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
  55         rdsdebug("conn %p id %p handling event %u (%s)\n", conn, cm_id,
  56                  event->event, rdma_event_msg(event->event));
  57
  58         if (cm_id->device->node_type == RDMA_NODE_IB_CA)
  59                 trans = &rds_ib_transport;
  60
  61         /* Prevent shutdown from tearing down the connection
  62          * while we're executing. */
  63         if (conn) {
  64                 mutex_lock(&conn->c_cm_lock);
  65
  66                 /* If the connection is being shut down, bail out
  67                  * right away. We return 0 so cm_id doesn't get
  68                  * destroyed prematurely */
  69                 if (rds_conn_state(conn) == RDS_CONN_DISCONNECTING) {
  70                         /* Reject incoming connections while we're tearing
  71                          * down an existing one. */
  72                         if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST)
  73                                 ret = 1;
  74                         goto out;
  75                 }
  76         }
  77
  78         switch (event->event) {
  79         case RDMA_CM_EVENT_CONNECT_REQUEST:
  80                 ret = trans->cm_handle_connect(cm_id, event, isv6);
  81                 break;
  82
  83         case RDMA_CM_EVENT_ADDR_RESOLVED:
  84                 /* XXX do we need to clean up if this fails? */
  85                 ret = rdma_resolve_route(cm_id,
  86                                          RDS_RDMA_RESOLVE_TIMEOUT_MS);
  87                 break;
  88
  89         case RDMA_CM_EVENT_ROUTE_RESOLVED:
  90                 /* Connection could have been dropped so make sure the
  91                  * cm_id is valid before proceeding
  92                  */
  93                 if (conn) {
  94                         struct rds_ib_connection *ibic;
  95
  96                         ibic = conn->c_transport_data;
  97                         if (ibic && ibic->i_cm_id == cm_id)
  98                                 ret = trans->cm_initiate_connect(cm_id, isv6);
  99                         else
 100                                 rds_conn_drop(conn);
 101                 }
 102                 break;
 103
 104         case RDMA_CM_EVENT_ESTABLISHED:
 105                 trans->cm_connect_complete(conn, event);
 106                 break;
 107
 108         case RDMA_CM_EVENT_REJECTED:
 109                 rdsdebug("Connection rejected: %s\n",
 110                          rdma_reject_msg(cm_id, event->status));
 111                 /* FALLTHROUGH */
 112         case RDMA_CM_EVENT_ADDR_ERROR:
 113         case RDMA_CM_EVENT_ROUTE_ERROR:
 114         case RDMA_CM_EVENT_CONNECT_ERROR:
 115         case RDMA_CM_EVENT_UNREACHABLE:
 116         case RDMA_CM_EVENT_DEVICE_REMOVAL:
 117         case RDMA_CM_EVENT_ADDR_CHANGE:
 118                 if (conn)
 119                         rds_conn_drop(conn);
 120                 break;
 121
 122         case RDMA_CM_EVENT_DISCONNECTED:
 123                 rdsdebug("DISCONNECT event - dropping connection "
 124                          "%pI6c->%pI6c\n", &conn->c_laddr,
 125                          &conn->c_faddr);
 126                 rds_conn_drop(conn);
 127                 break;
 128
 129         case RDMA_CM_EVENT_TIMEWAIT_EXIT:
 130                 if (conn) {
 131                         pr_info("RDS: RDMA_CM_EVENT_TIMEWAIT_EXIT event: dropping connection %pI6c->%pI6c\n",
 132                                 &conn->c_laddr, &conn->c_faddr);
 133                         rds_conn_drop(conn);
 134                 }
 135                 break;
 136
 137         default:
 138                 /* things like device disconnect? */
 139                 printk(KERN_ERR "RDS: unknown event %u (%s)!\n",
 140                        event->event, rdma_event_msg(event->event));
 141                 break;
 142         }
 143
 144 out:
 145         if (conn)
 146                 mutex_unlock(&conn->c_cm_lock);
 147
 148         rdsdebug("id %p event %u (%s) handling ret %d\n", cm_id, event->event,
 149                  rdma_event_msg(event->event), ret);
 150
 151         return ret;
 152 }
 153
 154 int rds_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 155                               struct rdma_cm_event *event)
 156 {
 157         return rds_rdma_cm_event_handler_cmn(cm_id, event, false);
 158 }
 159
 160 #if IS_ENABLED(CONFIG_IPV6)
 161 int rds6_rdma_cm_event_handler(struct rdma_cm_id *cm_id,
 162                                struct rdma_cm_event *event)
 163 {
 164         return rds_rdma_cm_event_handler_cmn(cm_id, event, true);
 165 }
 166 #endif
 167
 168 static int rds_rdma_listen_init_common(rdma_cm_event_handler handler,
 169                                        struct sockaddr *sa,
 170                                        struct rdma_cm_id **ret_cm_id)
 171 {
 172         struct rdma_cm_id *cm_id;
 173         int ret;
 174
 175         cm_id = rdma_create_id(&init_net, handler, NULL,
 176                                RDMA_PS_TCP, IB_QPT_RC);
 177         if (IS_ERR(cm_id)) {
 178                 ret = PTR_ERR(cm_id);
 179                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 180                        "rdma_create_id() returned %d\n", ret);
 181                 return ret;
 182         }
 183
 184         /*
 185          * XXX I bet this binds the cm_id to a device.  If we want to support
 186          * fail-over we'll have to take this into consideration.
 187          */
 188         ret = rdma_bind_addr(cm_id, sa);
 189         if (ret) {
 190                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 191                        "rdma_bind_addr() returned %d\n", ret);
 192                 goto out;
 193         }
 194
 195         ret = rdma_listen(cm_id, 128);
 196         if (ret) {
 197                 printk(KERN_ERR "RDS/RDMA: failed to setup listener, "
 198                        "rdma_listen() returned %d\n", ret);
 199                 goto out;
 200         }
 201
 202         rdsdebug("cm %p listening on port %u\n", cm_id, RDS_PORT);
 203
 204         *ret_cm_id = cm_id;
 205         cm_id = NULL;
 206 out:
 207         if (cm_id)
 208                 rdma_destroy_id(cm_id);
 209         return ret;
 210 }
 211
 212 /* Initialize the RDS RDMA listeners.  We create two listeners for
 213  * compatibility reason.  The one on RDS_PORT is used for IPv4
 214  * requests only.  The one on RDS_CM_PORT is used for IPv6 requests
 215  * only.  So only IPv6 enabled RDS module will communicate using this
 216  * port.
 217  */
 218 static int rds_rdma_listen_init(void)
 219 {
 220         int ret;
 221 #if IS_ENABLED(CONFIG_IPV6)
 222         struct sockaddr_in6 sin6;
 223 #endif
 224         struct sockaddr_in sin;
 225
 226         sin.sin_family = PF_INET;
 227         sin.sin_addr.s_addr = htonl(INADDR_ANY);
 228         sin.sin_port = htons(RDS_PORT);
 229         ret = rds_rdma_listen_init_common(rds_rdma_cm_event_handler,
 230                                           (struct sockaddr *)&sin,
 231                                           &rds_rdma_listen_id);
 232         if (ret != 0)
 233                 return ret;
 234
 235 #if IS_ENABLED(CONFIG_IPV6)
 236         sin6.sin6_family = PF_INET6;
 237         sin6.sin6_addr = in6addr_any;
 238         sin6.sin6_port = htons(RDS_CM_PORT);
 239         sin6.sin6_scope_id = 0;
 240         sin6.sin6_flowinfo = 0;
 241         ret = rds_rdma_listen_init_common(rds6_rdma_cm_event_handler,
 242                                           (struct sockaddr *)&sin6,
 243                                           &rds6_rdma_listen_id);
 244         /* Keep going even when IPv6 is not enabled in the system. */
 245         if (ret != 0)
 246                 rdsdebug("Cannot set up IPv6 RDMA listener\n");
 247 #endif
 248         return 0;
 249 }
 250
 251 static void rds_rdma_listen_stop(void)
 252 {
 253         if (rds_rdma_listen_id) {
 254                 rdsdebug("cm %p\n", rds_rdma_listen_id);
 255                 rdma_destroy_id(rds_rdma_listen_id);
 256                 rds_rdma_listen_id = NULL;
 257         }
 258 #if IS_ENABLED(CONFIG_IPV6)
 259         if (rds6_rdma_listen_id) {
 260                 rdsdebug("cm %p\n", rds6_rdma_listen_id);
 261                 rdma_destroy_id(rds6_rdma_listen_id);
 262                 rds6_rdma_listen_id = NULL;
 263         }
 264 #endif
 265 }
 266
 267 static int rds_rdma_init(void)
 268 {
 269         int ret;
 270
 271         ret = rds_ib_init();
 272         if (ret)
 273                 goto out;
 274
 275         ret = rds_rdma_listen_init();
 276         if (ret)
 277                 rds_ib_exit();
 278 out:
 279         return ret;
 280 }
 281 module_init(rds_rdma_init);
 282
 283 static void rds_rdma_exit(void)
 284 {
 285         /* stop listening first to ensure no new connections are attempted */
 286         rds_rdma_listen_stop();
 287         rds_ib_exit();
 288 }
 289 module_exit(rds_rdma_exit);
 290
 291 MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
 292 MODULE_DESCRIPTION("RDS: IB transport");
 293 MODULE_LICENSE("Dual BSD/GPL");