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 <net/arp.h> 19 #include <net/ndisc.h> 20 #include <net/gro.h> 21 #include <net/ipv6_stubs.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/tun_proto.h> 29 #include <net/vxlan.h> 30 #include <net/nexthop.h> 31 32 #if IS_ENABLED(CONFIG_IPV6) 33 #include <net/ip6_tunnel.h> 34 #include <net/ip6_checksum.h> 35 #endif 36 37 #include "vxlan_private.h" 38 39 #define VXLAN_VERSION "0.1" 40 41 #define FDB_AGE_DEFAULT 300 /* 5 min */ 42 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ 43 44 /* UDP port for VXLAN traffic. 45 * The IANA assigned port is 4789, but the Linux default is 8472 46 * for compatibility with early adopters. 47 */ 48 static unsigned short vxlan_port __read_mostly = 8472; 49 module_param_named(udp_port, vxlan_port, ushort, 0444); 50 MODULE_PARM_DESC(udp_port, "Destination UDP port"); 51 52 static bool log_ecn_error = true; 53 module_param(log_ecn_error, bool, 0644); 54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 55 56 unsigned int vxlan_net_id; 57 58 const u8 all_zeros_mac[ETH_ALEN + 2]; 59 static struct rtnl_link_ops vxlan_link_ops; 60 61 static int vxlan_sock_add(struct vxlan_dev *vxlan); 62 63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan); 64 65 /* salt for hash table */ 66 static u32 vxlan_salt __read_mostly; 67 68 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) 69 { 70 return vs->flags & VXLAN_F_COLLECT_METADATA || 71 ip_tunnel_collect_metadata(); 72 } 73 74 /* Find VXLAN socket based on network namespace, address family, UDP port, 75 * enabled unshareable flags and socket device binding (see l3mdev with 76 * non-default VRF). 77 */ 78 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, 79 __be16 port, u32 flags, int ifindex) 80 { 81 struct vxlan_sock *vs; 82 83 flags &= VXLAN_F_RCV_FLAGS; 84 85 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { 86 if (inet_sk(vs->sock->sk)->inet_sport == port && 87 vxlan_get_sk_family(vs) == family && 88 vs->flags == flags && 89 vs->sock->sk->sk_bound_dev_if == ifindex) 90 return vs; 91 } 92 return NULL; 93 } 94 95 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, 96 int ifindex, __be32 vni, 97 struct vxlan_vni_node **vninode) 98 { 99 struct vxlan_vni_node *vnode; 100 struct vxlan_dev_node *node; 101 102 /* For flow based devices, map all packets to VNI 0 */ 103 if (vs->flags & VXLAN_F_COLLECT_METADATA && 104 !(vs->flags & VXLAN_F_VNIFILTER)) 105 vni = 0; 106 107 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) { 108 if (!node->vxlan) 109 continue; 110 vnode = NULL; 111 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 112 vnode = vxlan_vnifilter_lookup(node->vxlan, vni); 113 if (!vnode) 114 continue; 115 } else if (node->vxlan->default_dst.remote_vni != vni) { 116 continue; 117 } 118 119 if (IS_ENABLED(CONFIG_IPV6)) { 120 const struct vxlan_config *cfg = &node->vxlan->cfg; 121 122 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) && 123 cfg->remote_ifindex != ifindex) 124 continue; 125 } 126 127 if (vninode) 128 *vninode = vnode; 129 return node->vxlan; 130 } 131 132 return NULL; 133 } 134 135 /* Look up VNI in a per net namespace table */ 136 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex, 137 __be32 vni, sa_family_t family, 138 __be16 port, u32 flags) 139 { 140 struct vxlan_sock *vs; 141 142 vs = vxlan_find_sock(net, family, port, flags, ifindex); 143 if (!vs) 144 return NULL; 145 146 return vxlan_vs_find_vni(vs, ifindex, vni, NULL); 147 } 148 149 /* Fill in neighbour message in skbuff. */ 150 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, 151 const struct vxlan_fdb *fdb, 152 u32 portid, u32 seq, int type, unsigned int flags, 153 const struct vxlan_rdst *rdst) 154 { 155 unsigned long now = jiffies; 156 struct nda_cacheinfo ci; 157 bool send_ip, send_eth; 158 struct nlmsghdr *nlh; 159 struct nexthop *nh; 160 struct ndmsg *ndm; 161 int nh_family; 162 u32 nh_id; 163 164 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); 165 if (nlh == NULL) 166 return -EMSGSIZE; 167 168 ndm = nlmsg_data(nlh); 169 memset(ndm, 0, sizeof(*ndm)); 170 171 send_eth = send_ip = true; 172 173 rcu_read_lock(); 174 nh = rcu_dereference(fdb->nh); 175 if (nh) { 176 nh_family = nexthop_get_family(nh); 177 nh_id = nh->id; 178 } 179 rcu_read_unlock(); 180 181 if (type == RTM_GETNEIGH) { 182 if (rdst) { 183 send_ip = !vxlan_addr_any(&rdst->remote_ip); 184 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET; 185 } else if (nh) { 186 ndm->ndm_family = nh_family; 187 } 188 send_eth = !is_zero_ether_addr(fdb->eth_addr); 189 } else 190 ndm->ndm_family = AF_BRIDGE; 191 ndm->ndm_state = fdb->state; 192 ndm->ndm_ifindex = vxlan->dev->ifindex; 193 ndm->ndm_flags = fdb->flags; 194 if (rdst && rdst->offloaded) 195 ndm->ndm_flags |= NTF_OFFLOADED; 196 ndm->ndm_type = RTN_UNICAST; 197 198 if (!net_eq(dev_net(vxlan->dev), vxlan->net) && 199 nla_put_s32(skb, NDA_LINK_NETNSID, 200 peernet2id(dev_net(vxlan->dev), vxlan->net))) 201 goto nla_put_failure; 202 203 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) 204 goto nla_put_failure; 205 if (nh) { 206 if (nla_put_u32(skb, NDA_NH_ID, nh_id)) 207 goto nla_put_failure; 208 } else if (rdst) { 209 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, 210 &rdst->remote_ip)) 211 goto nla_put_failure; 212 213 if (rdst->remote_port && 214 rdst->remote_port != vxlan->cfg.dst_port && 215 nla_put_be16(skb, NDA_PORT, rdst->remote_port)) 216 goto nla_put_failure; 217 if (rdst->remote_vni != vxlan->default_dst.remote_vni && 218 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) 219 goto nla_put_failure; 220 if (rdst->remote_ifindex && 221 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) 222 goto nla_put_failure; 223 } 224 225 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni && 226 nla_put_u32(skb, NDA_SRC_VNI, 227 be32_to_cpu(fdb->vni))) 228 goto nla_put_failure; 229 230 ci.ndm_used = jiffies_to_clock_t(now - fdb->used); 231 ci.ndm_confirmed = 0; 232 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); 233 ci.ndm_refcnt = 0; 234 235 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) 236 goto nla_put_failure; 237 238 nlmsg_end(skb, nlh); 239 return 0; 240 241 nla_put_failure: 242 nlmsg_cancel(skb, nlh); 243 return -EMSGSIZE; 244 } 245 246 static inline size_t vxlan_nlmsg_size(void) 247 { 248 return NLMSG_ALIGN(sizeof(struct ndmsg)) 249 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ 250 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ 251 + nla_total_size(sizeof(__be16)) /* NDA_PORT */ 252 + nla_total_size(sizeof(__be32)) /* NDA_VNI */ 253 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ 254 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ 255 + nla_total_size(sizeof(struct nda_cacheinfo)); 256 } 257 258 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 259 struct vxlan_rdst *rd, int type) 260 { 261 struct net *net = dev_net(vxlan->dev); 262 struct sk_buff *skb; 263 int err = -ENOBUFS; 264 265 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); 266 if (skb == NULL) 267 goto errout; 268 269 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); 270 if (err < 0) { 271 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ 272 WARN_ON(err == -EMSGSIZE); 273 kfree_skb(skb); 274 goto errout; 275 } 276 277 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 278 return; 279 errout: 280 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 281 } 282 283 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 284 const struct vxlan_fdb *fdb, 285 const struct vxlan_rdst *rd, 286 struct netlink_ext_ack *extack, 287 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 288 { 289 fdb_info->info.dev = vxlan->dev; 290 fdb_info->info.extack = extack; 291 fdb_info->remote_ip = rd->remote_ip; 292 fdb_info->remote_port = rd->remote_port; 293 fdb_info->remote_vni = rd->remote_vni; 294 fdb_info->remote_ifindex = rd->remote_ifindex; 295 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN); 296 fdb_info->vni = fdb->vni; 297 fdb_info->offloaded = rd->offloaded; 298 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 299 } 300 301 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 302 struct vxlan_fdb *fdb, 303 struct vxlan_rdst *rd, 304 bool adding, 305 struct netlink_ext_ack *extack) 306 { 307 struct switchdev_notifier_vxlan_fdb_info info; 308 enum switchdev_notifier_type notifier_type; 309 int ret; 310 311 if (WARN_ON(!rd)) 312 return 0; 313 314 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 315 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 316 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 317 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 318 &info.info, extack); 319 return notifier_to_errno(ret); 320 } 321 322 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 323 struct vxlan_rdst *rd, int type, bool swdev_notify, 324 struct netlink_ext_ack *extack) 325 { 326 int err; 327 328 if (swdev_notify && rd) { 329 switch (type) { 330 case RTM_NEWNEIGH: 331 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 332 true, extack); 333 if (err) 334 return err; 335 break; 336 case RTM_DELNEIGH: 337 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 338 false, extack); 339 break; 340 } 341 } 342 343 __vxlan_fdb_notify(vxlan, fdb, rd, type); 344 return 0; 345 } 346 347 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 348 { 349 struct vxlan_dev *vxlan = netdev_priv(dev); 350 struct vxlan_fdb f = { 351 .state = NUD_STALE, 352 }; 353 struct vxlan_rdst remote = { 354 .remote_ip = *ipa, /* goes to NDA_DST */ 355 .remote_vni = cpu_to_be32(VXLAN_N_VID), 356 }; 357 358 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 359 } 360 361 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 362 { 363 struct vxlan_fdb f = { 364 .state = NUD_STALE, 365 }; 366 struct vxlan_rdst remote = { }; 367 368 memcpy(f.eth_addr, eth_addr, ETH_ALEN); 369 370 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 371 } 372 373 /* Hash Ethernet address */ 374 static u32 eth_hash(const unsigned char *addr) 375 { 376 u64 value = get_unaligned((u64 *)addr); 377 378 /* only want 6 bytes */ 379 #ifdef __BIG_ENDIAN 380 value >>= 16; 381 #else 382 value <<= 16; 383 #endif 384 return hash_64(value, FDB_HASH_BITS); 385 } 386 387 u32 eth_vni_hash(const unsigned char *addr, __be32 vni) 388 { 389 /* use 1 byte of OUI and 3 bytes of NIC */ 390 u32 key = get_unaligned((u32 *)(addr + 2)); 391 392 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1); 393 } 394 395 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni) 396 { 397 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) 398 return eth_vni_hash(mac, vni); 399 else 400 return eth_hash(mac); 401 } 402 403 /* Hash chain to use given mac address */ 404 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, 405 const u8 *mac, __be32 vni) 406 { 407 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)]; 408 } 409 410 /* Look up Ethernet address in forwarding table */ 411 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, 412 const u8 *mac, __be32 vni) 413 { 414 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni); 415 struct vxlan_fdb *f; 416 417 hlist_for_each_entry_rcu(f, head, hlist) { 418 if (ether_addr_equal(mac, f->eth_addr)) { 419 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 420 if (vni == f->vni) 421 return f; 422 } else { 423 return f; 424 } 425 } 426 } 427 428 return NULL; 429 } 430 431 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 432 const u8 *mac, __be32 vni) 433 { 434 struct vxlan_fdb *f; 435 436 f = __vxlan_find_mac(vxlan, mac, vni); 437 if (f && f->used != jiffies) 438 f->used = jiffies; 439 440 return f; 441 } 442 443 /* caller should hold vxlan->hash_lock */ 444 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 445 union vxlan_addr *ip, __be16 port, 446 __be32 vni, __u32 ifindex) 447 { 448 struct vxlan_rdst *rd; 449 450 list_for_each_entry(rd, &f->remotes, list) { 451 if (vxlan_addr_equal(&rd->remote_ip, ip) && 452 rd->remote_port == port && 453 rd->remote_vni == vni && 454 rd->remote_ifindex == ifindex) 455 return rd; 456 } 457 458 return NULL; 459 } 460 461 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 462 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 463 { 464 struct vxlan_dev *vxlan = netdev_priv(dev); 465 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 466 struct vxlan_rdst *rdst; 467 struct vxlan_fdb *f; 468 int rc = 0; 469 470 if (is_multicast_ether_addr(mac) || 471 is_zero_ether_addr(mac)) 472 return -EINVAL; 473 474 ether_addr_copy(eth_addr, mac); 475 476 rcu_read_lock(); 477 478 f = __vxlan_find_mac(vxlan, eth_addr, vni); 479 if (!f) { 480 rc = -ENOENT; 481 goto out; 482 } 483 484 rdst = first_remote_rcu(f); 485 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 486 487 out: 488 rcu_read_unlock(); 489 return rc; 490 } 491 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 492 493 static int vxlan_fdb_notify_one(struct notifier_block *nb, 494 const struct vxlan_dev *vxlan, 495 const struct vxlan_fdb *f, 496 const struct vxlan_rdst *rdst, 497 struct netlink_ext_ack *extack) 498 { 499 struct switchdev_notifier_vxlan_fdb_info fdb_info; 500 int rc; 501 502 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 503 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 504 &fdb_info); 505 return notifier_to_errno(rc); 506 } 507 508 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 509 struct notifier_block *nb, 510 struct netlink_ext_ack *extack) 511 { 512 struct vxlan_dev *vxlan; 513 struct vxlan_rdst *rdst; 514 struct vxlan_fdb *f; 515 unsigned int h; 516 int rc = 0; 517 518 if (!netif_is_vxlan(dev)) 519 return -EINVAL; 520 vxlan = netdev_priv(dev); 521 522 for (h = 0; h < FDB_HASH_SIZE; ++h) { 523 spin_lock_bh(&vxlan->hash_lock[h]); 524 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) { 525 if (f->vni == vni) { 526 list_for_each_entry(rdst, &f->remotes, list) { 527 rc = vxlan_fdb_notify_one(nb, vxlan, 528 f, rdst, 529 extack); 530 if (rc) 531 goto unlock; 532 } 533 } 534 } 535 spin_unlock_bh(&vxlan->hash_lock[h]); 536 } 537 return 0; 538 539 unlock: 540 spin_unlock_bh(&vxlan->hash_lock[h]); 541 return rc; 542 } 543 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 544 545 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 546 { 547 struct vxlan_dev *vxlan; 548 struct vxlan_rdst *rdst; 549 struct vxlan_fdb *f; 550 unsigned int h; 551 552 if (!netif_is_vxlan(dev)) 553 return; 554 vxlan = netdev_priv(dev); 555 556 for (h = 0; h < FDB_HASH_SIZE; ++h) { 557 spin_lock_bh(&vxlan->hash_lock[h]); 558 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) 559 if (f->vni == vni) 560 list_for_each_entry(rdst, &f->remotes, list) 561 rdst->offloaded = false; 562 spin_unlock_bh(&vxlan->hash_lock[h]); 563 } 564 565 } 566 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 567 568 /* Replace destination of unicast mac */ 569 static int vxlan_fdb_replace(struct vxlan_fdb *f, 570 union vxlan_addr *ip, __be16 port, __be32 vni, 571 __u32 ifindex, struct vxlan_rdst *oldrd) 572 { 573 struct vxlan_rdst *rd; 574 575 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 576 if (rd) 577 return 0; 578 579 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 580 if (!rd) 581 return 0; 582 583 *oldrd = *rd; 584 dst_cache_reset(&rd->dst_cache); 585 rd->remote_ip = *ip; 586 rd->remote_port = port; 587 rd->remote_vni = vni; 588 rd->remote_ifindex = ifindex; 589 rd->offloaded = false; 590 return 1; 591 } 592 593 /* Add/update destinations for multicast */ 594 static int vxlan_fdb_append(struct vxlan_fdb *f, 595 union vxlan_addr *ip, __be16 port, __be32 vni, 596 __u32 ifindex, struct vxlan_rdst **rdp) 597 { 598 struct vxlan_rdst *rd; 599 600 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 601 if (rd) 602 return 0; 603 604 rd = kmalloc(sizeof(*rd), GFP_ATOMIC); 605 if (rd == NULL) 606 return -ENOMEM; 607 608 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { 609 kfree(rd); 610 return -ENOMEM; 611 } 612 613 rd->remote_ip = *ip; 614 rd->remote_port = port; 615 rd->offloaded = false; 616 rd->remote_vni = vni; 617 rd->remote_ifindex = ifindex; 618 619 list_add_tail_rcu(&rd->list, &f->remotes); 620 621 *rdp = rd; 622 return 1; 623 } 624 625 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol) 626 { 627 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr; 628 629 /* Need to have Next Protocol set for interfaces in GPE mode. */ 630 if (!gpe->np_applied) 631 return false; 632 /* "The initial version is 0. If a receiver does not support the 633 * version indicated it MUST drop the packet. 634 */ 635 if (gpe->version != 0) 636 return false; 637 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 638 * processing MUST occur." However, we don't implement OAM 639 * processing, thus drop the packet. 640 */ 641 if (gpe->oam_flag) 642 return false; 643 644 *protocol = tun_p_to_eth_p(gpe->next_protocol); 645 if (!*protocol) 646 return false; 647 648 return true; 649 } 650 651 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 652 unsigned int off, 653 struct vxlanhdr *vh, size_t hdrlen, 654 __be32 vni_field, 655 struct gro_remcsum *grc, 656 bool nopartial) 657 { 658 size_t start, offset; 659 660 if (skb->remcsum_offload) 661 return vh; 662 663 if (!NAPI_GRO_CB(skb)->csum_valid) 664 return NULL; 665 666 start = vxlan_rco_start(vni_field); 667 offset = start + vxlan_rco_offset(vni_field); 668 669 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 670 start, offset, grc, nopartial); 671 672 skb->remcsum_offload = 1; 673 674 return vh; 675 } 676 677 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk, 678 struct list_head *head, 679 struct sk_buff *skb, 680 struct gro_remcsum *grc) 681 { 682 struct sk_buff *p; 683 struct vxlanhdr *vh, *vh2; 684 unsigned int hlen, off_vx; 685 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); 686 __be32 flags; 687 688 skb_gro_remcsum_init(grc); 689 690 off_vx = skb_gro_offset(skb); 691 hlen = off_vx + sizeof(*vh); 692 vh = skb_gro_header(skb, hlen, off_vx); 693 if (unlikely(!vh)) 694 return NULL; 695 696 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 697 698 flags = vh->vx_flags; 699 700 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 701 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 702 vh->vx_vni, grc, 703 !!(vs->flags & 704 VXLAN_F_REMCSUM_NOPARTIAL)); 705 706 if (!vh) 707 return NULL; 708 } 709 710 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 711 712 list_for_each_entry(p, head, list) { 713 if (!NAPI_GRO_CB(p)->same_flow) 714 continue; 715 716 vh2 = (struct vxlanhdr *)(p->data + off_vx); 717 if (vh->vx_flags != vh2->vx_flags || 718 vh->vx_vni != vh2->vx_vni) { 719 NAPI_GRO_CB(p)->same_flow = 0; 720 continue; 721 } 722 } 723 724 return vh; 725 } 726 727 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 728 struct list_head *head, 729 struct sk_buff *skb) 730 { 731 struct sk_buff *pp = NULL; 732 struct gro_remcsum grc; 733 int flush = 1; 734 735 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) { 736 pp = call_gro_receive(eth_gro_receive, head, skb); 737 flush = 0; 738 } 739 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 740 return pp; 741 } 742 743 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk, 744 struct list_head *head, 745 struct sk_buff *skb) 746 { 747 const struct packet_offload *ptype; 748 struct sk_buff *pp = NULL; 749 struct gro_remcsum grc; 750 struct vxlanhdr *vh; 751 __be16 protocol; 752 int flush = 1; 753 754 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc); 755 if (vh) { 756 if (!vxlan_parse_gpe_proto(vh, &protocol)) 757 goto out; 758 ptype = gro_find_receive_by_type(protocol); 759 if (!ptype) 760 goto out; 761 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb); 762 flush = 0; 763 } 764 out: 765 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 766 return pp; 767 } 768 769 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 770 { 771 /* Sets 'skb->inner_mac_header' since we are always called with 772 * 'skb->encapsulation' set. 773 */ 774 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 775 } 776 777 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 778 { 779 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff); 780 const struct packet_offload *ptype; 781 int err = -ENOSYS; 782 __be16 protocol; 783 784 if (!vxlan_parse_gpe_proto(vh, &protocol)) 785 return err; 786 ptype = gro_find_complete_by_type(protocol); 787 if (ptype) 788 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 789 return err; 790 } 791 792 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 793 __u16 state, __be32 src_vni, 794 __u16 ndm_flags) 795 { 796 struct vxlan_fdb *f; 797 798 f = kmalloc(sizeof(*f), GFP_ATOMIC); 799 if (!f) 800 return NULL; 801 f->state = state; 802 f->flags = ndm_flags; 803 f->updated = f->used = jiffies; 804 f->vni = src_vni; 805 f->nh = NULL; 806 RCU_INIT_POINTER(f->vdev, vxlan); 807 INIT_LIST_HEAD(&f->nh_list); 808 INIT_LIST_HEAD(&f->remotes); 809 memcpy(f->eth_addr, mac, ETH_ALEN); 810 811 return f; 812 } 813 814 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac, 815 __be32 src_vni, struct vxlan_fdb *f) 816 { 817 ++vxlan->addrcnt; 818 hlist_add_head_rcu(&f->hlist, 819 vxlan_fdb_head(vxlan, mac, src_vni)); 820 } 821 822 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 823 u32 nhid, struct netlink_ext_ack *extack) 824 { 825 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 826 struct nexthop *nh; 827 int err = -EINVAL; 828 829 if (old_nh && old_nh->id == nhid) 830 return 0; 831 832 nh = nexthop_find_by_id(vxlan->net, nhid); 833 if (!nh) { 834 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 835 goto err_inval; 836 } 837 838 if (!nexthop_get(nh)) { 839 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 840 nh = NULL; 841 goto err_inval; 842 } 843 if (!nexthop_is_fdb(nh)) { 844 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 845 goto err_inval; 846 } 847 848 if (!nexthop_is_multipath(nh)) { 849 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 850 goto err_inval; 851 } 852 853 /* check nexthop group family */ 854 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 855 case AF_INET: 856 if (!nexthop_has_v4(nh)) { 857 err = -EAFNOSUPPORT; 858 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 859 goto err_inval; 860 } 861 break; 862 case AF_INET6: 863 if (nexthop_has_v4(nh)) { 864 err = -EAFNOSUPPORT; 865 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 866 goto err_inval; 867 } 868 } 869 870 if (old_nh) { 871 list_del_rcu(&fdb->nh_list); 872 nexthop_put(old_nh); 873 } 874 rcu_assign_pointer(fdb->nh, nh); 875 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 876 return 1; 877 878 err_inval: 879 if (nh) 880 nexthop_put(nh); 881 return err; 882 } 883 884 int vxlan_fdb_create(struct vxlan_dev *vxlan, 885 const u8 *mac, union vxlan_addr *ip, 886 __u16 state, __be16 port, __be32 src_vni, 887 __be32 vni, __u32 ifindex, __u16 ndm_flags, 888 u32 nhid, struct vxlan_fdb **fdb, 889 struct netlink_ext_ack *extack) 890 { 891 struct vxlan_rdst *rd = NULL; 892 struct vxlan_fdb *f; 893 int rc; 894 895 if (vxlan->cfg.addrmax && 896 vxlan->addrcnt >= vxlan->cfg.addrmax) 897 return -ENOSPC; 898 899 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 900 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 901 if (!f) 902 return -ENOMEM; 903 904 if (nhid) 905 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 906 else 907 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 908 if (rc < 0) 909 goto errout; 910 911 *fdb = f; 912 913 return 0; 914 915 errout: 916 kfree(f); 917 return rc; 918 } 919 920 static void __vxlan_fdb_free(struct vxlan_fdb *f) 921 { 922 struct vxlan_rdst *rd, *nd; 923 struct nexthop *nh; 924 925 nh = rcu_dereference_raw(f->nh); 926 if (nh) { 927 rcu_assign_pointer(f->nh, NULL); 928 rcu_assign_pointer(f->vdev, NULL); 929 nexthop_put(nh); 930 } 931 932 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 933 dst_cache_destroy(&rd->dst_cache); 934 kfree(rd); 935 } 936 kfree(f); 937 } 938 939 static void vxlan_fdb_free(struct rcu_head *head) 940 { 941 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 942 943 __vxlan_fdb_free(f); 944 } 945 946 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 947 bool do_notify, bool swdev_notify) 948 { 949 struct vxlan_rdst *rd; 950 951 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr); 952 953 --vxlan->addrcnt; 954 if (do_notify) { 955 if (rcu_access_pointer(f->nh)) 956 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 957 swdev_notify, NULL); 958 else 959 list_for_each_entry(rd, &f->remotes, list) 960 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 961 swdev_notify, NULL); 962 } 963 964 hlist_del_rcu(&f->hlist); 965 list_del_rcu(&f->nh_list); 966 call_rcu(&f->rcu, vxlan_fdb_free); 967 } 968 969 static void vxlan_dst_free(struct rcu_head *head) 970 { 971 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 972 973 dst_cache_destroy(&rd->dst_cache); 974 kfree(rd); 975 } 976 977 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 978 union vxlan_addr *ip, 979 __u16 state, __u16 flags, 980 __be16 port, __be32 vni, 981 __u32 ifindex, __u16 ndm_flags, 982 struct vxlan_fdb *f, u32 nhid, 983 bool swdev_notify, 984 struct netlink_ext_ack *extack) 985 { 986 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 987 struct vxlan_rdst *rd = NULL; 988 struct vxlan_rdst oldrd; 989 int notify = 0; 990 int rc = 0; 991 int err; 992 993 if (nhid && !rcu_access_pointer(f->nh)) { 994 NL_SET_ERR_MSG(extack, 995 "Cannot replace an existing non nexthop fdb with a nexthop"); 996 return -EOPNOTSUPP; 997 } 998 999 if (nhid && (flags & NLM_F_APPEND)) { 1000 NL_SET_ERR_MSG(extack, 1001 "Cannot append to a nexthop fdb"); 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 f->updated = jiffies; 1013 notify = 1; 1014 } 1015 if (f->flags != fdb_flags) { 1016 f->flags = fdb_flags; 1017 f->updated = jiffies; 1018 notify = 1; 1019 } 1020 } 1021 1022 if ((flags & NLM_F_REPLACE)) { 1023 /* Only change unicasts */ 1024 if (!(is_multicast_ether_addr(f->eth_addr) || 1025 is_zero_ether_addr(f->eth_addr))) { 1026 if (nhid) { 1027 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1028 if (rc < 0) 1029 return rc; 1030 } else { 1031 rc = vxlan_fdb_replace(f, ip, port, vni, 1032 ifindex, &oldrd); 1033 } 1034 notify |= rc; 1035 } else { 1036 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1037 return -EOPNOTSUPP; 1038 } 1039 } 1040 if ((flags & NLM_F_APPEND) && 1041 (is_multicast_ether_addr(f->eth_addr) || 1042 is_zero_ether_addr(f->eth_addr))) { 1043 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1044 1045 if (rc < 0) 1046 return rc; 1047 notify |= rc; 1048 } 1049 1050 if (ndm_flags & NTF_USE) 1051 f->used = jiffies; 1052 1053 if (notify) { 1054 if (rd == NULL) 1055 rd = first_remote_rtnl(f); 1056 1057 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1058 swdev_notify, extack); 1059 if (err) 1060 goto err_notify; 1061 } 1062 1063 return 0; 1064 1065 err_notify: 1066 if (nhid) 1067 return err; 1068 if ((flags & NLM_F_REPLACE) && rc) 1069 *rd = oldrd; 1070 else if ((flags & NLM_F_APPEND) && rc) { 1071 list_del_rcu(&rd->list); 1072 call_rcu(&rd->rcu, vxlan_dst_free); 1073 } 1074 return err; 1075 } 1076 1077 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1078 const u8 *mac, union vxlan_addr *ip, 1079 __u16 state, __u16 flags, 1080 __be16 port, __be32 src_vni, __be32 vni, 1081 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1082 bool swdev_notify, 1083 struct netlink_ext_ack *extack) 1084 { 1085 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1086 struct vxlan_fdb *f; 1087 int rc; 1088 1089 /* Disallow replace to add a multicast entry */ 1090 if ((flags & NLM_F_REPLACE) && 1091 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1092 return -EOPNOTSUPP; 1093 1094 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1095 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1096 vni, ifindex, fdb_flags, nhid, &f, extack); 1097 if (rc < 0) 1098 return rc; 1099 1100 vxlan_fdb_insert(vxlan, mac, src_vni, f); 1101 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1102 swdev_notify, extack); 1103 if (rc) 1104 goto err_notify; 1105 1106 return 0; 1107 1108 err_notify: 1109 vxlan_fdb_destroy(vxlan, f, false, false); 1110 return rc; 1111 } 1112 1113 /* Add new entry to forwarding table -- assumes lock held */ 1114 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1115 const u8 *mac, union vxlan_addr *ip, 1116 __u16 state, __u16 flags, 1117 __be16 port, __be32 src_vni, __be32 vni, 1118 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1119 bool swdev_notify, 1120 struct netlink_ext_ack *extack) 1121 { 1122 struct vxlan_fdb *f; 1123 1124 f = __vxlan_find_mac(vxlan, mac, src_vni); 1125 if (f) { 1126 if (flags & NLM_F_EXCL) { 1127 netdev_dbg(vxlan->dev, 1128 "lost race to create %pM\n", mac); 1129 return -EEXIST; 1130 } 1131 1132 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1133 vni, ifindex, ndm_flags, f, 1134 nhid, swdev_notify, extack); 1135 } else { 1136 if (!(flags & NLM_F_CREATE)) 1137 return -ENOENT; 1138 1139 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1140 port, src_vni, vni, ifindex, 1141 ndm_flags, nhid, swdev_notify, 1142 extack); 1143 } 1144 } 1145 1146 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1147 struct vxlan_rdst *rd, bool swdev_notify) 1148 { 1149 list_del_rcu(&rd->list); 1150 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1151 call_rcu(&rd->rcu, vxlan_dst_free); 1152 } 1153 1154 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1155 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1156 __be32 *vni, u32 *ifindex, u32 *nhid, 1157 struct netlink_ext_ack *extack) 1158 { 1159 struct net *net = dev_net(vxlan->dev); 1160 int err; 1161 1162 if (tb[NDA_NH_ID] && 1163 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1164 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1165 return -EINVAL; 1166 } 1167 1168 if (tb[NDA_DST]) { 1169 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1170 if (err) { 1171 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1172 return err; 1173 } 1174 } else { 1175 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1176 1177 if (remote->sa.sa_family == AF_INET) { 1178 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1179 ip->sa.sa_family = AF_INET; 1180 #if IS_ENABLED(CONFIG_IPV6) 1181 } else { 1182 ip->sin6.sin6_addr = in6addr_any; 1183 ip->sa.sa_family = AF_INET6; 1184 #endif 1185 } 1186 } 1187 1188 if (tb[NDA_PORT]) { 1189 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1190 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1191 return -EINVAL; 1192 } 1193 *port = nla_get_be16(tb[NDA_PORT]); 1194 } else { 1195 *port = vxlan->cfg.dst_port; 1196 } 1197 1198 if (tb[NDA_VNI]) { 1199 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1200 NL_SET_ERR_MSG(extack, "Invalid vni"); 1201 return -EINVAL; 1202 } 1203 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1204 } else { 1205 *vni = vxlan->default_dst.remote_vni; 1206 } 1207 1208 if (tb[NDA_SRC_VNI]) { 1209 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1210 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1211 return -EINVAL; 1212 } 1213 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1214 } else { 1215 *src_vni = vxlan->default_dst.remote_vni; 1216 } 1217 1218 if (tb[NDA_IFINDEX]) { 1219 struct net_device *tdev; 1220 1221 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1222 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1223 return -EINVAL; 1224 } 1225 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1226 tdev = __dev_get_by_index(net, *ifindex); 1227 if (!tdev) { 1228 NL_SET_ERR_MSG(extack, "Device not found"); 1229 return -EADDRNOTAVAIL; 1230 } 1231 } else { 1232 *ifindex = 0; 1233 } 1234 1235 if (tb[NDA_NH_ID]) 1236 *nhid = nla_get_u32(tb[NDA_NH_ID]); 1237 else 1238 *nhid = 0; 1239 1240 return 0; 1241 } 1242 1243 /* Add static entry (via netlink) */ 1244 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1245 struct net_device *dev, 1246 const unsigned char *addr, u16 vid, u16 flags, 1247 struct netlink_ext_ack *extack) 1248 { 1249 struct vxlan_dev *vxlan = netdev_priv(dev); 1250 /* struct net *net = dev_net(vxlan->dev); */ 1251 union vxlan_addr ip; 1252 __be16 port; 1253 __be32 src_vni, vni; 1254 u32 ifindex, nhid; 1255 u32 hash_index; 1256 int err; 1257 1258 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1259 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1260 ndm->ndm_state); 1261 return -EINVAL; 1262 } 1263 1264 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1265 return -EINVAL; 1266 1267 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1268 &nhid, extack); 1269 if (err) 1270 return err; 1271 1272 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1273 return -EAFNOSUPPORT; 1274 1275 hash_index = fdb_head_index(vxlan, addr, src_vni); 1276 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1277 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1278 port, src_vni, vni, ifindex, 1279 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1280 nhid, true, extack); 1281 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1282 1283 return err; 1284 } 1285 1286 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1287 const unsigned char *addr, union vxlan_addr ip, 1288 __be16 port, __be32 src_vni, __be32 vni, 1289 u32 ifindex, bool swdev_notify) 1290 { 1291 struct vxlan_rdst *rd = NULL; 1292 struct vxlan_fdb *f; 1293 int err = -ENOENT; 1294 1295 f = vxlan_find_mac(vxlan, addr, src_vni); 1296 if (!f) 1297 return err; 1298 1299 if (!vxlan_addr_any(&ip)) { 1300 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1301 if (!rd) 1302 goto out; 1303 } 1304 1305 /* remove a destination if it's not the only one on the list, 1306 * otherwise destroy the fdb entry 1307 */ 1308 if (rd && !list_is_singular(&f->remotes)) { 1309 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1310 goto out; 1311 } 1312 1313 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1314 1315 out: 1316 return 0; 1317 } 1318 1319 /* Delete entry (via netlink) */ 1320 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1321 struct net_device *dev, 1322 const unsigned char *addr, u16 vid, 1323 struct netlink_ext_ack *extack) 1324 { 1325 struct vxlan_dev *vxlan = netdev_priv(dev); 1326 union vxlan_addr ip; 1327 __be32 src_vni, vni; 1328 u32 ifindex, nhid; 1329 u32 hash_index; 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 hash_index = fdb_head_index(vxlan, addr, src_vni); 1339 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1340 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1341 true); 1342 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1343 1344 return err; 1345 } 1346 1347 /* Dump forwarding table */ 1348 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1349 struct net_device *dev, 1350 struct net_device *filter_dev, int *idx) 1351 { 1352 struct vxlan_dev *vxlan = netdev_priv(dev); 1353 unsigned int h; 1354 int err = 0; 1355 1356 for (h = 0; h < FDB_HASH_SIZE; ++h) { 1357 struct vxlan_fdb *f; 1358 1359 rcu_read_lock(); 1360 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { 1361 struct vxlan_rdst *rd; 1362 1363 if (rcu_access_pointer(f->nh)) { 1364 if (*idx < cb->args[2]) 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, 1370 NLM_F_MULTI, NULL); 1371 if (err < 0) { 1372 rcu_read_unlock(); 1373 goto out; 1374 } 1375 skip_nh: 1376 *idx += 1; 1377 continue; 1378 } 1379 1380 list_for_each_entry_rcu(rd, &f->remotes, list) { 1381 if (*idx < cb->args[2]) 1382 goto skip; 1383 1384 err = vxlan_fdb_info(skb, vxlan, f, 1385 NETLINK_CB(cb->skb).portid, 1386 cb->nlh->nlmsg_seq, 1387 RTM_NEWNEIGH, 1388 NLM_F_MULTI, rd); 1389 if (err < 0) { 1390 rcu_read_unlock(); 1391 goto out; 1392 } 1393 skip: 1394 *idx += 1; 1395 } 1396 } 1397 rcu_read_unlock(); 1398 } 1399 out: 1400 return err; 1401 } 1402 1403 static int vxlan_fdb_get(struct sk_buff *skb, 1404 struct nlattr *tb[], 1405 struct net_device *dev, 1406 const unsigned char *addr, 1407 u16 vid, u32 portid, u32 seq, 1408 struct netlink_ext_ack *extack) 1409 { 1410 struct vxlan_dev *vxlan = netdev_priv(dev); 1411 struct vxlan_fdb *f; 1412 __be32 vni; 1413 int err; 1414 1415 if (tb[NDA_VNI]) 1416 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1417 else 1418 vni = vxlan->default_dst.remote_vni; 1419 1420 rcu_read_lock(); 1421 1422 f = __vxlan_find_mac(vxlan, addr, vni); 1423 if (!f) { 1424 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1425 err = -ENOENT; 1426 goto errout; 1427 } 1428 1429 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1430 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1431 errout: 1432 rcu_read_unlock(); 1433 return err; 1434 } 1435 1436 /* Watch incoming packets to learn mapping between Ethernet address 1437 * and Tunnel endpoint. 1438 * Return true if packet is bogus and should be dropped. 1439 */ 1440 static bool vxlan_snoop(struct net_device *dev, 1441 union vxlan_addr *src_ip, const u8 *src_mac, 1442 u32 src_ifindex, __be32 vni) 1443 { 1444 struct vxlan_dev *vxlan = netdev_priv(dev); 1445 struct vxlan_fdb *f; 1446 u32 ifindex = 0; 1447 1448 /* Ignore packets from invalid src-address */ 1449 if (!is_valid_ether_addr(src_mac)) 1450 return true; 1451 1452 #if IS_ENABLED(CONFIG_IPV6) 1453 if (src_ip->sa.sa_family == AF_INET6 && 1454 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1455 ifindex = src_ifindex; 1456 #endif 1457 1458 f = vxlan_find_mac(vxlan, src_mac, vni); 1459 if (likely(f)) { 1460 struct vxlan_rdst *rdst = first_remote_rcu(f); 1461 1462 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1463 rdst->remote_ifindex == ifindex)) 1464 return false; 1465 1466 /* Don't migrate static entries, drop packets */ 1467 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1468 return true; 1469 1470 /* Don't override an fdb with nexthop with a learnt entry */ 1471 if (rcu_access_pointer(f->nh)) 1472 return true; 1473 1474 if (net_ratelimit()) 1475 netdev_info(dev, 1476 "%pM migrated from %pIS to %pIS\n", 1477 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1478 1479 rdst->remote_ip = *src_ip; 1480 f->updated = jiffies; 1481 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1482 } else { 1483 u32 hash_index = fdb_head_index(vxlan, src_mac, vni); 1484 1485 /* learned new entry */ 1486 spin_lock(&vxlan->hash_lock[hash_index]); 1487 1488 /* close off race between vxlan_flush and incoming packets */ 1489 if (netif_running(dev)) 1490 vxlan_fdb_update(vxlan, src_mac, src_ip, 1491 NUD_REACHABLE, 1492 NLM_F_EXCL|NLM_F_CREATE, 1493 vxlan->cfg.dst_port, 1494 vni, 1495 vxlan->default_dst.remote_vni, 1496 ifindex, NTF_SELF, 0, true, NULL); 1497 spin_unlock(&vxlan->hash_lock[hash_index]); 1498 } 1499 1500 return false; 1501 } 1502 1503 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1504 { 1505 struct vxlan_net *vn; 1506 1507 if (!vs) 1508 return false; 1509 if (!refcount_dec_and_test(&vs->refcnt)) 1510 return false; 1511 1512 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); 1513 spin_lock(&vn->sock_lock); 1514 hlist_del_rcu(&vs->hlist); 1515 udp_tunnel_notify_del_rx_port(vs->sock, 1516 (vs->flags & VXLAN_F_GPE) ? 1517 UDP_TUNNEL_TYPE_VXLAN_GPE : 1518 UDP_TUNNEL_TYPE_VXLAN); 1519 spin_unlock(&vn->sock_lock); 1520 1521 return true; 1522 } 1523 1524 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1525 { 1526 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1527 #if IS_ENABLED(CONFIG_IPV6) 1528 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1529 1530 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1531 #endif 1532 1533 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1534 synchronize_net(); 1535 1536 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1537 vxlan_vs_del_vnigrp(vxlan); 1538 else 1539 vxlan_vs_del_dev(vxlan); 1540 1541 if (__vxlan_sock_release_prep(sock4)) { 1542 udp_tunnel_sock_release(sock4->sock); 1543 kfree(sock4); 1544 } 1545 1546 #if IS_ENABLED(CONFIG_IPV6) 1547 if (__vxlan_sock_release_prep(sock6)) { 1548 udp_tunnel_sock_release(sock6->sock); 1549 kfree(sock6); 1550 } 1551 #endif 1552 } 1553 1554 static bool vxlan_remcsum(struct vxlanhdr *unparsed, 1555 struct sk_buff *skb, u32 vxflags) 1556 { 1557 size_t start, offset; 1558 1559 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1560 goto out; 1561 1562 start = vxlan_rco_start(unparsed->vx_vni); 1563 offset = start + vxlan_rco_offset(unparsed->vx_vni); 1564 1565 if (!pskb_may_pull(skb, offset + sizeof(u16))) 1566 return false; 1567 1568 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1569 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1570 out: 1571 unparsed->vx_flags &= ~VXLAN_HF_RCO; 1572 unparsed->vx_vni &= VXLAN_VNI_MASK; 1573 return true; 1574 } 1575 1576 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed, 1577 struct sk_buff *skb, u32 vxflags, 1578 struct vxlan_metadata *md) 1579 { 1580 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed; 1581 struct metadata_dst *tun_dst; 1582 1583 if (!(unparsed->vx_flags & VXLAN_HF_GBP)) 1584 goto out; 1585 1586 md->gbp = ntohs(gbp->policy_id); 1587 1588 tun_dst = (struct metadata_dst *)skb_dst(skb); 1589 if (tun_dst) { 1590 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1591 tun_dst->u.tun_info.key.tun_flags); 1592 tun_dst->u.tun_info.options_len = sizeof(*md); 1593 } 1594 if (gbp->dont_learn) 1595 md->gbp |= VXLAN_GBP_DONT_LEARN; 1596 1597 if (gbp->policy_applied) 1598 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1599 1600 /* In flow-based mode, GBP is carried in dst_metadata */ 1601 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1602 skb->mark = md->gbp; 1603 out: 1604 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS; 1605 } 1606 1607 static bool vxlan_set_mac(struct vxlan_dev *vxlan, 1608 struct vxlan_sock *vs, 1609 struct sk_buff *skb, __be32 vni) 1610 { 1611 union vxlan_addr saddr; 1612 u32 ifindex = skb->dev->ifindex; 1613 1614 skb_reset_mac_header(skb); 1615 skb->protocol = eth_type_trans(skb, vxlan->dev); 1616 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1617 1618 /* Ignore packet loops (and multicast echo) */ 1619 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1620 return false; 1621 1622 /* Get address from the outer IP header */ 1623 if (vxlan_get_sk_family(vs) == AF_INET) { 1624 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1625 saddr.sa.sa_family = AF_INET; 1626 #if IS_ENABLED(CONFIG_IPV6) 1627 } else { 1628 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1629 saddr.sa.sa_family = AF_INET6; 1630 #endif 1631 } 1632 1633 if ((vxlan->cfg.flags & VXLAN_F_LEARN) && 1634 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni)) 1635 return false; 1636 1637 return true; 1638 } 1639 1640 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1641 struct sk_buff *skb) 1642 { 1643 int err = 0; 1644 1645 if (vxlan_get_sk_family(vs) == AF_INET) 1646 err = IP_ECN_decapsulate(oiph, skb); 1647 #if IS_ENABLED(CONFIG_IPV6) 1648 else 1649 err = IP6_ECN_decapsulate(oiph, skb); 1650 #endif 1651 1652 if (unlikely(err) && log_ecn_error) { 1653 if (vxlan_get_sk_family(vs) == AF_INET) 1654 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1655 &((struct iphdr *)oiph)->saddr, 1656 ((struct iphdr *)oiph)->tos); 1657 else 1658 net_info_ratelimited("non-ECT from %pI6\n", 1659 &((struct ipv6hdr *)oiph)->saddr); 1660 } 1661 return err <= 1; 1662 } 1663 1664 /* Callback from net/ipv4/udp.c to receive packets */ 1665 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1666 { 1667 struct vxlan_vni_node *vninode = NULL; 1668 struct vxlan_dev *vxlan; 1669 struct vxlan_sock *vs; 1670 struct vxlanhdr unparsed; 1671 struct vxlan_metadata _md; 1672 struct vxlan_metadata *md = &_md; 1673 __be16 protocol = htons(ETH_P_TEB); 1674 enum skb_drop_reason reason; 1675 bool raw_proto = false; 1676 void *oiph; 1677 __be32 vni = 0; 1678 int nh; 1679 1680 /* Need UDP and VXLAN header to be present */ 1681 reason = pskb_may_pull_reason(skb, VXLAN_HLEN); 1682 if (reason) 1683 goto drop; 1684 1685 unparsed = *vxlan_hdr(skb); 1686 /* VNI flag always required to be set */ 1687 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1688 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1689 ntohl(vxlan_hdr(skb)->vx_flags), 1690 ntohl(vxlan_hdr(skb)->vx_vni)); 1691 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1692 /* Return non vxlan pkt */ 1693 goto drop; 1694 } 1695 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1696 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1697 1698 vs = rcu_dereference_sk_user_data(sk); 1699 if (!vs) 1700 goto drop; 1701 1702 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1703 1704 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1705 if (!vxlan) { 1706 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND; 1707 goto drop; 1708 } 1709 1710 /* For backwards compatibility, only allow reserved fields to be 1711 * used by VXLAN extensions if explicitly requested. 1712 */ 1713 if (vs->flags & VXLAN_F_GPE) { 1714 if (!vxlan_parse_gpe_proto(&unparsed, &protocol)) 1715 goto drop; 1716 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS; 1717 raw_proto = true; 1718 } 1719 1720 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1721 !net_eq(vxlan->net, dev_net(vxlan->dev)))) { 1722 reason = SKB_DROP_REASON_NOMEM; 1723 goto drop; 1724 } 1725 1726 if (vs->flags & VXLAN_F_REMCSUM_RX) 1727 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1728 goto drop; 1729 1730 if (vxlan_collect_metadata(vs)) { 1731 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1732 struct metadata_dst *tun_dst; 1733 1734 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1735 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1736 key32_to_tunnel_id(vni), sizeof(*md)); 1737 1738 if (!tun_dst) { 1739 reason = SKB_DROP_REASON_NOMEM; 1740 goto drop; 1741 } 1742 1743 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1744 1745 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1746 } else { 1747 memset(md, 0, sizeof(*md)); 1748 } 1749 1750 if (vs->flags & VXLAN_F_GBP) 1751 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1752 /* Note that GBP and GPE can never be active together. This is 1753 * ensured in vxlan_dev_configure. 1754 */ 1755 1756 if (unparsed.vx_flags || unparsed.vx_vni) { 1757 /* If there are any unprocessed flags remaining treat 1758 * this as a malformed packet. This behavior diverges from 1759 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1760 * in reserved fields are to be ignored. The approach here 1761 * maintains compatibility with previous stack code, and also 1762 * is more robust and provides a little more security in 1763 * adding extensions to VXLAN. 1764 */ 1765 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1766 goto drop; 1767 } 1768 1769 if (!raw_proto) { 1770 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1771 goto drop; 1772 } else { 1773 skb_reset_mac_header(skb); 1774 skb->dev = vxlan->dev; 1775 skb->pkt_type = PACKET_HOST; 1776 } 1777 1778 /* Save offset of outer header relative to skb->head, 1779 * because we are going to reset the network header to the inner header 1780 * and might change skb->head. 1781 */ 1782 nh = skb_network_header(skb) - skb->head; 1783 1784 skb_reset_network_header(skb); 1785 1786 reason = pskb_inet_may_pull_reason(skb); 1787 if (reason) { 1788 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1789 DEV_STATS_INC(vxlan->dev, rx_errors); 1790 vxlan_vnifilter_count(vxlan, vni, vninode, 1791 VXLAN_VNI_STATS_RX_ERRORS, 0); 1792 goto drop; 1793 } 1794 1795 /* Get the outer header. */ 1796 oiph = skb->head + nh; 1797 1798 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1799 reason = SKB_DROP_REASON_IP_TUNNEL_ECN; 1800 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1801 DEV_STATS_INC(vxlan->dev, rx_errors); 1802 vxlan_vnifilter_count(vxlan, vni, vninode, 1803 VXLAN_VNI_STATS_RX_ERRORS, 0); 1804 goto drop; 1805 } 1806 1807 rcu_read_lock(); 1808 1809 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1810 rcu_read_unlock(); 1811 dev_core_stats_rx_dropped_inc(vxlan->dev); 1812 vxlan_vnifilter_count(vxlan, vni, vninode, 1813 VXLAN_VNI_STATS_RX_DROPS, 0); 1814 reason = SKB_DROP_REASON_DEV_READY; 1815 goto drop; 1816 } 1817 1818 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1819 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1820 gro_cells_receive(&vxlan->gro_cells, skb); 1821 1822 rcu_read_unlock(); 1823 1824 return 0; 1825 1826 drop: 1827 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED; 1828 /* Consume bad packet */ 1829 kfree_skb_reason(skb, reason); 1830 return 0; 1831 } 1832 1833 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1834 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1835 { 1836 struct vxlan_dev *vxlan; 1837 struct vxlan_sock *vs; 1838 struct vxlanhdr *hdr; 1839 __be32 vni; 1840 1841 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1842 return -EINVAL; 1843 1844 hdr = vxlan_hdr(skb); 1845 1846 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1847 return -EINVAL; 1848 1849 vs = rcu_dereference_sk_user_data(sk); 1850 if (!vs) 1851 return -ENOENT; 1852 1853 vni = vxlan_vni(hdr->vx_vni); 1854 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1855 if (!vxlan) 1856 return -ENOENT; 1857 1858 return 0; 1859 } 1860 1861 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1862 { 1863 struct vxlan_dev *vxlan = netdev_priv(dev); 1864 struct arphdr *parp; 1865 u8 *arpptr, *sha; 1866 __be32 sip, tip; 1867 struct neighbour *n; 1868 1869 if (dev->flags & IFF_NOARP) 1870 goto out; 1871 1872 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1873 dev_core_stats_tx_dropped_inc(dev); 1874 vxlan_vnifilter_count(vxlan, vni, NULL, 1875 VXLAN_VNI_STATS_TX_DROPS, 0); 1876 goto out; 1877 } 1878 parp = arp_hdr(skb); 1879 1880 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1881 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1882 parp->ar_pro != htons(ETH_P_IP) || 1883 parp->ar_op != htons(ARPOP_REQUEST) || 1884 parp->ar_hln != dev->addr_len || 1885 parp->ar_pln != 4) 1886 goto out; 1887 arpptr = (u8 *)parp + sizeof(struct arphdr); 1888 sha = arpptr; 1889 arpptr += dev->addr_len; /* sha */ 1890 memcpy(&sip, arpptr, sizeof(sip)); 1891 arpptr += sizeof(sip); 1892 arpptr += dev->addr_len; /* tha */ 1893 memcpy(&tip, arpptr, sizeof(tip)); 1894 1895 if (ipv4_is_loopback(tip) || 1896 ipv4_is_multicast(tip)) 1897 goto out; 1898 1899 n = neigh_lookup(&arp_tbl, &tip, dev); 1900 1901 if (n) { 1902 struct vxlan_fdb *f; 1903 struct sk_buff *reply; 1904 1905 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1906 neigh_release(n); 1907 goto out; 1908 } 1909 1910 f = vxlan_find_mac(vxlan, n->ha, vni); 1911 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1912 /* bridge-local neighbor */ 1913 neigh_release(n); 1914 goto out; 1915 } 1916 1917 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1918 n->ha, sha); 1919 1920 neigh_release(n); 1921 1922 if (reply == NULL) 1923 goto out; 1924 1925 skb_reset_mac_header(reply); 1926 __skb_pull(reply, skb_network_offset(reply)); 1927 reply->ip_summed = CHECKSUM_UNNECESSARY; 1928 reply->pkt_type = PACKET_HOST; 1929 1930 if (netif_rx(reply) == NET_RX_DROP) { 1931 dev_core_stats_rx_dropped_inc(dev); 1932 vxlan_vnifilter_count(vxlan, vni, NULL, 1933 VXLAN_VNI_STATS_RX_DROPS, 0); 1934 } 1935 1936 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1937 union vxlan_addr ipa = { 1938 .sin.sin_addr.s_addr = tip, 1939 .sin.sin_family = AF_INET, 1940 }; 1941 1942 vxlan_ip_miss(dev, &ipa); 1943 } 1944 out: 1945 consume_skb(skb); 1946 return NETDEV_TX_OK; 1947 } 1948 1949 #if IS_ENABLED(CONFIG_IPV6) 1950 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1951 struct neighbour *n, bool isrouter) 1952 { 1953 struct net_device *dev = request->dev; 1954 struct sk_buff *reply; 1955 struct nd_msg *ns, *na; 1956 struct ipv6hdr *pip6; 1957 u8 *daddr; 1958 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1959 int ns_olen; 1960 int i, len; 1961 1962 if (dev == NULL || !pskb_may_pull(request, request->len)) 1963 return NULL; 1964 1965 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1966 sizeof(*na) + na_olen + dev->needed_tailroom; 1967 reply = alloc_skb(len, GFP_ATOMIC); 1968 if (reply == NULL) 1969 return NULL; 1970 1971 reply->protocol = htons(ETH_P_IPV6); 1972 reply->dev = dev; 1973 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1974 skb_push(reply, sizeof(struct ethhdr)); 1975 skb_reset_mac_header(reply); 1976 1977 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1978 1979 daddr = eth_hdr(request)->h_source; 1980 ns_olen = request->len - skb_network_offset(request) - 1981 sizeof(struct ipv6hdr) - sizeof(*ns); 1982 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1983 if (!ns->opt[i + 1]) { 1984 kfree_skb(reply); 1985 return NULL; 1986 } 1987 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1988 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1989 break; 1990 } 1991 } 1992 1993 /* Ethernet header */ 1994 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1995 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1996 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1997 reply->protocol = htons(ETH_P_IPV6); 1998 1999 skb_pull(reply, sizeof(struct ethhdr)); 2000 skb_reset_network_header(reply); 2001 skb_put(reply, sizeof(struct ipv6hdr)); 2002 2003 /* IPv6 header */ 2004 2005 pip6 = ipv6_hdr(reply); 2006 memset(pip6, 0, sizeof(struct ipv6hdr)); 2007 pip6->version = 6; 2008 pip6->priority = ipv6_hdr(request)->priority; 2009 pip6->nexthdr = IPPROTO_ICMPV6; 2010 pip6->hop_limit = 255; 2011 pip6->daddr = ipv6_hdr(request)->saddr; 2012 pip6->saddr = *(struct in6_addr *)n->primary_key; 2013 2014 skb_pull(reply, sizeof(struct ipv6hdr)); 2015 skb_reset_transport_header(reply); 2016 2017 /* Neighbor Advertisement */ 2018 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2019 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2020 na->icmph.icmp6_router = isrouter; 2021 na->icmph.icmp6_override = 1; 2022 na->icmph.icmp6_solicited = 1; 2023 na->target = ns->target; 2024 ether_addr_copy(&na->opt[2], n->ha); 2025 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2026 na->opt[1] = na_olen >> 3; 2027 2028 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2029 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2030 csum_partial(na, sizeof(*na)+na_olen, 0)); 2031 2032 pip6->payload_len = htons(sizeof(*na)+na_olen); 2033 2034 skb_push(reply, sizeof(struct ipv6hdr)); 2035 2036 reply->ip_summed = CHECKSUM_UNNECESSARY; 2037 2038 return reply; 2039 } 2040 2041 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2042 { 2043 struct vxlan_dev *vxlan = netdev_priv(dev); 2044 const struct in6_addr *daddr; 2045 const struct ipv6hdr *iphdr; 2046 struct inet6_dev *in6_dev; 2047 struct neighbour *n; 2048 struct nd_msg *msg; 2049 2050 rcu_read_lock(); 2051 in6_dev = __in6_dev_get(dev); 2052 if (!in6_dev) 2053 goto out; 2054 2055 iphdr = ipv6_hdr(skb); 2056 daddr = &iphdr->daddr; 2057 msg = (struct nd_msg *)(iphdr + 1); 2058 2059 if (ipv6_addr_loopback(daddr) || 2060 ipv6_addr_is_multicast(&msg->target)) 2061 goto out; 2062 2063 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2064 2065 if (n) { 2066 struct vxlan_fdb *f; 2067 struct sk_buff *reply; 2068 2069 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2070 neigh_release(n); 2071 goto out; 2072 } 2073 2074 f = vxlan_find_mac(vxlan, n->ha, vni); 2075 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2076 /* bridge-local neighbor */ 2077 neigh_release(n); 2078 goto out; 2079 } 2080 2081 reply = vxlan_na_create(skb, n, 2082 !!(f ? f->flags & NTF_ROUTER : 0)); 2083 2084 neigh_release(n); 2085 2086 if (reply == NULL) 2087 goto out; 2088 2089 if (netif_rx(reply) == NET_RX_DROP) { 2090 dev_core_stats_rx_dropped_inc(dev); 2091 vxlan_vnifilter_count(vxlan, vni, NULL, 2092 VXLAN_VNI_STATS_RX_DROPS, 0); 2093 } 2094 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2095 union vxlan_addr ipa = { 2096 .sin6.sin6_addr = msg->target, 2097 .sin6.sin6_family = AF_INET6, 2098 }; 2099 2100 vxlan_ip_miss(dev, &ipa); 2101 } 2102 2103 out: 2104 rcu_read_unlock(); 2105 consume_skb(skb); 2106 return NETDEV_TX_OK; 2107 } 2108 #endif 2109 2110 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2111 { 2112 struct vxlan_dev *vxlan = netdev_priv(dev); 2113 struct neighbour *n; 2114 2115 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2116 return false; 2117 2118 n = NULL; 2119 switch (ntohs(eth_hdr(skb)->h_proto)) { 2120 case ETH_P_IP: 2121 { 2122 struct iphdr *pip; 2123 2124 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2125 return false; 2126 pip = ip_hdr(skb); 2127 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2128 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2129 union vxlan_addr ipa = { 2130 .sin.sin_addr.s_addr = pip->daddr, 2131 .sin.sin_family = AF_INET, 2132 }; 2133 2134 vxlan_ip_miss(dev, &ipa); 2135 return false; 2136 } 2137 2138 break; 2139 } 2140 #if IS_ENABLED(CONFIG_IPV6) 2141 case ETH_P_IPV6: 2142 { 2143 struct ipv6hdr *pip6; 2144 2145 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2146 return false; 2147 pip6 = ipv6_hdr(skb); 2148 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2149 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2150 union vxlan_addr ipa = { 2151 .sin6.sin6_addr = pip6->daddr, 2152 .sin6.sin6_family = AF_INET6, 2153 }; 2154 2155 vxlan_ip_miss(dev, &ipa); 2156 return false; 2157 } 2158 2159 break; 2160 } 2161 #endif 2162 default: 2163 return false; 2164 } 2165 2166 if (n) { 2167 bool diff; 2168 2169 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2170 if (diff) { 2171 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2172 dev->addr_len); 2173 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2174 } 2175 neigh_release(n); 2176 return diff; 2177 } 2178 2179 return false; 2180 } 2181 2182 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2183 { 2184 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2185 2186 gpe->np_applied = 1; 2187 gpe->next_protocol = tun_p_from_eth_p(protocol); 2188 if (!gpe->next_protocol) 2189 return -EPFNOSUPPORT; 2190 return 0; 2191 } 2192 2193 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2194 int iphdr_len, __be32 vni, 2195 struct vxlan_metadata *md, u32 vxflags, 2196 bool udp_sum) 2197 { 2198 struct vxlanhdr *vxh; 2199 int min_headroom; 2200 int err; 2201 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2202 __be16 inner_protocol = htons(ETH_P_TEB); 2203 2204 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2205 skb->ip_summed == CHECKSUM_PARTIAL) { 2206 int csum_start = skb_checksum_start_offset(skb); 2207 2208 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2209 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2210 (skb->csum_offset == offsetof(struct udphdr, check) || 2211 skb->csum_offset == offsetof(struct tcphdr, check))) 2212 type |= SKB_GSO_TUNNEL_REMCSUM; 2213 } 2214 2215 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2216 + VXLAN_HLEN + iphdr_len; 2217 2218 /* Need space for new headers (invalidates iph ptr) */ 2219 err = skb_cow_head(skb, min_headroom); 2220 if (unlikely(err)) 2221 return err; 2222 2223 err = iptunnel_handle_offloads(skb, type); 2224 if (err) 2225 return err; 2226 2227 vxh = __skb_push(skb, sizeof(*vxh)); 2228 vxh->vx_flags = VXLAN_HF_VNI; 2229 vxh->vx_vni = vxlan_vni_field(vni); 2230 2231 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2232 unsigned int start; 2233 2234 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2235 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2236 vxh->vx_flags |= VXLAN_HF_RCO; 2237 2238 if (!skb_is_gso(skb)) { 2239 skb->ip_summed = CHECKSUM_NONE; 2240 skb->encapsulation = 0; 2241 } 2242 } 2243 2244 if (vxflags & VXLAN_F_GBP) 2245 vxlan_build_gbp_hdr(vxh, md); 2246 if (vxflags & VXLAN_F_GPE) { 2247 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2248 if (err < 0) 2249 return err; 2250 inner_protocol = skb->protocol; 2251 } 2252 2253 skb_set_inner_protocol(skb, inner_protocol); 2254 return 0; 2255 } 2256 2257 /* Bypass encapsulation if the destination is local */ 2258 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2259 struct vxlan_dev *dst_vxlan, __be32 vni, 2260 bool snoop) 2261 { 2262 union vxlan_addr loopback; 2263 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2264 struct net_device *dev; 2265 int len = skb->len; 2266 2267 skb->pkt_type = PACKET_HOST; 2268 skb->encapsulation = 0; 2269 skb->dev = dst_vxlan->dev; 2270 __skb_pull(skb, skb_network_offset(skb)); 2271 2272 if (remote_ip->sa.sa_family == AF_INET) { 2273 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2274 loopback.sa.sa_family = AF_INET; 2275 #if IS_ENABLED(CONFIG_IPV6) 2276 } else { 2277 loopback.sin6.sin6_addr = in6addr_loopback; 2278 loopback.sa.sa_family = AF_INET6; 2279 #endif 2280 } 2281 2282 rcu_read_lock(); 2283 dev = skb->dev; 2284 if (unlikely(!(dev->flags & IFF_UP))) { 2285 kfree_skb(skb); 2286 goto drop; 2287 } 2288 2289 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2290 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2291 2292 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2293 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2294 2295 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2296 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2297 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2298 len); 2299 } else { 2300 drop: 2301 dev_core_stats_rx_dropped_inc(dev); 2302 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2303 VXLAN_VNI_STATS_RX_DROPS, 0); 2304 } 2305 rcu_read_unlock(); 2306 } 2307 2308 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2309 struct vxlan_dev *vxlan, 2310 int addr_family, 2311 __be16 dst_port, int dst_ifindex, __be32 vni, 2312 struct dst_entry *dst, 2313 u32 rt_flags) 2314 { 2315 #if IS_ENABLED(CONFIG_IPV6) 2316 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2317 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2318 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2319 */ 2320 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2321 #endif 2322 /* Bypass encapsulation if the destination is local */ 2323 if (rt_flags & RTCF_LOCAL && 2324 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2325 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2326 struct vxlan_dev *dst_vxlan; 2327 2328 dst_release(dst); 2329 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2330 addr_family, dst_port, 2331 vxlan->cfg.flags); 2332 if (!dst_vxlan) { 2333 DEV_STATS_INC(dev, tx_errors); 2334 vxlan_vnifilter_count(vxlan, vni, NULL, 2335 VXLAN_VNI_STATS_TX_ERRORS, 0); 2336 kfree_skb(skb); 2337 2338 return -ENOENT; 2339 } 2340 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2341 return 1; 2342 } 2343 2344 return 0; 2345 } 2346 2347 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2348 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2349 { 2350 struct dst_cache *dst_cache; 2351 struct ip_tunnel_info *info; 2352 struct ip_tunnel_key *pkey; 2353 struct ip_tunnel_key key; 2354 struct vxlan_dev *vxlan = netdev_priv(dev); 2355 const struct iphdr *old_iph; 2356 struct vxlan_metadata _md; 2357 struct vxlan_metadata *md = &_md; 2358 unsigned int pkt_len = skb->len; 2359 __be16 src_port = 0, dst_port; 2360 struct dst_entry *ndst = NULL; 2361 int addr_family; 2362 __u8 tos, ttl; 2363 int ifindex; 2364 int err; 2365 u32 flags = vxlan->cfg.flags; 2366 bool use_cache; 2367 bool udp_sum = false; 2368 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2369 bool no_eth_encap; 2370 __be32 vni = 0; 2371 2372 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2373 if (skb_vlan_inet_prepare(skb, no_eth_encap)) 2374 goto drop; 2375 2376 old_iph = ip_hdr(skb); 2377 2378 info = skb_tunnel_info(skb); 2379 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2380 2381 if (rdst) { 2382 memset(&key, 0, sizeof(key)); 2383 pkey = &key; 2384 2385 if (vxlan_addr_any(&rdst->remote_ip)) { 2386 if (did_rsc) { 2387 /* short-circuited back to local bridge */ 2388 vxlan_encap_bypass(skb, vxlan, vxlan, 2389 default_vni, true); 2390 return; 2391 } 2392 goto drop; 2393 } 2394 2395 addr_family = vxlan->cfg.saddr.sa.sa_family; 2396 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2397 vni = (rdst->remote_vni) ? : default_vni; 2398 ifindex = rdst->remote_ifindex; 2399 2400 if (addr_family == AF_INET) { 2401 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2402 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2403 } else { 2404 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2405 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2406 } 2407 2408 dst_cache = &rdst->dst_cache; 2409 md->gbp = skb->mark; 2410 if (flags & VXLAN_F_TTL_INHERIT) { 2411 ttl = ip_tunnel_get_ttl(old_iph, skb); 2412 } else { 2413 ttl = vxlan->cfg.ttl; 2414 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2415 ttl = 1; 2416 } 2417 tos = vxlan->cfg.tos; 2418 if (tos == 1) 2419 tos = ip_tunnel_get_dsfield(old_iph, skb); 2420 if (tos && !info) 2421 use_cache = false; 2422 2423 if (addr_family == AF_INET) 2424 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2425 else 2426 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2427 #if IS_ENABLED(CONFIG_IPV6) 2428 switch (vxlan->cfg.label_policy) { 2429 case VXLAN_LABEL_FIXED: 2430 key.label = vxlan->cfg.label; 2431 break; 2432 case VXLAN_LABEL_INHERIT: 2433 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2434 break; 2435 default: 2436 DEBUG_NET_WARN_ON_ONCE(1); 2437 goto drop; 2438 } 2439 #endif 2440 } else { 2441 if (!info) { 2442 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2443 dev->name); 2444 goto drop; 2445 } 2446 pkey = &info->key; 2447 addr_family = ip_tunnel_info_af(info); 2448 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2449 vni = tunnel_id_to_key32(info->key.tun_id); 2450 ifindex = 0; 2451 dst_cache = &info->dst_cache; 2452 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2453 if (info->options_len < sizeof(*md)) 2454 goto drop; 2455 md = ip_tunnel_info_opts(info); 2456 } 2457 ttl = info->key.ttl; 2458 tos = info->key.tos; 2459 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2460 } 2461 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2462 vxlan->cfg.port_max, true); 2463 2464 rcu_read_lock(); 2465 if (addr_family == AF_INET) { 2466 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2467 struct rtable *rt; 2468 __be16 df = 0; 2469 __be32 saddr; 2470 2471 if (!ifindex) 2472 ifindex = sock4->sock->sk->sk_bound_dev_if; 2473 2474 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2475 &saddr, pkey, src_port, dst_port, 2476 tos, use_cache ? dst_cache : NULL); 2477 if (IS_ERR(rt)) { 2478 err = PTR_ERR(rt); 2479 goto tx_error; 2480 } 2481 2482 if (!info) { 2483 /* Bypass encapsulation if the destination is local */ 2484 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2485 dst_port, ifindex, vni, 2486 &rt->dst, rt->rt_flags); 2487 if (err) 2488 goto out_unlock; 2489 2490 if (vxlan->cfg.df == VXLAN_DF_SET) { 2491 df = htons(IP_DF); 2492 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2493 struct ethhdr *eth = eth_hdr(skb); 2494 2495 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2496 (ntohs(eth->h_proto) == ETH_P_IP && 2497 old_iph->frag_off & htons(IP_DF))) 2498 df = htons(IP_DF); 2499 } 2500 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2501 info->key.tun_flags)) { 2502 df = htons(IP_DF); 2503 } 2504 2505 ndst = &rt->dst; 2506 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2507 netif_is_any_bridge_port(dev)); 2508 if (err < 0) { 2509 goto tx_error; 2510 } else if (err) { 2511 if (info) { 2512 struct ip_tunnel_info *unclone; 2513 2514 unclone = skb_tunnel_info_unclone(skb); 2515 if (unlikely(!unclone)) 2516 goto tx_error; 2517 2518 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2519 unclone->key.u.ipv4.dst = saddr; 2520 } 2521 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2522 dst_release(ndst); 2523 goto out_unlock; 2524 } 2525 2526 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2527 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2528 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2529 vni, md, flags, udp_sum); 2530 if (err < 0) 2531 goto tx_error; 2532 2533 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2534 pkey->u.ipv4.dst, tos, ttl, df, 2535 src_port, dst_port, xnet, !udp_sum); 2536 #if IS_ENABLED(CONFIG_IPV6) 2537 } else { 2538 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2539 struct in6_addr saddr; 2540 2541 if (!ifindex) 2542 ifindex = sock6->sock->sk->sk_bound_dev_if; 2543 2544 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2545 ifindex, &saddr, pkey, 2546 src_port, dst_port, tos, 2547 use_cache ? dst_cache : NULL); 2548 if (IS_ERR(ndst)) { 2549 err = PTR_ERR(ndst); 2550 ndst = NULL; 2551 goto tx_error; 2552 } 2553 2554 if (!info) { 2555 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2556 2557 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2558 dst_port, ifindex, vni, 2559 ndst, rt6i_flags); 2560 if (err) 2561 goto out_unlock; 2562 } 2563 2564 err = skb_tunnel_check_pmtu(skb, ndst, 2565 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2566 netif_is_any_bridge_port(dev)); 2567 if (err < 0) { 2568 goto tx_error; 2569 } else if (err) { 2570 if (info) { 2571 struct ip_tunnel_info *unclone; 2572 2573 unclone = skb_tunnel_info_unclone(skb); 2574 if (unlikely(!unclone)) 2575 goto tx_error; 2576 2577 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2578 unclone->key.u.ipv6.dst = saddr; 2579 } 2580 2581 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2582 dst_release(ndst); 2583 goto out_unlock; 2584 } 2585 2586 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2587 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2588 skb_scrub_packet(skb, xnet); 2589 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2590 vni, md, flags, udp_sum); 2591 if (err < 0) 2592 goto tx_error; 2593 2594 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2595 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2596 pkey->label, src_port, dst_port, !udp_sum); 2597 #endif 2598 } 2599 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2600 out_unlock: 2601 rcu_read_unlock(); 2602 return; 2603 2604 drop: 2605 dev_core_stats_tx_dropped_inc(dev); 2606 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2607 dev_kfree_skb(skb); 2608 return; 2609 2610 tx_error: 2611 rcu_read_unlock(); 2612 if (err == -ELOOP) 2613 DEV_STATS_INC(dev, collisions); 2614 else if (err == -ENETUNREACH) 2615 DEV_STATS_INC(dev, tx_carrier_errors); 2616 dst_release(ndst); 2617 DEV_STATS_INC(dev, tx_errors); 2618 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2619 kfree_skb(skb); 2620 } 2621 2622 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2623 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2624 { 2625 struct vxlan_rdst nh_rdst; 2626 struct nexthop *nh; 2627 bool do_xmit; 2628 u32 hash; 2629 2630 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2631 hash = skb_get_hash(skb); 2632 2633 rcu_read_lock(); 2634 nh = rcu_dereference(f->nh); 2635 if (!nh) { 2636 rcu_read_unlock(); 2637 goto drop; 2638 } 2639 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2640 rcu_read_unlock(); 2641 2642 if (likely(do_xmit)) 2643 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2644 else 2645 goto drop; 2646 2647 return; 2648 2649 drop: 2650 dev_core_stats_tx_dropped_inc(dev); 2651 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2652 VXLAN_VNI_STATS_TX_DROPS, 0); 2653 dev_kfree_skb(skb); 2654 } 2655 2656 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2657 u32 nhid, __be32 vni) 2658 { 2659 struct vxlan_dev *vxlan = netdev_priv(dev); 2660 struct vxlan_rdst nh_rdst; 2661 struct nexthop *nh; 2662 bool do_xmit; 2663 u32 hash; 2664 2665 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2666 hash = skb_get_hash(skb); 2667 2668 rcu_read_lock(); 2669 nh = nexthop_find_by_id(dev_net(dev), nhid); 2670 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2671 rcu_read_unlock(); 2672 goto drop; 2673 } 2674 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2675 rcu_read_unlock(); 2676 2677 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2678 goto drop; 2679 2680 if (likely(do_xmit)) 2681 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2682 else 2683 goto drop; 2684 2685 return NETDEV_TX_OK; 2686 2687 drop: 2688 dev_core_stats_tx_dropped_inc(dev); 2689 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2690 VXLAN_VNI_STATS_TX_DROPS, 0); 2691 dev_kfree_skb(skb); 2692 return NETDEV_TX_OK; 2693 } 2694 2695 /* Transmit local packets over Vxlan 2696 * 2697 * Outer IP header inherits ECN and DF from inner header. 2698 * Outer UDP destination is the VXLAN assigned port. 2699 * source port is based on hash of flow 2700 */ 2701 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2702 { 2703 struct vxlan_dev *vxlan = netdev_priv(dev); 2704 struct vxlan_rdst *rdst, *fdst = NULL; 2705 const struct ip_tunnel_info *info; 2706 struct vxlan_fdb *f; 2707 struct ethhdr *eth; 2708 __be32 vni = 0; 2709 u32 nhid = 0; 2710 bool did_rsc; 2711 2712 info = skb_tunnel_info(skb); 2713 2714 skb_reset_mac_header(skb); 2715 2716 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2717 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2718 info->mode & IP_TUNNEL_INFO_TX) { 2719 vni = tunnel_id_to_key32(info->key.tun_id); 2720 nhid = info->key.nhid; 2721 } else { 2722 if (info && info->mode & IP_TUNNEL_INFO_TX) 2723 vxlan_xmit_one(skb, dev, vni, NULL, false); 2724 else 2725 kfree_skb(skb); 2726 return NETDEV_TX_OK; 2727 } 2728 } 2729 2730 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2731 eth = eth_hdr(skb); 2732 if (ntohs(eth->h_proto) == ETH_P_ARP) 2733 return arp_reduce(dev, skb, vni); 2734 #if IS_ENABLED(CONFIG_IPV6) 2735 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2736 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2737 sizeof(struct nd_msg)) && 2738 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2739 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2740 2741 if (m->icmph.icmp6_code == 0 && 2742 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2743 return neigh_reduce(dev, skb, vni); 2744 } 2745 #endif 2746 } 2747 2748 if (nhid) 2749 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2750 2751 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2752 struct vxlan_mdb_entry *mdb_entry; 2753 2754 rcu_read_lock(); 2755 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2756 if (mdb_entry) { 2757 netdev_tx_t ret; 2758 2759 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2760 rcu_read_unlock(); 2761 return ret; 2762 } 2763 rcu_read_unlock(); 2764 } 2765 2766 eth = eth_hdr(skb); 2767 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2768 did_rsc = false; 2769 2770 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2771 (ntohs(eth->h_proto) == ETH_P_IP || 2772 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2773 did_rsc = route_shortcircuit(dev, skb); 2774 if (did_rsc) 2775 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2776 } 2777 2778 if (f == NULL) { 2779 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2780 if (f == NULL) { 2781 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2782 !is_multicast_ether_addr(eth->h_dest)) 2783 vxlan_fdb_miss(vxlan, eth->h_dest); 2784 2785 dev_core_stats_tx_dropped_inc(dev); 2786 vxlan_vnifilter_count(vxlan, vni, NULL, 2787 VXLAN_VNI_STATS_TX_DROPS, 0); 2788 kfree_skb(skb); 2789 return NETDEV_TX_OK; 2790 } 2791 } 2792 2793 if (rcu_access_pointer(f->nh)) { 2794 vxlan_xmit_nh(skb, dev, f, 2795 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2796 } else { 2797 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2798 struct sk_buff *skb1; 2799 2800 if (!fdst) { 2801 fdst = rdst; 2802 continue; 2803 } 2804 skb1 = skb_clone(skb, GFP_ATOMIC); 2805 if (skb1) 2806 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2807 } 2808 if (fdst) 2809 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2810 else 2811 kfree_skb(skb); 2812 } 2813 2814 return NETDEV_TX_OK; 2815 } 2816 2817 /* Walk the forwarding table and purge stale entries */ 2818 static void vxlan_cleanup(struct timer_list *t) 2819 { 2820 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2821 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2822 unsigned int h; 2823 2824 if (!netif_running(vxlan->dev)) 2825 return; 2826 2827 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2828 struct hlist_node *p, *n; 2829 2830 spin_lock(&vxlan->hash_lock[h]); 2831 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2832 struct vxlan_fdb *f 2833 = container_of(p, struct vxlan_fdb, hlist); 2834 unsigned long timeout; 2835 2836 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2837 continue; 2838 2839 if (f->flags & NTF_EXT_LEARNED) 2840 continue; 2841 2842 timeout = f->used + vxlan->cfg.age_interval * HZ; 2843 if (time_before_eq(timeout, jiffies)) { 2844 netdev_dbg(vxlan->dev, 2845 "garbage collect %pM\n", 2846 f->eth_addr); 2847 f->state = NUD_STALE; 2848 vxlan_fdb_destroy(vxlan, f, true, true); 2849 } else if (time_before(timeout, next_timer)) 2850 next_timer = timeout; 2851 } 2852 spin_unlock(&vxlan->hash_lock[h]); 2853 } 2854 2855 mod_timer(&vxlan->age_timer, next_timer); 2856 } 2857 2858 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2859 { 2860 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2861 2862 spin_lock(&vn->sock_lock); 2863 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2864 #if IS_ENABLED(CONFIG_IPV6) 2865 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2866 #endif 2867 spin_unlock(&vn->sock_lock); 2868 } 2869 2870 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2871 struct vxlan_dev_node *node) 2872 { 2873 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2874 __be32 vni = vxlan->default_dst.remote_vni; 2875 2876 node->vxlan = vxlan; 2877 spin_lock(&vn->sock_lock); 2878 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2879 spin_unlock(&vn->sock_lock); 2880 } 2881 2882 /* Setup stats when device is created */ 2883 static int vxlan_init(struct net_device *dev) 2884 { 2885 struct vxlan_dev *vxlan = netdev_priv(dev); 2886 int err; 2887 2888 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2889 vxlan_vnigroup_init(vxlan); 2890 2891 err = gro_cells_init(&vxlan->gro_cells, dev); 2892 if (err) 2893 goto err_vnigroup_uninit; 2894 2895 err = vxlan_mdb_init(vxlan); 2896 if (err) 2897 goto err_gro_cells_destroy; 2898 2899 netdev_lockdep_set_classes(dev); 2900 return 0; 2901 2902 err_gro_cells_destroy: 2903 gro_cells_destroy(&vxlan->gro_cells); 2904 err_vnigroup_uninit: 2905 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2906 vxlan_vnigroup_uninit(vxlan); 2907 return err; 2908 } 2909 2910 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2911 { 2912 struct vxlan_fdb *f; 2913 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2914 2915 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2916 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2917 if (f) 2918 vxlan_fdb_destroy(vxlan, f, true, true); 2919 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2920 } 2921 2922 static void vxlan_uninit(struct net_device *dev) 2923 { 2924 struct vxlan_dev *vxlan = netdev_priv(dev); 2925 2926 vxlan_mdb_fini(vxlan); 2927 2928 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2929 vxlan_vnigroup_uninit(vxlan); 2930 2931 gro_cells_destroy(&vxlan->gro_cells); 2932 2933 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2934 } 2935 2936 /* Start ageing timer and join group when device is brought up */ 2937 static int vxlan_open(struct net_device *dev) 2938 { 2939 struct vxlan_dev *vxlan = netdev_priv(dev); 2940 int ret; 2941 2942 ret = vxlan_sock_add(vxlan); 2943 if (ret < 0) 2944 return ret; 2945 2946 ret = vxlan_multicast_join(vxlan); 2947 if (ret) { 2948 vxlan_sock_release(vxlan); 2949 return ret; 2950 } 2951 2952 if (vxlan->cfg.age_interval) 2953 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2954 2955 return ret; 2956 } 2957 2958 struct vxlan_fdb_flush_desc { 2959 bool ignore_default_entry; 2960 unsigned long state; 2961 unsigned long state_mask; 2962 unsigned long flags; 2963 unsigned long flags_mask; 2964 __be32 src_vni; 2965 u32 nhid; 2966 __be32 vni; 2967 __be16 port; 2968 union vxlan_addr dst_ip; 2969 }; 2970 2971 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2972 const struct vxlan_dev *vxlan) 2973 { 2974 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2975 } 2976 2977 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2978 { 2979 struct nexthop *nh = rtnl_dereference(f->nh); 2980 2981 return nh && nh->id == nhid; 2982 } 2983 2984 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2985 const struct vxlan_dev *vxlan, 2986 const struct vxlan_fdb_flush_desc *desc) 2987 { 2988 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2989 return false; 2990 2991 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2992 return false; 2993 2994 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2995 return false; 2996 2997 if (desc->src_vni && f->vni != desc->src_vni) 2998 return false; 2999 3000 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 3001 return false; 3002 3003 return true; 3004 } 3005 3006 static bool 3007 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 3008 { 3009 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 3010 } 3011 3012 static bool 3013 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3014 const struct vxlan_rdst *rd) 3015 { 3016 if (desc->vni && rd->remote_vni != desc->vni) 3017 return false; 3018 3019 if (desc->port && rd->remote_port != desc->port) 3020 return false; 3021 3022 if (desc->dst_ip.sa.sa_family && 3023 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3024 return false; 3025 3026 return true; 3027 } 3028 3029 static void 3030 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3031 const struct vxlan_fdb_flush_desc *desc, 3032 bool *p_destroy_fdb) 3033 { 3034 bool remotes_flushed = false; 3035 struct vxlan_rdst *rd, *tmp; 3036 3037 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3038 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3039 continue; 3040 3041 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3042 remotes_flushed = true; 3043 } 3044 3045 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3046 } 3047 3048 /* Purge the forwarding table */ 3049 static void vxlan_flush(struct vxlan_dev *vxlan, 3050 const struct vxlan_fdb_flush_desc *desc) 3051 { 3052 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3053 unsigned int h; 3054 3055 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3056 struct hlist_node *p, *n; 3057 3058 spin_lock_bh(&vxlan->hash_lock[h]); 3059 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3060 struct vxlan_fdb *f 3061 = container_of(p, struct vxlan_fdb, hlist); 3062 3063 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3064 continue; 3065 3066 if (match_remotes) { 3067 bool destroy_fdb = false; 3068 3069 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3070 &destroy_fdb); 3071 3072 if (!destroy_fdb) 3073 continue; 3074 } 3075 3076 vxlan_fdb_destroy(vxlan, f, true, true); 3077 } 3078 spin_unlock_bh(&vxlan->hash_lock[h]); 3079 } 3080 } 3081 3082 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3083 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3084 [NDA_NH_ID] = { .type = NLA_U32 }, 3085 [NDA_VNI] = { .type = NLA_U32 }, 3086 [NDA_PORT] = { .type = NLA_U16 }, 3087 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3088 sizeof(struct in6_addr)), 3089 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3090 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3091 }; 3092 3093 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3094 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3095 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3096 NTF_ROUTER) 3097 3098 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3099 struct netlink_ext_ack *extack) 3100 { 3101 struct vxlan_dev *vxlan = netdev_priv(dev); 3102 struct vxlan_fdb_flush_desc desc = {}; 3103 struct ndmsg *ndm = nlmsg_data(nlh); 3104 struct nlattr *tb[NDA_MAX + 1]; 3105 u8 ndm_flags; 3106 int err; 3107 3108 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3109 3110 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3111 extack); 3112 if (err) 3113 return err; 3114 3115 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3116 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3117 return -EINVAL; 3118 } 3119 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3120 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3121 return -EINVAL; 3122 } 3123 3124 desc.state = ndm->ndm_state; 3125 desc.flags = ndm_flags; 3126 3127 if (tb[NDA_NDM_STATE_MASK]) 3128 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3129 3130 if (tb[NDA_NDM_FLAGS_MASK]) 3131 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3132 3133 if (tb[NDA_SRC_VNI]) 3134 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3135 3136 if (tb[NDA_NH_ID]) 3137 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3138 3139 if (tb[NDA_VNI]) 3140 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3141 3142 if (tb[NDA_PORT]) 3143 desc.port = nla_get_be16(tb[NDA_PORT]); 3144 3145 if (tb[NDA_DST]) { 3146 union vxlan_addr ip; 3147 3148 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3149 if (err) { 3150 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3151 "Unsupported address family"); 3152 return err; 3153 } 3154 desc.dst_ip = ip; 3155 } 3156 3157 vxlan_flush(vxlan, &desc); 3158 3159 return 0; 3160 } 3161 3162 /* Cleanup timer and forwarding table on shutdown */ 3163 static int vxlan_stop(struct net_device *dev) 3164 { 3165 struct vxlan_dev *vxlan = netdev_priv(dev); 3166 struct vxlan_fdb_flush_desc desc = { 3167 /* Default entry is deleted at vxlan_uninit. */ 3168 .ignore_default_entry = true, 3169 .state = 0, 3170 .state_mask = NUD_PERMANENT | NUD_NOARP, 3171 }; 3172 3173 vxlan_multicast_leave(vxlan); 3174 3175 del_timer_sync(&vxlan->age_timer); 3176 3177 vxlan_flush(vxlan, &desc); 3178 vxlan_sock_release(vxlan); 3179 3180 return 0; 3181 } 3182 3183 /* Stub, nothing needs to be done. */ 3184 static void vxlan_set_multicast_list(struct net_device *dev) 3185 { 3186 } 3187 3188 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3189 { 3190 struct vxlan_dev *vxlan = netdev_priv(dev); 3191 struct vxlan_rdst *dst = &vxlan->default_dst; 3192 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3193 dst->remote_ifindex); 3194 3195 /* This check is different than dev->max_mtu, because it looks at 3196 * the lowerdev->mtu, rather than the static dev->max_mtu 3197 */ 3198 if (lowerdev) { 3199 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3200 if (new_mtu > max_mtu) 3201 return -EINVAL; 3202 } 3203 3204 WRITE_ONCE(dev->mtu, new_mtu); 3205 return 0; 3206 } 3207 3208 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3209 { 3210 struct vxlan_dev *vxlan = netdev_priv(dev); 3211 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3212 __be16 sport, dport; 3213 3214 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3215 vxlan->cfg.port_max, true); 3216 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3217 3218 if (ip_tunnel_info_af(info) == AF_INET) { 3219 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3220 struct rtable *rt; 3221 3222 if (!sock4) 3223 return -EIO; 3224 3225 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3226 &info->key.u.ipv4.src, 3227 &info->key, 3228 sport, dport, info->key.tos, 3229 &info->dst_cache); 3230 if (IS_ERR(rt)) 3231 return PTR_ERR(rt); 3232 ip_rt_put(rt); 3233 } else { 3234 #if IS_ENABLED(CONFIG_IPV6) 3235 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3236 struct dst_entry *ndst; 3237 3238 if (!sock6) 3239 return -EIO; 3240 3241 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3242 0, &info->key.u.ipv6.src, 3243 &info->key, 3244 sport, dport, info->key.tos, 3245 &info->dst_cache); 3246 if (IS_ERR(ndst)) 3247 return PTR_ERR(ndst); 3248 dst_release(ndst); 3249 #else /* !CONFIG_IPV6 */ 3250 return -EPFNOSUPPORT; 3251 #endif 3252 } 3253 info->key.tp_src = sport; 3254 info->key.tp_dst = dport; 3255 return 0; 3256 } 3257 3258 static const struct net_device_ops vxlan_netdev_ether_ops = { 3259 .ndo_init = vxlan_init, 3260 .ndo_uninit = vxlan_uninit, 3261 .ndo_open = vxlan_open, 3262 .ndo_stop = vxlan_stop, 3263 .ndo_start_xmit = vxlan_xmit, 3264 .ndo_set_rx_mode = vxlan_set_multicast_list, 3265 .ndo_change_mtu = vxlan_change_mtu, 3266 .ndo_validate_addr = eth_validate_addr, 3267 .ndo_set_mac_address = eth_mac_addr, 3268 .ndo_fdb_add = vxlan_fdb_add, 3269 .ndo_fdb_del = vxlan_fdb_delete, 3270 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3271 .ndo_fdb_dump = vxlan_fdb_dump, 3272 .ndo_fdb_get = vxlan_fdb_get, 3273 .ndo_mdb_add = vxlan_mdb_add, 3274 .ndo_mdb_del = vxlan_mdb_del, 3275 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3276 .ndo_mdb_dump = vxlan_mdb_dump, 3277 .ndo_mdb_get = vxlan_mdb_get, 3278 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3279 }; 3280 3281 static const struct net_device_ops vxlan_netdev_raw_ops = { 3282 .ndo_init = vxlan_init, 3283 .ndo_uninit = vxlan_uninit, 3284 .ndo_open = vxlan_open, 3285 .ndo_stop = vxlan_stop, 3286 .ndo_start_xmit = vxlan_xmit, 3287 .ndo_change_mtu = vxlan_change_mtu, 3288 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3289 }; 3290 3291 /* Info for udev, that this is a virtual tunnel endpoint */ 3292 static const struct device_type vxlan_type = { 3293 .name = "vxlan", 3294 }; 3295 3296 /* Calls the ndo_udp_tunnel_add of the caller in order to 3297 * supply the listening VXLAN udp ports. Callers are expected 3298 * to implement the ndo_udp_tunnel_add. 3299 */ 3300 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3301 { 3302 struct vxlan_sock *vs; 3303 struct net *net = dev_net(dev); 3304 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3305 unsigned int i; 3306 3307 spin_lock(&vn->sock_lock); 3308 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3309 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3310 unsigned short type; 3311 3312 if (vs->flags & VXLAN_F_GPE) 3313 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3314 else 3315 type = UDP_TUNNEL_TYPE_VXLAN; 3316 3317 if (push) 3318 udp_tunnel_push_rx_port(dev, vs->sock, type); 3319 else 3320 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3321 } 3322 } 3323 spin_unlock(&vn->sock_lock); 3324 } 3325 3326 /* Initialize the device structure. */ 3327 static void vxlan_setup(struct net_device *dev) 3328 { 3329 struct vxlan_dev *vxlan = netdev_priv(dev); 3330 unsigned int h; 3331 3332 eth_hw_addr_random(dev); 3333 ether_setup(dev); 3334 3335 dev->needs_free_netdev = true; 3336 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3337 3338 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3339 dev->features |= NETIF_F_RXCSUM; 3340 dev->features |= NETIF_F_GSO_SOFTWARE; 3341 3342 dev->vlan_features = dev->features; 3343 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3344 dev->hw_features |= NETIF_F_RXCSUM; 3345 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3346 netif_keep_dst(dev); 3347 dev->priv_flags |= IFF_NO_QUEUE; 3348 dev->change_proto_down = true; 3349 dev->lltx = true; 3350 3351 /* MTU range: 68 - 65535 */ 3352 dev->min_mtu = ETH_MIN_MTU; 3353 dev->max_mtu = ETH_MAX_MTU; 3354 3355 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3356 INIT_LIST_HEAD(&vxlan->next); 3357 3358 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3359 3360 vxlan->dev = dev; 3361 3362 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3363 spin_lock_init(&vxlan->hash_lock[h]); 3364 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3365 } 3366 } 3367 3368 static void vxlan_ether_setup(struct net_device *dev) 3369 { 3370 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3371 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3372 dev->netdev_ops = &vxlan_netdev_ether_ops; 3373 } 3374 3375 static void vxlan_raw_setup(struct net_device *dev) 3376 { 3377 dev->header_ops = NULL; 3378 dev->type = ARPHRD_NONE; 3379 dev->hard_header_len = 0; 3380 dev->addr_len = 0; 3381 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3382 dev->netdev_ops = &vxlan_netdev_raw_ops; 3383 } 3384 3385 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3386 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3387 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3388 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3389 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3390 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3391 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3392 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3393 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3394 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3395 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3396 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3397 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3398 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3399 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3400 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3401 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3402 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3403 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3404 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3405 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3406 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3407 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3408 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3409 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3410 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3411 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3412 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3413 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3414 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3415 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3416 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3417 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3418 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3419 }; 3420 3421 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3422 struct netlink_ext_ack *extack) 3423 { 3424 if (tb[IFLA_ADDRESS]) { 3425 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3426 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3427 "Provided link layer address is not Ethernet"); 3428 return -EINVAL; 3429 } 3430 3431 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3432 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3433 "Provided Ethernet address is not unicast"); 3434 return -EADDRNOTAVAIL; 3435 } 3436 } 3437 3438 if (tb[IFLA_MTU]) { 3439 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3440 3441 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3442 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3443 "MTU must be between 68 and 65535"); 3444 return -EINVAL; 3445 } 3446 } 3447 3448 if (!data) { 3449 NL_SET_ERR_MSG(extack, 3450 "Required attributes not provided to perform the operation"); 3451 return -EINVAL; 3452 } 3453 3454 if (data[IFLA_VXLAN_ID]) { 3455 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3456 3457 if (id >= VXLAN_N_VID) { 3458 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3459 "VXLAN ID must be lower than 16777216"); 3460 return -ERANGE; 3461 } 3462 } 3463 3464 if (data[IFLA_VXLAN_PORT_RANGE]) { 3465 const struct ifla_vxlan_port_range *p 3466 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3467 3468 if (ntohs(p->high) < ntohs(p->low)) { 3469 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3470 "Invalid source port range"); 3471 return -EINVAL; 3472 } 3473 } 3474 3475 if (data[IFLA_VXLAN_DF]) { 3476 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3477 3478 if (df < 0 || df > VXLAN_DF_MAX) { 3479 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3480 "Invalid DF attribute"); 3481 return -EINVAL; 3482 } 3483 } 3484 3485 return 0; 3486 } 3487 3488 static void vxlan_get_drvinfo(struct net_device *netdev, 3489 struct ethtool_drvinfo *drvinfo) 3490 { 3491 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3492 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3493 } 3494 3495 static int vxlan_get_link_ksettings(struct net_device *dev, 3496 struct ethtool_link_ksettings *cmd) 3497 { 3498 struct vxlan_dev *vxlan = netdev_priv(dev); 3499 struct vxlan_rdst *dst = &vxlan->default_dst; 3500 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3501 dst->remote_ifindex); 3502 3503 if (!lowerdev) { 3504 cmd->base.duplex = DUPLEX_UNKNOWN; 3505 cmd->base.port = PORT_OTHER; 3506 cmd->base.speed = SPEED_UNKNOWN; 3507 3508 return 0; 3509 } 3510 3511 return __ethtool_get_link_ksettings(lowerdev, cmd); 3512 } 3513 3514 static const struct ethtool_ops vxlan_ethtool_ops = { 3515 .get_drvinfo = vxlan_get_drvinfo, 3516 .get_link = ethtool_op_get_link, 3517 .get_link_ksettings = vxlan_get_link_ksettings, 3518 }; 3519 3520 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3521 __be16 port, u32 flags, int ifindex) 3522 { 3523 struct socket *sock; 3524 struct udp_port_cfg udp_conf; 3525 int err; 3526 3527 memset(&udp_conf, 0, sizeof(udp_conf)); 3528 3529 if (ipv6) { 3530 udp_conf.family = AF_INET6; 3531 udp_conf.use_udp6_rx_checksums = 3532 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3533 udp_conf.ipv6_v6only = 1; 3534 } else { 3535 udp_conf.family = AF_INET; 3536 } 3537 3538 udp_conf.local_udp_port = port; 3539 udp_conf.bind_ifindex = ifindex; 3540 3541 /* Open UDP socket */ 3542 err = udp_sock_create(net, &udp_conf, &sock); 3543 if (err < 0) 3544 return ERR_PTR(err); 3545 3546 udp_allow_gso(sock->sk); 3547 return sock; 3548 } 3549 3550 /* Create new listen socket if needed */ 3551 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3552 __be16 port, u32 flags, 3553 int ifindex) 3554 { 3555 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3556 struct vxlan_sock *vs; 3557 struct socket *sock; 3558 unsigned int h; 3559 struct udp_tunnel_sock_cfg tunnel_cfg; 3560 3561 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3562 if (!vs) 3563 return ERR_PTR(-ENOMEM); 3564 3565 for (h = 0; h < VNI_HASH_SIZE; ++h) 3566 INIT_HLIST_HEAD(&vs->vni_list[h]); 3567 3568 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3569 if (IS_ERR(sock)) { 3570 kfree(vs); 3571 return ERR_CAST(sock); 3572 } 3573 3574 vs->sock = sock; 3575 refcount_set(&vs->refcnt, 1); 3576 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3577 3578 spin_lock(&vn->sock_lock); 3579 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3580 udp_tunnel_notify_add_rx_port(sock, 3581 (vs->flags & VXLAN_F_GPE) ? 3582 UDP_TUNNEL_TYPE_VXLAN_GPE : 3583 UDP_TUNNEL_TYPE_VXLAN); 3584 spin_unlock(&vn->sock_lock); 3585 3586 /* Mark socket as an encapsulation socket. */ 3587 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3588 tunnel_cfg.sk_user_data = vs; 3589 tunnel_cfg.encap_type = 1; 3590 tunnel_cfg.encap_rcv = vxlan_rcv; 3591 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3592 tunnel_cfg.encap_destroy = NULL; 3593 if (vs->flags & VXLAN_F_GPE) { 3594 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3595 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3596 } else { 3597 tunnel_cfg.gro_receive = vxlan_gro_receive; 3598 tunnel_cfg.gro_complete = vxlan_gro_complete; 3599 } 3600 3601 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3602 3603 return vs; 3604 } 3605 3606 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3607 { 3608 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3609 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3610 struct vxlan_sock *vs = NULL; 3611 struct vxlan_dev_node *node; 3612 int l3mdev_index = 0; 3613 3614 if (vxlan->cfg.remote_ifindex) 3615 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3616 vxlan->net, vxlan->cfg.remote_ifindex); 3617 3618 if (!vxlan->cfg.no_share) { 3619 spin_lock(&vn->sock_lock); 3620 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3621 vxlan->cfg.dst_port, vxlan->cfg.flags, 3622 l3mdev_index); 3623 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3624 spin_unlock(&vn->sock_lock); 3625 return -EBUSY; 3626 } 3627 spin_unlock(&vn->sock_lock); 3628 } 3629 if (!vs) 3630 vs = vxlan_socket_create(vxlan->net, ipv6, 3631 vxlan->cfg.dst_port, vxlan->cfg.flags, 3632 l3mdev_index); 3633 if (IS_ERR(vs)) 3634 return PTR_ERR(vs); 3635 #if IS_ENABLED(CONFIG_IPV6) 3636 if (ipv6) { 3637 rcu_assign_pointer(vxlan->vn6_sock, vs); 3638 node = &vxlan->hlist6; 3639 } else 3640 #endif 3641 { 3642 rcu_assign_pointer(vxlan->vn4_sock, vs); 3643 node = &vxlan->hlist4; 3644 } 3645 3646 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3647 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3648 else 3649 vxlan_vs_add_dev(vs, vxlan, node); 3650 3651 return 0; 3652 } 3653 3654 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3655 { 3656 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3657 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3658 bool ipv4 = !ipv6 || metadata; 3659 int ret = 0; 3660 3661 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3662 #if IS_ENABLED(CONFIG_IPV6) 3663 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3664 if (ipv6) { 3665 ret = __vxlan_sock_add(vxlan, true); 3666 if (ret < 0 && ret != -EAFNOSUPPORT) 3667 ipv4 = false; 3668 } 3669 #endif 3670 if (ipv4) 3671 ret = __vxlan_sock_add(vxlan, false); 3672 if (ret < 0) 3673 vxlan_sock_release(vxlan); 3674 return ret; 3675 } 3676 3677 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3678 struct vxlan_config *conf, __be32 vni) 3679 { 3680 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3681 struct vxlan_dev *tmp; 3682 3683 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3684 if (tmp == vxlan) 3685 continue; 3686 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3687 if (!vxlan_vnifilter_lookup(tmp, vni)) 3688 continue; 3689 } else if (tmp->cfg.vni != vni) { 3690 continue; 3691 } 3692 if (tmp->cfg.dst_port != conf->dst_port) 3693 continue; 3694 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3695 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3696 continue; 3697 3698 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3699 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3700 continue; 3701 3702 return -EEXIST; 3703 } 3704 3705 return 0; 3706 } 3707 3708 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3709 struct net_device **lower, 3710 struct vxlan_dev *old, 3711 struct netlink_ext_ack *extack) 3712 { 3713 bool use_ipv6 = false; 3714 3715 if (conf->flags & VXLAN_F_GPE) { 3716 /* For now, allow GPE only together with 3717 * COLLECT_METADATA. This can be relaxed later; in such 3718 * case, the other side of the PtP link will have to be 3719 * provided. 3720 */ 3721 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3722 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3723 NL_SET_ERR_MSG(extack, 3724 "VXLAN GPE does not support this combination of attributes"); 3725 return -EINVAL; 3726 } 3727 } 3728 3729 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3730 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3731 conf->remote_ip.sa.sa_family = AF_INET; 3732 conf->saddr.sa.sa_family = AF_INET; 3733 } else if (!conf->remote_ip.sa.sa_family) { 3734 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3735 } else if (!conf->saddr.sa.sa_family) { 3736 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3737 } 3738 3739 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3740 NL_SET_ERR_MSG(extack, 3741 "Local and remote address must be from the same family"); 3742 return -EINVAL; 3743 } 3744 3745 if (vxlan_addr_multicast(&conf->saddr)) { 3746 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3747 return -EINVAL; 3748 } 3749 3750 if (conf->saddr.sa.sa_family == AF_INET6) { 3751 if (!IS_ENABLED(CONFIG_IPV6)) { 3752 NL_SET_ERR_MSG(extack, 3753 "IPv6 support not enabled in the kernel"); 3754 return -EPFNOSUPPORT; 3755 } 3756 use_ipv6 = true; 3757 conf->flags |= VXLAN_F_IPV6; 3758 3759 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3760 int local_type = 3761 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3762 int remote_type = 3763 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3764 3765 if (local_type & IPV6_ADDR_LINKLOCAL) { 3766 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3767 (remote_type != IPV6_ADDR_ANY)) { 3768 NL_SET_ERR_MSG(extack, 3769 "Invalid combination of local and remote address scopes"); 3770 return -EINVAL; 3771 } 3772 3773 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3774 } else { 3775 if (remote_type == 3776 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3777 NL_SET_ERR_MSG(extack, 3778 "Invalid combination of local and remote address scopes"); 3779 return -EINVAL; 3780 } 3781 3782 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3783 } 3784 } 3785 } 3786 3787 if (conf->label && !use_ipv6) { 3788 NL_SET_ERR_MSG(extack, 3789 "Label attribute only applies to IPv6 VXLAN devices"); 3790 return -EINVAL; 3791 } 3792 3793 if (conf->label_policy && !use_ipv6) { 3794 NL_SET_ERR_MSG(extack, 3795 "Label policy only applies to IPv6 VXLAN devices"); 3796 return -EINVAL; 3797 } 3798 3799 if (conf->remote_ifindex) { 3800 struct net_device *lowerdev; 3801 3802 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3803 if (!lowerdev) { 3804 NL_SET_ERR_MSG(extack, 3805 "Invalid local interface, device not found"); 3806 return -ENODEV; 3807 } 3808 3809 #if IS_ENABLED(CONFIG_IPV6) 3810 if (use_ipv6) { 3811 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3812 3813 if (idev && idev->cnf.disable_ipv6) { 3814 NL_SET_ERR_MSG(extack, 3815 "IPv6 support disabled by administrator"); 3816 return -EPERM; 3817 } 3818 } 3819 #endif 3820 3821 *lower = lowerdev; 3822 } else { 3823 if (vxlan_addr_multicast(&conf->remote_ip)) { 3824 NL_SET_ERR_MSG(extack, 3825 "Local interface required for multicast remote destination"); 3826 3827 return -EINVAL; 3828 } 3829 3830 #if IS_ENABLED(CONFIG_IPV6) 3831 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3832 NL_SET_ERR_MSG(extack, 3833 "Local interface required for link-local local/remote addresses"); 3834 return -EINVAL; 3835 } 3836 #endif 3837 3838 *lower = NULL; 3839 } 3840 3841 if (!conf->dst_port) { 3842 if (conf->flags & VXLAN_F_GPE) 3843 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3844 else 3845 conf->dst_port = htons(vxlan_port); 3846 } 3847 3848 if (!conf->age_interval) 3849 conf->age_interval = FDB_AGE_DEFAULT; 3850 3851 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3852 NL_SET_ERR_MSG(extack, 3853 "A VXLAN device with the specified VNI already exists"); 3854 return -EEXIST; 3855 } 3856 3857 return 0; 3858 } 3859 3860 static void vxlan_config_apply(struct net_device *dev, 3861 struct vxlan_config *conf, 3862 struct net_device *lowerdev, 3863 struct net *src_net, 3864 bool changelink) 3865 { 3866 struct vxlan_dev *vxlan = netdev_priv(dev); 3867 struct vxlan_rdst *dst = &vxlan->default_dst; 3868 unsigned short needed_headroom = ETH_HLEN; 3869 int max_mtu = ETH_MAX_MTU; 3870 u32 flags = conf->flags; 3871 3872 if (!changelink) { 3873 if (flags & VXLAN_F_GPE) 3874 vxlan_raw_setup(dev); 3875 else 3876 vxlan_ether_setup(dev); 3877 3878 if (conf->mtu) 3879 dev->mtu = conf->mtu; 3880 3881 vxlan->net = src_net; 3882 } 3883 3884 dst->remote_vni = conf->vni; 3885 3886 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3887 3888 if (lowerdev) { 3889 dst->remote_ifindex = conf->remote_ifindex; 3890 3891 netif_inherit_tso_max(dev, lowerdev); 3892 3893 needed_headroom = lowerdev->hard_header_len; 3894 needed_headroom += lowerdev->needed_headroom; 3895 3896 dev->needed_tailroom = lowerdev->needed_tailroom; 3897 3898 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3899 if (max_mtu < ETH_MIN_MTU) 3900 max_mtu = ETH_MIN_MTU; 3901 3902 if (!changelink && !conf->mtu) 3903 dev->mtu = max_mtu; 3904 } 3905 3906 if (dev->mtu > max_mtu) 3907 dev->mtu = max_mtu; 3908 3909 if (flags & VXLAN_F_COLLECT_METADATA) 3910 flags |= VXLAN_F_IPV6; 3911 needed_headroom += vxlan_headroom(flags); 3912 dev->needed_headroom = needed_headroom; 3913 3914 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3915 } 3916 3917 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3918 struct vxlan_config *conf, bool changelink, 3919 struct netlink_ext_ack *extack) 3920 { 3921 struct vxlan_dev *vxlan = netdev_priv(dev); 3922 struct net_device *lowerdev; 3923 int ret; 3924 3925 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3926 if (ret) 3927 return ret; 3928 3929 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3930 3931 return 0; 3932 } 3933 3934 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3935 struct vxlan_config *conf, 3936 struct netlink_ext_ack *extack) 3937 { 3938 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3939 struct vxlan_dev *vxlan = netdev_priv(dev); 3940 struct net_device *remote_dev = NULL; 3941 struct vxlan_fdb *f = NULL; 3942 bool unregister = false; 3943 struct vxlan_rdst *dst; 3944 int err; 3945 3946 dst = &vxlan->default_dst; 3947 err = vxlan_dev_configure(net, dev, conf, false, extack); 3948 if (err) 3949 return err; 3950 3951 dev->ethtool_ops = &vxlan_ethtool_ops; 3952 3953 /* create an fdb entry for a valid default destination */ 3954 if (!vxlan_addr_any(&dst->remote_ip)) { 3955 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3956 &dst->remote_ip, 3957 NUD_REACHABLE | NUD_PERMANENT, 3958 vxlan->cfg.dst_port, 3959 dst->remote_vni, 3960 dst->remote_vni, 3961 dst->remote_ifindex, 3962 NTF_SELF, 0, &f, extack); 3963 if (err) 3964 return err; 3965 } 3966 3967 err = register_netdevice(dev); 3968 if (err) 3969 goto errout; 3970 unregister = true; 3971 3972 if (dst->remote_ifindex) { 3973 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3974 if (!remote_dev) { 3975 err = -ENODEV; 3976 goto errout; 3977 } 3978 3979 err = netdev_upper_dev_link(remote_dev, dev, extack); 3980 if (err) 3981 goto errout; 3982 } 3983 3984 err = rtnl_configure_link(dev, NULL, 0, NULL); 3985 if (err < 0) 3986 goto unlink; 3987 3988 if (f) { 3989 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3990 3991 /* notify default fdb entry */ 3992 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3993 RTM_NEWNEIGH, true, extack); 3994 if (err) { 3995 vxlan_fdb_destroy(vxlan, f, false, false); 3996 if (remote_dev) 3997 netdev_upper_dev_unlink(remote_dev, dev); 3998 goto unregister; 3999 } 4000 } 4001 4002 list_add(&vxlan->next, &vn->vxlan_list); 4003 if (remote_dev) 4004 dst->remote_dev = remote_dev; 4005 return 0; 4006 unlink: 4007 if (remote_dev) 4008 netdev_upper_dev_unlink(remote_dev, dev); 4009 errout: 4010 /* unregister_netdevice() destroys the default FDB entry with deletion 4011 * notification. But the addition notification was not sent yet, so 4012 * destroy the entry by hand here. 4013 */ 4014 if (f) 4015 __vxlan_fdb_free(f); 4016 unregister: 4017 if (unregister) 4018 unregister_netdevice(dev); 4019 return err; 4020 } 4021 4022 /* Set/clear flags based on attribute */ 4023 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4024 int attrtype, unsigned long mask, bool changelink, 4025 bool changelink_supported, 4026 struct netlink_ext_ack *extack) 4027 { 4028 unsigned long flags; 4029 4030 if (!tb[attrtype]) 4031 return 0; 4032 4033 if (changelink && !changelink_supported) { 4034 vxlan_flag_attr_error(attrtype, extack); 4035 return -EOPNOTSUPP; 4036 } 4037 4038 if (vxlan_policy[attrtype].type == NLA_FLAG) 4039 flags = conf->flags | mask; 4040 else if (nla_get_u8(tb[attrtype])) 4041 flags = conf->flags | mask; 4042 else 4043 flags = conf->flags & ~mask; 4044 4045 conf->flags = flags; 4046 4047 return 0; 4048 } 4049 4050 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4051 struct net_device *dev, struct vxlan_config *conf, 4052 bool changelink, struct netlink_ext_ack *extack) 4053 { 4054 struct vxlan_dev *vxlan = netdev_priv(dev); 4055 int err = 0; 4056 4057 memset(conf, 0, sizeof(*conf)); 4058 4059 /* if changelink operation, start with old existing cfg */ 4060 if (changelink) 4061 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4062 4063 if (data[IFLA_VXLAN_ID]) { 4064 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4065 4066 if (changelink && (vni != conf->vni)) { 4067 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4068 return -EOPNOTSUPP; 4069 } 4070 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4071 } 4072 4073 if (data[IFLA_VXLAN_GROUP]) { 4074 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4075 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4076 return -EOPNOTSUPP; 4077 } 4078 4079 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4080 conf->remote_ip.sa.sa_family = AF_INET; 4081 } else if (data[IFLA_VXLAN_GROUP6]) { 4082 if (!IS_ENABLED(CONFIG_IPV6)) { 4083 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4084 return -EPFNOSUPPORT; 4085 } 4086 4087 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4088 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4089 return -EOPNOTSUPP; 4090 } 4091 4092 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4093 conf->remote_ip.sa.sa_family = AF_INET6; 4094 } 4095 4096 if (data[IFLA_VXLAN_LOCAL]) { 4097 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4098 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4099 return -EOPNOTSUPP; 4100 } 4101 4102 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4103 conf->saddr.sa.sa_family = AF_INET; 4104 } else if (data[IFLA_VXLAN_LOCAL6]) { 4105 if (!IS_ENABLED(CONFIG_IPV6)) { 4106 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4107 return -EPFNOSUPPORT; 4108 } 4109 4110 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4111 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4112 return -EOPNOTSUPP; 4113 } 4114 4115 /* TODO: respect scope id */ 4116 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4117 conf->saddr.sa.sa_family = AF_INET6; 4118 } 4119 4120 if (data[IFLA_VXLAN_LINK]) 4121 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4122 4123 if (data[IFLA_VXLAN_TOS]) 4124 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4125 4126 if (data[IFLA_VXLAN_TTL]) 4127 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4128 4129 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4130 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4131 VXLAN_F_TTL_INHERIT, changelink, false, 4132 extack); 4133 if (err) 4134 return err; 4135 4136 } 4137 4138 if (data[IFLA_VXLAN_LABEL]) 4139 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4140 IPV6_FLOWLABEL_MASK; 4141 if (data[IFLA_VXLAN_LABEL_POLICY]) 4142 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4143 4144 if (data[IFLA_VXLAN_LEARNING]) { 4145 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4146 VXLAN_F_LEARN, changelink, true, 4147 extack); 4148 if (err) 4149 return err; 4150 } else if (!changelink) { 4151 /* default to learn on a new device */ 4152 conf->flags |= VXLAN_F_LEARN; 4153 } 4154 4155 if (data[IFLA_VXLAN_AGEING]) 4156 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4157 4158 if (data[IFLA_VXLAN_PROXY]) { 4159 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4160 VXLAN_F_PROXY, changelink, false, 4161 extack); 4162 if (err) 4163 return err; 4164 } 4165 4166 if (data[IFLA_VXLAN_RSC]) { 4167 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4168 VXLAN_F_RSC, changelink, false, 4169 extack); 4170 if (err) 4171 return err; 4172 } 4173 4174 if (data[IFLA_VXLAN_L2MISS]) { 4175 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4176 VXLAN_F_L2MISS, changelink, false, 4177 extack); 4178 if (err) 4179 return err; 4180 } 4181 4182 if (data[IFLA_VXLAN_L3MISS]) { 4183 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4184 VXLAN_F_L3MISS, changelink, false, 4185 extack); 4186 if (err) 4187 return err; 4188 } 4189 4190 if (data[IFLA_VXLAN_LIMIT]) { 4191 if (changelink) { 4192 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4193 "Cannot change limit"); 4194 return -EOPNOTSUPP; 4195 } 4196 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4197 } 4198 4199 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4200 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4201 VXLAN_F_COLLECT_METADATA, changelink, false, 4202 extack); 4203 if (err) 4204 return err; 4205 } 4206 4207 if (data[IFLA_VXLAN_PORT_RANGE]) { 4208 if (!changelink) { 4209 const struct ifla_vxlan_port_range *p 4210 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4211 conf->port_min = ntohs(p->low); 4212 conf->port_max = ntohs(p->high); 4213 } else { 4214 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4215 "Cannot change port range"); 4216 return -EOPNOTSUPP; 4217 } 4218 } 4219 4220 if (data[IFLA_VXLAN_PORT]) { 4221 if (changelink) { 4222 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4223 "Cannot change port"); 4224 return -EOPNOTSUPP; 4225 } 4226 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4227 } 4228 4229 if (data[IFLA_VXLAN_UDP_CSUM]) { 4230 if (changelink) { 4231 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4232 "Cannot change UDP_CSUM flag"); 4233 return -EOPNOTSUPP; 4234 } 4235 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4236 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4237 } 4238 4239 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4240 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4241 VXLAN_F_LOCALBYPASS, changelink, 4242 true, extack); 4243 if (err) 4244 return err; 4245 } else if (!changelink) { 4246 /* default to local bypass on a new device */ 4247 conf->flags |= VXLAN_F_LOCALBYPASS; 4248 } 4249 4250 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4251 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4252 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4253 false, extack); 4254 if (err) 4255 return err; 4256 } 4257 4258 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4259 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4260 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4261 false, extack); 4262 if (err) 4263 return err; 4264 } 4265 4266 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4267 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4268 VXLAN_F_REMCSUM_TX, changelink, false, 4269 extack); 4270 if (err) 4271 return err; 4272 } 4273 4274 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4275 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4276 VXLAN_F_REMCSUM_RX, changelink, false, 4277 extack); 4278 if (err) 4279 return err; 4280 } 4281 4282 if (data[IFLA_VXLAN_GBP]) { 4283 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4284 VXLAN_F_GBP, changelink, false, extack); 4285 if (err) 4286 return err; 4287 } 4288 4289 if (data[IFLA_VXLAN_GPE]) { 4290 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4291 VXLAN_F_GPE, changelink, false, 4292 extack); 4293 if (err) 4294 return err; 4295 } 4296 4297 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4298 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4299 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4300 false, extack); 4301 if (err) 4302 return err; 4303 } 4304 4305 if (tb[IFLA_MTU]) { 4306 if (changelink) { 4307 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4308 "Cannot change mtu"); 4309 return -EOPNOTSUPP; 4310 } 4311 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4312 } 4313 4314 if (data[IFLA_VXLAN_DF]) 4315 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4316 4317 if (data[IFLA_VXLAN_VNIFILTER]) { 4318 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4319 VXLAN_F_VNIFILTER, changelink, false, 4320 extack); 4321 if (err) 4322 return err; 4323 4324 if ((conf->flags & VXLAN_F_VNIFILTER) && 4325 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4326 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4327 "vxlan vnifilter only valid in collect metadata mode"); 4328 return -EINVAL; 4329 } 4330 } 4331 4332 return 0; 4333 } 4334 4335 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4336 struct nlattr *tb[], struct nlattr *data[], 4337 struct netlink_ext_ack *extack) 4338 { 4339 struct vxlan_config conf; 4340 int err; 4341 4342 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4343 if (err) 4344 return err; 4345 4346 return __vxlan_dev_create(src_net, dev, &conf, extack); 4347 } 4348 4349 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4350 struct nlattr *data[], 4351 struct netlink_ext_ack *extack) 4352 { 4353 struct vxlan_dev *vxlan = netdev_priv(dev); 4354 struct net_device *lowerdev; 4355 struct vxlan_config conf; 4356 struct vxlan_rdst *dst; 4357 int err; 4358 4359 dst = &vxlan->default_dst; 4360 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4361 if (err) 4362 return err; 4363 4364 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4365 vxlan, extack); 4366 if (err) 4367 return err; 4368 4369 if (dst->remote_dev == lowerdev) 4370 lowerdev = NULL; 4371 4372 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4373 extack); 4374 if (err) 4375 return err; 4376 4377 /* handle default dst entry */ 4378 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4379 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4380 4381 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4382 if (!vxlan_addr_any(&conf.remote_ip)) { 4383 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4384 &conf.remote_ip, 4385 NUD_REACHABLE | NUD_PERMANENT, 4386 NLM_F_APPEND | NLM_F_CREATE, 4387 vxlan->cfg.dst_port, 4388 conf.vni, conf.vni, 4389 conf.remote_ifindex, 4390 NTF_SELF, 0, true, extack); 4391 if (err) { 4392 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4393 netdev_adjacent_change_abort(dst->remote_dev, 4394 lowerdev, dev); 4395 return err; 4396 } 4397 } 4398 if (!vxlan_addr_any(&dst->remote_ip)) 4399 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4400 dst->remote_ip, 4401 vxlan->cfg.dst_port, 4402 dst->remote_vni, 4403 dst->remote_vni, 4404 dst->remote_ifindex, 4405 true); 4406 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4407 4408 /* If vni filtering device, also update fdb entries of 4409 * all vnis that were using default remote ip 4410 */ 4411 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4412 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4413 &conf.remote_ip, extack); 4414 if (err) { 4415 netdev_adjacent_change_abort(dst->remote_dev, 4416 lowerdev, dev); 4417 return err; 4418 } 4419 } 4420 } 4421 4422 if (conf.age_interval != vxlan->cfg.age_interval) 4423 mod_timer(&vxlan->age_timer, jiffies); 4424 4425 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4426 if (lowerdev && lowerdev != dst->remote_dev) 4427 dst->remote_dev = lowerdev; 4428 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4429 return 0; 4430 } 4431 4432 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4433 { 4434 struct vxlan_dev *vxlan = netdev_priv(dev); 4435 struct vxlan_fdb_flush_desc desc = { 4436 /* Default entry is deleted at vxlan_uninit. */ 4437 .ignore_default_entry = true, 4438 }; 4439 4440 vxlan_flush(vxlan, &desc); 4441 4442 list_del(&vxlan->next); 4443 unregister_netdevice_queue(dev, head); 4444 if (vxlan->default_dst.remote_dev) 4445 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4446 } 4447 4448 static size_t vxlan_get_size(const struct net_device *dev) 4449 { 4450 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4451 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4452 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4453 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4454 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4455 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4456 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4457 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4458 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4459 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4460 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4461 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4462 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4463 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4464 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4465 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4466 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4467 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4468 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4469 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4470 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4471 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4472 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4473 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4474 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4475 /* IFLA_VXLAN_PORT_RANGE */ 4476 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4477 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4478 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4479 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4480 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4481 0; 4482 } 4483 4484 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4485 { 4486 const struct vxlan_dev *vxlan = netdev_priv(dev); 4487 const struct vxlan_rdst *dst = &vxlan->default_dst; 4488 struct ifla_vxlan_port_range ports = { 4489 .low = htons(vxlan->cfg.port_min), 4490 .high = htons(vxlan->cfg.port_max), 4491 }; 4492 4493 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4494 goto nla_put_failure; 4495 4496 if (!vxlan_addr_any(&dst->remote_ip)) { 4497 if (dst->remote_ip.sa.sa_family == AF_INET) { 4498 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4499 dst->remote_ip.sin.sin_addr.s_addr)) 4500 goto nla_put_failure; 4501 #if IS_ENABLED(CONFIG_IPV6) 4502 } else { 4503 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4504 &dst->remote_ip.sin6.sin6_addr)) 4505 goto nla_put_failure; 4506 #endif 4507 } 4508 } 4509 4510 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4511 goto nla_put_failure; 4512 4513 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4514 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4515 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4516 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4517 goto nla_put_failure; 4518 #if IS_ENABLED(CONFIG_IPV6) 4519 } else { 4520 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4521 &vxlan->cfg.saddr.sin6.sin6_addr)) 4522 goto nla_put_failure; 4523 #endif 4524 } 4525 } 4526 4527 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4528 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4529 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4530 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4531 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4532 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4533 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4534 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4535 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4536 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4537 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4538 nla_put_u8(skb, IFLA_VXLAN_RSC, 4539 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4540 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4541 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4542 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4543 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4544 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4545 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4546 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4547 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4548 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4549 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4550 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4551 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4552 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4553 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4554 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4555 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4556 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4557 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4558 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4559 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4560 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4561 goto nla_put_failure; 4562 4563 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4564 goto nla_put_failure; 4565 4566 if (vxlan->cfg.flags & VXLAN_F_GBP && 4567 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4568 goto nla_put_failure; 4569 4570 if (vxlan->cfg.flags & VXLAN_F_GPE && 4571 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4572 goto nla_put_failure; 4573 4574 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4575 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4576 goto nla_put_failure; 4577 4578 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4579 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4580 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4581 goto nla_put_failure; 4582 4583 return 0; 4584 4585 nla_put_failure: 4586 return -EMSGSIZE; 4587 } 4588 4589 static struct net *vxlan_get_link_net(const struct net_device *dev) 4590 { 4591 struct vxlan_dev *vxlan = netdev_priv(dev); 4592 4593 return READ_ONCE(vxlan->net); 4594 } 4595 4596 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4597 .kind = "vxlan", 4598 .maxtype = IFLA_VXLAN_MAX, 4599 .policy = vxlan_policy, 4600 .priv_size = sizeof(struct vxlan_dev), 4601 .setup = vxlan_setup, 4602 .validate = vxlan_validate, 4603 .newlink = vxlan_newlink, 4604 .changelink = vxlan_changelink, 4605 .dellink = vxlan_dellink, 4606 .get_size = vxlan_get_size, 4607 .fill_info = vxlan_fill_info, 4608 .get_link_net = vxlan_get_link_net, 4609 }; 4610 4611 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4612 u8 name_assign_type, 4613 struct vxlan_config *conf) 4614 { 4615 struct nlattr *tb[IFLA_MAX + 1]; 4616 struct net_device *dev; 4617 int err; 4618 4619 memset(&tb, 0, sizeof(tb)); 4620 4621 dev = rtnl_create_link(net, name, name_assign_type, 4622 &vxlan_link_ops, tb, NULL); 4623 if (IS_ERR(dev)) 4624 return dev; 4625 4626 err = __vxlan_dev_create(net, dev, conf, NULL); 4627 if (err < 0) { 4628 free_netdev(dev); 4629 return ERR_PTR(err); 4630 } 4631 4632 err = rtnl_configure_link(dev, NULL, 0, NULL); 4633 if (err < 0) { 4634 LIST_HEAD(list_kill); 4635 4636 vxlan_dellink(dev, &list_kill); 4637 unregister_netdevice_many(&list_kill); 4638 return ERR_PTR(err); 4639 } 4640 4641 return dev; 4642 } 4643 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4644 4645 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4646 struct net_device *dev) 4647 { 4648 struct vxlan_dev *vxlan, *next; 4649 LIST_HEAD(list_kill); 4650 4651 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4652 struct vxlan_rdst *dst = &vxlan->default_dst; 4653 4654 /* In case we created vxlan device with carrier 4655 * and we loose the carrier due to module unload 4656 * we also need to remove vxlan device. In other 4657 * cases, it's not necessary and remote_ifindex 4658 * is 0 here, so no matches. 4659 */ 4660 if (dst->remote_ifindex == dev->ifindex) 4661 vxlan_dellink(vxlan->dev, &list_kill); 4662 } 4663 4664 unregister_netdevice_many(&list_kill); 4665 } 4666 4667 static int vxlan_netdevice_event(struct notifier_block *unused, 4668 unsigned long event, void *ptr) 4669 { 4670 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4671 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4672 4673 if (event == NETDEV_UNREGISTER) 4674 vxlan_handle_lowerdev_unregister(vn, dev); 4675 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4676 vxlan_offload_rx_ports(dev, true); 4677 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4678 vxlan_offload_rx_ports(dev, false); 4679 4680 return NOTIFY_DONE; 4681 } 4682 4683 static struct notifier_block vxlan_notifier_block __read_mostly = { 4684 .notifier_call = vxlan_netdevice_event, 4685 }; 4686 4687 static void 4688 vxlan_fdb_offloaded_set(struct net_device *dev, 4689 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4690 { 4691 struct vxlan_dev *vxlan = netdev_priv(dev); 4692 struct vxlan_rdst *rdst; 4693 struct vxlan_fdb *f; 4694 u32 hash_index; 4695 4696 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4697 4698 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4699 4700 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4701 if (!f) 4702 goto out; 4703 4704 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4705 fdb_info->remote_port, 4706 fdb_info->remote_vni, 4707 fdb_info->remote_ifindex); 4708 if (!rdst) 4709 goto out; 4710 4711 rdst->offloaded = fdb_info->offloaded; 4712 4713 out: 4714 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4715 } 4716 4717 static int 4718 vxlan_fdb_external_learn_add(struct net_device *dev, 4719 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4720 { 4721 struct vxlan_dev *vxlan = netdev_priv(dev); 4722 struct netlink_ext_ack *extack; 4723 u32 hash_index; 4724 int err; 4725 4726 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4727 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4728 4729 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4730 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4731 NUD_REACHABLE, 4732 NLM_F_CREATE | NLM_F_REPLACE, 4733 fdb_info->remote_port, 4734 fdb_info->vni, 4735 fdb_info->remote_vni, 4736 fdb_info->remote_ifindex, 4737 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4738 0, false, extack); 4739 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4740 4741 return err; 4742 } 4743 4744 static int 4745 vxlan_fdb_external_learn_del(struct net_device *dev, 4746 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4747 { 4748 struct vxlan_dev *vxlan = netdev_priv(dev); 4749 struct vxlan_fdb *f; 4750 u32 hash_index; 4751 int err = 0; 4752 4753 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4754 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4755 4756 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4757 if (!f) 4758 err = -ENOENT; 4759 else if (f->flags & NTF_EXT_LEARNED) 4760 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4761 fdb_info->remote_ip, 4762 fdb_info->remote_port, 4763 fdb_info->vni, 4764 fdb_info->remote_vni, 4765 fdb_info->remote_ifindex, 4766 false); 4767 4768 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4769 4770 return err; 4771 } 4772 4773 static int vxlan_switchdev_event(struct notifier_block *unused, 4774 unsigned long event, void *ptr) 4775 { 4776 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4777 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4778 int err = 0; 4779 4780 switch (event) { 4781 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4782 vxlan_fdb_offloaded_set(dev, ptr); 4783 break; 4784 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4785 fdb_info = ptr; 4786 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4787 if (err) { 4788 err = notifier_from_errno(err); 4789 break; 4790 } 4791 fdb_info->offloaded = true; 4792 vxlan_fdb_offloaded_set(dev, fdb_info); 4793 break; 4794 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4795 fdb_info = ptr; 4796 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4797 if (err) { 4798 err = notifier_from_errno(err); 4799 break; 4800 } 4801 fdb_info->offloaded = false; 4802 vxlan_fdb_offloaded_set(dev, fdb_info); 4803 break; 4804 } 4805 4806 return err; 4807 } 4808 4809 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4810 .notifier_call = vxlan_switchdev_event, 4811 }; 4812 4813 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4814 { 4815 struct vxlan_fdb *fdb; 4816 struct vxlan_dev *vxlan; 4817 u32 hash_index; 4818 4819 rcu_read_lock(); 4820 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4821 vxlan = rcu_dereference(fdb->vdev); 4822 WARN_ON(!vxlan); 4823 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4824 vxlan->default_dst.remote_vni); 4825 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4826 if (!hlist_unhashed(&fdb->hlist)) 4827 vxlan_fdb_destroy(vxlan, fdb, false, false); 4828 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4829 } 4830 rcu_read_unlock(); 4831 } 4832 4833 static int vxlan_nexthop_event(struct notifier_block *nb, 4834 unsigned long event, void *ptr) 4835 { 4836 struct nh_notifier_info *info = ptr; 4837 struct nexthop *nh; 4838 4839 if (event != NEXTHOP_EVENT_DEL) 4840 return NOTIFY_DONE; 4841 4842 nh = nexthop_find_by_id(info->net, info->id); 4843 if (!nh) 4844 return NOTIFY_DONE; 4845 4846 vxlan_fdb_nh_flush(nh); 4847 4848 return NOTIFY_DONE; 4849 } 4850 4851 static __net_init int vxlan_init_net(struct net *net) 4852 { 4853 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4854 unsigned int h; 4855 4856 INIT_LIST_HEAD(&vn->vxlan_list); 4857 spin_lock_init(&vn->sock_lock); 4858 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4859 4860 for (h = 0; h < PORT_HASH_SIZE; ++h) 4861 INIT_HLIST_HEAD(&vn->sock_list[h]); 4862 4863 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4864 NULL); 4865 } 4866 4867 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4868 struct list_head *dev_to_kill) 4869 { 4870 struct vxlan_dev *vxlan, *next; 4871 4872 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4873 vxlan_dellink(vxlan->dev, dev_to_kill); 4874 } 4875 4876 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4877 struct list_head *dev_to_kill) 4878 { 4879 struct net *net; 4880 4881 ASSERT_RTNL(); 4882 list_for_each_entry(net, net_list, exit_list) { 4883 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4884 4885 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4886 4887 vxlan_destroy_tunnels(vn, dev_to_kill); 4888 } 4889 } 4890 4891 static void __net_exit vxlan_exit_net(struct net *net) 4892 { 4893 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4894 unsigned int h; 4895 4896 for (h = 0; h < PORT_HASH_SIZE; ++h) 4897 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4898 } 4899 4900 static struct pernet_operations vxlan_net_ops = { 4901 .init = vxlan_init_net, 4902 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4903 .exit = vxlan_exit_net, 4904 .id = &vxlan_net_id, 4905 .size = sizeof(struct vxlan_net), 4906 }; 4907 4908 static int __init vxlan_init_module(void) 4909 { 4910 int rc; 4911 4912 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4913 4914 rc = register_pernet_subsys(&vxlan_net_ops); 4915 if (rc) 4916 goto out1; 4917 4918 rc = register_netdevice_notifier(&vxlan_notifier_block); 4919 if (rc) 4920 goto out2; 4921 4922 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4923 if (rc) 4924 goto out3; 4925 4926 rc = rtnl_link_register(&vxlan_link_ops); 4927 if (rc) 4928 goto out4; 4929 4930 rc = vxlan_vnifilter_init(); 4931 if (rc) 4932 goto out5; 4933 4934 return 0; 4935 out5: 4936 rtnl_link_unregister(&vxlan_link_ops); 4937 out4: 4938 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4939 out3: 4940 unregister_netdevice_notifier(&vxlan_notifier_block); 4941 out2: 4942 unregister_pernet_subsys(&vxlan_net_ops); 4943 out1: 4944 return rc; 4945 } 4946 late_initcall(vxlan_init_module); 4947 4948 static void __exit vxlan_cleanup_module(void) 4949 { 4950 vxlan_vnifilter_uninit(); 4951 rtnl_link_unregister(&vxlan_link_ops); 4952 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4953 unregister_netdevice_notifier(&vxlan_notifier_block); 4954 unregister_pernet_subsys(&vxlan_net_ops); 4955 /* rcu_barrier() is called by netns */ 4956 } 4957 module_exit(vxlan_cleanup_module); 4958 4959 MODULE_LICENSE("GPL"); 4960 MODULE_VERSION(VXLAN_VERSION); 4961 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4962 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4963 MODULE_ALIAS_RTNL_LINK("vxlan"); 4964