listen_init(struct net *net) |
70041088 AG |
196 | { |
197 | struct sockaddr_in sin; | |
198 | struct socket *sock = NULL; | |
199 | int ret; | |
200 | ||
467fa153 | 201 | ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock); |
70041088 AG |
202 | if (ret < 0) |
203 | goto out; | |
204 | ||
4a17fd52 | 205 | sock->sk->sk_reuse = SK_CAN_REUSE; |
70041088 AG |
206 | rds_tcp_nonagle(sock); |
207 | ||
208 | write_lock_bh(&sock->sk->sk_callback_lock); | |
209 | sock->sk->sk_user_data = sock->sk->sk_data_ready; | |
210 | sock->sk->sk_data_ready = rds_tcp_listen_data_ready; | |
211 | write_unlock_bh(&sock->sk->sk_callback_lock); | |
212 | ||
01728371 | 213 | sin.sin_family = PF_INET; |
70041088 AG |
214 | sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY); |
215 | sin.sin_port = (__force u16)htons(RDS_TCP_PORT); | |
216 | ||
217 | ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); | |
218 | if (ret < 0) | |
219 | goto out; | |
220 | ||
221 | ret = sock->ops->listen(sock, 64); | |
222 | if (ret < 0) | |
223 | goto out; | |
224 | ||
467fa153 | 225 | return sock; |
70041088 AG |
226 | out: |
227 | if (sock) | |
228 | sock_release(sock); | |
467fa153 | 229 | return NULL; |
70041088 AG |
230 | } |
231 | ||
467fa153 | 232 | void rds_tcp_listen_stop(struct socket *sock) |
70041088 | 233 | { |
70041088 AG |
234 | struct sock *sk; |
235 | ||
8690bfa1 | 236 | if (!sock) |
70041088 AG |
237 | return; |
238 | ||
239 | sk = sock->sk; | |
240 | ||
241 | /* serialize with and prevent further callbacks */ | |
242 | lock_sock(sk); | |
243 | write_lock_bh(&sk->sk_callback_lock); | |
244 | if (sk->sk_user_data) { | |
245 | sk->sk_data_ready = sk->sk_user_data; | |
246 | sk->sk_user_data = NULL; | |
247 | } | |
248 | write_unlock_bh(&sk->sk_callback_lock); | |
249 | release_sock(sk); | |
250 | ||
251 | /* wait for accepts to stop and close the socket */ | |
252 | flush_workqueue(rds_wq); | |
253 | sock_release(sock); | |
70041088 | 254 | } |