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 if (rcu_access_pointer(f->nh) && 1000 !(state & (NUD_PERMANENT | NUD_NOARP))) { 1001 NL_SET_ERR_MSG(extack, "Cannot make a nexthop fdb dynamic"); 1002 return -EOPNOTSUPP; 1003 } 1004 1005 /* Do not allow an externally learned entry to take over an entry added 1006 * by the user. 1007 */ 1008 if (!(fdb_flags & NTF_EXT_LEARNED) || 1009 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 1010 if (f->state != state) { 1011 f->state = state; 1012 notify = 1; 1013 } 1014 if (f->flags != fdb_flags) { 1015 f->flags = fdb_flags; 1016 notify = 1; 1017 } 1018 } 1019 1020 if ((flags & NLM_F_REPLACE)) { 1021 /* Only change unicasts */ 1022 if (!(is_multicast_ether_addr(f->key.eth_addr) || 1023 is_zero_ether_addr(f->key.eth_addr))) { 1024 if (nhid) { 1025 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1026 if (rc < 0) 1027 return rc; 1028 } else { 1029 rc = vxlan_fdb_replace(f, ip, port, vni, 1030 ifindex, &oldrd); 1031 } 1032 notify |= rc; 1033 } else { 1034 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1035 return -EOPNOTSUPP; 1036 } 1037 } 1038 if ((flags & NLM_F_APPEND) && 1039 (is_multicast_ether_addr(f->key.eth_addr) || 1040 is_zero_ether_addr(f->key.eth_addr))) { 1041 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1042 1043 if (rc < 0) 1044 return rc; 1045 notify |= rc; 1046 } 1047 1048 if (ndm_flags & NTF_USE) 1049 WRITE_ONCE(f->updated, jiffies); 1050 1051 if (notify) { 1052 if (rd == NULL) 1053 rd = first_remote_rtnl(f); 1054 1055 WRITE_ONCE(f->updated, jiffies); 1056 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1057 swdev_notify, extack); 1058 if (err) 1059 goto err_notify; 1060 } 1061 1062 return 0; 1063 1064 err_notify: 1065 if (nhid) 1066 return err; 1067 if ((flags & NLM_F_REPLACE) && rc) 1068 *rd = oldrd; 1069 else if ((flags & NLM_F_APPEND) && rc) { 1070 list_del_rcu(&rd->list); 1071 call_rcu(&rd->rcu, vxlan_dst_free); 1072 } 1073 return err; 1074 } 1075 1076 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1077 const u8 *mac, union vxlan_addr *ip, 1078 __u16 state, __u16 flags, 1079 __be16 port, __be32 src_vni, __be32 vni, 1080 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1081 bool swdev_notify, 1082 struct netlink_ext_ack *extack) 1083 { 1084 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1085 struct vxlan_fdb *f; 1086 int rc; 1087 1088 /* Disallow replace to add a multicast entry */ 1089 if ((flags & NLM_F_REPLACE) && 1090 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1091 return -EOPNOTSUPP; 1092 1093 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1094 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1095 vni, ifindex, fdb_flags, nhid, &f, extack); 1096 if (rc < 0) 1097 return rc; 1098 1099 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1100 swdev_notify, extack); 1101 if (rc) 1102 goto err_notify; 1103 1104 return 0; 1105 1106 err_notify: 1107 vxlan_fdb_destroy(vxlan, f, false, false); 1108 return rc; 1109 } 1110 1111 /* Add new entry to forwarding table -- assumes lock held */ 1112 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1113 const u8 *mac, union vxlan_addr *ip, 1114 __u16 state, __u16 flags, 1115 __be16 port, __be32 src_vni, __be32 vni, 1116 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1117 bool swdev_notify, 1118 struct netlink_ext_ack *extack) 1119 { 1120 struct vxlan_fdb *f; 1121 1122 f = vxlan_find_mac(vxlan, mac, src_vni); 1123 if (f) { 1124 if (flags & NLM_F_EXCL) { 1125 netdev_dbg(vxlan->dev, 1126 "lost race to create %pM\n", mac); 1127 return -EEXIST; 1128 } 1129 1130 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1131 vni, ifindex, ndm_flags, f, 1132 nhid, swdev_notify, extack); 1133 } else { 1134 if (!(flags & NLM_F_CREATE)) 1135 return -ENOENT; 1136 1137 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1138 port, src_vni, vni, ifindex, 1139 ndm_flags, nhid, swdev_notify, 1140 extack); 1141 } 1142 } 1143 1144 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1145 struct vxlan_rdst *rd, bool swdev_notify) 1146 { 1147 list_del_rcu(&rd->list); 1148 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1149 call_rcu(&rd->rcu, vxlan_dst_free); 1150 } 1151 1152 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1153 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1154 __be32 *vni, u32 *ifindex, u32 *nhid, 1155 struct netlink_ext_ack *extack) 1156 { 1157 struct net *net = dev_net(vxlan->dev); 1158 int err; 1159 1160 if (tb[NDA_NH_ID] && 1161 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1162 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1163 return -EINVAL; 1164 } 1165 1166 if (tb[NDA_DST]) { 1167 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1168 if (err) { 1169 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1170 return err; 1171 } 1172 } else { 1173 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1174 1175 if (remote->sa.sa_family == AF_INET) { 1176 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1177 ip->sa.sa_family = AF_INET; 1178 #if IS_ENABLED(CONFIG_IPV6) 1179 } else { 1180 ip->sin6.sin6_addr = in6addr_any; 1181 ip->sa.sa_family = AF_INET6; 1182 #endif 1183 } 1184 } 1185 1186 if (tb[NDA_PORT]) { 1187 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1188 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1189 return -EINVAL; 1190 } 1191 *port = nla_get_be16(tb[NDA_PORT]); 1192 } else { 1193 *port = vxlan->cfg.dst_port; 1194 } 1195 1196 if (tb[NDA_VNI]) { 1197 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1198 NL_SET_ERR_MSG(extack, "Invalid vni"); 1199 return -EINVAL; 1200 } 1201 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1202 } else { 1203 *vni = vxlan->default_dst.remote_vni; 1204 } 1205 1206 if (tb[NDA_SRC_VNI]) { 1207 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1208 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1209 return -EINVAL; 1210 } 1211 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1212 } else { 1213 *src_vni = vxlan->default_dst.remote_vni; 1214 } 1215 1216 if (tb[NDA_IFINDEX]) { 1217 struct net_device *tdev; 1218 1219 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1220 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1221 return -EINVAL; 1222 } 1223 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1224 tdev = __dev_get_by_index(net, *ifindex); 1225 if (!tdev) { 1226 NL_SET_ERR_MSG(extack, "Device not found"); 1227 return -EADDRNOTAVAIL; 1228 } 1229 } else { 1230 *ifindex = 0; 1231 } 1232 1233 *nhid = nla_get_u32_default(tb[NDA_NH_ID], 0); 1234 1235 return 0; 1236 } 1237 1238 /* Add static entry (via netlink) */ 1239 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1240 struct net_device *dev, 1241 const unsigned char *addr, u16 vid, u16 flags, 1242 bool *notified, struct netlink_ext_ack *extack) 1243 { 1244 struct vxlan_dev *vxlan = netdev_priv(dev); 1245 /* struct net *net = dev_net(vxlan->dev); */ 1246 union vxlan_addr ip; 1247 __be16 port; 1248 __be32 src_vni, vni; 1249 u32 ifindex, nhid; 1250 int err; 1251 1252 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1253 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1254 ndm->ndm_state); 1255 return -EINVAL; 1256 } 1257 1258 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1259 return -EINVAL; 1260 1261 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1262 &nhid, extack); 1263 if (err) 1264 return err; 1265 1266 if (nhid && !(ndm->ndm_state & (NUD_PERMANENT | NUD_NOARP))) { 1267 NL_SET_ERR_MSG(extack, "A nexthop fdb cannot be dynamic"); 1268 return -EINVAL; 1269 } 1270 1271 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1272 return -EAFNOSUPPORT; 1273 1274 spin_lock_bh(&vxlan->hash_lock); 1275 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1276 port, src_vni, vni, ifindex, 1277 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1278 nhid, true, extack); 1279 spin_unlock_bh(&vxlan->hash_lock); 1280 1281 if (!err) 1282 *notified = true; 1283 1284 return err; 1285 } 1286 1287 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1288 const unsigned char *addr, union vxlan_addr ip, 1289 __be16 port, __be32 src_vni, __be32 vni, 1290 u32 ifindex, bool swdev_notify) 1291 { 1292 struct vxlan_rdst *rd = NULL; 1293 struct vxlan_fdb *f; 1294 int err = -ENOENT; 1295 1296 f = vxlan_find_mac(vxlan, addr, src_vni); 1297 if (!f) 1298 return err; 1299 1300 if (!vxlan_addr_any(&ip)) { 1301 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1302 if (!rd) 1303 goto out; 1304 } 1305 1306 /* remove a destination if it's not the only one on the list, 1307 * otherwise destroy the fdb entry 1308 */ 1309 if (rd && !list_is_singular(&f->remotes)) { 1310 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1311 goto out; 1312 } 1313 1314 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1315 1316 out: 1317 return 0; 1318 } 1319 1320 /* Delete entry (via netlink) */ 1321 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1322 struct net_device *dev, 1323 const unsigned char *addr, u16 vid, bool *notified, 1324 struct netlink_ext_ack *extack) 1325 { 1326 struct vxlan_dev *vxlan = netdev_priv(dev); 1327 union vxlan_addr ip; 1328 __be32 src_vni, vni; 1329 u32 ifindex, nhid; 1330 __be16 port; 1331 int err; 1332 1333 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1334 &nhid, extack); 1335 if (err) 1336 return err; 1337 1338 spin_lock_bh(&vxlan->hash_lock); 1339 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1340 true); 1341 spin_unlock_bh(&vxlan->hash_lock); 1342 1343 if (!err) 1344 *notified = true; 1345 1346 return err; 1347 } 1348 1349 /* Dump forwarding table */ 1350 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1351 struct net_device *dev, 1352 struct net_device *filter_dev, int *idx) 1353 { 1354 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx; 1355 struct vxlan_dev *vxlan = netdev_priv(dev); 1356 struct vxlan_fdb *f; 1357 int err = 0; 1358 1359 rcu_read_lock(); 1360 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 1361 struct vxlan_rdst *rd; 1362 1363 if (rcu_access_pointer(f->nh)) { 1364 if (*idx < ctx->fdb_idx) 1365 goto skip_nh; 1366 err = vxlan_fdb_info(skb, vxlan, f, 1367 NETLINK_CB(cb->skb).portid, 1368 cb->nlh->nlmsg_seq, 1369 RTM_NEWNEIGH, NLM_F_MULTI, NULL); 1370 if (err < 0) { 1371 rcu_read_unlock(); 1372 goto out; 1373 } 1374 skip_nh: 1375 *idx += 1; 1376 continue; 1377 } 1378 1379 list_for_each_entry_rcu(rd, &f->remotes, list) { 1380 if (*idx < ctx->fdb_idx) 1381 goto skip; 1382 1383 err = vxlan_fdb_info(skb, vxlan, f, 1384 NETLINK_CB(cb->skb).portid, 1385 cb->nlh->nlmsg_seq, 1386 RTM_NEWNEIGH, NLM_F_MULTI, rd); 1387 if (err < 0) { 1388 rcu_read_unlock(); 1389 goto out; 1390 } 1391 skip: 1392 *idx += 1; 1393 } 1394 } 1395 rcu_read_unlock(); 1396 out: 1397 return err; 1398 } 1399 1400 static int vxlan_fdb_get(struct sk_buff *skb, 1401 struct nlattr *tb[], 1402 struct net_device *dev, 1403 const unsigned char *addr, 1404 u16 vid, u32 portid, u32 seq, 1405 struct netlink_ext_ack *extack) 1406 { 1407 struct vxlan_dev *vxlan = netdev_priv(dev); 1408 struct vxlan_fdb *f; 1409 __be32 vni; 1410 int err; 1411 1412 if (tb[NDA_VNI]) 1413 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1414 else 1415 vni = vxlan->default_dst.remote_vni; 1416 1417 rcu_read_lock(); 1418 1419 f = vxlan_find_mac_rcu(vxlan, addr, vni); 1420 if (!f) { 1421 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1422 err = -ENOENT; 1423 goto errout; 1424 } 1425 1426 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1427 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1428 errout: 1429 rcu_read_unlock(); 1430 return err; 1431 } 1432 1433 /* Watch incoming packets to learn mapping between Ethernet address 1434 * and Tunnel endpoint. 1435 */ 1436 static enum skb_drop_reason vxlan_snoop(struct net_device *dev, 1437 union vxlan_addr *src_ip, 1438 const u8 *src_mac, u32 src_ifindex, 1439 __be32 vni) 1440 { 1441 struct vxlan_dev *vxlan = netdev_priv(dev); 1442 struct vxlan_fdb *f; 1443 u32 ifindex = 0; 1444 1445 /* Ignore packets from invalid src-address */ 1446 if (!is_valid_ether_addr(src_mac)) 1447 return SKB_DROP_REASON_MAC_INVALID_SOURCE; 1448 1449 #if IS_ENABLED(CONFIG_IPV6) 1450 if (src_ip->sa.sa_family == AF_INET6 && 1451 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1452 ifindex = src_ifindex; 1453 #endif 1454 1455 f = vxlan_find_mac_rcu(vxlan, src_mac, vni); 1456 if (likely(f)) { 1457 struct vxlan_rdst *rdst = first_remote_rcu(f); 1458 unsigned long now = jiffies; 1459 1460 if (READ_ONCE(f->updated) != now) 1461 WRITE_ONCE(f->updated, now); 1462 1463 /* Don't override an fdb with nexthop with a learnt entry */ 1464 if (rcu_access_pointer(f->nh)) 1465 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1466 1467 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1468 rdst->remote_ifindex == ifindex)) 1469 return SKB_NOT_DROPPED_YET; 1470 1471 /* Don't migrate static entries, drop packets */ 1472 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1473 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1474 1475 if (net_ratelimit()) 1476 netdev_info(dev, 1477 "%pM migrated from %pIS to %pIS\n", 1478 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1479 1480 rdst->remote_ip = *src_ip; 1481 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1482 } else { 1483 /* learned new entry */ 1484 spin_lock(&vxlan->hash_lock); 1485 1486 /* close off race between vxlan_flush and incoming packets */ 1487 if (netif_running(dev)) 1488 vxlan_fdb_update(vxlan, src_mac, src_ip, 1489 NUD_REACHABLE, 1490 NLM_F_EXCL|NLM_F_CREATE, 1491 vxlan->cfg.dst_port, 1492 vni, 1493 vxlan->default_dst.remote_vni, 1494 ifindex, NTF_SELF, 0, true, NULL); 1495 spin_unlock(&vxlan->hash_lock); 1496 } 1497 1498 return SKB_NOT_DROPPED_YET; 1499 } 1500 1501 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1502 { 1503 ASSERT_RTNL(); 1504 1505 if (!vs) 1506 return false; 1507 if (!refcount_dec_and_test(&vs->refcnt)) 1508 return false; 1509 1510 hlist_del_rcu(&vs->hlist); 1511 udp_tunnel_notify_del_rx_port(vs->sk, 1512 (vs->flags & VXLAN_F_GPE) ? 1513 UDP_TUNNEL_TYPE_VXLAN_GPE : 1514 UDP_TUNNEL_TYPE_VXLAN); 1515 1516 return true; 1517 } 1518 1519 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1520 { 1521 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1522 #if IS_ENABLED(CONFIG_IPV6) 1523 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1524 1525 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1526 #endif 1527 1528 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1529 1530 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1531 vxlan_vs_del_vnigrp(vxlan); 1532 else 1533 vxlan_vs_del_dev(vxlan); 1534 1535 if (__vxlan_sock_release_prep(sock4)) { 1536 udp_tunnel_sock_release(sock4->sk); 1537 kfree_rcu(sock4, rcu); 1538 } 1539 1540 #if IS_ENABLED(CONFIG_IPV6) 1541 if (__vxlan_sock_release_prep(sock6)) { 1542 udp_tunnel_sock_release(sock6->sk); 1543 kfree_rcu(sock6, rcu); 1544 } 1545 #endif 1546 } 1547 1548 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags) 1549 { 1550 const struct vxlanhdr *vh = vxlan_hdr(skb); 1551 enum skb_drop_reason reason; 1552 size_t start, offset; 1553 1554 if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1555 return SKB_NOT_DROPPED_YET; 1556 1557 start = vxlan_rco_start(vh->vx_vni); 1558 offset = start + vxlan_rco_offset(vh->vx_vni); 1559 1560 reason = pskb_may_pull_reason(skb, offset + sizeof(u16)); 1561 if (reason) 1562 return reason; 1563 1564 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1565 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1566 return SKB_NOT_DROPPED_YET; 1567 } 1568 1569 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags, 1570 struct vxlan_metadata *md) 1571 { 1572 const struct vxlanhdr *vh = vxlan_hdr(skb); 1573 const struct vxlanhdr_gbp *gbp; 1574 struct metadata_dst *tun_dst; 1575 1576 gbp = (const struct vxlanhdr_gbp *)vh; 1577 1578 if (!(vh->vx_flags & VXLAN_HF_GBP)) 1579 return; 1580 1581 md->gbp = ntohs(gbp->policy_id); 1582 1583 tun_dst = (struct metadata_dst *)skb_dst(skb); 1584 if (tun_dst) { 1585 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1586 tun_dst->u.tun_info.key.tun_flags); 1587 tun_dst->u.tun_info.options_len = sizeof(*md); 1588 } 1589 if (gbp->dont_learn) 1590 md->gbp |= VXLAN_GBP_DONT_LEARN; 1591 1592 if (gbp->policy_applied) 1593 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1594 1595 /* In flow-based mode, GBP is carried in dst_metadata */ 1596 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1597 skb->mark = md->gbp; 1598 } 1599 1600 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan, 1601 struct vxlan_sock *vs, 1602 struct sk_buff *skb, __be32 vni) 1603 { 1604 union vxlan_addr saddr; 1605 u32 ifindex = skb->dev->ifindex; 1606 1607 skb_reset_mac_header(skb); 1608 skb->protocol = eth_type_trans(skb, vxlan->dev); 1609 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1610 1611 /* Ignore packet loops (and multicast echo) */ 1612 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1613 return SKB_DROP_REASON_LOCAL_MAC; 1614 1615 /* Get address from the outer IP header */ 1616 if (vxlan_get_sk_family(vs) == AF_INET) { 1617 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1618 saddr.sa.sa_family = AF_INET; 1619 #if IS_ENABLED(CONFIG_IPV6) 1620 } else { 1621 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1622 saddr.sa.sa_family = AF_INET6; 1623 #endif 1624 } 1625 1626 if (!(vxlan->cfg.flags & VXLAN_F_LEARN)) 1627 return SKB_NOT_DROPPED_YET; 1628 1629 return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, 1630 ifindex, vni); 1631 } 1632 1633 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1634 struct sk_buff *skb) 1635 { 1636 int err = 0; 1637 1638 if (vxlan_get_sk_family(vs) == AF_INET) 1639 err = IP_ECN_decapsulate(oiph, skb); 1640 #if IS_ENABLED(CONFIG_IPV6) 1641 else 1642 err = IP6_ECN_decapsulate(oiph, skb); 1643 #endif 1644 1645 if (unlikely(err) && log_ecn_error) { 1646 if (vxlan_get_sk_family(vs) == AF_INET) 1647 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1648 &((struct iphdr *)oiph)->saddr, 1649 ((struct iphdr *)oiph)->tos); 1650 else 1651 net_info_ratelimited("non-ECT from %pI6\n", 1652 &((struct ipv6hdr *)oiph)->saddr); 1653 } 1654 return err <= 1; 1655 } 1656 1657 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1658 { 1659 struct vxlan_vni_node *vninode = NULL; 1660 const struct vxlanhdr *vh; 1661 struct vxlan_dev *vxlan; 1662 struct vxlan_sock *vs; 1663 struct vxlan_metadata _md; 1664 struct vxlan_metadata *md = &_md; 1665 __be16 protocol = htons(ETH_P_TEB); 1666 enum skb_drop_reason reason; 1667 bool raw_proto = false; 1668 void *oiph; 1669 __be32 vni = 0; 1670 int nh; 1671 1672 /* Need UDP and VXLAN header to be present */ 1673 reason = pskb_may_pull_reason(skb, VXLAN_HLEN); 1674 if (reason) 1675 goto drop; 1676 1677 vh = vxlan_hdr(skb); 1678 /* VNI flag always required to be set */ 1679 if (!(vh->vx_flags & VXLAN_HF_VNI)) { 1680 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1681 ntohl(vh->vx_flags), ntohl(vh->vx_vni)); 1682 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1683 /* Return non vxlan pkt */ 1684 goto drop; 1685 } 1686 1687 vs = rcu_dereference_sk_user_data(sk); 1688 if (!vs) 1689 goto drop; 1690 1691 vni = vxlan_vni(vh->vx_vni); 1692 1693 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1694 if (!vxlan) { 1695 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND; 1696 goto drop; 1697 } 1698 1699 if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags || 1700 vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) { 1701 /* If the header uses bits besides those enabled by the 1702 * netdevice configuration, treat this as a malformed packet. 1703 * This behavior diverges from VXLAN RFC (RFC7348) which 1704 * stipulates that bits in reserved in reserved fields are to be 1705 * ignored. The approach here maintains compatibility with 1706 * previous stack code, and also is more robust and provides a 1707 * little more security in adding extensions to VXLAN. 1708 */ 1709 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1710 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1711 DEV_STATS_INC(vxlan->dev, rx_errors); 1712 vxlan_vnifilter_count(vxlan, vni, vninode, 1713 VXLAN_VNI_STATS_RX_ERRORS, 0); 1714 goto drop; 1715 } 1716 1717 if (vxlan->cfg.flags & VXLAN_F_GPE) { 1718 if (!vxlan_parse_gpe_proto(vh, &protocol)) 1719 goto drop; 1720 raw_proto = true; 1721 } 1722 1723 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1724 !net_eq(vxlan->net, dev_net(vxlan->dev)))) { 1725 reason = SKB_DROP_REASON_NOMEM; 1726 goto drop; 1727 } 1728 1729 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) { 1730 reason = vxlan_remcsum(skb, vxlan->cfg.flags); 1731 if (unlikely(reason)) 1732 goto drop; 1733 } 1734 1735 if (vxlan_collect_metadata(vs)) { 1736 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1737 struct metadata_dst *tun_dst; 1738 1739 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1740 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1741 key32_to_tunnel_id(vni), sizeof(*md)); 1742 1743 if (!tun_dst) { 1744 reason = SKB_DROP_REASON_NOMEM; 1745 goto drop; 1746 } 1747 1748 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1749 1750 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1751 } else { 1752 memset(md, 0, sizeof(*md)); 1753 } 1754 1755 if (vxlan->cfg.flags & VXLAN_F_GBP) 1756 vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md); 1757 /* Note that GBP and GPE can never be active together. This is 1758 * ensured in vxlan_dev_configure. 1759 */ 1760 1761 if (!raw_proto) { 1762 reason = vxlan_set_mac(vxlan, vs, skb, vni); 1763 if (reason) 1764 goto drop; 1765 } else { 1766 skb_reset_mac_header(skb); 1767 skb->dev = vxlan->dev; 1768 skb->pkt_type = PACKET_HOST; 1769 } 1770 1771 /* Save offset of outer header relative to skb->head, 1772 * because we are going to reset the network header to the inner header 1773 * and might change skb->head. 1774 */ 1775 nh = skb_network_header(skb) - skb->head; 1776 1777 skb_reset_network_header(skb); 1778 1779 reason = pskb_inet_may_pull_reason(skb); 1780 if (reason) { 1781 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1782 DEV_STATS_INC(vxlan->dev, rx_errors); 1783 vxlan_vnifilter_count(vxlan, vni, vninode, 1784 VXLAN_VNI_STATS_RX_ERRORS, 0); 1785 goto drop; 1786 } 1787 1788 /* Get the outer header. */ 1789 oiph = skb->head + nh; 1790 1791 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1792 reason = SKB_DROP_REASON_IP_TUNNEL_ECN; 1793 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1794 DEV_STATS_INC(vxlan->dev, rx_errors); 1795 vxlan_vnifilter_count(vxlan, vni, vninode, 1796 VXLAN_VNI_STATS_RX_ERRORS, 0); 1797 goto drop; 1798 } 1799 1800 rcu_read_lock(); 1801 1802 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1803 rcu_read_unlock(); 1804 dev_dstats_rx_dropped(vxlan->dev); 1805 vxlan_vnifilter_count(vxlan, vni, vninode, 1806 VXLAN_VNI_STATS_RX_DROPS, 0); 1807 reason = SKB_DROP_REASON_DEV_READY; 1808 goto drop; 1809 } 1810 1811 dev_dstats_rx_add(vxlan->dev, skb->len); 1812 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1813 gro_cells_receive(&vxlan->gro_cells, skb); 1814 1815 rcu_read_unlock(); 1816 1817 return 0; 1818 1819 drop: 1820 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED; 1821 /* Consume bad packet */ 1822 kfree_skb_reason(skb, reason); 1823 return 0; 1824 } 1825 1826 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1827 { 1828 struct vxlan_dev *vxlan; 1829 struct vxlan_sock *vs; 1830 struct vxlanhdr *hdr; 1831 __be32 vni; 1832 1833 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1834 return -EINVAL; 1835 1836 hdr = vxlan_hdr(skb); 1837 1838 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1839 return -EINVAL; 1840 1841 vs = rcu_dereference_sk_user_data(sk); 1842 if (!vs) 1843 return -ENOENT; 1844 1845 vni = vxlan_vni(hdr->vx_vni); 1846 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1847 if (!vxlan) 1848 return -ENOENT; 1849 1850 return 0; 1851 } 1852 1853 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1854 { 1855 struct neigh_table *tbl = arp_table(dev_net(dev)); 1856 struct vxlan_dev *vxlan = netdev_priv(dev); 1857 struct neighbour *n; 1858 struct arphdr *parp; 1859 u8 *arpptr, *sha; 1860 __be32 sip, tip; 1861 1862 if (dev->flags & IFF_NOARP) 1863 goto out; 1864 1865 if (!pskb_network_may_pull(skb, arp_hdr_len(dev))) { 1866 dev_dstats_tx_dropped(dev); 1867 vxlan_vnifilter_count(vxlan, vni, NULL, 1868 VXLAN_VNI_STATS_TX_DROPS, 0); 1869 goto out; 1870 } 1871 parp = arp_hdr(skb); 1872 1873 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1874 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1875 parp->ar_pro != htons(ETH_P_IP) || 1876 parp->ar_op != htons(ARPOP_REQUEST) || 1877 parp->ar_hln != dev->addr_len || 1878 parp->ar_pln != 4) 1879 goto out; 1880 arpptr = (u8 *)parp + sizeof(struct arphdr); 1881 sha = arpptr; 1882 arpptr += dev->addr_len; /* sha */ 1883 memcpy(&sip, arpptr, sizeof(sip)); 1884 arpptr += sizeof(sip); 1885 arpptr += dev->addr_len; /* tha */ 1886 memcpy(&tip, arpptr, sizeof(tip)); 1887 1888 if (ipv4_is_loopback(tip) || 1889 ipv4_is_multicast(tip)) 1890 goto out; 1891 1892 n = neigh_lookup(tbl, &tip, dev); 1893 1894 if (n) { 1895 struct vxlan_rdst *rdst = NULL; 1896 u8 ha[ETH_ALEN] __aligned(2); 1897 struct vxlan_fdb *f; 1898 struct sk_buff *reply; 1899 1900 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1901 neigh_release(n); 1902 goto out; 1903 } 1904 1905 neigh_ha_snapshot(ha, n, n->dev); 1906 1907 rcu_read_lock(); 1908 f = vxlan_find_mac_tx(vxlan, ha, vni); 1909 if (f) 1910 rdst = first_remote_rcu(f); 1911 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 1912 /* bridge-local neighbor */ 1913 neigh_release(n); 1914 rcu_read_unlock(); 1915 goto out; 1916 } 1917 rcu_read_unlock(); 1918 1919 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1920 ha, sha); 1921 1922 neigh_release(n); 1923 1924 if (reply == NULL) 1925 goto out; 1926 1927 skb_reset_mac_header(reply); 1928 __skb_pull(reply, skb_network_offset(reply)); 1929 reply->ip_summed = CHECKSUM_UNNECESSARY; 1930 reply->pkt_type = PACKET_HOST; 1931 1932 if (netif_rx(reply) == NET_RX_DROP) { 1933 dev_dstats_rx_dropped(dev); 1934 vxlan_vnifilter_count(vxlan, vni, NULL, 1935 VXLAN_VNI_STATS_RX_DROPS, 0); 1936 } 1937 1938 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1939 union vxlan_addr ipa = { 1940 .sin.sin_addr.s_addr = tip, 1941 .sin.sin_family = AF_INET, 1942 }; 1943 1944 vxlan_ip_miss(dev, &ipa); 1945 } 1946 out: 1947 consume_skb(skb); 1948 return NETDEV_TX_OK; 1949 } 1950 1951 #if IS_ENABLED(CONFIG_IPV6) 1952 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1953 struct neighbour *n, u8 *ha, 1954 bool isrouter) 1955 { 1956 struct net_device *dev = request->dev; 1957 struct sk_buff *reply; 1958 struct nd_msg *ns, *na; 1959 struct ipv6hdr *pip6; 1960 u8 *daddr; 1961 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1962 int headroom; 1963 int ns_olen; 1964 int i, len; 1965 1966 if (dev == NULL || !pskb_may_pull(request, request->len)) 1967 return NULL; 1968 1969 headroom = LL_RESERVED_SPACE(dev); 1970 len = headroom + sizeof(struct ipv6hdr) + 1971 sizeof(*na) + na_olen + dev->needed_tailroom; 1972 reply = alloc_skb(len, GFP_ATOMIC); 1973 if (reply == NULL) 1974 return NULL; 1975 1976 reply->protocol = htons(ETH_P_IPV6); 1977 reply->dev = dev; 1978 skb_reserve(reply, headroom); 1979 skb_push(reply, sizeof(struct ethhdr)); 1980 skb_reset_mac_header(reply); 1981 1982 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1983 1984 daddr = eth_hdr(request)->h_source; 1985 ns_olen = request->len - skb_network_offset(request) - 1986 sizeof(struct ipv6hdr) - sizeof(*ns); 1987 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1988 if (!ns->opt[i + 1] || i + (ns->opt[i + 1] << 3) > ns_olen) { 1989 kfree_skb(reply); 1990 return NULL; 1991 } 1992 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1993 if ((ns->opt[i + 1] << 3) >= 1994 sizeof(struct nd_opt_hdr) + ETH_ALEN) 1995 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1996 break; 1997 } 1998 } 1999 2000 /* Ethernet header */ 2001 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 2002 ether_addr_copy(eth_hdr(reply)->h_source, ha); 2003 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 2004 reply->protocol = htons(ETH_P_IPV6); 2005 2006 skb_pull(reply, sizeof(struct ethhdr)); 2007 skb_reset_network_header(reply); 2008 skb_put(reply, sizeof(struct ipv6hdr)); 2009 2010 /* IPv6 header */ 2011 2012 pip6 = ipv6_hdr(reply); 2013 memset(pip6, 0, sizeof(struct ipv6hdr)); 2014 pip6->version = 6; 2015 pip6->priority = ipv6_hdr(request)->priority; 2016 pip6->nexthdr = IPPROTO_ICMPV6; 2017 pip6->hop_limit = 255; 2018 pip6->daddr = ipv6_hdr(request)->saddr; 2019 pip6->saddr = *(struct in6_addr *)n->primary_key; 2020 2021 skb_pull(reply, sizeof(struct ipv6hdr)); 2022 skb_reset_transport_header(reply); 2023 2024 /* Neighbor Advertisement */ 2025 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2026 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2027 na->icmph.icmp6_router = isrouter; 2028 na->icmph.icmp6_override = 1; 2029 na->icmph.icmp6_solicited = 1; 2030 na->target = ns->target; 2031 ether_addr_copy(&na->opt[2], ha); 2032 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2033 na->opt[1] = na_olen >> 3; 2034 2035 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2036 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2037 csum_partial(na, sizeof(*na)+na_olen, 0)); 2038 2039 pip6->payload_len = htons(sizeof(*na)+na_olen); 2040 2041 skb_push(reply, sizeof(struct ipv6hdr)); 2042 2043 reply->ip_summed = CHECKSUM_UNNECESSARY; 2044 2045 return reply; 2046 } 2047 2048 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2049 { 2050 struct vxlan_dev *vxlan = netdev_priv(dev); 2051 const struct in6_addr *daddr; 2052 const struct ipv6hdr *iphdr; 2053 struct neighbour *n; 2054 struct nd_msg *msg; 2055 2056 rcu_read_lock(); 2057 if (unlikely(!ipv6_mod_enabled())) 2058 goto out; 2059 2060 iphdr = ipv6_hdr(skb); 2061 daddr = &iphdr->daddr; 2062 msg = (struct nd_msg *)(iphdr + 1); 2063 2064 if (ipv6_addr_loopback(daddr) || 2065 ipv6_addr_is_multicast(&msg->target)) 2066 goto out; 2067 2068 n = neigh_lookup(nd_table(dev_net(dev)), &msg->target, dev); 2069 2070 if (n) { 2071 struct vxlan_rdst *rdst = NULL; 2072 u8 ha[ETH_ALEN] __aligned(2); 2073 struct vxlan_fdb *f; 2074 struct sk_buff *reply; 2075 2076 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2077 neigh_release(n); 2078 goto out; 2079 } 2080 2081 neigh_ha_snapshot(ha, n, n->dev); 2082 f = vxlan_find_mac_tx(vxlan, ha, vni); 2083 if (f) 2084 rdst = first_remote_rcu(f); 2085 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 2086 /* bridge-local neighbor */ 2087 neigh_release(n); 2088 goto out; 2089 } 2090 2091 reply = vxlan_na_create(skb, n, ha, 2092 !!(f ? f->flags & NTF_ROUTER : 0)); 2093 2094 neigh_release(n); 2095 2096 if (reply == NULL) 2097 goto out; 2098 2099 if (netif_rx(reply) == NET_RX_DROP) { 2100 dev_dstats_rx_dropped(dev); 2101 vxlan_vnifilter_count(vxlan, vni, NULL, 2102 VXLAN_VNI_STATS_RX_DROPS, 0); 2103 } 2104 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2105 union vxlan_addr ipa = { 2106 .sin6.sin6_addr = msg->target, 2107 .sin6.sin6_family = AF_INET6, 2108 }; 2109 2110 vxlan_ip_miss(dev, &ipa); 2111 } 2112 2113 out: 2114 rcu_read_unlock(); 2115 consume_skb(skb); 2116 return NETDEV_TX_OK; 2117 } 2118 #endif 2119 2120 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2121 { 2122 struct vxlan_dev *vxlan = netdev_priv(dev); 2123 struct neigh_table *tbl; 2124 struct neighbour *n; 2125 2126 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2127 return false; 2128 2129 n = NULL; 2130 switch (ntohs(eth_hdr(skb)->h_proto)) { 2131 case ETH_P_IP: 2132 { 2133 struct iphdr *pip; 2134 2135 if (!pskb_network_may_pull(skb, sizeof(struct iphdr))) 2136 return false; 2137 2138 tbl = arp_table(dev_net(dev)); 2139 pip = ip_hdr(skb); 2140 n = neigh_lookup(tbl, &pip->daddr, dev); 2141 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2142 union vxlan_addr ipa = { 2143 .sin.sin_addr.s_addr = pip->daddr, 2144 .sin.sin_family = AF_INET, 2145 }; 2146 2147 vxlan_ip_miss(dev, &ipa); 2148 return false; 2149 } 2150 2151 break; 2152 } 2153 #if IS_ENABLED(CONFIG_IPV6) 2154 case ETH_P_IPV6: 2155 { 2156 struct ipv6hdr *pip6; 2157 2158 /* check if ipv6.disable=1 set during boot was set 2159 * during booting so nd_tbl is not initialized 2160 */ 2161 if (!ipv6_mod_enabled()) 2162 return false; 2163 if (!pskb_network_may_pull(skb, sizeof(struct ipv6hdr))) 2164 return false; 2165 2166 tbl = nd_table(dev_net(dev)); 2167 pip6 = ipv6_hdr(skb); 2168 n = neigh_lookup(tbl, &pip6->daddr, dev); 2169 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2170 union vxlan_addr ipa = { 2171 .sin6.sin6_addr = pip6->daddr, 2172 .sin6.sin6_family = AF_INET6, 2173 }; 2174 2175 vxlan_ip_miss(dev, &ipa); 2176 return false; 2177 } 2178 2179 break; 2180 } 2181 #endif 2182 default: 2183 return false; 2184 } 2185 2186 if (n) { 2187 u8 haddr[ETH_ALEN]; 2188 bool diff; 2189 2190 neigh_ha_snapshot(haddr, n, dev); 2191 diff = !ether_addr_equal_unaligned(eth_hdr(skb)->h_dest, haddr); 2192 if (diff) { 2193 if (skb_cow_head(skb, 0)) { 2194 neigh_release(n); 2195 return false; 2196 } 2197 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2198 dev->addr_len); 2199 memcpy(eth_hdr(skb)->h_dest, haddr, dev->addr_len); 2200 } 2201 neigh_release(n); 2202 return diff; 2203 } 2204 2205 return false; 2206 } 2207 2208 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2209 { 2210 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2211 2212 gpe->np_applied = 1; 2213 gpe->next_protocol = tun_p_from_eth_p(protocol); 2214 if (!gpe->next_protocol) 2215 return -EPFNOSUPPORT; 2216 return 0; 2217 } 2218 2219 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2220 int iphdr_len, __be32 vni, 2221 struct vxlan_metadata *md, u32 vxflags, 2222 bool udp_sum) 2223 { 2224 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2225 __be16 inner_protocol = htons(ETH_P_TEB); 2226 struct vxlanhdr *vxh; 2227 bool double_encap; 2228 int min_headroom; 2229 int err; 2230 2231 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2232 skb->ip_summed == CHECKSUM_PARTIAL) { 2233 int csum_start = skb_checksum_start_offset(skb); 2234 2235 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2236 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2237 (skb->csum_offset == offsetof(struct udphdr, check) || 2238 skb->csum_offset == offsetof(struct tcphdr, check))) 2239 type |= SKB_GSO_TUNNEL_REMCSUM; 2240 } 2241 2242 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2243 + VXLAN_HLEN + iphdr_len; 2244 2245 /* Need space for new headers (invalidates iph ptr) */ 2246 err = skb_cow_head(skb, min_headroom); 2247 if (unlikely(err)) 2248 return err; 2249 2250 double_encap = udp_tunnel_handle_partial(skb); 2251 err = iptunnel_handle_offloads(skb, type); 2252 if (err) 2253 return err; 2254 2255 vxh = __skb_push(skb, sizeof(*vxh)); 2256 vxh->vx_flags = VXLAN_HF_VNI; 2257 vxh->vx_vni = vxlan_vni_field(vni); 2258 2259 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2260 unsigned int start; 2261 2262 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2263 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2264 vxh->vx_flags |= VXLAN_HF_RCO; 2265 2266 if (!skb_is_gso(skb)) { 2267 skb->ip_summed = CHECKSUM_NONE; 2268 skb->encapsulation = 0; 2269 } 2270 } 2271 2272 if (vxflags & VXLAN_F_GBP) 2273 vxlan_build_gbp_hdr(vxh, md); 2274 if (vxflags & VXLAN_F_GPE) { 2275 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2276 if (err < 0) 2277 return err; 2278 inner_protocol = skb->protocol; 2279 } 2280 2281 udp_tunnel_set_inner_protocol(skb, double_encap, inner_protocol); 2282 return 0; 2283 } 2284 2285 /* Bypass encapsulation if the destination is local */ 2286 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2287 struct vxlan_dev *dst_vxlan, __be32 vni, 2288 bool snoop) 2289 { 2290 union vxlan_addr loopback; 2291 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2292 unsigned int len = skb->len; 2293 struct net_device *dev; 2294 2295 skb->pkt_type = PACKET_HOST; 2296 skb->encapsulation = 0; 2297 skb->dev = dst_vxlan->dev; 2298 __skb_pull(skb, skb_network_offset(skb)); 2299 2300 if (remote_ip->sa.sa_family == AF_INET) { 2301 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2302 loopback.sa.sa_family = AF_INET; 2303 #if IS_ENABLED(CONFIG_IPV6) 2304 } else { 2305 loopback.sin6.sin6_addr = in6addr_loopback; 2306 loopback.sa.sa_family = AF_INET6; 2307 #endif 2308 } 2309 2310 rcu_read_lock(); 2311 dev = skb->dev; 2312 if (unlikely(!(dev->flags & IFF_UP))) { 2313 kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY); 2314 goto drop; 2315 } 2316 2317 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2318 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2319 2320 dev_dstats_tx_add(src_vxlan->dev, len); 2321 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2322 2323 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2324 dev_dstats_rx_add(dst_vxlan->dev, len); 2325 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2326 len); 2327 } else { 2328 drop: 2329 dev_dstats_rx_dropped(dev); 2330 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2331 VXLAN_VNI_STATS_RX_DROPS, 0); 2332 } 2333 rcu_read_unlock(); 2334 } 2335 2336 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2337 struct vxlan_dev *vxlan, 2338 int addr_family, 2339 __be16 dst_port, int dst_ifindex, __be32 vni, 2340 struct dst_entry *dst, 2341 u32 rt_flags) 2342 { 2343 #if IS_ENABLED(CONFIG_IPV6) 2344 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2345 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2346 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2347 */ 2348 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2349 #endif 2350 /* Bypass encapsulation if the destination is local */ 2351 if (rt_flags & RTCF_LOCAL && 2352 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2353 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2354 struct vxlan_dev *dst_vxlan; 2355 2356 dst_release(dst); 2357 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2358 addr_family, dst_port, 2359 vxlan->cfg.flags); 2360 if (!dst_vxlan) { 2361 DEV_STATS_INC(dev, tx_errors); 2362 vxlan_vnifilter_count(vxlan, vni, NULL, 2363 VXLAN_VNI_STATS_TX_ERRORS, 0); 2364 kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND); 2365 2366 return -ENOENT; 2367 } 2368 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2369 return 1; 2370 } 2371 2372 return 0; 2373 } 2374 2375 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2376 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2377 { 2378 struct dst_cache *dst_cache; 2379 struct ip_tunnel_info *info; 2380 struct ip_tunnel_key *pkey; 2381 struct ip_tunnel_key key; 2382 struct vxlan_dev *vxlan = netdev_priv(dev); 2383 const struct iphdr *old_iph; 2384 struct vxlan_metadata _md = {}; 2385 struct vxlan_metadata *md = &_md; 2386 unsigned int pkt_len = skb->len; 2387 __be16 src_port = 0, dst_port; 2388 struct dst_entry *ndst = NULL; 2389 int addr_family; 2390 __u8 tos, ttl; 2391 int ifindex; 2392 int err = 0; 2393 u32 flags = vxlan->cfg.flags; 2394 bool use_cache; 2395 bool udp_sum = false; 2396 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2397 enum skb_drop_reason reason; 2398 bool no_eth_encap; 2399 __be32 vni = 0; 2400 2401 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2402 reason = skb_vlan_inet_prepare(skb, no_eth_encap); 2403 if (reason) 2404 goto drop; 2405 2406 reason = SKB_DROP_REASON_NOT_SPECIFIED; 2407 old_iph = ip_hdr(skb); 2408 2409 info = skb_tunnel_info(skb); 2410 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2411 2412 if (rdst) { 2413 memset(&key, 0, sizeof(key)); 2414 pkey = &key; 2415 2416 if (vxlan_addr_any(&rdst->remote_ip)) { 2417 if (did_rsc) { 2418 /* short-circuited back to local bridge */ 2419 vxlan_encap_bypass(skb, vxlan, vxlan, 2420 default_vni, true); 2421 return; 2422 } 2423 goto drop; 2424 } 2425 2426 addr_family = vxlan->cfg.saddr.sa.sa_family; 2427 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2428 vni = (rdst->remote_vni) ? : default_vni; 2429 ifindex = rdst->remote_ifindex; 2430 2431 if (addr_family == AF_INET) { 2432 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2433 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2434 } else { 2435 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2436 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2437 } 2438 2439 dst_cache = &rdst->dst_cache; 2440 md->gbp = skb->mark; 2441 if (flags & VXLAN_F_TTL_INHERIT) { 2442 ttl = ip_tunnel_get_ttl(old_iph, skb); 2443 } else { 2444 ttl = vxlan->cfg.ttl; 2445 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2446 ttl = 1; 2447 } 2448 tos = vxlan->cfg.tos; 2449 if (tos == 1) 2450 tos = ip_tunnel_get_dsfield(old_iph, skb); 2451 if (tos && !info) 2452 use_cache = false; 2453 2454 if (addr_family == AF_INET) 2455 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2456 else 2457 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2458 #if IS_ENABLED(CONFIG_IPV6) 2459 switch (vxlan->cfg.label_policy) { 2460 case VXLAN_LABEL_FIXED: 2461 key.label = vxlan->cfg.label; 2462 break; 2463 case VXLAN_LABEL_INHERIT: 2464 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2465 break; 2466 default: 2467 DEBUG_NET_WARN_ON_ONCE(1); 2468 goto drop; 2469 } 2470 #endif 2471 } else { 2472 if (!info) { 2473 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2474 dev->name); 2475 goto drop; 2476 } 2477 pkey = &info->key; 2478 addr_family = ip_tunnel_info_af(info); 2479 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2480 vni = tunnel_id_to_key32(info->key.tun_id); 2481 ifindex = 0; 2482 dst_cache = &info->dst_cache; 2483 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2484 if (info->options_len < sizeof(*md)) 2485 goto drop; 2486 md = ip_tunnel_info_opts(info); 2487 } 2488 ttl = info->key.ttl; 2489 tos = info->key.tos; 2490 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2491 } 2492 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2493 vxlan->cfg.port_max, true); 2494 2495 rcu_read_lock(); 2496 if (addr_family == AF_INET) { 2497 struct vxlan_sock *sock4; 2498 u16 ipcb_flags = 0; 2499 struct rtable *rt; 2500 __be16 df = 0; 2501 __be32 saddr; 2502 2503 sock4 = rcu_dereference(vxlan->vn4_sock); 2504 if (unlikely(!sock4)) { 2505 reason = SKB_DROP_REASON_DEV_READY; 2506 goto tx_error; 2507 } 2508 2509 if (!ifindex) 2510 ifindex = sock4->sk->sk_bound_dev_if; 2511 2512 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2513 &saddr, pkey, src_port, dst_port, 2514 tos, use_cache ? dst_cache : NULL); 2515 if (IS_ERR(rt)) { 2516 err = PTR_ERR(rt); 2517 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2518 goto tx_error; 2519 } 2520 2521 if (flags & VXLAN_F_MC_ROUTE) 2522 ipcb_flags |= IPSKB_MCROUTE; 2523 2524 if (!info) { 2525 /* Bypass encapsulation if the destination is local */ 2526 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2527 dst_port, ifindex, vni, 2528 &rt->dst, rt->rt_flags); 2529 if (err) 2530 goto out_unlock; 2531 2532 if (vxlan->cfg.df == VXLAN_DF_SET) { 2533 df = htons(IP_DF); 2534 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2535 struct ethhdr *eth = eth_hdr(skb); 2536 2537 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2538 (ntohs(eth->h_proto) == ETH_P_IP && 2539 old_iph->frag_off & htons(IP_DF))) 2540 df = htons(IP_DF); 2541 } 2542 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2543 info->key.tun_flags)) { 2544 df = htons(IP_DF); 2545 } 2546 2547 ndst = &rt->dst; 2548 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2549 netif_is_any_bridge_port(dev)); 2550 if (err < 0) { 2551 goto tx_error; 2552 } else if (err) { 2553 if (info) { 2554 struct ip_tunnel_info *unclone; 2555 2556 unclone = skb_tunnel_info_unclone(skb); 2557 if (unlikely(!unclone)) 2558 goto tx_error; 2559 2560 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2561 unclone->key.u.ipv4.dst = saddr; 2562 } 2563 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2564 dst_release(ndst); 2565 goto out_unlock; 2566 } 2567 2568 tos = ip_tunnel_ecn_encap(tos, ip_hdr(skb), skb); 2569 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2570 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2571 vni, md, flags, udp_sum); 2572 if (err < 0) { 2573 reason = SKB_DROP_REASON_NOMEM; 2574 goto tx_error; 2575 } 2576 2577 udp_tunnel_xmit_skb(rt, sock4->sk, skb, saddr, 2578 pkey->u.ipv4.dst, tos, ttl, df, 2579 src_port, dst_port, xnet, !udp_sum, 2580 ipcb_flags); 2581 #if IS_ENABLED(CONFIG_IPV6) 2582 } else { 2583 struct vxlan_sock *sock6; 2584 struct in6_addr saddr; 2585 u16 ip6cb_flags = 0; 2586 2587 sock6 = rcu_dereference(vxlan->vn6_sock); 2588 if (unlikely(!sock6)) { 2589 reason = SKB_DROP_REASON_DEV_READY; 2590 goto tx_error; 2591 } 2592 2593 if (!ifindex) 2594 ifindex = sock6->sk->sk_bound_dev_if; 2595 2596 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sk, 2597 ifindex, &saddr, pkey, 2598 src_port, dst_port, tos, 2599 use_cache ? dst_cache : NULL); 2600 if (IS_ERR(ndst)) { 2601 err = PTR_ERR(ndst); 2602 ndst = NULL; 2603 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2604 goto tx_error; 2605 } 2606 2607 if (flags & VXLAN_F_MC_ROUTE) 2608 ip6cb_flags |= IP6SKB_MCROUTE; 2609 2610 if (!info) { 2611 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2612 2613 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2614 dst_port, ifindex, vni, 2615 ndst, rt6i_flags); 2616 if (err) 2617 goto out_unlock; 2618 } 2619 2620 err = skb_tunnel_check_pmtu(skb, ndst, 2621 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2622 netif_is_any_bridge_port(dev)); 2623 if (err < 0) { 2624 goto tx_error; 2625 } else if (err) { 2626 if (info) { 2627 struct ip_tunnel_info *unclone; 2628 2629 unclone = skb_tunnel_info_unclone(skb); 2630 if (unlikely(!unclone)) 2631 goto tx_error; 2632 2633 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2634 unclone->key.u.ipv6.dst = saddr; 2635 } 2636 2637 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2638 dst_release(ndst); 2639 goto out_unlock; 2640 } 2641 2642 tos = ip_tunnel_ecn_encap(tos, ip_hdr(skb), skb); 2643 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2644 skb_scrub_packet(skb, xnet); 2645 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2646 vni, md, flags, udp_sum); 2647 if (err < 0) { 2648 reason = SKB_DROP_REASON_NOMEM; 2649 goto tx_error; 2650 } 2651 2652 udp_tunnel6_xmit_skb(ndst, sock6->sk, skb, dev, 2653 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2654 pkey->label, src_port, dst_port, !udp_sum, 2655 ip6cb_flags); 2656 #endif 2657 } 2658 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2659 out_unlock: 2660 rcu_read_unlock(); 2661 return; 2662 2663 drop: 2664 dev_dstats_tx_dropped(dev); 2665 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2666 kfree_skb_reason(skb, reason); 2667 return; 2668 2669 tx_error: 2670 rcu_read_unlock(); 2671 if (err == -ELOOP) 2672 DEV_STATS_INC(dev, collisions); 2673 else if (err == -ENETUNREACH) 2674 DEV_STATS_INC(dev, tx_carrier_errors); 2675 dst_release(ndst); 2676 DEV_STATS_INC(dev, tx_errors); 2677 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2678 kfree_skb_reason(skb, reason); 2679 } 2680 2681 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2682 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2683 { 2684 struct vxlan_rdst nh_rdst; 2685 struct nexthop *nh; 2686 bool do_xmit; 2687 u32 hash; 2688 2689 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2690 hash = skb_get_hash(skb); 2691 2692 nh = rcu_dereference(f->nh); 2693 if (!nh) 2694 goto drop; 2695 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2696 2697 if (likely(do_xmit)) 2698 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2699 else 2700 goto drop; 2701 2702 return; 2703 2704 drop: 2705 dev_dstats_tx_dropped(dev); 2706 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2707 VXLAN_VNI_STATS_TX_DROPS, 0); 2708 dev_kfree_skb(skb); 2709 } 2710 2711 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2712 u32 nhid, __be32 vni) 2713 { 2714 struct vxlan_dev *vxlan = netdev_priv(dev); 2715 struct vxlan_rdst nh_rdst; 2716 struct nexthop *nh; 2717 bool do_xmit; 2718 u32 hash; 2719 2720 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2721 hash = skb_get_hash(skb); 2722 2723 rcu_read_lock(); 2724 nh = nexthop_find_by_id(dev_net(dev), nhid); 2725 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2726 rcu_read_unlock(); 2727 goto drop; 2728 } 2729 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2730 rcu_read_unlock(); 2731 2732 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2733 goto drop; 2734 2735 if (likely(do_xmit)) 2736 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2737 else 2738 goto drop; 2739 2740 return NETDEV_TX_OK; 2741 2742 drop: 2743 dev_dstats_tx_dropped(dev); 2744 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2745 VXLAN_VNI_STATS_TX_DROPS, 0); 2746 dev_kfree_skb(skb); 2747 return NETDEV_TX_OK; 2748 } 2749 2750 /* Transmit local packets over Vxlan 2751 * 2752 * Outer IP header inherits ECN and DF from inner header. 2753 * Outer UDP destination is the VXLAN assigned port. 2754 * source port is based on hash of flow 2755 */ 2756 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2757 { 2758 struct vxlan_dev *vxlan = netdev_priv(dev); 2759 struct vxlan_rdst *rdst, *fdst = NULL; 2760 const struct ip_tunnel_info *info; 2761 struct vxlan_fdb *f; 2762 struct ethhdr *eth; 2763 __be32 vni = 0; 2764 u32 nhid = 0; 2765 bool did_rsc; 2766 2767 info = skb_tunnel_info(skb); 2768 2769 skb_reset_mac_header(skb); 2770 2771 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2772 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2773 info->mode & IP_TUNNEL_INFO_TX) { 2774 vni = tunnel_id_to_key32(info->key.tun_id); 2775 nhid = info->key.nhid; 2776 } else { 2777 if (info && info->mode & IP_TUNNEL_INFO_TX) 2778 vxlan_xmit_one(skb, dev, vni, NULL, false); 2779 else 2780 kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO); 2781 return NETDEV_TX_OK; 2782 } 2783 } 2784 2785 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2786 eth = eth_hdr(skb); 2787 if (ntohs(eth->h_proto) == ETH_P_ARP) 2788 return arp_reduce(dev, skb, vni); 2789 #if IS_ENABLED(CONFIG_IPV6) 2790 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2791 pskb_network_may_pull(skb, sizeof(struct ipv6hdr) + 2792 sizeof(struct nd_msg)) && 2793 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2794 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2795 2796 if (m->icmph.icmp6_code == 0 && 2797 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2798 return neigh_reduce(dev, skb, vni); 2799 } 2800 #endif 2801 } 2802 2803 if (nhid) 2804 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2805 2806 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2807 struct vxlan_mdb_entry *mdb_entry; 2808 2809 rcu_read_lock(); 2810 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2811 if (mdb_entry) { 2812 netdev_tx_t ret; 2813 2814 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2815 rcu_read_unlock(); 2816 return ret; 2817 } 2818 rcu_read_unlock(); 2819 } 2820 2821 eth = eth_hdr(skb); 2822 rcu_read_lock(); 2823 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2824 did_rsc = false; 2825 2826 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2827 (ntohs(eth->h_proto) == ETH_P_IP || 2828 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2829 did_rsc = route_shortcircuit(dev, skb); 2830 eth = eth_hdr(skb); 2831 if (did_rsc) 2832 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2833 } 2834 2835 if (f == NULL) { 2836 f = vxlan_find_mac_tx(vxlan, all_zeros_mac, vni); 2837 if (f == NULL) { 2838 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2839 !is_multicast_ether_addr(eth->h_dest)) 2840 vxlan_fdb_miss(vxlan, eth->h_dest); 2841 2842 dev_dstats_tx_dropped(dev); 2843 vxlan_vnifilter_count(vxlan, vni, NULL, 2844 VXLAN_VNI_STATS_TX_DROPS, 0); 2845 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2846 goto out; 2847 } 2848 } 2849 2850 if (rcu_access_pointer(f->nh)) { 2851 vxlan_xmit_nh(skb, dev, f, 2852 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2853 } else { 2854 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2855 struct sk_buff *skb1; 2856 2857 if (!fdst) { 2858 fdst = rdst; 2859 continue; 2860 } 2861 skb1 = skb_clone(skb, GFP_ATOMIC); 2862 if (skb1) 2863 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2864 } 2865 if (fdst) 2866 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2867 else 2868 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2869 } 2870 2871 out: 2872 rcu_read_unlock(); 2873 return NETDEV_TX_OK; 2874 } 2875 2876 /* Walk the forwarding table and purge stale entries */ 2877 static void vxlan_cleanup(struct timer_list *t) 2878 { 2879 struct vxlan_dev *vxlan = timer_container_of(vxlan, t, age_timer); 2880 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2881 struct vxlan_fdb *f; 2882 2883 if (!netif_running(vxlan->dev)) 2884 return; 2885 2886 rcu_read_lock(); 2887 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 2888 unsigned long timeout; 2889 2890 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2891 continue; 2892 2893 if (f->flags & NTF_EXT_LEARNED) 2894 continue; 2895 2896 timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ; 2897 if (time_before_eq(timeout, jiffies)) { 2898 spin_lock(&vxlan->hash_lock); 2899 if (!hlist_unhashed(&f->fdb_node)) { 2900 netdev_dbg(vxlan->dev, "garbage collect %pM\n", 2901 f->key.eth_addr); 2902 f->state = NUD_STALE; 2903 vxlan_fdb_destroy(vxlan, f, true, true); 2904 } 2905 spin_unlock(&vxlan->hash_lock); 2906 } else if (time_before(timeout, next_timer)) { 2907 next_timer = timeout; 2908 } 2909 } 2910 rcu_read_unlock(); 2911 2912 mod_timer(&vxlan->age_timer, next_timer); 2913 } 2914 2915 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2916 { 2917 ASSERT_RTNL(); 2918 2919 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2920 #if IS_ENABLED(CONFIG_IPV6) 2921 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2922 #endif 2923 } 2924 2925 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2926 struct vxlan_dev_node *node) 2927 { 2928 __be32 vni = vxlan->default_dst.remote_vni; 2929 2930 ASSERT_RTNL(); 2931 2932 node->vxlan = vxlan; 2933 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2934 } 2935 2936 /* Setup stats when device is created */ 2937 static int vxlan_init(struct net_device *dev) 2938 { 2939 struct vxlan_dev *vxlan = netdev_priv(dev); 2940 int err; 2941 2942 err = rhashtable_init(&vxlan->fdb_hash_tbl, &vxlan_fdb_rht_params); 2943 if (err) 2944 return err; 2945 2946 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 2947 err = vxlan_vnigroup_init(vxlan); 2948 if (err) 2949 goto err_rhashtable_destroy; 2950 } 2951 2952 err = gro_cells_init(&vxlan->gro_cells, dev); 2953 if (err) 2954 goto err_vnigroup_uninit; 2955 2956 err = vxlan_mdb_init(vxlan); 2957 if (err) 2958 goto err_gro_cells_destroy; 2959 2960 netdev_lockdep_set_classes(dev); 2961 return 0; 2962 2963 err_gro_cells_destroy: 2964 gro_cells_destroy(&vxlan->gro_cells); 2965 err_vnigroup_uninit: 2966 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2967 vxlan_vnigroup_uninit(vxlan); 2968 err_rhashtable_destroy: 2969 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2970 return err; 2971 } 2972 2973 static void vxlan_uninit(struct net_device *dev) 2974 { 2975 struct vxlan_dev *vxlan = netdev_priv(dev); 2976 2977 vxlan_mdb_fini(vxlan); 2978 2979 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2980 vxlan_vnigroup_uninit(vxlan); 2981 2982 gro_cells_destroy(&vxlan->gro_cells); 2983 2984 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2985 } 2986 2987 /* Start ageing timer and join group when device is brought up */ 2988 static int vxlan_open(struct net_device *dev) 2989 { 2990 struct vxlan_dev *vxlan = netdev_priv(dev); 2991 int ret; 2992 2993 ret = vxlan_sock_add(vxlan); 2994 if (ret < 0) 2995 return ret; 2996 2997 ret = vxlan_multicast_join(vxlan); 2998 if (ret) { 2999 vxlan_sock_release(vxlan); 3000 return ret; 3001 } 3002 3003 if (vxlan->cfg.age_interval) 3004 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 3005 3006 return ret; 3007 } 3008 3009 struct vxlan_fdb_flush_desc { 3010 bool ignore_default_entry; 3011 unsigned long state; 3012 unsigned long state_mask; 3013 unsigned long flags; 3014 unsigned long flags_mask; 3015 __be32 src_vni; 3016 u32 nhid; 3017 __be32 vni; 3018 __be16 port; 3019 union vxlan_addr dst_ip; 3020 }; 3021 3022 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 3023 const struct vxlan_dev *vxlan) 3024 { 3025 return is_zero_ether_addr(f->key.eth_addr) && 3026 f->key.vni == vxlan->cfg.vni; 3027 } 3028 3029 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 3030 { 3031 struct nexthop *nh = rtnl_dereference(f->nh); 3032 3033 return nh && nh->id == nhid; 3034 } 3035 3036 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 3037 const struct vxlan_dev *vxlan, 3038 const struct vxlan_fdb_flush_desc *desc) 3039 { 3040 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 3041 return false; 3042 3043 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 3044 return false; 3045 3046 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 3047 return false; 3048 3049 if (desc->src_vni && f->key.vni != desc->src_vni) 3050 return false; 3051 3052 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 3053 return false; 3054 3055 return true; 3056 } 3057 3058 static bool 3059 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 3060 { 3061 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 3062 } 3063 3064 static bool 3065 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3066 const struct vxlan_rdst *rd) 3067 { 3068 if (desc->vni && rd->remote_vni != desc->vni) 3069 return false; 3070 3071 if (desc->port && rd->remote_port != desc->port) 3072 return false; 3073 3074 if (desc->dst_ip.sa.sa_family && 3075 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3076 return false; 3077 3078 return true; 3079 } 3080 3081 static void 3082 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3083 const struct vxlan_fdb_flush_desc *desc, 3084 bool *p_destroy_fdb) 3085 { 3086 struct vxlan_rdst *rd, *tmp; 3087 3088 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3089 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3090 continue; 3091 3092 if (list_is_singular(&f->remotes)) { 3093 *p_destroy_fdb = true; 3094 return; 3095 } 3096 3097 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3098 } 3099 } 3100 3101 /* Purge the forwarding table */ 3102 static void vxlan_flush(struct vxlan_dev *vxlan, 3103 const struct vxlan_fdb_flush_desc *desc) 3104 { 3105 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3106 struct vxlan_fdb *f; 3107 3108 rcu_read_lock(); 3109 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 3110 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3111 continue; 3112 3113 spin_lock_bh(&vxlan->hash_lock); 3114 if (hlist_unhashed(&f->fdb_node)) 3115 goto unlock; 3116 3117 if (match_remotes) { 3118 bool destroy_fdb = false; 3119 3120 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3121 &destroy_fdb); 3122 3123 if (!destroy_fdb) 3124 goto unlock; 3125 } 3126 3127 vxlan_fdb_destroy(vxlan, f, true, true); 3128 unlock: 3129 spin_unlock_bh(&vxlan->hash_lock); 3130 } 3131 rcu_read_unlock(); 3132 } 3133 3134 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3135 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3136 [NDA_NH_ID] = { .type = NLA_U32 }, 3137 [NDA_VNI] = { .type = NLA_U32 }, 3138 [NDA_PORT] = { .type = NLA_U16 }, 3139 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3140 sizeof(struct in6_addr)), 3141 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3142 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3143 }; 3144 3145 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3146 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3147 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3148 NTF_ROUTER) 3149 3150 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3151 struct netlink_ext_ack *extack) 3152 { 3153 struct vxlan_dev *vxlan = netdev_priv(dev); 3154 struct vxlan_fdb_flush_desc desc = {}; 3155 struct ndmsg *ndm = nlmsg_data(nlh); 3156 struct nlattr *tb[NDA_MAX + 1]; 3157 u8 ndm_flags; 3158 int err; 3159 3160 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3161 3162 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3163 extack); 3164 if (err) 3165 return err; 3166 3167 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3168 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3169 return -EINVAL; 3170 } 3171 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3172 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3173 return -EINVAL; 3174 } 3175 3176 desc.state = ndm->ndm_state; 3177 desc.flags = ndm_flags; 3178 3179 if (tb[NDA_NDM_STATE_MASK]) 3180 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3181 3182 if (tb[NDA_NDM_FLAGS_MASK]) 3183 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3184 3185 if (tb[NDA_SRC_VNI]) 3186 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3187 3188 if (tb[NDA_NH_ID]) 3189 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3190 3191 if (tb[NDA_VNI]) 3192 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3193 3194 if (tb[NDA_PORT]) 3195 desc.port = nla_get_be16(tb[NDA_PORT]); 3196 3197 if (tb[NDA_DST]) { 3198 union vxlan_addr ip; 3199 3200 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3201 if (err) { 3202 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3203 "Unsupported address family"); 3204 return err; 3205 } 3206 desc.dst_ip = ip; 3207 } 3208 3209 vxlan_flush(vxlan, &desc); 3210 3211 return 0; 3212 } 3213 3214 /* Cleanup timer and forwarding table on shutdown */ 3215 static int vxlan_stop(struct net_device *dev) 3216 { 3217 struct vxlan_dev *vxlan = netdev_priv(dev); 3218 struct vxlan_fdb_flush_desc desc = { 3219 /* Default entry is deleted at vxlan_dellink. */ 3220 .ignore_default_entry = true, 3221 .state = 0, 3222 .state_mask = NUD_PERMANENT | NUD_NOARP, 3223 }; 3224 3225 vxlan_multicast_leave(vxlan); 3226 3227 timer_delete_sync(&vxlan->age_timer); 3228 3229 vxlan_flush(vxlan, &desc); 3230 vxlan_sock_release(vxlan); 3231 3232 return 0; 3233 } 3234 3235 /* Stub, nothing needs to be done. */ 3236 static void vxlan_set_multicast_list(struct net_device *dev) 3237 { 3238 } 3239 3240 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3241 { 3242 struct vxlan_dev *vxlan = netdev_priv(dev); 3243 struct vxlan_rdst *dst = &vxlan->default_dst; 3244 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3245 dst->remote_ifindex); 3246 3247 /* This check is different than dev->max_mtu, because it looks at 3248 * the lowerdev->mtu, rather than the static dev->max_mtu 3249 */ 3250 if (lowerdev) { 3251 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3252 if (new_mtu > max_mtu) 3253 return -EINVAL; 3254 } 3255 3256 WRITE_ONCE(dev->mtu, new_mtu); 3257 return 0; 3258 } 3259 3260 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3261 { 3262 struct vxlan_dev *vxlan = netdev_priv(dev); 3263 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3264 __be16 sport, dport; 3265 3266 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3267 vxlan->cfg.port_max, true); 3268 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3269 3270 if (ip_tunnel_info_af(info) == AF_INET) { 3271 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3272 struct rtable *rt; 3273 3274 if (!sock4) 3275 return -EIO; 3276 3277 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3278 &info->key.u.ipv4.src, 3279 &info->key, 3280 sport, dport, info->key.tos, 3281 &info->dst_cache); 3282 if (IS_ERR(rt)) 3283 return PTR_ERR(rt); 3284 ip_rt_put(rt); 3285 } else { 3286 #if IS_ENABLED(CONFIG_IPV6) 3287 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3288 struct dst_entry *ndst; 3289 3290 if (!sock6) 3291 return -EIO; 3292 3293 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sk, 3294 0, &info->key.u.ipv6.src, 3295 &info->key, 3296 sport, dport, info->key.tos, 3297 &info->dst_cache); 3298 if (IS_ERR(ndst)) 3299 return PTR_ERR(ndst); 3300 dst_release(ndst); 3301 #else /* !CONFIG_IPV6 */ 3302 return -EPFNOSUPPORT; 3303 #endif 3304 } 3305 info->key.tp_src = sport; 3306 info->key.tp_dst = dport; 3307 return 0; 3308 } 3309 3310 static const struct net_device_ops vxlan_netdev_ether_ops = { 3311 .ndo_init = vxlan_init, 3312 .ndo_uninit = vxlan_uninit, 3313 .ndo_open = vxlan_open, 3314 .ndo_stop = vxlan_stop, 3315 .ndo_start_xmit = vxlan_xmit, 3316 .ndo_set_rx_mode = vxlan_set_multicast_list, 3317 .ndo_change_mtu = vxlan_change_mtu, 3318 .ndo_validate_addr = eth_validate_addr, 3319 .ndo_set_mac_address = eth_mac_addr, 3320 .ndo_fdb_add = vxlan_fdb_add, 3321 .ndo_fdb_del = vxlan_fdb_delete, 3322 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3323 .ndo_fdb_dump = vxlan_fdb_dump, 3324 .ndo_fdb_get = vxlan_fdb_get, 3325 .ndo_mdb_add = vxlan_mdb_add, 3326 .ndo_mdb_del = vxlan_mdb_del, 3327 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3328 .ndo_mdb_dump = vxlan_mdb_dump, 3329 .ndo_mdb_get = vxlan_mdb_get, 3330 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3331 }; 3332 3333 static const struct net_device_ops vxlan_netdev_raw_ops = { 3334 .ndo_init = vxlan_init, 3335 .ndo_uninit = vxlan_uninit, 3336 .ndo_open = vxlan_open, 3337 .ndo_stop = vxlan_stop, 3338 .ndo_start_xmit = vxlan_xmit, 3339 .ndo_change_mtu = vxlan_change_mtu, 3340 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3341 }; 3342 3343 /* Info for udev, that this is a virtual tunnel endpoint */ 3344 static const struct device_type vxlan_type = { 3345 .name = "vxlan", 3346 }; 3347 3348 /* Calls the ndo_udp_tunnel_add of the caller in order to 3349 * supply the listening VXLAN udp ports. Callers are expected 3350 * to implement the ndo_udp_tunnel_add. 3351 */ 3352 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3353 { 3354 struct vxlan_sock *vs; 3355 struct net *net = dev_net(dev); 3356 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3357 unsigned int i; 3358 3359 ASSERT_RTNL(); 3360 3361 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3362 hlist_for_each_entry(vs, &vn->sock_list[i], hlist) { 3363 unsigned short type; 3364 3365 if (vs->flags & VXLAN_F_GPE) 3366 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3367 else 3368 type = UDP_TUNNEL_TYPE_VXLAN; 3369 3370 if (push) 3371 udp_tunnel_push_rx_port(dev, vs->sk, type); 3372 else 3373 udp_tunnel_drop_rx_port(dev, vs->sk, type); 3374 } 3375 } 3376 } 3377 3378 /* Initialize the device structure. */ 3379 static void vxlan_setup(struct net_device *dev) 3380 { 3381 struct vxlan_dev *vxlan = netdev_priv(dev); 3382 3383 eth_hw_addr_random(dev); 3384 ether_setup(dev); 3385 3386 dev->needs_free_netdev = true; 3387 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3388 3389 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3390 dev->features |= NETIF_F_RXCSUM; 3391 dev->features |= NETIF_F_GSO_SOFTWARE; 3392 3393 /* Partial features are disabled by default. */ 3394 dev->vlan_features = dev->features; 3395 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3396 dev->hw_features |= NETIF_F_RXCSUM; 3397 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3398 dev->hw_features |= UDP_TUNNEL_PARTIAL_FEATURES; 3399 dev->hw_features |= NETIF_F_GSO_PARTIAL; 3400 3401 dev->hw_enc_features = dev->hw_features; 3402 dev->gso_partial_features = UDP_TUNNEL_PARTIAL_FEATURES; 3403 dev->mangleid_features = NETIF_F_GSO_PARTIAL; 3404 3405 netif_keep_dst(dev); 3406 netif_set_tso_max_size(dev, GSO_MAX_SIZE); 3407 3408 dev->priv_flags |= IFF_NO_QUEUE; 3409 dev->change_proto_down = true; 3410 dev->lltx = true; 3411 3412 /* MTU range: 68 - 65535 */ 3413 dev->min_mtu = ETH_MIN_MTU; 3414 dev->max_mtu = ETH_MAX_MTU; 3415 3416 dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS; 3417 INIT_LIST_HEAD(&vxlan->next); 3418 spin_lock_init(&vxlan->hash_lock); 3419 3420 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3421 3422 vxlan->dev = dev; 3423 3424 INIT_HLIST_HEAD(&vxlan->fdb_list); 3425 } 3426 3427 static void vxlan_ether_setup(struct net_device *dev) 3428 { 3429 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3430 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3431 dev->netdev_ops = &vxlan_netdev_ether_ops; 3432 } 3433 3434 static void vxlan_raw_setup(struct net_device *dev) 3435 { 3436 dev->header_ops = NULL; 3437 dev->type = ARPHRD_NONE; 3438 dev->hard_header_len = 0; 3439 dev->addr_len = 0; 3440 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3441 dev->netdev_ops = &vxlan_netdev_raw_ops; 3442 } 3443 3444 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3445 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3446 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3447 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3448 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3449 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3450 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3451 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3452 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3453 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3454 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3455 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3456 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3457 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3458 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3459 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3460 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3461 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3462 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3463 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3464 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3465 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3466 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3467 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3468 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3469 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3470 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3471 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3472 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3473 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3474 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3475 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3476 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3477 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3478 [IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)), 3479 [IFLA_VXLAN_MC_ROUTE] = NLA_POLICY_MAX(NLA_U8, 1), 3480 }; 3481 3482 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3483 struct netlink_ext_ack *extack) 3484 { 3485 if (tb[IFLA_ADDRESS]) { 3486 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3487 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3488 "Provided link layer address is not Ethernet"); 3489 return -EINVAL; 3490 } 3491 3492 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3493 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3494 "Provided Ethernet address is not unicast"); 3495 return -EADDRNOTAVAIL; 3496 } 3497 } 3498 3499 if (tb[IFLA_MTU]) { 3500 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3501 3502 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3503 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3504 "MTU must be between 68 and 65535"); 3505 return -EINVAL; 3506 } 3507 } 3508 3509 if (!data) { 3510 NL_SET_ERR_MSG(extack, 3511 "Required attributes not provided to perform the operation"); 3512 return -EINVAL; 3513 } 3514 3515 if (data[IFLA_VXLAN_ID]) { 3516 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3517 3518 if (id >= VXLAN_N_VID) { 3519 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3520 "VXLAN ID must be lower than 16777216"); 3521 return -ERANGE; 3522 } 3523 } 3524 3525 if (data[IFLA_VXLAN_PORT_RANGE]) { 3526 const struct ifla_vxlan_port_range *p 3527 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3528 3529 if (ntohs(p->high) < ntohs(p->low)) { 3530 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3531 "Invalid source port range"); 3532 return -EINVAL; 3533 } 3534 } 3535 3536 if (data[IFLA_VXLAN_DF]) { 3537 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3538 3539 if (df < 0 || df > VXLAN_DF_MAX) { 3540 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3541 "Invalid DF attribute"); 3542 return -EINVAL; 3543 } 3544 } 3545 3546 return 0; 3547 } 3548 3549 static void vxlan_get_drvinfo(struct net_device *netdev, 3550 struct ethtool_drvinfo *drvinfo) 3551 { 3552 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3553 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3554 } 3555 3556 static int vxlan_get_link_ksettings(struct net_device *dev, 3557 struct ethtool_link_ksettings *cmd) 3558 { 3559 struct vxlan_dev *vxlan = netdev_priv(dev); 3560 struct vxlan_rdst *dst = &vxlan->default_dst; 3561 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3562 dst->remote_ifindex); 3563 3564 if (!lowerdev) { 3565 cmd->base.duplex = DUPLEX_UNKNOWN; 3566 cmd->base.port = PORT_OTHER; 3567 cmd->base.speed = SPEED_UNKNOWN; 3568 3569 return 0; 3570 } 3571 3572 return __ethtool_get_link_ksettings(lowerdev, cmd); 3573 } 3574 3575 static const struct ethtool_ops vxlan_ethtool_ops = { 3576 .get_drvinfo = vxlan_get_drvinfo, 3577 .get_link = ethtool_op_get_link, 3578 .get_link_ksettings = vxlan_get_link_ksettings, 3579 }; 3580 3581 static struct sock *vxlan_create_sock(struct net *net, bool ipv6, 3582 __be16 port, u32 flags, int ifindex) 3583 { 3584 struct socket *sock; 3585 struct udp_port_cfg udp_conf; 3586 int err; 3587 3588 memset(&udp_conf, 0, sizeof(udp_conf)); 3589 3590 if (ipv6) { 3591 udp_conf.family = AF_INET6; 3592 udp_conf.use_udp6_rx_checksums = 3593 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3594 udp_conf.ipv6_v6only = 1; 3595 } else { 3596 udp_conf.family = AF_INET; 3597 } 3598 3599 udp_conf.local_udp_port = port; 3600 udp_conf.bind_ifindex = ifindex; 3601 3602 /* Open UDP socket */ 3603 err = udp_sock_create(net, &udp_conf, &sock); 3604 if (err < 0) 3605 return ERR_PTR(err); 3606 3607 udp_allow_gso(sock->sk); 3608 return sock->sk; 3609 } 3610 3611 /* Create new listen socket if needed */ 3612 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3613 __be16 port, u32 flags, 3614 int ifindex) 3615 { 3616 struct udp_tunnel_sock_cfg tunnel_cfg; 3617 struct vxlan_sock *vs; 3618 struct sock *sk; 3619 unsigned int h; 3620 3621 ASSERT_RTNL(); 3622 3623 vs = kzalloc_obj(*vs); 3624 if (!vs) 3625 return ERR_PTR(-ENOMEM); 3626 3627 for (h = 0; h < VNI_HASH_SIZE; ++h) 3628 INIT_HLIST_HEAD(&vs->vni_list[h]); 3629 3630 sk = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3631 if (IS_ERR(sk)) { 3632 kfree(vs); 3633 return ERR_CAST(sk); 3634 } 3635 3636 vs->sk = sk; 3637 refcount_set(&vs->refcnt, 1); 3638 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3639 3640 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3641 udp_tunnel_notify_add_rx_port(sk, 3642 (vs->flags & VXLAN_F_GPE) ? 3643 UDP_TUNNEL_TYPE_VXLAN_GPE : 3644 UDP_TUNNEL_TYPE_VXLAN); 3645 3646 /* Mark socket as an encapsulation socket. */ 3647 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3648 tunnel_cfg.sk_user_data = vs; 3649 tunnel_cfg.encap_type = 1; 3650 tunnel_cfg.encap_rcv = vxlan_rcv; 3651 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3652 tunnel_cfg.encap_destroy = NULL; 3653 if (vs->flags & VXLAN_F_GPE) { 3654 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3655 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3656 } else { 3657 tunnel_cfg.gro_receive = vxlan_gro_receive; 3658 tunnel_cfg.gro_complete = vxlan_gro_complete; 3659 } 3660 3661 setup_udp_tunnel_sock(net, sk, &tunnel_cfg); 3662 3663 return vs; 3664 } 3665 3666 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3667 { 3668 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3669 struct vxlan_sock *vs = NULL; 3670 struct vxlan_dev_node *node; 3671 int l3mdev_index = 0; 3672 3673 ASSERT_RTNL(); 3674 3675 if (vxlan->cfg.remote_ifindex) 3676 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3677 vxlan->net, vxlan->cfg.remote_ifindex); 3678 3679 if (!vxlan->cfg.no_share) { 3680 rcu_read_lock(); 3681 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3682 vxlan->cfg.dst_port, vxlan->cfg.flags, 3683 l3mdev_index); 3684 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3685 rcu_read_unlock(); 3686 return -EBUSY; 3687 } 3688 rcu_read_unlock(); 3689 } 3690 if (!vs) 3691 vs = vxlan_socket_create(vxlan->net, ipv6, 3692 vxlan->cfg.dst_port, vxlan->cfg.flags, 3693 l3mdev_index); 3694 if (IS_ERR(vs)) 3695 return PTR_ERR(vs); 3696 #if IS_ENABLED(CONFIG_IPV6) 3697 if (ipv6) { 3698 rcu_assign_pointer(vxlan->vn6_sock, vs); 3699 node = &vxlan->hlist6; 3700 } else 3701 #endif 3702 { 3703 rcu_assign_pointer(vxlan->vn4_sock, vs); 3704 node = &vxlan->hlist4; 3705 } 3706 3707 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3708 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3709 else 3710 vxlan_vs_add_dev(vs, vxlan, node); 3711 3712 return 0; 3713 } 3714 3715 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3716 { 3717 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3718 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3719 bool ipv4 = !ipv6 || metadata; 3720 int ret = 0; 3721 3722 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3723 #if IS_ENABLED(CONFIG_IPV6) 3724 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3725 if (ipv6) { 3726 ret = __vxlan_sock_add(vxlan, true); 3727 if (ret < 0 && ret != -EAFNOSUPPORT) 3728 ipv4 = false; 3729 } 3730 #endif 3731 if (ipv4) 3732 ret = __vxlan_sock_add(vxlan, false); 3733 if (ret < 0) 3734 vxlan_sock_release(vxlan); 3735 return ret; 3736 } 3737 3738 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3739 struct vxlan_config *conf, __be32 vni) 3740 { 3741 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3742 struct vxlan_dev *tmp; 3743 3744 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3745 if (tmp == vxlan) 3746 continue; 3747 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3748 if (!vxlan_vnifilter_lookup(tmp, vni)) 3749 continue; 3750 } else if (tmp->cfg.vni != vni) { 3751 continue; 3752 } 3753 if (tmp->cfg.dst_port != conf->dst_port) 3754 continue; 3755 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3756 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3757 continue; 3758 3759 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3760 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3761 continue; 3762 3763 return -EEXIST; 3764 } 3765 3766 return 0; 3767 } 3768 3769 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3770 struct net_device **lower, 3771 struct vxlan_dev *old, 3772 struct netlink_ext_ack *extack) 3773 { 3774 bool use_ipv6 = false; 3775 3776 if (conf->flags & VXLAN_F_GPE) { 3777 /* For now, allow GPE only together with 3778 * COLLECT_METADATA. This can be relaxed later; in such 3779 * case, the other side of the PtP link will have to be 3780 * provided. 3781 */ 3782 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3783 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3784 NL_SET_ERR_MSG(extack, 3785 "VXLAN GPE does not support this combination of attributes"); 3786 return -EINVAL; 3787 } 3788 } 3789 3790 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3791 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3792 conf->remote_ip.sa.sa_family = AF_INET; 3793 conf->saddr.sa.sa_family = AF_INET; 3794 } else if (!conf->remote_ip.sa.sa_family) { 3795 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3796 } else if (!conf->saddr.sa.sa_family) { 3797 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3798 } 3799 3800 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3801 NL_SET_ERR_MSG(extack, 3802 "Local and remote address must be from the same family"); 3803 return -EINVAL; 3804 } 3805 3806 if (vxlan_addr_multicast(&conf->saddr)) { 3807 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3808 return -EINVAL; 3809 } 3810 3811 if (conf->saddr.sa.sa_family == AF_INET6) { 3812 if (!IS_ENABLED(CONFIG_IPV6)) { 3813 NL_SET_ERR_MSG(extack, 3814 "IPv6 support not enabled in the kernel"); 3815 return -EPFNOSUPPORT; 3816 } 3817 use_ipv6 = true; 3818 conf->flags |= VXLAN_F_IPV6; 3819 3820 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3821 int local_type = 3822 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3823 int remote_type = 3824 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3825 3826 if (local_type & IPV6_ADDR_LINKLOCAL) { 3827 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3828 (remote_type != IPV6_ADDR_ANY)) { 3829 NL_SET_ERR_MSG(extack, 3830 "Invalid combination of local and remote address scopes"); 3831 return -EINVAL; 3832 } 3833 3834 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3835 } else { 3836 if (remote_type == 3837 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3838 NL_SET_ERR_MSG(extack, 3839 "Invalid combination of local and remote address scopes"); 3840 return -EINVAL; 3841 } 3842 3843 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3844 } 3845 } 3846 } 3847 3848 if (conf->label && !use_ipv6) { 3849 NL_SET_ERR_MSG(extack, 3850 "Label attribute only applies to IPv6 VXLAN devices"); 3851 return -EINVAL; 3852 } 3853 3854 if (conf->label_policy && !use_ipv6) { 3855 NL_SET_ERR_MSG(extack, 3856 "Label policy only applies to IPv6 VXLAN devices"); 3857 return -EINVAL; 3858 } 3859 3860 if (conf->remote_ifindex) { 3861 struct net_device *lowerdev; 3862 3863 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3864 if (!lowerdev) { 3865 NL_SET_ERR_MSG(extack, 3866 "Invalid local interface, device not found"); 3867 return -ENODEV; 3868 } 3869 3870 #if IS_ENABLED(CONFIG_IPV6) 3871 if (use_ipv6) { 3872 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3873 3874 if (idev && idev->cnf.disable_ipv6) { 3875 NL_SET_ERR_MSG(extack, 3876 "IPv6 support disabled by administrator"); 3877 return -EPERM; 3878 } 3879 } 3880 #endif 3881 3882 *lower = lowerdev; 3883 } else { 3884 if (vxlan_addr_multicast(&conf->remote_ip)) { 3885 NL_SET_ERR_MSG(extack, 3886 "Local interface required for multicast remote destination"); 3887 3888 return -EINVAL; 3889 } 3890 3891 #if IS_ENABLED(CONFIG_IPV6) 3892 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3893 NL_SET_ERR_MSG(extack, 3894 "Local interface required for link-local local/remote addresses"); 3895 return -EINVAL; 3896 } 3897 #endif 3898 3899 *lower = NULL; 3900 } 3901 3902 if (!conf->dst_port) { 3903 if (conf->flags & VXLAN_F_GPE) 3904 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3905 else 3906 conf->dst_port = htons(vxlan_port); 3907 } 3908 3909 if (!conf->age_interval) 3910 conf->age_interval = FDB_AGE_DEFAULT; 3911 3912 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3913 NL_SET_ERR_MSG(extack, 3914 "A VXLAN device with the specified VNI already exists"); 3915 return -EEXIST; 3916 } 3917 3918 return 0; 3919 } 3920 3921 static void vxlan_config_apply(struct net_device *dev, 3922 struct vxlan_config *conf, 3923 struct net_device *lowerdev, 3924 struct net *src_net, 3925 bool changelink) 3926 { 3927 struct vxlan_dev *vxlan = netdev_priv(dev); 3928 struct vxlan_rdst *dst = &vxlan->default_dst; 3929 unsigned short needed_headroom = ETH_HLEN; 3930 int max_mtu = ETH_MAX_MTU; 3931 u32 flags = conf->flags; 3932 3933 if (!changelink) { 3934 if (flags & VXLAN_F_GPE) 3935 vxlan_raw_setup(dev); 3936 else 3937 vxlan_ether_setup(dev); 3938 3939 if (conf->mtu) 3940 dev->mtu = conf->mtu; 3941 3942 vxlan->net = src_net; 3943 } 3944 3945 dst->remote_vni = conf->vni; 3946 3947 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3948 3949 if (lowerdev) { 3950 dst->remote_ifindex = conf->remote_ifindex; 3951 3952 netif_inherit_tso_max(dev, lowerdev); 3953 3954 needed_headroom = lowerdev->hard_header_len; 3955 needed_headroom += lowerdev->needed_headroom; 3956 3957 dev->needed_tailroom = lowerdev->needed_tailroom; 3958 3959 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3960 if (max_mtu < ETH_MIN_MTU) 3961 max_mtu = ETH_MIN_MTU; 3962 3963 if (!changelink && !conf->mtu) 3964 dev->mtu = max_mtu; 3965 } 3966 3967 if (dev->mtu > max_mtu) 3968 dev->mtu = max_mtu; 3969 3970 if (flags & VXLAN_F_COLLECT_METADATA) 3971 flags |= VXLAN_F_IPV6; 3972 needed_headroom += vxlan_headroom(flags); 3973 dev->needed_headroom = needed_headroom; 3974 3975 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3976 } 3977 3978 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3979 struct vxlan_config *conf, 3980 struct netlink_ext_ack *extack) 3981 { 3982 struct vxlan_dev *vxlan = netdev_priv(dev); 3983 struct net_device *lowerdev; 3984 int ret; 3985 3986 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3987 if (ret) 3988 return ret; 3989 3990 vxlan_config_apply(dev, conf, lowerdev, src_net, false); 3991 3992 return 0; 3993 } 3994 3995 static int vxlan_dev_create(struct net *net, struct net_device *dev, 3996 struct vxlan_config *conf, 3997 struct netlink_ext_ack *extack) 3998 { 3999 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4000 struct vxlan_dev *vxlan = netdev_priv(dev); 4001 struct net_device *remote_dev = NULL; 4002 struct vxlan_rdst *dst; 4003 int err; 4004 4005 dst = &vxlan->default_dst; 4006 err = vxlan_dev_configure(net, dev, conf, extack); 4007 if (err) 4008 return err; 4009 4010 dev->ethtool_ops = &vxlan_ethtool_ops; 4011 4012 err = register_netdevice(dev); 4013 if (err) 4014 return err; 4015 4016 if (dst->remote_ifindex) { 4017 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 4018 if (!remote_dev) { 4019 err = -ENODEV; 4020 goto unregister; 4021 } 4022 4023 err = netdev_upper_dev_link(remote_dev, dev, extack); 4024 if (err) 4025 goto unregister; 4026 4027 dst->remote_dev = remote_dev; 4028 } 4029 4030 err = rtnl_configure_link(dev, NULL, 0, NULL); 4031 if (err < 0) 4032 goto unlink; 4033 4034 /* create an fdb entry for a valid default destination */ 4035 if (!vxlan_addr_any(&dst->remote_ip)) { 4036 spin_lock_bh(&vxlan->hash_lock); 4037 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4038 &dst->remote_ip, 4039 NUD_REACHABLE | NUD_PERMANENT, 4040 NLM_F_EXCL | NLM_F_CREATE, 4041 vxlan->cfg.dst_port, 4042 dst->remote_vni, 4043 dst->remote_vni, 4044 dst->remote_ifindex, 4045 NTF_SELF, 0, true, extack); 4046 spin_unlock_bh(&vxlan->hash_lock); 4047 if (err) 4048 goto unlink; 4049 } 4050 4051 list_add(&vxlan->next, &vn->vxlan_list); 4052 4053 return 0; 4054 4055 unlink: 4056 if (remote_dev) 4057 netdev_upper_dev_unlink(remote_dev, dev); 4058 unregister: 4059 unregister_netdevice(dev); 4060 return err; 4061 } 4062 4063 /* Set/clear flags based on attribute */ 4064 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4065 int attrtype, unsigned long mask, bool changelink, 4066 bool changelink_supported, 4067 struct netlink_ext_ack *extack) 4068 { 4069 unsigned long flags; 4070 4071 if (!tb[attrtype]) 4072 return 0; 4073 4074 if (changelink && !changelink_supported) { 4075 vxlan_flag_attr_error(attrtype, extack); 4076 return -EOPNOTSUPP; 4077 } 4078 4079 if (vxlan_policy[attrtype].type == NLA_FLAG) 4080 flags = conf->flags | mask; 4081 else if (nla_get_u8(tb[attrtype])) 4082 flags = conf->flags | mask; 4083 else 4084 flags = conf->flags & ~mask; 4085 4086 conf->flags = flags; 4087 4088 return 0; 4089 } 4090 4091 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4092 struct net_device *dev, struct vxlan_config *conf, 4093 bool changelink, struct netlink_ext_ack *extack) 4094 { 4095 struct vxlanhdr used_bits = { 4096 .vx_flags = VXLAN_HF_VNI, 4097 .vx_vni = VXLAN_VNI_MASK, 4098 }; 4099 struct vxlan_dev *vxlan = netdev_priv(dev); 4100 int err = 0; 4101 4102 memset(conf, 0, sizeof(*conf)); 4103 4104 /* if changelink operation, start with old existing cfg */ 4105 if (changelink) 4106 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4107 4108 if (data[IFLA_VXLAN_ID]) { 4109 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4110 4111 if (changelink && (vni != conf->vni)) { 4112 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4113 return -EOPNOTSUPP; 4114 } 4115 conf->vni = vni; 4116 } 4117 4118 if (data[IFLA_VXLAN_GROUP]) { 4119 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4120 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4121 return -EOPNOTSUPP; 4122 } 4123 4124 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4125 conf->remote_ip.sa.sa_family = AF_INET; 4126 } else if (data[IFLA_VXLAN_GROUP6]) { 4127 if (!IS_ENABLED(CONFIG_IPV6)) { 4128 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4129 return -EPFNOSUPPORT; 4130 } 4131 4132 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4133 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4134 return -EOPNOTSUPP; 4135 } 4136 4137 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4138 conf->remote_ip.sa.sa_family = AF_INET6; 4139 } 4140 4141 if (data[IFLA_VXLAN_LOCAL]) { 4142 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4143 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4144 return -EOPNOTSUPP; 4145 } 4146 4147 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4148 conf->saddr.sa.sa_family = AF_INET; 4149 } else if (data[IFLA_VXLAN_LOCAL6]) { 4150 if (!IS_ENABLED(CONFIG_IPV6)) { 4151 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4152 return -EPFNOSUPPORT; 4153 } 4154 4155 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4156 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4157 return -EOPNOTSUPP; 4158 } 4159 4160 /* TODO: respect scope id */ 4161 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4162 conf->saddr.sa.sa_family = AF_INET6; 4163 } 4164 4165 if (data[IFLA_VXLAN_LINK]) 4166 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4167 4168 if (data[IFLA_VXLAN_TOS]) 4169 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4170 4171 if (data[IFLA_VXLAN_TTL]) 4172 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4173 4174 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4175 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4176 VXLAN_F_TTL_INHERIT, changelink, false, 4177 extack); 4178 if (err) 4179 return err; 4180 4181 } 4182 4183 if (data[IFLA_VXLAN_LABEL]) 4184 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4185 IPV6_FLOWLABEL_MASK; 4186 if (data[IFLA_VXLAN_LABEL_POLICY]) 4187 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4188 4189 if (data[IFLA_VXLAN_LEARNING]) { 4190 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4191 VXLAN_F_LEARN, changelink, true, 4192 extack); 4193 if (err) 4194 return err; 4195 } else if (!changelink) { 4196 /* default to learn on a new device */ 4197 conf->flags |= VXLAN_F_LEARN; 4198 } 4199 4200 if (data[IFLA_VXLAN_AGEING]) 4201 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4202 4203 if (data[IFLA_VXLAN_PROXY]) { 4204 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4205 VXLAN_F_PROXY, changelink, false, 4206 extack); 4207 if (err) 4208 return err; 4209 } 4210 4211 if (data[IFLA_VXLAN_RSC]) { 4212 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4213 VXLAN_F_RSC, changelink, false, 4214 extack); 4215 if (err) 4216 return err; 4217 } 4218 4219 if (data[IFLA_VXLAN_L2MISS]) { 4220 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4221 VXLAN_F_L2MISS, changelink, false, 4222 extack); 4223 if (err) 4224 return err; 4225 } 4226 4227 if (data[IFLA_VXLAN_L3MISS]) { 4228 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4229 VXLAN_F_L3MISS, changelink, false, 4230 extack); 4231 if (err) 4232 return err; 4233 } 4234 4235 if (data[IFLA_VXLAN_LIMIT]) { 4236 if (changelink) { 4237 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4238 "Cannot change limit"); 4239 return -EOPNOTSUPP; 4240 } 4241 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4242 } 4243 4244 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4245 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4246 VXLAN_F_COLLECT_METADATA, changelink, false, 4247 extack); 4248 if (err) 4249 return err; 4250 } 4251 4252 if (data[IFLA_VXLAN_PORT_RANGE]) { 4253 if (!changelink) { 4254 const struct ifla_vxlan_port_range *p 4255 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4256 conf->port_min = ntohs(p->low); 4257 conf->port_max = ntohs(p->high); 4258 } else { 4259 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4260 "Cannot change port range"); 4261 return -EOPNOTSUPP; 4262 } 4263 } 4264 4265 if (data[IFLA_VXLAN_PORT]) { 4266 if (changelink) { 4267 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4268 "Cannot change port"); 4269 return -EOPNOTSUPP; 4270 } 4271 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4272 } 4273 4274 if (data[IFLA_VXLAN_UDP_CSUM]) { 4275 if (changelink) { 4276 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4277 "Cannot change UDP_CSUM flag"); 4278 return -EOPNOTSUPP; 4279 } 4280 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4281 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4282 } 4283 4284 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4285 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4286 VXLAN_F_LOCALBYPASS, changelink, 4287 true, extack); 4288 if (err) 4289 return err; 4290 } else if (!changelink) { 4291 /* default to local bypass on a new device */ 4292 conf->flags |= VXLAN_F_LOCALBYPASS; 4293 } 4294 4295 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4296 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4297 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4298 false, extack); 4299 if (err) 4300 return err; 4301 } 4302 4303 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4304 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4305 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4306 false, extack); 4307 if (err) 4308 return err; 4309 } 4310 4311 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4312 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4313 VXLAN_F_REMCSUM_TX, changelink, false, 4314 extack); 4315 if (err) 4316 return err; 4317 } 4318 4319 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4320 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4321 VXLAN_F_REMCSUM_RX, changelink, false, 4322 extack); 4323 if (err) 4324 return err; 4325 used_bits.vx_flags |= VXLAN_HF_RCO; 4326 used_bits.vx_vni |= ~VXLAN_VNI_MASK; 4327 } 4328 4329 if (data[IFLA_VXLAN_GBP]) { 4330 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4331 VXLAN_F_GBP, changelink, false, extack); 4332 if (err) 4333 return err; 4334 used_bits.vx_flags |= VXLAN_GBP_USED_BITS; 4335 } 4336 4337 if (data[IFLA_VXLAN_GPE]) { 4338 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4339 VXLAN_F_GPE, changelink, false, 4340 extack); 4341 if (err) 4342 return err; 4343 4344 used_bits.vx_flags |= VXLAN_GPE_USED_BITS; 4345 } 4346 4347 if (data[IFLA_VXLAN_RESERVED_BITS]) { 4348 struct vxlanhdr reserved_bits; 4349 4350 if (changelink) { 4351 NL_SET_ERR_MSG_ATTR(extack, 4352 data[IFLA_VXLAN_RESERVED_BITS], 4353 "Cannot change reserved_bits"); 4354 return -EOPNOTSUPP; 4355 } 4356 4357 nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS], 4358 sizeof(reserved_bits)); 4359 if (used_bits.vx_flags & reserved_bits.vx_flags || 4360 used_bits.vx_vni & reserved_bits.vx_vni) { 4361 __be64 ub_be64, rb_be64; 4362 4363 memcpy(&ub_be64, &used_bits, sizeof(ub_be64)); 4364 memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64)); 4365 4366 NL_SET_ERR_MSG_ATTR_FMT(extack, 4367 data[IFLA_VXLAN_RESERVED_BITS], 4368 "Used bits %#018llx cannot overlap reserved bits %#018llx", 4369 be64_to_cpu(ub_be64), 4370 be64_to_cpu(rb_be64)); 4371 return -EINVAL; 4372 } 4373 4374 conf->reserved_bits = reserved_bits; 4375 } else { 4376 /* For backwards compatibility, only allow reserved fields to be 4377 * used by VXLAN extensions if explicitly requested. 4378 */ 4379 conf->reserved_bits = (struct vxlanhdr) { 4380 .vx_flags = ~used_bits.vx_flags, 4381 .vx_vni = ~used_bits.vx_vni, 4382 }; 4383 } 4384 4385 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4386 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4387 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4388 false, extack); 4389 if (err) 4390 return err; 4391 } 4392 4393 if (data[IFLA_VXLAN_MC_ROUTE]) { 4394 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_MC_ROUTE, 4395 VXLAN_F_MC_ROUTE, changelink, 4396 true, extack); 4397 if (err) 4398 return err; 4399 } 4400 4401 if (tb[IFLA_MTU]) { 4402 if (changelink) { 4403 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4404 "Cannot change mtu"); 4405 return -EOPNOTSUPP; 4406 } 4407 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4408 } 4409 4410 if (data[IFLA_VXLAN_DF]) 4411 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4412 4413 if (data[IFLA_VXLAN_VNIFILTER]) { 4414 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4415 VXLAN_F_VNIFILTER, changelink, false, 4416 extack); 4417 if (err) 4418 return err; 4419 4420 if ((conf->flags & VXLAN_F_VNIFILTER) && 4421 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4422 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4423 "vxlan vnifilter only valid in collect metadata mode"); 4424 return -EINVAL; 4425 } 4426 } 4427 4428 return 0; 4429 } 4430 4431 static int vxlan_newlink(struct net_device *dev, 4432 struct rtnl_newlink_params *params, 4433 struct netlink_ext_ack *extack) 4434 { 4435 struct net *link_net = rtnl_newlink_link_net(params); 4436 struct nlattr **data = params->data; 4437 struct nlattr **tb = params->tb; 4438 struct vxlan_config conf; 4439 int err; 4440 4441 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4442 if (err) 4443 return err; 4444 4445 return vxlan_dev_create(link_net, dev, &conf, extack); 4446 } 4447 4448 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4449 struct nlattr *data[], 4450 struct netlink_ext_ack *extack) 4451 { 4452 struct vxlan_dev *vxlan = netdev_priv(dev); 4453 bool rem_ip_changed, change_igmp; 4454 struct net_device *lowerdev; 4455 struct vxlan_config conf; 4456 struct vxlan_rdst *dst; 4457 int err; 4458 4459 if (!rtnl_dev_link_net_capable(dev, vxlan->net)) 4460 return -EPERM; 4461 4462 dst = &vxlan->default_dst; 4463 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4464 if (err) 4465 return err; 4466 4467 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4468 vxlan, extack); 4469 if (err) 4470 return err; 4471 4472 if (dst->remote_dev == lowerdev) 4473 lowerdev = NULL; 4474 4475 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4476 extack); 4477 if (err) 4478 return err; 4479 4480 rem_ip_changed = !vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip); 4481 change_igmp = vxlan->dev->flags & IFF_UP && 4482 (rem_ip_changed || 4483 dst->remote_ifindex != conf.remote_ifindex); 4484 4485 /* handle default dst entry */ 4486 if (rem_ip_changed) { 4487 spin_lock_bh(&vxlan->hash_lock); 4488 if (!vxlan_addr_any(&conf.remote_ip)) { 4489 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4490 &conf.remote_ip, 4491 NUD_REACHABLE | NUD_PERMANENT, 4492 NLM_F_APPEND | NLM_F_CREATE, 4493 vxlan->cfg.dst_port, 4494 conf.vni, conf.vni, 4495 conf.remote_ifindex, 4496 NTF_SELF, 0, true, extack); 4497 if (err) { 4498 spin_unlock_bh(&vxlan->hash_lock); 4499 netdev_adjacent_change_abort(dst->remote_dev, 4500 lowerdev, dev); 4501 return err; 4502 } 4503 } 4504 if (!vxlan_addr_any(&dst->remote_ip)) 4505 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4506 dst->remote_ip, 4507 vxlan->cfg.dst_port, 4508 dst->remote_vni, 4509 dst->remote_vni, 4510 dst->remote_ifindex, 4511 true); 4512 spin_unlock_bh(&vxlan->hash_lock); 4513 4514 /* If vni filtering device, also update fdb entries of 4515 * all vnis that were using default remote ip 4516 */ 4517 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4518 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4519 &conf.remote_ip, extack); 4520 if (err) { 4521 netdev_adjacent_change_abort(dst->remote_dev, 4522 lowerdev, dev); 4523 return err; 4524 } 4525 } 4526 } 4527 4528 if (change_igmp && vxlan_addr_multicast(&dst->remote_ip)) 4529 err = vxlan_multicast_leave(vxlan); 4530 4531 if (netif_running(dev) && conf.age_interval != vxlan->cfg.age_interval) 4532 mod_timer(&vxlan->age_timer, jiffies); 4533 4534 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4535 if (lowerdev && lowerdev != dst->remote_dev) 4536 dst->remote_dev = lowerdev; 4537 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4538 4539 if (!err && change_igmp && 4540 vxlan_addr_multicast(&dst->remote_ip)) 4541 err = vxlan_multicast_join(vxlan); 4542 4543 return err; 4544 } 4545 4546 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4547 { 4548 struct vxlan_dev *vxlan = netdev_priv(dev); 4549 struct vxlan_fdb_flush_desc desc = {}; 4550 4551 vxlan_flush(vxlan, &desc); 4552 4553 list_del(&vxlan->next); 4554 unregister_netdevice_queue(dev, head); 4555 if (vxlan->default_dst.remote_dev) 4556 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4557 } 4558 4559 static size_t vxlan_get_size(const struct net_device *dev) 4560 { 4561 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4562 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4563 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4564 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4565 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4566 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4567 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4568 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4569 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4570 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4571 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4572 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4573 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4574 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4575 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4576 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4577 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4578 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4579 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4580 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4581 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4582 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4583 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4584 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4585 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4586 /* IFLA_VXLAN_PORT_RANGE */ 4587 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4588 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4589 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4590 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4591 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4592 /* IFLA_VXLAN_RESERVED_BITS */ 4593 nla_total_size(sizeof(struct vxlanhdr)) + 4594 0; 4595 } 4596 4597 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4598 { 4599 const struct vxlan_dev *vxlan = netdev_priv(dev); 4600 const struct vxlan_rdst *dst = &vxlan->default_dst; 4601 struct ifla_vxlan_port_range ports = { 4602 .low = htons(vxlan->cfg.port_min), 4603 .high = htons(vxlan->cfg.port_max), 4604 }; 4605 4606 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4607 goto nla_put_failure; 4608 4609 if (!vxlan_addr_any(&dst->remote_ip)) { 4610 if (dst->remote_ip.sa.sa_family == AF_INET) { 4611 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4612 dst->remote_ip.sin.sin_addr.s_addr)) 4613 goto nla_put_failure; 4614 #if IS_ENABLED(CONFIG_IPV6) 4615 } else { 4616 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4617 &dst->remote_ip.sin6.sin6_addr)) 4618 goto nla_put_failure; 4619 #endif 4620 } 4621 } 4622 4623 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4624 goto nla_put_failure; 4625 4626 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4627 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4628 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4629 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4630 goto nla_put_failure; 4631 #if IS_ENABLED(CONFIG_IPV6) 4632 } else { 4633 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4634 &vxlan->cfg.saddr.sin6.sin6_addr)) 4635 goto nla_put_failure; 4636 #endif 4637 } 4638 } 4639 4640 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4641 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4642 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4643 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4644 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4645 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4646 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4647 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4648 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4649 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4650 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4651 nla_put_u8(skb, IFLA_VXLAN_RSC, 4652 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4653 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4654 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4655 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4656 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4657 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4658 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4659 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4660 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4661 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4662 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4663 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4664 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4665 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4666 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4667 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4668 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4669 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4670 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4671 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4672 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4673 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4674 goto nla_put_failure; 4675 4676 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4677 goto nla_put_failure; 4678 4679 if (vxlan->cfg.flags & VXLAN_F_GBP && 4680 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4681 goto nla_put_failure; 4682 4683 if (vxlan->cfg.flags & VXLAN_F_GPE && 4684 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4685 goto nla_put_failure; 4686 4687 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4688 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4689 goto nla_put_failure; 4690 4691 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4692 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4693 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4694 goto nla_put_failure; 4695 4696 if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS, 4697 sizeof(vxlan->cfg.reserved_bits), 4698 &vxlan->cfg.reserved_bits)) 4699 goto nla_put_failure; 4700 4701 return 0; 4702 4703 nla_put_failure: 4704 return -EMSGSIZE; 4705 } 4706 4707 static struct net *vxlan_get_link_net(const struct net_device *dev) 4708 { 4709 struct vxlan_dev *vxlan = netdev_priv(dev); 4710 4711 return READ_ONCE(vxlan->net); 4712 } 4713 4714 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4715 .kind = "vxlan", 4716 .maxtype = IFLA_VXLAN_MAX, 4717 .policy = vxlan_policy, 4718 .priv_size = sizeof(struct vxlan_dev), 4719 .setup = vxlan_setup, 4720 .validate = vxlan_validate, 4721 .newlink = vxlan_newlink, 4722 .changelink = vxlan_changelink, 4723 .dellink = vxlan_dellink, 4724 .get_size = vxlan_get_size, 4725 .fill_info = vxlan_fill_info, 4726 .get_link_net = vxlan_get_link_net, 4727 }; 4728 4729 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4730 struct net_device *dev) 4731 { 4732 struct vxlan_dev *vxlan, *next; 4733 LIST_HEAD(list_kill); 4734 4735 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4736 struct vxlan_rdst *dst = &vxlan->default_dst; 4737 4738 /* In case we created vxlan device with carrier 4739 * and we loose the carrier due to module unload 4740 * we also need to remove vxlan device. In other 4741 * cases, it's not necessary and remote_ifindex 4742 * is 0 here, so no matches. 4743 */ 4744 if (dst->remote_ifindex == dev->ifindex) 4745 vxlan_dellink(vxlan->dev, &list_kill); 4746 } 4747 4748 unregister_netdevice_many(&list_kill); 4749 } 4750 4751 static int vxlan_netdevice_event(struct notifier_block *unused, 4752 unsigned long event, void *ptr) 4753 { 4754 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4755 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4756 4757 if (event == NETDEV_UNREGISTER) 4758 vxlan_handle_lowerdev_unregister(vn, dev); 4759 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4760 vxlan_offload_rx_ports(dev, true); 4761 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4762 vxlan_offload_rx_ports(dev, false); 4763 4764 return NOTIFY_DONE; 4765 } 4766 4767 static struct notifier_block vxlan_notifier_block __read_mostly = { 4768 .notifier_call = vxlan_netdevice_event, 4769 }; 4770 4771 static void 4772 vxlan_fdb_offloaded_set(struct net_device *dev, 4773 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4774 { 4775 struct vxlan_dev *vxlan = netdev_priv(dev); 4776 struct vxlan_rdst *rdst; 4777 struct vxlan_fdb *f; 4778 4779 spin_lock_bh(&vxlan->hash_lock); 4780 4781 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4782 if (!f) 4783 goto out; 4784 4785 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4786 fdb_info->remote_port, 4787 fdb_info->remote_vni, 4788 fdb_info->remote_ifindex); 4789 if (!rdst) 4790 goto out; 4791 4792 rdst->offloaded = fdb_info->offloaded; 4793 4794 out: 4795 spin_unlock_bh(&vxlan->hash_lock); 4796 } 4797 4798 static int 4799 vxlan_fdb_external_learn_add(struct net_device *dev, 4800 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4801 { 4802 struct vxlan_dev *vxlan = netdev_priv(dev); 4803 struct netlink_ext_ack *extack; 4804 int err; 4805 4806 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4807 4808 spin_lock_bh(&vxlan->hash_lock); 4809 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4810 NUD_REACHABLE, 4811 NLM_F_CREATE | NLM_F_REPLACE, 4812 fdb_info->remote_port, 4813 fdb_info->vni, 4814 fdb_info->remote_vni, 4815 fdb_info->remote_ifindex, 4816 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4817 0, false, extack); 4818 spin_unlock_bh(&vxlan->hash_lock); 4819 4820 return err; 4821 } 4822 4823 static int 4824 vxlan_fdb_external_learn_del(struct net_device *dev, 4825 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4826 { 4827 struct vxlan_dev *vxlan = netdev_priv(dev); 4828 struct vxlan_fdb *f; 4829 int err = 0; 4830 4831 spin_lock_bh(&vxlan->hash_lock); 4832 4833 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4834 if (!f) 4835 err = -ENOENT; 4836 else if (f->flags & NTF_EXT_LEARNED) 4837 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4838 fdb_info->remote_ip, 4839 fdb_info->remote_port, 4840 fdb_info->vni, 4841 fdb_info->remote_vni, 4842 fdb_info->remote_ifindex, 4843 false); 4844 4845 spin_unlock_bh(&vxlan->hash_lock); 4846 4847 return err; 4848 } 4849 4850 static int vxlan_switchdev_event(struct notifier_block *unused, 4851 unsigned long event, void *ptr) 4852 { 4853 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4854 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4855 int err = 0; 4856 4857 switch (event) { 4858 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4859 vxlan_fdb_offloaded_set(dev, ptr); 4860 break; 4861 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4862 fdb_info = ptr; 4863 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4864 if (err) { 4865 err = notifier_from_errno(err); 4866 break; 4867 } 4868 fdb_info->offloaded = true; 4869 vxlan_fdb_offloaded_set(dev, fdb_info); 4870 break; 4871 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4872 fdb_info = ptr; 4873 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4874 if (err) { 4875 err = notifier_from_errno(err); 4876 break; 4877 } 4878 fdb_info->offloaded = false; 4879 vxlan_fdb_offloaded_set(dev, fdb_info); 4880 break; 4881 } 4882 4883 return err; 4884 } 4885 4886 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4887 .notifier_call = vxlan_switchdev_event, 4888 }; 4889 4890 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4891 { 4892 struct vxlan_fdb *fdb; 4893 struct vxlan_dev *vxlan; 4894 4895 rcu_read_lock(); 4896 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4897 vxlan = rcu_dereference(fdb->vdev); 4898 WARN_ON(!vxlan); 4899 spin_lock_bh(&vxlan->hash_lock); 4900 if (!hlist_unhashed(&fdb->fdb_node)) 4901 vxlan_fdb_destroy(vxlan, fdb, false, false); 4902 spin_unlock_bh(&vxlan->hash_lock); 4903 } 4904 rcu_read_unlock(); 4905 } 4906 4907 static int vxlan_nexthop_event(struct notifier_block *nb, 4908 unsigned long event, void *ptr) 4909 { 4910 struct nh_notifier_info *info = ptr; 4911 struct nexthop *nh; 4912 4913 if (event != NEXTHOP_EVENT_DEL) 4914 return NOTIFY_DONE; 4915 4916 nh = nexthop_find_by_id(info->net, info->id); 4917 if (!nh) 4918 return NOTIFY_DONE; 4919 4920 vxlan_fdb_nh_flush(nh); 4921 4922 return NOTIFY_DONE; 4923 } 4924 4925 static __net_init int vxlan_init_net(struct net *net) 4926 { 4927 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4928 unsigned int h; 4929 4930 INIT_LIST_HEAD(&vn->vxlan_list); 4931 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4932 4933 for (h = 0; h < PORT_HASH_SIZE; ++h) 4934 INIT_HLIST_HEAD(&vn->sock_list[h]); 4935 4936 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4937 NULL); 4938 } 4939 4940 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4941 struct list_head *dev_to_kill) 4942 { 4943 struct vxlan_dev *vxlan, *next; 4944 4945 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4946 vxlan_dellink(vxlan->dev, dev_to_kill); 4947 } 4948 4949 static void __net_exit vxlan_exit_rtnl(struct net *net, 4950 struct list_head *dev_to_kill) 4951 { 4952 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4953 4954 ASSERT_RTNL_NET(net); 4955 4956 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4957 vxlan_destroy_tunnels(vn, dev_to_kill); 4958 } 4959 4960 static void __net_exit vxlan_exit_net(struct net *net) 4961 { 4962 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4963 unsigned int h; 4964 4965 for (h = 0; h < PORT_HASH_SIZE; ++h) 4966 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4967 } 4968 4969 static struct pernet_operations vxlan_net_ops = { 4970 .init = vxlan_init_net, 4971 .exit_rtnl = vxlan_exit_rtnl, 4972 .exit = vxlan_exit_net, 4973 .id = &vxlan_net_id, 4974 .size = sizeof(struct vxlan_net), 4975 }; 4976 4977 static int __init vxlan_init_module(void) 4978 { 4979 int rc; 4980 4981 rc = register_pernet_subsys(&vxlan_net_ops); 4982 if (rc) 4983 goto out1; 4984 4985 rc = register_netdevice_notifier(&vxlan_notifier_block); 4986 if (rc) 4987 goto out2; 4988 4989 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4990 if (rc) 4991 goto out3; 4992 4993 rc = rtnl_link_register(&vxlan_link_ops); 4994 if (rc) 4995 goto out4; 4996 4997 rc = vxlan_vnifilter_init(); 4998 if (rc) 4999 goto out5; 5000 5001 return 0; 5002 out5: 5003 rtnl_link_unregister(&vxlan_link_ops); 5004 out4: 5005 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5006 out3: 5007 unregister_netdevice_notifier(&vxlan_notifier_block); 5008 out2: 5009 unregister_pernet_subsys(&vxlan_net_ops); 5010 out1: 5011 return rc; 5012 } 5013 late_initcall(vxlan_init_module); 5014 5015 static void __exit vxlan_cleanup_module(void) 5016 { 5017 vxlan_vnifilter_uninit(); 5018 rtnl_link_unregister(&vxlan_link_ops); 5019 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5020 unregister_netdevice_notifier(&vxlan_notifier_block); 5021 unregister_pernet_subsys(&vxlan_net_ops); 5022 /* rcu_barrier() is called by netns */ 5023 } 5024 module_exit(vxlan_cleanup_module); 5025 5026 MODULE_LICENSE("GPL"); 5027 MODULE_VERSION(VXLAN_VERSION); 5028 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 5029 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 5030 MODULE_ALIAS_RTNL_LINK("vxlan"); 5031