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 if (err < 0) 281 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 282 } 283 284 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 285 const struct vxlan_fdb *fdb, 286 const struct vxlan_rdst *rd, 287 struct netlink_ext_ack *extack, 288 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 289 { 290 fdb_info->info.dev = vxlan->dev; 291 fdb_info->info.extack = extack; 292 fdb_info->remote_ip = rd->remote_ip; 293 fdb_info->remote_port = rd->remote_port; 294 fdb_info->remote_vni = rd->remote_vni; 295 fdb_info->remote_ifindex = rd->remote_ifindex; 296 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN); 297 fdb_info->vni = fdb->vni; 298 fdb_info->offloaded = rd->offloaded; 299 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 300 } 301 302 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 303 struct vxlan_fdb *fdb, 304 struct vxlan_rdst *rd, 305 bool adding, 306 struct netlink_ext_ack *extack) 307 { 308 struct switchdev_notifier_vxlan_fdb_info info; 309 enum switchdev_notifier_type notifier_type; 310 int ret; 311 312 if (WARN_ON(!rd)) 313 return 0; 314 315 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 316 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 317 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 318 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 319 &info.info, extack); 320 return notifier_to_errno(ret); 321 } 322 323 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 324 struct vxlan_rdst *rd, int type, bool swdev_notify, 325 struct netlink_ext_ack *extack) 326 { 327 int err; 328 329 if (swdev_notify && rd) { 330 switch (type) { 331 case RTM_NEWNEIGH: 332 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 333 true, extack); 334 if (err) 335 return err; 336 break; 337 case RTM_DELNEIGH: 338 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 339 false, extack); 340 break; 341 } 342 } 343 344 __vxlan_fdb_notify(vxlan, fdb, rd, type); 345 return 0; 346 } 347 348 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 349 { 350 struct vxlan_dev *vxlan = netdev_priv(dev); 351 struct vxlan_fdb f = { 352 .state = NUD_STALE, 353 }; 354 struct vxlan_rdst remote = { 355 .remote_ip = *ipa, /* goes to NDA_DST */ 356 .remote_vni = cpu_to_be32(VXLAN_N_VID), 357 }; 358 359 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 360 } 361 362 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 363 { 364 struct vxlan_fdb f = { 365 .state = NUD_STALE, 366 }; 367 struct vxlan_rdst remote = { }; 368 369 memcpy(f.eth_addr, eth_addr, ETH_ALEN); 370 371 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 372 } 373 374 /* Hash Ethernet address */ 375 static u32 eth_hash(const unsigned char *addr) 376 { 377 u64 value = get_unaligned((u64 *)addr); 378 379 /* only want 6 bytes */ 380 #ifdef __BIG_ENDIAN 381 value >>= 16; 382 #else 383 value <<= 16; 384 #endif 385 return hash_64(value, FDB_HASH_BITS); 386 } 387 388 u32 eth_vni_hash(const unsigned char *addr, __be32 vni) 389 { 390 /* use 1 byte of OUI and 3 bytes of NIC */ 391 u32 key = get_unaligned((u32 *)(addr + 2)); 392 393 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1); 394 } 395 396 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni) 397 { 398 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) 399 return eth_vni_hash(mac, vni); 400 else 401 return eth_hash(mac); 402 } 403 404 /* Hash chain to use given mac address */ 405 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, 406 const u8 *mac, __be32 vni) 407 { 408 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)]; 409 } 410 411 /* Look up Ethernet address in forwarding table */ 412 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, 413 const u8 *mac, __be32 vni) 414 { 415 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni); 416 struct vxlan_fdb *f; 417 418 hlist_for_each_entry_rcu(f, head, hlist) { 419 if (ether_addr_equal(mac, f->eth_addr)) { 420 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 421 if (vni == f->vni) 422 return f; 423 } else { 424 return f; 425 } 426 } 427 } 428 429 return NULL; 430 } 431 432 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 433 const u8 *mac, __be32 vni) 434 { 435 struct vxlan_fdb *f; 436 437 f = __vxlan_find_mac(vxlan, mac, vni); 438 if (f && f->used != jiffies) 439 f->used = jiffies; 440 441 return f; 442 } 443 444 /* caller should hold vxlan->hash_lock */ 445 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 446 union vxlan_addr *ip, __be16 port, 447 __be32 vni, __u32 ifindex) 448 { 449 struct vxlan_rdst *rd; 450 451 list_for_each_entry(rd, &f->remotes, list) { 452 if (vxlan_addr_equal(&rd->remote_ip, ip) && 453 rd->remote_port == port && 454 rd->remote_vni == vni && 455 rd->remote_ifindex == ifindex) 456 return rd; 457 } 458 459 return NULL; 460 } 461 462 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 463 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 464 { 465 struct vxlan_dev *vxlan = netdev_priv(dev); 466 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 467 struct vxlan_rdst *rdst; 468 struct vxlan_fdb *f; 469 int rc = 0; 470 471 if (is_multicast_ether_addr(mac) || 472 is_zero_ether_addr(mac)) 473 return -EINVAL; 474 475 ether_addr_copy(eth_addr, mac); 476 477 rcu_read_lock(); 478 479 f = __vxlan_find_mac(vxlan, eth_addr, vni); 480 if (!f) { 481 rc = -ENOENT; 482 goto out; 483 } 484 485 rdst = first_remote_rcu(f); 486 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 487 488 out: 489 rcu_read_unlock(); 490 return rc; 491 } 492 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 493 494 static int vxlan_fdb_notify_one(struct notifier_block *nb, 495 const struct vxlan_dev *vxlan, 496 const struct vxlan_fdb *f, 497 const struct vxlan_rdst *rdst, 498 struct netlink_ext_ack *extack) 499 { 500 struct switchdev_notifier_vxlan_fdb_info fdb_info; 501 int rc; 502 503 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 504 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 505 &fdb_info); 506 return notifier_to_errno(rc); 507 } 508 509 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 510 struct notifier_block *nb, 511 struct netlink_ext_ack *extack) 512 { 513 struct vxlan_dev *vxlan; 514 struct vxlan_rdst *rdst; 515 struct vxlan_fdb *f; 516 unsigned int h; 517 int rc = 0; 518 519 if (!netif_is_vxlan(dev)) 520 return -EINVAL; 521 vxlan = netdev_priv(dev); 522 523 for (h = 0; h < FDB_HASH_SIZE; ++h) { 524 spin_lock_bh(&vxlan->hash_lock[h]); 525 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) { 526 if (f->vni == vni) { 527 list_for_each_entry(rdst, &f->remotes, list) { 528 rc = vxlan_fdb_notify_one(nb, vxlan, 529 f, rdst, 530 extack); 531 if (rc) 532 goto unlock; 533 } 534 } 535 } 536 spin_unlock_bh(&vxlan->hash_lock[h]); 537 } 538 return 0; 539 540 unlock: 541 spin_unlock_bh(&vxlan->hash_lock[h]); 542 return rc; 543 } 544 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 545 546 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 547 { 548 struct vxlan_dev *vxlan; 549 struct vxlan_rdst *rdst; 550 struct vxlan_fdb *f; 551 unsigned int h; 552 553 if (!netif_is_vxlan(dev)) 554 return; 555 vxlan = netdev_priv(dev); 556 557 for (h = 0; h < FDB_HASH_SIZE; ++h) { 558 spin_lock_bh(&vxlan->hash_lock[h]); 559 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) 560 if (f->vni == vni) 561 list_for_each_entry(rdst, &f->remotes, list) 562 rdst->offloaded = false; 563 spin_unlock_bh(&vxlan->hash_lock[h]); 564 } 565 566 } 567 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 568 569 /* Replace destination of unicast mac */ 570 static int vxlan_fdb_replace(struct vxlan_fdb *f, 571 union vxlan_addr *ip, __be16 port, __be32 vni, 572 __u32 ifindex, struct vxlan_rdst *oldrd) 573 { 574 struct vxlan_rdst *rd; 575 576 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 577 if (rd) 578 return 0; 579 580 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 581 if (!rd) 582 return 0; 583 584 *oldrd = *rd; 585 dst_cache_reset(&rd->dst_cache); 586 rd->remote_ip = *ip; 587 rd->remote_port = port; 588 rd->remote_vni = vni; 589 rd->remote_ifindex = ifindex; 590 rd->offloaded = false; 591 return 1; 592 } 593 594 /* Add/update destinations for multicast */ 595 static int vxlan_fdb_append(struct vxlan_fdb *f, 596 union vxlan_addr *ip, __be16 port, __be32 vni, 597 __u32 ifindex, struct vxlan_rdst **rdp) 598 { 599 struct vxlan_rdst *rd; 600 601 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 602 if (rd) 603 return 0; 604 605 rd = kmalloc(sizeof(*rd), GFP_ATOMIC); 606 if (rd == NULL) 607 return -ENOMEM; 608 609 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { 610 kfree(rd); 611 return -ENOMEM; 612 } 613 614 rd->remote_ip = *ip; 615 rd->remote_port = port; 616 rd->offloaded = false; 617 rd->remote_vni = vni; 618 rd->remote_ifindex = ifindex; 619 620 list_add_tail_rcu(&rd->list, &f->remotes); 621 622 *rdp = rd; 623 return 1; 624 } 625 626 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol) 627 { 628 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr; 629 630 /* Need to have Next Protocol set for interfaces in GPE mode. */ 631 if (!gpe->np_applied) 632 return false; 633 /* "The initial version is 0. If a receiver does not support the 634 * version indicated it MUST drop the packet. 635 */ 636 if (gpe->version != 0) 637 return false; 638 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 639 * processing MUST occur." However, we don't implement OAM 640 * processing, thus drop the packet. 641 */ 642 if (gpe->oam_flag) 643 return false; 644 645 *protocol = tun_p_to_eth_p(gpe->next_protocol); 646 if (!*protocol) 647 return false; 648 649 return true; 650 } 651 652 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 653 unsigned int off, 654 struct vxlanhdr *vh, size_t hdrlen, 655 __be32 vni_field, 656 struct gro_remcsum *grc, 657 bool nopartial) 658 { 659 size_t start, offset; 660 661 if (skb->remcsum_offload) 662 return vh; 663 664 if (!NAPI_GRO_CB(skb)->csum_valid) 665 return NULL; 666 667 start = vxlan_rco_start(vni_field); 668 offset = start + vxlan_rco_offset(vni_field); 669 670 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 671 start, offset, grc, nopartial); 672 673 skb->remcsum_offload = 1; 674 675 return vh; 676 } 677 678 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk, 679 struct list_head *head, 680 struct sk_buff *skb, 681 struct gro_remcsum *grc) 682 { 683 struct sk_buff *p; 684 struct vxlanhdr *vh, *vh2; 685 unsigned int hlen, off_vx; 686 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); 687 __be32 flags; 688 689 skb_gro_remcsum_init(grc); 690 691 off_vx = skb_gro_offset(skb); 692 hlen = off_vx + sizeof(*vh); 693 vh = skb_gro_header(skb, hlen, off_vx); 694 if (unlikely(!vh)) 695 return NULL; 696 697 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 698 699 flags = vh->vx_flags; 700 701 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 702 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 703 vh->vx_vni, grc, 704 !!(vs->flags & 705 VXLAN_F_REMCSUM_NOPARTIAL)); 706 707 if (!vh) 708 return NULL; 709 } 710 711 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 712 713 list_for_each_entry(p, head, list) { 714 if (!NAPI_GRO_CB(p)->same_flow) 715 continue; 716 717 vh2 = (struct vxlanhdr *)(p->data + off_vx); 718 if (vh->vx_flags != vh2->vx_flags || 719 vh->vx_vni != vh2->vx_vni) { 720 NAPI_GRO_CB(p)->same_flow = 0; 721 continue; 722 } 723 } 724 725 return vh; 726 } 727 728 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 729 struct list_head *head, 730 struct sk_buff *skb) 731 { 732 struct sk_buff *pp = NULL; 733 struct gro_remcsum grc; 734 int flush = 1; 735 736 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) { 737 pp = call_gro_receive(eth_gro_receive, head, skb); 738 flush = 0; 739 } 740 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 741 return pp; 742 } 743 744 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk, 745 struct list_head *head, 746 struct sk_buff *skb) 747 { 748 const struct packet_offload *ptype; 749 struct sk_buff *pp = NULL; 750 struct gro_remcsum grc; 751 struct vxlanhdr *vh; 752 __be16 protocol; 753 int flush = 1; 754 755 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc); 756 if (vh) { 757 if (!vxlan_parse_gpe_proto(vh, &protocol)) 758 goto out; 759 ptype = gro_find_receive_by_type(protocol); 760 if (!ptype) 761 goto out; 762 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb); 763 flush = 0; 764 } 765 out: 766 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 767 return pp; 768 } 769 770 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 771 { 772 /* Sets 'skb->inner_mac_header' since we are always called with 773 * 'skb->encapsulation' set. 774 */ 775 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 776 } 777 778 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 779 { 780 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff); 781 const struct packet_offload *ptype; 782 int err = -ENOSYS; 783 __be16 protocol; 784 785 if (!vxlan_parse_gpe_proto(vh, &protocol)) 786 return err; 787 ptype = gro_find_complete_by_type(protocol); 788 if (ptype) 789 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 790 return err; 791 } 792 793 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 794 __u16 state, __be32 src_vni, 795 __u16 ndm_flags) 796 { 797 struct vxlan_fdb *f; 798 799 f = kmalloc(sizeof(*f), GFP_ATOMIC); 800 if (!f) 801 return NULL; 802 f->state = state; 803 f->flags = ndm_flags; 804 f->updated = f->used = jiffies; 805 f->vni = src_vni; 806 f->nh = NULL; 807 RCU_INIT_POINTER(f->vdev, vxlan); 808 INIT_LIST_HEAD(&f->nh_list); 809 INIT_LIST_HEAD(&f->remotes); 810 memcpy(f->eth_addr, mac, ETH_ALEN); 811 812 return f; 813 } 814 815 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac, 816 __be32 src_vni, struct vxlan_fdb *f) 817 { 818 ++vxlan->addrcnt; 819 hlist_add_head_rcu(&f->hlist, 820 vxlan_fdb_head(vxlan, mac, src_vni)); 821 } 822 823 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 824 u32 nhid, struct netlink_ext_ack *extack) 825 { 826 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 827 struct nexthop *nh; 828 int err = -EINVAL; 829 830 if (old_nh && old_nh->id == nhid) 831 return 0; 832 833 nh = nexthop_find_by_id(vxlan->net, nhid); 834 if (!nh) { 835 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 836 goto err_inval; 837 } 838 839 if (!nexthop_get(nh)) { 840 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 841 nh = NULL; 842 goto err_inval; 843 } 844 if (!nexthop_is_fdb(nh)) { 845 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 846 goto err_inval; 847 } 848 849 if (!nexthop_is_multipath(nh)) { 850 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 851 goto err_inval; 852 } 853 854 /* check nexthop group family */ 855 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 856 case AF_INET: 857 if (!nexthop_has_v4(nh)) { 858 err = -EAFNOSUPPORT; 859 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 860 goto err_inval; 861 } 862 break; 863 case AF_INET6: 864 if (nexthop_has_v4(nh)) { 865 err = -EAFNOSUPPORT; 866 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 867 goto err_inval; 868 } 869 } 870 871 if (old_nh) { 872 list_del_rcu(&fdb->nh_list); 873 nexthop_put(old_nh); 874 } 875 rcu_assign_pointer(fdb->nh, nh); 876 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 877 return 1; 878 879 err_inval: 880 if (nh) 881 nexthop_put(nh); 882 return err; 883 } 884 885 int vxlan_fdb_create(struct vxlan_dev *vxlan, 886 const u8 *mac, union vxlan_addr *ip, 887 __u16 state, __be16 port, __be32 src_vni, 888 __be32 vni, __u32 ifindex, __u16 ndm_flags, 889 u32 nhid, struct vxlan_fdb **fdb, 890 struct netlink_ext_ack *extack) 891 { 892 struct vxlan_rdst *rd = NULL; 893 struct vxlan_fdb *f; 894 int rc; 895 896 if (vxlan->cfg.addrmax && 897 vxlan->addrcnt >= vxlan->cfg.addrmax) 898 return -ENOSPC; 899 900 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 901 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 902 if (!f) 903 return -ENOMEM; 904 905 if (nhid) 906 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 907 else 908 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 909 if (rc < 0) 910 goto errout; 911 912 *fdb = f; 913 914 return 0; 915 916 errout: 917 kfree(f); 918 return rc; 919 } 920 921 static void __vxlan_fdb_free(struct vxlan_fdb *f) 922 { 923 struct vxlan_rdst *rd, *nd; 924 struct nexthop *nh; 925 926 nh = rcu_dereference_raw(f->nh); 927 if (nh) { 928 rcu_assign_pointer(f->nh, NULL); 929 rcu_assign_pointer(f->vdev, NULL); 930 nexthop_put(nh); 931 } 932 933 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 934 dst_cache_destroy(&rd->dst_cache); 935 kfree(rd); 936 } 937 kfree(f); 938 } 939 940 static void vxlan_fdb_free(struct rcu_head *head) 941 { 942 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 943 944 __vxlan_fdb_free(f); 945 } 946 947 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 948 bool do_notify, bool swdev_notify) 949 { 950 struct vxlan_rdst *rd; 951 952 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr); 953 954 --vxlan->addrcnt; 955 if (do_notify) { 956 if (rcu_access_pointer(f->nh)) 957 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 958 swdev_notify, NULL); 959 else 960 list_for_each_entry(rd, &f->remotes, list) 961 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 962 swdev_notify, NULL); 963 } 964 965 hlist_del_rcu(&f->hlist); 966 list_del_rcu(&f->nh_list); 967 call_rcu(&f->rcu, vxlan_fdb_free); 968 } 969 970 static void vxlan_dst_free(struct rcu_head *head) 971 { 972 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 973 974 dst_cache_destroy(&rd->dst_cache); 975 kfree(rd); 976 } 977 978 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 979 union vxlan_addr *ip, 980 __u16 state, __u16 flags, 981 __be16 port, __be32 vni, 982 __u32 ifindex, __u16 ndm_flags, 983 struct vxlan_fdb *f, u32 nhid, 984 bool swdev_notify, 985 struct netlink_ext_ack *extack) 986 { 987 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 988 struct vxlan_rdst *rd = NULL; 989 struct vxlan_rdst oldrd; 990 int notify = 0; 991 int rc = 0; 992 int err; 993 994 if (nhid && !rcu_access_pointer(f->nh)) { 995 NL_SET_ERR_MSG(extack, 996 "Cannot replace an existing non nexthop fdb with a nexthop"); 997 return -EOPNOTSUPP; 998 } 999 1000 if (nhid && (flags & NLM_F_APPEND)) { 1001 NL_SET_ERR_MSG(extack, 1002 "Cannot append to a nexthop fdb"); 1003 return -EOPNOTSUPP; 1004 } 1005 1006 /* Do not allow an externally learned entry to take over an entry added 1007 * by the user. 1008 */ 1009 if (!(fdb_flags & NTF_EXT_LEARNED) || 1010 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 1011 if (f->state != state) { 1012 f->state = state; 1013 f->updated = jiffies; 1014 notify = 1; 1015 } 1016 if (f->flags != fdb_flags) { 1017 f->flags = fdb_flags; 1018 f->updated = jiffies; 1019 notify = 1; 1020 } 1021 } 1022 1023 if ((flags & NLM_F_REPLACE)) { 1024 /* Only change unicasts */ 1025 if (!(is_multicast_ether_addr(f->eth_addr) || 1026 is_zero_ether_addr(f->eth_addr))) { 1027 if (nhid) { 1028 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1029 if (rc < 0) 1030 return rc; 1031 } else { 1032 rc = vxlan_fdb_replace(f, ip, port, vni, 1033 ifindex, &oldrd); 1034 } 1035 notify |= rc; 1036 } else { 1037 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1038 return -EOPNOTSUPP; 1039 } 1040 } 1041 if ((flags & NLM_F_APPEND) && 1042 (is_multicast_ether_addr(f->eth_addr) || 1043 is_zero_ether_addr(f->eth_addr))) { 1044 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1045 1046 if (rc < 0) 1047 return rc; 1048 notify |= rc; 1049 } 1050 1051 if (ndm_flags & NTF_USE) 1052 f->used = jiffies; 1053 1054 if (notify) { 1055 if (rd == NULL) 1056 rd = first_remote_rtnl(f); 1057 1058 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1059 swdev_notify, extack); 1060 if (err) 1061 goto err_notify; 1062 } 1063 1064 return 0; 1065 1066 err_notify: 1067 if (nhid) 1068 return err; 1069 if ((flags & NLM_F_REPLACE) && rc) 1070 *rd = oldrd; 1071 else if ((flags & NLM_F_APPEND) && rc) { 1072 list_del_rcu(&rd->list); 1073 call_rcu(&rd->rcu, vxlan_dst_free); 1074 } 1075 return err; 1076 } 1077 1078 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1079 const u8 *mac, union vxlan_addr *ip, 1080 __u16 state, __u16 flags, 1081 __be16 port, __be32 src_vni, __be32 vni, 1082 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1083 bool swdev_notify, 1084 struct netlink_ext_ack *extack) 1085 { 1086 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1087 struct vxlan_fdb *f; 1088 int rc; 1089 1090 /* Disallow replace to add a multicast entry */ 1091 if ((flags & NLM_F_REPLACE) && 1092 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1093 return -EOPNOTSUPP; 1094 1095 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1096 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1097 vni, ifindex, fdb_flags, nhid, &f, extack); 1098 if (rc < 0) 1099 return rc; 1100 1101 vxlan_fdb_insert(vxlan, mac, src_vni, f); 1102 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1103 swdev_notify, extack); 1104 if (rc) 1105 goto err_notify; 1106 1107 return 0; 1108 1109 err_notify: 1110 vxlan_fdb_destroy(vxlan, f, false, false); 1111 return rc; 1112 } 1113 1114 /* Add new entry to forwarding table -- assumes lock held */ 1115 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1116 const u8 *mac, union vxlan_addr *ip, 1117 __u16 state, __u16 flags, 1118 __be16 port, __be32 src_vni, __be32 vni, 1119 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1120 bool swdev_notify, 1121 struct netlink_ext_ack *extack) 1122 { 1123 struct vxlan_fdb *f; 1124 1125 f = __vxlan_find_mac(vxlan, mac, src_vni); 1126 if (f) { 1127 if (flags & NLM_F_EXCL) { 1128 netdev_dbg(vxlan->dev, 1129 "lost race to create %pM\n", mac); 1130 return -EEXIST; 1131 } 1132 1133 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1134 vni, ifindex, ndm_flags, f, 1135 nhid, swdev_notify, extack); 1136 } else { 1137 if (!(flags & NLM_F_CREATE)) 1138 return -ENOENT; 1139 1140 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1141 port, src_vni, vni, ifindex, 1142 ndm_flags, nhid, swdev_notify, 1143 extack); 1144 } 1145 } 1146 1147 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1148 struct vxlan_rdst *rd, bool swdev_notify) 1149 { 1150 list_del_rcu(&rd->list); 1151 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1152 call_rcu(&rd->rcu, vxlan_dst_free); 1153 } 1154 1155 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1156 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1157 __be32 *vni, u32 *ifindex, u32 *nhid, 1158 struct netlink_ext_ack *extack) 1159 { 1160 struct net *net = dev_net(vxlan->dev); 1161 int err; 1162 1163 if (tb[NDA_NH_ID] && 1164 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1165 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1166 return -EINVAL; 1167 } 1168 1169 if (tb[NDA_DST]) { 1170 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1171 if (err) { 1172 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1173 return err; 1174 } 1175 } else { 1176 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1177 1178 if (remote->sa.sa_family == AF_INET) { 1179 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1180 ip->sa.sa_family = AF_INET; 1181 #if IS_ENABLED(CONFIG_IPV6) 1182 } else { 1183 ip->sin6.sin6_addr = in6addr_any; 1184 ip->sa.sa_family = AF_INET6; 1185 #endif 1186 } 1187 } 1188 1189 if (tb[NDA_PORT]) { 1190 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1191 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1192 return -EINVAL; 1193 } 1194 *port = nla_get_be16(tb[NDA_PORT]); 1195 } else { 1196 *port = vxlan->cfg.dst_port; 1197 } 1198 1199 if (tb[NDA_VNI]) { 1200 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1201 NL_SET_ERR_MSG(extack, "Invalid vni"); 1202 return -EINVAL; 1203 } 1204 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1205 } else { 1206 *vni = vxlan->default_dst.remote_vni; 1207 } 1208 1209 if (tb[NDA_SRC_VNI]) { 1210 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1211 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1212 return -EINVAL; 1213 } 1214 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1215 } else { 1216 *src_vni = vxlan->default_dst.remote_vni; 1217 } 1218 1219 if (tb[NDA_IFINDEX]) { 1220 struct net_device *tdev; 1221 1222 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1223 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1224 return -EINVAL; 1225 } 1226 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1227 tdev = __dev_get_by_index(net, *ifindex); 1228 if (!tdev) { 1229 NL_SET_ERR_MSG(extack, "Device not found"); 1230 return -EADDRNOTAVAIL; 1231 } 1232 } else { 1233 *ifindex = 0; 1234 } 1235 1236 if (tb[NDA_NH_ID]) 1237 *nhid = nla_get_u32(tb[NDA_NH_ID]); 1238 else 1239 *nhid = 0; 1240 1241 return 0; 1242 } 1243 1244 /* Add static entry (via netlink) */ 1245 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1246 struct net_device *dev, 1247 const unsigned char *addr, u16 vid, u16 flags, 1248 struct netlink_ext_ack *extack) 1249 { 1250 struct vxlan_dev *vxlan = netdev_priv(dev); 1251 /* struct net *net = dev_net(vxlan->dev); */ 1252 union vxlan_addr ip; 1253 __be16 port; 1254 __be32 src_vni, vni; 1255 u32 ifindex, nhid; 1256 u32 hash_index; 1257 int err; 1258 1259 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1260 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1261 ndm->ndm_state); 1262 return -EINVAL; 1263 } 1264 1265 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1266 return -EINVAL; 1267 1268 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1269 &nhid, extack); 1270 if (err) 1271 return err; 1272 1273 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1274 return -EAFNOSUPPORT; 1275 1276 hash_index = fdb_head_index(vxlan, addr, src_vni); 1277 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1278 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1279 port, src_vni, vni, ifindex, 1280 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1281 nhid, true, extack); 1282 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1283 1284 return err; 1285 } 1286 1287 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1288 const unsigned char *addr, union vxlan_addr ip, 1289 __be16 port, __be32 src_vni, __be32 vni, 1290 u32 ifindex, bool swdev_notify) 1291 { 1292 struct vxlan_rdst *rd = NULL; 1293 struct vxlan_fdb *f; 1294 int err = -ENOENT; 1295 1296 f = vxlan_find_mac(vxlan, addr, src_vni); 1297 if (!f) 1298 return err; 1299 1300 if (!vxlan_addr_any(&ip)) { 1301 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1302 if (!rd) 1303 goto out; 1304 } 1305 1306 /* remove a destination if it's not the only one on the list, 1307 * otherwise destroy the fdb entry 1308 */ 1309 if (rd && !list_is_singular(&f->remotes)) { 1310 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1311 goto out; 1312 } 1313 1314 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1315 1316 out: 1317 return 0; 1318 } 1319 1320 /* Delete entry (via netlink) */ 1321 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1322 struct net_device *dev, 1323 const unsigned char *addr, u16 vid, 1324 struct netlink_ext_ack *extack) 1325 { 1326 struct vxlan_dev *vxlan = netdev_priv(dev); 1327 union vxlan_addr ip; 1328 __be32 src_vni, vni; 1329 u32 ifindex, nhid; 1330 u32 hash_index; 1331 __be16 port; 1332 int err; 1333 1334 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1335 &nhid, extack); 1336 if (err) 1337 return err; 1338 1339 hash_index = fdb_head_index(vxlan, addr, src_vni); 1340 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1341 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1342 true); 1343 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1344 1345 return err; 1346 } 1347 1348 /* Dump forwarding table */ 1349 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1350 struct net_device *dev, 1351 struct net_device *filter_dev, int *idx) 1352 { 1353 struct vxlan_dev *vxlan = netdev_priv(dev); 1354 unsigned int h; 1355 int err = 0; 1356 1357 for (h = 0; h < FDB_HASH_SIZE; ++h) { 1358 struct vxlan_fdb *f; 1359 1360 rcu_read_lock(); 1361 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { 1362 struct vxlan_rdst *rd; 1363 1364 if (rcu_access_pointer(f->nh)) { 1365 if (*idx < cb->args[2]) 1366 goto skip_nh; 1367 err = vxlan_fdb_info(skb, vxlan, f, 1368 NETLINK_CB(cb->skb).portid, 1369 cb->nlh->nlmsg_seq, 1370 RTM_NEWNEIGH, 1371 NLM_F_MULTI, NULL); 1372 if (err < 0) { 1373 rcu_read_unlock(); 1374 goto out; 1375 } 1376 skip_nh: 1377 *idx += 1; 1378 continue; 1379 } 1380 1381 list_for_each_entry_rcu(rd, &f->remotes, list) { 1382 if (*idx < cb->args[2]) 1383 goto skip; 1384 1385 err = vxlan_fdb_info(skb, vxlan, f, 1386 NETLINK_CB(cb->skb).portid, 1387 cb->nlh->nlmsg_seq, 1388 RTM_NEWNEIGH, 1389 NLM_F_MULTI, rd); 1390 if (err < 0) { 1391 rcu_read_unlock(); 1392 goto out; 1393 } 1394 skip: 1395 *idx += 1; 1396 } 1397 } 1398 rcu_read_unlock(); 1399 } 1400 out: 1401 return err; 1402 } 1403 1404 static int vxlan_fdb_get(struct sk_buff *skb, 1405 struct nlattr *tb[], 1406 struct net_device *dev, 1407 const unsigned char *addr, 1408 u16 vid, u32 portid, u32 seq, 1409 struct netlink_ext_ack *extack) 1410 { 1411 struct vxlan_dev *vxlan = netdev_priv(dev); 1412 struct vxlan_fdb *f; 1413 __be32 vni; 1414 int err; 1415 1416 if (tb[NDA_VNI]) 1417 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1418 else 1419 vni = vxlan->default_dst.remote_vni; 1420 1421 rcu_read_lock(); 1422 1423 f = __vxlan_find_mac(vxlan, addr, vni); 1424 if (!f) { 1425 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1426 err = -ENOENT; 1427 goto errout; 1428 } 1429 1430 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1431 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1432 errout: 1433 rcu_read_unlock(); 1434 return err; 1435 } 1436 1437 /* Watch incoming packets to learn mapping between Ethernet address 1438 * and Tunnel endpoint. 1439 * Return true if packet is bogus and should be dropped. 1440 */ 1441 static bool vxlan_snoop(struct net_device *dev, 1442 union vxlan_addr *src_ip, const u8 *src_mac, 1443 u32 src_ifindex, __be32 vni) 1444 { 1445 struct vxlan_dev *vxlan = netdev_priv(dev); 1446 struct vxlan_fdb *f; 1447 u32 ifindex = 0; 1448 1449 #if IS_ENABLED(CONFIG_IPV6) 1450 if (src_ip->sa.sa_family == AF_INET6 && 1451 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1452 ifindex = src_ifindex; 1453 #endif 1454 1455 f = vxlan_find_mac(vxlan, src_mac, vni); 1456 if (likely(f)) { 1457 struct vxlan_rdst *rdst = first_remote_rcu(f); 1458 1459 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1460 rdst->remote_ifindex == ifindex)) 1461 return false; 1462 1463 /* Don't migrate static entries, drop packets */ 1464 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1465 return true; 1466 1467 /* Don't override an fdb with nexthop with a learnt entry */ 1468 if (rcu_access_pointer(f->nh)) 1469 return true; 1470 1471 if (net_ratelimit()) 1472 netdev_info(dev, 1473 "%pM migrated from %pIS to %pIS\n", 1474 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1475 1476 rdst->remote_ip = *src_ip; 1477 f->updated = jiffies; 1478 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1479 } else { 1480 u32 hash_index = fdb_head_index(vxlan, src_mac, vni); 1481 1482 /* learned new entry */ 1483 spin_lock(&vxlan->hash_lock[hash_index]); 1484 1485 /* close off race between vxlan_flush and incoming packets */ 1486 if (netif_running(dev)) 1487 vxlan_fdb_update(vxlan, src_mac, src_ip, 1488 NUD_REACHABLE, 1489 NLM_F_EXCL|NLM_F_CREATE, 1490 vxlan->cfg.dst_port, 1491 vni, 1492 vxlan->default_dst.remote_vni, 1493 ifindex, NTF_SELF, 0, true, NULL); 1494 spin_unlock(&vxlan->hash_lock[hash_index]); 1495 } 1496 1497 return false; 1498 } 1499 1500 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1501 { 1502 struct vxlan_net *vn; 1503 1504 if (!vs) 1505 return false; 1506 if (!refcount_dec_and_test(&vs->refcnt)) 1507 return false; 1508 1509 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); 1510 spin_lock(&vn->sock_lock); 1511 hlist_del_rcu(&vs->hlist); 1512 udp_tunnel_notify_del_rx_port(vs->sock, 1513 (vs->flags & VXLAN_F_GPE) ? 1514 UDP_TUNNEL_TYPE_VXLAN_GPE : 1515 UDP_TUNNEL_TYPE_VXLAN); 1516 spin_unlock(&vn->sock_lock); 1517 1518 return true; 1519 } 1520 1521 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1522 { 1523 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1524 #if IS_ENABLED(CONFIG_IPV6) 1525 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1526 1527 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1528 #endif 1529 1530 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1531 synchronize_net(); 1532 1533 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1534 vxlan_vs_del_vnigrp(vxlan); 1535 else 1536 vxlan_vs_del_dev(vxlan); 1537 1538 if (__vxlan_sock_release_prep(sock4)) { 1539 udp_tunnel_sock_release(sock4->sock); 1540 kfree(sock4); 1541 } 1542 1543 #if IS_ENABLED(CONFIG_IPV6) 1544 if (__vxlan_sock_release_prep(sock6)) { 1545 udp_tunnel_sock_release(sock6->sock); 1546 kfree(sock6); 1547 } 1548 #endif 1549 } 1550 1551 static bool vxlan_remcsum(struct vxlanhdr *unparsed, 1552 struct sk_buff *skb, u32 vxflags) 1553 { 1554 size_t start, offset; 1555 1556 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1557 goto out; 1558 1559 start = vxlan_rco_start(unparsed->vx_vni); 1560 offset = start + vxlan_rco_offset(unparsed->vx_vni); 1561 1562 if (!pskb_may_pull(skb, offset + sizeof(u16))) 1563 return false; 1564 1565 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1566 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1567 out: 1568 unparsed->vx_flags &= ~VXLAN_HF_RCO; 1569 unparsed->vx_vni &= VXLAN_VNI_MASK; 1570 return true; 1571 } 1572 1573 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed, 1574 struct sk_buff *skb, u32 vxflags, 1575 struct vxlan_metadata *md) 1576 { 1577 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed; 1578 struct metadata_dst *tun_dst; 1579 1580 if (!(unparsed->vx_flags & VXLAN_HF_GBP)) 1581 goto out; 1582 1583 md->gbp = ntohs(gbp->policy_id); 1584 1585 tun_dst = (struct metadata_dst *)skb_dst(skb); 1586 if (tun_dst) { 1587 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT; 1588 tun_dst->u.tun_info.options_len = sizeof(*md); 1589 } 1590 if (gbp->dont_learn) 1591 md->gbp |= VXLAN_GBP_DONT_LEARN; 1592 1593 if (gbp->policy_applied) 1594 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1595 1596 /* In flow-based mode, GBP is carried in dst_metadata */ 1597 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1598 skb->mark = md->gbp; 1599 out: 1600 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS; 1601 } 1602 1603 static bool vxlan_set_mac(struct vxlan_dev *vxlan, 1604 struct vxlan_sock *vs, 1605 struct sk_buff *skb, __be32 vni) 1606 { 1607 union vxlan_addr saddr; 1608 u32 ifindex = skb->dev->ifindex; 1609 1610 skb_reset_mac_header(skb); 1611 skb->protocol = eth_type_trans(skb, vxlan->dev); 1612 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1613 1614 /* Ignore packet loops (and multicast echo) */ 1615 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1616 return false; 1617 1618 /* Ignore packets from invalid src-address */ 1619 if (!is_valid_ether_addr(eth_hdr(skb)->h_source)) 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 1678 /* Need UDP and VXLAN header to be present */ 1679 if (!pskb_may_pull(skb, VXLAN_HLEN)) 1680 goto drop; 1681 1682 unparsed = *vxlan_hdr(skb); 1683 /* VNI flag always required to be set */ 1684 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1685 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1686 ntohl(vxlan_hdr(skb)->vx_flags), 1687 ntohl(vxlan_hdr(skb)->vx_vni)); 1688 /* Return non vxlan pkt */ 1689 goto drop; 1690 } 1691 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1692 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1693 1694 vs = rcu_dereference_sk_user_data(sk); 1695 if (!vs) 1696 goto drop; 1697 1698 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1699 1700 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1701 if (!vxlan) 1702 goto drop; 1703 1704 /* For backwards compatibility, only allow reserved fields to be 1705 * used by VXLAN extensions if explicitly requested. 1706 */ 1707 if (vs->flags & VXLAN_F_GPE) { 1708 if (!vxlan_parse_gpe_proto(&unparsed, &protocol)) 1709 goto drop; 1710 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS; 1711 raw_proto = true; 1712 } 1713 1714 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1715 !net_eq(vxlan->net, dev_net(vxlan->dev)))) 1716 goto drop; 1717 1718 if (vs->flags & VXLAN_F_REMCSUM_RX) 1719 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1720 goto drop; 1721 1722 if (vxlan_collect_metadata(vs)) { 1723 struct metadata_dst *tun_dst; 1724 1725 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY, 1726 key32_to_tunnel_id(vni), sizeof(*md)); 1727 1728 if (!tun_dst) 1729 goto drop; 1730 1731 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1732 1733 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1734 } else { 1735 memset(md, 0, sizeof(*md)); 1736 } 1737 1738 if (vs->flags & VXLAN_F_GBP) 1739 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1740 /* Note that GBP and GPE can never be active together. This is 1741 * ensured in vxlan_dev_configure. 1742 */ 1743 1744 if (unparsed.vx_flags || unparsed.vx_vni) { 1745 /* If there are any unprocessed flags remaining treat 1746 * this as a malformed packet. This behavior diverges from 1747 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1748 * in reserved fields are to be ignored. The approach here 1749 * maintains compatibility with previous stack code, and also 1750 * is more robust and provides a little more security in 1751 * adding extensions to VXLAN. 1752 */ 1753 goto drop; 1754 } 1755 1756 if (!raw_proto) { 1757 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1758 goto drop; 1759 } else { 1760 skb_reset_mac_header(skb); 1761 skb->dev = vxlan->dev; 1762 skb->pkt_type = PACKET_HOST; 1763 } 1764 1765 oiph = skb_network_header(skb); 1766 skb_reset_network_header(skb); 1767 1768 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1769 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1770 DEV_STATS_INC(vxlan->dev, rx_errors); 1771 vxlan_vnifilter_count(vxlan, vni, vninode, 1772 VXLAN_VNI_STATS_RX_ERRORS, 0); 1773 goto drop; 1774 } 1775 1776 rcu_read_lock(); 1777 1778 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1779 rcu_read_unlock(); 1780 dev_core_stats_rx_dropped_inc(vxlan->dev); 1781 vxlan_vnifilter_count(vxlan, vni, vninode, 1782 VXLAN_VNI_STATS_RX_DROPS, 0); 1783 goto drop; 1784 } 1785 1786 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1787 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1788 gro_cells_receive(&vxlan->gro_cells, skb); 1789 1790 rcu_read_unlock(); 1791 1792 return 0; 1793 1794 drop: 1795 /* Consume bad packet */ 1796 kfree_skb(skb); 1797 return 0; 1798 } 1799 1800 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1801 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1802 { 1803 struct vxlan_dev *vxlan; 1804 struct vxlan_sock *vs; 1805 struct vxlanhdr *hdr; 1806 __be32 vni; 1807 1808 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1809 return -EINVAL; 1810 1811 hdr = vxlan_hdr(skb); 1812 1813 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1814 return -EINVAL; 1815 1816 vs = rcu_dereference_sk_user_data(sk); 1817 if (!vs) 1818 return -ENOENT; 1819 1820 vni = vxlan_vni(hdr->vx_vni); 1821 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1822 if (!vxlan) 1823 return -ENOENT; 1824 1825 return 0; 1826 } 1827 1828 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1829 { 1830 struct vxlan_dev *vxlan = netdev_priv(dev); 1831 struct arphdr *parp; 1832 u8 *arpptr, *sha; 1833 __be32 sip, tip; 1834 struct neighbour *n; 1835 1836 if (dev->flags & IFF_NOARP) 1837 goto out; 1838 1839 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1840 dev_core_stats_tx_dropped_inc(dev); 1841 vxlan_vnifilter_count(vxlan, vni, NULL, 1842 VXLAN_VNI_STATS_TX_DROPS, 0); 1843 goto out; 1844 } 1845 parp = arp_hdr(skb); 1846 1847 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1848 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1849 parp->ar_pro != htons(ETH_P_IP) || 1850 parp->ar_op != htons(ARPOP_REQUEST) || 1851 parp->ar_hln != dev->addr_len || 1852 parp->ar_pln != 4) 1853 goto out; 1854 arpptr = (u8 *)parp + sizeof(struct arphdr); 1855 sha = arpptr; 1856 arpptr += dev->addr_len; /* sha */ 1857 memcpy(&sip, arpptr, sizeof(sip)); 1858 arpptr += sizeof(sip); 1859 arpptr += dev->addr_len; /* tha */ 1860 memcpy(&tip, arpptr, sizeof(tip)); 1861 1862 if (ipv4_is_loopback(tip) || 1863 ipv4_is_multicast(tip)) 1864 goto out; 1865 1866 n = neigh_lookup(&arp_tbl, &tip, dev); 1867 1868 if (n) { 1869 struct vxlan_fdb *f; 1870 struct sk_buff *reply; 1871 1872 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1873 neigh_release(n); 1874 goto out; 1875 } 1876 1877 f = vxlan_find_mac(vxlan, n->ha, vni); 1878 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1879 /* bridge-local neighbor */ 1880 neigh_release(n); 1881 goto out; 1882 } 1883 1884 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1885 n->ha, sha); 1886 1887 neigh_release(n); 1888 1889 if (reply == NULL) 1890 goto out; 1891 1892 skb_reset_mac_header(reply); 1893 __skb_pull(reply, skb_network_offset(reply)); 1894 reply->ip_summed = CHECKSUM_UNNECESSARY; 1895 reply->pkt_type = PACKET_HOST; 1896 1897 if (netif_rx(reply) == NET_RX_DROP) { 1898 dev_core_stats_rx_dropped_inc(dev); 1899 vxlan_vnifilter_count(vxlan, vni, NULL, 1900 VXLAN_VNI_STATS_RX_DROPS, 0); 1901 } 1902 1903 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1904 union vxlan_addr ipa = { 1905 .sin.sin_addr.s_addr = tip, 1906 .sin.sin_family = AF_INET, 1907 }; 1908 1909 vxlan_ip_miss(dev, &ipa); 1910 } 1911 out: 1912 consume_skb(skb); 1913 return NETDEV_TX_OK; 1914 } 1915 1916 #if IS_ENABLED(CONFIG_IPV6) 1917 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1918 struct neighbour *n, bool isrouter) 1919 { 1920 struct net_device *dev = request->dev; 1921 struct sk_buff *reply; 1922 struct nd_msg *ns, *na; 1923 struct ipv6hdr *pip6; 1924 u8 *daddr; 1925 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1926 int ns_olen; 1927 int i, len; 1928 1929 if (dev == NULL || !pskb_may_pull(request, request->len)) 1930 return NULL; 1931 1932 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1933 sizeof(*na) + na_olen + dev->needed_tailroom; 1934 reply = alloc_skb(len, GFP_ATOMIC); 1935 if (reply == NULL) 1936 return NULL; 1937 1938 reply->protocol = htons(ETH_P_IPV6); 1939 reply->dev = dev; 1940 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1941 skb_push(reply, sizeof(struct ethhdr)); 1942 skb_reset_mac_header(reply); 1943 1944 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1945 1946 daddr = eth_hdr(request)->h_source; 1947 ns_olen = request->len - skb_network_offset(request) - 1948 sizeof(struct ipv6hdr) - sizeof(*ns); 1949 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1950 if (!ns->opt[i + 1]) { 1951 kfree_skb(reply); 1952 return NULL; 1953 } 1954 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1955 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1956 break; 1957 } 1958 } 1959 1960 /* Ethernet header */ 1961 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1962 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1963 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1964 reply->protocol = htons(ETH_P_IPV6); 1965 1966 skb_pull(reply, sizeof(struct ethhdr)); 1967 skb_reset_network_header(reply); 1968 skb_put(reply, sizeof(struct ipv6hdr)); 1969 1970 /* IPv6 header */ 1971 1972 pip6 = ipv6_hdr(reply); 1973 memset(pip6, 0, sizeof(struct ipv6hdr)); 1974 pip6->version = 6; 1975 pip6->priority = ipv6_hdr(request)->priority; 1976 pip6->nexthdr = IPPROTO_ICMPV6; 1977 pip6->hop_limit = 255; 1978 pip6->daddr = ipv6_hdr(request)->saddr; 1979 pip6->saddr = *(struct in6_addr *)n->primary_key; 1980 1981 skb_pull(reply, sizeof(struct ipv6hdr)); 1982 skb_reset_transport_header(reply); 1983 1984 /* Neighbor Advertisement */ 1985 na = skb_put_zero(reply, sizeof(*na) + na_olen); 1986 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 1987 na->icmph.icmp6_router = isrouter; 1988 na->icmph.icmp6_override = 1; 1989 na->icmph.icmp6_solicited = 1; 1990 na->target = ns->target; 1991 ether_addr_copy(&na->opt[2], n->ha); 1992 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 1993 na->opt[1] = na_olen >> 3; 1994 1995 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 1996 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 1997 csum_partial(na, sizeof(*na)+na_olen, 0)); 1998 1999 pip6->payload_len = htons(sizeof(*na)+na_olen); 2000 2001 skb_push(reply, sizeof(struct ipv6hdr)); 2002 2003 reply->ip_summed = CHECKSUM_UNNECESSARY; 2004 2005 return reply; 2006 } 2007 2008 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2009 { 2010 struct vxlan_dev *vxlan = netdev_priv(dev); 2011 const struct in6_addr *daddr; 2012 const struct ipv6hdr *iphdr; 2013 struct inet6_dev *in6_dev; 2014 struct neighbour *n; 2015 struct nd_msg *msg; 2016 2017 rcu_read_lock(); 2018 in6_dev = __in6_dev_get(dev); 2019 if (!in6_dev) 2020 goto out; 2021 2022 iphdr = ipv6_hdr(skb); 2023 daddr = &iphdr->daddr; 2024 msg = (struct nd_msg *)(iphdr + 1); 2025 2026 if (ipv6_addr_loopback(daddr) || 2027 ipv6_addr_is_multicast(&msg->target)) 2028 goto out; 2029 2030 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2031 2032 if (n) { 2033 struct vxlan_fdb *f; 2034 struct sk_buff *reply; 2035 2036 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2037 neigh_release(n); 2038 goto out; 2039 } 2040 2041 f = vxlan_find_mac(vxlan, n->ha, vni); 2042 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2043 /* bridge-local neighbor */ 2044 neigh_release(n); 2045 goto out; 2046 } 2047 2048 reply = vxlan_na_create(skb, n, 2049 !!(f ? f->flags & NTF_ROUTER : 0)); 2050 2051 neigh_release(n); 2052 2053 if (reply == NULL) 2054 goto out; 2055 2056 if (netif_rx(reply) == NET_RX_DROP) { 2057 dev_core_stats_rx_dropped_inc(dev); 2058 vxlan_vnifilter_count(vxlan, vni, NULL, 2059 VXLAN_VNI_STATS_RX_DROPS, 0); 2060 } 2061 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2062 union vxlan_addr ipa = { 2063 .sin6.sin6_addr = msg->target, 2064 .sin6.sin6_family = AF_INET6, 2065 }; 2066 2067 vxlan_ip_miss(dev, &ipa); 2068 } 2069 2070 out: 2071 rcu_read_unlock(); 2072 consume_skb(skb); 2073 return NETDEV_TX_OK; 2074 } 2075 #endif 2076 2077 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2078 { 2079 struct vxlan_dev *vxlan = netdev_priv(dev); 2080 struct neighbour *n; 2081 2082 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2083 return false; 2084 2085 n = NULL; 2086 switch (ntohs(eth_hdr(skb)->h_proto)) { 2087 case ETH_P_IP: 2088 { 2089 struct iphdr *pip; 2090 2091 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2092 return false; 2093 pip = ip_hdr(skb); 2094 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2095 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2096 union vxlan_addr ipa = { 2097 .sin.sin_addr.s_addr = pip->daddr, 2098 .sin.sin_family = AF_INET, 2099 }; 2100 2101 vxlan_ip_miss(dev, &ipa); 2102 return false; 2103 } 2104 2105 break; 2106 } 2107 #if IS_ENABLED(CONFIG_IPV6) 2108 case ETH_P_IPV6: 2109 { 2110 struct ipv6hdr *pip6; 2111 2112 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2113 return false; 2114 pip6 = ipv6_hdr(skb); 2115 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2116 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2117 union vxlan_addr ipa = { 2118 .sin6.sin6_addr = pip6->daddr, 2119 .sin6.sin6_family = AF_INET6, 2120 }; 2121 2122 vxlan_ip_miss(dev, &ipa); 2123 return false; 2124 } 2125 2126 break; 2127 } 2128 #endif 2129 default: 2130 return false; 2131 } 2132 2133 if (n) { 2134 bool diff; 2135 2136 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2137 if (diff) { 2138 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2139 dev->addr_len); 2140 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2141 } 2142 neigh_release(n); 2143 return diff; 2144 } 2145 2146 return false; 2147 } 2148 2149 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2150 { 2151 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2152 2153 gpe->np_applied = 1; 2154 gpe->next_protocol = tun_p_from_eth_p(protocol); 2155 if (!gpe->next_protocol) 2156 return -EPFNOSUPPORT; 2157 return 0; 2158 } 2159 2160 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2161 int iphdr_len, __be32 vni, 2162 struct vxlan_metadata *md, u32 vxflags, 2163 bool udp_sum) 2164 { 2165 struct vxlanhdr *vxh; 2166 int min_headroom; 2167 int err; 2168 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2169 __be16 inner_protocol = htons(ETH_P_TEB); 2170 2171 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2172 skb->ip_summed == CHECKSUM_PARTIAL) { 2173 int csum_start = skb_checksum_start_offset(skb); 2174 2175 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2176 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2177 (skb->csum_offset == offsetof(struct udphdr, check) || 2178 skb->csum_offset == offsetof(struct tcphdr, check))) 2179 type |= SKB_GSO_TUNNEL_REMCSUM; 2180 } 2181 2182 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2183 + VXLAN_HLEN + iphdr_len; 2184 2185 /* Need space for new headers (invalidates iph ptr) */ 2186 err = skb_cow_head(skb, min_headroom); 2187 if (unlikely(err)) 2188 return err; 2189 2190 err = iptunnel_handle_offloads(skb, type); 2191 if (err) 2192 return err; 2193 2194 vxh = __skb_push(skb, sizeof(*vxh)); 2195 vxh->vx_flags = VXLAN_HF_VNI; 2196 vxh->vx_vni = vxlan_vni_field(vni); 2197 2198 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2199 unsigned int start; 2200 2201 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2202 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2203 vxh->vx_flags |= VXLAN_HF_RCO; 2204 2205 if (!skb_is_gso(skb)) { 2206 skb->ip_summed = CHECKSUM_NONE; 2207 skb->encapsulation = 0; 2208 } 2209 } 2210 2211 if (vxflags & VXLAN_F_GBP) 2212 vxlan_build_gbp_hdr(vxh, md); 2213 if (vxflags & VXLAN_F_GPE) { 2214 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2215 if (err < 0) 2216 return err; 2217 inner_protocol = skb->protocol; 2218 } 2219 2220 skb_set_inner_protocol(skb, inner_protocol); 2221 return 0; 2222 } 2223 2224 /* Bypass encapsulation if the destination is local */ 2225 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2226 struct vxlan_dev *dst_vxlan, __be32 vni, 2227 bool snoop) 2228 { 2229 union vxlan_addr loopback; 2230 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2231 struct net_device *dev; 2232 int len = skb->len; 2233 2234 skb->pkt_type = PACKET_HOST; 2235 skb->encapsulation = 0; 2236 skb->dev = dst_vxlan->dev; 2237 __skb_pull(skb, skb_network_offset(skb)); 2238 2239 if (remote_ip->sa.sa_family == AF_INET) { 2240 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2241 loopback.sa.sa_family = AF_INET; 2242 #if IS_ENABLED(CONFIG_IPV6) 2243 } else { 2244 loopback.sin6.sin6_addr = in6addr_loopback; 2245 loopback.sa.sa_family = AF_INET6; 2246 #endif 2247 } 2248 2249 rcu_read_lock(); 2250 dev = skb->dev; 2251 if (unlikely(!(dev->flags & IFF_UP))) { 2252 kfree_skb(skb); 2253 goto drop; 2254 } 2255 2256 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2257 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2258 2259 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2260 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2261 2262 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2263 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2264 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2265 len); 2266 } else { 2267 drop: 2268 dev_core_stats_rx_dropped_inc(dev); 2269 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2270 VXLAN_VNI_STATS_RX_DROPS, 0); 2271 } 2272 rcu_read_unlock(); 2273 } 2274 2275 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2276 struct vxlan_dev *vxlan, 2277 int addr_family, 2278 __be16 dst_port, int dst_ifindex, __be32 vni, 2279 struct dst_entry *dst, 2280 u32 rt_flags) 2281 { 2282 #if IS_ENABLED(CONFIG_IPV6) 2283 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2284 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2285 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2286 */ 2287 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2288 #endif 2289 /* Bypass encapsulation if the destination is local */ 2290 if (rt_flags & RTCF_LOCAL && 2291 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2292 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2293 struct vxlan_dev *dst_vxlan; 2294 2295 dst_release(dst); 2296 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2297 addr_family, dst_port, 2298 vxlan->cfg.flags); 2299 if (!dst_vxlan) { 2300 DEV_STATS_INC(dev, tx_errors); 2301 vxlan_vnifilter_count(vxlan, vni, NULL, 2302 VXLAN_VNI_STATS_TX_ERRORS, 0); 2303 kfree_skb(skb); 2304 2305 return -ENOENT; 2306 } 2307 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2308 return 1; 2309 } 2310 2311 return 0; 2312 } 2313 2314 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2315 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2316 { 2317 struct dst_cache *dst_cache; 2318 struct ip_tunnel_info *info; 2319 struct ip_tunnel_key *pkey; 2320 struct ip_tunnel_key key; 2321 struct vxlan_dev *vxlan = netdev_priv(dev); 2322 const struct iphdr *old_iph = ip_hdr(skb); 2323 struct vxlan_metadata _md; 2324 struct vxlan_metadata *md = &_md; 2325 unsigned int pkt_len = skb->len; 2326 __be16 src_port = 0, dst_port; 2327 struct dst_entry *ndst = NULL; 2328 int addr_family; 2329 __u8 tos, ttl; 2330 int ifindex; 2331 int err; 2332 u32 flags = vxlan->cfg.flags; 2333 bool use_cache; 2334 bool udp_sum = false; 2335 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2336 __be32 vni = 0; 2337 2338 info = skb_tunnel_info(skb); 2339 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2340 2341 if (rdst) { 2342 memset(&key, 0, sizeof(key)); 2343 pkey = &key; 2344 2345 if (vxlan_addr_any(&rdst->remote_ip)) { 2346 if (did_rsc) { 2347 /* short-circuited back to local bridge */ 2348 vxlan_encap_bypass(skb, vxlan, vxlan, 2349 default_vni, true); 2350 return; 2351 } 2352 goto drop; 2353 } 2354 2355 addr_family = vxlan->cfg.saddr.sa.sa_family; 2356 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2357 vni = (rdst->remote_vni) ? : default_vni; 2358 ifindex = rdst->remote_ifindex; 2359 2360 if (addr_family == AF_INET) { 2361 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2362 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2363 } else { 2364 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2365 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2366 } 2367 2368 dst_cache = &rdst->dst_cache; 2369 md->gbp = skb->mark; 2370 if (flags & VXLAN_F_TTL_INHERIT) { 2371 ttl = ip_tunnel_get_ttl(old_iph, skb); 2372 } else { 2373 ttl = vxlan->cfg.ttl; 2374 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2375 ttl = 1; 2376 } 2377 tos = vxlan->cfg.tos; 2378 if (tos == 1) 2379 tos = ip_tunnel_get_dsfield(old_iph, skb); 2380 if (tos && !info) 2381 use_cache = false; 2382 2383 if (addr_family == AF_INET) 2384 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2385 else 2386 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2387 #if IS_ENABLED(CONFIG_IPV6) 2388 switch (vxlan->cfg.label_policy) { 2389 case VXLAN_LABEL_FIXED: 2390 key.label = vxlan->cfg.label; 2391 break; 2392 case VXLAN_LABEL_INHERIT: 2393 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2394 break; 2395 default: 2396 DEBUG_NET_WARN_ON_ONCE(1); 2397 goto drop; 2398 } 2399 #endif 2400 } else { 2401 if (!info) { 2402 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2403 dev->name); 2404 goto drop; 2405 } 2406 pkey = &info->key; 2407 addr_family = ip_tunnel_info_af(info); 2408 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2409 vni = tunnel_id_to_key32(info->key.tun_id); 2410 ifindex = 0; 2411 dst_cache = &info->dst_cache; 2412 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) { 2413 if (info->options_len < sizeof(*md)) 2414 goto drop; 2415 md = ip_tunnel_info_opts(info); 2416 } 2417 ttl = info->key.ttl; 2418 tos = info->key.tos; 2419 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 2420 } 2421 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2422 vxlan->cfg.port_max, true); 2423 2424 rcu_read_lock(); 2425 if (addr_family == AF_INET) { 2426 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2427 struct rtable *rt; 2428 __be16 df = 0; 2429 __be32 saddr; 2430 2431 if (!ifindex) 2432 ifindex = sock4->sock->sk->sk_bound_dev_if; 2433 2434 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2435 &saddr, pkey, src_port, dst_port, 2436 tos, use_cache ? dst_cache : NULL); 2437 if (IS_ERR(rt)) { 2438 err = PTR_ERR(rt); 2439 goto tx_error; 2440 } 2441 2442 if (!info) { 2443 /* Bypass encapsulation if the destination is local */ 2444 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2445 dst_port, ifindex, vni, 2446 &rt->dst, rt->rt_flags); 2447 if (err) 2448 goto out_unlock; 2449 2450 if (vxlan->cfg.df == VXLAN_DF_SET) { 2451 df = htons(IP_DF); 2452 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2453 struct ethhdr *eth = eth_hdr(skb); 2454 2455 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2456 (ntohs(eth->h_proto) == ETH_P_IP && 2457 old_iph->frag_off & htons(IP_DF))) 2458 df = htons(IP_DF); 2459 } 2460 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) { 2461 df = htons(IP_DF); 2462 } 2463 2464 ndst = &rt->dst; 2465 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2466 netif_is_any_bridge_port(dev)); 2467 if (err < 0) { 2468 goto tx_error; 2469 } else if (err) { 2470 if (info) { 2471 struct ip_tunnel_info *unclone; 2472 2473 unclone = skb_tunnel_info_unclone(skb); 2474 if (unlikely(!unclone)) 2475 goto tx_error; 2476 2477 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2478 unclone->key.u.ipv4.dst = saddr; 2479 } 2480 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2481 dst_release(ndst); 2482 goto out_unlock; 2483 } 2484 2485 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2486 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2487 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2488 vni, md, flags, udp_sum); 2489 if (err < 0) 2490 goto tx_error; 2491 2492 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2493 pkey->u.ipv4.dst, tos, ttl, df, 2494 src_port, dst_port, xnet, !udp_sum); 2495 #if IS_ENABLED(CONFIG_IPV6) 2496 } else { 2497 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2498 struct in6_addr saddr; 2499 2500 if (!ifindex) 2501 ifindex = sock6->sock->sk->sk_bound_dev_if; 2502 2503 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2504 ifindex, &saddr, pkey, 2505 src_port, dst_port, tos, 2506 use_cache ? dst_cache : NULL); 2507 if (IS_ERR(ndst)) { 2508 err = PTR_ERR(ndst); 2509 ndst = NULL; 2510 goto tx_error; 2511 } 2512 2513 if (!info) { 2514 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; 2515 2516 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2517 dst_port, ifindex, vni, 2518 ndst, rt6i_flags); 2519 if (err) 2520 goto out_unlock; 2521 } 2522 2523 err = skb_tunnel_check_pmtu(skb, ndst, 2524 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2525 netif_is_any_bridge_port(dev)); 2526 if (err < 0) { 2527 goto tx_error; 2528 } else if (err) { 2529 if (info) { 2530 struct ip_tunnel_info *unclone; 2531 2532 unclone = skb_tunnel_info_unclone(skb); 2533 if (unlikely(!unclone)) 2534 goto tx_error; 2535 2536 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2537 unclone->key.u.ipv6.dst = saddr; 2538 } 2539 2540 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2541 dst_release(ndst); 2542 goto out_unlock; 2543 } 2544 2545 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2546 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2547 skb_scrub_packet(skb, xnet); 2548 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2549 vni, md, flags, udp_sum); 2550 if (err < 0) 2551 goto tx_error; 2552 2553 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2554 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2555 pkey->label, src_port, dst_port, !udp_sum); 2556 #endif 2557 } 2558 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2559 out_unlock: 2560 rcu_read_unlock(); 2561 return; 2562 2563 drop: 2564 dev_core_stats_tx_dropped_inc(dev); 2565 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2566 dev_kfree_skb(skb); 2567 return; 2568 2569 tx_error: 2570 rcu_read_unlock(); 2571 if (err == -ELOOP) 2572 DEV_STATS_INC(dev, collisions); 2573 else if (err == -ENETUNREACH) 2574 DEV_STATS_INC(dev, tx_carrier_errors); 2575 dst_release(ndst); 2576 DEV_STATS_INC(dev, tx_errors); 2577 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2578 kfree_skb(skb); 2579 } 2580 2581 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2582 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2583 { 2584 struct vxlan_rdst nh_rdst; 2585 struct nexthop *nh; 2586 bool do_xmit; 2587 u32 hash; 2588 2589 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2590 hash = skb_get_hash(skb); 2591 2592 rcu_read_lock(); 2593 nh = rcu_dereference(f->nh); 2594 if (!nh) { 2595 rcu_read_unlock(); 2596 goto drop; 2597 } 2598 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2599 rcu_read_unlock(); 2600 2601 if (likely(do_xmit)) 2602 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2603 else 2604 goto drop; 2605 2606 return; 2607 2608 drop: 2609 dev_core_stats_tx_dropped_inc(dev); 2610 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2611 VXLAN_VNI_STATS_TX_DROPS, 0); 2612 dev_kfree_skb(skb); 2613 } 2614 2615 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2616 u32 nhid, __be32 vni) 2617 { 2618 struct vxlan_dev *vxlan = netdev_priv(dev); 2619 struct vxlan_rdst nh_rdst; 2620 struct nexthop *nh; 2621 bool do_xmit; 2622 u32 hash; 2623 2624 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2625 hash = skb_get_hash(skb); 2626 2627 rcu_read_lock(); 2628 nh = nexthop_find_by_id(dev_net(dev), nhid); 2629 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2630 rcu_read_unlock(); 2631 goto drop; 2632 } 2633 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2634 rcu_read_unlock(); 2635 2636 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2637 goto drop; 2638 2639 if (likely(do_xmit)) 2640 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2641 else 2642 goto drop; 2643 2644 return NETDEV_TX_OK; 2645 2646 drop: 2647 dev_core_stats_tx_dropped_inc(dev); 2648 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2649 VXLAN_VNI_STATS_TX_DROPS, 0); 2650 dev_kfree_skb(skb); 2651 return NETDEV_TX_OK; 2652 } 2653 2654 /* Transmit local packets over Vxlan 2655 * 2656 * Outer IP header inherits ECN and DF from inner header. 2657 * Outer UDP destination is the VXLAN assigned port. 2658 * source port is based on hash of flow 2659 */ 2660 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2661 { 2662 struct vxlan_dev *vxlan = netdev_priv(dev); 2663 struct vxlan_rdst *rdst, *fdst = NULL; 2664 const struct ip_tunnel_info *info; 2665 bool did_rsc = false; 2666 struct vxlan_fdb *f; 2667 struct ethhdr *eth; 2668 __be32 vni = 0; 2669 u32 nhid = 0; 2670 2671 info = skb_tunnel_info(skb); 2672 2673 skb_reset_mac_header(skb); 2674 2675 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2676 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2677 info->mode & IP_TUNNEL_INFO_TX) { 2678 vni = tunnel_id_to_key32(info->key.tun_id); 2679 nhid = info->key.nhid; 2680 } else { 2681 if (info && info->mode & IP_TUNNEL_INFO_TX) 2682 vxlan_xmit_one(skb, dev, vni, NULL, false); 2683 else 2684 kfree_skb(skb); 2685 return NETDEV_TX_OK; 2686 } 2687 } 2688 2689 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2690 eth = eth_hdr(skb); 2691 if (ntohs(eth->h_proto) == ETH_P_ARP) 2692 return arp_reduce(dev, skb, vni); 2693 #if IS_ENABLED(CONFIG_IPV6) 2694 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2695 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2696 sizeof(struct nd_msg)) && 2697 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2698 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2699 2700 if (m->icmph.icmp6_code == 0 && 2701 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2702 return neigh_reduce(dev, skb, vni); 2703 } 2704 #endif 2705 } 2706 2707 if (nhid) 2708 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2709 2710 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2711 struct vxlan_mdb_entry *mdb_entry; 2712 2713 rcu_read_lock(); 2714 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2715 if (mdb_entry) { 2716 netdev_tx_t ret; 2717 2718 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2719 rcu_read_unlock(); 2720 return ret; 2721 } 2722 rcu_read_unlock(); 2723 } 2724 2725 eth = eth_hdr(skb); 2726 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2727 did_rsc = false; 2728 2729 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2730 (ntohs(eth->h_proto) == ETH_P_IP || 2731 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2732 did_rsc = route_shortcircuit(dev, skb); 2733 if (did_rsc) 2734 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2735 } 2736 2737 if (f == NULL) { 2738 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2739 if (f == NULL) { 2740 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2741 !is_multicast_ether_addr(eth->h_dest)) 2742 vxlan_fdb_miss(vxlan, eth->h_dest); 2743 2744 dev_core_stats_tx_dropped_inc(dev); 2745 vxlan_vnifilter_count(vxlan, vni, NULL, 2746 VXLAN_VNI_STATS_TX_DROPS, 0); 2747 kfree_skb(skb); 2748 return NETDEV_TX_OK; 2749 } 2750 } 2751 2752 if (rcu_access_pointer(f->nh)) { 2753 vxlan_xmit_nh(skb, dev, f, 2754 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2755 } else { 2756 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2757 struct sk_buff *skb1; 2758 2759 if (!fdst) { 2760 fdst = rdst; 2761 continue; 2762 } 2763 skb1 = skb_clone(skb, GFP_ATOMIC); 2764 if (skb1) 2765 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2766 } 2767 if (fdst) 2768 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2769 else 2770 kfree_skb(skb); 2771 } 2772 2773 return NETDEV_TX_OK; 2774 } 2775 2776 /* Walk the forwarding table and purge stale entries */ 2777 static void vxlan_cleanup(struct timer_list *t) 2778 { 2779 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2780 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2781 unsigned int h; 2782 2783 if (!netif_running(vxlan->dev)) 2784 return; 2785 2786 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2787 struct hlist_node *p, *n; 2788 2789 spin_lock(&vxlan->hash_lock[h]); 2790 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2791 struct vxlan_fdb *f 2792 = container_of(p, struct vxlan_fdb, hlist); 2793 unsigned long timeout; 2794 2795 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2796 continue; 2797 2798 if (f->flags & NTF_EXT_LEARNED) 2799 continue; 2800 2801 timeout = f->used + vxlan->cfg.age_interval * HZ; 2802 if (time_before_eq(timeout, jiffies)) { 2803 netdev_dbg(vxlan->dev, 2804 "garbage collect %pM\n", 2805 f->eth_addr); 2806 f->state = NUD_STALE; 2807 vxlan_fdb_destroy(vxlan, f, true, true); 2808 } else if (time_before(timeout, next_timer)) 2809 next_timer = timeout; 2810 } 2811 spin_unlock(&vxlan->hash_lock[h]); 2812 } 2813 2814 mod_timer(&vxlan->age_timer, next_timer); 2815 } 2816 2817 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2818 { 2819 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2820 2821 spin_lock(&vn->sock_lock); 2822 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2823 #if IS_ENABLED(CONFIG_IPV6) 2824 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2825 #endif 2826 spin_unlock(&vn->sock_lock); 2827 } 2828 2829 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2830 struct vxlan_dev_node *node) 2831 { 2832 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2833 __be32 vni = vxlan->default_dst.remote_vni; 2834 2835 node->vxlan = vxlan; 2836 spin_lock(&vn->sock_lock); 2837 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2838 spin_unlock(&vn->sock_lock); 2839 } 2840 2841 /* Setup stats when device is created */ 2842 static int vxlan_init(struct net_device *dev) 2843 { 2844 struct vxlan_dev *vxlan = netdev_priv(dev); 2845 int err; 2846 2847 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2848 vxlan_vnigroup_init(vxlan); 2849 2850 err = gro_cells_init(&vxlan->gro_cells, dev); 2851 if (err) 2852 goto err_vnigroup_uninit; 2853 2854 err = vxlan_mdb_init(vxlan); 2855 if (err) 2856 goto err_gro_cells_destroy; 2857 2858 netdev_lockdep_set_classes(dev); 2859 return 0; 2860 2861 err_gro_cells_destroy: 2862 gro_cells_destroy(&vxlan->gro_cells); 2863 err_vnigroup_uninit: 2864 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2865 vxlan_vnigroup_uninit(vxlan); 2866 return err; 2867 } 2868 2869 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2870 { 2871 struct vxlan_fdb *f; 2872 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2873 2874 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2875 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2876 if (f) 2877 vxlan_fdb_destroy(vxlan, f, true, true); 2878 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2879 } 2880 2881 static void vxlan_uninit(struct net_device *dev) 2882 { 2883 struct vxlan_dev *vxlan = netdev_priv(dev); 2884 2885 vxlan_mdb_fini(vxlan); 2886 2887 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2888 vxlan_vnigroup_uninit(vxlan); 2889 2890 gro_cells_destroy(&vxlan->gro_cells); 2891 2892 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2893 } 2894 2895 /* Start ageing timer and join group when device is brought up */ 2896 static int vxlan_open(struct net_device *dev) 2897 { 2898 struct vxlan_dev *vxlan = netdev_priv(dev); 2899 int ret; 2900 2901 ret = vxlan_sock_add(vxlan); 2902 if (ret < 0) 2903 return ret; 2904 2905 ret = vxlan_multicast_join(vxlan); 2906 if (ret) { 2907 vxlan_sock_release(vxlan); 2908 return ret; 2909 } 2910 2911 if (vxlan->cfg.age_interval) 2912 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2913 2914 return ret; 2915 } 2916 2917 struct vxlan_fdb_flush_desc { 2918 bool ignore_default_entry; 2919 unsigned long state; 2920 unsigned long state_mask; 2921 unsigned long flags; 2922 unsigned long flags_mask; 2923 __be32 src_vni; 2924 u32 nhid; 2925 __be32 vni; 2926 __be16 port; 2927 union vxlan_addr dst_ip; 2928 }; 2929 2930 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2931 const struct vxlan_dev *vxlan) 2932 { 2933 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2934 } 2935 2936 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2937 { 2938 struct nexthop *nh = rtnl_dereference(f->nh); 2939 2940 return nh && nh->id == nhid; 2941 } 2942 2943 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2944 const struct vxlan_dev *vxlan, 2945 const struct vxlan_fdb_flush_desc *desc) 2946 { 2947 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2948 return false; 2949 2950 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2951 return false; 2952 2953 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2954 return false; 2955 2956 if (desc->src_vni && f->vni != desc->src_vni) 2957 return false; 2958 2959 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 2960 return false; 2961 2962 return true; 2963 } 2964 2965 static bool 2966 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 2967 { 2968 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 2969 } 2970 2971 static bool 2972 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 2973 const struct vxlan_rdst *rd) 2974 { 2975 if (desc->vni && rd->remote_vni != desc->vni) 2976 return false; 2977 2978 if (desc->port && rd->remote_port != desc->port) 2979 return false; 2980 2981 if (desc->dst_ip.sa.sa_family && 2982 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 2983 return false; 2984 2985 return true; 2986 } 2987 2988 static void 2989 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 2990 const struct vxlan_fdb_flush_desc *desc, 2991 bool *p_destroy_fdb) 2992 { 2993 bool remotes_flushed = false; 2994 struct vxlan_rdst *rd, *tmp; 2995 2996 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 2997 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 2998 continue; 2999 3000 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3001 remotes_flushed = true; 3002 } 3003 3004 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3005 } 3006 3007 /* Purge the forwarding table */ 3008 static void vxlan_flush(struct vxlan_dev *vxlan, 3009 const struct vxlan_fdb_flush_desc *desc) 3010 { 3011 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3012 unsigned int h; 3013 3014 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3015 struct hlist_node *p, *n; 3016 3017 spin_lock_bh(&vxlan->hash_lock[h]); 3018 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3019 struct vxlan_fdb *f 3020 = container_of(p, struct vxlan_fdb, hlist); 3021 3022 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3023 continue; 3024 3025 if (match_remotes) { 3026 bool destroy_fdb = false; 3027 3028 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3029 &destroy_fdb); 3030 3031 if (!destroy_fdb) 3032 continue; 3033 } 3034 3035 vxlan_fdb_destroy(vxlan, f, true, true); 3036 } 3037 spin_unlock_bh(&vxlan->hash_lock[h]); 3038 } 3039 } 3040 3041 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3042 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3043 [NDA_NH_ID] = { .type = NLA_U32 }, 3044 [NDA_VNI] = { .type = NLA_U32 }, 3045 [NDA_PORT] = { .type = NLA_U16 }, 3046 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3047 sizeof(struct in6_addr)), 3048 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3049 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3050 }; 3051 3052 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3053 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3054 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3055 NTF_ROUTER) 3056 3057 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3058 struct netlink_ext_ack *extack) 3059 { 3060 struct vxlan_dev *vxlan = netdev_priv(dev); 3061 struct vxlan_fdb_flush_desc desc = {}; 3062 struct ndmsg *ndm = nlmsg_data(nlh); 3063 struct nlattr *tb[NDA_MAX + 1]; 3064 u8 ndm_flags; 3065 int err; 3066 3067 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3068 3069 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3070 extack); 3071 if (err) 3072 return err; 3073 3074 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3075 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3076 return -EINVAL; 3077 } 3078 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3079 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3080 return -EINVAL; 3081 } 3082 3083 desc.state = ndm->ndm_state; 3084 desc.flags = ndm_flags; 3085 3086 if (tb[NDA_NDM_STATE_MASK]) 3087 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3088 3089 if (tb[NDA_NDM_FLAGS_MASK]) 3090 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3091 3092 if (tb[NDA_SRC_VNI]) 3093 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3094 3095 if (tb[NDA_NH_ID]) 3096 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3097 3098 if (tb[NDA_VNI]) 3099 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3100 3101 if (tb[NDA_PORT]) 3102 desc.port = nla_get_be16(tb[NDA_PORT]); 3103 3104 if (tb[NDA_DST]) { 3105 union vxlan_addr ip; 3106 3107 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3108 if (err) { 3109 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3110 "Unsupported address family"); 3111 return err; 3112 } 3113 desc.dst_ip = ip; 3114 } 3115 3116 vxlan_flush(vxlan, &desc); 3117 3118 return 0; 3119 } 3120 3121 /* Cleanup timer and forwarding table on shutdown */ 3122 static int vxlan_stop(struct net_device *dev) 3123 { 3124 struct vxlan_dev *vxlan = netdev_priv(dev); 3125 struct vxlan_fdb_flush_desc desc = { 3126 /* Default entry is deleted at vxlan_uninit. */ 3127 .ignore_default_entry = true, 3128 .state = 0, 3129 .state_mask = NUD_PERMANENT | NUD_NOARP, 3130 }; 3131 3132 vxlan_multicast_leave(vxlan); 3133 3134 del_timer_sync(&vxlan->age_timer); 3135 3136 vxlan_flush(vxlan, &desc); 3137 vxlan_sock_release(vxlan); 3138 3139 return 0; 3140 } 3141 3142 /* Stub, nothing needs to be done. */ 3143 static void vxlan_set_multicast_list(struct net_device *dev) 3144 { 3145 } 3146 3147 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3148 { 3149 struct vxlan_dev *vxlan = netdev_priv(dev); 3150 struct vxlan_rdst *dst = &vxlan->default_dst; 3151 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3152 dst->remote_ifindex); 3153 3154 /* This check is different than dev->max_mtu, because it looks at 3155 * the lowerdev->mtu, rather than the static dev->max_mtu 3156 */ 3157 if (lowerdev) { 3158 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3159 if (new_mtu > max_mtu) 3160 return -EINVAL; 3161 } 3162 3163 dev->mtu = new_mtu; 3164 return 0; 3165 } 3166 3167 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3168 { 3169 struct vxlan_dev *vxlan = netdev_priv(dev); 3170 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3171 __be16 sport, dport; 3172 3173 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3174 vxlan->cfg.port_max, true); 3175 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3176 3177 if (ip_tunnel_info_af(info) == AF_INET) { 3178 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3179 struct rtable *rt; 3180 3181 if (!sock4) 3182 return -EIO; 3183 3184 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3185 &info->key.u.ipv4.src, 3186 &info->key, 3187 sport, dport, info->key.tos, 3188 &info->dst_cache); 3189 if (IS_ERR(rt)) 3190 return PTR_ERR(rt); 3191 ip_rt_put(rt); 3192 } else { 3193 #if IS_ENABLED(CONFIG_IPV6) 3194 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3195 struct dst_entry *ndst; 3196 3197 if (!sock6) 3198 return -EIO; 3199 3200 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3201 0, &info->key.u.ipv6.src, 3202 &info->key, 3203 sport, dport, info->key.tos, 3204 &info->dst_cache); 3205 if (IS_ERR(ndst)) 3206 return PTR_ERR(ndst); 3207 dst_release(ndst); 3208 #else /* !CONFIG_IPV6 */ 3209 return -EPFNOSUPPORT; 3210 #endif 3211 } 3212 info->key.tp_src = sport; 3213 info->key.tp_dst = dport; 3214 return 0; 3215 } 3216 3217 static const struct net_device_ops vxlan_netdev_ether_ops = { 3218 .ndo_init = vxlan_init, 3219 .ndo_uninit = vxlan_uninit, 3220 .ndo_open = vxlan_open, 3221 .ndo_stop = vxlan_stop, 3222 .ndo_start_xmit = vxlan_xmit, 3223 .ndo_set_rx_mode = vxlan_set_multicast_list, 3224 .ndo_change_mtu = vxlan_change_mtu, 3225 .ndo_validate_addr = eth_validate_addr, 3226 .ndo_set_mac_address = eth_mac_addr, 3227 .ndo_fdb_add = vxlan_fdb_add, 3228 .ndo_fdb_del = vxlan_fdb_delete, 3229 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3230 .ndo_fdb_dump = vxlan_fdb_dump, 3231 .ndo_fdb_get = vxlan_fdb_get, 3232 .ndo_mdb_add = vxlan_mdb_add, 3233 .ndo_mdb_del = vxlan_mdb_del, 3234 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3235 .ndo_mdb_dump = vxlan_mdb_dump, 3236 .ndo_mdb_get = vxlan_mdb_get, 3237 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3238 }; 3239 3240 static const struct net_device_ops vxlan_netdev_raw_ops = { 3241 .ndo_init = vxlan_init, 3242 .ndo_uninit = vxlan_uninit, 3243 .ndo_open = vxlan_open, 3244 .ndo_stop = vxlan_stop, 3245 .ndo_start_xmit = vxlan_xmit, 3246 .ndo_change_mtu = vxlan_change_mtu, 3247 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3248 }; 3249 3250 /* Info for udev, that this is a virtual tunnel endpoint */ 3251 static const struct device_type vxlan_type = { 3252 .name = "vxlan", 3253 }; 3254 3255 /* Calls the ndo_udp_tunnel_add of the caller in order to 3256 * supply the listening VXLAN udp ports. Callers are expected 3257 * to implement the ndo_udp_tunnel_add. 3258 */ 3259 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3260 { 3261 struct vxlan_sock *vs; 3262 struct net *net = dev_net(dev); 3263 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3264 unsigned int i; 3265 3266 spin_lock(&vn->sock_lock); 3267 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3268 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3269 unsigned short type; 3270 3271 if (vs->flags & VXLAN_F_GPE) 3272 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3273 else 3274 type = UDP_TUNNEL_TYPE_VXLAN; 3275 3276 if (push) 3277 udp_tunnel_push_rx_port(dev, vs->sock, type); 3278 else 3279 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3280 } 3281 } 3282 spin_unlock(&vn->sock_lock); 3283 } 3284 3285 /* Initialize the device structure. */ 3286 static void vxlan_setup(struct net_device *dev) 3287 { 3288 struct vxlan_dev *vxlan = netdev_priv(dev); 3289 unsigned int h; 3290 3291 eth_hw_addr_random(dev); 3292 ether_setup(dev); 3293 3294 dev->needs_free_netdev = true; 3295 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3296 3297 dev->features |= NETIF_F_LLTX; 3298 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3299 dev->features |= NETIF_F_RXCSUM; 3300 dev->features |= NETIF_F_GSO_SOFTWARE; 3301 3302 dev->vlan_features = dev->features; 3303 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3304 dev->hw_features |= NETIF_F_RXCSUM; 3305 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3306 netif_keep_dst(dev); 3307 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN; 3308 3309 /* MTU range: 68 - 65535 */ 3310 dev->min_mtu = ETH_MIN_MTU; 3311 dev->max_mtu = ETH_MAX_MTU; 3312 3313 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3314 INIT_LIST_HEAD(&vxlan->next); 3315 3316 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3317 3318 vxlan->dev = dev; 3319 3320 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3321 spin_lock_init(&vxlan->hash_lock[h]); 3322 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3323 } 3324 } 3325 3326 static void vxlan_ether_setup(struct net_device *dev) 3327 { 3328 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3329 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3330 dev->netdev_ops = &vxlan_netdev_ether_ops; 3331 } 3332 3333 static void vxlan_raw_setup(struct net_device *dev) 3334 { 3335 dev->header_ops = NULL; 3336 dev->type = ARPHRD_NONE; 3337 dev->hard_header_len = 0; 3338 dev->addr_len = 0; 3339 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3340 dev->netdev_ops = &vxlan_netdev_raw_ops; 3341 } 3342 3343 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3344 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3345 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3346 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3347 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3348 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3349 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3350 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3351 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3352 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3353 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3354 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3355 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3356 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3357 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3358 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3359 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3360 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3361 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3362 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3363 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3364 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3365 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3366 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3367 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3368 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3369 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3370 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3371 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3372 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3373 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3374 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3375 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3376 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3377 }; 3378 3379 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3380 struct netlink_ext_ack *extack) 3381 { 3382 if (tb[IFLA_ADDRESS]) { 3383 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3384 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3385 "Provided link layer address is not Ethernet"); 3386 return -EINVAL; 3387 } 3388 3389 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3390 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3391 "Provided Ethernet address is not unicast"); 3392 return -EADDRNOTAVAIL; 3393 } 3394 } 3395 3396 if (tb[IFLA_MTU]) { 3397 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3398 3399 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3400 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3401 "MTU must be between 68 and 65535"); 3402 return -EINVAL; 3403 } 3404 } 3405 3406 if (!data) { 3407 NL_SET_ERR_MSG(extack, 3408 "Required attributes not provided to perform the operation"); 3409 return -EINVAL; 3410 } 3411 3412 if (data[IFLA_VXLAN_ID]) { 3413 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3414 3415 if (id >= VXLAN_N_VID) { 3416 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3417 "VXLAN ID must be lower than 16777216"); 3418 return -ERANGE; 3419 } 3420 } 3421 3422 if (data[IFLA_VXLAN_PORT_RANGE]) { 3423 const struct ifla_vxlan_port_range *p 3424 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3425 3426 if (ntohs(p->high) < ntohs(p->low)) { 3427 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3428 "Invalid source port range"); 3429 return -EINVAL; 3430 } 3431 } 3432 3433 if (data[IFLA_VXLAN_DF]) { 3434 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3435 3436 if (df < 0 || df > VXLAN_DF_MAX) { 3437 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3438 "Invalid DF attribute"); 3439 return -EINVAL; 3440 } 3441 } 3442 3443 return 0; 3444 } 3445 3446 static void vxlan_get_drvinfo(struct net_device *netdev, 3447 struct ethtool_drvinfo *drvinfo) 3448 { 3449 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3450 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3451 } 3452 3453 static int vxlan_get_link_ksettings(struct net_device *dev, 3454 struct ethtool_link_ksettings *cmd) 3455 { 3456 struct vxlan_dev *vxlan = netdev_priv(dev); 3457 struct vxlan_rdst *dst = &vxlan->default_dst; 3458 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3459 dst->remote_ifindex); 3460 3461 if (!lowerdev) { 3462 cmd->base.duplex = DUPLEX_UNKNOWN; 3463 cmd->base.port = PORT_OTHER; 3464 cmd->base.speed = SPEED_UNKNOWN; 3465 3466 return 0; 3467 } 3468 3469 return __ethtool_get_link_ksettings(lowerdev, cmd); 3470 } 3471 3472 static const struct ethtool_ops vxlan_ethtool_ops = { 3473 .get_drvinfo = vxlan_get_drvinfo, 3474 .get_link = ethtool_op_get_link, 3475 .get_link_ksettings = vxlan_get_link_ksettings, 3476 }; 3477 3478 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3479 __be16 port, u32 flags, int ifindex) 3480 { 3481 struct socket *sock; 3482 struct udp_port_cfg udp_conf; 3483 int err; 3484 3485 memset(&udp_conf, 0, sizeof(udp_conf)); 3486 3487 if (ipv6) { 3488 udp_conf.family = AF_INET6; 3489 udp_conf.use_udp6_rx_checksums = 3490 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3491 udp_conf.ipv6_v6only = 1; 3492 } else { 3493 udp_conf.family = AF_INET; 3494 } 3495 3496 udp_conf.local_udp_port = port; 3497 udp_conf.bind_ifindex = ifindex; 3498 3499 /* Open UDP socket */ 3500 err = udp_sock_create(net, &udp_conf, &sock); 3501 if (err < 0) 3502 return ERR_PTR(err); 3503 3504 udp_allow_gso(sock->sk); 3505 return sock; 3506 } 3507 3508 /* Create new listen socket if needed */ 3509 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3510 __be16 port, u32 flags, 3511 int ifindex) 3512 { 3513 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3514 struct vxlan_sock *vs; 3515 struct socket *sock; 3516 unsigned int h; 3517 struct udp_tunnel_sock_cfg tunnel_cfg; 3518 3519 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3520 if (!vs) 3521 return ERR_PTR(-ENOMEM); 3522 3523 for (h = 0; h < VNI_HASH_SIZE; ++h) 3524 INIT_HLIST_HEAD(&vs->vni_list[h]); 3525 3526 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3527 if (IS_ERR(sock)) { 3528 kfree(vs); 3529 return ERR_CAST(sock); 3530 } 3531 3532 vs->sock = sock; 3533 refcount_set(&vs->refcnt, 1); 3534 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3535 3536 spin_lock(&vn->sock_lock); 3537 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3538 udp_tunnel_notify_add_rx_port(sock, 3539 (vs->flags & VXLAN_F_GPE) ? 3540 UDP_TUNNEL_TYPE_VXLAN_GPE : 3541 UDP_TUNNEL_TYPE_VXLAN); 3542 spin_unlock(&vn->sock_lock); 3543 3544 /* Mark socket as an encapsulation socket. */ 3545 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3546 tunnel_cfg.sk_user_data = vs; 3547 tunnel_cfg.encap_type = 1; 3548 tunnel_cfg.encap_rcv = vxlan_rcv; 3549 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3550 tunnel_cfg.encap_destroy = NULL; 3551 if (vs->flags & VXLAN_F_GPE) { 3552 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3553 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3554 } else { 3555 tunnel_cfg.gro_receive = vxlan_gro_receive; 3556 tunnel_cfg.gro_complete = vxlan_gro_complete; 3557 } 3558 3559 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3560 3561 return vs; 3562 } 3563 3564 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3565 { 3566 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3567 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3568 struct vxlan_sock *vs = NULL; 3569 struct vxlan_dev_node *node; 3570 int l3mdev_index = 0; 3571 3572 if (vxlan->cfg.remote_ifindex) 3573 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3574 vxlan->net, vxlan->cfg.remote_ifindex); 3575 3576 if (!vxlan->cfg.no_share) { 3577 spin_lock(&vn->sock_lock); 3578 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3579 vxlan->cfg.dst_port, vxlan->cfg.flags, 3580 l3mdev_index); 3581 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3582 spin_unlock(&vn->sock_lock); 3583 return -EBUSY; 3584 } 3585 spin_unlock(&vn->sock_lock); 3586 } 3587 if (!vs) 3588 vs = vxlan_socket_create(vxlan->net, ipv6, 3589 vxlan->cfg.dst_port, vxlan->cfg.flags, 3590 l3mdev_index); 3591 if (IS_ERR(vs)) 3592 return PTR_ERR(vs); 3593 #if IS_ENABLED(CONFIG_IPV6) 3594 if (ipv6) { 3595 rcu_assign_pointer(vxlan->vn6_sock, vs); 3596 node = &vxlan->hlist6; 3597 } else 3598 #endif 3599 { 3600 rcu_assign_pointer(vxlan->vn4_sock, vs); 3601 node = &vxlan->hlist4; 3602 } 3603 3604 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3605 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3606 else 3607 vxlan_vs_add_dev(vs, vxlan, node); 3608 3609 return 0; 3610 } 3611 3612 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3613 { 3614 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3615 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3616 bool ipv4 = !ipv6 || metadata; 3617 int ret = 0; 3618 3619 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3620 #if IS_ENABLED(CONFIG_IPV6) 3621 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3622 if (ipv6) { 3623 ret = __vxlan_sock_add(vxlan, true); 3624 if (ret < 0 && ret != -EAFNOSUPPORT) 3625 ipv4 = false; 3626 } 3627 #endif 3628 if (ipv4) 3629 ret = __vxlan_sock_add(vxlan, false); 3630 if (ret < 0) 3631 vxlan_sock_release(vxlan); 3632 return ret; 3633 } 3634 3635 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3636 struct vxlan_config *conf, __be32 vni) 3637 { 3638 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3639 struct vxlan_dev *tmp; 3640 3641 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3642 if (tmp == vxlan) 3643 continue; 3644 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3645 if (!vxlan_vnifilter_lookup(tmp, vni)) 3646 continue; 3647 } else if (tmp->cfg.vni != vni) { 3648 continue; 3649 } 3650 if (tmp->cfg.dst_port != conf->dst_port) 3651 continue; 3652 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3653 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3654 continue; 3655 3656 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3657 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3658 continue; 3659 3660 return -EEXIST; 3661 } 3662 3663 return 0; 3664 } 3665 3666 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3667 struct net_device **lower, 3668 struct vxlan_dev *old, 3669 struct netlink_ext_ack *extack) 3670 { 3671 bool use_ipv6 = false; 3672 3673 if (conf->flags & VXLAN_F_GPE) { 3674 /* For now, allow GPE only together with 3675 * COLLECT_METADATA. This can be relaxed later; in such 3676 * case, the other side of the PtP link will have to be 3677 * provided. 3678 */ 3679 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3680 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3681 NL_SET_ERR_MSG(extack, 3682 "VXLAN GPE does not support this combination of attributes"); 3683 return -EINVAL; 3684 } 3685 } 3686 3687 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3688 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3689 conf->remote_ip.sa.sa_family = AF_INET; 3690 conf->saddr.sa.sa_family = AF_INET; 3691 } else if (!conf->remote_ip.sa.sa_family) { 3692 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3693 } else if (!conf->saddr.sa.sa_family) { 3694 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3695 } 3696 3697 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3698 NL_SET_ERR_MSG(extack, 3699 "Local and remote address must be from the same family"); 3700 return -EINVAL; 3701 } 3702 3703 if (vxlan_addr_multicast(&conf->saddr)) { 3704 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3705 return -EINVAL; 3706 } 3707 3708 if (conf->saddr.sa.sa_family == AF_INET6) { 3709 if (!IS_ENABLED(CONFIG_IPV6)) { 3710 NL_SET_ERR_MSG(extack, 3711 "IPv6 support not enabled in the kernel"); 3712 return -EPFNOSUPPORT; 3713 } 3714 use_ipv6 = true; 3715 conf->flags |= VXLAN_F_IPV6; 3716 3717 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3718 int local_type = 3719 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3720 int remote_type = 3721 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3722 3723 if (local_type & IPV6_ADDR_LINKLOCAL) { 3724 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3725 (remote_type != IPV6_ADDR_ANY)) { 3726 NL_SET_ERR_MSG(extack, 3727 "Invalid combination of local and remote address scopes"); 3728 return -EINVAL; 3729 } 3730 3731 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3732 } else { 3733 if (remote_type == 3734 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3735 NL_SET_ERR_MSG(extack, 3736 "Invalid combination of local and remote address scopes"); 3737 return -EINVAL; 3738 } 3739 3740 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3741 } 3742 } 3743 } 3744 3745 if (conf->label && !use_ipv6) { 3746 NL_SET_ERR_MSG(extack, 3747 "Label attribute only applies to IPv6 VXLAN devices"); 3748 return -EINVAL; 3749 } 3750 3751 if (conf->label_policy && !use_ipv6) { 3752 NL_SET_ERR_MSG(extack, 3753 "Label policy only applies to IPv6 VXLAN devices"); 3754 return -EINVAL; 3755 } 3756 3757 if (conf->remote_ifindex) { 3758 struct net_device *lowerdev; 3759 3760 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3761 if (!lowerdev) { 3762 NL_SET_ERR_MSG(extack, 3763 "Invalid local interface, device not found"); 3764 return -ENODEV; 3765 } 3766 3767 #if IS_ENABLED(CONFIG_IPV6) 3768 if (use_ipv6) { 3769 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3770 3771 if (idev && idev->cnf.disable_ipv6) { 3772 NL_SET_ERR_MSG(extack, 3773 "IPv6 support disabled by administrator"); 3774 return -EPERM; 3775 } 3776 } 3777 #endif 3778 3779 *lower = lowerdev; 3780 } else { 3781 if (vxlan_addr_multicast(&conf->remote_ip)) { 3782 NL_SET_ERR_MSG(extack, 3783 "Local interface required for multicast remote destination"); 3784 3785 return -EINVAL; 3786 } 3787 3788 #if IS_ENABLED(CONFIG_IPV6) 3789 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3790 NL_SET_ERR_MSG(extack, 3791 "Local interface required for link-local local/remote addresses"); 3792 return -EINVAL; 3793 } 3794 #endif 3795 3796 *lower = NULL; 3797 } 3798 3799 if (!conf->dst_port) { 3800 if (conf->flags & VXLAN_F_GPE) 3801 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3802 else 3803 conf->dst_port = htons(vxlan_port); 3804 } 3805 3806 if (!conf->age_interval) 3807 conf->age_interval = FDB_AGE_DEFAULT; 3808 3809 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3810 NL_SET_ERR_MSG(extack, 3811 "A VXLAN device with the specified VNI already exists"); 3812 return -EEXIST; 3813 } 3814 3815 return 0; 3816 } 3817 3818 static void vxlan_config_apply(struct net_device *dev, 3819 struct vxlan_config *conf, 3820 struct net_device *lowerdev, 3821 struct net *src_net, 3822 bool changelink) 3823 { 3824 struct vxlan_dev *vxlan = netdev_priv(dev); 3825 struct vxlan_rdst *dst = &vxlan->default_dst; 3826 unsigned short needed_headroom = ETH_HLEN; 3827 int max_mtu = ETH_MAX_MTU; 3828 u32 flags = conf->flags; 3829 3830 if (!changelink) { 3831 if (flags & VXLAN_F_GPE) 3832 vxlan_raw_setup(dev); 3833 else 3834 vxlan_ether_setup(dev); 3835 3836 if (conf->mtu) 3837 dev->mtu = conf->mtu; 3838 3839 vxlan->net = src_net; 3840 } 3841 3842 dst->remote_vni = conf->vni; 3843 3844 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3845 3846 if (lowerdev) { 3847 dst->remote_ifindex = conf->remote_ifindex; 3848 3849 netif_inherit_tso_max(dev, lowerdev); 3850 3851 needed_headroom = lowerdev->hard_header_len; 3852 needed_headroom += lowerdev->needed_headroom; 3853 3854 dev->needed_tailroom = lowerdev->needed_tailroom; 3855 3856 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3857 if (max_mtu < ETH_MIN_MTU) 3858 max_mtu = ETH_MIN_MTU; 3859 3860 if (!changelink && !conf->mtu) 3861 dev->mtu = max_mtu; 3862 } 3863 3864 if (dev->mtu > max_mtu) 3865 dev->mtu = max_mtu; 3866 3867 if (flags & VXLAN_F_COLLECT_METADATA) 3868 flags |= VXLAN_F_IPV6; 3869 needed_headroom += vxlan_headroom(flags); 3870 dev->needed_headroom = needed_headroom; 3871 3872 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3873 } 3874 3875 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3876 struct vxlan_config *conf, bool changelink, 3877 struct netlink_ext_ack *extack) 3878 { 3879 struct vxlan_dev *vxlan = netdev_priv(dev); 3880 struct net_device *lowerdev; 3881 int ret; 3882 3883 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3884 if (ret) 3885 return ret; 3886 3887 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3888 3889 return 0; 3890 } 3891 3892 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3893 struct vxlan_config *conf, 3894 struct netlink_ext_ack *extack) 3895 { 3896 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3897 struct vxlan_dev *vxlan = netdev_priv(dev); 3898 struct net_device *remote_dev = NULL; 3899 struct vxlan_fdb *f = NULL; 3900 bool unregister = false; 3901 struct vxlan_rdst *dst; 3902 int err; 3903 3904 dst = &vxlan->default_dst; 3905 err = vxlan_dev_configure(net, dev, conf, false, extack); 3906 if (err) 3907 return err; 3908 3909 dev->ethtool_ops = &vxlan_ethtool_ops; 3910 3911 /* create an fdb entry for a valid default destination */ 3912 if (!vxlan_addr_any(&dst->remote_ip)) { 3913 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3914 &dst->remote_ip, 3915 NUD_REACHABLE | NUD_PERMANENT, 3916 vxlan->cfg.dst_port, 3917 dst->remote_vni, 3918 dst->remote_vni, 3919 dst->remote_ifindex, 3920 NTF_SELF, 0, &f, extack); 3921 if (err) 3922 return err; 3923 } 3924 3925 err = register_netdevice(dev); 3926 if (err) 3927 goto errout; 3928 unregister = true; 3929 3930 if (dst->remote_ifindex) { 3931 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3932 if (!remote_dev) { 3933 err = -ENODEV; 3934 goto errout; 3935 } 3936 3937 err = netdev_upper_dev_link(remote_dev, dev, extack); 3938 if (err) 3939 goto errout; 3940 } 3941 3942 err = rtnl_configure_link(dev, NULL, 0, NULL); 3943 if (err < 0) 3944 goto unlink; 3945 3946 if (f) { 3947 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3948 3949 /* notify default fdb entry */ 3950 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3951 RTM_NEWNEIGH, true, extack); 3952 if (err) { 3953 vxlan_fdb_destroy(vxlan, f, false, false); 3954 if (remote_dev) 3955 netdev_upper_dev_unlink(remote_dev, dev); 3956 goto unregister; 3957 } 3958 } 3959 3960 list_add(&vxlan->next, &vn->vxlan_list); 3961 if (remote_dev) 3962 dst->remote_dev = remote_dev; 3963 return 0; 3964 unlink: 3965 if (remote_dev) 3966 netdev_upper_dev_unlink(remote_dev, dev); 3967 errout: 3968 /* unregister_netdevice() destroys the default FDB entry with deletion 3969 * notification. But the addition notification was not sent yet, so 3970 * destroy the entry by hand here. 3971 */ 3972 if (f) 3973 __vxlan_fdb_free(f); 3974 unregister: 3975 if (unregister) 3976 unregister_netdevice(dev); 3977 return err; 3978 } 3979 3980 /* Set/clear flags based on attribute */ 3981 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 3982 int attrtype, unsigned long mask, bool changelink, 3983 bool changelink_supported, 3984 struct netlink_ext_ack *extack) 3985 { 3986 unsigned long flags; 3987 3988 if (!tb[attrtype]) 3989 return 0; 3990 3991 if (changelink && !changelink_supported) { 3992 vxlan_flag_attr_error(attrtype, extack); 3993 return -EOPNOTSUPP; 3994 } 3995 3996 if (vxlan_policy[attrtype].type == NLA_FLAG) 3997 flags = conf->flags | mask; 3998 else if (nla_get_u8(tb[attrtype])) 3999 flags = conf->flags | mask; 4000 else 4001 flags = conf->flags & ~mask; 4002 4003 conf->flags = flags; 4004 4005 return 0; 4006 } 4007 4008 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4009 struct net_device *dev, struct vxlan_config *conf, 4010 bool changelink, struct netlink_ext_ack *extack) 4011 { 4012 struct vxlan_dev *vxlan = netdev_priv(dev); 4013 int err = 0; 4014 4015 memset(conf, 0, sizeof(*conf)); 4016 4017 /* if changelink operation, start with old existing cfg */ 4018 if (changelink) 4019 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4020 4021 if (data[IFLA_VXLAN_ID]) { 4022 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4023 4024 if (changelink && (vni != conf->vni)) { 4025 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4026 return -EOPNOTSUPP; 4027 } 4028 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4029 } 4030 4031 if (data[IFLA_VXLAN_GROUP]) { 4032 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4033 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4034 return -EOPNOTSUPP; 4035 } 4036 4037 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4038 conf->remote_ip.sa.sa_family = AF_INET; 4039 } else if (data[IFLA_VXLAN_GROUP6]) { 4040 if (!IS_ENABLED(CONFIG_IPV6)) { 4041 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4042 return -EPFNOSUPPORT; 4043 } 4044 4045 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4046 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4047 return -EOPNOTSUPP; 4048 } 4049 4050 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4051 conf->remote_ip.sa.sa_family = AF_INET6; 4052 } 4053 4054 if (data[IFLA_VXLAN_LOCAL]) { 4055 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4056 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4057 return -EOPNOTSUPP; 4058 } 4059 4060 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4061 conf->saddr.sa.sa_family = AF_INET; 4062 } else if (data[IFLA_VXLAN_LOCAL6]) { 4063 if (!IS_ENABLED(CONFIG_IPV6)) { 4064 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4065 return -EPFNOSUPPORT; 4066 } 4067 4068 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4069 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4070 return -EOPNOTSUPP; 4071 } 4072 4073 /* TODO: respect scope id */ 4074 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4075 conf->saddr.sa.sa_family = AF_INET6; 4076 } 4077 4078 if (data[IFLA_VXLAN_LINK]) 4079 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4080 4081 if (data[IFLA_VXLAN_TOS]) 4082 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4083 4084 if (data[IFLA_VXLAN_TTL]) 4085 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4086 4087 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4088 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4089 VXLAN_F_TTL_INHERIT, changelink, false, 4090 extack); 4091 if (err) 4092 return err; 4093 4094 } 4095 4096 if (data[IFLA_VXLAN_LABEL]) 4097 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4098 IPV6_FLOWLABEL_MASK; 4099 if (data[IFLA_VXLAN_LABEL_POLICY]) 4100 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4101 4102 if (data[IFLA_VXLAN_LEARNING]) { 4103 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4104 VXLAN_F_LEARN, changelink, true, 4105 extack); 4106 if (err) 4107 return err; 4108 } else if (!changelink) { 4109 /* default to learn on a new device */ 4110 conf->flags |= VXLAN_F_LEARN; 4111 } 4112 4113 if (data[IFLA_VXLAN_AGEING]) 4114 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4115 4116 if (data[IFLA_VXLAN_PROXY]) { 4117 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4118 VXLAN_F_PROXY, changelink, false, 4119 extack); 4120 if (err) 4121 return err; 4122 } 4123 4124 if (data[IFLA_VXLAN_RSC]) { 4125 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4126 VXLAN_F_RSC, changelink, false, 4127 extack); 4128 if (err) 4129 return err; 4130 } 4131 4132 if (data[IFLA_VXLAN_L2MISS]) { 4133 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4134 VXLAN_F_L2MISS, changelink, false, 4135 extack); 4136 if (err) 4137 return err; 4138 } 4139 4140 if (data[IFLA_VXLAN_L3MISS]) { 4141 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4142 VXLAN_F_L3MISS, changelink, false, 4143 extack); 4144 if (err) 4145 return err; 4146 } 4147 4148 if (data[IFLA_VXLAN_LIMIT]) { 4149 if (changelink) { 4150 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4151 "Cannot change limit"); 4152 return -EOPNOTSUPP; 4153 } 4154 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4155 } 4156 4157 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4158 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4159 VXLAN_F_COLLECT_METADATA, changelink, false, 4160 extack); 4161 if (err) 4162 return err; 4163 } 4164 4165 if (data[IFLA_VXLAN_PORT_RANGE]) { 4166 if (!changelink) { 4167 const struct ifla_vxlan_port_range *p 4168 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4169 conf->port_min = ntohs(p->low); 4170 conf->port_max = ntohs(p->high); 4171 } else { 4172 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4173 "Cannot change port range"); 4174 return -EOPNOTSUPP; 4175 } 4176 } 4177 4178 if (data[IFLA_VXLAN_PORT]) { 4179 if (changelink) { 4180 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4181 "Cannot change port"); 4182 return -EOPNOTSUPP; 4183 } 4184 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4185 } 4186 4187 if (data[IFLA_VXLAN_UDP_CSUM]) { 4188 if (changelink) { 4189 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4190 "Cannot change UDP_CSUM flag"); 4191 return -EOPNOTSUPP; 4192 } 4193 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4194 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4195 } 4196 4197 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4198 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4199 VXLAN_F_LOCALBYPASS, changelink, 4200 true, extack); 4201 if (err) 4202 return err; 4203 } else if (!changelink) { 4204 /* default to local bypass on a new device */ 4205 conf->flags |= VXLAN_F_LOCALBYPASS; 4206 } 4207 4208 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4209 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4210 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4211 false, extack); 4212 if (err) 4213 return err; 4214 } 4215 4216 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4217 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4218 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4219 false, extack); 4220 if (err) 4221 return err; 4222 } 4223 4224 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4225 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4226 VXLAN_F_REMCSUM_TX, changelink, false, 4227 extack); 4228 if (err) 4229 return err; 4230 } 4231 4232 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4233 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4234 VXLAN_F_REMCSUM_RX, changelink, false, 4235 extack); 4236 if (err) 4237 return err; 4238 } 4239 4240 if (data[IFLA_VXLAN_GBP]) { 4241 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4242 VXLAN_F_GBP, changelink, false, extack); 4243 if (err) 4244 return err; 4245 } 4246 4247 if (data[IFLA_VXLAN_GPE]) { 4248 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4249 VXLAN_F_GPE, changelink, false, 4250 extack); 4251 if (err) 4252 return err; 4253 } 4254 4255 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4256 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4257 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4258 false, extack); 4259 if (err) 4260 return err; 4261 } 4262 4263 if (tb[IFLA_MTU]) { 4264 if (changelink) { 4265 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4266 "Cannot change mtu"); 4267 return -EOPNOTSUPP; 4268 } 4269 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4270 } 4271 4272 if (data[IFLA_VXLAN_DF]) 4273 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4274 4275 if (data[IFLA_VXLAN_VNIFILTER]) { 4276 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4277 VXLAN_F_VNIFILTER, changelink, false, 4278 extack); 4279 if (err) 4280 return err; 4281 4282 if ((conf->flags & VXLAN_F_VNIFILTER) && 4283 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4284 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4285 "vxlan vnifilter only valid in collect metadata mode"); 4286 return -EINVAL; 4287 } 4288 } 4289 4290 return 0; 4291 } 4292 4293 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4294 struct nlattr *tb[], struct nlattr *data[], 4295 struct netlink_ext_ack *extack) 4296 { 4297 struct vxlan_config conf; 4298 int err; 4299 4300 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4301 if (err) 4302 return err; 4303 4304 return __vxlan_dev_create(src_net, dev, &conf, extack); 4305 } 4306 4307 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4308 struct nlattr *data[], 4309 struct netlink_ext_ack *extack) 4310 { 4311 struct vxlan_dev *vxlan = netdev_priv(dev); 4312 struct net_device *lowerdev; 4313 struct vxlan_config conf; 4314 struct vxlan_rdst *dst; 4315 int err; 4316 4317 dst = &vxlan->default_dst; 4318 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4319 if (err) 4320 return err; 4321 4322 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4323 vxlan, extack); 4324 if (err) 4325 return err; 4326 4327 if (dst->remote_dev == lowerdev) 4328 lowerdev = NULL; 4329 4330 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4331 extack); 4332 if (err) 4333 return err; 4334 4335 /* handle default dst entry */ 4336 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4337 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4338 4339 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4340 if (!vxlan_addr_any(&conf.remote_ip)) { 4341 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4342 &conf.remote_ip, 4343 NUD_REACHABLE | NUD_PERMANENT, 4344 NLM_F_APPEND | NLM_F_CREATE, 4345 vxlan->cfg.dst_port, 4346 conf.vni, conf.vni, 4347 conf.remote_ifindex, 4348 NTF_SELF, 0, true, extack); 4349 if (err) { 4350 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4351 netdev_adjacent_change_abort(dst->remote_dev, 4352 lowerdev, dev); 4353 return err; 4354 } 4355 } 4356 if (!vxlan_addr_any(&dst->remote_ip)) 4357 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4358 dst->remote_ip, 4359 vxlan->cfg.dst_port, 4360 dst->remote_vni, 4361 dst->remote_vni, 4362 dst->remote_ifindex, 4363 true); 4364 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4365 4366 /* If vni filtering device, also update fdb entries of 4367 * all vnis that were using default remote ip 4368 */ 4369 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4370 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4371 &conf.remote_ip, extack); 4372 if (err) { 4373 netdev_adjacent_change_abort(dst->remote_dev, 4374 lowerdev, dev); 4375 return err; 4376 } 4377 } 4378 } 4379 4380 if (conf.age_interval != vxlan->cfg.age_interval) 4381 mod_timer(&vxlan->age_timer, jiffies); 4382 4383 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4384 if (lowerdev && lowerdev != dst->remote_dev) 4385 dst->remote_dev = lowerdev; 4386 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4387 return 0; 4388 } 4389 4390 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4391 { 4392 struct vxlan_dev *vxlan = netdev_priv(dev); 4393 struct vxlan_fdb_flush_desc desc = { 4394 /* Default entry is deleted at vxlan_uninit. */ 4395 .ignore_default_entry = true, 4396 }; 4397 4398 vxlan_flush(vxlan, &desc); 4399 4400 list_del(&vxlan->next); 4401 unregister_netdevice_queue(dev, head); 4402 if (vxlan->default_dst.remote_dev) 4403 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4404 } 4405 4406 static size_t vxlan_get_size(const struct net_device *dev) 4407 { 4408 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4409 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4410 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4411 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4412 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4413 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4414 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4415 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4416 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4417 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4418 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4419 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4420 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4421 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4422 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4423 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4424 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4425 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4426 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4427 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4428 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4429 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4430 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4431 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4432 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4433 /* IFLA_VXLAN_PORT_RANGE */ 4434 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4435 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4436 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4437 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4438 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4439 0; 4440 } 4441 4442 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4443 { 4444 const struct vxlan_dev *vxlan = netdev_priv(dev); 4445 const struct vxlan_rdst *dst = &vxlan->default_dst; 4446 struct ifla_vxlan_port_range ports = { 4447 .low = htons(vxlan->cfg.port_min), 4448 .high = htons(vxlan->cfg.port_max), 4449 }; 4450 4451 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4452 goto nla_put_failure; 4453 4454 if (!vxlan_addr_any(&dst->remote_ip)) { 4455 if (dst->remote_ip.sa.sa_family == AF_INET) { 4456 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4457 dst->remote_ip.sin.sin_addr.s_addr)) 4458 goto nla_put_failure; 4459 #if IS_ENABLED(CONFIG_IPV6) 4460 } else { 4461 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4462 &dst->remote_ip.sin6.sin6_addr)) 4463 goto nla_put_failure; 4464 #endif 4465 } 4466 } 4467 4468 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4469 goto nla_put_failure; 4470 4471 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4472 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4473 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4474 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4475 goto nla_put_failure; 4476 #if IS_ENABLED(CONFIG_IPV6) 4477 } else { 4478 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4479 &vxlan->cfg.saddr.sin6.sin6_addr)) 4480 goto nla_put_failure; 4481 #endif 4482 } 4483 } 4484 4485 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4486 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4487 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4488 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4489 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4490 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4491 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4492 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4493 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4494 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4495 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4496 nla_put_u8(skb, IFLA_VXLAN_RSC, 4497 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4498 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4499 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4500 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4501 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4502 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4503 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4504 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4505 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4506 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4507 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4508 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4509 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4510 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4511 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4512 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4513 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4514 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4515 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4516 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4517 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4518 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4519 goto nla_put_failure; 4520 4521 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4522 goto nla_put_failure; 4523 4524 if (vxlan->cfg.flags & VXLAN_F_GBP && 4525 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4526 goto nla_put_failure; 4527 4528 if (vxlan->cfg.flags & VXLAN_F_GPE && 4529 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4530 goto nla_put_failure; 4531 4532 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4533 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4534 goto nla_put_failure; 4535 4536 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4537 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4538 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4539 goto nla_put_failure; 4540 4541 return 0; 4542 4543 nla_put_failure: 4544 return -EMSGSIZE; 4545 } 4546 4547 static struct net *vxlan_get_link_net(const struct net_device *dev) 4548 { 4549 struct vxlan_dev *vxlan = netdev_priv(dev); 4550 4551 return vxlan->net; 4552 } 4553 4554 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4555 .kind = "vxlan", 4556 .maxtype = IFLA_VXLAN_MAX, 4557 .policy = vxlan_policy, 4558 .priv_size = sizeof(struct vxlan_dev), 4559 .setup = vxlan_setup, 4560 .validate = vxlan_validate, 4561 .newlink = vxlan_newlink, 4562 .changelink = vxlan_changelink, 4563 .dellink = vxlan_dellink, 4564 .get_size = vxlan_get_size, 4565 .fill_info = vxlan_fill_info, 4566 .get_link_net = vxlan_get_link_net, 4567 }; 4568 4569 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4570 u8 name_assign_type, 4571 struct vxlan_config *conf) 4572 { 4573 struct nlattr *tb[IFLA_MAX + 1]; 4574 struct net_device *dev; 4575 int err; 4576 4577 memset(&tb, 0, sizeof(tb)); 4578 4579 dev = rtnl_create_link(net, name, name_assign_type, 4580 &vxlan_link_ops, tb, NULL); 4581 if (IS_ERR(dev)) 4582 return dev; 4583 4584 err = __vxlan_dev_create(net, dev, conf, NULL); 4585 if (err < 0) { 4586 free_netdev(dev); 4587 return ERR_PTR(err); 4588 } 4589 4590 err = rtnl_configure_link(dev, NULL, 0, NULL); 4591 if (err < 0) { 4592 LIST_HEAD(list_kill); 4593 4594 vxlan_dellink(dev, &list_kill); 4595 unregister_netdevice_many(&list_kill); 4596 return ERR_PTR(err); 4597 } 4598 4599 return dev; 4600 } 4601 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4602 4603 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4604 struct net_device *dev) 4605 { 4606 struct vxlan_dev *vxlan, *next; 4607 LIST_HEAD(list_kill); 4608 4609 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4610 struct vxlan_rdst *dst = &vxlan->default_dst; 4611 4612 /* In case we created vxlan device with carrier 4613 * and we loose the carrier due to module unload 4614 * we also need to remove vxlan device. In other 4615 * cases, it's not necessary and remote_ifindex 4616 * is 0 here, so no matches. 4617 */ 4618 if (dst->remote_ifindex == dev->ifindex) 4619 vxlan_dellink(vxlan->dev, &list_kill); 4620 } 4621 4622 unregister_netdevice_many(&list_kill); 4623 } 4624 4625 static int vxlan_netdevice_event(struct notifier_block *unused, 4626 unsigned long event, void *ptr) 4627 { 4628 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4629 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4630 4631 if (event == NETDEV_UNREGISTER) 4632 vxlan_handle_lowerdev_unregister(vn, dev); 4633 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4634 vxlan_offload_rx_ports(dev, true); 4635 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4636 vxlan_offload_rx_ports(dev, false); 4637 4638 return NOTIFY_DONE; 4639 } 4640 4641 static struct notifier_block vxlan_notifier_block __read_mostly = { 4642 .notifier_call = vxlan_netdevice_event, 4643 }; 4644 4645 static void 4646 vxlan_fdb_offloaded_set(struct net_device *dev, 4647 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4648 { 4649 struct vxlan_dev *vxlan = netdev_priv(dev); 4650 struct vxlan_rdst *rdst; 4651 struct vxlan_fdb *f; 4652 u32 hash_index; 4653 4654 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4655 4656 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4657 4658 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4659 if (!f) 4660 goto out; 4661 4662 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4663 fdb_info->remote_port, 4664 fdb_info->remote_vni, 4665 fdb_info->remote_ifindex); 4666 if (!rdst) 4667 goto out; 4668 4669 rdst->offloaded = fdb_info->offloaded; 4670 4671 out: 4672 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4673 } 4674 4675 static int 4676 vxlan_fdb_external_learn_add(struct net_device *dev, 4677 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4678 { 4679 struct vxlan_dev *vxlan = netdev_priv(dev); 4680 struct netlink_ext_ack *extack; 4681 u32 hash_index; 4682 int err; 4683 4684 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4685 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4686 4687 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4688 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4689 NUD_REACHABLE, 4690 NLM_F_CREATE | NLM_F_REPLACE, 4691 fdb_info->remote_port, 4692 fdb_info->vni, 4693 fdb_info->remote_vni, 4694 fdb_info->remote_ifindex, 4695 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4696 0, false, extack); 4697 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4698 4699 return err; 4700 } 4701 4702 static int 4703 vxlan_fdb_external_learn_del(struct net_device *dev, 4704 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4705 { 4706 struct vxlan_dev *vxlan = netdev_priv(dev); 4707 struct vxlan_fdb *f; 4708 u32 hash_index; 4709 int err = 0; 4710 4711 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4712 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4713 4714 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4715 if (!f) 4716 err = -ENOENT; 4717 else if (f->flags & NTF_EXT_LEARNED) 4718 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4719 fdb_info->remote_ip, 4720 fdb_info->remote_port, 4721 fdb_info->vni, 4722 fdb_info->remote_vni, 4723 fdb_info->remote_ifindex, 4724 false); 4725 4726 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4727 4728 return err; 4729 } 4730 4731 static int vxlan_switchdev_event(struct notifier_block *unused, 4732 unsigned long event, void *ptr) 4733 { 4734 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4735 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4736 int err = 0; 4737 4738 switch (event) { 4739 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4740 vxlan_fdb_offloaded_set(dev, ptr); 4741 break; 4742 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4743 fdb_info = ptr; 4744 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4745 if (err) { 4746 err = notifier_from_errno(err); 4747 break; 4748 } 4749 fdb_info->offloaded = true; 4750 vxlan_fdb_offloaded_set(dev, fdb_info); 4751 break; 4752 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4753 fdb_info = ptr; 4754 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4755 if (err) { 4756 err = notifier_from_errno(err); 4757 break; 4758 } 4759 fdb_info->offloaded = false; 4760 vxlan_fdb_offloaded_set(dev, fdb_info); 4761 break; 4762 } 4763 4764 return err; 4765 } 4766 4767 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4768 .notifier_call = vxlan_switchdev_event, 4769 }; 4770 4771 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4772 { 4773 struct vxlan_fdb *fdb; 4774 struct vxlan_dev *vxlan; 4775 u32 hash_index; 4776 4777 rcu_read_lock(); 4778 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4779 vxlan = rcu_dereference(fdb->vdev); 4780 WARN_ON(!vxlan); 4781 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4782 vxlan->default_dst.remote_vni); 4783 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4784 if (!hlist_unhashed(&fdb->hlist)) 4785 vxlan_fdb_destroy(vxlan, fdb, false, false); 4786 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4787 } 4788 rcu_read_unlock(); 4789 } 4790 4791 static int vxlan_nexthop_event(struct notifier_block *nb, 4792 unsigned long event, void *ptr) 4793 { 4794 struct nh_notifier_info *info = ptr; 4795 struct nexthop *nh; 4796 4797 if (event != NEXTHOP_EVENT_DEL) 4798 return NOTIFY_DONE; 4799 4800 nh = nexthop_find_by_id(info->net, info->id); 4801 if (!nh) 4802 return NOTIFY_DONE; 4803 4804 vxlan_fdb_nh_flush(nh); 4805 4806 return NOTIFY_DONE; 4807 } 4808 4809 static __net_init int vxlan_init_net(struct net *net) 4810 { 4811 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4812 unsigned int h; 4813 4814 INIT_LIST_HEAD(&vn->vxlan_list); 4815 spin_lock_init(&vn->sock_lock); 4816 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4817 4818 for (h = 0; h < PORT_HASH_SIZE; ++h) 4819 INIT_HLIST_HEAD(&vn->sock_list[h]); 4820 4821 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4822 NULL); 4823 } 4824 4825 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4826 struct list_head *dev_to_kill) 4827 { 4828 struct vxlan_dev *vxlan, *next; 4829 4830 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4831 vxlan_dellink(vxlan->dev, dev_to_kill); 4832 } 4833 4834 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4835 struct list_head *dev_to_kill) 4836 { 4837 struct net *net; 4838 4839 ASSERT_RTNL(); 4840 list_for_each_entry(net, net_list, exit_list) { 4841 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4842 4843 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4844 4845 vxlan_destroy_tunnels(vn, dev_to_kill); 4846 } 4847 } 4848 4849 static void __net_exit vxlan_exit_net(struct net *net) 4850 { 4851 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4852 unsigned int h; 4853 4854 for (h = 0; h < PORT_HASH_SIZE; ++h) 4855 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4856 } 4857 4858 static struct pernet_operations vxlan_net_ops = { 4859 .init = vxlan_init_net, 4860 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4861 .exit = vxlan_exit_net, 4862 .id = &vxlan_net_id, 4863 .size = sizeof(struct vxlan_net), 4864 }; 4865 4866 static int __init vxlan_init_module(void) 4867 { 4868 int rc; 4869 4870 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4871 4872 rc = register_pernet_subsys(&vxlan_net_ops); 4873 if (rc) 4874 goto out1; 4875 4876 rc = register_netdevice_notifier(&vxlan_notifier_block); 4877 if (rc) 4878 goto out2; 4879 4880 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4881 if (rc) 4882 goto out3; 4883 4884 rc = rtnl_link_register(&vxlan_link_ops); 4885 if (rc) 4886 goto out4; 4887 4888 vxlan_vnifilter_init(); 4889 4890 return 0; 4891 out4: 4892 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4893 out3: 4894 unregister_netdevice_notifier(&vxlan_notifier_block); 4895 out2: 4896 unregister_pernet_subsys(&vxlan_net_ops); 4897 out1: 4898 return rc; 4899 } 4900 late_initcall(vxlan_init_module); 4901 4902 static void __exit vxlan_cleanup_module(void) 4903 { 4904 vxlan_vnifilter_uninit(); 4905 rtnl_link_unregister(&vxlan_link_ops); 4906 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4907 unregister_netdevice_notifier(&vxlan_notifier_block); 4908 unregister_pernet_subsys(&vxlan_net_ops); 4909 /* rcu_barrier() is called by netns */ 4910 } 4911 module_exit(vxlan_cleanup_module); 4912 4913 MODULE_LICENSE("GPL"); 4914 MODULE_VERSION(VXLAN_VERSION); 4915 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4916 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4917 MODULE_ALIAS_RTNL_LINK("vxlan"); 4918