net/rds/iw.c

   1 /*
   2  * Copyright (c) 2006 Oracle.  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/kernel.h>
  34 #include <linux/in.h>
  35 #include <linux/if.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/inetdevice.h>
  38 #include <linux/if_arp.h>
  39 #include <linux/delay.h>
  40
  41 #include "rds.h"
  42 #include "iw.h"
  43
  44 unsigned int fastreg_pool_size = RDS_FASTREG_POOL_SIZE;
  45 unsigned int fastreg_message_size = RDS_FASTREG_SIZE + 1; /* +1 allows for unaligned MRs */
  46
  47 module_param(fastreg_pool_size, int, 0444);
  48 MODULE_PARM_DESC(fastreg_pool_size, " Max number of fastreg MRs per device");
  49 module_param(fastreg_message_size, int, 0444);
  50 MODULE_PARM_DESC(fastreg_message_size, " Max size of a RDMA transfer (fastreg MRs)");
  51
  52 struct list_head rds_iw_devices;
  53
  54 /* NOTE: if also grabbing iwdev lock, grab this first */
  55 DEFINE_SPINLOCK(iw_nodev_conns_lock);
  56 LIST_HEAD(iw_nodev_conns);
  57
  58 void rds_iw_add_one(struct ib_device *device)
  59 {
  60         struct rds_iw_device *rds_iwdev;
  61         struct ib_device_attr *dev_attr;
  62
  63         /* Only handle iwarp devices */
  64         if (device->node_type != RDMA_NODE_RNIC)
  65                 return;
  66
  67         dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
  68         if (!dev_attr)
  69                 return;
  70
  71         if (ib_query_device(device, dev_attr)) {
  72                 rdsdebug("Query device failed for %s\n", device->name);
  73                 goto free_attr;
  74         }
  75
  76         rds_iwdev = kmalloc(sizeof *rds_iwdev, GFP_KERNEL);
  77         if (!rds_iwdev)
  78                 goto free_attr;
  79
  80         spin_lock_init(&rds_iwdev->spinlock);
  81
  82         rds_iwdev->dma_local_lkey = !!(dev_attr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY);
  83         rds_iwdev->max_wrs = dev_attr->max_qp_wr;
  84         rds_iwdev->max_sge = min(dev_attr->max_sge, RDS_IW_MAX_SGE);
  85
  86         rds_iwdev->dev = device;
  87         rds_iwdev->pd = ib_alloc_pd(device);
  88         if (IS_ERR(rds_iwdev->pd))
  89                 goto free_dev;
  90
  91         if (!rds_iwdev->dma_local_lkey) {
  92                 rds_iwdev->mr = ib_get_dma_mr(rds_iwdev->pd,
  93                                         IB_ACCESS_REMOTE_READ |
  94                                         IB_ACCESS_REMOTE_WRITE |
  95                                         IB_ACCESS_LOCAL_WRITE);
  96                 if (IS_ERR(rds_iwdev->mr))
  97                         goto err_pd;
  98         } else
  99                 rds_iwdev->mr = NULL;
 100
 101         rds_iwdev->mr_pool = rds_iw_create_mr_pool(rds_iwdev);
 102         if (IS_ERR(rds_iwdev->mr_pool)) {
 103                 rds_iwdev->mr_pool = NULL;
 104                 goto err_mr;
 105         }
 106
 107         INIT_LIST_HEAD(&rds_iwdev->cm_id_list);
 108         INIT_LIST_HEAD(&rds_iwdev->conn_list);
 109         list_add_tail(&rds_iwdev->list, &rds_iw_devices);
 110
 111         ib_set_client_data(device, &rds_iw_client, rds_iwdev);
 112
 113         goto free_attr;
 114
 115 err_mr:
 116         if (rds_iwdev->mr)
 117                 ib_dereg_mr(rds_iwdev->mr);
 118 err_pd:
 119         ib_dealloc_pd(rds_iwdev->pd);
 120 free_dev:
 121         kfree(rds_iwdev);
 122 free_attr:
 123         kfree(dev_attr);
 124 }
 125
 126 void rds_iw_remove_one(struct ib_device *device)
 127 {
 128         struct rds_iw_device *rds_iwdev;
 129         struct rds_iw_cm_id *i_cm_id, *next;
 130
 131         rds_iwdev = ib_get_client_data(device, &rds_iw_client);
 132         if (!rds_iwdev)
 133                 return;
 134
 135         spin_lock_irq(&rds_iwdev->spinlock);
 136         list_for_each_entry_safe(i_cm_id, next, &rds_iwdev->cm_id_list, list) {
 137                 list_del(&i_cm_id->list);
 138                 kfree(i_cm_id);
 139         }
 140         spin_unlock_irq(&rds_iwdev->spinlock);
 141
 142         rds_iw_destroy_conns(rds_iwdev);
 143
 144         if (rds_iwdev->mr_pool)
 145                 rds_iw_destroy_mr_pool(rds_iwdev->mr_pool);
 146
 147         if (rds_iwdev->mr)
 148                 ib_dereg_mr(rds_iwdev->mr);
 149
 150         while (ib_dealloc_pd(rds_iwdev->pd)) {
 151                 rdsdebug("Failed to dealloc pd %p\n", rds_iwdev->pd);
 152                 msleep(1);
 153         }
 154
 155         list_del(&rds_iwdev->list);
 156         kfree(rds_iwdev);
 157 }
 158
 159 struct ib_client rds_iw_client = {
 160         .name   = "rds_iw",
 161         .add    = rds_iw_add_one,
 162         .remove = rds_iw_remove_one
 163 };
 164
 165 static int rds_iw_conn_info_visitor(struct rds_connection *conn,
 166                                     void *buffer)
 167 {
 168         struct rds_info_rdma_connection *iinfo = buffer;
 169         struct rds_iw_connection *ic;
 170
 171         /* We will only ever look at IB transports */
 172         if (conn->c_trans != &rds_iw_transport)
 173                 return 0;
 174
 175         iinfo->src_addr = conn->c_laddr;
 176         iinfo->dst_addr = conn->c_faddr;
 177
 178         memset(&iinfo->src_gid, 0, sizeof(iinfo->src_gid));
 179         memset(&iinfo->dst_gid, 0, sizeof(iinfo->dst_gid));
 180         if (rds_conn_state(conn) == RDS_CONN_UP) {
 181                 struct rds_iw_device *rds_iwdev;
 182                 struct rdma_dev_addr *dev_addr;
 183
 184                 ic = conn->c_transport_data;
 185                 dev_addr = &ic->i_cm_id->route.addr.dev_addr;
 186
 187                 ib_addr_get_sgid(dev_addr, (union ib_gid *) &iinfo->src_gid);
 188                 ib_addr_get_dgid(dev_addr, (union ib_gid *) &iinfo->dst_gid);
 189
 190                 rds_iwdev = ib_get_client_data(ic->i_cm_id->device, &rds_iw_client);
 191                 iinfo->max_send_wr = ic->i_send_ring.w_nr;
 192                 iinfo->max_recv_wr = ic->i_recv_ring.w_nr;
 193                 iinfo->max_send_sge = rds_iwdev->max_sge;
 194                 rds_iw_get_mr_info(rds_iwdev, iinfo);
 195         }
 196         return 1;
 197 }
 198
 199 static void rds_iw_ic_info(struct socket *sock, unsigned int len,
 200                            struct rds_info_iterator *iter,
 201                            struct rds_info_lengths *lens)
 202 {
 203         rds_for_each_conn_info(sock, len, iter, lens,
 204                                 rds_iw_conn_info_visitor,
 205                                 sizeof(struct rds_info_rdma_connection));
 206 }
 207
 208
 209 /*
 210  * Early RDS/IB was built to only bind to an address if there is an IPoIB
 211  * device with that address set.
 212  *
 213  * If it were me, I'd advocate for something more flexible.  Sending and
 214  * receiving should be device-agnostic.  Transports would try and maintain
 215  * connections between peers who have messages queued.  Userspace would be
 216  * allowed to influence which paths have priority.  We could call userspace
 217  * asserting this policy "routing".
 218  */
 219 static int rds_iw_laddr_check(__be32 addr)
 220 {
 221         int ret;
 222         struct rdma_cm_id *cm_id;
 223         struct sockaddr_in sin;
 224
 225         /* Create a CMA ID and try to bind it. This catches both
 226          * IB and iWARP capable NICs.
 227          */
 228         cm_id = rdma_create_id(NULL, NULL, RDMA_PS_TCP);
 229         if (IS_ERR(cm_id))
 230                 return PTR_ERR(cm_id);
 231
 232         memset(&sin, 0, sizeof(sin));
 233         sin.sin_family = AF_INET;
 234         sin.sin_addr.s_addr = addr;
 235
 236         /* rdma_bind_addr will only succeed for IB & iWARP devices */
 237         ret = rdma_bind_addr(cm_id, (struct sockaddr *)&sin);
 238         /* due to this, we will claim to support IB devices unless we
 239            check node_type. */
 240         if (ret || cm_id->device->node_type != RDMA_NODE_RNIC)
 241                 ret = -EADDRNOTAVAIL;
 242
 243         rdsdebug("addr %pI4 ret %d node type %d\n",
 244                 &addr, ret,
 245                 cm_id->device ? cm_id->device->node_type : -1);
 246
 247         rdma_destroy_id(cm_id);
 248
 249         return ret;
 250 }
 251
 252 void rds_iw_exit(void)
 253 {
 254         rds_info_deregister_func(RDS_INFO_IWARP_CONNECTIONS, rds_iw_ic_info);
 255         rds_iw_destroy_nodev_conns();
 256         ib_unregister_client(&rds_iw_client);
 257         rds_iw_sysctl_exit();
 258         rds_iw_recv_exit();
 259         rds_trans_unregister(&rds_iw_transport);
 260 }
 261
 262 struct rds_transport rds_iw_transport = {
 263         .laddr_check            = rds_iw_laddr_check,
 264         .xmit_complete          = rds_iw_xmit_complete,
 265         .xmit                   = rds_iw_xmit,
 266         .xmit_cong_map          = NULL,
 267         .xmit_rdma              = rds_iw_xmit_rdma,
 268         .recv                   = rds_iw_recv,
 269         .conn_alloc             = rds_iw_conn_alloc,
 270         .conn_free              = rds_iw_conn_free,
 271         .conn_connect           = rds_iw_conn_connect,
 272         .conn_shutdown          = rds_iw_conn_shutdown,
 273         .inc_copy_to_user       = rds_iw_inc_copy_to_user,
 274         .inc_purge              = rds_iw_inc_purge,
 275         .inc_free               = rds_iw_inc_free,
 276         .cm_initiate_connect    = rds_iw_cm_initiate_connect,
 277         .cm_handle_connect      = rds_iw_cm_handle_connect,
 278         .cm_connect_complete    = rds_iw_cm_connect_complete,
 279         .stats_info_copy        = rds_iw_stats_info_copy,
 280         .exit                   = rds_iw_exit,
 281         .get_mr                 = rds_iw_get_mr,
 282         .sync_mr                = rds_iw_sync_mr,
 283         .free_mr                = rds_iw_free_mr,
 284         .flush_mrs              = rds_iw_flush_mrs,
 285         .t_owner                = THIS_MODULE,
 286         .t_name                 = "iwarp",
 287         .t_type                 = RDS_TRANS_IWARP,
 288         .t_prefer_loopback      = 1,
 289 };
 290
 291 int __init rds_iw_init(void)
 292 {
 293         int ret;
 294
 295         INIT_LIST_HEAD(&rds_iw_devices);
 296
 297         ret = ib_register_client(&rds_iw_client);
 298         if (ret)
 299                 goto out;
 300
 301         ret = rds_iw_sysctl_init();
 302         if (ret)
 303                 goto out_ibreg;
 304
 305         ret = rds_iw_recv_init();
 306         if (ret)
 307                 goto out_sysctl;
 308
 309         ret = rds_trans_register(&rds_iw_transport);
 310         if (ret)
 311                 goto out_recv;
 312
 313         rds_info_register_func(RDS_INFO_IWARP_CONNECTIONS, rds_iw_ic_info);
 314
 315         goto out;
 316
 317 out_recv:
 318         rds_iw_recv_exit();
 319 out_sysctl:
 320         rds_iw_sysctl_exit();
 321 out_ibreg:
 322         ib_unregister_client(&rds_iw_client);
 323 out:
 324         return ret;
 325 }
 326
 327 MODULE_LICENSE("GPL");
 328