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 ++vxlan->dev->stats.rx_frame_errors; 1770 ++vxlan->dev->stats.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->stats.tx_dropped++; 1841 goto out; 1842 } 1843 parp = arp_hdr(skb); 1844 1845 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1846 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1847 parp->ar_pro != htons(ETH_P_IP) || 1848 parp->ar_op != htons(ARPOP_REQUEST) || 1849 parp->ar_hln != dev->addr_len || 1850 parp->ar_pln != 4) 1851 goto out; 1852 arpptr = (u8 *)parp + sizeof(struct arphdr); 1853 sha = arpptr; 1854 arpptr += dev->addr_len; /* sha */ 1855 memcpy(&sip, arpptr, sizeof(sip)); 1856 arpptr += sizeof(sip); 1857 arpptr += dev->addr_len; /* tha */ 1858 memcpy(&tip, arpptr, sizeof(tip)); 1859 1860 if (ipv4_is_loopback(tip) || 1861 ipv4_is_multicast(tip)) 1862 goto out; 1863 1864 n = neigh_lookup(&arp_tbl, &tip, dev); 1865 1866 if (n) { 1867 struct vxlan_fdb *f; 1868 struct sk_buff *reply; 1869 1870 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1871 neigh_release(n); 1872 goto out; 1873 } 1874 1875 f = vxlan_find_mac(vxlan, n->ha, vni); 1876 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1877 /* bridge-local neighbor */ 1878 neigh_release(n); 1879 goto out; 1880 } 1881 1882 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1883 n->ha, sha); 1884 1885 neigh_release(n); 1886 1887 if (reply == NULL) 1888 goto out; 1889 1890 skb_reset_mac_header(reply); 1891 __skb_pull(reply, skb_network_offset(reply)); 1892 reply->ip_summed = CHECKSUM_UNNECESSARY; 1893 reply->pkt_type = PACKET_HOST; 1894 1895 if (netif_rx(reply) == NET_RX_DROP) { 1896 dev->stats.rx_dropped++; 1897 vxlan_vnifilter_count(vxlan, vni, NULL, 1898 VXLAN_VNI_STATS_RX_DROPS, 0); 1899 } 1900 1901 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1902 union vxlan_addr ipa = { 1903 .sin.sin_addr.s_addr = tip, 1904 .sin.sin_family = AF_INET, 1905 }; 1906 1907 vxlan_ip_miss(dev, &ipa); 1908 } 1909 out: 1910 consume_skb(skb); 1911 return NETDEV_TX_OK; 1912 } 1913 1914 #if IS_ENABLED(CONFIG_IPV6) 1915 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1916 struct neighbour *n, bool isrouter) 1917 { 1918 struct net_device *dev = request->dev; 1919 struct sk_buff *reply; 1920 struct nd_msg *ns, *na; 1921 struct ipv6hdr *pip6; 1922 u8 *daddr; 1923 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1924 int ns_olen; 1925 int i, len; 1926 1927 if (dev == NULL || !pskb_may_pull(request, request->len)) 1928 return NULL; 1929 1930 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1931 sizeof(*na) + na_olen + dev->needed_tailroom; 1932 reply = alloc_skb(len, GFP_ATOMIC); 1933 if (reply == NULL) 1934 return NULL; 1935 1936 reply->protocol = htons(ETH_P_IPV6); 1937 reply->dev = dev; 1938 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1939 skb_push(reply, sizeof(struct ethhdr)); 1940 skb_reset_mac_header(reply); 1941 1942 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1943 1944 daddr = eth_hdr(request)->h_source; 1945 ns_olen = request->len - skb_network_offset(request) - 1946 sizeof(struct ipv6hdr) - sizeof(*ns); 1947 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1948 if (!ns->opt[i + 1]) { 1949 kfree_skb(reply); 1950 return NULL; 1951 } 1952 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1953 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1954 break; 1955 } 1956 } 1957 1958 /* Ethernet header */ 1959 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1960 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1961 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1962 reply->protocol = htons(ETH_P_IPV6); 1963 1964 skb_pull(reply, sizeof(struct ethhdr)); 1965 skb_reset_network_header(reply); 1966 skb_put(reply, sizeof(struct ipv6hdr)); 1967 1968 /* IPv6 header */ 1969 1970 pip6 = ipv6_hdr(reply); 1971 memset(pip6, 0, sizeof(struct ipv6hdr)); 1972 pip6->version = 6; 1973 pip6->priority = ipv6_hdr(request)->priority; 1974 pip6->nexthdr = IPPROTO_ICMPV6; 1975 pip6->hop_limit = 255; 1976 pip6->daddr = ipv6_hdr(request)->saddr; 1977 pip6->saddr = *(struct in6_addr *)n->primary_key; 1978 1979 skb_pull(reply, sizeof(struct ipv6hdr)); 1980 skb_reset_transport_header(reply); 1981 1982 /* Neighbor Advertisement */ 1983 na = skb_put_zero(reply, sizeof(*na) + na_olen); 1984 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 1985 na->icmph.icmp6_router = isrouter; 1986 na->icmph.icmp6_override = 1; 1987 na->icmph.icmp6_solicited = 1; 1988 na->target = ns->target; 1989 ether_addr_copy(&na->opt[2], n->ha); 1990 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 1991 na->opt[1] = na_olen >> 3; 1992 1993 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 1994 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 1995 csum_partial(na, sizeof(*na)+na_olen, 0)); 1996 1997 pip6->payload_len = htons(sizeof(*na)+na_olen); 1998 1999 skb_push(reply, sizeof(struct ipv6hdr)); 2000 2001 reply->ip_summed = CHECKSUM_UNNECESSARY; 2002 2003 return reply; 2004 } 2005 2006 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2007 { 2008 struct vxlan_dev *vxlan = netdev_priv(dev); 2009 const struct in6_addr *daddr; 2010 const struct ipv6hdr *iphdr; 2011 struct inet6_dev *in6_dev; 2012 struct neighbour *n; 2013 struct nd_msg *msg; 2014 2015 rcu_read_lock(); 2016 in6_dev = __in6_dev_get(dev); 2017 if (!in6_dev) 2018 goto out; 2019 2020 iphdr = ipv6_hdr(skb); 2021 daddr = &iphdr->daddr; 2022 msg = (struct nd_msg *)(iphdr + 1); 2023 2024 if (ipv6_addr_loopback(daddr) || 2025 ipv6_addr_is_multicast(&msg->target)) 2026 goto out; 2027 2028 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2029 2030 if (n) { 2031 struct vxlan_fdb *f; 2032 struct sk_buff *reply; 2033 2034 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2035 neigh_release(n); 2036 goto out; 2037 } 2038 2039 f = vxlan_find_mac(vxlan, n->ha, vni); 2040 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2041 /* bridge-local neighbor */ 2042 neigh_release(n); 2043 goto out; 2044 } 2045 2046 reply = vxlan_na_create(skb, n, 2047 !!(f ? f->flags & NTF_ROUTER : 0)); 2048 2049 neigh_release(n); 2050 2051 if (reply == NULL) 2052 goto out; 2053 2054 if (netif_rx(reply) == NET_RX_DROP) { 2055 dev->stats.rx_dropped++; 2056 vxlan_vnifilter_count(vxlan, vni, NULL, 2057 VXLAN_VNI_STATS_RX_DROPS, 0); 2058 } 2059 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2060 union vxlan_addr ipa = { 2061 .sin6.sin6_addr = msg->target, 2062 .sin6.sin6_family = AF_INET6, 2063 }; 2064 2065 vxlan_ip_miss(dev, &ipa); 2066 } 2067 2068 out: 2069 rcu_read_unlock(); 2070 consume_skb(skb); 2071 return NETDEV_TX_OK; 2072 } 2073 #endif 2074 2075 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2076 { 2077 struct vxlan_dev *vxlan = netdev_priv(dev); 2078 struct neighbour *n; 2079 2080 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2081 return false; 2082 2083 n = NULL; 2084 switch (ntohs(eth_hdr(skb)->h_proto)) { 2085 case ETH_P_IP: 2086 { 2087 struct iphdr *pip; 2088 2089 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2090 return false; 2091 pip = ip_hdr(skb); 2092 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2093 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2094 union vxlan_addr ipa = { 2095 .sin.sin_addr.s_addr = pip->daddr, 2096 .sin.sin_family = AF_INET, 2097 }; 2098 2099 vxlan_ip_miss(dev, &ipa); 2100 return false; 2101 } 2102 2103 break; 2104 } 2105 #if IS_ENABLED(CONFIG_IPV6) 2106 case ETH_P_IPV6: 2107 { 2108 struct ipv6hdr *pip6; 2109 2110 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2111 return false; 2112 pip6 = ipv6_hdr(skb); 2113 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2114 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2115 union vxlan_addr ipa = { 2116 .sin6.sin6_addr = pip6->daddr, 2117 .sin6.sin6_family = AF_INET6, 2118 }; 2119 2120 vxlan_ip_miss(dev, &ipa); 2121 return false; 2122 } 2123 2124 break; 2125 } 2126 #endif 2127 default: 2128 return false; 2129 } 2130 2131 if (n) { 2132 bool diff; 2133 2134 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2135 if (diff) { 2136 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2137 dev->addr_len); 2138 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2139 } 2140 neigh_release(n); 2141 return diff; 2142 } 2143 2144 return false; 2145 } 2146 2147 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2148 { 2149 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2150 2151 gpe->np_applied = 1; 2152 gpe->next_protocol = tun_p_from_eth_p(protocol); 2153 if (!gpe->next_protocol) 2154 return -EPFNOSUPPORT; 2155 return 0; 2156 } 2157 2158 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2159 int iphdr_len, __be32 vni, 2160 struct vxlan_metadata *md, u32 vxflags, 2161 bool udp_sum) 2162 { 2163 struct vxlanhdr *vxh; 2164 int min_headroom; 2165 int err; 2166 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2167 __be16 inner_protocol = htons(ETH_P_TEB); 2168 2169 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2170 skb->ip_summed == CHECKSUM_PARTIAL) { 2171 int csum_start = skb_checksum_start_offset(skb); 2172 2173 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2174 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2175 (skb->csum_offset == offsetof(struct udphdr, check) || 2176 skb->csum_offset == offsetof(struct tcphdr, check))) 2177 type |= SKB_GSO_TUNNEL_REMCSUM; 2178 } 2179 2180 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2181 + VXLAN_HLEN + iphdr_len; 2182 2183 /* Need space for new headers (invalidates iph ptr) */ 2184 err = skb_cow_head(skb, min_headroom); 2185 if (unlikely(err)) 2186 return err; 2187 2188 err = iptunnel_handle_offloads(skb, type); 2189 if (err) 2190 return err; 2191 2192 vxh = __skb_push(skb, sizeof(*vxh)); 2193 vxh->vx_flags = VXLAN_HF_VNI; 2194 vxh->vx_vni = vxlan_vni_field(vni); 2195 2196 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2197 unsigned int start; 2198 2199 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2200 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2201 vxh->vx_flags |= VXLAN_HF_RCO; 2202 2203 if (!skb_is_gso(skb)) { 2204 skb->ip_summed = CHECKSUM_NONE; 2205 skb->encapsulation = 0; 2206 } 2207 } 2208 2209 if (vxflags & VXLAN_F_GBP) 2210 vxlan_build_gbp_hdr(vxh, md); 2211 if (vxflags & VXLAN_F_GPE) { 2212 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2213 if (err < 0) 2214 return err; 2215 inner_protocol = skb->protocol; 2216 } 2217 2218 skb_set_inner_protocol(skb, inner_protocol); 2219 return 0; 2220 } 2221 2222 /* Bypass encapsulation if the destination is local */ 2223 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2224 struct vxlan_dev *dst_vxlan, __be32 vni, 2225 bool snoop) 2226 { 2227 union vxlan_addr loopback; 2228 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2229 struct net_device *dev; 2230 int len = skb->len; 2231 2232 skb->pkt_type = PACKET_HOST; 2233 skb->encapsulation = 0; 2234 skb->dev = dst_vxlan->dev; 2235 __skb_pull(skb, skb_network_offset(skb)); 2236 2237 if (remote_ip->sa.sa_family == AF_INET) { 2238 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2239 loopback.sa.sa_family = AF_INET; 2240 #if IS_ENABLED(CONFIG_IPV6) 2241 } else { 2242 loopback.sin6.sin6_addr = in6addr_loopback; 2243 loopback.sa.sa_family = AF_INET6; 2244 #endif 2245 } 2246 2247 rcu_read_lock(); 2248 dev = skb->dev; 2249 if (unlikely(!(dev->flags & IFF_UP))) { 2250 kfree_skb(skb); 2251 goto drop; 2252 } 2253 2254 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2255 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2256 2257 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2258 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2259 2260 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2261 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2262 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2263 len); 2264 } else { 2265 drop: 2266 dev->stats.rx_dropped++; 2267 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2268 VXLAN_VNI_STATS_RX_DROPS, 0); 2269 } 2270 rcu_read_unlock(); 2271 } 2272 2273 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2274 struct vxlan_dev *vxlan, 2275 int addr_family, 2276 __be16 dst_port, int dst_ifindex, __be32 vni, 2277 struct dst_entry *dst, 2278 u32 rt_flags) 2279 { 2280 #if IS_ENABLED(CONFIG_IPV6) 2281 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2282 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2283 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2284 */ 2285 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2286 #endif 2287 /* Bypass encapsulation if the destination is local */ 2288 if (rt_flags & RTCF_LOCAL && 2289 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2290 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2291 struct vxlan_dev *dst_vxlan; 2292 2293 dst_release(dst); 2294 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2295 addr_family, dst_port, 2296 vxlan->cfg.flags); 2297 if (!dst_vxlan) { 2298 dev->stats.tx_errors++; 2299 vxlan_vnifilter_count(vxlan, vni, NULL, 2300 VXLAN_VNI_STATS_TX_ERRORS, 0); 2301 kfree_skb(skb); 2302 2303 return -ENOENT; 2304 } 2305 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2306 return 1; 2307 } 2308 2309 return 0; 2310 } 2311 2312 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2313 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2314 { 2315 struct dst_cache *dst_cache; 2316 struct ip_tunnel_info *info; 2317 struct ip_tunnel_key *pkey; 2318 struct ip_tunnel_key key; 2319 struct vxlan_dev *vxlan = netdev_priv(dev); 2320 const struct iphdr *old_iph = ip_hdr(skb); 2321 struct vxlan_metadata _md; 2322 struct vxlan_metadata *md = &_md; 2323 unsigned int pkt_len = skb->len; 2324 __be16 src_port = 0, dst_port; 2325 struct dst_entry *ndst = NULL; 2326 int addr_family; 2327 __u8 tos, ttl; 2328 int ifindex; 2329 int err; 2330 u32 flags = vxlan->cfg.flags; 2331 bool use_cache; 2332 bool udp_sum = false; 2333 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2334 __be32 vni = 0; 2335 2336 info = skb_tunnel_info(skb); 2337 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2338 2339 if (rdst) { 2340 memset(&key, 0, sizeof(key)); 2341 pkey = &key; 2342 2343 if (vxlan_addr_any(&rdst->remote_ip)) { 2344 if (did_rsc) { 2345 /* short-circuited back to local bridge */ 2346 vxlan_encap_bypass(skb, vxlan, vxlan, 2347 default_vni, true); 2348 return; 2349 } 2350 goto drop; 2351 } 2352 2353 addr_family = vxlan->cfg.saddr.sa.sa_family; 2354 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2355 vni = (rdst->remote_vni) ? : default_vni; 2356 ifindex = rdst->remote_ifindex; 2357 2358 if (addr_family == AF_INET) { 2359 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2360 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2361 } else { 2362 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2363 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2364 } 2365 2366 dst_cache = &rdst->dst_cache; 2367 md->gbp = skb->mark; 2368 if (flags & VXLAN_F_TTL_INHERIT) { 2369 ttl = ip_tunnel_get_ttl(old_iph, skb); 2370 } else { 2371 ttl = vxlan->cfg.ttl; 2372 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2373 ttl = 1; 2374 } 2375 tos = vxlan->cfg.tos; 2376 if (tos == 1) 2377 tos = ip_tunnel_get_dsfield(old_iph, skb); 2378 if (tos && !info) 2379 use_cache = false; 2380 2381 if (addr_family == AF_INET) 2382 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2383 else 2384 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2385 #if IS_ENABLED(CONFIG_IPV6) 2386 switch (vxlan->cfg.label_policy) { 2387 case VXLAN_LABEL_FIXED: 2388 key.label = vxlan->cfg.label; 2389 break; 2390 case VXLAN_LABEL_INHERIT: 2391 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2392 break; 2393 default: 2394 DEBUG_NET_WARN_ON_ONCE(1); 2395 goto drop; 2396 } 2397 #endif 2398 } else { 2399 if (!info) { 2400 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2401 dev->name); 2402 goto drop; 2403 } 2404 pkey = &info->key; 2405 addr_family = ip_tunnel_info_af(info); 2406 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2407 vni = tunnel_id_to_key32(info->key.tun_id); 2408 ifindex = 0; 2409 dst_cache = &info->dst_cache; 2410 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) { 2411 if (info->options_len < sizeof(*md)) 2412 goto drop; 2413 md = ip_tunnel_info_opts(info); 2414 } 2415 ttl = info->key.ttl; 2416 tos = info->key.tos; 2417 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 2418 } 2419 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2420 vxlan->cfg.port_max, true); 2421 2422 rcu_read_lock(); 2423 if (addr_family == AF_INET) { 2424 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2425 struct rtable *rt; 2426 __be16 df = 0; 2427 __be32 saddr; 2428 2429 if (!ifindex) 2430 ifindex = sock4->sock->sk->sk_bound_dev_if; 2431 2432 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2433 &saddr, pkey, src_port, dst_port, 2434 tos, use_cache ? dst_cache : NULL); 2435 if (IS_ERR(rt)) { 2436 err = PTR_ERR(rt); 2437 goto tx_error; 2438 } 2439 2440 if (!info) { 2441 /* Bypass encapsulation if the destination is local */ 2442 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2443 dst_port, ifindex, vni, 2444 &rt->dst, rt->rt_flags); 2445 if (err) 2446 goto out_unlock; 2447 2448 if (vxlan->cfg.df == VXLAN_DF_SET) { 2449 df = htons(IP_DF); 2450 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2451 struct ethhdr *eth = eth_hdr(skb); 2452 2453 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2454 (ntohs(eth->h_proto) == ETH_P_IP && 2455 old_iph->frag_off & htons(IP_DF))) 2456 df = htons(IP_DF); 2457 } 2458 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) { 2459 df = htons(IP_DF); 2460 } 2461 2462 ndst = &rt->dst; 2463 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2464 netif_is_any_bridge_port(dev)); 2465 if (err < 0) { 2466 goto tx_error; 2467 } else if (err) { 2468 if (info) { 2469 struct ip_tunnel_info *unclone; 2470 2471 unclone = skb_tunnel_info_unclone(skb); 2472 if (unlikely(!unclone)) 2473 goto tx_error; 2474 2475 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2476 unclone->key.u.ipv4.dst = saddr; 2477 } 2478 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2479 dst_release(ndst); 2480 goto out_unlock; 2481 } 2482 2483 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2484 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2485 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2486 vni, md, flags, udp_sum); 2487 if (err < 0) 2488 goto tx_error; 2489 2490 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2491 pkey->u.ipv4.dst, tos, ttl, df, 2492 src_port, dst_port, xnet, !udp_sum); 2493 #if IS_ENABLED(CONFIG_IPV6) 2494 } else { 2495 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2496 struct in6_addr saddr; 2497 2498 if (!ifindex) 2499 ifindex = sock6->sock->sk->sk_bound_dev_if; 2500 2501 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2502 ifindex, &saddr, pkey, 2503 src_port, dst_port, tos, 2504 use_cache ? dst_cache : NULL); 2505 if (IS_ERR(ndst)) { 2506 err = PTR_ERR(ndst); 2507 ndst = NULL; 2508 goto tx_error; 2509 } 2510 2511 if (!info) { 2512 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; 2513 2514 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2515 dst_port, ifindex, vni, 2516 ndst, rt6i_flags); 2517 if (err) 2518 goto out_unlock; 2519 } 2520 2521 err = skb_tunnel_check_pmtu(skb, ndst, 2522 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2523 netif_is_any_bridge_port(dev)); 2524 if (err < 0) { 2525 goto tx_error; 2526 } else if (err) { 2527 if (info) { 2528 struct ip_tunnel_info *unclone; 2529 2530 unclone = skb_tunnel_info_unclone(skb); 2531 if (unlikely(!unclone)) 2532 goto tx_error; 2533 2534 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2535 unclone->key.u.ipv6.dst = saddr; 2536 } 2537 2538 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2539 dst_release(ndst); 2540 goto out_unlock; 2541 } 2542 2543 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2544 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2545 skb_scrub_packet(skb, xnet); 2546 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2547 vni, md, flags, udp_sum); 2548 if (err < 0) 2549 goto tx_error; 2550 2551 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2552 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2553 pkey->label, src_port, dst_port, !udp_sum); 2554 #endif 2555 } 2556 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2557 out_unlock: 2558 rcu_read_unlock(); 2559 return; 2560 2561 drop: 2562 dev->stats.tx_dropped++; 2563 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2564 dev_kfree_skb(skb); 2565 return; 2566 2567 tx_error: 2568 rcu_read_unlock(); 2569 if (err == -ELOOP) 2570 dev->stats.collisions++; 2571 else if (err == -ENETUNREACH) 2572 dev->stats.tx_carrier_errors++; 2573 dst_release(ndst); 2574 dev->stats.tx_errors++; 2575 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2576 kfree_skb(skb); 2577 } 2578 2579 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2580 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2581 { 2582 struct vxlan_rdst nh_rdst; 2583 struct nexthop *nh; 2584 bool do_xmit; 2585 u32 hash; 2586 2587 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2588 hash = skb_get_hash(skb); 2589 2590 rcu_read_lock(); 2591 nh = rcu_dereference(f->nh); 2592 if (!nh) { 2593 rcu_read_unlock(); 2594 goto drop; 2595 } 2596 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2597 rcu_read_unlock(); 2598 2599 if (likely(do_xmit)) 2600 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2601 else 2602 goto drop; 2603 2604 return; 2605 2606 drop: 2607 dev->stats.tx_dropped++; 2608 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2609 VXLAN_VNI_STATS_TX_DROPS, 0); 2610 dev_kfree_skb(skb); 2611 } 2612 2613 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2614 u32 nhid, __be32 vni) 2615 { 2616 struct vxlan_dev *vxlan = netdev_priv(dev); 2617 struct vxlan_rdst nh_rdst; 2618 struct nexthop *nh; 2619 bool do_xmit; 2620 u32 hash; 2621 2622 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2623 hash = skb_get_hash(skb); 2624 2625 rcu_read_lock(); 2626 nh = nexthop_find_by_id(dev_net(dev), nhid); 2627 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2628 rcu_read_unlock(); 2629 goto drop; 2630 } 2631 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2632 rcu_read_unlock(); 2633 2634 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2635 goto drop; 2636 2637 if (likely(do_xmit)) 2638 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2639 else 2640 goto drop; 2641 2642 return NETDEV_TX_OK; 2643 2644 drop: 2645 dev->stats.tx_dropped++; 2646 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2647 VXLAN_VNI_STATS_TX_DROPS, 0); 2648 dev_kfree_skb(skb); 2649 return NETDEV_TX_OK; 2650 } 2651 2652 /* Transmit local packets over Vxlan 2653 * 2654 * Outer IP header inherits ECN and DF from inner header. 2655 * Outer UDP destination is the VXLAN assigned port. 2656 * source port is based on hash of flow 2657 */ 2658 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2659 { 2660 struct vxlan_dev *vxlan = netdev_priv(dev); 2661 struct vxlan_rdst *rdst, *fdst = NULL; 2662 const struct ip_tunnel_info *info; 2663 bool did_rsc = false; 2664 struct vxlan_fdb *f; 2665 struct ethhdr *eth; 2666 __be32 vni = 0; 2667 u32 nhid = 0; 2668 2669 info = skb_tunnel_info(skb); 2670 2671 skb_reset_mac_header(skb); 2672 2673 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2674 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2675 info->mode & IP_TUNNEL_INFO_TX) { 2676 vni = tunnel_id_to_key32(info->key.tun_id); 2677 nhid = info->key.nhid; 2678 } else { 2679 if (info && info->mode & IP_TUNNEL_INFO_TX) 2680 vxlan_xmit_one(skb, dev, vni, NULL, false); 2681 else 2682 kfree_skb(skb); 2683 return NETDEV_TX_OK; 2684 } 2685 } 2686 2687 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2688 eth = eth_hdr(skb); 2689 if (ntohs(eth->h_proto) == ETH_P_ARP) 2690 return arp_reduce(dev, skb, vni); 2691 #if IS_ENABLED(CONFIG_IPV6) 2692 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2693 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2694 sizeof(struct nd_msg)) && 2695 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2696 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2697 2698 if (m->icmph.icmp6_code == 0 && 2699 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2700 return neigh_reduce(dev, skb, vni); 2701 } 2702 #endif 2703 } 2704 2705 if (nhid) 2706 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2707 2708 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2709 struct vxlan_mdb_entry *mdb_entry; 2710 2711 rcu_read_lock(); 2712 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2713 if (mdb_entry) { 2714 netdev_tx_t ret; 2715 2716 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2717 rcu_read_unlock(); 2718 return ret; 2719 } 2720 rcu_read_unlock(); 2721 } 2722 2723 eth = eth_hdr(skb); 2724 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2725 did_rsc = false; 2726 2727 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2728 (ntohs(eth->h_proto) == ETH_P_IP || 2729 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2730 did_rsc = route_shortcircuit(dev, skb); 2731 if (did_rsc) 2732 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2733 } 2734 2735 if (f == NULL) { 2736 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2737 if (f == NULL) { 2738 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2739 !is_multicast_ether_addr(eth->h_dest)) 2740 vxlan_fdb_miss(vxlan, eth->h_dest); 2741 2742 dev->stats.tx_dropped++; 2743 vxlan_vnifilter_count(vxlan, vni, NULL, 2744 VXLAN_VNI_STATS_TX_DROPS, 0); 2745 kfree_skb(skb); 2746 return NETDEV_TX_OK; 2747 } 2748 } 2749 2750 if (rcu_access_pointer(f->nh)) { 2751 vxlan_xmit_nh(skb, dev, f, 2752 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2753 } else { 2754 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2755 struct sk_buff *skb1; 2756 2757 if (!fdst) { 2758 fdst = rdst; 2759 continue; 2760 } 2761 skb1 = skb_clone(skb, GFP_ATOMIC); 2762 if (skb1) 2763 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2764 } 2765 if (fdst) 2766 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2767 else 2768 kfree_skb(skb); 2769 } 2770 2771 return NETDEV_TX_OK; 2772 } 2773 2774 /* Walk the forwarding table and purge stale entries */ 2775 static void vxlan_cleanup(struct timer_list *t) 2776 { 2777 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2778 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2779 unsigned int h; 2780 2781 if (!netif_running(vxlan->dev)) 2782 return; 2783 2784 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2785 struct hlist_node *p, *n; 2786 2787 spin_lock(&vxlan->hash_lock[h]); 2788 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2789 struct vxlan_fdb *f 2790 = container_of(p, struct vxlan_fdb, hlist); 2791 unsigned long timeout; 2792 2793 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2794 continue; 2795 2796 if (f->flags & NTF_EXT_LEARNED) 2797 continue; 2798 2799 timeout = f->used + vxlan->cfg.age_interval * HZ; 2800 if (time_before_eq(timeout, jiffies)) { 2801 netdev_dbg(vxlan->dev, 2802 "garbage collect %pM\n", 2803 f->eth_addr); 2804 f->state = NUD_STALE; 2805 vxlan_fdb_destroy(vxlan, f, true, true); 2806 } else if (time_before(timeout, next_timer)) 2807 next_timer = timeout; 2808 } 2809 spin_unlock(&vxlan->hash_lock[h]); 2810 } 2811 2812 mod_timer(&vxlan->age_timer, next_timer); 2813 } 2814 2815 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2816 { 2817 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2818 2819 spin_lock(&vn->sock_lock); 2820 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2821 #if IS_ENABLED(CONFIG_IPV6) 2822 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2823 #endif 2824 spin_unlock(&vn->sock_lock); 2825 } 2826 2827 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2828 struct vxlan_dev_node *node) 2829 { 2830 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2831 __be32 vni = vxlan->default_dst.remote_vni; 2832 2833 node->vxlan = vxlan; 2834 spin_lock(&vn->sock_lock); 2835 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2836 spin_unlock(&vn->sock_lock); 2837 } 2838 2839 /* Setup stats when device is created */ 2840 static int vxlan_init(struct net_device *dev) 2841 { 2842 struct vxlan_dev *vxlan = netdev_priv(dev); 2843 int err; 2844 2845 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2846 vxlan_vnigroup_init(vxlan); 2847 2848 err = gro_cells_init(&vxlan->gro_cells, dev); 2849 if (err) 2850 goto err_vnigroup_uninit; 2851 2852 err = vxlan_mdb_init(vxlan); 2853 if (err) 2854 goto err_gro_cells_destroy; 2855 2856 netdev_lockdep_set_classes(dev); 2857 return 0; 2858 2859 err_gro_cells_destroy: 2860 gro_cells_destroy(&vxlan->gro_cells); 2861 err_vnigroup_uninit: 2862 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2863 vxlan_vnigroup_uninit(vxlan); 2864 return err; 2865 } 2866 2867 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2868 { 2869 struct vxlan_fdb *f; 2870 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2871 2872 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2873 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2874 if (f) 2875 vxlan_fdb_destroy(vxlan, f, true, true); 2876 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2877 } 2878 2879 static void vxlan_uninit(struct net_device *dev) 2880 { 2881 struct vxlan_dev *vxlan = netdev_priv(dev); 2882 2883 vxlan_mdb_fini(vxlan); 2884 2885 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2886 vxlan_vnigroup_uninit(vxlan); 2887 2888 gro_cells_destroy(&vxlan->gro_cells); 2889 2890 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2891 } 2892 2893 /* Start ageing timer and join group when device is brought up */ 2894 static int vxlan_open(struct net_device *dev) 2895 { 2896 struct vxlan_dev *vxlan = netdev_priv(dev); 2897 int ret; 2898 2899 ret = vxlan_sock_add(vxlan); 2900 if (ret < 0) 2901 return ret; 2902 2903 ret = vxlan_multicast_join(vxlan); 2904 if (ret) { 2905 vxlan_sock_release(vxlan); 2906 return ret; 2907 } 2908 2909 if (vxlan->cfg.age_interval) 2910 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2911 2912 return ret; 2913 } 2914 2915 struct vxlan_fdb_flush_desc { 2916 bool ignore_default_entry; 2917 unsigned long state; 2918 unsigned long state_mask; 2919 unsigned long flags; 2920 unsigned long flags_mask; 2921 __be32 src_vni; 2922 u32 nhid; 2923 __be32 vni; 2924 __be16 port; 2925 union vxlan_addr dst_ip; 2926 }; 2927 2928 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2929 const struct vxlan_dev *vxlan) 2930 { 2931 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2932 } 2933 2934 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2935 { 2936 struct nexthop *nh = rtnl_dereference(f->nh); 2937 2938 return nh && nh->id == nhid; 2939 } 2940 2941 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2942 const struct vxlan_dev *vxlan, 2943 const struct vxlan_fdb_flush_desc *desc) 2944 { 2945 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2946 return false; 2947 2948 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2949 return false; 2950 2951 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2952 return false; 2953 2954 if (desc->src_vni && f->vni != desc->src_vni) 2955 return false; 2956 2957 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 2958 return false; 2959 2960 return true; 2961 } 2962 2963 static bool 2964 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 2965 { 2966 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 2967 } 2968 2969 static bool 2970 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 2971 const struct vxlan_rdst *rd) 2972 { 2973 if (desc->vni && rd->remote_vni != desc->vni) 2974 return false; 2975 2976 if (desc->port && rd->remote_port != desc->port) 2977 return false; 2978 2979 if (desc->dst_ip.sa.sa_family && 2980 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 2981 return false; 2982 2983 return true; 2984 } 2985 2986 static void 2987 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 2988 const struct vxlan_fdb_flush_desc *desc, 2989 bool *p_destroy_fdb) 2990 { 2991 bool remotes_flushed = false; 2992 struct vxlan_rdst *rd, *tmp; 2993 2994 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 2995 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 2996 continue; 2997 2998 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 2999 remotes_flushed = true; 3000 } 3001 3002 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3003 } 3004 3005 /* Purge the forwarding table */ 3006 static void vxlan_flush(struct vxlan_dev *vxlan, 3007 const struct vxlan_fdb_flush_desc *desc) 3008 { 3009 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3010 unsigned int h; 3011 3012 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3013 struct hlist_node *p, *n; 3014 3015 spin_lock_bh(&vxlan->hash_lock[h]); 3016 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3017 struct vxlan_fdb *f 3018 = container_of(p, struct vxlan_fdb, hlist); 3019 3020 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3021 continue; 3022 3023 if (match_remotes) { 3024 bool destroy_fdb = false; 3025 3026 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3027 &destroy_fdb); 3028 3029 if (!destroy_fdb) 3030 continue; 3031 } 3032 3033 vxlan_fdb_destroy(vxlan, f, true, true); 3034 } 3035 spin_unlock_bh(&vxlan->hash_lock[h]); 3036 } 3037 } 3038 3039 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3040 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3041 [NDA_NH_ID] = { .type = NLA_U32 }, 3042 [NDA_VNI] = { .type = NLA_U32 }, 3043 [NDA_PORT] = { .type = NLA_U16 }, 3044 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3045 sizeof(struct in6_addr)), 3046 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3047 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3048 }; 3049 3050 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3051 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3052 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3053 NTF_ROUTER) 3054 3055 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3056 struct netlink_ext_ack *extack) 3057 { 3058 struct vxlan_dev *vxlan = netdev_priv(dev); 3059 struct vxlan_fdb_flush_desc desc = {}; 3060 struct ndmsg *ndm = nlmsg_data(nlh); 3061 struct nlattr *tb[NDA_MAX + 1]; 3062 u8 ndm_flags; 3063 int err; 3064 3065 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3066 3067 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3068 extack); 3069 if (err) 3070 return err; 3071 3072 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3073 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3074 return -EINVAL; 3075 } 3076 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3077 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3078 return -EINVAL; 3079 } 3080 3081 desc.state = ndm->ndm_state; 3082 desc.flags = ndm_flags; 3083 3084 if (tb[NDA_NDM_STATE_MASK]) 3085 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3086 3087 if (tb[NDA_NDM_FLAGS_MASK]) 3088 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3089 3090 if (tb[NDA_SRC_VNI]) 3091 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3092 3093 if (tb[NDA_NH_ID]) 3094 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3095 3096 if (tb[NDA_VNI]) 3097 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3098 3099 if (tb[NDA_PORT]) 3100 desc.port = nla_get_be16(tb[NDA_PORT]); 3101 3102 if (tb[NDA_DST]) { 3103 union vxlan_addr ip; 3104 3105 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3106 if (err) { 3107 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3108 "Unsupported address family"); 3109 return err; 3110 } 3111 desc.dst_ip = ip; 3112 } 3113 3114 vxlan_flush(vxlan, &desc); 3115 3116 return 0; 3117 } 3118 3119 /* Cleanup timer and forwarding table on shutdown */ 3120 static int vxlan_stop(struct net_device *dev) 3121 { 3122 struct vxlan_dev *vxlan = netdev_priv(dev); 3123 struct vxlan_fdb_flush_desc desc = { 3124 /* Default entry is deleted at vxlan_uninit. */ 3125 .ignore_default_entry = true, 3126 .state = 0, 3127 .state_mask = NUD_PERMANENT | NUD_NOARP, 3128 }; 3129 3130 vxlan_multicast_leave(vxlan); 3131 3132 del_timer_sync(&vxlan->age_timer); 3133 3134 vxlan_flush(vxlan, &desc); 3135 vxlan_sock_release(vxlan); 3136 3137 return 0; 3138 } 3139 3140 /* Stub, nothing needs to be done. */ 3141 static void vxlan_set_multicast_list(struct net_device *dev) 3142 { 3143 } 3144 3145 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3146 { 3147 struct vxlan_dev *vxlan = netdev_priv(dev); 3148 struct vxlan_rdst *dst = &vxlan->default_dst; 3149 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3150 dst->remote_ifindex); 3151 3152 /* This check is different than dev->max_mtu, because it looks at 3153 * the lowerdev->mtu, rather than the static dev->max_mtu 3154 */ 3155 if (lowerdev) { 3156 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3157 if (new_mtu > max_mtu) 3158 return -EINVAL; 3159 } 3160 3161 dev->mtu = new_mtu; 3162 return 0; 3163 } 3164 3165 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3166 { 3167 struct vxlan_dev *vxlan = netdev_priv(dev); 3168 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3169 __be16 sport, dport; 3170 3171 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3172 vxlan->cfg.port_max, true); 3173 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3174 3175 if (ip_tunnel_info_af(info) == AF_INET) { 3176 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3177 struct rtable *rt; 3178 3179 if (!sock4) 3180 return -EIO; 3181 3182 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3183 &info->key.u.ipv4.src, 3184 &info->key, 3185 sport, dport, info->key.tos, 3186 &info->dst_cache); 3187 if (IS_ERR(rt)) 3188 return PTR_ERR(rt); 3189 ip_rt_put(rt); 3190 } else { 3191 #if IS_ENABLED(CONFIG_IPV6) 3192 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3193 struct dst_entry *ndst; 3194 3195 if (!sock6) 3196 return -EIO; 3197 3198 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3199 0, &info->key.u.ipv6.src, 3200 &info->key, 3201 sport, dport, info->key.tos, 3202 &info->dst_cache); 3203 if (IS_ERR(ndst)) 3204 return PTR_ERR(ndst); 3205 dst_release(ndst); 3206 #else /* !CONFIG_IPV6 */ 3207 return -EPFNOSUPPORT; 3208 #endif 3209 } 3210 info->key.tp_src = sport; 3211 info->key.tp_dst = dport; 3212 return 0; 3213 } 3214 3215 static const struct net_device_ops vxlan_netdev_ether_ops = { 3216 .ndo_init = vxlan_init, 3217 .ndo_uninit = vxlan_uninit, 3218 .ndo_open = vxlan_open, 3219 .ndo_stop = vxlan_stop, 3220 .ndo_start_xmit = vxlan_xmit, 3221 .ndo_set_rx_mode = vxlan_set_multicast_list, 3222 .ndo_change_mtu = vxlan_change_mtu, 3223 .ndo_validate_addr = eth_validate_addr, 3224 .ndo_set_mac_address = eth_mac_addr, 3225 .ndo_fdb_add = vxlan_fdb_add, 3226 .ndo_fdb_del = vxlan_fdb_delete, 3227 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3228 .ndo_fdb_dump = vxlan_fdb_dump, 3229 .ndo_fdb_get = vxlan_fdb_get, 3230 .ndo_mdb_add = vxlan_mdb_add, 3231 .ndo_mdb_del = vxlan_mdb_del, 3232 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3233 .ndo_mdb_dump = vxlan_mdb_dump, 3234 .ndo_mdb_get = vxlan_mdb_get, 3235 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3236 }; 3237 3238 static const struct net_device_ops vxlan_netdev_raw_ops = { 3239 .ndo_init = vxlan_init, 3240 .ndo_uninit = vxlan_uninit, 3241 .ndo_open = vxlan_open, 3242 .ndo_stop = vxlan_stop, 3243 .ndo_start_xmit = vxlan_xmit, 3244 .ndo_change_mtu = vxlan_change_mtu, 3245 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3246 }; 3247 3248 /* Info for udev, that this is a virtual tunnel endpoint */ 3249 static const struct device_type vxlan_type = { 3250 .name = "vxlan", 3251 }; 3252 3253 /* Calls the ndo_udp_tunnel_add of the caller in order to 3254 * supply the listening VXLAN udp ports. Callers are expected 3255 * to implement the ndo_udp_tunnel_add. 3256 */ 3257 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3258 { 3259 struct vxlan_sock *vs; 3260 struct net *net = dev_net(dev); 3261 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3262 unsigned int i; 3263 3264 spin_lock(&vn->sock_lock); 3265 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3266 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3267 unsigned short type; 3268 3269 if (vs->flags & VXLAN_F_GPE) 3270 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3271 else 3272 type = UDP_TUNNEL_TYPE_VXLAN; 3273 3274 if (push) 3275 udp_tunnel_push_rx_port(dev, vs->sock, type); 3276 else 3277 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3278 } 3279 } 3280 spin_unlock(&vn->sock_lock); 3281 } 3282 3283 /* Initialize the device structure. */ 3284 static void vxlan_setup(struct net_device *dev) 3285 { 3286 struct vxlan_dev *vxlan = netdev_priv(dev); 3287 unsigned int h; 3288 3289 eth_hw_addr_random(dev); 3290 ether_setup(dev); 3291 3292 dev->needs_free_netdev = true; 3293 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3294 3295 dev->features |= NETIF_F_LLTX; 3296 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3297 dev->features |= NETIF_F_RXCSUM; 3298 dev->features |= NETIF_F_GSO_SOFTWARE; 3299 3300 dev->vlan_features = dev->features; 3301 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3302 dev->hw_features |= NETIF_F_RXCSUM; 3303 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3304 netif_keep_dst(dev); 3305 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN; 3306 3307 /* MTU range: 68 - 65535 */ 3308 dev->min_mtu = ETH_MIN_MTU; 3309 dev->max_mtu = ETH_MAX_MTU; 3310 3311 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3312 INIT_LIST_HEAD(&vxlan->next); 3313 3314 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3315 3316 vxlan->dev = dev; 3317 3318 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3319 spin_lock_init(&vxlan->hash_lock[h]); 3320 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3321 } 3322 } 3323 3324 static void vxlan_ether_setup(struct net_device *dev) 3325 { 3326 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3327 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3328 dev->netdev_ops = &vxlan_netdev_ether_ops; 3329 } 3330 3331 static void vxlan_raw_setup(struct net_device *dev) 3332 { 3333 dev->header_ops = NULL; 3334 dev->type = ARPHRD_NONE; 3335 dev->hard_header_len = 0; 3336 dev->addr_len = 0; 3337 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3338 dev->netdev_ops = &vxlan_netdev_raw_ops; 3339 } 3340 3341 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3342 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3343 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3344 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3345 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3346 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3347 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3348 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3349 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3350 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3351 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3352 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3353 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3354 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3355 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3356 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3357 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3358 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3359 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3360 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3361 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3362 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3363 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3364 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3365 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3366 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3367 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3368 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3369 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3370 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3371 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3372 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3373 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3374 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3375 }; 3376 3377 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3378 struct netlink_ext_ack *extack) 3379 { 3380 if (tb[IFLA_ADDRESS]) { 3381 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3382 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3383 "Provided link layer address is not Ethernet"); 3384 return -EINVAL; 3385 } 3386 3387 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3388 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3389 "Provided Ethernet address is not unicast"); 3390 return -EADDRNOTAVAIL; 3391 } 3392 } 3393 3394 if (tb[IFLA_MTU]) { 3395 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3396 3397 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3398 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3399 "MTU must be between 68 and 65535"); 3400 return -EINVAL; 3401 } 3402 } 3403 3404 if (!data) { 3405 NL_SET_ERR_MSG(extack, 3406 "Required attributes not provided to perform the operation"); 3407 return -EINVAL; 3408 } 3409 3410 if (data[IFLA_VXLAN_ID]) { 3411 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3412 3413 if (id >= VXLAN_N_VID) { 3414 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3415 "VXLAN ID must be lower than 16777216"); 3416 return -ERANGE; 3417 } 3418 } 3419 3420 if (data[IFLA_VXLAN_PORT_RANGE]) { 3421 const struct ifla_vxlan_port_range *p 3422 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3423 3424 if (ntohs(p->high) < ntohs(p->low)) { 3425 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3426 "Invalid source port range"); 3427 return -EINVAL; 3428 } 3429 } 3430 3431 if (data[IFLA_VXLAN_DF]) { 3432 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3433 3434 if (df < 0 || df > VXLAN_DF_MAX) { 3435 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3436 "Invalid DF attribute"); 3437 return -EINVAL; 3438 } 3439 } 3440 3441 return 0; 3442 } 3443 3444 static void vxlan_get_drvinfo(struct net_device *netdev, 3445 struct ethtool_drvinfo *drvinfo) 3446 { 3447 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3448 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3449 } 3450 3451 static int vxlan_get_link_ksettings(struct net_device *dev, 3452 struct ethtool_link_ksettings *cmd) 3453 { 3454 struct vxlan_dev *vxlan = netdev_priv(dev); 3455 struct vxlan_rdst *dst = &vxlan->default_dst; 3456 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3457 dst->remote_ifindex); 3458 3459 if (!lowerdev) { 3460 cmd->base.duplex = DUPLEX_UNKNOWN; 3461 cmd->base.port = PORT_OTHER; 3462 cmd->base.speed = SPEED_UNKNOWN; 3463 3464 return 0; 3465 } 3466 3467 return __ethtool_get_link_ksettings(lowerdev, cmd); 3468 } 3469 3470 static const struct ethtool_ops vxlan_ethtool_ops = { 3471 .get_drvinfo = vxlan_get_drvinfo, 3472 .get_link = ethtool_op_get_link, 3473 .get_link_ksettings = vxlan_get_link_ksettings, 3474 }; 3475 3476 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3477 __be16 port, u32 flags, int ifindex) 3478 { 3479 struct socket *sock; 3480 struct udp_port_cfg udp_conf; 3481 int err; 3482 3483 memset(&udp_conf, 0, sizeof(udp_conf)); 3484 3485 if (ipv6) { 3486 udp_conf.family = AF_INET6; 3487 udp_conf.use_udp6_rx_checksums = 3488 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3489 udp_conf.ipv6_v6only = 1; 3490 } else { 3491 udp_conf.family = AF_INET; 3492 } 3493 3494 udp_conf.local_udp_port = port; 3495 udp_conf.bind_ifindex = ifindex; 3496 3497 /* Open UDP socket */ 3498 err = udp_sock_create(net, &udp_conf, &sock); 3499 if (err < 0) 3500 return ERR_PTR(err); 3501 3502 udp_allow_gso(sock->sk); 3503 return sock; 3504 } 3505 3506 /* Create new listen socket if needed */ 3507 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3508 __be16 port, u32 flags, 3509 int ifindex) 3510 { 3511 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3512 struct vxlan_sock *vs; 3513 struct socket *sock; 3514 unsigned int h; 3515 struct udp_tunnel_sock_cfg tunnel_cfg; 3516 3517 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3518 if (!vs) 3519 return ERR_PTR(-ENOMEM); 3520 3521 for (h = 0; h < VNI_HASH_SIZE; ++h) 3522 INIT_HLIST_HEAD(&vs->vni_list[h]); 3523 3524 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3525 if (IS_ERR(sock)) { 3526 kfree(vs); 3527 return ERR_CAST(sock); 3528 } 3529 3530 vs->sock = sock; 3531 refcount_set(&vs->refcnt, 1); 3532 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3533 3534 spin_lock(&vn->sock_lock); 3535 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3536 udp_tunnel_notify_add_rx_port(sock, 3537 (vs->flags & VXLAN_F_GPE) ? 3538 UDP_TUNNEL_TYPE_VXLAN_GPE : 3539 UDP_TUNNEL_TYPE_VXLAN); 3540 spin_unlock(&vn->sock_lock); 3541 3542 /* Mark socket as an encapsulation socket. */ 3543 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3544 tunnel_cfg.sk_user_data = vs; 3545 tunnel_cfg.encap_type = 1; 3546 tunnel_cfg.encap_rcv = vxlan_rcv; 3547 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3548 tunnel_cfg.encap_destroy = NULL; 3549 if (vs->flags & VXLAN_F_GPE) { 3550 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3551 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3552 } else { 3553 tunnel_cfg.gro_receive = vxlan_gro_receive; 3554 tunnel_cfg.gro_complete = vxlan_gro_complete; 3555 } 3556 3557 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3558 3559 return vs; 3560 } 3561 3562 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3563 { 3564 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3565 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3566 struct vxlan_sock *vs = NULL; 3567 struct vxlan_dev_node *node; 3568 int l3mdev_index = 0; 3569 3570 if (vxlan->cfg.remote_ifindex) 3571 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3572 vxlan->net, vxlan->cfg.remote_ifindex); 3573 3574 if (!vxlan->cfg.no_share) { 3575 spin_lock(&vn->sock_lock); 3576 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3577 vxlan->cfg.dst_port, vxlan->cfg.flags, 3578 l3mdev_index); 3579 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3580 spin_unlock(&vn->sock_lock); 3581 return -EBUSY; 3582 } 3583 spin_unlock(&vn->sock_lock); 3584 } 3585 if (!vs) 3586 vs = vxlan_socket_create(vxlan->net, ipv6, 3587 vxlan->cfg.dst_port, vxlan->cfg.flags, 3588 l3mdev_index); 3589 if (IS_ERR(vs)) 3590 return PTR_ERR(vs); 3591 #if IS_ENABLED(CONFIG_IPV6) 3592 if (ipv6) { 3593 rcu_assign_pointer(vxlan->vn6_sock, vs); 3594 node = &vxlan->hlist6; 3595 } else 3596 #endif 3597 { 3598 rcu_assign_pointer(vxlan->vn4_sock, vs); 3599 node = &vxlan->hlist4; 3600 } 3601 3602 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3603 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3604 else 3605 vxlan_vs_add_dev(vs, vxlan, node); 3606 3607 return 0; 3608 } 3609 3610 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3611 { 3612 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3613 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3614 bool ipv4 = !ipv6 || metadata; 3615 int ret = 0; 3616 3617 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3618 #if IS_ENABLED(CONFIG_IPV6) 3619 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3620 if (ipv6) { 3621 ret = __vxlan_sock_add(vxlan, true); 3622 if (ret < 0 && ret != -EAFNOSUPPORT) 3623 ipv4 = false; 3624 } 3625 #endif 3626 if (ipv4) 3627 ret = __vxlan_sock_add(vxlan, false); 3628 if (ret < 0) 3629 vxlan_sock_release(vxlan); 3630 return ret; 3631 } 3632 3633 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3634 struct vxlan_config *conf, __be32 vni) 3635 { 3636 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3637 struct vxlan_dev *tmp; 3638 3639 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3640 if (tmp == vxlan) 3641 continue; 3642 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3643 if (!vxlan_vnifilter_lookup(tmp, vni)) 3644 continue; 3645 } else if (tmp->cfg.vni != vni) { 3646 continue; 3647 } 3648 if (tmp->cfg.dst_port != conf->dst_port) 3649 continue; 3650 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3651 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3652 continue; 3653 3654 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3655 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3656 continue; 3657 3658 return -EEXIST; 3659 } 3660 3661 return 0; 3662 } 3663 3664 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3665 struct net_device **lower, 3666 struct vxlan_dev *old, 3667 struct netlink_ext_ack *extack) 3668 { 3669 bool use_ipv6 = false; 3670 3671 if (conf->flags & VXLAN_F_GPE) { 3672 /* For now, allow GPE only together with 3673 * COLLECT_METADATA. This can be relaxed later; in such 3674 * case, the other side of the PtP link will have to be 3675 * provided. 3676 */ 3677 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3678 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3679 NL_SET_ERR_MSG(extack, 3680 "VXLAN GPE does not support this combination of attributes"); 3681 return -EINVAL; 3682 } 3683 } 3684 3685 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3686 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3687 conf->remote_ip.sa.sa_family = AF_INET; 3688 conf->saddr.sa.sa_family = AF_INET; 3689 } else if (!conf->remote_ip.sa.sa_family) { 3690 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3691 } else if (!conf->saddr.sa.sa_family) { 3692 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3693 } 3694 3695 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3696 NL_SET_ERR_MSG(extack, 3697 "Local and remote address must be from the same family"); 3698 return -EINVAL; 3699 } 3700 3701 if (vxlan_addr_multicast(&conf->saddr)) { 3702 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3703 return -EINVAL; 3704 } 3705 3706 if (conf->saddr.sa.sa_family == AF_INET6) { 3707 if (!IS_ENABLED(CONFIG_IPV6)) { 3708 NL_SET_ERR_MSG(extack, 3709 "IPv6 support not enabled in the kernel"); 3710 return -EPFNOSUPPORT; 3711 } 3712 use_ipv6 = true; 3713 conf->flags |= VXLAN_F_IPV6; 3714 3715 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3716 int local_type = 3717 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3718 int remote_type = 3719 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3720 3721 if (local_type & IPV6_ADDR_LINKLOCAL) { 3722 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3723 (remote_type != IPV6_ADDR_ANY)) { 3724 NL_SET_ERR_MSG(extack, 3725 "Invalid combination of local and remote address scopes"); 3726 return -EINVAL; 3727 } 3728 3729 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3730 } else { 3731 if (remote_type == 3732 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3733 NL_SET_ERR_MSG(extack, 3734 "Invalid combination of local and remote address scopes"); 3735 return -EINVAL; 3736 } 3737 3738 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3739 } 3740 } 3741 } 3742 3743 if (conf->label && !use_ipv6) { 3744 NL_SET_ERR_MSG(extack, 3745 "Label attribute only applies to IPv6 VXLAN devices"); 3746 return -EINVAL; 3747 } 3748 3749 if (conf->label_policy && !use_ipv6) { 3750 NL_SET_ERR_MSG(extack, 3751 "Label policy only applies to IPv6 VXLAN devices"); 3752 return -EINVAL; 3753 } 3754 3755 if (conf->remote_ifindex) { 3756 struct net_device *lowerdev; 3757 3758 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3759 if (!lowerdev) { 3760 NL_SET_ERR_MSG(extack, 3761 "Invalid local interface, device not found"); 3762 return -ENODEV; 3763 } 3764 3765 #if IS_ENABLED(CONFIG_IPV6) 3766 if (use_ipv6) { 3767 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3768 3769 if (idev && idev->cnf.disable_ipv6) { 3770 NL_SET_ERR_MSG(extack, 3771 "IPv6 support disabled by administrator"); 3772 return -EPERM; 3773 } 3774 } 3775 #endif 3776 3777 *lower = lowerdev; 3778 } else { 3779 if (vxlan_addr_multicast(&conf->remote_ip)) { 3780 NL_SET_ERR_MSG(extack, 3781 "Local interface required for multicast remote destination"); 3782 3783 return -EINVAL; 3784 } 3785 3786 #if IS_ENABLED(CONFIG_IPV6) 3787 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3788 NL_SET_ERR_MSG(extack, 3789 "Local interface required for link-local local/remote addresses"); 3790 return -EINVAL; 3791 } 3792 #endif 3793 3794 *lower = NULL; 3795 } 3796 3797 if (!conf->dst_port) { 3798 if (conf->flags & VXLAN_F_GPE) 3799 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3800 else 3801 conf->dst_port = htons(vxlan_port); 3802 } 3803 3804 if (!conf->age_interval) 3805 conf->age_interval = FDB_AGE_DEFAULT; 3806 3807 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3808 NL_SET_ERR_MSG(extack, 3809 "A VXLAN device with the specified VNI already exists"); 3810 return -EEXIST; 3811 } 3812 3813 return 0; 3814 } 3815 3816 static void vxlan_config_apply(struct net_device *dev, 3817 struct vxlan_config *conf, 3818 struct net_device *lowerdev, 3819 struct net *src_net, 3820 bool changelink) 3821 { 3822 struct vxlan_dev *vxlan = netdev_priv(dev); 3823 struct vxlan_rdst *dst = &vxlan->default_dst; 3824 unsigned short needed_headroom = ETH_HLEN; 3825 int max_mtu = ETH_MAX_MTU; 3826 u32 flags = conf->flags; 3827 3828 if (!changelink) { 3829 if (flags & VXLAN_F_GPE) 3830 vxlan_raw_setup(dev); 3831 else 3832 vxlan_ether_setup(dev); 3833 3834 if (conf->mtu) 3835 dev->mtu = conf->mtu; 3836 3837 vxlan->net = src_net; 3838 } 3839 3840 dst->remote_vni = conf->vni; 3841 3842 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3843 3844 if (lowerdev) { 3845 dst->remote_ifindex = conf->remote_ifindex; 3846 3847 netif_inherit_tso_max(dev, lowerdev); 3848 3849 needed_headroom = lowerdev->hard_header_len; 3850 needed_headroom += lowerdev->needed_headroom; 3851 3852 dev->needed_tailroom = lowerdev->needed_tailroom; 3853 3854 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3855 if (max_mtu < ETH_MIN_MTU) 3856 max_mtu = ETH_MIN_MTU; 3857 3858 if (!changelink && !conf->mtu) 3859 dev->mtu = max_mtu; 3860 } 3861 3862 if (dev->mtu > max_mtu) 3863 dev->mtu = max_mtu; 3864 3865 if (flags & VXLAN_F_COLLECT_METADATA) 3866 flags |= VXLAN_F_IPV6; 3867 needed_headroom += vxlan_headroom(flags); 3868 dev->needed_headroom = needed_headroom; 3869 3870 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3871 } 3872 3873 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3874 struct vxlan_config *conf, bool changelink, 3875 struct netlink_ext_ack *extack) 3876 { 3877 struct vxlan_dev *vxlan = netdev_priv(dev); 3878 struct net_device *lowerdev; 3879 int ret; 3880 3881 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3882 if (ret) 3883 return ret; 3884 3885 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3886 3887 return 0; 3888 } 3889 3890 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3891 struct vxlan_config *conf, 3892 struct netlink_ext_ack *extack) 3893 { 3894 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3895 struct vxlan_dev *vxlan = netdev_priv(dev); 3896 struct net_device *remote_dev = NULL; 3897 struct vxlan_fdb *f = NULL; 3898 bool unregister = false; 3899 struct vxlan_rdst *dst; 3900 int err; 3901 3902 dst = &vxlan->default_dst; 3903 err = vxlan_dev_configure(net, dev, conf, false, extack); 3904 if (err) 3905 return err; 3906 3907 dev->ethtool_ops = &vxlan_ethtool_ops; 3908 3909 /* create an fdb entry for a valid default destination */ 3910 if (!vxlan_addr_any(&dst->remote_ip)) { 3911 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3912 &dst->remote_ip, 3913 NUD_REACHABLE | NUD_PERMANENT, 3914 vxlan->cfg.dst_port, 3915 dst->remote_vni, 3916 dst->remote_vni, 3917 dst->remote_ifindex, 3918 NTF_SELF, 0, &f, extack); 3919 if (err) 3920 return err; 3921 } 3922 3923 err = register_netdevice(dev); 3924 if (err) 3925 goto errout; 3926 unregister = true; 3927 3928 if (dst->remote_ifindex) { 3929 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3930 if (!remote_dev) { 3931 err = -ENODEV; 3932 goto errout; 3933 } 3934 3935 err = netdev_upper_dev_link(remote_dev, dev, extack); 3936 if (err) 3937 goto errout; 3938 } 3939 3940 err = rtnl_configure_link(dev, NULL, 0, NULL); 3941 if (err < 0) 3942 goto unlink; 3943 3944 if (f) { 3945 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3946 3947 /* notify default fdb entry */ 3948 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3949 RTM_NEWNEIGH, true, extack); 3950 if (err) { 3951 vxlan_fdb_destroy(vxlan, f, false, false); 3952 if (remote_dev) 3953 netdev_upper_dev_unlink(remote_dev, dev); 3954 goto unregister; 3955 } 3956 } 3957 3958 list_add(&vxlan->next, &vn->vxlan_list); 3959 if (remote_dev) 3960 dst->remote_dev = remote_dev; 3961 return 0; 3962 unlink: 3963 if (remote_dev) 3964 netdev_upper_dev_unlink(remote_dev, dev); 3965 errout: 3966 /* unregister_netdevice() destroys the default FDB entry with deletion 3967 * notification. But the addition notification was not sent yet, so 3968 * destroy the entry by hand here. 3969 */ 3970 if (f) 3971 __vxlan_fdb_free(f); 3972 unregister: 3973 if (unregister) 3974 unregister_netdevice(dev); 3975 return err; 3976 } 3977 3978 /* Set/clear flags based on attribute */ 3979 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 3980 int attrtype, unsigned long mask, bool changelink, 3981 bool changelink_supported, 3982 struct netlink_ext_ack *extack) 3983 { 3984 unsigned long flags; 3985 3986 if (!tb[attrtype]) 3987 return 0; 3988 3989 if (changelink && !changelink_supported) { 3990 vxlan_flag_attr_error(attrtype, extack); 3991 return -EOPNOTSUPP; 3992 } 3993 3994 if (vxlan_policy[attrtype].type == NLA_FLAG) 3995 flags = conf->flags | mask; 3996 else if (nla_get_u8(tb[attrtype])) 3997 flags = conf->flags | mask; 3998 else 3999 flags = conf->flags & ~mask; 4000 4001 conf->flags = flags; 4002 4003 return 0; 4004 } 4005 4006 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4007 struct net_device *dev, struct vxlan_config *conf, 4008 bool changelink, struct netlink_ext_ack *extack) 4009 { 4010 struct vxlan_dev *vxlan = netdev_priv(dev); 4011 int err = 0; 4012 4013 memset(conf, 0, sizeof(*conf)); 4014 4015 /* if changelink operation, start with old existing cfg */ 4016 if (changelink) 4017 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4018 4019 if (data[IFLA_VXLAN_ID]) { 4020 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4021 4022 if (changelink && (vni != conf->vni)) { 4023 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4024 return -EOPNOTSUPP; 4025 } 4026 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4027 } 4028 4029 if (data[IFLA_VXLAN_GROUP]) { 4030 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4031 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4032 return -EOPNOTSUPP; 4033 } 4034 4035 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4036 conf->remote_ip.sa.sa_family = AF_INET; 4037 } else if (data[IFLA_VXLAN_GROUP6]) { 4038 if (!IS_ENABLED(CONFIG_IPV6)) { 4039 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4040 return -EPFNOSUPPORT; 4041 } 4042 4043 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4044 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4045 return -EOPNOTSUPP; 4046 } 4047 4048 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4049 conf->remote_ip.sa.sa_family = AF_INET6; 4050 } 4051 4052 if (data[IFLA_VXLAN_LOCAL]) { 4053 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4054 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4055 return -EOPNOTSUPP; 4056 } 4057 4058 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4059 conf->saddr.sa.sa_family = AF_INET; 4060 } else if (data[IFLA_VXLAN_LOCAL6]) { 4061 if (!IS_ENABLED(CONFIG_IPV6)) { 4062 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4063 return -EPFNOSUPPORT; 4064 } 4065 4066 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4067 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4068 return -EOPNOTSUPP; 4069 } 4070 4071 /* TODO: respect scope id */ 4072 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4073 conf->saddr.sa.sa_family = AF_INET6; 4074 } 4075 4076 if (data[IFLA_VXLAN_LINK]) 4077 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4078 4079 if (data[IFLA_VXLAN_TOS]) 4080 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4081 4082 if (data[IFLA_VXLAN_TTL]) 4083 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4084 4085 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4086 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4087 VXLAN_F_TTL_INHERIT, changelink, false, 4088 extack); 4089 if (err) 4090 return err; 4091 4092 } 4093 4094 if (data[IFLA_VXLAN_LABEL]) 4095 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4096 IPV6_FLOWLABEL_MASK; 4097 if (data[IFLA_VXLAN_LABEL_POLICY]) 4098 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4099 4100 if (data[IFLA_VXLAN_LEARNING]) { 4101 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4102 VXLAN_F_LEARN, changelink, true, 4103 extack); 4104 if (err) 4105 return err; 4106 } else if (!changelink) { 4107 /* default to learn on a new device */ 4108 conf->flags |= VXLAN_F_LEARN; 4109 } 4110 4111 if (data[IFLA_VXLAN_AGEING]) 4112 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4113 4114 if (data[IFLA_VXLAN_PROXY]) { 4115 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4116 VXLAN_F_PROXY, changelink, false, 4117 extack); 4118 if (err) 4119 return err; 4120 } 4121 4122 if (data[IFLA_VXLAN_RSC]) { 4123 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4124 VXLAN_F_RSC, changelink, false, 4125 extack); 4126 if (err) 4127 return err; 4128 } 4129 4130 if (data[IFLA_VXLAN_L2MISS]) { 4131 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4132 VXLAN_F_L2MISS, changelink, false, 4133 extack); 4134 if (err) 4135 return err; 4136 } 4137 4138 if (data[IFLA_VXLAN_L3MISS]) { 4139 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4140 VXLAN_F_L3MISS, changelink, false, 4141 extack); 4142 if (err) 4143 return err; 4144 } 4145 4146 if (data[IFLA_VXLAN_LIMIT]) { 4147 if (changelink) { 4148 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4149 "Cannot change limit"); 4150 return -EOPNOTSUPP; 4151 } 4152 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4153 } 4154 4155 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4156 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4157 VXLAN_F_COLLECT_METADATA, changelink, false, 4158 extack); 4159 if (err) 4160 return err; 4161 } 4162 4163 if (data[IFLA_VXLAN_PORT_RANGE]) { 4164 if (!changelink) { 4165 const struct ifla_vxlan_port_range *p 4166 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4167 conf->port_min = ntohs(p->low); 4168 conf->port_max = ntohs(p->high); 4169 } else { 4170 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4171 "Cannot change port range"); 4172 return -EOPNOTSUPP; 4173 } 4174 } 4175 4176 if (data[IFLA_VXLAN_PORT]) { 4177 if (changelink) { 4178 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4179 "Cannot change port"); 4180 return -EOPNOTSUPP; 4181 } 4182 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4183 } 4184 4185 if (data[IFLA_VXLAN_UDP_CSUM]) { 4186 if (changelink) { 4187 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4188 "Cannot change UDP_CSUM flag"); 4189 return -EOPNOTSUPP; 4190 } 4191 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4192 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4193 } 4194 4195 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4196 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4197 VXLAN_F_LOCALBYPASS, changelink, 4198 true, extack); 4199 if (err) 4200 return err; 4201 } else if (!changelink) { 4202 /* default to local bypass on a new device */ 4203 conf->flags |= VXLAN_F_LOCALBYPASS; 4204 } 4205 4206 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4207 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4208 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4209 false, extack); 4210 if (err) 4211 return err; 4212 } 4213 4214 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4215 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4216 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4217 false, extack); 4218 if (err) 4219 return err; 4220 } 4221 4222 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4223 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4224 VXLAN_F_REMCSUM_TX, changelink, false, 4225 extack); 4226 if (err) 4227 return err; 4228 } 4229 4230 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4231 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4232 VXLAN_F_REMCSUM_RX, changelink, false, 4233 extack); 4234 if (err) 4235 return err; 4236 } 4237 4238 if (data[IFLA_VXLAN_GBP]) { 4239 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4240 VXLAN_F_GBP, changelink, false, extack); 4241 if (err) 4242 return err; 4243 } 4244 4245 if (data[IFLA_VXLAN_GPE]) { 4246 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4247 VXLAN_F_GPE, changelink, false, 4248 extack); 4249 if (err) 4250 return err; 4251 } 4252 4253 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4254 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4255 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4256 false, extack); 4257 if (err) 4258 return err; 4259 } 4260 4261 if (tb[IFLA_MTU]) { 4262 if (changelink) { 4263 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4264 "Cannot change mtu"); 4265 return -EOPNOTSUPP; 4266 } 4267 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4268 } 4269 4270 if (data[IFLA_VXLAN_DF]) 4271 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4272 4273 if (data[IFLA_VXLAN_VNIFILTER]) { 4274 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4275 VXLAN_F_VNIFILTER, changelink, false, 4276 extack); 4277 if (err) 4278 return err; 4279 4280 if ((conf->flags & VXLAN_F_VNIFILTER) && 4281 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4282 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4283 "vxlan vnifilter only valid in collect metadata mode"); 4284 return -EINVAL; 4285 } 4286 } 4287 4288 return 0; 4289 } 4290 4291 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4292 struct nlattr *tb[], struct nlattr *data[], 4293 struct netlink_ext_ack *extack) 4294 { 4295 struct vxlan_config conf; 4296 int err; 4297 4298 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4299 if (err) 4300 return err; 4301 4302 return __vxlan_dev_create(src_net, dev, &conf, extack); 4303 } 4304 4305 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4306 struct nlattr *data[], 4307 struct netlink_ext_ack *extack) 4308 { 4309 struct vxlan_dev *vxlan = netdev_priv(dev); 4310 struct net_device *lowerdev; 4311 struct vxlan_config conf; 4312 struct vxlan_rdst *dst; 4313 int err; 4314 4315 dst = &vxlan->default_dst; 4316 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4317 if (err) 4318 return err; 4319 4320 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4321 vxlan, extack); 4322 if (err) 4323 return err; 4324 4325 if (dst->remote_dev == lowerdev) 4326 lowerdev = NULL; 4327 4328 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4329 extack); 4330 if (err) 4331 return err; 4332 4333 /* handle default dst entry */ 4334 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4335 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4336 4337 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4338 if (!vxlan_addr_any(&conf.remote_ip)) { 4339 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4340 &conf.remote_ip, 4341 NUD_REACHABLE | NUD_PERMANENT, 4342 NLM_F_APPEND | NLM_F_CREATE, 4343 vxlan->cfg.dst_port, 4344 conf.vni, conf.vni, 4345 conf.remote_ifindex, 4346 NTF_SELF, 0, true, extack); 4347 if (err) { 4348 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4349 netdev_adjacent_change_abort(dst->remote_dev, 4350 lowerdev, dev); 4351 return err; 4352 } 4353 } 4354 if (!vxlan_addr_any(&dst->remote_ip)) 4355 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4356 dst->remote_ip, 4357 vxlan->cfg.dst_port, 4358 dst->remote_vni, 4359 dst->remote_vni, 4360 dst->remote_ifindex, 4361 true); 4362 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4363 4364 /* If vni filtering device, also update fdb entries of 4365 * all vnis that were using default remote ip 4366 */ 4367 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4368 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4369 &conf.remote_ip, extack); 4370 if (err) { 4371 netdev_adjacent_change_abort(dst->remote_dev, 4372 lowerdev, dev); 4373 return err; 4374 } 4375 } 4376 } 4377 4378 if (conf.age_interval != vxlan->cfg.age_interval) 4379 mod_timer(&vxlan->age_timer, jiffies); 4380 4381 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4382 if (lowerdev && lowerdev != dst->remote_dev) 4383 dst->remote_dev = lowerdev; 4384 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4385 return 0; 4386 } 4387 4388 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4389 { 4390 struct vxlan_dev *vxlan = netdev_priv(dev); 4391 struct vxlan_fdb_flush_desc desc = { 4392 /* Default entry is deleted at vxlan_uninit. */ 4393 .ignore_default_entry = true, 4394 }; 4395 4396 vxlan_flush(vxlan, &desc); 4397 4398 list_del(&vxlan->next); 4399 unregister_netdevice_queue(dev, head); 4400 if (vxlan->default_dst.remote_dev) 4401 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4402 } 4403 4404 static size_t vxlan_get_size(const struct net_device *dev) 4405 { 4406 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4407 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4408 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4409 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4410 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4411 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4412 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4413 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4414 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4415 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4416 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4417 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4418 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4419 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4420 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4421 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4422 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4423 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4424 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4425 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4426 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4427 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4428 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4429 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4430 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4431 /* IFLA_VXLAN_PORT_RANGE */ 4432 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4433 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4434 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4435 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4436 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4437 0; 4438 } 4439 4440 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4441 { 4442 const struct vxlan_dev *vxlan = netdev_priv(dev); 4443 const struct vxlan_rdst *dst = &vxlan->default_dst; 4444 struct ifla_vxlan_port_range ports = { 4445 .low = htons(vxlan->cfg.port_min), 4446 .high = htons(vxlan->cfg.port_max), 4447 }; 4448 4449 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4450 goto nla_put_failure; 4451 4452 if (!vxlan_addr_any(&dst->remote_ip)) { 4453 if (dst->remote_ip.sa.sa_family == AF_INET) { 4454 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4455 dst->remote_ip.sin.sin_addr.s_addr)) 4456 goto nla_put_failure; 4457 #if IS_ENABLED(CONFIG_IPV6) 4458 } else { 4459 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4460 &dst->remote_ip.sin6.sin6_addr)) 4461 goto nla_put_failure; 4462 #endif 4463 } 4464 } 4465 4466 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4467 goto nla_put_failure; 4468 4469 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4470 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4471 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4472 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4473 goto nla_put_failure; 4474 #if IS_ENABLED(CONFIG_IPV6) 4475 } else { 4476 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4477 &vxlan->cfg.saddr.sin6.sin6_addr)) 4478 goto nla_put_failure; 4479 #endif 4480 } 4481 } 4482 4483 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4484 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4485 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4486 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4487 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4488 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4489 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4490 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4491 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4492 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4493 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4494 nla_put_u8(skb, IFLA_VXLAN_RSC, 4495 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4496 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4497 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4498 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4499 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4500 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4501 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4502 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4503 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4504 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4505 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4506 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4507 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4508 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4509 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4510 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4511 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4512 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4513 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4514 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4515 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4516 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4517 goto nla_put_failure; 4518 4519 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4520 goto nla_put_failure; 4521 4522 if (vxlan->cfg.flags & VXLAN_F_GBP && 4523 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4524 goto nla_put_failure; 4525 4526 if (vxlan->cfg.flags & VXLAN_F_GPE && 4527 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4528 goto nla_put_failure; 4529 4530 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4531 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4532 goto nla_put_failure; 4533 4534 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4535 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4536 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4537 goto nla_put_failure; 4538 4539 return 0; 4540 4541 nla_put_failure: 4542 return -EMSGSIZE; 4543 } 4544 4545 static struct net *vxlan_get_link_net(const struct net_device *dev) 4546 { 4547 struct vxlan_dev *vxlan = netdev_priv(dev); 4548 4549 return vxlan->net; 4550 } 4551 4552 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4553 .kind = "vxlan", 4554 .maxtype = IFLA_VXLAN_MAX, 4555 .policy = vxlan_policy, 4556 .priv_size = sizeof(struct vxlan_dev), 4557 .setup = vxlan_setup, 4558 .validate = vxlan_validate, 4559 .newlink = vxlan_newlink, 4560 .changelink = vxlan_changelink, 4561 .dellink = vxlan_dellink, 4562 .get_size = vxlan_get_size, 4563 .fill_info = vxlan_fill_info, 4564 .get_link_net = vxlan_get_link_net, 4565 }; 4566 4567 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4568 u8 name_assign_type, 4569 struct vxlan_config *conf) 4570 { 4571 struct nlattr *tb[IFLA_MAX + 1]; 4572 struct net_device *dev; 4573 int err; 4574 4575 memset(&tb, 0, sizeof(tb)); 4576 4577 dev = rtnl_create_link(net, name, name_assign_type, 4578 &vxlan_link_ops, tb, NULL); 4579 if (IS_ERR(dev)) 4580 return dev; 4581 4582 err = __vxlan_dev_create(net, dev, conf, NULL); 4583 if (err < 0) { 4584 free_netdev(dev); 4585 return ERR_PTR(err); 4586 } 4587 4588 err = rtnl_configure_link(dev, NULL, 0, NULL); 4589 if (err < 0) { 4590 LIST_HEAD(list_kill); 4591 4592 vxlan_dellink(dev, &list_kill); 4593 unregister_netdevice_many(&list_kill); 4594 return ERR_PTR(err); 4595 } 4596 4597 return dev; 4598 } 4599 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4600 4601 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4602 struct net_device *dev) 4603 { 4604 struct vxlan_dev *vxlan, *next; 4605 LIST_HEAD(list_kill); 4606 4607 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4608 struct vxlan_rdst *dst = &vxlan->default_dst; 4609 4610 /* In case we created vxlan device with carrier 4611 * and we loose the carrier due to module unload 4612 * we also need to remove vxlan device. In other 4613 * cases, it's not necessary and remote_ifindex 4614 * is 0 here, so no matches. 4615 */ 4616 if (dst->remote_ifindex == dev->ifindex) 4617 vxlan_dellink(vxlan->dev, &list_kill); 4618 } 4619 4620 unregister_netdevice_many(&list_kill); 4621 } 4622 4623 static int vxlan_netdevice_event(struct notifier_block *unused, 4624 unsigned long event, void *ptr) 4625 { 4626 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4627 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4628 4629 if (event == NETDEV_UNREGISTER) 4630 vxlan_handle_lowerdev_unregister(vn, dev); 4631 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4632 vxlan_offload_rx_ports(dev, true); 4633 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4634 vxlan_offload_rx_ports(dev, false); 4635 4636 return NOTIFY_DONE; 4637 } 4638 4639 static struct notifier_block vxlan_notifier_block __read_mostly = { 4640 .notifier_call = vxlan_netdevice_event, 4641 }; 4642 4643 static void 4644 vxlan_fdb_offloaded_set(struct net_device *dev, 4645 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4646 { 4647 struct vxlan_dev *vxlan = netdev_priv(dev); 4648 struct vxlan_rdst *rdst; 4649 struct vxlan_fdb *f; 4650 u32 hash_index; 4651 4652 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4653 4654 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4655 4656 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4657 if (!f) 4658 goto out; 4659 4660 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4661 fdb_info->remote_port, 4662 fdb_info->remote_vni, 4663 fdb_info->remote_ifindex); 4664 if (!rdst) 4665 goto out; 4666 4667 rdst->offloaded = fdb_info->offloaded; 4668 4669 out: 4670 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4671 } 4672 4673 static int 4674 vxlan_fdb_external_learn_add(struct net_device *dev, 4675 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4676 { 4677 struct vxlan_dev *vxlan = netdev_priv(dev); 4678 struct netlink_ext_ack *extack; 4679 u32 hash_index; 4680 int err; 4681 4682 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4683 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4684 4685 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4686 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4687 NUD_REACHABLE, 4688 NLM_F_CREATE | NLM_F_REPLACE, 4689 fdb_info->remote_port, 4690 fdb_info->vni, 4691 fdb_info->remote_vni, 4692 fdb_info->remote_ifindex, 4693 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4694 0, false, extack); 4695 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4696 4697 return err; 4698 } 4699 4700 static int 4701 vxlan_fdb_external_learn_del(struct net_device *dev, 4702 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4703 { 4704 struct vxlan_dev *vxlan = netdev_priv(dev); 4705 struct vxlan_fdb *f; 4706 u32 hash_index; 4707 int err = 0; 4708 4709 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4710 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4711 4712 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4713 if (!f) 4714 err = -ENOENT; 4715 else if (f->flags & NTF_EXT_LEARNED) 4716 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4717 fdb_info->remote_ip, 4718 fdb_info->remote_port, 4719 fdb_info->vni, 4720 fdb_info->remote_vni, 4721 fdb_info->remote_ifindex, 4722 false); 4723 4724 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4725 4726 return err; 4727 } 4728 4729 static int vxlan_switchdev_event(struct notifier_block *unused, 4730 unsigned long event, void *ptr) 4731 { 4732 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4733 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4734 int err = 0; 4735 4736 switch (event) { 4737 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4738 vxlan_fdb_offloaded_set(dev, ptr); 4739 break; 4740 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4741 fdb_info = ptr; 4742 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4743 if (err) { 4744 err = notifier_from_errno(err); 4745 break; 4746 } 4747 fdb_info->offloaded = true; 4748 vxlan_fdb_offloaded_set(dev, fdb_info); 4749 break; 4750 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4751 fdb_info = ptr; 4752 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4753 if (err) { 4754 err = notifier_from_errno(err); 4755 break; 4756 } 4757 fdb_info->offloaded = false; 4758 vxlan_fdb_offloaded_set(dev, fdb_info); 4759 break; 4760 } 4761 4762 return err; 4763 } 4764 4765 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4766 .notifier_call = vxlan_switchdev_event, 4767 }; 4768 4769 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4770 { 4771 struct vxlan_fdb *fdb; 4772 struct vxlan_dev *vxlan; 4773 u32 hash_index; 4774 4775 rcu_read_lock(); 4776 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4777 vxlan = rcu_dereference(fdb->vdev); 4778 WARN_ON(!vxlan); 4779 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4780 vxlan->default_dst.remote_vni); 4781 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4782 if (!hlist_unhashed(&fdb->hlist)) 4783 vxlan_fdb_destroy(vxlan, fdb, false, false); 4784 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4785 } 4786 rcu_read_unlock(); 4787 } 4788 4789 static int vxlan_nexthop_event(struct notifier_block *nb, 4790 unsigned long event, void *ptr) 4791 { 4792 struct nh_notifier_info *info = ptr; 4793 struct nexthop *nh; 4794 4795 if (event != NEXTHOP_EVENT_DEL) 4796 return NOTIFY_DONE; 4797 4798 nh = nexthop_find_by_id(info->net, info->id); 4799 if (!nh) 4800 return NOTIFY_DONE; 4801 4802 vxlan_fdb_nh_flush(nh); 4803 4804 return NOTIFY_DONE; 4805 } 4806 4807 static __net_init int vxlan_init_net(struct net *net) 4808 { 4809 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4810 unsigned int h; 4811 4812 INIT_LIST_HEAD(&vn->vxlan_list); 4813 spin_lock_init(&vn->sock_lock); 4814 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4815 4816 for (h = 0; h < PORT_HASH_SIZE; ++h) 4817 INIT_HLIST_HEAD(&vn->sock_list[h]); 4818 4819 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4820 NULL); 4821 } 4822 4823 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4824 struct list_head *dev_to_kill) 4825 { 4826 struct vxlan_dev *vxlan, *next; 4827 4828 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4829 vxlan_dellink(vxlan->dev, dev_to_kill); 4830 } 4831 4832 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4833 struct list_head *dev_to_kill) 4834 { 4835 struct net *net; 4836 4837 ASSERT_RTNL(); 4838 list_for_each_entry(net, net_list, exit_list) { 4839 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4840 4841 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4842 4843 vxlan_destroy_tunnels(vn, dev_to_kill); 4844 } 4845 } 4846 4847 static void __net_exit vxlan_exit_net(struct net *net) 4848 { 4849 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4850 unsigned int h; 4851 4852 for (h = 0; h < PORT_HASH_SIZE; ++h) 4853 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4854 } 4855 4856 static struct pernet_operations vxlan_net_ops = { 4857 .init = vxlan_init_net, 4858 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4859 .exit = vxlan_exit_net, 4860 .id = &vxlan_net_id, 4861 .size = sizeof(struct vxlan_net), 4862 }; 4863 4864 static int __init vxlan_init_module(void) 4865 { 4866 int rc; 4867 4868 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4869 4870 rc = register_pernet_subsys(&vxlan_net_ops); 4871 if (rc) 4872 goto out1; 4873 4874 rc = register_netdevice_notifier(&vxlan_notifier_block); 4875 if (rc) 4876 goto out2; 4877 4878 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4879 if (rc) 4880 goto out3; 4881 4882 rc = rtnl_link_register(&vxlan_link_ops); 4883 if (rc) 4884 goto out4; 4885 4886 vxlan_vnifilter_init(); 4887 4888 return 0; 4889 out4: 4890 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4891 out3: 4892 unregister_netdevice_notifier(&vxlan_notifier_block); 4893 out2: 4894 unregister_pernet_subsys(&vxlan_net_ops); 4895 out1: 4896 return rc; 4897 } 4898 late_initcall(vxlan_init_module); 4899 4900 static void __exit vxlan_cleanup_module(void) 4901 { 4902 vxlan_vnifilter_uninit(); 4903 rtnl_link_unregister(&vxlan_link_ops); 4904 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4905 unregister_netdevice_notifier(&vxlan_notifier_block); 4906 unregister_pernet_subsys(&vxlan_net_ops); 4907 /* rcu_barrier() is called by netns */ 4908 } 4909 module_exit(vxlan_cleanup_module); 4910 4911 MODULE_LICENSE("GPL"); 4912 MODULE_VERSION(VXLAN_VERSION); 4913 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4914 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4915 MODULE_ALIAS_RTNL_LINK("vxlan"); 4916