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 <linux/rhashtable.h> 19 #include <net/arp.h> 20 #include <net/ndisc.h> 21 #include <net/gro.h> 22 #include <net/ipv6_stubs.h> 23 #include <net/ip.h> 24 #include <net/icmp.h> 25 #include <net/rtnetlink.h> 26 #include <net/inet_ecn.h> 27 #include <net/net_namespace.h> 28 #include <net/netns/generic.h> 29 #include <net/netdev_lock.h> 30 #include <net/tun_proto.h> 31 #include <net/vxlan.h> 32 #include <net/nexthop.h> 33 34 #if IS_ENABLED(CONFIG_IPV6) 35 #include <net/ip6_tunnel.h> 36 #include <net/ip6_checksum.h> 37 #endif 38 39 #include "vxlan_private.h" 40 41 #define VXLAN_VERSION "0.1" 42 43 #define FDB_AGE_DEFAULT 300 /* 5 min */ 44 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ 45 46 /* UDP port for VXLAN traffic. 47 * The IANA assigned port is 4789, but the Linux default is 8472 48 * for compatibility with early adopters. 49 */ 50 static unsigned short vxlan_port __read_mostly = 8472; 51 module_param_named(udp_port, vxlan_port, ushort, 0444); 52 MODULE_PARM_DESC(udp_port, "Destination UDP port"); 53 54 static bool log_ecn_error = true; 55 module_param(log_ecn_error, bool, 0644); 56 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 57 58 unsigned int vxlan_net_id; 59 60 const u8 all_zeros_mac[ETH_ALEN + 2]; 61 static struct rtnl_link_ops vxlan_link_ops; 62 63 static int vxlan_sock_add(struct vxlan_dev *vxlan); 64 65 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan); 66 67 static const struct rhashtable_params vxlan_fdb_rht_params = { 68 .head_offset = offsetof(struct vxlan_fdb, rhnode), 69 .key_offset = offsetof(struct vxlan_fdb, key), 70 .key_len = sizeof(struct vxlan_fdb_key), 71 .automatic_shrinking = true, 72 }; 73 74 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) 75 { 76 return vs->flags & VXLAN_F_COLLECT_METADATA || 77 ip_tunnel_collect_metadata(); 78 } 79 80 /* Find VXLAN socket based on network namespace, address family, UDP port, 81 * enabled unshareable flags and socket device binding (see l3mdev with 82 * non-default VRF). 83 */ 84 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, 85 __be16 port, u32 flags, int ifindex) 86 { 87 struct vxlan_sock *vs; 88 89 flags &= VXLAN_F_RCV_FLAGS; 90 91 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { 92 if (inet_sk(vs->sock->sk)->inet_sport == port && 93 vxlan_get_sk_family(vs) == family && 94 vs->flags == flags && 95 vs->sock->sk->sk_bound_dev_if == ifindex) 96 return vs; 97 } 98 return NULL; 99 } 100 101 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, 102 int ifindex, __be32 vni, 103 struct vxlan_vni_node **vninode) 104 { 105 struct vxlan_vni_node *vnode; 106 struct vxlan_dev_node *node; 107 108 /* For flow based devices, map all packets to VNI 0 */ 109 if (vs->flags & VXLAN_F_COLLECT_METADATA && 110 !(vs->flags & VXLAN_F_VNIFILTER)) 111 vni = 0; 112 113 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) { 114 if (!node->vxlan) 115 continue; 116 vnode = NULL; 117 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 118 vnode = vxlan_vnifilter_lookup(node->vxlan, vni); 119 if (!vnode) 120 continue; 121 } else if (node->vxlan->default_dst.remote_vni != vni) { 122 continue; 123 } 124 125 if (IS_ENABLED(CONFIG_IPV6)) { 126 const struct vxlan_config *cfg = &node->vxlan->cfg; 127 128 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) && 129 cfg->remote_ifindex != ifindex) 130 continue; 131 } 132 133 if (vninode) 134 *vninode = vnode; 135 return node->vxlan; 136 } 137 138 return NULL; 139 } 140 141 /* Look up VNI in a per net namespace table */ 142 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex, 143 __be32 vni, sa_family_t family, 144 __be16 port, u32 flags) 145 { 146 struct vxlan_sock *vs; 147 148 vs = vxlan_find_sock(net, family, port, flags, ifindex); 149 if (!vs) 150 return NULL; 151 152 return vxlan_vs_find_vni(vs, ifindex, vni, NULL); 153 } 154 155 /* Fill in neighbour message in skbuff. */ 156 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, 157 const struct vxlan_fdb *fdb, 158 u32 portid, u32 seq, int type, unsigned int flags, 159 const struct vxlan_rdst *rdst) 160 { 161 unsigned long now = jiffies; 162 struct nda_cacheinfo ci; 163 bool send_ip, send_eth; 164 struct nlmsghdr *nlh; 165 struct nexthop *nh; 166 struct ndmsg *ndm; 167 int nh_family; 168 u32 nh_id; 169 170 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); 171 if (nlh == NULL) 172 return -EMSGSIZE; 173 174 ndm = nlmsg_data(nlh); 175 memset(ndm, 0, sizeof(*ndm)); 176 177 send_eth = send_ip = true; 178 179 rcu_read_lock(); 180 nh = rcu_dereference(fdb->nh); 181 if (nh) { 182 nh_family = nexthop_get_family(nh); 183 nh_id = nh->id; 184 } 185 rcu_read_unlock(); 186 187 if (type == RTM_GETNEIGH) { 188 if (rdst) { 189 send_ip = !vxlan_addr_any(&rdst->remote_ip); 190 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET; 191 } else if (nh) { 192 ndm->ndm_family = nh_family; 193 } 194 send_eth = !is_zero_ether_addr(fdb->key.eth_addr); 195 } else 196 ndm->ndm_family = AF_BRIDGE; 197 ndm->ndm_state = fdb->state; 198 ndm->ndm_ifindex = vxlan->dev->ifindex; 199 ndm->ndm_flags = fdb->flags; 200 if (rdst && rdst->offloaded) 201 ndm->ndm_flags |= NTF_OFFLOADED; 202 ndm->ndm_type = RTN_UNICAST; 203 204 if (!net_eq(dev_net(vxlan->dev), vxlan->net) && 205 nla_put_s32(skb, NDA_LINK_NETNSID, 206 peernet2id(dev_net(vxlan->dev), vxlan->net))) 207 goto nla_put_failure; 208 209 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->key.eth_addr)) 210 goto nla_put_failure; 211 if (nh) { 212 if (nla_put_u32(skb, NDA_NH_ID, nh_id)) 213 goto nla_put_failure; 214 } else if (rdst) { 215 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, 216 &rdst->remote_ip)) 217 goto nla_put_failure; 218 219 if (rdst->remote_port && 220 rdst->remote_port != vxlan->cfg.dst_port && 221 nla_put_be16(skb, NDA_PORT, rdst->remote_port)) 222 goto nla_put_failure; 223 if (rdst->remote_vni != vxlan->default_dst.remote_vni && 224 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) 225 goto nla_put_failure; 226 if (rdst->remote_ifindex && 227 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) 228 goto nla_put_failure; 229 } 230 231 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->key.vni && 232 nla_put_u32(skb, NDA_SRC_VNI, 233 be32_to_cpu(fdb->key.vni))) 234 goto nla_put_failure; 235 236 ci.ndm_used = jiffies_to_clock_t(now - READ_ONCE(fdb->used)); 237 ci.ndm_confirmed = 0; 238 ci.ndm_updated = jiffies_to_clock_t(now - READ_ONCE(fdb->updated)); 239 ci.ndm_refcnt = 0; 240 241 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) 242 goto nla_put_failure; 243 244 nlmsg_end(skb, nlh); 245 return 0; 246 247 nla_put_failure: 248 nlmsg_cancel(skb, nlh); 249 return -EMSGSIZE; 250 } 251 252 static inline size_t vxlan_nlmsg_size(void) 253 { 254 return NLMSG_ALIGN(sizeof(struct ndmsg)) 255 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ 256 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ 257 + nla_total_size(sizeof(__be16)) /* NDA_PORT */ 258 + nla_total_size(sizeof(__be32)) /* NDA_VNI */ 259 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ 260 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ 261 + nla_total_size(sizeof(struct nda_cacheinfo)); 262 } 263 264 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 265 struct vxlan_rdst *rd, int type) 266 { 267 struct net *net = dev_net(vxlan->dev); 268 struct sk_buff *skb; 269 int err = -ENOBUFS; 270 271 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); 272 if (skb == NULL) 273 goto errout; 274 275 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); 276 if (err < 0) { 277 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ 278 WARN_ON(err == -EMSGSIZE); 279 kfree_skb(skb); 280 goto errout; 281 } 282 283 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 284 return; 285 errout: 286 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 287 } 288 289 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 290 const struct vxlan_fdb *fdb, 291 const struct vxlan_rdst *rd, 292 struct netlink_ext_ack *extack, 293 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 294 { 295 fdb_info->info.dev = vxlan->dev; 296 fdb_info->info.extack = extack; 297 fdb_info->remote_ip = rd->remote_ip; 298 fdb_info->remote_port = rd->remote_port; 299 fdb_info->remote_vni = rd->remote_vni; 300 fdb_info->remote_ifindex = rd->remote_ifindex; 301 memcpy(fdb_info->eth_addr, fdb->key.eth_addr, ETH_ALEN); 302 fdb_info->vni = fdb->key.vni; 303 fdb_info->offloaded = rd->offloaded; 304 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 305 } 306 307 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 308 struct vxlan_fdb *fdb, 309 struct vxlan_rdst *rd, 310 bool adding, 311 struct netlink_ext_ack *extack) 312 { 313 struct switchdev_notifier_vxlan_fdb_info info; 314 enum switchdev_notifier_type notifier_type; 315 int ret; 316 317 if (WARN_ON(!rd)) 318 return 0; 319 320 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 321 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 322 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 323 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 324 &info.info, extack); 325 return notifier_to_errno(ret); 326 } 327 328 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 329 struct vxlan_rdst *rd, int type, bool swdev_notify, 330 struct netlink_ext_ack *extack) 331 { 332 int err; 333 334 if (swdev_notify && rd) { 335 switch (type) { 336 case RTM_NEWNEIGH: 337 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 338 true, extack); 339 if (err) 340 return err; 341 break; 342 case RTM_DELNEIGH: 343 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 344 false, extack); 345 break; 346 } 347 } 348 349 __vxlan_fdb_notify(vxlan, fdb, rd, type); 350 return 0; 351 } 352 353 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 354 { 355 struct vxlan_dev *vxlan = netdev_priv(dev); 356 struct vxlan_fdb f = { 357 .state = NUD_STALE, 358 }; 359 struct vxlan_rdst remote = { 360 .remote_ip = *ipa, /* goes to NDA_DST */ 361 .remote_vni = cpu_to_be32(VXLAN_N_VID), 362 }; 363 364 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 365 } 366 367 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 368 { 369 struct vxlan_fdb f = { 370 .state = NUD_STALE, 371 }; 372 struct vxlan_rdst remote = { }; 373 374 memcpy(f.key.eth_addr, eth_addr, ETH_ALEN); 375 376 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 377 } 378 379 /* Look up Ethernet address in forwarding table */ 380 static struct vxlan_fdb *vxlan_find_mac_rcu(struct vxlan_dev *vxlan, 381 const u8 *mac, __be32 vni) 382 { 383 struct vxlan_fdb_key key; 384 385 memset(&key, 0, sizeof(key)); 386 memcpy(key.eth_addr, mac, sizeof(key.eth_addr)); 387 if (!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) 388 key.vni = vxlan->default_dst.remote_vni; 389 else 390 key.vni = vni; 391 392 return rhashtable_lookup(&vxlan->fdb_hash_tbl, &key, 393 vxlan_fdb_rht_params); 394 } 395 396 static struct vxlan_fdb *vxlan_find_mac_tx(struct vxlan_dev *vxlan, 397 const u8 *mac, __be32 vni) 398 { 399 struct vxlan_fdb *f; 400 401 f = vxlan_find_mac_rcu(vxlan, mac, vni); 402 if (f) { 403 unsigned long now = jiffies; 404 405 if (READ_ONCE(f->used) != now) 406 WRITE_ONCE(f->used, now); 407 } 408 409 return f; 410 } 411 412 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 413 const u8 *mac, __be32 vni) 414 { 415 struct vxlan_fdb *f; 416 417 lockdep_assert_held_once(&vxlan->hash_lock); 418 419 rcu_read_lock(); 420 f = vxlan_find_mac_rcu(vxlan, mac, vni); 421 rcu_read_unlock(); 422 423 return f; 424 } 425 426 /* caller should hold vxlan->hash_lock */ 427 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 428 union vxlan_addr *ip, __be16 port, 429 __be32 vni, __u32 ifindex) 430 { 431 struct vxlan_rdst *rd; 432 433 list_for_each_entry(rd, &f->remotes, list) { 434 if (vxlan_addr_equal(&rd->remote_ip, ip) && 435 rd->remote_port == port && 436 rd->remote_vni == vni && 437 rd->remote_ifindex == ifindex) 438 return rd; 439 } 440 441 return NULL; 442 } 443 444 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 445 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 446 { 447 struct vxlan_dev *vxlan = netdev_priv(dev); 448 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 449 struct vxlan_rdst *rdst = NULL; 450 struct vxlan_fdb *f; 451 int rc = 0; 452 453 if (is_multicast_ether_addr(mac) || 454 is_zero_ether_addr(mac)) 455 return -EINVAL; 456 457 ether_addr_copy(eth_addr, mac); 458 459 rcu_read_lock(); 460 461 f = vxlan_find_mac_rcu(vxlan, eth_addr, vni); 462 if (f) 463 rdst = first_remote_rcu(f); 464 if (!rdst) { 465 rc = -ENOENT; 466 goto out; 467 } 468 469 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 470 471 out: 472 rcu_read_unlock(); 473 return rc; 474 } 475 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 476 477 static int vxlan_fdb_notify_one(struct notifier_block *nb, 478 const struct vxlan_dev *vxlan, 479 const struct vxlan_fdb *f, 480 const struct vxlan_rdst *rdst, 481 struct netlink_ext_ack *extack) 482 { 483 struct switchdev_notifier_vxlan_fdb_info fdb_info; 484 int rc; 485 486 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 487 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 488 &fdb_info); 489 return notifier_to_errno(rc); 490 } 491 492 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 493 struct notifier_block *nb, 494 struct netlink_ext_ack *extack) 495 { 496 struct vxlan_dev *vxlan; 497 struct vxlan_rdst *rdst; 498 struct vxlan_fdb *f; 499 int rc = 0; 500 501 if (!netif_is_vxlan(dev)) 502 return -EINVAL; 503 vxlan = netdev_priv(dev); 504 505 spin_lock_bh(&vxlan->hash_lock); 506 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) { 507 if (f->key.vni == vni) { 508 list_for_each_entry(rdst, &f->remotes, list) { 509 rc = vxlan_fdb_notify_one(nb, vxlan, f, rdst, 510 extack); 511 if (rc) 512 goto unlock; 513 } 514 } 515 } 516 spin_unlock_bh(&vxlan->hash_lock); 517 return 0; 518 519 unlock: 520 spin_unlock_bh(&vxlan->hash_lock); 521 return rc; 522 } 523 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 524 525 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 526 { 527 struct vxlan_dev *vxlan; 528 struct vxlan_rdst *rdst; 529 struct vxlan_fdb *f; 530 531 if (!netif_is_vxlan(dev)) 532 return; 533 vxlan = netdev_priv(dev); 534 535 spin_lock_bh(&vxlan->hash_lock); 536 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) { 537 if (f->key.vni == vni) { 538 list_for_each_entry(rdst, &f->remotes, list) 539 rdst->offloaded = false; 540 } 541 } 542 spin_unlock_bh(&vxlan->hash_lock); 543 544 } 545 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 546 547 /* Replace destination of unicast mac */ 548 static int vxlan_fdb_replace(struct vxlan_fdb *f, 549 union vxlan_addr *ip, __be16 port, __be32 vni, 550 __u32 ifindex, struct vxlan_rdst *oldrd) 551 { 552 struct vxlan_rdst *rd; 553 554 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 555 if (rd) 556 return 0; 557 558 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 559 if (!rd) 560 return 0; 561 562 *oldrd = *rd; 563 dst_cache_reset(&rd->dst_cache); 564 rd->remote_ip = *ip; 565 rd->remote_port = port; 566 rd->remote_vni = vni; 567 rd->remote_ifindex = ifindex; 568 rd->offloaded = false; 569 return 1; 570 } 571 572 /* Add/update destinations for multicast */ 573 static int vxlan_fdb_append(struct vxlan_fdb *f, 574 union vxlan_addr *ip, __be16 port, __be32 vni, 575 __u32 ifindex, struct vxlan_rdst **rdp) 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 = kmalloc(sizeof(*rd), GFP_ATOMIC); 584 if (rd == NULL) 585 return -ENOMEM; 586 587 /* The driver can work correctly without a dst cache, so do not treat 588 * dst cache initialization errors as fatal. 589 */ 590 dst_cache_init(&rd->dst_cache, GFP_ATOMIC | __GFP_NOWARN); 591 592 rd->remote_ip = *ip; 593 rd->remote_port = port; 594 rd->offloaded = false; 595 rd->remote_vni = vni; 596 rd->remote_ifindex = ifindex; 597 598 list_add_tail_rcu(&rd->list, &f->remotes); 599 600 *rdp = rd; 601 return 1; 602 } 603 604 static bool vxlan_parse_gpe_proto(const struct vxlanhdr *hdr, __be16 *protocol) 605 { 606 const struct vxlanhdr_gpe *gpe = (const struct vxlanhdr_gpe *)hdr; 607 608 /* Need to have Next Protocol set for interfaces in GPE mode. */ 609 if (!gpe->np_applied) 610 return false; 611 /* "The initial version is 0. If a receiver does not support the 612 * version indicated it MUST drop the packet. 613 */ 614 if (gpe->version != 0) 615 return false; 616 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 617 * processing MUST occur." However, we don't implement OAM 618 * processing, thus drop the packet. 619 */ 620 if (gpe->oam_flag) 621 return false; 622 623 *protocol = tun_p_to_eth_p(gpe->next_protocol); 624 if (!*protocol) 625 return false; 626 627 return true; 628 } 629 630 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 631 unsigned int off, 632 struct vxlanhdr *vh, size_t hdrlen, 633 __be32 vni_field, 634 struct gro_remcsum *grc, 635 bool nopartial) 636 { 637 size_t start, offset; 638 639 if (skb->remcsum_offload) 640 return vh; 641 642 if (!NAPI_GRO_CB(skb)->csum_valid) 643 return NULL; 644 645 start = vxlan_rco_start(vni_field); 646 offset = start + vxlan_rco_offset(vni_field); 647 648 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 649 start, offset, grc, nopartial); 650 651 skb->remcsum_offload = 1; 652 653 return vh; 654 } 655 656 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk, 657 struct list_head *head, 658 struct sk_buff *skb, 659 struct gro_remcsum *grc) 660 { 661 struct sk_buff *p; 662 struct vxlanhdr *vh, *vh2; 663 unsigned int hlen, off_vx; 664 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); 665 __be32 flags; 666 667 skb_gro_remcsum_init(grc); 668 669 off_vx = skb_gro_offset(skb); 670 hlen = off_vx + sizeof(*vh); 671 vh = skb_gro_header(skb, hlen, off_vx); 672 if (unlikely(!vh)) 673 return NULL; 674 675 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 676 677 flags = vh->vx_flags; 678 679 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 680 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 681 vh->vx_vni, grc, 682 !!(vs->flags & 683 VXLAN_F_REMCSUM_NOPARTIAL)); 684 685 if (!vh) 686 return NULL; 687 } 688 689 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 690 691 list_for_each_entry(p, head, list) { 692 if (!NAPI_GRO_CB(p)->same_flow) 693 continue; 694 695 vh2 = (struct vxlanhdr *)(p->data + off_vx); 696 if (vh->vx_flags != vh2->vx_flags || 697 vh->vx_vni != vh2->vx_vni) { 698 NAPI_GRO_CB(p)->same_flow = 0; 699 continue; 700 } 701 } 702 703 return vh; 704 } 705 706 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 707 struct list_head *head, 708 struct sk_buff *skb) 709 { 710 struct sk_buff *pp = NULL; 711 struct gro_remcsum grc; 712 int flush = 1; 713 714 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) { 715 pp = call_gro_receive(eth_gro_receive, head, skb); 716 flush = 0; 717 } 718 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 719 return pp; 720 } 721 722 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk, 723 struct list_head *head, 724 struct sk_buff *skb) 725 { 726 const struct packet_offload *ptype; 727 struct sk_buff *pp = NULL; 728 struct gro_remcsum grc; 729 struct vxlanhdr *vh; 730 __be16 protocol; 731 int flush = 1; 732 733 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc); 734 if (vh) { 735 if (!vxlan_parse_gpe_proto(vh, &protocol)) 736 goto out; 737 ptype = gro_find_receive_by_type(protocol); 738 if (!ptype) 739 goto out; 740 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb); 741 flush = 0; 742 } 743 out: 744 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 745 return pp; 746 } 747 748 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 749 { 750 /* Sets 'skb->inner_mac_header' since we are always called with 751 * 'skb->encapsulation' set. 752 */ 753 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 754 } 755 756 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 757 { 758 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff); 759 const struct packet_offload *ptype; 760 int err = -ENOSYS; 761 __be16 protocol; 762 763 if (!vxlan_parse_gpe_proto(vh, &protocol)) 764 return err; 765 ptype = gro_find_complete_by_type(protocol); 766 if (ptype) 767 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 768 return err; 769 } 770 771 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 772 __u16 state, __be32 src_vni, 773 __u16 ndm_flags) 774 { 775 struct vxlan_fdb *f; 776 777 f = kmalloc(sizeof(*f), GFP_ATOMIC); 778 if (!f) 779 return NULL; 780 memset(&f->key, 0, sizeof(f->key)); 781 f->state = state; 782 f->flags = ndm_flags; 783 f->updated = f->used = jiffies; 784 f->key.vni = src_vni; 785 f->nh = NULL; 786 RCU_INIT_POINTER(f->vdev, vxlan); 787 INIT_LIST_HEAD(&f->nh_list); 788 INIT_LIST_HEAD(&f->remotes); 789 memcpy(f->key.eth_addr, mac, ETH_ALEN); 790 791 return f; 792 } 793 794 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 795 u32 nhid, struct netlink_ext_ack *extack) 796 { 797 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 798 struct nexthop *nh; 799 int err = -EINVAL; 800 801 if (old_nh && old_nh->id == nhid) 802 return 0; 803 804 nh = nexthop_find_by_id(vxlan->net, nhid); 805 if (!nh) { 806 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 807 goto err_inval; 808 } 809 810 if (!nexthop_get(nh)) { 811 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 812 nh = NULL; 813 goto err_inval; 814 } 815 if (!nexthop_is_fdb(nh)) { 816 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 817 goto err_inval; 818 } 819 820 if (!nexthop_is_multipath(nh)) { 821 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 822 goto err_inval; 823 } 824 825 /* check nexthop group family */ 826 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 827 case AF_INET: 828 if (!nexthop_has_v4(nh)) { 829 err = -EAFNOSUPPORT; 830 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 831 goto err_inval; 832 } 833 break; 834 case AF_INET6: 835 if (nexthop_has_v4(nh)) { 836 err = -EAFNOSUPPORT; 837 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 838 goto err_inval; 839 } 840 } 841 842 if (old_nh) { 843 list_del_rcu(&fdb->nh_list); 844 nexthop_put(old_nh); 845 } 846 rcu_assign_pointer(fdb->nh, nh); 847 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 848 return 1; 849 850 err_inval: 851 if (nh) 852 nexthop_put(nh); 853 return err; 854 } 855 856 int vxlan_fdb_create(struct vxlan_dev *vxlan, 857 const u8 *mac, union vxlan_addr *ip, 858 __u16 state, __be16 port, __be32 src_vni, 859 __be32 vni, __u32 ifindex, __u16 ndm_flags, 860 u32 nhid, struct vxlan_fdb **fdb, 861 struct netlink_ext_ack *extack) 862 { 863 struct vxlan_rdst *rd = NULL; 864 struct vxlan_fdb *f; 865 int rc; 866 867 if (vxlan->cfg.addrmax && 868 vxlan->addrcnt >= vxlan->cfg.addrmax) 869 return -ENOSPC; 870 871 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 872 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 873 if (!f) 874 return -ENOMEM; 875 876 if (nhid) 877 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 878 else 879 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 880 if (rc < 0) 881 goto errout; 882 883 rc = rhashtable_lookup_insert_fast(&vxlan->fdb_hash_tbl, &f->rhnode, 884 vxlan_fdb_rht_params); 885 if (rc) 886 goto destroy_remote; 887 888 ++vxlan->addrcnt; 889 hlist_add_head_rcu(&f->fdb_node, &vxlan->fdb_list); 890 891 *fdb = f; 892 893 return 0; 894 895 destroy_remote: 896 if (rcu_access_pointer(f->nh)) { 897 list_del_rcu(&f->nh_list); 898 nexthop_put(rtnl_dereference(f->nh)); 899 } else { 900 list_del(&rd->list); 901 dst_cache_destroy(&rd->dst_cache); 902 kfree(rd); 903 } 904 errout: 905 kfree(f); 906 return rc; 907 } 908 909 static void __vxlan_fdb_free(struct vxlan_fdb *f) 910 { 911 struct vxlan_rdst *rd, *nd; 912 struct nexthop *nh; 913 914 nh = rcu_dereference_raw(f->nh); 915 if (nh) { 916 rcu_assign_pointer(f->nh, NULL); 917 rcu_assign_pointer(f->vdev, NULL); 918 nexthop_put(nh); 919 } 920 921 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 922 dst_cache_destroy(&rd->dst_cache); 923 kfree(rd); 924 } 925 kfree(f); 926 } 927 928 static void vxlan_fdb_free(struct rcu_head *head) 929 { 930 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 931 932 __vxlan_fdb_free(f); 933 } 934 935 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 936 bool do_notify, bool swdev_notify) 937 { 938 struct vxlan_rdst *rd; 939 940 netdev_dbg(vxlan->dev, "delete %pM\n", f->key.eth_addr); 941 942 --vxlan->addrcnt; 943 if (do_notify) { 944 if (rcu_access_pointer(f->nh)) 945 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 946 swdev_notify, NULL); 947 else 948 list_for_each_entry(rd, &f->remotes, list) 949 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 950 swdev_notify, NULL); 951 } 952 953 hlist_del_init_rcu(&f->fdb_node); 954 rhashtable_remove_fast(&vxlan->fdb_hash_tbl, &f->rhnode, 955 vxlan_fdb_rht_params); 956 list_del_rcu(&f->nh_list); 957 call_rcu(&f->rcu, vxlan_fdb_free); 958 } 959 960 static void vxlan_dst_free(struct rcu_head *head) 961 { 962 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 963 964 dst_cache_destroy(&rd->dst_cache); 965 kfree(rd); 966 } 967 968 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 969 union vxlan_addr *ip, 970 __u16 state, __u16 flags, 971 __be16 port, __be32 vni, 972 __u32 ifindex, __u16 ndm_flags, 973 struct vxlan_fdb *f, u32 nhid, 974 bool swdev_notify, 975 struct netlink_ext_ack *extack) 976 { 977 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 978 struct vxlan_rdst *rd = NULL; 979 struct vxlan_rdst oldrd; 980 int notify = 0; 981 int rc = 0; 982 int err; 983 984 if (nhid && !rcu_access_pointer(f->nh)) { 985 NL_SET_ERR_MSG(extack, 986 "Cannot replace an existing non nexthop fdb with a nexthop"); 987 return -EOPNOTSUPP; 988 } 989 990 if (nhid && (flags & NLM_F_APPEND)) { 991 NL_SET_ERR_MSG(extack, 992 "Cannot append to a nexthop fdb"); 993 return -EOPNOTSUPP; 994 } 995 996 /* Do not allow an externally learned entry to take over an entry added 997 * by the user. 998 */ 999 if (!(fdb_flags & NTF_EXT_LEARNED) || 1000 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 1001 if (f->state != state) { 1002 f->state = state; 1003 notify = 1; 1004 } 1005 if (f->flags != fdb_flags) { 1006 f->flags = fdb_flags; 1007 notify = 1; 1008 } 1009 } 1010 1011 if ((flags & NLM_F_REPLACE)) { 1012 /* Only change unicasts */ 1013 if (!(is_multicast_ether_addr(f->key.eth_addr) || 1014 is_zero_ether_addr(f->key.eth_addr))) { 1015 if (nhid) { 1016 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 1017 if (rc < 0) 1018 return rc; 1019 } else { 1020 rc = vxlan_fdb_replace(f, ip, port, vni, 1021 ifindex, &oldrd); 1022 } 1023 notify |= rc; 1024 } else { 1025 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 1026 return -EOPNOTSUPP; 1027 } 1028 } 1029 if ((flags & NLM_F_APPEND) && 1030 (is_multicast_ether_addr(f->key.eth_addr) || 1031 is_zero_ether_addr(f->key.eth_addr))) { 1032 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 1033 1034 if (rc < 0) 1035 return rc; 1036 notify |= rc; 1037 } 1038 1039 if (ndm_flags & NTF_USE) 1040 WRITE_ONCE(f->updated, jiffies); 1041 1042 if (notify) { 1043 if (rd == NULL) 1044 rd = first_remote_rtnl(f); 1045 1046 WRITE_ONCE(f->updated, jiffies); 1047 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 1048 swdev_notify, extack); 1049 if (err) 1050 goto err_notify; 1051 } 1052 1053 return 0; 1054 1055 err_notify: 1056 if (nhid) 1057 return err; 1058 if ((flags & NLM_F_REPLACE) && rc) 1059 *rd = oldrd; 1060 else if ((flags & NLM_F_APPEND) && rc) { 1061 list_del_rcu(&rd->list); 1062 call_rcu(&rd->rcu, vxlan_dst_free); 1063 } 1064 return err; 1065 } 1066 1067 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1068 const u8 *mac, union vxlan_addr *ip, 1069 __u16 state, __u16 flags, 1070 __be16 port, __be32 src_vni, __be32 vni, 1071 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1072 bool swdev_notify, 1073 struct netlink_ext_ack *extack) 1074 { 1075 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1076 struct vxlan_fdb *f; 1077 int rc; 1078 1079 /* Disallow replace to add a multicast entry */ 1080 if ((flags & NLM_F_REPLACE) && 1081 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1082 return -EOPNOTSUPP; 1083 1084 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1085 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1086 vni, ifindex, fdb_flags, nhid, &f, extack); 1087 if (rc < 0) 1088 return rc; 1089 1090 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1091 swdev_notify, extack); 1092 if (rc) 1093 goto err_notify; 1094 1095 return 0; 1096 1097 err_notify: 1098 vxlan_fdb_destroy(vxlan, f, false, false); 1099 return rc; 1100 } 1101 1102 /* Add new entry to forwarding table -- assumes lock held */ 1103 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1104 const u8 *mac, union vxlan_addr *ip, 1105 __u16 state, __u16 flags, 1106 __be16 port, __be32 src_vni, __be32 vni, 1107 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1108 bool swdev_notify, 1109 struct netlink_ext_ack *extack) 1110 { 1111 struct vxlan_fdb *f; 1112 1113 f = vxlan_find_mac(vxlan, mac, src_vni); 1114 if (f) { 1115 if (flags & NLM_F_EXCL) { 1116 netdev_dbg(vxlan->dev, 1117 "lost race to create %pM\n", mac); 1118 return -EEXIST; 1119 } 1120 1121 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1122 vni, ifindex, ndm_flags, f, 1123 nhid, swdev_notify, extack); 1124 } else { 1125 if (!(flags & NLM_F_CREATE)) 1126 return -ENOENT; 1127 1128 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1129 port, src_vni, vni, ifindex, 1130 ndm_flags, nhid, swdev_notify, 1131 extack); 1132 } 1133 } 1134 1135 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1136 struct vxlan_rdst *rd, bool swdev_notify) 1137 { 1138 list_del_rcu(&rd->list); 1139 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1140 call_rcu(&rd->rcu, vxlan_dst_free); 1141 } 1142 1143 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1144 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1145 __be32 *vni, u32 *ifindex, u32 *nhid, 1146 struct netlink_ext_ack *extack) 1147 { 1148 struct net *net = dev_net(vxlan->dev); 1149 int err; 1150 1151 if (tb[NDA_NH_ID] && 1152 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1153 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1154 return -EINVAL; 1155 } 1156 1157 if (tb[NDA_DST]) { 1158 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1159 if (err) { 1160 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1161 return err; 1162 } 1163 } else { 1164 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1165 1166 if (remote->sa.sa_family == AF_INET) { 1167 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1168 ip->sa.sa_family = AF_INET; 1169 #if IS_ENABLED(CONFIG_IPV6) 1170 } else { 1171 ip->sin6.sin6_addr = in6addr_any; 1172 ip->sa.sa_family = AF_INET6; 1173 #endif 1174 } 1175 } 1176 1177 if (tb[NDA_PORT]) { 1178 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1179 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1180 return -EINVAL; 1181 } 1182 *port = nla_get_be16(tb[NDA_PORT]); 1183 } else { 1184 *port = vxlan->cfg.dst_port; 1185 } 1186 1187 if (tb[NDA_VNI]) { 1188 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1189 NL_SET_ERR_MSG(extack, "Invalid vni"); 1190 return -EINVAL; 1191 } 1192 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1193 } else { 1194 *vni = vxlan->default_dst.remote_vni; 1195 } 1196 1197 if (tb[NDA_SRC_VNI]) { 1198 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1199 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1200 return -EINVAL; 1201 } 1202 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1203 } else { 1204 *src_vni = vxlan->default_dst.remote_vni; 1205 } 1206 1207 if (tb[NDA_IFINDEX]) { 1208 struct net_device *tdev; 1209 1210 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1211 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1212 return -EINVAL; 1213 } 1214 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1215 tdev = __dev_get_by_index(net, *ifindex); 1216 if (!tdev) { 1217 NL_SET_ERR_MSG(extack, "Device not found"); 1218 return -EADDRNOTAVAIL; 1219 } 1220 } else { 1221 *ifindex = 0; 1222 } 1223 1224 *nhid = nla_get_u32_default(tb[NDA_NH_ID], 0); 1225 1226 return 0; 1227 } 1228 1229 /* Add static entry (via netlink) */ 1230 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1231 struct net_device *dev, 1232 const unsigned char *addr, u16 vid, u16 flags, 1233 bool *notified, struct netlink_ext_ack *extack) 1234 { 1235 struct vxlan_dev *vxlan = netdev_priv(dev); 1236 /* struct net *net = dev_net(vxlan->dev); */ 1237 union vxlan_addr ip; 1238 __be16 port; 1239 __be32 src_vni, vni; 1240 u32 ifindex, nhid; 1241 int err; 1242 1243 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1244 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1245 ndm->ndm_state); 1246 return -EINVAL; 1247 } 1248 1249 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1250 return -EINVAL; 1251 1252 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1253 &nhid, extack); 1254 if (err) 1255 return err; 1256 1257 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1258 return -EAFNOSUPPORT; 1259 1260 spin_lock_bh(&vxlan->hash_lock); 1261 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1262 port, src_vni, vni, ifindex, 1263 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1264 nhid, true, extack); 1265 spin_unlock_bh(&vxlan->hash_lock); 1266 1267 if (!err) 1268 *notified = true; 1269 1270 return err; 1271 } 1272 1273 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1274 const unsigned char *addr, union vxlan_addr ip, 1275 __be16 port, __be32 src_vni, __be32 vni, 1276 u32 ifindex, bool swdev_notify) 1277 { 1278 struct vxlan_rdst *rd = NULL; 1279 struct vxlan_fdb *f; 1280 int err = -ENOENT; 1281 1282 f = vxlan_find_mac(vxlan, addr, src_vni); 1283 if (!f) 1284 return err; 1285 1286 if (!vxlan_addr_any(&ip)) { 1287 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1288 if (!rd) 1289 goto out; 1290 } 1291 1292 /* remove a destination if it's not the only one on the list, 1293 * otherwise destroy the fdb entry 1294 */ 1295 if (rd && !list_is_singular(&f->remotes)) { 1296 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1297 goto out; 1298 } 1299 1300 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1301 1302 out: 1303 return 0; 1304 } 1305 1306 /* Delete entry (via netlink) */ 1307 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1308 struct net_device *dev, 1309 const unsigned char *addr, u16 vid, bool *notified, 1310 struct netlink_ext_ack *extack) 1311 { 1312 struct vxlan_dev *vxlan = netdev_priv(dev); 1313 union vxlan_addr ip; 1314 __be32 src_vni, vni; 1315 u32 ifindex, nhid; 1316 __be16 port; 1317 int err; 1318 1319 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1320 &nhid, extack); 1321 if (err) 1322 return err; 1323 1324 spin_lock_bh(&vxlan->hash_lock); 1325 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1326 true); 1327 spin_unlock_bh(&vxlan->hash_lock); 1328 1329 if (!err) 1330 *notified = true; 1331 1332 return err; 1333 } 1334 1335 /* Dump forwarding table */ 1336 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1337 struct net_device *dev, 1338 struct net_device *filter_dev, int *idx) 1339 { 1340 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx; 1341 struct vxlan_dev *vxlan = netdev_priv(dev); 1342 struct vxlan_fdb *f; 1343 int err = 0; 1344 1345 rcu_read_lock(); 1346 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 1347 struct vxlan_rdst *rd; 1348 1349 if (rcu_access_pointer(f->nh)) { 1350 if (*idx < ctx->fdb_idx) 1351 goto skip_nh; 1352 err = vxlan_fdb_info(skb, vxlan, f, 1353 NETLINK_CB(cb->skb).portid, 1354 cb->nlh->nlmsg_seq, 1355 RTM_NEWNEIGH, NLM_F_MULTI, NULL); 1356 if (err < 0) { 1357 rcu_read_unlock(); 1358 goto out; 1359 } 1360 skip_nh: 1361 *idx += 1; 1362 continue; 1363 } 1364 1365 list_for_each_entry_rcu(rd, &f->remotes, list) { 1366 if (*idx < ctx->fdb_idx) 1367 goto skip; 1368 1369 err = vxlan_fdb_info(skb, vxlan, f, 1370 NETLINK_CB(cb->skb).portid, 1371 cb->nlh->nlmsg_seq, 1372 RTM_NEWNEIGH, NLM_F_MULTI, rd); 1373 if (err < 0) { 1374 rcu_read_unlock(); 1375 goto out; 1376 } 1377 skip: 1378 *idx += 1; 1379 } 1380 } 1381 rcu_read_unlock(); 1382 out: 1383 return err; 1384 } 1385 1386 static int vxlan_fdb_get(struct sk_buff *skb, 1387 struct nlattr *tb[], 1388 struct net_device *dev, 1389 const unsigned char *addr, 1390 u16 vid, u32 portid, u32 seq, 1391 struct netlink_ext_ack *extack) 1392 { 1393 struct vxlan_dev *vxlan = netdev_priv(dev); 1394 struct vxlan_fdb *f; 1395 __be32 vni; 1396 int err; 1397 1398 if (tb[NDA_VNI]) 1399 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1400 else 1401 vni = vxlan->default_dst.remote_vni; 1402 1403 rcu_read_lock(); 1404 1405 f = vxlan_find_mac_rcu(vxlan, addr, vni); 1406 if (!f) { 1407 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1408 err = -ENOENT; 1409 goto errout; 1410 } 1411 1412 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1413 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1414 errout: 1415 rcu_read_unlock(); 1416 return err; 1417 } 1418 1419 /* Watch incoming packets to learn mapping between Ethernet address 1420 * and Tunnel endpoint. 1421 */ 1422 static enum skb_drop_reason vxlan_snoop(struct net_device *dev, 1423 union vxlan_addr *src_ip, 1424 const u8 *src_mac, u32 src_ifindex, 1425 __be32 vni) 1426 { 1427 struct vxlan_dev *vxlan = netdev_priv(dev); 1428 struct vxlan_fdb *f; 1429 u32 ifindex = 0; 1430 1431 /* Ignore packets from invalid src-address */ 1432 if (!is_valid_ether_addr(src_mac)) 1433 return SKB_DROP_REASON_MAC_INVALID_SOURCE; 1434 1435 #if IS_ENABLED(CONFIG_IPV6) 1436 if (src_ip->sa.sa_family == AF_INET6 && 1437 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1438 ifindex = src_ifindex; 1439 #endif 1440 1441 f = vxlan_find_mac_rcu(vxlan, src_mac, vni); 1442 if (likely(f)) { 1443 struct vxlan_rdst *rdst = first_remote_rcu(f); 1444 unsigned long now = jiffies; 1445 1446 if (READ_ONCE(f->updated) != now) 1447 WRITE_ONCE(f->updated, now); 1448 1449 /* Don't override an fdb with nexthop with a learnt entry */ 1450 if (rcu_access_pointer(f->nh)) 1451 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1452 1453 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1454 rdst->remote_ifindex == ifindex)) 1455 return SKB_NOT_DROPPED_YET; 1456 1457 /* Don't migrate static entries, drop packets */ 1458 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1459 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS; 1460 1461 if (net_ratelimit()) 1462 netdev_info(dev, 1463 "%pM migrated from %pIS to %pIS\n", 1464 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1465 1466 rdst->remote_ip = *src_ip; 1467 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1468 } else { 1469 /* learned new entry */ 1470 spin_lock(&vxlan->hash_lock); 1471 1472 /* close off race between vxlan_flush and incoming packets */ 1473 if (netif_running(dev)) 1474 vxlan_fdb_update(vxlan, src_mac, src_ip, 1475 NUD_REACHABLE, 1476 NLM_F_EXCL|NLM_F_CREATE, 1477 vxlan->cfg.dst_port, 1478 vni, 1479 vxlan->default_dst.remote_vni, 1480 ifindex, NTF_SELF, 0, true, NULL); 1481 spin_unlock(&vxlan->hash_lock); 1482 } 1483 1484 return SKB_NOT_DROPPED_YET; 1485 } 1486 1487 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1488 { 1489 ASSERT_RTNL(); 1490 1491 if (!vs) 1492 return false; 1493 if (!refcount_dec_and_test(&vs->refcnt)) 1494 return false; 1495 1496 hlist_del_rcu(&vs->hlist); 1497 udp_tunnel_notify_del_rx_port(vs->sock, 1498 (vs->flags & VXLAN_F_GPE) ? 1499 UDP_TUNNEL_TYPE_VXLAN_GPE : 1500 UDP_TUNNEL_TYPE_VXLAN); 1501 1502 return true; 1503 } 1504 1505 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1506 { 1507 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1508 #if IS_ENABLED(CONFIG_IPV6) 1509 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1510 1511 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1512 #endif 1513 1514 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1515 synchronize_net(); 1516 1517 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1518 vxlan_vs_del_vnigrp(vxlan); 1519 else 1520 vxlan_vs_del_dev(vxlan); 1521 1522 if (__vxlan_sock_release_prep(sock4)) { 1523 udp_tunnel_sock_release(sock4->sock); 1524 kfree(sock4); 1525 } 1526 1527 #if IS_ENABLED(CONFIG_IPV6) 1528 if (__vxlan_sock_release_prep(sock6)) { 1529 udp_tunnel_sock_release(sock6->sock); 1530 kfree(sock6); 1531 } 1532 #endif 1533 } 1534 1535 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags) 1536 { 1537 const struct vxlanhdr *vh = vxlan_hdr(skb); 1538 enum skb_drop_reason reason; 1539 size_t start, offset; 1540 1541 if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1542 return SKB_NOT_DROPPED_YET; 1543 1544 start = vxlan_rco_start(vh->vx_vni); 1545 offset = start + vxlan_rco_offset(vh->vx_vni); 1546 1547 reason = pskb_may_pull_reason(skb, offset + sizeof(u16)); 1548 if (reason) 1549 return reason; 1550 1551 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1552 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1553 return SKB_NOT_DROPPED_YET; 1554 } 1555 1556 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags, 1557 struct vxlan_metadata *md) 1558 { 1559 const struct vxlanhdr *vh = vxlan_hdr(skb); 1560 const struct vxlanhdr_gbp *gbp; 1561 struct metadata_dst *tun_dst; 1562 1563 gbp = (const struct vxlanhdr_gbp *)vh; 1564 1565 if (!(vh->vx_flags & VXLAN_HF_GBP)) 1566 return; 1567 1568 md->gbp = ntohs(gbp->policy_id); 1569 1570 tun_dst = (struct metadata_dst *)skb_dst(skb); 1571 if (tun_dst) { 1572 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, 1573 tun_dst->u.tun_info.key.tun_flags); 1574 tun_dst->u.tun_info.options_len = sizeof(*md); 1575 } 1576 if (gbp->dont_learn) 1577 md->gbp |= VXLAN_GBP_DONT_LEARN; 1578 1579 if (gbp->policy_applied) 1580 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1581 1582 /* In flow-based mode, GBP is carried in dst_metadata */ 1583 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1584 skb->mark = md->gbp; 1585 } 1586 1587 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan, 1588 struct vxlan_sock *vs, 1589 struct sk_buff *skb, __be32 vni) 1590 { 1591 union vxlan_addr saddr; 1592 u32 ifindex = skb->dev->ifindex; 1593 1594 skb_reset_mac_header(skb); 1595 skb->protocol = eth_type_trans(skb, vxlan->dev); 1596 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1597 1598 /* Ignore packet loops (and multicast echo) */ 1599 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1600 return SKB_DROP_REASON_LOCAL_MAC; 1601 1602 /* Get address from the outer IP header */ 1603 if (vxlan_get_sk_family(vs) == AF_INET) { 1604 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1605 saddr.sa.sa_family = AF_INET; 1606 #if IS_ENABLED(CONFIG_IPV6) 1607 } else { 1608 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1609 saddr.sa.sa_family = AF_INET6; 1610 #endif 1611 } 1612 1613 if (!(vxlan->cfg.flags & VXLAN_F_LEARN)) 1614 return SKB_NOT_DROPPED_YET; 1615 1616 return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, 1617 ifindex, vni); 1618 } 1619 1620 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1621 struct sk_buff *skb) 1622 { 1623 int err = 0; 1624 1625 if (vxlan_get_sk_family(vs) == AF_INET) 1626 err = IP_ECN_decapsulate(oiph, skb); 1627 #if IS_ENABLED(CONFIG_IPV6) 1628 else 1629 err = IP6_ECN_decapsulate(oiph, skb); 1630 #endif 1631 1632 if (unlikely(err) && log_ecn_error) { 1633 if (vxlan_get_sk_family(vs) == AF_INET) 1634 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1635 &((struct iphdr *)oiph)->saddr, 1636 ((struct iphdr *)oiph)->tos); 1637 else 1638 net_info_ratelimited("non-ECT from %pI6\n", 1639 &((struct ipv6hdr *)oiph)->saddr); 1640 } 1641 return err <= 1; 1642 } 1643 1644 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1645 { 1646 struct vxlan_vni_node *vninode = NULL; 1647 const struct vxlanhdr *vh; 1648 struct vxlan_dev *vxlan; 1649 struct vxlan_sock *vs; 1650 struct vxlan_metadata _md; 1651 struct vxlan_metadata *md = &_md; 1652 __be16 protocol = htons(ETH_P_TEB); 1653 enum skb_drop_reason reason; 1654 bool raw_proto = false; 1655 void *oiph; 1656 __be32 vni = 0; 1657 int nh; 1658 1659 /* Need UDP and VXLAN header to be present */ 1660 reason = pskb_may_pull_reason(skb, VXLAN_HLEN); 1661 if (reason) 1662 goto drop; 1663 1664 vh = vxlan_hdr(skb); 1665 /* VNI flag always required to be set */ 1666 if (!(vh->vx_flags & VXLAN_HF_VNI)) { 1667 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1668 ntohl(vh->vx_flags), ntohl(vh->vx_vni)); 1669 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1670 /* Return non vxlan pkt */ 1671 goto drop; 1672 } 1673 1674 vs = rcu_dereference_sk_user_data(sk); 1675 if (!vs) 1676 goto drop; 1677 1678 vni = vxlan_vni(vh->vx_vni); 1679 1680 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1681 if (!vxlan) { 1682 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND; 1683 goto drop; 1684 } 1685 1686 if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags || 1687 vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) { 1688 /* If the header uses bits besides those enabled by the 1689 * netdevice configuration, treat this as a malformed packet. 1690 * This behavior diverges from VXLAN RFC (RFC7348) which 1691 * stipulates that bits in reserved in reserved fields are to be 1692 * ignored. The approach here maintains compatibility with 1693 * previous stack code, and also is more robust and provides a 1694 * little more security in adding extensions to VXLAN. 1695 */ 1696 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR; 1697 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1698 DEV_STATS_INC(vxlan->dev, rx_errors); 1699 vxlan_vnifilter_count(vxlan, vni, vninode, 1700 VXLAN_VNI_STATS_RX_ERRORS, 0); 1701 goto drop; 1702 } 1703 1704 if (vxlan->cfg.flags & VXLAN_F_GPE) { 1705 if (!vxlan_parse_gpe_proto(vh, &protocol)) 1706 goto drop; 1707 raw_proto = true; 1708 } 1709 1710 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1711 !net_eq(vxlan->net, dev_net(vxlan->dev)))) { 1712 reason = SKB_DROP_REASON_NOMEM; 1713 goto drop; 1714 } 1715 1716 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) { 1717 reason = vxlan_remcsum(skb, vxlan->cfg.flags); 1718 if (unlikely(reason)) 1719 goto drop; 1720 } 1721 1722 if (vxlan_collect_metadata(vs)) { 1723 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 1724 struct metadata_dst *tun_dst; 1725 1726 __set_bit(IP_TUNNEL_KEY_BIT, flags); 1727 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags, 1728 key32_to_tunnel_id(vni), sizeof(*md)); 1729 1730 if (!tun_dst) { 1731 reason = SKB_DROP_REASON_NOMEM; 1732 goto drop; 1733 } 1734 1735 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1736 1737 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1738 } else { 1739 memset(md, 0, sizeof(*md)); 1740 } 1741 1742 if (vxlan->cfg.flags & VXLAN_F_GBP) 1743 vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md); 1744 /* Note that GBP and GPE can never be active together. This is 1745 * ensured in vxlan_dev_configure. 1746 */ 1747 1748 if (!raw_proto) { 1749 reason = vxlan_set_mac(vxlan, vs, skb, vni); 1750 if (reason) 1751 goto drop; 1752 } else { 1753 skb_reset_mac_header(skb); 1754 skb->dev = vxlan->dev; 1755 skb->pkt_type = PACKET_HOST; 1756 } 1757 1758 /* Save offset of outer header relative to skb->head, 1759 * because we are going to reset the network header to the inner header 1760 * and might change skb->head. 1761 */ 1762 nh = skb_network_header(skb) - skb->head; 1763 1764 skb_reset_network_header(skb); 1765 1766 reason = pskb_inet_may_pull_reason(skb); 1767 if (reason) { 1768 DEV_STATS_INC(vxlan->dev, rx_length_errors); 1769 DEV_STATS_INC(vxlan->dev, rx_errors); 1770 vxlan_vnifilter_count(vxlan, vni, vninode, 1771 VXLAN_VNI_STATS_RX_ERRORS, 0); 1772 goto drop; 1773 } 1774 1775 /* Get the outer header. */ 1776 oiph = skb->head + nh; 1777 1778 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1779 reason = SKB_DROP_REASON_IP_TUNNEL_ECN; 1780 DEV_STATS_INC(vxlan->dev, rx_frame_errors); 1781 DEV_STATS_INC(vxlan->dev, rx_errors); 1782 vxlan_vnifilter_count(vxlan, vni, vninode, 1783 VXLAN_VNI_STATS_RX_ERRORS, 0); 1784 goto drop; 1785 } 1786 1787 rcu_read_lock(); 1788 1789 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1790 rcu_read_unlock(); 1791 dev_dstats_rx_dropped(vxlan->dev); 1792 vxlan_vnifilter_count(vxlan, vni, vninode, 1793 VXLAN_VNI_STATS_RX_DROPS, 0); 1794 reason = SKB_DROP_REASON_DEV_READY; 1795 goto drop; 1796 } 1797 1798 dev_dstats_rx_add(vxlan->dev, skb->len); 1799 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1800 gro_cells_receive(&vxlan->gro_cells, skb); 1801 1802 rcu_read_unlock(); 1803 1804 return 0; 1805 1806 drop: 1807 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED; 1808 /* Consume bad packet */ 1809 kfree_skb_reason(skb, reason); 1810 return 0; 1811 } 1812 1813 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1814 { 1815 struct vxlan_dev *vxlan; 1816 struct vxlan_sock *vs; 1817 struct vxlanhdr *hdr; 1818 __be32 vni; 1819 1820 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1821 return -EINVAL; 1822 1823 hdr = vxlan_hdr(skb); 1824 1825 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1826 return -EINVAL; 1827 1828 vs = rcu_dereference_sk_user_data(sk); 1829 if (!vs) 1830 return -ENOENT; 1831 1832 vni = vxlan_vni(hdr->vx_vni); 1833 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1834 if (!vxlan) 1835 return -ENOENT; 1836 1837 return 0; 1838 } 1839 1840 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1841 { 1842 struct vxlan_dev *vxlan = netdev_priv(dev); 1843 struct arphdr *parp; 1844 u8 *arpptr, *sha; 1845 __be32 sip, tip; 1846 struct neighbour *n; 1847 1848 if (dev->flags & IFF_NOARP) 1849 goto out; 1850 1851 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1852 dev_dstats_tx_dropped(dev); 1853 vxlan_vnifilter_count(vxlan, vni, NULL, 1854 VXLAN_VNI_STATS_TX_DROPS, 0); 1855 goto out; 1856 } 1857 parp = arp_hdr(skb); 1858 1859 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1860 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1861 parp->ar_pro != htons(ETH_P_IP) || 1862 parp->ar_op != htons(ARPOP_REQUEST) || 1863 parp->ar_hln != dev->addr_len || 1864 parp->ar_pln != 4) 1865 goto out; 1866 arpptr = (u8 *)parp + sizeof(struct arphdr); 1867 sha = arpptr; 1868 arpptr += dev->addr_len; /* sha */ 1869 memcpy(&sip, arpptr, sizeof(sip)); 1870 arpptr += sizeof(sip); 1871 arpptr += dev->addr_len; /* tha */ 1872 memcpy(&tip, arpptr, sizeof(tip)); 1873 1874 if (ipv4_is_loopback(tip) || 1875 ipv4_is_multicast(tip)) 1876 goto out; 1877 1878 n = neigh_lookup(&arp_tbl, &tip, dev); 1879 1880 if (n) { 1881 struct vxlan_rdst *rdst = NULL; 1882 struct vxlan_fdb *f; 1883 struct sk_buff *reply; 1884 1885 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1886 neigh_release(n); 1887 goto out; 1888 } 1889 1890 rcu_read_lock(); 1891 f = vxlan_find_mac_tx(vxlan, n->ha, vni); 1892 if (f) 1893 rdst = first_remote_rcu(f); 1894 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 1895 /* bridge-local neighbor */ 1896 neigh_release(n); 1897 rcu_read_unlock(); 1898 goto out; 1899 } 1900 rcu_read_unlock(); 1901 1902 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1903 n->ha, sha); 1904 1905 neigh_release(n); 1906 1907 if (reply == NULL) 1908 goto out; 1909 1910 skb_reset_mac_header(reply); 1911 __skb_pull(reply, skb_network_offset(reply)); 1912 reply->ip_summed = CHECKSUM_UNNECESSARY; 1913 reply->pkt_type = PACKET_HOST; 1914 1915 if (netif_rx(reply) == NET_RX_DROP) { 1916 dev_dstats_rx_dropped(dev); 1917 vxlan_vnifilter_count(vxlan, vni, NULL, 1918 VXLAN_VNI_STATS_RX_DROPS, 0); 1919 } 1920 1921 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1922 union vxlan_addr ipa = { 1923 .sin.sin_addr.s_addr = tip, 1924 .sin.sin_family = AF_INET, 1925 }; 1926 1927 vxlan_ip_miss(dev, &ipa); 1928 } 1929 out: 1930 consume_skb(skb); 1931 return NETDEV_TX_OK; 1932 } 1933 1934 #if IS_ENABLED(CONFIG_IPV6) 1935 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1936 struct neighbour *n, bool isrouter) 1937 { 1938 struct net_device *dev = request->dev; 1939 struct sk_buff *reply; 1940 struct nd_msg *ns, *na; 1941 struct ipv6hdr *pip6; 1942 u8 *daddr; 1943 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1944 int ns_olen; 1945 int i, len; 1946 1947 if (dev == NULL || !pskb_may_pull(request, request->len)) 1948 return NULL; 1949 1950 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1951 sizeof(*na) + na_olen + dev->needed_tailroom; 1952 reply = alloc_skb(len, GFP_ATOMIC); 1953 if (reply == NULL) 1954 return NULL; 1955 1956 reply->protocol = htons(ETH_P_IPV6); 1957 reply->dev = dev; 1958 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1959 skb_push(reply, sizeof(struct ethhdr)); 1960 skb_reset_mac_header(reply); 1961 1962 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1963 1964 daddr = eth_hdr(request)->h_source; 1965 ns_olen = request->len - skb_network_offset(request) - 1966 sizeof(struct ipv6hdr) - sizeof(*ns); 1967 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1968 if (!ns->opt[i + 1]) { 1969 kfree_skb(reply); 1970 return NULL; 1971 } 1972 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1973 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1974 break; 1975 } 1976 } 1977 1978 /* Ethernet header */ 1979 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1980 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1981 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1982 reply->protocol = htons(ETH_P_IPV6); 1983 1984 skb_pull(reply, sizeof(struct ethhdr)); 1985 skb_reset_network_header(reply); 1986 skb_put(reply, sizeof(struct ipv6hdr)); 1987 1988 /* IPv6 header */ 1989 1990 pip6 = ipv6_hdr(reply); 1991 memset(pip6, 0, sizeof(struct ipv6hdr)); 1992 pip6->version = 6; 1993 pip6->priority = ipv6_hdr(request)->priority; 1994 pip6->nexthdr = IPPROTO_ICMPV6; 1995 pip6->hop_limit = 255; 1996 pip6->daddr = ipv6_hdr(request)->saddr; 1997 pip6->saddr = *(struct in6_addr *)n->primary_key; 1998 1999 skb_pull(reply, sizeof(struct ipv6hdr)); 2000 skb_reset_transport_header(reply); 2001 2002 /* Neighbor Advertisement */ 2003 na = skb_put_zero(reply, sizeof(*na) + na_olen); 2004 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 2005 na->icmph.icmp6_router = isrouter; 2006 na->icmph.icmp6_override = 1; 2007 na->icmph.icmp6_solicited = 1; 2008 na->target = ns->target; 2009 ether_addr_copy(&na->opt[2], n->ha); 2010 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 2011 na->opt[1] = na_olen >> 3; 2012 2013 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 2014 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 2015 csum_partial(na, sizeof(*na)+na_olen, 0)); 2016 2017 pip6->payload_len = htons(sizeof(*na)+na_olen); 2018 2019 skb_push(reply, sizeof(struct ipv6hdr)); 2020 2021 reply->ip_summed = CHECKSUM_UNNECESSARY; 2022 2023 return reply; 2024 } 2025 2026 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 2027 { 2028 struct vxlan_dev *vxlan = netdev_priv(dev); 2029 const struct in6_addr *daddr; 2030 const struct ipv6hdr *iphdr; 2031 struct inet6_dev *in6_dev; 2032 struct neighbour *n; 2033 struct nd_msg *msg; 2034 2035 rcu_read_lock(); 2036 in6_dev = __in6_dev_get(dev); 2037 if (!in6_dev) 2038 goto out; 2039 2040 iphdr = ipv6_hdr(skb); 2041 daddr = &iphdr->daddr; 2042 msg = (struct nd_msg *)(iphdr + 1); 2043 2044 if (ipv6_addr_loopback(daddr) || 2045 ipv6_addr_is_multicast(&msg->target)) 2046 goto out; 2047 2048 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 2049 2050 if (n) { 2051 struct vxlan_rdst *rdst = NULL; 2052 struct vxlan_fdb *f; 2053 struct sk_buff *reply; 2054 2055 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 2056 neigh_release(n); 2057 goto out; 2058 } 2059 2060 f = vxlan_find_mac_tx(vxlan, n->ha, vni); 2061 if (f) 2062 rdst = first_remote_rcu(f); 2063 if (rdst && vxlan_addr_any(&rdst->remote_ip)) { 2064 /* bridge-local neighbor */ 2065 neigh_release(n); 2066 goto out; 2067 } 2068 2069 reply = vxlan_na_create(skb, n, 2070 !!(f ? f->flags & NTF_ROUTER : 0)); 2071 2072 neigh_release(n); 2073 2074 if (reply == NULL) 2075 goto out; 2076 2077 if (netif_rx(reply) == NET_RX_DROP) { 2078 dev_dstats_rx_dropped(dev); 2079 vxlan_vnifilter_count(vxlan, vni, NULL, 2080 VXLAN_VNI_STATS_RX_DROPS, 0); 2081 } 2082 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2083 union vxlan_addr ipa = { 2084 .sin6.sin6_addr = msg->target, 2085 .sin6.sin6_family = AF_INET6, 2086 }; 2087 2088 vxlan_ip_miss(dev, &ipa); 2089 } 2090 2091 out: 2092 rcu_read_unlock(); 2093 consume_skb(skb); 2094 return NETDEV_TX_OK; 2095 } 2096 #endif 2097 2098 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2099 { 2100 struct vxlan_dev *vxlan = netdev_priv(dev); 2101 struct neighbour *n; 2102 2103 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2104 return false; 2105 2106 n = NULL; 2107 switch (ntohs(eth_hdr(skb)->h_proto)) { 2108 case ETH_P_IP: 2109 { 2110 struct iphdr *pip; 2111 2112 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2113 return false; 2114 pip = ip_hdr(skb); 2115 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2116 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2117 union vxlan_addr ipa = { 2118 .sin.sin_addr.s_addr = pip->daddr, 2119 .sin.sin_family = AF_INET, 2120 }; 2121 2122 vxlan_ip_miss(dev, &ipa); 2123 return false; 2124 } 2125 2126 break; 2127 } 2128 #if IS_ENABLED(CONFIG_IPV6) 2129 case ETH_P_IPV6: 2130 { 2131 struct ipv6hdr *pip6; 2132 2133 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2134 return false; 2135 pip6 = ipv6_hdr(skb); 2136 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2137 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2138 union vxlan_addr ipa = { 2139 .sin6.sin6_addr = pip6->daddr, 2140 .sin6.sin6_family = AF_INET6, 2141 }; 2142 2143 vxlan_ip_miss(dev, &ipa); 2144 return false; 2145 } 2146 2147 break; 2148 } 2149 #endif 2150 default: 2151 return false; 2152 } 2153 2154 if (n) { 2155 bool diff; 2156 2157 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2158 if (diff) { 2159 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2160 dev->addr_len); 2161 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2162 } 2163 neigh_release(n); 2164 return diff; 2165 } 2166 2167 return false; 2168 } 2169 2170 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol) 2171 { 2172 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2173 2174 gpe->np_applied = 1; 2175 gpe->next_protocol = tun_p_from_eth_p(protocol); 2176 if (!gpe->next_protocol) 2177 return -EPFNOSUPPORT; 2178 return 0; 2179 } 2180 2181 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2182 int iphdr_len, __be32 vni, 2183 struct vxlan_metadata *md, u32 vxflags, 2184 bool udp_sum) 2185 { 2186 struct vxlanhdr *vxh; 2187 int min_headroom; 2188 int err; 2189 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2190 __be16 inner_protocol = htons(ETH_P_TEB); 2191 2192 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2193 skb->ip_summed == CHECKSUM_PARTIAL) { 2194 int csum_start = skb_checksum_start_offset(skb); 2195 2196 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2197 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2198 (skb->csum_offset == offsetof(struct udphdr, check) || 2199 skb->csum_offset == offsetof(struct tcphdr, check))) 2200 type |= SKB_GSO_TUNNEL_REMCSUM; 2201 } 2202 2203 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2204 + VXLAN_HLEN + iphdr_len; 2205 2206 /* Need space for new headers (invalidates iph ptr) */ 2207 err = skb_cow_head(skb, min_headroom); 2208 if (unlikely(err)) 2209 return err; 2210 2211 err = iptunnel_handle_offloads(skb, type); 2212 if (err) 2213 return err; 2214 2215 vxh = __skb_push(skb, sizeof(*vxh)); 2216 vxh->vx_flags = VXLAN_HF_VNI; 2217 vxh->vx_vni = vxlan_vni_field(vni); 2218 2219 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2220 unsigned int start; 2221 2222 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2223 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2224 vxh->vx_flags |= VXLAN_HF_RCO; 2225 2226 if (!skb_is_gso(skb)) { 2227 skb->ip_summed = CHECKSUM_NONE; 2228 skb->encapsulation = 0; 2229 } 2230 } 2231 2232 if (vxflags & VXLAN_F_GBP) 2233 vxlan_build_gbp_hdr(vxh, md); 2234 if (vxflags & VXLAN_F_GPE) { 2235 err = vxlan_build_gpe_hdr(vxh, skb->protocol); 2236 if (err < 0) 2237 return err; 2238 inner_protocol = skb->protocol; 2239 } 2240 2241 skb_set_inner_protocol(skb, inner_protocol); 2242 return 0; 2243 } 2244 2245 /* Bypass encapsulation if the destination is local */ 2246 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2247 struct vxlan_dev *dst_vxlan, __be32 vni, 2248 bool snoop) 2249 { 2250 union vxlan_addr loopback; 2251 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2252 unsigned int len = skb->len; 2253 struct net_device *dev; 2254 2255 skb->pkt_type = PACKET_HOST; 2256 skb->encapsulation = 0; 2257 skb->dev = dst_vxlan->dev; 2258 __skb_pull(skb, skb_network_offset(skb)); 2259 2260 if (remote_ip->sa.sa_family == AF_INET) { 2261 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2262 loopback.sa.sa_family = AF_INET; 2263 #if IS_ENABLED(CONFIG_IPV6) 2264 } else { 2265 loopback.sin6.sin6_addr = in6addr_loopback; 2266 loopback.sa.sa_family = AF_INET6; 2267 #endif 2268 } 2269 2270 rcu_read_lock(); 2271 dev = skb->dev; 2272 if (unlikely(!(dev->flags & IFF_UP))) { 2273 kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY); 2274 goto drop; 2275 } 2276 2277 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2278 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2279 2280 dev_dstats_tx_add(src_vxlan->dev, len); 2281 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2282 2283 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2284 dev_dstats_rx_add(dst_vxlan->dev, len); 2285 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2286 len); 2287 } else { 2288 drop: 2289 dev_dstats_rx_dropped(dev); 2290 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2291 VXLAN_VNI_STATS_RX_DROPS, 0); 2292 } 2293 rcu_read_unlock(); 2294 } 2295 2296 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2297 struct vxlan_dev *vxlan, 2298 int addr_family, 2299 __be16 dst_port, int dst_ifindex, __be32 vni, 2300 struct dst_entry *dst, 2301 u32 rt_flags) 2302 { 2303 #if IS_ENABLED(CONFIG_IPV6) 2304 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2305 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2306 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2307 */ 2308 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2309 #endif 2310 /* Bypass encapsulation if the destination is local */ 2311 if (rt_flags & RTCF_LOCAL && 2312 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && 2313 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) { 2314 struct vxlan_dev *dst_vxlan; 2315 2316 dst_release(dst); 2317 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2318 addr_family, dst_port, 2319 vxlan->cfg.flags); 2320 if (!dst_vxlan) { 2321 DEV_STATS_INC(dev, tx_errors); 2322 vxlan_vnifilter_count(vxlan, vni, NULL, 2323 VXLAN_VNI_STATS_TX_ERRORS, 0); 2324 kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND); 2325 2326 return -ENOENT; 2327 } 2328 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2329 return 1; 2330 } 2331 2332 return 0; 2333 } 2334 2335 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2336 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2337 { 2338 struct dst_cache *dst_cache; 2339 struct ip_tunnel_info *info; 2340 struct ip_tunnel_key *pkey; 2341 struct ip_tunnel_key key; 2342 struct vxlan_dev *vxlan = netdev_priv(dev); 2343 const struct iphdr *old_iph; 2344 struct vxlan_metadata _md; 2345 struct vxlan_metadata *md = &_md; 2346 unsigned int pkt_len = skb->len; 2347 __be16 src_port = 0, dst_port; 2348 struct dst_entry *ndst = NULL; 2349 int addr_family; 2350 __u8 tos, ttl; 2351 int ifindex; 2352 int err; 2353 u32 flags = vxlan->cfg.flags; 2354 bool use_cache; 2355 bool udp_sum = false; 2356 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2357 enum skb_drop_reason reason; 2358 bool no_eth_encap; 2359 __be32 vni = 0; 2360 2361 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB); 2362 reason = skb_vlan_inet_prepare(skb, no_eth_encap); 2363 if (reason) 2364 goto drop; 2365 2366 reason = SKB_DROP_REASON_NOT_SPECIFIED; 2367 old_iph = ip_hdr(skb); 2368 2369 info = skb_tunnel_info(skb); 2370 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2371 2372 if (rdst) { 2373 memset(&key, 0, sizeof(key)); 2374 pkey = &key; 2375 2376 if (vxlan_addr_any(&rdst->remote_ip)) { 2377 if (did_rsc) { 2378 /* short-circuited back to local bridge */ 2379 vxlan_encap_bypass(skb, vxlan, vxlan, 2380 default_vni, true); 2381 return; 2382 } 2383 goto drop; 2384 } 2385 2386 addr_family = vxlan->cfg.saddr.sa.sa_family; 2387 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2388 vni = (rdst->remote_vni) ? : default_vni; 2389 ifindex = rdst->remote_ifindex; 2390 2391 if (addr_family == AF_INET) { 2392 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2393 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2394 } else { 2395 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2396 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2397 } 2398 2399 dst_cache = &rdst->dst_cache; 2400 md->gbp = skb->mark; 2401 if (flags & VXLAN_F_TTL_INHERIT) { 2402 ttl = ip_tunnel_get_ttl(old_iph, skb); 2403 } else { 2404 ttl = vxlan->cfg.ttl; 2405 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2406 ttl = 1; 2407 } 2408 tos = vxlan->cfg.tos; 2409 if (tos == 1) 2410 tos = ip_tunnel_get_dsfield(old_iph, skb); 2411 if (tos && !info) 2412 use_cache = false; 2413 2414 if (addr_family == AF_INET) 2415 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2416 else 2417 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2418 #if IS_ENABLED(CONFIG_IPV6) 2419 switch (vxlan->cfg.label_policy) { 2420 case VXLAN_LABEL_FIXED: 2421 key.label = vxlan->cfg.label; 2422 break; 2423 case VXLAN_LABEL_INHERIT: 2424 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2425 break; 2426 default: 2427 DEBUG_NET_WARN_ON_ONCE(1); 2428 goto drop; 2429 } 2430 #endif 2431 } else { 2432 if (!info) { 2433 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2434 dev->name); 2435 goto drop; 2436 } 2437 pkey = &info->key; 2438 addr_family = ip_tunnel_info_af(info); 2439 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2440 vni = tunnel_id_to_key32(info->key.tun_id); 2441 ifindex = 0; 2442 dst_cache = &info->dst_cache; 2443 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2444 if (info->options_len < sizeof(*md)) 2445 goto drop; 2446 md = ip_tunnel_info_opts(info); 2447 } 2448 ttl = info->key.ttl; 2449 tos = info->key.tos; 2450 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2451 } 2452 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2453 vxlan->cfg.port_max, true); 2454 2455 rcu_read_lock(); 2456 if (addr_family == AF_INET) { 2457 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2458 u16 ipcb_flags = 0; 2459 struct rtable *rt; 2460 __be16 df = 0; 2461 __be32 saddr; 2462 2463 if (!ifindex) 2464 ifindex = sock4->sock->sk->sk_bound_dev_if; 2465 2466 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2467 &saddr, pkey, src_port, dst_port, 2468 tos, use_cache ? dst_cache : NULL); 2469 if (IS_ERR(rt)) { 2470 err = PTR_ERR(rt); 2471 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2472 goto tx_error; 2473 } 2474 2475 if (flags & VXLAN_F_MC_ROUTE) 2476 ipcb_flags |= IPSKB_MCROUTE; 2477 2478 if (!info) { 2479 /* Bypass encapsulation if the destination is local */ 2480 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2481 dst_port, ifindex, vni, 2482 &rt->dst, rt->rt_flags); 2483 if (err) 2484 goto out_unlock; 2485 2486 if (vxlan->cfg.df == VXLAN_DF_SET) { 2487 df = htons(IP_DF); 2488 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2489 struct ethhdr *eth = eth_hdr(skb); 2490 2491 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2492 (ntohs(eth->h_proto) == ETH_P_IP && 2493 old_iph->frag_off & htons(IP_DF))) 2494 df = htons(IP_DF); 2495 } 2496 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2497 info->key.tun_flags)) { 2498 df = htons(IP_DF); 2499 } 2500 2501 ndst = &rt->dst; 2502 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2503 netif_is_any_bridge_port(dev)); 2504 if (err < 0) { 2505 goto tx_error; 2506 } else if (err) { 2507 if (info) { 2508 struct ip_tunnel_info *unclone; 2509 2510 unclone = skb_tunnel_info_unclone(skb); 2511 if (unlikely(!unclone)) 2512 goto tx_error; 2513 2514 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2515 unclone->key.u.ipv4.dst = saddr; 2516 } 2517 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2518 dst_release(ndst); 2519 goto out_unlock; 2520 } 2521 2522 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2523 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2524 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2525 vni, md, flags, udp_sum); 2526 if (err < 0) { 2527 reason = SKB_DROP_REASON_NOMEM; 2528 goto tx_error; 2529 } 2530 2531 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2532 pkey->u.ipv4.dst, tos, ttl, df, 2533 src_port, dst_port, xnet, !udp_sum, 2534 ipcb_flags); 2535 #if IS_ENABLED(CONFIG_IPV6) 2536 } else { 2537 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2538 struct in6_addr saddr; 2539 u16 ip6cb_flags = 0; 2540 2541 if (!ifindex) 2542 ifindex = sock6->sock->sk->sk_bound_dev_if; 2543 2544 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2545 ifindex, &saddr, pkey, 2546 src_port, dst_port, tos, 2547 use_cache ? dst_cache : NULL); 2548 if (IS_ERR(ndst)) { 2549 err = PTR_ERR(ndst); 2550 ndst = NULL; 2551 reason = SKB_DROP_REASON_IP_OUTNOROUTES; 2552 goto tx_error; 2553 } 2554 2555 if (flags & VXLAN_F_MC_ROUTE) 2556 ip6cb_flags |= IP6SKB_MCROUTE; 2557 2558 if (!info) { 2559 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2560 2561 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2562 dst_port, ifindex, vni, 2563 ndst, rt6i_flags); 2564 if (err) 2565 goto out_unlock; 2566 } 2567 2568 err = skb_tunnel_check_pmtu(skb, ndst, 2569 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2570 netif_is_any_bridge_port(dev)); 2571 if (err < 0) { 2572 goto tx_error; 2573 } else if (err) { 2574 if (info) { 2575 struct ip_tunnel_info *unclone; 2576 2577 unclone = skb_tunnel_info_unclone(skb); 2578 if (unlikely(!unclone)) 2579 goto tx_error; 2580 2581 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2582 unclone->key.u.ipv6.dst = saddr; 2583 } 2584 2585 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2586 dst_release(ndst); 2587 goto out_unlock; 2588 } 2589 2590 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2591 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2592 skb_scrub_packet(skb, xnet); 2593 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2594 vni, md, flags, udp_sum); 2595 if (err < 0) { 2596 reason = SKB_DROP_REASON_NOMEM; 2597 goto tx_error; 2598 } 2599 2600 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2601 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2602 pkey->label, src_port, dst_port, !udp_sum, 2603 ip6cb_flags); 2604 #endif 2605 } 2606 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2607 out_unlock: 2608 rcu_read_unlock(); 2609 return; 2610 2611 drop: 2612 dev_dstats_tx_dropped(dev); 2613 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2614 kfree_skb_reason(skb, reason); 2615 return; 2616 2617 tx_error: 2618 rcu_read_unlock(); 2619 if (err == -ELOOP) 2620 DEV_STATS_INC(dev, collisions); 2621 else if (err == -ENETUNREACH) 2622 DEV_STATS_INC(dev, tx_carrier_errors); 2623 dst_release(ndst); 2624 DEV_STATS_INC(dev, tx_errors); 2625 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2626 kfree_skb_reason(skb, reason); 2627 } 2628 2629 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2630 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2631 { 2632 struct vxlan_rdst nh_rdst; 2633 struct nexthop *nh; 2634 bool do_xmit; 2635 u32 hash; 2636 2637 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2638 hash = skb_get_hash(skb); 2639 2640 nh = rcu_dereference(f->nh); 2641 if (!nh) 2642 goto drop; 2643 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2644 2645 if (likely(do_xmit)) 2646 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2647 else 2648 goto drop; 2649 2650 return; 2651 2652 drop: 2653 dev_dstats_tx_dropped(dev); 2654 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2655 VXLAN_VNI_STATS_TX_DROPS, 0); 2656 dev_kfree_skb(skb); 2657 } 2658 2659 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev, 2660 u32 nhid, __be32 vni) 2661 { 2662 struct vxlan_dev *vxlan = netdev_priv(dev); 2663 struct vxlan_rdst nh_rdst; 2664 struct nexthop *nh; 2665 bool do_xmit; 2666 u32 hash; 2667 2668 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2669 hash = skb_get_hash(skb); 2670 2671 rcu_read_lock(); 2672 nh = nexthop_find_by_id(dev_net(dev), nhid); 2673 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) { 2674 rcu_read_unlock(); 2675 goto drop; 2676 } 2677 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2678 rcu_read_unlock(); 2679 2680 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family) 2681 goto drop; 2682 2683 if (likely(do_xmit)) 2684 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false); 2685 else 2686 goto drop; 2687 2688 return NETDEV_TX_OK; 2689 2690 drop: 2691 dev_dstats_tx_dropped(dev); 2692 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2693 VXLAN_VNI_STATS_TX_DROPS, 0); 2694 dev_kfree_skb(skb); 2695 return NETDEV_TX_OK; 2696 } 2697 2698 /* Transmit local packets over Vxlan 2699 * 2700 * Outer IP header inherits ECN and DF from inner header. 2701 * Outer UDP destination is the VXLAN assigned port. 2702 * source port is based on hash of flow 2703 */ 2704 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2705 { 2706 struct vxlan_dev *vxlan = netdev_priv(dev); 2707 struct vxlan_rdst *rdst, *fdst = NULL; 2708 const struct ip_tunnel_info *info; 2709 struct vxlan_fdb *f; 2710 struct ethhdr *eth; 2711 __be32 vni = 0; 2712 u32 nhid = 0; 2713 bool did_rsc; 2714 2715 info = skb_tunnel_info(skb); 2716 2717 skb_reset_mac_header(skb); 2718 2719 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2720 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2721 info->mode & IP_TUNNEL_INFO_TX) { 2722 vni = tunnel_id_to_key32(info->key.tun_id); 2723 nhid = info->key.nhid; 2724 } else { 2725 if (info && info->mode & IP_TUNNEL_INFO_TX) 2726 vxlan_xmit_one(skb, dev, vni, NULL, false); 2727 else 2728 kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO); 2729 return NETDEV_TX_OK; 2730 } 2731 } 2732 2733 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2734 eth = eth_hdr(skb); 2735 if (ntohs(eth->h_proto) == ETH_P_ARP) 2736 return arp_reduce(dev, skb, vni); 2737 #if IS_ENABLED(CONFIG_IPV6) 2738 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2739 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2740 sizeof(struct nd_msg)) && 2741 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2742 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2743 2744 if (m->icmph.icmp6_code == 0 && 2745 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2746 return neigh_reduce(dev, skb, vni); 2747 } 2748 #endif 2749 } 2750 2751 if (nhid) 2752 return vxlan_xmit_nhid(skb, dev, nhid, vni); 2753 2754 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2755 struct vxlan_mdb_entry *mdb_entry; 2756 2757 rcu_read_lock(); 2758 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2759 if (mdb_entry) { 2760 netdev_tx_t ret; 2761 2762 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2763 rcu_read_unlock(); 2764 return ret; 2765 } 2766 rcu_read_unlock(); 2767 } 2768 2769 eth = eth_hdr(skb); 2770 rcu_read_lock(); 2771 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2772 did_rsc = false; 2773 2774 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2775 (ntohs(eth->h_proto) == ETH_P_IP || 2776 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2777 did_rsc = route_shortcircuit(dev, skb); 2778 if (did_rsc) 2779 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni); 2780 } 2781 2782 if (f == NULL) { 2783 f = vxlan_find_mac_tx(vxlan, all_zeros_mac, vni); 2784 if (f == NULL) { 2785 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2786 !is_multicast_ether_addr(eth->h_dest)) 2787 vxlan_fdb_miss(vxlan, eth->h_dest); 2788 2789 dev_dstats_tx_dropped(dev); 2790 vxlan_vnifilter_count(vxlan, vni, NULL, 2791 VXLAN_VNI_STATS_TX_DROPS, 0); 2792 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2793 goto out; 2794 } 2795 } 2796 2797 if (rcu_access_pointer(f->nh)) { 2798 vxlan_xmit_nh(skb, dev, f, 2799 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2800 } else { 2801 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2802 struct sk_buff *skb1; 2803 2804 if (!fdst) { 2805 fdst = rdst; 2806 continue; 2807 } 2808 skb1 = skb_clone(skb, GFP_ATOMIC); 2809 if (skb1) 2810 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2811 } 2812 if (fdst) 2813 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2814 else 2815 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET); 2816 } 2817 2818 out: 2819 rcu_read_unlock(); 2820 return NETDEV_TX_OK; 2821 } 2822 2823 /* Walk the forwarding table and purge stale entries */ 2824 static void vxlan_cleanup(struct timer_list *t) 2825 { 2826 struct vxlan_dev *vxlan = timer_container_of(vxlan, t, age_timer); 2827 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2828 struct vxlan_fdb *f; 2829 2830 if (!netif_running(vxlan->dev)) 2831 return; 2832 2833 rcu_read_lock(); 2834 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 2835 unsigned long timeout; 2836 2837 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2838 continue; 2839 2840 if (f->flags & NTF_EXT_LEARNED) 2841 continue; 2842 2843 timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ; 2844 if (time_before_eq(timeout, jiffies)) { 2845 spin_lock(&vxlan->hash_lock); 2846 if (!hlist_unhashed(&f->fdb_node)) { 2847 netdev_dbg(vxlan->dev, "garbage collect %pM\n", 2848 f->key.eth_addr); 2849 f->state = NUD_STALE; 2850 vxlan_fdb_destroy(vxlan, f, true, true); 2851 } 2852 spin_unlock(&vxlan->hash_lock); 2853 } else if (time_before(timeout, next_timer)) { 2854 next_timer = timeout; 2855 } 2856 } 2857 rcu_read_unlock(); 2858 2859 mod_timer(&vxlan->age_timer, next_timer); 2860 } 2861 2862 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2863 { 2864 ASSERT_RTNL(); 2865 2866 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2867 #if IS_ENABLED(CONFIG_IPV6) 2868 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2869 #endif 2870 } 2871 2872 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2873 struct vxlan_dev_node *node) 2874 { 2875 __be32 vni = vxlan->default_dst.remote_vni; 2876 2877 ASSERT_RTNL(); 2878 2879 node->vxlan = vxlan; 2880 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2881 } 2882 2883 /* Setup stats when device is created */ 2884 static int vxlan_init(struct net_device *dev) 2885 { 2886 struct vxlan_dev *vxlan = netdev_priv(dev); 2887 int err; 2888 2889 err = rhashtable_init(&vxlan->fdb_hash_tbl, &vxlan_fdb_rht_params); 2890 if (err) 2891 return err; 2892 2893 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 2894 err = vxlan_vnigroup_init(vxlan); 2895 if (err) 2896 goto err_rhashtable_destroy; 2897 } 2898 2899 err = gro_cells_init(&vxlan->gro_cells, dev); 2900 if (err) 2901 goto err_vnigroup_uninit; 2902 2903 err = vxlan_mdb_init(vxlan); 2904 if (err) 2905 goto err_gro_cells_destroy; 2906 2907 netdev_lockdep_set_classes(dev); 2908 return 0; 2909 2910 err_gro_cells_destroy: 2911 gro_cells_destroy(&vxlan->gro_cells); 2912 err_vnigroup_uninit: 2913 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2914 vxlan_vnigroup_uninit(vxlan); 2915 err_rhashtable_destroy: 2916 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2917 return err; 2918 } 2919 2920 static void vxlan_uninit(struct net_device *dev) 2921 { 2922 struct vxlan_dev *vxlan = netdev_priv(dev); 2923 2924 vxlan_mdb_fini(vxlan); 2925 2926 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2927 vxlan_vnigroup_uninit(vxlan); 2928 2929 gro_cells_destroy(&vxlan->gro_cells); 2930 2931 rhashtable_destroy(&vxlan->fdb_hash_tbl); 2932 } 2933 2934 /* Start ageing timer and join group when device is brought up */ 2935 static int vxlan_open(struct net_device *dev) 2936 { 2937 struct vxlan_dev *vxlan = netdev_priv(dev); 2938 int ret; 2939 2940 ret = vxlan_sock_add(vxlan); 2941 if (ret < 0) 2942 return ret; 2943 2944 ret = vxlan_multicast_join(vxlan); 2945 if (ret) { 2946 vxlan_sock_release(vxlan); 2947 return ret; 2948 } 2949 2950 if (vxlan->cfg.age_interval) 2951 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2952 2953 return ret; 2954 } 2955 2956 struct vxlan_fdb_flush_desc { 2957 bool ignore_default_entry; 2958 unsigned long state; 2959 unsigned long state_mask; 2960 unsigned long flags; 2961 unsigned long flags_mask; 2962 __be32 src_vni; 2963 u32 nhid; 2964 __be32 vni; 2965 __be16 port; 2966 union vxlan_addr dst_ip; 2967 }; 2968 2969 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f, 2970 const struct vxlan_dev *vxlan) 2971 { 2972 return is_zero_ether_addr(f->key.eth_addr) && 2973 f->key.vni == vxlan->cfg.vni; 2974 } 2975 2976 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid) 2977 { 2978 struct nexthop *nh = rtnl_dereference(f->nh); 2979 2980 return nh && nh->id == nhid; 2981 } 2982 2983 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f, 2984 const struct vxlan_dev *vxlan, 2985 const struct vxlan_fdb_flush_desc *desc) 2986 { 2987 if (desc->state_mask && (f->state & desc->state_mask) != desc->state) 2988 return false; 2989 2990 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags) 2991 return false; 2992 2993 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan)) 2994 return false; 2995 2996 if (desc->src_vni && f->key.vni != desc->src_vni) 2997 return false; 2998 2999 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid)) 3000 return false; 3001 3002 return true; 3003 } 3004 3005 static bool 3006 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc) 3007 { 3008 return desc->vni || desc->port || desc->dst_ip.sa.sa_family; 3009 } 3010 3011 static bool 3012 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc, 3013 const struct vxlan_rdst *rd) 3014 { 3015 if (desc->vni && rd->remote_vni != desc->vni) 3016 return false; 3017 3018 if (desc->port && rd->remote_port != desc->port) 3019 return false; 3020 3021 if (desc->dst_ip.sa.sa_family && 3022 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip)) 3023 return false; 3024 3025 return true; 3026 } 3027 3028 static void 3029 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan, 3030 const struct vxlan_fdb_flush_desc *desc, 3031 bool *p_destroy_fdb) 3032 { 3033 bool remotes_flushed = false; 3034 struct vxlan_rdst *rd, *tmp; 3035 3036 list_for_each_entry_safe(rd, tmp, &f->remotes, list) { 3037 if (!vxlan_fdb_flush_remote_matches(desc, rd)) 3038 continue; 3039 3040 vxlan_fdb_dst_destroy(vxlan, f, rd, true); 3041 remotes_flushed = true; 3042 } 3043 3044 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes); 3045 } 3046 3047 /* Purge the forwarding table */ 3048 static void vxlan_flush(struct vxlan_dev *vxlan, 3049 const struct vxlan_fdb_flush_desc *desc) 3050 { 3051 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc); 3052 struct vxlan_fdb *f; 3053 3054 rcu_read_lock(); 3055 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) { 3056 if (!vxlan_fdb_flush_matches(f, vxlan, desc)) 3057 continue; 3058 3059 spin_lock_bh(&vxlan->hash_lock); 3060 if (hlist_unhashed(&f->fdb_node)) 3061 goto unlock; 3062 3063 if (match_remotes) { 3064 bool destroy_fdb = false; 3065 3066 vxlan_fdb_flush_match_remotes(f, vxlan, desc, 3067 &destroy_fdb); 3068 3069 if (!destroy_fdb) 3070 goto unlock; 3071 } 3072 3073 vxlan_fdb_destroy(vxlan, f, true, true); 3074 unlock: 3075 spin_unlock_bh(&vxlan->hash_lock); 3076 } 3077 rcu_read_unlock(); 3078 } 3079 3080 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = { 3081 [NDA_SRC_VNI] = { .type = NLA_U32 }, 3082 [NDA_NH_ID] = { .type = NLA_U32 }, 3083 [NDA_VNI] = { .type = NLA_U32 }, 3084 [NDA_PORT] = { .type = NLA_U16 }, 3085 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr), 3086 sizeof(struct in6_addr)), 3087 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 }, 3088 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 }, 3089 }; 3090 3091 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF) 3092 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP) 3093 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \ 3094 NTF_ROUTER) 3095 3096 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev, 3097 struct netlink_ext_ack *extack) 3098 { 3099 struct vxlan_dev *vxlan = netdev_priv(dev); 3100 struct vxlan_fdb_flush_desc desc = {}; 3101 struct ndmsg *ndm = nlmsg_data(nlh); 3102 struct nlattr *tb[NDA_MAX + 1]; 3103 u8 ndm_flags; 3104 int err; 3105 3106 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS; 3107 3108 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy, 3109 extack); 3110 if (err) 3111 return err; 3112 3113 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) { 3114 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set"); 3115 return -EINVAL; 3116 } 3117 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) { 3118 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set"); 3119 return -EINVAL; 3120 } 3121 3122 desc.state = ndm->ndm_state; 3123 desc.flags = ndm_flags; 3124 3125 if (tb[NDA_NDM_STATE_MASK]) 3126 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]); 3127 3128 if (tb[NDA_NDM_FLAGS_MASK]) 3129 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]); 3130 3131 if (tb[NDA_SRC_VNI]) 3132 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 3133 3134 if (tb[NDA_NH_ID]) 3135 desc.nhid = nla_get_u32(tb[NDA_NH_ID]); 3136 3137 if (tb[NDA_VNI]) 3138 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 3139 3140 if (tb[NDA_PORT]) 3141 desc.port = nla_get_be16(tb[NDA_PORT]); 3142 3143 if (tb[NDA_DST]) { 3144 union vxlan_addr ip; 3145 3146 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]); 3147 if (err) { 3148 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST], 3149 "Unsupported address family"); 3150 return err; 3151 } 3152 desc.dst_ip = ip; 3153 } 3154 3155 vxlan_flush(vxlan, &desc); 3156 3157 return 0; 3158 } 3159 3160 /* Cleanup timer and forwarding table on shutdown */ 3161 static int vxlan_stop(struct net_device *dev) 3162 { 3163 struct vxlan_dev *vxlan = netdev_priv(dev); 3164 struct vxlan_fdb_flush_desc desc = { 3165 /* Default entry is deleted at vxlan_dellink. */ 3166 .ignore_default_entry = true, 3167 .state = 0, 3168 .state_mask = NUD_PERMANENT | NUD_NOARP, 3169 }; 3170 3171 vxlan_multicast_leave(vxlan); 3172 3173 timer_delete_sync(&vxlan->age_timer); 3174 3175 vxlan_flush(vxlan, &desc); 3176 vxlan_sock_release(vxlan); 3177 3178 return 0; 3179 } 3180 3181 /* Stub, nothing needs to be done. */ 3182 static void vxlan_set_multicast_list(struct net_device *dev) 3183 { 3184 } 3185 3186 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3187 { 3188 struct vxlan_dev *vxlan = netdev_priv(dev); 3189 struct vxlan_rdst *dst = &vxlan->default_dst; 3190 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3191 dst->remote_ifindex); 3192 3193 /* This check is different than dev->max_mtu, because it looks at 3194 * the lowerdev->mtu, rather than the static dev->max_mtu 3195 */ 3196 if (lowerdev) { 3197 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags); 3198 if (new_mtu > max_mtu) 3199 return -EINVAL; 3200 } 3201 3202 WRITE_ONCE(dev->mtu, new_mtu); 3203 return 0; 3204 } 3205 3206 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3207 { 3208 struct vxlan_dev *vxlan = netdev_priv(dev); 3209 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3210 __be16 sport, dport; 3211 3212 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3213 vxlan->cfg.port_max, true); 3214 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3215 3216 if (ip_tunnel_info_af(info) == AF_INET) { 3217 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3218 struct rtable *rt; 3219 3220 if (!sock4) 3221 return -EIO; 3222 3223 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0, 3224 &info->key.u.ipv4.src, 3225 &info->key, 3226 sport, dport, info->key.tos, 3227 &info->dst_cache); 3228 if (IS_ERR(rt)) 3229 return PTR_ERR(rt); 3230 ip_rt_put(rt); 3231 } else { 3232 #if IS_ENABLED(CONFIG_IPV6) 3233 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3234 struct dst_entry *ndst; 3235 3236 if (!sock6) 3237 return -EIO; 3238 3239 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 3240 0, &info->key.u.ipv6.src, 3241 &info->key, 3242 sport, dport, info->key.tos, 3243 &info->dst_cache); 3244 if (IS_ERR(ndst)) 3245 return PTR_ERR(ndst); 3246 dst_release(ndst); 3247 #else /* !CONFIG_IPV6 */ 3248 return -EPFNOSUPPORT; 3249 #endif 3250 } 3251 info->key.tp_src = sport; 3252 info->key.tp_dst = dport; 3253 return 0; 3254 } 3255 3256 static const struct net_device_ops vxlan_netdev_ether_ops = { 3257 .ndo_init = vxlan_init, 3258 .ndo_uninit = vxlan_uninit, 3259 .ndo_open = vxlan_open, 3260 .ndo_stop = vxlan_stop, 3261 .ndo_start_xmit = vxlan_xmit, 3262 .ndo_set_rx_mode = vxlan_set_multicast_list, 3263 .ndo_change_mtu = vxlan_change_mtu, 3264 .ndo_validate_addr = eth_validate_addr, 3265 .ndo_set_mac_address = eth_mac_addr, 3266 .ndo_fdb_add = vxlan_fdb_add, 3267 .ndo_fdb_del = vxlan_fdb_delete, 3268 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk, 3269 .ndo_fdb_dump = vxlan_fdb_dump, 3270 .ndo_fdb_get = vxlan_fdb_get, 3271 .ndo_mdb_add = vxlan_mdb_add, 3272 .ndo_mdb_del = vxlan_mdb_del, 3273 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk, 3274 .ndo_mdb_dump = vxlan_mdb_dump, 3275 .ndo_mdb_get = vxlan_mdb_get, 3276 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3277 }; 3278 3279 static const struct net_device_ops vxlan_netdev_raw_ops = { 3280 .ndo_init = vxlan_init, 3281 .ndo_uninit = vxlan_uninit, 3282 .ndo_open = vxlan_open, 3283 .ndo_stop = vxlan_stop, 3284 .ndo_start_xmit = vxlan_xmit, 3285 .ndo_change_mtu = vxlan_change_mtu, 3286 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3287 }; 3288 3289 /* Info for udev, that this is a virtual tunnel endpoint */ 3290 static const struct device_type vxlan_type = { 3291 .name = "vxlan", 3292 }; 3293 3294 /* Calls the ndo_udp_tunnel_add of the caller in order to 3295 * supply the listening VXLAN udp ports. Callers are expected 3296 * to implement the ndo_udp_tunnel_add. 3297 */ 3298 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3299 { 3300 struct vxlan_sock *vs; 3301 struct net *net = dev_net(dev); 3302 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3303 unsigned int i; 3304 3305 ASSERT_RTNL(); 3306 3307 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3308 hlist_for_each_entry(vs, &vn->sock_list[i], hlist) { 3309 unsigned short type; 3310 3311 if (vs->flags & VXLAN_F_GPE) 3312 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3313 else 3314 type = UDP_TUNNEL_TYPE_VXLAN; 3315 3316 if (push) 3317 udp_tunnel_push_rx_port(dev, vs->sock, type); 3318 else 3319 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3320 } 3321 } 3322 } 3323 3324 /* Initialize the device structure. */ 3325 static void vxlan_setup(struct net_device *dev) 3326 { 3327 struct vxlan_dev *vxlan = netdev_priv(dev); 3328 3329 eth_hw_addr_random(dev); 3330 ether_setup(dev); 3331 3332 dev->needs_free_netdev = true; 3333 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3334 3335 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3336 dev->features |= NETIF_F_RXCSUM; 3337 dev->features |= NETIF_F_GSO_SOFTWARE; 3338 3339 dev->vlan_features = dev->features; 3340 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3341 dev->hw_features |= NETIF_F_RXCSUM; 3342 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3343 netif_keep_dst(dev); 3344 dev->priv_flags |= IFF_NO_QUEUE; 3345 dev->change_proto_down = true; 3346 dev->lltx = true; 3347 3348 /* MTU range: 68 - 65535 */ 3349 dev->min_mtu = ETH_MIN_MTU; 3350 dev->max_mtu = ETH_MAX_MTU; 3351 3352 dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS; 3353 INIT_LIST_HEAD(&vxlan->next); 3354 spin_lock_init(&vxlan->hash_lock); 3355 3356 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3357 3358 vxlan->dev = dev; 3359 3360 INIT_HLIST_HEAD(&vxlan->fdb_list); 3361 } 3362 3363 static void vxlan_ether_setup(struct net_device *dev) 3364 { 3365 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3366 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3367 dev->netdev_ops = &vxlan_netdev_ether_ops; 3368 } 3369 3370 static void vxlan_raw_setup(struct net_device *dev) 3371 { 3372 dev->header_ops = NULL; 3373 dev->type = ARPHRD_NONE; 3374 dev->hard_header_len = 0; 3375 dev->addr_len = 0; 3376 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3377 dev->netdev_ops = &vxlan_netdev_raw_ops; 3378 } 3379 3380 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3381 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS }, 3382 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3383 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3384 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3385 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3386 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3387 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3388 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3389 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3390 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3391 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3392 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3393 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3394 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3395 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3396 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3397 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3398 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3399 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3400 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3401 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3402 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3403 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3404 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3405 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3406 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3407 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3408 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3409 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3410 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3411 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3412 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1), 3413 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX), 3414 [IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)), 3415 [IFLA_VXLAN_MC_ROUTE] = NLA_POLICY_MAX(NLA_U8, 1), 3416 }; 3417 3418 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3419 struct netlink_ext_ack *extack) 3420 { 3421 if (tb[IFLA_ADDRESS]) { 3422 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3423 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3424 "Provided link layer address is not Ethernet"); 3425 return -EINVAL; 3426 } 3427 3428 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3429 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3430 "Provided Ethernet address is not unicast"); 3431 return -EADDRNOTAVAIL; 3432 } 3433 } 3434 3435 if (tb[IFLA_MTU]) { 3436 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3437 3438 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3439 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3440 "MTU must be between 68 and 65535"); 3441 return -EINVAL; 3442 } 3443 } 3444 3445 if (!data) { 3446 NL_SET_ERR_MSG(extack, 3447 "Required attributes not provided to perform the operation"); 3448 return -EINVAL; 3449 } 3450 3451 if (data[IFLA_VXLAN_ID]) { 3452 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3453 3454 if (id >= VXLAN_N_VID) { 3455 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3456 "VXLAN ID must be lower than 16777216"); 3457 return -ERANGE; 3458 } 3459 } 3460 3461 if (data[IFLA_VXLAN_PORT_RANGE]) { 3462 const struct ifla_vxlan_port_range *p 3463 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3464 3465 if (ntohs(p->high) < ntohs(p->low)) { 3466 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3467 "Invalid source port range"); 3468 return -EINVAL; 3469 } 3470 } 3471 3472 if (data[IFLA_VXLAN_DF]) { 3473 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3474 3475 if (df < 0 || df > VXLAN_DF_MAX) { 3476 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3477 "Invalid DF attribute"); 3478 return -EINVAL; 3479 } 3480 } 3481 3482 return 0; 3483 } 3484 3485 static void vxlan_get_drvinfo(struct net_device *netdev, 3486 struct ethtool_drvinfo *drvinfo) 3487 { 3488 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3489 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3490 } 3491 3492 static int vxlan_get_link_ksettings(struct net_device *dev, 3493 struct ethtool_link_ksettings *cmd) 3494 { 3495 struct vxlan_dev *vxlan = netdev_priv(dev); 3496 struct vxlan_rdst *dst = &vxlan->default_dst; 3497 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3498 dst->remote_ifindex); 3499 3500 if (!lowerdev) { 3501 cmd->base.duplex = DUPLEX_UNKNOWN; 3502 cmd->base.port = PORT_OTHER; 3503 cmd->base.speed = SPEED_UNKNOWN; 3504 3505 return 0; 3506 } 3507 3508 return __ethtool_get_link_ksettings(lowerdev, cmd); 3509 } 3510 3511 static const struct ethtool_ops vxlan_ethtool_ops = { 3512 .get_drvinfo = vxlan_get_drvinfo, 3513 .get_link = ethtool_op_get_link, 3514 .get_link_ksettings = vxlan_get_link_ksettings, 3515 }; 3516 3517 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3518 __be16 port, u32 flags, int ifindex) 3519 { 3520 struct socket *sock; 3521 struct udp_port_cfg udp_conf; 3522 int err; 3523 3524 memset(&udp_conf, 0, sizeof(udp_conf)); 3525 3526 if (ipv6) { 3527 udp_conf.family = AF_INET6; 3528 udp_conf.use_udp6_rx_checksums = 3529 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3530 udp_conf.ipv6_v6only = 1; 3531 } else { 3532 udp_conf.family = AF_INET; 3533 } 3534 3535 udp_conf.local_udp_port = port; 3536 udp_conf.bind_ifindex = ifindex; 3537 3538 /* Open UDP socket */ 3539 err = udp_sock_create(net, &udp_conf, &sock); 3540 if (err < 0) 3541 return ERR_PTR(err); 3542 3543 udp_allow_gso(sock->sk); 3544 return sock; 3545 } 3546 3547 /* Create new listen socket if needed */ 3548 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3549 __be16 port, u32 flags, 3550 int ifindex) 3551 { 3552 struct vxlan_sock *vs; 3553 struct socket *sock; 3554 unsigned int h; 3555 struct udp_tunnel_sock_cfg tunnel_cfg; 3556 3557 ASSERT_RTNL(); 3558 3559 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3560 if (!vs) 3561 return ERR_PTR(-ENOMEM); 3562 3563 for (h = 0; h < VNI_HASH_SIZE; ++h) 3564 INIT_HLIST_HEAD(&vs->vni_list[h]); 3565 3566 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3567 if (IS_ERR(sock)) { 3568 kfree(vs); 3569 return ERR_CAST(sock); 3570 } 3571 3572 vs->sock = sock; 3573 refcount_set(&vs->refcnt, 1); 3574 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3575 3576 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3577 udp_tunnel_notify_add_rx_port(sock, 3578 (vs->flags & VXLAN_F_GPE) ? 3579 UDP_TUNNEL_TYPE_VXLAN_GPE : 3580 UDP_TUNNEL_TYPE_VXLAN); 3581 3582 /* Mark socket as an encapsulation socket. */ 3583 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3584 tunnel_cfg.sk_user_data = vs; 3585 tunnel_cfg.encap_type = 1; 3586 tunnel_cfg.encap_rcv = vxlan_rcv; 3587 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3588 tunnel_cfg.encap_destroy = NULL; 3589 if (vs->flags & VXLAN_F_GPE) { 3590 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive; 3591 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete; 3592 } else { 3593 tunnel_cfg.gro_receive = vxlan_gro_receive; 3594 tunnel_cfg.gro_complete = vxlan_gro_complete; 3595 } 3596 3597 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3598 3599 return vs; 3600 } 3601 3602 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3603 { 3604 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3605 struct vxlan_sock *vs = NULL; 3606 struct vxlan_dev_node *node; 3607 int l3mdev_index = 0; 3608 3609 ASSERT_RTNL(); 3610 3611 if (vxlan->cfg.remote_ifindex) 3612 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3613 vxlan->net, vxlan->cfg.remote_ifindex); 3614 3615 if (!vxlan->cfg.no_share) { 3616 rcu_read_lock(); 3617 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3618 vxlan->cfg.dst_port, vxlan->cfg.flags, 3619 l3mdev_index); 3620 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3621 rcu_read_unlock(); 3622 return -EBUSY; 3623 } 3624 rcu_read_unlock(); 3625 } 3626 if (!vs) 3627 vs = vxlan_socket_create(vxlan->net, ipv6, 3628 vxlan->cfg.dst_port, vxlan->cfg.flags, 3629 l3mdev_index); 3630 if (IS_ERR(vs)) 3631 return PTR_ERR(vs); 3632 #if IS_ENABLED(CONFIG_IPV6) 3633 if (ipv6) { 3634 rcu_assign_pointer(vxlan->vn6_sock, vs); 3635 node = &vxlan->hlist6; 3636 } else 3637 #endif 3638 { 3639 rcu_assign_pointer(vxlan->vn4_sock, vs); 3640 node = &vxlan->hlist4; 3641 } 3642 3643 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3644 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3645 else 3646 vxlan_vs_add_dev(vs, vxlan, node); 3647 3648 return 0; 3649 } 3650 3651 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3652 { 3653 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3654 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3655 bool ipv4 = !ipv6 || metadata; 3656 int ret = 0; 3657 3658 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3659 #if IS_ENABLED(CONFIG_IPV6) 3660 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3661 if (ipv6) { 3662 ret = __vxlan_sock_add(vxlan, true); 3663 if (ret < 0 && ret != -EAFNOSUPPORT) 3664 ipv4 = false; 3665 } 3666 #endif 3667 if (ipv4) 3668 ret = __vxlan_sock_add(vxlan, false); 3669 if (ret < 0) 3670 vxlan_sock_release(vxlan); 3671 return ret; 3672 } 3673 3674 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3675 struct vxlan_config *conf, __be32 vni) 3676 { 3677 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3678 struct vxlan_dev *tmp; 3679 3680 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3681 if (tmp == vxlan) 3682 continue; 3683 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3684 if (!vxlan_vnifilter_lookup(tmp, vni)) 3685 continue; 3686 } else if (tmp->cfg.vni != vni) { 3687 continue; 3688 } 3689 if (tmp->cfg.dst_port != conf->dst_port) 3690 continue; 3691 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3692 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3693 continue; 3694 3695 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3696 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3697 continue; 3698 3699 return -EEXIST; 3700 } 3701 3702 return 0; 3703 } 3704 3705 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3706 struct net_device **lower, 3707 struct vxlan_dev *old, 3708 struct netlink_ext_ack *extack) 3709 { 3710 bool use_ipv6 = false; 3711 3712 if (conf->flags & VXLAN_F_GPE) { 3713 /* For now, allow GPE only together with 3714 * COLLECT_METADATA. This can be relaxed later; in such 3715 * case, the other side of the PtP link will have to be 3716 * provided. 3717 */ 3718 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3719 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3720 NL_SET_ERR_MSG(extack, 3721 "VXLAN GPE does not support this combination of attributes"); 3722 return -EINVAL; 3723 } 3724 } 3725 3726 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3727 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3728 conf->remote_ip.sa.sa_family = AF_INET; 3729 conf->saddr.sa.sa_family = AF_INET; 3730 } else if (!conf->remote_ip.sa.sa_family) { 3731 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3732 } else if (!conf->saddr.sa.sa_family) { 3733 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3734 } 3735 3736 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3737 NL_SET_ERR_MSG(extack, 3738 "Local and remote address must be from the same family"); 3739 return -EINVAL; 3740 } 3741 3742 if (vxlan_addr_multicast(&conf->saddr)) { 3743 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3744 return -EINVAL; 3745 } 3746 3747 if (conf->saddr.sa.sa_family == AF_INET6) { 3748 if (!IS_ENABLED(CONFIG_IPV6)) { 3749 NL_SET_ERR_MSG(extack, 3750 "IPv6 support not enabled in the kernel"); 3751 return -EPFNOSUPPORT; 3752 } 3753 use_ipv6 = true; 3754 conf->flags |= VXLAN_F_IPV6; 3755 3756 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3757 int local_type = 3758 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3759 int remote_type = 3760 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3761 3762 if (local_type & IPV6_ADDR_LINKLOCAL) { 3763 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3764 (remote_type != IPV6_ADDR_ANY)) { 3765 NL_SET_ERR_MSG(extack, 3766 "Invalid combination of local and remote address scopes"); 3767 return -EINVAL; 3768 } 3769 3770 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3771 } else { 3772 if (remote_type == 3773 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3774 NL_SET_ERR_MSG(extack, 3775 "Invalid combination of local and remote address scopes"); 3776 return -EINVAL; 3777 } 3778 3779 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3780 } 3781 } 3782 } 3783 3784 if (conf->label && !use_ipv6) { 3785 NL_SET_ERR_MSG(extack, 3786 "Label attribute only applies to IPv6 VXLAN devices"); 3787 return -EINVAL; 3788 } 3789 3790 if (conf->label_policy && !use_ipv6) { 3791 NL_SET_ERR_MSG(extack, 3792 "Label policy only applies to IPv6 VXLAN devices"); 3793 return -EINVAL; 3794 } 3795 3796 if (conf->remote_ifindex) { 3797 struct net_device *lowerdev; 3798 3799 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3800 if (!lowerdev) { 3801 NL_SET_ERR_MSG(extack, 3802 "Invalid local interface, device not found"); 3803 return -ENODEV; 3804 } 3805 3806 #if IS_ENABLED(CONFIG_IPV6) 3807 if (use_ipv6) { 3808 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3809 3810 if (idev && idev->cnf.disable_ipv6) { 3811 NL_SET_ERR_MSG(extack, 3812 "IPv6 support disabled by administrator"); 3813 return -EPERM; 3814 } 3815 } 3816 #endif 3817 3818 *lower = lowerdev; 3819 } else { 3820 if (vxlan_addr_multicast(&conf->remote_ip)) { 3821 NL_SET_ERR_MSG(extack, 3822 "Local interface required for multicast remote destination"); 3823 3824 return -EINVAL; 3825 } 3826 3827 #if IS_ENABLED(CONFIG_IPV6) 3828 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3829 NL_SET_ERR_MSG(extack, 3830 "Local interface required for link-local local/remote addresses"); 3831 return -EINVAL; 3832 } 3833 #endif 3834 3835 *lower = NULL; 3836 } 3837 3838 if (!conf->dst_port) { 3839 if (conf->flags & VXLAN_F_GPE) 3840 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3841 else 3842 conf->dst_port = htons(vxlan_port); 3843 } 3844 3845 if (!conf->age_interval) 3846 conf->age_interval = FDB_AGE_DEFAULT; 3847 3848 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3849 NL_SET_ERR_MSG(extack, 3850 "A VXLAN device with the specified VNI already exists"); 3851 return -EEXIST; 3852 } 3853 3854 return 0; 3855 } 3856 3857 static void vxlan_config_apply(struct net_device *dev, 3858 struct vxlan_config *conf, 3859 struct net_device *lowerdev, 3860 struct net *src_net, 3861 bool changelink) 3862 { 3863 struct vxlan_dev *vxlan = netdev_priv(dev); 3864 struct vxlan_rdst *dst = &vxlan->default_dst; 3865 unsigned short needed_headroom = ETH_HLEN; 3866 int max_mtu = ETH_MAX_MTU; 3867 u32 flags = conf->flags; 3868 3869 if (!changelink) { 3870 if (flags & VXLAN_F_GPE) 3871 vxlan_raw_setup(dev); 3872 else 3873 vxlan_ether_setup(dev); 3874 3875 if (conf->mtu) 3876 dev->mtu = conf->mtu; 3877 3878 vxlan->net = src_net; 3879 } 3880 3881 dst->remote_vni = conf->vni; 3882 3883 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3884 3885 if (lowerdev) { 3886 dst->remote_ifindex = conf->remote_ifindex; 3887 3888 netif_inherit_tso_max(dev, lowerdev); 3889 3890 needed_headroom = lowerdev->hard_header_len; 3891 needed_headroom += lowerdev->needed_headroom; 3892 3893 dev->needed_tailroom = lowerdev->needed_tailroom; 3894 3895 max_mtu = lowerdev->mtu - vxlan_headroom(flags); 3896 if (max_mtu < ETH_MIN_MTU) 3897 max_mtu = ETH_MIN_MTU; 3898 3899 if (!changelink && !conf->mtu) 3900 dev->mtu = max_mtu; 3901 } 3902 3903 if (dev->mtu > max_mtu) 3904 dev->mtu = max_mtu; 3905 3906 if (flags & VXLAN_F_COLLECT_METADATA) 3907 flags |= VXLAN_F_IPV6; 3908 needed_headroom += vxlan_headroom(flags); 3909 dev->needed_headroom = needed_headroom; 3910 3911 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3912 } 3913 3914 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3915 struct vxlan_config *conf, 3916 struct netlink_ext_ack *extack) 3917 { 3918 struct vxlan_dev *vxlan = netdev_priv(dev); 3919 struct net_device *lowerdev; 3920 int ret; 3921 3922 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3923 if (ret) 3924 return ret; 3925 3926 vxlan_config_apply(dev, conf, lowerdev, src_net, false); 3927 3928 return 0; 3929 } 3930 3931 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3932 struct vxlan_config *conf, 3933 struct netlink_ext_ack *extack) 3934 { 3935 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3936 struct vxlan_dev *vxlan = netdev_priv(dev); 3937 struct net_device *remote_dev = NULL; 3938 struct vxlan_rdst *dst; 3939 int err; 3940 3941 dst = &vxlan->default_dst; 3942 err = vxlan_dev_configure(net, dev, conf, extack); 3943 if (err) 3944 return err; 3945 3946 dev->ethtool_ops = &vxlan_ethtool_ops; 3947 3948 err = register_netdevice(dev); 3949 if (err) 3950 return err; 3951 3952 if (dst->remote_ifindex) { 3953 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3954 if (!remote_dev) { 3955 err = -ENODEV; 3956 goto unregister; 3957 } 3958 3959 err = netdev_upper_dev_link(remote_dev, dev, extack); 3960 if (err) 3961 goto unregister; 3962 3963 dst->remote_dev = remote_dev; 3964 } 3965 3966 err = rtnl_configure_link(dev, NULL, 0, NULL); 3967 if (err < 0) 3968 goto unlink; 3969 3970 /* create an fdb entry for a valid default destination */ 3971 if (!vxlan_addr_any(&dst->remote_ip)) { 3972 spin_lock_bh(&vxlan->hash_lock); 3973 err = vxlan_fdb_update(vxlan, all_zeros_mac, 3974 &dst->remote_ip, 3975 NUD_REACHABLE | NUD_PERMANENT, 3976 NLM_F_EXCL | NLM_F_CREATE, 3977 vxlan->cfg.dst_port, 3978 dst->remote_vni, 3979 dst->remote_vni, 3980 dst->remote_ifindex, 3981 NTF_SELF, 0, true, extack); 3982 spin_unlock_bh(&vxlan->hash_lock); 3983 if (err) 3984 goto unlink; 3985 } 3986 3987 list_add(&vxlan->next, &vn->vxlan_list); 3988 3989 return 0; 3990 3991 unlink: 3992 if (remote_dev) 3993 netdev_upper_dev_unlink(remote_dev, dev); 3994 unregister: 3995 unregister_netdevice(dev); 3996 return err; 3997 } 3998 3999 /* Set/clear flags based on attribute */ 4000 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 4001 int attrtype, unsigned long mask, bool changelink, 4002 bool changelink_supported, 4003 struct netlink_ext_ack *extack) 4004 { 4005 unsigned long flags; 4006 4007 if (!tb[attrtype]) 4008 return 0; 4009 4010 if (changelink && !changelink_supported) { 4011 vxlan_flag_attr_error(attrtype, extack); 4012 return -EOPNOTSUPP; 4013 } 4014 4015 if (vxlan_policy[attrtype].type == NLA_FLAG) 4016 flags = conf->flags | mask; 4017 else if (nla_get_u8(tb[attrtype])) 4018 flags = conf->flags | mask; 4019 else 4020 flags = conf->flags & ~mask; 4021 4022 conf->flags = flags; 4023 4024 return 0; 4025 } 4026 4027 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 4028 struct net_device *dev, struct vxlan_config *conf, 4029 bool changelink, struct netlink_ext_ack *extack) 4030 { 4031 struct vxlanhdr used_bits = { 4032 .vx_flags = VXLAN_HF_VNI, 4033 .vx_vni = VXLAN_VNI_MASK, 4034 }; 4035 struct vxlan_dev *vxlan = netdev_priv(dev); 4036 int err = 0; 4037 4038 memset(conf, 0, sizeof(*conf)); 4039 4040 /* if changelink operation, start with old existing cfg */ 4041 if (changelink) 4042 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 4043 4044 if (data[IFLA_VXLAN_ID]) { 4045 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 4046 4047 if (changelink && (vni != conf->vni)) { 4048 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 4049 return -EOPNOTSUPP; 4050 } 4051 conf->vni = vni; 4052 } 4053 4054 if (data[IFLA_VXLAN_GROUP]) { 4055 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 4056 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 4057 return -EOPNOTSUPP; 4058 } 4059 4060 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 4061 conf->remote_ip.sa.sa_family = AF_INET; 4062 } else if (data[IFLA_VXLAN_GROUP6]) { 4063 if (!IS_ENABLED(CONFIG_IPV6)) { 4064 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 4065 return -EPFNOSUPPORT; 4066 } 4067 4068 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 4069 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 4070 return -EOPNOTSUPP; 4071 } 4072 4073 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 4074 conf->remote_ip.sa.sa_family = AF_INET6; 4075 } 4076 4077 if (data[IFLA_VXLAN_LOCAL]) { 4078 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 4079 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 4080 return -EOPNOTSUPP; 4081 } 4082 4083 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 4084 conf->saddr.sa.sa_family = AF_INET; 4085 } else if (data[IFLA_VXLAN_LOCAL6]) { 4086 if (!IS_ENABLED(CONFIG_IPV6)) { 4087 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 4088 return -EPFNOSUPPORT; 4089 } 4090 4091 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 4092 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 4093 return -EOPNOTSUPP; 4094 } 4095 4096 /* TODO: respect scope id */ 4097 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 4098 conf->saddr.sa.sa_family = AF_INET6; 4099 } 4100 4101 if (data[IFLA_VXLAN_LINK]) 4102 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 4103 4104 if (data[IFLA_VXLAN_TOS]) 4105 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 4106 4107 if (data[IFLA_VXLAN_TTL]) 4108 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 4109 4110 if (data[IFLA_VXLAN_TTL_INHERIT]) { 4111 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 4112 VXLAN_F_TTL_INHERIT, changelink, false, 4113 extack); 4114 if (err) 4115 return err; 4116 4117 } 4118 4119 if (data[IFLA_VXLAN_LABEL]) 4120 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 4121 IPV6_FLOWLABEL_MASK; 4122 if (data[IFLA_VXLAN_LABEL_POLICY]) 4123 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]); 4124 4125 if (data[IFLA_VXLAN_LEARNING]) { 4126 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 4127 VXLAN_F_LEARN, changelink, true, 4128 extack); 4129 if (err) 4130 return err; 4131 } else if (!changelink) { 4132 /* default to learn on a new device */ 4133 conf->flags |= VXLAN_F_LEARN; 4134 } 4135 4136 if (data[IFLA_VXLAN_AGEING]) 4137 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 4138 4139 if (data[IFLA_VXLAN_PROXY]) { 4140 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 4141 VXLAN_F_PROXY, changelink, false, 4142 extack); 4143 if (err) 4144 return err; 4145 } 4146 4147 if (data[IFLA_VXLAN_RSC]) { 4148 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 4149 VXLAN_F_RSC, changelink, false, 4150 extack); 4151 if (err) 4152 return err; 4153 } 4154 4155 if (data[IFLA_VXLAN_L2MISS]) { 4156 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 4157 VXLAN_F_L2MISS, changelink, false, 4158 extack); 4159 if (err) 4160 return err; 4161 } 4162 4163 if (data[IFLA_VXLAN_L3MISS]) { 4164 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 4165 VXLAN_F_L3MISS, changelink, false, 4166 extack); 4167 if (err) 4168 return err; 4169 } 4170 4171 if (data[IFLA_VXLAN_LIMIT]) { 4172 if (changelink) { 4173 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 4174 "Cannot change limit"); 4175 return -EOPNOTSUPP; 4176 } 4177 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 4178 } 4179 4180 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 4181 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 4182 VXLAN_F_COLLECT_METADATA, changelink, false, 4183 extack); 4184 if (err) 4185 return err; 4186 } 4187 4188 if (data[IFLA_VXLAN_PORT_RANGE]) { 4189 if (!changelink) { 4190 const struct ifla_vxlan_port_range *p 4191 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4192 conf->port_min = ntohs(p->low); 4193 conf->port_max = ntohs(p->high); 4194 } else { 4195 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4196 "Cannot change port range"); 4197 return -EOPNOTSUPP; 4198 } 4199 } 4200 4201 if (data[IFLA_VXLAN_PORT]) { 4202 if (changelink) { 4203 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4204 "Cannot change port"); 4205 return -EOPNOTSUPP; 4206 } 4207 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4208 } 4209 4210 if (data[IFLA_VXLAN_UDP_CSUM]) { 4211 if (changelink) { 4212 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4213 "Cannot change UDP_CSUM flag"); 4214 return -EOPNOTSUPP; 4215 } 4216 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4217 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4218 } 4219 4220 if (data[IFLA_VXLAN_LOCALBYPASS]) { 4221 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS, 4222 VXLAN_F_LOCALBYPASS, changelink, 4223 true, extack); 4224 if (err) 4225 return err; 4226 } else if (!changelink) { 4227 /* default to local bypass on a new device */ 4228 conf->flags |= VXLAN_F_LOCALBYPASS; 4229 } 4230 4231 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4232 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4233 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4234 false, extack); 4235 if (err) 4236 return err; 4237 } 4238 4239 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4240 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4241 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4242 false, extack); 4243 if (err) 4244 return err; 4245 } 4246 4247 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4248 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4249 VXLAN_F_REMCSUM_TX, changelink, false, 4250 extack); 4251 if (err) 4252 return err; 4253 } 4254 4255 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4256 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4257 VXLAN_F_REMCSUM_RX, changelink, false, 4258 extack); 4259 if (err) 4260 return err; 4261 used_bits.vx_flags |= VXLAN_HF_RCO; 4262 used_bits.vx_vni |= ~VXLAN_VNI_MASK; 4263 } 4264 4265 if (data[IFLA_VXLAN_GBP]) { 4266 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4267 VXLAN_F_GBP, changelink, false, extack); 4268 if (err) 4269 return err; 4270 used_bits.vx_flags |= VXLAN_GBP_USED_BITS; 4271 } 4272 4273 if (data[IFLA_VXLAN_GPE]) { 4274 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4275 VXLAN_F_GPE, changelink, false, 4276 extack); 4277 if (err) 4278 return err; 4279 4280 used_bits.vx_flags |= VXLAN_GPE_USED_BITS; 4281 } 4282 4283 if (data[IFLA_VXLAN_RESERVED_BITS]) { 4284 struct vxlanhdr reserved_bits; 4285 4286 if (changelink) { 4287 NL_SET_ERR_MSG_ATTR(extack, 4288 data[IFLA_VXLAN_RESERVED_BITS], 4289 "Cannot change reserved_bits"); 4290 return -EOPNOTSUPP; 4291 } 4292 4293 nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS], 4294 sizeof(reserved_bits)); 4295 if (used_bits.vx_flags & reserved_bits.vx_flags || 4296 used_bits.vx_vni & reserved_bits.vx_vni) { 4297 __be64 ub_be64, rb_be64; 4298 4299 memcpy(&ub_be64, &used_bits, sizeof(ub_be64)); 4300 memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64)); 4301 4302 NL_SET_ERR_MSG_ATTR_FMT(extack, 4303 data[IFLA_VXLAN_RESERVED_BITS], 4304 "Used bits %#018llx cannot overlap reserved bits %#018llx", 4305 be64_to_cpu(ub_be64), 4306 be64_to_cpu(rb_be64)); 4307 return -EINVAL; 4308 } 4309 4310 conf->reserved_bits = reserved_bits; 4311 } else { 4312 /* For backwards compatibility, only allow reserved fields to be 4313 * used by VXLAN extensions if explicitly requested. 4314 */ 4315 conf->reserved_bits = (struct vxlanhdr) { 4316 .vx_flags = ~used_bits.vx_flags, 4317 .vx_vni = ~used_bits.vx_vni, 4318 }; 4319 } 4320 4321 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4322 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4323 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4324 false, extack); 4325 if (err) 4326 return err; 4327 } 4328 4329 if (data[IFLA_VXLAN_MC_ROUTE]) { 4330 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_MC_ROUTE, 4331 VXLAN_F_MC_ROUTE, changelink, 4332 true, extack); 4333 if (err) 4334 return err; 4335 } 4336 4337 if (tb[IFLA_MTU]) { 4338 if (changelink) { 4339 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4340 "Cannot change mtu"); 4341 return -EOPNOTSUPP; 4342 } 4343 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4344 } 4345 4346 if (data[IFLA_VXLAN_DF]) 4347 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4348 4349 if (data[IFLA_VXLAN_VNIFILTER]) { 4350 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4351 VXLAN_F_VNIFILTER, changelink, false, 4352 extack); 4353 if (err) 4354 return err; 4355 4356 if ((conf->flags & VXLAN_F_VNIFILTER) && 4357 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4358 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4359 "vxlan vnifilter only valid in collect metadata mode"); 4360 return -EINVAL; 4361 } 4362 } 4363 4364 return 0; 4365 } 4366 4367 static int vxlan_newlink(struct net_device *dev, 4368 struct rtnl_newlink_params *params, 4369 struct netlink_ext_ack *extack) 4370 { 4371 struct net *link_net = rtnl_newlink_link_net(params); 4372 struct nlattr **data = params->data; 4373 struct nlattr **tb = params->tb; 4374 struct vxlan_config conf; 4375 int err; 4376 4377 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4378 if (err) 4379 return err; 4380 4381 return __vxlan_dev_create(link_net, dev, &conf, extack); 4382 } 4383 4384 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4385 struct nlattr *data[], 4386 struct netlink_ext_ack *extack) 4387 { 4388 struct vxlan_dev *vxlan = netdev_priv(dev); 4389 bool rem_ip_changed, change_igmp; 4390 struct net_device *lowerdev; 4391 struct vxlan_config conf; 4392 struct vxlan_rdst *dst; 4393 int err; 4394 4395 dst = &vxlan->default_dst; 4396 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4397 if (err) 4398 return err; 4399 4400 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4401 vxlan, extack); 4402 if (err) 4403 return err; 4404 4405 if (dst->remote_dev == lowerdev) 4406 lowerdev = NULL; 4407 4408 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4409 extack); 4410 if (err) 4411 return err; 4412 4413 rem_ip_changed = !vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip); 4414 change_igmp = vxlan->dev->flags & IFF_UP && 4415 (rem_ip_changed || 4416 dst->remote_ifindex != conf.remote_ifindex); 4417 4418 /* handle default dst entry */ 4419 if (rem_ip_changed) { 4420 spin_lock_bh(&vxlan->hash_lock); 4421 if (!vxlan_addr_any(&conf.remote_ip)) { 4422 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4423 &conf.remote_ip, 4424 NUD_REACHABLE | NUD_PERMANENT, 4425 NLM_F_APPEND | NLM_F_CREATE, 4426 vxlan->cfg.dst_port, 4427 conf.vni, conf.vni, 4428 conf.remote_ifindex, 4429 NTF_SELF, 0, true, extack); 4430 if (err) { 4431 spin_unlock_bh(&vxlan->hash_lock); 4432 netdev_adjacent_change_abort(dst->remote_dev, 4433 lowerdev, dev); 4434 return err; 4435 } 4436 } 4437 if (!vxlan_addr_any(&dst->remote_ip)) 4438 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4439 dst->remote_ip, 4440 vxlan->cfg.dst_port, 4441 dst->remote_vni, 4442 dst->remote_vni, 4443 dst->remote_ifindex, 4444 true); 4445 spin_unlock_bh(&vxlan->hash_lock); 4446 4447 /* If vni filtering device, also update fdb entries of 4448 * all vnis that were using default remote ip 4449 */ 4450 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4451 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4452 &conf.remote_ip, extack); 4453 if (err) { 4454 netdev_adjacent_change_abort(dst->remote_dev, 4455 lowerdev, dev); 4456 return err; 4457 } 4458 } 4459 } 4460 4461 if (change_igmp && vxlan_addr_multicast(&dst->remote_ip)) 4462 err = vxlan_multicast_leave(vxlan); 4463 4464 if (conf.age_interval != vxlan->cfg.age_interval) 4465 mod_timer(&vxlan->age_timer, jiffies); 4466 4467 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4468 if (lowerdev && lowerdev != dst->remote_dev) 4469 dst->remote_dev = lowerdev; 4470 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4471 4472 if (!err && change_igmp && 4473 vxlan_addr_multicast(&dst->remote_ip)) 4474 err = vxlan_multicast_join(vxlan); 4475 4476 return err; 4477 } 4478 4479 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4480 { 4481 struct vxlan_dev *vxlan = netdev_priv(dev); 4482 struct vxlan_fdb_flush_desc desc = {}; 4483 4484 vxlan_flush(vxlan, &desc); 4485 4486 list_del(&vxlan->next); 4487 unregister_netdevice_queue(dev, head); 4488 if (vxlan->default_dst.remote_dev) 4489 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4490 } 4491 4492 static size_t vxlan_get_size(const struct net_device *dev) 4493 { 4494 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4495 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4496 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4497 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4498 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4499 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4500 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4501 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4502 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4503 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */ 4504 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4505 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4506 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4507 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4508 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4509 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4510 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4511 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4512 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4513 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4514 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4515 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4516 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4517 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4518 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */ 4519 /* IFLA_VXLAN_PORT_RANGE */ 4520 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4521 nla_total_size(0) + /* IFLA_VXLAN_GBP */ 4522 nla_total_size(0) + /* IFLA_VXLAN_GPE */ 4523 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */ 4524 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */ 4525 /* IFLA_VXLAN_RESERVED_BITS */ 4526 nla_total_size(sizeof(struct vxlanhdr)) + 4527 0; 4528 } 4529 4530 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4531 { 4532 const struct vxlan_dev *vxlan = netdev_priv(dev); 4533 const struct vxlan_rdst *dst = &vxlan->default_dst; 4534 struct ifla_vxlan_port_range ports = { 4535 .low = htons(vxlan->cfg.port_min), 4536 .high = htons(vxlan->cfg.port_max), 4537 }; 4538 4539 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4540 goto nla_put_failure; 4541 4542 if (!vxlan_addr_any(&dst->remote_ip)) { 4543 if (dst->remote_ip.sa.sa_family == AF_INET) { 4544 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4545 dst->remote_ip.sin.sin_addr.s_addr)) 4546 goto nla_put_failure; 4547 #if IS_ENABLED(CONFIG_IPV6) 4548 } else { 4549 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4550 &dst->remote_ip.sin6.sin6_addr)) 4551 goto nla_put_failure; 4552 #endif 4553 } 4554 } 4555 4556 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4557 goto nla_put_failure; 4558 4559 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4560 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4561 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4562 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4563 goto nla_put_failure; 4564 #if IS_ENABLED(CONFIG_IPV6) 4565 } else { 4566 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4567 &vxlan->cfg.saddr.sin6.sin6_addr)) 4568 goto nla_put_failure; 4569 #endif 4570 } 4571 } 4572 4573 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4574 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4575 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4576 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4577 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4578 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4579 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) || 4580 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4581 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4582 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4583 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4584 nla_put_u8(skb, IFLA_VXLAN_RSC, 4585 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4586 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4587 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4588 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4589 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4590 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4591 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4592 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4593 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4594 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4595 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4596 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4597 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4598 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4599 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4600 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4601 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4602 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4603 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4604 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) || 4605 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS, 4606 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS))) 4607 goto nla_put_failure; 4608 4609 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4610 goto nla_put_failure; 4611 4612 if (vxlan->cfg.flags & VXLAN_F_GBP && 4613 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4614 goto nla_put_failure; 4615 4616 if (vxlan->cfg.flags & VXLAN_F_GPE && 4617 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4618 goto nla_put_failure; 4619 4620 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4621 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4622 goto nla_put_failure; 4623 4624 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4625 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4626 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4627 goto nla_put_failure; 4628 4629 if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS, 4630 sizeof(vxlan->cfg.reserved_bits), 4631 &vxlan->cfg.reserved_bits)) 4632 goto nla_put_failure; 4633 4634 return 0; 4635 4636 nla_put_failure: 4637 return -EMSGSIZE; 4638 } 4639 4640 static struct net *vxlan_get_link_net(const struct net_device *dev) 4641 { 4642 struct vxlan_dev *vxlan = netdev_priv(dev); 4643 4644 return READ_ONCE(vxlan->net); 4645 } 4646 4647 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4648 .kind = "vxlan", 4649 .maxtype = IFLA_VXLAN_MAX, 4650 .policy = vxlan_policy, 4651 .priv_size = sizeof(struct vxlan_dev), 4652 .setup = vxlan_setup, 4653 .validate = vxlan_validate, 4654 .newlink = vxlan_newlink, 4655 .changelink = vxlan_changelink, 4656 .dellink = vxlan_dellink, 4657 .get_size = vxlan_get_size, 4658 .fill_info = vxlan_fill_info, 4659 .get_link_net = vxlan_get_link_net, 4660 }; 4661 4662 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4663 u8 name_assign_type, 4664 struct vxlan_config *conf) 4665 { 4666 struct nlattr *tb[IFLA_MAX + 1]; 4667 struct net_device *dev; 4668 int err; 4669 4670 memset(&tb, 0, sizeof(tb)); 4671 4672 dev = rtnl_create_link(net, name, name_assign_type, 4673 &vxlan_link_ops, tb, NULL); 4674 if (IS_ERR(dev)) 4675 return dev; 4676 4677 err = __vxlan_dev_create(net, dev, conf, NULL); 4678 if (err < 0) { 4679 free_netdev(dev); 4680 return ERR_PTR(err); 4681 } 4682 4683 err = rtnl_configure_link(dev, NULL, 0, NULL); 4684 if (err < 0) { 4685 LIST_HEAD(list_kill); 4686 4687 vxlan_dellink(dev, &list_kill); 4688 unregister_netdevice_many(&list_kill); 4689 return ERR_PTR(err); 4690 } 4691 4692 return dev; 4693 } 4694 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4695 4696 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4697 struct net_device *dev) 4698 { 4699 struct vxlan_dev *vxlan, *next; 4700 LIST_HEAD(list_kill); 4701 4702 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4703 struct vxlan_rdst *dst = &vxlan->default_dst; 4704 4705 /* In case we created vxlan device with carrier 4706 * and we loose the carrier due to module unload 4707 * we also need to remove vxlan device. In other 4708 * cases, it's not necessary and remote_ifindex 4709 * is 0 here, so no matches. 4710 */ 4711 if (dst->remote_ifindex == dev->ifindex) 4712 vxlan_dellink(vxlan->dev, &list_kill); 4713 } 4714 4715 unregister_netdevice_many(&list_kill); 4716 } 4717 4718 static int vxlan_netdevice_event(struct notifier_block *unused, 4719 unsigned long event, void *ptr) 4720 { 4721 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4722 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4723 4724 if (event == NETDEV_UNREGISTER) 4725 vxlan_handle_lowerdev_unregister(vn, dev); 4726 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4727 vxlan_offload_rx_ports(dev, true); 4728 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4729 vxlan_offload_rx_ports(dev, false); 4730 4731 return NOTIFY_DONE; 4732 } 4733 4734 static struct notifier_block vxlan_notifier_block __read_mostly = { 4735 .notifier_call = vxlan_netdevice_event, 4736 }; 4737 4738 static void 4739 vxlan_fdb_offloaded_set(struct net_device *dev, 4740 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4741 { 4742 struct vxlan_dev *vxlan = netdev_priv(dev); 4743 struct vxlan_rdst *rdst; 4744 struct vxlan_fdb *f; 4745 4746 spin_lock_bh(&vxlan->hash_lock); 4747 4748 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4749 if (!f) 4750 goto out; 4751 4752 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4753 fdb_info->remote_port, 4754 fdb_info->remote_vni, 4755 fdb_info->remote_ifindex); 4756 if (!rdst) 4757 goto out; 4758 4759 rdst->offloaded = fdb_info->offloaded; 4760 4761 out: 4762 spin_unlock_bh(&vxlan->hash_lock); 4763 } 4764 4765 static int 4766 vxlan_fdb_external_learn_add(struct net_device *dev, 4767 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4768 { 4769 struct vxlan_dev *vxlan = netdev_priv(dev); 4770 struct netlink_ext_ack *extack; 4771 int err; 4772 4773 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4774 4775 spin_lock_bh(&vxlan->hash_lock); 4776 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4777 NUD_REACHABLE, 4778 NLM_F_CREATE | NLM_F_REPLACE, 4779 fdb_info->remote_port, 4780 fdb_info->vni, 4781 fdb_info->remote_vni, 4782 fdb_info->remote_ifindex, 4783 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4784 0, false, extack); 4785 spin_unlock_bh(&vxlan->hash_lock); 4786 4787 return err; 4788 } 4789 4790 static int 4791 vxlan_fdb_external_learn_del(struct net_device *dev, 4792 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4793 { 4794 struct vxlan_dev *vxlan = netdev_priv(dev); 4795 struct vxlan_fdb *f; 4796 int err = 0; 4797 4798 spin_lock_bh(&vxlan->hash_lock); 4799 4800 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4801 if (!f) 4802 err = -ENOENT; 4803 else if (f->flags & NTF_EXT_LEARNED) 4804 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4805 fdb_info->remote_ip, 4806 fdb_info->remote_port, 4807 fdb_info->vni, 4808 fdb_info->remote_vni, 4809 fdb_info->remote_ifindex, 4810 false); 4811 4812 spin_unlock_bh(&vxlan->hash_lock); 4813 4814 return err; 4815 } 4816 4817 static int vxlan_switchdev_event(struct notifier_block *unused, 4818 unsigned long event, void *ptr) 4819 { 4820 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4821 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4822 int err = 0; 4823 4824 switch (event) { 4825 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4826 vxlan_fdb_offloaded_set(dev, ptr); 4827 break; 4828 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4829 fdb_info = ptr; 4830 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4831 if (err) { 4832 err = notifier_from_errno(err); 4833 break; 4834 } 4835 fdb_info->offloaded = true; 4836 vxlan_fdb_offloaded_set(dev, fdb_info); 4837 break; 4838 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4839 fdb_info = ptr; 4840 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4841 if (err) { 4842 err = notifier_from_errno(err); 4843 break; 4844 } 4845 fdb_info->offloaded = false; 4846 vxlan_fdb_offloaded_set(dev, fdb_info); 4847 break; 4848 } 4849 4850 return err; 4851 } 4852 4853 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4854 .notifier_call = vxlan_switchdev_event, 4855 }; 4856 4857 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4858 { 4859 struct vxlan_fdb *fdb; 4860 struct vxlan_dev *vxlan; 4861 4862 rcu_read_lock(); 4863 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4864 vxlan = rcu_dereference(fdb->vdev); 4865 WARN_ON(!vxlan); 4866 spin_lock_bh(&vxlan->hash_lock); 4867 if (!hlist_unhashed(&fdb->fdb_node)) 4868 vxlan_fdb_destroy(vxlan, fdb, false, false); 4869 spin_unlock_bh(&vxlan->hash_lock); 4870 } 4871 rcu_read_unlock(); 4872 } 4873 4874 static int vxlan_nexthop_event(struct notifier_block *nb, 4875 unsigned long event, void *ptr) 4876 { 4877 struct nh_notifier_info *info = ptr; 4878 struct nexthop *nh; 4879 4880 if (event != NEXTHOP_EVENT_DEL) 4881 return NOTIFY_DONE; 4882 4883 nh = nexthop_find_by_id(info->net, info->id); 4884 if (!nh) 4885 return NOTIFY_DONE; 4886 4887 vxlan_fdb_nh_flush(nh); 4888 4889 return NOTIFY_DONE; 4890 } 4891 4892 static __net_init int vxlan_init_net(struct net *net) 4893 { 4894 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4895 unsigned int h; 4896 4897 INIT_LIST_HEAD(&vn->vxlan_list); 4898 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4899 4900 for (h = 0; h < PORT_HASH_SIZE; ++h) 4901 INIT_HLIST_HEAD(&vn->sock_list[h]); 4902 4903 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4904 NULL); 4905 } 4906 4907 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn, 4908 struct list_head *dev_to_kill) 4909 { 4910 struct vxlan_dev *vxlan, *next; 4911 4912 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) 4913 vxlan_dellink(vxlan->dev, dev_to_kill); 4914 } 4915 4916 static void __net_exit vxlan_exit_rtnl(struct net *net, 4917 struct list_head *dev_to_kill) 4918 { 4919 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4920 4921 ASSERT_RTNL_NET(net); 4922 4923 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4924 vxlan_destroy_tunnels(vn, dev_to_kill); 4925 } 4926 4927 static void __net_exit vxlan_exit_net(struct net *net) 4928 { 4929 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4930 unsigned int h; 4931 4932 for (h = 0; h < PORT_HASH_SIZE; ++h) 4933 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4934 } 4935 4936 static struct pernet_operations vxlan_net_ops = { 4937 .init = vxlan_init_net, 4938 .exit_rtnl = vxlan_exit_rtnl, 4939 .exit = vxlan_exit_net, 4940 .id = &vxlan_net_id, 4941 .size = sizeof(struct vxlan_net), 4942 }; 4943 4944 static int __init vxlan_init_module(void) 4945 { 4946 int rc; 4947 4948 rc = register_pernet_subsys(&vxlan_net_ops); 4949 if (rc) 4950 goto out1; 4951 4952 rc = register_netdevice_notifier(&vxlan_notifier_block); 4953 if (rc) 4954 goto out2; 4955 4956 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4957 if (rc) 4958 goto out3; 4959 4960 rc = rtnl_link_register(&vxlan_link_ops); 4961 if (rc) 4962 goto out4; 4963 4964 rc = vxlan_vnifilter_init(); 4965 if (rc) 4966 goto out5; 4967 4968 return 0; 4969 out5: 4970 rtnl_link_unregister(&vxlan_link_ops); 4971 out4: 4972 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4973 out3: 4974 unregister_netdevice_notifier(&vxlan_notifier_block); 4975 out2: 4976 unregister_pernet_subsys(&vxlan_net_ops); 4977 out1: 4978 return rc; 4979 } 4980 late_initcall(vxlan_init_module); 4981 4982 static void __exit vxlan_cleanup_module(void) 4983 { 4984 vxlan_vnifilter_uninit(); 4985 rtnl_link_unregister(&vxlan_link_ops); 4986 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4987 unregister_netdevice_notifier(&vxlan_notifier_block); 4988 unregister_pernet_subsys(&vxlan_net_ops); 4989 /* rcu_barrier() is called by netns */ 4990 } 4991 module_exit(vxlan_cleanup_module); 4992 4993 MODULE_LICENSE("GPL"); 4994 MODULE_VERSION(VXLAN_VERSION); 4995 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4996 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4997 MODULE_ALIAS_RTNL_LINK("vxlan"); 4998