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 bool raw_proto = false; 1675 void *oiph; 1676 __be32 vni = 0; 1677 int nh; 1678 1679 /* Need UDP and VXLAN header to be present */ 1680 if (!pskb_may_pull(skb, VXLAN_HLEN)) 1681 goto drop; 1682 1683 unparsed = *vxlan_hdr(skb); 1684 /* VNI flag always required to be set */ 1685 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1686 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1687 ntohl(vxlan_hdr(skb)->vx_flags), 1688 ntohl(vxlan_hdr(skb)->vx_vni)); 1689 /* Return non vxlan pkt */ 1690 goto drop; 1691 } 1692 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1693 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1694 1695 vs = rcu_dereference_sk_user_data(sk); 1696 if (!vs) 1697 goto drop; 1698 1699 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1700 1701 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1702 if (!vxlan) 1703 goto drop; 1704 1705 /* For backwards compatibility, only allow reserved fields to be 1706 * used by VXLAN extensions if explicitly requested. 1707 */ 1708 if (vs->flags & VXLAN_F_GPE) { 1709 if (!vxlan_parse_gpe_proto(&unparsed, &protocol)) 1710 goto drop; 1711 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS; 1712 raw_proto = true; 1713 } 1714 1715 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1716 !net_eq(vxlan->net, dev_net(vxlan->dev)))) 1717 goto drop; 1718 1719 if (vs->flags & VXLAN_F_REMCSUM_RX) 1720 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1721 goto drop; 1722 1723 if (vxlan_collect_metadata(vs)) { 1724 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1725 struct metadata_dst *tun_dst; 1726 1727 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1728 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1729 key32_to_tunnel_id(vni), sizeof(*md)); 1730 1731 if (!tun_dst) 1732 goto drop; 1733 1734 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1735 1736 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1737 } else { 1738 memset(md, 0, sizeof(*md)); 1739 } 1740 1741 if (vs->flags & VXLAN_F_GBP) 1742 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1743 /* Note that GBP and GPE can never be active together. This is 1744 * ensured in vxlan_dev_configure. 1745 */ 1746 1747 if (unparsed.vx_flags || unparsed.vx_vni) { 1748 /* If there are any unprocessed flags remaining treat 1749 * this as a malformed packet. This behavior diverges from 1750 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1751 * in reserved fields are to be ignored. The approach here 1752 * maintains compatibility with previous stack code, and also 1753 * is more robust and provides a little more security in 1754 * adding extensions to VXLAN. 1755 */ 1756 goto drop; 1757 } 1758 1759 if (!raw_proto) { 1760 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1761 goto drop; 1762 } else { 1763 skb_reset_mac_header(skb); 1764 skb->dev = vxlan->dev; 1765 skb->pkt_type = PACKET_HOST; 1766 } 1767 1768 /* Save offset of outer header relative to skb->head, 1769 * because we are going to reset the network header to the inner header 1770 * and might change skb->head. 1771 */ 1772 nh = skb_network_header(skb) - skb->head; 1773 1774 skb_reset_network_header(skb); 1775 1776 if (!pskb_inet_may_pull(skb)) { 1777 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1778 DEV_STATS_INC(vxlan->dev, rx_errors); 1779 vxlan_vnifilter_count(vxlan, vni, vninode, 1780 VXLAN_VNI_STATS_RX_ERRORS, 0); 1781 goto drop; 1782 } 1783 1784 /* Get the outer header. */ 1785 oiph = skb->head + nh; 1786 1787 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1788 DEV_STATS_INC(vxlan->dev, rx_frame_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 rcu_read_lock(); 1796 1797 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1798 rcu_read_unlock(); 1799 dev_core_stats_rx_dropped_inc(vxlan->dev); 1800 vxlan_vnifilter_count(vxlan, vni, vninode, 1801 VXLAN_VNI_STATS_RX_DROPS, 0); 1802 goto drop; 1803 } 1804 1805 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1806 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1807 gro_cells_receive(&vxlan->gro_cells, skb); 1808 1809 rcu_read_unlock(); 1810 1811 return 0; 1812 1813 drop: 1814 /* Consume bad packet */ 1815 kfree_skb(skb); 1816 return 0; 1817 } 1818 1819 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1820 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1821 { 1822 struct vxlan_dev *vxlan; 1823 struct vxlan_sock *vs; 1824 struct vxlanhdr *hdr; 1825 __be32 vni; 1826 1827 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1828 return -EINVAL; 1829 1830 hdr = vxlan_hdr(skb); 1831 1832 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1833 return -EINVAL; 1834 1835 vs = rcu_dereference_sk_user_data(sk); 1836 if (!vs) 1837 return -ENOENT; 1838 1839 vni = vxlan_vni(hdr->vx_vni); 1840 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1841 if (!vxlan) 1842 return -ENOENT; 1843 1844 return 0; 1845 } 1846 1847 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1848 { 1849 struct vxlan_dev *vxlan = netdev_priv(dev); 1850 struct arphdr *parp; 1851 u8 *arpptr, *sha; 1852 __be32 sip, tip; 1853 struct neighbour *n; 1854 1855 if (dev->flags & IFF_NOARP) 1856 goto out; 1857 1858 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1859 dev_core_stats_tx_dropped_inc(dev); 1860 vxlan_vnifilter_count(vxlan, vni, NULL, 1861 VXLAN_VNI_STATS_TX_DROPS, 0); 1862 goto out; 1863 } 1864 parp = arp_hdr(skb); 1865 1866 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1867 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1868 parp->ar_pro != htons(ETH_P_IP) || 1869 parp->ar_op != htons(ARPOP_REQUEST) || 1870 parp->ar_hln != dev->addr_len || 1871 parp->ar_pln != 4) 1872 goto out; 1873 arpptr = (u8 *)parp + sizeof(struct arphdr); 1874 sha = arpptr; 1875 arpptr += dev->addr_len; /* sha */ 1876 memcpy(&sip, arpptr, sizeof(sip)); 1877 arpptr += sizeof(sip); 1878 arpptr += dev->addr_len; /* tha */ 1879 memcpy(&tip, arpptr, sizeof(tip)); 1880 1881 if (ipv4_is_loopback(tip) || 1882 ipv4_is_multicast(tip)) 1883 goto out; 1884 1885 n = neigh_lookup(&arp_tbl, &tip, dev); 1886 1887 if (n) { 1888 struct vxlan_fdb *f; 1889 struct sk_buff *reply; 1890 1891 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1892 neigh_release(n); 1893 goto out; 1894 } 1895 1896 f = vxlan_find_mac(vxlan, n->ha, vni); 1897 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1898 /* bridge-local neighbor */ 1899 neigh_release(n); 1900 goto out; 1901 } 1902 1903 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1904 n->ha, sha); 1905 1906 neigh_release(n); 1907 1908 if (reply == NULL) 1909 goto out; 1910 1911 skb_reset_mac_header(reply); 1912 __skb_pull(reply, skb_network_offset(reply)); 1913 reply->ip_summed = CHECKSUM_UNNECESSARY; 1914 reply->pkt_type = PACKET_HOST; 1915 1916 if (netif_rx(reply) == NET_RX_DROP) { 1917 dev_core_stats_rx_dropped_inc(dev); 1918 vxlan_vnifilter_count(vxlan, vni, NULL, 1919 VXLAN_VNI_STATS_RX_DROPS, 0); 1920 } 1921 1922 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1923 union vxlan_addr ipa = { 1924 .sin.sin_addr.s_addr = tip, 1925 .sin.sin_family = AF_INET, 1926 }; 1927 1928 vxlan_ip_miss(dev, &ipa); 1929 } 1930 out: 1931 consume_skb(skb); 1932 return NETDEV_TX_OK; 1933 } 1934 1935 #if IS_ENABLED(CONFIG_IPV6) 1936 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1937 struct neighbour *n, bool isrouter) 1938 { 1939 struct net_device *dev = request->dev; 1940 struct sk_buff *reply; 1941 struct nd_msg *ns, *na; 1942 struct ipv6hdr *pip6; 1943 u8 *daddr; 1944 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1945 int ns_olen; 1946 int i, len; 1947 1948 if (dev == NULL || !pskb_may_pull(request, request->len)) 1949 return NULL; 1950 1951 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1952 sizeof(*na) + na_olen + dev->needed_tailroom; 1953 reply = alloc_skb(len, GFP_ATOMIC); 1954 if (reply == NULL) 1955 return NULL; 1956 1957 reply->protocol = htons(ETH_P_IPV6); 1958 reply->dev = dev; 1959 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1960 skb_push(reply, sizeof(struct ethhdr)); 1961 skb_reset_mac_header(reply); 1962 1963 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1964 1965 daddr = eth_hdr(request)->h_source; 1966 ns_olen = request->len - skb_network_offset(request) - 1967 sizeof(struct ipv6hdr) - sizeof(*ns); 1968 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1969 if (!ns->opt[i + 1]) { 1970 kfree_skb(reply); 1971 return NULL; 1972 } 1973 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1974 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1975 break; 1976 } 1977 } 1978 1979 /* Ethernet header */ 1980 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1981 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1982 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1983 reply->protocol = htons(ETH_P_IPV6); 1984 1985 skb_pull(reply, sizeof(struct ethhdr)); 1986 skb_reset_network_header(reply); 1987 skb_put(reply, sizeof(struct ipv6hdr)); 1988 1989 /* IPv6 header */ 1990 1991 pip6 = ipv6_hdr(reply); 1992 memset(pip6, 0, sizeof(struct ipv6hdr)); 1993 pip6->version = 6; 1994 pip6->priority = ipv6_hdr(request)->priority; 1995 pip6->nexthdr = IPPROTO_ICMPV6; 1996 pip6->hop_limit = 255; 1997 pip6->daddr = ipv6_hdr(request)->saddr; 1998 pip6->saddr = *(struct in6_addr *)n->primary_key; 1999 2000 skb_pull(reply, sizeof(struct ipv6hdr)); 2001 skb_reset_transport_header(reply); 2002 2003 /* Neighbor Advertisement */ 2004 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2005 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2006 na->icmph.icmp6_router = isrouter; 2007 na->icmph.icmp6_override = 1; 2008 na->icmph.icmp6_solicited = 1; 2009 na->target = ns->target; 2010 ether_addr_copy(&na->opt[2], n->ha); 2011 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2012 na->opt[1] = na_olen >> 3; 2013 2014 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2015 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2016 csum_partial(na, sizeof(*na)+na_olen, 0)); 2017 2018 pip6->payload_len = htons(sizeof(*na)+na_olen); 2019 2020 skb_push(reply, sizeof(struct ipv6hdr)); 2021 2022 reply->ip_summed = CHECKSUM_UNNECESSARY; 2023 2024 return reply; 2025 } 2026 2027 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2028 { 2029 struct vxlan_dev *vxlan = netdev_priv(dev); 2030 const struct in6_addr *daddr; 2031 const struct ipv6hdr *iphdr; 2032 struct inet6_dev *in6_dev; 2033 struct neighbour *n; 2034 struct nd_msg *msg; 2035 2036 rcu_read_lock(); 2037 in6_dev = __in6_dev_get(dev); 2038 if (!in6_dev) 2039 goto out; 2040 2041 iphdr = ipv6_hdr(skb); 2042 daddr = &iphdr->daddr; 2043 msg = (struct nd_msg *)(iphdr + 1); 2044 2045 if (ipv6_addr_loopback(daddr) || 2046 ipv6_addr_is_multicast(&msg->target)) 2047 goto out; 2048 2049 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2050 2051 if (n) { 2052 struct vxlan_fdb *f; 2053 struct sk_buff *reply; 2054 2055 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2056 neigh_release(n); 2057 goto out; 2058 } 2059 2060 f = vxlan_find_mac(vxlan, n->ha, vni); 2061 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2062 /* bridge-local neighbor */ 2063 neigh_release(n); 2064 goto out; 2065 } 2066 2067 reply = vxlan_na_create(skb, n, 2068 !!(f ? f->flags & NTF_ROUTER : 0)); 2069 2070 neigh_release(n); 2071 2072 if (reply == NULL) 2073 goto out; 2074 2075 if (netif_rx(reply) == NET_RX_DROP) { 2076 dev_core_stats_rx_dropped_inc(dev); 2077 vxlan_vnifilter_count(vxlan, vni, NULL, 2078 VXLAN_VNI_STATS_RX_DROPS, 0); 2079 } 2080 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2081 union vxlan_addr ipa = { 2082 .sin6.sin6_addr = msg->target, 2083 .sin6.sin6_family = AF_INET6, 2084 }; 2085 2086 vxlan_ip_miss(dev, &ipa); 2087 } 2088 2089 out: 2090 rcu_read_unlock(); 2091 consume_skb(skb); 2092 return NETDEV_TX_OK; 2093 } 2094 #endif 2095 2096 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2097 { 2098 struct vxlan_dev *vxlan = netdev_priv(dev); 2099 struct neighbour *n; 2100 2101 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2102 return false; 2103 2104 n = NULL; 2105 switch (ntohs(eth_hdr(skb)->h_proto)) { 2106 case ETH_P_IP: 2107 { 2108 struct iphdr *pip; 2109 2110 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2111 return false; 2112 pip = ip_hdr(skb); 2113 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2114 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2115 union vxlan_addr ipa = { 2116 .sin.sin_addr.s_addr = pip->daddr, 2117 .sin.sin_family = AF_INET, 2118 }; 2119 2120 vxlan_ip_miss(dev, &ipa); 2121 return false; 2122 } 2123 2124 break; 2125 } 2126 #if IS_ENABLED(CONFIG_IPV6) 2127 case ETH_P_IPV6: 2128 { 2129 struct ipv6hdr *pip6; 2130 2131 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2132 return false; 2133 pip6 = ipv6_hdr(skb); 2134 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2135 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2136 union vxlan_addr ipa = { 2137 .sin6.sin6_addr = pip6->daddr, 2138 .sin6.sin6_family = AF_INET6, 2139 }; 2140 2141 vxlan_ip_miss(dev, &ipa); 2142 return false; 2143 } 2144 2145 break; 2146 } 2147 #endif 2148 default: 2149 return false; 2150 } 2151 2152 if (n) { 2153 bool diff; 2154 2155 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2156 if (diff) { 2157 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2158 dev->addr_len); 2159 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2160 } 2161 neigh_release(n); 2162 return diff; 2163 } 2164 2165 return false; 2166 } 2167 2168 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2169 { 2170 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2171 2172 gpe->np_applied = 1; 2173 gpe->next_protocol = tun_p_from_eth_p(protocol); 2174 if (!gpe->next_protocol) 2175 return -EPFNOSUPPORT; 2176 return 0; 2177 } 2178 2179 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2180 int iphdr_len, __be32 vni, 2181 struct vxlan_metadata *md, u32 vxflags, 2182 bool udp_sum) 2183 { 2184 struct vxlanhdr *vxh; 2185 int min_headroom; 2186 int err; 2187 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2188 __be16 inner_protocol = htons(ETH_P_TEB); 2189 2190 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2191 skb->ip_summed == CHECKSUM_PARTIAL) { 2192 int csum_start = skb_checksum_start_offset(skb); 2193 2194 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2195 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2196 (skb->csum_offset == offsetof(struct udphdr, check) || 2197 skb->csum_offset == offsetof(struct tcphdr, check))) 2198 type |= SKB_GSO_TUNNEL_REMCSUM; 2199 } 2200 2201 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2202 + VXLAN_HLEN + iphdr_len; 2203 2204 /* Need space for new headers (invalidates iph ptr) */ 2205 err = skb_cow_head(skb, min_headroom); 2206 if (unlikely(err)) 2207 return err; 2208 2209 err = iptunnel_handle_offloads(skb, type); 2210 if (err) 2211 return err; 2212 2213 vxh = __skb_push(skb, sizeof(*vxh)); 2214 vxh->vx_flags = VXLAN_HF_VNI; 2215 vxh->vx_vni = vxlan_vni_field(vni); 2216 2217 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2218 unsigned int start; 2219 2220 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2221 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2222 vxh->vx_flags |= VXLAN_HF_RCO; 2223 2224 if (!skb_is_gso(skb)) { 2225 skb->ip_summed = CHECKSUM_NONE; 2226 skb->encapsulation = 0; 2227 } 2228 } 2229 2230 if (vxflags & VXLAN_F_GBP) 2231 vxlan_build_gbp_hdr(vxh, md); 2232 if (vxflags & VXLAN_F_GPE) { 2233 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2234 if (err < 0) 2235 return err; 2236 inner_protocol = skb->protocol; 2237 } 2238 2239 skb_set_inner_protocol(skb, inner_protocol); 2240 return 0; 2241 } 2242 2243 /* Bypass encapsulation if the destination is local */ 2244 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2245 struct vxlan_dev *dst_vxlan, __be32 vni, 2246 bool snoop) 2247 { 2248 union vxlan_addr loopback; 2249 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2250 struct net_device *dev; 2251 int len = skb->len; 2252 2253 skb->pkt_type = PACKET_HOST; 2254 skb->encapsulation = 0; 2255 skb->dev = dst_vxlan->dev; 2256 __skb_pull(skb, skb_network_offset(skb)); 2257 2258 if (remote_ip->sa.sa_family == AF_INET) { 2259 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2260 loopback.sa.sa_family = AF_INET; 2261 #if IS_ENABLED(CONFIG_IPV6) 2262 } else { 2263 loopback.sin6.sin6_addr = in6addr_loopback; 2264 loopback.sa.sa_family = AF_INET6; 2265 #endif 2266 } 2267 2268 rcu_read_lock(); 2269 dev = skb->dev; 2270 if (unlikely(!(dev->flags & IFF_UP))) { 2271 kfree_skb(skb); 2272 goto drop; 2273 } 2274 2275 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2276 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2277 2278 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2279 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2280 2281 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2282 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2283 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2284 len); 2285 } else { 2286 drop: 2287 dev_core_stats_rx_dropped_inc(dev); 2288 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2289 VXLAN_VNI_STATS_RX_DROPS, 0); 2290 } 2291 rcu_read_unlock(); 2292 } 2293 2294 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2295 struct vxlan_dev *vxlan, 2296 int addr_family, 2297 __be16 dst_port, int dst_ifindex, __be32 vni, 2298 struct dst_entry *dst, 2299 u32 rt_flags) 2300 { 2301 #if IS_ENABLED(CONFIG_IPV6) 2302 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2303 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2304 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2305 */ 2306 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2307 #endif 2308 /* Bypass encapsulation if the destination is local */ 2309 if (rt_flags & RTCF_LOCAL && 2310 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2311 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2312 struct vxlan_dev *dst_vxlan; 2313 2314 dst_release(dst); 2315 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2316 addr_family, dst_port, 2317 vxlan->cfg.flags); 2318 if (!dst_vxlan) { 2319 DEV_STATS_INC(dev, tx_errors); 2320 vxlan_vnifilter_count(vxlan, vni, NULL, 2321 VXLAN_VNI_STATS_TX_ERRORS, 0); 2322 kfree_skb(skb); 2323 2324 return -ENOENT; 2325 } 2326 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2327 return 1; 2328 } 2329 2330 return 0; 2331 } 2332 2333 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2334 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2335 { 2336 struct dst_cache *dst_cache; 2337 struct ip_tunnel_info *info; 2338 struct ip_tunnel_key *pkey; 2339 struct ip_tunnel_key key; 2340 struct vxlan_dev *vxlan = netdev_priv(dev); 2341 const struct iphdr *old_iph; 2342 struct vxlan_metadata _md; 2343 struct vxlan_metadata *md = &_md; 2344 unsigned int pkt_len = skb->len; 2345 __be16 src_port = 0, dst_port; 2346 struct dst_entry *ndst = NULL; 2347 int addr_family; 2348 __u8 tos, ttl; 2349 int ifindex; 2350 int err; 2351 u32 flags = vxlan->cfg.flags; 2352 bool use_cache; 2353 bool udp_sum = false; 2354 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2355 bool no_eth_encap; 2356 __be32 vni = 0; 2357 2358 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2359 if (!skb_vlan_inet_prepare(skb, no_eth_encap)) 2360 goto drop; 2361 2362 old_iph = ip_hdr(skb); 2363 2364 info = skb_tunnel_info(skb); 2365 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2366 2367 if (rdst) { 2368 memset(&key, 0, sizeof(key)); 2369 pkey = &key; 2370 2371 if (vxlan_addr_any(&rdst->remote_ip)) { 2372 if (did_rsc) { 2373 /* short-circuited back to local bridge */ 2374 vxlan_encap_bypass(skb, vxlan, vxlan, 2375 default_vni, true); 2376 return; 2377 } 2378 goto drop; 2379 } 2380 2381 addr_family = vxlan->cfg.saddr.sa.sa_family; 2382 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2383 vni = (rdst->remote_vni) ? : default_vni; 2384 ifindex = rdst->remote_ifindex; 2385 2386 if (addr_family == AF_INET) { 2387 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2388 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2389 } else { 2390 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2391 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2392 } 2393 2394 dst_cache = &rdst->dst_cache; 2395 md->gbp = skb->mark; 2396 if (flags & VXLAN_F_TTL_INHERIT) { 2397 ttl = ip_tunnel_get_ttl(old_iph, skb); 2398 } else { 2399 ttl = vxlan->cfg.ttl; 2400 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2401 ttl = 1; 2402 } 2403 tos = vxlan->cfg.tos; 2404 if (tos == 1) 2405 tos = ip_tunnel_get_dsfield(old_iph, skb); 2406 if (tos && !info) 2407 use_cache = false; 2408 2409 if (addr_family == AF_INET) 2410 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2411 else 2412 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2413 #if IS_ENABLED(CONFIG_IPV6) 2414 switch (vxlan->cfg.label_policy) { 2415 case VXLAN_LABEL_FIXED: 2416 key.label = vxlan->cfg.label; 2417 break; 2418 case VXLAN_LABEL_INHERIT: 2419 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2420 break; 2421 default: 2422 DEBUG_NET_WARN_ON_ONCE(1); 2423 goto drop; 2424 } 2425 #endif 2426 } else { 2427 if (!info) { 2428 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2429 dev->name); 2430 goto drop; 2431 } 2432 pkey = &info->key; 2433 addr_family = ip_tunnel_info_af(info); 2434 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2435 vni = tunnel_id_to_key32(info->key.tun_id); 2436 ifindex = 0; 2437 dst_cache = &info->dst_cache; 2438 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2439 if (info->options_len < sizeof(*md)) 2440 goto drop; 2441 md = ip_tunnel_info_opts(info); 2442 } 2443 ttl = info->key.ttl; 2444 tos = info->key.tos; 2445 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2446 } 2447 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2448 vxlan->cfg.port_max, true); 2449 2450 rcu_read_lock(); 2451 if (addr_family == AF_INET) { 2452 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2453 struct rtable *rt; 2454 __be16 df = 0; 2455 __be32 saddr; 2456 2457 if (!ifindex) 2458 ifindex = sock4->sock->sk->sk_bound_dev_if; 2459 2460 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2461 &saddr, pkey, src_port, dst_port, 2462 tos, use_cache ? dst_cache : NULL); 2463 if (IS_ERR(rt)) { 2464 err = PTR_ERR(rt); 2465 goto tx_error; 2466 } 2467 2468 if (!info) { 2469 /* Bypass encapsulation if the destination is local */ 2470 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2471 dst_port, ifindex, vni, 2472 &rt->dst, rt->rt_flags); 2473 if (err) 2474 goto out_unlock; 2475 2476 if (vxlan->cfg.df == VXLAN_DF_SET) { 2477 df = htons(IP_DF); 2478 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2479 struct ethhdr *eth = eth_hdr(skb); 2480 2481 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2482 (ntohs(eth->h_proto) == ETH_P_IP && 2483 old_iph->frag_off & htons(IP_DF))) 2484 df = htons(IP_DF); 2485 } 2486 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2487 info->key.tun_flags)) { 2488 df = htons(IP_DF); 2489 } 2490 2491 ndst = &rt->dst; 2492 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2493 netif_is_any_bridge_port(dev)); 2494 if (err < 0) { 2495 goto tx_error; 2496 } else if (err) { 2497 if (info) { 2498 struct ip_tunnel_info *unclone; 2499 2500 unclone = skb_tunnel_info_unclone(skb); 2501 if (unlikely(!unclone)) 2502 goto tx_error; 2503 2504 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2505 unclone->key.u.ipv4.dst = saddr; 2506 } 2507 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2508 dst_release(ndst); 2509 goto out_unlock; 2510 } 2511 2512 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2513 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2514 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2515 vni, md, flags, udp_sum); 2516 if (err < 0) 2517 goto tx_error; 2518 2519 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2520 pkey->u.ipv4.dst, tos, ttl, df, 2521 src_port, dst_port, xnet, !udp_sum); 2522 #if IS_ENABLED(CONFIG_IPV6) 2523 } else { 2524 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2525 struct in6_addr saddr; 2526 2527 if (!ifindex) 2528 ifindex = sock6->sock->sk->sk_bound_dev_if; 2529 2530 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2531 ifindex, &saddr, pkey, 2532 src_port, dst_port, tos, 2533 use_cache ? dst_cache : NULL); 2534 if (IS_ERR(ndst)) { 2535 err = PTR_ERR(ndst); 2536 ndst = NULL; 2537 goto tx_error; 2538 } 2539 2540 if (!info) { 2541 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2542 2543 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2544 dst_port, ifindex, vni, 2545 ndst, rt6i_flags); 2546 if (err) 2547 goto out_unlock; 2548 } 2549 2550 err = skb_tunnel_check_pmtu(skb, ndst, 2551 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2552 netif_is_any_bridge_port(dev)); 2553 if (err < 0) { 2554 goto tx_error; 2555 } else if (err) { 2556 if (info) { 2557 struct ip_tunnel_info *unclone; 2558 2559 unclone = skb_tunnel_info_unclone(skb); 2560 if (unlikely(!unclone)) 2561 goto tx_error; 2562 2563 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2564 unclone->key.u.ipv6.dst = saddr; 2565 } 2566 2567 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2568 dst_release(ndst); 2569 goto out_unlock; 2570 } 2571 2572 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2573 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2574 skb_scrub_packet(skb, xnet); 2575 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2576 vni, md, flags, udp_sum); 2577 if (err < 0) 2578 goto tx_error; 2579 2580 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2581 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2582 pkey->label, src_port, dst_port, !udp_sum); 2583 #endif 2584 } 2585 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2586 out_unlock: 2587 rcu_read_unlock(); 2588 return; 2589 2590 drop: 2591 dev_core_stats_tx_dropped_inc(dev); 2592 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2593 dev_kfree_skb(skb); 2594 return; 2595 2596 tx_error: 2597 rcu_read_unlock(); 2598 if (err == -ELOOP) 2599 DEV_STATS_INC(dev, collisions); 2600 else if (err == -ENETUNREACH) 2601 DEV_STATS_INC(dev, tx_carrier_errors); 2602 dst_release(ndst); 2603 DEV_STATS_INC(dev, tx_errors); 2604 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2605 kfree_skb(skb); 2606 } 2607 2608 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2609 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2610 { 2611 struct vxlan_rdst nh_rdst; 2612 struct nexthop *nh; 2613 bool do_xmit; 2614 u32 hash; 2615 2616 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2617 hash = skb_get_hash(skb); 2618 2619 rcu_read_lock(); 2620 nh = rcu_dereference(f->nh); 2621 if (!nh) { 2622 rcu_read_unlock(); 2623 goto drop; 2624 } 2625 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2626 rcu_read_unlock(); 2627 2628 if (likely(do_xmit)) 2629 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2630 else 2631 goto drop; 2632 2633 return; 2634 2635 drop: 2636 dev_core_stats_tx_dropped_inc(dev); 2637 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2638 VXLAN_VNI_STATS_TX_DROPS, 0); 2639 dev_kfree_skb(skb); 2640 } 2641 2642 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2643 u32 nhid, __be32 vni) 2644 { 2645 struct vxlan_dev *vxlan = netdev_priv(dev); 2646 struct vxlan_rdst nh_rdst; 2647 struct nexthop *nh; 2648 bool do_xmit; 2649 u32 hash; 2650 2651 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2652 hash = skb_get_hash(skb); 2653 2654 rcu_read_lock(); 2655 nh = nexthop_find_by_id(dev_net(dev), nhid); 2656 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2657 rcu_read_unlock(); 2658 goto drop; 2659 } 2660 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2661 rcu_read_unlock(); 2662 2663 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2664 goto drop; 2665 2666 if (likely(do_xmit)) 2667 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2668 else 2669 goto drop; 2670 2671 return NETDEV_TX_OK; 2672 2673 drop: 2674 dev_core_stats_tx_dropped_inc(dev); 2675 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2676 VXLAN_VNI_STATS_TX_DROPS, 0); 2677 dev_kfree_skb(skb); 2678 return NETDEV_TX_OK; 2679 } 2680 2681 /* Transmit local packets over Vxlan 2682 * 2683 * Outer IP header inherits ECN and DF from inner header. 2684 * Outer UDP destination is the VXLAN assigned port. 2685 * source port is based on hash of flow 2686 */ 2687 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2688 { 2689 struct vxlan_dev *vxlan = netdev_priv(dev); 2690 struct vxlan_rdst *rdst, *fdst = NULL; 2691 const struct ip_tunnel_info *info; 2692 struct vxlan_fdb *f; 2693 struct ethhdr *eth; 2694 __be32 vni = 0; 2695 u32 nhid = 0; 2696 bool did_rsc; 2697 2698 info = skb_tunnel_info(skb); 2699 2700 skb_reset_mac_header(skb); 2701 2702 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2703 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2704 info->mode & IP_TUNNEL_INFO_TX) { 2705 vni = tunnel_id_to_key32(info->key.tun_id); 2706 nhid = info->key.nhid; 2707 } else { 2708 if (info && info->mode & IP_TUNNEL_INFO_TX) 2709 vxlan_xmit_one(skb, dev, vni, NULL, false); 2710 else 2711 kfree_skb(skb); 2712 return NETDEV_TX_OK; 2713 } 2714 } 2715 2716 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2717 eth = eth_hdr(skb); 2718 if (ntohs(eth->h_proto) == ETH_P_ARP) 2719 return arp_reduce(dev, skb, vni); 2720 #if IS_ENABLED(CONFIG_IPV6) 2721 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2722 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2723 sizeof(struct nd_msg)) && 2724 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2725 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2726 2727 if (m->icmph.icmp6_code == 0 && 2728 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2729 return neigh_reduce(dev, skb, vni); 2730 } 2731 #endif 2732 } 2733 2734 if (nhid) 2735 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2736 2737 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2738 struct vxlan_mdb_entry *mdb_entry; 2739 2740 rcu_read_lock(); 2741 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2742 if (mdb_entry) { 2743 netdev_tx_t ret; 2744 2745 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2746 rcu_read_unlock(); 2747 return ret; 2748 } 2749 rcu_read_unlock(); 2750 } 2751 2752 eth = eth_hdr(skb); 2753 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2754 did_rsc = false; 2755 2756 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2757 (ntohs(eth->h_proto) == ETH_P_IP || 2758 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2759 did_rsc = route_shortcircuit(dev, skb); 2760 if (did_rsc) 2761 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2762 } 2763 2764 if (f == NULL) { 2765 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2766 if (f == NULL) { 2767 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2768 !is_multicast_ether_addr(eth->h_dest)) 2769 vxlan_fdb_miss(vxlan, eth->h_dest); 2770 2771 dev_core_stats_tx_dropped_inc(dev); 2772 vxlan_vnifilter_count(vxlan, vni, NULL, 2773 VXLAN_VNI_STATS_TX_DROPS, 0); 2774 kfree_skb(skb); 2775 return NETDEV_TX_OK; 2776 } 2777 } 2778 2779 if (rcu_access_pointer(f->nh)) { 2780 vxlan_xmit_nh(skb, dev, f, 2781 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2782 } else { 2783 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2784 struct sk_buff *skb1; 2785 2786 if (!fdst) { 2787 fdst = rdst; 2788 continue; 2789 } 2790 skb1 = skb_clone(skb, GFP_ATOMIC); 2791 if (skb1) 2792 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2793 } 2794 if (fdst) 2795 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2796 else 2797 kfree_skb(skb); 2798 } 2799 2800 return NETDEV_TX_OK; 2801 } 2802 2803 /* Walk the forwarding table and purge stale entries */ 2804 static void vxlan_cleanup(struct timer_list *t) 2805 { 2806 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2807 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2808 unsigned int h; 2809 2810 if (!netif_running(vxlan->dev)) 2811 return; 2812 2813 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2814 struct hlist_node *p, *n; 2815 2816 spin_lock(&vxlan->hash_lock[h]); 2817 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2818 struct vxlan_fdb *f 2819 = container_of(p, struct vxlan_fdb, hlist); 2820 unsigned long timeout; 2821 2822 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2823 continue; 2824 2825 if (f->flags & NTF_EXT_LEARNED) 2826 continue; 2827 2828 timeout = f->used + vxlan->cfg.age_interval * HZ; 2829 if (time_before_eq(timeout, jiffies)) { 2830 netdev_dbg(vxlan->dev, 2831 "garbage collect %pM\n", 2832 f->eth_addr); 2833 f->state = NUD_STALE; 2834 vxlan_fdb_destroy(vxlan, f, true, true); 2835 } else if (time_before(timeout, next_timer)) 2836 next_timer = timeout; 2837 } 2838 spin_unlock(&vxlan->hash_lock[h]); 2839 } 2840 2841 mod_timer(&vxlan->age_timer, next_timer); 2842 } 2843 2844 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2845 { 2846 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2847 2848 spin_lock(&vn->sock_lock); 2849 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2850 #if IS_ENABLED(CONFIG_IPV6) 2851 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2852 #endif 2853 spin_unlock(&vn->sock_lock); 2854 } 2855 2856 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2857 struct vxlan_dev_node *node) 2858 { 2859 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2860 __be32 vni = vxlan->default_dst.remote_vni; 2861 2862 node->vxlan = vxlan; 2863 spin_lock(&vn->sock_lock); 2864 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2865 spin_unlock(&vn->sock_lock); 2866 } 2867 2868 /* Setup stats when device is created */ 2869 static int vxlan_init(struct net_device *dev) 2870 { 2871 struct vxlan_dev *vxlan = netdev_priv(dev); 2872 int err; 2873 2874 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2875 vxlan_vnigroup_init(vxlan); 2876 2877 err = gro_cells_init(&vxlan->gro_cells, dev); 2878 if (err) 2879 goto err_vnigroup_uninit; 2880 2881 err = vxlan_mdb_init(vxlan); 2882 if (err) 2883 goto err_gro_cells_destroy; 2884 2885 netdev_lockdep_set_classes(dev); 2886 return 0; 2887 2888 err_gro_cells_destroy: 2889 gro_cells_destroy(&vxlan->gro_cells); 2890 err_vnigroup_uninit: 2891 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2892 vxlan_vnigroup_uninit(vxlan); 2893 return err; 2894 } 2895 2896 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2897 { 2898 struct vxlan_fdb *f; 2899 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2900 2901 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2902 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2903 if (f) 2904 vxlan_fdb_destroy(vxlan, f, true, true); 2905 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2906 } 2907 2908 static void vxlan_uninit(struct net_device *dev) 2909 { 2910 struct vxlan_dev *vxlan = netdev_priv(dev); 2911 2912 vxlan_mdb_fini(vxlan); 2913 2914 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2915 vxlan_vnigroup_uninit(vxlan); 2916 2917 gro_cells_destroy(&vxlan->gro_cells); 2918 2919 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2920 } 2921 2922 /* Start ageing timer and join group when device is brought up */ 2923 static int vxlan_open(struct net_device *dev) 2924 { 2925 struct vxlan_dev *vxlan = netdev_priv(dev); 2926 int ret; 2927 2928 ret = vxlan_sock_add(vxlan); 2929 if (ret < 0) 2930 return ret; 2931 2932 ret = vxlan_multicast_join(vxlan); 2933 if (ret) { 2934 vxlan_sock_release(vxlan); 2935 return ret; 2936 } 2937 2938 if (vxlan->cfg.age_interval) 2939 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2940 2941 return ret; 2942 } 2943 2944 struct vxlan_fdb_flush_desc { 2945 bool ignore_default_entry; 2946 unsigned long state; 2947 unsigned long state_mask; 2948 unsigned long flags; 2949 unsigned long flags_mask; 2950 __be32 src_vni; 2951 u32 nhid; 2952 __be32 vni; 2953 __be16 port; 2954 union vxlan_addr dst_ip; 2955 }; 2956 2957 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2958 const struct vxlan_dev *vxlan) 2959 { 2960 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2961 } 2962 2963 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2964 { 2965 struct nexthop *nh = rtnl_dereference(f->nh); 2966 2967 return nh && nh->id == nhid; 2968 } 2969 2970 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2971 const struct vxlan_dev *vxlan, 2972 const struct vxlan_fdb_flush_desc *desc) 2973 { 2974 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2975 return false; 2976 2977 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2978 return false; 2979 2980 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2981 return false; 2982 2983 if (desc->src_vni && f->vni != desc->src_vni) 2984 return false; 2985 2986 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 2987 return false; 2988 2989 return true; 2990 } 2991 2992 static bool 2993 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 2994 { 2995 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 2996 } 2997 2998 static bool 2999 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3000 const struct vxlan_rdst *rd) 3001 { 3002 if (desc->vni && rd->remote_vni != desc->vni) 3003 return false; 3004 3005 if (desc->port && rd->remote_port != desc->port) 3006 return false; 3007 3008 if (desc->dst_ip.sa.sa_family && 3009 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3010 return false; 3011 3012 return true; 3013 } 3014 3015 static void 3016 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3017 const struct vxlan_fdb_flush_desc *desc, 3018 bool *p_destroy_fdb) 3019 { 3020 bool remotes_flushed = false; 3021 struct vxlan_rdst *rd, *tmp; 3022 3023 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3024 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3025 continue; 3026 3027 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3028 remotes_flushed = true; 3029 } 3030 3031 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3032 } 3033 3034 /* Purge the forwarding table */ 3035 static void vxlan_flush(struct vxlan_dev *vxlan, 3036 const struct vxlan_fdb_flush_desc *desc) 3037 { 3038 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3039 unsigned int h; 3040 3041 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3042 struct hlist_node *p, *n; 3043 3044 spin_lock_bh(&vxlan->hash_lock[h]); 3045 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3046 struct vxlan_fdb *f 3047 = container_of(p, struct vxlan_fdb, hlist); 3048 3049 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3050 continue; 3051 3052 if (match_remotes) { 3053 bool destroy_fdb = false; 3054 3055 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3056 &destroy_fdb); 3057 3058 if (!destroy_fdb) 3059 continue; 3060 } 3061 3062 vxlan_fdb_destroy(vxlan, f, true, true); 3063 } 3064 spin_unlock_bh(&vxlan->hash_lock[h]); 3065 } 3066 } 3067 3068 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3069 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3070 [NDA_NH_ID] = { .type = NLA_U32 }, 3071 [NDA_VNI] = { .type = NLA_U32 }, 3072 [NDA_PORT] = { .type = NLA_U16 }, 3073 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3074 sizeof(struct in6_addr)), 3075 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3076 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3077 }; 3078 3079 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3080 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3081 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3082 NTF_ROUTER) 3083 3084 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3085 struct netlink_ext_ack *extack) 3086 { 3087 struct vxlan_dev *vxlan = netdev_priv(dev); 3088 struct vxlan_fdb_flush_desc desc = {}; 3089 struct ndmsg *ndm = nlmsg_data(nlh); 3090 struct nlattr *tb[NDA_MAX + 1]; 3091 u8 ndm_flags; 3092 int err; 3093 3094 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3095 3096 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3097 extack); 3098 if (err) 3099 return err; 3100 3101 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3102 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3103 return -EINVAL; 3104 } 3105 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3106 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3107 return -EINVAL; 3108 } 3109 3110 desc.state = ndm->ndm_state; 3111 desc.flags = ndm_flags; 3112 3113 if (tb[NDA_NDM_STATE_MASK]) 3114 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3115 3116 if (tb[NDA_NDM_FLAGS_MASK]) 3117 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3118 3119 if (tb[NDA_SRC_VNI]) 3120 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3121 3122 if (tb[NDA_NH_ID]) 3123 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3124 3125 if (tb[NDA_VNI]) 3126 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3127 3128 if (tb[NDA_PORT]) 3129 desc.port = nla_get_be16(tb[NDA_PORT]); 3130 3131 if (tb[NDA_DST]) { 3132 union vxlan_addr ip; 3133 3134 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3135 if (err) { 3136 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3137 "Unsupported address family"); 3138 return err; 3139 } 3140 desc.dst_ip = ip; 3141 } 3142 3143 vxlan_flush(vxlan, &desc); 3144 3145 return 0; 3146 } 3147 3148 /* Cleanup timer and forwarding table on shutdown */ 3149 static int vxlan_stop(struct net_device *dev) 3150 { 3151 struct vxlan_dev *vxlan = netdev_priv(dev); 3152 struct vxlan_fdb_flush_desc desc = { 3153 /* Default entry is deleted at vxlan_uninit. */ 3154 .ignore_default_entry = true, 3155 .state = 0, 3156 .state_mask = NUD_PERMANENT | NUD_NOARP, 3157 }; 3158 3159 vxlan_multicast_leave(vxlan); 3160 3161 del_timer_sync(&vxlan->age_timer); 3162 3163 vxlan_flush(vxlan, &desc); 3164 vxlan_sock_release(vxlan); 3165 3166 return 0; 3167 } 3168 3169 /* Stub, nothing needs to be done. */ 3170 static void vxlan_set_multicast_list(struct net_device *dev) 3171 { 3172 } 3173 3174 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3175 { 3176 struct vxlan_dev *vxlan = netdev_priv(dev); 3177 struct vxlan_rdst *dst = &vxlan->default_dst; 3178 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3179 dst->remote_ifindex); 3180 3181 /* This check is different than dev->max_mtu, because it looks at 3182 * the lowerdev->mtu, rather than the static dev->max_mtu 3183 */ 3184 if (lowerdev) { 3185 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3186 if (new_mtu > max_mtu) 3187 return -EINVAL; 3188 } 3189 3190 WRITE_ONCE(dev->mtu, new_mtu); 3191 return 0; 3192 } 3193 3194 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3195 { 3196 struct vxlan_dev *vxlan = netdev_priv(dev); 3197 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3198 __be16 sport, dport; 3199 3200 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3201 vxlan->cfg.port_max, true); 3202 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3203 3204 if (ip_tunnel_info_af(info) == AF_INET) { 3205 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3206 struct rtable *rt; 3207 3208 if (!sock4) 3209 return -EIO; 3210 3211 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3212 &info->key.u.ipv4.src, 3213 &info->key, 3214 sport, dport, info->key.tos, 3215 &info->dst_cache); 3216 if (IS_ERR(rt)) 3217 return PTR_ERR(rt); 3218 ip_rt_put(rt); 3219 } else { 3220 #if IS_ENABLED(CONFIG_IPV6) 3221 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3222 struct dst_entry *ndst; 3223 3224 if (!sock6) 3225 return -EIO; 3226 3227 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3228 0, &info->key.u.ipv6.src, 3229 &info->key, 3230 sport, dport, info->key.tos, 3231 &info->dst_cache); 3232 if (IS_ERR(ndst)) 3233 return PTR_ERR(ndst); 3234 dst_release(ndst); 3235 #else /* !CONFIG_IPV6 */ 3236 return -EPFNOSUPPORT; 3237 #endif 3238 } 3239 info->key.tp_src = sport; 3240 info->key.tp_dst = dport; 3241 return 0; 3242 } 3243 3244 static const struct net_device_ops vxlan_netdev_ether_ops = { 3245 .ndo_init = vxlan_init, 3246 .ndo_uninit = vxlan_uninit, 3247 .ndo_open = vxlan_open, 3248 .ndo_stop = vxlan_stop, 3249 .ndo_start_xmit = vxlan_xmit, 3250 .ndo_set_rx_mode = vxlan_set_multicast_list, 3251 .ndo_change_mtu = vxlan_change_mtu, 3252 .ndo_validate_addr = eth_validate_addr, 3253 .ndo_set_mac_address = eth_mac_addr, 3254 .ndo_fdb_add = vxlan_fdb_add, 3255 .ndo_fdb_del = vxlan_fdb_delete, 3256 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3257 .ndo_fdb_dump = vxlan_fdb_dump, 3258 .ndo_fdb_get = vxlan_fdb_get, 3259 .ndo_mdb_add = vxlan_mdb_add, 3260 .ndo_mdb_del = vxlan_mdb_del, 3261 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3262 .ndo_mdb_dump = vxlan_mdb_dump, 3263 .ndo_mdb_get = vxlan_mdb_get, 3264 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3265 }; 3266 3267 static const struct net_device_ops vxlan_netdev_raw_ops = { 3268 .ndo_init = vxlan_init, 3269 .ndo_uninit = vxlan_uninit, 3270 .ndo_open = vxlan_open, 3271 .ndo_stop = vxlan_stop, 3272 .ndo_start_xmit = vxlan_xmit, 3273 .ndo_change_mtu = vxlan_change_mtu, 3274 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3275 }; 3276 3277 /* Info for udev, that this is a virtual tunnel endpoint */ 3278 static const struct device_type vxlan_type = { 3279 .name = "vxlan", 3280 }; 3281 3282 /* Calls the ndo_udp_tunnel_add of the caller in order to 3283 * supply the listening VXLAN udp ports. Callers are expected 3284 * to implement the ndo_udp_tunnel_add. 3285 */ 3286 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3287 { 3288 struct vxlan_sock *vs; 3289 struct net *net = dev_net(dev); 3290 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3291 unsigned int i; 3292 3293 spin_lock(&vn->sock_lock); 3294 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3295 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3296 unsigned short type; 3297 3298 if (vs->flags & VXLAN_F_GPE) 3299 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3300 else 3301 type = UDP_TUNNEL_TYPE_VXLAN; 3302 3303 if (push) 3304 udp_tunnel_push_rx_port(dev, vs->sock, type); 3305 else 3306 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3307 } 3308 } 3309 spin_unlock(&vn->sock_lock); 3310 } 3311 3312 /* Initialize the device structure. */ 3313 static void vxlan_setup(struct net_device *dev) 3314 { 3315 struct vxlan_dev *vxlan = netdev_priv(dev); 3316 unsigned int h; 3317 3318 eth_hw_addr_random(dev); 3319 ether_setup(dev); 3320 3321 dev->needs_free_netdev = true; 3322 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3323 3324 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3325 dev->features |= NETIF_F_RXCSUM; 3326 dev->features |= NETIF_F_GSO_SOFTWARE; 3327 3328 dev->vlan_features = dev->features; 3329 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3330 dev->hw_features |= NETIF_F_RXCSUM; 3331 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3332 netif_keep_dst(dev); 3333 dev->priv_flags |= IFF_NO_QUEUE; 3334 dev->change_proto_down = true; 3335 dev->lltx = true; 3336 3337 /* MTU range: 68 - 65535 */ 3338 dev->min_mtu = ETH_MIN_MTU; 3339 dev->max_mtu = ETH_MAX_MTU; 3340 3341 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3342 INIT_LIST_HEAD(&vxlan->next); 3343 3344 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3345 3346 vxlan->dev = dev; 3347 3348 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3349 spin_lock_init(&vxlan->hash_lock[h]); 3350 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3351 } 3352 } 3353 3354 static void vxlan_ether_setup(struct net_device *dev) 3355 { 3356 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3357 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3358 dev->netdev_ops = &vxlan_netdev_ether_ops; 3359 } 3360 3361 static void vxlan_raw_setup(struct net_device *dev) 3362 { 3363 dev->header_ops = NULL; 3364 dev->type = ARPHRD_NONE; 3365 dev->hard_header_len = 0; 3366 dev->addr_len = 0; 3367 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3368 dev->netdev_ops = &vxlan_netdev_raw_ops; 3369 } 3370 3371 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3372 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3373 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3374 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3375 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3376 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3377 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3378 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3379 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3380 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3381 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3382 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3383 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3384 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3385 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3386 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3387 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3388 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3389 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3390 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3391 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3392 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3393 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3394 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3395 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3396 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3397 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3398 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3399 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3400 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3401 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3402 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3403 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3404 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3405 }; 3406 3407 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3408 struct netlink_ext_ack *extack) 3409 { 3410 if (tb[IFLA_ADDRESS]) { 3411 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3412 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3413 "Provided link layer address is not Ethernet"); 3414 return -EINVAL; 3415 } 3416 3417 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3418 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3419 "Provided Ethernet address is not unicast"); 3420 return -EADDRNOTAVAIL; 3421 } 3422 } 3423 3424 if (tb[IFLA_MTU]) { 3425 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3426 3427 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3428 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3429 "MTU must be between 68 and 65535"); 3430 return -EINVAL; 3431 } 3432 } 3433 3434 if (!data) { 3435 NL_SET_ERR_MSG(extack, 3436 "Required attributes not provided to perform the operation"); 3437 return -EINVAL; 3438 } 3439 3440 if (data[IFLA_VXLAN_ID]) { 3441 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3442 3443 if (id >= VXLAN_N_VID) { 3444 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3445 "VXLAN ID must be lower than 16777216"); 3446 return -ERANGE; 3447 } 3448 } 3449 3450 if (data[IFLA_VXLAN_PORT_RANGE]) { 3451 const struct ifla_vxlan_port_range *p 3452 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3453 3454 if (ntohs(p->high) < ntohs(p->low)) { 3455 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3456 "Invalid source port range"); 3457 return -EINVAL; 3458 } 3459 } 3460 3461 if (data[IFLA_VXLAN_DF]) { 3462 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3463 3464 if (df < 0 || df > VXLAN_DF_MAX) { 3465 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3466 "Invalid DF attribute"); 3467 return -EINVAL; 3468 } 3469 } 3470 3471 return 0; 3472 } 3473 3474 static void vxlan_get_drvinfo(struct net_device *netdev, 3475 struct ethtool_drvinfo *drvinfo) 3476 { 3477 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3478 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3479 } 3480 3481 static int vxlan_get_link_ksettings(struct net_device *dev, 3482 struct ethtool_link_ksettings *cmd) 3483 { 3484 struct vxlan_dev *vxlan = netdev_priv(dev); 3485 struct vxlan_rdst *dst = &vxlan->default_dst; 3486 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3487 dst->remote_ifindex); 3488 3489 if (!lowerdev) { 3490 cmd->base.duplex = DUPLEX_UNKNOWN; 3491 cmd->base.port = PORT_OTHER; 3492 cmd->base.speed = SPEED_UNKNOWN; 3493 3494 return 0; 3495 } 3496 3497 return __ethtool_get_link_ksettings(lowerdev, cmd); 3498 } 3499 3500 static const struct ethtool_ops vxlan_ethtool_ops = { 3501 .get_drvinfo = vxlan_get_drvinfo, 3502 .get_link = ethtool_op_get_link, 3503 .get_link_ksettings = vxlan_get_link_ksettings, 3504 }; 3505 3506 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3507 __be16 port, u32 flags, int ifindex) 3508 { 3509 struct socket *sock; 3510 struct udp_port_cfg udp_conf; 3511 int err; 3512 3513 memset(&udp_conf, 0, sizeof(udp_conf)); 3514 3515 if (ipv6) { 3516 udp_conf.family = AF_INET6; 3517 udp_conf.use_udp6_rx_checksums = 3518 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3519 udp_conf.ipv6_v6only = 1; 3520 } else { 3521 udp_conf.family = AF_INET; 3522 } 3523 3524 udp_conf.local_udp_port = port; 3525 udp_conf.bind_ifindex = ifindex; 3526 3527 /* Open UDP socket */ 3528 err = udp_sock_create(net, &udp_conf, &sock); 3529 if (err < 0) 3530 return ERR_PTR(err); 3531 3532 udp_allow_gso(sock->sk); 3533 return sock; 3534 } 3535 3536 /* Create new listen socket if needed */ 3537 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3538 __be16 port, u32 flags, 3539 int ifindex) 3540 { 3541 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3542 struct vxlan_sock *vs; 3543 struct socket *sock; 3544 unsigned int h; 3545 struct udp_tunnel_sock_cfg tunnel_cfg; 3546 3547 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3548 if (!vs) 3549 return ERR_PTR(-ENOMEM); 3550 3551 for (h = 0; h < VNI_HASH_SIZE; ++h) 3552 INIT_HLIST_HEAD(&vs->vni_list[h]); 3553 3554 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3555 if (IS_ERR(sock)) { 3556 kfree(vs); 3557 return ERR_CAST(sock); 3558 } 3559 3560 vs->sock = sock; 3561 refcount_set(&vs->refcnt, 1); 3562 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3563 3564 spin_lock(&vn->sock_lock); 3565 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3566 udp_tunnel_notify_add_rx_port(sock, 3567 (vs->flags & VXLAN_F_GPE) ? 3568 UDP_TUNNEL_TYPE_VXLAN_GPE : 3569 UDP_TUNNEL_TYPE_VXLAN); 3570 spin_unlock(&vn->sock_lock); 3571 3572 /* Mark socket as an encapsulation socket. */ 3573 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3574 tunnel_cfg.sk_user_data = vs; 3575 tunnel_cfg.encap_type = 1; 3576 tunnel_cfg.encap_rcv = vxlan_rcv; 3577 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3578 tunnel_cfg.encap_destroy = NULL; 3579 if (vs->flags & VXLAN_F_GPE) { 3580 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3581 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3582 } else { 3583 tunnel_cfg.gro_receive = vxlan_gro_receive; 3584 tunnel_cfg.gro_complete = vxlan_gro_complete; 3585 } 3586 3587 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3588 3589 return vs; 3590 } 3591 3592 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3593 { 3594 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3595 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3596 struct vxlan_sock *vs = NULL; 3597 struct vxlan_dev_node *node; 3598 int l3mdev_index = 0; 3599 3600 if (vxlan->cfg.remote_ifindex) 3601 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3602 vxlan->net, vxlan->cfg.remote_ifindex); 3603 3604 if (!vxlan->cfg.no_share) { 3605 spin_lock(&vn->sock_lock); 3606 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3607 vxlan->cfg.dst_port, vxlan->cfg.flags, 3608 l3mdev_index); 3609 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3610 spin_unlock(&vn->sock_lock); 3611 return -EBUSY; 3612 } 3613 spin_unlock(&vn->sock_lock); 3614 } 3615 if (!vs) 3616 vs = vxlan_socket_create(vxlan->net, ipv6, 3617 vxlan->cfg.dst_port, vxlan->cfg.flags, 3618 l3mdev_index); 3619 if (IS_ERR(vs)) 3620 return PTR_ERR(vs); 3621 #if IS_ENABLED(CONFIG_IPV6) 3622 if (ipv6) { 3623 rcu_assign_pointer(vxlan->vn6_sock, vs); 3624 node = &vxlan->hlist6; 3625 } else 3626 #endif 3627 { 3628 rcu_assign_pointer(vxlan->vn4_sock, vs); 3629 node = &vxlan->hlist4; 3630 } 3631 3632 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3633 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3634 else 3635 vxlan_vs_add_dev(vs, vxlan, node); 3636 3637 return 0; 3638 } 3639 3640 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3641 { 3642 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3643 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3644 bool ipv4 = !ipv6 || metadata; 3645 int ret = 0; 3646 3647 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3648 #if IS_ENABLED(CONFIG_IPV6) 3649 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3650 if (ipv6) { 3651 ret = __vxlan_sock_add(vxlan, true); 3652 if (ret < 0 && ret != -EAFNOSUPPORT) 3653 ipv4 = false; 3654 } 3655 #endif 3656 if (ipv4) 3657 ret = __vxlan_sock_add(vxlan, false); 3658 if (ret < 0) 3659 vxlan_sock_release(vxlan); 3660 return ret; 3661 } 3662 3663 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3664 struct vxlan_config *conf, __be32 vni) 3665 { 3666 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3667 struct vxlan_dev *tmp; 3668 3669 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3670 if (tmp == vxlan) 3671 continue; 3672 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3673 if (!vxlan_vnifilter_lookup(tmp, vni)) 3674 continue; 3675 } else if (tmp->cfg.vni != vni) { 3676 continue; 3677 } 3678 if (tmp->cfg.dst_port != conf->dst_port) 3679 continue; 3680 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3681 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3682 continue; 3683 3684 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3685 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3686 continue; 3687 3688 return -EEXIST; 3689 } 3690 3691 return 0; 3692 } 3693 3694 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3695 struct net_device **lower, 3696 struct vxlan_dev *old, 3697 struct netlink_ext_ack *extack) 3698 { 3699 bool use_ipv6 = false; 3700 3701 if (conf->flags & VXLAN_F_GPE) { 3702 /* For now, allow GPE only together with 3703 * COLLECT_METADATA. This can be relaxed later; in such 3704 * case, the other side of the PtP link will have to be 3705 * provided. 3706 */ 3707 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3708 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3709 NL_SET_ERR_MSG(extack, 3710 "VXLAN GPE does not support this combination of attributes"); 3711 return -EINVAL; 3712 } 3713 } 3714 3715 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3716 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3717 conf->remote_ip.sa.sa_family = AF_INET; 3718 conf->saddr.sa.sa_family = AF_INET; 3719 } else if (!conf->remote_ip.sa.sa_family) { 3720 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3721 } else if (!conf->saddr.sa.sa_family) { 3722 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3723 } 3724 3725 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3726 NL_SET_ERR_MSG(extack, 3727 "Local and remote address must be from the same family"); 3728 return -EINVAL; 3729 } 3730 3731 if (vxlan_addr_multicast(&conf->saddr)) { 3732 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3733 return -EINVAL; 3734 } 3735 3736 if (conf->saddr.sa.sa_family == AF_INET6) { 3737 if (!IS_ENABLED(CONFIG_IPV6)) { 3738 NL_SET_ERR_MSG(extack, 3739 "IPv6 support not enabled in the kernel"); 3740 return -EPFNOSUPPORT; 3741 } 3742 use_ipv6 = true; 3743 conf->flags |= VXLAN_F_IPV6; 3744 3745 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3746 int local_type = 3747 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3748 int remote_type = 3749 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3750 3751 if (local_type & IPV6_ADDR_LINKLOCAL) { 3752 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3753 (remote_type != IPV6_ADDR_ANY)) { 3754 NL_SET_ERR_MSG(extack, 3755 "Invalid combination of local and remote address scopes"); 3756 return -EINVAL; 3757 } 3758 3759 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3760 } else { 3761 if (remote_type == 3762 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3763 NL_SET_ERR_MSG(extack, 3764 "Invalid combination of local and remote address scopes"); 3765 return -EINVAL; 3766 } 3767 3768 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3769 } 3770 } 3771 } 3772 3773 if (conf->label && !use_ipv6) { 3774 NL_SET_ERR_MSG(extack, 3775 "Label attribute only applies to IPv6 VXLAN devices"); 3776 return -EINVAL; 3777 } 3778 3779 if (conf->label_policy && !use_ipv6) { 3780 NL_SET_ERR_MSG(extack, 3781 "Label policy only applies to IPv6 VXLAN devices"); 3782 return -EINVAL; 3783 } 3784 3785 if (conf->remote_ifindex) { 3786 struct net_device *lowerdev; 3787 3788 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3789 if (!lowerdev) { 3790 NL_SET_ERR_MSG(extack, 3791 "Invalid local interface, device not found"); 3792 return -ENODEV; 3793 } 3794 3795 #if IS_ENABLED(CONFIG_IPV6) 3796 if (use_ipv6) { 3797 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3798 3799 if (idev && idev->cnf.disable_ipv6) { 3800 NL_SET_ERR_MSG(extack, 3801 "IPv6 support disabled by administrator"); 3802 return -EPERM; 3803 } 3804 } 3805 #endif 3806 3807 *lower = lowerdev; 3808 } else { 3809 if (vxlan_addr_multicast(&conf->remote_ip)) { 3810 NL_SET_ERR_MSG(extack, 3811 "Local interface required for multicast remote destination"); 3812 3813 return -EINVAL; 3814 } 3815 3816 #if IS_ENABLED(CONFIG_IPV6) 3817 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3818 NL_SET_ERR_MSG(extack, 3819 "Local interface required for link-local local/remote addresses"); 3820 return -EINVAL; 3821 } 3822 #endif 3823 3824 *lower = NULL; 3825 } 3826 3827 if (!conf->dst_port) { 3828 if (conf->flags & VXLAN_F_GPE) 3829 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3830 else 3831 conf->dst_port = htons(vxlan_port); 3832 } 3833 3834 if (!conf->age_interval) 3835 conf->age_interval = FDB_AGE_DEFAULT; 3836 3837 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3838 NL_SET_ERR_MSG(extack, 3839 "A VXLAN device with the specified VNI already exists"); 3840 return -EEXIST; 3841 } 3842 3843 return 0; 3844 } 3845 3846 static void vxlan_config_apply(struct net_device *dev, 3847 struct vxlan_config *conf, 3848 struct net_device *lowerdev, 3849 struct net *src_net, 3850 bool changelink) 3851 { 3852 struct vxlan_dev *vxlan = netdev_priv(dev); 3853 struct vxlan_rdst *dst = &vxlan->default_dst; 3854 unsigned short needed_headroom = ETH_HLEN; 3855 int max_mtu = ETH_MAX_MTU; 3856 u32 flags = conf->flags; 3857 3858 if (!changelink) { 3859 if (flags & VXLAN_F_GPE) 3860 vxlan_raw_setup(dev); 3861 else 3862 vxlan_ether_setup(dev); 3863 3864 if (conf->mtu) 3865 dev->mtu = conf->mtu; 3866 3867 vxlan->net = src_net; 3868 } 3869 3870 dst->remote_vni = conf->vni; 3871 3872 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3873 3874 if (lowerdev) { 3875 dst->remote_ifindex = conf->remote_ifindex; 3876 3877 netif_inherit_tso_max(dev, lowerdev); 3878 3879 needed_headroom = lowerdev->hard_header_len; 3880 needed_headroom += lowerdev->needed_headroom; 3881 3882 dev->needed_tailroom = lowerdev->needed_tailroom; 3883 3884 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3885 if (max_mtu < ETH_MIN_MTU) 3886 max_mtu = ETH_MIN_MTU; 3887 3888 if (!changelink && !conf->mtu) 3889 dev->mtu = max_mtu; 3890 } 3891 3892 if (dev->mtu > max_mtu) 3893 dev->mtu = max_mtu; 3894 3895 if (flags & VXLAN_F_COLLECT_METADATA) 3896 flags |= VXLAN_F_IPV6; 3897 needed_headroom += vxlan_headroom(flags); 3898 dev->needed_headroom = needed_headroom; 3899 3900 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3901 } 3902 3903 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3904 struct vxlan_config *conf, bool changelink, 3905 struct netlink_ext_ack *extack) 3906 { 3907 struct vxlan_dev *vxlan = netdev_priv(dev); 3908 struct net_device *lowerdev; 3909 int ret; 3910 3911 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3912 if (ret) 3913 return ret; 3914 3915 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3916 3917 return 0; 3918 } 3919 3920 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3921 struct vxlan_config *conf, 3922 struct netlink_ext_ack *extack) 3923 { 3924 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3925 struct vxlan_dev *vxlan = netdev_priv(dev); 3926 struct net_device *remote_dev = NULL; 3927 struct vxlan_fdb *f = NULL; 3928 bool unregister = false; 3929 struct vxlan_rdst *dst; 3930 int err; 3931 3932 dst = &vxlan->default_dst; 3933 err = vxlan_dev_configure(net, dev, conf, false, extack); 3934 if (err) 3935 return err; 3936 3937 dev->ethtool_ops = &vxlan_ethtool_ops; 3938 3939 /* create an fdb entry for a valid default destination */ 3940 if (!vxlan_addr_any(&dst->remote_ip)) { 3941 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3942 &dst->remote_ip, 3943 NUD_REACHABLE | NUD_PERMANENT, 3944 vxlan->cfg.dst_port, 3945 dst->remote_vni, 3946 dst->remote_vni, 3947 dst->remote_ifindex, 3948 NTF_SELF, 0, &f, extack); 3949 if (err) 3950 return err; 3951 } 3952 3953 err = register_netdevice(dev); 3954 if (err) 3955 goto errout; 3956 unregister = true; 3957 3958 if (dst->remote_ifindex) { 3959 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3960 if (!remote_dev) { 3961 err = -ENODEV; 3962 goto errout; 3963 } 3964 3965 err = netdev_upper_dev_link(remote_dev, dev, extack); 3966 if (err) 3967 goto errout; 3968 } 3969 3970 err = rtnl_configure_link(dev, NULL, 0, NULL); 3971 if (err < 0) 3972 goto unlink; 3973 3974 if (f) { 3975 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3976 3977 /* notify default fdb entry */ 3978 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3979 RTM_NEWNEIGH, true, extack); 3980 if (err) { 3981 vxlan_fdb_destroy(vxlan, f, false, false); 3982 if (remote_dev) 3983 netdev_upper_dev_unlink(remote_dev, dev); 3984 goto unregister; 3985 } 3986 } 3987 3988 list_add(&vxlan->next, &vn->vxlan_list); 3989 if (remote_dev) 3990 dst->remote_dev = remote_dev; 3991 return 0; 3992 unlink: 3993 if (remote_dev) 3994 netdev_upper_dev_unlink(remote_dev, dev); 3995 errout: 3996 /* unregister_netdevice() destroys the default FDB entry with deletion 3997 * notification. But the addition notification was not sent yet, so 3998 * destroy the entry by hand here. 3999 */ 4000 if (f) 4001 __vxlan_fdb_free(f); 4002 unregister: 4003 if (unregister) 4004 unregister_netdevice(dev); 4005 return err; 4006 } 4007 4008 /* Set/clear flags based on attribute */ 4009 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4010 int attrtype, unsigned long mask, bool changelink, 4011 bool changelink_supported, 4012 struct netlink_ext_ack *extack) 4013 { 4014 unsigned long flags; 4015 4016 if (!tb[attrtype]) 4017 return 0; 4018 4019 if (changelink && !changelink_supported) { 4020 vxlan_flag_attr_error(attrtype, extack); 4021 return -EOPNOTSUPP; 4022 } 4023 4024 if (vxlan_policy[attrtype].type == NLA_FLAG) 4025 flags = conf->flags | mask; 4026 else if (nla_get_u8(tb[attrtype])) 4027 flags = conf->flags | mask; 4028 else 4029 flags = conf->flags & ~mask; 4030 4031 conf->flags = flags; 4032 4033 return 0; 4034 } 4035 4036 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4037 struct net_device *dev, struct vxlan_config *conf, 4038 bool changelink, struct netlink_ext_ack *extack) 4039 { 4040 struct vxlan_dev *vxlan = netdev_priv(dev); 4041 int err = 0; 4042 4043 memset(conf, 0, sizeof(*conf)); 4044 4045 /* if changelink operation, start with old existing cfg */ 4046 if (changelink) 4047 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4048 4049 if (data[IFLA_VXLAN_ID]) { 4050 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4051 4052 if (changelink && (vni != conf->vni)) { 4053 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4054 return -EOPNOTSUPP; 4055 } 4056 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4057 } 4058 4059 if (data[IFLA_VXLAN_GROUP]) { 4060 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4061 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4062 return -EOPNOTSUPP; 4063 } 4064 4065 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4066 conf->remote_ip.sa.sa_family = AF_INET; 4067 } else if (data[IFLA_VXLAN_GROUP6]) { 4068 if (!IS_ENABLED(CONFIG_IPV6)) { 4069 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4070 return -EPFNOSUPPORT; 4071 } 4072 4073 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4074 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4075 return -EOPNOTSUPP; 4076 } 4077 4078 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4079 conf->remote_ip.sa.sa_family = AF_INET6; 4080 } 4081 4082 if (data[IFLA_VXLAN_LOCAL]) { 4083 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4084 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4085 return -EOPNOTSUPP; 4086 } 4087 4088 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4089 conf->saddr.sa.sa_family = AF_INET; 4090 } else if (data[IFLA_VXLAN_LOCAL6]) { 4091 if (!IS_ENABLED(CONFIG_IPV6)) { 4092 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4093 return -EPFNOSUPPORT; 4094 } 4095 4096 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4097 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4098 return -EOPNOTSUPP; 4099 } 4100 4101 /* TODO: respect scope id */ 4102 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4103 conf->saddr.sa.sa_family = AF_INET6; 4104 } 4105 4106 if (data[IFLA_VXLAN_LINK]) 4107 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4108 4109 if (data[IFLA_VXLAN_TOS]) 4110 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4111 4112 if (data[IFLA_VXLAN_TTL]) 4113 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4114 4115 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4116 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4117 VXLAN_F_TTL_INHERIT, changelink, false, 4118 extack); 4119 if (err) 4120 return err; 4121 4122 } 4123 4124 if (data[IFLA_VXLAN_LABEL]) 4125 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4126 IPV6_FLOWLABEL_MASK; 4127 if (data[IFLA_VXLAN_LABEL_POLICY]) 4128 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4129 4130 if (data[IFLA_VXLAN_LEARNING]) { 4131 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4132 VXLAN_F_LEARN, changelink, true, 4133 extack); 4134 if (err) 4135 return err; 4136 } else if (!changelink) { 4137 /* default to learn on a new device */ 4138 conf->flags |= VXLAN_F_LEARN; 4139 } 4140 4141 if (data[IFLA_VXLAN_AGEING]) 4142 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4143 4144 if (data[IFLA_VXLAN_PROXY]) { 4145 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4146 VXLAN_F_PROXY, changelink, false, 4147 extack); 4148 if (err) 4149 return err; 4150 } 4151 4152 if (data[IFLA_VXLAN_RSC]) { 4153 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4154 VXLAN_F_RSC, changelink, false, 4155 extack); 4156 if (err) 4157 return err; 4158 } 4159 4160 if (data[IFLA_VXLAN_L2MISS]) { 4161 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4162 VXLAN_F_L2MISS, changelink, false, 4163 extack); 4164 if (err) 4165 return err; 4166 } 4167 4168 if (data[IFLA_VXLAN_L3MISS]) { 4169 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4170 VXLAN_F_L3MISS, changelink, false, 4171 extack); 4172 if (err) 4173 return err; 4174 } 4175 4176 if (data[IFLA_VXLAN_LIMIT]) { 4177 if (changelink) { 4178 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4179 "Cannot change limit"); 4180 return -EOPNOTSUPP; 4181 } 4182 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4183 } 4184 4185 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4186 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4187 VXLAN_F_COLLECT_METADATA, changelink, false, 4188 extack); 4189 if (err) 4190 return err; 4191 } 4192 4193 if (data[IFLA_VXLAN_PORT_RANGE]) { 4194 if (!changelink) { 4195 const struct ifla_vxlan_port_range *p 4196 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4197 conf->port_min = ntohs(p->low); 4198 conf->port_max = ntohs(p->high); 4199 } else { 4200 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4201 "Cannot change port range"); 4202 return -EOPNOTSUPP; 4203 } 4204 } 4205 4206 if (data[IFLA_VXLAN_PORT]) { 4207 if (changelink) { 4208 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4209 "Cannot change port"); 4210 return -EOPNOTSUPP; 4211 } 4212 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4213 } 4214 4215 if (data[IFLA_VXLAN_UDP_CSUM]) { 4216 if (changelink) { 4217 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4218 "Cannot change UDP_CSUM flag"); 4219 return -EOPNOTSUPP; 4220 } 4221 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4222 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4223 } 4224 4225 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4226 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4227 VXLAN_F_LOCALBYPASS, changelink, 4228 true, extack); 4229 if (err) 4230 return err; 4231 } else if (!changelink) { 4232 /* default to local bypass on a new device */ 4233 conf->flags |= VXLAN_F_LOCALBYPASS; 4234 } 4235 4236 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4237 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4238 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4239 false, extack); 4240 if (err) 4241 return err; 4242 } 4243 4244 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4245 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4246 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4247 false, extack); 4248 if (err) 4249 return err; 4250 } 4251 4252 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4253 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4254 VXLAN_F_REMCSUM_TX, changelink, false, 4255 extack); 4256 if (err) 4257 return err; 4258 } 4259 4260 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4261 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4262 VXLAN_F_REMCSUM_RX, changelink, false, 4263 extack); 4264 if (err) 4265 return err; 4266 } 4267 4268 if (data[IFLA_VXLAN_GBP]) { 4269 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4270 VXLAN_F_GBP, changelink, false, extack); 4271 if (err) 4272 return err; 4273 } 4274 4275 if (data[IFLA_VXLAN_GPE]) { 4276 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4277 VXLAN_F_GPE, changelink, false, 4278 extack); 4279 if (err) 4280 return err; 4281 } 4282 4283 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4284 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4285 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4286 false, extack); 4287 if (err) 4288 return err; 4289 } 4290 4291 if (tb[IFLA_MTU]) { 4292 if (changelink) { 4293 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4294 "Cannot change mtu"); 4295 return -EOPNOTSUPP; 4296 } 4297 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4298 } 4299 4300 if (data[IFLA_VXLAN_DF]) 4301 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4302 4303 if (data[IFLA_VXLAN_VNIFILTER]) { 4304 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4305 VXLAN_F_VNIFILTER, changelink, false, 4306 extack); 4307 if (err) 4308 return err; 4309 4310 if ((conf->flags & VXLAN_F_VNIFILTER) && 4311 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4312 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4313 "vxlan vnifilter only valid in collect metadata mode"); 4314 return -EINVAL; 4315 } 4316 } 4317 4318 return 0; 4319 } 4320 4321 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4322 struct nlattr *tb[], struct nlattr *data[], 4323 struct netlink_ext_ack *extack) 4324 { 4325 struct vxlan_config conf; 4326 int err; 4327 4328 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4329 if (err) 4330 return err; 4331 4332 return __vxlan_dev_create(src_net, dev, &conf, extack); 4333 } 4334 4335 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4336 struct nlattr *data[], 4337 struct netlink_ext_ack *extack) 4338 { 4339 struct vxlan_dev *vxlan = netdev_priv(dev); 4340 struct net_device *lowerdev; 4341 struct vxlan_config conf; 4342 struct vxlan_rdst *dst; 4343 int err; 4344 4345 dst = &vxlan->default_dst; 4346 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4347 if (err) 4348 return err; 4349 4350 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4351 vxlan, extack); 4352 if (err) 4353 return err; 4354 4355 if (dst->remote_dev == lowerdev) 4356 lowerdev = NULL; 4357 4358 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4359 extack); 4360 if (err) 4361 return err; 4362 4363 /* handle default dst entry */ 4364 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4365 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4366 4367 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4368 if (!vxlan_addr_any(&conf.remote_ip)) { 4369 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4370 &conf.remote_ip, 4371 NUD_REACHABLE | NUD_PERMANENT, 4372 NLM_F_APPEND | NLM_F_CREATE, 4373 vxlan->cfg.dst_port, 4374 conf.vni, conf.vni, 4375 conf.remote_ifindex, 4376 NTF_SELF, 0, true, extack); 4377 if (err) { 4378 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4379 netdev_adjacent_change_abort(dst->remote_dev, 4380 lowerdev, dev); 4381 return err; 4382 } 4383 } 4384 if (!vxlan_addr_any(&dst->remote_ip)) 4385 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4386 dst->remote_ip, 4387 vxlan->cfg.dst_port, 4388 dst->remote_vni, 4389 dst->remote_vni, 4390 dst->remote_ifindex, 4391 true); 4392 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4393 4394 /* If vni filtering device, also update fdb entries of 4395 * all vnis that were using default remote ip 4396 */ 4397 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4398 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4399 &conf.remote_ip, extack); 4400 if (err) { 4401 netdev_adjacent_change_abort(dst->remote_dev, 4402 lowerdev, dev); 4403 return err; 4404 } 4405 } 4406 } 4407 4408 if (conf.age_interval != vxlan->cfg.age_interval) 4409 mod_timer(&vxlan->age_timer, jiffies); 4410 4411 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4412 if (lowerdev && lowerdev != dst->remote_dev) 4413 dst->remote_dev = lowerdev; 4414 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4415 return 0; 4416 } 4417 4418 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4419 { 4420 struct vxlan_dev *vxlan = netdev_priv(dev); 4421 struct vxlan_fdb_flush_desc desc = { 4422 /* Default entry is deleted at vxlan_uninit. */ 4423 .ignore_default_entry = true, 4424 }; 4425 4426 vxlan_flush(vxlan, &desc); 4427 4428 list_del(&vxlan->next); 4429 unregister_netdevice_queue(dev, head); 4430 if (vxlan->default_dst.remote_dev) 4431 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4432 } 4433 4434 static size_t vxlan_get_size(const struct net_device *dev) 4435 { 4436 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4437 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4438 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4439 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4440 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4441 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4442 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4443 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4444 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4445 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4446 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4447 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4448 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4449 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4450 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4451 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4452 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4453 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4454 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4455 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4456 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4457 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4458 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4459 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4460 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4461 /* IFLA_VXLAN_PORT_RANGE */ 4462 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4463 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4464 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4465 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4466 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4467 0; 4468 } 4469 4470 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4471 { 4472 const struct vxlan_dev *vxlan = netdev_priv(dev); 4473 const struct vxlan_rdst *dst = &vxlan->default_dst; 4474 struct ifla_vxlan_port_range ports = { 4475 .low = htons(vxlan->cfg.port_min), 4476 .high = htons(vxlan->cfg.port_max), 4477 }; 4478 4479 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4480 goto nla_put_failure; 4481 4482 if (!vxlan_addr_any(&dst->remote_ip)) { 4483 if (dst->remote_ip.sa.sa_family == AF_INET) { 4484 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4485 dst->remote_ip.sin.sin_addr.s_addr)) 4486 goto nla_put_failure; 4487 #if IS_ENABLED(CONFIG_IPV6) 4488 } else { 4489 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4490 &dst->remote_ip.sin6.sin6_addr)) 4491 goto nla_put_failure; 4492 #endif 4493 } 4494 } 4495 4496 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4497 goto nla_put_failure; 4498 4499 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4500 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4501 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4502 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4503 goto nla_put_failure; 4504 #if IS_ENABLED(CONFIG_IPV6) 4505 } else { 4506 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4507 &vxlan->cfg.saddr.sin6.sin6_addr)) 4508 goto nla_put_failure; 4509 #endif 4510 } 4511 } 4512 4513 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4514 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4515 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4516 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4517 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4518 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4519 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4520 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4521 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4522 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4523 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4524 nla_put_u8(skb, IFLA_VXLAN_RSC, 4525 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4526 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4527 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4528 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4529 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4530 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4531 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4532 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4533 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4534 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4535 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4536 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4537 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4538 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4539 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4540 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4541 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4542 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4543 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4544 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4545 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4546 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4547 goto nla_put_failure; 4548 4549 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4550 goto nla_put_failure; 4551 4552 if (vxlan->cfg.flags & VXLAN_F_GBP && 4553 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4554 goto nla_put_failure; 4555 4556 if (vxlan->cfg.flags & VXLAN_F_GPE && 4557 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4558 goto nla_put_failure; 4559 4560 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4561 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4562 goto nla_put_failure; 4563 4564 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4565 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4566 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4567 goto nla_put_failure; 4568 4569 return 0; 4570 4571 nla_put_failure: 4572 return -EMSGSIZE; 4573 } 4574 4575 static struct net *vxlan_get_link_net(const struct net_device *dev) 4576 { 4577 struct vxlan_dev *vxlan = netdev_priv(dev); 4578 4579 return READ_ONCE(vxlan->net); 4580 } 4581 4582 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4583 .kind = "vxlan", 4584 .maxtype = IFLA_VXLAN_MAX, 4585 .policy = vxlan_policy, 4586 .priv_size = sizeof(struct vxlan_dev), 4587 .setup = vxlan_setup, 4588 .validate = vxlan_validate, 4589 .newlink = vxlan_newlink, 4590 .changelink = vxlan_changelink, 4591 .dellink = vxlan_dellink, 4592 .get_size = vxlan_get_size, 4593 .fill_info = vxlan_fill_info, 4594 .get_link_net = vxlan_get_link_net, 4595 }; 4596 4597 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4598 u8 name_assign_type, 4599 struct vxlan_config *conf) 4600 { 4601 struct nlattr *tb[IFLA_MAX + 1]; 4602 struct net_device *dev; 4603 int err; 4604 4605 memset(&tb, 0, sizeof(tb)); 4606 4607 dev = rtnl_create_link(net, name, name_assign_type, 4608 &vxlan_link_ops, tb, NULL); 4609 if (IS_ERR(dev)) 4610 return dev; 4611 4612 err = __vxlan_dev_create(net, dev, conf, NULL); 4613 if (err < 0) { 4614 free_netdev(dev); 4615 return ERR_PTR(err); 4616 } 4617 4618 err = rtnl_configure_link(dev, NULL, 0, NULL); 4619 if (err < 0) { 4620 LIST_HEAD(list_kill); 4621 4622 vxlan_dellink(dev, &list_kill); 4623 unregister_netdevice_many(&list_kill); 4624 return ERR_PTR(err); 4625 } 4626 4627 return dev; 4628 } 4629 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4630 4631 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4632 struct net_device *dev) 4633 { 4634 struct vxlan_dev *vxlan, *next; 4635 LIST_HEAD(list_kill); 4636 4637 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4638 struct vxlan_rdst *dst = &vxlan->default_dst; 4639 4640 /* In case we created vxlan device with carrier 4641 * and we loose the carrier due to module unload 4642 * we also need to remove vxlan device. In other 4643 * cases, it's not necessary and remote_ifindex 4644 * is 0 here, so no matches. 4645 */ 4646 if (dst->remote_ifindex == dev->ifindex) 4647 vxlan_dellink(vxlan->dev, &list_kill); 4648 } 4649 4650 unregister_netdevice_many(&list_kill); 4651 } 4652 4653 static int vxlan_netdevice_event(struct notifier_block *unused, 4654 unsigned long event, void *ptr) 4655 { 4656 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4657 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4658 4659 if (event == NETDEV_UNREGISTER) 4660 vxlan_handle_lowerdev_unregister(vn, dev); 4661 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4662 vxlan_offload_rx_ports(dev, true); 4663 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4664 vxlan_offload_rx_ports(dev, false); 4665 4666 return NOTIFY_DONE; 4667 } 4668 4669 static struct notifier_block vxlan_notifier_block __read_mostly = { 4670 .notifier_call = vxlan_netdevice_event, 4671 }; 4672 4673 static void 4674 vxlan_fdb_offloaded_set(struct net_device *dev, 4675 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4676 { 4677 struct vxlan_dev *vxlan = netdev_priv(dev); 4678 struct vxlan_rdst *rdst; 4679 struct vxlan_fdb *f; 4680 u32 hash_index; 4681 4682 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4683 4684 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4685 4686 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4687 if (!f) 4688 goto out; 4689 4690 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4691 fdb_info->remote_port, 4692 fdb_info->remote_vni, 4693 fdb_info->remote_ifindex); 4694 if (!rdst) 4695 goto out; 4696 4697 rdst->offloaded = fdb_info->offloaded; 4698 4699 out: 4700 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4701 } 4702 4703 static int 4704 vxlan_fdb_external_learn_add(struct net_device *dev, 4705 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4706 { 4707 struct vxlan_dev *vxlan = netdev_priv(dev); 4708 struct netlink_ext_ack *extack; 4709 u32 hash_index; 4710 int err; 4711 4712 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4713 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4714 4715 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4716 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4717 NUD_REACHABLE, 4718 NLM_F_CREATE | NLM_F_REPLACE, 4719 fdb_info->remote_port, 4720 fdb_info->vni, 4721 fdb_info->remote_vni, 4722 fdb_info->remote_ifindex, 4723 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4724 0, false, extack); 4725 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4726 4727 return err; 4728 } 4729 4730 static int 4731 vxlan_fdb_external_learn_del(struct net_device *dev, 4732 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4733 { 4734 struct vxlan_dev *vxlan = netdev_priv(dev); 4735 struct vxlan_fdb *f; 4736 u32 hash_index; 4737 int err = 0; 4738 4739 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4740 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4741 4742 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4743 if (!f) 4744 err = -ENOENT; 4745 else if (f->flags & NTF_EXT_LEARNED) 4746 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4747 fdb_info->remote_ip, 4748 fdb_info->remote_port, 4749 fdb_info->vni, 4750 fdb_info->remote_vni, 4751 fdb_info->remote_ifindex, 4752 false); 4753 4754 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4755 4756 return err; 4757 } 4758 4759 static int vxlan_switchdev_event(struct notifier_block *unused, 4760 unsigned long event, void *ptr) 4761 { 4762 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4763 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4764 int err = 0; 4765 4766 switch (event) { 4767 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4768 vxlan_fdb_offloaded_set(dev, ptr); 4769 break; 4770 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4771 fdb_info = ptr; 4772 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4773 if (err) { 4774 err = notifier_from_errno(err); 4775 break; 4776 } 4777 fdb_info->offloaded = true; 4778 vxlan_fdb_offloaded_set(dev, fdb_info); 4779 break; 4780 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4781 fdb_info = ptr; 4782 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4783 if (err) { 4784 err = notifier_from_errno(err); 4785 break; 4786 } 4787 fdb_info->offloaded = false; 4788 vxlan_fdb_offloaded_set(dev, fdb_info); 4789 break; 4790 } 4791 4792 return err; 4793 } 4794 4795 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4796 .notifier_call = vxlan_switchdev_event, 4797 }; 4798 4799 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4800 { 4801 struct vxlan_fdb *fdb; 4802 struct vxlan_dev *vxlan; 4803 u32 hash_index; 4804 4805 rcu_read_lock(); 4806 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4807 vxlan = rcu_dereference(fdb->vdev); 4808 WARN_ON(!vxlan); 4809 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4810 vxlan->default_dst.remote_vni); 4811 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4812 if (!hlist_unhashed(&fdb->hlist)) 4813 vxlan_fdb_destroy(vxlan, fdb, false, false); 4814 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4815 } 4816 rcu_read_unlock(); 4817 } 4818 4819 static int vxlan_nexthop_event(struct notifier_block *nb, 4820 unsigned long event, void *ptr) 4821 { 4822 struct nh_notifier_info *info = ptr; 4823 struct nexthop *nh; 4824 4825 if (event != NEXTHOP_EVENT_DEL) 4826 return NOTIFY_DONE; 4827 4828 nh = nexthop_find_by_id(info->net, info->id); 4829 if (!nh) 4830 return NOTIFY_DONE; 4831 4832 vxlan_fdb_nh_flush(nh); 4833 4834 return NOTIFY_DONE; 4835 } 4836 4837 static __net_init int vxlan_init_net(struct net *net) 4838 { 4839 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4840 unsigned int h; 4841 4842 INIT_LIST_HEAD(&vn->vxlan_list); 4843 spin_lock_init(&vn->sock_lock); 4844 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4845 4846 for (h = 0; h < PORT_HASH_SIZE; ++h) 4847 INIT_HLIST_HEAD(&vn->sock_list[h]); 4848 4849 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4850 NULL); 4851 } 4852 4853 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4854 struct list_head *dev_to_kill) 4855 { 4856 struct vxlan_dev *vxlan, *next; 4857 4858 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4859 vxlan_dellink(vxlan->dev, dev_to_kill); 4860 } 4861 4862 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4863 struct list_head *dev_to_kill) 4864 { 4865 struct net *net; 4866 4867 ASSERT_RTNL(); 4868 list_for_each_entry(net, net_list, exit_list) { 4869 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4870 4871 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4872 4873 vxlan_destroy_tunnels(vn, dev_to_kill); 4874 } 4875 } 4876 4877 static void __net_exit vxlan_exit_net(struct net *net) 4878 { 4879 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4880 unsigned int h; 4881 4882 for (h = 0; h < PORT_HASH_SIZE; ++h) 4883 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4884 } 4885 4886 static struct pernet_operations vxlan_net_ops = { 4887 .init = vxlan_init_net, 4888 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4889 .exit = vxlan_exit_net, 4890 .id = &vxlan_net_id, 4891 .size = sizeof(struct vxlan_net), 4892 }; 4893 4894 static int __init vxlan_init_module(void) 4895 { 4896 int rc; 4897 4898 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4899 4900 rc = register_pernet_subsys(&vxlan_net_ops); 4901 if (rc) 4902 goto out1; 4903 4904 rc = register_netdevice_notifier(&vxlan_notifier_block); 4905 if (rc) 4906 goto out2; 4907 4908 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4909 if (rc) 4910 goto out3; 4911 4912 rc = rtnl_link_register(&vxlan_link_ops); 4913 if (rc) 4914 goto out4; 4915 4916 vxlan_vnifilter_init(); 4917 4918 return 0; 4919 out4: 4920 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4921 out3: 4922 unregister_netdevice_notifier(&vxlan_notifier_block); 4923 out2: 4924 unregister_pernet_subsys(&vxlan_net_ops); 4925 out1: 4926 return rc; 4927 } 4928 late_initcall(vxlan_init_module); 4929 4930 static void __exit vxlan_cleanup_module(void) 4931 { 4932 vxlan_vnifilter_uninit(); 4933 rtnl_link_unregister(&vxlan_link_ops); 4934 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4935 unregister_netdevice_notifier(&vxlan_notifier_block); 4936 unregister_pernet_subsys(&vxlan_net_ops); 4937 /* rcu_barrier() is called by netns */ 4938 } 4939 module_exit(vxlan_cleanup_module); 4940 4941 MODULE_LICENSE("GPL"); 4942 MODULE_VERSION(VXLAN_VERSION); 4943 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4944 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4945 MODULE_ALIAS_RTNL_LINK("vxlan"); 4946