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 int nh; 1678 1679 /* Need UDP and VXLAN header to be present */ 1680 if (!pskb_may_pull(skb, VXLAN_HLEN)) 1681 goto drop; 1682 1683 unparsed = *vxlan_hdr(skb); 1684 /* VNI flag always required to be set */ 1685 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1686 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1687 ntohl(vxlan_hdr(skb)->vx_flags), 1688 ntohl(vxlan_hdr(skb)->vx_vni)); 1689 /* Return non vxlan pkt */ 1690 goto drop; 1691 } 1692 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1693 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1694 1695 vs = rcu_dereference_sk_user_data(sk); 1696 if (!vs) 1697 goto drop; 1698 1699 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1700 1701 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1702 if (!vxlan) 1703 goto drop; 1704 1705 /* For backwards compatibility, only allow reserved fields to be 1706 * used by VXLAN extensions if explicitly requested. 1707 */ 1708 if (vs->flags & VXLAN_F_GPE) { 1709 if (!vxlan_parse_gpe_proto(&unparsed, &protocol)) 1710 goto drop; 1711 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS; 1712 raw_proto = true; 1713 } 1714 1715 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1716 !net_eq(vxlan->net, dev_net(vxlan->dev)))) 1717 goto drop; 1718 1719 if (vs->flags & VXLAN_F_REMCSUM_RX) 1720 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1721 goto drop; 1722 1723 if (vxlan_collect_metadata(vs)) { 1724 struct metadata_dst *tun_dst; 1725 1726 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY, 1727 key32_to_tunnel_id(vni), sizeof(*md)); 1728 1729 if (!tun_dst) 1730 goto drop; 1731 1732 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1733 1734 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1735 } else { 1736 memset(md, 0, sizeof(*md)); 1737 } 1738 1739 if (vs->flags & VXLAN_F_GBP) 1740 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1741 /* Note that GBP and GPE can never be active together. This is 1742 * ensured in vxlan_dev_configure. 1743 */ 1744 1745 if (unparsed.vx_flags || unparsed.vx_vni) { 1746 /* If there are any unprocessed flags remaining treat 1747 * this as a malformed packet. This behavior diverges from 1748 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1749 * in reserved fields are to be ignored. The approach here 1750 * maintains compatibility with previous stack code, and also 1751 * is more robust and provides a little more security in 1752 * adding extensions to VXLAN. 1753 */ 1754 goto drop; 1755 } 1756 1757 if (!raw_proto) { 1758 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1759 goto drop; 1760 } else { 1761 skb_reset_mac_header(skb); 1762 skb->dev = vxlan->dev; 1763 skb->pkt_type = PACKET_HOST; 1764 } 1765 1766 /* Save offset of outer header relative to skb->head, 1767 * because we are going to reset the network header to the inner header 1768 * and might change skb->head. 1769 */ 1770 nh = skb_network_header(skb) - skb->head; 1771 1772 skb_reset_network_header(skb); 1773 1774 if (!pskb_inet_may_pull(skb)) { 1775 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1776 DEV_STATS_INC(vxlan->dev, rx_errors); 1777 vxlan_vnifilter_count(vxlan, vni, vninode, 1778 VXLAN_VNI_STATS_RX_ERRORS, 0); 1779 goto drop; 1780 } 1781 1782 /* Get the outer header. */ 1783 oiph = skb->head + nh; 1784 1785 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1786 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1787 DEV_STATS_INC(vxlan->dev, rx_errors); 1788 vxlan_vnifilter_count(vxlan, vni, vninode, 1789 VXLAN_VNI_STATS_RX_ERRORS, 0); 1790 goto drop; 1791 } 1792 1793 rcu_read_lock(); 1794 1795 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1796 rcu_read_unlock(); 1797 dev_core_stats_rx_dropped_inc(vxlan->dev); 1798 vxlan_vnifilter_count(vxlan, vni, vninode, 1799 VXLAN_VNI_STATS_RX_DROPS, 0); 1800 goto drop; 1801 } 1802 1803 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1804 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1805 gro_cells_receive(&vxlan->gro_cells, skb); 1806 1807 rcu_read_unlock(); 1808 1809 return 0; 1810 1811 drop: 1812 /* Consume bad packet */ 1813 kfree_skb(skb); 1814 return 0; 1815 } 1816 1817 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1818 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1819 { 1820 struct vxlan_dev *vxlan; 1821 struct vxlan_sock *vs; 1822 struct vxlanhdr *hdr; 1823 __be32 vni; 1824 1825 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1826 return -EINVAL; 1827 1828 hdr = vxlan_hdr(skb); 1829 1830 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1831 return -EINVAL; 1832 1833 vs = rcu_dereference_sk_user_data(sk); 1834 if (!vs) 1835 return -ENOENT; 1836 1837 vni = vxlan_vni(hdr->vx_vni); 1838 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1839 if (!vxlan) 1840 return -ENOENT; 1841 1842 return 0; 1843 } 1844 1845 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1846 { 1847 struct vxlan_dev *vxlan = netdev_priv(dev); 1848 struct arphdr *parp; 1849 u8 *arpptr, *sha; 1850 __be32 sip, tip; 1851 struct neighbour *n; 1852 1853 if (dev->flags & IFF_NOARP) 1854 goto out; 1855 1856 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1857 dev_core_stats_tx_dropped_inc(dev); 1858 vxlan_vnifilter_count(vxlan, vni, NULL, 1859 VXLAN_VNI_STATS_TX_DROPS, 0); 1860 goto out; 1861 } 1862 parp = arp_hdr(skb); 1863 1864 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1865 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1866 parp->ar_pro != htons(ETH_P_IP) || 1867 parp->ar_op != htons(ARPOP_REQUEST) || 1868 parp->ar_hln != dev->addr_len || 1869 parp->ar_pln != 4) 1870 goto out; 1871 arpptr = (u8 *)parp + sizeof(struct arphdr); 1872 sha = arpptr; 1873 arpptr += dev->addr_len; /* sha */ 1874 memcpy(&sip, arpptr, sizeof(sip)); 1875 arpptr += sizeof(sip); 1876 arpptr += dev->addr_len; /* tha */ 1877 memcpy(&tip, arpptr, sizeof(tip)); 1878 1879 if (ipv4_is_loopback(tip) || 1880 ipv4_is_multicast(tip)) 1881 goto out; 1882 1883 n = neigh_lookup(&arp_tbl, &tip, dev); 1884 1885 if (n) { 1886 struct vxlan_fdb *f; 1887 struct sk_buff *reply; 1888 1889 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1890 neigh_release(n); 1891 goto out; 1892 } 1893 1894 f = vxlan_find_mac(vxlan, n->ha, vni); 1895 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1896 /* bridge-local neighbor */ 1897 neigh_release(n); 1898 goto out; 1899 } 1900 1901 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1902 n->ha, sha); 1903 1904 neigh_release(n); 1905 1906 if (reply == NULL) 1907 goto out; 1908 1909 skb_reset_mac_header(reply); 1910 __skb_pull(reply, skb_network_offset(reply)); 1911 reply->ip_summed = CHECKSUM_UNNECESSARY; 1912 reply->pkt_type = PACKET_HOST; 1913 1914 if (netif_rx(reply) == NET_RX_DROP) { 1915 dev_core_stats_rx_dropped_inc(dev); 1916 vxlan_vnifilter_count(vxlan, vni, NULL, 1917 VXLAN_VNI_STATS_RX_DROPS, 0); 1918 } 1919 1920 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1921 union vxlan_addr ipa = { 1922 .sin.sin_addr.s_addr = tip, 1923 .sin.sin_family = AF_INET, 1924 }; 1925 1926 vxlan_ip_miss(dev, &ipa); 1927 } 1928 out: 1929 consume_skb(skb); 1930 return NETDEV_TX_OK; 1931 } 1932 1933 #if IS_ENABLED(CONFIG_IPV6) 1934 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1935 struct neighbour *n, bool isrouter) 1936 { 1937 struct net_device *dev = request->dev; 1938 struct sk_buff *reply; 1939 struct nd_msg *ns, *na; 1940 struct ipv6hdr *pip6; 1941 u8 *daddr; 1942 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1943 int ns_olen; 1944 int i, len; 1945 1946 if (dev == NULL || !pskb_may_pull(request, request->len)) 1947 return NULL; 1948 1949 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1950 sizeof(*na) + na_olen + dev->needed_tailroom; 1951 reply = alloc_skb(len, GFP_ATOMIC); 1952 if (reply == NULL) 1953 return NULL; 1954 1955 reply->protocol = htons(ETH_P_IPV6); 1956 reply->dev = dev; 1957 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1958 skb_push(reply, sizeof(struct ethhdr)); 1959 skb_reset_mac_header(reply); 1960 1961 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1962 1963 daddr = eth_hdr(request)->h_source; 1964 ns_olen = request->len - skb_network_offset(request) - 1965 sizeof(struct ipv6hdr) - sizeof(*ns); 1966 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1967 if (!ns->opt[i + 1]) { 1968 kfree_skb(reply); 1969 return NULL; 1970 } 1971 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1972 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1973 break; 1974 } 1975 } 1976 1977 /* Ethernet header */ 1978 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1979 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1980 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1981 reply->protocol = htons(ETH_P_IPV6); 1982 1983 skb_pull(reply, sizeof(struct ethhdr)); 1984 skb_reset_network_header(reply); 1985 skb_put(reply, sizeof(struct ipv6hdr)); 1986 1987 /* IPv6 header */ 1988 1989 pip6 = ipv6_hdr(reply); 1990 memset(pip6, 0, sizeof(struct ipv6hdr)); 1991 pip6->version = 6; 1992 pip6->priority = ipv6_hdr(request)->priority; 1993 pip6->nexthdr = IPPROTO_ICMPV6; 1994 pip6->hop_limit = 255; 1995 pip6->daddr = ipv6_hdr(request)->saddr; 1996 pip6->saddr = *(struct in6_addr *)n->primary_key; 1997 1998 skb_pull(reply, sizeof(struct ipv6hdr)); 1999 skb_reset_transport_header(reply); 2000 2001 /* Neighbor Advertisement */ 2002 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2003 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2004 na->icmph.icmp6_router = isrouter; 2005 na->icmph.icmp6_override = 1; 2006 na->icmph.icmp6_solicited = 1; 2007 na->target = ns->target; 2008 ether_addr_copy(&na->opt[2], n->ha); 2009 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2010 na->opt[1] = na_olen >> 3; 2011 2012 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2013 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2014 csum_partial(na, sizeof(*na)+na_olen, 0)); 2015 2016 pip6->payload_len = htons(sizeof(*na)+na_olen); 2017 2018 skb_push(reply, sizeof(struct ipv6hdr)); 2019 2020 reply->ip_summed = CHECKSUM_UNNECESSARY; 2021 2022 return reply; 2023 } 2024 2025 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2026 { 2027 struct vxlan_dev *vxlan = netdev_priv(dev); 2028 const struct in6_addr *daddr; 2029 const struct ipv6hdr *iphdr; 2030 struct inet6_dev *in6_dev; 2031 struct neighbour *n; 2032 struct nd_msg *msg; 2033 2034 rcu_read_lock(); 2035 in6_dev = __in6_dev_get(dev); 2036 if (!in6_dev) 2037 goto out; 2038 2039 iphdr = ipv6_hdr(skb); 2040 daddr = &iphdr->daddr; 2041 msg = (struct nd_msg *)(iphdr + 1); 2042 2043 if (ipv6_addr_loopback(daddr) || 2044 ipv6_addr_is_multicast(&msg->target)) 2045 goto out; 2046 2047 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2048 2049 if (n) { 2050 struct vxlan_fdb *f; 2051 struct sk_buff *reply; 2052 2053 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2054 neigh_release(n); 2055 goto out; 2056 } 2057 2058 f = vxlan_find_mac(vxlan, n->ha, vni); 2059 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2060 /* bridge-local neighbor */ 2061 neigh_release(n); 2062 goto out; 2063 } 2064 2065 reply = vxlan_na_create(skb, n, 2066 !!(f ? f->flags & NTF_ROUTER : 0)); 2067 2068 neigh_release(n); 2069 2070 if (reply == NULL) 2071 goto out; 2072 2073 if (netif_rx(reply) == NET_RX_DROP) { 2074 dev_core_stats_rx_dropped_inc(dev); 2075 vxlan_vnifilter_count(vxlan, vni, NULL, 2076 VXLAN_VNI_STATS_RX_DROPS, 0); 2077 } 2078 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2079 union vxlan_addr ipa = { 2080 .sin6.sin6_addr = msg->target, 2081 .sin6.sin6_family = AF_INET6, 2082 }; 2083 2084 vxlan_ip_miss(dev, &ipa); 2085 } 2086 2087 out: 2088 rcu_read_unlock(); 2089 consume_skb(skb); 2090 return NETDEV_TX_OK; 2091 } 2092 #endif 2093 2094 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2095 { 2096 struct vxlan_dev *vxlan = netdev_priv(dev); 2097 struct neighbour *n; 2098 2099 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2100 return false; 2101 2102 n = NULL; 2103 switch (ntohs(eth_hdr(skb)->h_proto)) { 2104 case ETH_P_IP: 2105 { 2106 struct iphdr *pip; 2107 2108 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2109 return false; 2110 pip = ip_hdr(skb); 2111 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2112 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2113 union vxlan_addr ipa = { 2114 .sin.sin_addr.s_addr = pip->daddr, 2115 .sin.sin_family = AF_INET, 2116 }; 2117 2118 vxlan_ip_miss(dev, &ipa); 2119 return false; 2120 } 2121 2122 break; 2123 } 2124 #if IS_ENABLED(CONFIG_IPV6) 2125 case ETH_P_IPV6: 2126 { 2127 struct ipv6hdr *pip6; 2128 2129 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2130 return false; 2131 pip6 = ipv6_hdr(skb); 2132 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2133 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2134 union vxlan_addr ipa = { 2135 .sin6.sin6_addr = pip6->daddr, 2136 .sin6.sin6_family = AF_INET6, 2137 }; 2138 2139 vxlan_ip_miss(dev, &ipa); 2140 return false; 2141 } 2142 2143 break; 2144 } 2145 #endif 2146 default: 2147 return false; 2148 } 2149 2150 if (n) { 2151 bool diff; 2152 2153 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2154 if (diff) { 2155 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2156 dev->addr_len); 2157 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2158 } 2159 neigh_release(n); 2160 return diff; 2161 } 2162 2163 return false; 2164 } 2165 2166 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2167 { 2168 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2169 2170 gpe->np_applied = 1; 2171 gpe->next_protocol = tun_p_from_eth_p(protocol); 2172 if (!gpe->next_protocol) 2173 return -EPFNOSUPPORT; 2174 return 0; 2175 } 2176 2177 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2178 int iphdr_len, __be32 vni, 2179 struct vxlan_metadata *md, u32 vxflags, 2180 bool udp_sum) 2181 { 2182 struct vxlanhdr *vxh; 2183 int min_headroom; 2184 int err; 2185 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2186 __be16 inner_protocol = htons(ETH_P_TEB); 2187 2188 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2189 skb->ip_summed == CHECKSUM_PARTIAL) { 2190 int csum_start = skb_checksum_start_offset(skb); 2191 2192 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2193 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2194 (skb->csum_offset == offsetof(struct udphdr, check) || 2195 skb->csum_offset == offsetof(struct tcphdr, check))) 2196 type |= SKB_GSO_TUNNEL_REMCSUM; 2197 } 2198 2199 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2200 + VXLAN_HLEN + iphdr_len; 2201 2202 /* Need space for new headers (invalidates iph ptr) */ 2203 err = skb_cow_head(skb, min_headroom); 2204 if (unlikely(err)) 2205 return err; 2206 2207 err = iptunnel_handle_offloads(skb, type); 2208 if (err) 2209 return err; 2210 2211 vxh = __skb_push(skb, sizeof(*vxh)); 2212 vxh->vx_flags = VXLAN_HF_VNI; 2213 vxh->vx_vni = vxlan_vni_field(vni); 2214 2215 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2216 unsigned int start; 2217 2218 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2219 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2220 vxh->vx_flags |= VXLAN_HF_RCO; 2221 2222 if (!skb_is_gso(skb)) { 2223 skb->ip_summed = CHECKSUM_NONE; 2224 skb->encapsulation = 0; 2225 } 2226 } 2227 2228 if (vxflags & VXLAN_F_GBP) 2229 vxlan_build_gbp_hdr(vxh, md); 2230 if (vxflags & VXLAN_F_GPE) { 2231 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2232 if (err < 0) 2233 return err; 2234 inner_protocol = skb->protocol; 2235 } 2236 2237 skb_set_inner_protocol(skb, inner_protocol); 2238 return 0; 2239 } 2240 2241 /* Bypass encapsulation if the destination is local */ 2242 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2243 struct vxlan_dev *dst_vxlan, __be32 vni, 2244 bool snoop) 2245 { 2246 union vxlan_addr loopback; 2247 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2248 struct net_device *dev; 2249 int len = skb->len; 2250 2251 skb->pkt_type = PACKET_HOST; 2252 skb->encapsulation = 0; 2253 skb->dev = dst_vxlan->dev; 2254 __skb_pull(skb, skb_network_offset(skb)); 2255 2256 if (remote_ip->sa.sa_family == AF_INET) { 2257 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2258 loopback.sa.sa_family = AF_INET; 2259 #if IS_ENABLED(CONFIG_IPV6) 2260 } else { 2261 loopback.sin6.sin6_addr = in6addr_loopback; 2262 loopback.sa.sa_family = AF_INET6; 2263 #endif 2264 } 2265 2266 rcu_read_lock(); 2267 dev = skb->dev; 2268 if (unlikely(!(dev->flags & IFF_UP))) { 2269 kfree_skb(skb); 2270 goto drop; 2271 } 2272 2273 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2274 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2275 2276 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2277 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2278 2279 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2280 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2281 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2282 len); 2283 } else { 2284 drop: 2285 dev_core_stats_rx_dropped_inc(dev); 2286 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2287 VXLAN_VNI_STATS_RX_DROPS, 0); 2288 } 2289 rcu_read_unlock(); 2290 } 2291 2292 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2293 struct vxlan_dev *vxlan, 2294 int addr_family, 2295 __be16 dst_port, int dst_ifindex, __be32 vni, 2296 struct dst_entry *dst, 2297 u32 rt_flags) 2298 { 2299 #if IS_ENABLED(CONFIG_IPV6) 2300 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2301 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2302 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2303 */ 2304 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2305 #endif 2306 /* Bypass encapsulation if the destination is local */ 2307 if (rt_flags & RTCF_LOCAL && 2308 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2309 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2310 struct vxlan_dev *dst_vxlan; 2311 2312 dst_release(dst); 2313 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2314 addr_family, dst_port, 2315 vxlan->cfg.flags); 2316 if (!dst_vxlan) { 2317 DEV_STATS_INC(dev, tx_errors); 2318 vxlan_vnifilter_count(vxlan, vni, NULL, 2319 VXLAN_VNI_STATS_TX_ERRORS, 0); 2320 kfree_skb(skb); 2321 2322 return -ENOENT; 2323 } 2324 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2325 return 1; 2326 } 2327 2328 return 0; 2329 } 2330 2331 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2332 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2333 { 2334 struct dst_cache *dst_cache; 2335 struct ip_tunnel_info *info; 2336 struct ip_tunnel_key *pkey; 2337 struct ip_tunnel_key key; 2338 struct vxlan_dev *vxlan = netdev_priv(dev); 2339 const struct iphdr *old_iph = ip_hdr(skb); 2340 struct vxlan_metadata _md; 2341 struct vxlan_metadata *md = &_md; 2342 unsigned int pkt_len = skb->len; 2343 __be16 src_port = 0, dst_port; 2344 struct dst_entry *ndst = NULL; 2345 int addr_family; 2346 __u8 tos, ttl; 2347 int ifindex; 2348 int err; 2349 u32 flags = vxlan->cfg.flags; 2350 bool use_cache; 2351 bool udp_sum = false; 2352 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2353 __be32 vni = 0; 2354 2355 info = skb_tunnel_info(skb); 2356 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2357 2358 if (rdst) { 2359 memset(&key, 0, sizeof(key)); 2360 pkey = &key; 2361 2362 if (vxlan_addr_any(&rdst->remote_ip)) { 2363 if (did_rsc) { 2364 /* short-circuited back to local bridge */ 2365 vxlan_encap_bypass(skb, vxlan, vxlan, 2366 default_vni, true); 2367 return; 2368 } 2369 goto drop; 2370 } 2371 2372 addr_family = vxlan->cfg.saddr.sa.sa_family; 2373 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2374 vni = (rdst->remote_vni) ? : default_vni; 2375 ifindex = rdst->remote_ifindex; 2376 2377 if (addr_family == AF_INET) { 2378 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2379 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2380 } else { 2381 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2382 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2383 } 2384 2385 dst_cache = &rdst->dst_cache; 2386 md->gbp = skb->mark; 2387 if (flags & VXLAN_F_TTL_INHERIT) { 2388 ttl = ip_tunnel_get_ttl(old_iph, skb); 2389 } else { 2390 ttl = vxlan->cfg.ttl; 2391 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2392 ttl = 1; 2393 } 2394 tos = vxlan->cfg.tos; 2395 if (tos == 1) 2396 tos = ip_tunnel_get_dsfield(old_iph, skb); 2397 if (tos && !info) 2398 use_cache = false; 2399 2400 if (addr_family == AF_INET) 2401 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2402 else 2403 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2404 #if IS_ENABLED(CONFIG_IPV6) 2405 switch (vxlan->cfg.label_policy) { 2406 case VXLAN_LABEL_FIXED: 2407 key.label = vxlan->cfg.label; 2408 break; 2409 case VXLAN_LABEL_INHERIT: 2410 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2411 break; 2412 default: 2413 DEBUG_NET_WARN_ON_ONCE(1); 2414 goto drop; 2415 } 2416 #endif 2417 } else { 2418 if (!info) { 2419 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2420 dev->name); 2421 goto drop; 2422 } 2423 pkey = &info->key; 2424 addr_family = ip_tunnel_info_af(info); 2425 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2426 vni = tunnel_id_to_key32(info->key.tun_id); 2427 ifindex = 0; 2428 dst_cache = &info->dst_cache; 2429 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) { 2430 if (info->options_len < sizeof(*md)) 2431 goto drop; 2432 md = ip_tunnel_info_opts(info); 2433 } 2434 ttl = info->key.ttl; 2435 tos = info->key.tos; 2436 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 2437 } 2438 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2439 vxlan->cfg.port_max, true); 2440 2441 rcu_read_lock(); 2442 if (addr_family == AF_INET) { 2443 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2444 struct rtable *rt; 2445 __be16 df = 0; 2446 __be32 saddr; 2447 2448 if (!ifindex) 2449 ifindex = sock4->sock->sk->sk_bound_dev_if; 2450 2451 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2452 &saddr, pkey, src_port, dst_port, 2453 tos, use_cache ? dst_cache : NULL); 2454 if (IS_ERR(rt)) { 2455 err = PTR_ERR(rt); 2456 goto tx_error; 2457 } 2458 2459 if (!info) { 2460 /* Bypass encapsulation if the destination is local */ 2461 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2462 dst_port, ifindex, vni, 2463 &rt->dst, rt->rt_flags); 2464 if (err) 2465 goto out_unlock; 2466 2467 if (vxlan->cfg.df == VXLAN_DF_SET) { 2468 df = htons(IP_DF); 2469 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2470 struct ethhdr *eth = eth_hdr(skb); 2471 2472 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2473 (ntohs(eth->h_proto) == ETH_P_IP && 2474 old_iph->frag_off & htons(IP_DF))) 2475 df = htons(IP_DF); 2476 } 2477 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) { 2478 df = htons(IP_DF); 2479 } 2480 2481 ndst = &rt->dst; 2482 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2483 netif_is_any_bridge_port(dev)); 2484 if (err < 0) { 2485 goto tx_error; 2486 } else if (err) { 2487 if (info) { 2488 struct ip_tunnel_info *unclone; 2489 2490 unclone = skb_tunnel_info_unclone(skb); 2491 if (unlikely(!unclone)) 2492 goto tx_error; 2493 2494 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2495 unclone->key.u.ipv4.dst = saddr; 2496 } 2497 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2498 dst_release(ndst); 2499 goto out_unlock; 2500 } 2501 2502 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2503 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2504 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2505 vni, md, flags, udp_sum); 2506 if (err < 0) 2507 goto tx_error; 2508 2509 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2510 pkey->u.ipv4.dst, tos, ttl, df, 2511 src_port, dst_port, xnet, !udp_sum); 2512 #if IS_ENABLED(CONFIG_IPV6) 2513 } else { 2514 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2515 struct in6_addr saddr; 2516 2517 if (!ifindex) 2518 ifindex = sock6->sock->sk->sk_bound_dev_if; 2519 2520 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2521 ifindex, &saddr, pkey, 2522 src_port, dst_port, tos, 2523 use_cache ? dst_cache : NULL); 2524 if (IS_ERR(ndst)) { 2525 err = PTR_ERR(ndst); 2526 ndst = NULL; 2527 goto tx_error; 2528 } 2529 2530 if (!info) { 2531 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; 2532 2533 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2534 dst_port, ifindex, vni, 2535 ndst, rt6i_flags); 2536 if (err) 2537 goto out_unlock; 2538 } 2539 2540 err = skb_tunnel_check_pmtu(skb, ndst, 2541 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2542 netif_is_any_bridge_port(dev)); 2543 if (err < 0) { 2544 goto tx_error; 2545 } else if (err) { 2546 if (info) { 2547 struct ip_tunnel_info *unclone; 2548 2549 unclone = skb_tunnel_info_unclone(skb); 2550 if (unlikely(!unclone)) 2551 goto tx_error; 2552 2553 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2554 unclone->key.u.ipv6.dst = saddr; 2555 } 2556 2557 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2558 dst_release(ndst); 2559 goto out_unlock; 2560 } 2561 2562 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2563 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2564 skb_scrub_packet(skb, xnet); 2565 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2566 vni, md, flags, udp_sum); 2567 if (err < 0) 2568 goto tx_error; 2569 2570 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2571 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2572 pkey->label, src_port, dst_port, !udp_sum); 2573 #endif 2574 } 2575 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2576 out_unlock: 2577 rcu_read_unlock(); 2578 return; 2579 2580 drop: 2581 dev_core_stats_tx_dropped_inc(dev); 2582 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2583 dev_kfree_skb(skb); 2584 return; 2585 2586 tx_error: 2587 rcu_read_unlock(); 2588 if (err == -ELOOP) 2589 DEV_STATS_INC(dev, collisions); 2590 else if (err == -ENETUNREACH) 2591 DEV_STATS_INC(dev, tx_carrier_errors); 2592 dst_release(ndst); 2593 DEV_STATS_INC(dev, tx_errors); 2594 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2595 kfree_skb(skb); 2596 } 2597 2598 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2599 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2600 { 2601 struct vxlan_rdst nh_rdst; 2602 struct nexthop *nh; 2603 bool do_xmit; 2604 u32 hash; 2605 2606 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2607 hash = skb_get_hash(skb); 2608 2609 rcu_read_lock(); 2610 nh = rcu_dereference(f->nh); 2611 if (!nh) { 2612 rcu_read_unlock(); 2613 goto drop; 2614 } 2615 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2616 rcu_read_unlock(); 2617 2618 if (likely(do_xmit)) 2619 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2620 else 2621 goto drop; 2622 2623 return; 2624 2625 drop: 2626 dev_core_stats_tx_dropped_inc(dev); 2627 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2628 VXLAN_VNI_STATS_TX_DROPS, 0); 2629 dev_kfree_skb(skb); 2630 } 2631 2632 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2633 u32 nhid, __be32 vni) 2634 { 2635 struct vxlan_dev *vxlan = netdev_priv(dev); 2636 struct vxlan_rdst nh_rdst; 2637 struct nexthop *nh; 2638 bool do_xmit; 2639 u32 hash; 2640 2641 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2642 hash = skb_get_hash(skb); 2643 2644 rcu_read_lock(); 2645 nh = nexthop_find_by_id(dev_net(dev), nhid); 2646 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2647 rcu_read_unlock(); 2648 goto drop; 2649 } 2650 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2651 rcu_read_unlock(); 2652 2653 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2654 goto drop; 2655 2656 if (likely(do_xmit)) 2657 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2658 else 2659 goto drop; 2660 2661 return NETDEV_TX_OK; 2662 2663 drop: 2664 dev_core_stats_tx_dropped_inc(dev); 2665 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2666 VXLAN_VNI_STATS_TX_DROPS, 0); 2667 dev_kfree_skb(skb); 2668 return NETDEV_TX_OK; 2669 } 2670 2671 /* Transmit local packets over Vxlan 2672 * 2673 * Outer IP header inherits ECN and DF from inner header. 2674 * Outer UDP destination is the VXLAN assigned port. 2675 * source port is based on hash of flow 2676 */ 2677 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2678 { 2679 struct vxlan_dev *vxlan = netdev_priv(dev); 2680 struct vxlan_rdst *rdst, *fdst = NULL; 2681 const struct ip_tunnel_info *info; 2682 bool did_rsc = false; 2683 struct vxlan_fdb *f; 2684 struct ethhdr *eth; 2685 __be32 vni = 0; 2686 u32 nhid = 0; 2687 2688 info = skb_tunnel_info(skb); 2689 2690 skb_reset_mac_header(skb); 2691 2692 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2693 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2694 info->mode & IP_TUNNEL_INFO_TX) { 2695 vni = tunnel_id_to_key32(info->key.tun_id); 2696 nhid = info->key.nhid; 2697 } else { 2698 if (info && info->mode & IP_TUNNEL_INFO_TX) 2699 vxlan_xmit_one(skb, dev, vni, NULL, false); 2700 else 2701 kfree_skb(skb); 2702 return NETDEV_TX_OK; 2703 } 2704 } 2705 2706 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2707 eth = eth_hdr(skb); 2708 if (ntohs(eth->h_proto) == ETH_P_ARP) 2709 return arp_reduce(dev, skb, vni); 2710 #if IS_ENABLED(CONFIG_IPV6) 2711 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2712 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2713 sizeof(struct nd_msg)) && 2714 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2715 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2716 2717 if (m->icmph.icmp6_code == 0 && 2718 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2719 return neigh_reduce(dev, skb, vni); 2720 } 2721 #endif 2722 } 2723 2724 if (nhid) 2725 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2726 2727 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2728 struct vxlan_mdb_entry *mdb_entry; 2729 2730 rcu_read_lock(); 2731 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2732 if (mdb_entry) { 2733 netdev_tx_t ret; 2734 2735 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2736 rcu_read_unlock(); 2737 return ret; 2738 } 2739 rcu_read_unlock(); 2740 } 2741 2742 eth = eth_hdr(skb); 2743 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2744 did_rsc = false; 2745 2746 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2747 (ntohs(eth->h_proto) == ETH_P_IP || 2748 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2749 did_rsc = route_shortcircuit(dev, skb); 2750 if (did_rsc) 2751 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2752 } 2753 2754 if (f == NULL) { 2755 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2756 if (f == NULL) { 2757 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2758 !is_multicast_ether_addr(eth->h_dest)) 2759 vxlan_fdb_miss(vxlan, eth->h_dest); 2760 2761 dev_core_stats_tx_dropped_inc(dev); 2762 vxlan_vnifilter_count(vxlan, vni, NULL, 2763 VXLAN_VNI_STATS_TX_DROPS, 0); 2764 kfree_skb(skb); 2765 return NETDEV_TX_OK; 2766 } 2767 } 2768 2769 if (rcu_access_pointer(f->nh)) { 2770 vxlan_xmit_nh(skb, dev, f, 2771 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2772 } else { 2773 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2774 struct sk_buff *skb1; 2775 2776 if (!fdst) { 2777 fdst = rdst; 2778 continue; 2779 } 2780 skb1 = skb_clone(skb, GFP_ATOMIC); 2781 if (skb1) 2782 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2783 } 2784 if (fdst) 2785 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2786 else 2787 kfree_skb(skb); 2788 } 2789 2790 return NETDEV_TX_OK; 2791 } 2792 2793 /* Walk the forwarding table and purge stale entries */ 2794 static void vxlan_cleanup(struct timer_list *t) 2795 { 2796 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2797 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2798 unsigned int h; 2799 2800 if (!netif_running(vxlan->dev)) 2801 return; 2802 2803 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2804 struct hlist_node *p, *n; 2805 2806 spin_lock(&vxlan->hash_lock[h]); 2807 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2808 struct vxlan_fdb *f 2809 = container_of(p, struct vxlan_fdb, hlist); 2810 unsigned long timeout; 2811 2812 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2813 continue; 2814 2815 if (f->flags & NTF_EXT_LEARNED) 2816 continue; 2817 2818 timeout = f->used + vxlan->cfg.age_interval * HZ; 2819 if (time_before_eq(timeout, jiffies)) { 2820 netdev_dbg(vxlan->dev, 2821 "garbage collect %pM\n", 2822 f->eth_addr); 2823 f->state = NUD_STALE; 2824 vxlan_fdb_destroy(vxlan, f, true, true); 2825 } else if (time_before(timeout, next_timer)) 2826 next_timer = timeout; 2827 } 2828 spin_unlock(&vxlan->hash_lock[h]); 2829 } 2830 2831 mod_timer(&vxlan->age_timer, next_timer); 2832 } 2833 2834 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2835 { 2836 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2837 2838 spin_lock(&vn->sock_lock); 2839 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2840 #if IS_ENABLED(CONFIG_IPV6) 2841 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2842 #endif 2843 spin_unlock(&vn->sock_lock); 2844 } 2845 2846 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2847 struct vxlan_dev_node *node) 2848 { 2849 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2850 __be32 vni = vxlan->default_dst.remote_vni; 2851 2852 node->vxlan = vxlan; 2853 spin_lock(&vn->sock_lock); 2854 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2855 spin_unlock(&vn->sock_lock); 2856 } 2857 2858 /* Setup stats when device is created */ 2859 static int vxlan_init(struct net_device *dev) 2860 { 2861 struct vxlan_dev *vxlan = netdev_priv(dev); 2862 int err; 2863 2864 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2865 vxlan_vnigroup_init(vxlan); 2866 2867 err = gro_cells_init(&vxlan->gro_cells, dev); 2868 if (err) 2869 goto err_vnigroup_uninit; 2870 2871 err = vxlan_mdb_init(vxlan); 2872 if (err) 2873 goto err_gro_cells_destroy; 2874 2875 netdev_lockdep_set_classes(dev); 2876 return 0; 2877 2878 err_gro_cells_destroy: 2879 gro_cells_destroy(&vxlan->gro_cells); 2880 err_vnigroup_uninit: 2881 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2882 vxlan_vnigroup_uninit(vxlan); 2883 return err; 2884 } 2885 2886 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2887 { 2888 struct vxlan_fdb *f; 2889 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2890 2891 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2892 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2893 if (f) 2894 vxlan_fdb_destroy(vxlan, f, true, true); 2895 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2896 } 2897 2898 static void vxlan_uninit(struct net_device *dev) 2899 { 2900 struct vxlan_dev *vxlan = netdev_priv(dev); 2901 2902 vxlan_mdb_fini(vxlan); 2903 2904 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2905 vxlan_vnigroup_uninit(vxlan); 2906 2907 gro_cells_destroy(&vxlan->gro_cells); 2908 2909 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2910 } 2911 2912 /* Start ageing timer and join group when device is brought up */ 2913 static int vxlan_open(struct net_device *dev) 2914 { 2915 struct vxlan_dev *vxlan = netdev_priv(dev); 2916 int ret; 2917 2918 ret = vxlan_sock_add(vxlan); 2919 if (ret < 0) 2920 return ret; 2921 2922 ret = vxlan_multicast_join(vxlan); 2923 if (ret) { 2924 vxlan_sock_release(vxlan); 2925 return ret; 2926 } 2927 2928 if (vxlan->cfg.age_interval) 2929 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2930 2931 return ret; 2932 } 2933 2934 struct vxlan_fdb_flush_desc { 2935 bool ignore_default_entry; 2936 unsigned long state; 2937 unsigned long state_mask; 2938 unsigned long flags; 2939 unsigned long flags_mask; 2940 __be32 src_vni; 2941 u32 nhid; 2942 __be32 vni; 2943 __be16 port; 2944 union vxlan_addr dst_ip; 2945 }; 2946 2947 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2948 const struct vxlan_dev *vxlan) 2949 { 2950 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2951 } 2952 2953 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2954 { 2955 struct nexthop *nh = rtnl_dereference(f->nh); 2956 2957 return nh && nh->id == nhid; 2958 } 2959 2960 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2961 const struct vxlan_dev *vxlan, 2962 const struct vxlan_fdb_flush_desc *desc) 2963 { 2964 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2965 return false; 2966 2967 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2968 return false; 2969 2970 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2971 return false; 2972 2973 if (desc->src_vni && f->vni != desc->src_vni) 2974 return false; 2975 2976 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 2977 return false; 2978 2979 return true; 2980 } 2981 2982 static bool 2983 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 2984 { 2985 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 2986 } 2987 2988 static bool 2989 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 2990 const struct vxlan_rdst *rd) 2991 { 2992 if (desc->vni && rd->remote_vni != desc->vni) 2993 return false; 2994 2995 if (desc->port && rd->remote_port != desc->port) 2996 return false; 2997 2998 if (desc->dst_ip.sa.sa_family && 2999 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3000 return false; 3001 3002 return true; 3003 } 3004 3005 static void 3006 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3007 const struct vxlan_fdb_flush_desc *desc, 3008 bool *p_destroy_fdb) 3009 { 3010 bool remotes_flushed = false; 3011 struct vxlan_rdst *rd, *tmp; 3012 3013 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3014 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3015 continue; 3016 3017 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3018 remotes_flushed = true; 3019 } 3020 3021 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3022 } 3023 3024 /* Purge the forwarding table */ 3025 static void vxlan_flush(struct vxlan_dev *vxlan, 3026 const struct vxlan_fdb_flush_desc *desc) 3027 { 3028 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3029 unsigned int h; 3030 3031 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3032 struct hlist_node *p, *n; 3033 3034 spin_lock_bh(&vxlan->hash_lock[h]); 3035 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3036 struct vxlan_fdb *f 3037 = container_of(p, struct vxlan_fdb, hlist); 3038 3039 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3040 continue; 3041 3042 if (match_remotes) { 3043 bool destroy_fdb = false; 3044 3045 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3046 &destroy_fdb); 3047 3048 if (!destroy_fdb) 3049 continue; 3050 } 3051 3052 vxlan_fdb_destroy(vxlan, f, true, true); 3053 } 3054 spin_unlock_bh(&vxlan->hash_lock[h]); 3055 } 3056 } 3057 3058 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3059 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3060 [NDA_NH_ID] = { .type = NLA_U32 }, 3061 [NDA_VNI] = { .type = NLA_U32 }, 3062 [NDA_PORT] = { .type = NLA_U16 }, 3063 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3064 sizeof(struct in6_addr)), 3065 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3066 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3067 }; 3068 3069 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3070 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3071 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3072 NTF_ROUTER) 3073 3074 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3075 struct netlink_ext_ack *extack) 3076 { 3077 struct vxlan_dev *vxlan = netdev_priv(dev); 3078 struct vxlan_fdb_flush_desc desc = {}; 3079 struct ndmsg *ndm = nlmsg_data(nlh); 3080 struct nlattr *tb[NDA_MAX + 1]; 3081 u8 ndm_flags; 3082 int err; 3083 3084 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3085 3086 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3087 extack); 3088 if (err) 3089 return err; 3090 3091 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3092 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3093 return -EINVAL; 3094 } 3095 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3096 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3097 return -EINVAL; 3098 } 3099 3100 desc.state = ndm->ndm_state; 3101 desc.flags = ndm_flags; 3102 3103 if (tb[NDA_NDM_STATE_MASK]) 3104 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3105 3106 if (tb[NDA_NDM_FLAGS_MASK]) 3107 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3108 3109 if (tb[NDA_SRC_VNI]) 3110 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3111 3112 if (tb[NDA_NH_ID]) 3113 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3114 3115 if (tb[NDA_VNI]) 3116 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3117 3118 if (tb[NDA_PORT]) 3119 desc.port = nla_get_be16(tb[NDA_PORT]); 3120 3121 if (tb[NDA_DST]) { 3122 union vxlan_addr ip; 3123 3124 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3125 if (err) { 3126 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3127 "Unsupported address family"); 3128 return err; 3129 } 3130 desc.dst_ip = ip; 3131 } 3132 3133 vxlan_flush(vxlan, &desc); 3134 3135 return 0; 3136 } 3137 3138 /* Cleanup timer and forwarding table on shutdown */ 3139 static int vxlan_stop(struct net_device *dev) 3140 { 3141 struct vxlan_dev *vxlan = netdev_priv(dev); 3142 struct vxlan_fdb_flush_desc desc = { 3143 /* Default entry is deleted at vxlan_uninit. */ 3144 .ignore_default_entry = true, 3145 .state = 0, 3146 .state_mask = NUD_PERMANENT | NUD_NOARP, 3147 }; 3148 3149 vxlan_multicast_leave(vxlan); 3150 3151 del_timer_sync(&vxlan->age_timer); 3152 3153 vxlan_flush(vxlan, &desc); 3154 vxlan_sock_release(vxlan); 3155 3156 return 0; 3157 } 3158 3159 /* Stub, nothing needs to be done. */ 3160 static void vxlan_set_multicast_list(struct net_device *dev) 3161 { 3162 } 3163 3164 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3165 { 3166 struct vxlan_dev *vxlan = netdev_priv(dev); 3167 struct vxlan_rdst *dst = &vxlan->default_dst; 3168 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3169 dst->remote_ifindex); 3170 3171 /* This check is different than dev->max_mtu, because it looks at 3172 * the lowerdev->mtu, rather than the static dev->max_mtu 3173 */ 3174 if (lowerdev) { 3175 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3176 if (new_mtu > max_mtu) 3177 return -EINVAL; 3178 } 3179 3180 dev->mtu = new_mtu; 3181 return 0; 3182 } 3183 3184 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3185 { 3186 struct vxlan_dev *vxlan = netdev_priv(dev); 3187 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3188 __be16 sport, dport; 3189 3190 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3191 vxlan->cfg.port_max, true); 3192 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3193 3194 if (ip_tunnel_info_af(info) == AF_INET) { 3195 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3196 struct rtable *rt; 3197 3198 if (!sock4) 3199 return -EIO; 3200 3201 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3202 &info->key.u.ipv4.src, 3203 &info->key, 3204 sport, dport, info->key.tos, 3205 &info->dst_cache); 3206 if (IS_ERR(rt)) 3207 return PTR_ERR(rt); 3208 ip_rt_put(rt); 3209 } else { 3210 #if IS_ENABLED(CONFIG_IPV6) 3211 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3212 struct dst_entry *ndst; 3213 3214 if (!sock6) 3215 return -EIO; 3216 3217 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3218 0, &info->key.u.ipv6.src, 3219 &info->key, 3220 sport, dport, info->key.tos, 3221 &info->dst_cache); 3222 if (IS_ERR(ndst)) 3223 return PTR_ERR(ndst); 3224 dst_release(ndst); 3225 #else /* !CONFIG_IPV6 */ 3226 return -EPFNOSUPPORT; 3227 #endif 3228 } 3229 info->key.tp_src = sport; 3230 info->key.tp_dst = dport; 3231 return 0; 3232 } 3233 3234 static const struct net_device_ops vxlan_netdev_ether_ops = { 3235 .ndo_init = vxlan_init, 3236 .ndo_uninit = vxlan_uninit, 3237 .ndo_open = vxlan_open, 3238 .ndo_stop = vxlan_stop, 3239 .ndo_start_xmit = vxlan_xmit, 3240 .ndo_set_rx_mode = vxlan_set_multicast_list, 3241 .ndo_change_mtu = vxlan_change_mtu, 3242 .ndo_validate_addr = eth_validate_addr, 3243 .ndo_set_mac_address = eth_mac_addr, 3244 .ndo_fdb_add = vxlan_fdb_add, 3245 .ndo_fdb_del = vxlan_fdb_delete, 3246 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3247 .ndo_fdb_dump = vxlan_fdb_dump, 3248 .ndo_fdb_get = vxlan_fdb_get, 3249 .ndo_mdb_add = vxlan_mdb_add, 3250 .ndo_mdb_del = vxlan_mdb_del, 3251 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3252 .ndo_mdb_dump = vxlan_mdb_dump, 3253 .ndo_mdb_get = vxlan_mdb_get, 3254 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3255 }; 3256 3257 static const struct net_device_ops vxlan_netdev_raw_ops = { 3258 .ndo_init = vxlan_init, 3259 .ndo_uninit = vxlan_uninit, 3260 .ndo_open = vxlan_open, 3261 .ndo_stop = vxlan_stop, 3262 .ndo_start_xmit = vxlan_xmit, 3263 .ndo_change_mtu = vxlan_change_mtu, 3264 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3265 }; 3266 3267 /* Info for udev, that this is a virtual tunnel endpoint */ 3268 static const struct device_type vxlan_type = { 3269 .name = "vxlan", 3270 }; 3271 3272 /* Calls the ndo_udp_tunnel_add of the caller in order to 3273 * supply the listening VXLAN udp ports. Callers are expected 3274 * to implement the ndo_udp_tunnel_add. 3275 */ 3276 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3277 { 3278 struct vxlan_sock *vs; 3279 struct net *net = dev_net(dev); 3280 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3281 unsigned int i; 3282 3283 spin_lock(&vn->sock_lock); 3284 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3285 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3286 unsigned short type; 3287 3288 if (vs->flags & VXLAN_F_GPE) 3289 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3290 else 3291 type = UDP_TUNNEL_TYPE_VXLAN; 3292 3293 if (push) 3294 udp_tunnel_push_rx_port(dev, vs->sock, type); 3295 else 3296 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3297 } 3298 } 3299 spin_unlock(&vn->sock_lock); 3300 } 3301 3302 /* Initialize the device structure. */ 3303 static void vxlan_setup(struct net_device *dev) 3304 { 3305 struct vxlan_dev *vxlan = netdev_priv(dev); 3306 unsigned int h; 3307 3308 eth_hw_addr_random(dev); 3309 ether_setup(dev); 3310 3311 dev->needs_free_netdev = true; 3312 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3313 3314 dev->features |= NETIF_F_LLTX; 3315 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3316 dev->features |= NETIF_F_RXCSUM; 3317 dev->features |= NETIF_F_GSO_SOFTWARE; 3318 3319 dev->vlan_features = dev->features; 3320 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3321 dev->hw_features |= NETIF_F_RXCSUM; 3322 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3323 netif_keep_dst(dev); 3324 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN; 3325 3326 /* MTU range: 68 - 65535 */ 3327 dev->min_mtu = ETH_MIN_MTU; 3328 dev->max_mtu = ETH_MAX_MTU; 3329 3330 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3331 INIT_LIST_HEAD(&vxlan->next); 3332 3333 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3334 3335 vxlan->dev = dev; 3336 3337 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3338 spin_lock_init(&vxlan->hash_lock[h]); 3339 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3340 } 3341 } 3342 3343 static void vxlan_ether_setup(struct net_device *dev) 3344 { 3345 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3346 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3347 dev->netdev_ops = &vxlan_netdev_ether_ops; 3348 } 3349 3350 static void vxlan_raw_setup(struct net_device *dev) 3351 { 3352 dev->header_ops = NULL; 3353 dev->type = ARPHRD_NONE; 3354 dev->hard_header_len = 0; 3355 dev->addr_len = 0; 3356 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3357 dev->netdev_ops = &vxlan_netdev_raw_ops; 3358 } 3359 3360 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3361 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3362 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3363 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3364 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3365 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3366 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3367 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3368 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3369 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3370 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3371 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3372 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3373 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3374 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3375 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3376 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3377 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3378 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3379 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3380 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3381 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3382 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3383 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3384 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3385 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3386 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3387 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3388 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3389 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3390 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3391 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3392 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3393 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3394 }; 3395 3396 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3397 struct netlink_ext_ack *extack) 3398 { 3399 if (tb[IFLA_ADDRESS]) { 3400 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3401 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3402 "Provided link layer address is not Ethernet"); 3403 return -EINVAL; 3404 } 3405 3406 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3407 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3408 "Provided Ethernet address is not unicast"); 3409 return -EADDRNOTAVAIL; 3410 } 3411 } 3412 3413 if (tb[IFLA_MTU]) { 3414 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3415 3416 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3417 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3418 "MTU must be between 68 and 65535"); 3419 return -EINVAL; 3420 } 3421 } 3422 3423 if (!data) { 3424 NL_SET_ERR_MSG(extack, 3425 "Required attributes not provided to perform the operation"); 3426 return -EINVAL; 3427 } 3428 3429 if (data[IFLA_VXLAN_ID]) { 3430 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3431 3432 if (id >= VXLAN_N_VID) { 3433 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3434 "VXLAN ID must be lower than 16777216"); 3435 return -ERANGE; 3436 } 3437 } 3438 3439 if (data[IFLA_VXLAN_PORT_RANGE]) { 3440 const struct ifla_vxlan_port_range *p 3441 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3442 3443 if (ntohs(p->high) < ntohs(p->low)) { 3444 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3445 "Invalid source port range"); 3446 return -EINVAL; 3447 } 3448 } 3449 3450 if (data[IFLA_VXLAN_DF]) { 3451 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3452 3453 if (df < 0 || df > VXLAN_DF_MAX) { 3454 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3455 "Invalid DF attribute"); 3456 return -EINVAL; 3457 } 3458 } 3459 3460 return 0; 3461 } 3462 3463 static void vxlan_get_drvinfo(struct net_device *netdev, 3464 struct ethtool_drvinfo *drvinfo) 3465 { 3466 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3467 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3468 } 3469 3470 static int vxlan_get_link_ksettings(struct net_device *dev, 3471 struct ethtool_link_ksettings *cmd) 3472 { 3473 struct vxlan_dev *vxlan = netdev_priv(dev); 3474 struct vxlan_rdst *dst = &vxlan->default_dst; 3475 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3476 dst->remote_ifindex); 3477 3478 if (!lowerdev) { 3479 cmd->base.duplex = DUPLEX_UNKNOWN; 3480 cmd->base.port = PORT_OTHER; 3481 cmd->base.speed = SPEED_UNKNOWN; 3482 3483 return 0; 3484 } 3485 3486 return __ethtool_get_link_ksettings(lowerdev, cmd); 3487 } 3488 3489 static const struct ethtool_ops vxlan_ethtool_ops = { 3490 .get_drvinfo = vxlan_get_drvinfo, 3491 .get_link = ethtool_op_get_link, 3492 .get_link_ksettings = vxlan_get_link_ksettings, 3493 }; 3494 3495 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3496 __be16 port, u32 flags, int ifindex) 3497 { 3498 struct socket *sock; 3499 struct udp_port_cfg udp_conf; 3500 int err; 3501 3502 memset(&udp_conf, 0, sizeof(udp_conf)); 3503 3504 if (ipv6) { 3505 udp_conf.family = AF_INET6; 3506 udp_conf.use_udp6_rx_checksums = 3507 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3508 udp_conf.ipv6_v6only = 1; 3509 } else { 3510 udp_conf.family = AF_INET; 3511 } 3512 3513 udp_conf.local_udp_port = port; 3514 udp_conf.bind_ifindex = ifindex; 3515 3516 /* Open UDP socket */ 3517 err = udp_sock_create(net, &udp_conf, &sock); 3518 if (err < 0) 3519 return ERR_PTR(err); 3520 3521 udp_allow_gso(sock->sk); 3522 return sock; 3523 } 3524 3525 /* Create new listen socket if needed */ 3526 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3527 __be16 port, u32 flags, 3528 int ifindex) 3529 { 3530 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3531 struct vxlan_sock *vs; 3532 struct socket *sock; 3533 unsigned int h; 3534 struct udp_tunnel_sock_cfg tunnel_cfg; 3535 3536 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3537 if (!vs) 3538 return ERR_PTR(-ENOMEM); 3539 3540 for (h = 0; h < VNI_HASH_SIZE; ++h) 3541 INIT_HLIST_HEAD(&vs->vni_list[h]); 3542 3543 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3544 if (IS_ERR(sock)) { 3545 kfree(vs); 3546 return ERR_CAST(sock); 3547 } 3548 3549 vs->sock = sock; 3550 refcount_set(&vs->refcnt, 1); 3551 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3552 3553 spin_lock(&vn->sock_lock); 3554 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3555 udp_tunnel_notify_add_rx_port(sock, 3556 (vs->flags & VXLAN_F_GPE) ? 3557 UDP_TUNNEL_TYPE_VXLAN_GPE : 3558 UDP_TUNNEL_TYPE_VXLAN); 3559 spin_unlock(&vn->sock_lock); 3560 3561 /* Mark socket as an encapsulation socket. */ 3562 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3563 tunnel_cfg.sk_user_data = vs; 3564 tunnel_cfg.encap_type = 1; 3565 tunnel_cfg.encap_rcv = vxlan_rcv; 3566 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3567 tunnel_cfg.encap_destroy = NULL; 3568 if (vs->flags & VXLAN_F_GPE) { 3569 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3570 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3571 } else { 3572 tunnel_cfg.gro_receive = vxlan_gro_receive; 3573 tunnel_cfg.gro_complete = vxlan_gro_complete; 3574 } 3575 3576 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3577 3578 return vs; 3579 } 3580 3581 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3582 { 3583 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3584 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3585 struct vxlan_sock *vs = NULL; 3586 struct vxlan_dev_node *node; 3587 int l3mdev_index = 0; 3588 3589 if (vxlan->cfg.remote_ifindex) 3590 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3591 vxlan->net, vxlan->cfg.remote_ifindex); 3592 3593 if (!vxlan->cfg.no_share) { 3594 spin_lock(&vn->sock_lock); 3595 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3596 vxlan->cfg.dst_port, vxlan->cfg.flags, 3597 l3mdev_index); 3598 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3599 spin_unlock(&vn->sock_lock); 3600 return -EBUSY; 3601 } 3602 spin_unlock(&vn->sock_lock); 3603 } 3604 if (!vs) 3605 vs = vxlan_socket_create(vxlan->net, ipv6, 3606 vxlan->cfg.dst_port, vxlan->cfg.flags, 3607 l3mdev_index); 3608 if (IS_ERR(vs)) 3609 return PTR_ERR(vs); 3610 #if IS_ENABLED(CONFIG_IPV6) 3611 if (ipv6) { 3612 rcu_assign_pointer(vxlan->vn6_sock, vs); 3613 node = &vxlan->hlist6; 3614 } else 3615 #endif 3616 { 3617 rcu_assign_pointer(vxlan->vn4_sock, vs); 3618 node = &vxlan->hlist4; 3619 } 3620 3621 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3622 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3623 else 3624 vxlan_vs_add_dev(vs, vxlan, node); 3625 3626 return 0; 3627 } 3628 3629 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3630 { 3631 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3632 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3633 bool ipv4 = !ipv6 || metadata; 3634 int ret = 0; 3635 3636 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3637 #if IS_ENABLED(CONFIG_IPV6) 3638 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3639 if (ipv6) { 3640 ret = __vxlan_sock_add(vxlan, true); 3641 if (ret < 0 && ret != -EAFNOSUPPORT) 3642 ipv4 = false; 3643 } 3644 #endif 3645 if (ipv4) 3646 ret = __vxlan_sock_add(vxlan, false); 3647 if (ret < 0) 3648 vxlan_sock_release(vxlan); 3649 return ret; 3650 } 3651 3652 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3653 struct vxlan_config *conf, __be32 vni) 3654 { 3655 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3656 struct vxlan_dev *tmp; 3657 3658 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3659 if (tmp == vxlan) 3660 continue; 3661 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3662 if (!vxlan_vnifilter_lookup(tmp, vni)) 3663 continue; 3664 } else if (tmp->cfg.vni != vni) { 3665 continue; 3666 } 3667 if (tmp->cfg.dst_port != conf->dst_port) 3668 continue; 3669 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3670 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3671 continue; 3672 3673 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3674 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3675 continue; 3676 3677 return -EEXIST; 3678 } 3679 3680 return 0; 3681 } 3682 3683 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3684 struct net_device **lower, 3685 struct vxlan_dev *old, 3686 struct netlink_ext_ack *extack) 3687 { 3688 bool use_ipv6 = false; 3689 3690 if (conf->flags & VXLAN_F_GPE) { 3691 /* For now, allow GPE only together with 3692 * COLLECT_METADATA. This can be relaxed later; in such 3693 * case, the other side of the PtP link will have to be 3694 * provided. 3695 */ 3696 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3697 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3698 NL_SET_ERR_MSG(extack, 3699 "VXLAN GPE does not support this combination of attributes"); 3700 return -EINVAL; 3701 } 3702 } 3703 3704 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3705 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3706 conf->remote_ip.sa.sa_family = AF_INET; 3707 conf->saddr.sa.sa_family = AF_INET; 3708 } else if (!conf->remote_ip.sa.sa_family) { 3709 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3710 } else if (!conf->saddr.sa.sa_family) { 3711 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3712 } 3713 3714 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3715 NL_SET_ERR_MSG(extack, 3716 "Local and remote address must be from the same family"); 3717 return -EINVAL; 3718 } 3719 3720 if (vxlan_addr_multicast(&conf->saddr)) { 3721 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3722 return -EINVAL; 3723 } 3724 3725 if (conf->saddr.sa.sa_family == AF_INET6) { 3726 if (!IS_ENABLED(CONFIG_IPV6)) { 3727 NL_SET_ERR_MSG(extack, 3728 "IPv6 support not enabled in the kernel"); 3729 return -EPFNOSUPPORT; 3730 } 3731 use_ipv6 = true; 3732 conf->flags |= VXLAN_F_IPV6; 3733 3734 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3735 int local_type = 3736 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3737 int remote_type = 3738 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3739 3740 if (local_type & IPV6_ADDR_LINKLOCAL) { 3741 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3742 (remote_type != IPV6_ADDR_ANY)) { 3743 NL_SET_ERR_MSG(extack, 3744 "Invalid combination of local and remote address scopes"); 3745 return -EINVAL; 3746 } 3747 3748 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3749 } else { 3750 if (remote_type == 3751 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3752 NL_SET_ERR_MSG(extack, 3753 "Invalid combination of local and remote address scopes"); 3754 return -EINVAL; 3755 } 3756 3757 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3758 } 3759 } 3760 } 3761 3762 if (conf->label && !use_ipv6) { 3763 NL_SET_ERR_MSG(extack, 3764 "Label attribute only applies to IPv6 VXLAN devices"); 3765 return -EINVAL; 3766 } 3767 3768 if (conf->label_policy && !use_ipv6) { 3769 NL_SET_ERR_MSG(extack, 3770 "Label policy only applies to IPv6 VXLAN devices"); 3771 return -EINVAL; 3772 } 3773 3774 if (conf->remote_ifindex) { 3775 struct net_device *lowerdev; 3776 3777 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3778 if (!lowerdev) { 3779 NL_SET_ERR_MSG(extack, 3780 "Invalid local interface, device not found"); 3781 return -ENODEV; 3782 } 3783 3784 #if IS_ENABLED(CONFIG_IPV6) 3785 if (use_ipv6) { 3786 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3787 3788 if (idev && idev->cnf.disable_ipv6) { 3789 NL_SET_ERR_MSG(extack, 3790 "IPv6 support disabled by administrator"); 3791 return -EPERM; 3792 } 3793 } 3794 #endif 3795 3796 *lower = lowerdev; 3797 } else { 3798 if (vxlan_addr_multicast(&conf->remote_ip)) { 3799 NL_SET_ERR_MSG(extack, 3800 "Local interface required for multicast remote destination"); 3801 3802 return -EINVAL; 3803 } 3804 3805 #if IS_ENABLED(CONFIG_IPV6) 3806 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3807 NL_SET_ERR_MSG(extack, 3808 "Local interface required for link-local local/remote addresses"); 3809 return -EINVAL; 3810 } 3811 #endif 3812 3813 *lower = NULL; 3814 } 3815 3816 if (!conf->dst_port) { 3817 if (conf->flags & VXLAN_F_GPE) 3818 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3819 else 3820 conf->dst_port = htons(vxlan_port); 3821 } 3822 3823 if (!conf->age_interval) 3824 conf->age_interval = FDB_AGE_DEFAULT; 3825 3826 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3827 NL_SET_ERR_MSG(extack, 3828 "A VXLAN device with the specified VNI already exists"); 3829 return -EEXIST; 3830 } 3831 3832 return 0; 3833 } 3834 3835 static void vxlan_config_apply(struct net_device *dev, 3836 struct vxlan_config *conf, 3837 struct net_device *lowerdev, 3838 struct net *src_net, 3839 bool changelink) 3840 { 3841 struct vxlan_dev *vxlan = netdev_priv(dev); 3842 struct vxlan_rdst *dst = &vxlan->default_dst; 3843 unsigned short needed_headroom = ETH_HLEN; 3844 int max_mtu = ETH_MAX_MTU; 3845 u32 flags = conf->flags; 3846 3847 if (!changelink) { 3848 if (flags & VXLAN_F_GPE) 3849 vxlan_raw_setup(dev); 3850 else 3851 vxlan_ether_setup(dev); 3852 3853 if (conf->mtu) 3854 dev->mtu = conf->mtu; 3855 3856 vxlan->net = src_net; 3857 } 3858 3859 dst->remote_vni = conf->vni; 3860 3861 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3862 3863 if (lowerdev) { 3864 dst->remote_ifindex = conf->remote_ifindex; 3865 3866 netif_inherit_tso_max(dev, lowerdev); 3867 3868 needed_headroom = lowerdev->hard_header_len; 3869 needed_headroom += lowerdev->needed_headroom; 3870 3871 dev->needed_tailroom = lowerdev->needed_tailroom; 3872 3873 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3874 if (max_mtu < ETH_MIN_MTU) 3875 max_mtu = ETH_MIN_MTU; 3876 3877 if (!changelink && !conf->mtu) 3878 dev->mtu = max_mtu; 3879 } 3880 3881 if (dev->mtu > max_mtu) 3882 dev->mtu = max_mtu; 3883 3884 if (flags & VXLAN_F_COLLECT_METADATA) 3885 flags |= VXLAN_F_IPV6; 3886 needed_headroom += vxlan_headroom(flags); 3887 dev->needed_headroom = needed_headroom; 3888 3889 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3890 } 3891 3892 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3893 struct vxlan_config *conf, bool changelink, 3894 struct netlink_ext_ack *extack) 3895 { 3896 struct vxlan_dev *vxlan = netdev_priv(dev); 3897 struct net_device *lowerdev; 3898 int ret; 3899 3900 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3901 if (ret) 3902 return ret; 3903 3904 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3905 3906 return 0; 3907 } 3908 3909 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3910 struct vxlan_config *conf, 3911 struct netlink_ext_ack *extack) 3912 { 3913 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3914 struct vxlan_dev *vxlan = netdev_priv(dev); 3915 struct net_device *remote_dev = NULL; 3916 struct vxlan_fdb *f = NULL; 3917 bool unregister = false; 3918 struct vxlan_rdst *dst; 3919 int err; 3920 3921 dst = &vxlan->default_dst; 3922 err = vxlan_dev_configure(net, dev, conf, false, extack); 3923 if (err) 3924 return err; 3925 3926 dev->ethtool_ops = &vxlan_ethtool_ops; 3927 3928 /* create an fdb entry for a valid default destination */ 3929 if (!vxlan_addr_any(&dst->remote_ip)) { 3930 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3931 &dst->remote_ip, 3932 NUD_REACHABLE | NUD_PERMANENT, 3933 vxlan->cfg.dst_port, 3934 dst->remote_vni, 3935 dst->remote_vni, 3936 dst->remote_ifindex, 3937 NTF_SELF, 0, &f, extack); 3938 if (err) 3939 return err; 3940 } 3941 3942 err = register_netdevice(dev); 3943 if (err) 3944 goto errout; 3945 unregister = true; 3946 3947 if (dst->remote_ifindex) { 3948 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3949 if (!remote_dev) { 3950 err = -ENODEV; 3951 goto errout; 3952 } 3953 3954 err = netdev_upper_dev_link(remote_dev, dev, extack); 3955 if (err) 3956 goto errout; 3957 } 3958 3959 err = rtnl_configure_link(dev, NULL, 0, NULL); 3960 if (err < 0) 3961 goto unlink; 3962 3963 if (f) { 3964 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3965 3966 /* notify default fdb entry */ 3967 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3968 RTM_NEWNEIGH, true, extack); 3969 if (err) { 3970 vxlan_fdb_destroy(vxlan, f, false, false); 3971 if (remote_dev) 3972 netdev_upper_dev_unlink(remote_dev, dev); 3973 goto unregister; 3974 } 3975 } 3976 3977 list_add(&vxlan->next, &vn->vxlan_list); 3978 if (remote_dev) 3979 dst->remote_dev = remote_dev; 3980 return 0; 3981 unlink: 3982 if (remote_dev) 3983 netdev_upper_dev_unlink(remote_dev, dev); 3984 errout: 3985 /* unregister_netdevice() destroys the default FDB entry with deletion 3986 * notification. But the addition notification was not sent yet, so 3987 * destroy the entry by hand here. 3988 */ 3989 if (f) 3990 __vxlan_fdb_free(f); 3991 unregister: 3992 if (unregister) 3993 unregister_netdevice(dev); 3994 return err; 3995 } 3996 3997 /* Set/clear flags based on attribute */ 3998 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 3999 int attrtype, unsigned long mask, bool changelink, 4000 bool changelink_supported, 4001 struct netlink_ext_ack *extack) 4002 { 4003 unsigned long flags; 4004 4005 if (!tb[attrtype]) 4006 return 0; 4007 4008 if (changelink && !changelink_supported) { 4009 vxlan_flag_attr_error(attrtype, extack); 4010 return -EOPNOTSUPP; 4011 } 4012 4013 if (vxlan_policy[attrtype].type == NLA_FLAG) 4014 flags = conf->flags | mask; 4015 else if (nla_get_u8(tb[attrtype])) 4016 flags = conf->flags | mask; 4017 else 4018 flags = conf->flags & ~mask; 4019 4020 conf->flags = flags; 4021 4022 return 0; 4023 } 4024 4025 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4026 struct net_device *dev, struct vxlan_config *conf, 4027 bool changelink, struct netlink_ext_ack *extack) 4028 { 4029 struct vxlan_dev *vxlan = netdev_priv(dev); 4030 int err = 0; 4031 4032 memset(conf, 0, sizeof(*conf)); 4033 4034 /* if changelink operation, start with old existing cfg */ 4035 if (changelink) 4036 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4037 4038 if (data[IFLA_VXLAN_ID]) { 4039 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4040 4041 if (changelink && (vni != conf->vni)) { 4042 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4043 return -EOPNOTSUPP; 4044 } 4045 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4046 } 4047 4048 if (data[IFLA_VXLAN_GROUP]) { 4049 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4050 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4051 return -EOPNOTSUPP; 4052 } 4053 4054 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4055 conf->remote_ip.sa.sa_family = AF_INET; 4056 } else if (data[IFLA_VXLAN_GROUP6]) { 4057 if (!IS_ENABLED(CONFIG_IPV6)) { 4058 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4059 return -EPFNOSUPPORT; 4060 } 4061 4062 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4063 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4064 return -EOPNOTSUPP; 4065 } 4066 4067 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4068 conf->remote_ip.sa.sa_family = AF_INET6; 4069 } 4070 4071 if (data[IFLA_VXLAN_LOCAL]) { 4072 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4073 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4074 return -EOPNOTSUPP; 4075 } 4076 4077 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4078 conf->saddr.sa.sa_family = AF_INET; 4079 } else if (data[IFLA_VXLAN_LOCAL6]) { 4080 if (!IS_ENABLED(CONFIG_IPV6)) { 4081 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4082 return -EPFNOSUPPORT; 4083 } 4084 4085 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4086 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4087 return -EOPNOTSUPP; 4088 } 4089 4090 /* TODO: respect scope id */ 4091 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4092 conf->saddr.sa.sa_family = AF_INET6; 4093 } 4094 4095 if (data[IFLA_VXLAN_LINK]) 4096 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4097 4098 if (data[IFLA_VXLAN_TOS]) 4099 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4100 4101 if (data[IFLA_VXLAN_TTL]) 4102 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4103 4104 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4105 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4106 VXLAN_F_TTL_INHERIT, changelink, false, 4107 extack); 4108 if (err) 4109 return err; 4110 4111 } 4112 4113 if (data[IFLA_VXLAN_LABEL]) 4114 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4115 IPV6_FLOWLABEL_MASK; 4116 if (data[IFLA_VXLAN_LABEL_POLICY]) 4117 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4118 4119 if (data[IFLA_VXLAN_LEARNING]) { 4120 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4121 VXLAN_F_LEARN, changelink, true, 4122 extack); 4123 if (err) 4124 return err; 4125 } else if (!changelink) { 4126 /* default to learn on a new device */ 4127 conf->flags |= VXLAN_F_LEARN; 4128 } 4129 4130 if (data[IFLA_VXLAN_AGEING]) 4131 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4132 4133 if (data[IFLA_VXLAN_PROXY]) { 4134 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4135 VXLAN_F_PROXY, changelink, false, 4136 extack); 4137 if (err) 4138 return err; 4139 } 4140 4141 if (data[IFLA_VXLAN_RSC]) { 4142 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4143 VXLAN_F_RSC, changelink, false, 4144 extack); 4145 if (err) 4146 return err; 4147 } 4148 4149 if (data[IFLA_VXLAN_L2MISS]) { 4150 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4151 VXLAN_F_L2MISS, changelink, false, 4152 extack); 4153 if (err) 4154 return err; 4155 } 4156 4157 if (data[IFLA_VXLAN_L3MISS]) { 4158 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4159 VXLAN_F_L3MISS, changelink, false, 4160 extack); 4161 if (err) 4162 return err; 4163 } 4164 4165 if (data[IFLA_VXLAN_LIMIT]) { 4166 if (changelink) { 4167 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4168 "Cannot change limit"); 4169 return -EOPNOTSUPP; 4170 } 4171 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4172 } 4173 4174 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4175 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4176 VXLAN_F_COLLECT_METADATA, changelink, false, 4177 extack); 4178 if (err) 4179 return err; 4180 } 4181 4182 if (data[IFLA_VXLAN_PORT_RANGE]) { 4183 if (!changelink) { 4184 const struct ifla_vxlan_port_range *p 4185 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4186 conf->port_min = ntohs(p->low); 4187 conf->port_max = ntohs(p->high); 4188 } else { 4189 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4190 "Cannot change port range"); 4191 return -EOPNOTSUPP; 4192 } 4193 } 4194 4195 if (data[IFLA_VXLAN_PORT]) { 4196 if (changelink) { 4197 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4198 "Cannot change port"); 4199 return -EOPNOTSUPP; 4200 } 4201 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4202 } 4203 4204 if (data[IFLA_VXLAN_UDP_CSUM]) { 4205 if (changelink) { 4206 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4207 "Cannot change UDP_CSUM flag"); 4208 return -EOPNOTSUPP; 4209 } 4210 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4211 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4212 } 4213 4214 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4215 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4216 VXLAN_F_LOCALBYPASS, changelink, 4217 true, extack); 4218 if (err) 4219 return err; 4220 } else if (!changelink) { 4221 /* default to local bypass on a new device */ 4222 conf->flags |= VXLAN_F_LOCALBYPASS; 4223 } 4224 4225 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4226 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4227 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4228 false, extack); 4229 if (err) 4230 return err; 4231 } 4232 4233 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4234 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4235 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4236 false, extack); 4237 if (err) 4238 return err; 4239 } 4240 4241 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4242 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4243 VXLAN_F_REMCSUM_TX, changelink, false, 4244 extack); 4245 if (err) 4246 return err; 4247 } 4248 4249 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4250 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4251 VXLAN_F_REMCSUM_RX, changelink, false, 4252 extack); 4253 if (err) 4254 return err; 4255 } 4256 4257 if (data[IFLA_VXLAN_GBP]) { 4258 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4259 VXLAN_F_GBP, changelink, false, extack); 4260 if (err) 4261 return err; 4262 } 4263 4264 if (data[IFLA_VXLAN_GPE]) { 4265 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4266 VXLAN_F_GPE, changelink, false, 4267 extack); 4268 if (err) 4269 return err; 4270 } 4271 4272 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4273 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4274 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4275 false, extack); 4276 if (err) 4277 return err; 4278 } 4279 4280 if (tb[IFLA_MTU]) { 4281 if (changelink) { 4282 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4283 "Cannot change mtu"); 4284 return -EOPNOTSUPP; 4285 } 4286 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4287 } 4288 4289 if (data[IFLA_VXLAN_DF]) 4290 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4291 4292 if (data[IFLA_VXLAN_VNIFILTER]) { 4293 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4294 VXLAN_F_VNIFILTER, changelink, false, 4295 extack); 4296 if (err) 4297 return err; 4298 4299 if ((conf->flags & VXLAN_F_VNIFILTER) && 4300 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4301 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4302 "vxlan vnifilter only valid in collect metadata mode"); 4303 return -EINVAL; 4304 } 4305 } 4306 4307 return 0; 4308 } 4309 4310 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4311 struct nlattr *tb[], struct nlattr *data[], 4312 struct netlink_ext_ack *extack) 4313 { 4314 struct vxlan_config conf; 4315 int err; 4316 4317 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4318 if (err) 4319 return err; 4320 4321 return __vxlan_dev_create(src_net, dev, &conf, extack); 4322 } 4323 4324 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4325 struct nlattr *data[], 4326 struct netlink_ext_ack *extack) 4327 { 4328 struct vxlan_dev *vxlan = netdev_priv(dev); 4329 struct net_device *lowerdev; 4330 struct vxlan_config conf; 4331 struct vxlan_rdst *dst; 4332 int err; 4333 4334 dst = &vxlan->default_dst; 4335 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4336 if (err) 4337 return err; 4338 4339 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4340 vxlan, extack); 4341 if (err) 4342 return err; 4343 4344 if (dst->remote_dev == lowerdev) 4345 lowerdev = NULL; 4346 4347 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4348 extack); 4349 if (err) 4350 return err; 4351 4352 /* handle default dst entry */ 4353 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4354 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4355 4356 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4357 if (!vxlan_addr_any(&conf.remote_ip)) { 4358 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4359 &conf.remote_ip, 4360 NUD_REACHABLE | NUD_PERMANENT, 4361 NLM_F_APPEND | NLM_F_CREATE, 4362 vxlan->cfg.dst_port, 4363 conf.vni, conf.vni, 4364 conf.remote_ifindex, 4365 NTF_SELF, 0, true, extack); 4366 if (err) { 4367 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4368 netdev_adjacent_change_abort(dst->remote_dev, 4369 lowerdev, dev); 4370 return err; 4371 } 4372 } 4373 if (!vxlan_addr_any(&dst->remote_ip)) 4374 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4375 dst->remote_ip, 4376 vxlan->cfg.dst_port, 4377 dst->remote_vni, 4378 dst->remote_vni, 4379 dst->remote_ifindex, 4380 true); 4381 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4382 4383 /* If vni filtering device, also update fdb entries of 4384 * all vnis that were using default remote ip 4385 */ 4386 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4387 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4388 &conf.remote_ip, extack); 4389 if (err) { 4390 netdev_adjacent_change_abort(dst->remote_dev, 4391 lowerdev, dev); 4392 return err; 4393 } 4394 } 4395 } 4396 4397 if (conf.age_interval != vxlan->cfg.age_interval) 4398 mod_timer(&vxlan->age_timer, jiffies); 4399 4400 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4401 if (lowerdev && lowerdev != dst->remote_dev) 4402 dst->remote_dev = lowerdev; 4403 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4404 return 0; 4405 } 4406 4407 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4408 { 4409 struct vxlan_dev *vxlan = netdev_priv(dev); 4410 struct vxlan_fdb_flush_desc desc = { 4411 /* Default entry is deleted at vxlan_uninit. */ 4412 .ignore_default_entry = true, 4413 }; 4414 4415 vxlan_flush(vxlan, &desc); 4416 4417 list_del(&vxlan->next); 4418 unregister_netdevice_queue(dev, head); 4419 if (vxlan->default_dst.remote_dev) 4420 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4421 } 4422 4423 static size_t vxlan_get_size(const struct net_device *dev) 4424 { 4425 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4426 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4427 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4428 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4429 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4430 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4431 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4432 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4433 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4434 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4435 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4436 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4437 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4438 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4439 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4440 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4441 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4442 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4443 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4444 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4445 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4446 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4447 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4448 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4449 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4450 /* IFLA_VXLAN_PORT_RANGE */ 4451 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4452 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4453 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4454 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4455 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4456 0; 4457 } 4458 4459 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4460 { 4461 const struct vxlan_dev *vxlan = netdev_priv(dev); 4462 const struct vxlan_rdst *dst = &vxlan->default_dst; 4463 struct ifla_vxlan_port_range ports = { 4464 .low = htons(vxlan->cfg.port_min), 4465 .high = htons(vxlan->cfg.port_max), 4466 }; 4467 4468 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4469 goto nla_put_failure; 4470 4471 if (!vxlan_addr_any(&dst->remote_ip)) { 4472 if (dst->remote_ip.sa.sa_family == AF_INET) { 4473 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4474 dst->remote_ip.sin.sin_addr.s_addr)) 4475 goto nla_put_failure; 4476 #if IS_ENABLED(CONFIG_IPV6) 4477 } else { 4478 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4479 &dst->remote_ip.sin6.sin6_addr)) 4480 goto nla_put_failure; 4481 #endif 4482 } 4483 } 4484 4485 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4486 goto nla_put_failure; 4487 4488 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4489 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4490 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4491 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4492 goto nla_put_failure; 4493 #if IS_ENABLED(CONFIG_IPV6) 4494 } else { 4495 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4496 &vxlan->cfg.saddr.sin6.sin6_addr)) 4497 goto nla_put_failure; 4498 #endif 4499 } 4500 } 4501 4502 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4503 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4504 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4505 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4506 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4507 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4508 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4509 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4510 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4511 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4512 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4513 nla_put_u8(skb, IFLA_VXLAN_RSC, 4514 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4515 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4516 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4517 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4518 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4519 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4520 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4521 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4522 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4523 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4524 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4525 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4526 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4527 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4528 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4529 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4530 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4531 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4532 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4533 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4534 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4535 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4536 goto nla_put_failure; 4537 4538 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4539 goto nla_put_failure; 4540 4541 if (vxlan->cfg.flags & VXLAN_F_GBP && 4542 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4543 goto nla_put_failure; 4544 4545 if (vxlan->cfg.flags & VXLAN_F_GPE && 4546 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4547 goto nla_put_failure; 4548 4549 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4550 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4551 goto nla_put_failure; 4552 4553 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4554 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4555 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4556 goto nla_put_failure; 4557 4558 return 0; 4559 4560 nla_put_failure: 4561 return -EMSGSIZE; 4562 } 4563 4564 static struct net *vxlan_get_link_net(const struct net_device *dev) 4565 { 4566 struct vxlan_dev *vxlan = netdev_priv(dev); 4567 4568 return vxlan->net; 4569 } 4570 4571 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4572 .kind = "vxlan", 4573 .maxtype = IFLA_VXLAN_MAX, 4574 .policy = vxlan_policy, 4575 .priv_size = sizeof(struct vxlan_dev), 4576 .setup = vxlan_setup, 4577 .validate = vxlan_validate, 4578 .newlink = vxlan_newlink, 4579 .changelink = vxlan_changelink, 4580 .dellink = vxlan_dellink, 4581 .get_size = vxlan_get_size, 4582 .fill_info = vxlan_fill_info, 4583 .get_link_net = vxlan_get_link_net, 4584 }; 4585 4586 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4587 u8 name_assign_type, 4588 struct vxlan_config *conf) 4589 { 4590 struct nlattr *tb[IFLA_MAX + 1]; 4591 struct net_device *dev; 4592 int err; 4593 4594 memset(&tb, 0, sizeof(tb)); 4595 4596 dev = rtnl_create_link(net, name, name_assign_type, 4597 &vxlan_link_ops, tb, NULL); 4598 if (IS_ERR(dev)) 4599 return dev; 4600 4601 err = __vxlan_dev_create(net, dev, conf, NULL); 4602 if (err < 0) { 4603 free_netdev(dev); 4604 return ERR_PTR(err); 4605 } 4606 4607 err = rtnl_configure_link(dev, NULL, 0, NULL); 4608 if (err < 0) { 4609 LIST_HEAD(list_kill); 4610 4611 vxlan_dellink(dev, &list_kill); 4612 unregister_netdevice_many(&list_kill); 4613 return ERR_PTR(err); 4614 } 4615 4616 return dev; 4617 } 4618 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4619 4620 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4621 struct net_device *dev) 4622 { 4623 struct vxlan_dev *vxlan, *next; 4624 LIST_HEAD(list_kill); 4625 4626 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4627 struct vxlan_rdst *dst = &vxlan->default_dst; 4628 4629 /* In case we created vxlan device with carrier 4630 * and we loose the carrier due to module unload 4631 * we also need to remove vxlan device. In other 4632 * cases, it's not necessary and remote_ifindex 4633 * is 0 here, so no matches. 4634 */ 4635 if (dst->remote_ifindex == dev->ifindex) 4636 vxlan_dellink(vxlan->dev, &list_kill); 4637 } 4638 4639 unregister_netdevice_many(&list_kill); 4640 } 4641 4642 static int vxlan_netdevice_event(struct notifier_block *unused, 4643 unsigned long event, void *ptr) 4644 { 4645 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4646 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4647 4648 if (event == NETDEV_UNREGISTER) 4649 vxlan_handle_lowerdev_unregister(vn, dev); 4650 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4651 vxlan_offload_rx_ports(dev, true); 4652 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4653 vxlan_offload_rx_ports(dev, false); 4654 4655 return NOTIFY_DONE; 4656 } 4657 4658 static struct notifier_block vxlan_notifier_block __read_mostly = { 4659 .notifier_call = vxlan_netdevice_event, 4660 }; 4661 4662 static void 4663 vxlan_fdb_offloaded_set(struct net_device *dev, 4664 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4665 { 4666 struct vxlan_dev *vxlan = netdev_priv(dev); 4667 struct vxlan_rdst *rdst; 4668 struct vxlan_fdb *f; 4669 u32 hash_index; 4670 4671 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4672 4673 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4674 4675 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4676 if (!f) 4677 goto out; 4678 4679 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4680 fdb_info->remote_port, 4681 fdb_info->remote_vni, 4682 fdb_info->remote_ifindex); 4683 if (!rdst) 4684 goto out; 4685 4686 rdst->offloaded = fdb_info->offloaded; 4687 4688 out: 4689 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4690 } 4691 4692 static int 4693 vxlan_fdb_external_learn_add(struct net_device *dev, 4694 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4695 { 4696 struct vxlan_dev *vxlan = netdev_priv(dev); 4697 struct netlink_ext_ack *extack; 4698 u32 hash_index; 4699 int err; 4700 4701 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4702 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4703 4704 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4705 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4706 NUD_REACHABLE, 4707 NLM_F_CREATE | NLM_F_REPLACE, 4708 fdb_info->remote_port, 4709 fdb_info->vni, 4710 fdb_info->remote_vni, 4711 fdb_info->remote_ifindex, 4712 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4713 0, false, extack); 4714 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4715 4716 return err; 4717 } 4718 4719 static int 4720 vxlan_fdb_external_learn_del(struct net_device *dev, 4721 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4722 { 4723 struct vxlan_dev *vxlan = netdev_priv(dev); 4724 struct vxlan_fdb *f; 4725 u32 hash_index; 4726 int err = 0; 4727 4728 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4729 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4730 4731 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4732 if (!f) 4733 err = -ENOENT; 4734 else if (f->flags & NTF_EXT_LEARNED) 4735 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4736 fdb_info->remote_ip, 4737 fdb_info->remote_port, 4738 fdb_info->vni, 4739 fdb_info->remote_vni, 4740 fdb_info->remote_ifindex, 4741 false); 4742 4743 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4744 4745 return err; 4746 } 4747 4748 static int vxlan_switchdev_event(struct notifier_block *unused, 4749 unsigned long event, void *ptr) 4750 { 4751 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4752 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4753 int err = 0; 4754 4755 switch (event) { 4756 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4757 vxlan_fdb_offloaded_set(dev, ptr); 4758 break; 4759 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4760 fdb_info = ptr; 4761 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4762 if (err) { 4763 err = notifier_from_errno(err); 4764 break; 4765 } 4766 fdb_info->offloaded = true; 4767 vxlan_fdb_offloaded_set(dev, fdb_info); 4768 break; 4769 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4770 fdb_info = ptr; 4771 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4772 if (err) { 4773 err = notifier_from_errno(err); 4774 break; 4775 } 4776 fdb_info->offloaded = false; 4777 vxlan_fdb_offloaded_set(dev, fdb_info); 4778 break; 4779 } 4780 4781 return err; 4782 } 4783 4784 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4785 .notifier_call = vxlan_switchdev_event, 4786 }; 4787 4788 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4789 { 4790 struct vxlan_fdb *fdb; 4791 struct vxlan_dev *vxlan; 4792 u32 hash_index; 4793 4794 rcu_read_lock(); 4795 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4796 vxlan = rcu_dereference(fdb->vdev); 4797 WARN_ON(!vxlan); 4798 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4799 vxlan->default_dst.remote_vni); 4800 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4801 if (!hlist_unhashed(&fdb->hlist)) 4802 vxlan_fdb_destroy(vxlan, fdb, false, false); 4803 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4804 } 4805 rcu_read_unlock(); 4806 } 4807 4808 static int vxlan_nexthop_event(struct notifier_block *nb, 4809 unsigned long event, void *ptr) 4810 { 4811 struct nh_notifier_info *info = ptr; 4812 struct nexthop *nh; 4813 4814 if (event != NEXTHOP_EVENT_DEL) 4815 return NOTIFY_DONE; 4816 4817 nh = nexthop_find_by_id(info->net, info->id); 4818 if (!nh) 4819 return NOTIFY_DONE; 4820 4821 vxlan_fdb_nh_flush(nh); 4822 4823 return NOTIFY_DONE; 4824 } 4825 4826 static __net_init int vxlan_init_net(struct net *net) 4827 { 4828 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4829 unsigned int h; 4830 4831 INIT_LIST_HEAD(&vn->vxlan_list); 4832 spin_lock_init(&vn->sock_lock); 4833 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4834 4835 for (h = 0; h < PORT_HASH_SIZE; ++h) 4836 INIT_HLIST_HEAD(&vn->sock_list[h]); 4837 4838 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4839 NULL); 4840 } 4841 4842 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4843 struct list_head *dev_to_kill) 4844 { 4845 struct vxlan_dev *vxlan, *next; 4846 4847 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4848 vxlan_dellink(vxlan->dev, dev_to_kill); 4849 } 4850 4851 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4852 struct list_head *dev_to_kill) 4853 { 4854 struct net *net; 4855 4856 ASSERT_RTNL(); 4857 list_for_each_entry(net, net_list, exit_list) { 4858 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4859 4860 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4861 4862 vxlan_destroy_tunnels(vn, dev_to_kill); 4863 } 4864 } 4865 4866 static void __net_exit vxlan_exit_net(struct net *net) 4867 { 4868 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4869 unsigned int h; 4870 4871 for (h = 0; h < PORT_HASH_SIZE; ++h) 4872 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4873 } 4874 4875 static struct pernet_operations vxlan_net_ops = { 4876 .init = vxlan_init_net, 4877 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4878 .exit = vxlan_exit_net, 4879 .id = &vxlan_net_id, 4880 .size = sizeof(struct vxlan_net), 4881 }; 4882 4883 static int __init vxlan_init_module(void) 4884 { 4885 int rc; 4886 4887 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4888 4889 rc = register_pernet_subsys(&vxlan_net_ops); 4890 if (rc) 4891 goto out1; 4892 4893 rc = register_netdevice_notifier(&vxlan_notifier_block); 4894 if (rc) 4895 goto out2; 4896 4897 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4898 if (rc) 4899 goto out3; 4900 4901 rc = rtnl_link_register(&vxlan_link_ops); 4902 if (rc) 4903 goto out4; 4904 4905 vxlan_vnifilter_init(); 4906 4907 return 0; 4908 out4: 4909 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4910 out3: 4911 unregister_netdevice_notifier(&vxlan_notifier_block); 4912 out2: 4913 unregister_pernet_subsys(&vxlan_net_ops); 4914 out1: 4915 return rc; 4916 } 4917 late_initcall(vxlan_init_module); 4918 4919 static void __exit vxlan_cleanup_module(void) 4920 { 4921 vxlan_vnifilter_uninit(); 4922 rtnl_link_unregister(&vxlan_link_ops); 4923 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4924 unregister_netdevice_notifier(&vxlan_notifier_block); 4925 unregister_pernet_subsys(&vxlan_net_ops); 4926 /* rcu_barrier() is called by netns */ 4927 } 4928 module_exit(vxlan_cleanup_module); 4929 4930 MODULE_LICENSE("GPL"); 4931 MODULE_VERSION(VXLAN_VERSION); 4932 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4933 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4934 MODULE_ALIAS_RTNL_LINK("vxlan"); 4935