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/ip_tunnels.h>
  41 #include <net/addrconf.h>
  42 #include <net/xfrm.h>
  43 #include <net/net_namespace.h>
  44 #include <net/netns/generic.h>
  45 #include <linux/etherdevice.h>
  46
  47 static int xfrmi_dev_init(struct net_device *dev);
  48 static void xfrmi_dev_setup(struct net_device *dev);
  49 static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
  50 static unsigned int xfrmi_net_id __read_mostly;
  51 static const struct net_device_ops xfrmi_netdev_ops;
  52
  53 #define XFRMI_HASH_BITS 8
  54 #define XFRMI_HASH_SIZE BIT(XFRMI_HASH_BITS)
  55
  56 struct xfrmi_net {
  57         /* lists for storing interfaces in use */
  58         struct xfrm_if __rcu *xfrmi[XFRMI_HASH_SIZE];
  59 };
  60
  61 #define for_each_xfrmi_rcu(start, xi) \
  62         for (xi = rcu_dereference(start); xi; xi = rcu_dereference(xi->next))
  63
  64 static u32 xfrmi_hash(u32 if_id)
  65 {
  66         return hash_32(if_id, XFRMI_HASH_BITS);
  67 }
  68
  69 static struct xfrm_if *xfrmi_lookup(struct net *net, struct xfrm_state *x)
  70 {
  71         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
  72         struct xfrm_if *xi;
  73
  74         for_each_xfrmi_rcu(xfrmn->xfrmi[xfrmi_hash(x->if_id)], xi) {
  75                 if (x->if_id == xi->p.if_id &&
  76                     (xi->dev->flags & IFF_UP))
  77                         return xi;
  78         }
  79
  80         return NULL;
  81 }
  82
  83 static struct xfrm_if *xfrmi_decode_session(struct sk_buff *skb,
  84                                             unsigned short family)
  85 {
  86         struct net_device *dev;
  87         int ifindex = 0;
  88
  89         if (!secpath_exists(skb) || !skb->dev)
  90                 return NULL;
  91
  92         switch (family) {
  93         case AF_INET6:
  94                 ifindex = inet6_sdif(skb);
  95                 break;
  96         case AF_INET:
  97                 ifindex = inet_sdif(skb);
  98                 break;
  99         }
 100
 101         if (ifindex) {
 102                 struct net *net = xs_net(xfrm_input_state(skb));
 103
 104                 dev = dev_get_by_index_rcu(net, ifindex);
 105         } else {
 106                 dev = skb->dev;
 107         }
 108
 109         if (!dev || !(dev->flags & IFF_UP))
 110                 return NULL;
 111         if (dev->netdev_ops != &xfrmi_netdev_ops)
 112                 return NULL;
 113
 114         return netdev_priv(dev);
 115 }
 116
 117 static void xfrmi_link(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
 118 {
 119         struct xfrm_if __rcu **xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
 120
 121         rcu_assign_pointer(xi->next , rtnl_dereference(*xip));
 122         rcu_assign_pointer(*xip, xi);
 123 }
 124
 125 static void xfrmi_unlink(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
 126 {
 127         struct xfrm_if __rcu **xip;
 128         struct xfrm_if *iter;
 129
 130         for (xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
 131              (iter = rtnl_dereference(*xip)) != NULL;
 132              xip = &iter->next) {
 133                 if (xi == iter) {
 134                         rcu_assign_pointer(*xip, xi->next);
 135                         break;
 136                 }
 137         }
 138 }
 139
 140 static void xfrmi_dev_free(struct net_device *dev)
 141 {
 142         struct xfrm_if *xi = netdev_priv(dev);
 143
 144         gro_cells_destroy(&xi->gro_cells);
 145         free_percpu(dev->tstats);
 146 }
 147
 148 static int xfrmi_create(struct net_device *dev)
 149 {
 150         struct xfrm_if *xi = netdev_priv(dev);
 151         struct net *net = dev_net(dev);
 152         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 153         int err;
 154
 155         dev->rtnl_link_ops = &xfrmi_link_ops;
 156         err = register_netdevice(dev);
 157         if (err < 0)
 158                 goto out;
 159
 160         xfrmi_link(xfrmn, xi);
 161
 162         return 0;
 163
 164 out:
 165         return err;
 166 }
 167
 168 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
 169 {
 170         struct xfrm_if __rcu **xip;
 171         struct xfrm_if *xi;
 172         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 173
 174         for (xip = &xfrmn->xfrmi[xfrmi_hash(p->if_id)];
 175              (xi = rtnl_dereference(*xip)) != NULL;
 176              xip = &xi->next)
 177                 if (xi->p.if_id == p->if_id)
 178                         return xi;
 179
 180         return NULL;
 181 }
 182
 183 static void xfrmi_dev_uninit(struct net_device *dev)
 184 {
 185         struct xfrm_if *xi = netdev_priv(dev);
 186         struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
 187
 188         xfrmi_unlink(xfrmn, xi);
 189 }
 190
 191 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
 192 {
 193         skb->tstamp = 0;
 194         skb->pkt_type = PACKET_HOST;
 195         skb->skb_iif = 0;
 196         skb->ignore_df = 0;
 197         skb_dst_drop(skb);
 198         nf_reset_ct(skb);
 199         nf_reset_trace(skb);
 200
 201         if (!xnet)
 202                 return;
 203
 204         ipvs_reset(skb);
 205         secpath_reset(skb);
 206         skb_orphan(skb);
 207         skb->mark = 0;
 208 }
 209
 210 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
 211 {
 212         const struct xfrm_mode *inner_mode;
 213         struct net_device *dev;
 214         struct xfrm_state *x;
 215         struct xfrm_if *xi;
 216         bool xnet;
 217
 218         if (err && !secpath_exists(skb))
 219                 return 0;
 220
 221         x = xfrm_input_state(skb);
 222
 223         xi = xfrmi_lookup(xs_net(x), x);
 224         if (!xi)
 225                 return 1;
 226
 227         dev = xi->dev;
 228         skb->dev = dev;
 229
 230         if (err) {
 231                 dev->stats.rx_errors++;
 232                 dev->stats.rx_dropped++;
 233
 234                 return 0;
 235         }
 236
 237         xnet = !net_eq(xi->net, dev_net(skb->dev));
 238
 239         if (xnet) {
 240                 inner_mode = &x->inner_mode;
 241
 242                 if (x->sel.family == AF_UNSPEC) {
 243                         inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
 244                         if (inner_mode == NULL) {
 245                                 XFRM_INC_STATS(dev_net(skb->dev),
 246                                                LINUX_MIB_XFRMINSTATEMODEERROR);
 247                                 return -EINVAL;
 248                         }
 249                 }
 250
 251                 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
 252                                        inner_mode->family))
 253                         return -EPERM;
 254         }
 255
 256         xfrmi_scrub_packet(skb, xnet);
 257         dev_sw_netstats_rx_add(dev, skb->len);
 258
 259         return 0;
 260 }
 261
 262 static int
 263 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
 264 {
 265         struct xfrm_if *xi = netdev_priv(dev);
 266         struct net_device_stats *stats = &xi->dev->stats;
 267         struct dst_entry *dst = skb_dst(skb);
 268         unsigned int length = skb->len;
 269         struct net_device *tdev;
 270         struct xfrm_state *x;
 271         int err = -1;
 272         int mtu;
 273
 274         dst_hold(dst);
 275         dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, xi->p.if_id);
 276         if (IS_ERR(dst)) {
 277                 err = PTR_ERR(dst);
 278                 dst = NULL;
 279                 goto tx_err_link_failure;
 280         }
 281
 282         x = dst->xfrm;
 283         if (!x)
 284                 goto tx_err_link_failure;
 285
 286         if (x->if_id != xi->p.if_id)
 287                 goto tx_err_link_failure;
 288
 289         tdev = dst->dev;
 290
 291         if (tdev == dev) {
 292                 stats->collisions++;
 293                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 294                                      dev->name);
 295                 goto tx_err_dst_release;
 296         }
 297
 298         mtu = dst_mtu(dst);
 299         if ((!skb_is_gso(skb) && skb->len > mtu) ||
 300             (skb_is_gso(skb) && !skb_gso_validate_network_len(skb, mtu))) {
 301                 skb_dst_update_pmtu_no_confirm(skb, mtu);
 302
 303                 if (skb->protocol == htons(ETH_P_IPV6)) {
 304                         if (mtu < IPV6_MIN_MTU)
 305                                 mtu = IPV6_MIN_MTU;
 306
 307                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 308                 } else {
 309                         if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
 310                                 goto xmit;
 311                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 312                                       htonl(mtu));
 313                 }
 314
 315                 dst_release(dst);
 316                 return -EMSGSIZE;
 317         }
 318
 319 xmit:
 320         xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
 321         skb_dst_set(skb, dst);
 322         skb->dev = tdev;
 323
 324         err = dst_output(xi->net, skb->sk, skb);
 325         if (net_xmit_eval(err) == 0) {
 326                 dev_sw_netstats_tx_add(dev, 1, length);
 327         } else {
 328                 stats->tx_errors++;
 329                 stats->tx_aborted_errors++;
 330         }
 331
 332         return 0;
 333 tx_err_link_failure:
 334         stats->tx_carrier_errors++;
 335         dst_link_failure(skb);
 336 tx_err_dst_release:
 337         dst_release(dst);
 338         return err;
 339 }
 340
 341 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
 342 {
 343         struct xfrm_if *xi = netdev_priv(dev);
 344         struct net_device_stats *stats = &xi->dev->stats;
 345         struct dst_entry *dst = skb_dst(skb);
 346         struct flowi fl;
 347         int ret;
 348
 349         memset(&fl, 0, sizeof(fl));
 350
 351         switch (skb->protocol) {
 352         case htons(ETH_P_IPV6):
 353                 xfrm_decode_session(skb, &fl, AF_INET6);
 354                 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
 355                 if (!dst) {
 356                         fl.u.ip6.flowi6_oif = dev->ifindex;
 357                         fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
 358                         dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
 359                         if (dst->error) {
 360                                 dst_release(dst);
 361                                 stats->tx_carrier_errors++;
 362                                 goto tx_err;
 363                         }
 364                         skb_dst_set(skb, dst);
 365                 }
 366                 break;
 367         case htons(ETH_P_IP):
 368                 xfrm_decode_session(skb, &fl, AF_INET);
 369                 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 370                 if (!dst) {
 371                         struct rtable *rt;
 372
 373                         fl.u.ip4.flowi4_oif = dev->ifindex;
 374                         fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
 375                         rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
 376                         if (IS_ERR(rt)) {
 377                                 stats->tx_carrier_errors++;
 378                                 goto tx_err;
 379                         }
 380                         skb_dst_set(skb, &rt->dst);
 381                 }
 382                 break;
 383         default:
 384                 goto tx_err;
 385         }
 386
 387         fl.flowi_oif = xi->p.link;
 388
 389         ret = xfrmi_xmit2(skb, dev, &fl);
 390         if (ret < 0)
 391                 goto tx_err;
 392
 393         return NETDEV_TX_OK;
 394
 395 tx_err:
 396         stats->tx_errors++;
 397         stats->tx_dropped++;
 398         kfree_skb(skb);
 399         return NETDEV_TX_OK;
 400 }
 401
 402 static int xfrmi4_err(struct sk_buff *skb, u32 info)
 403 {
 404         const struct iphdr *iph = (const struct iphdr *)skb->data;
 405         struct net *net = dev_net(skb->dev);
 406         int protocol = iph->protocol;
 407         struct ip_comp_hdr *ipch;
 408         struct ip_esp_hdr *esph;
 409         struct ip_auth_hdr *ah ;
 410         struct xfrm_state *x;
 411         struct xfrm_if *xi;
 412         __be32 spi;
 413
 414         switch (protocol) {
 415         case IPPROTO_ESP:
 416                 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
 417                 spi = esph->spi;
 418                 break;
 419         case IPPROTO_AH:
 420                 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
 421                 spi = ah->spi;
 422                 break;
 423         case IPPROTO_COMP:
 424                 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
 425                 spi = htonl(ntohs(ipch->cpi));
 426                 break;
 427         default:
 428                 return 0;
 429         }
 430
 431         switch (icmp_hdr(skb)->type) {
 432         case ICMP_DEST_UNREACH:
 433                 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
 434                         return 0;
 435                 break;
 436         case ICMP_REDIRECT:
 437                 break;
 438         default:
 439                 return 0;
 440         }
 441
 442         x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
 443                               spi, protocol, AF_INET);
 444         if (!x)
 445                 return 0;
 446
 447         xi = xfrmi_lookup(net, x);
 448         if (!xi) {
 449                 xfrm_state_put(x);
 450                 return -1;
 451         }
 452
 453         if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
 454                 ipv4_update_pmtu(skb, net, info, 0, protocol);
 455         else
 456                 ipv4_redirect(skb, net, 0, protocol);
 457         xfrm_state_put(x);
 458
 459         return 0;
 460 }
 461
 462 static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 463                     u8 type, u8 code, int offset, __be32 info)
 464 {
 465         const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
 466         struct net *net = dev_net(skb->dev);
 467         int protocol = iph->nexthdr;
 468         struct ip_comp_hdr *ipch;
 469         struct ip_esp_hdr *esph;
 470         struct ip_auth_hdr *ah;
 471         struct xfrm_state *x;
 472         struct xfrm_if *xi;
 473         __be32 spi;
 474
 475         switch (protocol) {
 476         case IPPROTO_ESP:
 477                 esph = (struct ip_esp_hdr *)(skb->data + offset);
 478                 spi = esph->spi;
 479                 break;
 480         case IPPROTO_AH:
 481                 ah = (struct ip_auth_hdr *)(skb->data + offset);
 482                 spi = ah->spi;
 483                 break;
 484         case IPPROTO_COMP:
 485                 ipch = (struct ip_comp_hdr *)(skb->data + offset);
 486                 spi = htonl(ntohs(ipch->cpi));
 487                 break;
 488         default:
 489                 return 0;
 490         }
 491
 492         if (type != ICMPV6_PKT_TOOBIG &&
 493             type != NDISC_REDIRECT)
 494                 return 0;
 495
 496         x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
 497                               spi, protocol, AF_INET6);
 498         if (!x)
 499                 return 0;
 500
 501         xi = xfrmi_lookup(net, x);
 502         if (!xi) {
 503                 xfrm_state_put(x);
 504                 return -1;
 505         }
 506
 507         if (type == NDISC_REDIRECT)
 508                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 509                              sock_net_uid(net, NULL));
 510         else
 511                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 512         xfrm_state_put(x);
 513
 514         return 0;
 515 }
 516
 517 static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
 518 {
 519         if (xi->p.link != p->link)
 520                 return -EINVAL;
 521
 522         xi->p.if_id = p->if_id;
 523
 524         return 0;
 525 }
 526
 527 static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
 528 {
 529         struct net *net = xi->net;
 530         struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 531         int err;
 532
 533         xfrmi_unlink(xfrmn, xi);
 534         synchronize_net();
 535         err = xfrmi_change(xi, p);
 536         xfrmi_link(xfrmn, xi);
 537         netdev_state_change(xi->dev);
 538         return err;
 539 }
 540
 541 static int xfrmi_get_iflink(const struct net_device *dev)
 542 {
 543         struct xfrm_if *xi = netdev_priv(dev);
 544
 545         return xi->p.link;
 546 }
 547
 548 static const struct net_device_ops xfrmi_netdev_ops = {
 549         .ndo_init       = xfrmi_dev_init,
 550         .ndo_uninit     = xfrmi_dev_uninit,
 551         .ndo_start_xmit = xfrmi_xmit,
 552         .ndo_get_stats64 = dev_get_tstats64,
 553         .ndo_get_iflink = xfrmi_get_iflink,
 554 };
 555
 556 static void xfrmi_dev_setup(struct net_device *dev)
 557 {
 558         dev->netdev_ops         = &xfrmi_netdev_ops;
 559         dev->header_ops         = &ip_tunnel_header_ops;
 560         dev->type               = ARPHRD_NONE;
 561         dev->mtu                = ETH_DATA_LEN;
 562         dev->min_mtu            = ETH_MIN_MTU;
 563         dev->max_mtu            = IP_MAX_MTU;
 564         dev->flags              = IFF_NOARP;
 565         dev->needs_free_netdev  = true;
 566         dev->priv_destructor    = xfrmi_dev_free;
 567         netif_keep_dst(dev);
 568
 569         eth_broadcast_addr(dev->broadcast);
 570 }
 571
 572 #define XFRMI_FEATURES (NETIF_F_SG |            \
 573                         NETIF_F_FRAGLIST |      \
 574                         NETIF_F_GSO_SOFTWARE |  \
 575                         NETIF_F_HW_CSUM)
 576
 577 static int xfrmi_dev_init(struct net_device *dev)
 578 {
 579         struct xfrm_if *xi = netdev_priv(dev);
 580         struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
 581         int err;
 582
 583         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 584         if (!dev->tstats)
 585                 return -ENOMEM;
 586
 587         err = gro_cells_init(&xi->gro_cells, dev);
 588         if (err) {
 589                 free_percpu(dev->tstats);
 590                 return err;
 591         }
 592
 593         dev->features |= NETIF_F_LLTX;
 594         dev->features |= XFRMI_FEATURES;
 595         dev->hw_features |= XFRMI_FEATURES;
 596
 597         if (phydev) {
 598                 dev->needed_headroom = phydev->needed_headroom;
 599                 dev->needed_tailroom = phydev->needed_tailroom;
 600
 601                 if (is_zero_ether_addr(dev->dev_addr))
 602                         eth_hw_addr_inherit(dev, phydev);
 603                 if (is_zero_ether_addr(dev->broadcast))
 604                         memcpy(dev->broadcast, phydev->broadcast,
 605                                dev->addr_len);
 606         } else {
 607                 eth_hw_addr_random(dev);
 608                 eth_broadcast_addr(dev->broadcast);
 609         }
 610
 611         return 0;
 612 }
 613
 614 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
 615                          struct netlink_ext_ack *extack)
 616 {
 617         return 0;
 618 }
 619
 620 static void xfrmi_netlink_parms(struct nlattr *data[],
 621                                struct xfrm_if_parms *parms)
 622 {
 623         memset(parms, 0, sizeof(*parms));
 624
 625         if (!data)
 626                 return;
 627
 628         if (data[IFLA_XFRM_LINK])
 629                 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
 630
 631         if (data[IFLA_XFRM_IF_ID])
 632                 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
 633 }
 634
 635 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
 636                         struct nlattr *tb[], struct nlattr *data[],
 637                         struct netlink_ext_ack *extack)
 638 {
 639         struct net *net = dev_net(dev);
 640         struct xfrm_if_parms p;
 641         struct xfrm_if *xi;
 642         int err;
 643
 644         xfrmi_netlink_parms(data, &p);
 645         xi = xfrmi_locate(net, &p);
 646         if (xi)
 647                 return -EEXIST;
 648
 649         xi = netdev_priv(dev);
 650         xi->p = p;
 651         xi->net = net;
 652         xi->dev = dev;
 653
 654         err = xfrmi_create(dev);
 655         return err;
 656 }
 657
 658 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
 659 {
 660         unregister_netdevice_queue(dev, head);
 661 }
 662
 663 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
 664                            struct nlattr *data[],
 665                            struct netlink_ext_ack *extack)
 666 {
 667         struct xfrm_if *xi = netdev_priv(dev);
 668         struct net *net = xi->net;
 669         struct xfrm_if_parms p;
 670
 671         xfrmi_netlink_parms(data, &p);
 672         xi = xfrmi_locate(net, &p);
 673         if (!xi) {
 674                 xi = netdev_priv(dev);
 675         } else {
 676                 if (xi->dev != dev)
 677                         return -EEXIST;
 678         }
 679
 680         return xfrmi_update(xi, &p);
 681 }
 682
 683 static size_t xfrmi_get_size(const struct net_device *dev)
 684 {
 685         return
 686                 /* IFLA_XFRM_LINK */
 687                 nla_total_size(4) +
 688                 /* IFLA_XFRM_IF_ID */
 689                 nla_total_size(4) +
 690                 0;
 691 }
 692
 693 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
 694 {
 695         struct xfrm_if *xi = netdev_priv(dev);
 696         struct xfrm_if_parms *parm = &xi->p;
 697
 698         if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
 699             nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
 700                 goto nla_put_failure;
 701         return 0;
 702
 703 nla_put_failure:
 704         return -EMSGSIZE;
 705 }
 706
 707 static struct net *xfrmi_get_link_net(const struct net_device *dev)
 708 {
 709         struct xfrm_if *xi = netdev_priv(dev);
 710
 711         return xi->net;
 712 }
 713
 714 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
 715         [IFLA_XFRM_LINK]        = { .type = NLA_U32 },
 716         [IFLA_XFRM_IF_ID]       = { .type = NLA_U32 },
 717 };
 718
 719 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
 720         .kind           = "xfrm",
 721         .maxtype        = IFLA_XFRM_MAX,
 722         .policy         = xfrmi_policy,
 723         .priv_size      = sizeof(struct xfrm_if),
 724         .setup          = xfrmi_dev_setup,
 725         .validate       = xfrmi_validate,
 726         .newlink        = xfrmi_newlink,
 727         .dellink        = xfrmi_dellink,
 728         .changelink     = xfrmi_changelink,
 729         .get_size       = xfrmi_get_size,
 730         .fill_info      = xfrmi_fill_info,
 731         .get_link_net   = xfrmi_get_link_net,
 732 };
 733
 734 static void __net_exit xfrmi_exit_batch_net(struct list_head *net_exit_list)
 735 {
 736         struct net *net;
 737         LIST_HEAD(list);
 738
 739         rtnl_lock();
 740         list_for_each_entry(net, net_exit_list, exit_list) {
 741                 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
 742                 struct xfrm_if __rcu **xip;
 743                 struct xfrm_if *xi;
 744                 int i;
 745
 746                 for (i = 0; i < XFRMI_HASH_SIZE; i++) {
 747                         for (xip = &xfrmn->xfrmi[i];
 748                              (xi = rtnl_dereference(*xip)) != NULL;
 749                              xip = &xi->next)
 750                                 unregister_netdevice_queue(xi->dev, &list);
 751                 }
 752         }
 753         unregister_netdevice_many(&list);
 754         rtnl_unlock();
 755 }
 756
 757 static struct pernet_operations xfrmi_net_ops = {
 758         .exit_batch = xfrmi_exit_batch_net,
 759         .id   = &xfrmi_net_id,
 760         .size = sizeof(struct xfrmi_net),
 761 };
 762
 763 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
 764         .handler        =       xfrm6_rcv,
 765         .input_handler  =       xfrm_input,
 766         .cb_handler     =       xfrmi_rcv_cb,
 767         .err_handler    =       xfrmi6_err,
 768         .priority       =       10,
 769 };
 770
 771 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
 772         .handler        =       xfrm6_rcv,
 773         .input_handler  =       xfrm_input,
 774         .cb_handler     =       xfrmi_rcv_cb,
 775         .err_handler    =       xfrmi6_err,
 776         .priority       =       10,
 777 };
 778
 779 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
 780         .handler        =       xfrm6_rcv,
 781         .input_handler  =       xfrm_input,
 782         .cb_handler     =       xfrmi_rcv_cb,
 783         .err_handler    =       xfrmi6_err,
 784         .priority       =       10,
 785 };
 786
 787 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
 788 static int xfrmi6_rcv_tunnel(struct sk_buff *skb)
 789 {
 790         const xfrm_address_t *saddr;
 791         __be32 spi;
 792
 793         saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
 794         spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
 795
 796         return xfrm6_rcv_spi(skb, IPPROTO_IPV6, spi, NULL);
 797 }
 798
 799 static struct xfrm6_tunnel xfrmi_ipv6_handler __read_mostly = {
 800         .handler        =       xfrmi6_rcv_tunnel,
 801         .cb_handler     =       xfrmi_rcv_cb,
 802         .err_handler    =       xfrmi6_err,
 803         .priority       =       2,
 804 };
 805
 806 static struct xfrm6_tunnel xfrmi_ip6ip_handler __read_mostly = {
 807         .handler        =       xfrmi6_rcv_tunnel,
 808         .cb_handler     =       xfrmi_rcv_cb,
 809         .err_handler    =       xfrmi6_err,
 810         .priority       =       2,
 811 };
 812 #endif
 813
 814 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
 815         .handler        =       xfrm4_rcv,
 816         .input_handler  =       xfrm_input,
 817         .cb_handler     =       xfrmi_rcv_cb,
 818         .err_handler    =       xfrmi4_err,
 819         .priority       =       10,
 820 };
 821
 822 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
 823         .handler        =       xfrm4_rcv,
 824         .input_handler  =       xfrm_input,
 825         .cb_handler     =       xfrmi_rcv_cb,
 826         .err_handler    =       xfrmi4_err,
 827         .priority       =       10,
 828 };
 829
 830 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
 831         .handler        =       xfrm4_rcv,
 832         .input_handler  =       xfrm_input,
 833         .cb_handler     =       xfrmi_rcv_cb,
 834         .err_handler    =       xfrmi4_err,
 835         .priority       =       10,
 836 };
 837
 838 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
 839 static int xfrmi4_rcv_tunnel(struct sk_buff *skb)
 840 {
 841         return xfrm4_rcv_spi(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr);
 842 }
 843
 844 static struct xfrm_tunnel xfrmi_ipip_handler __read_mostly = {
 845         .handler        =       xfrmi4_rcv_tunnel,
 846         .cb_handler     =       xfrmi_rcv_cb,
 847         .err_handler    =       xfrmi4_err,
 848         .priority       =       3,
 849 };
 850
 851 static struct xfrm_tunnel xfrmi_ipip6_handler __read_mostly = {
 852         .handler        =       xfrmi4_rcv_tunnel,
 853         .cb_handler     =       xfrmi_rcv_cb,
 854         .err_handler    =       xfrmi4_err,
 855         .priority       =       2,
 856 };
 857 #endif
 858
 859 static int __init xfrmi4_init(void)
 860 {
 861         int err;
 862
 863         err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
 864         if (err < 0)
 865                 goto xfrm_proto_esp_failed;
 866         err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
 867         if (err < 0)
 868                 goto xfrm_proto_ah_failed;
 869         err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
 870         if (err < 0)
 871                 goto xfrm_proto_comp_failed;
 872 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
 873         err = xfrm4_tunnel_register(&xfrmi_ipip_handler, AF_INET);
 874         if (err < 0)
 875                 goto xfrm_tunnel_ipip_failed;
 876         err = xfrm4_tunnel_register(&xfrmi_ipip6_handler, AF_INET6);
 877         if (err < 0)
 878                 goto xfrm_tunnel_ipip6_failed;
 879 #endif
 880
 881         return 0;
 882
 883 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
 884 xfrm_tunnel_ipip6_failed:
 885         xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
 886 xfrm_tunnel_ipip_failed:
 887         xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
 888 #endif
 889 xfrm_proto_comp_failed:
 890         xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
 891 xfrm_proto_ah_failed:
 892         xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
 893 xfrm_proto_esp_failed:
 894         return err;
 895 }
 896
 897 static void xfrmi4_fini(void)
 898 {
 899 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
 900         xfrm4_tunnel_deregister(&xfrmi_ipip6_handler, AF_INET6);
 901         xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
 902 #endif
 903         xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
 904         xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
 905         xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
 906 }
 907
 908 static int __init xfrmi6_init(void)
 909 {
 910         int err;
 911
 912         err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
 913         if (err < 0)
 914                 goto xfrm_proto_esp_failed;
 915         err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
 916         if (err < 0)
 917                 goto xfrm_proto_ah_failed;
 918         err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
 919         if (err < 0)
 920                 goto xfrm_proto_comp_failed;
 921 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
 922         err = xfrm6_tunnel_register(&xfrmi_ipv6_handler, AF_INET6);
 923         if (err < 0)
 924                 goto xfrm_tunnel_ipv6_failed;
 925         err = xfrm6_tunnel_register(&xfrmi_ip6ip_handler, AF_INET);
 926         if (err < 0)
 927                 goto xfrm_tunnel_ip6ip_failed;
 928 #endif
 929
 930         return 0;
 931
 932 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
 933 xfrm_tunnel_ip6ip_failed:
 934         xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
 935 xfrm_tunnel_ipv6_failed:
 936         xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
 937 #endif
 938 xfrm_proto_comp_failed:
 939         xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
 940 xfrm_proto_ah_failed:
 941         xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
 942 xfrm_proto_esp_failed:
 943         return err;
 944 }
 945
 946 static void xfrmi6_fini(void)
 947 {
 948 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
 949         xfrm6_tunnel_deregister(&xfrmi_ip6ip_handler, AF_INET);
 950         xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
 951 #endif
 952         xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
 953         xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
 954         xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
 955 }
 956
 957 static const struct xfrm_if_cb xfrm_if_cb = {
 958         .decode_session =       xfrmi_decode_session,
 959 };
 960
 961 static int __init xfrmi_init(void)
 962 {
 963         const char *msg;
 964         int err;
 965
 966         pr_info("IPsec XFRM device driver\n");
 967
 968         msg = "tunnel device";
 969         err = register_pernet_device(&xfrmi_net_ops);
 970         if (err < 0)
 971                 goto pernet_dev_failed;
 972
 973         msg = "xfrm4 protocols";
 974         err = xfrmi4_init();
 975         if (err < 0)
 976                 goto xfrmi4_failed;
 977
 978         msg = "xfrm6 protocols";
 979         err = xfrmi6_init();
 980         if (err < 0)
 981                 goto xfrmi6_failed;
 982
 983
 984         msg = "netlink interface";
 985         err = rtnl_link_register(&xfrmi_link_ops);
 986         if (err < 0)
 987                 goto rtnl_link_failed;
 988
 989         xfrm_if_register_cb(&xfrm_if_cb);
 990
 991         return err;
 992
 993 rtnl_link_failed:
 994         xfrmi6_fini();
 995 xfrmi6_failed:
 996         xfrmi4_fini();
 997 xfrmi4_failed:
 998         unregister_pernet_device(&xfrmi_net_ops);
 999 pernet_dev_failed:
1000         pr_err("xfrmi init: failed to register %s\n", msg);
1001         return err;
1002 }
1003
1004 static void __exit xfrmi_fini(void)
1005 {
1006         xfrm_if_unregister_cb();
1007         rtnl_link_unregister(&xfrmi_link_ops);
1008         xfrmi4_fini();
1009         xfrmi6_fini();
1010         unregister_pernet_device(&xfrmi_net_ops);
1011 }
1012
1013 module_init(xfrmi_init);
1014 module_exit(xfrmi_fini);
1015 MODULE_LICENSE("GPL");
1016 MODULE_ALIAS_RTNL_LINK("xfrm");
1017 MODULE_ALIAS_NETDEV("xfrm0");
1018 MODULE_AUTHOR("Steffen Klassert");
1019 MODULE_DESCRIPTION("XFRM virtual interface");