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 - READ_ONCE(fdb->used)); 231 ci.ndm_confirmed = 0; 232 ci.ndm_updated = jiffies_to_clock_t(now - READ_ONCE(fdb->updated)); 233 ci.ndm_refcnt = 0; 234 235 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) 236 goto nla_put_failure; 237 238 nlmsg_end(skb, nlh); 239 return 0; 240 241 nla_put_failure: 242 nlmsg_cancel(skb, nlh); 243 return -EMSGSIZE; 244 } 245 246 static inline size_t vxlan_nlmsg_size(void) 247 { 248 return NLMSG_ALIGN(sizeof(struct ndmsg)) 249 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ 250 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ 251 + nla_total_size(sizeof(__be16)) /* NDA_PORT */ 252 + nla_total_size(sizeof(__be32)) /* NDA_VNI */ 253 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ 254 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ 255 + nla_total_size(sizeof(struct nda_cacheinfo)); 256 } 257 258 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 259 struct vxlan_rdst *rd, int type) 260 { 261 struct net *net = dev_net(vxlan->dev); 262 struct sk_buff *skb; 263 int err = -ENOBUFS; 264 265 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); 266 if (skb == NULL) 267 goto errout; 268 269 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); 270 if (err < 0) { 271 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ 272 WARN_ON(err == -EMSGSIZE); 273 kfree_skb(skb); 274 goto errout; 275 } 276 277 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 278 return; 279 errout: 280 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 281 } 282 283 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 284 const struct vxlan_fdb *fdb, 285 const struct vxlan_rdst *rd, 286 struct netlink_ext_ack *extack, 287 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 288 { 289 fdb_info->info.dev = vxlan->dev; 290 fdb_info->info.extack = extack; 291 fdb_info->remote_ip = rd->remote_ip; 292 fdb_info->remote_port = rd->remote_port; 293 fdb_info->remote_vni = rd->remote_vni; 294 fdb_info->remote_ifindex = rd->remote_ifindex; 295 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN); 296 fdb_info->vni = fdb->vni; 297 fdb_info->offloaded = rd->offloaded; 298 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 299 } 300 301 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 302 struct vxlan_fdb *fdb, 303 struct vxlan_rdst *rd, 304 bool adding, 305 struct netlink_ext_ack *extack) 306 { 307 struct switchdev_notifier_vxlan_fdb_info info; 308 enum switchdev_notifier_type notifier_type; 309 int ret; 310 311 if (WARN_ON(!rd)) 312 return 0; 313 314 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 315 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 316 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 317 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 318 &info.info, extack); 319 return notifier_to_errno(ret); 320 } 321 322 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 323 struct vxlan_rdst *rd, int type, bool swdev_notify, 324 struct netlink_ext_ack *extack) 325 { 326 int err; 327 328 if (swdev_notify && rd) { 329 switch (type) { 330 case RTM_NEWNEIGH: 331 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 332 true, extack); 333 if (err) 334 return err; 335 break; 336 case RTM_DELNEIGH: 337 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 338 false, extack); 339 break; 340 } 341 } 342 343 __vxlan_fdb_notify(vxlan, fdb, rd, type); 344 return 0; 345 } 346 347 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 348 { 349 struct vxlan_dev *vxlan = netdev_priv(dev); 350 struct vxlan_fdb f = { 351 .state = NUD_STALE, 352 }; 353 struct vxlan_rdst remote = { 354 .remote_ip = *ipa, /* goes to NDA_DST */ 355 .remote_vni = cpu_to_be32(VXLAN_N_VID), 356 }; 357 358 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 359 } 360 361 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 362 { 363 struct vxlan_fdb f = { 364 .state = NUD_STALE, 365 }; 366 struct vxlan_rdst remote = { }; 367 368 memcpy(f.eth_addr, eth_addr, ETH_ALEN); 369 370 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 371 } 372 373 /* Hash Ethernet address */ 374 static u32 eth_hash(const unsigned char *addr) 375 { 376 u64 value = get_unaligned((u64 *)addr); 377 378 /* only want 6 bytes */ 379 #ifdef __BIG_ENDIAN 380 value >>= 16; 381 #else 382 value <<= 16; 383 #endif 384 return hash_64(value, FDB_HASH_BITS); 385 } 386 387 u32 eth_vni_hash(const unsigned char *addr, __be32 vni) 388 { 389 /* use 1 byte of OUI and 3 bytes of NIC */ 390 u32 key = get_unaligned((u32 *)(addr + 2)); 391 392 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1); 393 } 394 395 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni) 396 { 397 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) 398 return eth_vni_hash(mac, vni); 399 else 400 return eth_hash(mac); 401 } 402 403 /* Hash chain to use given mac address */ 404 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, 405 const u8 *mac, __be32 vni) 406 { 407 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)]; 408 } 409 410 /* Look up Ethernet address in forwarding table */ 411 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, 412 const u8 *mac, __be32 vni) 413 { 414 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni); 415 struct vxlan_fdb *f; 416 417 hlist_for_each_entry_rcu(f, head, hlist) { 418 if (ether_addr_equal(mac, f->eth_addr)) { 419 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 420 if (vni == f->vni) 421 return f; 422 } else { 423 return f; 424 } 425 } 426 } 427 428 return NULL; 429 } 430 431 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 432 const u8 *mac, __be32 vni) 433 { 434 struct vxlan_fdb *f; 435 436 f = __vxlan_find_mac(vxlan, mac, vni); 437 if (f) { 438 unsigned long now = jiffies; 439 440 if (READ_ONCE(f->used) != now) 441 WRITE_ONCE(f->used, now); 442 } 443 444 return f; 445 } 446 447 /* caller should hold vxlan->hash_lock */ 448 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 449 union vxlan_addr *ip, __be16 port, 450 __be32 vni, __u32 ifindex) 451 { 452 struct vxlan_rdst *rd; 453 454 list_for_each_entry(rd, &f->remotes, list) { 455 if (vxlan_addr_equal(&rd->remote_ip, ip) && 456 rd->remote_port == port && 457 rd->remote_vni == vni && 458 rd->remote_ifindex == ifindex) 459 return rd; 460 } 461 462 return NULL; 463 } 464 465 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 466 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 467 { 468 struct vxlan_dev *vxlan = netdev_priv(dev); 469 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 470 struct vxlan_rdst *rdst; 471 struct vxlan_fdb *f; 472 int rc = 0; 473 474 if (is_multicast_ether_addr(mac) || 475 is_zero_ether_addr(mac)) 476 return -EINVAL; 477 478 ether_addr_copy(eth_addr, mac); 479 480 rcu_read_lock(); 481 482 f = __vxlan_find_mac(vxlan, eth_addr, vni); 483 if (!f) { 484 rc = -ENOENT; 485 goto out; 486 } 487 488 rdst = first_remote_rcu(f); 489 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 490 491 out: 492 rcu_read_unlock(); 493 return rc; 494 } 495 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 496 497 static int vxlan_fdb_notify_one(struct notifier_block *nb, 498 const struct vxlan_dev *vxlan, 499 const struct vxlan_fdb *f, 500 const struct vxlan_rdst *rdst, 501 struct netlink_ext_ack *extack) 502 { 503 struct switchdev_notifier_vxlan_fdb_info fdb_info; 504 int rc; 505 506 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 507 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 508 &fdb_info); 509 return notifier_to_errno(rc); 510 } 511 512 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 513 struct notifier_block *nb, 514 struct netlink_ext_ack *extack) 515 { 516 struct vxlan_dev *vxlan; 517 struct vxlan_rdst *rdst; 518 struct vxlan_fdb *f; 519 unsigned int h; 520 int rc = 0; 521 522 if (!netif_is_vxlan(dev)) 523 return -EINVAL; 524 vxlan = netdev_priv(dev); 525 526 for (h = 0; h < FDB_HASH_SIZE; ++h) { 527 spin_lock_bh(&vxlan->hash_lock[h]); 528 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) { 529 if (f->vni == vni) { 530 list_for_each_entry(rdst, &f->remotes, list) { 531 rc = vxlan_fdb_notify_one(nb, vxlan, 532 f, rdst, 533 extack); 534 if (rc) 535 goto unlock; 536 } 537 } 538 } 539 spin_unlock_bh(&vxlan->hash_lock[h]); 540 } 541 return 0; 542 543 unlock: 544 spin_unlock_bh(&vxlan->hash_lock[h]); 545 return rc; 546 } 547 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 548 549 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 550 { 551 struct vxlan_dev *vxlan; 552 struct vxlan_rdst *rdst; 553 struct vxlan_fdb *f; 554 unsigned int h; 555 556 if (!netif_is_vxlan(dev)) 557 return; 558 vxlan = netdev_priv(dev); 559 560 for (h = 0; h < FDB_HASH_SIZE; ++h) { 561 spin_lock_bh(&vxlan->hash_lock[h]); 562 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) 563 if (f->vni == vni) 564 list_for_each_entry(rdst, &f->remotes, list) 565 rdst->offloaded = false; 566 spin_unlock_bh(&vxlan->hash_lock[h]); 567 } 568 569 } 570 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 571 572 /* Replace destination of unicast mac */ 573 static int vxlan_fdb_replace(struct vxlan_fdb *f, 574 union vxlan_addr *ip, __be16 port, __be32 vni, 575 __u32 ifindex, struct vxlan_rdst *oldrd) 576 { 577 struct vxlan_rdst *rd; 578 579 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 580 if (rd) 581 return 0; 582 583 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 584 if (!rd) 585 return 0; 586 587 *oldrd = *rd; 588 dst_cache_reset(&rd->dst_cache); 589 rd->remote_ip = *ip; 590 rd->remote_port = port; 591 rd->remote_vni = vni; 592 rd->remote_ifindex = ifindex; 593 rd->offloaded = false; 594 return 1; 595 } 596 597 /* Add/update destinations for multicast */ 598 static int vxlan_fdb_append(struct vxlan_fdb *f, 599 union vxlan_addr *ip, __be16 port, __be32 vni, 600 __u32 ifindex, struct vxlan_rdst **rdp) 601 { 602 struct vxlan_rdst *rd; 603 604 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 605 if (rd) 606 return 0; 607 608 rd = kmalloc(sizeof(*rd), GFP_ATOMIC); 609 if (rd == NULL) 610 return -ENOMEM; 611 612 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { 613 kfree(rd); 614 return -ENOMEM; 615 } 616 617 rd->remote_ip = *ip; 618 rd->remote_port = port; 619 rd->offloaded = false; 620 rd->remote_vni = vni; 621 rd->remote_ifindex = ifindex; 622 623 list_add_tail_rcu(&rd->list, &f->remotes); 624 625 *rdp = rd; 626 return 1; 627 } 628 629 static bool vxlan_parse_gpe_proto(const struct vxlanhdr *hdr, __be16 *protocol) 630 { 631 const struct vxlanhdr_gpe *gpe = (const struct vxlanhdr_gpe *)hdr; 632 633 /* Need to have Next Protocol set for interfaces in GPE mode. */ 634 if (!gpe->np_applied) 635 return false; 636 /* "The initial version is 0. If a receiver does not support the 637 * version indicated it MUST drop the packet. 638 */ 639 if (gpe->version != 0) 640 return false; 641 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 642 * processing MUST occur." However, we don't implement OAM 643 * processing, thus drop the packet. 644 */ 645 if (gpe->oam_flag) 646 return false; 647 648 *protocol = tun_p_to_eth_p(gpe->next_protocol); 649 if (!*protocol) 650 return false; 651 652 return true; 653 } 654 655 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 656 unsigned int off, 657 struct vxlanhdr *vh, size_t hdrlen, 658 __be32 vni_field, 659 struct gro_remcsum *grc, 660 bool nopartial) 661 { 662 size_t start, offset; 663 664 if (skb->remcsum_offload) 665 return vh; 666 667 if (!NAPI_GRO_CB(skb)->csum_valid) 668 return NULL; 669 670 start = vxlan_rco_start(vni_field); 671 offset = start + vxlan_rco_offset(vni_field); 672 673 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 674 start, offset, grc, nopartial); 675 676 skb->remcsum_offload = 1; 677 678 return vh; 679 } 680 681 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk, 682 struct list_head *head, 683 struct sk_buff *skb, 684 struct gro_remcsum *grc) 685 { 686 struct sk_buff *p; 687 struct vxlanhdr *vh, *vh2; 688 unsigned int hlen, off_vx; 689 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); 690 __be32 flags; 691 692 skb_gro_remcsum_init(grc); 693 694 off_vx = skb_gro_offset(skb); 695 hlen = off_vx + sizeof(*vh); 696 vh = skb_gro_header(skb, hlen, off_vx); 697 if (unlikely(!vh)) 698 return NULL; 699 700 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 701 702 flags = vh->vx_flags; 703 704 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 705 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 706 vh->vx_vni, grc, 707 !!(vs->flags & 708 VXLAN_F_REMCSUM_NOPARTIAL)); 709 710 if (!vh) 711 return NULL; 712 } 713 714 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 715 716 list_for_each_entry(p, head, list) { 717 if (!NAPI_GRO_CB(p)->same_flow) 718 continue; 719 720 vh2 = (struct vxlanhdr *)(p->data + off_vx); 721 if (vh->vx_flags != vh2->vx_flags || 722 vh->vx_vni != vh2->vx_vni) { 723 NAPI_GRO_CB(p)->same_flow = 0; 724 continue; 725 } 726 } 727 728 return vh; 729 } 730 731 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 732 struct list_head *head, 733 struct sk_buff *skb) 734 { 735 struct sk_buff *pp = NULL; 736 struct gro_remcsum grc; 737 int flush = 1; 738 739 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) { 740 pp = call_gro_receive(eth_gro_receive, head, skb); 741 flush = 0; 742 } 743 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 744 return pp; 745 } 746 747 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk, 748 struct list_head *head, 749 struct sk_buff *skb) 750 { 751 const struct packet_offload *ptype; 752 struct sk_buff *pp = NULL; 753 struct gro_remcsum grc; 754 struct vxlanhdr *vh; 755 __be16 protocol; 756 int flush = 1; 757 758 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc); 759 if (vh) { 760 if (!vxlan_parse_gpe_proto(vh, &protocol)) 761 goto out; 762 ptype = gro_find_receive_by_type(protocol); 763 if (!ptype) 764 goto out; 765 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb); 766 flush = 0; 767 } 768 out: 769 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 770 return pp; 771 } 772 773 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 774 { 775 /* Sets 'skb->inner_mac_header' since we are always called with 776 * 'skb->encapsulation' set. 777 */ 778 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 779 } 780 781 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 782 { 783 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff); 784 const struct packet_offload *ptype; 785 int err = -ENOSYS; 786 __be16 protocol; 787 788 if (!vxlan_parse_gpe_proto(vh, &protocol)) 789 return err; 790 ptype = gro_find_complete_by_type(protocol); 791 if (ptype) 792 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 793 return err; 794 } 795 796 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 797 __u16 state, __be32 src_vni, 798 __u16 ndm_flags) 799 { 800 struct vxlan_fdb *f; 801 802 f = kmalloc(sizeof(*f), GFP_ATOMIC); 803 if (!f) 804 return NULL; 805 f->state = state; 806 f->flags = ndm_flags; 807 f->updated = f->used = jiffies; 808 f->vni = src_vni; 809 f->nh = NULL; 810 RCU_INIT_POINTER(f->vdev, vxlan); 811 INIT_LIST_HEAD(&f->nh_list); 812 INIT_LIST_HEAD(&f->remotes); 813 memcpy(f->eth_addr, mac, ETH_ALEN); 814 815 return f; 816 } 817 818 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac, 819 __be32 src_vni, struct vxlan_fdb *f) 820 { 821 ++vxlan->addrcnt; 822 hlist_add_head_rcu(&f->hlist, 823 vxlan_fdb_head(vxlan, mac, src_vni)); 824 } 825 826 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 827 u32 nhid, struct netlink_ext_ack *extack) 828 { 829 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 830 struct nexthop *nh; 831 int err = -EINVAL; 832 833 if (old_nh && old_nh->id == nhid) 834 return 0; 835 836 nh = nexthop_find_by_id(vxlan->net, nhid); 837 if (!nh) { 838 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 839 goto err_inval; 840 } 841 842 if (!nexthop_get(nh)) { 843 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 844 nh = NULL; 845 goto err_inval; 846 } 847 if (!nexthop_is_fdb(nh)) { 848 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 849 goto err_inval; 850 } 851 852 if (!nexthop_is_multipath(nh)) { 853 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 854 goto err_inval; 855 } 856 857 /* check nexthop group family */ 858 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 859 case AF_INET: 860 if (!nexthop_has_v4(nh)) { 861 err = -EAFNOSUPPORT; 862 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 863 goto err_inval; 864 } 865 break; 866 case AF_INET6: 867 if (nexthop_has_v4(nh)) { 868 err = -EAFNOSUPPORT; 869 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 870 goto err_inval; 871 } 872 } 873 874 if (old_nh) { 875 list_del_rcu(&fdb->nh_list); 876 nexthop_put(old_nh); 877 } 878 rcu_assign_pointer(fdb->nh, nh); 879 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 880 return 1; 881 882 err_inval: 883 if (nh) 884 nexthop_put(nh); 885 return err; 886 } 887 888 int vxlan_fdb_create(struct vxlan_dev *vxlan, 889 const u8 *mac, union vxlan_addr *ip, 890 __u16 state, __be16 port, __be32 src_vni, 891 __be32 vni, __u32 ifindex, __u16 ndm_flags, 892 u32 nhid, struct vxlan_fdb **fdb, 893 struct netlink_ext_ack *extack) 894 { 895 struct vxlan_rdst *rd = NULL; 896 struct vxlan_fdb *f; 897 int rc; 898 899 if (vxlan->cfg.addrmax && 900 vxlan->addrcnt >= vxlan->cfg.addrmax) 901 return -ENOSPC; 902 903 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 904 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 905 if (!f) 906 return -ENOMEM; 907 908 if (nhid) 909 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 910 else 911 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 912 if (rc < 0) 913 goto errout; 914 915 *fdb = f; 916 917 return 0; 918 919 errout: 920 kfree(f); 921 return rc; 922 } 923 924 static void __vxlan_fdb_free(struct vxlan_fdb *f) 925 { 926 struct vxlan_rdst *rd, *nd; 927 struct nexthop *nh; 928 929 nh = rcu_dereference_raw(f->nh); 930 if (nh) { 931 rcu_assign_pointer(f->nh, NULL); 932 rcu_assign_pointer(f->vdev, NULL); 933 nexthop_put(nh); 934 } 935 936 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 937 dst_cache_destroy(&rd->dst_cache); 938 kfree(rd); 939 } 940 kfree(f); 941 } 942 943 static void vxlan_fdb_free(struct rcu_head *head) 944 { 945 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 946 947 __vxlan_fdb_free(f); 948 } 949 950 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 951 bool do_notify, bool swdev_notify) 952 { 953 struct vxlan_rdst *rd; 954 955 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr); 956 957 --vxlan->addrcnt; 958 if (do_notify) { 959 if (rcu_access_pointer(f->nh)) 960 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 961 swdev_notify, NULL); 962 else 963 list_for_each_entry(rd, &f->remotes, list) 964 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 965 swdev_notify, NULL); 966 } 967 968 hlist_del_rcu(&f->hlist); 969 list_del_rcu(&f->nh_list); 970 call_rcu(&f->rcu, vxlan_fdb_free); 971 } 972 973 static void vxlan_dst_free(struct rcu_head *head) 974 { 975 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 976 977 dst_cache_destroy(&rd->dst_cache); 978 kfree(rd); 979 } 980 981 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 982 union vxlan_addr *ip, 983 __u16 state, __u16 flags, 984 __be16 port, __be32 vni, 985 __u32 ifindex, __u16 ndm_flags, 986 struct vxlan_fdb *f, u32 nhid, 987 bool swdev_notify, 988 struct netlink_ext_ack *extack) 989 { 990 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 991 struct vxlan_rdst *rd = NULL; 992 struct vxlan_rdst oldrd; 993 int notify = 0; 994 int rc = 0; 995 int err; 996 997 if (nhid && !rcu_access_pointer(f->nh)) { 998 NL_SET_ERR_MSG(extack, 999 "Cannot replace an existing non nexthop fdb with a nexthop"); 1000 return -EOPNOTSUPP; 1001 } 1002 1003 if (nhid && (flags & NLM_F_APPEND)) { 1004 NL_SET_ERR_MSG(extack, 1005 "Cannot append to a nexthop fdb"); 1006 return -EOPNOTSUPP; 1007 } 1008 1009 /* Do not allow an externally learned entry to take over an entry added 1010 * by the user. 1011 */ 1012 if (!(fdb_flags & NTF_EXT_LEARNED) || 1013 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 1014 if (f->state != state) { 1015 f->state = state; 1016 notify = 1; 1017 } 1018 if (f->flags != fdb_flags) { 1019 f->flags = fdb_flags; 1020 notify = 1; 1021 } 1022 } 1023 1024 if ((flags & NLM_F_REPLACE)) { 1025 /* Only change unicasts */ 1026 if (!(is_multicast_ether_addr(f->eth_addr) || 1027 is_zero_ether_addr(f->eth_addr))) { 1028 if (nhid) { 1029 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1030 if (rc < 0) 1031 return rc; 1032 } else { 1033 rc = vxlan_fdb_replace(f, ip, port, vni, 1034 ifindex, &oldrd); 1035 } 1036 notify |= rc; 1037 } else { 1038 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1039 return -EOPNOTSUPP; 1040 } 1041 } 1042 if ((flags & NLM_F_APPEND) && 1043 (is_multicast_ether_addr(f->eth_addr) || 1044 is_zero_ether_addr(f->eth_addr))) { 1045 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1046 1047 if (rc < 0) 1048 return rc; 1049 notify |= rc; 1050 } 1051 1052 if (ndm_flags & NTF_USE) 1053 WRITE_ONCE(f->updated, jiffies); 1054 1055 if (notify) { 1056 if (rd == NULL) 1057 rd = first_remote_rtnl(f); 1058 1059 WRITE_ONCE(f->updated, jiffies); 1060 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1061 swdev_notify, extack); 1062 if (err) 1063 goto err_notify; 1064 } 1065 1066 return 0; 1067 1068 err_notify: 1069 if (nhid) 1070 return err; 1071 if ((flags & NLM_F_REPLACE) && rc) 1072 *rd = oldrd; 1073 else if ((flags & NLM_F_APPEND) && rc) { 1074 list_del_rcu(&rd->list); 1075 call_rcu(&rd->rcu, vxlan_dst_free); 1076 } 1077 return err; 1078 } 1079 1080 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1081 const u8 *mac, union vxlan_addr *ip, 1082 __u16 state, __u16 flags, 1083 __be16 port, __be32 src_vni, __be32 vni, 1084 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1085 bool swdev_notify, 1086 struct netlink_ext_ack *extack) 1087 { 1088 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1089 struct vxlan_fdb *f; 1090 int rc; 1091 1092 /* Disallow replace to add a multicast entry */ 1093 if ((flags & NLM_F_REPLACE) && 1094 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1095 return -EOPNOTSUPP; 1096 1097 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1098 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1099 vni, ifindex, fdb_flags, nhid, &f, extack); 1100 if (rc < 0) 1101 return rc; 1102 1103 vxlan_fdb_insert(vxlan, mac, src_vni, f); 1104 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1105 swdev_notify, extack); 1106 if (rc) 1107 goto err_notify; 1108 1109 return 0; 1110 1111 err_notify: 1112 vxlan_fdb_destroy(vxlan, f, false, false); 1113 return rc; 1114 } 1115 1116 /* Add new entry to forwarding table -- assumes lock held */ 1117 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1118 const u8 *mac, union vxlan_addr *ip, 1119 __u16 state, __u16 flags, 1120 __be16 port, __be32 src_vni, __be32 vni, 1121 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1122 bool swdev_notify, 1123 struct netlink_ext_ack *extack) 1124 { 1125 struct vxlan_fdb *f; 1126 1127 f = __vxlan_find_mac(vxlan, mac, src_vni); 1128 if (f) { 1129 if (flags & NLM_F_EXCL) { 1130 netdev_dbg(vxlan->dev, 1131 "lost race to create %pM\n", mac); 1132 return -EEXIST; 1133 } 1134 1135 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1136 vni, ifindex, ndm_flags, f, 1137 nhid, swdev_notify, extack); 1138 } else { 1139 if (!(flags & NLM_F_CREATE)) 1140 return -ENOENT; 1141 1142 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1143 port, src_vni, vni, ifindex, 1144 ndm_flags, nhid, swdev_notify, 1145 extack); 1146 } 1147 } 1148 1149 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1150 struct vxlan_rdst *rd, bool swdev_notify) 1151 { 1152 list_del_rcu(&rd->list); 1153 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1154 call_rcu(&rd->rcu, vxlan_dst_free); 1155 } 1156 1157 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1158 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1159 __be32 *vni, u32 *ifindex, u32 *nhid, 1160 struct netlink_ext_ack *extack) 1161 { 1162 struct net *net = dev_net(vxlan->dev); 1163 int err; 1164 1165 if (tb[NDA_NH_ID] && 1166 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1167 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1168 return -EINVAL; 1169 } 1170 1171 if (tb[NDA_DST]) { 1172 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1173 if (err) { 1174 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1175 return err; 1176 } 1177 } else { 1178 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1179 1180 if (remote->sa.sa_family == AF_INET) { 1181 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1182 ip->sa.sa_family = AF_INET; 1183 #if IS_ENABLED(CONFIG_IPV6) 1184 } else { 1185 ip->sin6.sin6_addr = in6addr_any; 1186 ip->sa.sa_family = AF_INET6; 1187 #endif 1188 } 1189 } 1190 1191 if (tb[NDA_PORT]) { 1192 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1193 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1194 return -EINVAL; 1195 } 1196 *port = nla_get_be16(tb[NDA_PORT]); 1197 } else { 1198 *port = vxlan->cfg.dst_port; 1199 } 1200 1201 if (tb[NDA_VNI]) { 1202 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1203 NL_SET_ERR_MSG(extack, "Invalid vni"); 1204 return -EINVAL; 1205 } 1206 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1207 } else { 1208 *vni = vxlan->default_dst.remote_vni; 1209 } 1210 1211 if (tb[NDA_SRC_VNI]) { 1212 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1213 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1214 return -EINVAL; 1215 } 1216 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1217 } else { 1218 *src_vni = vxlan->default_dst.remote_vni; 1219 } 1220 1221 if (tb[NDA_IFINDEX]) { 1222 struct net_device *tdev; 1223 1224 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1225 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1226 return -EINVAL; 1227 } 1228 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1229 tdev = __dev_get_by_index(net, *ifindex); 1230 if (!tdev) { 1231 NL_SET_ERR_MSG(extack, "Device not found"); 1232 return -EADDRNOTAVAIL; 1233 } 1234 } else { 1235 *ifindex = 0; 1236 } 1237 1238 *nhid = nla_get_u32_default(tb[NDA_NH_ID], 0); 1239 1240 return 0; 1241 } 1242 1243 /* Add static entry (via netlink) */ 1244 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1245 struct net_device *dev, 1246 const unsigned char *addr, u16 vid, u16 flags, 1247 bool *notified, struct netlink_ext_ack *extack) 1248 { 1249 struct vxlan_dev *vxlan = netdev_priv(dev); 1250 /* struct net *net = dev_net(vxlan->dev); */ 1251 union vxlan_addr ip; 1252 __be16 port; 1253 __be32 src_vni, vni; 1254 u32 ifindex, nhid; 1255 u32 hash_index; 1256 int err; 1257 1258 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1259 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1260 ndm->ndm_state); 1261 return -EINVAL; 1262 } 1263 1264 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1265 return -EINVAL; 1266 1267 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1268 &nhid, extack); 1269 if (err) 1270 return err; 1271 1272 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1273 return -EAFNOSUPPORT; 1274 1275 hash_index = fdb_head_index(vxlan, addr, src_vni); 1276 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1277 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1278 port, src_vni, vni, ifindex, 1279 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1280 nhid, true, extack); 1281 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1282 1283 if (!err) 1284 *notified = true; 1285 1286 return err; 1287 } 1288 1289 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1290 const unsigned char *addr, union vxlan_addr ip, 1291 __be16 port, __be32 src_vni, __be32 vni, 1292 u32 ifindex, bool swdev_notify) 1293 { 1294 struct vxlan_rdst *rd = NULL; 1295 struct vxlan_fdb *f; 1296 int err = -ENOENT; 1297 1298 f = __vxlan_find_mac(vxlan, addr, src_vni); 1299 if (!f) 1300 return err; 1301 1302 if (!vxlan_addr_any(&ip)) { 1303 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1304 if (!rd) 1305 goto out; 1306 } 1307 1308 /* remove a destination if it's not the only one on the list, 1309 * otherwise destroy the fdb entry 1310 */ 1311 if (rd && !list_is_singular(&f->remotes)) { 1312 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1313 goto out; 1314 } 1315 1316 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1317 1318 out: 1319 return 0; 1320 } 1321 1322 /* Delete entry (via netlink) */ 1323 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1324 struct net_device *dev, 1325 const unsigned char *addr, u16 vid, bool *notified, 1326 struct netlink_ext_ack *extack) 1327 { 1328 struct vxlan_dev *vxlan = netdev_priv(dev); 1329 union vxlan_addr ip; 1330 __be32 src_vni, vni; 1331 u32 ifindex, nhid; 1332 u32 hash_index; 1333 __be16 port; 1334 int err; 1335 1336 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1337 &nhid, extack); 1338 if (err) 1339 return err; 1340 1341 hash_index = fdb_head_index(vxlan, addr, src_vni); 1342 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1343 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1344 true); 1345 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1346 1347 if (!err) 1348 *notified = true; 1349 1350 return err; 1351 } 1352 1353 /* Dump forwarding table */ 1354 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1355 struct net_device *dev, 1356 struct net_device *filter_dev, int *idx) 1357 { 1358 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx; 1359 struct vxlan_dev *vxlan = netdev_priv(dev); 1360 unsigned int h; 1361 int err = 0; 1362 1363 for (h = 0; h < FDB_HASH_SIZE; ++h) { 1364 struct vxlan_fdb *f; 1365 1366 rcu_read_lock(); 1367 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { 1368 struct vxlan_rdst *rd; 1369 1370 if (rcu_access_pointer(f->nh)) { 1371 if (*idx < ctx->fdb_idx) 1372 goto skip_nh; 1373 err = vxlan_fdb_info(skb, vxlan, f, 1374 NETLINK_CB(cb->skb).portid, 1375 cb->nlh->nlmsg_seq, 1376 RTM_NEWNEIGH, 1377 NLM_F_MULTI, NULL); 1378 if (err < 0) { 1379 rcu_read_unlock(); 1380 goto out; 1381 } 1382 skip_nh: 1383 *idx += 1; 1384 continue; 1385 } 1386 1387 list_for_each_entry_rcu(rd, &f->remotes, list) { 1388 if (*idx < ctx->fdb_idx) 1389 goto skip; 1390 1391 err = vxlan_fdb_info(skb, vxlan, f, 1392 NETLINK_CB(cb->skb).portid, 1393 cb->nlh->nlmsg_seq, 1394 RTM_NEWNEIGH, 1395 NLM_F_MULTI, rd); 1396 if (err < 0) { 1397 rcu_read_unlock(); 1398 goto out; 1399 } 1400 skip: 1401 *idx += 1; 1402 } 1403 } 1404 rcu_read_unlock(); 1405 } 1406 out: 1407 return err; 1408 } 1409 1410 static int vxlan_fdb_get(struct sk_buff *skb, 1411 struct nlattr *tb[], 1412 struct net_device *dev, 1413 const unsigned char *addr, 1414 u16 vid, u32 portid, u32 seq, 1415 struct netlink_ext_ack *extack) 1416 { 1417 struct vxlan_dev *vxlan = netdev_priv(dev); 1418 struct vxlan_fdb *f; 1419 __be32 vni; 1420 int err; 1421 1422 if (tb[NDA_VNI]) 1423 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1424 else 1425 vni = vxlan->default_dst.remote_vni; 1426 1427 rcu_read_lock(); 1428 1429 f = __vxlan_find_mac(vxlan, addr, vni); 1430 if (!f) { 1431 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1432 err = -ENOENT; 1433 goto errout; 1434 } 1435 1436 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1437 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1438 errout: 1439 rcu_read_unlock(); 1440 return err; 1441 } 1442 1443 /* Watch incoming packets to learn mapping between Ethernet address 1444 * and Tunnel endpoint. 1445 */ 1446 static enum skb_drop_reason vxlan_snoop(struct net_device *dev, 1447 union vxlan_addr *src_ip, 1448 const u8 *src_mac, u32 src_ifindex, 1449 __be32 vni) 1450 { 1451 struct vxlan_dev *vxlan = netdev_priv(dev); 1452 struct vxlan_fdb *f; 1453 u32 ifindex = 0; 1454 1455 /* Ignore packets from invalid src-address */ 1456 if (!is_valid_ether_addr(src_mac)) 1457 return SKB_DROP_REASON_MAC_INVALID_SOURCE; 1458 1459 #if IS_ENABLED(CONFIG_IPV6) 1460 if (src_ip->sa.sa_family == AF_INET6 && 1461 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1462 ifindex = src_ifindex; 1463 #endif 1464 1465 f = __vxlan_find_mac(vxlan, src_mac, vni); 1466 if (likely(f)) { 1467 struct vxlan_rdst *rdst = first_remote_rcu(f); 1468 unsigned long now = jiffies; 1469 1470 if (READ_ONCE(f->updated) != now) 1471 WRITE_ONCE(f->updated, now); 1472 1473 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1474 rdst->remote_ifindex == ifindex)) 1475 return SKB_NOT_DROPPED_YET; 1476 1477 /* Don't migrate static entries, drop packets */ 1478 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1479 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1480 1481 /* Don't override an fdb with nexthop with a learnt entry */ 1482 if (rcu_access_pointer(f->nh)) 1483 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1484 1485 if (net_ratelimit()) 1486 netdev_info(dev, 1487 "%pM migrated from %pIS to %pIS\n", 1488 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1489 1490 rdst->remote_ip = *src_ip; 1491 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1492 } else { 1493 u32 hash_index = fdb_head_index(vxlan, src_mac, vni); 1494 1495 /* learned new entry */ 1496 spin_lock(&vxlan->hash_lock[hash_index]); 1497 1498 /* close off race between vxlan_flush and incoming packets */ 1499 if (netif_running(dev)) 1500 vxlan_fdb_update(vxlan, src_mac, src_ip, 1501 NUD_REACHABLE, 1502 NLM_F_EXCL|NLM_F_CREATE, 1503 vxlan->cfg.dst_port, 1504 vni, 1505 vxlan->default_dst.remote_vni, 1506 ifindex, NTF_SELF, 0, true, NULL); 1507 spin_unlock(&vxlan->hash_lock[hash_index]); 1508 } 1509 1510 return SKB_NOT_DROPPED_YET; 1511 } 1512 1513 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1514 { 1515 struct vxlan_net *vn; 1516 1517 if (!vs) 1518 return false; 1519 if (!refcount_dec_and_test(&vs->refcnt)) 1520 return false; 1521 1522 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); 1523 spin_lock(&vn->sock_lock); 1524 hlist_del_rcu(&vs->hlist); 1525 udp_tunnel_notify_del_rx_port(vs->sock, 1526 (vs->flags & VXLAN_F_GPE) ? 1527 UDP_TUNNEL_TYPE_VXLAN_GPE : 1528 UDP_TUNNEL_TYPE_VXLAN); 1529 spin_unlock(&vn->sock_lock); 1530 1531 return true; 1532 } 1533 1534 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1535 { 1536 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1537 #if IS_ENABLED(CONFIG_IPV6) 1538 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1539 1540 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1541 #endif 1542 1543 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1544 synchronize_net(); 1545 1546 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1547 vxlan_vs_del_vnigrp(vxlan); 1548 else 1549 vxlan_vs_del_dev(vxlan); 1550 1551 if (__vxlan_sock_release_prep(sock4)) { 1552 udp_tunnel_sock_release(sock4->sock); 1553 kfree(sock4); 1554 } 1555 1556 #if IS_ENABLED(CONFIG_IPV6) 1557 if (__vxlan_sock_release_prep(sock6)) { 1558 udp_tunnel_sock_release(sock6->sock); 1559 kfree(sock6); 1560 } 1561 #endif 1562 } 1563 1564 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags) 1565 { 1566 const struct vxlanhdr *vh = vxlan_hdr(skb); 1567 enum skb_drop_reason reason; 1568 size_t start, offset; 1569 1570 if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1571 return SKB_NOT_DROPPED_YET; 1572 1573 start = vxlan_rco_start(vh->vx_vni); 1574 offset = start + vxlan_rco_offset(vh->vx_vni); 1575 1576 reason = pskb_may_pull_reason(skb, offset + sizeof(u16)); 1577 if (reason) 1578 return reason; 1579 1580 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1581 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1582 return SKB_NOT_DROPPED_YET; 1583 } 1584 1585 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags, 1586 struct vxlan_metadata *md) 1587 { 1588 const struct vxlanhdr *vh = vxlan_hdr(skb); 1589 const struct vxlanhdr_gbp *gbp; 1590 struct metadata_dst *tun_dst; 1591 1592 gbp = (const struct vxlanhdr_gbp *)vh; 1593 1594 if (!(vh->vx_flags & VXLAN_HF_GBP)) 1595 return; 1596 1597 md->gbp = ntohs(gbp->policy_id); 1598 1599 tun_dst = (struct metadata_dst *)skb_dst(skb); 1600 if (tun_dst) { 1601 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1602 tun_dst->u.tun_info.key.tun_flags); 1603 tun_dst->u.tun_info.options_len = sizeof(*md); 1604 } 1605 if (gbp->dont_learn) 1606 md->gbp |= VXLAN_GBP_DONT_LEARN; 1607 1608 if (gbp->policy_applied) 1609 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1610 1611 /* In flow-based mode, GBP is carried in dst_metadata */ 1612 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1613 skb->mark = md->gbp; 1614 } 1615 1616 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan, 1617 struct vxlan_sock *vs, 1618 struct sk_buff *skb, __be32 vni) 1619 { 1620 union vxlan_addr saddr; 1621 u32 ifindex = skb->dev->ifindex; 1622 1623 skb_reset_mac_header(skb); 1624 skb->protocol = eth_type_trans(skb, vxlan->dev); 1625 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1626 1627 /* Ignore packet loops (and multicast echo) */ 1628 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1629 return SKB_DROP_REASON_LOCAL_MAC; 1630 1631 /* Get address from the outer IP header */ 1632 if (vxlan_get_sk_family(vs) == AF_INET) { 1633 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1634 saddr.sa.sa_family = AF_INET; 1635 #if IS_ENABLED(CONFIG_IPV6) 1636 } else { 1637 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1638 saddr.sa.sa_family = AF_INET6; 1639 #endif 1640 } 1641 1642 if (!(vxlan->cfg.flags & VXLAN_F_LEARN)) 1643 return SKB_NOT_DROPPED_YET; 1644 1645 return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, 1646 ifindex, vni); 1647 } 1648 1649 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1650 struct sk_buff *skb) 1651 { 1652 int err = 0; 1653 1654 if (vxlan_get_sk_family(vs) == AF_INET) 1655 err = IP_ECN_decapsulate(oiph, skb); 1656 #if IS_ENABLED(CONFIG_IPV6) 1657 else 1658 err = IP6_ECN_decapsulate(oiph, skb); 1659 #endif 1660 1661 if (unlikely(err) && log_ecn_error) { 1662 if (vxlan_get_sk_family(vs) == AF_INET) 1663 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1664 &((struct iphdr *)oiph)->saddr, 1665 ((struct iphdr *)oiph)->tos); 1666 else 1667 net_info_ratelimited("non-ECT from %pI6\n", 1668 &((struct ipv6hdr *)oiph)->saddr); 1669 } 1670 return err <= 1; 1671 } 1672 1673 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1674 { 1675 struct vxlan_vni_node *vninode = NULL; 1676 const struct vxlanhdr *vh; 1677 struct vxlan_dev *vxlan; 1678 struct vxlan_sock *vs; 1679 struct vxlan_metadata _md; 1680 struct vxlan_metadata *md = &_md; 1681 __be16 protocol = htons(ETH_P_TEB); 1682 enum skb_drop_reason reason; 1683 bool raw_proto = false; 1684 void *oiph; 1685 __be32 vni = 0; 1686 int nh; 1687 1688 /* Need UDP and VXLAN header to be present */ 1689 reason = pskb_may_pull_reason(skb, VXLAN_HLEN); 1690 if (reason) 1691 goto drop; 1692 1693 vh = vxlan_hdr(skb); 1694 /* VNI flag always required to be set */ 1695 if (!(vh->vx_flags & VXLAN_HF_VNI)) { 1696 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1697 ntohl(vh->vx_flags), ntohl(vh->vx_vni)); 1698 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1699 /* Return non vxlan pkt */ 1700 goto drop; 1701 } 1702 1703 vs = rcu_dereference_sk_user_data(sk); 1704 if (!vs) 1705 goto drop; 1706 1707 vni = vxlan_vni(vh->vx_vni); 1708 1709 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1710 if (!vxlan) { 1711 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND; 1712 goto drop; 1713 } 1714 1715 if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags || 1716 vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) { 1717 /* If the header uses bits besides those enabled by the 1718 * netdevice configuration, treat this as a malformed packet. 1719 * This behavior diverges from VXLAN RFC (RFC7348) which 1720 * stipulates that bits in reserved in reserved fields are to be 1721 * ignored. The approach here maintains compatibility with 1722 * previous stack code, and also is more robust and provides a 1723 * little more security in adding extensions to VXLAN. 1724 */ 1725 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1726 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1727 DEV_STATS_INC(vxlan->dev, rx_errors); 1728 vxlan_vnifilter_count(vxlan, vni, vninode, 1729 VXLAN_VNI_STATS_RX_ERRORS, 0); 1730 goto drop; 1731 } 1732 1733 if (vxlan->cfg.flags & VXLAN_F_GPE) { 1734 if (!vxlan_parse_gpe_proto(vh, &protocol)) 1735 goto drop; 1736 raw_proto = true; 1737 } 1738 1739 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1740 !net_eq(vxlan->net, dev_net(vxlan->dev)))) { 1741 reason = SKB_DROP_REASON_NOMEM; 1742 goto drop; 1743 } 1744 1745 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) { 1746 reason = vxlan_remcsum(skb, vxlan->cfg.flags); 1747 if (unlikely(reason)) 1748 goto drop; 1749 } 1750 1751 if (vxlan_collect_metadata(vs)) { 1752 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1753 struct metadata_dst *tun_dst; 1754 1755 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1756 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1757 key32_to_tunnel_id(vni), sizeof(*md)); 1758 1759 if (!tun_dst) { 1760 reason = SKB_DROP_REASON_NOMEM; 1761 goto drop; 1762 } 1763 1764 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1765 1766 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1767 } else { 1768 memset(md, 0, sizeof(*md)); 1769 } 1770 1771 if (vxlan->cfg.flags & VXLAN_F_GBP) 1772 vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md); 1773 /* Note that GBP and GPE can never be active together. This is 1774 * ensured in vxlan_dev_configure. 1775 */ 1776 1777 if (!raw_proto) { 1778 reason = vxlan_set_mac(vxlan, vs, skb, vni); 1779 if (reason) 1780 goto drop; 1781 } else { 1782 skb_reset_mac_header(skb); 1783 skb->dev = vxlan->dev; 1784 skb->pkt_type = PACKET_HOST; 1785 } 1786 1787 /* Save offset of outer header relative to skb->head, 1788 * because we are going to reset the network header to the inner header 1789 * and might change skb->head. 1790 */ 1791 nh = skb_network_header(skb) - skb->head; 1792 1793 skb_reset_network_header(skb); 1794 1795 reason = pskb_inet_may_pull_reason(skb); 1796 if (reason) { 1797 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1798 DEV_STATS_INC(vxlan->dev, rx_errors); 1799 vxlan_vnifilter_count(vxlan, vni, vninode, 1800 VXLAN_VNI_STATS_RX_ERRORS, 0); 1801 goto drop; 1802 } 1803 1804 /* Get the outer header. */ 1805 oiph = skb->head + nh; 1806 1807 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1808 reason = SKB_DROP_REASON_IP_TUNNEL_ECN; 1809 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1810 DEV_STATS_INC(vxlan->dev, rx_errors); 1811 vxlan_vnifilter_count(vxlan, vni, vninode, 1812 VXLAN_VNI_STATS_RX_ERRORS, 0); 1813 goto drop; 1814 } 1815 1816 rcu_read_lock(); 1817 1818 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1819 rcu_read_unlock(); 1820 dev_dstats_rx_dropped(vxlan->dev); 1821 vxlan_vnifilter_count(vxlan, vni, vninode, 1822 VXLAN_VNI_STATS_RX_DROPS, 0); 1823 reason = SKB_DROP_REASON_DEV_READY; 1824 goto drop; 1825 } 1826 1827 dev_dstats_rx_add(vxlan->dev, skb->len); 1828 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1829 gro_cells_receive(&vxlan->gro_cells, skb); 1830 1831 rcu_read_unlock(); 1832 1833 return 0; 1834 1835 drop: 1836 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED; 1837 /* Consume bad packet */ 1838 kfree_skb_reason(skb, reason); 1839 return 0; 1840 } 1841 1842 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1843 { 1844 struct vxlan_dev *vxlan; 1845 struct vxlan_sock *vs; 1846 struct vxlanhdr *hdr; 1847 __be32 vni; 1848 1849 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1850 return -EINVAL; 1851 1852 hdr = vxlan_hdr(skb); 1853 1854 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1855 return -EINVAL; 1856 1857 vs = rcu_dereference_sk_user_data(sk); 1858 if (!vs) 1859 return -ENOENT; 1860 1861 vni = vxlan_vni(hdr->vx_vni); 1862 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1863 if (!vxlan) 1864 return -ENOENT; 1865 1866 return 0; 1867 } 1868 1869 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1870 { 1871 struct vxlan_dev *vxlan = netdev_priv(dev); 1872 struct arphdr *parp; 1873 u8 *arpptr, *sha; 1874 __be32 sip, tip; 1875 struct neighbour *n; 1876 1877 if (dev->flags & IFF_NOARP) 1878 goto out; 1879 1880 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1881 dev_dstats_tx_dropped(dev); 1882 vxlan_vnifilter_count(vxlan, vni, NULL, 1883 VXLAN_VNI_STATS_TX_DROPS, 0); 1884 goto out; 1885 } 1886 parp = arp_hdr(skb); 1887 1888 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1889 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1890 parp->ar_pro != htons(ETH_P_IP) || 1891 parp->ar_op != htons(ARPOP_REQUEST) || 1892 parp->ar_hln != dev->addr_len || 1893 parp->ar_pln != 4) 1894 goto out; 1895 arpptr = (u8 *)parp + sizeof(struct arphdr); 1896 sha = arpptr; 1897 arpptr += dev->addr_len; /* sha */ 1898 memcpy(&sip, arpptr, sizeof(sip)); 1899 arpptr += sizeof(sip); 1900 arpptr += dev->addr_len; /* tha */ 1901 memcpy(&tip, arpptr, sizeof(tip)); 1902 1903 if (ipv4_is_loopback(tip) || 1904 ipv4_is_multicast(tip)) 1905 goto out; 1906 1907 n = neigh_lookup(&arp_tbl, &tip, dev); 1908 1909 if (n) { 1910 struct vxlan_fdb *f; 1911 struct sk_buff *reply; 1912 1913 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1914 neigh_release(n); 1915 goto out; 1916 } 1917 1918 f = vxlan_find_mac(vxlan, n->ha, vni); 1919 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1920 /* bridge-local neighbor */ 1921 neigh_release(n); 1922 goto out; 1923 } 1924 1925 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1926 n->ha, sha); 1927 1928 neigh_release(n); 1929 1930 if (reply == NULL) 1931 goto out; 1932 1933 skb_reset_mac_header(reply); 1934 __skb_pull(reply, skb_network_offset(reply)); 1935 reply->ip_summed = CHECKSUM_UNNECESSARY; 1936 reply->pkt_type = PACKET_HOST; 1937 1938 if (netif_rx(reply) == NET_RX_DROP) { 1939 dev_dstats_rx_dropped(dev); 1940 vxlan_vnifilter_count(vxlan, vni, NULL, 1941 VXLAN_VNI_STATS_RX_DROPS, 0); 1942 } 1943 1944 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1945 union vxlan_addr ipa = { 1946 .sin.sin_addr.s_addr = tip, 1947 .sin.sin_family = AF_INET, 1948 }; 1949 1950 vxlan_ip_miss(dev, &ipa); 1951 } 1952 out: 1953 consume_skb(skb); 1954 return NETDEV_TX_OK; 1955 } 1956 1957 #if IS_ENABLED(CONFIG_IPV6) 1958 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1959 struct neighbour *n, bool isrouter) 1960 { 1961 struct net_device *dev = request->dev; 1962 struct sk_buff *reply; 1963 struct nd_msg *ns, *na; 1964 struct ipv6hdr *pip6; 1965 u8 *daddr; 1966 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1967 int ns_olen; 1968 int i, len; 1969 1970 if (dev == NULL || !pskb_may_pull(request, request->len)) 1971 return NULL; 1972 1973 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1974 sizeof(*na) + na_olen + dev->needed_tailroom; 1975 reply = alloc_skb(len, GFP_ATOMIC); 1976 if (reply == NULL) 1977 return NULL; 1978 1979 reply->protocol = htons(ETH_P_IPV6); 1980 reply->dev = dev; 1981 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1982 skb_push(reply, sizeof(struct ethhdr)); 1983 skb_reset_mac_header(reply); 1984 1985 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1986 1987 daddr = eth_hdr(request)->h_source; 1988 ns_olen = request->len - skb_network_offset(request) - 1989 sizeof(struct ipv6hdr) - sizeof(*ns); 1990 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1991 if (!ns->opt[i + 1]) { 1992 kfree_skb(reply); 1993 return NULL; 1994 } 1995 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1996 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1997 break; 1998 } 1999 } 2000 2001 /* Ethernet header */ 2002 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 2003 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 2004 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 2005 reply->protocol = htons(ETH_P_IPV6); 2006 2007 skb_pull(reply, sizeof(struct ethhdr)); 2008 skb_reset_network_header(reply); 2009 skb_put(reply, sizeof(struct ipv6hdr)); 2010 2011 /* IPv6 header */ 2012 2013 pip6 = ipv6_hdr(reply); 2014 memset(pip6, 0, sizeof(struct ipv6hdr)); 2015 pip6->version = 6; 2016 pip6->priority = ipv6_hdr(request)->priority; 2017 pip6->nexthdr = IPPROTO_ICMPV6; 2018 pip6->hop_limit = 255; 2019 pip6->daddr = ipv6_hdr(request)->saddr; 2020 pip6->saddr = *(struct in6_addr *)n->primary_key; 2021 2022 skb_pull(reply, sizeof(struct ipv6hdr)); 2023 skb_reset_transport_header(reply); 2024 2025 /* Neighbor Advertisement */ 2026 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2027 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2028 na->icmph.icmp6_router = isrouter; 2029 na->icmph.icmp6_override = 1; 2030 na->icmph.icmp6_solicited = 1; 2031 na->target = ns->target; 2032 ether_addr_copy(&na->opt[2], n->ha); 2033 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2034 na->opt[1] = na_olen >> 3; 2035 2036 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2037 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2038 csum_partial(na, sizeof(*na)+na_olen, 0)); 2039 2040 pip6->payload_len = htons(sizeof(*na)+na_olen); 2041 2042 skb_push(reply, sizeof(struct ipv6hdr)); 2043 2044 reply->ip_summed = CHECKSUM_UNNECESSARY; 2045 2046 return reply; 2047 } 2048 2049 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2050 { 2051 struct vxlan_dev *vxlan = netdev_priv(dev); 2052 const struct in6_addr *daddr; 2053 const struct ipv6hdr *iphdr; 2054 struct inet6_dev *in6_dev; 2055 struct neighbour *n; 2056 struct nd_msg *msg; 2057 2058 rcu_read_lock(); 2059 in6_dev = __in6_dev_get(dev); 2060 if (!in6_dev) 2061 goto out; 2062 2063 iphdr = ipv6_hdr(skb); 2064 daddr = &iphdr->daddr; 2065 msg = (struct nd_msg *)(iphdr + 1); 2066 2067 if (ipv6_addr_loopback(daddr) || 2068 ipv6_addr_is_multicast(&msg->target)) 2069 goto out; 2070 2071 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2072 2073 if (n) { 2074 struct vxlan_fdb *f; 2075 struct sk_buff *reply; 2076 2077 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2078 neigh_release(n); 2079 goto out; 2080 } 2081 2082 f = vxlan_find_mac(vxlan, n->ha, vni); 2083 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 2084 /* bridge-local neighbor */ 2085 neigh_release(n); 2086 goto out; 2087 } 2088 2089 reply = vxlan_na_create(skb, n, 2090 !!(f ? f->flags & NTF_ROUTER : 0)); 2091 2092 neigh_release(n); 2093 2094 if (reply == NULL) 2095 goto out; 2096 2097 if (netif_rx(reply) == NET_RX_DROP) { 2098 dev_dstats_rx_dropped(dev); 2099 vxlan_vnifilter_count(vxlan, vni, NULL, 2100 VXLAN_VNI_STATS_RX_DROPS, 0); 2101 } 2102 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2103 union vxlan_addr ipa = { 2104 .sin6.sin6_addr = msg->target, 2105 .sin6.sin6_family = AF_INET6, 2106 }; 2107 2108 vxlan_ip_miss(dev, &ipa); 2109 } 2110 2111 out: 2112 rcu_read_unlock(); 2113 consume_skb(skb); 2114 return NETDEV_TX_OK; 2115 } 2116 #endif 2117 2118 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2119 { 2120 struct vxlan_dev *vxlan = netdev_priv(dev); 2121 struct neighbour *n; 2122 2123 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2124 return false; 2125 2126 n = NULL; 2127 switch (ntohs(eth_hdr(skb)->h_proto)) { 2128 case ETH_P_IP: 2129 { 2130 struct iphdr *pip; 2131 2132 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2133 return false; 2134 pip = ip_hdr(skb); 2135 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2136 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2137 union vxlan_addr ipa = { 2138 .sin.sin_addr.s_addr = pip->daddr, 2139 .sin.sin_family = AF_INET, 2140 }; 2141 2142 vxlan_ip_miss(dev, &ipa); 2143 return false; 2144 } 2145 2146 break; 2147 } 2148 #if IS_ENABLED(CONFIG_IPV6) 2149 case ETH_P_IPV6: 2150 { 2151 struct ipv6hdr *pip6; 2152 2153 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2154 return false; 2155 pip6 = ipv6_hdr(skb); 2156 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2157 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2158 union vxlan_addr ipa = { 2159 .sin6.sin6_addr = pip6->daddr, 2160 .sin6.sin6_family = AF_INET6, 2161 }; 2162 2163 vxlan_ip_miss(dev, &ipa); 2164 return false; 2165 } 2166 2167 break; 2168 } 2169 #endif 2170 default: 2171 return false; 2172 } 2173 2174 if (n) { 2175 bool diff; 2176 2177 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2178 if (diff) { 2179 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2180 dev->addr_len); 2181 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2182 } 2183 neigh_release(n); 2184 return diff; 2185 } 2186 2187 return false; 2188 } 2189 2190 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2191 { 2192 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2193 2194 gpe->np_applied = 1; 2195 gpe->next_protocol = tun_p_from_eth_p(protocol); 2196 if (!gpe->next_protocol) 2197 return -EPFNOSUPPORT; 2198 return 0; 2199 } 2200 2201 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2202 int iphdr_len, __be32 vni, 2203 struct vxlan_metadata *md, u32 vxflags, 2204 bool udp_sum) 2205 { 2206 struct vxlanhdr *vxh; 2207 int min_headroom; 2208 int err; 2209 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2210 __be16 inner_protocol = htons(ETH_P_TEB); 2211 2212 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2213 skb->ip_summed == CHECKSUM_PARTIAL) { 2214 int csum_start = skb_checksum_start_offset(skb); 2215 2216 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2217 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2218 (skb->csum_offset == offsetof(struct udphdr, check) || 2219 skb->csum_offset == offsetof(struct tcphdr, check))) 2220 type |= SKB_GSO_TUNNEL_REMCSUM; 2221 } 2222 2223 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2224 + VXLAN_HLEN + iphdr_len; 2225 2226 /* Need space for new headers (invalidates iph ptr) */ 2227 err = skb_cow_head(skb, min_headroom); 2228 if (unlikely(err)) 2229 return err; 2230 2231 err = iptunnel_handle_offloads(skb, type); 2232 if (err) 2233 return err; 2234 2235 vxh = __skb_push(skb, sizeof(*vxh)); 2236 vxh->vx_flags = VXLAN_HF_VNI; 2237 vxh->vx_vni = vxlan_vni_field(vni); 2238 2239 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2240 unsigned int start; 2241 2242 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2243 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2244 vxh->vx_flags |= VXLAN_HF_RCO; 2245 2246 if (!skb_is_gso(skb)) { 2247 skb->ip_summed = CHECKSUM_NONE; 2248 skb->encapsulation = 0; 2249 } 2250 } 2251 2252 if (vxflags & VXLAN_F_GBP) 2253 vxlan_build_gbp_hdr(vxh, md); 2254 if (vxflags & VXLAN_F_GPE) { 2255 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2256 if (err < 0) 2257 return err; 2258 inner_protocol = skb->protocol; 2259 } 2260 2261 skb_set_inner_protocol(skb, inner_protocol); 2262 return 0; 2263 } 2264 2265 /* Bypass encapsulation if the destination is local */ 2266 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2267 struct vxlan_dev *dst_vxlan, __be32 vni, 2268 bool snoop) 2269 { 2270 union vxlan_addr loopback; 2271 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2272 unsigned int len = skb->len; 2273 struct net_device *dev; 2274 2275 skb->pkt_type = PACKET_HOST; 2276 skb->encapsulation = 0; 2277 skb->dev = dst_vxlan->dev; 2278 __skb_pull(skb, skb_network_offset(skb)); 2279 2280 if (remote_ip->sa.sa_family == AF_INET) { 2281 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2282 loopback.sa.sa_family = AF_INET; 2283 #if IS_ENABLED(CONFIG_IPV6) 2284 } else { 2285 loopback.sin6.sin6_addr = in6addr_loopback; 2286 loopback.sa.sa_family = AF_INET6; 2287 #endif 2288 } 2289 2290 rcu_read_lock(); 2291 dev = skb->dev; 2292 if (unlikely(!(dev->flags & IFF_UP))) { 2293 kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY); 2294 goto drop; 2295 } 2296 2297 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2298 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2299 2300 dev_dstats_tx_add(src_vxlan->dev, len); 2301 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2302 2303 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2304 dev_dstats_rx_add(dst_vxlan->dev, len); 2305 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2306 len); 2307 } else { 2308 drop: 2309 dev_dstats_rx_dropped(dev); 2310 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2311 VXLAN_VNI_STATS_RX_DROPS, 0); 2312 } 2313 rcu_read_unlock(); 2314 } 2315 2316 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2317 struct vxlan_dev *vxlan, 2318 int addr_family, 2319 __be16 dst_port, int dst_ifindex, __be32 vni, 2320 struct dst_entry *dst, 2321 u32 rt_flags) 2322 { 2323 #if IS_ENABLED(CONFIG_IPV6) 2324 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2325 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2326 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2327 */ 2328 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2329 #endif 2330 /* Bypass encapsulation if the destination is local */ 2331 if (rt_flags & RTCF_LOCAL && 2332 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2333 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2334 struct vxlan_dev *dst_vxlan; 2335 2336 dst_release(dst); 2337 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2338 addr_family, dst_port, 2339 vxlan->cfg.flags); 2340 if (!dst_vxlan) { 2341 DEV_STATS_INC(dev, tx_errors); 2342 vxlan_vnifilter_count(vxlan, vni, NULL, 2343 VXLAN_VNI_STATS_TX_ERRORS, 0); 2344 kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND); 2345 2346 return -ENOENT; 2347 } 2348 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2349 return 1; 2350 } 2351 2352 return 0; 2353 } 2354 2355 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2356 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2357 { 2358 struct dst_cache *dst_cache; 2359 struct ip_tunnel_info *info; 2360 struct ip_tunnel_key *pkey; 2361 struct ip_tunnel_key key; 2362 struct vxlan_dev *vxlan = netdev_priv(dev); 2363 const struct iphdr *old_iph; 2364 struct vxlan_metadata _md; 2365 struct vxlan_metadata *md = &_md; 2366 unsigned int pkt_len = skb->len; 2367 __be16 src_port = 0, dst_port; 2368 struct dst_entry *ndst = NULL; 2369 int addr_family; 2370 __u8 tos, ttl; 2371 int ifindex; 2372 int err; 2373 u32 flags = vxlan->cfg.flags; 2374 bool use_cache; 2375 bool udp_sum = false; 2376 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2377 enum skb_drop_reason reason; 2378 bool no_eth_encap; 2379 __be32 vni = 0; 2380 2381 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2382 reason = skb_vlan_inet_prepare(skb, no_eth_encap); 2383 if (reason) 2384 goto drop; 2385 2386 reason = SKB_DROP_REASON_NOT_SPECIFIED; 2387 old_iph = ip_hdr(skb); 2388 2389 info = skb_tunnel_info(skb); 2390 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2391 2392 if (rdst) { 2393 memset(&key, 0, sizeof(key)); 2394 pkey = &key; 2395 2396 if (vxlan_addr_any(&rdst->remote_ip)) { 2397 if (did_rsc) { 2398 /* short-circuited back to local bridge */ 2399 vxlan_encap_bypass(skb, vxlan, vxlan, 2400 default_vni, true); 2401 return; 2402 } 2403 goto drop; 2404 } 2405 2406 addr_family = vxlan->cfg.saddr.sa.sa_family; 2407 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2408 vni = (rdst->remote_vni) ? : default_vni; 2409 ifindex = rdst->remote_ifindex; 2410 2411 if (addr_family == AF_INET) { 2412 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2413 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2414 } else { 2415 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2416 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2417 } 2418 2419 dst_cache = &rdst->dst_cache; 2420 md->gbp = skb->mark; 2421 if (flags & VXLAN_F_TTL_INHERIT) { 2422 ttl = ip_tunnel_get_ttl(old_iph, skb); 2423 } else { 2424 ttl = vxlan->cfg.ttl; 2425 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2426 ttl = 1; 2427 } 2428 tos = vxlan->cfg.tos; 2429 if (tos == 1) 2430 tos = ip_tunnel_get_dsfield(old_iph, skb); 2431 if (tos && !info) 2432 use_cache = false; 2433 2434 if (addr_family == AF_INET) 2435 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2436 else 2437 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2438 #if IS_ENABLED(CONFIG_IPV6) 2439 switch (vxlan->cfg.label_policy) { 2440 case VXLAN_LABEL_FIXED: 2441 key.label = vxlan->cfg.label; 2442 break; 2443 case VXLAN_LABEL_INHERIT: 2444 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2445 break; 2446 default: 2447 DEBUG_NET_WARN_ON_ONCE(1); 2448 goto drop; 2449 } 2450 #endif 2451 } else { 2452 if (!info) { 2453 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2454 dev->name); 2455 goto drop; 2456 } 2457 pkey = &info->key; 2458 addr_family = ip_tunnel_info_af(info); 2459 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2460 vni = tunnel_id_to_key32(info->key.tun_id); 2461 ifindex = 0; 2462 dst_cache = &info->dst_cache; 2463 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2464 if (info->options_len < sizeof(*md)) 2465 goto drop; 2466 md = ip_tunnel_info_opts(info); 2467 } 2468 ttl = info->key.ttl; 2469 tos = info->key.tos; 2470 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2471 } 2472 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2473 vxlan->cfg.port_max, true); 2474 2475 rcu_read_lock(); 2476 if (addr_family == AF_INET) { 2477 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2478 struct rtable *rt; 2479 __be16 df = 0; 2480 __be32 saddr; 2481 2482 if (!ifindex) 2483 ifindex = sock4->sock->sk->sk_bound_dev_if; 2484 2485 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2486 &saddr, pkey, src_port, dst_port, 2487 tos, use_cache ? dst_cache : NULL); 2488 if (IS_ERR(rt)) { 2489 err = PTR_ERR(rt); 2490 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2491 goto tx_error; 2492 } 2493 2494 if (!info) { 2495 /* Bypass encapsulation if the destination is local */ 2496 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2497 dst_port, ifindex, vni, 2498 &rt->dst, rt->rt_flags); 2499 if (err) 2500 goto out_unlock; 2501 2502 if (vxlan->cfg.df == VXLAN_DF_SET) { 2503 df = htons(IP_DF); 2504 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2505 struct ethhdr *eth = eth_hdr(skb); 2506 2507 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2508 (ntohs(eth->h_proto) == ETH_P_IP && 2509 old_iph->frag_off & htons(IP_DF))) 2510 df = htons(IP_DF); 2511 } 2512 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2513 info->key.tun_flags)) { 2514 df = htons(IP_DF); 2515 } 2516 2517 ndst = &rt->dst; 2518 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2519 netif_is_any_bridge_port(dev)); 2520 if (err < 0) { 2521 goto tx_error; 2522 } else if (err) { 2523 if (info) { 2524 struct ip_tunnel_info *unclone; 2525 2526 unclone = skb_tunnel_info_unclone(skb); 2527 if (unlikely(!unclone)) 2528 goto tx_error; 2529 2530 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2531 unclone->key.u.ipv4.dst = saddr; 2532 } 2533 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2534 dst_release(ndst); 2535 goto out_unlock; 2536 } 2537 2538 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2539 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2540 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2541 vni, md, flags, udp_sum); 2542 if (err < 0) { 2543 reason = SKB_DROP_REASON_NOMEM; 2544 goto tx_error; 2545 } 2546 2547 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2548 pkey->u.ipv4.dst, tos, ttl, df, 2549 src_port, dst_port, xnet, !udp_sum); 2550 #if IS_ENABLED(CONFIG_IPV6) 2551 } else { 2552 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2553 struct in6_addr saddr; 2554 2555 if (!ifindex) 2556 ifindex = sock6->sock->sk->sk_bound_dev_if; 2557 2558 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2559 ifindex, &saddr, pkey, 2560 src_port, dst_port, tos, 2561 use_cache ? dst_cache : NULL); 2562 if (IS_ERR(ndst)) { 2563 err = PTR_ERR(ndst); 2564 ndst = NULL; 2565 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2566 goto tx_error; 2567 } 2568 2569 if (!info) { 2570 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2571 2572 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2573 dst_port, ifindex, vni, 2574 ndst, rt6i_flags); 2575 if (err) 2576 goto out_unlock; 2577 } 2578 2579 err = skb_tunnel_check_pmtu(skb, ndst, 2580 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2581 netif_is_any_bridge_port(dev)); 2582 if (err < 0) { 2583 goto tx_error; 2584 } else if (err) { 2585 if (info) { 2586 struct ip_tunnel_info *unclone; 2587 2588 unclone = skb_tunnel_info_unclone(skb); 2589 if (unlikely(!unclone)) 2590 goto tx_error; 2591 2592 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2593 unclone->key.u.ipv6.dst = saddr; 2594 } 2595 2596 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2597 dst_release(ndst); 2598 goto out_unlock; 2599 } 2600 2601 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2602 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2603 skb_scrub_packet(skb, xnet); 2604 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2605 vni, md, flags, udp_sum); 2606 if (err < 0) { 2607 reason = SKB_DROP_REASON_NOMEM; 2608 goto tx_error; 2609 } 2610 2611 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2612 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2613 pkey->label, src_port, dst_port, !udp_sum); 2614 #endif 2615 } 2616 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2617 out_unlock: 2618 rcu_read_unlock(); 2619 return; 2620 2621 drop: 2622 dev_dstats_tx_dropped(dev); 2623 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2624 kfree_skb_reason(skb, reason); 2625 return; 2626 2627 tx_error: 2628 rcu_read_unlock(); 2629 if (err == -ELOOP) 2630 DEV_STATS_INC(dev, collisions); 2631 else if (err == -ENETUNREACH) 2632 DEV_STATS_INC(dev, tx_carrier_errors); 2633 dst_release(ndst); 2634 DEV_STATS_INC(dev, tx_errors); 2635 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2636 kfree_skb_reason(skb, reason); 2637 } 2638 2639 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2640 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2641 { 2642 struct vxlan_rdst nh_rdst; 2643 struct nexthop *nh; 2644 bool do_xmit; 2645 u32 hash; 2646 2647 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2648 hash = skb_get_hash(skb); 2649 2650 rcu_read_lock(); 2651 nh = rcu_dereference(f->nh); 2652 if (!nh) { 2653 rcu_read_unlock(); 2654 goto drop; 2655 } 2656 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2657 rcu_read_unlock(); 2658 2659 if (likely(do_xmit)) 2660 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2661 else 2662 goto drop; 2663 2664 return; 2665 2666 drop: 2667 dev_dstats_tx_dropped(dev); 2668 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2669 VXLAN_VNI_STATS_TX_DROPS, 0); 2670 dev_kfree_skb(skb); 2671 } 2672 2673 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2674 u32 nhid, __be32 vni) 2675 { 2676 struct vxlan_dev *vxlan = netdev_priv(dev); 2677 struct vxlan_rdst nh_rdst; 2678 struct nexthop *nh; 2679 bool do_xmit; 2680 u32 hash; 2681 2682 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2683 hash = skb_get_hash(skb); 2684 2685 rcu_read_lock(); 2686 nh = nexthop_find_by_id(dev_net(dev), nhid); 2687 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2688 rcu_read_unlock(); 2689 goto drop; 2690 } 2691 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2692 rcu_read_unlock(); 2693 2694 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2695 goto drop; 2696 2697 if (likely(do_xmit)) 2698 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2699 else 2700 goto drop; 2701 2702 return NETDEV_TX_OK; 2703 2704 drop: 2705 dev_dstats_tx_dropped(dev); 2706 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2707 VXLAN_VNI_STATS_TX_DROPS, 0); 2708 dev_kfree_skb(skb); 2709 return NETDEV_TX_OK; 2710 } 2711 2712 /* Transmit local packets over Vxlan 2713 * 2714 * Outer IP header inherits ECN and DF from inner header. 2715 * Outer UDP destination is the VXLAN assigned port. 2716 * source port is based on hash of flow 2717 */ 2718 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2719 { 2720 struct vxlan_dev *vxlan = netdev_priv(dev); 2721 struct vxlan_rdst *rdst, *fdst = NULL; 2722 const struct ip_tunnel_info *info; 2723 struct vxlan_fdb *f; 2724 struct ethhdr *eth; 2725 __be32 vni = 0; 2726 u32 nhid = 0; 2727 bool did_rsc; 2728 2729 info = skb_tunnel_info(skb); 2730 2731 skb_reset_mac_header(skb); 2732 2733 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2734 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2735 info->mode & IP_TUNNEL_INFO_TX) { 2736 vni = tunnel_id_to_key32(info->key.tun_id); 2737 nhid = info->key.nhid; 2738 } else { 2739 if (info && info->mode & IP_TUNNEL_INFO_TX) 2740 vxlan_xmit_one(skb, dev, vni, NULL, false); 2741 else 2742 kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO); 2743 return NETDEV_TX_OK; 2744 } 2745 } 2746 2747 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2748 eth = eth_hdr(skb); 2749 if (ntohs(eth->h_proto) == ETH_P_ARP) 2750 return arp_reduce(dev, skb, vni); 2751 #if IS_ENABLED(CONFIG_IPV6) 2752 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2753 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2754 sizeof(struct nd_msg)) && 2755 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2756 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2757 2758 if (m->icmph.icmp6_code == 0 && 2759 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2760 return neigh_reduce(dev, skb, vni); 2761 } 2762 #endif 2763 } 2764 2765 if (nhid) 2766 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2767 2768 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2769 struct vxlan_mdb_entry *mdb_entry; 2770 2771 rcu_read_lock(); 2772 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2773 if (mdb_entry) { 2774 netdev_tx_t ret; 2775 2776 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2777 rcu_read_unlock(); 2778 return ret; 2779 } 2780 rcu_read_unlock(); 2781 } 2782 2783 eth = eth_hdr(skb); 2784 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2785 did_rsc = false; 2786 2787 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2788 (ntohs(eth->h_proto) == ETH_P_IP || 2789 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2790 did_rsc = route_shortcircuit(dev, skb); 2791 if (did_rsc) 2792 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2793 } 2794 2795 if (f == NULL) { 2796 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2797 if (f == NULL) { 2798 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2799 !is_multicast_ether_addr(eth->h_dest)) 2800 vxlan_fdb_miss(vxlan, eth->h_dest); 2801 2802 dev_dstats_tx_dropped(dev); 2803 vxlan_vnifilter_count(vxlan, vni, NULL, 2804 VXLAN_VNI_STATS_TX_DROPS, 0); 2805 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2806 return NETDEV_TX_OK; 2807 } 2808 } 2809 2810 if (rcu_access_pointer(f->nh)) { 2811 vxlan_xmit_nh(skb, dev, f, 2812 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2813 } else { 2814 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2815 struct sk_buff *skb1; 2816 2817 if (!fdst) { 2818 fdst = rdst; 2819 continue; 2820 } 2821 skb1 = skb_clone(skb, GFP_ATOMIC); 2822 if (skb1) 2823 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2824 } 2825 if (fdst) 2826 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2827 else 2828 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2829 } 2830 2831 return NETDEV_TX_OK; 2832 } 2833 2834 /* Walk the forwarding table and purge stale entries */ 2835 static void vxlan_cleanup(struct timer_list *t) 2836 { 2837 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2838 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2839 unsigned int h; 2840 2841 if (!netif_running(vxlan->dev)) 2842 return; 2843 2844 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2845 struct hlist_node *p, *n; 2846 2847 spin_lock(&vxlan->hash_lock[h]); 2848 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2849 struct vxlan_fdb *f 2850 = container_of(p, struct vxlan_fdb, hlist); 2851 unsigned long timeout; 2852 2853 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2854 continue; 2855 2856 if (f->flags & NTF_EXT_LEARNED) 2857 continue; 2858 2859 timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ; 2860 if (time_before_eq(timeout, jiffies)) { 2861 netdev_dbg(vxlan->dev, 2862 "garbage collect %pM\n", 2863 f->eth_addr); 2864 f->state = NUD_STALE; 2865 vxlan_fdb_destroy(vxlan, f, true, true); 2866 } else if (time_before(timeout, next_timer)) 2867 next_timer = timeout; 2868 } 2869 spin_unlock(&vxlan->hash_lock[h]); 2870 } 2871 2872 mod_timer(&vxlan->age_timer, next_timer); 2873 } 2874 2875 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2876 { 2877 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2878 2879 spin_lock(&vn->sock_lock); 2880 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2881 #if IS_ENABLED(CONFIG_IPV6) 2882 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2883 #endif 2884 spin_unlock(&vn->sock_lock); 2885 } 2886 2887 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2888 struct vxlan_dev_node *node) 2889 { 2890 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2891 __be32 vni = vxlan->default_dst.remote_vni; 2892 2893 node->vxlan = vxlan; 2894 spin_lock(&vn->sock_lock); 2895 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2896 spin_unlock(&vn->sock_lock); 2897 } 2898 2899 /* Setup stats when device is created */ 2900 static int vxlan_init(struct net_device *dev) 2901 { 2902 struct vxlan_dev *vxlan = netdev_priv(dev); 2903 int err; 2904 2905 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2906 vxlan_vnigroup_init(vxlan); 2907 2908 err = gro_cells_init(&vxlan->gro_cells, dev); 2909 if (err) 2910 goto err_vnigroup_uninit; 2911 2912 err = vxlan_mdb_init(vxlan); 2913 if (err) 2914 goto err_gro_cells_destroy; 2915 2916 netdev_lockdep_set_classes(dev); 2917 return 0; 2918 2919 err_gro_cells_destroy: 2920 gro_cells_destroy(&vxlan->gro_cells); 2921 err_vnigroup_uninit: 2922 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2923 vxlan_vnigroup_uninit(vxlan); 2924 return err; 2925 } 2926 2927 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2928 { 2929 struct vxlan_fdb *f; 2930 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2931 2932 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2933 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2934 if (f) 2935 vxlan_fdb_destroy(vxlan, f, true, true); 2936 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2937 } 2938 2939 static void vxlan_uninit(struct net_device *dev) 2940 { 2941 struct vxlan_dev *vxlan = netdev_priv(dev); 2942 2943 vxlan_mdb_fini(vxlan); 2944 2945 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2946 vxlan_vnigroup_uninit(vxlan); 2947 2948 gro_cells_destroy(&vxlan->gro_cells); 2949 2950 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2951 } 2952 2953 /* Start ageing timer and join group when device is brought up */ 2954 static int vxlan_open(struct net_device *dev) 2955 { 2956 struct vxlan_dev *vxlan = netdev_priv(dev); 2957 int ret; 2958 2959 ret = vxlan_sock_add(vxlan); 2960 if (ret < 0) 2961 return ret; 2962 2963 ret = vxlan_multicast_join(vxlan); 2964 if (ret) { 2965 vxlan_sock_release(vxlan); 2966 return ret; 2967 } 2968 2969 if (vxlan->cfg.age_interval) 2970 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2971 2972 return ret; 2973 } 2974 2975 struct vxlan_fdb_flush_desc { 2976 bool ignore_default_entry; 2977 unsigned long state; 2978 unsigned long state_mask; 2979 unsigned long flags; 2980 unsigned long flags_mask; 2981 __be32 src_vni; 2982 u32 nhid; 2983 __be32 vni; 2984 __be16 port; 2985 union vxlan_addr dst_ip; 2986 }; 2987 2988 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2989 const struct vxlan_dev *vxlan) 2990 { 2991 return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni; 2992 } 2993 2994 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2995 { 2996 struct nexthop *nh = rtnl_dereference(f->nh); 2997 2998 return nh && nh->id == nhid; 2999 } 3000 3001 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 3002 const struct vxlan_dev *vxlan, 3003 const struct vxlan_fdb_flush_desc *desc) 3004 { 3005 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 3006 return false; 3007 3008 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 3009 return false; 3010 3011 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 3012 return false; 3013 3014 if (desc->src_vni && f->vni != desc->src_vni) 3015 return false; 3016 3017 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 3018 return false; 3019 3020 return true; 3021 } 3022 3023 static bool 3024 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 3025 { 3026 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 3027 } 3028 3029 static bool 3030 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3031 const struct vxlan_rdst *rd) 3032 { 3033 if (desc->vni && rd->remote_vni != desc->vni) 3034 return false; 3035 3036 if (desc->port && rd->remote_port != desc->port) 3037 return false; 3038 3039 if (desc->dst_ip.sa.sa_family && 3040 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3041 return false; 3042 3043 return true; 3044 } 3045 3046 static void 3047 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3048 const struct vxlan_fdb_flush_desc *desc, 3049 bool *p_destroy_fdb) 3050 { 3051 bool remotes_flushed = false; 3052 struct vxlan_rdst *rd, *tmp; 3053 3054 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3055 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3056 continue; 3057 3058 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3059 remotes_flushed = true; 3060 } 3061 3062 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3063 } 3064 3065 /* Purge the forwarding table */ 3066 static void vxlan_flush(struct vxlan_dev *vxlan, 3067 const struct vxlan_fdb_flush_desc *desc) 3068 { 3069 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3070 unsigned int h; 3071 3072 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3073 struct hlist_node *p, *n; 3074 3075 spin_lock_bh(&vxlan->hash_lock[h]); 3076 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 3077 struct vxlan_fdb *f 3078 = container_of(p, struct vxlan_fdb, hlist); 3079 3080 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3081 continue; 3082 3083 if (match_remotes) { 3084 bool destroy_fdb = false; 3085 3086 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3087 &destroy_fdb); 3088 3089 if (!destroy_fdb) 3090 continue; 3091 } 3092 3093 vxlan_fdb_destroy(vxlan, f, true, true); 3094 } 3095 spin_unlock_bh(&vxlan->hash_lock[h]); 3096 } 3097 } 3098 3099 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3100 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3101 [NDA_NH_ID] = { .type = NLA_U32 }, 3102 [NDA_VNI] = { .type = NLA_U32 }, 3103 [NDA_PORT] = { .type = NLA_U16 }, 3104 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3105 sizeof(struct in6_addr)), 3106 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3107 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3108 }; 3109 3110 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3111 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3112 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3113 NTF_ROUTER) 3114 3115 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3116 struct netlink_ext_ack *extack) 3117 { 3118 struct vxlan_dev *vxlan = netdev_priv(dev); 3119 struct vxlan_fdb_flush_desc desc = {}; 3120 struct ndmsg *ndm = nlmsg_data(nlh); 3121 struct nlattr *tb[NDA_MAX + 1]; 3122 u8 ndm_flags; 3123 int err; 3124 3125 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3126 3127 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3128 extack); 3129 if (err) 3130 return err; 3131 3132 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3133 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3134 return -EINVAL; 3135 } 3136 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3137 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3138 return -EINVAL; 3139 } 3140 3141 desc.state = ndm->ndm_state; 3142 desc.flags = ndm_flags; 3143 3144 if (tb[NDA_NDM_STATE_MASK]) 3145 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3146 3147 if (tb[NDA_NDM_FLAGS_MASK]) 3148 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3149 3150 if (tb[NDA_SRC_VNI]) 3151 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3152 3153 if (tb[NDA_NH_ID]) 3154 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3155 3156 if (tb[NDA_VNI]) 3157 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3158 3159 if (tb[NDA_PORT]) 3160 desc.port = nla_get_be16(tb[NDA_PORT]); 3161 3162 if (tb[NDA_DST]) { 3163 union vxlan_addr ip; 3164 3165 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3166 if (err) { 3167 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3168 "Unsupported address family"); 3169 return err; 3170 } 3171 desc.dst_ip = ip; 3172 } 3173 3174 vxlan_flush(vxlan, &desc); 3175 3176 return 0; 3177 } 3178 3179 /* Cleanup timer and forwarding table on shutdown */ 3180 static int vxlan_stop(struct net_device *dev) 3181 { 3182 struct vxlan_dev *vxlan = netdev_priv(dev); 3183 struct vxlan_fdb_flush_desc desc = { 3184 /* Default entry is deleted at vxlan_uninit. */ 3185 .ignore_default_entry = true, 3186 .state = 0, 3187 .state_mask = NUD_PERMANENT | NUD_NOARP, 3188 }; 3189 3190 vxlan_multicast_leave(vxlan); 3191 3192 del_timer_sync(&vxlan->age_timer); 3193 3194 vxlan_flush(vxlan, &desc); 3195 vxlan_sock_release(vxlan); 3196 3197 return 0; 3198 } 3199 3200 /* Stub, nothing needs to be done. */ 3201 static void vxlan_set_multicast_list(struct net_device *dev) 3202 { 3203 } 3204 3205 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3206 { 3207 struct vxlan_dev *vxlan = netdev_priv(dev); 3208 struct vxlan_rdst *dst = &vxlan->default_dst; 3209 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3210 dst->remote_ifindex); 3211 3212 /* This check is different than dev->max_mtu, because it looks at 3213 * the lowerdev->mtu, rather than the static dev->max_mtu 3214 */ 3215 if (lowerdev) { 3216 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3217 if (new_mtu > max_mtu) 3218 return -EINVAL; 3219 } 3220 3221 WRITE_ONCE(dev->mtu, new_mtu); 3222 return 0; 3223 } 3224 3225 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3226 { 3227 struct vxlan_dev *vxlan = netdev_priv(dev); 3228 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3229 __be16 sport, dport; 3230 3231 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3232 vxlan->cfg.port_max, true); 3233 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3234 3235 if (ip_tunnel_info_af(info) == AF_INET) { 3236 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3237 struct rtable *rt; 3238 3239 if (!sock4) 3240 return -EIO; 3241 3242 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3243 &info->key.u.ipv4.src, 3244 &info->key, 3245 sport, dport, info->key.tos, 3246 &info->dst_cache); 3247 if (IS_ERR(rt)) 3248 return PTR_ERR(rt); 3249 ip_rt_put(rt); 3250 } else { 3251 #if IS_ENABLED(CONFIG_IPV6) 3252 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3253 struct dst_entry *ndst; 3254 3255 if (!sock6) 3256 return -EIO; 3257 3258 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3259 0, &info->key.u.ipv6.src, 3260 &info->key, 3261 sport, dport, info->key.tos, 3262 &info->dst_cache); 3263 if (IS_ERR(ndst)) 3264 return PTR_ERR(ndst); 3265 dst_release(ndst); 3266 #else /* !CONFIG_IPV6 */ 3267 return -EPFNOSUPPORT; 3268 #endif 3269 } 3270 info->key.tp_src = sport; 3271 info->key.tp_dst = dport; 3272 return 0; 3273 } 3274 3275 static const struct net_device_ops vxlan_netdev_ether_ops = { 3276 .ndo_init = vxlan_init, 3277 .ndo_uninit = vxlan_uninit, 3278 .ndo_open = vxlan_open, 3279 .ndo_stop = vxlan_stop, 3280 .ndo_start_xmit = vxlan_xmit, 3281 .ndo_set_rx_mode = vxlan_set_multicast_list, 3282 .ndo_change_mtu = vxlan_change_mtu, 3283 .ndo_validate_addr = eth_validate_addr, 3284 .ndo_set_mac_address = eth_mac_addr, 3285 .ndo_fdb_add = vxlan_fdb_add, 3286 .ndo_fdb_del = vxlan_fdb_delete, 3287 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3288 .ndo_fdb_dump = vxlan_fdb_dump, 3289 .ndo_fdb_get = vxlan_fdb_get, 3290 .ndo_mdb_add = vxlan_mdb_add, 3291 .ndo_mdb_del = vxlan_mdb_del, 3292 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3293 .ndo_mdb_dump = vxlan_mdb_dump, 3294 .ndo_mdb_get = vxlan_mdb_get, 3295 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3296 }; 3297 3298 static const struct net_device_ops vxlan_netdev_raw_ops = { 3299 .ndo_init = vxlan_init, 3300 .ndo_uninit = vxlan_uninit, 3301 .ndo_open = vxlan_open, 3302 .ndo_stop = vxlan_stop, 3303 .ndo_start_xmit = vxlan_xmit, 3304 .ndo_change_mtu = vxlan_change_mtu, 3305 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3306 }; 3307 3308 /* Info for udev, that this is a virtual tunnel endpoint */ 3309 static const struct device_type vxlan_type = { 3310 .name = "vxlan", 3311 }; 3312 3313 /* Calls the ndo_udp_tunnel_add of the caller in order to 3314 * supply the listening VXLAN udp ports. Callers are expected 3315 * to implement the ndo_udp_tunnel_add. 3316 */ 3317 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3318 { 3319 struct vxlan_sock *vs; 3320 struct net *net = dev_net(dev); 3321 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3322 unsigned int i; 3323 3324 spin_lock(&vn->sock_lock); 3325 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3326 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3327 unsigned short type; 3328 3329 if (vs->flags & VXLAN_F_GPE) 3330 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3331 else 3332 type = UDP_TUNNEL_TYPE_VXLAN; 3333 3334 if (push) 3335 udp_tunnel_push_rx_port(dev, vs->sock, type); 3336 else 3337 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3338 } 3339 } 3340 spin_unlock(&vn->sock_lock); 3341 } 3342 3343 /* Initialize the device structure. */ 3344 static void vxlan_setup(struct net_device *dev) 3345 { 3346 struct vxlan_dev *vxlan = netdev_priv(dev); 3347 unsigned int h; 3348 3349 eth_hw_addr_random(dev); 3350 ether_setup(dev); 3351 3352 dev->needs_free_netdev = true; 3353 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3354 3355 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3356 dev->features |= NETIF_F_RXCSUM; 3357 dev->features |= NETIF_F_GSO_SOFTWARE; 3358 3359 dev->vlan_features = dev->features; 3360 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3361 dev->hw_features |= NETIF_F_RXCSUM; 3362 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3363 netif_keep_dst(dev); 3364 dev->priv_flags |= IFF_NO_QUEUE; 3365 dev->change_proto_down = true; 3366 dev->lltx = true; 3367 3368 /* MTU range: 68 - 65535 */ 3369 dev->min_mtu = ETH_MIN_MTU; 3370 dev->max_mtu = ETH_MAX_MTU; 3371 3372 dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS; 3373 INIT_LIST_HEAD(&vxlan->next); 3374 3375 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3376 3377 vxlan->dev = dev; 3378 3379 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3380 spin_lock_init(&vxlan->hash_lock[h]); 3381 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3382 } 3383 } 3384 3385 static void vxlan_ether_setup(struct net_device *dev) 3386 { 3387 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3388 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3389 dev->netdev_ops = &vxlan_netdev_ether_ops; 3390 } 3391 3392 static void vxlan_raw_setup(struct net_device *dev) 3393 { 3394 dev->header_ops = NULL; 3395 dev->type = ARPHRD_NONE; 3396 dev->hard_header_len = 0; 3397 dev->addr_len = 0; 3398 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3399 dev->netdev_ops = &vxlan_netdev_raw_ops; 3400 } 3401 3402 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3403 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3404 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3405 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3406 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3407 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3408 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3409 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3410 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3411 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3412 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3413 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3414 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3415 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3416 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3417 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3418 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3419 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3420 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3421 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3422 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3423 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3424 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3425 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3426 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3427 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3428 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3429 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3430 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3431 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3432 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3433 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3434 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3435 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3436 [IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)), 3437 }; 3438 3439 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3440 struct netlink_ext_ack *extack) 3441 { 3442 if (tb[IFLA_ADDRESS]) { 3443 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3444 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3445 "Provided link layer address is not Ethernet"); 3446 return -EINVAL; 3447 } 3448 3449 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3450 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3451 "Provided Ethernet address is not unicast"); 3452 return -EADDRNOTAVAIL; 3453 } 3454 } 3455 3456 if (tb[IFLA_MTU]) { 3457 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3458 3459 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3460 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3461 "MTU must be between 68 and 65535"); 3462 return -EINVAL; 3463 } 3464 } 3465 3466 if (!data) { 3467 NL_SET_ERR_MSG(extack, 3468 "Required attributes not provided to perform the operation"); 3469 return -EINVAL; 3470 } 3471 3472 if (data[IFLA_VXLAN_ID]) { 3473 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3474 3475 if (id >= VXLAN_N_VID) { 3476 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3477 "VXLAN ID must be lower than 16777216"); 3478 return -ERANGE; 3479 } 3480 } 3481 3482 if (data[IFLA_VXLAN_PORT_RANGE]) { 3483 const struct ifla_vxlan_port_range *p 3484 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3485 3486 if (ntohs(p->high) < ntohs(p->low)) { 3487 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3488 "Invalid source port range"); 3489 return -EINVAL; 3490 } 3491 } 3492 3493 if (data[IFLA_VXLAN_DF]) { 3494 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3495 3496 if (df < 0 || df > VXLAN_DF_MAX) { 3497 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3498 "Invalid DF attribute"); 3499 return -EINVAL; 3500 } 3501 } 3502 3503 return 0; 3504 } 3505 3506 static void vxlan_get_drvinfo(struct net_device *netdev, 3507 struct ethtool_drvinfo *drvinfo) 3508 { 3509 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3510 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3511 } 3512 3513 static int vxlan_get_link_ksettings(struct net_device *dev, 3514 struct ethtool_link_ksettings *cmd) 3515 { 3516 struct vxlan_dev *vxlan = netdev_priv(dev); 3517 struct vxlan_rdst *dst = &vxlan->default_dst; 3518 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3519 dst->remote_ifindex); 3520 3521 if (!lowerdev) { 3522 cmd->base.duplex = DUPLEX_UNKNOWN; 3523 cmd->base.port = PORT_OTHER; 3524 cmd->base.speed = SPEED_UNKNOWN; 3525 3526 return 0; 3527 } 3528 3529 return __ethtool_get_link_ksettings(lowerdev, cmd); 3530 } 3531 3532 static const struct ethtool_ops vxlan_ethtool_ops = { 3533 .get_drvinfo = vxlan_get_drvinfo, 3534 .get_link = ethtool_op_get_link, 3535 .get_link_ksettings = vxlan_get_link_ksettings, 3536 }; 3537 3538 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3539 __be16 port, u32 flags, int ifindex) 3540 { 3541 struct socket *sock; 3542 struct udp_port_cfg udp_conf; 3543 int err; 3544 3545 memset(&udp_conf, 0, sizeof(udp_conf)); 3546 3547 if (ipv6) { 3548 udp_conf.family = AF_INET6; 3549 udp_conf.use_udp6_rx_checksums = 3550 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3551 udp_conf.ipv6_v6only = 1; 3552 } else { 3553 udp_conf.family = AF_INET; 3554 } 3555 3556 udp_conf.local_udp_port = port; 3557 udp_conf.bind_ifindex = ifindex; 3558 3559 /* Open UDP socket */ 3560 err = udp_sock_create(net, &udp_conf, &sock); 3561 if (err < 0) 3562 return ERR_PTR(err); 3563 3564 udp_allow_gso(sock->sk); 3565 return sock; 3566 } 3567 3568 /* Create new listen socket if needed */ 3569 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3570 __be16 port, u32 flags, 3571 int ifindex) 3572 { 3573 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3574 struct vxlan_sock *vs; 3575 struct socket *sock; 3576 unsigned int h; 3577 struct udp_tunnel_sock_cfg tunnel_cfg; 3578 3579 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3580 if (!vs) 3581 return ERR_PTR(-ENOMEM); 3582 3583 for (h = 0; h < VNI_HASH_SIZE; ++h) 3584 INIT_HLIST_HEAD(&vs->vni_list[h]); 3585 3586 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3587 if (IS_ERR(sock)) { 3588 kfree(vs); 3589 return ERR_CAST(sock); 3590 } 3591 3592 vs->sock = sock; 3593 refcount_set(&vs->refcnt, 1); 3594 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3595 3596 spin_lock(&vn->sock_lock); 3597 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3598 udp_tunnel_notify_add_rx_port(sock, 3599 (vs->flags & VXLAN_F_GPE) ? 3600 UDP_TUNNEL_TYPE_VXLAN_GPE : 3601 UDP_TUNNEL_TYPE_VXLAN); 3602 spin_unlock(&vn->sock_lock); 3603 3604 /* Mark socket as an encapsulation socket. */ 3605 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3606 tunnel_cfg.sk_user_data = vs; 3607 tunnel_cfg.encap_type = 1; 3608 tunnel_cfg.encap_rcv = vxlan_rcv; 3609 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3610 tunnel_cfg.encap_destroy = NULL; 3611 if (vs->flags & VXLAN_F_GPE) { 3612 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3613 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3614 } else { 3615 tunnel_cfg.gro_receive = vxlan_gro_receive; 3616 tunnel_cfg.gro_complete = vxlan_gro_complete; 3617 } 3618 3619 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3620 3621 return vs; 3622 } 3623 3624 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3625 { 3626 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3627 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3628 struct vxlan_sock *vs = NULL; 3629 struct vxlan_dev_node *node; 3630 int l3mdev_index = 0; 3631 3632 if (vxlan->cfg.remote_ifindex) 3633 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3634 vxlan->net, vxlan->cfg.remote_ifindex); 3635 3636 if (!vxlan->cfg.no_share) { 3637 spin_lock(&vn->sock_lock); 3638 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3639 vxlan->cfg.dst_port, vxlan->cfg.flags, 3640 l3mdev_index); 3641 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3642 spin_unlock(&vn->sock_lock); 3643 return -EBUSY; 3644 } 3645 spin_unlock(&vn->sock_lock); 3646 } 3647 if (!vs) 3648 vs = vxlan_socket_create(vxlan->net, ipv6, 3649 vxlan->cfg.dst_port, vxlan->cfg.flags, 3650 l3mdev_index); 3651 if (IS_ERR(vs)) 3652 return PTR_ERR(vs); 3653 #if IS_ENABLED(CONFIG_IPV6) 3654 if (ipv6) { 3655 rcu_assign_pointer(vxlan->vn6_sock, vs); 3656 node = &vxlan->hlist6; 3657 } else 3658 #endif 3659 { 3660 rcu_assign_pointer(vxlan->vn4_sock, vs); 3661 node = &vxlan->hlist4; 3662 } 3663 3664 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3665 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3666 else 3667 vxlan_vs_add_dev(vs, vxlan, node); 3668 3669 return 0; 3670 } 3671 3672 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3673 { 3674 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3675 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3676 bool ipv4 = !ipv6 || metadata; 3677 int ret = 0; 3678 3679 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3680 #if IS_ENABLED(CONFIG_IPV6) 3681 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3682 if (ipv6) { 3683 ret = __vxlan_sock_add(vxlan, true); 3684 if (ret < 0 && ret != -EAFNOSUPPORT) 3685 ipv4 = false; 3686 } 3687 #endif 3688 if (ipv4) 3689 ret = __vxlan_sock_add(vxlan, false); 3690 if (ret < 0) 3691 vxlan_sock_release(vxlan); 3692 return ret; 3693 } 3694 3695 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3696 struct vxlan_config *conf, __be32 vni) 3697 { 3698 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3699 struct vxlan_dev *tmp; 3700 3701 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3702 if (tmp == vxlan) 3703 continue; 3704 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3705 if (!vxlan_vnifilter_lookup(tmp, vni)) 3706 continue; 3707 } else if (tmp->cfg.vni != vni) { 3708 continue; 3709 } 3710 if (tmp->cfg.dst_port != conf->dst_port) 3711 continue; 3712 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3713 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3714 continue; 3715 3716 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3717 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3718 continue; 3719 3720 return -EEXIST; 3721 } 3722 3723 return 0; 3724 } 3725 3726 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3727 struct net_device **lower, 3728 struct vxlan_dev *old, 3729 struct netlink_ext_ack *extack) 3730 { 3731 bool use_ipv6 = false; 3732 3733 if (conf->flags & VXLAN_F_GPE) { 3734 /* For now, allow GPE only together with 3735 * COLLECT_METADATA. This can be relaxed later; in such 3736 * case, the other side of the PtP link will have to be 3737 * provided. 3738 */ 3739 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3740 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3741 NL_SET_ERR_MSG(extack, 3742 "VXLAN GPE does not support this combination of attributes"); 3743 return -EINVAL; 3744 } 3745 } 3746 3747 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3748 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3749 conf->remote_ip.sa.sa_family = AF_INET; 3750 conf->saddr.sa.sa_family = AF_INET; 3751 } else if (!conf->remote_ip.sa.sa_family) { 3752 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3753 } else if (!conf->saddr.sa.sa_family) { 3754 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3755 } 3756 3757 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3758 NL_SET_ERR_MSG(extack, 3759 "Local and remote address must be from the same family"); 3760 return -EINVAL; 3761 } 3762 3763 if (vxlan_addr_multicast(&conf->saddr)) { 3764 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3765 return -EINVAL; 3766 } 3767 3768 if (conf->saddr.sa.sa_family == AF_INET6) { 3769 if (!IS_ENABLED(CONFIG_IPV6)) { 3770 NL_SET_ERR_MSG(extack, 3771 "IPv6 support not enabled in the kernel"); 3772 return -EPFNOSUPPORT; 3773 } 3774 use_ipv6 = true; 3775 conf->flags |= VXLAN_F_IPV6; 3776 3777 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3778 int local_type = 3779 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3780 int remote_type = 3781 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3782 3783 if (local_type & IPV6_ADDR_LINKLOCAL) { 3784 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3785 (remote_type != IPV6_ADDR_ANY)) { 3786 NL_SET_ERR_MSG(extack, 3787 "Invalid combination of local and remote address scopes"); 3788 return -EINVAL; 3789 } 3790 3791 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3792 } else { 3793 if (remote_type == 3794 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3795 NL_SET_ERR_MSG(extack, 3796 "Invalid combination of local and remote address scopes"); 3797 return -EINVAL; 3798 } 3799 3800 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3801 } 3802 } 3803 } 3804 3805 if (conf->label && !use_ipv6) { 3806 NL_SET_ERR_MSG(extack, 3807 "Label attribute only applies to IPv6 VXLAN devices"); 3808 return -EINVAL; 3809 } 3810 3811 if (conf->label_policy && !use_ipv6) { 3812 NL_SET_ERR_MSG(extack, 3813 "Label policy only applies to IPv6 VXLAN devices"); 3814 return -EINVAL; 3815 } 3816 3817 if (conf->remote_ifindex) { 3818 struct net_device *lowerdev; 3819 3820 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3821 if (!lowerdev) { 3822 NL_SET_ERR_MSG(extack, 3823 "Invalid local interface, device not found"); 3824 return -ENODEV; 3825 } 3826 3827 #if IS_ENABLED(CONFIG_IPV6) 3828 if (use_ipv6) { 3829 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3830 3831 if (idev && idev->cnf.disable_ipv6) { 3832 NL_SET_ERR_MSG(extack, 3833 "IPv6 support disabled by administrator"); 3834 return -EPERM; 3835 } 3836 } 3837 #endif 3838 3839 *lower = lowerdev; 3840 } else { 3841 if (vxlan_addr_multicast(&conf->remote_ip)) { 3842 NL_SET_ERR_MSG(extack, 3843 "Local interface required for multicast remote destination"); 3844 3845 return -EINVAL; 3846 } 3847 3848 #if IS_ENABLED(CONFIG_IPV6) 3849 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3850 NL_SET_ERR_MSG(extack, 3851 "Local interface required for link-local local/remote addresses"); 3852 return -EINVAL; 3853 } 3854 #endif 3855 3856 *lower = NULL; 3857 } 3858 3859 if (!conf->dst_port) { 3860 if (conf->flags & VXLAN_F_GPE) 3861 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3862 else 3863 conf->dst_port = htons(vxlan_port); 3864 } 3865 3866 if (!conf->age_interval) 3867 conf->age_interval = FDB_AGE_DEFAULT; 3868 3869 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3870 NL_SET_ERR_MSG(extack, 3871 "A VXLAN device with the specified VNI already exists"); 3872 return -EEXIST; 3873 } 3874 3875 return 0; 3876 } 3877 3878 static void vxlan_config_apply(struct net_device *dev, 3879 struct vxlan_config *conf, 3880 struct net_device *lowerdev, 3881 struct net *src_net, 3882 bool changelink) 3883 { 3884 struct vxlan_dev *vxlan = netdev_priv(dev); 3885 struct vxlan_rdst *dst = &vxlan->default_dst; 3886 unsigned short needed_headroom = ETH_HLEN; 3887 int max_mtu = ETH_MAX_MTU; 3888 u32 flags = conf->flags; 3889 3890 if (!changelink) { 3891 if (flags & VXLAN_F_GPE) 3892 vxlan_raw_setup(dev); 3893 else 3894 vxlan_ether_setup(dev); 3895 3896 if (conf->mtu) 3897 dev->mtu = conf->mtu; 3898 3899 vxlan->net = src_net; 3900 } 3901 3902 dst->remote_vni = conf->vni; 3903 3904 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3905 3906 if (lowerdev) { 3907 dst->remote_ifindex = conf->remote_ifindex; 3908 3909 netif_inherit_tso_max(dev, lowerdev); 3910 3911 needed_headroom = lowerdev->hard_header_len; 3912 needed_headroom += lowerdev->needed_headroom; 3913 3914 dev->needed_tailroom = lowerdev->needed_tailroom; 3915 3916 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3917 if (max_mtu < ETH_MIN_MTU) 3918 max_mtu = ETH_MIN_MTU; 3919 3920 if (!changelink && !conf->mtu) 3921 dev->mtu = max_mtu; 3922 } 3923 3924 if (dev->mtu > max_mtu) 3925 dev->mtu = max_mtu; 3926 3927 if (flags & VXLAN_F_COLLECT_METADATA) 3928 flags |= VXLAN_F_IPV6; 3929 needed_headroom += vxlan_headroom(flags); 3930 dev->needed_headroom = needed_headroom; 3931 3932 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3933 } 3934 3935 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3936 struct vxlan_config *conf, bool changelink, 3937 struct netlink_ext_ack *extack) 3938 { 3939 struct vxlan_dev *vxlan = netdev_priv(dev); 3940 struct net_device *lowerdev; 3941 int ret; 3942 3943 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3944 if (ret) 3945 return ret; 3946 3947 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3948 3949 return 0; 3950 } 3951 3952 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3953 struct vxlan_config *conf, 3954 struct netlink_ext_ack *extack) 3955 { 3956 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3957 struct vxlan_dev *vxlan = netdev_priv(dev); 3958 struct net_device *remote_dev = NULL; 3959 struct vxlan_fdb *f = NULL; 3960 bool unregister = false; 3961 struct vxlan_rdst *dst; 3962 int err; 3963 3964 dst = &vxlan->default_dst; 3965 err = vxlan_dev_configure(net, dev, conf, false, extack); 3966 if (err) 3967 return err; 3968 3969 dev->ethtool_ops = &vxlan_ethtool_ops; 3970 3971 /* create an fdb entry for a valid default destination */ 3972 if (!vxlan_addr_any(&dst->remote_ip)) { 3973 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3974 &dst->remote_ip, 3975 NUD_REACHABLE | NUD_PERMANENT, 3976 vxlan->cfg.dst_port, 3977 dst->remote_vni, 3978 dst->remote_vni, 3979 dst->remote_ifindex, 3980 NTF_SELF, 0, &f, extack); 3981 if (err) 3982 return err; 3983 } 3984 3985 err = register_netdevice(dev); 3986 if (err) 3987 goto errout; 3988 unregister = true; 3989 3990 if (dst->remote_ifindex) { 3991 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3992 if (!remote_dev) { 3993 err = -ENODEV; 3994 goto errout; 3995 } 3996 3997 err = netdev_upper_dev_link(remote_dev, dev, extack); 3998 if (err) 3999 goto errout; 4000 } 4001 4002 err = rtnl_configure_link(dev, NULL, 0, NULL); 4003 if (err < 0) 4004 goto unlink; 4005 4006 if (f) { 4007 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 4008 4009 /* notify default fdb entry */ 4010 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 4011 RTM_NEWNEIGH, true, extack); 4012 if (err) { 4013 vxlan_fdb_destroy(vxlan, f, false, false); 4014 if (remote_dev) 4015 netdev_upper_dev_unlink(remote_dev, dev); 4016 goto unregister; 4017 } 4018 } 4019 4020 list_add(&vxlan->next, &vn->vxlan_list); 4021 if (remote_dev) 4022 dst->remote_dev = remote_dev; 4023 return 0; 4024 unlink: 4025 if (remote_dev) 4026 netdev_upper_dev_unlink(remote_dev, dev); 4027 errout: 4028 /* unregister_netdevice() destroys the default FDB entry with deletion 4029 * notification. But the addition notification was not sent yet, so 4030 * destroy the entry by hand here. 4031 */ 4032 if (f) 4033 __vxlan_fdb_free(f); 4034 unregister: 4035 if (unregister) 4036 unregister_netdevice(dev); 4037 return err; 4038 } 4039 4040 /* Set/clear flags based on attribute */ 4041 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4042 int attrtype, unsigned long mask, bool changelink, 4043 bool changelink_supported, 4044 struct netlink_ext_ack *extack) 4045 { 4046 unsigned long flags; 4047 4048 if (!tb[attrtype]) 4049 return 0; 4050 4051 if (changelink && !changelink_supported) { 4052 vxlan_flag_attr_error(attrtype, extack); 4053 return -EOPNOTSUPP; 4054 } 4055 4056 if (vxlan_policy[attrtype].type == NLA_FLAG) 4057 flags = conf->flags | mask; 4058 else if (nla_get_u8(tb[attrtype])) 4059 flags = conf->flags | mask; 4060 else 4061 flags = conf->flags & ~mask; 4062 4063 conf->flags = flags; 4064 4065 return 0; 4066 } 4067 4068 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4069 struct net_device *dev, struct vxlan_config *conf, 4070 bool changelink, struct netlink_ext_ack *extack) 4071 { 4072 struct vxlanhdr used_bits = { 4073 .vx_flags = VXLAN_HF_VNI, 4074 .vx_vni = VXLAN_VNI_MASK, 4075 }; 4076 struct vxlan_dev *vxlan = netdev_priv(dev); 4077 int err = 0; 4078 4079 memset(conf, 0, sizeof(*conf)); 4080 4081 /* if changelink operation, start with old existing cfg */ 4082 if (changelink) 4083 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4084 4085 if (data[IFLA_VXLAN_ID]) { 4086 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4087 4088 if (changelink && (vni != conf->vni)) { 4089 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4090 return -EOPNOTSUPP; 4091 } 4092 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4093 } 4094 4095 if (data[IFLA_VXLAN_GROUP]) { 4096 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4097 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4098 return -EOPNOTSUPP; 4099 } 4100 4101 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4102 conf->remote_ip.sa.sa_family = AF_INET; 4103 } else if (data[IFLA_VXLAN_GROUP6]) { 4104 if (!IS_ENABLED(CONFIG_IPV6)) { 4105 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4106 return -EPFNOSUPPORT; 4107 } 4108 4109 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4110 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4111 return -EOPNOTSUPP; 4112 } 4113 4114 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4115 conf->remote_ip.sa.sa_family = AF_INET6; 4116 } 4117 4118 if (data[IFLA_VXLAN_LOCAL]) { 4119 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4120 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4121 return -EOPNOTSUPP; 4122 } 4123 4124 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4125 conf->saddr.sa.sa_family = AF_INET; 4126 } else if (data[IFLA_VXLAN_LOCAL6]) { 4127 if (!IS_ENABLED(CONFIG_IPV6)) { 4128 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4129 return -EPFNOSUPPORT; 4130 } 4131 4132 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4133 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4134 return -EOPNOTSUPP; 4135 } 4136 4137 /* TODO: respect scope id */ 4138 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4139 conf->saddr.sa.sa_family = AF_INET6; 4140 } 4141 4142 if (data[IFLA_VXLAN_LINK]) 4143 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4144 4145 if (data[IFLA_VXLAN_TOS]) 4146 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4147 4148 if (data[IFLA_VXLAN_TTL]) 4149 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4150 4151 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4152 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4153 VXLAN_F_TTL_INHERIT, changelink, false, 4154 extack); 4155 if (err) 4156 return err; 4157 4158 } 4159 4160 if (data[IFLA_VXLAN_LABEL]) 4161 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4162 IPV6_FLOWLABEL_MASK; 4163 if (data[IFLA_VXLAN_LABEL_POLICY]) 4164 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4165 4166 if (data[IFLA_VXLAN_LEARNING]) { 4167 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4168 VXLAN_F_LEARN, changelink, true, 4169 extack); 4170 if (err) 4171 return err; 4172 } else if (!changelink) { 4173 /* default to learn on a new device */ 4174 conf->flags |= VXLAN_F_LEARN; 4175 } 4176 4177 if (data[IFLA_VXLAN_AGEING]) 4178 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4179 4180 if (data[IFLA_VXLAN_PROXY]) { 4181 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4182 VXLAN_F_PROXY, changelink, false, 4183 extack); 4184 if (err) 4185 return err; 4186 } 4187 4188 if (data[IFLA_VXLAN_RSC]) { 4189 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4190 VXLAN_F_RSC, changelink, false, 4191 extack); 4192 if (err) 4193 return err; 4194 } 4195 4196 if (data[IFLA_VXLAN_L2MISS]) { 4197 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4198 VXLAN_F_L2MISS, changelink, false, 4199 extack); 4200 if (err) 4201 return err; 4202 } 4203 4204 if (data[IFLA_VXLAN_L3MISS]) { 4205 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4206 VXLAN_F_L3MISS, changelink, false, 4207 extack); 4208 if (err) 4209 return err; 4210 } 4211 4212 if (data[IFLA_VXLAN_LIMIT]) { 4213 if (changelink) { 4214 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4215 "Cannot change limit"); 4216 return -EOPNOTSUPP; 4217 } 4218 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4219 } 4220 4221 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4222 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4223 VXLAN_F_COLLECT_METADATA, changelink, false, 4224 extack); 4225 if (err) 4226 return err; 4227 } 4228 4229 if (data[IFLA_VXLAN_PORT_RANGE]) { 4230 if (!changelink) { 4231 const struct ifla_vxlan_port_range *p 4232 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4233 conf->port_min = ntohs(p->low); 4234 conf->port_max = ntohs(p->high); 4235 } else { 4236 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4237 "Cannot change port range"); 4238 return -EOPNOTSUPP; 4239 } 4240 } 4241 4242 if (data[IFLA_VXLAN_PORT]) { 4243 if (changelink) { 4244 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4245 "Cannot change port"); 4246 return -EOPNOTSUPP; 4247 } 4248 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4249 } 4250 4251 if (data[IFLA_VXLAN_UDP_CSUM]) { 4252 if (changelink) { 4253 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4254 "Cannot change UDP_CSUM flag"); 4255 return -EOPNOTSUPP; 4256 } 4257 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4258 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4259 } 4260 4261 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4262 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4263 VXLAN_F_LOCALBYPASS, changelink, 4264 true, extack); 4265 if (err) 4266 return err; 4267 } else if (!changelink) { 4268 /* default to local bypass on a new device */ 4269 conf->flags |= VXLAN_F_LOCALBYPASS; 4270 } 4271 4272 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4273 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4274 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4275 false, extack); 4276 if (err) 4277 return err; 4278 } 4279 4280 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4281 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4282 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4283 false, extack); 4284 if (err) 4285 return err; 4286 } 4287 4288 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4289 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4290 VXLAN_F_REMCSUM_TX, changelink, false, 4291 extack); 4292 if (err) 4293 return err; 4294 } 4295 4296 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4297 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4298 VXLAN_F_REMCSUM_RX, changelink, false, 4299 extack); 4300 if (err) 4301 return err; 4302 used_bits.vx_flags |= VXLAN_HF_RCO; 4303 used_bits.vx_vni |= ~VXLAN_VNI_MASK; 4304 } 4305 4306 if (data[IFLA_VXLAN_GBP]) { 4307 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4308 VXLAN_F_GBP, changelink, false, extack); 4309 if (err) 4310 return err; 4311 used_bits.vx_flags |= VXLAN_GBP_USED_BITS; 4312 } 4313 4314 if (data[IFLA_VXLAN_GPE]) { 4315 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4316 VXLAN_F_GPE, changelink, false, 4317 extack); 4318 if (err) 4319 return err; 4320 4321 used_bits.vx_flags |= VXLAN_GPE_USED_BITS; 4322 } 4323 4324 if (data[IFLA_VXLAN_RESERVED_BITS]) { 4325 struct vxlanhdr reserved_bits; 4326 4327 if (changelink) { 4328 NL_SET_ERR_MSG_ATTR(extack, 4329 data[IFLA_VXLAN_RESERVED_BITS], 4330 "Cannot change reserved_bits"); 4331 return -EOPNOTSUPP; 4332 } 4333 4334 nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS], 4335 sizeof(reserved_bits)); 4336 if (used_bits.vx_flags & reserved_bits.vx_flags || 4337 used_bits.vx_vni & reserved_bits.vx_vni) { 4338 __be64 ub_be64, rb_be64; 4339 4340 memcpy(&ub_be64, &used_bits, sizeof(ub_be64)); 4341 memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64)); 4342 4343 NL_SET_ERR_MSG_ATTR_FMT(extack, 4344 data[IFLA_VXLAN_RESERVED_BITS], 4345 "Used bits %#018llx cannot overlap reserved bits %#018llx", 4346 be64_to_cpu(ub_be64), 4347 be64_to_cpu(rb_be64)); 4348 return -EINVAL; 4349 } 4350 4351 conf->reserved_bits = reserved_bits; 4352 } else { 4353 /* For backwards compatibility, only allow reserved fields to be 4354 * used by VXLAN extensions if explicitly requested. 4355 */ 4356 conf->reserved_bits = (struct vxlanhdr) { 4357 .vx_flags = ~used_bits.vx_flags, 4358 .vx_vni = ~used_bits.vx_vni, 4359 }; 4360 } 4361 4362 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4363 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4364 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4365 false, extack); 4366 if (err) 4367 return err; 4368 } 4369 4370 if (tb[IFLA_MTU]) { 4371 if (changelink) { 4372 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4373 "Cannot change mtu"); 4374 return -EOPNOTSUPP; 4375 } 4376 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4377 } 4378 4379 if (data[IFLA_VXLAN_DF]) 4380 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4381 4382 if (data[IFLA_VXLAN_VNIFILTER]) { 4383 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4384 VXLAN_F_VNIFILTER, changelink, false, 4385 extack); 4386 if (err) 4387 return err; 4388 4389 if ((conf->flags & VXLAN_F_VNIFILTER) && 4390 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4391 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4392 "vxlan vnifilter only valid in collect metadata mode"); 4393 return -EINVAL; 4394 } 4395 } 4396 4397 return 0; 4398 } 4399 4400 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4401 struct nlattr *tb[], struct nlattr *data[], 4402 struct netlink_ext_ack *extack) 4403 { 4404 struct vxlan_config conf; 4405 int err; 4406 4407 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4408 if (err) 4409 return err; 4410 4411 return __vxlan_dev_create(src_net, dev, &conf, extack); 4412 } 4413 4414 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4415 struct nlattr *data[], 4416 struct netlink_ext_ack *extack) 4417 { 4418 struct vxlan_dev *vxlan = netdev_priv(dev); 4419 struct net_device *lowerdev; 4420 struct vxlan_config conf; 4421 struct vxlan_rdst *dst; 4422 int err; 4423 4424 dst = &vxlan->default_dst; 4425 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4426 if (err) 4427 return err; 4428 4429 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4430 vxlan, extack); 4431 if (err) 4432 return err; 4433 4434 if (dst->remote_dev == lowerdev) 4435 lowerdev = NULL; 4436 4437 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4438 extack); 4439 if (err) 4440 return err; 4441 4442 /* handle default dst entry */ 4443 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4444 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4445 4446 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4447 if (!vxlan_addr_any(&conf.remote_ip)) { 4448 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4449 &conf.remote_ip, 4450 NUD_REACHABLE | NUD_PERMANENT, 4451 NLM_F_APPEND | NLM_F_CREATE, 4452 vxlan->cfg.dst_port, 4453 conf.vni, conf.vni, 4454 conf.remote_ifindex, 4455 NTF_SELF, 0, true, extack); 4456 if (err) { 4457 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4458 netdev_adjacent_change_abort(dst->remote_dev, 4459 lowerdev, dev); 4460 return err; 4461 } 4462 } 4463 if (!vxlan_addr_any(&dst->remote_ip)) 4464 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4465 dst->remote_ip, 4466 vxlan->cfg.dst_port, 4467 dst->remote_vni, 4468 dst->remote_vni, 4469 dst->remote_ifindex, 4470 true); 4471 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4472 4473 /* If vni filtering device, also update fdb entries of 4474 * all vnis that were using default remote ip 4475 */ 4476 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4477 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4478 &conf.remote_ip, extack); 4479 if (err) { 4480 netdev_adjacent_change_abort(dst->remote_dev, 4481 lowerdev, dev); 4482 return err; 4483 } 4484 } 4485 } 4486 4487 if (conf.age_interval != vxlan->cfg.age_interval) 4488 mod_timer(&vxlan->age_timer, jiffies); 4489 4490 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4491 if (lowerdev && lowerdev != dst->remote_dev) 4492 dst->remote_dev = lowerdev; 4493 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4494 return 0; 4495 } 4496 4497 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4498 { 4499 struct vxlan_dev *vxlan = netdev_priv(dev); 4500 struct vxlan_fdb_flush_desc desc = { 4501 /* Default entry is deleted at vxlan_uninit. */ 4502 .ignore_default_entry = true, 4503 }; 4504 4505 vxlan_flush(vxlan, &desc); 4506 4507 list_del(&vxlan->next); 4508 unregister_netdevice_queue(dev, head); 4509 if (vxlan->default_dst.remote_dev) 4510 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4511 } 4512 4513 static size_t vxlan_get_size(const struct net_device *dev) 4514 { 4515 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4516 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4517 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4518 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4519 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4520 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4521 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4522 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4523 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4524 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4525 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4526 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4527 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4528 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4529 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4530 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4531 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4532 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4533 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4534 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4535 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4536 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4537 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4538 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4540 /* IFLA_VXLAN_PORT_RANGE */ 4541 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4542 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4543 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4544 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4545 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4546 /* IFLA_VXLAN_RESERVED_BITS */ 4547 nla_total_size(sizeof(struct vxlanhdr)) + 4548 0; 4549 } 4550 4551 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4552 { 4553 const struct vxlan_dev *vxlan = netdev_priv(dev); 4554 const struct vxlan_rdst *dst = &vxlan->default_dst; 4555 struct ifla_vxlan_port_range ports = { 4556 .low = htons(vxlan->cfg.port_min), 4557 .high = htons(vxlan->cfg.port_max), 4558 }; 4559 4560 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4561 goto nla_put_failure; 4562 4563 if (!vxlan_addr_any(&dst->remote_ip)) { 4564 if (dst->remote_ip.sa.sa_family == AF_INET) { 4565 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4566 dst->remote_ip.sin.sin_addr.s_addr)) 4567 goto nla_put_failure; 4568 #if IS_ENABLED(CONFIG_IPV6) 4569 } else { 4570 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4571 &dst->remote_ip.sin6.sin6_addr)) 4572 goto nla_put_failure; 4573 #endif 4574 } 4575 } 4576 4577 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4578 goto nla_put_failure; 4579 4580 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4581 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4582 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4583 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4584 goto nla_put_failure; 4585 #if IS_ENABLED(CONFIG_IPV6) 4586 } else { 4587 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4588 &vxlan->cfg.saddr.sin6.sin6_addr)) 4589 goto nla_put_failure; 4590 #endif 4591 } 4592 } 4593 4594 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4595 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4596 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4597 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4598 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4599 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4600 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4601 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4602 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4603 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4604 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4605 nla_put_u8(skb, IFLA_VXLAN_RSC, 4606 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4607 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4608 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4609 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4610 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4611 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4612 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4613 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4614 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4615 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4616 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4617 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4618 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4619 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4620 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4621 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4622 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4623 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4624 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4625 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4626 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4627 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4628 goto nla_put_failure; 4629 4630 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4631 goto nla_put_failure; 4632 4633 if (vxlan->cfg.flags & VXLAN_F_GBP && 4634 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4635 goto nla_put_failure; 4636 4637 if (vxlan->cfg.flags & VXLAN_F_GPE && 4638 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4639 goto nla_put_failure; 4640 4641 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4642 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4643 goto nla_put_failure; 4644 4645 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4646 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4647 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4648 goto nla_put_failure; 4649 4650 if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS, 4651 sizeof(vxlan->cfg.reserved_bits), 4652 &vxlan->cfg.reserved_bits)) 4653 goto nla_put_failure; 4654 4655 return 0; 4656 4657 nla_put_failure: 4658 return -EMSGSIZE; 4659 } 4660 4661 static struct net *vxlan_get_link_net(const struct net_device *dev) 4662 { 4663 struct vxlan_dev *vxlan = netdev_priv(dev); 4664 4665 return READ_ONCE(vxlan->net); 4666 } 4667 4668 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4669 .kind = "vxlan", 4670 .maxtype = IFLA_VXLAN_MAX, 4671 .policy = vxlan_policy, 4672 .priv_size = sizeof(struct vxlan_dev), 4673 .setup = vxlan_setup, 4674 .validate = vxlan_validate, 4675 .newlink = vxlan_newlink, 4676 .changelink = vxlan_changelink, 4677 .dellink = vxlan_dellink, 4678 .get_size = vxlan_get_size, 4679 .fill_info = vxlan_fill_info, 4680 .get_link_net = vxlan_get_link_net, 4681 }; 4682 4683 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4684 u8 name_assign_type, 4685 struct vxlan_config *conf) 4686 { 4687 struct nlattr *tb[IFLA_MAX + 1]; 4688 struct net_device *dev; 4689 int err; 4690 4691 memset(&tb, 0, sizeof(tb)); 4692 4693 dev = rtnl_create_link(net, name, name_assign_type, 4694 &vxlan_link_ops, tb, NULL); 4695 if (IS_ERR(dev)) 4696 return dev; 4697 4698 err = __vxlan_dev_create(net, dev, conf, NULL); 4699 if (err < 0) { 4700 free_netdev(dev); 4701 return ERR_PTR(err); 4702 } 4703 4704 err = rtnl_configure_link(dev, NULL, 0, NULL); 4705 if (err < 0) { 4706 LIST_HEAD(list_kill); 4707 4708 vxlan_dellink(dev, &list_kill); 4709 unregister_netdevice_many(&list_kill); 4710 return ERR_PTR(err); 4711 } 4712 4713 return dev; 4714 } 4715 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4716 4717 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4718 struct net_device *dev) 4719 { 4720 struct vxlan_dev *vxlan, *next; 4721 LIST_HEAD(list_kill); 4722 4723 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4724 struct vxlan_rdst *dst = &vxlan->default_dst; 4725 4726 /* In case we created vxlan device with carrier 4727 * and we loose the carrier due to module unload 4728 * we also need to remove vxlan device. In other 4729 * cases, it's not necessary and remote_ifindex 4730 * is 0 here, so no matches. 4731 */ 4732 if (dst->remote_ifindex == dev->ifindex) 4733 vxlan_dellink(vxlan->dev, &list_kill); 4734 } 4735 4736 unregister_netdevice_many(&list_kill); 4737 } 4738 4739 static int vxlan_netdevice_event(struct notifier_block *unused, 4740 unsigned long event, void *ptr) 4741 { 4742 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4743 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4744 4745 if (event == NETDEV_UNREGISTER) 4746 vxlan_handle_lowerdev_unregister(vn, dev); 4747 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4748 vxlan_offload_rx_ports(dev, true); 4749 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4750 vxlan_offload_rx_ports(dev, false); 4751 4752 return NOTIFY_DONE; 4753 } 4754 4755 static struct notifier_block vxlan_notifier_block __read_mostly = { 4756 .notifier_call = vxlan_netdevice_event, 4757 }; 4758 4759 static void 4760 vxlan_fdb_offloaded_set(struct net_device *dev, 4761 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4762 { 4763 struct vxlan_dev *vxlan = netdev_priv(dev); 4764 struct vxlan_rdst *rdst; 4765 struct vxlan_fdb *f; 4766 u32 hash_index; 4767 4768 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4769 4770 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4771 4772 f = __vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4773 if (!f) 4774 goto out; 4775 4776 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4777 fdb_info->remote_port, 4778 fdb_info->remote_vni, 4779 fdb_info->remote_ifindex); 4780 if (!rdst) 4781 goto out; 4782 4783 rdst->offloaded = fdb_info->offloaded; 4784 4785 out: 4786 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4787 } 4788 4789 static int 4790 vxlan_fdb_external_learn_add(struct net_device *dev, 4791 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4792 { 4793 struct vxlan_dev *vxlan = netdev_priv(dev); 4794 struct netlink_ext_ack *extack; 4795 u32 hash_index; 4796 int err; 4797 4798 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4799 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4800 4801 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4802 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4803 NUD_REACHABLE, 4804 NLM_F_CREATE | NLM_F_REPLACE, 4805 fdb_info->remote_port, 4806 fdb_info->vni, 4807 fdb_info->remote_vni, 4808 fdb_info->remote_ifindex, 4809 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4810 0, false, extack); 4811 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4812 4813 return err; 4814 } 4815 4816 static int 4817 vxlan_fdb_external_learn_del(struct net_device *dev, 4818 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4819 { 4820 struct vxlan_dev *vxlan = netdev_priv(dev); 4821 struct vxlan_fdb *f; 4822 u32 hash_index; 4823 int err = 0; 4824 4825 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4826 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4827 4828 f = __vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4829 if (!f) 4830 err = -ENOENT; 4831 else if (f->flags & NTF_EXT_LEARNED) 4832 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4833 fdb_info->remote_ip, 4834 fdb_info->remote_port, 4835 fdb_info->vni, 4836 fdb_info->remote_vni, 4837 fdb_info->remote_ifindex, 4838 false); 4839 4840 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4841 4842 return err; 4843 } 4844 4845 static int vxlan_switchdev_event(struct notifier_block *unused, 4846 unsigned long event, void *ptr) 4847 { 4848 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4849 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4850 int err = 0; 4851 4852 switch (event) { 4853 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4854 vxlan_fdb_offloaded_set(dev, ptr); 4855 break; 4856 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4857 fdb_info = ptr; 4858 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4859 if (err) { 4860 err = notifier_from_errno(err); 4861 break; 4862 } 4863 fdb_info->offloaded = true; 4864 vxlan_fdb_offloaded_set(dev, fdb_info); 4865 break; 4866 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4867 fdb_info = ptr; 4868 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4869 if (err) { 4870 err = notifier_from_errno(err); 4871 break; 4872 } 4873 fdb_info->offloaded = false; 4874 vxlan_fdb_offloaded_set(dev, fdb_info); 4875 break; 4876 } 4877 4878 return err; 4879 } 4880 4881 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4882 .notifier_call = vxlan_switchdev_event, 4883 }; 4884 4885 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4886 { 4887 struct vxlan_fdb *fdb; 4888 struct vxlan_dev *vxlan; 4889 u32 hash_index; 4890 4891 rcu_read_lock(); 4892 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4893 vxlan = rcu_dereference(fdb->vdev); 4894 WARN_ON(!vxlan); 4895 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4896 vxlan->default_dst.remote_vni); 4897 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4898 if (!hlist_unhashed(&fdb->hlist)) 4899 vxlan_fdb_destroy(vxlan, fdb, false, false); 4900 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4901 } 4902 rcu_read_unlock(); 4903 } 4904 4905 static int vxlan_nexthop_event(struct notifier_block *nb, 4906 unsigned long event, void *ptr) 4907 { 4908 struct nh_notifier_info *info = ptr; 4909 struct nexthop *nh; 4910 4911 if (event != NEXTHOP_EVENT_DEL) 4912 return NOTIFY_DONE; 4913 4914 nh = nexthop_find_by_id(info->net, info->id); 4915 if (!nh) 4916 return NOTIFY_DONE; 4917 4918 vxlan_fdb_nh_flush(nh); 4919 4920 return NOTIFY_DONE; 4921 } 4922 4923 static __net_init int vxlan_init_net(struct net *net) 4924 { 4925 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4926 unsigned int h; 4927 4928 INIT_LIST_HEAD(&vn->vxlan_list); 4929 spin_lock_init(&vn->sock_lock); 4930 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4931 4932 for (h = 0; h < PORT_HASH_SIZE; ++h) 4933 INIT_HLIST_HEAD(&vn->sock_list[h]); 4934 4935 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4936 NULL); 4937 } 4938 4939 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4940 struct list_head *dev_to_kill) 4941 { 4942 struct vxlan_dev *vxlan, *next; 4943 4944 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4945 vxlan_dellink(vxlan->dev, dev_to_kill); 4946 } 4947 4948 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list, 4949 struct list_head *dev_to_kill) 4950 { 4951 struct net *net; 4952 4953 ASSERT_RTNL(); 4954 list_for_each_entry(net, net_list, exit_list) { 4955 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4956 4957 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4958 4959 vxlan_destroy_tunnels(vn, dev_to_kill); 4960 } 4961 } 4962 4963 static void __net_exit vxlan_exit_net(struct net *net) 4964 { 4965 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4966 unsigned int h; 4967 4968 for (h = 0; h < PORT_HASH_SIZE; ++h) 4969 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4970 } 4971 4972 static struct pernet_operations vxlan_net_ops = { 4973 .init = vxlan_init_net, 4974 .exit_batch_rtnl = vxlan_exit_batch_rtnl, 4975 .exit = vxlan_exit_net, 4976 .id = &vxlan_net_id, 4977 .size = sizeof(struct vxlan_net), 4978 }; 4979 4980 static int __init vxlan_init_module(void) 4981 { 4982 int rc; 4983 4984 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4985 4986 rc = register_pernet_subsys(&vxlan_net_ops); 4987 if (rc) 4988 goto out1; 4989 4990 rc = register_netdevice_notifier(&vxlan_notifier_block); 4991 if (rc) 4992 goto out2; 4993 4994 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4995 if (rc) 4996 goto out3; 4997 4998 rc = rtnl_link_register(&vxlan_link_ops); 4999 if (rc) 5000 goto out4; 5001 5002 rc = vxlan_vnifilter_init(); 5003 if (rc) 5004 goto out5; 5005 5006 return 0; 5007 out5: 5008 rtnl_link_unregister(&vxlan_link_ops); 5009 out4: 5010 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5011 out3: 5012 unregister_netdevice_notifier(&vxlan_notifier_block); 5013 out2: 5014 unregister_pernet_subsys(&vxlan_net_ops); 5015 out1: 5016 return rc; 5017 } 5018 late_initcall(vxlan_init_module); 5019 5020 static void __exit vxlan_cleanup_module(void) 5021 { 5022 vxlan_vnifilter_uninit(); 5023 rtnl_link_unregister(&vxlan_link_ops); 5024 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 5025 unregister_netdevice_notifier(&vxlan_notifier_block); 5026 unregister_pernet_subsys(&vxlan_net_ops); 5027 /* rcu_barrier() is called by netns */ 5028 } 5029 module_exit(vxlan_cleanup_module); 5030 5031 MODULE_LICENSE("GPL"); 5032 MODULE_VERSION(VXLAN_VERSION); 5033 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 5034 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 5035 MODULE_ALIAS_RTNL_LINK("vxlan"); 5036