socket: fix snd_delayed_lowbuf, use large buffer chunks for sockets
[cor.git] / net / rds / rdma_transport.c
blob5f741e51b4baa46f825b7a676969cde3159b2c21
1 /*
2 * Copyright (c) 2009, 2018 Oracle and/or its affiliates. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
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.
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.
33 #include <linux/module.h>
34 #include <rdma/rdma_cm.h>
36 #include "rds_single_path.h"
37 #include "rdma_transport.h"
38 #include "ib.h"
40 /* Global IPv4 and IPv6 RDS RDMA listener cm_id */
41 static struct rdma_cm_id *rds_rdma_listen_id;
42 #if IS_ENABLED(CONFIG_IPV6)
43 static struct rdma_cm_id *rds6_rdma_listen_id;
44 #endif
46 /* Per IB specification 7.7.3, service level is a 4-bit field. */
47 #define TOS_TO_SL(tos) ((tos) & 0xF)
49 static int rds_rdma_cm_event_handler_cmn(struct rdma_cm_id *cm_id,
50 struct rdma_cm_event *event,
51 bool isv6)
53 /* this can be null in the listening path */
54 struct rds_connection *conn = cm_id->context;
55 struct rds_transport *trans;
56 int ret = 0;
57 int *err;
58 u8 len;
60 rdsdebug("conn %p id %p handling event %u (%s)\n", conn, cm_id,
61 event->event, rdma_event_msg(event->event));
63 if (cm_id->device->node_type == RDMA_NODE_IB_CA)
64 trans = &rds_ib_transport;
66 /* Prevent shutdown from tearing down the connection
67 * while we're executing. */
68 if (conn) {
69 mutex_lock(&conn->c_cm_lock);
71 /* If the connection is being shut down, bail out
72 * right away. We return 0 so cm_id doesn't get
73 * destroyed prematurely */
74 if (rds_conn_state(conn) == RDS_CONN_DISCONNECTING) {
75 /* Reject incoming connections while we're tearing
76 * down an existing one. */
77 if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST)
78 ret = 1;
79 goto out;
83 switch (event->event) {
84 case RDMA_CM_EVENT_CONNECT_REQUEST:
85 ret = trans->cm_handle_connect(cm_id, event, isv6);
86 break;
88 case RDMA_CM_EVENT_ADDR_RESOLVED:
89 rdma_set_service_type(cm_id, conn->c_tos);
90 /* XXX do we need to clean up if this fails? */
91 ret = rdma_resolve_route(cm_id,
92 RDS_RDMA_RESOLVE_TIMEOUT_MS);
93 break;
95 case RDMA_CM_EVENT_ROUTE_RESOLVED:
96 /* Connection could have been dropped so make sure the
97 * cm_id is valid before proceeding
99 if (conn) {
100 struct rds_ib_connection *ibic;
102 ibic = conn->c_transport_data;
103 if (ibic && ibic->i_cm_id == cm_id) {
104 cm_id->route.path_rec[0].sl =
105 TOS_TO_SL(conn->c_tos);
106 ret = trans->cm_initiate_connect(cm_id, isv6);
107 } else {
108 rds_conn_drop(conn);
111 break;
113 case RDMA_CM_EVENT_ESTABLISHED:
114 if (conn)
115 trans->cm_connect_complete(conn, event);
116 break;
118 case RDMA_CM_EVENT_REJECTED:
119 if (!conn)
120 break;
121 err = (int *)rdma_consumer_reject_data(cm_id, event, &len);
122 if (!err ||
123 (err && len >= sizeof(*err) &&
124 ((*err) <= RDS_RDMA_REJ_INCOMPAT))) {
125 pr_warn("RDS/RDMA: conn <%pI6c, %pI6c> rejected, dropping connection\n",
126 &conn->c_laddr, &conn->c_faddr);
128 if (!conn->c_tos)
129 conn->c_proposed_version = RDS_PROTOCOL_COMPAT_VERSION;
131 rds_conn_drop(conn);
133 rdsdebug("Connection rejected: %s\n",
134 rdma_reject_msg(cm_id, event->status));
135 break;
136 case RDMA_CM_EVENT_ADDR_ERROR:
137 case RDMA_CM_EVENT_ROUTE_ERROR:
138 case RDMA_CM_EVENT_CONNECT_ERROR:
139 case RDMA_CM_EVENT_UNREACHABLE:
140 case RDMA_CM_EVENT_DEVICE_REMOVAL:
141 case RDMA_CM_EVENT_ADDR_CHANGE:
142 if (conn)
143 rds_conn_drop(conn);
144 break;
146 case RDMA_CM_EVENT_DISCONNECTED:
147 if (!conn)
148 break;
149 rdsdebug("DISCONNECT event - dropping connection "
150 "%pI6c->%pI6c\n", &conn->c_laddr,
151 &conn->c_faddr);
152 rds_conn_drop(conn);
153 break;
155 case RDMA_CM_EVENT_TIMEWAIT_EXIT:
156 if (conn) {
157 pr_info("RDS: RDMA_CM_EVENT_TIMEWAIT_EXIT event: dropping connection %pI6c->%pI6c\n",
158 &conn->c_laddr, &conn->c_faddr);
159 rds_conn_drop(conn);
161 break;
163 default:
164 /* things like device disconnect? */
165 printk(KERN_ERR "RDS: unknown event %u (%s)!\n",
166 event->event, rdma_event_msg(event->event));
167 break;
170 out:
171 if (conn)
172 mutex_unlock(&conn->c_cm_lock);
174 rdsdebug("id %p event %u (%s) handling ret %d\n", cm_id, event->event,
175 rdma_event_msg(event->event), ret);
177 return ret;
180 int rds_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
181 struct rdma_cm_event *event)
183 return rds_rdma_cm_event_handler_cmn(cm_id, event, false);
186 #if IS_ENABLED(CONFIG_IPV6)
187 int rds6_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
188 struct rdma_cm_event *event)
190 return rds_rdma_cm_event_handler_cmn(cm_id, event, true);
192 #endif
194 static int rds_rdma_listen_init_common(rdma_cm_event_handler handler,
195 struct sockaddr *sa,
196 struct rdma_cm_id **ret_cm_id)
198 struct rdma_cm_id *cm_id;
199 int ret;
201 cm_id = rdma_create_id(&init_net, handler, NULL,
202 RDMA_PS_TCP, IB_QPT_RC);
203 if (IS_ERR(cm_id)) {
204 ret = PTR_ERR(cm_id);
205 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
206 "rdma_create_id() returned %d\n", ret);
207 return ret;
211 * XXX I bet this binds the cm_id to a device. If we want to support
212 * fail-over we'll have to take this into consideration.
214 ret = rdma_bind_addr(cm_id, sa);
215 if (ret) {
216 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
217 "rdma_bind_addr() returned %d\n", ret);
218 goto out;
221 ret = rdma_listen(cm_id, 128);
222 if (ret) {
223 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
224 "rdma_listen() returned %d\n", ret);
225 goto out;
228 rdsdebug("cm %p listening on port %u\n", cm_id, RDS_PORT);
230 *ret_cm_id = cm_id;
231 cm_id = NULL;
232 out:
233 if (cm_id)
234 rdma_destroy_id(cm_id);
235 return ret;
238 /* Initialize the RDS RDMA listeners. We create two listeners for
239 * compatibility reason. The one on RDS_PORT is used for IPv4
240 * requests only. The one on RDS_CM_PORT is used for IPv6 requests
241 * only. So only IPv6 enabled RDS module will communicate using this
242 * port.
244 static int rds_rdma_listen_init(void)
246 int ret;
247 #if IS_ENABLED(CONFIG_IPV6)
248 struct sockaddr_in6 sin6;
249 #endif
250 struct sockaddr_in sin;
252 sin.sin_family = PF_INET;
253 sin.sin_addr.s_addr = htonl(INADDR_ANY);
254 sin.sin_port = htons(RDS_PORT);
255 ret = rds_rdma_listen_init_common(rds_rdma_cm_event_handler,
256 (struct sockaddr *)&sin,
257 &rds_rdma_listen_id);
258 if (ret != 0)
259 return ret;
261 #if IS_ENABLED(CONFIG_IPV6)
262 sin6.sin6_family = PF_INET6;
263 sin6.sin6_addr = in6addr_any;
264 sin6.sin6_port = htons(RDS_CM_PORT);
265 sin6.sin6_scope_id = 0;
266 sin6.sin6_flowinfo = 0;
267 ret = rds_rdma_listen_init_common(rds6_rdma_cm_event_handler,
268 (struct sockaddr *)&sin6,
269 &rds6_rdma_listen_id);
270 /* Keep going even when IPv6 is not enabled in the system. */
271 if (ret != 0)
272 rdsdebug("Cannot set up IPv6 RDMA listener\n");
273 #endif
274 return 0;
277 static void rds_rdma_listen_stop(void)
279 if (rds_rdma_listen_id) {
280 rdsdebug("cm %p\n", rds_rdma_listen_id);
281 rdma_destroy_id(rds_rdma_listen_id);
282 rds_rdma_listen_id = NULL;
284 #if IS_ENABLED(CONFIG_IPV6)
285 if (rds6_rdma_listen_id) {
286 rdsdebug("cm %p\n", rds6_rdma_listen_id);
287 rdma_destroy_id(rds6_rdma_listen_id);
288 rds6_rdma_listen_id = NULL;
290 #endif
293 static int rds_rdma_init(void)
295 int ret;
297 ret = rds_ib_init();
298 if (ret)
299 goto out;
301 ret = rds_rdma_listen_init();
302 if (ret)
303 rds_ib_exit();
304 out:
305 return ret;
307 module_init(rds_rdma_init);
309 static void rds_rdma_exit(void)
311 /* stop listening first to ensure no new connections are attempted */
312 rds_rdma_listen_stop();
313 rds_ib_exit();
315 module_exit(rds_rdma_exit);
317 MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
318 MODULE_DESCRIPTION("RDS: IB transport");
319 MODULE_LICENSE("Dual BSD/GPL");