net/rds/rdma_transport.c

   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 static int rds_rdma_cm_event_handler_cmn(struct rdma_cm_id *cm_id,
  47                                          struct rdma_cm_event *event,
  48                                          bool isv6)
  49 {
  50         /* this can be null in the listening path */
  51         struct rds_connection *conn = cm_id->context;
  52         struct rds_transport *trans;
  53         int ret = 0;
  54         int *err;
  55         u8 len;
  56
  57         rdsdebug("conn %p id %p handling event %u (%s)\n", conn, cm_id,
  58                  event->event, rdma_event_msg(event->event));
  59
  60         if (cm_id->device->node_type == RDMA_NODE_IB_CA)
  61                 trans = &rds_ib_transport;
  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                 ret = trans->cm_handle_connect(cm_id, event, isv6);
  83                 break;
  84
  85         case RDMA_CM_EVENT_ADDR_RESOLVED:
  86                 rdma_set_service_type(cm_id, conn->c_tos);
  87                 /* XXX do we need to clean up if this fails? */
  88                 ret = rdma_resolve_route(cm_id,
  89                                          RDS_RDMA_RESOLVE_TIMEOUT_MS);
  90                 break;
  91
  92         case RDMA_CM_EVENT_ROUTE_RESOLVED:
  93                 /* Connection could have been dropped so make sure the
  94                  * cm_id is valid before proceeding
  95                  */
  96                 if (conn) {
  97                         struct rds_ib_connection *ibic;
  98
  99                         ibic = conn->c_transport_data;
 100                         if (ibic && ibic->i_cm_id == cm_id)
 101                                 ret = trans->cm_initiate_connect(cm_id, isv6);
 102                         else
 103                                 rds_conn_drop(conn);
 104                 }
 105                 break;
 106
 107         case RDMA_CM_EVENT_ESTABLISHED:
 108                 trans->cm_connect_complete(conn, event);
 109                 break;
 110
 111         case RDMA_CM_EVENT_REJECTED:
 112                 if (!conn)
 113                         break;
 114                 err = (int *)rdma_consumer_reject_data(cm_id, event, &len);
 115                 if (!err || (err && ((*err) == RDS_RDMA_REJ_INCOMPAT))) {
 116                         pr_warn("RDS/RDMA: conn <%pI6c, %pI6c> rejected, dropping connection\n",
 117                                 &conn->c_laddr, &conn->c_faddr);
 118                         conn->c_proposed_version = RDS_PROTOCOL_COMPAT_VERSION;
 119                         conn->c_tos = 0;
 120                         rds_conn_drop(conn);
 121                 }
 122                 rdsdebug("Connection rejected: %s\n",
 123                          rdma_reject_msg(cm_id, event->status));
 124                 break;
 125                 /* FALLTHROUGH */
 126         case RDMA_CM_EVENT_ADDR_ERROR:
 127         case RDMA_CM_EVENT_ROUTE_ERROR:
 128         case RDMA_CM_EVENT_CONNECT_ERROR:
 129         case RDMA_CM_EVENT_UNREACHABLE:
 130         case RDMA_CM_EVENT_DEVICE_REMOVAL:
 131         case RDMA_CM_EVENT_ADDR_CHANGE:
 132                 if (conn)
 133                         rds_conn_drop(conn);
 134                 break;
 135
 136         case RDMA_CM_EVENT_DISCONNECTED:
 137                 rdsdebug("DISCONNECT event - dropping connection "
 138                          "%pI6c->%pI6c\n", &conn->c_laddr,
 139                          &conn->c_faddr);
 140                 rds_conn_drop(conn);
 141                 break;
 142
 143         case RDMA_CM_EVENT_TIMEWAIT_EXIT:
 144                 if (conn) {
 145                         pr_info("RDS: RDMA_CM_EVENT_TIMEWAIT_EXIT event: dropping connection %pI6c->%pI6c\n",
 146                                 &conn->c_laddr, &conn->c_faddr);
 147                         rds_conn_drop(conn);
 148                 }
 149                 break;
 150
 151         default:
 152                 /* things like device disconnect? */
 153                 printk(KERN_ERR "RDS: unknown event %u (%s)!\n",
 154                        event->event, rdma_event_msg(event->event));
 155                 break;
 156         }
 157
 158 out:
 159         if (conn)
 160                 mutex_unlock(&conn->c_cm_lock);
 161
 162         rdsdebug("id %p event %u (%s) handling ret %d\n", cm_id, event->event,
 163                  rdma_event_msg(event->event), ret);
 164
 165         return ret;
 166 }
 167
 168 int rds_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 169                               struct rdma_cm_event *event)
 170 {
 171         return rds_rdma_cm_event_handler_cmn(cm_id, event, false);
 172 }
 173
 174 #if IS_ENABLED(CONFIG_IPV6)
 175 int rds6_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 176                                struct rdma_cm_event *event)
 177 {
 178         return rds_rdma_cm_event_handler_cmn(cm_id, event, true);
 179 }
 180 #endif
 181
 182 static int rds_rdma_listen_init_common(rdma_cm_event_handler handler,
 183                                        struct sockaddr *sa,
 184                                        struct rdma_cm_id **ret_cm_id)
 185 {
 186         struct rdma_cm_id *cm_id;
 187         int ret;
 188
 189         cm_id = rdma_create_id(&init_net, handler, NULL,
 190                                RDMA_PS_TCP, IB_QPT_RC);
 191         if (IS_ERR(cm_id)) {
 192                 ret = PTR_ERR(cm_id);
 193                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 194                        "rdma_create_id() returned %d\n", ret);
 195                 return ret;
 196         }
 197
 198         /*
 199          * XXX I bet this binds the cm_id to a device.  If we want to support
 200          * fail-over we'll have to take this into consideration.
 201          */
 202         ret = rdma_bind_addr(cm_id, sa);
 203         if (ret) {
 204                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 205                        "rdma_bind_addr() returned %d\n", ret);
 206                 goto out;
 207         }
 208
 209         ret = rdma_listen(cm_id, 128);
 210         if (ret) {
 211                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 212                        "rdma_listen() returned %d\n", ret);
 213                 goto out;
 214         }
 215
 216         rdsdebug("cm %p listening on port %u\n", cm_id, RDS_PORT);
 217
 218         *ret_cm_id = cm_id;
 219         cm_id = NULL;
 220 out:
 221         if (cm_id)
 222                 rdma_destroy_id(cm_id);
 223         return ret;
 224 }
 225
 226 /* Initialize the RDS RDMA listeners.  We create two listeners for
 227  * compatibility reason.  The one on RDS_PORT is used for IPv4
 228  * requests only.  The one on RDS_CM_PORT is used for IPv6 requests
 229  * only.  So only IPv6 enabled RDS module will communicate using this
 230  * port.
 231  */
 232 static int rds_rdma_listen_init(void)
 233 {
 234         int ret;
 235 #if IS_ENABLED(CONFIG_IPV6)
 236         struct sockaddr_in6 sin6;
 237 #endif
 238         struct sockaddr_in sin;
 239
 240         sin.sin_family = PF_INET;
 241         sin.sin_addr.s_addr = htonl(INADDR_ANY);
 242         sin.sin_port = htons(RDS_PORT);
 243         ret = rds_rdma_listen_init_common(rds_rdma_cm_event_handler,
 244                                           (struct sockaddr *)&sin,
 245                                           &rds_rdma_listen_id);
 246         if (ret != 0)
 247                 return ret;
 248
 249 #if IS_ENABLED(CONFIG_IPV6)
 250         sin6.sin6_family = PF_INET6;
 251         sin6.sin6_addr = in6addr_any;
 252         sin6.sin6_port = htons(RDS_CM_PORT);
 253         sin6.sin6_scope_id = 0;
 254         sin6.sin6_flowinfo = 0;
 255         ret = rds_rdma_listen_init_common(rds6_rdma_cm_event_handler,
 256                                           (struct sockaddr *)&sin6,
 257                                           &rds6_rdma_listen_id);
 258         /* Keep going even when IPv6 is not enabled in the system. */
 259         if (ret != 0)
 260                 rdsdebug("Cannot set up IPv6 RDMA listener\n");
 261 #endif
 262         return 0;
 263 }
 264
 265 static void rds_rdma_listen_stop(void)
 266 {
 267         if (rds_rdma_listen_id) {
 268                 rdsdebug("cm %p\n", rds_rdma_listen_id);
 269                 rdma_destroy_id(rds_rdma_listen_id);
 270                 rds_rdma_listen_id = NULL;
 271         }
 272 #if IS_ENABLED(CONFIG_IPV6)
 273         if (rds6_rdma_listen_id) {
 274                 rdsdebug("cm %p\n", rds6_rdma_listen_id);
 275                 rdma_destroy_id(rds6_rdma_listen_id);
 276                 rds6_rdma_listen_id = NULL;
 277         }
 278 #endif
 279 }
 280
 281 static int rds_rdma_init(void)
 282 {
 283         int ret;
 284
 285         ret = rds_ib_init();
 286         if (ret)
 287                 goto out;
 288
 289         ret = rds_rdma_listen_init();
 290         if (ret)
 291                 rds_ib_exit();
 292 out:
 293         return ret;
 294 }
 295 module_init(rds_rdma_init);
 296
 297 static void rds_rdma_exit(void)
 298 {
 299         /* stop listening first to ensure no new connections are attempted */
 300         rds_rdma_listen_stop();
 301         rds_ib_exit();
 302 }
 303 module_exit(rds_rdma_exit);
 304
 305 MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
 306 MODULE_DESCRIPTION("RDS: IB transport");
 307 MODULE_LICENSE("Dual BSD/GPL");