drivers/infiniband/sw/rxe/rxe_net.c

   1 /*
   2  * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
   3  * Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
   4  *
   5  * This software is available to you under a choice of one of two
   6  * licenses.  You may choose to be licensed under the terms of the GNU
   7  * General Public License (GPL) Version 2, available from the file
   8  * COPYING in the main directory of this source tree, or the
   9  * OpenIB.org BSD license below:
  10  *
  11  *     Redistribution and use in source and binary forms, with or
  12  *     without modification, are permitted provided that the following
  13  *     conditions are met:
  14  *
  15  *      - Redistributions of source code must retain the above
  16  *        copyright notice, this list of conditions and the following
  17  *        disclaimer.
  18  *
  19  *      - Redistributions in binary form must reproduce the above
  20  *        copyright notice, this list of conditions and the following
  21  *        disclaimer in the documentation and/or other materials
  22  *        provided with the distribution.
  23  *
  24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  25  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  26  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  27  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  28  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  29  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  30  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  31  * SOFTWARE.
  32  */
  33
  34 #include <linux/skbuff.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/if.h>
  38 #include <linux/if_vlan.h>
  39 #include <net/udp_tunnel.h>
  40 #include <net/sch_generic.h>
  41 #include <linux/netfilter.h>
  42 #include <rdma/ib_addr.h>
  43
  44 #include "rxe.h"
  45 #include "rxe_net.h"
  46 #include "rxe_loc.h"
  47
  48 static LIST_HEAD(rxe_dev_list);
  49 static DEFINE_SPINLOCK(dev_list_lock); /* spinlock for device list */
  50
  51 struct rxe_dev *net_to_rxe(struct net_device *ndev)
  52 {
  53         struct rxe_dev *rxe;
  54         struct rxe_dev *found = NULL;
  55
  56         spin_lock_bh(&dev_list_lock);
  57         list_for_each_entry(rxe, &rxe_dev_list, list) {
  58                 if (rxe->ndev == ndev) {
  59                         found = rxe;
  60                         break;
  61                 }
  62         }
  63         spin_unlock_bh(&dev_list_lock);
  64
  65         return found;
  66 }
  67
  68 struct rxe_dev *get_rxe_by_name(const char *name)
  69 {
  70         struct rxe_dev *rxe;
  71         struct rxe_dev *found = NULL;
  72
  73         spin_lock_bh(&dev_list_lock);
  74         list_for_each_entry(rxe, &rxe_dev_list, list) {
  75                 if (!strcmp(name, rxe->ib_dev.name)) {
  76                         found = rxe;
  77                         break;
  78                 }
  79         }
  80         spin_unlock_bh(&dev_list_lock);
  81         return found;
  82 }
  83
  84
  85 static struct rxe_recv_sockets recv_sockets;
  86
  87 struct device *rxe_dma_device(struct rxe_dev *rxe)
  88 {
  89         struct net_device *ndev;
  90
  91         ndev = rxe->ndev;
  92
  93         if (is_vlan_dev(ndev))
  94                 ndev = vlan_dev_real_dev(ndev);
  95
  96         return ndev->dev.parent;
  97 }
  98
  99 int rxe_mcast_add(struct rxe_dev *rxe, union ib_gid *mgid)
 100 {
 101         int err;
 102         unsigned char ll_addr[ETH_ALEN];
 103
 104         ipv6_eth_mc_map((struct in6_addr *)mgid->raw, ll_addr);
 105         err = dev_mc_add(rxe->ndev, ll_addr);
 106
 107         return err;
 108 }
 109
 110 int rxe_mcast_delete(struct rxe_dev *rxe, union ib_gid *mgid)
 111 {
 112         int err;
 113         unsigned char ll_addr[ETH_ALEN];
 114
 115         ipv6_eth_mc_map((struct in6_addr *)mgid->raw, ll_addr);
 116         err = dev_mc_del(rxe->ndev, ll_addr);
 117
 118         return err;
 119 }
 120
 121 static struct dst_entry *rxe_find_route4(struct net_device *ndev,
 122                                   struct in_addr *saddr,
 123                                   struct in_addr *daddr)
 124 {
 125         struct rtable *rt;
 126         struct flowi4 fl = { { 0 } };
 127
 128         memset(&fl, 0, sizeof(fl));
 129         fl.flowi4_oif = ndev->ifindex;
 130         memcpy(&fl.saddr, saddr, sizeof(*saddr));
 131         memcpy(&fl.daddr, daddr, sizeof(*daddr));
 132         fl.flowi4_proto = IPPROTO_UDP;
 133
 134         rt = ip_route_output_key(&init_net, &fl);
 135         if (IS_ERR(rt)) {
 136                 pr_err_ratelimited("no route to %pI4\n", &daddr->s_addr);
 137                 return NULL;
 138         }
 139
 140         return &rt->dst;
 141 }
 142
 143 #if IS_ENABLED(CONFIG_IPV6)
 144 static struct dst_entry *rxe_find_route6(struct net_device *ndev,
 145                                          struct in6_addr *saddr,
 146                                          struct in6_addr *daddr)
 147 {
 148         struct dst_entry *ndst;
 149         struct flowi6 fl6 = { { 0 } };
 150
 151         memset(&fl6, 0, sizeof(fl6));
 152         fl6.flowi6_oif = ndev->ifindex;
 153         memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 154         memcpy(&fl6.daddr, daddr, sizeof(*daddr));
 155         fl6.flowi6_proto = IPPROTO_UDP;
 156
 157         if (unlikely(ipv6_stub->ipv6_dst_lookup(sock_net(recv_sockets.sk6->sk),
 158                                                 recv_sockets.sk6->sk, &ndst, &fl6))) {
 159                 pr_err_ratelimited("no route to %pI6\n", daddr);
 160                 goto put;
 161         }
 162
 163         if (unlikely(ndst->error)) {
 164                 pr_err("no route to %pI6\n", daddr);
 165                 goto put;
 166         }
 167
 168         return ndst;
 169 put:
 170         dst_release(ndst);
 171         return NULL;
 172 }
 173
 174 #else
 175
 176 static struct dst_entry *rxe_find_route6(struct net_device *ndev,
 177                                          struct in6_addr *saddr,
 178                                          struct in6_addr *daddr)
 179 {
 180         return NULL;
 181 }
 182
 183 #endif
 184
 185 static struct dst_entry *rxe_find_route(struct rxe_dev *rxe,
 186                                         struct rxe_qp *qp,
 187                                         struct rxe_av *av)
 188 {
 189         struct dst_entry *dst = NULL;
 190
 191         if (qp_type(qp) == IB_QPT_RC)
 192                 dst = sk_dst_get(qp->sk->sk);
 193
 194         if (!dst || !dst_check(dst, qp->dst_cookie)) {
 195                 if (dst)
 196                         dst_release(dst);
 197
 198                 if (av->network_type == RDMA_NETWORK_IPV4) {
 199                         struct in_addr *saddr;
 200                         struct in_addr *daddr;
 201
 202                         saddr = &av->sgid_addr._sockaddr_in.sin_addr;
 203                         daddr = &av->dgid_addr._sockaddr_in.sin_addr;
 204                         dst = rxe_find_route4(rxe->ndev, saddr, daddr);
 205                 } else if (av->network_type == RDMA_NETWORK_IPV6) {
 206                         struct in6_addr *saddr6;
 207                         struct in6_addr *daddr6;
 208
 209                         saddr6 = &av->sgid_addr._sockaddr_in6.sin6_addr;
 210                         daddr6 = &av->dgid_addr._sockaddr_in6.sin6_addr;
 211                         dst = rxe_find_route6(rxe->ndev, saddr6, daddr6);
 212 #if IS_ENABLED(CONFIG_IPV6)
 213                         if (dst)
 214                                 qp->dst_cookie =
 215                                         rt6_get_cookie((struct rt6_info *)dst);
 216 #endif
 217                 }
 218         }
 219
 220         return dst;
 221 }
 222
 223 static int rxe_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
 224 {
 225         struct udphdr *udph;
 226         struct net_device *ndev = skb->dev;
 227         struct rxe_dev *rxe = net_to_rxe(ndev);
 228         struct rxe_pkt_info *pkt = SKB_TO_PKT(skb);
 229
 230         if (!rxe)
 231                 goto drop;
 232
 233         if (skb_linearize(skb)) {
 234                 pr_err("skb_linearize failed\n");
 235                 goto drop;
 236         }
 237
 238         udph = udp_hdr(skb);
 239         pkt->rxe = rxe;
 240         pkt->port_num = 1;
 241         pkt->hdr = (u8 *)(udph + 1);
 242         pkt->mask = RXE_GRH_MASK;
 243         pkt->paylen = be16_to_cpu(udph->len) - sizeof(*udph);
 244
 245         return rxe_rcv(skb);
 246 drop:
 247         kfree_skb(skb);
 248         return 0;
 249 }
 250
 251 static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port,
 252                                            bool ipv6)
 253 {
 254         int err;
 255         struct socket *sock;
 256         struct udp_port_cfg udp_cfg = { };
 257         struct udp_tunnel_sock_cfg tnl_cfg = { };
 258
 259         if (ipv6) {
 260                 udp_cfg.family = AF_INET6;
 261                 udp_cfg.ipv6_v6only = 1;
 262         } else {
 263                 udp_cfg.family = AF_INET;
 264         }
 265
 266         udp_cfg.local_udp_port = port;
 267
 268         /* Create UDP socket */
 269         err = udp_sock_create(net, &udp_cfg, &sock);
 270         if (err < 0) {
 271                 pr_err("failed to create udp socket. err = %d\n", err);
 272                 return ERR_PTR(err);
 273         }
 274
 275         tnl_cfg.encap_type = 1;
 276         tnl_cfg.encap_rcv = rxe_udp_encap_recv;
 277
 278         /* Setup UDP tunnel */
 279         setup_udp_tunnel_sock(net, sock, &tnl_cfg);
 280
 281         return sock;
 282 }
 283
 284 void rxe_release_udp_tunnel(struct socket *sk)
 285 {
 286         if (sk)
 287                 udp_tunnel_sock_release(sk);
 288 }
 289
 290 static void prepare_udp_hdr(struct sk_buff *skb, __be16 src_port,
 291                             __be16 dst_port)
 292 {
 293         struct udphdr *udph;
 294
 295         __skb_push(skb, sizeof(*udph));
 296         skb_reset_transport_header(skb);
 297         udph = udp_hdr(skb);
 298
 299         udph->dest = dst_port;
 300         udph->source = src_port;
 301         udph->len = htons(skb->len);
 302         udph->check = 0;
 303 }
 304
 305 static void prepare_ipv4_hdr(struct dst_entry *dst, struct sk_buff *skb,
 306                              __be32 saddr, __be32 daddr, __u8 proto,
 307                              __u8 tos, __u8 ttl, __be16 df, bool xnet)
 308 {
 309         struct iphdr *iph;
 310
 311         skb_scrub_packet(skb, xnet);
 312
 313         skb_clear_hash(skb);
 314         skb_dst_set(skb, dst_clone(dst));
 315         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 316
 317         skb_push(skb, sizeof(struct iphdr));
 318         skb_reset_network_header(skb);
 319
 320         iph = ip_hdr(skb);
 321
 322         iph->version    =       IPVERSION;
 323         iph->ihl        =       sizeof(struct iphdr) >> 2;
 324         iph->frag_off   =       df;
 325         iph->protocol   =       proto;
 326         iph->tos        =       tos;
 327         iph->daddr      =       daddr;
 328         iph->saddr      =       saddr;
 329         iph->ttl        =       ttl;
 330         __ip_select_ident(dev_net(dst->dev), iph,
 331                           skb_shinfo(skb)->gso_segs ?: 1);
 332         iph->tot_len = htons(skb->len);
 333         ip_send_check(iph);
 334 }
 335
 336 static void prepare_ipv6_hdr(struct dst_entry *dst, struct sk_buff *skb,
 337                              struct in6_addr *saddr, struct in6_addr *daddr,
 338                              __u8 proto, __u8 prio, __u8 ttl)
 339 {
 340         struct ipv6hdr *ip6h;
 341
 342         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 343         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED
 344                             | IPSKB_REROUTED);
 345         skb_dst_set(skb, dst_clone(dst));
 346
 347         __skb_push(skb, sizeof(*ip6h));
 348         skb_reset_network_header(skb);
 349         ip6h              = ipv6_hdr(skb);
 350         ip6_flow_hdr(ip6h, prio, htonl(0));
 351         ip6h->payload_len = htons(skb->len);
 352         ip6h->nexthdr     = proto;
 353         ip6h->hop_limit   = ttl;
 354         ip6h->daddr       = *daddr;
 355         ip6h->saddr       = *saddr;
 356         ip6h->payload_len = htons(skb->len - sizeof(*ip6h));
 357 }
 358
 359 static int prepare4(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 360                     struct sk_buff *skb, struct rxe_av *av)
 361 {
 362         struct rxe_qp *qp = pkt->qp;
 363         struct dst_entry *dst;
 364         bool xnet = false;
 365         __be16 df = htons(IP_DF);
 366         struct in_addr *saddr = &av->sgid_addr._sockaddr_in.sin_addr;
 367         struct in_addr *daddr = &av->dgid_addr._sockaddr_in.sin_addr;
 368
 369         dst = rxe_find_route(rxe, qp, av);
 370         if (!dst) {
 371                 pr_err("Host not reachable\n");
 372                 return -EHOSTUNREACH;
 373         }
 374
 375         if (!memcmp(saddr, daddr, sizeof(*daddr)))
 376                 pkt->mask |= RXE_LOOPBACK_MASK;
 377
 378         prepare_udp_hdr(skb, htons(RXE_ROCE_V2_SPORT),
 379                         htons(ROCE_V2_UDP_DPORT));
 380
 381         prepare_ipv4_hdr(dst, skb, saddr->s_addr, daddr->s_addr, IPPROTO_UDP,
 382                          av->grh.traffic_class, av->grh.hop_limit, df, xnet);
 383
 384         if (qp_type(qp) == IB_QPT_RC)
 385                 sk_dst_set(qp->sk->sk, dst);
 386         else
 387                 dst_release(dst);
 388
 389         return 0;
 390 }
 391
 392 static int prepare6(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 393                     struct sk_buff *skb, struct rxe_av *av)
 394 {
 395         struct rxe_qp *qp = pkt->qp;
 396         struct dst_entry *dst;
 397         struct in6_addr *saddr = &av->sgid_addr._sockaddr_in6.sin6_addr;
 398         struct in6_addr *daddr = &av->dgid_addr._sockaddr_in6.sin6_addr;
 399
 400         dst = rxe_find_route(rxe, qp, av);
 401         if (!dst) {
 402                 pr_err("Host not reachable\n");
 403                 return -EHOSTUNREACH;
 404         }
 405
 406         if (!memcmp(saddr, daddr, sizeof(*daddr)))
 407                 pkt->mask |= RXE_LOOPBACK_MASK;
 408
 409         prepare_udp_hdr(skb, htons(RXE_ROCE_V2_SPORT),
 410                         htons(ROCE_V2_UDP_DPORT));
 411
 412         prepare_ipv6_hdr(dst, skb, saddr, daddr, IPPROTO_UDP,
 413                          av->grh.traffic_class,
 414                          av->grh.hop_limit);
 415
 416         if (qp_type(qp) == IB_QPT_RC)
 417                 sk_dst_set(qp->sk->sk, dst);
 418         else
 419                 dst_release(dst);
 420
 421         return 0;
 422 }
 423
 424 int rxe_prepare(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 425                 struct sk_buff *skb, u32 *crc)
 426 {
 427         int err = 0;
 428         struct rxe_av *av = rxe_get_av(pkt);
 429
 430         if (av->network_type == RDMA_NETWORK_IPV4)
 431                 err = prepare4(rxe, pkt, skb, av);
 432         else if (av->network_type == RDMA_NETWORK_IPV6)
 433                 err = prepare6(rxe, pkt, skb, av);
 434
 435         *crc = rxe_icrc_hdr(pkt, skb);
 436
 437         return err;
 438 }
 439
 440 static void rxe_skb_tx_dtor(struct sk_buff *skb)
 441 {
 442         struct sock *sk = skb->sk;
 443         struct rxe_qp *qp = sk->sk_user_data;
 444         int skb_out = atomic_dec_return(&qp->skb_out);
 445
 446         if (unlikely(qp->need_req_skb &&
 447                      skb_out < RXE_INFLIGHT_SKBS_PER_QP_LOW))
 448                 rxe_run_task(&qp->req.task, 1);
 449
 450         rxe_drop_ref(qp);
 451 }
 452
 453 int rxe_send(struct rxe_dev *rxe, struct rxe_pkt_info *pkt, struct sk_buff *skb)
 454 {
 455         struct rxe_av *av;
 456         int err;
 457
 458         av = rxe_get_av(pkt);
 459
 460         skb->destructor = rxe_skb_tx_dtor;
 461         skb->sk = pkt->qp->sk->sk;
 462
 463         rxe_add_ref(pkt->qp);
 464         atomic_inc(&pkt->qp->skb_out);
 465
 466         if (av->network_type == RDMA_NETWORK_IPV4) {
 467                 err = ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 468         } else if (av->network_type == RDMA_NETWORK_IPV6) {
 469                 err = ip6_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 470         } else {
 471                 pr_err("Unknown layer 3 protocol: %d\n", av->network_type);
 472                 atomic_dec(&pkt->qp->skb_out);
 473                 rxe_drop_ref(pkt->qp);
 474                 kfree_skb(skb);
 475                 return -EINVAL;
 476         }
 477
 478         if (unlikely(net_xmit_eval(err))) {
 479                 pr_debug("error sending packet: %d\n", err);
 480                 return -EAGAIN;
 481         }
 482
 483         return 0;
 484 }
 485
 486 int rxe_loopback(struct sk_buff *skb)
 487 {
 488         return rxe_rcv(skb);
 489 }
 490
 491 static inline int addr_same(struct rxe_dev *rxe, struct rxe_av *av)
 492 {
 493         return rxe->port.port_guid == av->grh.dgid.global.interface_id;
 494 }
 495
 496 struct sk_buff *rxe_init_packet(struct rxe_dev *rxe, struct rxe_av *av,
 497                                 int paylen, struct rxe_pkt_info *pkt)
 498 {
 499         unsigned int hdr_len;
 500         struct sk_buff *skb;
 501
 502         if (av->network_type == RDMA_NETWORK_IPV4)
 503                 hdr_len = ETH_HLEN + sizeof(struct udphdr) +
 504                         sizeof(struct iphdr);
 505         else
 506                 hdr_len = ETH_HLEN + sizeof(struct udphdr) +
 507                         sizeof(struct ipv6hdr);
 508
 509         skb = alloc_skb(paylen + hdr_len + LL_RESERVED_SPACE(rxe->ndev),
 510                         GFP_ATOMIC);
 511         if (unlikely(!skb))
 512                 return NULL;
 513
 514         skb_reserve(skb, hdr_len + LL_RESERVED_SPACE(rxe->ndev));
 515
 516         skb->dev        = rxe->ndev;
 517         if (av->network_type == RDMA_NETWORK_IPV4)
 518                 skb->protocol = htons(ETH_P_IP);
 519         else
 520                 skb->protocol = htons(ETH_P_IPV6);
 521
 522         pkt->rxe        = rxe;
 523         pkt->port_num   = 1;
 524         pkt->hdr        = skb_put(skb, paylen);
 525         pkt->mask       |= RXE_GRH_MASK;
 526
 527         memset(pkt->hdr, 0, paylen);
 528
 529         return skb;
 530 }
 531
 532 /*
 533  * this is required by rxe_cfg to match rxe devices in
 534  * /sys/class/infiniband up with their underlying ethernet devices
 535  */
 536 const char *rxe_parent_name(struct rxe_dev *rxe, unsigned int port_num)
 537 {
 538         return rxe->ndev->name;
 539 }
 540
 541 enum rdma_link_layer rxe_link_layer(struct rxe_dev *rxe, unsigned int port_num)
 542 {
 543         return IB_LINK_LAYER_ETHERNET;
 544 }
 545
 546 struct rxe_dev *rxe_net_add(struct net_device *ndev)
 547 {
 548         int err;
 549         struct rxe_dev *rxe = NULL;
 550
 551         rxe = (struct rxe_dev *)ib_alloc_device(sizeof(*rxe));
 552         if (!rxe)
 553                 return NULL;
 554
 555         rxe->ndev = ndev;
 556
 557         err = rxe_add(rxe, ndev->mtu);
 558         if (err) {
 559                 ib_dealloc_device(&rxe->ib_dev);
 560                 return NULL;
 561         }
 562
 563         spin_lock_bh(&dev_list_lock);
 564         list_add_tail(&rxe->list, &rxe_dev_list);
 565         spin_unlock_bh(&dev_list_lock);
 566         return rxe;
 567 }
 568
 569 void rxe_remove_all(void)
 570 {
 571         spin_lock_bh(&dev_list_lock);
 572         while (!list_empty(&rxe_dev_list)) {
 573                 struct rxe_dev *rxe =
 574                         list_first_entry(&rxe_dev_list, struct rxe_dev, list);
 575
 576                 list_del(&rxe->list);
 577                 spin_unlock_bh(&dev_list_lock);
 578                 rxe_remove(rxe);
 579                 spin_lock_bh(&dev_list_lock);
 580         }
 581         spin_unlock_bh(&dev_list_lock);
 582 }
 583 EXPORT_SYMBOL(rxe_remove_all);
 584
 585 static void rxe_port_event(struct rxe_dev *rxe,
 586                            enum ib_event_type event)
 587 {
 588         struct ib_event ev;
 589
 590         ev.device = &rxe->ib_dev;
 591         ev.element.port_num = 1;
 592         ev.event = event;
 593
 594         ib_dispatch_event(&ev);
 595 }
 596
 597 /* Caller must hold net_info_lock */
 598 void rxe_port_up(struct rxe_dev *rxe)
 599 {
 600         struct rxe_port *port;
 601
 602         port = &rxe->port;
 603         port->attr.state = IB_PORT_ACTIVE;
 604         port->attr.phys_state = IB_PHYS_STATE_LINK_UP;
 605
 606         rxe_port_event(rxe, IB_EVENT_PORT_ACTIVE);
 607         pr_info("set %s active\n", rxe->ib_dev.name);
 608 }
 609
 610 /* Caller must hold net_info_lock */
 611 void rxe_port_down(struct rxe_dev *rxe)
 612 {
 613         struct rxe_port *port;
 614
 615         port = &rxe->port;
 616         port->attr.state = IB_PORT_DOWN;
 617         port->attr.phys_state = IB_PHYS_STATE_LINK_DOWN;
 618
 619         rxe_port_event(rxe, IB_EVENT_PORT_ERR);
 620         pr_info("set %s down\n", rxe->ib_dev.name);
 621 }
 622
 623 static int rxe_notify(struct notifier_block *not_blk,
 624                       unsigned long event,
 625                       void *arg)
 626 {
 627         struct net_device *ndev = netdev_notifier_info_to_dev(arg);
 628         struct rxe_dev *rxe = net_to_rxe(ndev);
 629
 630         if (!rxe)
 631                 goto out;
 632
 633         switch (event) {
 634         case NETDEV_UNREGISTER:
 635                 list_del(&rxe->list);
 636                 rxe_remove(rxe);
 637                 break;
 638         case NETDEV_UP:
 639                 rxe_port_up(rxe);
 640                 break;
 641         case NETDEV_DOWN:
 642                 rxe_port_down(rxe);
 643                 break;
 644         case NETDEV_CHANGEMTU:
 645                 pr_info("%s changed mtu to %d\n", ndev->name, ndev->mtu);
 646                 rxe_set_mtu(rxe, ndev->mtu);
 647                 break;
 648         case NETDEV_CHANGE:
 649                 if (netif_running(ndev) && netif_carrier_ok(ndev))
 650                         rxe_port_up(rxe);
 651                 else
 652                         rxe_port_down(rxe);
 653                 break;
 654         case NETDEV_REBOOT:
 655         case NETDEV_GOING_DOWN:
 656         case NETDEV_CHANGEADDR:
 657         case NETDEV_CHANGENAME:
 658         case NETDEV_FEAT_CHANGE:
 659         default:
 660                 pr_info("ignoring netdev event = %ld for %s\n",
 661                         event, ndev->name);
 662                 break;
 663         }
 664 out:
 665         return NOTIFY_OK;
 666 }
 667
 668 struct notifier_block rxe_net_notifier = {
 669         .notifier_call = rxe_notify,
 670 };
 671
 672 static int rxe_net_ipv4_init(void)
 673 {
 674         recv_sockets.sk4 = rxe_setup_udp_tunnel(&init_net,
 675                                 htons(ROCE_V2_UDP_DPORT), false);
 676         if (IS_ERR(recv_sockets.sk4)) {
 677                 recv_sockets.sk4 = NULL;
 678                 pr_err("Failed to create IPv4 UDP tunnel\n");
 679                 return -1;
 680         }
 681
 682         return 0;
 683 }
 684
 685 static int rxe_net_ipv6_init(void)
 686 {
 687 #if IS_ENABLED(CONFIG_IPV6)
 688
 689         recv_sockets.sk6 = rxe_setup_udp_tunnel(&init_net,
 690                                                 htons(ROCE_V2_UDP_DPORT), true);
 691         if (IS_ERR(recv_sockets.sk6)) {
 692                 recv_sockets.sk6 = NULL;
 693                 pr_err("Failed to create IPv6 UDP tunnel\n");
 694                 return -1;
 695         }
 696 #endif
 697         return 0;
 698 }
 699
 700 void rxe_net_exit(void)
 701 {
 702         rxe_release_udp_tunnel(recv_sockets.sk6);
 703         rxe_release_udp_tunnel(recv_sockets.sk4);
 704         unregister_netdevice_notifier(&rxe_net_notifier);
 705 }
 706
 707 int rxe_net_init(void)
 708 {
 709         int err;
 710
 711         recv_sockets.sk6 = NULL;
 712
 713         err = rxe_net_ipv4_init();
 714         if (err)
 715                 return err;
 716         err = rxe_net_ipv6_init();
 717         if (err)
 718                 goto err_out;
 719         err = register_netdevice_notifier(&rxe_net_notifier);
 720         if (err) {
 721                 pr_err("Failed to register netdev notifier\n");
 722                 goto err_out;
 723         }
 724         return 0;
 725 err_out:
 726         rxe_net_exit();
 727         return err;
 728 }