1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * VXLAN: Virtual eXtensible Local Area Network 4 * 5 * Copyright (c) 2012-2013 Vyatta Inc. 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/kernel.h> 11 #include <linux/module.h> 12 #include <linux/errno.h> 13 #include <linux/slab.h> 14 #include <linux/udp.h> 15 #include <linux/igmp.h> 16 #include <linux/if_ether.h> 17 #include <linux/ethtool.h> 18 #include <net/arp.h> 19 #include <net/ndisc.h> 20 #include <net/gro.h> 21 #include <net/ipv6_stubs.h> 22 #include <net/ip.h> 23 #include <net/icmp.h> 24 #include <net/rtnetlink.h> 25 #include <net/inet_ecn.h> 26 #include <net/net_namespace.h> 27 #include <net/netns/generic.h> 28 #include <net/tun_proto.h> 29 #include <net/vxlan.h> 30 #include <net/nexthop.h> 31 32 #if IS_ENABLED(CONFIG_IPV6) 33 #include <net/ip6_tunnel.h> 34 #include <net/ip6_checksum.h> 35 #endif 36 37 #include "vxlan_private.h" 38 39 #define VXLAN_VERSION "0.1" 40 41 #define FDB_AGE_DEFAULT 300 /* 5 min */ 42 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ 43 44 /* UDP port for VXLAN traffic. 45 * The IANA assigned port is 4789, but the Linux default is 8472 46 * for compatibility with early adopters. 47 */ 48 static unsigned short vxlan_port __read_mostly = 8472; 49 module_param_named(udp_port, vxlan_port, ushort, 0444); 50 MODULE_PARM_DESC(udp_port, "Destination UDP port"); 51 52 static bool log_ecn_error = true; 53 module_param(log_ecn_error, bool, 0644); 54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 55 56 unsigned int vxlan_net_id; 57 58 const u8 all_zeros_mac[ETH_ALEN + 2]; 59 static struct rtnl_link_ops vxlan_link_ops; 60 61 static int vxlan_sock_add(struct vxlan_dev *vxlan); 62 63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan); 64 65 /* salt for hash table */ 66 static u32 vxlan_salt __read_mostly; 67 68 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) 69 { 70 return vs->flags & VXLAN_F_COLLECT_METADATA || 71 ip_tunnel_collect_metadata(); 72 } 73 74 /* Find VXLAN socket based on network namespace, address family, UDP port, 75 * enabled unshareable flags and socket device binding (see l3mdev with 76 * non-default VRF). 77 */ 78 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, 79 __be16 port, u32 flags, int ifindex) 80 { 81 struct vxlan_sock *vs; 82 83 flags &= VXLAN_F_RCV_FLAGS; 84 85 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { 86 if (inet_sk(vs->sock->sk)->inet_sport == port && 87 vxlan_get_sk_family(vs) == family && 88 vs->flags == flags && 89 vs->sock->sk->sk_bound_dev_if == ifindex) 90 return vs; 91 } 92 return NULL; 93 } 94 95 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, 96 int ifindex, __be32 vni, 97 struct vxlan_vni_node **vninode) 98 { 99 struct vxlan_vni_node *vnode; 100 struct vxlan_dev_node *node; 101 102 /* For flow based devices, map all packets to VNI 0 */ 103 if (vs->flags & VXLAN_F_COLLECT_METADATA && 104 !(vs->flags & VXLAN_F_VNIFILTER)) 105 vni = 0; 106 107 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) { 108 if (!node->vxlan) 109 continue; 110 vnode = NULL; 111 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 112 vnode = vxlan_vnifilter_lookup(node->vxlan, vni); 113 if (!vnode) 114 continue; 115 } else if (node->vxlan->default_dst.remote_vni != vni) { 116 continue; 117 } 118 119 if (IS_ENABLED(CONFIG_IPV6)) { 120 const struct vxlan_config *cfg = &node->vxlan->cfg; 121 122 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) && 123 cfg->remote_ifindex != ifindex) 124 continue; 125 } 126 127 if (vninode) 128 *vninode = vnode; 129 return node->vxlan; 130 } 131 132 return NULL; 133 } 134 135 /* Look up VNI in a per net namespace table */ 136 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex, 137 __be32 vni, sa_family_t family, 138 __be16 port, u32 flags) 139 { 140 struct vxlan_sock *vs; 141 142 vs = vxlan_find_sock(net, family, port, flags, ifindex); 143 if (!vs) 144 return NULL; 145 146 return vxlan_vs_find_vni(vs, ifindex, vni, NULL); 147 } 148 149 /* Fill in neighbour message in skbuff. */ 150 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, 151 const struct vxlan_fdb *fdb, 152 u32 portid, u32 seq, int type, unsigned int flags, 153 const struct vxlan_rdst *rdst) 154 { 155 unsigned long now = jiffies; 156 struct nda_cacheinfo ci; 157 bool send_ip, send_eth; 158 struct nlmsghdr *nlh; 159 struct nexthop *nh; 160 struct ndmsg *ndm; 161 int nh_family; 162 u32 nh_id; 163 164 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); 165 if (nlh == NULL) 166 return -EMSGSIZE; 167 168 ndm = nlmsg_data(nlh); 169 memset(ndm, 0, sizeof(*ndm)); 170 171 send_eth = send_ip = true; 172 173 rcu_read_lock(); 174 nh = rcu_dereference(fdb->nh); 175 if (nh) { 176 nh_family = nexthop_get_family(nh); 177 nh_id = nh->id; 178 } 179 rcu_read_unlock(); 180 181 if (type == RTM_GETNEIGH) { 182 if (rdst) { 183 send_ip = !vxlan_addr_any(&rdst->remote_ip); 184 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET; 185 } else if (nh) { 186 ndm->ndm_family = nh_family; 187 } 188 send_eth = !is_zero_ether_addr(fdb->eth_addr); 189 } else 190 ndm->ndm_family = AF_BRIDGE; 191 ndm->ndm_state = fdb->state; 192 ndm->ndm_ifindex = vxlan->dev->ifindex; 193 ndm->ndm_flags = fdb->flags; 194 if (rdst && rdst->offloaded) 195 ndm->ndm_flags |= NTF_OFFLOADED; 196 ndm->ndm_type = RTN_UNICAST; 197 198 if (!net_eq(dev_net(vxlan->dev), vxlan->net) && 199 nla_put_s32(skb, NDA_LINK_NETNSID, 200 peernet2id(dev_net(vxlan->dev), vxlan->net))) 201 goto nla_put_failure; 202 203 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) 204 goto nla_put_failure; 205 if (nh) { 206 if (nla_put_u32(skb, NDA_NH_ID, nh_id)) 207 goto nla_put_failure; 208 } else if (rdst) { 209 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, 210 &rdst->remote_ip)) 211 goto nla_put_failure; 212 213 if (rdst->remote_port && 214 rdst->remote_port != vxlan->cfg.dst_port && 215 nla_put_be16(skb, NDA_PORT, rdst->remote_port)) 216 goto nla_put_failure; 217 if (rdst->remote_vni != vxlan->default_dst.remote_vni && 218 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) 219 goto nla_put_failure; 220 if (rdst->remote_ifindex && 221 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) 222 goto nla_put_failure; 223 } 224 225 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni && 226 nla_put_u32(skb, NDA_SRC_VNI, 227 be32_to_cpu(fdb->vni))) 228 goto nla_put_failure; 229 230 ci.ndm_used = jiffies_to_clock_t(now - fdb->used); 231 ci.ndm_confirmed = 0; 232 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); 233 ci.ndm_refcnt = 0; 234 235 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) 236 goto nla_put_failure; 237 238 nlmsg_end(skb, nlh); 239 return 0; 240 241 nla_put_failure: 242 nlmsg_cancel(skb, nlh); 243 return -EMSGSIZE; 244 } 245 246 static inline size_t vxlan_nlmsg_size(void) 247 { 248 return NLMSG_ALIGN(sizeof(struct ndmsg)) 249 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ 250 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ 251 + nla_total_size(sizeof(__be16)) /* NDA_PORT */ 252 + nla_total_size(sizeof(__be32)) /* NDA_VNI */ 253 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ 254 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ 255 + nla_total_size(sizeof(struct nda_cacheinfo)); 256 } 257 258 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 259 struct vxlan_rdst *rd, int type) 260 { 261 struct net *net = dev_net(vxlan->dev); 262 struct sk_buff *skb; 263 int err = -ENOBUFS; 264 265 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); 266 if (skb == NULL) 267 goto errout; 268 269 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); 270 if (err < 0) { 271 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ 272 WARN_ON(err == -EMSGSIZE); 273 kfree_skb(skb); 274 goto errout; 275 } 276 277 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 278 return; 279 errout: 280 if (err < 0) 281 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 282 } 283 284 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan, 285 const struct vxlan_fdb *fdb, 286 const struct vxlan_rdst *rd, 287 struct netlink_ext_ack *extack, 288 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 289 { 290 fdb_info->info.dev = vxlan->dev; 291 fdb_info->info.extack = extack; 292 fdb_info->remote_ip = rd->remote_ip; 293 fdb_info->remote_port = rd->remote_port; 294 fdb_info->remote_vni = rd->remote_vni; 295 fdb_info->remote_ifindex = rd->remote_ifindex; 296 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN); 297 fdb_info->vni = fdb->vni; 298 fdb_info->offloaded = rd->offloaded; 299 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER; 300 } 301 302 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan, 303 struct vxlan_fdb *fdb, 304 struct vxlan_rdst *rd, 305 bool adding, 306 struct netlink_ext_ack *extack) 307 { 308 struct switchdev_notifier_vxlan_fdb_info info; 309 enum switchdev_notifier_type notifier_type; 310 int ret; 311 312 if (WARN_ON(!rd)) 313 return 0; 314 315 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE 316 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE; 317 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info); 318 ret = call_switchdev_notifiers(notifier_type, vxlan->dev, 319 &info.info, extack); 320 return notifier_to_errno(ret); 321 } 322 323 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 324 struct vxlan_rdst *rd, int type, bool swdev_notify, 325 struct netlink_ext_ack *extack) 326 { 327 int err; 328 329 if (swdev_notify && rd) { 330 switch (type) { 331 case RTM_NEWNEIGH: 332 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 333 true, extack); 334 if (err) 335 return err; 336 break; 337 case RTM_DELNEIGH: 338 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, 339 false, extack); 340 break; 341 } 342 } 343 344 __vxlan_fdb_notify(vxlan, fdb, rd, type); 345 return 0; 346 } 347 348 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) 349 { 350 struct vxlan_dev *vxlan = netdev_priv(dev); 351 struct vxlan_fdb f = { 352 .state = NUD_STALE, 353 }; 354 struct vxlan_rdst remote = { 355 .remote_ip = *ipa, /* goes to NDA_DST */ 356 .remote_vni = cpu_to_be32(VXLAN_N_VID), 357 }; 358 359 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 360 } 361 362 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) 363 { 364 struct vxlan_fdb f = { 365 .state = NUD_STALE, 366 }; 367 struct vxlan_rdst remote = { }; 368 369 memcpy(f.eth_addr, eth_addr, ETH_ALEN); 370 371 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL); 372 } 373 374 /* Hash Ethernet address */ 375 static u32 eth_hash(const unsigned char *addr) 376 { 377 u64 value = get_unaligned((u64 *)addr); 378 379 /* only want 6 bytes */ 380 #ifdef __BIG_ENDIAN 381 value >>= 16; 382 #else 383 value <<= 16; 384 #endif 385 return hash_64(value, FDB_HASH_BITS); 386 } 387 388 u32 eth_vni_hash(const unsigned char *addr, __be32 vni) 389 { 390 /* use 1 byte of OUI and 3 bytes of NIC */ 391 u32 key = get_unaligned((u32 *)(addr + 2)); 392 393 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1); 394 } 395 396 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni) 397 { 398 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) 399 return eth_vni_hash(mac, vni); 400 else 401 return eth_hash(mac); 402 } 403 404 /* Hash chain to use given mac address */ 405 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, 406 const u8 *mac, __be32 vni) 407 { 408 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)]; 409 } 410 411 /* Look up Ethernet address in forwarding table */ 412 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, 413 const u8 *mac, __be32 vni) 414 { 415 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni); 416 struct vxlan_fdb *f; 417 418 hlist_for_each_entry_rcu(f, head, hlist) { 419 if (ether_addr_equal(mac, f->eth_addr)) { 420 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 421 if (vni == f->vni) 422 return f; 423 } else { 424 return f; 425 } 426 } 427 } 428 429 return NULL; 430 } 431 432 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, 433 const u8 *mac, __be32 vni) 434 { 435 struct vxlan_fdb *f; 436 437 f = __vxlan_find_mac(vxlan, mac, vni); 438 if (f && f->used != jiffies) 439 f->used = jiffies; 440 441 return f; 442 } 443 444 /* caller should hold vxlan->hash_lock */ 445 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, 446 union vxlan_addr *ip, __be16 port, 447 __be32 vni, __u32 ifindex) 448 { 449 struct vxlan_rdst *rd; 450 451 list_for_each_entry(rd, &f->remotes, list) { 452 if (vxlan_addr_equal(&rd->remote_ip, ip) && 453 rd->remote_port == port && 454 rd->remote_vni == vni && 455 rd->remote_ifindex == ifindex) 456 return rd; 457 } 458 459 return NULL; 460 } 461 462 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni, 463 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 464 { 465 struct vxlan_dev *vxlan = netdev_priv(dev); 466 u8 eth_addr[ETH_ALEN + 2] = { 0 }; 467 struct vxlan_rdst *rdst; 468 struct vxlan_fdb *f; 469 int rc = 0; 470 471 if (is_multicast_ether_addr(mac) || 472 is_zero_ether_addr(mac)) 473 return -EINVAL; 474 475 ether_addr_copy(eth_addr, mac); 476 477 rcu_read_lock(); 478 479 f = __vxlan_find_mac(vxlan, eth_addr, vni); 480 if (!f) { 481 rc = -ENOENT; 482 goto out; 483 } 484 485 rdst = first_remote_rcu(f); 486 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info); 487 488 out: 489 rcu_read_unlock(); 490 return rc; 491 } 492 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc); 493 494 static int vxlan_fdb_notify_one(struct notifier_block *nb, 495 const struct vxlan_dev *vxlan, 496 const struct vxlan_fdb *f, 497 const struct vxlan_rdst *rdst, 498 struct netlink_ext_ack *extack) 499 { 500 struct switchdev_notifier_vxlan_fdb_info fdb_info; 501 int rc; 502 503 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info); 504 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE, 505 &fdb_info); 506 return notifier_to_errno(rc); 507 } 508 509 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni, 510 struct notifier_block *nb, 511 struct netlink_ext_ack *extack) 512 { 513 struct vxlan_dev *vxlan; 514 struct vxlan_rdst *rdst; 515 struct vxlan_fdb *f; 516 unsigned int h; 517 int rc = 0; 518 519 if (!netif_is_vxlan(dev)) 520 return -EINVAL; 521 vxlan = netdev_priv(dev); 522 523 for (h = 0; h < FDB_HASH_SIZE; ++h) { 524 spin_lock_bh(&vxlan->hash_lock[h]); 525 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) { 526 if (f->vni == vni) { 527 list_for_each_entry(rdst, &f->remotes, list) { 528 rc = vxlan_fdb_notify_one(nb, vxlan, 529 f, rdst, 530 extack); 531 if (rc) 532 goto unlock; 533 } 534 } 535 } 536 spin_unlock_bh(&vxlan->hash_lock[h]); 537 } 538 return 0; 539 540 unlock: 541 spin_unlock_bh(&vxlan->hash_lock[h]); 542 return rc; 543 } 544 EXPORT_SYMBOL_GPL(vxlan_fdb_replay); 545 546 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni) 547 { 548 struct vxlan_dev *vxlan; 549 struct vxlan_rdst *rdst; 550 struct vxlan_fdb *f; 551 unsigned int h; 552 553 if (!netif_is_vxlan(dev)) 554 return; 555 vxlan = netdev_priv(dev); 556 557 for (h = 0; h < FDB_HASH_SIZE; ++h) { 558 spin_lock_bh(&vxlan->hash_lock[h]); 559 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) 560 if (f->vni == vni) 561 list_for_each_entry(rdst, &f->remotes, list) 562 rdst->offloaded = false; 563 spin_unlock_bh(&vxlan->hash_lock[h]); 564 } 565 566 } 567 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload); 568 569 /* Replace destination of unicast mac */ 570 static int vxlan_fdb_replace(struct vxlan_fdb *f, 571 union vxlan_addr *ip, __be16 port, __be32 vni, 572 __u32 ifindex, struct vxlan_rdst *oldrd) 573 { 574 struct vxlan_rdst *rd; 575 576 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 577 if (rd) 578 return 0; 579 580 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); 581 if (!rd) 582 return 0; 583 584 *oldrd = *rd; 585 dst_cache_reset(&rd->dst_cache); 586 rd->remote_ip = *ip; 587 rd->remote_port = port; 588 rd->remote_vni = vni; 589 rd->remote_ifindex = ifindex; 590 rd->offloaded = false; 591 return 1; 592 } 593 594 /* Add/update destinations for multicast */ 595 static int vxlan_fdb_append(struct vxlan_fdb *f, 596 union vxlan_addr *ip, __be16 port, __be32 vni, 597 __u32 ifindex, struct vxlan_rdst **rdp) 598 { 599 struct vxlan_rdst *rd; 600 601 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); 602 if (rd) 603 return 0; 604 605 rd = kmalloc(sizeof(*rd), GFP_ATOMIC); 606 if (rd == NULL) 607 return -ENOMEM; 608 609 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { 610 kfree(rd); 611 return -ENOMEM; 612 } 613 614 rd->remote_ip = *ip; 615 rd->remote_port = port; 616 rd->offloaded = false; 617 rd->remote_vni = vni; 618 rd->remote_ifindex = ifindex; 619 620 list_add_tail_rcu(&rd->list, &f->remotes); 621 622 *rdp = rd; 623 return 1; 624 } 625 626 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, 627 unsigned int off, 628 struct vxlanhdr *vh, size_t hdrlen, 629 __be32 vni_field, 630 struct gro_remcsum *grc, 631 bool nopartial) 632 { 633 size_t start, offset; 634 635 if (skb->remcsum_offload) 636 return vh; 637 638 if (!NAPI_GRO_CB(skb)->csum_valid) 639 return NULL; 640 641 start = vxlan_rco_start(vni_field); 642 offset = start + vxlan_rco_offset(vni_field); 643 644 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, 645 start, offset, grc, nopartial); 646 647 skb->remcsum_offload = 1; 648 649 return vh; 650 } 651 652 static struct sk_buff *vxlan_gro_receive(struct sock *sk, 653 struct list_head *head, 654 struct sk_buff *skb) 655 { 656 struct sk_buff *pp = NULL; 657 struct sk_buff *p; 658 struct vxlanhdr *vh, *vh2; 659 unsigned int hlen, off_vx; 660 int flush = 1; 661 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); 662 __be32 flags; 663 struct gro_remcsum grc; 664 665 skb_gro_remcsum_init(&grc); 666 667 off_vx = skb_gro_offset(skb); 668 hlen = off_vx + sizeof(*vh); 669 vh = skb_gro_header(skb, hlen, off_vx); 670 if (unlikely(!vh)) 671 goto out; 672 673 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); 674 675 flags = vh->vx_flags; 676 677 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { 678 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), 679 vh->vx_vni, &grc, 680 !!(vs->flags & 681 VXLAN_F_REMCSUM_NOPARTIAL)); 682 683 if (!vh) 684 goto out; 685 } 686 687 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ 688 689 list_for_each_entry(p, head, list) { 690 if (!NAPI_GRO_CB(p)->same_flow) 691 continue; 692 693 vh2 = (struct vxlanhdr *)(p->data + off_vx); 694 if (vh->vx_flags != vh2->vx_flags || 695 vh->vx_vni != vh2->vx_vni) { 696 NAPI_GRO_CB(p)->same_flow = 0; 697 continue; 698 } 699 } 700 701 pp = call_gro_receive(eth_gro_receive, head, skb); 702 flush = 0; 703 704 out: 705 skb_gro_flush_final_remcsum(skb, pp, flush, &grc); 706 707 return pp; 708 } 709 710 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) 711 { 712 /* Sets 'skb->inner_mac_header' since we are always called with 713 * 'skb->encapsulation' set. 714 */ 715 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); 716 } 717 718 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac, 719 __u16 state, __be32 src_vni, 720 __u16 ndm_flags) 721 { 722 struct vxlan_fdb *f; 723 724 f = kmalloc(sizeof(*f), GFP_ATOMIC); 725 if (!f) 726 return NULL; 727 f->state = state; 728 f->flags = ndm_flags; 729 f->updated = f->used = jiffies; 730 f->vni = src_vni; 731 f->nh = NULL; 732 RCU_INIT_POINTER(f->vdev, vxlan); 733 INIT_LIST_HEAD(&f->nh_list); 734 INIT_LIST_HEAD(&f->remotes); 735 memcpy(f->eth_addr, mac, ETH_ALEN); 736 737 return f; 738 } 739 740 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac, 741 __be32 src_vni, struct vxlan_fdb *f) 742 { 743 ++vxlan->addrcnt; 744 hlist_add_head_rcu(&f->hlist, 745 vxlan_fdb_head(vxlan, mac, src_vni)); 746 } 747 748 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, 749 u32 nhid, struct netlink_ext_ack *extack) 750 { 751 struct nexthop *old_nh = rtnl_dereference(fdb->nh); 752 struct nexthop *nh; 753 int err = -EINVAL; 754 755 if (old_nh && old_nh->id == nhid) 756 return 0; 757 758 nh = nexthop_find_by_id(vxlan->net, nhid); 759 if (!nh) { 760 NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); 761 goto err_inval; 762 } 763 764 if (!nexthop_get(nh)) { 765 NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); 766 nh = NULL; 767 goto err_inval; 768 } 769 if (!nexthop_is_fdb(nh)) { 770 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop"); 771 goto err_inval; 772 } 773 774 if (!nexthop_is_multipath(nh)) { 775 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group"); 776 goto err_inval; 777 } 778 779 /* check nexthop group family */ 780 switch (vxlan->default_dst.remote_ip.sa.sa_family) { 781 case AF_INET: 782 if (!nexthop_has_v4(nh)) { 783 err = -EAFNOSUPPORT; 784 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 785 goto err_inval; 786 } 787 break; 788 case AF_INET6: 789 if (nexthop_has_v4(nh)) { 790 err = -EAFNOSUPPORT; 791 NL_SET_ERR_MSG(extack, "Nexthop group family not supported"); 792 goto err_inval; 793 } 794 } 795 796 if (old_nh) { 797 list_del_rcu(&fdb->nh_list); 798 nexthop_put(old_nh); 799 } 800 rcu_assign_pointer(fdb->nh, nh); 801 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list); 802 return 1; 803 804 err_inval: 805 if (nh) 806 nexthop_put(nh); 807 return err; 808 } 809 810 int vxlan_fdb_create(struct vxlan_dev *vxlan, 811 const u8 *mac, union vxlan_addr *ip, 812 __u16 state, __be16 port, __be32 src_vni, 813 __be32 vni, __u32 ifindex, __u16 ndm_flags, 814 u32 nhid, struct vxlan_fdb **fdb, 815 struct netlink_ext_ack *extack) 816 { 817 struct vxlan_rdst *rd = NULL; 818 struct vxlan_fdb *f; 819 int rc; 820 821 if (vxlan->cfg.addrmax && 822 vxlan->addrcnt >= vxlan->cfg.addrmax) 823 return -ENOSPC; 824 825 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 826 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags); 827 if (!f) 828 return -ENOMEM; 829 830 if (nhid) 831 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 832 else 833 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 834 if (rc < 0) 835 goto errout; 836 837 *fdb = f; 838 839 return 0; 840 841 errout: 842 kfree(f); 843 return rc; 844 } 845 846 static void __vxlan_fdb_free(struct vxlan_fdb *f) 847 { 848 struct vxlan_rdst *rd, *nd; 849 struct nexthop *nh; 850 851 nh = rcu_dereference_raw(f->nh); 852 if (nh) { 853 rcu_assign_pointer(f->nh, NULL); 854 rcu_assign_pointer(f->vdev, NULL); 855 nexthop_put(nh); 856 } 857 858 list_for_each_entry_safe(rd, nd, &f->remotes, list) { 859 dst_cache_destroy(&rd->dst_cache); 860 kfree(rd); 861 } 862 kfree(f); 863 } 864 865 static void vxlan_fdb_free(struct rcu_head *head) 866 { 867 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); 868 869 __vxlan_fdb_free(f); 870 } 871 872 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 873 bool do_notify, bool swdev_notify) 874 { 875 struct vxlan_rdst *rd; 876 877 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr); 878 879 --vxlan->addrcnt; 880 if (do_notify) { 881 if (rcu_access_pointer(f->nh)) 882 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH, 883 swdev_notify, NULL); 884 else 885 list_for_each_entry(rd, &f->remotes, list) 886 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, 887 swdev_notify, NULL); 888 } 889 890 hlist_del_rcu(&f->hlist); 891 list_del_rcu(&f->nh_list); 892 call_rcu(&f->rcu, vxlan_fdb_free); 893 } 894 895 static void vxlan_dst_free(struct rcu_head *head) 896 { 897 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu); 898 899 dst_cache_destroy(&rd->dst_cache); 900 kfree(rd); 901 } 902 903 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan, 904 union vxlan_addr *ip, 905 __u16 state, __u16 flags, 906 __be16 port, __be32 vni, 907 __u32 ifindex, __u16 ndm_flags, 908 struct vxlan_fdb *f, u32 nhid, 909 bool swdev_notify, 910 struct netlink_ext_ack *extack) 911 { 912 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 913 struct vxlan_rdst *rd = NULL; 914 struct vxlan_rdst oldrd; 915 int notify = 0; 916 int rc = 0; 917 int err; 918 919 if (nhid && !rcu_access_pointer(f->nh)) { 920 NL_SET_ERR_MSG(extack, 921 "Cannot replace an existing non nexthop fdb with a nexthop"); 922 return -EOPNOTSUPP; 923 } 924 925 if (nhid && (flags & NLM_F_APPEND)) { 926 NL_SET_ERR_MSG(extack, 927 "Cannot append to a nexthop fdb"); 928 return -EOPNOTSUPP; 929 } 930 931 /* Do not allow an externally learned entry to take over an entry added 932 * by the user. 933 */ 934 if (!(fdb_flags & NTF_EXT_LEARNED) || 935 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) { 936 if (f->state != state) { 937 f->state = state; 938 f->updated = jiffies; 939 notify = 1; 940 } 941 if (f->flags != fdb_flags) { 942 f->flags = fdb_flags; 943 f->updated = jiffies; 944 notify = 1; 945 } 946 } 947 948 if ((flags & NLM_F_REPLACE)) { 949 /* Only change unicasts */ 950 if (!(is_multicast_ether_addr(f->eth_addr) || 951 is_zero_ether_addr(f->eth_addr))) { 952 if (nhid) { 953 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack); 954 if (rc < 0) 955 return rc; 956 } else { 957 rc = vxlan_fdb_replace(f, ip, port, vni, 958 ifindex, &oldrd); 959 } 960 notify |= rc; 961 } else { 962 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries"); 963 return -EOPNOTSUPP; 964 } 965 } 966 if ((flags & NLM_F_APPEND) && 967 (is_multicast_ether_addr(f->eth_addr) || 968 is_zero_ether_addr(f->eth_addr))) { 969 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); 970 971 if (rc < 0) 972 return rc; 973 notify |= rc; 974 } 975 976 if (ndm_flags & NTF_USE) 977 f->used = jiffies; 978 979 if (notify) { 980 if (rd == NULL) 981 rd = first_remote_rtnl(f); 982 983 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH, 984 swdev_notify, extack); 985 if (err) 986 goto err_notify; 987 } 988 989 return 0; 990 991 err_notify: 992 if (nhid) 993 return err; 994 if ((flags & NLM_F_REPLACE) && rc) 995 *rd = oldrd; 996 else if ((flags & NLM_F_APPEND) && rc) { 997 list_del_rcu(&rd->list); 998 call_rcu(&rd->rcu, vxlan_dst_free); 999 } 1000 return err; 1001 } 1002 1003 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan, 1004 const u8 *mac, union vxlan_addr *ip, 1005 __u16 state, __u16 flags, 1006 __be16 port, __be32 src_vni, __be32 vni, 1007 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1008 bool swdev_notify, 1009 struct netlink_ext_ack *extack) 1010 { 1011 __u16 fdb_flags = (ndm_flags & ~NTF_USE); 1012 struct vxlan_fdb *f; 1013 int rc; 1014 1015 /* Disallow replace to add a multicast entry */ 1016 if ((flags & NLM_F_REPLACE) && 1017 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) 1018 return -EOPNOTSUPP; 1019 1020 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); 1021 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni, 1022 vni, ifindex, fdb_flags, nhid, &f, extack); 1023 if (rc < 0) 1024 return rc; 1025 1026 vxlan_fdb_insert(vxlan, mac, src_vni, f); 1027 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH, 1028 swdev_notify, extack); 1029 if (rc) 1030 goto err_notify; 1031 1032 return 0; 1033 1034 err_notify: 1035 vxlan_fdb_destroy(vxlan, f, false, false); 1036 return rc; 1037 } 1038 1039 /* Add new entry to forwarding table -- assumes lock held */ 1040 int vxlan_fdb_update(struct vxlan_dev *vxlan, 1041 const u8 *mac, union vxlan_addr *ip, 1042 __u16 state, __u16 flags, 1043 __be16 port, __be32 src_vni, __be32 vni, 1044 __u32 ifindex, __u16 ndm_flags, u32 nhid, 1045 bool swdev_notify, 1046 struct netlink_ext_ack *extack) 1047 { 1048 struct vxlan_fdb *f; 1049 1050 f = __vxlan_find_mac(vxlan, mac, src_vni); 1051 if (f) { 1052 if (flags & NLM_F_EXCL) { 1053 netdev_dbg(vxlan->dev, 1054 "lost race to create %pM\n", mac); 1055 return -EEXIST; 1056 } 1057 1058 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port, 1059 vni, ifindex, ndm_flags, f, 1060 nhid, swdev_notify, extack); 1061 } else { 1062 if (!(flags & NLM_F_CREATE)) 1063 return -ENOENT; 1064 1065 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags, 1066 port, src_vni, vni, ifindex, 1067 ndm_flags, nhid, swdev_notify, 1068 extack); 1069 } 1070 } 1071 1072 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f, 1073 struct vxlan_rdst *rd, bool swdev_notify) 1074 { 1075 list_del_rcu(&rd->list); 1076 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL); 1077 call_rcu(&rd->rcu, vxlan_dst_free); 1078 } 1079 1080 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, 1081 union vxlan_addr *ip, __be16 *port, __be32 *src_vni, 1082 __be32 *vni, u32 *ifindex, u32 *nhid, 1083 struct netlink_ext_ack *extack) 1084 { 1085 struct net *net = dev_net(vxlan->dev); 1086 int err; 1087 1088 if (tb[NDA_NH_ID] && 1089 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) { 1090 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID"); 1091 return -EINVAL; 1092 } 1093 1094 if (tb[NDA_DST]) { 1095 err = vxlan_nla_get_addr(ip, tb[NDA_DST]); 1096 if (err) { 1097 NL_SET_ERR_MSG(extack, "Unsupported address family"); 1098 return err; 1099 } 1100 } else { 1101 union vxlan_addr *remote = &vxlan->default_dst.remote_ip; 1102 1103 if (remote->sa.sa_family == AF_INET) { 1104 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); 1105 ip->sa.sa_family = AF_INET; 1106 #if IS_ENABLED(CONFIG_IPV6) 1107 } else { 1108 ip->sin6.sin6_addr = in6addr_any; 1109 ip->sa.sa_family = AF_INET6; 1110 #endif 1111 } 1112 } 1113 1114 if (tb[NDA_PORT]) { 1115 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) { 1116 NL_SET_ERR_MSG(extack, "Invalid vxlan port"); 1117 return -EINVAL; 1118 } 1119 *port = nla_get_be16(tb[NDA_PORT]); 1120 } else { 1121 *port = vxlan->cfg.dst_port; 1122 } 1123 1124 if (tb[NDA_VNI]) { 1125 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) { 1126 NL_SET_ERR_MSG(extack, "Invalid vni"); 1127 return -EINVAL; 1128 } 1129 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1130 } else { 1131 *vni = vxlan->default_dst.remote_vni; 1132 } 1133 1134 if (tb[NDA_SRC_VNI]) { 1135 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) { 1136 NL_SET_ERR_MSG(extack, "Invalid src vni"); 1137 return -EINVAL; 1138 } 1139 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI])); 1140 } else { 1141 *src_vni = vxlan->default_dst.remote_vni; 1142 } 1143 1144 if (tb[NDA_IFINDEX]) { 1145 struct net_device *tdev; 1146 1147 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) { 1148 NL_SET_ERR_MSG(extack, "Invalid ifindex"); 1149 return -EINVAL; 1150 } 1151 *ifindex = nla_get_u32(tb[NDA_IFINDEX]); 1152 tdev = __dev_get_by_index(net, *ifindex); 1153 if (!tdev) { 1154 NL_SET_ERR_MSG(extack, "Device not found"); 1155 return -EADDRNOTAVAIL; 1156 } 1157 } else { 1158 *ifindex = 0; 1159 } 1160 1161 if (tb[NDA_NH_ID]) 1162 *nhid = nla_get_u32(tb[NDA_NH_ID]); 1163 else 1164 *nhid = 0; 1165 1166 return 0; 1167 } 1168 1169 /* Add static entry (via netlink) */ 1170 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1171 struct net_device *dev, 1172 const unsigned char *addr, u16 vid, u16 flags, 1173 struct netlink_ext_ack *extack) 1174 { 1175 struct vxlan_dev *vxlan = netdev_priv(dev); 1176 /* struct net *net = dev_net(vxlan->dev); */ 1177 union vxlan_addr ip; 1178 __be16 port; 1179 __be32 src_vni, vni; 1180 u32 ifindex, nhid; 1181 u32 hash_index; 1182 int err; 1183 1184 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { 1185 pr_info("RTM_NEWNEIGH with invalid state %#x\n", 1186 ndm->ndm_state); 1187 return -EINVAL; 1188 } 1189 1190 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID])) 1191 return -EINVAL; 1192 1193 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1194 &nhid, extack); 1195 if (err) 1196 return err; 1197 1198 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) 1199 return -EAFNOSUPPORT; 1200 1201 hash_index = fdb_head_index(vxlan, addr, src_vni); 1202 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1203 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags, 1204 port, src_vni, vni, ifindex, 1205 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER, 1206 nhid, true, extack); 1207 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1208 1209 return err; 1210 } 1211 1212 int __vxlan_fdb_delete(struct vxlan_dev *vxlan, 1213 const unsigned char *addr, union vxlan_addr ip, 1214 __be16 port, __be32 src_vni, __be32 vni, 1215 u32 ifindex, bool swdev_notify) 1216 { 1217 struct vxlan_rdst *rd = NULL; 1218 struct vxlan_fdb *f; 1219 int err = -ENOENT; 1220 1221 f = vxlan_find_mac(vxlan, addr, src_vni); 1222 if (!f) 1223 return err; 1224 1225 if (!vxlan_addr_any(&ip)) { 1226 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); 1227 if (!rd) 1228 goto out; 1229 } 1230 1231 /* remove a destination if it's not the only one on the list, 1232 * otherwise destroy the fdb entry 1233 */ 1234 if (rd && !list_is_singular(&f->remotes)) { 1235 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify); 1236 goto out; 1237 } 1238 1239 vxlan_fdb_destroy(vxlan, f, true, swdev_notify); 1240 1241 out: 1242 return 0; 1243 } 1244 1245 /* Delete entry (via netlink) */ 1246 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], 1247 struct net_device *dev, 1248 const unsigned char *addr, u16 vid, 1249 struct netlink_ext_ack *extack) 1250 { 1251 struct vxlan_dev *vxlan = netdev_priv(dev); 1252 union vxlan_addr ip; 1253 __be32 src_vni, vni; 1254 u32 ifindex, nhid; 1255 u32 hash_index; 1256 __be16 port; 1257 int err; 1258 1259 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex, 1260 &nhid, extack); 1261 if (err) 1262 return err; 1263 1264 hash_index = fdb_head_index(vxlan, addr, src_vni); 1265 spin_lock_bh(&vxlan->hash_lock[hash_index]); 1266 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex, 1267 true); 1268 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 1269 1270 return err; 1271 } 1272 1273 /* Dump forwarding table */ 1274 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, 1275 struct net_device *dev, 1276 struct net_device *filter_dev, int *idx) 1277 { 1278 struct vxlan_dev *vxlan = netdev_priv(dev); 1279 unsigned int h; 1280 int err = 0; 1281 1282 for (h = 0; h < FDB_HASH_SIZE; ++h) { 1283 struct vxlan_fdb *f; 1284 1285 rcu_read_lock(); 1286 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { 1287 struct vxlan_rdst *rd; 1288 1289 if (rcu_access_pointer(f->nh)) { 1290 if (*idx < cb->args[2]) 1291 goto skip_nh; 1292 err = vxlan_fdb_info(skb, vxlan, f, 1293 NETLINK_CB(cb->skb).portid, 1294 cb->nlh->nlmsg_seq, 1295 RTM_NEWNEIGH, 1296 NLM_F_MULTI, NULL); 1297 if (err < 0) { 1298 rcu_read_unlock(); 1299 goto out; 1300 } 1301 skip_nh: 1302 *idx += 1; 1303 continue; 1304 } 1305 1306 list_for_each_entry_rcu(rd, &f->remotes, list) { 1307 if (*idx < cb->args[2]) 1308 goto skip; 1309 1310 err = vxlan_fdb_info(skb, vxlan, f, 1311 NETLINK_CB(cb->skb).portid, 1312 cb->nlh->nlmsg_seq, 1313 RTM_NEWNEIGH, 1314 NLM_F_MULTI, rd); 1315 if (err < 0) { 1316 rcu_read_unlock(); 1317 goto out; 1318 } 1319 skip: 1320 *idx += 1; 1321 } 1322 } 1323 rcu_read_unlock(); 1324 } 1325 out: 1326 return err; 1327 } 1328 1329 static int vxlan_fdb_get(struct sk_buff *skb, 1330 struct nlattr *tb[], 1331 struct net_device *dev, 1332 const unsigned char *addr, 1333 u16 vid, u32 portid, u32 seq, 1334 struct netlink_ext_ack *extack) 1335 { 1336 struct vxlan_dev *vxlan = netdev_priv(dev); 1337 struct vxlan_fdb *f; 1338 __be32 vni; 1339 int err; 1340 1341 if (tb[NDA_VNI]) 1342 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); 1343 else 1344 vni = vxlan->default_dst.remote_vni; 1345 1346 rcu_read_lock(); 1347 1348 f = __vxlan_find_mac(vxlan, addr, vni); 1349 if (!f) { 1350 NL_SET_ERR_MSG(extack, "Fdb entry not found"); 1351 err = -ENOENT; 1352 goto errout; 1353 } 1354 1355 err = vxlan_fdb_info(skb, vxlan, f, portid, seq, 1356 RTM_NEWNEIGH, 0, first_remote_rcu(f)); 1357 errout: 1358 rcu_read_unlock(); 1359 return err; 1360 } 1361 1362 /* Watch incoming packets to learn mapping between Ethernet address 1363 * and Tunnel endpoint. 1364 * Return true if packet is bogus and should be dropped. 1365 */ 1366 static bool vxlan_snoop(struct net_device *dev, 1367 union vxlan_addr *src_ip, const u8 *src_mac, 1368 u32 src_ifindex, __be32 vni) 1369 { 1370 struct vxlan_dev *vxlan = netdev_priv(dev); 1371 struct vxlan_fdb *f; 1372 u32 ifindex = 0; 1373 1374 #if IS_ENABLED(CONFIG_IPV6) 1375 if (src_ip->sa.sa_family == AF_INET6 && 1376 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL)) 1377 ifindex = src_ifindex; 1378 #endif 1379 1380 f = vxlan_find_mac(vxlan, src_mac, vni); 1381 if (likely(f)) { 1382 struct vxlan_rdst *rdst = first_remote_rcu(f); 1383 1384 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) && 1385 rdst->remote_ifindex == ifindex)) 1386 return false; 1387 1388 /* Don't migrate static entries, drop packets */ 1389 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 1390 return true; 1391 1392 /* Don't override an fdb with nexthop with a learnt entry */ 1393 if (rcu_access_pointer(f->nh)) 1394 return true; 1395 1396 if (net_ratelimit()) 1397 netdev_info(dev, 1398 "%pM migrated from %pIS to %pIS\n", 1399 src_mac, &rdst->remote_ip.sa, &src_ip->sa); 1400 1401 rdst->remote_ip = *src_ip; 1402 f->updated = jiffies; 1403 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL); 1404 } else { 1405 u32 hash_index = fdb_head_index(vxlan, src_mac, vni); 1406 1407 /* learned new entry */ 1408 spin_lock(&vxlan->hash_lock[hash_index]); 1409 1410 /* close off race between vxlan_flush and incoming packets */ 1411 if (netif_running(dev)) 1412 vxlan_fdb_update(vxlan, src_mac, src_ip, 1413 NUD_REACHABLE, 1414 NLM_F_EXCL|NLM_F_CREATE, 1415 vxlan->cfg.dst_port, 1416 vni, 1417 vxlan->default_dst.remote_vni, 1418 ifindex, NTF_SELF, 0, true, NULL); 1419 spin_unlock(&vxlan->hash_lock[hash_index]); 1420 } 1421 1422 return false; 1423 } 1424 1425 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) 1426 { 1427 struct vxlan_net *vn; 1428 1429 if (!vs) 1430 return false; 1431 if (!refcount_dec_and_test(&vs->refcnt)) 1432 return false; 1433 1434 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); 1435 spin_lock(&vn->sock_lock); 1436 hlist_del_rcu(&vs->hlist); 1437 udp_tunnel_notify_del_rx_port(vs->sock, 1438 (vs->flags & VXLAN_F_GPE) ? 1439 UDP_TUNNEL_TYPE_VXLAN_GPE : 1440 UDP_TUNNEL_TYPE_VXLAN); 1441 spin_unlock(&vn->sock_lock); 1442 1443 return true; 1444 } 1445 1446 static void vxlan_sock_release(struct vxlan_dev *vxlan) 1447 { 1448 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock); 1449 #if IS_ENABLED(CONFIG_IPV6) 1450 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock); 1451 1452 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 1453 #endif 1454 1455 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 1456 synchronize_net(); 1457 1458 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 1459 vxlan_vs_del_vnigrp(vxlan); 1460 else 1461 vxlan_vs_del_dev(vxlan); 1462 1463 if (__vxlan_sock_release_prep(sock4)) { 1464 udp_tunnel_sock_release(sock4->sock); 1465 kfree(sock4); 1466 } 1467 1468 #if IS_ENABLED(CONFIG_IPV6) 1469 if (__vxlan_sock_release_prep(sock6)) { 1470 udp_tunnel_sock_release(sock6->sock); 1471 kfree(sock6); 1472 } 1473 #endif 1474 } 1475 1476 static bool vxlan_remcsum(struct vxlanhdr *unparsed, 1477 struct sk_buff *skb, u32 vxflags) 1478 { 1479 size_t start, offset; 1480 1481 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) 1482 goto out; 1483 1484 start = vxlan_rco_start(unparsed->vx_vni); 1485 offset = start + vxlan_rco_offset(unparsed->vx_vni); 1486 1487 if (!pskb_may_pull(skb, offset + sizeof(u16))) 1488 return false; 1489 1490 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, 1491 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); 1492 out: 1493 unparsed->vx_flags &= ~VXLAN_HF_RCO; 1494 unparsed->vx_vni &= VXLAN_VNI_MASK; 1495 return true; 1496 } 1497 1498 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed, 1499 struct sk_buff *skb, u32 vxflags, 1500 struct vxlan_metadata *md) 1501 { 1502 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed; 1503 struct metadata_dst *tun_dst; 1504 1505 if (!(unparsed->vx_flags & VXLAN_HF_GBP)) 1506 goto out; 1507 1508 md->gbp = ntohs(gbp->policy_id); 1509 1510 tun_dst = (struct metadata_dst *)skb_dst(skb); 1511 if (tun_dst) { 1512 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT; 1513 tun_dst->u.tun_info.options_len = sizeof(*md); 1514 } 1515 if (gbp->dont_learn) 1516 md->gbp |= VXLAN_GBP_DONT_LEARN; 1517 1518 if (gbp->policy_applied) 1519 md->gbp |= VXLAN_GBP_POLICY_APPLIED; 1520 1521 /* In flow-based mode, GBP is carried in dst_metadata */ 1522 if (!(vxflags & VXLAN_F_COLLECT_METADATA)) 1523 skb->mark = md->gbp; 1524 out: 1525 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS; 1526 } 1527 1528 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed, 1529 __be16 *protocol, 1530 struct sk_buff *skb, u32 vxflags) 1531 { 1532 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed; 1533 1534 /* Need to have Next Protocol set for interfaces in GPE mode. */ 1535 if (!gpe->np_applied) 1536 return false; 1537 /* "The initial version is 0. If a receiver does not support the 1538 * version indicated it MUST drop the packet. 1539 */ 1540 if (gpe->version != 0) 1541 return false; 1542 /* "When the O bit is set to 1, the packet is an OAM packet and OAM 1543 * processing MUST occur." However, we don't implement OAM 1544 * processing, thus drop the packet. 1545 */ 1546 if (gpe->oam_flag) 1547 return false; 1548 1549 *protocol = tun_p_to_eth_p(gpe->next_protocol); 1550 if (!*protocol) 1551 return false; 1552 1553 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS; 1554 return true; 1555 } 1556 1557 static bool vxlan_set_mac(struct vxlan_dev *vxlan, 1558 struct vxlan_sock *vs, 1559 struct sk_buff *skb, __be32 vni) 1560 { 1561 union vxlan_addr saddr; 1562 u32 ifindex = skb->dev->ifindex; 1563 1564 skb_reset_mac_header(skb); 1565 skb->protocol = eth_type_trans(skb, vxlan->dev); 1566 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 1567 1568 /* Ignore packet loops (and multicast echo) */ 1569 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) 1570 return false; 1571 1572 /* Get address from the outer IP header */ 1573 if (vxlan_get_sk_family(vs) == AF_INET) { 1574 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; 1575 saddr.sa.sa_family = AF_INET; 1576 #if IS_ENABLED(CONFIG_IPV6) 1577 } else { 1578 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; 1579 saddr.sa.sa_family = AF_INET6; 1580 #endif 1581 } 1582 1583 if ((vxlan->cfg.flags & VXLAN_F_LEARN) && 1584 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni)) 1585 return false; 1586 1587 return true; 1588 } 1589 1590 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, 1591 struct sk_buff *skb) 1592 { 1593 int err = 0; 1594 1595 if (vxlan_get_sk_family(vs) == AF_INET) 1596 err = IP_ECN_decapsulate(oiph, skb); 1597 #if IS_ENABLED(CONFIG_IPV6) 1598 else 1599 err = IP6_ECN_decapsulate(oiph, skb); 1600 #endif 1601 1602 if (unlikely(err) && log_ecn_error) { 1603 if (vxlan_get_sk_family(vs) == AF_INET) 1604 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 1605 &((struct iphdr *)oiph)->saddr, 1606 ((struct iphdr *)oiph)->tos); 1607 else 1608 net_info_ratelimited("non-ECT from %pI6\n", 1609 &((struct ipv6hdr *)oiph)->saddr); 1610 } 1611 return err <= 1; 1612 } 1613 1614 /* Callback from net/ipv4/udp.c to receive packets */ 1615 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) 1616 { 1617 struct vxlan_vni_node *vninode = NULL; 1618 struct vxlan_dev *vxlan; 1619 struct vxlan_sock *vs; 1620 struct vxlanhdr unparsed; 1621 struct vxlan_metadata _md; 1622 struct vxlan_metadata *md = &_md; 1623 __be16 protocol = htons(ETH_P_TEB); 1624 bool raw_proto = false; 1625 void *oiph; 1626 __be32 vni = 0; 1627 1628 /* Need UDP and VXLAN header to be present */ 1629 if (!pskb_may_pull(skb, VXLAN_HLEN)) 1630 goto drop; 1631 1632 unparsed = *vxlan_hdr(skb); 1633 /* VNI flag always required to be set */ 1634 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { 1635 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", 1636 ntohl(vxlan_hdr(skb)->vx_flags), 1637 ntohl(vxlan_hdr(skb)->vx_vni)); 1638 /* Return non vxlan pkt */ 1639 goto drop; 1640 } 1641 unparsed.vx_flags &= ~VXLAN_HF_VNI; 1642 unparsed.vx_vni &= ~VXLAN_VNI_MASK; 1643 1644 vs = rcu_dereference_sk_user_data(sk); 1645 if (!vs) 1646 goto drop; 1647 1648 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); 1649 1650 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode); 1651 if (!vxlan) 1652 goto drop; 1653 1654 /* For backwards compatibility, only allow reserved fields to be 1655 * used by VXLAN extensions if explicitly requested. 1656 */ 1657 if (vs->flags & VXLAN_F_GPE) { 1658 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags)) 1659 goto drop; 1660 raw_proto = true; 1661 } 1662 1663 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, 1664 !net_eq(vxlan->net, dev_net(vxlan->dev)))) 1665 goto drop; 1666 1667 if (vs->flags & VXLAN_F_REMCSUM_RX) 1668 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags))) 1669 goto drop; 1670 1671 if (vxlan_collect_metadata(vs)) { 1672 struct metadata_dst *tun_dst; 1673 1674 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY, 1675 key32_to_tunnel_id(vni), sizeof(*md)); 1676 1677 if (!tun_dst) 1678 goto drop; 1679 1680 md = ip_tunnel_info_opts(&tun_dst->u.tun_info); 1681 1682 skb_dst_set(skb, (struct dst_entry *)tun_dst); 1683 } else { 1684 memset(md, 0, sizeof(*md)); 1685 } 1686 1687 if (vs->flags & VXLAN_F_GBP) 1688 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); 1689 /* Note that GBP and GPE can never be active together. This is 1690 * ensured in vxlan_dev_configure. 1691 */ 1692 1693 if (unparsed.vx_flags || unparsed.vx_vni) { 1694 /* If there are any unprocessed flags remaining treat 1695 * this as a malformed packet. This behavior diverges from 1696 * VXLAN RFC (RFC7348) which stipulates that bits in reserved 1697 * in reserved fields are to be ignored. The approach here 1698 * maintains compatibility with previous stack code, and also 1699 * is more robust and provides a little more security in 1700 * adding extensions to VXLAN. 1701 */ 1702 goto drop; 1703 } 1704 1705 if (!raw_proto) { 1706 if (!vxlan_set_mac(vxlan, vs, skb, vni)) 1707 goto drop; 1708 } else { 1709 skb_reset_mac_header(skb); 1710 skb->dev = vxlan->dev; 1711 skb->pkt_type = PACKET_HOST; 1712 } 1713 1714 oiph = skb_network_header(skb); 1715 skb_reset_network_header(skb); 1716 1717 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { 1718 ++vxlan->dev->stats.rx_frame_errors; 1719 ++vxlan->dev->stats.rx_errors; 1720 vxlan_vnifilter_count(vxlan, vni, vninode, 1721 VXLAN_VNI_STATS_RX_ERRORS, 0); 1722 goto drop; 1723 } 1724 1725 rcu_read_lock(); 1726 1727 if (unlikely(!(vxlan->dev->flags & IFF_UP))) { 1728 rcu_read_unlock(); 1729 dev_core_stats_rx_dropped_inc(vxlan->dev); 1730 vxlan_vnifilter_count(vxlan, vni, vninode, 1731 VXLAN_VNI_STATS_RX_DROPS, 0); 1732 goto drop; 1733 } 1734 1735 dev_sw_netstats_rx_add(vxlan->dev, skb->len); 1736 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len); 1737 gro_cells_receive(&vxlan->gro_cells, skb); 1738 1739 rcu_read_unlock(); 1740 1741 return 0; 1742 1743 drop: 1744 /* Consume bad packet */ 1745 kfree_skb(skb); 1746 return 0; 1747 } 1748 1749 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */ 1750 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb) 1751 { 1752 struct vxlan_dev *vxlan; 1753 struct vxlan_sock *vs; 1754 struct vxlanhdr *hdr; 1755 __be32 vni; 1756 1757 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN)) 1758 return -EINVAL; 1759 1760 hdr = vxlan_hdr(skb); 1761 1762 if (!(hdr->vx_flags & VXLAN_HF_VNI)) 1763 return -EINVAL; 1764 1765 vs = rcu_dereference_sk_user_data(sk); 1766 if (!vs) 1767 return -ENOENT; 1768 1769 vni = vxlan_vni(hdr->vx_vni); 1770 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL); 1771 if (!vxlan) 1772 return -ENOENT; 1773 1774 return 0; 1775 } 1776 1777 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1778 { 1779 struct vxlan_dev *vxlan = netdev_priv(dev); 1780 struct arphdr *parp; 1781 u8 *arpptr, *sha; 1782 __be32 sip, tip; 1783 struct neighbour *n; 1784 1785 if (dev->flags & IFF_NOARP) 1786 goto out; 1787 1788 if (!pskb_may_pull(skb, arp_hdr_len(dev))) { 1789 dev->stats.tx_dropped++; 1790 goto out; 1791 } 1792 parp = arp_hdr(skb); 1793 1794 if ((parp->ar_hrd != htons(ARPHRD_ETHER) && 1795 parp->ar_hrd != htons(ARPHRD_IEEE802)) || 1796 parp->ar_pro != htons(ETH_P_IP) || 1797 parp->ar_op != htons(ARPOP_REQUEST) || 1798 parp->ar_hln != dev->addr_len || 1799 parp->ar_pln != 4) 1800 goto out; 1801 arpptr = (u8 *)parp + sizeof(struct arphdr); 1802 sha = arpptr; 1803 arpptr += dev->addr_len; /* sha */ 1804 memcpy(&sip, arpptr, sizeof(sip)); 1805 arpptr += sizeof(sip); 1806 arpptr += dev->addr_len; /* tha */ 1807 memcpy(&tip, arpptr, sizeof(tip)); 1808 1809 if (ipv4_is_loopback(tip) || 1810 ipv4_is_multicast(tip)) 1811 goto out; 1812 1813 n = neigh_lookup(&arp_tbl, &tip, dev); 1814 1815 if (n) { 1816 struct vxlan_fdb *f; 1817 struct sk_buff *reply; 1818 1819 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1820 neigh_release(n); 1821 goto out; 1822 } 1823 1824 f = vxlan_find_mac(vxlan, n->ha, vni); 1825 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1826 /* bridge-local neighbor */ 1827 neigh_release(n); 1828 goto out; 1829 } 1830 1831 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, 1832 n->ha, sha); 1833 1834 neigh_release(n); 1835 1836 if (reply == NULL) 1837 goto out; 1838 1839 skb_reset_mac_header(reply); 1840 __skb_pull(reply, skb_network_offset(reply)); 1841 reply->ip_summed = CHECKSUM_UNNECESSARY; 1842 reply->pkt_type = PACKET_HOST; 1843 1844 if (netif_rx(reply) == NET_RX_DROP) { 1845 dev->stats.rx_dropped++; 1846 vxlan_vnifilter_count(vxlan, vni, NULL, 1847 VXLAN_VNI_STATS_RX_DROPS, 0); 1848 } 1849 1850 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 1851 union vxlan_addr ipa = { 1852 .sin.sin_addr.s_addr = tip, 1853 .sin.sin_family = AF_INET, 1854 }; 1855 1856 vxlan_ip_miss(dev, &ipa); 1857 } 1858 out: 1859 consume_skb(skb); 1860 return NETDEV_TX_OK; 1861 } 1862 1863 #if IS_ENABLED(CONFIG_IPV6) 1864 static struct sk_buff *vxlan_na_create(struct sk_buff *request, 1865 struct neighbour *n, bool isrouter) 1866 { 1867 struct net_device *dev = request->dev; 1868 struct sk_buff *reply; 1869 struct nd_msg *ns, *na; 1870 struct ipv6hdr *pip6; 1871 u8 *daddr; 1872 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 1873 int ns_olen; 1874 int i, len; 1875 1876 if (dev == NULL || !pskb_may_pull(request, request->len)) 1877 return NULL; 1878 1879 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 1880 sizeof(*na) + na_olen + dev->needed_tailroom; 1881 reply = alloc_skb(len, GFP_ATOMIC); 1882 if (reply == NULL) 1883 return NULL; 1884 1885 reply->protocol = htons(ETH_P_IPV6); 1886 reply->dev = dev; 1887 skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); 1888 skb_push(reply, sizeof(struct ethhdr)); 1889 skb_reset_mac_header(reply); 1890 1891 ns = (struct nd_msg *)(ipv6_hdr(request) + 1); 1892 1893 daddr = eth_hdr(request)->h_source; 1894 ns_olen = request->len - skb_network_offset(request) - 1895 sizeof(struct ipv6hdr) - sizeof(*ns); 1896 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { 1897 if (!ns->opt[i + 1]) { 1898 kfree_skb(reply); 1899 return NULL; 1900 } 1901 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 1902 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 1903 break; 1904 } 1905 } 1906 1907 /* Ethernet header */ 1908 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 1909 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 1910 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 1911 reply->protocol = htons(ETH_P_IPV6); 1912 1913 skb_pull(reply, sizeof(struct ethhdr)); 1914 skb_reset_network_header(reply); 1915 skb_put(reply, sizeof(struct ipv6hdr)); 1916 1917 /* IPv6 header */ 1918 1919 pip6 = ipv6_hdr(reply); 1920 memset(pip6, 0, sizeof(struct ipv6hdr)); 1921 pip6->version = 6; 1922 pip6->priority = ipv6_hdr(request)->priority; 1923 pip6->nexthdr = IPPROTO_ICMPV6; 1924 pip6->hop_limit = 255; 1925 pip6->daddr = ipv6_hdr(request)->saddr; 1926 pip6->saddr = *(struct in6_addr *)n->primary_key; 1927 1928 skb_pull(reply, sizeof(struct ipv6hdr)); 1929 skb_reset_transport_header(reply); 1930 1931 /* Neighbor Advertisement */ 1932 na = skb_put_zero(reply, sizeof(*na) + na_olen); 1933 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 1934 na->icmph.icmp6_router = isrouter; 1935 na->icmph.icmp6_override = 1; 1936 na->icmph.icmp6_solicited = 1; 1937 na->target = ns->target; 1938 ether_addr_copy(&na->opt[2], n->ha); 1939 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 1940 na->opt[1] = na_olen >> 3; 1941 1942 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 1943 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, 1944 csum_partial(na, sizeof(*na)+na_olen, 0)); 1945 1946 pip6->payload_len = htons(sizeof(*na)+na_olen); 1947 1948 skb_push(reply, sizeof(struct ipv6hdr)); 1949 1950 reply->ip_summed = CHECKSUM_UNNECESSARY; 1951 1952 return reply; 1953 } 1954 1955 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni) 1956 { 1957 struct vxlan_dev *vxlan = netdev_priv(dev); 1958 const struct in6_addr *daddr; 1959 const struct ipv6hdr *iphdr; 1960 struct inet6_dev *in6_dev; 1961 struct neighbour *n; 1962 struct nd_msg *msg; 1963 1964 rcu_read_lock(); 1965 in6_dev = __in6_dev_get(dev); 1966 if (!in6_dev) 1967 goto out; 1968 1969 iphdr = ipv6_hdr(skb); 1970 daddr = &iphdr->daddr; 1971 msg = (struct nd_msg *)(iphdr + 1); 1972 1973 if (ipv6_addr_loopback(daddr) || 1974 ipv6_addr_is_multicast(&msg->target)) 1975 goto out; 1976 1977 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); 1978 1979 if (n) { 1980 struct vxlan_fdb *f; 1981 struct sk_buff *reply; 1982 1983 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) { 1984 neigh_release(n); 1985 goto out; 1986 } 1987 1988 f = vxlan_find_mac(vxlan, n->ha, vni); 1989 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { 1990 /* bridge-local neighbor */ 1991 neigh_release(n); 1992 goto out; 1993 } 1994 1995 reply = vxlan_na_create(skb, n, 1996 !!(f ? f->flags & NTF_ROUTER : 0)); 1997 1998 neigh_release(n); 1999 2000 if (reply == NULL) 2001 goto out; 2002 2003 if (netif_rx(reply) == NET_RX_DROP) { 2004 dev->stats.rx_dropped++; 2005 vxlan_vnifilter_count(vxlan, vni, NULL, 2006 VXLAN_VNI_STATS_RX_DROPS, 0); 2007 } 2008 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) { 2009 union vxlan_addr ipa = { 2010 .sin6.sin6_addr = msg->target, 2011 .sin6.sin6_family = AF_INET6, 2012 }; 2013 2014 vxlan_ip_miss(dev, &ipa); 2015 } 2016 2017 out: 2018 rcu_read_unlock(); 2019 consume_skb(skb); 2020 return NETDEV_TX_OK; 2021 } 2022 #endif 2023 2024 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) 2025 { 2026 struct vxlan_dev *vxlan = netdev_priv(dev); 2027 struct neighbour *n; 2028 2029 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) 2030 return false; 2031 2032 n = NULL; 2033 switch (ntohs(eth_hdr(skb)->h_proto)) { 2034 case ETH_P_IP: 2035 { 2036 struct iphdr *pip; 2037 2038 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 2039 return false; 2040 pip = ip_hdr(skb); 2041 n = neigh_lookup(&arp_tbl, &pip->daddr, dev); 2042 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2043 union vxlan_addr ipa = { 2044 .sin.sin_addr.s_addr = pip->daddr, 2045 .sin.sin_family = AF_INET, 2046 }; 2047 2048 vxlan_ip_miss(dev, &ipa); 2049 return false; 2050 } 2051 2052 break; 2053 } 2054 #if IS_ENABLED(CONFIG_IPV6) 2055 case ETH_P_IPV6: 2056 { 2057 struct ipv6hdr *pip6; 2058 2059 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 2060 return false; 2061 pip6 = ipv6_hdr(skb); 2062 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); 2063 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) { 2064 union vxlan_addr ipa = { 2065 .sin6.sin6_addr = pip6->daddr, 2066 .sin6.sin6_family = AF_INET6, 2067 }; 2068 2069 vxlan_ip_miss(dev, &ipa); 2070 return false; 2071 } 2072 2073 break; 2074 } 2075 #endif 2076 default: 2077 return false; 2078 } 2079 2080 if (n) { 2081 bool diff; 2082 2083 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); 2084 if (diff) { 2085 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, 2086 dev->addr_len); 2087 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); 2088 } 2089 neigh_release(n); 2090 return diff; 2091 } 2092 2093 return false; 2094 } 2095 2096 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags, 2097 struct vxlan_metadata *md) 2098 { 2099 struct vxlanhdr_gbp *gbp; 2100 2101 if (!md->gbp) 2102 return; 2103 2104 gbp = (struct vxlanhdr_gbp *)vxh; 2105 vxh->vx_flags |= VXLAN_HF_GBP; 2106 2107 if (md->gbp & VXLAN_GBP_DONT_LEARN) 2108 gbp->dont_learn = 1; 2109 2110 if (md->gbp & VXLAN_GBP_POLICY_APPLIED) 2111 gbp->policy_applied = 1; 2112 2113 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK); 2114 } 2115 2116 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags, 2117 __be16 protocol) 2118 { 2119 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; 2120 2121 gpe->np_applied = 1; 2122 gpe->next_protocol = tun_p_from_eth_p(protocol); 2123 if (!gpe->next_protocol) 2124 return -EPFNOSUPPORT; 2125 return 0; 2126 } 2127 2128 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, 2129 int iphdr_len, __be32 vni, 2130 struct vxlan_metadata *md, u32 vxflags, 2131 bool udp_sum) 2132 { 2133 struct vxlanhdr *vxh; 2134 int min_headroom; 2135 int err; 2136 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; 2137 __be16 inner_protocol = htons(ETH_P_TEB); 2138 2139 if ((vxflags & VXLAN_F_REMCSUM_TX) && 2140 skb->ip_summed == CHECKSUM_PARTIAL) { 2141 int csum_start = skb_checksum_start_offset(skb); 2142 2143 if (csum_start <= VXLAN_MAX_REMCSUM_START && 2144 !(csum_start & VXLAN_RCO_SHIFT_MASK) && 2145 (skb->csum_offset == offsetof(struct udphdr, check) || 2146 skb->csum_offset == offsetof(struct tcphdr, check))) 2147 type |= SKB_GSO_TUNNEL_REMCSUM; 2148 } 2149 2150 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len 2151 + VXLAN_HLEN + iphdr_len; 2152 2153 /* Need space for new headers (invalidates iph ptr) */ 2154 err = skb_cow_head(skb, min_headroom); 2155 if (unlikely(err)) 2156 return err; 2157 2158 err = iptunnel_handle_offloads(skb, type); 2159 if (err) 2160 return err; 2161 2162 vxh = __skb_push(skb, sizeof(*vxh)); 2163 vxh->vx_flags = VXLAN_HF_VNI; 2164 vxh->vx_vni = vxlan_vni_field(vni); 2165 2166 if (type & SKB_GSO_TUNNEL_REMCSUM) { 2167 unsigned int start; 2168 2169 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); 2170 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); 2171 vxh->vx_flags |= VXLAN_HF_RCO; 2172 2173 if (!skb_is_gso(skb)) { 2174 skb->ip_summed = CHECKSUM_NONE; 2175 skb->encapsulation = 0; 2176 } 2177 } 2178 2179 if (vxflags & VXLAN_F_GBP) 2180 vxlan_build_gbp_hdr(vxh, vxflags, md); 2181 if (vxflags & VXLAN_F_GPE) { 2182 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol); 2183 if (err < 0) 2184 return err; 2185 inner_protocol = skb->protocol; 2186 } 2187 2188 skb_set_inner_protocol(skb, inner_protocol); 2189 return 0; 2190 } 2191 2192 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev, 2193 struct vxlan_sock *sock4, 2194 struct sk_buff *skb, int oif, u8 tos, 2195 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport, 2196 __u8 flow_flags, struct dst_cache *dst_cache, 2197 const struct ip_tunnel_info *info) 2198 { 2199 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 2200 struct rtable *rt = NULL; 2201 struct flowi4 fl4; 2202 2203 if (!sock4) 2204 return ERR_PTR(-EIO); 2205 2206 if (tos && !info) 2207 use_cache = false; 2208 if (use_cache) { 2209 rt = dst_cache_get_ip4(dst_cache, saddr); 2210 if (rt) 2211 return rt; 2212 } 2213 2214 memset(&fl4, 0, sizeof(fl4)); 2215 fl4.flowi4_oif = oif; 2216 fl4.flowi4_tos = RT_TOS(tos); 2217 fl4.flowi4_mark = skb->mark; 2218 fl4.flowi4_proto = IPPROTO_UDP; 2219 fl4.daddr = daddr; 2220 fl4.saddr = *saddr; 2221 fl4.fl4_dport = dport; 2222 fl4.fl4_sport = sport; 2223 fl4.flowi4_flags = flow_flags; 2224 2225 rt = ip_route_output_key(vxlan->net, &fl4); 2226 if (!IS_ERR(rt)) { 2227 if (rt->dst.dev == dev) { 2228 netdev_dbg(dev, "circular route to %pI4\n", &daddr); 2229 ip_rt_put(rt); 2230 return ERR_PTR(-ELOOP); 2231 } 2232 2233 *saddr = fl4.saddr; 2234 if (use_cache) 2235 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr); 2236 } else { 2237 netdev_dbg(dev, "no route to %pI4\n", &daddr); 2238 return ERR_PTR(-ENETUNREACH); 2239 } 2240 return rt; 2241 } 2242 2243 #if IS_ENABLED(CONFIG_IPV6) 2244 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan, 2245 struct net_device *dev, 2246 struct vxlan_sock *sock6, 2247 struct sk_buff *skb, int oif, u8 tos, 2248 __be32 label, 2249 const struct in6_addr *daddr, 2250 struct in6_addr *saddr, 2251 __be16 dport, __be16 sport, 2252 struct dst_cache *dst_cache, 2253 const struct ip_tunnel_info *info) 2254 { 2255 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 2256 struct dst_entry *ndst; 2257 struct flowi6 fl6; 2258 2259 if (!sock6) 2260 return ERR_PTR(-EIO); 2261 2262 if (tos && !info) 2263 use_cache = false; 2264 if (use_cache) { 2265 ndst = dst_cache_get_ip6(dst_cache, saddr); 2266 if (ndst) 2267 return ndst; 2268 } 2269 2270 memset(&fl6, 0, sizeof(fl6)); 2271 fl6.flowi6_oif = oif; 2272 fl6.daddr = *daddr; 2273 fl6.saddr = *saddr; 2274 fl6.flowlabel = ip6_make_flowinfo(tos, label); 2275 fl6.flowi6_mark = skb->mark; 2276 fl6.flowi6_proto = IPPROTO_UDP; 2277 fl6.fl6_dport = dport; 2278 fl6.fl6_sport = sport; 2279 2280 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk, 2281 &fl6, NULL); 2282 if (IS_ERR(ndst)) { 2283 netdev_dbg(dev, "no route to %pI6\n", daddr); 2284 return ERR_PTR(-ENETUNREACH); 2285 } 2286 2287 if (unlikely(ndst->dev == dev)) { 2288 netdev_dbg(dev, "circular route to %pI6\n", daddr); 2289 dst_release(ndst); 2290 return ERR_PTR(-ELOOP); 2291 } 2292 2293 *saddr = fl6.saddr; 2294 if (use_cache) 2295 dst_cache_set_ip6(dst_cache, ndst, saddr); 2296 return ndst; 2297 } 2298 #endif 2299 2300 /* Bypass encapsulation if the destination is local */ 2301 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, 2302 struct vxlan_dev *dst_vxlan, __be32 vni, 2303 bool snoop) 2304 { 2305 struct pcpu_sw_netstats *tx_stats, *rx_stats; 2306 union vxlan_addr loopback; 2307 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; 2308 struct net_device *dev; 2309 int len = skb->len; 2310 2311 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats); 2312 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats); 2313 skb->pkt_type = PACKET_HOST; 2314 skb->encapsulation = 0; 2315 skb->dev = dst_vxlan->dev; 2316 __skb_pull(skb, skb_network_offset(skb)); 2317 2318 if (remote_ip->sa.sa_family == AF_INET) { 2319 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 2320 loopback.sa.sa_family = AF_INET; 2321 #if IS_ENABLED(CONFIG_IPV6) 2322 } else { 2323 loopback.sin6.sin6_addr = in6addr_loopback; 2324 loopback.sa.sa_family = AF_INET6; 2325 #endif 2326 } 2327 2328 rcu_read_lock(); 2329 dev = skb->dev; 2330 if (unlikely(!(dev->flags & IFF_UP))) { 2331 kfree_skb(skb); 2332 goto drop; 2333 } 2334 2335 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop) 2336 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni); 2337 2338 u64_stats_update_begin(&tx_stats->syncp); 2339 u64_stats_inc(&tx_stats->tx_packets); 2340 u64_stats_add(&tx_stats->tx_bytes, len); 2341 u64_stats_update_end(&tx_stats->syncp); 2342 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len); 2343 2344 if (__netif_rx(skb) == NET_RX_SUCCESS) { 2345 u64_stats_update_begin(&rx_stats->syncp); 2346 u64_stats_inc(&rx_stats->rx_packets); 2347 u64_stats_add(&rx_stats->rx_bytes, len); 2348 u64_stats_update_end(&rx_stats->syncp); 2349 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX, 2350 len); 2351 } else { 2352 drop: 2353 dev->stats.rx_dropped++; 2354 vxlan_vnifilter_count(dst_vxlan, vni, NULL, 2355 VXLAN_VNI_STATS_RX_DROPS, 0); 2356 } 2357 rcu_read_unlock(); 2358 } 2359 2360 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, 2361 struct vxlan_dev *vxlan, 2362 union vxlan_addr *daddr, 2363 __be16 dst_port, int dst_ifindex, __be32 vni, 2364 struct dst_entry *dst, 2365 u32 rt_flags) 2366 { 2367 #if IS_ENABLED(CONFIG_IPV6) 2368 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of 2369 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple 2370 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry. 2371 */ 2372 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL); 2373 #endif 2374 /* Bypass encapsulation if the destination is local */ 2375 if (rt_flags & RTCF_LOCAL && 2376 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { 2377 struct vxlan_dev *dst_vxlan; 2378 2379 dst_release(dst); 2380 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, 2381 daddr->sa.sa_family, dst_port, 2382 vxlan->cfg.flags); 2383 if (!dst_vxlan) { 2384 dev->stats.tx_errors++; 2385 vxlan_vnifilter_count(vxlan, vni, NULL, 2386 VXLAN_VNI_STATS_TX_ERRORS, 0); 2387 kfree_skb(skb); 2388 2389 return -ENOENT; 2390 } 2391 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); 2392 return 1; 2393 } 2394 2395 return 0; 2396 } 2397 2398 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2399 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2400 { 2401 struct dst_cache *dst_cache; 2402 struct ip_tunnel_info *info; 2403 struct vxlan_dev *vxlan = netdev_priv(dev); 2404 const struct iphdr *old_iph = ip_hdr(skb); 2405 union vxlan_addr *dst; 2406 union vxlan_addr remote_ip, local_ip; 2407 struct vxlan_metadata _md; 2408 struct vxlan_metadata *md = &_md; 2409 unsigned int pkt_len = skb->len; 2410 __be16 src_port = 0, dst_port; 2411 struct dst_entry *ndst = NULL; 2412 __u8 tos, ttl, flow_flags = 0; 2413 int ifindex; 2414 int err; 2415 u32 flags = vxlan->cfg.flags; 2416 bool udp_sum = false; 2417 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2418 __be32 vni = 0; 2419 #if IS_ENABLED(CONFIG_IPV6) 2420 __be32 label; 2421 #endif 2422 2423 info = skb_tunnel_info(skb); 2424 2425 if (rdst) { 2426 dst = &rdst->remote_ip; 2427 if (vxlan_addr_any(dst)) { 2428 if (did_rsc) { 2429 /* short-circuited back to local bridge */ 2430 vxlan_encap_bypass(skb, vxlan, vxlan, 2431 default_vni, true); 2432 return; 2433 } 2434 goto drop; 2435 } 2436 2437 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2438 vni = (rdst->remote_vni) ? : default_vni; 2439 ifindex = rdst->remote_ifindex; 2440 local_ip = vxlan->cfg.saddr; 2441 dst_cache = &rdst->dst_cache; 2442 md->gbp = skb->mark; 2443 if (flags & VXLAN_F_TTL_INHERIT) { 2444 ttl = ip_tunnel_get_ttl(old_iph, skb); 2445 } else { 2446 ttl = vxlan->cfg.ttl; 2447 if (!ttl && vxlan_addr_multicast(dst)) 2448 ttl = 1; 2449 } 2450 2451 tos = vxlan->cfg.tos; 2452 if (tos == 1) 2453 tos = ip_tunnel_get_dsfield(old_iph, skb); 2454 2455 if (dst->sa.sa_family == AF_INET) 2456 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2457 else 2458 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2459 #if IS_ENABLED(CONFIG_IPV6) 2460 label = vxlan->cfg.label; 2461 #endif 2462 } else { 2463 if (!info) { 2464 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2465 dev->name); 2466 goto drop; 2467 } 2468 remote_ip.sa.sa_family = ip_tunnel_info_af(info); 2469 if (remote_ip.sa.sa_family == AF_INET) { 2470 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst; 2471 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src; 2472 } else { 2473 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst; 2474 local_ip.sin6.sin6_addr = info->key.u.ipv6.src; 2475 } 2476 dst = &remote_ip; 2477 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2478 flow_flags = info->key.flow_flags; 2479 vni = tunnel_id_to_key32(info->key.tun_id); 2480 ifindex = 0; 2481 dst_cache = &info->dst_cache; 2482 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) { 2483 if (info->options_len < sizeof(*md)) 2484 goto drop; 2485 md = ip_tunnel_info_opts(info); 2486 } 2487 ttl = info->key.ttl; 2488 tos = info->key.tos; 2489 #if IS_ENABLED(CONFIG_IPV6) 2490 label = info->key.label; 2491 #endif 2492 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 2493 } 2494 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2495 vxlan->cfg.port_max, true); 2496 2497 rcu_read_lock(); 2498 if (dst->sa.sa_family == AF_INET) { 2499 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2500 struct rtable *rt; 2501 __be16 df = 0; 2502 2503 if (!ifindex) 2504 ifindex = sock4->sock->sk->sk_bound_dev_if; 2505 2506 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos, 2507 dst->sin.sin_addr.s_addr, 2508 &local_ip.sin.sin_addr.s_addr, 2509 dst_port, src_port, flow_flags, 2510 dst_cache, info); 2511 if (IS_ERR(rt)) { 2512 err = PTR_ERR(rt); 2513 goto tx_error; 2514 } 2515 2516 if (!info) { 2517 /* Bypass encapsulation if the destination is local */ 2518 err = encap_bypass_if_local(skb, dev, vxlan, dst, 2519 dst_port, ifindex, vni, 2520 &rt->dst, rt->rt_flags); 2521 if (err) 2522 goto out_unlock; 2523 2524 if (vxlan->cfg.df == VXLAN_DF_SET) { 2525 df = htons(IP_DF); 2526 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2527 struct ethhdr *eth = eth_hdr(skb); 2528 2529 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2530 (ntohs(eth->h_proto) == ETH_P_IP && 2531 old_iph->frag_off & htons(IP_DF))) 2532 df = htons(IP_DF); 2533 } 2534 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) { 2535 df = htons(IP_DF); 2536 } 2537 2538 ndst = &rt->dst; 2539 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM, 2540 netif_is_any_bridge_port(dev)); 2541 if (err < 0) { 2542 goto tx_error; 2543 } else if (err) { 2544 if (info) { 2545 struct ip_tunnel_info *unclone; 2546 struct in_addr src, dst; 2547 2548 unclone = skb_tunnel_info_unclone(skb); 2549 if (unlikely(!unclone)) 2550 goto tx_error; 2551 2552 src = remote_ip.sin.sin_addr; 2553 dst = local_ip.sin.sin_addr; 2554 unclone->key.u.ipv4.src = src.s_addr; 2555 unclone->key.u.ipv4.dst = dst.s_addr; 2556 } 2557 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2558 dst_release(ndst); 2559 goto out_unlock; 2560 } 2561 2562 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2563 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2564 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2565 vni, md, flags, udp_sum); 2566 if (err < 0) 2567 goto tx_error; 2568 2569 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr, 2570 dst->sin.sin_addr.s_addr, tos, ttl, df, 2571 src_port, dst_port, xnet, !udp_sum); 2572 #if IS_ENABLED(CONFIG_IPV6) 2573 } else { 2574 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2575 2576 if (!ifindex) 2577 ifindex = sock6->sock->sk->sk_bound_dev_if; 2578 2579 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos, 2580 label, &dst->sin6.sin6_addr, 2581 &local_ip.sin6.sin6_addr, 2582 dst_port, src_port, 2583 dst_cache, info); 2584 if (IS_ERR(ndst)) { 2585 err = PTR_ERR(ndst); 2586 ndst = NULL; 2587 goto tx_error; 2588 } 2589 2590 if (!info) { 2591 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; 2592 2593 err = encap_bypass_if_local(skb, dev, vxlan, dst, 2594 dst_port, ifindex, vni, 2595 ndst, rt6i_flags); 2596 if (err) 2597 goto out_unlock; 2598 } 2599 2600 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM, 2601 netif_is_any_bridge_port(dev)); 2602 if (err < 0) { 2603 goto tx_error; 2604 } else if (err) { 2605 if (info) { 2606 struct ip_tunnel_info *unclone; 2607 struct in6_addr src, dst; 2608 2609 unclone = skb_tunnel_info_unclone(skb); 2610 if (unlikely(!unclone)) 2611 goto tx_error; 2612 2613 src = remote_ip.sin6.sin6_addr; 2614 dst = local_ip.sin6.sin6_addr; 2615 unclone->key.u.ipv6.src = src; 2616 unclone->key.u.ipv6.dst = dst; 2617 } 2618 2619 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2620 dst_release(ndst); 2621 goto out_unlock; 2622 } 2623 2624 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2625 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2626 skb_scrub_packet(skb, xnet); 2627 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2628 vni, md, flags, udp_sum); 2629 if (err < 0) 2630 goto tx_error; 2631 2632 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2633 &local_ip.sin6.sin6_addr, 2634 &dst->sin6.sin6_addr, tos, ttl, 2635 label, src_port, dst_port, !udp_sum); 2636 #endif 2637 } 2638 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2639 out_unlock: 2640 rcu_read_unlock(); 2641 return; 2642 2643 drop: 2644 dev->stats.tx_dropped++; 2645 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2646 dev_kfree_skb(skb); 2647 return; 2648 2649 tx_error: 2650 rcu_read_unlock(); 2651 if (err == -ELOOP) 2652 dev->stats.collisions++; 2653 else if (err == -ENETUNREACH) 2654 dev->stats.tx_carrier_errors++; 2655 dst_release(ndst); 2656 dev->stats.tx_errors++; 2657 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2658 kfree_skb(skb); 2659 } 2660 2661 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev, 2662 struct vxlan_fdb *f, __be32 vni, bool did_rsc) 2663 { 2664 struct vxlan_rdst nh_rdst; 2665 struct nexthop *nh; 2666 bool do_xmit; 2667 u32 hash; 2668 2669 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst)); 2670 hash = skb_get_hash(skb); 2671 2672 rcu_read_lock(); 2673 nh = rcu_dereference(f->nh); 2674 if (!nh) { 2675 rcu_read_unlock(); 2676 goto drop; 2677 } 2678 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst); 2679 rcu_read_unlock(); 2680 2681 if (likely(do_xmit)) 2682 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc); 2683 else 2684 goto drop; 2685 2686 return; 2687 2688 drop: 2689 dev->stats.tx_dropped++; 2690 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL, 2691 VXLAN_VNI_STATS_TX_DROPS, 0); 2692 dev_kfree_skb(skb); 2693 } 2694 2695 /* Transmit local packets over Vxlan 2696 * 2697 * Outer IP header inherits ECN and DF from inner header. 2698 * Outer UDP destination is the VXLAN assigned port. 2699 * source port is based on hash of flow 2700 */ 2701 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) 2702 { 2703 struct vxlan_dev *vxlan = netdev_priv(dev); 2704 struct vxlan_rdst *rdst, *fdst = NULL; 2705 const struct ip_tunnel_info *info; 2706 bool did_rsc = false; 2707 struct vxlan_fdb *f; 2708 struct ethhdr *eth; 2709 __be32 vni = 0; 2710 2711 info = skb_tunnel_info(skb); 2712 2713 skb_reset_mac_header(skb); 2714 2715 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) { 2716 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE && 2717 info->mode & IP_TUNNEL_INFO_TX) { 2718 vni = tunnel_id_to_key32(info->key.tun_id); 2719 } else { 2720 if (info && info->mode & IP_TUNNEL_INFO_TX) 2721 vxlan_xmit_one(skb, dev, vni, NULL, false); 2722 else 2723 kfree_skb(skb); 2724 return NETDEV_TX_OK; 2725 } 2726 } 2727 2728 if (vxlan->cfg.flags & VXLAN_F_PROXY) { 2729 eth = eth_hdr(skb); 2730 if (ntohs(eth->h_proto) == ETH_P_ARP) 2731 return arp_reduce(dev, skb, vni); 2732 #if IS_ENABLED(CONFIG_IPV6) 2733 else if (ntohs(eth->h_proto) == ETH_P_IPV6 && 2734 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 2735 sizeof(struct nd_msg)) && 2736 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 2737 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1); 2738 2739 if (m->icmph.icmp6_code == 0 && 2740 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) 2741 return neigh_reduce(dev, skb, vni); 2742 } 2743 #endif 2744 } 2745 2746 if (vxlan->cfg.flags & VXLAN_F_MDB) { 2747 struct vxlan_mdb_entry *mdb_entry; 2748 2749 rcu_read_lock(); 2750 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni); 2751 if (mdb_entry) { 2752 netdev_tx_t ret; 2753 2754 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb); 2755 rcu_read_unlock(); 2756 return ret; 2757 } 2758 rcu_read_unlock(); 2759 } 2760 2761 eth = eth_hdr(skb); 2762 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2763 did_rsc = false; 2764 2765 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) && 2766 (ntohs(eth->h_proto) == ETH_P_IP || 2767 ntohs(eth->h_proto) == ETH_P_IPV6)) { 2768 did_rsc = route_shortcircuit(dev, skb); 2769 if (did_rsc) 2770 f = vxlan_find_mac(vxlan, eth->h_dest, vni); 2771 } 2772 2773 if (f == NULL) { 2774 f = vxlan_find_mac(vxlan, all_zeros_mac, vni); 2775 if (f == NULL) { 2776 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) && 2777 !is_multicast_ether_addr(eth->h_dest)) 2778 vxlan_fdb_miss(vxlan, eth->h_dest); 2779 2780 dev->stats.tx_dropped++; 2781 vxlan_vnifilter_count(vxlan, vni, NULL, 2782 VXLAN_VNI_STATS_TX_DROPS, 0); 2783 kfree_skb(skb); 2784 return NETDEV_TX_OK; 2785 } 2786 } 2787 2788 if (rcu_access_pointer(f->nh)) { 2789 vxlan_xmit_nh(skb, dev, f, 2790 (vni ? : vxlan->default_dst.remote_vni), did_rsc); 2791 } else { 2792 list_for_each_entry_rcu(rdst, &f->remotes, list) { 2793 struct sk_buff *skb1; 2794 2795 if (!fdst) { 2796 fdst = rdst; 2797 continue; 2798 } 2799 skb1 = skb_clone(skb, GFP_ATOMIC); 2800 if (skb1) 2801 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc); 2802 } 2803 if (fdst) 2804 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc); 2805 else 2806 kfree_skb(skb); 2807 } 2808 2809 return NETDEV_TX_OK; 2810 } 2811 2812 /* Walk the forwarding table and purge stale entries */ 2813 static void vxlan_cleanup(struct timer_list *t) 2814 { 2815 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer); 2816 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; 2817 unsigned int h; 2818 2819 if (!netif_running(vxlan->dev)) 2820 return; 2821 2822 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2823 struct hlist_node *p, *n; 2824 2825 spin_lock(&vxlan->hash_lock[h]); 2826 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2827 struct vxlan_fdb *f 2828 = container_of(p, struct vxlan_fdb, hlist); 2829 unsigned long timeout; 2830 2831 if (f->state & (NUD_PERMANENT | NUD_NOARP)) 2832 continue; 2833 2834 if (f->flags & NTF_EXT_LEARNED) 2835 continue; 2836 2837 timeout = f->used + vxlan->cfg.age_interval * HZ; 2838 if (time_before_eq(timeout, jiffies)) { 2839 netdev_dbg(vxlan->dev, 2840 "garbage collect %pM\n", 2841 f->eth_addr); 2842 f->state = NUD_STALE; 2843 vxlan_fdb_destroy(vxlan, f, true, true); 2844 } else if (time_before(timeout, next_timer)) 2845 next_timer = timeout; 2846 } 2847 spin_unlock(&vxlan->hash_lock[h]); 2848 } 2849 2850 mod_timer(&vxlan->age_timer, next_timer); 2851 } 2852 2853 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan) 2854 { 2855 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2856 2857 spin_lock(&vn->sock_lock); 2858 hlist_del_init_rcu(&vxlan->hlist4.hlist); 2859 #if IS_ENABLED(CONFIG_IPV6) 2860 hlist_del_init_rcu(&vxlan->hlist6.hlist); 2861 #endif 2862 spin_unlock(&vn->sock_lock); 2863 } 2864 2865 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan, 2866 struct vxlan_dev_node *node) 2867 { 2868 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 2869 __be32 vni = vxlan->default_dst.remote_vni; 2870 2871 node->vxlan = vxlan; 2872 spin_lock(&vn->sock_lock); 2873 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni)); 2874 spin_unlock(&vn->sock_lock); 2875 } 2876 2877 /* Setup stats when device is created */ 2878 static int vxlan_init(struct net_device *dev) 2879 { 2880 struct vxlan_dev *vxlan = netdev_priv(dev); 2881 int err; 2882 2883 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2884 vxlan_vnigroup_init(vxlan); 2885 2886 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 2887 if (!dev->tstats) { 2888 err = -ENOMEM; 2889 goto err_vnigroup_uninit; 2890 } 2891 2892 err = gro_cells_init(&vxlan->gro_cells, dev); 2893 if (err) 2894 goto err_free_percpu; 2895 2896 err = vxlan_mdb_init(vxlan); 2897 if (err) 2898 goto err_gro_cells_destroy; 2899 2900 return 0; 2901 2902 err_gro_cells_destroy: 2903 gro_cells_destroy(&vxlan->gro_cells); 2904 err_free_percpu: 2905 free_percpu(dev->tstats); 2906 err_vnigroup_uninit: 2907 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2908 vxlan_vnigroup_uninit(vxlan); 2909 return err; 2910 } 2911 2912 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni) 2913 { 2914 struct vxlan_fdb *f; 2915 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni); 2916 2917 spin_lock_bh(&vxlan->hash_lock[hash_index]); 2918 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni); 2919 if (f) 2920 vxlan_fdb_destroy(vxlan, f, true, true); 2921 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 2922 } 2923 2924 static void vxlan_uninit(struct net_device *dev) 2925 { 2926 struct vxlan_dev *vxlan = netdev_priv(dev); 2927 2928 vxlan_mdb_fini(vxlan); 2929 2930 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) 2931 vxlan_vnigroup_uninit(vxlan); 2932 2933 gro_cells_destroy(&vxlan->gro_cells); 2934 2935 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni); 2936 2937 free_percpu(dev->tstats); 2938 } 2939 2940 /* Start ageing timer and join group when device is brought up */ 2941 static int vxlan_open(struct net_device *dev) 2942 { 2943 struct vxlan_dev *vxlan = netdev_priv(dev); 2944 int ret; 2945 2946 ret = vxlan_sock_add(vxlan); 2947 if (ret < 0) 2948 return ret; 2949 2950 ret = vxlan_multicast_join(vxlan); 2951 if (ret) { 2952 vxlan_sock_release(vxlan); 2953 return ret; 2954 } 2955 2956 if (vxlan->cfg.age_interval) 2957 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); 2958 2959 return ret; 2960 } 2961 2962 /* Purge the forwarding table */ 2963 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all) 2964 { 2965 unsigned int h; 2966 2967 for (h = 0; h < FDB_HASH_SIZE; ++h) { 2968 struct hlist_node *p, *n; 2969 2970 spin_lock_bh(&vxlan->hash_lock[h]); 2971 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { 2972 struct vxlan_fdb *f 2973 = container_of(p, struct vxlan_fdb, hlist); 2974 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP))) 2975 continue; 2976 /* the all_zeros_mac entry is deleted at vxlan_uninit */ 2977 if (is_zero_ether_addr(f->eth_addr) && 2978 f->vni == vxlan->cfg.vni) 2979 continue; 2980 vxlan_fdb_destroy(vxlan, f, true, true); 2981 } 2982 spin_unlock_bh(&vxlan->hash_lock[h]); 2983 } 2984 } 2985 2986 /* Cleanup timer and forwarding table on shutdown */ 2987 static int vxlan_stop(struct net_device *dev) 2988 { 2989 struct vxlan_dev *vxlan = netdev_priv(dev); 2990 2991 vxlan_multicast_leave(vxlan); 2992 2993 del_timer_sync(&vxlan->age_timer); 2994 2995 vxlan_flush(vxlan, false); 2996 vxlan_sock_release(vxlan); 2997 2998 return 0; 2999 } 3000 3001 /* Stub, nothing needs to be done. */ 3002 static void vxlan_set_multicast_list(struct net_device *dev) 3003 { 3004 } 3005 3006 static int vxlan_change_mtu(struct net_device *dev, int new_mtu) 3007 { 3008 struct vxlan_dev *vxlan = netdev_priv(dev); 3009 struct vxlan_rdst *dst = &vxlan->default_dst; 3010 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3011 dst->remote_ifindex); 3012 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6); 3013 3014 /* This check is different than dev->max_mtu, because it looks at 3015 * the lowerdev->mtu, rather than the static dev->max_mtu 3016 */ 3017 if (lowerdev) { 3018 int max_mtu = lowerdev->mtu - 3019 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM); 3020 if (new_mtu > max_mtu) 3021 return -EINVAL; 3022 } 3023 3024 dev->mtu = new_mtu; 3025 return 0; 3026 } 3027 3028 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) 3029 { 3030 struct vxlan_dev *vxlan = netdev_priv(dev); 3031 struct ip_tunnel_info *info = skb_tunnel_info(skb); 3032 __be16 sport, dport; 3033 3034 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 3035 vxlan->cfg.port_max, true); 3036 dport = info->key.tp_dst ? : vxlan->cfg.dst_port; 3037 3038 if (ip_tunnel_info_af(info) == AF_INET) { 3039 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 3040 struct rtable *rt; 3041 3042 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos, 3043 info->key.u.ipv4.dst, 3044 &info->key.u.ipv4.src, dport, sport, 3045 info->key.flow_flags, &info->dst_cache, 3046 info); 3047 if (IS_ERR(rt)) 3048 return PTR_ERR(rt); 3049 ip_rt_put(rt); 3050 } else { 3051 #if IS_ENABLED(CONFIG_IPV6) 3052 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 3053 struct dst_entry *ndst; 3054 3055 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos, 3056 info->key.label, &info->key.u.ipv6.dst, 3057 &info->key.u.ipv6.src, dport, sport, 3058 &info->dst_cache, info); 3059 if (IS_ERR(ndst)) 3060 return PTR_ERR(ndst); 3061 dst_release(ndst); 3062 #else /* !CONFIG_IPV6 */ 3063 return -EPFNOSUPPORT; 3064 #endif 3065 } 3066 info->key.tp_src = sport; 3067 info->key.tp_dst = dport; 3068 return 0; 3069 } 3070 3071 static const struct net_device_ops vxlan_netdev_ether_ops = { 3072 .ndo_init = vxlan_init, 3073 .ndo_uninit = vxlan_uninit, 3074 .ndo_open = vxlan_open, 3075 .ndo_stop = vxlan_stop, 3076 .ndo_start_xmit = vxlan_xmit, 3077 .ndo_get_stats64 = dev_get_tstats64, 3078 .ndo_set_rx_mode = vxlan_set_multicast_list, 3079 .ndo_change_mtu = vxlan_change_mtu, 3080 .ndo_validate_addr = eth_validate_addr, 3081 .ndo_set_mac_address = eth_mac_addr, 3082 .ndo_fdb_add = vxlan_fdb_add, 3083 .ndo_fdb_del = vxlan_fdb_delete, 3084 .ndo_fdb_dump = vxlan_fdb_dump, 3085 .ndo_fdb_get = vxlan_fdb_get, 3086 .ndo_mdb_add = vxlan_mdb_add, 3087 .ndo_mdb_del = vxlan_mdb_del, 3088 .ndo_mdb_dump = vxlan_mdb_dump, 3089 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3090 }; 3091 3092 static const struct net_device_ops vxlan_netdev_raw_ops = { 3093 .ndo_init = vxlan_init, 3094 .ndo_uninit = vxlan_uninit, 3095 .ndo_open = vxlan_open, 3096 .ndo_stop = vxlan_stop, 3097 .ndo_start_xmit = vxlan_xmit, 3098 .ndo_get_stats64 = dev_get_tstats64, 3099 .ndo_change_mtu = vxlan_change_mtu, 3100 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, 3101 }; 3102 3103 /* Info for udev, that this is a virtual tunnel endpoint */ 3104 static struct device_type vxlan_type = { 3105 .name = "vxlan", 3106 }; 3107 3108 /* Calls the ndo_udp_tunnel_add of the caller in order to 3109 * supply the listening VXLAN udp ports. Callers are expected 3110 * to implement the ndo_udp_tunnel_add. 3111 */ 3112 static void vxlan_offload_rx_ports(struct net_device *dev, bool push) 3113 { 3114 struct vxlan_sock *vs; 3115 struct net *net = dev_net(dev); 3116 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3117 unsigned int i; 3118 3119 spin_lock(&vn->sock_lock); 3120 for (i = 0; i < PORT_HASH_SIZE; ++i) { 3121 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { 3122 unsigned short type; 3123 3124 if (vs->flags & VXLAN_F_GPE) 3125 type = UDP_TUNNEL_TYPE_VXLAN_GPE; 3126 else 3127 type = UDP_TUNNEL_TYPE_VXLAN; 3128 3129 if (push) 3130 udp_tunnel_push_rx_port(dev, vs->sock, type); 3131 else 3132 udp_tunnel_drop_rx_port(dev, vs->sock, type); 3133 } 3134 } 3135 spin_unlock(&vn->sock_lock); 3136 } 3137 3138 /* Initialize the device structure. */ 3139 static void vxlan_setup(struct net_device *dev) 3140 { 3141 struct vxlan_dev *vxlan = netdev_priv(dev); 3142 unsigned int h; 3143 3144 eth_hw_addr_random(dev); 3145 ether_setup(dev); 3146 3147 dev->needs_free_netdev = true; 3148 SET_NETDEV_DEVTYPE(dev, &vxlan_type); 3149 3150 dev->features |= NETIF_F_LLTX; 3151 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3152 dev->features |= NETIF_F_RXCSUM; 3153 dev->features |= NETIF_F_GSO_SOFTWARE; 3154 3155 dev->vlan_features = dev->features; 3156 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST; 3157 dev->hw_features |= NETIF_F_RXCSUM; 3158 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 3159 netif_keep_dst(dev); 3160 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN; 3161 3162 /* MTU range: 68 - 65535 */ 3163 dev->min_mtu = ETH_MIN_MTU; 3164 dev->max_mtu = ETH_MAX_MTU; 3165 3166 INIT_LIST_HEAD(&vxlan->next); 3167 3168 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE); 3169 3170 vxlan->dev = dev; 3171 3172 for (h = 0; h < FDB_HASH_SIZE; ++h) { 3173 spin_lock_init(&vxlan->hash_lock[h]); 3174 INIT_HLIST_HEAD(&vxlan->fdb_head[h]); 3175 } 3176 } 3177 3178 static void vxlan_ether_setup(struct net_device *dev) 3179 { 3180 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 3181 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 3182 dev->netdev_ops = &vxlan_netdev_ether_ops; 3183 } 3184 3185 static void vxlan_raw_setup(struct net_device *dev) 3186 { 3187 dev->header_ops = NULL; 3188 dev->type = ARPHRD_NONE; 3189 dev->hard_header_len = 0; 3190 dev->addr_len = 0; 3191 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 3192 dev->netdev_ops = &vxlan_netdev_raw_ops; 3193 } 3194 3195 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { 3196 [IFLA_VXLAN_ID] = { .type = NLA_U32 }, 3197 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) }, 3198 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, 3199 [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, 3200 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) }, 3201 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, 3202 [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, 3203 [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, 3204 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, 3205 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, 3206 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, 3207 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, 3208 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, 3209 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, 3210 [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, 3211 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, 3212 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, 3213 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, 3214 [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, 3215 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, 3216 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, 3217 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, 3218 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, 3219 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, 3220 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, 3221 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, 3222 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, 3223 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG }, 3224 [IFLA_VXLAN_DF] = { .type = NLA_U8 }, 3225 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 }, 3226 }; 3227 3228 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[], 3229 struct netlink_ext_ack *extack) 3230 { 3231 if (tb[IFLA_ADDRESS]) { 3232 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { 3233 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3234 "Provided link layer address is not Ethernet"); 3235 return -EINVAL; 3236 } 3237 3238 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { 3239 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS], 3240 "Provided Ethernet address is not unicast"); 3241 return -EADDRNOTAVAIL; 3242 } 3243 } 3244 3245 if (tb[IFLA_MTU]) { 3246 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 3247 3248 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) { 3249 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 3250 "MTU must be between 68 and 65535"); 3251 return -EINVAL; 3252 } 3253 } 3254 3255 if (!data) { 3256 NL_SET_ERR_MSG(extack, 3257 "Required attributes not provided to perform the operation"); 3258 return -EINVAL; 3259 } 3260 3261 if (data[IFLA_VXLAN_ID]) { 3262 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); 3263 3264 if (id >= VXLAN_N_VID) { 3265 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID], 3266 "VXLAN ID must be lower than 16777216"); 3267 return -ERANGE; 3268 } 3269 } 3270 3271 if (data[IFLA_VXLAN_PORT_RANGE]) { 3272 const struct ifla_vxlan_port_range *p 3273 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 3274 3275 if (ntohs(p->high) < ntohs(p->low)) { 3276 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE], 3277 "Invalid source port range"); 3278 return -EINVAL; 3279 } 3280 } 3281 3282 if (data[IFLA_VXLAN_DF]) { 3283 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]); 3284 3285 if (df < 0 || df > VXLAN_DF_MAX) { 3286 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF], 3287 "Invalid DF attribute"); 3288 return -EINVAL; 3289 } 3290 } 3291 3292 return 0; 3293 } 3294 3295 static void vxlan_get_drvinfo(struct net_device *netdev, 3296 struct ethtool_drvinfo *drvinfo) 3297 { 3298 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); 3299 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); 3300 } 3301 3302 static int vxlan_get_link_ksettings(struct net_device *dev, 3303 struct ethtool_link_ksettings *cmd) 3304 { 3305 struct vxlan_dev *vxlan = netdev_priv(dev); 3306 struct vxlan_rdst *dst = &vxlan->default_dst; 3307 struct net_device *lowerdev = __dev_get_by_index(vxlan->net, 3308 dst->remote_ifindex); 3309 3310 if (!lowerdev) { 3311 cmd->base.duplex = DUPLEX_UNKNOWN; 3312 cmd->base.port = PORT_OTHER; 3313 cmd->base.speed = SPEED_UNKNOWN; 3314 3315 return 0; 3316 } 3317 3318 return __ethtool_get_link_ksettings(lowerdev, cmd); 3319 } 3320 3321 static const struct ethtool_ops vxlan_ethtool_ops = { 3322 .get_drvinfo = vxlan_get_drvinfo, 3323 .get_link = ethtool_op_get_link, 3324 .get_link_ksettings = vxlan_get_link_ksettings, 3325 }; 3326 3327 static struct socket *vxlan_create_sock(struct net *net, bool ipv6, 3328 __be16 port, u32 flags, int ifindex) 3329 { 3330 struct socket *sock; 3331 struct udp_port_cfg udp_conf; 3332 int err; 3333 3334 memset(&udp_conf, 0, sizeof(udp_conf)); 3335 3336 if (ipv6) { 3337 udp_conf.family = AF_INET6; 3338 udp_conf.use_udp6_rx_checksums = 3339 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); 3340 udp_conf.ipv6_v6only = 1; 3341 } else { 3342 udp_conf.family = AF_INET; 3343 } 3344 3345 udp_conf.local_udp_port = port; 3346 udp_conf.bind_ifindex = ifindex; 3347 3348 /* Open UDP socket */ 3349 err = udp_sock_create(net, &udp_conf, &sock); 3350 if (err < 0) 3351 return ERR_PTR(err); 3352 3353 udp_allow_gso(sock->sk); 3354 return sock; 3355 } 3356 3357 /* Create new listen socket if needed */ 3358 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, 3359 __be16 port, u32 flags, 3360 int ifindex) 3361 { 3362 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3363 struct vxlan_sock *vs; 3364 struct socket *sock; 3365 unsigned int h; 3366 struct udp_tunnel_sock_cfg tunnel_cfg; 3367 3368 vs = kzalloc(sizeof(*vs), GFP_KERNEL); 3369 if (!vs) 3370 return ERR_PTR(-ENOMEM); 3371 3372 for (h = 0; h < VNI_HASH_SIZE; ++h) 3373 INIT_HLIST_HEAD(&vs->vni_list[h]); 3374 3375 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex); 3376 if (IS_ERR(sock)) { 3377 kfree(vs); 3378 return ERR_CAST(sock); 3379 } 3380 3381 vs->sock = sock; 3382 refcount_set(&vs->refcnt, 1); 3383 vs->flags = (flags & VXLAN_F_RCV_FLAGS); 3384 3385 spin_lock(&vn->sock_lock); 3386 hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); 3387 udp_tunnel_notify_add_rx_port(sock, 3388 (vs->flags & VXLAN_F_GPE) ? 3389 UDP_TUNNEL_TYPE_VXLAN_GPE : 3390 UDP_TUNNEL_TYPE_VXLAN); 3391 spin_unlock(&vn->sock_lock); 3392 3393 /* Mark socket as an encapsulation socket. */ 3394 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 3395 tunnel_cfg.sk_user_data = vs; 3396 tunnel_cfg.encap_type = 1; 3397 tunnel_cfg.encap_rcv = vxlan_rcv; 3398 tunnel_cfg.encap_err_lookup = vxlan_err_lookup; 3399 tunnel_cfg.encap_destroy = NULL; 3400 tunnel_cfg.gro_receive = vxlan_gro_receive; 3401 tunnel_cfg.gro_complete = vxlan_gro_complete; 3402 3403 setup_udp_tunnel_sock(net, sock, &tunnel_cfg); 3404 3405 return vs; 3406 } 3407 3408 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) 3409 { 3410 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); 3411 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3412 struct vxlan_sock *vs = NULL; 3413 struct vxlan_dev_node *node; 3414 int l3mdev_index = 0; 3415 3416 if (vxlan->cfg.remote_ifindex) 3417 l3mdev_index = l3mdev_master_upper_ifindex_by_index( 3418 vxlan->net, vxlan->cfg.remote_ifindex); 3419 3420 if (!vxlan->cfg.no_share) { 3421 spin_lock(&vn->sock_lock); 3422 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, 3423 vxlan->cfg.dst_port, vxlan->cfg.flags, 3424 l3mdev_index); 3425 if (vs && !refcount_inc_not_zero(&vs->refcnt)) { 3426 spin_unlock(&vn->sock_lock); 3427 return -EBUSY; 3428 } 3429 spin_unlock(&vn->sock_lock); 3430 } 3431 if (!vs) 3432 vs = vxlan_socket_create(vxlan->net, ipv6, 3433 vxlan->cfg.dst_port, vxlan->cfg.flags, 3434 l3mdev_index); 3435 if (IS_ERR(vs)) 3436 return PTR_ERR(vs); 3437 #if IS_ENABLED(CONFIG_IPV6) 3438 if (ipv6) { 3439 rcu_assign_pointer(vxlan->vn6_sock, vs); 3440 node = &vxlan->hlist6; 3441 } else 3442 #endif 3443 { 3444 rcu_assign_pointer(vxlan->vn4_sock, vs); 3445 node = &vxlan->hlist4; 3446 } 3447 3448 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER)) 3449 vxlan_vs_add_vnigrp(vxlan, vs, ipv6); 3450 else 3451 vxlan_vs_add_dev(vs, vxlan, node); 3452 3453 return 0; 3454 } 3455 3456 static int vxlan_sock_add(struct vxlan_dev *vxlan) 3457 { 3458 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA; 3459 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata; 3460 bool ipv4 = !ipv6 || metadata; 3461 int ret = 0; 3462 3463 RCU_INIT_POINTER(vxlan->vn4_sock, NULL); 3464 #if IS_ENABLED(CONFIG_IPV6) 3465 RCU_INIT_POINTER(vxlan->vn6_sock, NULL); 3466 if (ipv6) { 3467 ret = __vxlan_sock_add(vxlan, true); 3468 if (ret < 0 && ret != -EAFNOSUPPORT) 3469 ipv4 = false; 3470 } 3471 #endif 3472 if (ipv4) 3473 ret = __vxlan_sock_add(vxlan, false); 3474 if (ret < 0) 3475 vxlan_sock_release(vxlan); 3476 return ret; 3477 } 3478 3479 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan, 3480 struct vxlan_config *conf, __be32 vni) 3481 { 3482 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); 3483 struct vxlan_dev *tmp; 3484 3485 list_for_each_entry(tmp, &vn->vxlan_list, next) { 3486 if (tmp == vxlan) 3487 continue; 3488 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) { 3489 if (!vxlan_vnifilter_lookup(tmp, vni)) 3490 continue; 3491 } else if (tmp->cfg.vni != vni) { 3492 continue; 3493 } 3494 if (tmp->cfg.dst_port != conf->dst_port) 3495 continue; 3496 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) != 3497 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6))) 3498 continue; 3499 3500 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) && 3501 tmp->cfg.remote_ifindex != conf->remote_ifindex) 3502 continue; 3503 3504 return -EEXIST; 3505 } 3506 3507 return 0; 3508 } 3509 3510 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf, 3511 struct net_device **lower, 3512 struct vxlan_dev *old, 3513 struct netlink_ext_ack *extack) 3514 { 3515 bool use_ipv6 = false; 3516 3517 if (conf->flags & VXLAN_F_GPE) { 3518 /* For now, allow GPE only together with 3519 * COLLECT_METADATA. This can be relaxed later; in such 3520 * case, the other side of the PtP link will have to be 3521 * provided. 3522 */ 3523 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) || 3524 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3525 NL_SET_ERR_MSG(extack, 3526 "VXLAN GPE does not support this combination of attributes"); 3527 return -EINVAL; 3528 } 3529 } 3530 3531 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) { 3532 /* Unless IPv6 is explicitly requested, assume IPv4 */ 3533 conf->remote_ip.sa.sa_family = AF_INET; 3534 conf->saddr.sa.sa_family = AF_INET; 3535 } else if (!conf->remote_ip.sa.sa_family) { 3536 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family; 3537 } else if (!conf->saddr.sa.sa_family) { 3538 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family; 3539 } 3540 3541 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) { 3542 NL_SET_ERR_MSG(extack, 3543 "Local and remote address must be from the same family"); 3544 return -EINVAL; 3545 } 3546 3547 if (vxlan_addr_multicast(&conf->saddr)) { 3548 NL_SET_ERR_MSG(extack, "Local address cannot be multicast"); 3549 return -EINVAL; 3550 } 3551 3552 if (conf->saddr.sa.sa_family == AF_INET6) { 3553 if (!IS_ENABLED(CONFIG_IPV6)) { 3554 NL_SET_ERR_MSG(extack, 3555 "IPv6 support not enabled in the kernel"); 3556 return -EPFNOSUPPORT; 3557 } 3558 use_ipv6 = true; 3559 conf->flags |= VXLAN_F_IPV6; 3560 3561 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) { 3562 int local_type = 3563 ipv6_addr_type(&conf->saddr.sin6.sin6_addr); 3564 int remote_type = 3565 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr); 3566 3567 if (local_type & IPV6_ADDR_LINKLOCAL) { 3568 if (!(remote_type & IPV6_ADDR_LINKLOCAL) && 3569 (remote_type != IPV6_ADDR_ANY)) { 3570 NL_SET_ERR_MSG(extack, 3571 "Invalid combination of local and remote address scopes"); 3572 return -EINVAL; 3573 } 3574 3575 conf->flags |= VXLAN_F_IPV6_LINKLOCAL; 3576 } else { 3577 if (remote_type == 3578 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) { 3579 NL_SET_ERR_MSG(extack, 3580 "Invalid combination of local and remote address scopes"); 3581 return -EINVAL; 3582 } 3583 3584 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL; 3585 } 3586 } 3587 } 3588 3589 if (conf->label && !use_ipv6) { 3590 NL_SET_ERR_MSG(extack, 3591 "Label attribute only applies to IPv6 VXLAN devices"); 3592 return -EINVAL; 3593 } 3594 3595 if (conf->remote_ifindex) { 3596 struct net_device *lowerdev; 3597 3598 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); 3599 if (!lowerdev) { 3600 NL_SET_ERR_MSG(extack, 3601 "Invalid local interface, device not found"); 3602 return -ENODEV; 3603 } 3604 3605 #if IS_ENABLED(CONFIG_IPV6) 3606 if (use_ipv6) { 3607 struct inet6_dev *idev = __in6_dev_get(lowerdev); 3608 3609 if (idev && idev->cnf.disable_ipv6) { 3610 NL_SET_ERR_MSG(extack, 3611 "IPv6 support disabled by administrator"); 3612 return -EPERM; 3613 } 3614 } 3615 #endif 3616 3617 *lower = lowerdev; 3618 } else { 3619 if (vxlan_addr_multicast(&conf->remote_ip)) { 3620 NL_SET_ERR_MSG(extack, 3621 "Local interface required for multicast remote destination"); 3622 3623 return -EINVAL; 3624 } 3625 3626 #if IS_ENABLED(CONFIG_IPV6) 3627 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) { 3628 NL_SET_ERR_MSG(extack, 3629 "Local interface required for link-local local/remote addresses"); 3630 return -EINVAL; 3631 } 3632 #endif 3633 3634 *lower = NULL; 3635 } 3636 3637 if (!conf->dst_port) { 3638 if (conf->flags & VXLAN_F_GPE) 3639 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT); 3640 else 3641 conf->dst_port = htons(vxlan_port); 3642 } 3643 3644 if (!conf->age_interval) 3645 conf->age_interval = FDB_AGE_DEFAULT; 3646 3647 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) { 3648 NL_SET_ERR_MSG(extack, 3649 "A VXLAN device with the specified VNI already exists"); 3650 return -EEXIST; 3651 } 3652 3653 return 0; 3654 } 3655 3656 static void vxlan_config_apply(struct net_device *dev, 3657 struct vxlan_config *conf, 3658 struct net_device *lowerdev, 3659 struct net *src_net, 3660 bool changelink) 3661 { 3662 struct vxlan_dev *vxlan = netdev_priv(dev); 3663 struct vxlan_rdst *dst = &vxlan->default_dst; 3664 unsigned short needed_headroom = ETH_HLEN; 3665 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6); 3666 int max_mtu = ETH_MAX_MTU; 3667 3668 if (!changelink) { 3669 if (conf->flags & VXLAN_F_GPE) 3670 vxlan_raw_setup(dev); 3671 else 3672 vxlan_ether_setup(dev); 3673 3674 if (conf->mtu) 3675 dev->mtu = conf->mtu; 3676 3677 vxlan->net = src_net; 3678 } 3679 3680 dst->remote_vni = conf->vni; 3681 3682 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); 3683 3684 if (lowerdev) { 3685 dst->remote_ifindex = conf->remote_ifindex; 3686 3687 netif_inherit_tso_max(dev, lowerdev); 3688 3689 needed_headroom = lowerdev->hard_header_len; 3690 needed_headroom += lowerdev->needed_headroom; 3691 3692 dev->needed_tailroom = lowerdev->needed_tailroom; 3693 3694 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : 3695 VXLAN_HEADROOM); 3696 if (max_mtu < ETH_MIN_MTU) 3697 max_mtu = ETH_MIN_MTU; 3698 3699 if (!changelink && !conf->mtu) 3700 dev->mtu = max_mtu; 3701 } 3702 3703 if (dev->mtu > max_mtu) 3704 dev->mtu = max_mtu; 3705 3706 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA) 3707 needed_headroom += VXLAN6_HEADROOM; 3708 else 3709 needed_headroom += VXLAN_HEADROOM; 3710 dev->needed_headroom = needed_headroom; 3711 3712 memcpy(&vxlan->cfg, conf, sizeof(*conf)); 3713 } 3714 3715 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, 3716 struct vxlan_config *conf, bool changelink, 3717 struct netlink_ext_ack *extack) 3718 { 3719 struct vxlan_dev *vxlan = netdev_priv(dev); 3720 struct net_device *lowerdev; 3721 int ret; 3722 3723 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack); 3724 if (ret) 3725 return ret; 3726 3727 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink); 3728 3729 return 0; 3730 } 3731 3732 static int __vxlan_dev_create(struct net *net, struct net_device *dev, 3733 struct vxlan_config *conf, 3734 struct netlink_ext_ack *extack) 3735 { 3736 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 3737 struct vxlan_dev *vxlan = netdev_priv(dev); 3738 struct net_device *remote_dev = NULL; 3739 struct vxlan_fdb *f = NULL; 3740 bool unregister = false; 3741 struct vxlan_rdst *dst; 3742 int err; 3743 3744 dst = &vxlan->default_dst; 3745 err = vxlan_dev_configure(net, dev, conf, false, extack); 3746 if (err) 3747 return err; 3748 3749 dev->ethtool_ops = &vxlan_ethtool_ops; 3750 3751 /* create an fdb entry for a valid default destination */ 3752 if (!vxlan_addr_any(&dst->remote_ip)) { 3753 err = vxlan_fdb_create(vxlan, all_zeros_mac, 3754 &dst->remote_ip, 3755 NUD_REACHABLE | NUD_PERMANENT, 3756 vxlan->cfg.dst_port, 3757 dst->remote_vni, 3758 dst->remote_vni, 3759 dst->remote_ifindex, 3760 NTF_SELF, 0, &f, extack); 3761 if (err) 3762 return err; 3763 } 3764 3765 err = register_netdevice(dev); 3766 if (err) 3767 goto errout; 3768 unregister = true; 3769 3770 if (dst->remote_ifindex) { 3771 remote_dev = __dev_get_by_index(net, dst->remote_ifindex); 3772 if (!remote_dev) { 3773 err = -ENODEV; 3774 goto errout; 3775 } 3776 3777 err = netdev_upper_dev_link(remote_dev, dev, extack); 3778 if (err) 3779 goto errout; 3780 } 3781 3782 err = rtnl_configure_link(dev, NULL, 0, NULL); 3783 if (err < 0) 3784 goto unlink; 3785 3786 if (f) { 3787 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f); 3788 3789 /* notify default fdb entry */ 3790 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), 3791 RTM_NEWNEIGH, true, extack); 3792 if (err) { 3793 vxlan_fdb_destroy(vxlan, f, false, false); 3794 if (remote_dev) 3795 netdev_upper_dev_unlink(remote_dev, dev); 3796 goto unregister; 3797 } 3798 } 3799 3800 list_add(&vxlan->next, &vn->vxlan_list); 3801 if (remote_dev) 3802 dst->remote_dev = remote_dev; 3803 return 0; 3804 unlink: 3805 if (remote_dev) 3806 netdev_upper_dev_unlink(remote_dev, dev); 3807 errout: 3808 /* unregister_netdevice() destroys the default FDB entry with deletion 3809 * notification. But the addition notification was not sent yet, so 3810 * destroy the entry by hand here. 3811 */ 3812 if (f) 3813 __vxlan_fdb_free(f); 3814 unregister: 3815 if (unregister) 3816 unregister_netdevice(dev); 3817 return err; 3818 } 3819 3820 /* Set/clear flags based on attribute */ 3821 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[], 3822 int attrtype, unsigned long mask, bool changelink, 3823 bool changelink_supported, 3824 struct netlink_ext_ack *extack) 3825 { 3826 unsigned long flags; 3827 3828 if (!tb[attrtype]) 3829 return 0; 3830 3831 if (changelink && !changelink_supported) { 3832 vxlan_flag_attr_error(attrtype, extack); 3833 return -EOPNOTSUPP; 3834 } 3835 3836 if (vxlan_policy[attrtype].type == NLA_FLAG) 3837 flags = conf->flags | mask; 3838 else if (nla_get_u8(tb[attrtype])) 3839 flags = conf->flags | mask; 3840 else 3841 flags = conf->flags & ~mask; 3842 3843 conf->flags = flags; 3844 3845 return 0; 3846 } 3847 3848 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[], 3849 struct net_device *dev, struct vxlan_config *conf, 3850 bool changelink, struct netlink_ext_ack *extack) 3851 { 3852 struct vxlan_dev *vxlan = netdev_priv(dev); 3853 int err = 0; 3854 3855 memset(conf, 0, sizeof(*conf)); 3856 3857 /* if changelink operation, start with old existing cfg */ 3858 if (changelink) 3859 memcpy(conf, &vxlan->cfg, sizeof(*conf)); 3860 3861 if (data[IFLA_VXLAN_ID]) { 3862 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 3863 3864 if (changelink && (vni != conf->vni)) { 3865 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI"); 3866 return -EOPNOTSUPP; 3867 } 3868 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); 3869 } 3870 3871 if (data[IFLA_VXLAN_GROUP]) { 3872 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) { 3873 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group"); 3874 return -EOPNOTSUPP; 3875 } 3876 3877 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); 3878 conf->remote_ip.sa.sa_family = AF_INET; 3879 } else if (data[IFLA_VXLAN_GROUP6]) { 3880 if (!IS_ENABLED(CONFIG_IPV6)) { 3881 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel"); 3882 return -EPFNOSUPPORT; 3883 } 3884 3885 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) { 3886 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group"); 3887 return -EOPNOTSUPP; 3888 } 3889 3890 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); 3891 conf->remote_ip.sa.sa_family = AF_INET6; 3892 } 3893 3894 if (data[IFLA_VXLAN_LOCAL]) { 3895 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) { 3896 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old"); 3897 return -EOPNOTSUPP; 3898 } 3899 3900 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); 3901 conf->saddr.sa.sa_family = AF_INET; 3902 } else if (data[IFLA_VXLAN_LOCAL6]) { 3903 if (!IS_ENABLED(CONFIG_IPV6)) { 3904 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel"); 3905 return -EPFNOSUPPORT; 3906 } 3907 3908 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) { 3909 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old"); 3910 return -EOPNOTSUPP; 3911 } 3912 3913 /* TODO: respect scope id */ 3914 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); 3915 conf->saddr.sa.sa_family = AF_INET6; 3916 } 3917 3918 if (data[IFLA_VXLAN_LINK]) 3919 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); 3920 3921 if (data[IFLA_VXLAN_TOS]) 3922 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); 3923 3924 if (data[IFLA_VXLAN_TTL]) 3925 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); 3926 3927 if (data[IFLA_VXLAN_TTL_INHERIT]) { 3928 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT, 3929 VXLAN_F_TTL_INHERIT, changelink, false, 3930 extack); 3931 if (err) 3932 return err; 3933 3934 } 3935 3936 if (data[IFLA_VXLAN_LABEL]) 3937 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & 3938 IPV6_FLOWLABEL_MASK; 3939 3940 if (data[IFLA_VXLAN_LEARNING]) { 3941 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING, 3942 VXLAN_F_LEARN, changelink, true, 3943 extack); 3944 if (err) 3945 return err; 3946 } else if (!changelink) { 3947 /* default to learn on a new device */ 3948 conf->flags |= VXLAN_F_LEARN; 3949 } 3950 3951 if (data[IFLA_VXLAN_AGEING]) 3952 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); 3953 3954 if (data[IFLA_VXLAN_PROXY]) { 3955 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY, 3956 VXLAN_F_PROXY, changelink, false, 3957 extack); 3958 if (err) 3959 return err; 3960 } 3961 3962 if (data[IFLA_VXLAN_RSC]) { 3963 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC, 3964 VXLAN_F_RSC, changelink, false, 3965 extack); 3966 if (err) 3967 return err; 3968 } 3969 3970 if (data[IFLA_VXLAN_L2MISS]) { 3971 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS, 3972 VXLAN_F_L2MISS, changelink, false, 3973 extack); 3974 if (err) 3975 return err; 3976 } 3977 3978 if (data[IFLA_VXLAN_L3MISS]) { 3979 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS, 3980 VXLAN_F_L3MISS, changelink, false, 3981 extack); 3982 if (err) 3983 return err; 3984 } 3985 3986 if (data[IFLA_VXLAN_LIMIT]) { 3987 if (changelink) { 3988 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT], 3989 "Cannot change limit"); 3990 return -EOPNOTSUPP; 3991 } 3992 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); 3993 } 3994 3995 if (data[IFLA_VXLAN_COLLECT_METADATA]) { 3996 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA, 3997 VXLAN_F_COLLECT_METADATA, changelink, false, 3998 extack); 3999 if (err) 4000 return err; 4001 } 4002 4003 if (data[IFLA_VXLAN_PORT_RANGE]) { 4004 if (!changelink) { 4005 const struct ifla_vxlan_port_range *p 4006 = nla_data(data[IFLA_VXLAN_PORT_RANGE]); 4007 conf->port_min = ntohs(p->low); 4008 conf->port_max = ntohs(p->high); 4009 } else { 4010 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE], 4011 "Cannot change port range"); 4012 return -EOPNOTSUPP; 4013 } 4014 } 4015 4016 if (data[IFLA_VXLAN_PORT]) { 4017 if (changelink) { 4018 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT], 4019 "Cannot change port"); 4020 return -EOPNOTSUPP; 4021 } 4022 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); 4023 } 4024 4025 if (data[IFLA_VXLAN_UDP_CSUM]) { 4026 if (changelink) { 4027 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM], 4028 "Cannot change UDP_CSUM flag"); 4029 return -EOPNOTSUPP; 4030 } 4031 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) 4032 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX; 4033 } 4034 4035 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) { 4036 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4037 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink, 4038 false, extack); 4039 if (err) 4040 return err; 4041 } 4042 4043 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) { 4044 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4045 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink, 4046 false, extack); 4047 if (err) 4048 return err; 4049 } 4050 4051 if (data[IFLA_VXLAN_REMCSUM_TX]) { 4052 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX, 4053 VXLAN_F_REMCSUM_TX, changelink, false, 4054 extack); 4055 if (err) 4056 return err; 4057 } 4058 4059 if (data[IFLA_VXLAN_REMCSUM_RX]) { 4060 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX, 4061 VXLAN_F_REMCSUM_RX, changelink, false, 4062 extack); 4063 if (err) 4064 return err; 4065 } 4066 4067 if (data[IFLA_VXLAN_GBP]) { 4068 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP, 4069 VXLAN_F_GBP, changelink, false, extack); 4070 if (err) 4071 return err; 4072 } 4073 4074 if (data[IFLA_VXLAN_GPE]) { 4075 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE, 4076 VXLAN_F_GPE, changelink, false, 4077 extack); 4078 if (err) 4079 return err; 4080 } 4081 4082 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) { 4083 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL, 4084 VXLAN_F_REMCSUM_NOPARTIAL, changelink, 4085 false, extack); 4086 if (err) 4087 return err; 4088 } 4089 4090 if (tb[IFLA_MTU]) { 4091 if (changelink) { 4092 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU], 4093 "Cannot change mtu"); 4094 return -EOPNOTSUPP; 4095 } 4096 conf->mtu = nla_get_u32(tb[IFLA_MTU]); 4097 } 4098 4099 if (data[IFLA_VXLAN_DF]) 4100 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]); 4101 4102 if (data[IFLA_VXLAN_VNIFILTER]) { 4103 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER, 4104 VXLAN_F_VNIFILTER, changelink, false, 4105 extack); 4106 if (err) 4107 return err; 4108 4109 if ((conf->flags & VXLAN_F_VNIFILTER) && 4110 !(conf->flags & VXLAN_F_COLLECT_METADATA)) { 4111 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER], 4112 "vxlan vnifilter only valid in collect metadata mode"); 4113 return -EINVAL; 4114 } 4115 } 4116 4117 return 0; 4118 } 4119 4120 static int vxlan_newlink(struct net *src_net, struct net_device *dev, 4121 struct nlattr *tb[], struct nlattr *data[], 4122 struct netlink_ext_ack *extack) 4123 { 4124 struct vxlan_config conf; 4125 int err; 4126 4127 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack); 4128 if (err) 4129 return err; 4130 4131 return __vxlan_dev_create(src_net, dev, &conf, extack); 4132 } 4133 4134 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[], 4135 struct nlattr *data[], 4136 struct netlink_ext_ack *extack) 4137 { 4138 struct vxlan_dev *vxlan = netdev_priv(dev); 4139 struct net_device *lowerdev; 4140 struct vxlan_config conf; 4141 struct vxlan_rdst *dst; 4142 int err; 4143 4144 dst = &vxlan->default_dst; 4145 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack); 4146 if (err) 4147 return err; 4148 4149 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev, 4150 vxlan, extack); 4151 if (err) 4152 return err; 4153 4154 if (dst->remote_dev == lowerdev) 4155 lowerdev = NULL; 4156 4157 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev, 4158 extack); 4159 if (err) 4160 return err; 4161 4162 /* handle default dst entry */ 4163 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) { 4164 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni); 4165 4166 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4167 if (!vxlan_addr_any(&conf.remote_ip)) { 4168 err = vxlan_fdb_update(vxlan, all_zeros_mac, 4169 &conf.remote_ip, 4170 NUD_REACHABLE | NUD_PERMANENT, 4171 NLM_F_APPEND | NLM_F_CREATE, 4172 vxlan->cfg.dst_port, 4173 conf.vni, conf.vni, 4174 conf.remote_ifindex, 4175 NTF_SELF, 0, true, extack); 4176 if (err) { 4177 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4178 netdev_adjacent_change_abort(dst->remote_dev, 4179 lowerdev, dev); 4180 return err; 4181 } 4182 } 4183 if (!vxlan_addr_any(&dst->remote_ip)) 4184 __vxlan_fdb_delete(vxlan, all_zeros_mac, 4185 dst->remote_ip, 4186 vxlan->cfg.dst_port, 4187 dst->remote_vni, 4188 dst->remote_vni, 4189 dst->remote_ifindex, 4190 true); 4191 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4192 4193 /* If vni filtering device, also update fdb entries of 4194 * all vnis that were using default remote ip 4195 */ 4196 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) { 4197 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip, 4198 &conf.remote_ip, extack); 4199 if (err) { 4200 netdev_adjacent_change_abort(dst->remote_dev, 4201 lowerdev, dev); 4202 return err; 4203 } 4204 } 4205 } 4206 4207 if (conf.age_interval != vxlan->cfg.age_interval) 4208 mod_timer(&vxlan->age_timer, jiffies); 4209 4210 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev); 4211 if (lowerdev && lowerdev != dst->remote_dev) 4212 dst->remote_dev = lowerdev; 4213 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true); 4214 return 0; 4215 } 4216 4217 static void vxlan_dellink(struct net_device *dev, struct list_head *head) 4218 { 4219 struct vxlan_dev *vxlan = netdev_priv(dev); 4220 4221 vxlan_flush(vxlan, true); 4222 4223 list_del(&vxlan->next); 4224 unregister_netdevice_queue(dev, head); 4225 if (vxlan->default_dst.remote_dev) 4226 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev); 4227 } 4228 4229 static size_t vxlan_get_size(const struct net_device *dev) 4230 { 4231 4232 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ 4233 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ 4234 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ 4235 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ 4236 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ 4237 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */ 4238 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ 4239 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */ 4240 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ 4241 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ 4242 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ 4243 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ 4244 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ 4245 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ 4246 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ 4247 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ 4248 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ 4249 nla_total_size(sizeof(struct ifla_vxlan_port_range)) + 4250 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ 4251 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ 4252 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ 4253 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ 4254 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ 4255 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ 4256 0; 4257 } 4258 4259 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) 4260 { 4261 const struct vxlan_dev *vxlan = netdev_priv(dev); 4262 const struct vxlan_rdst *dst = &vxlan->default_dst; 4263 struct ifla_vxlan_port_range ports = { 4264 .low = htons(vxlan->cfg.port_min), 4265 .high = htons(vxlan->cfg.port_max), 4266 }; 4267 4268 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) 4269 goto nla_put_failure; 4270 4271 if (!vxlan_addr_any(&dst->remote_ip)) { 4272 if (dst->remote_ip.sa.sa_family == AF_INET) { 4273 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, 4274 dst->remote_ip.sin.sin_addr.s_addr)) 4275 goto nla_put_failure; 4276 #if IS_ENABLED(CONFIG_IPV6) 4277 } else { 4278 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, 4279 &dst->remote_ip.sin6.sin6_addr)) 4280 goto nla_put_failure; 4281 #endif 4282 } 4283 } 4284 4285 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) 4286 goto nla_put_failure; 4287 4288 if (!vxlan_addr_any(&vxlan->cfg.saddr)) { 4289 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { 4290 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, 4291 vxlan->cfg.saddr.sin.sin_addr.s_addr)) 4292 goto nla_put_failure; 4293 #if IS_ENABLED(CONFIG_IPV6) 4294 } else { 4295 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, 4296 &vxlan->cfg.saddr.sin6.sin6_addr)) 4297 goto nla_put_failure; 4298 #endif 4299 } 4300 } 4301 4302 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || 4303 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT, 4304 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) || 4305 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || 4306 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) || 4307 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || 4308 nla_put_u8(skb, IFLA_VXLAN_LEARNING, 4309 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) || 4310 nla_put_u8(skb, IFLA_VXLAN_PROXY, 4311 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) || 4312 nla_put_u8(skb, IFLA_VXLAN_RSC, 4313 !!(vxlan->cfg.flags & VXLAN_F_RSC)) || 4314 nla_put_u8(skb, IFLA_VXLAN_L2MISS, 4315 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) || 4316 nla_put_u8(skb, IFLA_VXLAN_L3MISS, 4317 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) || 4318 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, 4319 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) || 4320 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || 4321 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || 4322 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || 4323 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, 4324 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || 4325 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, 4326 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || 4327 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, 4328 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || 4329 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, 4330 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) || 4331 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, 4332 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX))) 4333 goto nla_put_failure; 4334 4335 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) 4336 goto nla_put_failure; 4337 4338 if (vxlan->cfg.flags & VXLAN_F_GBP && 4339 nla_put_flag(skb, IFLA_VXLAN_GBP)) 4340 goto nla_put_failure; 4341 4342 if (vxlan->cfg.flags & VXLAN_F_GPE && 4343 nla_put_flag(skb, IFLA_VXLAN_GPE)) 4344 goto nla_put_failure; 4345 4346 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL && 4347 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) 4348 goto nla_put_failure; 4349 4350 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER && 4351 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER, 4352 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))) 4353 goto nla_put_failure; 4354 4355 return 0; 4356 4357 nla_put_failure: 4358 return -EMSGSIZE; 4359 } 4360 4361 static struct net *vxlan_get_link_net(const struct net_device *dev) 4362 { 4363 struct vxlan_dev *vxlan = netdev_priv(dev); 4364 4365 return vxlan->net; 4366 } 4367 4368 static struct rtnl_link_ops vxlan_link_ops __read_mostly = { 4369 .kind = "vxlan", 4370 .maxtype = IFLA_VXLAN_MAX, 4371 .policy = vxlan_policy, 4372 .priv_size = sizeof(struct vxlan_dev), 4373 .setup = vxlan_setup, 4374 .validate = vxlan_validate, 4375 .newlink = vxlan_newlink, 4376 .changelink = vxlan_changelink, 4377 .dellink = vxlan_dellink, 4378 .get_size = vxlan_get_size, 4379 .fill_info = vxlan_fill_info, 4380 .get_link_net = vxlan_get_link_net, 4381 }; 4382 4383 struct net_device *vxlan_dev_create(struct net *net, const char *name, 4384 u8 name_assign_type, 4385 struct vxlan_config *conf) 4386 { 4387 struct nlattr *tb[IFLA_MAX + 1]; 4388 struct net_device *dev; 4389 int err; 4390 4391 memset(&tb, 0, sizeof(tb)); 4392 4393 dev = rtnl_create_link(net, name, name_assign_type, 4394 &vxlan_link_ops, tb, NULL); 4395 if (IS_ERR(dev)) 4396 return dev; 4397 4398 err = __vxlan_dev_create(net, dev, conf, NULL); 4399 if (err < 0) { 4400 free_netdev(dev); 4401 return ERR_PTR(err); 4402 } 4403 4404 err = rtnl_configure_link(dev, NULL, 0, NULL); 4405 if (err < 0) { 4406 LIST_HEAD(list_kill); 4407 4408 vxlan_dellink(dev, &list_kill); 4409 unregister_netdevice_many(&list_kill); 4410 return ERR_PTR(err); 4411 } 4412 4413 return dev; 4414 } 4415 EXPORT_SYMBOL_GPL(vxlan_dev_create); 4416 4417 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, 4418 struct net_device *dev) 4419 { 4420 struct vxlan_dev *vxlan, *next; 4421 LIST_HEAD(list_kill); 4422 4423 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4424 struct vxlan_rdst *dst = &vxlan->default_dst; 4425 4426 /* In case we created vxlan device with carrier 4427 * and we loose the carrier due to module unload 4428 * we also need to remove vxlan device. In other 4429 * cases, it's not necessary and remote_ifindex 4430 * is 0 here, so no matches. 4431 */ 4432 if (dst->remote_ifindex == dev->ifindex) 4433 vxlan_dellink(vxlan->dev, &list_kill); 4434 } 4435 4436 unregister_netdevice_many(&list_kill); 4437 } 4438 4439 static int vxlan_netdevice_event(struct notifier_block *unused, 4440 unsigned long event, void *ptr) 4441 { 4442 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4443 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); 4444 4445 if (event == NETDEV_UNREGISTER) 4446 vxlan_handle_lowerdev_unregister(vn, dev); 4447 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) 4448 vxlan_offload_rx_ports(dev, true); 4449 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO) 4450 vxlan_offload_rx_ports(dev, false); 4451 4452 return NOTIFY_DONE; 4453 } 4454 4455 static struct notifier_block vxlan_notifier_block __read_mostly = { 4456 .notifier_call = vxlan_netdevice_event, 4457 }; 4458 4459 static void 4460 vxlan_fdb_offloaded_set(struct net_device *dev, 4461 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4462 { 4463 struct vxlan_dev *vxlan = netdev_priv(dev); 4464 struct vxlan_rdst *rdst; 4465 struct vxlan_fdb *f; 4466 u32 hash_index; 4467 4468 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4469 4470 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4471 4472 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4473 if (!f) 4474 goto out; 4475 4476 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip, 4477 fdb_info->remote_port, 4478 fdb_info->remote_vni, 4479 fdb_info->remote_ifindex); 4480 if (!rdst) 4481 goto out; 4482 4483 rdst->offloaded = fdb_info->offloaded; 4484 4485 out: 4486 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4487 } 4488 4489 static int 4490 vxlan_fdb_external_learn_add(struct net_device *dev, 4491 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4492 { 4493 struct vxlan_dev *vxlan = netdev_priv(dev); 4494 struct netlink_ext_ack *extack; 4495 u32 hash_index; 4496 int err; 4497 4498 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4499 extack = switchdev_notifier_info_to_extack(&fdb_info->info); 4500 4501 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4502 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip, 4503 NUD_REACHABLE, 4504 NLM_F_CREATE | NLM_F_REPLACE, 4505 fdb_info->remote_port, 4506 fdb_info->vni, 4507 fdb_info->remote_vni, 4508 fdb_info->remote_ifindex, 4509 NTF_USE | NTF_SELF | NTF_EXT_LEARNED, 4510 0, false, extack); 4511 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4512 4513 return err; 4514 } 4515 4516 static int 4517 vxlan_fdb_external_learn_del(struct net_device *dev, 4518 struct switchdev_notifier_vxlan_fdb_info *fdb_info) 4519 { 4520 struct vxlan_dev *vxlan = netdev_priv(dev); 4521 struct vxlan_fdb *f; 4522 u32 hash_index; 4523 int err = 0; 4524 4525 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni); 4526 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4527 4528 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni); 4529 if (!f) 4530 err = -ENOENT; 4531 else if (f->flags & NTF_EXT_LEARNED) 4532 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr, 4533 fdb_info->remote_ip, 4534 fdb_info->remote_port, 4535 fdb_info->vni, 4536 fdb_info->remote_vni, 4537 fdb_info->remote_ifindex, 4538 false); 4539 4540 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4541 4542 return err; 4543 } 4544 4545 static int vxlan_switchdev_event(struct notifier_block *unused, 4546 unsigned long event, void *ptr) 4547 { 4548 struct net_device *dev = switchdev_notifier_info_to_dev(ptr); 4549 struct switchdev_notifier_vxlan_fdb_info *fdb_info; 4550 int err = 0; 4551 4552 switch (event) { 4553 case SWITCHDEV_VXLAN_FDB_OFFLOADED: 4554 vxlan_fdb_offloaded_set(dev, ptr); 4555 break; 4556 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE: 4557 fdb_info = ptr; 4558 err = vxlan_fdb_external_learn_add(dev, fdb_info); 4559 if (err) { 4560 err = notifier_from_errno(err); 4561 break; 4562 } 4563 fdb_info->offloaded = true; 4564 vxlan_fdb_offloaded_set(dev, fdb_info); 4565 break; 4566 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE: 4567 fdb_info = ptr; 4568 err = vxlan_fdb_external_learn_del(dev, fdb_info); 4569 if (err) { 4570 err = notifier_from_errno(err); 4571 break; 4572 } 4573 fdb_info->offloaded = false; 4574 vxlan_fdb_offloaded_set(dev, fdb_info); 4575 break; 4576 } 4577 4578 return err; 4579 } 4580 4581 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = { 4582 .notifier_call = vxlan_switchdev_event, 4583 }; 4584 4585 static void vxlan_fdb_nh_flush(struct nexthop *nh) 4586 { 4587 struct vxlan_fdb *fdb; 4588 struct vxlan_dev *vxlan; 4589 u32 hash_index; 4590 4591 rcu_read_lock(); 4592 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) { 4593 vxlan = rcu_dereference(fdb->vdev); 4594 WARN_ON(!vxlan); 4595 hash_index = fdb_head_index(vxlan, fdb->eth_addr, 4596 vxlan->default_dst.remote_vni); 4597 spin_lock_bh(&vxlan->hash_lock[hash_index]); 4598 if (!hlist_unhashed(&fdb->hlist)) 4599 vxlan_fdb_destroy(vxlan, fdb, false, false); 4600 spin_unlock_bh(&vxlan->hash_lock[hash_index]); 4601 } 4602 rcu_read_unlock(); 4603 } 4604 4605 static int vxlan_nexthop_event(struct notifier_block *nb, 4606 unsigned long event, void *ptr) 4607 { 4608 struct nh_notifier_info *info = ptr; 4609 struct nexthop *nh; 4610 4611 if (event != NEXTHOP_EVENT_DEL) 4612 return NOTIFY_DONE; 4613 4614 nh = nexthop_find_by_id(info->net, info->id); 4615 if (!nh) 4616 return NOTIFY_DONE; 4617 4618 vxlan_fdb_nh_flush(nh); 4619 4620 return NOTIFY_DONE; 4621 } 4622 4623 static __net_init int vxlan_init_net(struct net *net) 4624 { 4625 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4626 unsigned int h; 4627 4628 INIT_LIST_HEAD(&vn->vxlan_list); 4629 spin_lock_init(&vn->sock_lock); 4630 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event; 4631 4632 for (h = 0; h < PORT_HASH_SIZE; ++h) 4633 INIT_HLIST_HEAD(&vn->sock_list[h]); 4634 4635 return register_nexthop_notifier(net, &vn->nexthop_notifier_block, 4636 NULL); 4637 } 4638 4639 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head) 4640 { 4641 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4642 struct vxlan_dev *vxlan, *next; 4643 struct net_device *dev, *aux; 4644 4645 for_each_netdev_safe(net, dev, aux) 4646 if (dev->rtnl_link_ops == &vxlan_link_ops) 4647 unregister_netdevice_queue(dev, head); 4648 4649 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { 4650 /* If vxlan->dev is in the same netns, it has already been added 4651 * to the list by the previous loop. 4652 */ 4653 if (!net_eq(dev_net(vxlan->dev), net)) 4654 unregister_netdevice_queue(vxlan->dev, head); 4655 } 4656 4657 } 4658 4659 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list) 4660 { 4661 struct net *net; 4662 LIST_HEAD(list); 4663 unsigned int h; 4664 4665 list_for_each_entry(net, net_list, exit_list) { 4666 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4667 4668 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block); 4669 } 4670 rtnl_lock(); 4671 list_for_each_entry(net, net_list, exit_list) 4672 vxlan_destroy_tunnels(net, &list); 4673 4674 unregister_netdevice_many(&list); 4675 rtnl_unlock(); 4676 4677 list_for_each_entry(net, net_list, exit_list) { 4678 struct vxlan_net *vn = net_generic(net, vxlan_net_id); 4679 4680 for (h = 0; h < PORT_HASH_SIZE; ++h) 4681 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h])); 4682 } 4683 } 4684 4685 static struct pernet_operations vxlan_net_ops = { 4686 .init = vxlan_init_net, 4687 .exit_batch = vxlan_exit_batch_net, 4688 .id = &vxlan_net_id, 4689 .size = sizeof(struct vxlan_net), 4690 }; 4691 4692 static int __init vxlan_init_module(void) 4693 { 4694 int rc; 4695 4696 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); 4697 4698 rc = register_pernet_subsys(&vxlan_net_ops); 4699 if (rc) 4700 goto out1; 4701 4702 rc = register_netdevice_notifier(&vxlan_notifier_block); 4703 if (rc) 4704 goto out2; 4705 4706 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block); 4707 if (rc) 4708 goto out3; 4709 4710 rc = rtnl_link_register(&vxlan_link_ops); 4711 if (rc) 4712 goto out4; 4713 4714 vxlan_vnifilter_init(); 4715 4716 return 0; 4717 out4: 4718 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4719 out3: 4720 unregister_netdevice_notifier(&vxlan_notifier_block); 4721 out2: 4722 unregister_pernet_subsys(&vxlan_net_ops); 4723 out1: 4724 return rc; 4725 } 4726 late_initcall(vxlan_init_module); 4727 4728 static void __exit vxlan_cleanup_module(void) 4729 { 4730 vxlan_vnifilter_uninit(); 4731 rtnl_link_unregister(&vxlan_link_ops); 4732 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block); 4733 unregister_netdevice_notifier(&vxlan_notifier_block); 4734 unregister_pernet_subsys(&vxlan_net_ops); 4735 /* rcu_barrier() is called by netns */ 4736 } 4737 module_exit(vxlan_cleanup_module); 4738 4739 MODULE_LICENSE("GPL"); 4740 MODULE_VERSION(VXLAN_VERSION); 4741 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); 4742 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); 4743 MODULE_ALIAS_RTNL_LINK("vxlan"); 4744