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 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1588 tun_dst->u.tun_info.key.tun_flags); 1589 tun_dst->u.tun_info.options_len = sizeof(*md); 1590 } 1591 if (gbp->dont_learn) 1592 md->gbp |= VXLAN_GBP_DONT_LEARN; 1593 1594 if (gbp->policy_applied) 1595 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1596 1597 /* In flow-based mode, GBP is carried in dst_metadata */ 1598 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1599 skb->mark = md->gbp; 1600 out: 1601 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS; 1602 } 1603 1604 static bool vxlan_set_mac(struct vxlan_dev *vxlan, 1605 struct vxlan_sock *vs, 1606 struct sk_buff *skb, __be32 vni) 1607 { 1608 union vxlan_addr saddr; 1609 u32 ifindex = skb->dev->ifindex; 1610 1611 skb_reset_mac_header(skb); 1612 skb->protocol = eth_type_trans(skb, vxlan->dev); 1613 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1614 1615 /* Ignore packet loops (and multicast echo) */ 1616 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1617 return false; 1618 1619 /* Ignore packets from invalid src-address */ 1620 if (!is_valid_ether_addr(eth_hdr(skb)->h_source)) 1621 return false; 1622 1623 /* Get address from the outer IP header */ 1624 if (vxlan_get_sk_family(vs) == AF_INET) { 1625 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1626 saddr.sa.sa_family = AF_INET; 1627 #if IS_ENABLED(CONFIG_IPV6) 1628 } else { 1629 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1630 saddr.sa.sa_family = AF_INET6; 1631 #endif 1632 } 1633 1634 if ((vxlan->cfg.flags & VXLAN_F_LEARN) && 1635 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni)) 1636 return false; 1637 1638 return true; 1639 } 1640 1641 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1642 struct sk_buff *skb) 1643 { 1644 int err = 0; 1645 1646 if (vxlan_get_sk_family(vs) == AF_INET) 1647 err = IP_ECN_decapsulate(oiph, skb); 1648 #if IS_ENABLED(CONFIG_IPV6) 1649 else 1650 err = IP6_ECN_decapsulate(oiph, skb); 1651 #endif 1652 1653 if (unlikely(err) && log_ecn_error) { 1654 if (vxlan_get_sk_family(vs) == AF_INET) 1655 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1656 &((struct iphdr *)oiph)->saddr, 1657 ((struct iphdr *)oiph)->tos); 1658 else 1659 net_info_ratelimited("non-ECT from %pI6\n", 1660 &((struct ipv6hdr *)oiph)->saddr); 1661 } 1662 return err <= 1; 1663 } 1664 1665 /* Callback from net/ipv4/udp.c to receive packets */ 1666 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1667 { 1668 struct vxlan_vni_node *vninode = NULL; 1669 struct vxlan_dev *vxlan; 1670 struct vxlan_sock *vs; 1671 struct vxlanhdr unparsed; 1672 struct vxlan_metadata _md; 1673 struct vxlan_metadata *md = &_md; 1674 __be16 protocol = htons(ETH_P_TEB); 1675 bool raw_proto = false; 1676 void *oiph; 1677 __be32 vni = 0; 1678 int nh; 1679 1680 /* Need UDP and VXLAN header to be present */ 1681 if (!pskb_may_pull(skb, VXLAN_HLEN)) 1682 goto drop; 1683 1684 unparsed = *vxlan_hdr(skb); 1685 /* VNI flag always required to be set */ 1686 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1687 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1688 ntohl(vxlan_hdr(skb)->vx_flags), 1689 ntohl(vxlan_hdr(skb)->vx_vni)); 1690 /* Return non vxlan pkt */ 1691 goto drop; 1692 } 1693 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1694 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1695 1696 vs = rcu_dereference_sk_user_data(sk); 1697 if (!vs) 1698 goto drop; 1699 1700 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1701 1702 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1703 if (!vxlan) 1704 goto drop; 1705 1706 /* For backwards compatibility, only allow reserved fields to be 1707 * used by VXLAN extensions if explicitly requested. 1708 */ 1709 if (vs->flags & VXLAN_F_GPE) { 1710 if (!vxlan_parse_gpe_proto(&unparsed, &protocol)) 1711 goto drop; 1712 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS; 1713 raw_proto = true; 1714 } 1715 1716 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1717 !net_eq(vxlan->net, dev_net(vxlan->dev)))) 1718 goto drop; 1719 1720 if (vs->flags & VXLAN_F_REMCSUM_RX) 1721 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1722 goto drop; 1723 1724 if (vxlan_collect_metadata(vs)) { 1725 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1726 struct metadata_dst *tun_dst; 1727 1728 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1729 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1730 key32_to_tunnel_id(vni), sizeof(*md)); 1731 1732 if (!tun_dst) 1733 goto drop; 1734 1735 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1736 1737 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1738 } else { 1739 memset(md, 0, sizeof(*md)); 1740 } 1741 1742 if (vs->flags & VXLAN_F_GBP) 1743 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1744 /* Note that GBP and GPE can never be active together. This is 1745 * ensured in vxlan_dev_configure. 1746 */ 1747 1748 if (unparsed.vx_flags || unparsed.vx_vni) { 1749 /* If there are any unprocessed flags remaining treat 1750 * this as a malformed packet. This behavior diverges from 1751 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1752 * in reserved fields are to be ignored. The approach here 1753 * maintains compatibility with previous stack code, and also 1754 * is more robust and provides a little more security in 1755 * adding extensions to VXLAN. 1756 */ 1757 goto drop; 1758 } 1759 1760 if (!raw_proto) { 1761 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1762 goto drop; 1763 } else { 1764 skb_reset_mac_header(skb); 1765 skb->dev = vxlan->dev; 1766 skb->pkt_type = PACKET_HOST; 1767 } 1768 1769 /* Save offset of outer header relative to skb->head, 1770 * because we are going to reset the network header to the inner header 1771 * and might change skb->head. 1772 */ 1773 nh = skb_network_header(skb) - skb->head; 1774 1775 skb_reset_network_header(skb); 1776 1777 if (!pskb_inet_may_pull(skb)) { 1778 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1779 DEV_STATS_INC(vxlan->dev, rx_errors); 1780 vxlan_vnifilter_count(vxlan, vni, vninode, 1781 VXLAN_VNI_STATS_RX_ERRORS, 0); 1782 goto drop; 1783 } 1784 1785 /* Get the outer header. */ 1786 oiph = skb->head + nh; 1787 1788 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1789 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1790 DEV_STATS_INC(vxlan->dev, rx_errors); 1791 vxlan_vnifilter_count(vxlan, vni, vninode, 1792 VXLAN_VNI_STATS_RX_ERRORS, 0); 1793 goto drop; 1794 } 1795 1796 rcu_read_lock(); 1797 1798 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1799 rcu_read_unlock(); 1800 dev_core_stats_rx_dropped_inc(vxlan->dev); 1801 vxlan_vnifilter_count(vxlan, vni, vninode, 1802 VXLAN_VNI_STATS_RX_DROPS, 0); 1803 goto drop; 1804 } 1805 1806 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1807 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1808 gro_cells_receive(&vxlan->gro_cells, skb); 1809 1810 rcu_read_unlock(); 1811 1812 return 0; 1813 1814 drop: 1815 /* Consume bad packet */ 1816 kfree_skb(skb); 1817 return 0; 1818 } 1819 1820 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1821 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1822 { 1823 struct vxlan_dev *vxlan; 1824 struct vxlan_sock *vs; 1825 struct vxlanhdr *hdr; 1826 __be32 vni; 1827 1828 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1829 return -EINVAL; 1830 1831 hdr = vxlan_hdr(skb); 1832 1833 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1834 return -EINVAL; 1835 1836 vs = rcu_dereference_sk_user_data(sk); 1837 if (!vs) 1838 return -ENOENT; 1839 1840 vni = vxlan_vni(hdr->vx_vni); 1841 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1842 if (!vxlan) 1843 return -ENOENT; 1844 1845 return 0; 1846 } 1847 1848 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1849 { 1850 struct vxlan_dev *vxlan = netdev_priv(dev); 1851 struct arphdr *parp; 1852 u8 *arpptr, *sha; 1853 __be32 sip, tip; 1854 struct neighbour *n; 1855 1856 if (dev->flags & IFF_NOARP) 1857 goto out; 1858 1859 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1860 dev_core_stats_tx_dropped_inc(dev); 1861 vxlan_vnifilter_count(vxlan, vni, NULL, 1862 VXLAN_VNI_STATS_TX_DROPS, 0); 1863 goto out; 1864 } 1865 parp = arp_hdr(skb); 1866 1867 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1868 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1869 parp->ar_pro != htons(ETH_P_IP) || 1870 parp->ar_op != htons(ARPOP_REQUEST) || 1871 parp->ar_hln != dev->addr_len || 1872 parp->ar_pln != 4) 1873 goto out; 1874 arpptr = (u8 *)parp + sizeof(struct arphdr); 1875 sha = arpptr; 1876 arpptr += dev->addr_len; /* sha */ 1877 memcpy(&sip, arpptr, sizeof(sip)); 1878 arpptr += sizeof(sip); 1879 arpptr += dev->addr_len; /* tha */ 1880 memcpy(&tip, arpptr, sizeof(tip)); 1881 1882 if (ipv4_is_loopback(tip) || 1883 ipv4_is_multicast(tip)) 1884 goto out; 1885 1886 n = neigh_lookup(&arp_tbl, &tip, dev); 1887 1888 if (n) { 1889 struct vxlan_fdb *f; 1890 struct sk_buff *reply; 1891 1892 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1893 neigh_release(n); 1894 goto out; 1895 } 1896 1897 f = vxlan_find_mac(vxlan, n->ha, vni); 1898 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1899 /* bridge-local neighbor */ 1900 neigh_release(n); 1901 goto out; 1902 } 1903 1904 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1905 n->ha, sha); 1906 1907 neigh_release(n); 1908 1909 if (reply == NULL) 1910 goto out; 1911 1912 skb_reset_mac_header(reply); 1913 __skb_pull(reply, skb_network_offset(reply)); 1914 reply->ip_summed = CHECKSUM_UNNECESSARY; 1915 reply->pkt_type = PACKET_HOST; 1916 1917 if (netif_rx(reply) == NET_RX_DROP) { 1918 dev_core_stats_rx_dropped_inc(dev); 1919 vxlan_vnifilter_count(vxlan, vni, NULL, 1920 VXLAN_VNI_STATS_RX_DROPS, 0); 1921 } 1922 1923 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1924 union vxlan_addr ipa = { 1925 .sin.sin_addr.s_addr = tip, 1926 .sin.sin_family = AF_INET, 1927 }; 1928 1929 vxlan_ip_miss(dev, &ipa); 1930 } 1931 out: 1932 consume_skb(skb); 1933 return NETDEV_TX_OK; 1934 } 1935 1936 #if IS_ENABLED(CONFIG_IPV6) 1937 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1938 struct neighbour *n, bool isrouter) 1939 { 1940 struct net_device *dev = request->dev; 1941 struct sk_buff *reply; 1942 struct nd_msg *ns, *na; 1943 struct ipv6hdr *pip6; 1944 u8 *daddr; 1945 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1946 int ns_olen; 1947 int i, len; 1948 1949 if (dev == NULL || !pskb_may_pull(request, request->len)) 1950 return NULL; 1951 1952 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1953 sizeof(*na) + na_olen + dev->needed_tailroom; 1954 reply = alloc_skb(len, GFP_ATOMIC); 1955 if (reply == NULL) 1956 return NULL; 1957 1958 reply->protocol = htons(ETH_P_IPV6); 1959 reply->dev = dev; 1960 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1961 skb_push(reply, sizeof(struct ethhdr)); 1962 skb_reset_mac_header(reply); 1963 1964 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1965 1966 daddr = eth_hdr(request)->h_source; 1967 ns_olen = request->len - skb_network_offset(request) - 1968 sizeof(struct ipv6hdr) - sizeof(*ns); 1969 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1970 if (!ns->opt[i + 1]) { 1971 kfree_skb(reply); 1972 return NULL; 1973 } 1974 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1975 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1976 break; 1977 } 1978 } 1979 1980 /* Ethernet header */ 1981 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1982 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1983 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1984 reply->protocol = htons(ETH_P_IPV6); 1985 1986 skb_pull(reply, sizeof(struct ethhdr)); 1987 skb_reset_network_header(reply); 1988 skb_put(reply, sizeof(struct ipv6hdr)); 1989 1990 /* IPv6 header */ 1991 1992 pip6 = ipv6_hdr(reply); 1993 memset(pip6, 0, sizeof(struct ipv6hdr)); 1994 pip6->version = 6; 1995 pip6->priority = ipv6_hdr(request)->priority; 1996 pip6->nexthdr = IPPROTO_ICMPV6; 1997 pip6->hop_limit = 255; 1998 pip6->daddr = ipv6_hdr(request)->saddr; 1999 pip6->saddr = *(struct in6_addr *)n->primary_key; 2000 2001 skb_pull(reply, sizeof(struct ipv6hdr)); 2002 skb_reset_transport_header(reply); 2003 2004 /* Neighbor Advertisement */ 2005 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2006 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2007 na->icmph.icmp6_router = isrouter; 2008 na->icmph.icmp6_override = 1; 2009 na->icmph.icmp6_solicited = 1; 2010 na->target = ns->target; 2011 ether_addr_copy(&na->opt[2], n->ha); 2012 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2013 na->opt[1] = na_olen >> 3; 2014 2015 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2016 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2017 csum_partial(na, sizeof(*na)+na_olen, 0)); 2018 2019 pip6->payload_len = htons(sizeof(*na)+na_olen); 2020 2021 skb_push(reply, sizeof(struct ipv6hdr)); 2022 2023 reply->ip_summed = CHECKSUM_UNNECESSARY; 2024 2025 return reply; 2026 } 2027 2028 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2029 { 2030 struct vxlan_dev *vxlan = netdev_priv(dev); 2031 const struct in6_addr *daddr; 2032 const struct ipv6hdr *iphdr; 2033 struct inet6_dev *in6_dev; 2034 struct neighbour *n; 2035 struct nd_msg *msg; 2036 2037 rcu_read_lock(); 2038 in6_dev = __in6_dev_get(dev); 2039 if (!in6_dev) 2040 goto out; 2041 2042 iphdr = ipv6_hdr(skb); 2043 daddr = &iphdr->daddr; 2044 msg = (struct nd_msg *)(iphdr + 1); 2045 2046 if (ipv6_addr_loopback(daddr) || 2047 ipv6_addr_is_multicast(&msg->target)) 2048 goto out; 2049 2050 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2051 2052 if (n) { 2053 struct vxlan_fdb *f; 2054 struct sk_buff *reply; 2055 2056 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2057 neigh_release(n); 2058 goto out; 2059 } 2060 2061 f = vxlan_find_mac(vxlan, n->ha, vni); 2062 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2063 /* bridge-local neighbor */ 2064 neigh_release(n); 2065 goto out; 2066 } 2067 2068 reply = vxlan_na_create(skb, n, 2069 !!(f ? f->flags & NTF_ROUTER : 0)); 2070 2071 neigh_release(n); 2072 2073 if (reply == NULL) 2074 goto out; 2075 2076 if (netif_rx(reply) == NET_RX_DROP) { 2077 dev_core_stats_rx_dropped_inc(dev); 2078 vxlan_vnifilter_count(vxlan, vni, NULL, 2079 VXLAN_VNI_STATS_RX_DROPS, 0); 2080 } 2081 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2082 union vxlan_addr ipa = { 2083 .sin6.sin6_addr = msg->target, 2084 .sin6.sin6_family = AF_INET6, 2085 }; 2086 2087 vxlan_ip_miss(dev, &ipa); 2088 } 2089 2090 out: 2091 rcu_read_unlock(); 2092 consume_skb(skb); 2093 return NETDEV_TX_OK; 2094 } 2095 #endif 2096 2097 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2098 { 2099 struct vxlan_dev *vxlan = netdev_priv(dev); 2100 struct neighbour *n; 2101 2102 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2103 return false; 2104 2105 n = NULL; 2106 switch (ntohs(eth_hdr(skb)->h_proto)) { 2107 case ETH_P_IP: 2108 { 2109 struct iphdr *pip; 2110 2111 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2112 return false; 2113 pip = ip_hdr(skb); 2114 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2115 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2116 union vxlan_addr ipa = { 2117 .sin.sin_addr.s_addr = pip->daddr, 2118 .sin.sin_family = AF_INET, 2119 }; 2120 2121 vxlan_ip_miss(dev, &ipa); 2122 return false; 2123 } 2124 2125 break; 2126 } 2127 #if IS_ENABLED(CONFIG_IPV6) 2128 case ETH_P_IPV6: 2129 { 2130 struct ipv6hdr *pip6; 2131 2132 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2133 return false; 2134 pip6 = ipv6_hdr(skb); 2135 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2136 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2137 union vxlan_addr ipa = { 2138 .sin6.sin6_addr = pip6->daddr, 2139 .sin6.sin6_family = AF_INET6, 2140 }; 2141 2142 vxlan_ip_miss(dev, &ipa); 2143 return false; 2144 } 2145 2146 break; 2147 } 2148 #endif 2149 default: 2150 return false; 2151 } 2152 2153 if (n) { 2154 bool diff; 2155 2156 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2157 if (diff) { 2158 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2159 dev->addr_len); 2160 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2161 } 2162 neigh_release(n); 2163 return diff; 2164 } 2165 2166 return false; 2167 } 2168 2169 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2170 { 2171 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2172 2173 gpe->np_applied = 1; 2174 gpe->next_protocol = tun_p_from_eth_p(protocol); 2175 if (!gpe->next_protocol) 2176 return -EPFNOSUPPORT; 2177 return 0; 2178 } 2179 2180 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2181 int iphdr_len, __be32 vni, 2182 struct vxlan_metadata *md, u32 vxflags, 2183 bool udp_sum) 2184 { 2185 struct vxlanhdr *vxh; 2186 int min_headroom; 2187 int err; 2188 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2189 __be16 inner_protocol = htons(ETH_P_TEB); 2190 2191 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2192 skb->ip_summed == CHECKSUM_PARTIAL) { 2193 int csum_start = skb_checksum_start_offset(skb); 2194 2195 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2196 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2197 (skb->csum_offset == offsetof(struct udphdr, check) || 2198 skb->csum_offset == offsetof(struct tcphdr, check))) 2199 type |= SKB_GSO_TUNNEL_REMCSUM; 2200 } 2201 2202 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2203 + VXLAN_HLEN + iphdr_len; 2204 2205 /* Need space for new headers (invalidates iph ptr) */ 2206 err = skb_cow_head(skb, min_headroom); 2207 if (unlikely(err)) 2208 return err; 2209 2210 err = iptunnel_handle_offloads(skb, type); 2211 if (err) 2212 return err; 2213 2214 vxh = __skb_push(skb, sizeof(*vxh)); 2215 vxh->vx_flags = VXLAN_HF_VNI; 2216 vxh->vx_vni = vxlan_vni_field(vni); 2217 2218 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2219 unsigned int start; 2220 2221 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2222 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2223 vxh->vx_flags |= VXLAN_HF_RCO; 2224 2225 if (!skb_is_gso(skb)) { 2226 skb->ip_summed = CHECKSUM_NONE; 2227 skb->encapsulation = 0; 2228 } 2229 } 2230 2231 if (vxflags & VXLAN_F_GBP) 2232 vxlan_build_gbp_hdr(vxh, md); 2233 if (vxflags & VXLAN_F_GPE) { 2234 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2235 if (err < 0) 2236 return err; 2237 inner_protocol = skb->protocol; 2238 } 2239 2240 skb_set_inner_protocol(skb, inner_protocol); 2241 return 0; 2242 } 2243 2244 /* Bypass encapsulation if the destination is local */ 2245 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2246 struct vxlan_dev *dst_vxlan, __be32 vni, 2247 bool snoop) 2248 { 2249 union vxlan_addr loopback; 2250 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2251 struct net_device *dev; 2252 int len = skb->len; 2253 2254 skb->pkt_type = PACKET_HOST; 2255 skb->encapsulation = 0; 2256 skb->dev = dst_vxlan->dev; 2257 __skb_pull(skb, skb_network_offset(skb)); 2258 2259 if (remote_ip->sa.sa_family == AF_INET) { 2260 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2261 loopback.sa.sa_family = AF_INET; 2262 #if IS_ENABLED(CONFIG_IPV6) 2263 } else { 2264 loopback.sin6.sin6_addr = in6addr_loopback; 2265 loopback.sa.sa_family = AF_INET6; 2266 #endif 2267 } 2268 2269 rcu_read_lock(); 2270 dev = skb->dev; 2271 if (unlikely(!(dev->flags & IFF_UP))) { 2272 kfree_skb(skb); 2273 goto drop; 2274 } 2275 2276 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2277 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2278 2279 dev_sw_netstats_tx_add(src_vxlan->dev, 1, len); 2280 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2281 2282 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2283 dev_sw_netstats_rx_add(dst_vxlan->dev, len); 2284 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2285 len); 2286 } else { 2287 drop: 2288 dev_core_stats_rx_dropped_inc(dev); 2289 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2290 VXLAN_VNI_STATS_RX_DROPS, 0); 2291 } 2292 rcu_read_unlock(); 2293 } 2294 2295 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2296 struct vxlan_dev *vxlan, 2297 int addr_family, 2298 __be16 dst_port, int dst_ifindex, __be32 vni, 2299 struct dst_entry *dst, 2300 u32 rt_flags) 2301 { 2302 #if IS_ENABLED(CONFIG_IPV6) 2303 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2304 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2305 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2306 */ 2307 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2308 #endif 2309 /* Bypass encapsulation if the destination is local */ 2310 if (rt_flags & RTCF_LOCAL && 2311 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2312 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2313 struct vxlan_dev *dst_vxlan; 2314 2315 dst_release(dst); 2316 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2317 addr_family, dst_port, 2318 vxlan->cfg.flags); 2319 if (!dst_vxlan) { 2320 DEV_STATS_INC(dev, tx_errors); 2321 vxlan_vnifilter_count(vxlan, vni, NULL, 2322 VXLAN_VNI_STATS_TX_ERRORS, 0); 2323 kfree_skb(skb); 2324 2325 return -ENOENT; 2326 } 2327 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2328 return 1; 2329 } 2330 2331 return 0; 2332 } 2333 2334 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2335 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2336 { 2337 struct dst_cache *dst_cache; 2338 struct ip_tunnel_info *info; 2339 struct ip_tunnel_key *pkey; 2340 struct ip_tunnel_key key; 2341 struct vxlan_dev *vxlan = netdev_priv(dev); 2342 const struct iphdr *old_iph = ip_hdr(skb); 2343 struct vxlan_metadata _md; 2344 struct vxlan_metadata *md = &_md; 2345 unsigned int pkt_len = skb->len; 2346 __be16 src_port = 0, dst_port; 2347 struct dst_entry *ndst = NULL; 2348 int addr_family; 2349 __u8 tos, ttl; 2350 int ifindex; 2351 int err; 2352 u32 flags = vxlan->cfg.flags; 2353 bool use_cache; 2354 bool udp_sum = false; 2355 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2356 __be32 vni = 0; 2357 2358 info = skb_tunnel_info(skb); 2359 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2360 2361 if (rdst) { 2362 memset(&key, 0, sizeof(key)); 2363 pkey = &key; 2364 2365 if (vxlan_addr_any(&rdst->remote_ip)) { 2366 if (did_rsc) { 2367 /* short-circuited back to local bridge */ 2368 vxlan_encap_bypass(skb, vxlan, vxlan, 2369 default_vni, true); 2370 return; 2371 } 2372 goto drop; 2373 } 2374 2375 addr_family = vxlan->cfg.saddr.sa.sa_family; 2376 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2377 vni = (rdst->remote_vni) ? : default_vni; 2378 ifindex = rdst->remote_ifindex; 2379 2380 if (addr_family == AF_INET) { 2381 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2382 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2383 } else { 2384 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2385 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2386 } 2387 2388 dst_cache = &rdst->dst_cache; 2389 md->gbp = skb->mark; 2390 if (flags & VXLAN_F_TTL_INHERIT) { 2391 ttl = ip_tunnel_get_ttl(old_iph, skb); 2392 } else { 2393 ttl = vxlan->cfg.ttl; 2394 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2395 ttl = 1; 2396 } 2397 tos = vxlan->cfg.tos; 2398 if (tos == 1) 2399 tos = ip_tunnel_get_dsfield(old_iph, skb); 2400 if (tos && !info) 2401 use_cache = false; 2402 2403 if (addr_family == AF_INET) 2404 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2405 else 2406 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2407 #if IS_ENABLED(CONFIG_IPV6) 2408 switch (vxlan->cfg.label_policy) { 2409 case VXLAN_LABEL_FIXED: 2410 key.label = vxlan->cfg.label; 2411 break; 2412 case VXLAN_LABEL_INHERIT: 2413 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2414 break; 2415 default: 2416 DEBUG_NET_WARN_ON_ONCE(1); 2417 goto drop; 2418 } 2419 #endif 2420 } else { 2421 if (!info) { 2422 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2423 dev->name); 2424 goto drop; 2425 } 2426 pkey = &info->key; 2427 addr_family = ip_tunnel_info_af(info); 2428 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2429 vni = tunnel_id_to_key32(info->key.tun_id); 2430 ifindex = 0; 2431 dst_cache = &info->dst_cache; 2432 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2433 if (info->options_len < sizeof(*md)) 2434 goto drop; 2435 md = ip_tunnel_info_opts(info); 2436 } 2437 ttl = info->key.ttl; 2438 tos = info->key.tos; 2439 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2440 } 2441 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2442 vxlan->cfg.port_max, true); 2443 2444 rcu_read_lock(); 2445 if (addr_family == AF_INET) { 2446 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2447 struct rtable *rt; 2448 __be16 df = 0; 2449 __be32 saddr; 2450 2451 if (!ifindex) 2452 ifindex = sock4->sock->sk->sk_bound_dev_if; 2453 2454 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2455 &saddr, pkey, src_port, dst_port, 2456 tos, use_cache ? dst_cache : NULL); 2457 if (IS_ERR(rt)) { 2458 err = PTR_ERR(rt); 2459 goto tx_error; 2460 } 2461 2462 if (!info) { 2463 /* Bypass encapsulation if the destination is local */ 2464 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2465 dst_port, ifindex, vni, 2466 &rt->dst, rt->rt_flags); 2467 if (err) 2468 goto out_unlock; 2469 2470 if (vxlan->cfg.df == VXLAN_DF_SET) { 2471 df = htons(IP_DF); 2472 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2473 struct ethhdr *eth = eth_hdr(skb); 2474 2475 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2476 (ntohs(eth->h_proto) == ETH_P_IP && 2477 old_iph->frag_off & htons(IP_DF))) 2478 df = htons(IP_DF); 2479 } 2480 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2481 info->key.tun_flags)) { 2482 df = htons(IP_DF); 2483 } 2484 2485 ndst = &rt->dst; 2486 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2487 netif_is_any_bridge_port(dev)); 2488 if (err < 0) { 2489 goto tx_error; 2490 } else if (err) { 2491 if (info) { 2492 struct ip_tunnel_info *unclone; 2493 2494 unclone = skb_tunnel_info_unclone(skb); 2495 if (unlikely(!unclone)) 2496 goto tx_error; 2497 2498 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2499 unclone->key.u.ipv4.dst = saddr; 2500 } 2501 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2502 dst_release(ndst); 2503 goto out_unlock; 2504 } 2505 2506 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2507 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2508 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2509 vni, md, flags, udp_sum); 2510 if (err < 0) 2511 goto tx_error; 2512 2513 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2514 pkey->u.ipv4.dst, tos, ttl, df, 2515 src_port, dst_port, xnet, !udp_sum); 2516 #if IS_ENABLED(CONFIG_IPV6) 2517 } else { 2518 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2519 struct in6_addr saddr; 2520 2521 if (!ifindex) 2522 ifindex = sock6->sock->sk->sk_bound_dev_if; 2523 2524 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2525 ifindex, &saddr, pkey, 2526 src_port, dst_port, tos, 2527 use_cache ? dst_cache : NULL); 2528 if (IS_ERR(ndst)) { 2529 err = PTR_ERR(ndst); 2530 ndst = NULL; 2531 goto tx_error; 2532 } 2533 2534 if (!info) { 2535 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2536 2537 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2538 dst_port, ifindex, vni, 2539 ndst, rt6i_flags); 2540 if (err) 2541 goto out_unlock; 2542 } 2543 2544 err = skb_tunnel_check_pmtu(skb, ndst, 2545 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2546 netif_is_any_bridge_port(dev)); 2547 if (err < 0) { 2548 goto tx_error; 2549 } else if (err) { 2550 if (info) { 2551 struct ip_tunnel_info *unclone; 2552 2553 unclone = skb_tunnel_info_unclone(skb); 2554 if (unlikely(!unclone)) 2555 goto tx_error; 2556 2557 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2558 unclone->key.u.ipv6.dst = saddr; 2559 } 2560 2561 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2562 dst_release(ndst); 2563 goto out_unlock; 2564 } 2565 2566 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2567 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2568 skb_scrub_packet(skb, xnet); 2569 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2570 vni, md, flags, udp_sum); 2571 if (err < 0) 2572 goto tx_error; 2573 2574 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2575 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2576 pkey->label, src_port, dst_port, !udp_sum); 2577 #endif 2578 } 2579 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2580 out_unlock: 2581 rcu_read_unlock(); 2582 return; 2583 2584 drop: 2585 dev_core_stats_tx_dropped_inc(dev); 2586 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2587 dev_kfree_skb(skb); 2588 return; 2589 2590 tx_error: 2591 rcu_read_unlock(); 2592 if (err == -ELOOP) 2593 DEV_STATS_INC(dev, collisions); 2594 else if (err == -ENETUNREACH) 2595 DEV_STATS_INC(dev, tx_carrier_errors); 2596 dst_release(ndst); 2597 DEV_STATS_INC(dev, tx_errors); 2598 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2599 kfree_skb(skb); 2600 } 2601 2602 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2603 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2604 { 2605 struct vxlan_rdst nh_rdst; 2606 struct nexthop *nh; 2607 bool do_xmit; 2608 u32 hash; 2609 2610 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2611 hash = skb_get_hash(skb); 2612 2613 rcu_read_lock(); 2614 nh = rcu_dereference(f->nh); 2615 if (!nh) { 2616 rcu_read_unlock(); 2617 goto drop; 2618 } 2619 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2620 rcu_read_unlock(); 2621 2622 if (likely(do_xmit)) 2623 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2624 else 2625 goto drop; 2626 2627 return; 2628 2629 drop: 2630 dev_core_stats_tx_dropped_inc(dev); 2631 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2632 VXLAN_VNI_STATS_TX_DROPS, 0); 2633 dev_kfree_skb(skb); 2634 } 2635 2636 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2637 u32 nhid, __be32 vni) 2638 { 2639 struct vxlan_dev *vxlan = netdev_priv(dev); 2640 struct vxlan_rdst nh_rdst; 2641 struct nexthop *nh; 2642 bool do_xmit; 2643 u32 hash; 2644 2645 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2646 hash = skb_get_hash(skb); 2647 2648 rcu_read_lock(); 2649 nh = nexthop_find_by_id(dev_net(dev), nhid); 2650 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2651 rcu_read_unlock(); 2652 goto drop; 2653 } 2654 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2655 rcu_read_unlock(); 2656 2657 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2658 goto drop; 2659 2660 if (likely(do_xmit)) 2661 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2662 else 2663 goto drop; 2664 2665 return NETDEV_TX_OK; 2666 2667 drop: 2668 dev_core_stats_tx_dropped_inc(dev); 2669 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2670 VXLAN_VNI_STATS_TX_DROPS, 0); 2671 dev_kfree_skb(skb); 2672 return NETDEV_TX_OK; 2673 } 2674 2675 /* Transmit local packets over Vxlan 2676 * 2677 * Outer IP header inherits ECN and DF from inner header. 2678 * Outer UDP destination is the VXLAN assigned port. 2679 * source port is based on hash of flow 2680 */ 2681 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2682 { 2683 struct vxlan_dev *vxlan = netdev_priv(dev); 2684 struct vxlan_rdst *rdst, *fdst = NULL; 2685 const struct ip_tunnel_info *info; 2686 bool did_rsc = false; 2687 struct vxlan_fdb *f; 2688 struct ethhdr *eth; 2689 __be32 vni = 0; 2690 u32 nhid = 0; 2691 2692 info = skb_tunnel_info(skb); 2693 2694 skb_reset_mac_header(skb); 2695 2696 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2697 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2698 info->mode & IP_TUNNEL_INFO_TX) { 2699 vni = tunnel_id_to_key32(info->key.tun_id); 2700 nhid = info->key.nhid; 2701 } else { 2702 if (info && info->mode & IP_TUNNEL_INFO_TX) 2703 vxlan_xmit_one(skb, dev, vni, NULL, false); 2704 else 2705 kfree_skb(skb); 2706 return NETDEV_TX_OK; 2707 } 2708 } 2709 2710 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2711 eth = eth_hdr(skb); 2712 if (ntohs(eth->h_proto) == ETH_P_ARP) 2713 return arp_reduce(dev, skb, vni); 2714 #if IS_ENABLED(CONFIG_IPV6) 2715 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2716 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2717 sizeof(struct nd_msg)) && 2718 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2719 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2720 2721 if (m->icmph.icmp6_code == 0 && 2722 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2723 return neigh_reduce(dev, skb, vni); 2724 } 2725 #endif 2726 } 2727 2728 if (nhid) 2729 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2730 2731 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2732 struct vxlan_mdb_entry *mdb_entry; 2733 2734 rcu_read_lock(); 2735 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2736 if (mdb_entry) { 2737 netdev_tx_t ret; 2738 2739 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2740 rcu_read_unlock(); 2741 return ret; 2742 } 2743 rcu_read_unlock(); 2744 } 2745 2746 eth = eth_hdr(skb); 2747 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2748 did_rsc = false; 2749 2750 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2751 (ntohs(eth->h_proto) == ETH_P_IP || 2752 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2753 did_rsc = route_shortcircuit(dev, skb); 2754 if (did_rsc) 2755 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2756 } 2757 2758 if (f == NULL) { 2759 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2760 if (f == NULL) { 2761 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2762 !is_multicast_ether_addr(eth->h_dest)) 2763 vxlan_fdb_miss(vxlan, eth->h_dest); 2764 2765 dev_core_stats_tx_dropped_inc(dev); 2766 vxlan_vnifilter_count(vxlan, vni, NULL, 2767 VXLAN_VNI_STATS_TX_DROPS, 0); 2768 kfree_skb(skb); 2769 return NETDEV_TX_OK; 2770 } 2771 } 2772 2773 if (rcu_access_pointer(f->nh)) { 2774 vxlan_xmit_nh(skb, dev, f, 2775 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2776 } else { 2777 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2778 struct sk_buff *skb1; 2779 2780 if (!fdst) { 2781 fdst = rdst; 2782 continue; 2783 } 2784 skb1 = skb_clone(skb, GFP_ATOMIC); 2785 if (skb1) 2786 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2787 } 2788 if (fdst) 2789 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2790 else 2791 kfree_skb(skb); 2792 } 2793 2794 return NETDEV_TX_OK; 2795 } 2796 2797 /* Walk the forwarding table and purge stale entries */ 2798 static void vxlan_cleanup(struct timer_list *t) 2799 { 2800 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2801 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2802 unsigned int h; 2803 2804 if (!netif_running(vxlan->dev)) 2805 return; 2806 2807 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2808 struct hlist_node *p, *n; 2809 2810 spin_lock(&vxlan->hash_lock[h]); 2811 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2812 struct vxlan_fdb *f 2813 = container_of(p, struct vxlan_fdb, hlist); 2814 unsigned long timeout; 2815 2816 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2817 continue; 2818 2819 if (f->flags & NTF_EXT_LEARNED) 2820 continue; 2821 2822 timeout = f->used + vxlan->cfg.age_interval * HZ; 2823 if (time_before_eq(timeout, jiffies)) { 2824 netdev_dbg(vxlan->dev, 2825 "garbage collect %pM\n", 2826 f->eth_addr); 2827 f->state = NUD_STALE; 2828 vxlan_fdb_destroy(vxlan, f, true, true); 2829 } else if (time_before(timeout, next_timer)) 2830 next_timer = timeout; 2831 } 2832 spin_unlock(&vxlan->hash_lock[h]); 2833 } 2834 2835 mod_timer(&vxlan->age_timer, next_timer); 2836 } 2837 2838 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2839 { 2840 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2841 2842 spin_lock(&vn->sock_lock); 2843 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2844 #if IS_ENABLED(CONFIG_IPV6) 2845 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2846 #endif 2847 spin_unlock(&vn->sock_lock); 2848 } 2849 2850 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2851 struct vxlan_dev_node *node) 2852 { 2853 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2854 __be32 vni = vxlan->default_dst.remote_vni; 2855 2856 node->vxlan = vxlan; 2857 spin_lock(&vn->sock_lock); 2858 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2859 spin_unlock(&vn->sock_lock); 2860 } 2861 2862 /* Setup stats when device is created */ 2863 static int vxlan_init(struct net_device *dev) 2864 { 2865 struct vxlan_dev *vxlan = netdev_priv(dev); 2866 int err; 2867 2868 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2869 vxlan_vnigroup_init(vxlan); 2870 2871 err = gro_cells_init(&vxlan->gro_cells, dev); 2872 if (err) 2873 goto err_vnigroup_uninit; 2874 2875 err = vxlan_mdb_init(vxlan); 2876 if (err) 2877 goto err_gro_cells_destroy; 2878 2879 netdev_lockdep_set_classes(dev); 2880 return 0; 2881 2882 err_gro_cells_destroy: 2883 gro_cells_destroy(&vxlan->gro_cells); 2884 err_vnigroup_uninit: 2885 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2886 vxlan_vnigroup_uninit(vxlan); 2887 return err; 2888 } 2889 2890 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2891 { 2892 struct vxlan_fdb *f; 2893 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2894 2895 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2896 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2897 if (f) 2898 vxlan_fdb_destroy(vxlan, f, true, true); 2899 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2900 } 2901 2902 static void vxlan_uninit(struct net_device *dev) 2903 { 2904 struct vxlan_dev *vxlan = netdev_priv(dev); 2905 2906 vxlan_mdb_fini(vxlan); 2907 2908 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2909 vxlan_vnigroup_uninit(vxlan); 2910 2911 gro_cells_destroy(&vxlan->gro_cells); 2912 2913 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2914 } 2915 2916 /* Start ageing timer and join group when device is brought up */ 2917 static int vxlan_open(struct net_device *dev) 2918 { 2919 struct vxlan_dev *vxlan = netdev_priv(dev); 2920 int ret; 2921 2922 ret = vxlan_sock_add(vxlan); 2923 if (ret < 0) 2924 return ret; 2925 2926 ret = vxlan_multicast_join(vxlan); 2927 if (ret) { 2928 vxlan_sock_release(vxlan); 2929 return ret; 2930 } 2931 2932 if (vxlan->cfg.age_interval) 2933 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2934 2935 return ret; 2936 } 2937 2938 struct vxlan_fdb_flush_desc { 2939 bool ignore_default_entry; 2940 unsigned long state; 2941 unsigned long state_mask; 2942 unsigned long flags; 2943 unsigned long flags_mask; 2944 __be32 src_vni; 2945 u32 nhid; 2946 __be32 vni; 2947 __be16 port; 2948 union vxlan_addr dst_ip; 2949 }; 2950 2951 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2952 const struct vxlan_dev *vxlan) 2953 { 2954 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2955 } 2956 2957 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2958 { 2959 struct nexthop *nh = rtnl_dereference(f->nh); 2960 2961 return nh && nh->id == nhid; 2962 } 2963 2964 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2965 const struct vxlan_dev *vxlan, 2966 const struct vxlan_fdb_flush_desc *desc) 2967 { 2968 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2969 return false; 2970 2971 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2972 return false; 2973 2974 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2975 return false; 2976 2977 if (desc->src_vni && f->vni != desc->src_vni) 2978 return false; 2979 2980 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 2981 return false; 2982 2983 return true; 2984 } 2985 2986 static bool 2987 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 2988 { 2989 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 2990 } 2991 2992 static bool 2993 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 2994 const struct vxlan_rdst *rd) 2995 { 2996 if (desc->vni && rd->remote_vni != desc->vni) 2997 return false; 2998 2999 if (desc->port && rd->remote_port != desc->port) 3000 return false; 3001 3002 if (desc->dst_ip.sa.sa_family && 3003 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3004 return false; 3005 3006 return true; 3007 } 3008 3009 static void 3010 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3011 const struct vxlan_fdb_flush_desc *desc, 3012 bool *p_destroy_fdb) 3013 { 3014 bool remotes_flushed = false; 3015 struct vxlan_rdst *rd, *tmp; 3016 3017 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3018 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3019 continue; 3020 3021 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3022 remotes_flushed = true; 3023 } 3024 3025 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3026 } 3027 3028 /* Purge the forwarding table */ 3029 static void vxlan_flush(struct vxlan_dev *vxlan, 3030 const struct vxlan_fdb_flush_desc *desc) 3031 { 3032 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3033 unsigned int h; 3034 3035 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3036 struct hlist_node *p, *n; 3037 3038 spin_lock_bh(&vxlan->hash_lock[h]); 3039 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3040 struct vxlan_fdb *f 3041 = container_of(p, struct vxlan_fdb, hlist); 3042 3043 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3044 continue; 3045 3046 if (match_remotes) { 3047 bool destroy_fdb = false; 3048 3049 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3050 &destroy_fdb); 3051 3052 if (!destroy_fdb) 3053 continue; 3054 } 3055 3056 vxlan_fdb_destroy(vxlan, f, true, true); 3057 } 3058 spin_unlock_bh(&vxlan->hash_lock[h]); 3059 } 3060 } 3061 3062 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3063 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3064 [NDA_NH_ID] = { .type = NLA_U32 }, 3065 [NDA_VNI] = { .type = NLA_U32 }, 3066 [NDA_PORT] = { .type = NLA_U16 }, 3067 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3068 sizeof(struct in6_addr)), 3069 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3070 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3071 }; 3072 3073 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3074 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3075 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3076 NTF_ROUTER) 3077 3078 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3079 struct netlink_ext_ack *extack) 3080 { 3081 struct vxlan_dev *vxlan = netdev_priv(dev); 3082 struct vxlan_fdb_flush_desc desc = {}; 3083 struct ndmsg *ndm = nlmsg_data(nlh); 3084 struct nlattr *tb[NDA_MAX + 1]; 3085 u8 ndm_flags; 3086 int err; 3087 3088 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3089 3090 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3091 extack); 3092 if (err) 3093 return err; 3094 3095 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3096 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3097 return -EINVAL; 3098 } 3099 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3100 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3101 return -EINVAL; 3102 } 3103 3104 desc.state = ndm->ndm_state; 3105 desc.flags = ndm_flags; 3106 3107 if (tb[NDA_NDM_STATE_MASK]) 3108 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3109 3110 if (tb[NDA_NDM_FLAGS_MASK]) 3111 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3112 3113 if (tb[NDA_SRC_VNI]) 3114 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3115 3116 if (tb[NDA_NH_ID]) 3117 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3118 3119 if (tb[NDA_VNI]) 3120 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3121 3122 if (tb[NDA_PORT]) 3123 desc.port = nla_get_be16(tb[NDA_PORT]); 3124 3125 if (tb[NDA_DST]) { 3126 union vxlan_addr ip; 3127 3128 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3129 if (err) { 3130 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3131 "Unsupported address family"); 3132 return err; 3133 } 3134 desc.dst_ip = ip; 3135 } 3136 3137 vxlan_flush(vxlan, &desc); 3138 3139 return 0; 3140 } 3141 3142 /* Cleanup timer and forwarding table on shutdown */ 3143 static int vxlan_stop(struct net_device *dev) 3144 { 3145 struct vxlan_dev *vxlan = netdev_priv(dev); 3146 struct vxlan_fdb_flush_desc desc = { 3147 /* Default entry is deleted at vxlan_uninit. */ 3148 .ignore_default_entry = true, 3149 .state = 0, 3150 .state_mask = NUD_PERMANENT | NUD_NOARP, 3151 }; 3152 3153 vxlan_multicast_leave(vxlan); 3154 3155 del_timer_sync(&vxlan->age_timer); 3156 3157 vxlan_flush(vxlan, &desc); 3158 vxlan_sock_release(vxlan); 3159 3160 return 0; 3161 } 3162 3163 /* Stub, nothing needs to be done. */ 3164 static void vxlan_set_multicast_list(struct net_device *dev) 3165 { 3166 } 3167 3168 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3169 { 3170 struct vxlan_dev *vxlan = netdev_priv(dev); 3171 struct vxlan_rdst *dst = &vxlan->default_dst; 3172 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3173 dst->remote_ifindex); 3174 3175 /* This check is different than dev->max_mtu, because it looks at 3176 * the lowerdev->mtu, rather than the static dev->max_mtu 3177 */ 3178 if (lowerdev) { 3179 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3180 if (new_mtu > max_mtu) 3181 return -EINVAL; 3182 } 3183 3184 WRITE_ONCE(dev->mtu, new_mtu); 3185 return 0; 3186 } 3187 3188 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3189 { 3190 struct vxlan_dev *vxlan = netdev_priv(dev); 3191 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3192 __be16 sport, dport; 3193 3194 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3195 vxlan->cfg.port_max, true); 3196 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3197 3198 if (ip_tunnel_info_af(info) == AF_INET) { 3199 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3200 struct rtable *rt; 3201 3202 if (!sock4) 3203 return -EIO; 3204 3205 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3206 &info->key.u.ipv4.src, 3207 &info->key, 3208 sport, dport, info->key.tos, 3209 &info->dst_cache); 3210 if (IS_ERR(rt)) 3211 return PTR_ERR(rt); 3212 ip_rt_put(rt); 3213 } else { 3214 #if IS_ENABLED(CONFIG_IPV6) 3215 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3216 struct dst_entry *ndst; 3217 3218 if (!sock6) 3219 return -EIO; 3220 3221 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3222 0, &info->key.u.ipv6.src, 3223 &info->key, 3224 sport, dport, info->key.tos, 3225 &info->dst_cache); 3226 if (IS_ERR(ndst)) 3227 return PTR_ERR(ndst); 3228 dst_release(ndst); 3229 #else /* !CONFIG_IPV6 */ 3230 return -EPFNOSUPPORT; 3231 #endif 3232 } 3233 info->key.tp_src = sport; 3234 info->key.tp_dst = dport; 3235 return 0; 3236 } 3237 3238 static const struct net_device_ops vxlan_netdev_ether_ops = { 3239 .ndo_init = vxlan_init, 3240 .ndo_uninit = vxlan_uninit, 3241 .ndo_open = vxlan_open, 3242 .ndo_stop = vxlan_stop, 3243 .ndo_start_xmit = vxlan_xmit, 3244 .ndo_set_rx_mode = vxlan_set_multicast_list, 3245 .ndo_change_mtu = vxlan_change_mtu, 3246 .ndo_validate_addr = eth_validate_addr, 3247 .ndo_set_mac_address = eth_mac_addr, 3248 .ndo_fdb_add = vxlan_fdb_add, 3249 .ndo_fdb_del = vxlan_fdb_delete, 3250 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3251 .ndo_fdb_dump = vxlan_fdb_dump, 3252 .ndo_fdb_get = vxlan_fdb_get, 3253 .ndo_mdb_add = vxlan_mdb_add, 3254 .ndo_mdb_del = vxlan_mdb_del, 3255 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3256 .ndo_mdb_dump = vxlan_mdb_dump, 3257 .ndo_mdb_get = vxlan_mdb_get, 3258 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3259 }; 3260 3261 static const struct net_device_ops vxlan_netdev_raw_ops = { 3262 .ndo_init = vxlan_init, 3263 .ndo_uninit = vxlan_uninit, 3264 .ndo_open = vxlan_open, 3265 .ndo_stop = vxlan_stop, 3266 .ndo_start_xmit = vxlan_xmit, 3267 .ndo_change_mtu = vxlan_change_mtu, 3268 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3269 }; 3270 3271 /* Info for udev, that this is a virtual tunnel endpoint */ 3272 static const struct device_type vxlan_type = { 3273 .name = "vxlan", 3274 }; 3275 3276 /* Calls the ndo_udp_tunnel_add of the caller in order to 3277 * supply the listening VXLAN udp ports. Callers are expected 3278 * to implement the ndo_udp_tunnel_add. 3279 */ 3280 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3281 { 3282 struct vxlan_sock *vs; 3283 struct net *net = dev_net(dev); 3284 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3285 unsigned int i; 3286 3287 spin_lock(&vn->sock_lock); 3288 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3289 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3290 unsigned short type; 3291 3292 if (vs->flags & VXLAN_F_GPE) 3293 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3294 else 3295 type = UDP_TUNNEL_TYPE_VXLAN; 3296 3297 if (push) 3298 udp_tunnel_push_rx_port(dev, vs->sock, type); 3299 else 3300 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3301 } 3302 } 3303 spin_unlock(&vn->sock_lock); 3304 } 3305 3306 /* Initialize the device structure. */ 3307 static void vxlan_setup(struct net_device *dev) 3308 { 3309 struct vxlan_dev *vxlan = netdev_priv(dev); 3310 unsigned int h; 3311 3312 eth_hw_addr_random(dev); 3313 ether_setup(dev); 3314 3315 dev->needs_free_netdev = true; 3316 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3317 3318 dev->features |= NETIF_F_LLTX; 3319 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3320 dev->features |= NETIF_F_RXCSUM; 3321 dev->features |= NETIF_F_GSO_SOFTWARE; 3322 3323 dev->vlan_features = dev->features; 3324 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3325 dev->hw_features |= NETIF_F_RXCSUM; 3326 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3327 netif_keep_dst(dev); 3328 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN; 3329 3330 /* MTU range: 68 - 65535 */ 3331 dev->min_mtu = ETH_MIN_MTU; 3332 dev->max_mtu = ETH_MAX_MTU; 3333 3334 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 3335 INIT_LIST_HEAD(&vxlan->next); 3336 3337 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3338 3339 vxlan->dev = dev; 3340 3341 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3342 spin_lock_init(&vxlan->hash_lock[h]); 3343 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3344 } 3345 } 3346 3347 static void vxlan_ether_setup(struct net_device *dev) 3348 { 3349 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3350 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3351 dev->netdev_ops = &vxlan_netdev_ether_ops; 3352 } 3353 3354 static void vxlan_raw_setup(struct net_device *dev) 3355 { 3356 dev->header_ops = NULL; 3357 dev->type = ARPHRD_NONE; 3358 dev->hard_header_len = 0; 3359 dev->addr_len = 0; 3360 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3361 dev->netdev_ops = &vxlan_netdev_raw_ops; 3362 } 3363 3364 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3365 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3366 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3367 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3368 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3369 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3370 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3371 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3372 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3373 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3374 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3375 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3376 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3377 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3378 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3379 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3380 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3381 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3382 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3383 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3384 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3385 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3386 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3387 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3388 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3389 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3390 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3391 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3392 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3393 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3394 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3395 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3396 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3397 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3398 }; 3399 3400 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3401 struct netlink_ext_ack *extack) 3402 { 3403 if (tb[IFLA_ADDRESS]) { 3404 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3405 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3406 "Provided link layer address is not Ethernet"); 3407 return -EINVAL; 3408 } 3409 3410 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3411 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3412 "Provided Ethernet address is not unicast"); 3413 return -EADDRNOTAVAIL; 3414 } 3415 } 3416 3417 if (tb[IFLA_MTU]) { 3418 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3419 3420 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3421 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3422 "MTU must be between 68 and 65535"); 3423 return -EINVAL; 3424 } 3425 } 3426 3427 if (!data) { 3428 NL_SET_ERR_MSG(extack, 3429 "Required attributes not provided to perform the operation"); 3430 return -EINVAL; 3431 } 3432 3433 if (data[IFLA_VXLAN_ID]) { 3434 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3435 3436 if (id >= VXLAN_N_VID) { 3437 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3438 "VXLAN ID must be lower than 16777216"); 3439 return -ERANGE; 3440 } 3441 } 3442 3443 if (data[IFLA_VXLAN_PORT_RANGE]) { 3444 const struct ifla_vxlan_port_range *p 3445 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3446 3447 if (ntohs(p->high) < ntohs(p->low)) { 3448 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3449 "Invalid source port range"); 3450 return -EINVAL; 3451 } 3452 } 3453 3454 if (data[IFLA_VXLAN_DF]) { 3455 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3456 3457 if (df < 0 || df > VXLAN_DF_MAX) { 3458 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3459 "Invalid DF attribute"); 3460 return -EINVAL; 3461 } 3462 } 3463 3464 return 0; 3465 } 3466 3467 static void vxlan_get_drvinfo(struct net_device *netdev, 3468 struct ethtool_drvinfo *drvinfo) 3469 { 3470 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3471 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3472 } 3473 3474 static int vxlan_get_link_ksettings(struct net_device *dev, 3475 struct ethtool_link_ksettings *cmd) 3476 { 3477 struct vxlan_dev *vxlan = netdev_priv(dev); 3478 struct vxlan_rdst *dst = &vxlan->default_dst; 3479 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3480 dst->remote_ifindex); 3481 3482 if (!lowerdev) { 3483 cmd->base.duplex = DUPLEX_UNKNOWN; 3484 cmd->base.port = PORT_OTHER; 3485 cmd->base.speed = SPEED_UNKNOWN; 3486 3487 return 0; 3488 } 3489 3490 return __ethtool_get_link_ksettings(lowerdev, cmd); 3491 } 3492 3493 static const struct ethtool_ops vxlan_ethtool_ops = { 3494 .get_drvinfo = vxlan_get_drvinfo, 3495 .get_link = ethtool_op_get_link, 3496 .get_link_ksettings = vxlan_get_link_ksettings, 3497 }; 3498 3499 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3500 __be16 port, u32 flags, int ifindex) 3501 { 3502 struct socket *sock; 3503 struct udp_port_cfg udp_conf; 3504 int err; 3505 3506 memset(&udp_conf, 0, sizeof(udp_conf)); 3507 3508 if (ipv6) { 3509 udp_conf.family = AF_INET6; 3510 udp_conf.use_udp6_rx_checksums = 3511 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3512 udp_conf.ipv6_v6only = 1; 3513 } else { 3514 udp_conf.family = AF_INET; 3515 } 3516 3517 udp_conf.local_udp_port = port; 3518 udp_conf.bind_ifindex = ifindex; 3519 3520 /* Open UDP socket */ 3521 err = udp_sock_create(net, &udp_conf, &sock); 3522 if (err < 0) 3523 return ERR_PTR(err); 3524 3525 udp_allow_gso(sock->sk); 3526 return sock; 3527 } 3528 3529 /* Create new listen socket if needed */ 3530 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3531 __be16 port, u32 flags, 3532 int ifindex) 3533 { 3534 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3535 struct vxlan_sock *vs; 3536 struct socket *sock; 3537 unsigned int h; 3538 struct udp_tunnel_sock_cfg tunnel_cfg; 3539 3540 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3541 if (!vs) 3542 return ERR_PTR(-ENOMEM); 3543 3544 for (h = 0; h < VNI_HASH_SIZE; ++h) 3545 INIT_HLIST_HEAD(&vs->vni_list[h]); 3546 3547 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3548 if (IS_ERR(sock)) { 3549 kfree(vs); 3550 return ERR_CAST(sock); 3551 } 3552 3553 vs->sock = sock; 3554 refcount_set(&vs->refcnt, 1); 3555 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3556 3557 spin_lock(&vn->sock_lock); 3558 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3559 udp_tunnel_notify_add_rx_port(sock, 3560 (vs->flags & VXLAN_F_GPE) ? 3561 UDP_TUNNEL_TYPE_VXLAN_GPE : 3562 UDP_TUNNEL_TYPE_VXLAN); 3563 spin_unlock(&vn->sock_lock); 3564 3565 /* Mark socket as an encapsulation socket. */ 3566 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3567 tunnel_cfg.sk_user_data = vs; 3568 tunnel_cfg.encap_type = 1; 3569 tunnel_cfg.encap_rcv = vxlan_rcv; 3570 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3571 tunnel_cfg.encap_destroy = NULL; 3572 if (vs->flags & VXLAN_F_GPE) { 3573 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3574 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3575 } else { 3576 tunnel_cfg.gro_receive = vxlan_gro_receive; 3577 tunnel_cfg.gro_complete = vxlan_gro_complete; 3578 } 3579 3580 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3581 3582 return vs; 3583 } 3584 3585 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3586 { 3587 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3588 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3589 struct vxlan_sock *vs = NULL; 3590 struct vxlan_dev_node *node; 3591 int l3mdev_index = 0; 3592 3593 if (vxlan->cfg.remote_ifindex) 3594 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3595 vxlan->net, vxlan->cfg.remote_ifindex); 3596 3597 if (!vxlan->cfg.no_share) { 3598 spin_lock(&vn->sock_lock); 3599 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3600 vxlan->cfg.dst_port, vxlan->cfg.flags, 3601 l3mdev_index); 3602 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3603 spin_unlock(&vn->sock_lock); 3604 return -EBUSY; 3605 } 3606 spin_unlock(&vn->sock_lock); 3607 } 3608 if (!vs) 3609 vs = vxlan_socket_create(vxlan->net, ipv6, 3610 vxlan->cfg.dst_port, vxlan->cfg.flags, 3611 l3mdev_index); 3612 if (IS_ERR(vs)) 3613 return PTR_ERR(vs); 3614 #if IS_ENABLED(CONFIG_IPV6) 3615 if (ipv6) { 3616 rcu_assign_pointer(vxlan->vn6_sock, vs); 3617 node = &vxlan->hlist6; 3618 } else 3619 #endif 3620 { 3621 rcu_assign_pointer(vxlan->vn4_sock, vs); 3622 node = &vxlan->hlist4; 3623 } 3624 3625 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3626 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3627 else 3628 vxlan_vs_add_dev(vs, vxlan, node); 3629 3630 return 0; 3631 } 3632 3633 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3634 { 3635 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3636 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3637 bool ipv4 = !ipv6 || metadata; 3638 int ret = 0; 3639 3640 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3641 #if IS_ENABLED(CONFIG_IPV6) 3642 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3643 if (ipv6) { 3644 ret = __vxlan_sock_add(vxlan, true); 3645 if (ret < 0 && ret != -EAFNOSUPPORT) 3646 ipv4 = false; 3647 } 3648 #endif 3649 if (ipv4) 3650 ret = __vxlan_sock_add(vxlan, false); 3651 if (ret < 0) 3652 vxlan_sock_release(vxlan); 3653 return ret; 3654 } 3655 3656 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3657 struct vxlan_config *conf, __be32 vni) 3658 { 3659 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3660 struct vxlan_dev *tmp; 3661 3662 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3663 if (tmp == vxlan) 3664 continue; 3665 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3666 if (!vxlan_vnifilter_lookup(tmp, vni)) 3667 continue; 3668 } else if (tmp->cfg.vni != vni) { 3669 continue; 3670 } 3671 if (tmp->cfg.dst_port != conf->dst_port) 3672 continue; 3673 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3674 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3675 continue; 3676 3677 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3678 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3679 continue; 3680 3681 return -EEXIST; 3682 } 3683 3684 return 0; 3685 } 3686 3687 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3688 struct net_device **lower, 3689 struct vxlan_dev *old, 3690 struct netlink_ext_ack *extack) 3691 { 3692 bool use_ipv6 = false; 3693 3694 if (conf->flags & VXLAN_F_GPE) { 3695 /* For now, allow GPE only together with 3696 * COLLECT_METADATA. This can be relaxed later; in such 3697 * case, the other side of the PtP link will have to be 3698 * provided. 3699 */ 3700 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3701 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3702 NL_SET_ERR_MSG(extack, 3703 "VXLAN GPE does not support this combination of attributes"); 3704 return -EINVAL; 3705 } 3706 } 3707 3708 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3709 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3710 conf->remote_ip.sa.sa_family = AF_INET; 3711 conf->saddr.sa.sa_family = AF_INET; 3712 } else if (!conf->remote_ip.sa.sa_family) { 3713 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3714 } else if (!conf->saddr.sa.sa_family) { 3715 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3716 } 3717 3718 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3719 NL_SET_ERR_MSG(extack, 3720 "Local and remote address must be from the same family"); 3721 return -EINVAL; 3722 } 3723 3724 if (vxlan_addr_multicast(&conf->saddr)) { 3725 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3726 return -EINVAL; 3727 } 3728 3729 if (conf->saddr.sa.sa_family == AF_INET6) { 3730 if (!IS_ENABLED(CONFIG_IPV6)) { 3731 NL_SET_ERR_MSG(extack, 3732 "IPv6 support not enabled in the kernel"); 3733 return -EPFNOSUPPORT; 3734 } 3735 use_ipv6 = true; 3736 conf->flags |= VXLAN_F_IPV6; 3737 3738 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3739 int local_type = 3740 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3741 int remote_type = 3742 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3743 3744 if (local_type & IPV6_ADDR_LINKLOCAL) { 3745 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3746 (remote_type != IPV6_ADDR_ANY)) { 3747 NL_SET_ERR_MSG(extack, 3748 "Invalid combination of local and remote address scopes"); 3749 return -EINVAL; 3750 } 3751 3752 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3753 } else { 3754 if (remote_type == 3755 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3756 NL_SET_ERR_MSG(extack, 3757 "Invalid combination of local and remote address scopes"); 3758 return -EINVAL; 3759 } 3760 3761 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3762 } 3763 } 3764 } 3765 3766 if (conf->label && !use_ipv6) { 3767 NL_SET_ERR_MSG(extack, 3768 "Label attribute only applies to IPv6 VXLAN devices"); 3769 return -EINVAL; 3770 } 3771 3772 if (conf->label_policy && !use_ipv6) { 3773 NL_SET_ERR_MSG(extack, 3774 "Label policy only applies to IPv6 VXLAN devices"); 3775 return -EINVAL; 3776 } 3777 3778 if (conf->remote_ifindex) { 3779 struct net_device *lowerdev; 3780 3781 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3782 if (!lowerdev) { 3783 NL_SET_ERR_MSG(extack, 3784 "Invalid local interface, device not found"); 3785 return -ENODEV; 3786 } 3787 3788 #if IS_ENABLED(CONFIG_IPV6) 3789 if (use_ipv6) { 3790 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3791 3792 if (idev && idev->cnf.disable_ipv6) { 3793 NL_SET_ERR_MSG(extack, 3794 "IPv6 support disabled by administrator"); 3795 return -EPERM; 3796 } 3797 } 3798 #endif 3799 3800 *lower = lowerdev; 3801 } else { 3802 if (vxlan_addr_multicast(&conf->remote_ip)) { 3803 NL_SET_ERR_MSG(extack, 3804 "Local interface required for multicast remote destination"); 3805 3806 return -EINVAL; 3807 } 3808 3809 #if IS_ENABLED(CONFIG_IPV6) 3810 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3811 NL_SET_ERR_MSG(extack, 3812 "Local interface required for link-local local/remote addresses"); 3813 return -EINVAL; 3814 } 3815 #endif 3816 3817 *lower = NULL; 3818 } 3819 3820 if (!conf->dst_port) { 3821 if (conf->flags & VXLAN_F_GPE) 3822 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3823 else 3824 conf->dst_port = htons(vxlan_port); 3825 } 3826 3827 if (!conf->age_interval) 3828 conf->age_interval = FDB_AGE_DEFAULT; 3829 3830 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3831 NL_SET_ERR_MSG(extack, 3832 "A VXLAN device with the specified VNI already exists"); 3833 return -EEXIST; 3834 } 3835 3836 return 0; 3837 } 3838 3839 static void vxlan_config_apply(struct net_device *dev, 3840 struct vxlan_config *conf, 3841 struct net_device *lowerdev, 3842 struct net *src_net, 3843 bool changelink) 3844 { 3845 struct vxlan_dev *vxlan = netdev_priv(dev); 3846 struct vxlan_rdst *dst = &vxlan->default_dst; 3847 unsigned short needed_headroom = ETH_HLEN; 3848 int max_mtu = ETH_MAX_MTU; 3849 u32 flags = conf->flags; 3850 3851 if (!changelink) { 3852 if (flags & VXLAN_F_GPE) 3853 vxlan_raw_setup(dev); 3854 else 3855 vxlan_ether_setup(dev); 3856 3857 if (conf->mtu) 3858 dev->mtu = conf->mtu; 3859 3860 vxlan->net = src_net; 3861 } 3862 3863 dst->remote_vni = conf->vni; 3864 3865 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3866 3867 if (lowerdev) { 3868 dst->remote_ifindex = conf->remote_ifindex; 3869 3870 netif_inherit_tso_max(dev, lowerdev); 3871 3872 needed_headroom = lowerdev->hard_header_len; 3873 needed_headroom += lowerdev->needed_headroom; 3874 3875 dev->needed_tailroom = lowerdev->needed_tailroom; 3876 3877 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3878 if (max_mtu < ETH_MIN_MTU) 3879 max_mtu = ETH_MIN_MTU; 3880 3881 if (!changelink && !conf->mtu) 3882 dev->mtu = max_mtu; 3883 } 3884 3885 if (dev->mtu > max_mtu) 3886 dev->mtu = max_mtu; 3887 3888 if (flags & VXLAN_F_COLLECT_METADATA) 3889 flags |= VXLAN_F_IPV6; 3890 needed_headroom += vxlan_headroom(flags); 3891 dev->needed_headroom = needed_headroom; 3892 3893 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3894 } 3895 3896 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3897 struct vxlan_config *conf, bool changelink, 3898 struct netlink_ext_ack *extack) 3899 { 3900 struct vxlan_dev *vxlan = netdev_priv(dev); 3901 struct net_device *lowerdev; 3902 int ret; 3903 3904 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3905 if (ret) 3906 return ret; 3907 3908 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3909 3910 return 0; 3911 } 3912 3913 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3914 struct vxlan_config *conf, 3915 struct netlink_ext_ack *extack) 3916 { 3917 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3918 struct vxlan_dev *vxlan = netdev_priv(dev); 3919 struct net_device *remote_dev = NULL; 3920 struct vxlan_fdb *f = NULL; 3921 bool unregister = false; 3922 struct vxlan_rdst *dst; 3923 int err; 3924 3925 dst = &vxlan->default_dst; 3926 err = vxlan_dev_configure(net, dev, conf, false, extack); 3927 if (err) 3928 return err; 3929 3930 dev->ethtool_ops = &vxlan_ethtool_ops; 3931 3932 /* create an fdb entry for a valid default destination */ 3933 if (!vxlan_addr_any(&dst->remote_ip)) { 3934 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3935 &dst->remote_ip, 3936 NUD_REACHABLE | NUD_PERMANENT, 3937 vxlan->cfg.dst_port, 3938 dst->remote_vni, 3939 dst->remote_vni, 3940 dst->remote_ifindex, 3941 NTF_SELF, 0, &f, extack); 3942 if (err) 3943 return err; 3944 } 3945 3946 err = register_netdevice(dev); 3947 if (err) 3948 goto errout; 3949 unregister = true; 3950 3951 if (dst->remote_ifindex) { 3952 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3953 if (!remote_dev) { 3954 err = -ENODEV; 3955 goto errout; 3956 } 3957 3958 err = netdev_upper_dev_link(remote_dev, dev, extack); 3959 if (err) 3960 goto errout; 3961 } 3962 3963 err = rtnl_configure_link(dev, NULL, 0, NULL); 3964 if (err < 0) 3965 goto unlink; 3966 3967 if (f) { 3968 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3969 3970 /* notify default fdb entry */ 3971 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3972 RTM_NEWNEIGH, true, extack); 3973 if (err) { 3974 vxlan_fdb_destroy(vxlan, f, false, false); 3975 if (remote_dev) 3976 netdev_upper_dev_unlink(remote_dev, dev); 3977 goto unregister; 3978 } 3979 } 3980 3981 list_add(&vxlan->next, &vn->vxlan_list); 3982 if (remote_dev) 3983 dst->remote_dev = remote_dev; 3984 return 0; 3985 unlink: 3986 if (remote_dev) 3987 netdev_upper_dev_unlink(remote_dev, dev); 3988 errout: 3989 /* unregister_netdevice() destroys the default FDB entry with deletion 3990 * notification. But the addition notification was not sent yet, so 3991 * destroy the entry by hand here. 3992 */ 3993 if (f) 3994 __vxlan_fdb_free(f); 3995 unregister: 3996 if (unregister) 3997 unregister_netdevice(dev); 3998 return err; 3999 } 4000 4001 /* Set/clear flags based on attribute */ 4002 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4003 int attrtype, unsigned long mask, bool changelink, 4004 bool changelink_supported, 4005 struct netlink_ext_ack *extack) 4006 { 4007 unsigned long flags; 4008 4009 if (!tb[attrtype]) 4010 return 0; 4011 4012 if (changelink && !changelink_supported) { 4013 vxlan_flag_attr_error(attrtype, extack); 4014 return -EOPNOTSUPP; 4015 } 4016 4017 if (vxlan_policy[attrtype].type == NLA_FLAG) 4018 flags = conf->flags | mask; 4019 else if (nla_get_u8(tb[attrtype])) 4020 flags = conf->flags | mask; 4021 else 4022 flags = conf->flags & ~mask; 4023 4024 conf->flags = flags; 4025 4026 return 0; 4027 } 4028 4029 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4030 struct net_device *dev, struct vxlan_config *conf, 4031 bool changelink, struct netlink_ext_ack *extack) 4032 { 4033 struct vxlan_dev *vxlan = netdev_priv(dev); 4034 int err = 0; 4035 4036 memset(conf, 0, sizeof(*conf)); 4037 4038 /* if changelink operation, start with old existing cfg */ 4039 if (changelink) 4040 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4041 4042 if (data[IFLA_VXLAN_ID]) { 4043 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4044 4045 if (changelink && (vni != conf->vni)) { 4046 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4047 return -EOPNOTSUPP; 4048 } 4049 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4050 } 4051 4052 if (data[IFLA_VXLAN_GROUP]) { 4053 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4054 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4055 return -EOPNOTSUPP; 4056 } 4057 4058 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4059 conf->remote_ip.sa.sa_family = AF_INET; 4060 } else if (data[IFLA_VXLAN_GROUP6]) { 4061 if (!IS_ENABLED(CONFIG_IPV6)) { 4062 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4063 return -EPFNOSUPPORT; 4064 } 4065 4066 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4067 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4068 return -EOPNOTSUPP; 4069 } 4070 4071 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4072 conf->remote_ip.sa.sa_family = AF_INET6; 4073 } 4074 4075 if (data[IFLA_VXLAN_LOCAL]) { 4076 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4077 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4078 return -EOPNOTSUPP; 4079 } 4080 4081 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4082 conf->saddr.sa.sa_family = AF_INET; 4083 } else if (data[IFLA_VXLAN_LOCAL6]) { 4084 if (!IS_ENABLED(CONFIG_IPV6)) { 4085 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4086 return -EPFNOSUPPORT; 4087 } 4088 4089 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4090 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4091 return -EOPNOTSUPP; 4092 } 4093 4094 /* TODO: respect scope id */ 4095 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4096 conf->saddr.sa.sa_family = AF_INET6; 4097 } 4098 4099 if (data[IFLA_VXLAN_LINK]) 4100 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4101 4102 if (data[IFLA_VXLAN_TOS]) 4103 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4104 4105 if (data[IFLA_VXLAN_TTL]) 4106 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4107 4108 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4109 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4110 VXLAN_F_TTL_INHERIT, changelink, false, 4111 extack); 4112 if (err) 4113 return err; 4114 4115 } 4116 4117 if (data[IFLA_VXLAN_LABEL]) 4118 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4119 IPV6_FLOWLABEL_MASK; 4120 if (data[IFLA_VXLAN_LABEL_POLICY]) 4121 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4122 4123 if (data[IFLA_VXLAN_LEARNING]) { 4124 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4125 VXLAN_F_LEARN, changelink, true, 4126 extack); 4127 if (err) 4128 return err; 4129 } else if (!changelink) { 4130 /* default to learn on a new device */ 4131 conf->flags |= VXLAN_F_LEARN; 4132 } 4133 4134 if (data[IFLA_VXLAN_AGEING]) 4135 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4136 4137 if (data[IFLA_VXLAN_PROXY]) { 4138 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4139 VXLAN_F_PROXY, changelink, false, 4140 extack); 4141 if (err) 4142 return err; 4143 } 4144 4145 if (data[IFLA_VXLAN_RSC]) { 4146 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4147 VXLAN_F_RSC, changelink, false, 4148 extack); 4149 if (err) 4150 return err; 4151 } 4152 4153 if (data[IFLA_VXLAN_L2MISS]) { 4154 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4155 VXLAN_F_L2MISS, changelink, false, 4156 extack); 4157 if (err) 4158 return err; 4159 } 4160 4161 if (data[IFLA_VXLAN_L3MISS]) { 4162 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4163 VXLAN_F_L3MISS, changelink, false, 4164 extack); 4165 if (err) 4166 return err; 4167 } 4168 4169 if (data[IFLA_VXLAN_LIMIT]) { 4170 if (changelink) { 4171 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4172 "Cannot change limit"); 4173 return -EOPNOTSUPP; 4174 } 4175 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4176 } 4177 4178 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4179 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4180 VXLAN_F_COLLECT_METADATA, changelink, false, 4181 extack); 4182 if (err) 4183 return err; 4184 } 4185 4186 if (data[IFLA_VXLAN_PORT_RANGE]) { 4187 if (!changelink) { 4188 const struct ifla_vxlan_port_range *p 4189 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4190 conf->port_min = ntohs(p->low); 4191 conf->port_max = ntohs(p->high); 4192 } else { 4193 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4194 "Cannot change port range"); 4195 return -EOPNOTSUPP; 4196 } 4197 } 4198 4199 if (data[IFLA_VXLAN_PORT]) { 4200 if (changelink) { 4201 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4202 "Cannot change port"); 4203 return -EOPNOTSUPP; 4204 } 4205 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4206 } 4207 4208 if (data[IFLA_VXLAN_UDP_CSUM]) { 4209 if (changelink) { 4210 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4211 "Cannot change UDP_CSUM flag"); 4212 return -EOPNOTSUPP; 4213 } 4214 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4215 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4216 } 4217 4218 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4219 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4220 VXLAN_F_LOCALBYPASS, changelink, 4221 true, extack); 4222 if (err) 4223 return err; 4224 } else if (!changelink) { 4225 /* default to local bypass on a new device */ 4226 conf->flags |= VXLAN_F_LOCALBYPASS; 4227 } 4228 4229 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4230 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4231 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4232 false, extack); 4233 if (err) 4234 return err; 4235 } 4236 4237 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4238 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4239 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4240 false, extack); 4241 if (err) 4242 return err; 4243 } 4244 4245 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4246 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4247 VXLAN_F_REMCSUM_TX, changelink, false, 4248 extack); 4249 if (err) 4250 return err; 4251 } 4252 4253 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4254 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4255 VXLAN_F_REMCSUM_RX, changelink, false, 4256 extack); 4257 if (err) 4258 return err; 4259 } 4260 4261 if (data[IFLA_VXLAN_GBP]) { 4262 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4263 VXLAN_F_GBP, changelink, false, extack); 4264 if (err) 4265 return err; 4266 } 4267 4268 if (data[IFLA_VXLAN_GPE]) { 4269 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4270 VXLAN_F_GPE, changelink, false, 4271 extack); 4272 if (err) 4273 return err; 4274 } 4275 4276 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4277 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4278 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4279 false, extack); 4280 if (err) 4281 return err; 4282 } 4283 4284 if (tb[IFLA_MTU]) { 4285 if (changelink) { 4286 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4287 "Cannot change mtu"); 4288 return -EOPNOTSUPP; 4289 } 4290 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4291 } 4292 4293 if (data[IFLA_VXLAN_DF]) 4294 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4295 4296 if (data[IFLA_VXLAN_VNIFILTER]) { 4297 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4298 VXLAN_F_VNIFILTER, changelink, false, 4299 extack); 4300 if (err) 4301 return err; 4302 4303 if ((conf->flags & VXLAN_F_VNIFILTER) && 4304 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4305 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4306 "vxlan vnifilter only valid in collect metadata mode"); 4307 return -EINVAL; 4308 } 4309 } 4310 4311 return 0; 4312 } 4313 4314 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4315 struct nlattr *tb[], struct nlattr *data[], 4316 struct netlink_ext_ack *extack) 4317 { 4318 struct vxlan_config conf; 4319 int err; 4320 4321 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4322 if (err) 4323 return err; 4324 4325 return __vxlan_dev_create(src_net, dev, &conf, extack); 4326 } 4327 4328 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4329 struct nlattr *data[], 4330 struct netlink_ext_ack *extack) 4331 { 4332 struct vxlan_dev *vxlan = netdev_priv(dev); 4333 struct net_device *lowerdev; 4334 struct vxlan_config conf; 4335 struct vxlan_rdst *dst; 4336 int err; 4337 4338 dst = &vxlan->default_dst; 4339 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4340 if (err) 4341 return err; 4342 4343 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4344 vxlan, extack); 4345 if (err) 4346 return err; 4347 4348 if (dst->remote_dev == lowerdev) 4349 lowerdev = NULL; 4350 4351 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4352 extack); 4353 if (err) 4354 return err; 4355 4356 /* handle default dst entry */ 4357 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4358 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4359 4360 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4361 if (!vxlan_addr_any(&conf.remote_ip)) { 4362 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4363 &conf.remote_ip, 4364 NUD_REACHABLE | NUD_PERMANENT, 4365 NLM_F_APPEND | NLM_F_CREATE, 4366 vxlan->cfg.dst_port, 4367 conf.vni, conf.vni, 4368 conf.remote_ifindex, 4369 NTF_SELF, 0, true, extack); 4370 if (err) { 4371 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4372 netdev_adjacent_change_abort(dst->remote_dev, 4373 lowerdev, dev); 4374 return err; 4375 } 4376 } 4377 if (!vxlan_addr_any(&dst->remote_ip)) 4378 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4379 dst->remote_ip, 4380 vxlan->cfg.dst_port, 4381 dst->remote_vni, 4382 dst->remote_vni, 4383 dst->remote_ifindex, 4384 true); 4385 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4386 4387 /* If vni filtering device, also update fdb entries of 4388 * all vnis that were using default remote ip 4389 */ 4390 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4391 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4392 &conf.remote_ip, extack); 4393 if (err) { 4394 netdev_adjacent_change_abort(dst->remote_dev, 4395 lowerdev, dev); 4396 return err; 4397 } 4398 } 4399 } 4400 4401 if (conf.age_interval != vxlan->cfg.age_interval) 4402 mod_timer(&vxlan->age_timer, jiffies); 4403 4404 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4405 if (lowerdev && lowerdev != dst->remote_dev) 4406 dst->remote_dev = lowerdev; 4407 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4408 return 0; 4409 } 4410 4411 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4412 { 4413 struct vxlan_dev *vxlan = netdev_priv(dev); 4414 struct vxlan_fdb_flush_desc desc = { 4415 /* Default entry is deleted at vxlan_uninit. */ 4416 .ignore_default_entry = true, 4417 }; 4418 4419 vxlan_flush(vxlan, &desc); 4420 4421 list_del(&vxlan->next); 4422 unregister_netdevice_queue(dev, head); 4423 if (vxlan->default_dst.remote_dev) 4424 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4425 } 4426 4427 static size_t vxlan_get_size(const struct net_device *dev) 4428 { 4429 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4430 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4431 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4432 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4433 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4434 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4435 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4436 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4437 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4438 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4439 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4440 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4441 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4442 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4443 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4444 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4445 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4446 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4447 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4448 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4449 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4450 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4451 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4452 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4453 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4454 /* IFLA_VXLAN_PORT_RANGE */ 4455 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4456 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4457 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4458 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4459 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4460 0; 4461 } 4462 4463 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4464 { 4465 const struct vxlan_dev *vxlan = netdev_priv(dev); 4466 const struct vxlan_rdst *dst = &vxlan->default_dst; 4467 struct ifla_vxlan_port_range ports = { 4468 .low = htons(vxlan->cfg.port_min), 4469 .high = htons(vxlan->cfg.port_max), 4470 }; 4471 4472 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4473 goto nla_put_failure; 4474 4475 if (!vxlan_addr_any(&dst->remote_ip)) { 4476 if (dst->remote_ip.sa.sa_family == AF_INET) { 4477 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4478 dst->remote_ip.sin.sin_addr.s_addr)) 4479 goto nla_put_failure; 4480 #if IS_ENABLED(CONFIG_IPV6) 4481 } else { 4482 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4483 &dst->remote_ip.sin6.sin6_addr)) 4484 goto nla_put_failure; 4485 #endif 4486 } 4487 } 4488 4489 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4490 goto nla_put_failure; 4491 4492 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4493 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4494 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4495 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4496 goto nla_put_failure; 4497 #if IS_ENABLED(CONFIG_IPV6) 4498 } else { 4499 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4500 &vxlan->cfg.saddr.sin6.sin6_addr)) 4501 goto nla_put_failure; 4502 #endif 4503 } 4504 } 4505 4506 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4507 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4508 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4509 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4510 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4511 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4512 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4513 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4514 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4515 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4516 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4517 nla_put_u8(skb, IFLA_VXLAN_RSC, 4518 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4519 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4520 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4521 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4522 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4523 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4524 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4525 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4526 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4527 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4528 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4529 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4530 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4531 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4532 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4533 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4534 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4535 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4536 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4537 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4538 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4539 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4540 goto nla_put_failure; 4541 4542 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4543 goto nla_put_failure; 4544 4545 if (vxlan->cfg.flags & VXLAN_F_GBP && 4546 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4547 goto nla_put_failure; 4548 4549 if (vxlan->cfg.flags & VXLAN_F_GPE && 4550 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4551 goto nla_put_failure; 4552 4553 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4554 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4555 goto nla_put_failure; 4556 4557 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4558 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4559 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4560 goto nla_put_failure; 4561 4562 return 0; 4563 4564 nla_put_failure: 4565 return -EMSGSIZE; 4566 } 4567 4568 static struct net *vxlan_get_link_net(const struct net_device *dev) 4569 { 4570 struct vxlan_dev *vxlan = netdev_priv(dev); 4571 4572 return READ_ONCE(vxlan->net); 4573 } 4574 4575 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4576 .kind = "vxlan", 4577 .maxtype = IFLA_VXLAN_MAX, 4578 .policy = vxlan_policy, 4579 .priv_size = sizeof(struct vxlan_dev), 4580 .setup = vxlan_setup, 4581 .validate = vxlan_validate, 4582 .newlink = vxlan_newlink, 4583 .changelink = vxlan_changelink, 4584 .dellink = vxlan_dellink, 4585 .get_size = vxlan_get_size, 4586 .fill_info = vxlan_fill_info, 4587 .get_link_net = vxlan_get_link_net, 4588 }; 4589 4590 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4591 u8 name_assign_type, 4592 struct vxlan_config *conf) 4593 { 4594 struct nlattr *tb[IFLA_MAX + 1]; 4595 struct net_device *dev; 4596 int err; 4597 4598 memset(&tb, 0, sizeof(tb)); 4599 4600 dev = rtnl_create_link(net, name, name_assign_type, 4601 &vxlan_link_ops, tb, NULL); 4602 if (IS_ERR(dev)) 4603 return dev; 4604 4605 err = __vxlan_dev_create(net, dev, conf, NULL); 4606 if (err < 0) { 4607 free_netdev(dev); 4608 return ERR_PTR(err); 4609 } 4610 4611 err = rtnl_configure_link(dev, NULL, 0, NULL); 4612 if (err < 0) { 4613 LIST_HEAD(list_kill); 4614 4615 vxlan_dellink(dev, &list_kill); 4616 unregister_netdevice_many(&list_kill); 4617 return ERR_PTR(err); 4618 } 4619 4620 return dev; 4621 } 4622 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4623 4624 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4625 struct net_device *dev) 4626 { 4627 struct vxlan_dev *vxlan, *next; 4628 LIST_HEAD(list_kill); 4629 4630 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4631 struct vxlan_rdst *dst = &vxlan->default_dst; 4632 4633 /* In case we created vxlan device with carrier 4634 * and we loose the carrier due to module unload 4635 * we also need to remove vxlan device. In other 4636 * cases, it's not necessary and remote_ifindex 4637 * is 0 here, so no matches. 4638 */ 4639 if (dst->remote_ifindex == dev->ifindex) 4640 vxlan_dellink(vxlan->dev, &list_kill); 4641 } 4642 4643 unregister_netdevice_many(&list_kill); 4644 } 4645 4646 static int vxlan_netdevice_event(struct notifier_block *unused, 4647 unsigned long event, void *ptr) 4648 { 4649 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4650 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4651 4652 if (event == NETDEV_UNREGISTER) 4653 vxlan_handle_lowerdev_unregister(vn, dev); 4654 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4655 vxlan_offload_rx_ports(dev, true); 4656 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4657 vxlan_offload_rx_ports(dev, false); 4658 4659 return NOTIFY_DONE; 4660 } 4661 4662 static struct notifier_block vxlan_notifier_block __read_mostly = { 4663 .notifier_call = vxlan_netdevice_event, 4664 }; 4665 4666 static void 4667 vxlan_fdb_offloaded_set(struct net_device *dev, 4668 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4669 { 4670 struct vxlan_dev *vxlan = netdev_priv(dev); 4671 struct vxlan_rdst *rdst; 4672 struct vxlan_fdb *f; 4673 u32 hash_index; 4674 4675 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4676 4677 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4678 4679 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4680 if (!f) 4681 goto out; 4682 4683 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4684 fdb_info->remote_port, 4685 fdb_info->remote_vni, 4686 fdb_info->remote_ifindex); 4687 if (!rdst) 4688 goto out; 4689 4690 rdst->offloaded = fdb_info->offloaded; 4691 4692 out: 4693 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4694 } 4695 4696 static int 4697 vxlan_fdb_external_learn_add(struct net_device *dev, 4698 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4699 { 4700 struct vxlan_dev *vxlan = netdev_priv(dev); 4701 struct netlink_ext_ack *extack; 4702 u32 hash_index; 4703 int err; 4704 4705 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4706 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4707 4708 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4709 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4710 NUD_REACHABLE, 4711 NLM_F_CREATE | NLM_F_REPLACE, 4712 fdb_info->remote_port, 4713 fdb_info->vni, 4714 fdb_info->remote_vni, 4715 fdb_info->remote_ifindex, 4716 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4717 0, false, extack); 4718 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4719 4720 return err; 4721 } 4722 4723 static int 4724 vxlan_fdb_external_learn_del(struct net_device *dev, 4725 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4726 { 4727 struct vxlan_dev *vxlan = netdev_priv(dev); 4728 struct vxlan_fdb *f; 4729 u32 hash_index; 4730 int err = 0; 4731 4732 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4733 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4734 4735 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4736 if (!f) 4737 err = -ENOENT; 4738 else if (f->flags & NTF_EXT_LEARNED) 4739 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4740 fdb_info->remote_ip, 4741 fdb_info->remote_port, 4742 fdb_info->vni, 4743 fdb_info->remote_vni, 4744 fdb_info->remote_ifindex, 4745 false); 4746 4747 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4748 4749 return err; 4750 } 4751 4752 static int vxlan_switchdev_event(struct notifier_block *unused, 4753 unsigned long event, void *ptr) 4754 { 4755 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4756 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4757 int err = 0; 4758 4759 switch (event) { 4760 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4761 vxlan_fdb_offloaded_set(dev, ptr); 4762 break; 4763 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4764 fdb_info = ptr; 4765 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4766 if (err) { 4767 err = notifier_from_errno(err); 4768 break; 4769 } 4770 fdb_info->offloaded = true; 4771 vxlan_fdb_offloaded_set(dev, fdb_info); 4772 break; 4773 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4774 fdb_info = ptr; 4775 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4776 if (err) { 4777 err = notifier_from_errno(err); 4778 break; 4779 } 4780 fdb_info->offloaded = false; 4781 vxlan_fdb_offloaded_set(dev, fdb_info); 4782 break; 4783 } 4784 4785 return err; 4786 } 4787 4788 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4789 .notifier_call = vxlan_switchdev_event, 4790 }; 4791 4792 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4793 { 4794 struct vxlan_fdb *fdb; 4795 struct vxlan_dev *vxlan; 4796 u32 hash_index; 4797 4798 rcu_read_lock(); 4799 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4800 vxlan = rcu_dereference(fdb->vdev); 4801 WARN_ON(!vxlan); 4802 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4803 vxlan->default_dst.remote_vni); 4804 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4805 if (!hlist_unhashed(&fdb->hlist)) 4806 vxlan_fdb_destroy(vxlan, fdb, false, false); 4807 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4808 } 4809 rcu_read_unlock(); 4810 } 4811 4812 static int vxlan_nexthop_event(struct notifier_block *nb, 4813 unsigned long event, void *ptr) 4814 { 4815 struct nh_notifier_info *info = ptr; 4816 struct nexthop *nh; 4817 4818 if (event != NEXTHOP_EVENT_DEL) 4819 return NOTIFY_DONE; 4820 4821 nh = nexthop_find_by_id(info->net, info->id); 4822 if (!nh) 4823 return NOTIFY_DONE; 4824 4825 vxlan_fdb_nh_flush(nh); 4826 4827 return NOTIFY_DONE; 4828 } 4829 4830 static __net_init int vxlan_init_net(struct net *net) 4831 { 4832 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4833 unsigned int h; 4834 4835 INIT_LIST_HEAD(&vn->vxlan_list); 4836 spin_lock_init(&vn->sock_lock); 4837 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4838 4839 for (h = 0; h < PORT_HASH_SIZE; ++h) 4840 INIT_HLIST_HEAD(&vn->sock_list[h]); 4841 4842 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4843 NULL); 4844 } 4845 4846 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4847 struct list_head *dev_to_kill) 4848 { 4849 struct vxlan_dev *vxlan, *next; 4850 4851 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4852 vxlan_dellink(vxlan->dev, dev_to_kill); 4853 } 4854 4855 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4856 struct list_head *dev_to_kill) 4857 { 4858 struct net *net; 4859 4860 ASSERT_RTNL(); 4861 list_for_each_entry(net, net_list, exit_list) { 4862 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4863 4864 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4865 4866 vxlan_destroy_tunnels(vn, dev_to_kill); 4867 } 4868 } 4869 4870 static void __net_exit vxlan_exit_net(struct net *net) 4871 { 4872 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4873 unsigned int h; 4874 4875 for (h = 0; h < PORT_HASH_SIZE; ++h) 4876 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4877 } 4878 4879 static struct pernet_operations vxlan_net_ops = { 4880 .init = vxlan_init_net, 4881 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4882 .exit = vxlan_exit_net, 4883 .id = &vxlan_net_id, 4884 .size = sizeof(struct vxlan_net), 4885 }; 4886 4887 static int __init vxlan_init_module(void) 4888 { 4889 int rc; 4890 4891 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4892 4893 rc = register_pernet_subsys(&vxlan_net_ops); 4894 if (rc) 4895 goto out1; 4896 4897 rc = register_netdevice_notifier(&vxlan_notifier_block); 4898 if (rc) 4899 goto out2; 4900 4901 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4902 if (rc) 4903 goto out3; 4904 4905 rc = rtnl_link_register(&vxlan_link_ops); 4906 if (rc) 4907 goto out4; 4908 4909 vxlan_vnifilter_init(); 4910 4911 return 0; 4912 out4: 4913 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4914 out3: 4915 unregister_netdevice_notifier(&vxlan_notifier_block); 4916 out2: 4917 unregister_pernet_subsys(&vxlan_net_ops); 4918 out1: 4919 return rc; 4920 } 4921 late_initcall(vxlan_init_module); 4922 4923 static void __exit vxlan_cleanup_module(void) 4924 { 4925 vxlan_vnifilter_uninit(); 4926 rtnl_link_unregister(&vxlan_link_ops); 4927 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4928 unregister_netdevice_notifier(&vxlan_notifier_block); 4929 unregister_pernet_subsys(&vxlan_net_ops); 4930 /* rcu_barrier() is called by netns */ 4931 } 4932 module_exit(vxlan_cleanup_module); 4933 4934 MODULE_LICENSE("GPL"); 4935 MODULE_VERSION(VXLAN_VERSION); 4936 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4937 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4938 MODULE_ALIAS_RTNL_LINK("vxlan"); 4939