net/xfrm/xfrm_interface.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *      XFRM virtual interface
   4  *
   5  *      Copyright (C) 2018 secunet Security Networks AG
   6  *
   7  *      Author:
   8  *      Steffen Klassert <steffen.klassert@secunet.com>
   9  */
  10
  11 #include <linux/module.h>
  12 #include <linux/capability.h>
  13 #include <linux/errno.h>
  14 #include <linux/types.h>
  15 #include <linux/sockios.h>
  16 #include <linux/icmp.h>
  17 #include <linux/if.h>
  18 #include <linux/in.h>
  19 #include <linux/ip.h>
  20 #include <linux/net.h>
  21 #include <linux/in6.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/if_link.h>
  24 #include <linux/if_arp.h>
  25 #include <linux/icmpv6.h>
  26 #include <linux/init.h>
  27 #include <linux/route.h>
  28 #include <linux/rtnetlink.h>
  29 #include <linux/netfilter_ipv6.h>
  30 #include <linux/slab.h>
  31 #include <linux/hash.h>
  32
  33 #include <linux/uaccess.h>
  34 #include <linux/atomic.h>
  35
  36 #include <net/icmp.h>
  37 #include <net/ip.h>
  38 #include <net/ipv6.h>
  39 #include <net/ip6_route.h>
  40 #include <net/addrconf.h>
  41 #include <net/xfrm.h>
  42 #include <net/net_namespace.h>
  43 #include <net/netns/generic.h>
  44 #include <linux/etherdevice.h>
  45
  46 static int xfrmi_dev_init(struct net_device *dev);
  47 static void xfrmi_dev_setup(struct net_device *dev);
  48 static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
  49 static unsigned int xfrmi_net_id __read_mostly;
  50
  51 struct xfrmi_net {
  52         /* lists for storing interfaces in use */
  53         struct xfrm_if __rcu *xfrmi[1];
  54 };
  55
  56 #define for_each_xfrmi_rcu(start, xi) \
  57         for (xi = rcu_dereference(start); xi; xi = rcu_dereference(xi->next))
  58
  59 static struct xfrm_if *xfrmi_lookup(struct net *net, struct xfrm_state *x)
  60 {
  61         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
  62         struct xfrm_if *xi;
  63
  64         for_each_xfrmi_rcu(xfrmn->xfrmi[0], xi) {
  65                 if (x->if_id == xi->p.if_id &&
  66                     (xi->dev->flags & IFF_UP))
  67                         return xi;
  68         }
  69
  70         return NULL;
  71 }
  72
  73 static struct xfrm_if *xfrmi_decode_session(struct sk_buff *skb,
  74                                             unsigned short family)
  75 {
  76         struct xfrmi_net *xfrmn;
  77         struct xfrm_if *xi;
  78         int ifindex = 0;
  79
  80         if (!secpath_exists(skb) || !skb->dev)
  81                 return NULL;
  82
  83         switch (family) {
  84         case AF_INET6:
  85                 ifindex = inet6_sdif(skb);
  86                 break;
  87         case AF_INET:
  88                 ifindex = inet_sdif(skb);
  89                 break;
  90         }
  91         if (!ifindex)
  92                 ifindex = skb->dev->ifindex;
  93
  94         xfrmn = net_generic(xs_net(xfrm_input_state(skb)), xfrmi_net_id);
  95
  96         for_each_xfrmi_rcu(xfrmn->xfrmi[0], xi) {
  97                 if (ifindex == xi->dev->ifindex &&
  98                         (xi->dev->flags & IFF_UP))
  99                                 return xi;
 100         }
 101
 102         return NULL;
 103 }
 104
 105 static void xfrmi_link(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
 106 {
 107         struct xfrm_if __rcu **xip = &xfrmn->xfrmi[0];
 108
 109         rcu_assign_pointer(xi->next , rtnl_dereference(*xip));
 110         rcu_assign_pointer(*xip, xi);
 111 }
 112
 113 static void xfrmi_unlink(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
 114 {
 115         struct xfrm_if __rcu **xip;
 116         struct xfrm_if *iter;
 117
 118         for (xip = &xfrmn->xfrmi[0];
 119              (iter = rtnl_dereference(*xip)) != NULL;
 120              xip = &iter->next) {
 121                 if (xi == iter) {
 122                         rcu_assign_pointer(*xip, xi->next);
 123                         break;
 124                 }
 125         }
 126 }
 127
 128 static void xfrmi_dev_free(struct net_device *dev)
 129 {
 130         struct xfrm_if *xi = netdev_priv(dev);
 131
 132         gro_cells_destroy(&xi->gro_cells);
 133         free_percpu(dev->tstats);
 134 }
 135
 136 static int xfrmi_create(struct net_device *dev)
 137 {
 138         struct xfrm_if *xi = netdev_priv(dev);
 139         struct net *net = dev_net(dev);
 140         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 141         int err;
 142
 143         dev->rtnl_link_ops = &xfrmi_link_ops;
 144         err = register_netdevice(dev);
 145         if (err < 0)
 146                 goto out;
 147
 148         dev_hold(dev);
 149         xfrmi_link(xfrmn, xi);
 150
 151         return 0;
 152
 153 out:
 154         return err;
 155 }
 156
 157 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
 158 {
 159         struct xfrm_if __rcu **xip;
 160         struct xfrm_if *xi;
 161         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 162
 163         for (xip = &xfrmn->xfrmi[0];
 164              (xi = rtnl_dereference(*xip)) != NULL;
 165              xip = &xi->next)
 166                 if (xi->p.if_id == p->if_id)
 167                         return xi;
 168
 169         return NULL;
 170 }
 171
 172 static void xfrmi_dev_uninit(struct net_device *dev)
 173 {
 174         struct xfrm_if *xi = netdev_priv(dev);
 175         struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
 176
 177         xfrmi_unlink(xfrmn, xi);
 178         dev_put(dev);
 179 }
 180
 181 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
 182 {
 183         skb->tstamp = 0;
 184         skb->pkt_type = PACKET_HOST;
 185         skb->skb_iif = 0;
 186         skb->ignore_df = 0;
 187         skb_dst_drop(skb);
 188         nf_reset_ct(skb);
 189         nf_reset_trace(skb);
 190
 191         if (!xnet)
 192                 return;
 193
 194         ipvs_reset(skb);
 195         secpath_reset(skb);
 196         skb_orphan(skb);
 197         skb->mark = 0;
 198 }
 199
 200 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
 201 {
 202         const struct xfrm_mode *inner_mode;
 203         struct pcpu_sw_netstats *tstats;
 204         struct net_device *dev;
 205         struct xfrm_state *x;
 206         struct xfrm_if *xi;
 207         bool xnet;
 208
 209         if (err && !secpath_exists(skb))
 210                 return 0;
 211
 212         x = xfrm_input_state(skb);
 213
 214         xi = xfrmi_lookup(xs_net(x), x);
 215         if (!xi)
 216                 return 1;
 217
 218         dev = xi->dev;
 219         skb->dev = dev;
 220
 221         if (err) {
 222                 dev->stats.rx_errors++;
 223                 dev->stats.rx_dropped++;
 224
 225                 return 0;
 226         }
 227
 228         xnet = !net_eq(xi->net, dev_net(skb->dev));
 229
 230         if (xnet) {
 231                 inner_mode = &x->inner_mode;
 232
 233                 if (x->sel.family == AF_UNSPEC) {
 234                         inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
 235                         if (inner_mode == NULL) {
 236                                 XFRM_INC_STATS(dev_net(skb->dev),
 237                                                LINUX_MIB_XFRMINSTATEMODEERROR);
 238                                 return -EINVAL;
 239                         }
 240                 }
 241
 242                 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
 243                                        inner_mode->family))
 244                         return -EPERM;
 245         }
 246
 247         xfrmi_scrub_packet(skb, xnet);
 248
 249         tstats = this_cpu_ptr(dev->tstats);
 250
 251         u64_stats_update_begin(&tstats->syncp);
 252         tstats->rx_packets++;
 253         tstats->rx_bytes += skb->len;
 254         u64_stats_update_end(&tstats->syncp);
 255
 256         return 0;
 257 }
 258
 259 static int
 260 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
 261 {
 262         struct xfrm_if *xi = netdev_priv(dev);
 263         struct net_device_stats *stats = &xi->dev->stats;
 264         struct dst_entry *dst = skb_dst(skb);
 265         unsigned int length = skb->len;
 266         struct net_device *tdev;
 267         struct xfrm_state *x;
 268         int err = -1;
 269         int mtu;
 270
 271         dst_hold(dst);
 272         dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, xi->p.if_id);
 273         if (IS_ERR(dst)) {
 274                 err = PTR_ERR(dst);
 275                 dst = NULL;
 276                 goto tx_err_link_failure;
 277         }
 278
 279         x = dst->xfrm;
 280         if (!x)
 281                 goto tx_err_link_failure;
 282
 283         if (x->if_id != xi->p.if_id)
 284                 goto tx_err_link_failure;
 285
 286         tdev = dst->dev;
 287
 288         if (tdev == dev) {
 289                 stats->collisions++;
 290                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 291                                      dev->name);
 292                 goto tx_err_dst_release;
 293         }
 294
 295         mtu = dst_mtu(dst);
 296         if (!skb->ignore_df && skb->len > mtu) {
 297                 skb_dst_update_pmtu_no_confirm(skb, mtu);
 298
 299                 if (skb->protocol == htons(ETH_P_IPV6)) {
 300                         if (mtu < IPV6_MIN_MTU)
 301                                 mtu = IPV6_MIN_MTU;
 302
 303                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 304                 } else {
 305                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 306                                       htonl(mtu));
 307                 }
 308
 309                 dst_release(dst);
 310                 return -EMSGSIZE;
 311         }
 312
 313         xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
 314         skb_dst_set(skb, dst);
 315         skb->dev = tdev;
 316
 317         err = dst_output(xi->net, skb->sk, skb);
 318         if (net_xmit_eval(err) == 0) {
 319                 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
 320
 321                 u64_stats_update_begin(&tstats->syncp);
 322                 tstats->tx_bytes += length;
 323                 tstats->tx_packets++;
 324                 u64_stats_update_end(&tstats->syncp);
 325         } else {
 326                 stats->tx_errors++;
 327                 stats->tx_aborted_errors++;
 328         }
 329
 330         return 0;
 331 tx_err_link_failure:
 332         stats->tx_carrier_errors++;
 333         dst_link_failure(skb);
 334 tx_err_dst_release:
 335         dst_release(dst);
 336         return err;
 337 }
 338
 339 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
 340 {
 341         struct xfrm_if *xi = netdev_priv(dev);
 342         struct net_device_stats *stats = &xi->dev->stats;
 343         struct dst_entry *dst = skb_dst(skb);
 344         struct flowi fl;
 345         int ret;
 346
 347         memset(&fl, 0, sizeof(fl));
 348
 349         switch (skb->protocol) {
 350         case htons(ETH_P_IPV6):
 351                 xfrm_decode_session(skb, &fl, AF_INET6);
 352                 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
 353                 if (!dst) {
 354                         fl.u.ip6.flowi6_oif = dev->ifindex;
 355                         fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
 356                         dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
 357                         if (dst->error) {
 358                                 dst_release(dst);
 359                                 stats->tx_carrier_errors++;
 360                                 goto tx_err;
 361                         }
 362                         skb_dst_set(skb, dst);
 363                 }
 364                 break;
 365         case htons(ETH_P_IP):
 366                 xfrm_decode_session(skb, &fl, AF_INET);
 367                 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 368                 if (!dst) {
 369                         struct rtable *rt;
 370
 371                         fl.u.ip4.flowi4_oif = dev->ifindex;
 372                         fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
 373                         rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
 374                         if (IS_ERR(rt)) {
 375                                 stats->tx_carrier_errors++;
 376                                 goto tx_err;
 377                         }
 378                         skb_dst_set(skb, &rt->dst);
 379                 }
 380                 break;
 381         default:
 382                 goto tx_err;
 383         }
 384
 385         fl.flowi_oif = xi->p.link;
 386
 387         ret = xfrmi_xmit2(skb, dev, &fl);
 388         if (ret < 0)
 389                 goto tx_err;
 390
 391         return NETDEV_TX_OK;
 392
 393 tx_err:
 394         stats->tx_errors++;
 395         stats->tx_dropped++;
 396         kfree_skb(skb);
 397         return NETDEV_TX_OK;
 398 }
 399
 400 static int xfrmi4_err(struct sk_buff *skb, u32 info)
 401 {
 402         const struct iphdr *iph = (const struct iphdr *)skb->data;
 403         struct net *net = dev_net(skb->dev);
 404         int protocol = iph->protocol;
 405         struct ip_comp_hdr *ipch;
 406         struct ip_esp_hdr *esph;
 407         struct ip_auth_hdr *ah ;
 408         struct xfrm_state *x;
 409         struct xfrm_if *xi;
 410         __be32 spi;
 411
 412         switch (protocol) {
 413         case IPPROTO_ESP:
 414                 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
 415                 spi = esph->spi;
 416                 break;
 417         case IPPROTO_AH:
 418                 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
 419                 spi = ah->spi;
 420                 break;
 421         case IPPROTO_COMP:
 422                 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
 423                 spi = htonl(ntohs(ipch->cpi));
 424                 break;
 425         default:
 426                 return 0;
 427         }
 428
 429         switch (icmp_hdr(skb)->type) {
 430         case ICMP_DEST_UNREACH:
 431                 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
 432                         return 0;
 433         case ICMP_REDIRECT:
 434                 break;
 435         default:
 436                 return 0;
 437         }
 438
 439         x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
 440                               spi, protocol, AF_INET);
 441         if (!x)
 442                 return 0;
 443
 444         xi = xfrmi_lookup(net, x);
 445         if (!xi) {
 446                 xfrm_state_put(x);
 447                 return -1;
 448         }
 449
 450         if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
 451                 ipv4_update_pmtu(skb, net, info, 0, protocol);
 452         else
 453                 ipv4_redirect(skb, net, 0, protocol);
 454         xfrm_state_put(x);
 455
 456         return 0;
 457 }
 458
 459 static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 460                     u8 type, u8 code, int offset, __be32 info)
 461 {
 462         const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
 463         struct net *net = dev_net(skb->dev);
 464         int protocol = iph->nexthdr;
 465         struct ip_comp_hdr *ipch;
 466         struct ip_esp_hdr *esph;
 467         struct ip_auth_hdr *ah;
 468         struct xfrm_state *x;
 469         struct xfrm_if *xi;
 470         __be32 spi;
 471
 472         switch (protocol) {
 473         case IPPROTO_ESP:
 474                 esph = (struct ip_esp_hdr *)(skb->data + offset);
 475                 spi = esph->spi;
 476                 break;
 477         case IPPROTO_AH:
 478                 ah = (struct ip_auth_hdr *)(skb->data + offset);
 479                 spi = ah->spi;
 480                 break;
 481         case IPPROTO_COMP:
 482                 ipch = (struct ip_comp_hdr *)(skb->data + offset);
 483                 spi = htonl(ntohs(ipch->cpi));
 484                 break;
 485         default:
 486                 return 0;
 487         }
 488
 489         if (type != ICMPV6_PKT_TOOBIG &&
 490             type != NDISC_REDIRECT)
 491                 return 0;
 492
 493         x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
 494                               spi, protocol, AF_INET6);
 495         if (!x)
 496                 return 0;
 497
 498         xi = xfrmi_lookup(net, x);
 499         if (!xi) {
 500                 xfrm_state_put(x);
 501                 return -1;
 502         }
 503
 504         if (type == NDISC_REDIRECT)
 505                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 506                              sock_net_uid(net, NULL));
 507         else
 508                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 509         xfrm_state_put(x);
 510
 511         return 0;
 512 }
 513
 514 static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
 515 {
 516         if (xi->p.link != p->link)
 517                 return -EINVAL;
 518
 519         xi->p.if_id = p->if_id;
 520
 521         return 0;
 522 }
 523
 524 static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
 525 {
 526         struct net *net = xi->net;
 527         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 528         int err;
 529
 530         xfrmi_unlink(xfrmn, xi);
 531         synchronize_net();
 532         err = xfrmi_change(xi, p);
 533         xfrmi_link(xfrmn, xi);
 534         netdev_state_change(xi->dev);
 535         return err;
 536 }
 537
 538 static void xfrmi_get_stats64(struct net_device *dev,
 539                                struct rtnl_link_stats64 *s)
 540 {
 541         int cpu;
 542
 543         for_each_possible_cpu(cpu) {
 544                 struct pcpu_sw_netstats *stats;
 545                 struct pcpu_sw_netstats tmp;
 546                 int start;
 547
 548                 stats = per_cpu_ptr(dev->tstats, cpu);
 549                 do {
 550                         start = u64_stats_fetch_begin_irq(&stats->syncp);
 551                         tmp.rx_packets = stats->rx_packets;
 552                         tmp.rx_bytes   = stats->rx_bytes;
 553                         tmp.tx_packets = stats->tx_packets;
 554                         tmp.tx_bytes   = stats->tx_bytes;
 555                 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
 556
 557                 s->rx_packets += tmp.rx_packets;
 558                 s->rx_bytes   += tmp.rx_bytes;
 559                 s->tx_packets += tmp.tx_packets;
 560                 s->tx_bytes   += tmp.tx_bytes;
 561         }
 562
 563         s->rx_dropped = dev->stats.rx_dropped;
 564         s->tx_dropped = dev->stats.tx_dropped;
 565 }
 566
 567 static int xfrmi_get_iflink(const struct net_device *dev)
 568 {
 569         struct xfrm_if *xi = netdev_priv(dev);
 570
 571         return xi->p.link;
 572 }
 573
 574
 575 static const struct net_device_ops xfrmi_netdev_ops = {
 576         .ndo_init       = xfrmi_dev_init,
 577         .ndo_uninit     = xfrmi_dev_uninit,
 578         .ndo_start_xmit = xfrmi_xmit,
 579         .ndo_get_stats64 = xfrmi_get_stats64,
 580         .ndo_get_iflink = xfrmi_get_iflink,
 581 };
 582
 583 static void xfrmi_dev_setup(struct net_device *dev)
 584 {
 585         dev->netdev_ops         = &xfrmi_netdev_ops;
 586         dev->type               = ARPHRD_NONE;
 587         dev->mtu                = ETH_DATA_LEN;
 588         dev->min_mtu            = ETH_MIN_MTU;
 589         dev->max_mtu            = IP_MAX_MTU;
 590         dev->flags              = IFF_NOARP;
 591         dev->needs_free_netdev  = true;
 592         dev->priv_destructor    = xfrmi_dev_free;
 593         netif_keep_dst(dev);
 594
 595         eth_broadcast_addr(dev->broadcast);
 596 }
 597
 598 static int xfrmi_dev_init(struct net_device *dev)
 599 {
 600         struct xfrm_if *xi = netdev_priv(dev);
 601         struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
 602         int err;
 603
 604         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 605         if (!dev->tstats)
 606                 return -ENOMEM;
 607
 608         err = gro_cells_init(&xi->gro_cells, dev);
 609         if (err) {
 610                 free_percpu(dev->tstats);
 611                 return err;
 612         }
 613
 614         dev->features |= NETIF_F_LLTX;
 615
 616         if (phydev) {
 617                 dev->needed_headroom = phydev->needed_headroom;
 618                 dev->needed_tailroom = phydev->needed_tailroom;
 619
 620                 if (is_zero_ether_addr(dev->dev_addr))
 621                         eth_hw_addr_inherit(dev, phydev);
 622                 if (is_zero_ether_addr(dev->broadcast))
 623                         memcpy(dev->broadcast, phydev->broadcast,
 624                                dev->addr_len);
 625         } else {
 626                 eth_hw_addr_random(dev);
 627                 eth_broadcast_addr(dev->broadcast);
 628         }
 629
 630         return 0;
 631 }
 632
 633 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
 634                          struct netlink_ext_ack *extack)
 635 {
 636         return 0;
 637 }
 638
 639 static void xfrmi_netlink_parms(struct nlattr *data[],
 640                                struct xfrm_if_parms *parms)
 641 {
 642         memset(parms, 0, sizeof(*parms));
 643
 644         if (!data)
 645                 return;
 646
 647         if (data[IFLA_XFRM_LINK])
 648                 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
 649
 650         if (data[IFLA_XFRM_IF_ID])
 651                 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
 652 }
 653
 654 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
 655                         struct nlattr *tb[], struct nlattr *data[],
 656                         struct netlink_ext_ack *extack)
 657 {
 658         struct net *net = dev_net(dev);
 659         struct xfrm_if_parms p;
 660         struct xfrm_if *xi;
 661         int err;
 662
 663         xfrmi_netlink_parms(data, &p);
 664         xi = xfrmi_locate(net, &p);
 665         if (xi)
 666                 return -EEXIST;
 667
 668         xi = netdev_priv(dev);
 669         xi->p = p;
 670         xi->net = net;
 671         xi->dev = dev;
 672
 673         err = xfrmi_create(dev);
 674         return err;
 675 }
 676
 677 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
 678 {
 679         unregister_netdevice_queue(dev, head);
 680 }
 681
 682 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
 683                            struct nlattr *data[],
 684                            struct netlink_ext_ack *extack)
 685 {
 686         struct xfrm_if *xi = netdev_priv(dev);
 687         struct net *net = xi->net;
 688         struct xfrm_if_parms p;
 689
 690         xfrmi_netlink_parms(data, &p);
 691         xi = xfrmi_locate(net, &p);
 692         if (!xi) {
 693                 xi = netdev_priv(dev);
 694         } else {
 695                 if (xi->dev != dev)
 696                         return -EEXIST;
 697         }
 698
 699         return xfrmi_update(xi, &p);
 700 }
 701
 702 static size_t xfrmi_get_size(const struct net_device *dev)
 703 {
 704         return
 705                 /* IFLA_XFRM_LINK */
 706                 nla_total_size(4) +
 707                 /* IFLA_XFRM_IF_ID */
 708                 nla_total_size(4) +
 709                 0;
 710 }
 711
 712 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
 713 {
 714         struct xfrm_if *xi = netdev_priv(dev);
 715         struct xfrm_if_parms *parm = &xi->p;
 716
 717         if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
 718             nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
 719                 goto nla_put_failure;
 720         return 0;
 721
 722 nla_put_failure:
 723         return -EMSGSIZE;
 724 }
 725
 726 static struct net *xfrmi_get_link_net(const struct net_device *dev)
 727 {
 728         struct xfrm_if *xi = netdev_priv(dev);
 729
 730         return xi->net;
 731 }
 732
 733 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
 734         [IFLA_XFRM_LINK]        = { .type = NLA_U32 },
 735         [IFLA_XFRM_IF_ID]       = { .type = NLA_U32 },
 736 };
 737
 738 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
 739         .kind           = "xfrm",
 740         .maxtype        = IFLA_XFRM_MAX,
 741         .policy         = xfrmi_policy,
 742         .priv_size      = sizeof(struct xfrm_if),
 743         .setup          = xfrmi_dev_setup,
 744         .validate       = xfrmi_validate,
 745         .newlink        = xfrmi_newlink,
 746         .dellink        = xfrmi_dellink,
 747         .changelink     = xfrmi_changelink,
 748         .get_size       = xfrmi_get_size,
 749         .fill_info      = xfrmi_fill_info,
 750         .get_link_net   = xfrmi_get_link_net,
 751 };
 752
 753 static struct pernet_operations xfrmi_net_ops = {
 754         .id   = &xfrmi_net_id,
 755         .size = sizeof(struct xfrmi_net),
 756 };
 757
 758 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
 759         .handler        =       xfrm6_rcv,
 760         .cb_handler     =       xfrmi_rcv_cb,
 761         .err_handler    =       xfrmi6_err,
 762         .priority       =       10,
 763 };
 764
 765 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
 766         .handler        =       xfrm6_rcv,
 767         .cb_handler     =       xfrmi_rcv_cb,
 768         .err_handler    =       xfrmi6_err,
 769         .priority       =       10,
 770 };
 771
 772 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
 773         .handler        =       xfrm6_rcv,
 774         .cb_handler     =       xfrmi_rcv_cb,
 775         .err_handler    =       xfrmi6_err,
 776         .priority       =       10,
 777 };
 778
 779 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
 780         .handler        =       xfrm4_rcv,
 781         .input_handler  =       xfrm_input,
 782         .cb_handler     =       xfrmi_rcv_cb,
 783         .err_handler    =       xfrmi4_err,
 784         .priority       =       10,
 785 };
 786
 787 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
 788         .handler        =       xfrm4_rcv,
 789         .input_handler  =       xfrm_input,
 790         .cb_handler     =       xfrmi_rcv_cb,
 791         .err_handler    =       xfrmi4_err,
 792         .priority       =       10,
 793 };
 794
 795 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
 796         .handler        =       xfrm4_rcv,
 797         .input_handler  =       xfrm_input,
 798         .cb_handler     =       xfrmi_rcv_cb,
 799         .err_handler    =       xfrmi4_err,
 800         .priority       =       10,
 801 };
 802
 803 static int __init xfrmi4_init(void)
 804 {
 805         int err;
 806
 807         err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
 808         if (err < 0)
 809                 goto xfrm_proto_esp_failed;
 810         err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
 811         if (err < 0)
 812                 goto xfrm_proto_ah_failed;
 813         err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
 814         if (err < 0)
 815                 goto xfrm_proto_comp_failed;
 816
 817         return 0;
 818
 819 xfrm_proto_comp_failed:
 820         xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
 821 xfrm_proto_ah_failed:
 822         xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
 823 xfrm_proto_esp_failed:
 824         return err;
 825 }
 826
 827 static void xfrmi4_fini(void)
 828 {
 829         xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
 830         xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
 831         xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
 832 }
 833
 834 static int __init xfrmi6_init(void)
 835 {
 836         int err;
 837
 838         err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
 839         if (err < 0)
 840                 goto xfrm_proto_esp_failed;
 841         err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
 842         if (err < 0)
 843                 goto xfrm_proto_ah_failed;
 844         err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
 845         if (err < 0)
 846                 goto xfrm_proto_comp_failed;
 847
 848         return 0;
 849
 850 xfrm_proto_comp_failed:
 851         xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
 852 xfrm_proto_ah_failed:
 853         xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
 854 xfrm_proto_esp_failed:
 855         return err;
 856 }
 857
 858 static void xfrmi6_fini(void)
 859 {
 860         xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
 861         xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
 862         xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
 863 }
 864
 865 static const struct xfrm_if_cb xfrm_if_cb = {
 866         .decode_session =       xfrmi_decode_session,
 867 };
 868
 869 static int __init xfrmi_init(void)
 870 {
 871         const char *msg;
 872         int err;
 873
 874         pr_info("IPsec XFRM device driver\n");
 875
 876         msg = "tunnel device";
 877         err = register_pernet_device(&xfrmi_net_ops);
 878         if (err < 0)
 879                 goto pernet_dev_failed;
 880
 881         msg = "xfrm4 protocols";
 882         err = xfrmi4_init();
 883         if (err < 0)
 884                 goto xfrmi4_failed;
 885
 886         msg = "xfrm6 protocols";
 887         err = xfrmi6_init();
 888         if (err < 0)
 889                 goto xfrmi6_failed;
 890
 891
 892         msg = "netlink interface";
 893         err = rtnl_link_register(&xfrmi_link_ops);
 894         if (err < 0)
 895                 goto rtnl_link_failed;
 896
 897         xfrm_if_register_cb(&xfrm_if_cb);
 898
 899         return err;
 900
 901 rtnl_link_failed:
 902         xfrmi6_fini();
 903 xfrmi6_failed:
 904         xfrmi4_fini();
 905 xfrmi4_failed:
 906         unregister_pernet_device(&xfrmi_net_ops);
 907 pernet_dev_failed:
 908         pr_err("xfrmi init: failed to register %s\n", msg);
 909         return err;
 910 }
 911
 912 static void __exit xfrmi_fini(void)
 913 {
 914         xfrm_if_unregister_cb();
 915         rtnl_link_unregister(&xfrmi_link_ops);
 916         xfrmi4_fini();
 917         xfrmi6_fini();
 918         unregister_pernet_device(&xfrmi_net_ops);
 919 }
 920
 921 module_init(xfrmi_init);
 922 module_exit(xfrmi_fini);
 923 MODULE_LICENSE("GPL");
 924 MODULE_ALIAS_RTNL_LINK("xfrm");
 925 MODULE_ALIAS_NETDEV("xfrm0");
 926 MODULE_AUTHOR("Steffen Klassert");
 927 MODULE_DESCRIPTION("XFRM virtual interface");