1 /* 2 * Copyright (c) 2009, 2018 Oracle and/or its affiliates. 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/module.h> 34 #include <rdma/rdma_cm.h> 35 36 #include "rds_single_path.h" 37 #include "rdma_transport.h" 38 #include "ib.h" 39 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 45 46 /* Per IB specification 7.7.3, service level is a 4-bit field. */ 47 #define TOS_TO_SL(tos) ((tos) & 0xF) 48 49 static int rds_rdma_cm_event_handler_cmn(struct rdma_cm_id *cm_id, 50 struct rdma_cm_event *event, 51 bool isv6) 52 { 53 /* this can be null in the listening path */ 54 struct rds_connection *conn = cm_id->context; 55 struct rds_transport *trans = &rds_ib_transport; 56 int ret = 0; 57 int *err; 58 u8 len; 59 60 rdsdebug("conn %p id %p handling event %u (%s)\n", conn, cm_id, 61 event->event, rdma_event_msg(event->event)); 62 63 /* Prevent shutdown from tearing down the connection 64 * while we're executing. */ 65 if (conn) { 66 mutex_lock(&conn->c_cm_lock); 67 68 /* If the connection is being shut down, bail out 69 * right away. We return 0 so cm_id doesn't get 70 * destroyed prematurely */ 71 if (rds_conn_state(conn) == RDS_CONN_DISCONNECTING) { 72 /* Reject incoming connections while we're tearing 73 * down an existing one. */ 74 if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST) 75 ret = 1; 76 goto out; 77 } 78 } 79 80 switch (event->event) { 81 case RDMA_CM_EVENT_CONNECT_REQUEST: 82 /* Only the IB transport is supported, but RDS listens on 83 * every RDMA device: reject a request that arrived on any 84 * other kind. A non-zero return has rdma_cm destroy the 85 * request's id, which is what we want here and only here. 86 */ 87 if (cm_id->device->node_type != RDMA_NODE_IB_CA) { 88 ret = 1; 89 break; 90 } 91 ret = trans->cm_handle_connect(cm_id, event, isv6); 92 break; 93 94 case RDMA_CM_EVENT_ADDR_RESOLVED: 95 if (conn) { 96 /* The address resolved to a device RDS has no 97 * transport for. Do not go on to resolve a route: 98 * an iWARP route leaves cm_id->route.path_rec 99 * unset, which the ROUTE_RESOLVED case below 100 * dereferences. Drop the connection instead; the 101 * id is still ic->i_cm_id, so return 0 and let the 102 * shutdown destroy it. 103 */ 104 if (cm_id->device->node_type != RDMA_NODE_IB_CA) { 105 rds_conn_drop(conn); 106 break; 107 } 108 rdma_set_service_type(cm_id, conn->c_tos); 109 rdma_set_min_rnr_timer(cm_id, IB_RNR_TIMER_000_32); 110 ret = rdma_resolve_route(cm_id, 111 RDS_RDMA_RESOLVE_TIMEOUT_MS); 112 if (ret) { 113 /* A non-zero return has the rdma_cm destroy 114 * the id, but it is still ic->i_cm_id, which 115 * the connection's shutdown would then 116 * disconnect and destroy again. Drop the 117 * connection and keep the id for that 118 * shutdown. 119 */ 120 rds_conn_drop(conn); 121 ret = 0; 122 } 123 } 124 break; 125 126 case RDMA_CM_EVENT_ROUTE_RESOLVED: 127 /* Connection could have been dropped so make sure the 128 * cm_id is valid before proceeding 129 */ 130 if (conn) { 131 struct rds_ib_connection *ibic; 132 133 ibic = conn->c_transport_data; 134 if (ibic && ibic->i_cm_id == cm_id) { 135 cm_id->route.path_rec[0].sl = 136 TOS_TO_SL(conn->c_tos); 137 ret = trans->cm_initiate_connect(cm_id, isv6); 138 } else { 139 rds_conn_drop(conn); 140 } 141 } 142 break; 143 144 case RDMA_CM_EVENT_ESTABLISHED: 145 if (conn) 146 trans->cm_connect_complete(conn, event); 147 break; 148 149 case RDMA_CM_EVENT_REJECTED: 150 if (!conn) 151 break; 152 err = (int *)rdma_consumer_reject_data(cm_id, event, &len); 153 if (!err || 154 (err && len >= sizeof(*err) && 155 ((*err) <= RDS_RDMA_REJ_INCOMPAT))) { 156 pr_warn("RDS/RDMA: conn <%pI6c, %pI6c> rejected, dropping connection\n", 157 &conn->c_laddr, &conn->c_faddr); 158 159 if (!conn->c_tos) 160 conn->c_proposed_version = RDS_PROTOCOL_COMPAT_VERSION; 161 162 rds_conn_drop(conn); 163 } 164 rdsdebug("Connection rejected: %s\n", 165 rdma_reject_msg(cm_id, event->status)); 166 break; 167 case RDMA_CM_EVENT_ADDR_ERROR: 168 case RDMA_CM_EVENT_ROUTE_ERROR: 169 case RDMA_CM_EVENT_CONNECT_ERROR: 170 case RDMA_CM_EVENT_UNREACHABLE: 171 case RDMA_CM_EVENT_DEVICE_REMOVAL: 172 case RDMA_CM_EVENT_ADDR_CHANGE: 173 if (conn) 174 rds_conn_drop(conn); 175 break; 176 177 case RDMA_CM_EVENT_DISCONNECTED: 178 if (!conn) 179 break; 180 rdsdebug("DISCONNECT event - dropping connection " 181 "%pI6c->%pI6c\n", &conn->c_laddr, 182 &conn->c_faddr); 183 rds_conn_drop(conn); 184 break; 185 186 case RDMA_CM_EVENT_TIMEWAIT_EXIT: 187 if (conn) { 188 pr_info("RDS: RDMA_CM_EVENT_TIMEWAIT_EXIT event: dropping connection %pI6c->%pI6c\n", 189 &conn->c_laddr, &conn->c_faddr); 190 rds_conn_drop(conn); 191 } 192 break; 193 194 default: 195 /* things like device disconnect? */ 196 printk(KERN_ERR "RDS: unknown event %u (%s)!\n", 197 event->event, rdma_event_msg(event->event)); 198 break; 199 } 200 201 out: 202 if (conn) 203 mutex_unlock(&conn->c_cm_lock); 204 205 rdsdebug("id %p event %u (%s) handling ret %d\n", cm_id, event->event, 206 rdma_event_msg(event->event), ret); 207 208 return ret; 209 } 210 211 int rds_rdma_cm_event_handler(struct rdma_cm_id *cm_id, 212 struct rdma_cm_event *event) 213 { 214 return rds_rdma_cm_event_handler_cmn(cm_id, event, false); 215 } 216 217 #if IS_ENABLED(CONFIG_IPV6) 218 int rds6_rdma_cm_event_handler(struct rdma_cm_id *cm_id, 219 struct rdma_cm_event *event) 220 { 221 return rds_rdma_cm_event_handler_cmn(cm_id, event, true); 222 } 223 #endif 224 225 static int rds_rdma_listen_init_common(rdma_cm_event_handler handler, 226 struct sockaddr *sa, 227 struct rdma_cm_id **ret_cm_id) 228 { 229 struct rdma_cm_id *cm_id; 230 int ret; 231 232 cm_id = rdma_create_id(&init_net, handler, NULL, 233 RDMA_PS_TCP, IB_QPT_RC); 234 if (IS_ERR(cm_id)) { 235 ret = PTR_ERR(cm_id); 236 printk(KERN_ERR "RDS/RDMA: failed to setup listener, " 237 "rdma_create_id() returned %d\n", ret); 238 return ret; 239 } 240 241 /* Only the IB transport is left, so only listen on IB devices */ 242 ret = rdma_restrict_node_type(cm_id, RDMA_NODE_IB_CA); 243 if (ret) { 244 pr_err("RDS/RDMA: failed to setup listener, rdma_restrict_node_type() returned %d\n", 245 ret); 246 goto out; 247 } 248 249 /* 250 * XXX I bet this binds the cm_id to a device. If we want to support 251 * fail-over we'll have to take this into consideration. 252 */ 253 ret = rdma_bind_addr(cm_id, sa); 254 if (ret) { 255 printk(KERN_ERR "RDS/RDMA: failed to setup listener, " 256 "rdma_bind_addr() returned %d\n", ret); 257 goto out; 258 } 259 260 ret = rdma_listen(cm_id, 128); 261 if (ret) { 262 printk(KERN_ERR "RDS/RDMA: failed to setup listener, " 263 "rdma_listen() returned %d\n", ret); 264 goto out; 265 } 266 267 rdsdebug("cm %p listening on port %u\n", cm_id, 268 ntohs(sa->sa_family == AF_INET6 ? 269 ((struct sockaddr_in6 *)sa)->sin6_port : 270 ((struct sockaddr_in *)sa)->sin_port)); 271 272 *ret_cm_id = cm_id; 273 cm_id = NULL; 274 out: 275 if (cm_id) 276 rdma_destroy_id(cm_id); 277 return ret; 278 } 279 280 /* Initialize the RDS RDMA listeners. We create two listeners for 281 * compatibility reason. The one on RDS_PORT is used for IPv4 282 * requests only. The one on RDS_CM_PORT is used for IPv6 requests 283 * only. So only IPv6 enabled RDS module will communicate using this 284 * port. 285 */ 286 static int rds_rdma_listen_init(void) 287 { 288 int ret; 289 #if IS_ENABLED(CONFIG_IPV6) 290 struct sockaddr_in6 sin6; 291 #endif 292 struct sockaddr_in sin; 293 294 sin.sin_family = PF_INET; 295 sin.sin_addr.s_addr = htonl(INADDR_ANY); 296 sin.sin_port = htons(RDS_PORT); 297 ret = rds_rdma_listen_init_common(rds_rdma_cm_event_handler, 298 (struct sockaddr *)&sin, 299 &rds_rdma_listen_id); 300 if (ret != 0) 301 return ret; 302 303 #if IS_ENABLED(CONFIG_IPV6) 304 sin6.sin6_family = PF_INET6; 305 sin6.sin6_addr = in6addr_any; 306 sin6.sin6_port = htons(RDS_CM_PORT); 307 sin6.sin6_scope_id = 0; 308 sin6.sin6_flowinfo = 0; 309 ret = rds_rdma_listen_init_common(rds6_rdma_cm_event_handler, 310 (struct sockaddr *)&sin6, 311 &rds6_rdma_listen_id); 312 /* Keep going even when IPv6 is not enabled in the system. */ 313 if (ret != 0) 314 rdsdebug("Cannot set up IPv6 RDMA listener\n"); 315 #endif 316 return 0; 317 } 318 319 static void rds_rdma_listen_stop(void) 320 { 321 if (rds_rdma_listen_id) { 322 rdsdebug("cm %p\n", rds_rdma_listen_id); 323 rdma_destroy_id(rds_rdma_listen_id); 324 rds_rdma_listen_id = NULL; 325 } 326 #if IS_ENABLED(CONFIG_IPV6) 327 if (rds6_rdma_listen_id) { 328 rdsdebug("cm %p\n", rds6_rdma_listen_id); 329 rdma_destroy_id(rds6_rdma_listen_id); 330 rds6_rdma_listen_id = NULL; 331 } 332 #endif 333 } 334 335 static int __init rds_rdma_init(void) 336 { 337 int ret; 338 339 ret = rds_ib_init(); 340 if (ret) 341 goto out; 342 343 ret = rds_rdma_listen_init(); 344 if (ret) 345 rds_ib_exit(); 346 out: 347 return ret; 348 } 349 module_init(rds_rdma_init); 350 351 static void __exit rds_rdma_exit(void) 352 { 353 /* stop listening first to ensure no new connections are attempted */ 354 rds_rdma_listen_stop(); 355 rds_ib_exit(); 356 } 357 module_exit(rds_rdma_exit); 358 359 MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>"); 360 MODULE_DESCRIPTION("RDS: IB transport"); 361 MODULE_LICENSE("Dual BSD/GPL"); 362