1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * VXLAN: Virtual eXtensible Local Area Network 4 * 5 * Copyright (c) 2012-2013 Vyatta Inc. 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/kernel.h> 11 #include <linux/module.h> 12 #include <linux/errno.h> 13 #include <linux/slab.h> 14 #include <linux/udp.h> 15 #include <linux/igmp.h> 16 #include <linux/if_ether.h> 17 #include <linux/ethtool.h> 18 #include <linux/rhashtable.h> 19 #include <net/arp.h> 20 #include <net/ndisc.h> 21 #include <net/gro.h> 22 #include <net/ip.h> 23 #include <net/icmp.h> 24 #include <net/rtnetlink.h> 25 #include <net/inet_ecn.h> 26 #include <net/net_namespace.h> 27 #include <net/netns/generic.h> 28 #include <net/netdev_lock.h> 29 #include <net/tun_proto.h> 30 #include <net/vxlan.h> 31 #include <net/nexthop.h> 32 33 #if IS_ENABLED(CONFIG_IPV6) 34 #include <net/ip6_tunnel.h> 35 #include <net/ip6_checksum.h> 36 #endif 37 38 #include "vxlan_private.h" 39 40 #define VXLAN_VERSION "0.1" 41 42 #define FDB_AGE_DEFAULT 300 /* 5 min */ 43 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ 44 45 /* UDP port for VXLAN traffic. 46 * The IANA assigned port is 4789, but the Linux default is 8472 47 * for compatibility with early adopters. 48 */ 49 static unsigned short vxlan_port __read_mostly = 8472; 50 module_param_named(udp_port, vxlan_port, ushort, 0444); 51 MODULE_PARM_DESC(udp_port, "Destination UDP port"); 52 53 static bool log_ecn_error = true; 54 module_param(log_ecn_error, bool, 0644); 55 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 56 57 unsigned int vxlan_net_id; 58 59 const u8 all_zeros_mac[ETH_ALEN + 2]; 60 static struct rtnl_link_ops vxlan_link_ops; 61 62 static int vxlan_sock_add(struct vxlan_dev *vxlan); 63 64 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan); 65 66 static const struct rhashtable_params vxlan_fdb_rht_params = { 67 .head_offset = offsetof(struct vxlan_fdb, rhnode), 68 .key_offset = offsetof(struct vxlan_fdb, key), 69 .key_len = sizeof(struct vxlan_fdb_key), 70 .automatic_shrinking = true, 71 }; 72 73 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) 74 { 75 return vs->flags & VXLAN_F_COLLECT_METADATA || 76 ip_tunnel_collect_metadata(); 77 } 78 79 /* Find VXLAN socket based on network namespace, address family, UDP port, 80 * enabled unshareable flags and socket device binding (see l3mdev with 81 * non-default VRF). 82 */ 83 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, 84 __be16 port, u32 flags, int ifindex) 85 { 86 struct vxlan_sock *vs; 87 88 flags &= VXLAN_F_RCV_FLAGS; 89 90 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { 91 if (inet_sk(vs->sk)->inet_sport == port && 92 vxlan_get_sk_family(vs) == family && 93 vs->flags == flags && 94 vs->sk->sk_bound_dev_if == ifindex) 95 return vs; 96 } 97 return NULL; 98 } 99 100 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, 101 int ifindex, __be32 vni, 102 struct vxlan_vni_node **vninode) 103 { 104 struct vxlan_vni_node *vnode; 105 struct vxlan_dev_node *node; 106 107 /* For flow based devices, map all packets to VNI 0 */ 108 if (vs->flags & VXLAN_F_COLLECT_METADATA && 109 !(vs->flags & VXLAN_F_VNIFILTER)) 110 vni = 0; 111 112 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) { 113 if (!node->vxlan) 114 continue; 115 vnode = NULL; 116 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 117 vnode = vxlan_vnifilter_lookup(node->vxlan, vni); 118 if (!vnode) 119 continue; 120 } else if (node->vxlan->default_dst.remote_vni != vni) { 121 continue; 122 } 123 124 if (IS_ENABLED(CONFIG_IPV6)) { 125 const struct vxlan_config *cfg = &node->vxlan->cfg; 126 127 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) && 128 cfg->remote_ifindex != ifindex) 129 continue; 130 } 131 132 if (vninode) 133 *vninode = vnode; 134 return node->vxlan; 135 } 136 137 return NULL; 138 } 139 140 /* Look up VNI in a per net namespace table */ 141 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex, 142 __be32 vni, sa_family_t family, 143 __be16 port, u32 flags) 144 { 145 struct vxlan_sock *vs; 146 147 vs = vxlan_find_sock(net, family, port, flags, ifindex); 148 if (!vs) 149 return NULL; 150 151 return vxlan_vs_find_vni(vs, ifindex, vni, NULL); 152 } 153 154 /* Fill in neighbour message in skbuff. */ 155 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, 156 const struct vxlan_fdb *fdb, 157 u32 portid, u32 seq, int type, unsigned int flags, 158 const struct vxlan_rdst *rdst) 159 { 160 unsigned long now = jiffies; 161 struct nda_cacheinfo ci; 162 bool send_ip, send_eth; 163 struct nlmsghdr *nlh; 164 struct nexthop *nh; 165 struct ndmsg *ndm; 166 int nh_family; 167 u32 nh_id; 168 169 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); 170 if (nlh == NULL) 171 return -EMSGSIZE; 172 173 ndm = nlmsg_data(nlh); 174 memset(ndm, 0, sizeof(*ndm)); 175 176 send_eth = send_ip = true; 177 178 rcu_read_lock(); 179 nh = rcu_dereference(fdb->nh); 180 if (nh) { 181 nh_family = nexthop_get_family(nh); 182 nh_id = nh->id; 183 } 184 rcu_read_unlock(); 185 186 if (type == RTM_GETNEIGH) { 187 if (rdst) { 188 send_ip = !vxlan_addr_any(&rdst->remote_ip); 189 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET; 190 } else if (nh) { 191 ndm->ndm_family = nh_family; 192 } 193 send_eth = !is_zero_ether_addr(fdb->key.eth_addr); 194 } else 195 ndm->ndm_family = AF_BRIDGE; 196 ndm->ndm_state = fdb->state; 197 ndm->ndm_ifindex = vxlan->dev->ifindex; 198 ndm->ndm_flags = fdb->flags; 199 if (rdst && rdst->offloaded) 200 ndm->ndm_flags |= NTF_OFFLOADED; 201 ndm->ndm_type = RTN_UNICAST; 202 203 if (!net_eq(dev_net(vxlan->dev), vxlan->net) && 204 nla_put_s32(skb, NDA_LINK_NETNSID, 205 peernet2id(dev_net(vxlan->dev), vxlan->net))) 206 goto nla_put_failure; 207 208 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->key.eth_addr)) 209 goto nla_put_failure; 210 if (nh) { 211 if (nla_put_u32(skb, NDA_NH_ID, nh_id)) 212 goto nla_put_failure; 213 } else if (rdst) { 214 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, 215 &rdst->remote_ip)) 216 goto nla_put_failure; 217 218 if (rdst->remote_port && 219 rdst->remote_port != vxlan->cfg.dst_port && 220 nla_put_be16(skb, NDA_PORT, rdst->remote_port)) 221 goto nla_put_failure; 222 if (rdst->remote_vni != vxlan->default_dst.remote_vni && 223 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) 224 goto nla_put_failure; 225 if (rdst->remote_ifindex && 226 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) 227 goto nla_put_failure; 228 } 229 230 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->key.vni && 231 nla_put_u32(skb, NDA_SRC_VNI, 232 be32_to_cpu(fdb->key.vni))) 233 goto nla_put_failure; 234 235 ci.ndm_used = jiffies_to_clock_t(now - READ_ONCE(fdb->used)); 236 ci.ndm_confirmed = 0; 237 ci.ndm_updated = jiffies_to_clock_t(now - READ_ONCE(fdb->updated)); 238 ci.ndm_refcnt = 0; 239 240 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) 241 goto nla_put_failure; 242 243 nlmsg_end(skb, nlh); 244 return 0; 245 246 nla_put_failure: 247 nlmsg_cancel(skb, nlh); 248 return -EMSGSIZE; 249 } 250 251 static inline size_t vxlan_nlmsg_size(void) 252 { 253 return NLMSG_ALIGN(sizeof(struct ndmsg)) 254 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ 255 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ 256 + nla_total_size(sizeof(__be16)) /* NDA_PORT */ 257 + nla_total_size(sizeof(__be32)) /* NDA_VNI */ 258 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ 259 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ 260 + nla_total_size(sizeof(struct nda_cacheinfo)); 261 } 262 263 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 264 struct vxlan_rdst *rd, int type) 265 { 266 struct net *net = dev_net(vxlan->dev); 267 struct sk_buff *skb; 268 int err = -ENOBUFS; 269 270 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); 271 if (skb == NULL) 272 goto errout; 273 274 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); 275 if (err < 0) { 276 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ 277 WARN_ON(err == -EMSGSIZE); 278 kfree_skb(skb); 279 goto errout; 280 } 281 282 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 283 return; 284 errout: 285 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 286 } 287 288 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 289 const struct vxlan_fdb *fdb, 290 const struct vxlan_rdst *rd, 291 struct netlink_ext_ack *extack, 292 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 293 { 294 fdb_info->info.dev = vxlan->dev; 295 fdb_info->info.extack = extack; 296 fdb_info->remote_ip = rd->remote_ip; 297 fdb_info->remote_port = rd->remote_port; 298 fdb_info->remote_vni = rd->remote_vni; 299 fdb_info->remote_ifindex = rd->remote_ifindex; 300 memcpy(fdb_info->eth_addr, fdb->key.eth_addr, ETH_ALEN); 301 fdb_info->vni = fdb->key.vni; 302 fdb_info->offloaded = rd->offloaded; 303 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 304 } 305 306 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 307 struct vxlan_fdb *fdb, 308 struct vxlan_rdst *rd, 309 bool adding, 310 struct netlink_ext_ack *extack) 311 { 312 struct switchdev_notifier_vxlan_fdb_info info; 313 enum switchdev_notifier_type notifier_type; 314 int ret; 315 316 if (WARN_ON(!rd)) 317 return 0; 318 319 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 320 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 321 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 322 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 323 &info.info, extack); 324 return notifier_to_errno(ret); 325 } 326 327 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 328 struct vxlan_rdst *rd, int type, bool swdev_notify, 329 struct netlink_ext_ack *extack) 330 { 331 int err; 332 333 if (swdev_notify && rd) { 334 switch (type) { 335 case RTM_NEWNEIGH: 336 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 337 true, extack); 338 if (err) 339 return err; 340 break; 341 case RTM_DELNEIGH: 342 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 343 false, extack); 344 break; 345 } 346 } 347 348 __vxlan_fdb_notify(vxlan, fdb, rd, type); 349 return 0; 350 } 351 352 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 353 { 354 struct vxlan_dev *vxlan = netdev_priv(dev); 355 struct vxlan_fdb f = { 356 .state = NUD_STALE, 357 }; 358 struct vxlan_rdst remote = { 359 .remote_ip = *ipa, /* goes to NDA_DST */ 360 .remote_vni = cpu_to_be32(VXLAN_N_VID), 361 }; 362 363 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 364 } 365 366 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 367 { 368 struct vxlan_fdb f = { 369 .state = NUD_STALE, 370 }; 371 struct vxlan_rdst remote = { }; 372 373 memcpy(f.key.eth_addr, eth_addr, ETH_ALEN); 374 375 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 376 } 377 378 /* Look up Ethernet address in forwarding table */ 379 static struct vxlan_fdb *vxlan_find_mac_rcu(struct vxlan_dev *vxlan, 380 const u8 *mac, __be32 vni) 381 { 382 struct vxlan_fdb_key key; 383 384 memset(&key, 0, sizeof(key)); 385 memcpy(key.eth_addr, mac, sizeof(key.eth_addr)); 386 if (!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) 387 key.vni = vxlan->default_dst.remote_vni; 388 else 389 key.vni = vni; 390 391 return rhashtable_lookup(&vxlan->fdb_hash_tbl, &key, 392 vxlan_fdb_rht_params); 393 } 394 395 static struct vxlan_fdb *vxlan_find_mac_tx(struct vxlan_dev *vxlan, 396 const u8 *mac, __be32 vni) 397 { 398 struct vxlan_fdb *f; 399 400 f = vxlan_find_mac_rcu(vxlan, mac, vni); 401 if (f) { 402 unsigned long now = jiffies; 403 404 if (READ_ONCE(f->used) != now) 405 WRITE_ONCE(f->used, now); 406 } 407 408 return f; 409 } 410 411 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 412 const u8 *mac, __be32 vni) 413 { 414 struct vxlan_fdb *f; 415 416 lockdep_assert_held_once(&vxlan->hash_lock); 417 418 rcu_read_lock(); 419 f = vxlan_find_mac_rcu(vxlan, mac, vni); 420 rcu_read_unlock(); 421 422 return f; 423 } 424 425 /* caller should hold vxlan->hash_lock */ 426 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 427 union vxlan_addr *ip, __be16 port, 428 __be32 vni, __u32 ifindex) 429 { 430 struct vxlan_rdst *rd; 431 432 list_for_each_entry(rd, &f->remotes, list) { 433 if (vxlan_addr_equal(&rd->remote_ip, ip) && 434 rd->remote_port == port && 435 rd->remote_vni == vni && 436 rd->remote_ifindex == ifindex) 437 return rd; 438 } 439 440 return NULL; 441 } 442 443 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 444 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 445 { 446 struct vxlan_dev *vxlan = netdev_priv(dev); 447 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 448 struct vxlan_rdst *rdst = NULL; 449 struct vxlan_fdb *f; 450 int rc = 0; 451 452 if (is_multicast_ether_addr(mac) || 453 is_zero_ether_addr(mac)) 454 return -EINVAL; 455 456 ether_addr_copy(eth_addr, mac); 457 458 rcu_read_lock(); 459 460 f = vxlan_find_mac_rcu(vxlan, eth_addr, vni); 461 if (f) 462 rdst = first_remote_rcu(f); 463 if (!rdst) { 464 rc = -ENOENT; 465 goto out; 466 } 467 468 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 469 470 out: 471 rcu_read_unlock(); 472 return rc; 473 } 474 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 475 476 static int vxlan_fdb_notify_one(struct notifier_block *nb, 477 const struct vxlan_dev *vxlan, 478 const struct vxlan_fdb *f, 479 const struct vxlan_rdst *rdst, 480 struct netlink_ext_ack *extack) 481 { 482 struct switchdev_notifier_vxlan_fdb_info fdb_info; 483 int rc; 484 485 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 486 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 487 &fdb_info); 488 return notifier_to_errno(rc); 489 } 490 491 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 492 struct notifier_block *nb, 493 struct netlink_ext_ack *extack) 494 { 495 struct vxlan_dev *vxlan; 496 struct vxlan_rdst *rdst; 497 struct vxlan_fdb *f; 498 int rc = 0; 499 500 if (!netif_is_vxlan(dev)) 501 return -EINVAL; 502 vxlan = netdev_priv(dev); 503 504 spin_lock_bh(&vxlan->hash_lock); 505 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) { 506 if (f->key.vni == vni) { 507 list_for_each_entry(rdst, &f->remotes, list) { 508 rc = vxlan_fdb_notify_one(nb, vxlan, f, rdst, 509 extack); 510 if (rc) 511 goto unlock; 512 } 513 } 514 } 515 spin_unlock_bh(&vxlan->hash_lock); 516 return 0; 517 518 unlock: 519 spin_unlock_bh(&vxlan->hash_lock); 520 return rc; 521 } 522 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 523 524 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 525 { 526 struct vxlan_dev *vxlan; 527 struct vxlan_rdst *rdst; 528 struct vxlan_fdb *f; 529 530 if (!netif_is_vxlan(dev)) 531 return; 532 vxlan = netdev_priv(dev); 533 534 spin_lock_bh(&vxlan->hash_lock); 535 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) { 536 if (f->key.vni == vni) { 537 list_for_each_entry(rdst, &f->remotes, list) 538 rdst->offloaded = false; 539 } 540 } 541 spin_unlock_bh(&vxlan->hash_lock); 542 543 } 544 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 545 546 /* Replace destination of unicast mac */ 547 static int vxlan_fdb_replace(struct vxlan_fdb *f, 548 union vxlan_addr *ip, __be16 port, __be32 vni, 549 __u32 ifindex, struct vxlan_rdst *oldrd) 550 { 551 struct vxlan_rdst *rd; 552 553 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 554 if (rd) 555 return 0; 556 557 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 558 if (!rd) 559 return 0; 560 561 *oldrd = *rd; 562 dst_cache_reset(&rd->dst_cache); 563 rd->remote_ip = *ip; 564 rd->remote_port = port; 565 rd->remote_vni = vni; 566 rd->remote_ifindex = ifindex; 567 rd->offloaded = false; 568 return 1; 569 } 570 571 /* Add/update destinations for multicast */ 572 static int vxlan_fdb_append(struct vxlan_fdb *f, 573 union vxlan_addr *ip, __be16 port, __be32 vni, 574 __u32 ifindex, struct vxlan_rdst **rdp) 575 { 576 struct vxlan_rdst *rd; 577 578 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 579 if (rd) 580 return 0; 581 582 rd = kmalloc_obj(*rd, GFP_ATOMIC); 583 if (rd == NULL) 584 return -ENOMEM; 585 586 /* The driver can work correctly without a dst cache, so do not treat 587 * dst cache initialization errors as fatal. 588 */ 589 dst_cache_init(&rd->dst_cache, GFP_ATOMIC | __GFP_NOWARN); 590 591 rd->remote_ip = *ip; 592 rd->remote_port = port; 593 rd->offloaded = false; 594 rd->remote_vni = vni; 595 rd->remote_ifindex = ifindex; 596 597 list_add_tail_rcu(&rd->list, &f->remotes); 598 599 *rdp = rd; 600 return 1; 601 } 602 603 static bool vxlan_parse_gpe_proto(const struct vxlanhdr *hdr, __be16 *protocol) 604 { 605 const struct vxlanhdr_gpe *gpe = (const struct vxlanhdr_gpe *)hdr; 606 607 /* Need to have Next Protocol set for interfaces in GPE mode. */ 608 if (!gpe->np_applied) 609 return false; 610 /* "The initial version is 0. If a receiver does not support the 611 * version indicated it MUST drop the packet. 612 */ 613 if (gpe->version != 0) 614 return false; 615 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 616 * processing MUST occur." However, we don't implement OAM 617 * processing, thus drop the packet. 618 */ 619 if (gpe->oam_flag) 620 return false; 621 622 *protocol = tun_p_to_eth_p(gpe->next_protocol); 623 if (!*protocol) 624 return false; 625 626 return true; 627 } 628 629 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 630 unsigned int off, 631 struct vxlanhdr *vh, size_t hdrlen, 632 __be32 vni_field, 633 struct gro_remcsum *grc, 634 bool nopartial) 635 { 636 size_t start, offset; 637 638 if (skb->remcsum_offload) 639 return vh; 640 641 if (!NAPI_GRO_CB(skb)->csum_valid) 642 return NULL; 643 644 start = vxlan_rco_start(vni_field); 645 offset = start + vxlan_rco_offset(vni_field); 646 647 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 648 start, offset, grc, nopartial); 649 650 skb->remcsum_offload = 1; 651 652 return vh; 653 } 654 655 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk, 656 struct list_head *head, 657 struct sk_buff *skb, 658 struct gro_remcsum *grc) 659 { 660 struct vxlanhdr *vh, *vh2; 661 unsigned int hlen, off_vx; 662 struct vxlan_sock *vs; 663 struct sk_buff *p; 664 __be32 flags; 665 666 skb_gro_remcsum_init(grc); 667 668 vs = rcu_dereference_sk_user_data(sk); 669 if (!vs) 670 return NULL; 671 672 off_vx = skb_gro_offset(skb); 673 hlen = off_vx + sizeof(*vh); 674 vh = skb_gro_header(skb, hlen, off_vx); 675 if (unlikely(!vh)) 676 return NULL; 677 678 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 679 680 flags = vh->vx_flags; 681 682 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 683 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 684 vh->vx_vni, grc, 685 !!(vs->flags & 686 VXLAN_F_REMCSUM_NOPARTIAL)); 687 688 if (!vh) 689 return NULL; 690 } 691 692 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 693 694 list_for_each_entry(p, head, list) { 695 if (!NAPI_GRO_CB(p)->same_flow) 696 continue; 697 698 vh2 = (struct vxlanhdr *)(p->data + off_vx); 699 if (vh->vx_flags != vh2->vx_flags || 700 vh->vx_vni != vh2->vx_vni) { 701 NAPI_GRO_CB(p)->same_flow = 0; 702 continue; 703 } 704 } 705 706 return vh; 707 } 708 709 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 710 struct list_head *head, 711 struct sk_buff *skb) 712 { 713 struct sk_buff *pp = NULL; 714 struct gro_remcsum grc; 715 int flush = 1; 716 717 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) { 718 pp = call_gro_receive(eth_gro_receive, head, skb); 719 flush = 0; 720 } 721 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 722 return pp; 723 } 724 725 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk, 726 struct list_head *head, 727 struct sk_buff *skb) 728 { 729 const struct packet_offload *ptype; 730 struct sk_buff *pp = NULL; 731 struct gro_remcsum grc; 732 struct vxlanhdr *vh; 733 __be16 protocol; 734 int flush = 1; 735 736 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc); 737 if (vh) { 738 if (!vxlan_parse_gpe_proto(vh, &protocol)) 739 goto out; 740 ptype = gro_find_receive_by_type(protocol); 741 if (!ptype) 742 goto out; 743 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb); 744 flush = 0; 745 } 746 out: 747 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 748 return pp; 749 } 750 751 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 752 { 753 /* Sets 'skb->inner_mac_header' since we are always called with 754 * 'skb->encapsulation' set. 755 */ 756 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 757 } 758 759 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 760 { 761 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff); 762 const struct packet_offload *ptype; 763 int err = -ENOSYS; 764 __be16 protocol; 765 766 if (!vxlan_parse_gpe_proto(vh, &protocol)) 767 return err; 768 ptype = gro_find_complete_by_type(protocol); 769 if (ptype) 770 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 771 return err; 772 } 773 774 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 775 __u16 state, __be32 src_vni, 776 __u16 ndm_flags) 777 { 778 struct vxlan_fdb *f; 779 780 f = kmalloc_obj(*f, GFP_ATOMIC); 781 if (!f) 782 return NULL; 783 memset(&f->key, 0, sizeof(f->key)); 784 f->state = state; 785 f->flags = ndm_flags; 786 f->updated = f->used = jiffies; 787 f->key.vni = src_vni; 788 f->nh = NULL; 789 RCU_INIT_POINTER(f->vdev, vxlan); 790 INIT_LIST_HEAD(&f->nh_list); 791 INIT_LIST_HEAD(&f->remotes); 792 memcpy(f->key.eth_addr, mac, ETH_ALEN); 793 794 return f; 795 } 796 797 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 798 u32 nhid, struct netlink_ext_ack *extack) 799 { 800 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 801 struct nexthop *nh; 802 int err = -EINVAL; 803 804 if (old_nh && old_nh->id == nhid) 805 return 0; 806 807 nh = nexthop_find_by_id(vxlan->net, nhid); 808 if (!nh) { 809 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 810 goto err_inval; 811 } 812 813 if (!nexthop_get(nh)) { 814 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 815 nh = NULL; 816 goto err_inval; 817 } 818 if (!nexthop_is_fdb(nh)) { 819 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 820 goto err_inval; 821 } 822 823 if (!nexthop_is_multipath(nh)) { 824 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 825 goto err_inval; 826 } 827 828 /* check nexthop group family */ 829 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 830 case AF_INET: 831 if (!nexthop_has_v4(nh)) { 832 err = -EAFNOSUPPORT; 833 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 834 goto err_inval; 835 } 836 break; 837 case AF_INET6: 838 if (nexthop_has_v4(nh)) { 839 err = -EAFNOSUPPORT; 840 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 841 goto err_inval; 842 } 843 } 844 845 if (old_nh) { 846 list_del_rcu(&fdb->nh_list); 847 nexthop_put(old_nh); 848 } 849 rcu_assign_pointer(fdb->nh, nh); 850 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 851 return 1; 852 853 err_inval: 854 if (nh) 855 nexthop_put(nh); 856 return err; 857 } 858 859 int vxlan_fdb_create(struct vxlan_dev *vxlan, 860 const u8 *mac, union vxlan_addr *ip, 861 __u16 state, __be16 port, __be32 src_vni, 862 __be32 vni, __u32 ifindex, __u16 ndm_flags, 863 u32 nhid, struct vxlan_fdb **fdb, 864 struct netlink_ext_ack *extack) 865 { 866 struct vxlan_rdst *rd = NULL; 867 struct vxlan_fdb *f; 868 int rc; 869 870 if (vxlan->cfg.addrmax && 871 vxlan->addrcnt >= vxlan->cfg.addrmax) 872 return -ENOSPC; 873 874 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 875 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 876 if (!f) 877 return -ENOMEM; 878 879 if (nhid) 880 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 881 else 882 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 883 if (rc < 0) 884 goto errout; 885 886 rc = rhashtable_lookup_insert_fast(&vxlan->fdb_hash_tbl, &f->rhnode, 887 vxlan_fdb_rht_params); 888 if (rc) 889 goto destroy_remote; 890 891 ++vxlan->addrcnt; 892 hlist_add_head_rcu(&f->fdb_node, &vxlan->fdb_list); 893 894 *fdb = f; 895 896 return 0; 897 898 destroy_remote: 899 if (rcu_access_pointer(f->nh)) { 900 list_del_rcu(&f->nh_list); 901 nexthop_put(rtnl_dereference(f->nh)); 902 } else { 903 list_del(&rd->list); 904 dst_cache_destroy(&rd->dst_cache); 905 kfree(rd); 906 } 907 errout: 908 kfree(f); 909 return rc; 910 } 911 912 static void __vxlan_fdb_free(struct vxlan_fdb *f) 913 { 914 struct vxlan_rdst *rd, *nd; 915 struct nexthop *nh; 916 917 nh = rcu_dereference_raw(f->nh); 918 if (nh) { 919 rcu_assign_pointer(f->nh, NULL); 920 rcu_assign_pointer(f->vdev, NULL); 921 nexthop_put(nh); 922 } 923 924 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 925 dst_cache_destroy(&rd->dst_cache); 926 kfree(rd); 927 } 928 kfree(f); 929 } 930 931 static void vxlan_fdb_free(struct rcu_head *head) 932 { 933 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 934 935 __vxlan_fdb_free(f); 936 } 937 938 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 939 bool do_notify, bool swdev_notify) 940 { 941 struct vxlan_rdst *rd; 942 943 netdev_dbg(vxlan->dev, "delete %pM\n", f->key.eth_addr); 944 945 --vxlan->addrcnt; 946 if (do_notify) { 947 if (rcu_access_pointer(f->nh)) 948 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 949 swdev_notify, NULL); 950 else 951 list_for_each_entry(rd, &f->remotes, list) 952 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 953 swdev_notify, NULL); 954 } 955 956 hlist_del_init_rcu(&f->fdb_node); 957 rhashtable_remove_fast(&vxlan->fdb_hash_tbl, &f->rhnode, 958 vxlan_fdb_rht_params); 959 list_del_rcu(&f->nh_list); 960 call_rcu(&f->rcu, vxlan_fdb_free); 961 } 962 963 static void vxlan_dst_free(struct rcu_head *head) 964 { 965 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 966 967 dst_cache_destroy(&rd->dst_cache); 968 kfree(rd); 969 } 970 971 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 972 union vxlan_addr *ip, 973 __u16 state, __u16 flags, 974 __be16 port, __be32 vni, 975 __u32 ifindex, __u16 ndm_flags, 976 struct vxlan_fdb *f, u32 nhid, 977 bool swdev_notify, 978 struct netlink_ext_ack *extack) 979 { 980 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 981 struct vxlan_rdst *rd = NULL; 982 struct vxlan_rdst oldrd; 983 int notify = 0; 984 int rc = 0; 985 int err; 986 987 if (nhid && !rcu_access_pointer(f->nh)) { 988 NL_SET_ERR_MSG(extack, 989 "Cannot replace an existing non nexthop fdb with a nexthop"); 990 return -EOPNOTSUPP; 991 } 992 993 if (nhid && (flags & NLM_F_APPEND)) { 994 NL_SET_ERR_MSG(extack, 995 "Cannot append to a nexthop fdb"); 996 return -EOPNOTSUPP; 997 } 998 999 /* Do not allow an externally learned entry to take over an entry added 1000 * by the user. 1001 */ 1002 if (!(fdb_flags & NTF_EXT_LEARNED) || 1003 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 1004 if (f->state != state) { 1005 f->state = state; 1006 notify = 1; 1007 } 1008 if (f->flags != fdb_flags) { 1009 f->flags = fdb_flags; 1010 notify = 1; 1011 } 1012 } 1013 1014 if ((flags & NLM_F_REPLACE)) { 1015 /* Only change unicasts */ 1016 if (!(is_multicast_ether_addr(f->key.eth_addr) || 1017 is_zero_ether_addr(f->key.eth_addr))) { 1018 if (nhid) { 1019 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1020 if (rc < 0) 1021 return rc; 1022 } else { 1023 rc = vxlan_fdb_replace(f, ip, port, vni, 1024 ifindex, &oldrd); 1025 } 1026 notify |= rc; 1027 } else { 1028 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1029 return -EOPNOTSUPP; 1030 } 1031 } 1032 if ((flags & NLM_F_APPEND) && 1033 (is_multicast_ether_addr(f->key.eth_addr) || 1034 is_zero_ether_addr(f->key.eth_addr))) { 1035 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1036 1037 if (rc < 0) 1038 return rc; 1039 notify |= rc; 1040 } 1041 1042 if (ndm_flags & NTF_USE) 1043 WRITE_ONCE(f->updated, jiffies); 1044 1045 if (notify) { 1046 if (rd == NULL) 1047 rd = first_remote_rtnl(f); 1048 1049 WRITE_ONCE(f->updated, jiffies); 1050 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1051 swdev_notify, extack); 1052 if (err) 1053 goto err_notify; 1054 } 1055 1056 return 0; 1057 1058 err_notify: 1059 if (nhid) 1060 return err; 1061 if ((flags & NLM_F_REPLACE) && rc) 1062 *rd = oldrd; 1063 else if ((flags & NLM_F_APPEND) && rc) { 1064 list_del_rcu(&rd->list); 1065 call_rcu(&rd->rcu, vxlan_dst_free); 1066 } 1067 return err; 1068 } 1069 1070 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1071 const u8 *mac, union vxlan_addr *ip, 1072 __u16 state, __u16 flags, 1073 __be16 port, __be32 src_vni, __be32 vni, 1074 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1075 bool swdev_notify, 1076 struct netlink_ext_ack *extack) 1077 { 1078 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1079 struct vxlan_fdb *f; 1080 int rc; 1081 1082 /* Disallow replace to add a multicast entry */ 1083 if ((flags & NLM_F_REPLACE) && 1084 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1085 return -EOPNOTSUPP; 1086 1087 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1088 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1089 vni, ifindex, fdb_flags, nhid, &f, extack); 1090 if (rc < 0) 1091 return rc; 1092 1093 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1094 swdev_notify, extack); 1095 if (rc) 1096 goto err_notify; 1097 1098 return 0; 1099 1100 err_notify: 1101 vxlan_fdb_destroy(vxlan, f, false, false); 1102 return rc; 1103 } 1104 1105 /* Add new entry to forwarding table -- assumes lock held */ 1106 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1107 const u8 *mac, union vxlan_addr *ip, 1108 __u16 state, __u16 flags, 1109 __be16 port, __be32 src_vni, __be32 vni, 1110 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1111 bool swdev_notify, 1112 struct netlink_ext_ack *extack) 1113 { 1114 struct vxlan_fdb *f; 1115 1116 f = vxlan_find_mac(vxlan, mac, src_vni); 1117 if (f) { 1118 if (flags & NLM_F_EXCL) { 1119 netdev_dbg(vxlan->dev, 1120 "lost race to create %pM\n", mac); 1121 return -EEXIST; 1122 } 1123 1124 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1125 vni, ifindex, ndm_flags, f, 1126 nhid, swdev_notify, extack); 1127 } else { 1128 if (!(flags & NLM_F_CREATE)) 1129 return -ENOENT; 1130 1131 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1132 port, src_vni, vni, ifindex, 1133 ndm_flags, nhid, swdev_notify, 1134 extack); 1135 } 1136 } 1137 1138 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1139 struct vxlan_rdst *rd, bool swdev_notify) 1140 { 1141 list_del_rcu(&rd->list); 1142 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1143 call_rcu(&rd->rcu, vxlan_dst_free); 1144 } 1145 1146 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1147 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1148 __be32 *vni, u32 *ifindex, u32 *nhid, 1149 struct netlink_ext_ack *extack) 1150 { 1151 struct net *net = dev_net(vxlan->dev); 1152 int err; 1153 1154 if (tb[NDA_NH_ID] && 1155 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1156 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1157 return -EINVAL; 1158 } 1159 1160 if (tb[NDA_DST]) { 1161 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1162 if (err) { 1163 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1164 return err; 1165 } 1166 } else { 1167 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1168 1169 if (remote->sa.sa_family == AF_INET) { 1170 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1171 ip->sa.sa_family = AF_INET; 1172 #if IS_ENABLED(CONFIG_IPV6) 1173 } else { 1174 ip->sin6.sin6_addr = in6addr_any; 1175 ip->sa.sa_family = AF_INET6; 1176 #endif 1177 } 1178 } 1179 1180 if (tb[NDA_PORT]) { 1181 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1182 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1183 return -EINVAL; 1184 } 1185 *port = nla_get_be16(tb[NDA_PORT]); 1186 } else { 1187 *port = vxlan->cfg.dst_port; 1188 } 1189 1190 if (tb[NDA_VNI]) { 1191 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1192 NL_SET_ERR_MSG(extack, "Invalid vni"); 1193 return -EINVAL; 1194 } 1195 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1196 } else { 1197 *vni = vxlan->default_dst.remote_vni; 1198 } 1199 1200 if (tb[NDA_SRC_VNI]) { 1201 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1202 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1203 return -EINVAL; 1204 } 1205 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1206 } else { 1207 *src_vni = vxlan->default_dst.remote_vni; 1208 } 1209 1210 if (tb[NDA_IFINDEX]) { 1211 struct net_device *tdev; 1212 1213 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1214 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1215 return -EINVAL; 1216 } 1217 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1218 tdev = __dev_get_by_index(net, *ifindex); 1219 if (!tdev) { 1220 NL_SET_ERR_MSG(extack, "Device not found"); 1221 return -EADDRNOTAVAIL; 1222 } 1223 } else { 1224 *ifindex = 0; 1225 } 1226 1227 *nhid = nla_get_u32_default(tb[NDA_NH_ID], 0); 1228 1229 return 0; 1230 } 1231 1232 /* Add static entry (via netlink) */ 1233 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1234 struct net_device *dev, 1235 const unsigned char *addr, u16 vid, u16 flags, 1236 bool *notified, struct netlink_ext_ack *extack) 1237 { 1238 struct vxlan_dev *vxlan = netdev_priv(dev); 1239 /* struct net *net = dev_net(vxlan->dev); */ 1240 union vxlan_addr ip; 1241 __be16 port; 1242 __be32 src_vni, vni; 1243 u32 ifindex, nhid; 1244 int err; 1245 1246 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1247 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1248 ndm->ndm_state); 1249 return -EINVAL; 1250 } 1251 1252 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1253 return -EINVAL; 1254 1255 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1256 &nhid, extack); 1257 if (err) 1258 return err; 1259 1260 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1261 return -EAFNOSUPPORT; 1262 1263 spin_lock_bh(&vxlan->hash_lock); 1264 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1265 port, src_vni, vni, ifindex, 1266 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1267 nhid, true, extack); 1268 spin_unlock_bh(&vxlan->hash_lock); 1269 1270 if (!err) 1271 *notified = true; 1272 1273 return err; 1274 } 1275 1276 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1277 const unsigned char *addr, union vxlan_addr ip, 1278 __be16 port, __be32 src_vni, __be32 vni, 1279 u32 ifindex, bool swdev_notify) 1280 { 1281 struct vxlan_rdst *rd = NULL; 1282 struct vxlan_fdb *f; 1283 int err = -ENOENT; 1284 1285 f = vxlan_find_mac(vxlan, addr, src_vni); 1286 if (!f) 1287 return err; 1288 1289 if (!vxlan_addr_any(&ip)) { 1290 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1291 if (!rd) 1292 goto out; 1293 } 1294 1295 /* remove a destination if it's not the only one on the list, 1296 * otherwise destroy the fdb entry 1297 */ 1298 if (rd && !list_is_singular(&f->remotes)) { 1299 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1300 goto out; 1301 } 1302 1303 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1304 1305 out: 1306 return 0; 1307 } 1308 1309 /* Delete entry (via netlink) */ 1310 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1311 struct net_device *dev, 1312 const unsigned char *addr, u16 vid, bool *notified, 1313 struct netlink_ext_ack *extack) 1314 { 1315 struct vxlan_dev *vxlan = netdev_priv(dev); 1316 union vxlan_addr ip; 1317 __be32 src_vni, vni; 1318 u32 ifindex, nhid; 1319 __be16 port; 1320 int err; 1321 1322 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1323 &nhid, extack); 1324 if (err) 1325 return err; 1326 1327 spin_lock_bh(&vxlan->hash_lock); 1328 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1329 true); 1330 spin_unlock_bh(&vxlan->hash_lock); 1331 1332 if (!err) 1333 *notified = true; 1334 1335 return err; 1336 } 1337 1338 /* Dump forwarding table */ 1339 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1340 struct net_device *dev, 1341 struct net_device *filter_dev, int *idx) 1342 { 1343 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx; 1344 struct vxlan_dev *vxlan = netdev_priv(dev); 1345 struct vxlan_fdb *f; 1346 int err = 0; 1347 1348 rcu_read_lock(); 1349 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 1350 struct vxlan_rdst *rd; 1351 1352 if (rcu_access_pointer(f->nh)) { 1353 if (*idx < ctx->fdb_idx) 1354 goto skip_nh; 1355 err = vxlan_fdb_info(skb, vxlan, f, 1356 NETLINK_CB(cb->skb).portid, 1357 cb->nlh->nlmsg_seq, 1358 RTM_NEWNEIGH, NLM_F_MULTI, NULL); 1359 if (err < 0) { 1360 rcu_read_unlock(); 1361 goto out; 1362 } 1363 skip_nh: 1364 *idx += 1; 1365 continue; 1366 } 1367 1368 list_for_each_entry_rcu(rd, &f->remotes, list) { 1369 if (*idx < ctx->fdb_idx) 1370 goto skip; 1371 1372 err = vxlan_fdb_info(skb, vxlan, f, 1373 NETLINK_CB(cb->skb).portid, 1374 cb->nlh->nlmsg_seq, 1375 RTM_NEWNEIGH, NLM_F_MULTI, rd); 1376 if (err < 0) { 1377 rcu_read_unlock(); 1378 goto out; 1379 } 1380 skip: 1381 *idx += 1; 1382 } 1383 } 1384 rcu_read_unlock(); 1385 out: 1386 return err; 1387 } 1388 1389 static int vxlan_fdb_get(struct sk_buff *skb, 1390 struct nlattr *tb[], 1391 struct net_device *dev, 1392 const unsigned char *addr, 1393 u16 vid, u32 portid, u32 seq, 1394 struct netlink_ext_ack *extack) 1395 { 1396 struct vxlan_dev *vxlan = netdev_priv(dev); 1397 struct vxlan_fdb *f; 1398 __be32 vni; 1399 int err; 1400 1401 if (tb[NDA_VNI]) 1402 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1403 else 1404 vni = vxlan->default_dst.remote_vni; 1405 1406 rcu_read_lock(); 1407 1408 f = vxlan_find_mac_rcu(vxlan, addr, vni); 1409 if (!f) { 1410 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1411 err = -ENOENT; 1412 goto errout; 1413 } 1414 1415 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1416 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1417 errout: 1418 rcu_read_unlock(); 1419 return err; 1420 } 1421 1422 /* Watch incoming packets to learn mapping between Ethernet address 1423 * and Tunnel endpoint. 1424 */ 1425 static enum skb_drop_reason vxlan_snoop(struct net_device *dev, 1426 union vxlan_addr *src_ip, 1427 const u8 *src_mac, u32 src_ifindex, 1428 __be32 vni) 1429 { 1430 struct vxlan_dev *vxlan = netdev_priv(dev); 1431 struct vxlan_fdb *f; 1432 u32 ifindex = 0; 1433 1434 /* Ignore packets from invalid src-address */ 1435 if (!is_valid_ether_addr(src_mac)) 1436 return SKB_DROP_REASON_MAC_INVALID_SOURCE; 1437 1438 #if IS_ENABLED(CONFIG_IPV6) 1439 if (src_ip->sa.sa_family == AF_INET6 && 1440 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1441 ifindex = src_ifindex; 1442 #endif 1443 1444 f = vxlan_find_mac_rcu(vxlan, src_mac, vni); 1445 if (likely(f)) { 1446 struct vxlan_rdst *rdst = first_remote_rcu(f); 1447 unsigned long now = jiffies; 1448 1449 if (READ_ONCE(f->updated) != now) 1450 WRITE_ONCE(f->updated, now); 1451 1452 /* Don't override an fdb with nexthop with a learnt entry */ 1453 if (rcu_access_pointer(f->nh)) 1454 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1455 1456 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1457 rdst->remote_ifindex == ifindex)) 1458 return SKB_NOT_DROPPED_YET; 1459 1460 /* Don't migrate static entries, drop packets */ 1461 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1462 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1463 1464 if (net_ratelimit()) 1465 netdev_info(dev, 1466 "%pM migrated from %pIS to %pIS\n", 1467 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1468 1469 rdst->remote_ip = *src_ip; 1470 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1471 } else { 1472 /* learned new entry */ 1473 spin_lock(&vxlan->hash_lock); 1474 1475 /* close off race between vxlan_flush and incoming packets */ 1476 if (netif_running(dev)) 1477 vxlan_fdb_update(vxlan, src_mac, src_ip, 1478 NUD_REACHABLE, 1479 NLM_F_EXCL|NLM_F_CREATE, 1480 vxlan->cfg.dst_port, 1481 vni, 1482 vxlan->default_dst.remote_vni, 1483 ifindex, NTF_SELF, 0, true, NULL); 1484 spin_unlock(&vxlan->hash_lock); 1485 } 1486 1487 return SKB_NOT_DROPPED_YET; 1488 } 1489 1490 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1491 { 1492 ASSERT_RTNL(); 1493 1494 if (!vs) 1495 return false; 1496 if (!refcount_dec_and_test(&vs->refcnt)) 1497 return false; 1498 1499 hlist_del_rcu(&vs->hlist); 1500 udp_tunnel_notify_del_rx_port(vs->sk, 1501 (vs->flags & VXLAN_F_GPE) ? 1502 UDP_TUNNEL_TYPE_VXLAN_GPE : 1503 UDP_TUNNEL_TYPE_VXLAN); 1504 1505 return true; 1506 } 1507 1508 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1509 { 1510 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1511 #if IS_ENABLED(CONFIG_IPV6) 1512 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1513 1514 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1515 #endif 1516 1517 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1518 synchronize_net(); 1519 1520 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1521 vxlan_vs_del_vnigrp(vxlan); 1522 else 1523 vxlan_vs_del_dev(vxlan); 1524 1525 if (__vxlan_sock_release_prep(sock4)) { 1526 udp_tunnel_sock_release(sock4->sk); 1527 kfree_rcu(sock4, rcu); 1528 } 1529 1530 #if IS_ENABLED(CONFIG_IPV6) 1531 if (__vxlan_sock_release_prep(sock6)) { 1532 udp_tunnel_sock_release(sock6->sk); 1533 kfree_rcu(sock6, rcu); 1534 } 1535 #endif 1536 } 1537 1538 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags) 1539 { 1540 const struct vxlanhdr *vh = vxlan_hdr(skb); 1541 enum skb_drop_reason reason; 1542 size_t start, offset; 1543 1544 if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1545 return SKB_NOT_DROPPED_YET; 1546 1547 start = vxlan_rco_start(vh->vx_vni); 1548 offset = start + vxlan_rco_offset(vh->vx_vni); 1549 1550 reason = pskb_may_pull_reason(skb, offset + sizeof(u16)); 1551 if (reason) 1552 return reason; 1553 1554 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1555 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1556 return SKB_NOT_DROPPED_YET; 1557 } 1558 1559 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags, 1560 struct vxlan_metadata *md) 1561 { 1562 const struct vxlanhdr *vh = vxlan_hdr(skb); 1563 const struct vxlanhdr_gbp *gbp; 1564 struct metadata_dst *tun_dst; 1565 1566 gbp = (const struct vxlanhdr_gbp *)vh; 1567 1568 if (!(vh->vx_flags & VXLAN_HF_GBP)) 1569 return; 1570 1571 md->gbp = ntohs(gbp->policy_id); 1572 1573 tun_dst = (struct metadata_dst *)skb_dst(skb); 1574 if (tun_dst) { 1575 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1576 tun_dst->u.tun_info.key.tun_flags); 1577 tun_dst->u.tun_info.options_len = sizeof(*md); 1578 } 1579 if (gbp->dont_learn) 1580 md->gbp |= VXLAN_GBP_DONT_LEARN; 1581 1582 if (gbp->policy_applied) 1583 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1584 1585 /* In flow-based mode, GBP is carried in dst_metadata */ 1586 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1587 skb->mark = md->gbp; 1588 } 1589 1590 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan, 1591 struct vxlan_sock *vs, 1592 struct sk_buff *skb, __be32 vni) 1593 { 1594 union vxlan_addr saddr; 1595 u32 ifindex = skb->dev->ifindex; 1596 1597 skb_reset_mac_header(skb); 1598 skb->protocol = eth_type_trans(skb, vxlan->dev); 1599 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1600 1601 /* Ignore packet loops (and multicast echo) */ 1602 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1603 return SKB_DROP_REASON_LOCAL_MAC; 1604 1605 /* Get address from the outer IP header */ 1606 if (vxlan_get_sk_family(vs) == AF_INET) { 1607 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1608 saddr.sa.sa_family = AF_INET; 1609 #if IS_ENABLED(CONFIG_IPV6) 1610 } else { 1611 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1612 saddr.sa.sa_family = AF_INET6; 1613 #endif 1614 } 1615 1616 if (!(vxlan->cfg.flags & VXLAN_F_LEARN)) 1617 return SKB_NOT_DROPPED_YET; 1618 1619 return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, 1620 ifindex, vni); 1621 } 1622 1623 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1624 struct sk_buff *skb) 1625 { 1626 int err = 0; 1627 1628 if (vxlan_get_sk_family(vs) == AF_INET) 1629 err = IP_ECN_decapsulate(oiph, skb); 1630 #if IS_ENABLED(CONFIG_IPV6) 1631 else 1632 err = IP6_ECN_decapsulate(oiph, skb); 1633 #endif 1634 1635 if (unlikely(err) && log_ecn_error) { 1636 if (vxlan_get_sk_family(vs) == AF_INET) 1637 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1638 &((struct iphdr *)oiph)->saddr, 1639 ((struct iphdr *)oiph)->tos); 1640 else 1641 net_info_ratelimited("non-ECT from %pI6\n", 1642 &((struct ipv6hdr *)oiph)->saddr); 1643 } 1644 return err <= 1; 1645 } 1646 1647 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1648 { 1649 struct vxlan_vni_node *vninode = NULL; 1650 const struct vxlanhdr *vh; 1651 struct vxlan_dev *vxlan; 1652 struct vxlan_sock *vs; 1653 struct vxlan_metadata _md; 1654 struct vxlan_metadata *md = &_md; 1655 __be16 protocol = htons(ETH_P_TEB); 1656 enum skb_drop_reason reason; 1657 bool raw_proto = false; 1658 void *oiph; 1659 __be32 vni = 0; 1660 int nh; 1661 1662 /* Need UDP and VXLAN header to be present */ 1663 reason = pskb_may_pull_reason(skb, VXLAN_HLEN); 1664 if (reason) 1665 goto drop; 1666 1667 vh = vxlan_hdr(skb); 1668 /* VNI flag always required to be set */ 1669 if (!(vh->vx_flags & VXLAN_HF_VNI)) { 1670 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1671 ntohl(vh->vx_flags), ntohl(vh->vx_vni)); 1672 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1673 /* Return non vxlan pkt */ 1674 goto drop; 1675 } 1676 1677 vs = rcu_dereference_sk_user_data(sk); 1678 if (!vs) 1679 goto drop; 1680 1681 vni = vxlan_vni(vh->vx_vni); 1682 1683 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1684 if (!vxlan) { 1685 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND; 1686 goto drop; 1687 } 1688 1689 if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags || 1690 vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) { 1691 /* If the header uses bits besides those enabled by the 1692 * netdevice configuration, treat this as a malformed packet. 1693 * This behavior diverges from VXLAN RFC (RFC7348) which 1694 * stipulates that bits in reserved in reserved fields are to be 1695 * ignored. The approach here maintains compatibility with 1696 * previous stack code, and also is more robust and provides a 1697 * little more security in adding extensions to VXLAN. 1698 */ 1699 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1700 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1701 DEV_STATS_INC(vxlan->dev, rx_errors); 1702 vxlan_vnifilter_count(vxlan, vni, vninode, 1703 VXLAN_VNI_STATS_RX_ERRORS, 0); 1704 goto drop; 1705 } 1706 1707 if (vxlan->cfg.flags & VXLAN_F_GPE) { 1708 if (!vxlan_parse_gpe_proto(vh, &protocol)) 1709 goto drop; 1710 raw_proto = true; 1711 } 1712 1713 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1714 !net_eq(vxlan->net, dev_net(vxlan->dev)))) { 1715 reason = SKB_DROP_REASON_NOMEM; 1716 goto drop; 1717 } 1718 1719 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) { 1720 reason = vxlan_remcsum(skb, vxlan->cfg.flags); 1721 if (unlikely(reason)) 1722 goto drop; 1723 } 1724 1725 if (vxlan_collect_metadata(vs)) { 1726 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1727 struct metadata_dst *tun_dst; 1728 1729 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1730 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1731 key32_to_tunnel_id(vni), sizeof(*md)); 1732 1733 if (!tun_dst) { 1734 reason = SKB_DROP_REASON_NOMEM; 1735 goto drop; 1736 } 1737 1738 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1739 1740 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1741 } else { 1742 memset(md, 0, sizeof(*md)); 1743 } 1744 1745 if (vxlan->cfg.flags & VXLAN_F_GBP) 1746 vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md); 1747 /* Note that GBP and GPE can never be active together. This is 1748 * ensured in vxlan_dev_configure. 1749 */ 1750 1751 if (!raw_proto) { 1752 reason = vxlan_set_mac(vxlan, vs, skb, vni); 1753 if (reason) 1754 goto drop; 1755 } else { 1756 skb_reset_mac_header(skb); 1757 skb->dev = vxlan->dev; 1758 skb->pkt_type = PACKET_HOST; 1759 } 1760 1761 /* Save offset of outer header relative to skb->head, 1762 * because we are going to reset the network header to the inner header 1763 * and might change skb->head. 1764 */ 1765 nh = skb_network_header(skb) - skb->head; 1766 1767 skb_reset_network_header(skb); 1768 1769 reason = pskb_inet_may_pull_reason(skb); 1770 if (reason) { 1771 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1772 DEV_STATS_INC(vxlan->dev, rx_errors); 1773 vxlan_vnifilter_count(vxlan, vni, vninode, 1774 VXLAN_VNI_STATS_RX_ERRORS, 0); 1775 goto drop; 1776 } 1777 1778 /* Get the outer header. */ 1779 oiph = skb->head + nh; 1780 1781 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1782 reason = SKB_DROP_REASON_IP_TUNNEL_ECN; 1783 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1784 DEV_STATS_INC(vxlan->dev, rx_errors); 1785 vxlan_vnifilter_count(vxlan, vni, vninode, 1786 VXLAN_VNI_STATS_RX_ERRORS, 0); 1787 goto drop; 1788 } 1789 1790 rcu_read_lock(); 1791 1792 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1793 rcu_read_unlock(); 1794 dev_dstats_rx_dropped(vxlan->dev); 1795 vxlan_vnifilter_count(vxlan, vni, vninode, 1796 VXLAN_VNI_STATS_RX_DROPS, 0); 1797 reason = SKB_DROP_REASON_DEV_READY; 1798 goto drop; 1799 } 1800 1801 dev_dstats_rx_add(vxlan->dev, skb->len); 1802 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1803 gro_cells_receive(&vxlan->gro_cells, skb); 1804 1805 rcu_read_unlock(); 1806 1807 return 0; 1808 1809 drop: 1810 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED; 1811 /* Consume bad packet */ 1812 kfree_skb_reason(skb, reason); 1813 return 0; 1814 } 1815 1816 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1817 { 1818 struct vxlan_dev *vxlan; 1819 struct vxlan_sock *vs; 1820 struct vxlanhdr *hdr; 1821 __be32 vni; 1822 1823 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1824 return -EINVAL; 1825 1826 hdr = vxlan_hdr(skb); 1827 1828 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1829 return -EINVAL; 1830 1831 vs = rcu_dereference_sk_user_data(sk); 1832 if (!vs) 1833 return -ENOENT; 1834 1835 vni = vxlan_vni(hdr->vx_vni); 1836 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1837 if (!vxlan) 1838 return -ENOENT; 1839 1840 return 0; 1841 } 1842 1843 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1844 { 1845 struct vxlan_dev *vxlan = netdev_priv(dev); 1846 struct arphdr *parp; 1847 u8 *arpptr, *sha; 1848 __be32 sip, tip; 1849 struct neighbour *n; 1850 1851 if (dev->flags & IFF_NOARP) 1852 goto out; 1853 1854 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1855 dev_dstats_tx_dropped(dev); 1856 vxlan_vnifilter_count(vxlan, vni, NULL, 1857 VXLAN_VNI_STATS_TX_DROPS, 0); 1858 goto out; 1859 } 1860 parp = arp_hdr(skb); 1861 1862 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1863 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1864 parp->ar_pro != htons(ETH_P_IP) || 1865 parp->ar_op != htons(ARPOP_REQUEST) || 1866 parp->ar_hln != dev->addr_len || 1867 parp->ar_pln != 4) 1868 goto out; 1869 arpptr = (u8 *)parp + sizeof(struct arphdr); 1870 sha = arpptr; 1871 arpptr += dev->addr_len; /* sha */ 1872 memcpy(&sip, arpptr, sizeof(sip)); 1873 arpptr += sizeof(sip); 1874 arpptr += dev->addr_len; /* tha */ 1875 memcpy(&tip, arpptr, sizeof(tip)); 1876 1877 if (ipv4_is_loopback(tip) || 1878 ipv4_is_multicast(tip)) 1879 goto out; 1880 1881 n = neigh_lookup(&arp_tbl, &tip, dev); 1882 1883 if (n) { 1884 struct vxlan_rdst *rdst = NULL; 1885 struct vxlan_fdb *f; 1886 struct sk_buff *reply; 1887 1888 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1889 neigh_release(n); 1890 goto out; 1891 } 1892 1893 rcu_read_lock(); 1894 f = vxlan_find_mac_tx(vxlan, n->ha, vni); 1895 if (f) 1896 rdst = first_remote_rcu(f); 1897 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 1898 /* bridge-local neighbor */ 1899 neigh_release(n); 1900 rcu_read_unlock(); 1901 goto out; 1902 } 1903 rcu_read_unlock(); 1904 1905 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1906 n->ha, sha); 1907 1908 neigh_release(n); 1909 1910 if (reply == NULL) 1911 goto out; 1912 1913 skb_reset_mac_header(reply); 1914 __skb_pull(reply, skb_network_offset(reply)); 1915 reply->ip_summed = CHECKSUM_UNNECESSARY; 1916 reply->pkt_type = PACKET_HOST; 1917 1918 if (netif_rx(reply) == NET_RX_DROP) { 1919 dev_dstats_rx_dropped(dev); 1920 vxlan_vnifilter_count(vxlan, vni, NULL, 1921 VXLAN_VNI_STATS_RX_DROPS, 0); 1922 } 1923 1924 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1925 union vxlan_addr ipa = { 1926 .sin.sin_addr.s_addr = tip, 1927 .sin.sin_family = AF_INET, 1928 }; 1929 1930 vxlan_ip_miss(dev, &ipa); 1931 } 1932 out: 1933 consume_skb(skb); 1934 return NETDEV_TX_OK; 1935 } 1936 1937 #if IS_ENABLED(CONFIG_IPV6) 1938 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1939 struct neighbour *n, bool isrouter) 1940 { 1941 struct net_device *dev = request->dev; 1942 struct sk_buff *reply; 1943 struct nd_msg *ns, *na; 1944 struct ipv6hdr *pip6; 1945 u8 *daddr; 1946 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1947 int ns_olen; 1948 int i, len; 1949 1950 if (dev == NULL || !pskb_may_pull(request, request->len)) 1951 return NULL; 1952 1953 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1954 sizeof(*na) + na_olen + dev->needed_tailroom; 1955 reply = alloc_skb(len, GFP_ATOMIC); 1956 if (reply == NULL) 1957 return NULL; 1958 1959 reply->protocol = htons(ETH_P_IPV6); 1960 reply->dev = dev; 1961 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1962 skb_push(reply, sizeof(struct ethhdr)); 1963 skb_reset_mac_header(reply); 1964 1965 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1966 1967 daddr = eth_hdr(request)->h_source; 1968 ns_olen = request->len - skb_network_offset(request) - 1969 sizeof(struct ipv6hdr) - sizeof(*ns); 1970 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1971 if (!ns->opt[i + 1] || i + (ns->opt[i + 1] << 3) > ns_olen) { 1972 kfree_skb(reply); 1973 return NULL; 1974 } 1975 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1976 if ((ns->opt[i + 1] << 3) >= 1977 sizeof(struct nd_opt_hdr) + ETH_ALEN) 1978 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1979 break; 1980 } 1981 } 1982 1983 /* Ethernet header */ 1984 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1985 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1986 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1987 reply->protocol = htons(ETH_P_IPV6); 1988 1989 skb_pull(reply, sizeof(struct ethhdr)); 1990 skb_reset_network_header(reply); 1991 skb_put(reply, sizeof(struct ipv6hdr)); 1992 1993 /* IPv6 header */ 1994 1995 pip6 = ipv6_hdr(reply); 1996 memset(pip6, 0, sizeof(struct ipv6hdr)); 1997 pip6->version = 6; 1998 pip6->priority = ipv6_hdr(request)->priority; 1999 pip6->nexthdr = IPPROTO_ICMPV6; 2000 pip6->hop_limit = 255; 2001 pip6->daddr = ipv6_hdr(request)->saddr; 2002 pip6->saddr = *(struct in6_addr *)n->primary_key; 2003 2004 skb_pull(reply, sizeof(struct ipv6hdr)); 2005 skb_reset_transport_header(reply); 2006 2007 /* Neighbor Advertisement */ 2008 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2009 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2010 na->icmph.icmp6_router = isrouter; 2011 na->icmph.icmp6_override = 1; 2012 na->icmph.icmp6_solicited = 1; 2013 na->target = ns->target; 2014 ether_addr_copy(&na->opt[2], n->ha); 2015 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2016 na->opt[1] = na_olen >> 3; 2017 2018 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2019 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2020 csum_partial(na, sizeof(*na)+na_olen, 0)); 2021 2022 pip6->payload_len = htons(sizeof(*na)+na_olen); 2023 2024 skb_push(reply, sizeof(struct ipv6hdr)); 2025 2026 reply->ip_summed = CHECKSUM_UNNECESSARY; 2027 2028 return reply; 2029 } 2030 2031 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2032 { 2033 struct vxlan_dev *vxlan = netdev_priv(dev); 2034 const struct in6_addr *daddr; 2035 const struct ipv6hdr *iphdr; 2036 struct neighbour *n; 2037 struct nd_msg *msg; 2038 2039 rcu_read_lock(); 2040 if (unlikely(!ipv6_mod_enabled())) 2041 goto out; 2042 2043 iphdr = ipv6_hdr(skb); 2044 daddr = &iphdr->daddr; 2045 msg = (struct nd_msg *)(iphdr + 1); 2046 2047 if (ipv6_addr_loopback(daddr) || 2048 ipv6_addr_is_multicast(&msg->target)) 2049 goto out; 2050 2051 n = neigh_lookup(&nd_tbl, &msg->target, dev); 2052 2053 if (n) { 2054 struct vxlan_rdst *rdst = NULL; 2055 struct vxlan_fdb *f; 2056 struct sk_buff *reply; 2057 2058 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2059 neigh_release(n); 2060 goto out; 2061 } 2062 2063 f = vxlan_find_mac_tx(vxlan, n->ha, vni); 2064 if (f) 2065 rdst = first_remote_rcu(f); 2066 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 2067 /* bridge-local neighbor */ 2068 neigh_release(n); 2069 goto out; 2070 } 2071 2072 reply = vxlan_na_create(skb, n, 2073 !!(f ? f->flags & NTF_ROUTER : 0)); 2074 2075 neigh_release(n); 2076 2077 if (reply == NULL) 2078 goto out; 2079 2080 if (netif_rx(reply) == NET_RX_DROP) { 2081 dev_dstats_rx_dropped(dev); 2082 vxlan_vnifilter_count(vxlan, vni, NULL, 2083 VXLAN_VNI_STATS_RX_DROPS, 0); 2084 } 2085 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2086 union vxlan_addr ipa = { 2087 .sin6.sin6_addr = msg->target, 2088 .sin6.sin6_family = AF_INET6, 2089 }; 2090 2091 vxlan_ip_miss(dev, &ipa); 2092 } 2093 2094 out: 2095 rcu_read_unlock(); 2096 consume_skb(skb); 2097 return NETDEV_TX_OK; 2098 } 2099 #endif 2100 2101 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2102 { 2103 struct vxlan_dev *vxlan = netdev_priv(dev); 2104 struct neighbour *n; 2105 2106 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2107 return false; 2108 2109 n = NULL; 2110 switch (ntohs(eth_hdr(skb)->h_proto)) { 2111 case ETH_P_IP: 2112 { 2113 struct iphdr *pip; 2114 2115 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2116 return false; 2117 pip = ip_hdr(skb); 2118 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2119 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2120 union vxlan_addr ipa = { 2121 .sin.sin_addr.s_addr = pip->daddr, 2122 .sin.sin_family = AF_INET, 2123 }; 2124 2125 vxlan_ip_miss(dev, &ipa); 2126 return false; 2127 } 2128 2129 break; 2130 } 2131 #if IS_ENABLED(CONFIG_IPV6) 2132 case ETH_P_IPV6: 2133 { 2134 struct ipv6hdr *pip6; 2135 2136 /* check if ipv6.disable=1 set during boot was set 2137 * during booting so nd_tbl is not initialized 2138 */ 2139 if (!ipv6_mod_enabled()) 2140 return false; 2141 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2142 return false; 2143 pip6 = ipv6_hdr(skb); 2144 n = neigh_lookup(&nd_tbl, &pip6->daddr, dev); 2145 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2146 union vxlan_addr ipa = { 2147 .sin6.sin6_addr = pip6->daddr, 2148 .sin6.sin6_family = AF_INET6, 2149 }; 2150 2151 vxlan_ip_miss(dev, &ipa); 2152 return false; 2153 } 2154 2155 break; 2156 } 2157 #endif 2158 default: 2159 return false; 2160 } 2161 2162 if (n) { 2163 bool diff; 2164 2165 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2166 if (diff) { 2167 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2168 dev->addr_len); 2169 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2170 } 2171 neigh_release(n); 2172 return diff; 2173 } 2174 2175 return false; 2176 } 2177 2178 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2179 { 2180 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2181 2182 gpe->np_applied = 1; 2183 gpe->next_protocol = tun_p_from_eth_p(protocol); 2184 if (!gpe->next_protocol) 2185 return -EPFNOSUPPORT; 2186 return 0; 2187 } 2188 2189 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2190 int iphdr_len, __be32 vni, 2191 struct vxlan_metadata *md, u32 vxflags, 2192 bool udp_sum) 2193 { 2194 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2195 __be16 inner_protocol = htons(ETH_P_TEB); 2196 struct vxlanhdr *vxh; 2197 bool double_encap; 2198 int min_headroom; 2199 int err; 2200 2201 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2202 skb->ip_summed == CHECKSUM_PARTIAL) { 2203 int csum_start = skb_checksum_start_offset(skb); 2204 2205 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2206 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2207 (skb->csum_offset == offsetof(struct udphdr, check) || 2208 skb->csum_offset == offsetof(struct tcphdr, check))) 2209 type |= SKB_GSO_TUNNEL_REMCSUM; 2210 } 2211 2212 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2213 + VXLAN_HLEN + iphdr_len; 2214 2215 /* Need space for new headers (invalidates iph ptr) */ 2216 err = skb_cow_head(skb, min_headroom); 2217 if (unlikely(err)) 2218 return err; 2219 2220 double_encap = udp_tunnel_handle_partial(skb); 2221 err = iptunnel_handle_offloads(skb, type); 2222 if (err) 2223 return err; 2224 2225 vxh = __skb_push(skb, sizeof(*vxh)); 2226 vxh->vx_flags = VXLAN_HF_VNI; 2227 vxh->vx_vni = vxlan_vni_field(vni); 2228 2229 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2230 unsigned int start; 2231 2232 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2233 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2234 vxh->vx_flags |= VXLAN_HF_RCO; 2235 2236 if (!skb_is_gso(skb)) { 2237 skb->ip_summed = CHECKSUM_NONE; 2238 skb->encapsulation = 0; 2239 } 2240 } 2241 2242 if (vxflags & VXLAN_F_GBP) 2243 vxlan_build_gbp_hdr(vxh, md); 2244 if (vxflags & VXLAN_F_GPE) { 2245 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2246 if (err < 0) 2247 return err; 2248 inner_protocol = skb->protocol; 2249 } 2250 2251 udp_tunnel_set_inner_protocol(skb, double_encap, inner_protocol); 2252 return 0; 2253 } 2254 2255 /* Bypass encapsulation if the destination is local */ 2256 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2257 struct vxlan_dev *dst_vxlan, __be32 vni, 2258 bool snoop) 2259 { 2260 union vxlan_addr loopback; 2261 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2262 unsigned int len = skb->len; 2263 struct net_device *dev; 2264 2265 skb->pkt_type = PACKET_HOST; 2266 skb->encapsulation = 0; 2267 skb->dev = dst_vxlan->dev; 2268 __skb_pull(skb, skb_network_offset(skb)); 2269 2270 if (remote_ip->sa.sa_family == AF_INET) { 2271 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2272 loopback.sa.sa_family = AF_INET; 2273 #if IS_ENABLED(CONFIG_IPV6) 2274 } else { 2275 loopback.sin6.sin6_addr = in6addr_loopback; 2276 loopback.sa.sa_family = AF_INET6; 2277 #endif 2278 } 2279 2280 rcu_read_lock(); 2281 dev = skb->dev; 2282 if (unlikely(!(dev->flags & IFF_UP))) { 2283 kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY); 2284 goto drop; 2285 } 2286 2287 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2288 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2289 2290 dev_dstats_tx_add(src_vxlan->dev, len); 2291 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2292 2293 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2294 dev_dstats_rx_add(dst_vxlan->dev, len); 2295 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2296 len); 2297 } else { 2298 drop: 2299 dev_dstats_rx_dropped(dev); 2300 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2301 VXLAN_VNI_STATS_RX_DROPS, 0); 2302 } 2303 rcu_read_unlock(); 2304 } 2305 2306 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2307 struct vxlan_dev *vxlan, 2308 int addr_family, 2309 __be16 dst_port, int dst_ifindex, __be32 vni, 2310 struct dst_entry *dst, 2311 u32 rt_flags) 2312 { 2313 #if IS_ENABLED(CONFIG_IPV6) 2314 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2315 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2316 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2317 */ 2318 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2319 #endif 2320 /* Bypass encapsulation if the destination is local */ 2321 if (rt_flags & RTCF_LOCAL && 2322 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2323 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2324 struct vxlan_dev *dst_vxlan; 2325 2326 dst_release(dst); 2327 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2328 addr_family, dst_port, 2329 vxlan->cfg.flags); 2330 if (!dst_vxlan) { 2331 DEV_STATS_INC(dev, tx_errors); 2332 vxlan_vnifilter_count(vxlan, vni, NULL, 2333 VXLAN_VNI_STATS_TX_ERRORS, 0); 2334 kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND); 2335 2336 return -ENOENT; 2337 } 2338 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2339 return 1; 2340 } 2341 2342 return 0; 2343 } 2344 2345 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2346 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2347 { 2348 struct dst_cache *dst_cache; 2349 struct ip_tunnel_info *info; 2350 struct ip_tunnel_key *pkey; 2351 struct ip_tunnel_key key; 2352 struct vxlan_dev *vxlan = netdev_priv(dev); 2353 const struct iphdr *old_iph; 2354 struct vxlan_metadata _md; 2355 struct vxlan_metadata *md = &_md; 2356 unsigned int pkt_len = skb->len; 2357 __be16 src_port = 0, dst_port; 2358 struct dst_entry *ndst = NULL; 2359 int addr_family; 2360 __u8 tos, ttl; 2361 int ifindex; 2362 int err = 0; 2363 u32 flags = vxlan->cfg.flags; 2364 bool use_cache; 2365 bool udp_sum = false; 2366 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2367 enum skb_drop_reason reason; 2368 bool no_eth_encap; 2369 __be32 vni = 0; 2370 2371 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2372 reason = skb_vlan_inet_prepare(skb, no_eth_encap); 2373 if (reason) 2374 goto drop; 2375 2376 reason = SKB_DROP_REASON_NOT_SPECIFIED; 2377 old_iph = ip_hdr(skb); 2378 2379 info = skb_tunnel_info(skb); 2380 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2381 2382 if (rdst) { 2383 memset(&key, 0, sizeof(key)); 2384 pkey = &key; 2385 2386 if (vxlan_addr_any(&rdst->remote_ip)) { 2387 if (did_rsc) { 2388 /* short-circuited back to local bridge */ 2389 vxlan_encap_bypass(skb, vxlan, vxlan, 2390 default_vni, true); 2391 return; 2392 } 2393 goto drop; 2394 } 2395 2396 addr_family = vxlan->cfg.saddr.sa.sa_family; 2397 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2398 vni = (rdst->remote_vni) ? : default_vni; 2399 ifindex = rdst->remote_ifindex; 2400 2401 if (addr_family == AF_INET) { 2402 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2403 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2404 } else { 2405 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2406 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2407 } 2408 2409 dst_cache = &rdst->dst_cache; 2410 md->gbp = skb->mark; 2411 if (flags & VXLAN_F_TTL_INHERIT) { 2412 ttl = ip_tunnel_get_ttl(old_iph, skb); 2413 } else { 2414 ttl = vxlan->cfg.ttl; 2415 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2416 ttl = 1; 2417 } 2418 tos = vxlan->cfg.tos; 2419 if (tos == 1) 2420 tos = ip_tunnel_get_dsfield(old_iph, skb); 2421 if (tos && !info) 2422 use_cache = false; 2423 2424 if (addr_family == AF_INET) 2425 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2426 else 2427 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2428 #if IS_ENABLED(CONFIG_IPV6) 2429 switch (vxlan->cfg.label_policy) { 2430 case VXLAN_LABEL_FIXED: 2431 key.label = vxlan->cfg.label; 2432 break; 2433 case VXLAN_LABEL_INHERIT: 2434 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2435 break; 2436 default: 2437 DEBUG_NET_WARN_ON_ONCE(1); 2438 goto drop; 2439 } 2440 #endif 2441 } else { 2442 if (!info) { 2443 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2444 dev->name); 2445 goto drop; 2446 } 2447 pkey = &info->key; 2448 addr_family = ip_tunnel_info_af(info); 2449 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2450 vni = tunnel_id_to_key32(info->key.tun_id); 2451 ifindex = 0; 2452 dst_cache = &info->dst_cache; 2453 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2454 if (info->options_len < sizeof(*md)) 2455 goto drop; 2456 md = ip_tunnel_info_opts(info); 2457 } 2458 ttl = info->key.ttl; 2459 tos = info->key.tos; 2460 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2461 } 2462 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2463 vxlan->cfg.port_max, true); 2464 2465 rcu_read_lock(); 2466 if (addr_family == AF_INET) { 2467 struct vxlan_sock *sock4; 2468 u16 ipcb_flags = 0; 2469 struct rtable *rt; 2470 __be16 df = 0; 2471 __be32 saddr; 2472 2473 sock4 = rcu_dereference(vxlan->vn4_sock); 2474 if (unlikely(!sock4)) { 2475 reason = SKB_DROP_REASON_DEV_READY; 2476 goto tx_error; 2477 } 2478 2479 if (!ifindex) 2480 ifindex = sock4->sk->sk_bound_dev_if; 2481 2482 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2483 &saddr, pkey, src_port, dst_port, 2484 tos, use_cache ? dst_cache : NULL); 2485 if (IS_ERR(rt)) { 2486 err = PTR_ERR(rt); 2487 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2488 goto tx_error; 2489 } 2490 2491 if (flags & VXLAN_F_MC_ROUTE) 2492 ipcb_flags |= IPSKB_MCROUTE; 2493 2494 if (!info) { 2495 /* Bypass encapsulation if the destination is local */ 2496 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2497 dst_port, ifindex, vni, 2498 &rt->dst, rt->rt_flags); 2499 if (err) 2500 goto out_unlock; 2501 2502 if (vxlan->cfg.df == VXLAN_DF_SET) { 2503 df = htons(IP_DF); 2504 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2505 struct ethhdr *eth = eth_hdr(skb); 2506 2507 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2508 (ntohs(eth->h_proto) == ETH_P_IP && 2509 old_iph->frag_off & htons(IP_DF))) 2510 df = htons(IP_DF); 2511 } 2512 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2513 info->key.tun_flags)) { 2514 df = htons(IP_DF); 2515 } 2516 2517 ndst = &rt->dst; 2518 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2519 netif_is_any_bridge_port(dev)); 2520 if (err < 0) { 2521 goto tx_error; 2522 } else if (err) { 2523 if (info) { 2524 struct ip_tunnel_info *unclone; 2525 2526 unclone = skb_tunnel_info_unclone(skb); 2527 if (unlikely(!unclone)) 2528 goto tx_error; 2529 2530 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2531 unclone->key.u.ipv4.dst = saddr; 2532 } 2533 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2534 dst_release(ndst); 2535 goto out_unlock; 2536 } 2537 2538 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2539 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2540 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2541 vni, md, flags, udp_sum); 2542 if (err < 0) { 2543 reason = SKB_DROP_REASON_NOMEM; 2544 goto tx_error; 2545 } 2546 2547 udp_tunnel_xmit_skb(rt, sock4->sk, skb, saddr, 2548 pkey->u.ipv4.dst, tos, ttl, df, 2549 src_port, dst_port, xnet, !udp_sum, 2550 ipcb_flags); 2551 #if IS_ENABLED(CONFIG_IPV6) 2552 } else { 2553 struct vxlan_sock *sock6; 2554 struct in6_addr saddr; 2555 u16 ip6cb_flags = 0; 2556 2557 sock6 = rcu_dereference(vxlan->vn6_sock); 2558 if (unlikely(!sock6)) { 2559 reason = SKB_DROP_REASON_DEV_READY; 2560 goto tx_error; 2561 } 2562 2563 if (!ifindex) 2564 ifindex = sock6->sk->sk_bound_dev_if; 2565 2566 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sk, 2567 ifindex, &saddr, pkey, 2568 src_port, dst_port, tos, 2569 use_cache ? dst_cache : NULL); 2570 if (IS_ERR(ndst)) { 2571 err = PTR_ERR(ndst); 2572 ndst = NULL; 2573 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2574 goto tx_error; 2575 } 2576 2577 if (flags & VXLAN_F_MC_ROUTE) 2578 ip6cb_flags |= IP6SKB_MCROUTE; 2579 2580 if (!info) { 2581 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2582 2583 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2584 dst_port, ifindex, vni, 2585 ndst, rt6i_flags); 2586 if (err) 2587 goto out_unlock; 2588 } 2589 2590 err = skb_tunnel_check_pmtu(skb, ndst, 2591 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2592 netif_is_any_bridge_port(dev)); 2593 if (err < 0) { 2594 goto tx_error; 2595 } else if (err) { 2596 if (info) { 2597 struct ip_tunnel_info *unclone; 2598 2599 unclone = skb_tunnel_info_unclone(skb); 2600 if (unlikely(!unclone)) 2601 goto tx_error; 2602 2603 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2604 unclone->key.u.ipv6.dst = saddr; 2605 } 2606 2607 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2608 dst_release(ndst); 2609 goto out_unlock; 2610 } 2611 2612 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2613 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2614 skb_scrub_packet(skb, xnet); 2615 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2616 vni, md, flags, udp_sum); 2617 if (err < 0) { 2618 reason = SKB_DROP_REASON_NOMEM; 2619 goto tx_error; 2620 } 2621 2622 udp_tunnel6_xmit_skb(ndst, sock6->sk, skb, dev, 2623 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2624 pkey->label, src_port, dst_port, !udp_sum, 2625 ip6cb_flags); 2626 #endif 2627 } 2628 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2629 out_unlock: 2630 rcu_read_unlock(); 2631 return; 2632 2633 drop: 2634 dev_dstats_tx_dropped(dev); 2635 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2636 kfree_skb_reason(skb, reason); 2637 return; 2638 2639 tx_error: 2640 rcu_read_unlock(); 2641 if (err == -ELOOP) 2642 DEV_STATS_INC(dev, collisions); 2643 else if (err == -ENETUNREACH) 2644 DEV_STATS_INC(dev, tx_carrier_errors); 2645 dst_release(ndst); 2646 DEV_STATS_INC(dev, tx_errors); 2647 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2648 kfree_skb_reason(skb, reason); 2649 } 2650 2651 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2652 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2653 { 2654 struct vxlan_rdst nh_rdst; 2655 struct nexthop *nh; 2656 bool do_xmit; 2657 u32 hash; 2658 2659 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2660 hash = skb_get_hash(skb); 2661 2662 nh = rcu_dereference(f->nh); 2663 if (!nh) 2664 goto drop; 2665 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2666 2667 if (likely(do_xmit)) 2668 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2669 else 2670 goto drop; 2671 2672 return; 2673 2674 drop: 2675 dev_dstats_tx_dropped(dev); 2676 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2677 VXLAN_VNI_STATS_TX_DROPS, 0); 2678 dev_kfree_skb(skb); 2679 } 2680 2681 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2682 u32 nhid, __be32 vni) 2683 { 2684 struct vxlan_dev *vxlan = netdev_priv(dev); 2685 struct vxlan_rdst nh_rdst; 2686 struct nexthop *nh; 2687 bool do_xmit; 2688 u32 hash; 2689 2690 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2691 hash = skb_get_hash(skb); 2692 2693 rcu_read_lock(); 2694 nh = nexthop_find_by_id(dev_net(dev), nhid); 2695 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2696 rcu_read_unlock(); 2697 goto drop; 2698 } 2699 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2700 rcu_read_unlock(); 2701 2702 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2703 goto drop; 2704 2705 if (likely(do_xmit)) 2706 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2707 else 2708 goto drop; 2709 2710 return NETDEV_TX_OK; 2711 2712 drop: 2713 dev_dstats_tx_dropped(dev); 2714 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2715 VXLAN_VNI_STATS_TX_DROPS, 0); 2716 dev_kfree_skb(skb); 2717 return NETDEV_TX_OK; 2718 } 2719 2720 /* Transmit local packets over Vxlan 2721 * 2722 * Outer IP header inherits ECN and DF from inner header. 2723 * Outer UDP destination is the VXLAN assigned port. 2724 * source port is based on hash of flow 2725 */ 2726 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2727 { 2728 struct vxlan_dev *vxlan = netdev_priv(dev); 2729 struct vxlan_rdst *rdst, *fdst = NULL; 2730 const struct ip_tunnel_info *info; 2731 struct vxlan_fdb *f; 2732 struct ethhdr *eth; 2733 __be32 vni = 0; 2734 u32 nhid = 0; 2735 bool did_rsc; 2736 2737 info = skb_tunnel_info(skb); 2738 2739 skb_reset_mac_header(skb); 2740 2741 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2742 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2743 info->mode & IP_TUNNEL_INFO_TX) { 2744 vni = tunnel_id_to_key32(info->key.tun_id); 2745 nhid = info->key.nhid; 2746 } else { 2747 if (info && info->mode & IP_TUNNEL_INFO_TX) 2748 vxlan_xmit_one(skb, dev, vni, NULL, false); 2749 else 2750 kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO); 2751 return NETDEV_TX_OK; 2752 } 2753 } 2754 2755 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2756 eth = eth_hdr(skb); 2757 if (ntohs(eth->h_proto) == ETH_P_ARP) 2758 return arp_reduce(dev, skb, vni); 2759 #if IS_ENABLED(CONFIG_IPV6) 2760 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2761 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2762 sizeof(struct nd_msg)) && 2763 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2764 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2765 2766 if (m->icmph.icmp6_code == 0 && 2767 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2768 return neigh_reduce(dev, skb, vni); 2769 } 2770 #endif 2771 } 2772 2773 if (nhid) 2774 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2775 2776 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2777 struct vxlan_mdb_entry *mdb_entry; 2778 2779 rcu_read_lock(); 2780 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2781 if (mdb_entry) { 2782 netdev_tx_t ret; 2783 2784 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2785 rcu_read_unlock(); 2786 return ret; 2787 } 2788 rcu_read_unlock(); 2789 } 2790 2791 eth = eth_hdr(skb); 2792 rcu_read_lock(); 2793 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2794 did_rsc = false; 2795 2796 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2797 (ntohs(eth->h_proto) == ETH_P_IP || 2798 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2799 did_rsc = route_shortcircuit(dev, skb); 2800 if (did_rsc) 2801 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2802 } 2803 2804 if (f == NULL) { 2805 f = vxlan_find_mac_tx(vxlan, all_zeros_mac, vni); 2806 if (f == NULL) { 2807 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2808 !is_multicast_ether_addr(eth->h_dest)) 2809 vxlan_fdb_miss(vxlan, eth->h_dest); 2810 2811 dev_dstats_tx_dropped(dev); 2812 vxlan_vnifilter_count(vxlan, vni, NULL, 2813 VXLAN_VNI_STATS_TX_DROPS, 0); 2814 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2815 goto out; 2816 } 2817 } 2818 2819 if (rcu_access_pointer(f->nh)) { 2820 vxlan_xmit_nh(skb, dev, f, 2821 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2822 } else { 2823 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2824 struct sk_buff *skb1; 2825 2826 if (!fdst) { 2827 fdst = rdst; 2828 continue; 2829 } 2830 skb1 = skb_clone(skb, GFP_ATOMIC); 2831 if (skb1) 2832 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2833 } 2834 if (fdst) 2835 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2836 else 2837 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2838 } 2839 2840 out: 2841 rcu_read_unlock(); 2842 return NETDEV_TX_OK; 2843 } 2844 2845 /* Walk the forwarding table and purge stale entries */ 2846 static void vxlan_cleanup(struct timer_list *t) 2847 { 2848 struct vxlan_dev *vxlan = timer_container_of(vxlan, t, age_timer); 2849 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2850 struct vxlan_fdb *f; 2851 2852 if (!netif_running(vxlan->dev)) 2853 return; 2854 2855 rcu_read_lock(); 2856 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 2857 unsigned long timeout; 2858 2859 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2860 continue; 2861 2862 if (f->flags & NTF_EXT_LEARNED) 2863 continue; 2864 2865 timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ; 2866 if (time_before_eq(timeout, jiffies)) { 2867 spin_lock(&vxlan->hash_lock); 2868 if (!hlist_unhashed(&f->fdb_node)) { 2869 netdev_dbg(vxlan->dev, "garbage collect %pM\n", 2870 f->key.eth_addr); 2871 f->state = NUD_STALE; 2872 vxlan_fdb_destroy(vxlan, f, true, true); 2873 } 2874 spin_unlock(&vxlan->hash_lock); 2875 } else if (time_before(timeout, next_timer)) { 2876 next_timer = timeout; 2877 } 2878 } 2879 rcu_read_unlock(); 2880 2881 mod_timer(&vxlan->age_timer, next_timer); 2882 } 2883 2884 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2885 { 2886 ASSERT_RTNL(); 2887 2888 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2889 #if IS_ENABLED(CONFIG_IPV6) 2890 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2891 #endif 2892 } 2893 2894 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2895 struct vxlan_dev_node *node) 2896 { 2897 __be32 vni = vxlan->default_dst.remote_vni; 2898 2899 ASSERT_RTNL(); 2900 2901 node->vxlan = vxlan; 2902 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2903 } 2904 2905 /* Setup stats when device is created */ 2906 static int vxlan_init(struct net_device *dev) 2907 { 2908 struct vxlan_dev *vxlan = netdev_priv(dev); 2909 int err; 2910 2911 err = rhashtable_init(&vxlan->fdb_hash_tbl, &vxlan_fdb_rht_params); 2912 if (err) 2913 return err; 2914 2915 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 2916 err = vxlan_vnigroup_init(vxlan); 2917 if (err) 2918 goto err_rhashtable_destroy; 2919 } 2920 2921 err = gro_cells_init(&vxlan->gro_cells, dev); 2922 if (err) 2923 goto err_vnigroup_uninit; 2924 2925 err = vxlan_mdb_init(vxlan); 2926 if (err) 2927 goto err_gro_cells_destroy; 2928 2929 netdev_lockdep_set_classes(dev); 2930 return 0; 2931 2932 err_gro_cells_destroy: 2933 gro_cells_destroy(&vxlan->gro_cells); 2934 err_vnigroup_uninit: 2935 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2936 vxlan_vnigroup_uninit(vxlan); 2937 err_rhashtable_destroy: 2938 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2939 return err; 2940 } 2941 2942 static void vxlan_uninit(struct net_device *dev) 2943 { 2944 struct vxlan_dev *vxlan = netdev_priv(dev); 2945 2946 vxlan_mdb_fini(vxlan); 2947 2948 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2949 vxlan_vnigroup_uninit(vxlan); 2950 2951 gro_cells_destroy(&vxlan->gro_cells); 2952 2953 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2954 } 2955 2956 /* Start ageing timer and join group when device is brought up */ 2957 static int vxlan_open(struct net_device *dev) 2958 { 2959 struct vxlan_dev *vxlan = netdev_priv(dev); 2960 int ret; 2961 2962 ret = vxlan_sock_add(vxlan); 2963 if (ret < 0) 2964 return ret; 2965 2966 ret = vxlan_multicast_join(vxlan); 2967 if (ret) { 2968 vxlan_sock_release(vxlan); 2969 return ret; 2970 } 2971 2972 if (vxlan->cfg.age_interval) 2973 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2974 2975 return ret; 2976 } 2977 2978 struct vxlan_fdb_flush_desc { 2979 bool ignore_default_entry; 2980 unsigned long state; 2981 unsigned long state_mask; 2982 unsigned long flags; 2983 unsigned long flags_mask; 2984 __be32 src_vni; 2985 u32 nhid; 2986 __be32 vni; 2987 __be16 port; 2988 union vxlan_addr dst_ip; 2989 }; 2990 2991 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2992 const struct vxlan_dev *vxlan) 2993 { 2994 return is_zero_ether_addr(f->key.eth_addr) && 2995 f->key.vni == vxlan->cfg.vni; 2996 } 2997 2998 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2999 { 3000 struct nexthop *nh = rtnl_dereference(f->nh); 3001 3002 return nh && nh->id == nhid; 3003 } 3004 3005 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 3006 const struct vxlan_dev *vxlan, 3007 const struct vxlan_fdb_flush_desc *desc) 3008 { 3009 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 3010 return false; 3011 3012 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 3013 return false; 3014 3015 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 3016 return false; 3017 3018 if (desc->src_vni && f->key.vni != desc->src_vni) 3019 return false; 3020 3021 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 3022 return false; 3023 3024 return true; 3025 } 3026 3027 static bool 3028 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 3029 { 3030 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 3031 } 3032 3033 static bool 3034 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3035 const struct vxlan_rdst *rd) 3036 { 3037 if (desc->vni && rd->remote_vni != desc->vni) 3038 return false; 3039 3040 if (desc->port && rd->remote_port != desc->port) 3041 return false; 3042 3043 if (desc->dst_ip.sa.sa_family && 3044 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3045 return false; 3046 3047 return true; 3048 } 3049 3050 static void 3051 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3052 const struct vxlan_fdb_flush_desc *desc, 3053 bool *p_destroy_fdb) 3054 { 3055 bool remotes_flushed = false; 3056 struct vxlan_rdst *rd, *tmp; 3057 3058 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3059 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3060 continue; 3061 3062 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3063 remotes_flushed = true; 3064 } 3065 3066 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3067 } 3068 3069 /* Purge the forwarding table */ 3070 static void vxlan_flush(struct vxlan_dev *vxlan, 3071 const struct vxlan_fdb_flush_desc *desc) 3072 { 3073 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3074 struct vxlan_fdb *f; 3075 3076 rcu_read_lock(); 3077 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 3078 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3079 continue; 3080 3081 spin_lock_bh(&vxlan->hash_lock); 3082 if (hlist_unhashed(&f->fdb_node)) 3083 goto unlock; 3084 3085 if (match_remotes) { 3086 bool destroy_fdb = false; 3087 3088 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3089 &destroy_fdb); 3090 3091 if (!destroy_fdb) 3092 goto unlock; 3093 } 3094 3095 vxlan_fdb_destroy(vxlan, f, true, true); 3096 unlock: 3097 spin_unlock_bh(&vxlan->hash_lock); 3098 } 3099 rcu_read_unlock(); 3100 } 3101 3102 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3103 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3104 [NDA_NH_ID] = { .type = NLA_U32 }, 3105 [NDA_VNI] = { .type = NLA_U32 }, 3106 [NDA_PORT] = { .type = NLA_U16 }, 3107 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3108 sizeof(struct in6_addr)), 3109 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3110 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3111 }; 3112 3113 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3114 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3115 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3116 NTF_ROUTER) 3117 3118 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3119 struct netlink_ext_ack *extack) 3120 { 3121 struct vxlan_dev *vxlan = netdev_priv(dev); 3122 struct vxlan_fdb_flush_desc desc = {}; 3123 struct ndmsg *ndm = nlmsg_data(nlh); 3124 struct nlattr *tb[NDA_MAX + 1]; 3125 u8 ndm_flags; 3126 int err; 3127 3128 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3129 3130 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3131 extack); 3132 if (err) 3133 return err; 3134 3135 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3136 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3137 return -EINVAL; 3138 } 3139 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3140 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3141 return -EINVAL; 3142 } 3143 3144 desc.state = ndm->ndm_state; 3145 desc.flags = ndm_flags; 3146 3147 if (tb[NDA_NDM_STATE_MASK]) 3148 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3149 3150 if (tb[NDA_NDM_FLAGS_MASK]) 3151 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3152 3153 if (tb[NDA_SRC_VNI]) 3154 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3155 3156 if (tb[NDA_NH_ID]) 3157 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3158 3159 if (tb[NDA_VNI]) 3160 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3161 3162 if (tb[NDA_PORT]) 3163 desc.port = nla_get_be16(tb[NDA_PORT]); 3164 3165 if (tb[NDA_DST]) { 3166 union vxlan_addr ip; 3167 3168 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3169 if (err) { 3170 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3171 "Unsupported address family"); 3172 return err; 3173 } 3174 desc.dst_ip = ip; 3175 } 3176 3177 vxlan_flush(vxlan, &desc); 3178 3179 return 0; 3180 } 3181 3182 /* Cleanup timer and forwarding table on shutdown */ 3183 static int vxlan_stop(struct net_device *dev) 3184 { 3185 struct vxlan_dev *vxlan = netdev_priv(dev); 3186 struct vxlan_fdb_flush_desc desc = { 3187 /* Default entry is deleted at vxlan_dellink. */ 3188 .ignore_default_entry = true, 3189 .state = 0, 3190 .state_mask = NUD_PERMANENT | NUD_NOARP, 3191 }; 3192 3193 vxlan_multicast_leave(vxlan); 3194 3195 timer_delete_sync(&vxlan->age_timer); 3196 3197 vxlan_flush(vxlan, &desc); 3198 vxlan_sock_release(vxlan); 3199 3200 return 0; 3201 } 3202 3203 /* Stub, nothing needs to be done. */ 3204 static void vxlan_set_multicast_list(struct net_device *dev) 3205 { 3206 } 3207 3208 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3209 { 3210 struct vxlan_dev *vxlan = netdev_priv(dev); 3211 struct vxlan_rdst *dst = &vxlan->default_dst; 3212 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3213 dst->remote_ifindex); 3214 3215 /* This check is different than dev->max_mtu, because it looks at 3216 * the lowerdev->mtu, rather than the static dev->max_mtu 3217 */ 3218 if (lowerdev) { 3219 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3220 if (new_mtu > max_mtu) 3221 return -EINVAL; 3222 } 3223 3224 WRITE_ONCE(dev->mtu, new_mtu); 3225 return 0; 3226 } 3227 3228 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3229 { 3230 struct vxlan_dev *vxlan = netdev_priv(dev); 3231 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3232 __be16 sport, dport; 3233 3234 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3235 vxlan->cfg.port_max, true); 3236 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3237 3238 if (ip_tunnel_info_af(info) == AF_INET) { 3239 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3240 struct rtable *rt; 3241 3242 if (!sock4) 3243 return -EIO; 3244 3245 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3246 &info->key.u.ipv4.src, 3247 &info->key, 3248 sport, dport, info->key.tos, 3249 &info->dst_cache); 3250 if (IS_ERR(rt)) 3251 return PTR_ERR(rt); 3252 ip_rt_put(rt); 3253 } else { 3254 #if IS_ENABLED(CONFIG_IPV6) 3255 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3256 struct dst_entry *ndst; 3257 3258 if (!sock6) 3259 return -EIO; 3260 3261 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sk, 3262 0, &info->key.u.ipv6.src, 3263 &info->key, 3264 sport, dport, info->key.tos, 3265 &info->dst_cache); 3266 if (IS_ERR(ndst)) 3267 return PTR_ERR(ndst); 3268 dst_release(ndst); 3269 #else /* !CONFIG_IPV6 */ 3270 return -EPFNOSUPPORT; 3271 #endif 3272 } 3273 info->key.tp_src = sport; 3274 info->key.tp_dst = dport; 3275 return 0; 3276 } 3277 3278 static const struct net_device_ops vxlan_netdev_ether_ops = { 3279 .ndo_init = vxlan_init, 3280 .ndo_uninit = vxlan_uninit, 3281 .ndo_open = vxlan_open, 3282 .ndo_stop = vxlan_stop, 3283 .ndo_start_xmit = vxlan_xmit, 3284 .ndo_set_rx_mode = vxlan_set_multicast_list, 3285 .ndo_change_mtu = vxlan_change_mtu, 3286 .ndo_validate_addr = eth_validate_addr, 3287 .ndo_set_mac_address = eth_mac_addr, 3288 .ndo_fdb_add = vxlan_fdb_add, 3289 .ndo_fdb_del = vxlan_fdb_delete, 3290 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3291 .ndo_fdb_dump = vxlan_fdb_dump, 3292 .ndo_fdb_get = vxlan_fdb_get, 3293 .ndo_mdb_add = vxlan_mdb_add, 3294 .ndo_mdb_del = vxlan_mdb_del, 3295 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3296 .ndo_mdb_dump = vxlan_mdb_dump, 3297 .ndo_mdb_get = vxlan_mdb_get, 3298 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3299 }; 3300 3301 static const struct net_device_ops vxlan_netdev_raw_ops = { 3302 .ndo_init = vxlan_init, 3303 .ndo_uninit = vxlan_uninit, 3304 .ndo_open = vxlan_open, 3305 .ndo_stop = vxlan_stop, 3306 .ndo_start_xmit = vxlan_xmit, 3307 .ndo_change_mtu = vxlan_change_mtu, 3308 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3309 }; 3310 3311 /* Info for udev, that this is a virtual tunnel endpoint */ 3312 static const struct device_type vxlan_type = { 3313 .name = "vxlan", 3314 }; 3315 3316 /* Calls the ndo_udp_tunnel_add of the caller in order to 3317 * supply the listening VXLAN udp ports. Callers are expected 3318 * to implement the ndo_udp_tunnel_add. 3319 */ 3320 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3321 { 3322 struct vxlan_sock *vs; 3323 struct net *net = dev_net(dev); 3324 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3325 unsigned int i; 3326 3327 ASSERT_RTNL(); 3328 3329 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3330 hlist_for_each_entry(vs, &vn->sock_list[i], hlist) { 3331 unsigned short type; 3332 3333 if (vs->flags & VXLAN_F_GPE) 3334 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3335 else 3336 type = UDP_TUNNEL_TYPE_VXLAN; 3337 3338 if (push) 3339 udp_tunnel_push_rx_port(dev, vs->sk, type); 3340 else 3341 udp_tunnel_drop_rx_port(dev, vs->sk, type); 3342 } 3343 } 3344 } 3345 3346 /* Initialize the device structure. */ 3347 static void vxlan_setup(struct net_device *dev) 3348 { 3349 struct vxlan_dev *vxlan = netdev_priv(dev); 3350 3351 eth_hw_addr_random(dev); 3352 ether_setup(dev); 3353 3354 dev->needs_free_netdev = true; 3355 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3356 3357 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3358 dev->features |= NETIF_F_RXCSUM; 3359 dev->features |= NETIF_F_GSO_SOFTWARE; 3360 3361 /* Partial features are disabled by default. */ 3362 dev->vlan_features = dev->features; 3363 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3364 dev->hw_features |= NETIF_F_RXCSUM; 3365 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3366 dev->hw_features |= UDP_TUNNEL_PARTIAL_FEATURES; 3367 dev->hw_features |= NETIF_F_GSO_PARTIAL; 3368 3369 dev->hw_enc_features = dev->hw_features; 3370 dev->gso_partial_features = UDP_TUNNEL_PARTIAL_FEATURES; 3371 dev->mangleid_features = NETIF_F_GSO_PARTIAL; 3372 3373 netif_keep_dst(dev); 3374 dev->priv_flags |= IFF_NO_QUEUE; 3375 dev->change_proto_down = true; 3376 dev->lltx = true; 3377 3378 /* MTU range: 68 - 65535 */ 3379 dev->min_mtu = ETH_MIN_MTU; 3380 dev->max_mtu = ETH_MAX_MTU; 3381 3382 dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS; 3383 INIT_LIST_HEAD(&vxlan->next); 3384 spin_lock_init(&vxlan->hash_lock); 3385 3386 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3387 3388 vxlan->dev = dev; 3389 3390 INIT_HLIST_HEAD(&vxlan->fdb_list); 3391 } 3392 3393 static void vxlan_ether_setup(struct net_device *dev) 3394 { 3395 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3396 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3397 dev->netdev_ops = &vxlan_netdev_ether_ops; 3398 } 3399 3400 static void vxlan_raw_setup(struct net_device *dev) 3401 { 3402 dev->header_ops = NULL; 3403 dev->type = ARPHRD_NONE; 3404 dev->hard_header_len = 0; 3405 dev->addr_len = 0; 3406 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3407 dev->netdev_ops = &vxlan_netdev_raw_ops; 3408 } 3409 3410 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3411 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3412 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3413 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3414 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3415 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3416 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3417 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3418 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3419 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3420 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3421 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3422 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3423 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3424 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3425 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3426 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3427 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3428 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3429 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3430 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3431 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3432 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3433 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3434 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3435 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3436 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3437 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3438 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3439 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3440 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3441 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3442 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3443 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3444 [IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)), 3445 [IFLA_VXLAN_MC_ROUTE] = NLA_POLICY_MAX(NLA_U8, 1), 3446 }; 3447 3448 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3449 struct netlink_ext_ack *extack) 3450 { 3451 if (tb[IFLA_ADDRESS]) { 3452 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3453 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3454 "Provided link layer address is not Ethernet"); 3455 return -EINVAL; 3456 } 3457 3458 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3459 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3460 "Provided Ethernet address is not unicast"); 3461 return -EADDRNOTAVAIL; 3462 } 3463 } 3464 3465 if (tb[IFLA_MTU]) { 3466 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3467 3468 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3469 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3470 "MTU must be between 68 and 65535"); 3471 return -EINVAL; 3472 } 3473 } 3474 3475 if (!data) { 3476 NL_SET_ERR_MSG(extack, 3477 "Required attributes not provided to perform the operation"); 3478 return -EINVAL; 3479 } 3480 3481 if (data[IFLA_VXLAN_ID]) { 3482 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3483 3484 if (id >= VXLAN_N_VID) { 3485 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3486 "VXLAN ID must be lower than 16777216"); 3487 return -ERANGE; 3488 } 3489 } 3490 3491 if (data[IFLA_VXLAN_PORT_RANGE]) { 3492 const struct ifla_vxlan_port_range *p 3493 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3494 3495 if (ntohs(p->high) < ntohs(p->low)) { 3496 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3497 "Invalid source port range"); 3498 return -EINVAL; 3499 } 3500 } 3501 3502 if (data[IFLA_VXLAN_DF]) { 3503 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3504 3505 if (df < 0 || df > VXLAN_DF_MAX) { 3506 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3507 "Invalid DF attribute"); 3508 return -EINVAL; 3509 } 3510 } 3511 3512 return 0; 3513 } 3514 3515 static void vxlan_get_drvinfo(struct net_device *netdev, 3516 struct ethtool_drvinfo *drvinfo) 3517 { 3518 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3519 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3520 } 3521 3522 static int vxlan_get_link_ksettings(struct net_device *dev, 3523 struct ethtool_link_ksettings *cmd) 3524 { 3525 struct vxlan_dev *vxlan = netdev_priv(dev); 3526 struct vxlan_rdst *dst = &vxlan->default_dst; 3527 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3528 dst->remote_ifindex); 3529 3530 if (!lowerdev) { 3531 cmd->base.duplex = DUPLEX_UNKNOWN; 3532 cmd->base.port = PORT_OTHER; 3533 cmd->base.speed = SPEED_UNKNOWN; 3534 3535 return 0; 3536 } 3537 3538 return __ethtool_get_link_ksettings(lowerdev, cmd); 3539 } 3540 3541 static const struct ethtool_ops vxlan_ethtool_ops = { 3542 .get_drvinfo = vxlan_get_drvinfo, 3543 .get_link = ethtool_op_get_link, 3544 .get_link_ksettings = vxlan_get_link_ksettings, 3545 }; 3546 3547 static struct sock *vxlan_create_sock(struct net *net, bool ipv6, 3548 __be16 port, u32 flags, int ifindex) 3549 { 3550 struct socket *sock; 3551 struct udp_port_cfg udp_conf; 3552 int err; 3553 3554 memset(&udp_conf, 0, sizeof(udp_conf)); 3555 3556 if (ipv6) { 3557 udp_conf.family = AF_INET6; 3558 udp_conf.use_udp6_rx_checksums = 3559 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3560 udp_conf.ipv6_v6only = 1; 3561 } else { 3562 udp_conf.family = AF_INET; 3563 } 3564 3565 udp_conf.local_udp_port = port; 3566 udp_conf.bind_ifindex = ifindex; 3567 3568 /* Open UDP socket */ 3569 err = udp_sock_create(net, &udp_conf, &sock); 3570 if (err < 0) 3571 return ERR_PTR(err); 3572 3573 udp_allow_gso(sock->sk); 3574 return sock->sk; 3575 } 3576 3577 /* Create new listen socket if needed */ 3578 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3579 __be16 port, u32 flags, 3580 int ifindex) 3581 { 3582 struct udp_tunnel_sock_cfg tunnel_cfg; 3583 struct vxlan_sock *vs; 3584 struct sock *sk; 3585 unsigned int h; 3586 3587 ASSERT_RTNL(); 3588 3589 vs = kzalloc_obj(*vs); 3590 if (!vs) 3591 return ERR_PTR(-ENOMEM); 3592 3593 for (h = 0; h < VNI_HASH_SIZE; ++h) 3594 INIT_HLIST_HEAD(&vs->vni_list[h]); 3595 3596 sk = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3597 if (IS_ERR(sk)) { 3598 kfree(vs); 3599 return ERR_CAST(sk); 3600 } 3601 3602 vs->sk = sk; 3603 refcount_set(&vs->refcnt, 1); 3604 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3605 3606 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3607 udp_tunnel_notify_add_rx_port(sk, 3608 (vs->flags & VXLAN_F_GPE) ? 3609 UDP_TUNNEL_TYPE_VXLAN_GPE : 3610 UDP_TUNNEL_TYPE_VXLAN); 3611 3612 /* Mark socket as an encapsulation socket. */ 3613 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3614 tunnel_cfg.sk_user_data = vs; 3615 tunnel_cfg.encap_type = 1; 3616 tunnel_cfg.encap_rcv = vxlan_rcv; 3617 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3618 tunnel_cfg.encap_destroy = NULL; 3619 if (vs->flags & VXLAN_F_GPE) { 3620 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3621 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3622 } else { 3623 tunnel_cfg.gro_receive = vxlan_gro_receive; 3624 tunnel_cfg.gro_complete = vxlan_gro_complete; 3625 } 3626 3627 setup_udp_tunnel_sock(net, sk, &tunnel_cfg); 3628 3629 return vs; 3630 } 3631 3632 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3633 { 3634 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3635 struct vxlan_sock *vs = NULL; 3636 struct vxlan_dev_node *node; 3637 int l3mdev_index = 0; 3638 3639 ASSERT_RTNL(); 3640 3641 if (vxlan->cfg.remote_ifindex) 3642 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3643 vxlan->net, vxlan->cfg.remote_ifindex); 3644 3645 if (!vxlan->cfg.no_share) { 3646 rcu_read_lock(); 3647 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3648 vxlan->cfg.dst_port, vxlan->cfg.flags, 3649 l3mdev_index); 3650 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3651 rcu_read_unlock(); 3652 return -EBUSY; 3653 } 3654 rcu_read_unlock(); 3655 } 3656 if (!vs) 3657 vs = vxlan_socket_create(vxlan->net, ipv6, 3658 vxlan->cfg.dst_port, vxlan->cfg.flags, 3659 l3mdev_index); 3660 if (IS_ERR(vs)) 3661 return PTR_ERR(vs); 3662 #if IS_ENABLED(CONFIG_IPV6) 3663 if (ipv6) { 3664 rcu_assign_pointer(vxlan->vn6_sock, vs); 3665 node = &vxlan->hlist6; 3666 } else 3667 #endif 3668 { 3669 rcu_assign_pointer(vxlan->vn4_sock, vs); 3670 node = &vxlan->hlist4; 3671 } 3672 3673 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3674 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3675 else 3676 vxlan_vs_add_dev(vs, vxlan, node); 3677 3678 return 0; 3679 } 3680 3681 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3682 { 3683 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3684 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3685 bool ipv4 = !ipv6 || metadata; 3686 int ret = 0; 3687 3688 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3689 #if IS_ENABLED(CONFIG_IPV6) 3690 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3691 if (ipv6) { 3692 ret = __vxlan_sock_add(vxlan, true); 3693 if (ret < 0 && ret != -EAFNOSUPPORT) 3694 ipv4 = false; 3695 } 3696 #endif 3697 if (ipv4) 3698 ret = __vxlan_sock_add(vxlan, false); 3699 if (ret < 0) 3700 vxlan_sock_release(vxlan); 3701 return ret; 3702 } 3703 3704 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3705 struct vxlan_config *conf, __be32 vni) 3706 { 3707 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3708 struct vxlan_dev *tmp; 3709 3710 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3711 if (tmp == vxlan) 3712 continue; 3713 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3714 if (!vxlan_vnifilter_lookup(tmp, vni)) 3715 continue; 3716 } else if (tmp->cfg.vni != vni) { 3717 continue; 3718 } 3719 if (tmp->cfg.dst_port != conf->dst_port) 3720 continue; 3721 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3722 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3723 continue; 3724 3725 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3726 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3727 continue; 3728 3729 return -EEXIST; 3730 } 3731 3732 return 0; 3733 } 3734 3735 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3736 struct net_device **lower, 3737 struct vxlan_dev *old, 3738 struct netlink_ext_ack *extack) 3739 { 3740 bool use_ipv6 = false; 3741 3742 if (conf->flags & VXLAN_F_GPE) { 3743 /* For now, allow GPE only together with 3744 * COLLECT_METADATA. This can be relaxed later; in such 3745 * case, the other side of the PtP link will have to be 3746 * provided. 3747 */ 3748 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3749 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3750 NL_SET_ERR_MSG(extack, 3751 "VXLAN GPE does not support this combination of attributes"); 3752 return -EINVAL; 3753 } 3754 } 3755 3756 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3757 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3758 conf->remote_ip.sa.sa_family = AF_INET; 3759 conf->saddr.sa.sa_family = AF_INET; 3760 } else if (!conf->remote_ip.sa.sa_family) { 3761 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3762 } else if (!conf->saddr.sa.sa_family) { 3763 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3764 } 3765 3766 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3767 NL_SET_ERR_MSG(extack, 3768 "Local and remote address must be from the same family"); 3769 return -EINVAL; 3770 } 3771 3772 if (vxlan_addr_multicast(&conf->saddr)) { 3773 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3774 return -EINVAL; 3775 } 3776 3777 if (conf->saddr.sa.sa_family == AF_INET6) { 3778 if (!IS_ENABLED(CONFIG_IPV6)) { 3779 NL_SET_ERR_MSG(extack, 3780 "IPv6 support not enabled in the kernel"); 3781 return -EPFNOSUPPORT; 3782 } 3783 use_ipv6 = true; 3784 conf->flags |= VXLAN_F_IPV6; 3785 3786 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3787 int local_type = 3788 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3789 int remote_type = 3790 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3791 3792 if (local_type & IPV6_ADDR_LINKLOCAL) { 3793 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3794 (remote_type != IPV6_ADDR_ANY)) { 3795 NL_SET_ERR_MSG(extack, 3796 "Invalid combination of local and remote address scopes"); 3797 return -EINVAL; 3798 } 3799 3800 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3801 } else { 3802 if (remote_type == 3803 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3804 NL_SET_ERR_MSG(extack, 3805 "Invalid combination of local and remote address scopes"); 3806 return -EINVAL; 3807 } 3808 3809 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3810 } 3811 } 3812 } 3813 3814 if (conf->label && !use_ipv6) { 3815 NL_SET_ERR_MSG(extack, 3816 "Label attribute only applies to IPv6 VXLAN devices"); 3817 return -EINVAL; 3818 } 3819 3820 if (conf->label_policy && !use_ipv6) { 3821 NL_SET_ERR_MSG(extack, 3822 "Label policy only applies to IPv6 VXLAN devices"); 3823 return -EINVAL; 3824 } 3825 3826 if (conf->remote_ifindex) { 3827 struct net_device *lowerdev; 3828 3829 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3830 if (!lowerdev) { 3831 NL_SET_ERR_MSG(extack, 3832 "Invalid local interface, device not found"); 3833 return -ENODEV; 3834 } 3835 3836 #if IS_ENABLED(CONFIG_IPV6) 3837 if (use_ipv6) { 3838 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3839 3840 if (idev && idev->cnf.disable_ipv6) { 3841 NL_SET_ERR_MSG(extack, 3842 "IPv6 support disabled by administrator"); 3843 return -EPERM; 3844 } 3845 } 3846 #endif 3847 3848 *lower = lowerdev; 3849 } else { 3850 if (vxlan_addr_multicast(&conf->remote_ip)) { 3851 NL_SET_ERR_MSG(extack, 3852 "Local interface required for multicast remote destination"); 3853 3854 return -EINVAL; 3855 } 3856 3857 #if IS_ENABLED(CONFIG_IPV6) 3858 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3859 NL_SET_ERR_MSG(extack, 3860 "Local interface required for link-local local/remote addresses"); 3861 return -EINVAL; 3862 } 3863 #endif 3864 3865 *lower = NULL; 3866 } 3867 3868 if (!conf->dst_port) { 3869 if (conf->flags & VXLAN_F_GPE) 3870 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3871 else 3872 conf->dst_port = htons(vxlan_port); 3873 } 3874 3875 if (!conf->age_interval) 3876 conf->age_interval = FDB_AGE_DEFAULT; 3877 3878 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3879 NL_SET_ERR_MSG(extack, 3880 "A VXLAN device with the specified VNI already exists"); 3881 return -EEXIST; 3882 } 3883 3884 return 0; 3885 } 3886 3887 static void vxlan_config_apply(struct net_device *dev, 3888 struct vxlan_config *conf, 3889 struct net_device *lowerdev, 3890 struct net *src_net, 3891 bool changelink) 3892 { 3893 struct vxlan_dev *vxlan = netdev_priv(dev); 3894 struct vxlan_rdst *dst = &vxlan->default_dst; 3895 unsigned short needed_headroom = ETH_HLEN; 3896 int max_mtu = ETH_MAX_MTU; 3897 u32 flags = conf->flags; 3898 3899 if (!changelink) { 3900 if (flags & VXLAN_F_GPE) 3901 vxlan_raw_setup(dev); 3902 else 3903 vxlan_ether_setup(dev); 3904 3905 if (conf->mtu) 3906 dev->mtu = conf->mtu; 3907 3908 vxlan->net = src_net; 3909 } 3910 3911 dst->remote_vni = conf->vni; 3912 3913 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3914 3915 if (lowerdev) { 3916 dst->remote_ifindex = conf->remote_ifindex; 3917 3918 netif_inherit_tso_max(dev, lowerdev); 3919 3920 needed_headroom = lowerdev->hard_header_len; 3921 needed_headroom += lowerdev->needed_headroom; 3922 3923 dev->needed_tailroom = lowerdev->needed_tailroom; 3924 3925 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3926 if (max_mtu < ETH_MIN_MTU) 3927 max_mtu = ETH_MIN_MTU; 3928 3929 if (!changelink && !conf->mtu) 3930 dev->mtu = max_mtu; 3931 } 3932 3933 if (dev->mtu > max_mtu) 3934 dev->mtu = max_mtu; 3935 3936 if (flags & VXLAN_F_COLLECT_METADATA) 3937 flags |= VXLAN_F_IPV6; 3938 needed_headroom += vxlan_headroom(flags); 3939 dev->needed_headroom = needed_headroom; 3940 3941 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3942 } 3943 3944 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3945 struct vxlan_config *conf, 3946 struct netlink_ext_ack *extack) 3947 { 3948 struct vxlan_dev *vxlan = netdev_priv(dev); 3949 struct net_device *lowerdev; 3950 int ret; 3951 3952 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3953 if (ret) 3954 return ret; 3955 3956 vxlan_config_apply(dev, conf, lowerdev, src_net, false); 3957 3958 return 0; 3959 } 3960 3961 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3962 struct vxlan_config *conf, 3963 struct netlink_ext_ack *extack) 3964 { 3965 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3966 struct vxlan_dev *vxlan = netdev_priv(dev); 3967 struct net_device *remote_dev = NULL; 3968 struct vxlan_rdst *dst; 3969 int err; 3970 3971 dst = &vxlan->default_dst; 3972 err = vxlan_dev_configure(net, dev, conf, extack); 3973 if (err) 3974 return err; 3975 3976 dev->ethtool_ops = &vxlan_ethtool_ops; 3977 3978 err = register_netdevice(dev); 3979 if (err) 3980 return err; 3981 3982 if (dst->remote_ifindex) { 3983 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3984 if (!remote_dev) { 3985 err = -ENODEV; 3986 goto unregister; 3987 } 3988 3989 err = netdev_upper_dev_link(remote_dev, dev, extack); 3990 if (err) 3991 goto unregister; 3992 3993 dst->remote_dev = remote_dev; 3994 } 3995 3996 err = rtnl_configure_link(dev, NULL, 0, NULL); 3997 if (err < 0) 3998 goto unlink; 3999 4000 /* create an fdb entry for a valid default destination */ 4001 if (!vxlan_addr_any(&dst->remote_ip)) { 4002 spin_lock_bh(&vxlan->hash_lock); 4003 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4004 &dst->remote_ip, 4005 NUD_REACHABLE | NUD_PERMANENT, 4006 NLM_F_EXCL | NLM_F_CREATE, 4007 vxlan->cfg.dst_port, 4008 dst->remote_vni, 4009 dst->remote_vni, 4010 dst->remote_ifindex, 4011 NTF_SELF, 0, true, extack); 4012 spin_unlock_bh(&vxlan->hash_lock); 4013 if (err) 4014 goto unlink; 4015 } 4016 4017 list_add(&vxlan->next, &vn->vxlan_list); 4018 4019 return 0; 4020 4021 unlink: 4022 if (remote_dev) 4023 netdev_upper_dev_unlink(remote_dev, dev); 4024 unregister: 4025 unregister_netdevice(dev); 4026 return err; 4027 } 4028 4029 /* Set/clear flags based on attribute */ 4030 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4031 int attrtype, unsigned long mask, bool changelink, 4032 bool changelink_supported, 4033 struct netlink_ext_ack *extack) 4034 { 4035 unsigned long flags; 4036 4037 if (!tb[attrtype]) 4038 return 0; 4039 4040 if (changelink && !changelink_supported) { 4041 vxlan_flag_attr_error(attrtype, extack); 4042 return -EOPNOTSUPP; 4043 } 4044 4045 if (vxlan_policy[attrtype].type == NLA_FLAG) 4046 flags = conf->flags | mask; 4047 else if (nla_get_u8(tb[attrtype])) 4048 flags = conf->flags | mask; 4049 else 4050 flags = conf->flags & ~mask; 4051 4052 conf->flags = flags; 4053 4054 return 0; 4055 } 4056 4057 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4058 struct net_device *dev, struct vxlan_config *conf, 4059 bool changelink, struct netlink_ext_ack *extack) 4060 { 4061 struct vxlanhdr used_bits = { 4062 .vx_flags = VXLAN_HF_VNI, 4063 .vx_vni = VXLAN_VNI_MASK, 4064 }; 4065 struct vxlan_dev *vxlan = netdev_priv(dev); 4066 int err = 0; 4067 4068 memset(conf, 0, sizeof(*conf)); 4069 4070 /* if changelink operation, start with old existing cfg */ 4071 if (changelink) 4072 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4073 4074 if (data[IFLA_VXLAN_ID]) { 4075 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4076 4077 if (changelink && (vni != conf->vni)) { 4078 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4079 return -EOPNOTSUPP; 4080 } 4081 conf->vni = vni; 4082 } 4083 4084 if (data[IFLA_VXLAN_GROUP]) { 4085 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4086 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4087 return -EOPNOTSUPP; 4088 } 4089 4090 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4091 conf->remote_ip.sa.sa_family = AF_INET; 4092 } else if (data[IFLA_VXLAN_GROUP6]) { 4093 if (!IS_ENABLED(CONFIG_IPV6)) { 4094 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4095 return -EPFNOSUPPORT; 4096 } 4097 4098 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4099 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4100 return -EOPNOTSUPP; 4101 } 4102 4103 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4104 conf->remote_ip.sa.sa_family = AF_INET6; 4105 } 4106 4107 if (data[IFLA_VXLAN_LOCAL]) { 4108 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4109 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4110 return -EOPNOTSUPP; 4111 } 4112 4113 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4114 conf->saddr.sa.sa_family = AF_INET; 4115 } else if (data[IFLA_VXLAN_LOCAL6]) { 4116 if (!IS_ENABLED(CONFIG_IPV6)) { 4117 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4118 return -EPFNOSUPPORT; 4119 } 4120 4121 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4122 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4123 return -EOPNOTSUPP; 4124 } 4125 4126 /* TODO: respect scope id */ 4127 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4128 conf->saddr.sa.sa_family = AF_INET6; 4129 } 4130 4131 if (data[IFLA_VXLAN_LINK]) 4132 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4133 4134 if (data[IFLA_VXLAN_TOS]) 4135 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4136 4137 if (data[IFLA_VXLAN_TTL]) 4138 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4139 4140 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4141 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4142 VXLAN_F_TTL_INHERIT, changelink, false, 4143 extack); 4144 if (err) 4145 return err; 4146 4147 } 4148 4149 if (data[IFLA_VXLAN_LABEL]) 4150 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4151 IPV6_FLOWLABEL_MASK; 4152 if (data[IFLA_VXLAN_LABEL_POLICY]) 4153 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4154 4155 if (data[IFLA_VXLAN_LEARNING]) { 4156 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4157 VXLAN_F_LEARN, changelink, true, 4158 extack); 4159 if (err) 4160 return err; 4161 } else if (!changelink) { 4162 /* default to learn on a new device */ 4163 conf->flags |= VXLAN_F_LEARN; 4164 } 4165 4166 if (data[IFLA_VXLAN_AGEING]) 4167 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4168 4169 if (data[IFLA_VXLAN_PROXY]) { 4170 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4171 VXLAN_F_PROXY, changelink, false, 4172 extack); 4173 if (err) 4174 return err; 4175 } 4176 4177 if (data[IFLA_VXLAN_RSC]) { 4178 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4179 VXLAN_F_RSC, changelink, false, 4180 extack); 4181 if (err) 4182 return err; 4183 } 4184 4185 if (data[IFLA_VXLAN_L2MISS]) { 4186 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4187 VXLAN_F_L2MISS, changelink, false, 4188 extack); 4189 if (err) 4190 return err; 4191 } 4192 4193 if (data[IFLA_VXLAN_L3MISS]) { 4194 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4195 VXLAN_F_L3MISS, changelink, false, 4196 extack); 4197 if (err) 4198 return err; 4199 } 4200 4201 if (data[IFLA_VXLAN_LIMIT]) { 4202 if (changelink) { 4203 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4204 "Cannot change limit"); 4205 return -EOPNOTSUPP; 4206 } 4207 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4208 } 4209 4210 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4211 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4212 VXLAN_F_COLLECT_METADATA, changelink, false, 4213 extack); 4214 if (err) 4215 return err; 4216 } 4217 4218 if (data[IFLA_VXLAN_PORT_RANGE]) { 4219 if (!changelink) { 4220 const struct ifla_vxlan_port_range *p 4221 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4222 conf->port_min = ntohs(p->low); 4223 conf->port_max = ntohs(p->high); 4224 } else { 4225 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4226 "Cannot change port range"); 4227 return -EOPNOTSUPP; 4228 } 4229 } 4230 4231 if (data[IFLA_VXLAN_PORT]) { 4232 if (changelink) { 4233 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4234 "Cannot change port"); 4235 return -EOPNOTSUPP; 4236 } 4237 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4238 } 4239 4240 if (data[IFLA_VXLAN_UDP_CSUM]) { 4241 if (changelink) { 4242 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4243 "Cannot change UDP_CSUM flag"); 4244 return -EOPNOTSUPP; 4245 } 4246 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4247 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4248 } 4249 4250 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4251 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4252 VXLAN_F_LOCALBYPASS, changelink, 4253 true, extack); 4254 if (err) 4255 return err; 4256 } else if (!changelink) { 4257 /* default to local bypass on a new device */ 4258 conf->flags |= VXLAN_F_LOCALBYPASS; 4259 } 4260 4261 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4262 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4263 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4264 false, extack); 4265 if (err) 4266 return err; 4267 } 4268 4269 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4270 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4271 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4272 false, extack); 4273 if (err) 4274 return err; 4275 } 4276 4277 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4278 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4279 VXLAN_F_REMCSUM_TX, changelink, false, 4280 extack); 4281 if (err) 4282 return err; 4283 } 4284 4285 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4286 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4287 VXLAN_F_REMCSUM_RX, changelink, false, 4288 extack); 4289 if (err) 4290 return err; 4291 used_bits.vx_flags |= VXLAN_HF_RCO; 4292 used_bits.vx_vni |= ~VXLAN_VNI_MASK; 4293 } 4294 4295 if (data[IFLA_VXLAN_GBP]) { 4296 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4297 VXLAN_F_GBP, changelink, false, extack); 4298 if (err) 4299 return err; 4300 used_bits.vx_flags |= VXLAN_GBP_USED_BITS; 4301 } 4302 4303 if (data[IFLA_VXLAN_GPE]) { 4304 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4305 VXLAN_F_GPE, changelink, false, 4306 extack); 4307 if (err) 4308 return err; 4309 4310 used_bits.vx_flags |= VXLAN_GPE_USED_BITS; 4311 } 4312 4313 if (data[IFLA_VXLAN_RESERVED_BITS]) { 4314 struct vxlanhdr reserved_bits; 4315 4316 if (changelink) { 4317 NL_SET_ERR_MSG_ATTR(extack, 4318 data[IFLA_VXLAN_RESERVED_BITS], 4319 "Cannot change reserved_bits"); 4320 return -EOPNOTSUPP; 4321 } 4322 4323 nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS], 4324 sizeof(reserved_bits)); 4325 if (used_bits.vx_flags & reserved_bits.vx_flags || 4326 used_bits.vx_vni & reserved_bits.vx_vni) { 4327 __be64 ub_be64, rb_be64; 4328 4329 memcpy(&ub_be64, &used_bits, sizeof(ub_be64)); 4330 memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64)); 4331 4332 NL_SET_ERR_MSG_ATTR_FMT(extack, 4333 data[IFLA_VXLAN_RESERVED_BITS], 4334 "Used bits %#018llx cannot overlap reserved bits %#018llx", 4335 be64_to_cpu(ub_be64), 4336 be64_to_cpu(rb_be64)); 4337 return -EINVAL; 4338 } 4339 4340 conf->reserved_bits = reserved_bits; 4341 } else { 4342 /* For backwards compatibility, only allow reserved fields to be 4343 * used by VXLAN extensions if explicitly requested. 4344 */ 4345 conf->reserved_bits = (struct vxlanhdr) { 4346 .vx_flags = ~used_bits.vx_flags, 4347 .vx_vni = ~used_bits.vx_vni, 4348 }; 4349 } 4350 4351 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4352 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4353 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4354 false, extack); 4355 if (err) 4356 return err; 4357 } 4358 4359 if (data[IFLA_VXLAN_MC_ROUTE]) { 4360 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_MC_ROUTE, 4361 VXLAN_F_MC_ROUTE, changelink, 4362 true, extack); 4363 if (err) 4364 return err; 4365 } 4366 4367 if (tb[IFLA_MTU]) { 4368 if (changelink) { 4369 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4370 "Cannot change mtu"); 4371 return -EOPNOTSUPP; 4372 } 4373 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4374 } 4375 4376 if (data[IFLA_VXLAN_DF]) 4377 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4378 4379 if (data[IFLA_VXLAN_VNIFILTER]) { 4380 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4381 VXLAN_F_VNIFILTER, changelink, false, 4382 extack); 4383 if (err) 4384 return err; 4385 4386 if ((conf->flags & VXLAN_F_VNIFILTER) && 4387 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4388 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4389 "vxlan vnifilter only valid in collect metadata mode"); 4390 return -EINVAL; 4391 } 4392 } 4393 4394 return 0; 4395 } 4396 4397 static int vxlan_newlink(struct net_device *dev, 4398 struct rtnl_newlink_params *params, 4399 struct netlink_ext_ack *extack) 4400 { 4401 struct net *link_net = rtnl_newlink_link_net(params); 4402 struct nlattr **data = params->data; 4403 struct nlattr **tb = params->tb; 4404 struct vxlan_config conf; 4405 int err; 4406 4407 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4408 if (err) 4409 return err; 4410 4411 return __vxlan_dev_create(link_net, dev, &conf, extack); 4412 } 4413 4414 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4415 struct nlattr *data[], 4416 struct netlink_ext_ack *extack) 4417 { 4418 struct vxlan_dev *vxlan = netdev_priv(dev); 4419 bool rem_ip_changed, change_igmp; 4420 struct net_device *lowerdev; 4421 struct vxlan_config conf; 4422 struct vxlan_rdst *dst; 4423 int err; 4424 4425 dst = &vxlan->default_dst; 4426 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4427 if (err) 4428 return err; 4429 4430 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4431 vxlan, extack); 4432 if (err) 4433 return err; 4434 4435 if (dst->remote_dev == lowerdev) 4436 lowerdev = NULL; 4437 4438 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4439 extack); 4440 if (err) 4441 return err; 4442 4443 rem_ip_changed = !vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip); 4444 change_igmp = vxlan->dev->flags & IFF_UP && 4445 (rem_ip_changed || 4446 dst->remote_ifindex != conf.remote_ifindex); 4447 4448 /* handle default dst entry */ 4449 if (rem_ip_changed) { 4450 spin_lock_bh(&vxlan->hash_lock); 4451 if (!vxlan_addr_any(&conf.remote_ip)) { 4452 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4453 &conf.remote_ip, 4454 NUD_REACHABLE | NUD_PERMANENT, 4455 NLM_F_APPEND | NLM_F_CREATE, 4456 vxlan->cfg.dst_port, 4457 conf.vni, conf.vni, 4458 conf.remote_ifindex, 4459 NTF_SELF, 0, true, extack); 4460 if (err) { 4461 spin_unlock_bh(&vxlan->hash_lock); 4462 netdev_adjacent_change_abort(dst->remote_dev, 4463 lowerdev, dev); 4464 return err; 4465 } 4466 } 4467 if (!vxlan_addr_any(&dst->remote_ip)) 4468 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4469 dst->remote_ip, 4470 vxlan->cfg.dst_port, 4471 dst->remote_vni, 4472 dst->remote_vni, 4473 dst->remote_ifindex, 4474 true); 4475 spin_unlock_bh(&vxlan->hash_lock); 4476 4477 /* If vni filtering device, also update fdb entries of 4478 * all vnis that were using default remote ip 4479 */ 4480 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4481 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4482 &conf.remote_ip, extack); 4483 if (err) { 4484 netdev_adjacent_change_abort(dst->remote_dev, 4485 lowerdev, dev); 4486 return err; 4487 } 4488 } 4489 } 4490 4491 if (change_igmp && vxlan_addr_multicast(&dst->remote_ip)) 4492 err = vxlan_multicast_leave(vxlan); 4493 4494 if (conf.age_interval != vxlan->cfg.age_interval) 4495 mod_timer(&vxlan->age_timer, jiffies); 4496 4497 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4498 if (lowerdev && lowerdev != dst->remote_dev) 4499 dst->remote_dev = lowerdev; 4500 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4501 4502 if (!err && change_igmp && 4503 vxlan_addr_multicast(&dst->remote_ip)) 4504 err = vxlan_multicast_join(vxlan); 4505 4506 return err; 4507 } 4508 4509 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4510 { 4511 struct vxlan_dev *vxlan = netdev_priv(dev); 4512 struct vxlan_fdb_flush_desc desc = {}; 4513 4514 vxlan_flush(vxlan, &desc); 4515 4516 list_del(&vxlan->next); 4517 unregister_netdevice_queue(dev, head); 4518 if (vxlan->default_dst.remote_dev) 4519 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4520 } 4521 4522 static size_t vxlan_get_size(const struct net_device *dev) 4523 { 4524 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4525 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4526 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4527 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4528 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4529 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4530 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4531 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4532 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4533 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4534 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4535 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4536 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4537 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4538 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4540 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4541 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4542 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4543 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4544 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4545 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4546 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4547 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4548 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4549 /* IFLA_VXLAN_PORT_RANGE */ 4550 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4551 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4552 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4553 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4554 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4555 /* IFLA_VXLAN_RESERVED_BITS */ 4556 nla_total_size(sizeof(struct vxlanhdr)) + 4557 0; 4558 } 4559 4560 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4561 { 4562 const struct vxlan_dev *vxlan = netdev_priv(dev); 4563 const struct vxlan_rdst *dst = &vxlan->default_dst; 4564 struct ifla_vxlan_port_range ports = { 4565 .low = htons(vxlan->cfg.port_min), 4566 .high = htons(vxlan->cfg.port_max), 4567 }; 4568 4569 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4570 goto nla_put_failure; 4571 4572 if (!vxlan_addr_any(&dst->remote_ip)) { 4573 if (dst->remote_ip.sa.sa_family == AF_INET) { 4574 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4575 dst->remote_ip.sin.sin_addr.s_addr)) 4576 goto nla_put_failure; 4577 #if IS_ENABLED(CONFIG_IPV6) 4578 } else { 4579 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4580 &dst->remote_ip.sin6.sin6_addr)) 4581 goto nla_put_failure; 4582 #endif 4583 } 4584 } 4585 4586 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4587 goto nla_put_failure; 4588 4589 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4590 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4591 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4592 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4593 goto nla_put_failure; 4594 #if IS_ENABLED(CONFIG_IPV6) 4595 } else { 4596 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4597 &vxlan->cfg.saddr.sin6.sin6_addr)) 4598 goto nla_put_failure; 4599 #endif 4600 } 4601 } 4602 4603 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4604 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4605 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4606 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4607 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4608 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4609 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4610 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4611 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4612 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4613 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4614 nla_put_u8(skb, IFLA_VXLAN_RSC, 4615 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4616 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4617 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4618 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4619 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4620 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4621 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4622 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4623 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4624 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4625 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4626 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4627 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4628 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4629 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4630 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4631 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4632 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4633 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4634 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4635 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4636 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4637 goto nla_put_failure; 4638 4639 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4640 goto nla_put_failure; 4641 4642 if (vxlan->cfg.flags & VXLAN_F_GBP && 4643 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4644 goto nla_put_failure; 4645 4646 if (vxlan->cfg.flags & VXLAN_F_GPE && 4647 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4648 goto nla_put_failure; 4649 4650 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4651 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4652 goto nla_put_failure; 4653 4654 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4655 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4656 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4657 goto nla_put_failure; 4658 4659 if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS, 4660 sizeof(vxlan->cfg.reserved_bits), 4661 &vxlan->cfg.reserved_bits)) 4662 goto nla_put_failure; 4663 4664 return 0; 4665 4666 nla_put_failure: 4667 return -EMSGSIZE; 4668 } 4669 4670 static struct net *vxlan_get_link_net(const struct net_device *dev) 4671 { 4672 struct vxlan_dev *vxlan = netdev_priv(dev); 4673 4674 return READ_ONCE(vxlan->net); 4675 } 4676 4677 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4678 .kind = "vxlan", 4679 .maxtype = IFLA_VXLAN_MAX, 4680 .policy = vxlan_policy, 4681 .priv_size = sizeof(struct vxlan_dev), 4682 .setup = vxlan_setup, 4683 .validate = vxlan_validate, 4684 .newlink = vxlan_newlink, 4685 .changelink = vxlan_changelink, 4686 .dellink = vxlan_dellink, 4687 .get_size = vxlan_get_size, 4688 .fill_info = vxlan_fill_info, 4689 .get_link_net = vxlan_get_link_net, 4690 }; 4691 4692 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4693 u8 name_assign_type, 4694 struct vxlan_config *conf) 4695 { 4696 struct nlattr *tb[IFLA_MAX + 1]; 4697 struct net_device *dev; 4698 int err; 4699 4700 memset(&tb, 0, sizeof(tb)); 4701 4702 dev = rtnl_create_link(net, name, name_assign_type, 4703 &vxlan_link_ops, tb, NULL); 4704 if (IS_ERR(dev)) 4705 return dev; 4706 4707 err = __vxlan_dev_create(net, dev, conf, NULL); 4708 if (err < 0) { 4709 free_netdev(dev); 4710 return ERR_PTR(err); 4711 } 4712 4713 err = rtnl_configure_link(dev, NULL, 0, NULL); 4714 if (err < 0) { 4715 LIST_HEAD(list_kill); 4716 4717 vxlan_dellink(dev, &list_kill); 4718 unregister_netdevice_many(&list_kill); 4719 return ERR_PTR(err); 4720 } 4721 4722 return dev; 4723 } 4724 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4725 4726 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4727 struct net_device *dev) 4728 { 4729 struct vxlan_dev *vxlan, *next; 4730 LIST_HEAD(list_kill); 4731 4732 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4733 struct vxlan_rdst *dst = &vxlan->default_dst; 4734 4735 /* In case we created vxlan device with carrier 4736 * and we loose the carrier due to module unload 4737 * we also need to remove vxlan device. In other 4738 * cases, it's not necessary and remote_ifindex 4739 * is 0 here, so no matches. 4740 */ 4741 if (dst->remote_ifindex == dev->ifindex) 4742 vxlan_dellink(vxlan->dev, &list_kill); 4743 } 4744 4745 unregister_netdevice_many(&list_kill); 4746 } 4747 4748 static int vxlan_netdevice_event(struct notifier_block *unused, 4749 unsigned long event, void *ptr) 4750 { 4751 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4752 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4753 4754 if (event == NETDEV_UNREGISTER) 4755 vxlan_handle_lowerdev_unregister(vn, dev); 4756 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4757 vxlan_offload_rx_ports(dev, true); 4758 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4759 vxlan_offload_rx_ports(dev, false); 4760 4761 return NOTIFY_DONE; 4762 } 4763 4764 static struct notifier_block vxlan_notifier_block __read_mostly = { 4765 .notifier_call = vxlan_netdevice_event, 4766 }; 4767 4768 static void 4769 vxlan_fdb_offloaded_set(struct net_device *dev, 4770 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4771 { 4772 struct vxlan_dev *vxlan = netdev_priv(dev); 4773 struct vxlan_rdst *rdst; 4774 struct vxlan_fdb *f; 4775 4776 spin_lock_bh(&vxlan->hash_lock); 4777 4778 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4779 if (!f) 4780 goto out; 4781 4782 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4783 fdb_info->remote_port, 4784 fdb_info->remote_vni, 4785 fdb_info->remote_ifindex); 4786 if (!rdst) 4787 goto out; 4788 4789 rdst->offloaded = fdb_info->offloaded; 4790 4791 out: 4792 spin_unlock_bh(&vxlan->hash_lock); 4793 } 4794 4795 static int 4796 vxlan_fdb_external_learn_add(struct net_device *dev, 4797 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4798 { 4799 struct vxlan_dev *vxlan = netdev_priv(dev); 4800 struct netlink_ext_ack *extack; 4801 int err; 4802 4803 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4804 4805 spin_lock_bh(&vxlan->hash_lock); 4806 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4807 NUD_REACHABLE, 4808 NLM_F_CREATE | NLM_F_REPLACE, 4809 fdb_info->remote_port, 4810 fdb_info->vni, 4811 fdb_info->remote_vni, 4812 fdb_info->remote_ifindex, 4813 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4814 0, false, extack); 4815 spin_unlock_bh(&vxlan->hash_lock); 4816 4817 return err; 4818 } 4819 4820 static int 4821 vxlan_fdb_external_learn_del(struct net_device *dev, 4822 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4823 { 4824 struct vxlan_dev *vxlan = netdev_priv(dev); 4825 struct vxlan_fdb *f; 4826 int err = 0; 4827 4828 spin_lock_bh(&vxlan->hash_lock); 4829 4830 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4831 if (!f) 4832 err = -ENOENT; 4833 else if (f->flags & NTF_EXT_LEARNED) 4834 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4835 fdb_info->remote_ip, 4836 fdb_info->remote_port, 4837 fdb_info->vni, 4838 fdb_info->remote_vni, 4839 fdb_info->remote_ifindex, 4840 false); 4841 4842 spin_unlock_bh(&vxlan->hash_lock); 4843 4844 return err; 4845 } 4846 4847 static int vxlan_switchdev_event(struct notifier_block *unused, 4848 unsigned long event, void *ptr) 4849 { 4850 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4851 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4852 int err = 0; 4853 4854 switch (event) { 4855 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4856 vxlan_fdb_offloaded_set(dev, ptr); 4857 break; 4858 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4859 fdb_info = ptr; 4860 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4861 if (err) { 4862 err = notifier_from_errno(err); 4863 break; 4864 } 4865 fdb_info->offloaded = true; 4866 vxlan_fdb_offloaded_set(dev, fdb_info); 4867 break; 4868 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4869 fdb_info = ptr; 4870 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4871 if (err) { 4872 err = notifier_from_errno(err); 4873 break; 4874 } 4875 fdb_info->offloaded = false; 4876 vxlan_fdb_offloaded_set(dev, fdb_info); 4877 break; 4878 } 4879 4880 return err; 4881 } 4882 4883 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4884 .notifier_call = vxlan_switchdev_event, 4885 }; 4886 4887 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4888 { 4889 struct vxlan_fdb *fdb; 4890 struct vxlan_dev *vxlan; 4891 4892 rcu_read_lock(); 4893 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4894 vxlan = rcu_dereference(fdb->vdev); 4895 WARN_ON(!vxlan); 4896 spin_lock_bh(&vxlan->hash_lock); 4897 if (!hlist_unhashed(&fdb->fdb_node)) 4898 vxlan_fdb_destroy(vxlan, fdb, false, false); 4899 spin_unlock_bh(&vxlan->hash_lock); 4900 } 4901 rcu_read_unlock(); 4902 } 4903 4904 static int vxlan_nexthop_event(struct notifier_block *nb, 4905 unsigned long event, void *ptr) 4906 { 4907 struct nh_notifier_info *info = ptr; 4908 struct nexthop *nh; 4909 4910 if (event != NEXTHOP_EVENT_DEL) 4911 return NOTIFY_DONE; 4912 4913 nh = nexthop_find_by_id(info->net, info->id); 4914 if (!nh) 4915 return NOTIFY_DONE; 4916 4917 vxlan_fdb_nh_flush(nh); 4918 4919 return NOTIFY_DONE; 4920 } 4921 4922 static __net_init int vxlan_init_net(struct net *net) 4923 { 4924 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4925 unsigned int h; 4926 4927 INIT_LIST_HEAD(&vn->vxlan_list); 4928 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4929 4930 for (h = 0; h < PORT_HASH_SIZE; ++h) 4931 INIT_HLIST_HEAD(&vn->sock_list[h]); 4932 4933 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4934 NULL); 4935 } 4936 4937 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4938 struct list_head *dev_to_kill) 4939 { 4940 struct vxlan_dev *vxlan, *next; 4941 4942 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4943 vxlan_dellink(vxlan->dev, dev_to_kill); 4944 } 4945 4946 static void __net_exit vxlan_exit_rtnl(struct net *net, 4947 struct list_head *dev_to_kill) 4948 { 4949 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4950 4951 ASSERT_RTNL_NET(net); 4952 4953 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4954 vxlan_destroy_tunnels(vn, dev_to_kill); 4955 } 4956 4957 static void __net_exit vxlan_exit_net(struct net *net) 4958 { 4959 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4960 unsigned int h; 4961 4962 for (h = 0; h < PORT_HASH_SIZE; ++h) 4963 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4964 } 4965 4966 static struct pernet_operations vxlan_net_ops = { 4967 .init = vxlan_init_net, 4968 .exit_rtnl = vxlan_exit_rtnl, 4969 .exit = vxlan_exit_net, 4970 .id = &vxlan_net_id, 4971 .size = sizeof(struct vxlan_net), 4972 }; 4973 4974 static int __init vxlan_init_module(void) 4975 { 4976 int rc; 4977 4978 rc = register_pernet_subsys(&vxlan_net_ops); 4979 if (rc) 4980 goto out1; 4981 4982 rc = register_netdevice_notifier(&vxlan_notifier_block); 4983 if (rc) 4984 goto out2; 4985 4986 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4987 if (rc) 4988 goto out3; 4989 4990 rc = rtnl_link_register(&vxlan_link_ops); 4991 if (rc) 4992 goto out4; 4993 4994 rc = vxlan_vnifilter_init(); 4995 if (rc) 4996 goto out5; 4997 4998 return 0; 4999 out5: 5000 rtnl_link_unregister(&vxlan_link_ops); 5001 out4: 5002 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5003 out3: 5004 unregister_netdevice_notifier(&vxlan_notifier_block); 5005 out2: 5006 unregister_pernet_subsys(&vxlan_net_ops); 5007 out1: 5008 return rc; 5009 } 5010 late_initcall(vxlan_init_module); 5011 5012 static void __exit vxlan_cleanup_module(void) 5013 { 5014 vxlan_vnifilter_uninit(); 5015 rtnl_link_unregister(&vxlan_link_ops); 5016 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5017 unregister_netdevice_notifier(&vxlan_notifier_block); 5018 unregister_pernet_subsys(&vxlan_net_ops); 5019 /* rcu_barrier() is called by netns */ 5020 } 5021 module_exit(vxlan_cleanup_module); 5022 5023 MODULE_LICENSE("GPL"); 5024 MODULE_VERSION(VXLAN_VERSION); 5025 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 5026 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 5027 MODULE_ALIAS_RTNL_LINK("vxlan"); 5028