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 /* 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;
  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         if (cm_id->device->node_type == RDMA_NODE_IB_CA)
  64                 trans = &rds_ib_transport;
  65
  66         /* Prevent shutdown from tearing down the connection
  67          * while we're executing. */
  68         if (conn) {
  69                 mutex_lock(&conn->c_cm_lock);
  70
  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;
  80                 }
  81         }
  82
  83         switch (event->event) {
  84         case RDMA_CM_EVENT_CONNECT_REQUEST:
  85                 ret = trans->cm_handle_connect(cm_id, event, isv6);
  86                 break;
  87
  88         case RDMA_CM_EVENT_ADDR_RESOLVED:
  89                 if (conn) {
  90                         rdma_set_service_type(cm_id, conn->c_tos);
  91                         rdma_set_min_rnr_timer(cm_id, IB_RNR_TIMER_000_32);
  92                         /* XXX do we need to clean up if this fails? */
  93                         ret = rdma_resolve_route(cm_id,
  94                                                  RDS_RDMA_RESOLVE_TIMEOUT_MS);
  95                 }
  96                 break;
  97
  98         case RDMA_CM_EVENT_ROUTE_RESOLVED:
  99                 /* Connection could have been dropped so make sure the
 100                  * cm_id is valid before proceeding
 101                  */
 102                 if (conn) {
 103                         struct rds_ib_connection *ibic;
 104
 105                         ibic = conn->c_transport_data;
 106                         if (ibic && ibic->i_cm_id == cm_id) {
 107                                 cm_id->route.path_rec[0].sl =
 108                                         TOS_TO_SL(conn->c_tos);
 109                                 ret = trans->cm_initiate_connect(cm_id, isv6);
 110                         } else {
 111                                 rds_conn_drop(conn);
 112                         }
 113                 }
 114                 break;
 115
 116         case RDMA_CM_EVENT_ESTABLISHED:
 117                 if (conn)
 118                         trans->cm_connect_complete(conn, event);
 119                 break;
 120
 121         case RDMA_CM_EVENT_REJECTED:
 122                 if (!conn)
 123                         break;
 124                 err = (int *)rdma_consumer_reject_data(cm_id, event, &len);
 125                 if (!err ||
 126                     (err && len >= sizeof(*err) &&
 127                      ((*err) <= RDS_RDMA_REJ_INCOMPAT))) {
 128                         pr_warn("RDS/RDMA: conn <%pI6c, %pI6c> rejected, dropping connection\n",
 129                                 &conn->c_laddr, &conn->c_faddr);
 130
 131                         if (!conn->c_tos)
 132                                 conn->c_proposed_version = RDS_PROTOCOL_COMPAT_VERSION;
 133
 134                         rds_conn_drop(conn);
 135                 }
 136                 rdsdebug("Connection rejected: %s\n",
 137                          rdma_reject_msg(cm_id, event->status));
 138                 break;
 139         case RDMA_CM_EVENT_ADDR_ERROR:
 140         case RDMA_CM_EVENT_ROUTE_ERROR:
 141         case RDMA_CM_EVENT_CONNECT_ERROR:
 142         case RDMA_CM_EVENT_UNREACHABLE:
 143         case RDMA_CM_EVENT_DEVICE_REMOVAL:
 144         case RDMA_CM_EVENT_ADDR_CHANGE:
 145                 if (conn)
 146                         rds_conn_drop(conn);
 147                 break;
 148
 149         case RDMA_CM_EVENT_DISCONNECTED:
 150                 if (!conn)
 151                         break;
 152                 rdsdebug("DISCONNECT event - dropping connection "
 153                          "%pI6c->%pI6c\n", &conn->c_laddr,
 154                          &conn->c_faddr);
 155                 rds_conn_drop(conn);
 156                 break;
 157
 158         case RDMA_CM_EVENT_TIMEWAIT_EXIT:
 159                 if (conn) {
 160                         pr_info("RDS: RDMA_CM_EVENT_TIMEWAIT_EXIT event: dropping connection %pI6c->%pI6c\n",
 161                                 &conn->c_laddr, &conn->c_faddr);
 162                         rds_conn_drop(conn);
 163                 }
 164                 break;
 165
 166         default:
 167                 /* things like device disconnect? */
 168                 printk(KERN_ERR "RDS: unknown event %u (%s)!\n",
 169                        event->event, rdma_event_msg(event->event));
 170                 break;
 171         }
 172
 173 out:
 174         if (conn)
 175                 mutex_unlock(&conn->c_cm_lock);
 176
 177         rdsdebug("id %p event %u (%s) handling ret %d\n", cm_id, event->event,
 178                  rdma_event_msg(event->event), ret);
 179
 180         return ret;
 181 }
 182
 183 int rds_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 184                               struct rdma_cm_event *event)
 185 {
 186         return rds_rdma_cm_event_handler_cmn(cm_id, event, false);
 187 }
 188
 189 #if IS_ENABLED(CONFIG_IPV6)
 190 int rds6_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 191                                struct rdma_cm_event *event)
 192 {
 193         return rds_rdma_cm_event_handler_cmn(cm_id, event, true);
 194 }
 195 #endif
 196
 197 static int rds_rdma_listen_init_common(rdma_cm_event_handler handler,
 198                                        struct sockaddr *sa,
 199                                        struct rdma_cm_id **ret_cm_id)
 200 {
 201         struct rdma_cm_id *cm_id;
 202         int ret;
 203
 204         cm_id = rdma_create_id(&init_net, handler, NULL,
 205                                RDMA_PS_TCP, IB_QPT_RC);
 206         if (IS_ERR(cm_id)) {
 207                 ret = PTR_ERR(cm_id);
 208                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 209                        "rdma_create_id() returned %d\n", ret);
 210                 return ret;
 211         }
 212
 213         /*
 214          * XXX I bet this binds the cm_id to a device.  If we want to support
 215          * fail-over we'll have to take this into consideration.
 216          */
 217         ret = rdma_bind_addr(cm_id, sa);
 218         if (ret) {
 219                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 220                        "rdma_bind_addr() returned %d\n", ret);
 221                 goto out;
 222         }
 223
 224         ret = rdma_listen(cm_id, 128);
 225         if (ret) {
 226                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 227                        "rdma_listen() returned %d\n", ret);
 228                 goto out;
 229         }
 230
 231         rdsdebug("cm %p listening on port %u\n", cm_id, RDS_PORT);
 232
 233         *ret_cm_id = cm_id;
 234         cm_id = NULL;
 235 out:
 236         if (cm_id)
 237                 rdma_destroy_id(cm_id);
 238         return ret;
 239 }
 240
 241 /* Initialize the RDS RDMA listeners.  We create two listeners for
 242  * compatibility reason.  The one on RDS_PORT is used for IPv4
 243  * requests only.  The one on RDS_CM_PORT is used for IPv6 requests
 244  * only.  So only IPv6 enabled RDS module will communicate using this
 245  * port.
 246  */
 247 static int rds_rdma_listen_init(void)
 248 {
 249         int ret;
 250 #if IS_ENABLED(CONFIG_IPV6)
 251         struct sockaddr_in6 sin6;
 252 #endif
 253         struct sockaddr_in sin;
 254
 255         sin.sin_family = PF_INET;
 256         sin.sin_addr.s_addr = htonl(INADDR_ANY);
 257         sin.sin_port = htons(RDS_PORT);
 258         ret = rds_rdma_listen_init_common(rds_rdma_cm_event_handler,
 259                                           (struct sockaddr *)&sin,
 260                                           &rds_rdma_listen_id);
 261         if (ret != 0)
 262                 return ret;
 263
 264 #if IS_ENABLED(CONFIG_IPV6)
 265         sin6.sin6_family = PF_INET6;
 266         sin6.sin6_addr = in6addr_any;
 267         sin6.sin6_port = htons(RDS_CM_PORT);
 268         sin6.sin6_scope_id = 0;
 269         sin6.sin6_flowinfo = 0;
 270         ret = rds_rdma_listen_init_common(rds6_rdma_cm_event_handler,
 271                                           (struct sockaddr *)&sin6,
 272                                           &rds6_rdma_listen_id);
 273         /* Keep going even when IPv6 is not enabled in the system. */
 274         if (ret != 0)
 275                 rdsdebug("Cannot set up IPv6 RDMA listener\n");
 276 #endif
 277         return 0;
 278 }
 279
 280 static void rds_rdma_listen_stop(void)
 281 {
 282         if (rds_rdma_listen_id) {
 283                 rdsdebug("cm %p\n", rds_rdma_listen_id);
 284                 rdma_destroy_id(rds_rdma_listen_id);
 285                 rds_rdma_listen_id = NULL;
 286         }
 287 #if IS_ENABLED(CONFIG_IPV6)
 288         if (rds6_rdma_listen_id) {
 289                 rdsdebug("cm %p\n", rds6_rdma_listen_id);
 290                 rdma_destroy_id(rds6_rdma_listen_id);
 291                 rds6_rdma_listen_id = NULL;
 292         }
 293 #endif
 294 }
 295
 296 static int __init rds_rdma_init(void)
 297 {
 298         int ret;
 299
 300         ret = rds_ib_init();
 301         if (ret)
 302                 goto out;
 303
 304         ret = rds_rdma_listen_init();
 305         if (ret)
 306                 rds_ib_exit();
 307 out:
 308         return ret;
 309 }
 310 module_init(rds_rdma_init);
 311
 312 static void __exit rds_rdma_exit(void)
 313 {
 314         /* stop listening first to ensure no new connections are attempted */
 315         rds_rdma_listen_stop();
 316         rds_ib_exit();
 317 }
 318 module_exit(rds_rdma_exit);
 319
 320 MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
 321 MODULE_DESCRIPTION("RDS: IB transport");
 322 MODULE_LICENSE("Dual BSD/GPL");