1 // SPDX-License-Identifier: GPL-2.0-only
3 * VXLAN: Virtual eXtensible Local Area Network
5 * Copyright (c) 2012-2013 Vyatta Inc.
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
19 #include <net/ndisc.h>
20 #include <net/ipv6_stubs.h>
23 #include <net/rtnetlink.h>
24 #include <net/inet_ecn.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27 #include <net/tun_proto.h>
28 #include <net/vxlan.h>
30 #if IS_ENABLED(CONFIG_IPV6)
31 #include <net/ip6_tunnel.h>
32 #include <net/ip6_checksum.h>
35 #define VXLAN_VERSION "0.1"
37 #define PORT_HASH_BITS 8
38 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
39 #define FDB_AGE_DEFAULT 300 /* 5 min */
40 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42 /* UDP port for VXLAN traffic.
43 * The IANA assigned port is 4789, but the Linux default is 8472
44 * for compatibility with early adopters.
46 static unsigned short vxlan_port __read_mostly = 8472;
47 module_param_named(udp_port, vxlan_port, ushort, 0444);
48 MODULE_PARM_DESC(udp_port, "Destination UDP port");
50 static bool log_ecn_error = true;
51 module_param(log_ecn_error, bool, 0644);
52 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54 static unsigned int vxlan_net_id;
55 static struct rtnl_link_ops vxlan_link_ops;
57 static const u8 all_zeros_mac[ETH_ALEN + 2];
59 static int vxlan_sock_add(struct vxlan_dev *vxlan);
61 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
63 /* per-network namespace private data for this module */
65 struct list_head vxlan_list;
66 struct hlist_head sock_list[PORT_HASH_SIZE];
70 /* Forwarding table entry */
72 struct hlist_node hlist; /* linked list of entries */
74 unsigned long updated; /* jiffies */
76 struct list_head remotes;
77 u8 eth_addr[ETH_ALEN];
78 u16 state; /* see ndm_state */
80 u16 flags; /* see ndm_flags and below */
83 #define NTF_VXLAN_ADDED_BY_USER 0x100
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly;
88 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
94 #if IS_ENABLED(CONFIG_IPV6)
96 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
98 if (a->sa.sa_family != b->sa.sa_family)
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
106 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
108 if (nla_len(nla) >= sizeof(struct in6_addr)) {
109 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
110 ip->sa.sa_family = AF_INET6;
112 } else if (nla_len(nla) >= sizeof(__be32)) {
113 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
114 ip->sa.sa_family = AF_INET;
117 return -EAFNOSUPPORT;
121 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
122 const union vxlan_addr *ip)
124 if (ip->sa.sa_family == AF_INET6)
125 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
127 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
130 #else /* !CONFIG_IPV6 */
133 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
135 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
138 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
141 return -EAFNOSUPPORT;
142 } else if (nla_len(nla) >= sizeof(__be32)) {
143 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
144 ip->sa.sa_family = AF_INET;
147 return -EAFNOSUPPORT;
151 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
152 const union vxlan_addr *ip)
154 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
158 /* Virtual Network hash table head */
159 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
161 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
164 /* Socket hash table head */
165 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
167 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
169 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
172 /* First remote destination for a forwarding entry.
173 * Guaranteed to be non-NULL because remotes are never deleted.
175 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
177 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
180 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
182 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
185 /* Find VXLAN socket based on network namespace, address family and UDP port
186 * and enabled unshareable flags.
188 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
189 __be16 port, u32 flags, int ifindex)
191 struct vxlan_sock *vs;
193 flags &= VXLAN_F_RCV_FLAGS;
195 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
196 if (inet_sk(vs->sock->sk)->inet_sport == port &&
197 vxlan_get_sk_family(vs) == family &&
198 vs->flags == flags &&
199 vs->sock->sk->sk_bound_dev_if == ifindex)
205 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
208 struct vxlan_dev_node *node;
210 /* For flow based devices, map all packets to VNI 0 */
211 if (vs->flags & VXLAN_F_COLLECT_METADATA)
214 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
215 if (node->vxlan->default_dst.remote_vni != vni)
218 if (IS_ENABLED(CONFIG_IPV6)) {
219 const struct vxlan_config *cfg = &node->vxlan->cfg;
221 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
222 cfg->remote_ifindex != ifindex)
232 /* Look up VNI in a per net namespace table */
233 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
234 __be32 vni, sa_family_t family,
235 __be16 port, u32 flags)
237 struct vxlan_sock *vs;
239 vs = vxlan_find_sock(net, family, port, flags, ifindex);
243 return vxlan_vs_find_vni(vs, ifindex, vni);
246 /* Fill in neighbour message in skbuff. */
247 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
248 const struct vxlan_fdb *fdb,
249 u32 portid, u32 seq, int type, unsigned int flags,
250 const struct vxlan_rdst *rdst)
252 unsigned long now = jiffies;
253 struct nda_cacheinfo ci;
254 struct nlmsghdr *nlh;
256 bool send_ip, send_eth;
258 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
262 ndm = nlmsg_data(nlh);
263 memset(ndm, 0, sizeof(*ndm));
265 send_eth = send_ip = true;
267 if (type == RTM_GETNEIGH) {
268 send_ip = !vxlan_addr_any(&rdst->remote_ip);
269 send_eth = !is_zero_ether_addr(fdb->eth_addr);
270 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
272 ndm->ndm_family = AF_BRIDGE;
273 ndm->ndm_state = fdb->state;
274 ndm->ndm_ifindex = vxlan->dev->ifindex;
275 ndm->ndm_flags = fdb->flags;
277 ndm->ndm_flags |= NTF_OFFLOADED;
278 ndm->ndm_type = RTN_UNICAST;
280 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
281 nla_put_s32(skb, NDA_LINK_NETNSID,
282 peernet2id(dev_net(vxlan->dev), vxlan->net)))
283 goto nla_put_failure;
285 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
286 goto nla_put_failure;
288 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
289 goto nla_put_failure;
291 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
292 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
293 goto nla_put_failure;
294 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
295 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
296 goto nla_put_failure;
297 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
298 nla_put_u32(skb, NDA_SRC_VNI,
299 be32_to_cpu(fdb->vni)))
300 goto nla_put_failure;
301 if (rdst->remote_ifindex &&
302 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
303 goto nla_put_failure;
305 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
306 ci.ndm_confirmed = 0;
307 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
310 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
311 goto nla_put_failure;
317 nlmsg_cancel(skb, nlh);
321 static inline size_t vxlan_nlmsg_size(void)
323 return NLMSG_ALIGN(sizeof(struct ndmsg))
324 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
325 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
326 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
327 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
328 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
329 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
330 + nla_total_size(sizeof(struct nda_cacheinfo));
333 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
334 struct vxlan_rdst *rd, int type)
336 struct net *net = dev_net(vxlan->dev);
340 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
344 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
346 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
347 WARN_ON(err == -EMSGSIZE);
352 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
356 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
359 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
360 const struct vxlan_fdb *fdb,
361 const struct vxlan_rdst *rd,
362 struct netlink_ext_ack *extack,
363 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
365 fdb_info->info.dev = vxlan->dev;
366 fdb_info->info.extack = extack;
367 fdb_info->remote_ip = rd->remote_ip;
368 fdb_info->remote_port = rd->remote_port;
369 fdb_info->remote_vni = rd->remote_vni;
370 fdb_info->remote_ifindex = rd->remote_ifindex;
371 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
372 fdb_info->vni = fdb->vni;
373 fdb_info->offloaded = rd->offloaded;
374 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
377 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
378 struct vxlan_fdb *fdb,
379 struct vxlan_rdst *rd,
381 struct netlink_ext_ack *extack)
383 struct switchdev_notifier_vxlan_fdb_info info;
384 enum switchdev_notifier_type notifier_type;
390 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
391 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
392 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
393 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
395 return notifier_to_errno(ret);
398 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
399 struct vxlan_rdst *rd, int type, bool swdev_notify,
400 struct netlink_ext_ack *extack)
407 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
413 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
419 __vxlan_fdb_notify(vxlan, fdb, rd, type);
423 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
425 struct vxlan_dev *vxlan = netdev_priv(dev);
426 struct vxlan_fdb f = {
429 struct vxlan_rdst remote = {
430 .remote_ip = *ipa, /* goes to NDA_DST */
431 .remote_vni = cpu_to_be32(VXLAN_N_VID),
434 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
437 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
439 struct vxlan_fdb f = {
442 struct vxlan_rdst remote = { };
444 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
446 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
449 /* Hash Ethernet address */
450 static u32 eth_hash(const unsigned char *addr)
452 u64 value = get_unaligned((u64 *)addr);
454 /* only want 6 bytes */
460 return hash_64(value, FDB_HASH_BITS);
463 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
465 /* use 1 byte of OUI and 3 bytes of NIC */
466 u32 key = get_unaligned((u32 *)(addr + 2));
468 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
471 static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
473 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
474 return eth_vni_hash(mac, vni);
476 return eth_hash(mac);
479 /* Hash chain to use given mac address */
480 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
481 const u8 *mac, __be32 vni)
483 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
486 /* Look up Ethernet address in forwarding table */
487 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
488 const u8 *mac, __be32 vni)
490 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
493 hlist_for_each_entry_rcu(f, head, hlist) {
494 if (ether_addr_equal(mac, f->eth_addr)) {
495 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
507 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
508 const u8 *mac, __be32 vni)
512 f = __vxlan_find_mac(vxlan, mac, vni);
513 if (f && f->used != jiffies)
519 /* caller should hold vxlan->hash_lock */
520 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
521 union vxlan_addr *ip, __be16 port,
522 __be32 vni, __u32 ifindex)
524 struct vxlan_rdst *rd;
526 list_for_each_entry(rd, &f->remotes, list) {
527 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
528 rd->remote_port == port &&
529 rd->remote_vni == vni &&
530 rd->remote_ifindex == ifindex)
537 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
538 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
540 struct vxlan_dev *vxlan = netdev_priv(dev);
541 u8 eth_addr[ETH_ALEN + 2] = { 0 };
542 struct vxlan_rdst *rdst;
546 if (is_multicast_ether_addr(mac) ||
547 is_zero_ether_addr(mac))
550 ether_addr_copy(eth_addr, mac);
554 f = __vxlan_find_mac(vxlan, eth_addr, vni);
560 rdst = first_remote_rcu(f);
561 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
567 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
569 static int vxlan_fdb_notify_one(struct notifier_block *nb,
570 const struct vxlan_dev *vxlan,
571 const struct vxlan_fdb *f,
572 const struct vxlan_rdst *rdst,
573 struct netlink_ext_ack *extack)
575 struct switchdev_notifier_vxlan_fdb_info fdb_info;
578 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
579 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
581 return notifier_to_errno(rc);
584 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
585 struct notifier_block *nb,
586 struct netlink_ext_ack *extack)
588 struct vxlan_dev *vxlan;
589 struct vxlan_rdst *rdst;
594 if (!netif_is_vxlan(dev))
596 vxlan = netdev_priv(dev);
598 for (h = 0; h < FDB_HASH_SIZE; ++h) {
599 spin_lock_bh(&vxlan->hash_lock[h]);
600 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
602 list_for_each_entry(rdst, &f->remotes, list) {
603 rc = vxlan_fdb_notify_one(nb, vxlan,
611 spin_unlock_bh(&vxlan->hash_lock[h]);
616 spin_unlock_bh(&vxlan->hash_lock[h]);
619 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
621 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
623 struct vxlan_dev *vxlan;
624 struct vxlan_rdst *rdst;
628 if (!netif_is_vxlan(dev))
630 vxlan = netdev_priv(dev);
632 for (h = 0; h < FDB_HASH_SIZE; ++h) {
633 spin_lock_bh(&vxlan->hash_lock[h]);
634 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
636 list_for_each_entry(rdst, &f->remotes, list)
637 rdst->offloaded = false;
638 spin_unlock_bh(&vxlan->hash_lock[h]);
642 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
644 /* Replace destination of unicast mac */
645 static int vxlan_fdb_replace(struct vxlan_fdb *f,
646 union vxlan_addr *ip, __be16 port, __be32 vni,
647 __u32 ifindex, struct vxlan_rdst *oldrd)
649 struct vxlan_rdst *rd;
651 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
655 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
660 dst_cache_reset(&rd->dst_cache);
662 rd->remote_port = port;
663 rd->remote_vni = vni;
664 rd->remote_ifindex = ifindex;
665 rd->offloaded = false;
669 /* Add/update destinations for multicast */
670 static int vxlan_fdb_append(struct vxlan_fdb *f,
671 union vxlan_addr *ip, __be16 port, __be32 vni,
672 __u32 ifindex, struct vxlan_rdst **rdp)
674 struct vxlan_rdst *rd;
676 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
680 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
684 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
690 rd->remote_port = port;
691 rd->offloaded = false;
692 rd->remote_vni = vni;
693 rd->remote_ifindex = ifindex;
695 list_add_tail_rcu(&rd->list, &f->remotes);
701 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
703 struct vxlanhdr *vh, size_t hdrlen,
705 struct gro_remcsum *grc,
708 size_t start, offset;
710 if (skb->remcsum_offload)
713 if (!NAPI_GRO_CB(skb)->csum_valid)
716 start = vxlan_rco_start(vni_field);
717 offset = start + vxlan_rco_offset(vni_field);
719 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
720 start, offset, grc, nopartial);
722 skb->remcsum_offload = 1;
727 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
728 struct list_head *head,
731 struct sk_buff *pp = NULL;
733 struct vxlanhdr *vh, *vh2;
734 unsigned int hlen, off_vx;
736 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
738 struct gro_remcsum grc;
740 skb_gro_remcsum_init(&grc);
742 off_vx = skb_gro_offset(skb);
743 hlen = off_vx + sizeof(*vh);
744 vh = skb_gro_header_fast(skb, off_vx);
745 if (skb_gro_header_hard(skb, hlen)) {
746 vh = skb_gro_header_slow(skb, hlen, off_vx);
751 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
753 flags = vh->vx_flags;
755 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
756 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
759 VXLAN_F_REMCSUM_NOPARTIAL));
765 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
767 list_for_each_entry(p, head, list) {
768 if (!NAPI_GRO_CB(p)->same_flow)
771 vh2 = (struct vxlanhdr *)(p->data + off_vx);
772 if (vh->vx_flags != vh2->vx_flags ||
773 vh->vx_vni != vh2->vx_vni) {
774 NAPI_GRO_CB(p)->same_flow = 0;
779 pp = call_gro_receive(eth_gro_receive, head, skb);
783 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
788 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
790 /* Sets 'skb->inner_mac_header' since we are always called with
791 * 'skb->encapsulation' set.
793 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
796 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
797 const u8 *mac, __u16 state,
798 __be32 src_vni, __u16 ndm_flags)
802 f = kmalloc(sizeof(*f), GFP_ATOMIC);
806 f->flags = ndm_flags;
807 f->updated = f->used = jiffies;
809 INIT_LIST_HEAD(&f->remotes);
810 memcpy(f->eth_addr, mac, ETH_ALEN);
815 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
816 __be32 src_vni, struct vxlan_fdb *f)
819 hlist_add_head_rcu(&f->hlist,
820 vxlan_fdb_head(vxlan, mac, src_vni));
823 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
824 const u8 *mac, union vxlan_addr *ip,
825 __u16 state, __be16 port, __be32 src_vni,
826 __be32 vni, __u32 ifindex, __u16 ndm_flags,
827 struct vxlan_fdb **fdb)
829 struct vxlan_rdst *rd = NULL;
833 if (vxlan->cfg.addrmax &&
834 vxlan->addrcnt >= vxlan->cfg.addrmax)
837 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
838 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
842 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
853 static void __vxlan_fdb_free(struct vxlan_fdb *f)
855 struct vxlan_rdst *rd, *nd;
857 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
858 dst_cache_destroy(&rd->dst_cache);
864 static void vxlan_fdb_free(struct rcu_head *head)
866 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
871 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
872 bool do_notify, bool swdev_notify)
874 struct vxlan_rdst *rd;
876 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
880 list_for_each_entry(rd, &f->remotes, list)
881 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
884 hlist_del_rcu(&f->hlist);
885 call_rcu(&f->rcu, vxlan_fdb_free);
888 static void vxlan_dst_free(struct rcu_head *head)
890 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
892 dst_cache_destroy(&rd->dst_cache);
896 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
897 union vxlan_addr *ip,
898 __u16 state, __u16 flags,
899 __be16 port, __be32 vni,
900 __u32 ifindex, __u16 ndm_flags,
903 struct netlink_ext_ack *extack)
905 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
906 struct vxlan_rdst *rd = NULL;
907 struct vxlan_rdst oldrd;
912 /* Do not allow an externally learned entry to take over an entry added
915 if (!(fdb_flags & NTF_EXT_LEARNED) ||
916 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
917 if (f->state != state) {
919 f->updated = jiffies;
922 if (f->flags != fdb_flags) {
923 f->flags = fdb_flags;
924 f->updated = jiffies;
929 if ((flags & NLM_F_REPLACE)) {
930 /* Only change unicasts */
931 if (!(is_multicast_ether_addr(f->eth_addr) ||
932 is_zero_ether_addr(f->eth_addr))) {
933 rc = vxlan_fdb_replace(f, ip, port, vni,
940 if ((flags & NLM_F_APPEND) &&
941 (is_multicast_ether_addr(f->eth_addr) ||
942 is_zero_ether_addr(f->eth_addr))) {
943 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
950 if (ndm_flags & NTF_USE)
955 rd = first_remote_rtnl(f);
957 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
958 swdev_notify, extack);
966 if ((flags & NLM_F_REPLACE) && rc)
968 else if ((flags & NLM_F_APPEND) && rc) {
969 list_del_rcu(&rd->list);
970 call_rcu(&rd->rcu, vxlan_dst_free);
975 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
976 const u8 *mac, union vxlan_addr *ip,
977 __u16 state, __u16 flags,
978 __be16 port, __be32 src_vni, __be32 vni,
979 __u32 ifindex, __u16 ndm_flags,
981 struct netlink_ext_ack *extack)
983 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
987 /* Disallow replace to add a multicast entry */
988 if ((flags & NLM_F_REPLACE) &&
989 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
992 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
993 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
994 vni, ifindex, fdb_flags, &f);
998 vxlan_fdb_insert(vxlan, mac, src_vni, f);
999 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1000 swdev_notify, extack);
1007 vxlan_fdb_destroy(vxlan, f, false, false);
1011 /* Add new entry to forwarding table -- assumes lock held */
1012 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1013 const u8 *mac, union vxlan_addr *ip,
1014 __u16 state, __u16 flags,
1015 __be16 port, __be32 src_vni, __be32 vni,
1016 __u32 ifindex, __u16 ndm_flags,
1018 struct netlink_ext_ack *extack)
1020 struct vxlan_fdb *f;
1022 f = __vxlan_find_mac(vxlan, mac, src_vni);
1024 if (flags & NLM_F_EXCL) {
1025 netdev_dbg(vxlan->dev,
1026 "lost race to create %pM\n", mac);
1030 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1031 vni, ifindex, ndm_flags, f,
1032 swdev_notify, extack);
1034 if (!(flags & NLM_F_CREATE))
1037 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1038 port, src_vni, vni, ifindex,
1039 ndm_flags, swdev_notify, extack);
1043 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1044 struct vxlan_rdst *rd, bool swdev_notify)
1046 list_del_rcu(&rd->list);
1047 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1048 call_rcu(&rd->rcu, vxlan_dst_free);
1051 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1052 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1053 __be32 *vni, u32 *ifindex)
1055 struct net *net = dev_net(vxlan->dev);
1059 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1063 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1064 if (remote->sa.sa_family == AF_INET) {
1065 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1066 ip->sa.sa_family = AF_INET;
1067 #if IS_ENABLED(CONFIG_IPV6)
1069 ip->sin6.sin6_addr = in6addr_any;
1070 ip->sa.sa_family = AF_INET6;
1076 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1078 *port = nla_get_be16(tb[NDA_PORT]);
1080 *port = vxlan->cfg.dst_port;
1084 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1086 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1088 *vni = vxlan->default_dst.remote_vni;
1091 if (tb[NDA_SRC_VNI]) {
1092 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1094 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1096 *src_vni = vxlan->default_dst.remote_vni;
1099 if (tb[NDA_IFINDEX]) {
1100 struct net_device *tdev;
1102 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1104 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1105 tdev = __dev_get_by_index(net, *ifindex);
1107 return -EADDRNOTAVAIL;
1115 /* Add static entry (via netlink) */
1116 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1117 struct net_device *dev,
1118 const unsigned char *addr, u16 vid, u16 flags,
1119 struct netlink_ext_ack *extack)
1121 struct vxlan_dev *vxlan = netdev_priv(dev);
1122 /* struct net *net = dev_net(vxlan->dev); */
1123 union vxlan_addr ip;
1125 __be32 src_vni, vni;
1130 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1131 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1136 if (tb[NDA_DST] == NULL)
1139 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1143 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1144 return -EAFNOSUPPORT;
1146 hash_index = fdb_head_index(vxlan, addr, src_vni);
1147 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1148 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1149 port, src_vni, vni, ifindex,
1150 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1152 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1157 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1158 const unsigned char *addr, union vxlan_addr ip,
1159 __be16 port, __be32 src_vni, __be32 vni,
1160 u32 ifindex, bool swdev_notify)
1162 struct vxlan_fdb *f;
1163 struct vxlan_rdst *rd = NULL;
1166 f = vxlan_find_mac(vxlan, addr, src_vni);
1170 if (!vxlan_addr_any(&ip)) {
1171 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1176 /* remove a destination if it's not the only one on the list,
1177 * otherwise destroy the fdb entry
1179 if (rd && !list_is_singular(&f->remotes)) {
1180 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1184 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1190 /* Delete entry (via netlink) */
1191 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1192 struct net_device *dev,
1193 const unsigned char *addr, u16 vid)
1195 struct vxlan_dev *vxlan = netdev_priv(dev);
1196 union vxlan_addr ip;
1197 __be32 src_vni, vni;
1203 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1207 hash_index = fdb_head_index(vxlan, addr, src_vni);
1208 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1209 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1211 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1216 /* Dump forwarding table */
1217 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1218 struct net_device *dev,
1219 struct net_device *filter_dev, int *idx)
1221 struct vxlan_dev *vxlan = netdev_priv(dev);
1225 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1226 struct vxlan_fdb *f;
1229 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1230 struct vxlan_rdst *rd;
1232 list_for_each_entry_rcu(rd, &f->remotes, list) {
1233 if (*idx < cb->args[2])
1236 err = vxlan_fdb_info(skb, vxlan, f,
1237 NETLINK_CB(cb->skb).portid,
1255 static int vxlan_fdb_get(struct sk_buff *skb,
1256 struct nlattr *tb[],
1257 struct net_device *dev,
1258 const unsigned char *addr,
1259 u16 vid, u32 portid, u32 seq,
1260 struct netlink_ext_ack *extack)
1262 struct vxlan_dev *vxlan = netdev_priv(dev);
1263 struct vxlan_fdb *f;
1268 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1270 vni = vxlan->default_dst.remote_vni;
1274 f = __vxlan_find_mac(vxlan, addr, vni);
1276 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1281 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1282 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1288 /* Watch incoming packets to learn mapping between Ethernet address
1289 * and Tunnel endpoint.
1290 * Return true if packet is bogus and should be dropped.
1292 static bool vxlan_snoop(struct net_device *dev,
1293 union vxlan_addr *src_ip, const u8 *src_mac,
1294 u32 src_ifindex, __be32 vni)
1296 struct vxlan_dev *vxlan = netdev_priv(dev);
1297 struct vxlan_fdb *f;
1300 #if IS_ENABLED(CONFIG_IPV6)
1301 if (src_ip->sa.sa_family == AF_INET6 &&
1302 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1303 ifindex = src_ifindex;
1306 f = vxlan_find_mac(vxlan, src_mac, vni);
1308 struct vxlan_rdst *rdst = first_remote_rcu(f);
1310 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1311 rdst->remote_ifindex == ifindex))
1314 /* Don't migrate static entries, drop packets */
1315 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1318 if (net_ratelimit())
1320 "%pM migrated from %pIS to %pIS\n",
1321 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1323 rdst->remote_ip = *src_ip;
1324 f->updated = jiffies;
1325 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1327 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1329 /* learned new entry */
1330 spin_lock(&vxlan->hash_lock[hash_index]);
1332 /* close off race between vxlan_flush and incoming packets */
1333 if (netif_running(dev))
1334 vxlan_fdb_update(vxlan, src_mac, src_ip,
1336 NLM_F_EXCL|NLM_F_CREATE,
1337 vxlan->cfg.dst_port,
1339 vxlan->default_dst.remote_vni,
1340 ifindex, NTF_SELF, true, NULL);
1341 spin_unlock(&vxlan->hash_lock[hash_index]);
1347 /* See if multicast group is already in use by other ID */
1348 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1350 struct vxlan_dev *vxlan;
1351 struct vxlan_sock *sock4;
1352 #if IS_ENABLED(CONFIG_IPV6)
1353 struct vxlan_sock *sock6;
1355 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1357 sock4 = rtnl_dereference(dev->vn4_sock);
1359 /* The vxlan_sock is only used by dev, leaving group has
1360 * no effect on other vxlan devices.
1362 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1364 #if IS_ENABLED(CONFIG_IPV6)
1365 sock6 = rtnl_dereference(dev->vn6_sock);
1366 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1370 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1371 if (!netif_running(vxlan->dev) || vxlan == dev)
1374 if (family == AF_INET &&
1375 rtnl_dereference(vxlan->vn4_sock) != sock4)
1377 #if IS_ENABLED(CONFIG_IPV6)
1378 if (family == AF_INET6 &&
1379 rtnl_dereference(vxlan->vn6_sock) != sock6)
1383 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1384 &dev->default_dst.remote_ip))
1387 if (vxlan->default_dst.remote_ifindex !=
1388 dev->default_dst.remote_ifindex)
1397 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1399 struct vxlan_net *vn;
1403 if (!refcount_dec_and_test(&vs->refcnt))
1406 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1407 spin_lock(&vn->sock_lock);
1408 hlist_del_rcu(&vs->hlist);
1409 udp_tunnel_notify_del_rx_port(vs->sock,
1410 (vs->flags & VXLAN_F_GPE) ?
1411 UDP_TUNNEL_TYPE_VXLAN_GPE :
1412 UDP_TUNNEL_TYPE_VXLAN);
1413 spin_unlock(&vn->sock_lock);
1418 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1420 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1421 #if IS_ENABLED(CONFIG_IPV6)
1422 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1424 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1427 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1430 vxlan_vs_del_dev(vxlan);
1432 if (__vxlan_sock_release_prep(sock4)) {
1433 udp_tunnel_sock_release(sock4->sock);
1437 #if IS_ENABLED(CONFIG_IPV6)
1438 if (__vxlan_sock_release_prep(sock6)) {
1439 udp_tunnel_sock_release(sock6->sock);
1445 /* Update multicast group membership when first VNI on
1446 * multicast address is brought up
1448 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1451 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1452 int ifindex = vxlan->default_dst.remote_ifindex;
1455 if (ip->sa.sa_family == AF_INET) {
1456 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1457 struct ip_mreqn mreq = {
1458 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1459 .imr_ifindex = ifindex,
1462 sk = sock4->sock->sk;
1464 ret = ip_mc_join_group(sk, &mreq);
1466 #if IS_ENABLED(CONFIG_IPV6)
1468 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1470 sk = sock6->sock->sk;
1472 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1473 &ip->sin6.sin6_addr);
1481 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1482 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1485 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1486 int ifindex = vxlan->default_dst.remote_ifindex;
1489 if (ip->sa.sa_family == AF_INET) {
1490 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1491 struct ip_mreqn mreq = {
1492 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1493 .imr_ifindex = ifindex,
1496 sk = sock4->sock->sk;
1498 ret = ip_mc_leave_group(sk, &mreq);
1500 #if IS_ENABLED(CONFIG_IPV6)
1502 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1504 sk = sock6->sock->sk;
1506 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1507 &ip->sin6.sin6_addr);
1515 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1516 struct sk_buff *skb, u32 vxflags)
1518 size_t start, offset;
1520 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1523 start = vxlan_rco_start(unparsed->vx_vni);
1524 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1526 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1529 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1530 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1532 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1533 unparsed->vx_vni &= VXLAN_VNI_MASK;
1537 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1538 struct sk_buff *skb, u32 vxflags,
1539 struct vxlan_metadata *md)
1541 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1542 struct metadata_dst *tun_dst;
1544 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1547 md->gbp = ntohs(gbp->policy_id);
1549 tun_dst = (struct metadata_dst *)skb_dst(skb);
1551 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1552 tun_dst->u.tun_info.options_len = sizeof(*md);
1554 if (gbp->dont_learn)
1555 md->gbp |= VXLAN_GBP_DONT_LEARN;
1557 if (gbp->policy_applied)
1558 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1560 /* In flow-based mode, GBP is carried in dst_metadata */
1561 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1562 skb->mark = md->gbp;
1564 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1567 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1569 struct sk_buff *skb, u32 vxflags)
1571 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1573 /* Need to have Next Protocol set for interfaces in GPE mode. */
1574 if (!gpe->np_applied)
1576 /* "The initial version is 0. If a receiver does not support the
1577 * version indicated it MUST drop the packet.
1579 if (gpe->version != 0)
1581 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1582 * processing MUST occur." However, we don't implement OAM
1583 * processing, thus drop the packet.
1588 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1592 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1596 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1597 struct vxlan_sock *vs,
1598 struct sk_buff *skb, __be32 vni)
1600 union vxlan_addr saddr;
1601 u32 ifindex = skb->dev->ifindex;
1603 skb_reset_mac_header(skb);
1604 skb->protocol = eth_type_trans(skb, vxlan->dev);
1605 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1607 /* Ignore packet loops (and multicast echo) */
1608 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1611 /* Get address from the outer IP header */
1612 if (vxlan_get_sk_family(vs) == AF_INET) {
1613 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1614 saddr.sa.sa_family = AF_INET;
1615 #if IS_ENABLED(CONFIG_IPV6)
1617 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1618 saddr.sa.sa_family = AF_INET6;
1622 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1623 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1629 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1630 struct sk_buff *skb)
1634 if (vxlan_get_sk_family(vs) == AF_INET)
1635 err = IP_ECN_decapsulate(oiph, skb);
1636 #if IS_ENABLED(CONFIG_IPV6)
1638 err = IP6_ECN_decapsulate(oiph, skb);
1641 if (unlikely(err) && log_ecn_error) {
1642 if (vxlan_get_sk_family(vs) == AF_INET)
1643 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1644 &((struct iphdr *)oiph)->saddr,
1645 ((struct iphdr *)oiph)->tos);
1647 net_info_ratelimited("non-ECT from %pI6\n",
1648 &((struct ipv6hdr *)oiph)->saddr);
1653 /* Callback from net/ipv4/udp.c to receive packets */
1654 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1656 struct pcpu_sw_netstats *stats;
1657 struct vxlan_dev *vxlan;
1658 struct vxlan_sock *vs;
1659 struct vxlanhdr unparsed;
1660 struct vxlan_metadata _md;
1661 struct vxlan_metadata *md = &_md;
1662 __be16 protocol = htons(ETH_P_TEB);
1663 bool raw_proto = false;
1667 /* Need UDP and VXLAN header to be present */
1668 if (!pskb_may_pull(skb, VXLAN_HLEN))
1671 unparsed = *vxlan_hdr(skb);
1672 /* VNI flag always required to be set */
1673 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1674 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1675 ntohl(vxlan_hdr(skb)->vx_flags),
1676 ntohl(vxlan_hdr(skb)->vx_vni));
1677 /* Return non vxlan pkt */
1680 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1681 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1683 vs = rcu_dereference_sk_user_data(sk);
1687 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1689 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1693 /* For backwards compatibility, only allow reserved fields to be
1694 * used by VXLAN extensions if explicitly requested.
1696 if (vs->flags & VXLAN_F_GPE) {
1697 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1702 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1703 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1706 if (vxlan_collect_metadata(vs)) {
1707 struct metadata_dst *tun_dst;
1709 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1710 key32_to_tunnel_id(vni), sizeof(*md));
1715 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1717 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1719 memset(md, 0, sizeof(*md));
1722 if (vs->flags & VXLAN_F_REMCSUM_RX)
1723 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1725 if (vs->flags & VXLAN_F_GBP)
1726 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1727 /* Note that GBP and GPE can never be active together. This is
1728 * ensured in vxlan_dev_configure.
1731 if (unparsed.vx_flags || unparsed.vx_vni) {
1732 /* If there are any unprocessed flags remaining treat
1733 * this as a malformed packet. This behavior diverges from
1734 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1735 * in reserved fields are to be ignored. The approach here
1736 * maintains compatibility with previous stack code, and also
1737 * is more robust and provides a little more security in
1738 * adding extensions to VXLAN.
1744 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1747 skb_reset_mac_header(skb);
1748 skb->dev = vxlan->dev;
1749 skb->pkt_type = PACKET_HOST;
1752 oiph = skb_network_header(skb);
1753 skb_reset_network_header(skb);
1755 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1756 ++vxlan->dev->stats.rx_frame_errors;
1757 ++vxlan->dev->stats.rx_errors;
1763 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1765 atomic_long_inc(&vxlan->dev->rx_dropped);
1769 stats = this_cpu_ptr(vxlan->dev->tstats);
1770 u64_stats_update_begin(&stats->syncp);
1771 stats->rx_packets++;
1772 stats->rx_bytes += skb->len;
1773 u64_stats_update_end(&stats->syncp);
1775 gro_cells_receive(&vxlan->gro_cells, skb);
1782 /* Consume bad packet */
1787 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1788 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1790 struct vxlan_dev *vxlan;
1791 struct vxlan_sock *vs;
1792 struct vxlanhdr *hdr;
1795 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1798 hdr = vxlan_hdr(skb);
1800 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1803 vs = rcu_dereference_sk_user_data(sk);
1807 vni = vxlan_vni(hdr->vx_vni);
1808 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1815 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1817 struct vxlan_dev *vxlan = netdev_priv(dev);
1818 struct arphdr *parp;
1821 struct neighbour *n;
1823 if (dev->flags & IFF_NOARP)
1826 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1827 dev->stats.tx_dropped++;
1830 parp = arp_hdr(skb);
1832 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1833 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1834 parp->ar_pro != htons(ETH_P_IP) ||
1835 parp->ar_op != htons(ARPOP_REQUEST) ||
1836 parp->ar_hln != dev->addr_len ||
1839 arpptr = (u8 *)parp + sizeof(struct arphdr);
1841 arpptr += dev->addr_len; /* sha */
1842 memcpy(&sip, arpptr, sizeof(sip));
1843 arpptr += sizeof(sip);
1844 arpptr += dev->addr_len; /* tha */
1845 memcpy(&tip, arpptr, sizeof(tip));
1847 if (ipv4_is_loopback(tip) ||
1848 ipv4_is_multicast(tip))
1851 n = neigh_lookup(&arp_tbl, &tip, dev);
1854 struct vxlan_fdb *f;
1855 struct sk_buff *reply;
1857 if (!(n->nud_state & NUD_CONNECTED)) {
1862 f = vxlan_find_mac(vxlan, n->ha, vni);
1863 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1864 /* bridge-local neighbor */
1869 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1877 skb_reset_mac_header(reply);
1878 __skb_pull(reply, skb_network_offset(reply));
1879 reply->ip_summed = CHECKSUM_UNNECESSARY;
1880 reply->pkt_type = PACKET_HOST;
1882 if (netif_rx_ni(reply) == NET_RX_DROP)
1883 dev->stats.rx_dropped++;
1884 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1885 union vxlan_addr ipa = {
1886 .sin.sin_addr.s_addr = tip,
1887 .sin.sin_family = AF_INET,
1890 vxlan_ip_miss(dev, &ipa);
1894 return NETDEV_TX_OK;
1897 #if IS_ENABLED(CONFIG_IPV6)
1898 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1899 struct neighbour *n, bool isrouter)
1901 struct net_device *dev = request->dev;
1902 struct sk_buff *reply;
1903 struct nd_msg *ns, *na;
1904 struct ipv6hdr *pip6;
1906 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1910 if (dev == NULL || !pskb_may_pull(request, request->len))
1913 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1914 sizeof(*na) + na_olen + dev->needed_tailroom;
1915 reply = alloc_skb(len, GFP_ATOMIC);
1919 reply->protocol = htons(ETH_P_IPV6);
1921 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1922 skb_push(reply, sizeof(struct ethhdr));
1923 skb_reset_mac_header(reply);
1925 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1927 daddr = eth_hdr(request)->h_source;
1928 ns_olen = request->len - skb_network_offset(request) -
1929 sizeof(struct ipv6hdr) - sizeof(*ns);
1930 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1931 if (!ns->opt[i + 1]) {
1935 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1936 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1941 /* Ethernet header */
1942 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1943 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1944 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1945 reply->protocol = htons(ETH_P_IPV6);
1947 skb_pull(reply, sizeof(struct ethhdr));
1948 skb_reset_network_header(reply);
1949 skb_put(reply, sizeof(struct ipv6hdr));
1953 pip6 = ipv6_hdr(reply);
1954 memset(pip6, 0, sizeof(struct ipv6hdr));
1956 pip6->priority = ipv6_hdr(request)->priority;
1957 pip6->nexthdr = IPPROTO_ICMPV6;
1958 pip6->hop_limit = 255;
1959 pip6->daddr = ipv6_hdr(request)->saddr;
1960 pip6->saddr = *(struct in6_addr *)n->primary_key;
1962 skb_pull(reply, sizeof(struct ipv6hdr));
1963 skb_reset_transport_header(reply);
1965 /* Neighbor Advertisement */
1966 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1967 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1968 na->icmph.icmp6_router = isrouter;
1969 na->icmph.icmp6_override = 1;
1970 na->icmph.icmp6_solicited = 1;
1971 na->target = ns->target;
1972 ether_addr_copy(&na->opt[2], n->ha);
1973 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1974 na->opt[1] = na_olen >> 3;
1976 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1977 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1978 csum_partial(na, sizeof(*na)+na_olen, 0));
1980 pip6->payload_len = htons(sizeof(*na)+na_olen);
1982 skb_push(reply, sizeof(struct ipv6hdr));
1984 reply->ip_summed = CHECKSUM_UNNECESSARY;
1989 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1991 struct vxlan_dev *vxlan = netdev_priv(dev);
1992 const struct in6_addr *daddr;
1993 const struct ipv6hdr *iphdr;
1994 struct inet6_dev *in6_dev;
1995 struct neighbour *n;
1999 in6_dev = __in6_dev_get(dev);
2003 iphdr = ipv6_hdr(skb);
2004 daddr = &iphdr->daddr;
2005 msg = (struct nd_msg *)(iphdr + 1);
2007 if (ipv6_addr_loopback(daddr) ||
2008 ipv6_addr_is_multicast(&msg->target))
2011 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2014 struct vxlan_fdb *f;
2015 struct sk_buff *reply;
2017 if (!(n->nud_state & NUD_CONNECTED)) {
2022 f = vxlan_find_mac(vxlan, n->ha, vni);
2023 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2024 /* bridge-local neighbor */
2029 reply = vxlan_na_create(skb, n,
2030 !!(f ? f->flags & NTF_ROUTER : 0));
2037 if (netif_rx_ni(reply) == NET_RX_DROP)
2038 dev->stats.rx_dropped++;
2040 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2041 union vxlan_addr ipa = {
2042 .sin6.sin6_addr = msg->target,
2043 .sin6.sin6_family = AF_INET6,
2046 vxlan_ip_miss(dev, &ipa);
2052 return NETDEV_TX_OK;
2056 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2058 struct vxlan_dev *vxlan = netdev_priv(dev);
2059 struct neighbour *n;
2061 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2065 switch (ntohs(eth_hdr(skb)->h_proto)) {
2070 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2073 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2074 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2075 union vxlan_addr ipa = {
2076 .sin.sin_addr.s_addr = pip->daddr,
2077 .sin.sin_family = AF_INET,
2080 vxlan_ip_miss(dev, &ipa);
2086 #if IS_ENABLED(CONFIG_IPV6)
2089 struct ipv6hdr *pip6;
2091 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2093 pip6 = ipv6_hdr(skb);
2094 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2095 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2096 union vxlan_addr ipa = {
2097 .sin6.sin6_addr = pip6->daddr,
2098 .sin6.sin6_family = AF_INET6,
2101 vxlan_ip_miss(dev, &ipa);
2115 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2117 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2119 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2128 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2129 struct vxlan_metadata *md)
2131 struct vxlanhdr_gbp *gbp;
2136 gbp = (struct vxlanhdr_gbp *)vxh;
2137 vxh->vx_flags |= VXLAN_HF_GBP;
2139 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2140 gbp->dont_learn = 1;
2142 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2143 gbp->policy_applied = 1;
2145 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2148 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2151 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2153 gpe->np_applied = 1;
2154 gpe->next_protocol = tun_p_from_eth_p(protocol);
2155 if (!gpe->next_protocol)
2156 return -EPFNOSUPPORT;
2160 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2161 int iphdr_len, __be32 vni,
2162 struct vxlan_metadata *md, u32 vxflags,
2165 struct vxlanhdr *vxh;
2168 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2169 __be16 inner_protocol = htons(ETH_P_TEB);
2171 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2172 skb->ip_summed == CHECKSUM_PARTIAL) {
2173 int csum_start = skb_checksum_start_offset(skb);
2175 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2176 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2177 (skb->csum_offset == offsetof(struct udphdr, check) ||
2178 skb->csum_offset == offsetof(struct tcphdr, check)))
2179 type |= SKB_GSO_TUNNEL_REMCSUM;
2182 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2183 + VXLAN_HLEN + iphdr_len;
2185 /* Need space for new headers (invalidates iph ptr) */
2186 err = skb_cow_head(skb, min_headroom);
2190 err = iptunnel_handle_offloads(skb, type);
2194 vxh = __skb_push(skb, sizeof(*vxh));
2195 vxh->vx_flags = VXLAN_HF_VNI;
2196 vxh->vx_vni = vxlan_vni_field(vni);
2198 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2201 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2202 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2203 vxh->vx_flags |= VXLAN_HF_RCO;
2205 if (!skb_is_gso(skb)) {
2206 skb->ip_summed = CHECKSUM_NONE;
2207 skb->encapsulation = 0;
2211 if (vxflags & VXLAN_F_GBP)
2212 vxlan_build_gbp_hdr(vxh, vxflags, md);
2213 if (vxflags & VXLAN_F_GPE) {
2214 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2217 inner_protocol = skb->protocol;
2220 skb_set_inner_protocol(skb, inner_protocol);
2224 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2225 struct vxlan_sock *sock4,
2226 struct sk_buff *skb, int oif, u8 tos,
2227 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2228 struct dst_cache *dst_cache,
2229 const struct ip_tunnel_info *info)
2231 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2232 struct rtable *rt = NULL;
2236 return ERR_PTR(-EIO);
2241 rt = dst_cache_get_ip4(dst_cache, saddr);
2246 memset(&fl4, 0, sizeof(fl4));
2247 fl4.flowi4_oif = oif;
2248 fl4.flowi4_tos = RT_TOS(tos);
2249 fl4.flowi4_mark = skb->mark;
2250 fl4.flowi4_proto = IPPROTO_UDP;
2253 fl4.fl4_dport = dport;
2254 fl4.fl4_sport = sport;
2256 rt = ip_route_output_key(vxlan->net, &fl4);
2258 if (rt->dst.dev == dev) {
2259 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2261 return ERR_PTR(-ELOOP);
2266 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2268 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2269 return ERR_PTR(-ENETUNREACH);
2274 #if IS_ENABLED(CONFIG_IPV6)
2275 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2276 struct net_device *dev,
2277 struct vxlan_sock *sock6,
2278 struct sk_buff *skb, int oif, u8 tos,
2280 const struct in6_addr *daddr,
2281 struct in6_addr *saddr,
2282 __be16 dport, __be16 sport,
2283 struct dst_cache *dst_cache,
2284 const struct ip_tunnel_info *info)
2286 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2287 struct dst_entry *ndst;
2291 return ERR_PTR(-EIO);
2296 ndst = dst_cache_get_ip6(dst_cache, saddr);
2301 memset(&fl6, 0, sizeof(fl6));
2302 fl6.flowi6_oif = oif;
2305 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2306 fl6.flowi6_mark = skb->mark;
2307 fl6.flowi6_proto = IPPROTO_UDP;
2308 fl6.fl6_dport = dport;
2309 fl6.fl6_sport = sport;
2311 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2313 if (unlikely(IS_ERR(ndst))) {
2314 netdev_dbg(dev, "no route to %pI6\n", daddr);
2315 return ERR_PTR(-ENETUNREACH);
2318 if (unlikely(ndst->dev == dev)) {
2319 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2321 return ERR_PTR(-ELOOP);
2326 dst_cache_set_ip6(dst_cache, ndst, saddr);
2331 /* Bypass encapsulation if the destination is local */
2332 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2333 struct vxlan_dev *dst_vxlan, __be32 vni)
2335 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2336 union vxlan_addr loopback;
2337 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2338 struct net_device *dev;
2341 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2342 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2343 skb->pkt_type = PACKET_HOST;
2344 skb->encapsulation = 0;
2345 skb->dev = dst_vxlan->dev;
2346 __skb_pull(skb, skb_network_offset(skb));
2348 if (remote_ip->sa.sa_family == AF_INET) {
2349 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2350 loopback.sa.sa_family = AF_INET;
2351 #if IS_ENABLED(CONFIG_IPV6)
2353 loopback.sin6.sin6_addr = in6addr_loopback;
2354 loopback.sa.sa_family = AF_INET6;
2360 if (unlikely(!(dev->flags & IFF_UP))) {
2365 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2366 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2368 u64_stats_update_begin(&tx_stats->syncp);
2369 tx_stats->tx_packets++;
2370 tx_stats->tx_bytes += len;
2371 u64_stats_update_end(&tx_stats->syncp);
2373 if (netif_rx(skb) == NET_RX_SUCCESS) {
2374 u64_stats_update_begin(&rx_stats->syncp);
2375 rx_stats->rx_packets++;
2376 rx_stats->rx_bytes += len;
2377 u64_stats_update_end(&rx_stats->syncp);
2380 dev->stats.rx_dropped++;
2385 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2386 struct vxlan_dev *vxlan,
2387 union vxlan_addr *daddr,
2388 __be16 dst_port, int dst_ifindex, __be32 vni,
2389 struct dst_entry *dst,
2392 #if IS_ENABLED(CONFIG_IPV6)
2393 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2394 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2395 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2397 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2399 /* Bypass encapsulation if the destination is local */
2400 if (rt_flags & RTCF_LOCAL &&
2401 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2402 struct vxlan_dev *dst_vxlan;
2405 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2406 daddr->sa.sa_family, dst_port,
2409 dev->stats.tx_errors++;
2414 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2421 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2422 __be32 default_vni, struct vxlan_rdst *rdst,
2425 struct dst_cache *dst_cache;
2426 struct ip_tunnel_info *info;
2427 struct vxlan_dev *vxlan = netdev_priv(dev);
2428 const struct iphdr *old_iph = ip_hdr(skb);
2429 union vxlan_addr *dst;
2430 union vxlan_addr remote_ip, local_ip;
2431 struct vxlan_metadata _md;
2432 struct vxlan_metadata *md = &_md;
2433 __be16 src_port = 0, dst_port;
2434 struct dst_entry *ndst = NULL;
2439 u32 flags = vxlan->cfg.flags;
2440 bool udp_sum = false;
2441 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2443 info = skb_tunnel_info(skb);
2446 dst = &rdst->remote_ip;
2447 if (vxlan_addr_any(dst)) {
2449 /* short-circuited back to local bridge */
2450 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2456 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2457 vni = (rdst->remote_vni) ? : default_vni;
2458 ifindex = rdst->remote_ifindex;
2459 local_ip = vxlan->cfg.saddr;
2460 dst_cache = &rdst->dst_cache;
2461 md->gbp = skb->mark;
2462 if (flags & VXLAN_F_TTL_INHERIT) {
2463 ttl = ip_tunnel_get_ttl(old_iph, skb);
2465 ttl = vxlan->cfg.ttl;
2466 if (!ttl && vxlan_addr_multicast(dst))
2470 tos = vxlan->cfg.tos;
2472 tos = ip_tunnel_get_dsfield(old_iph, skb);
2474 if (dst->sa.sa_family == AF_INET)
2475 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2477 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2478 label = vxlan->cfg.label;
2481 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2485 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2486 if (remote_ip.sa.sa_family == AF_INET) {
2487 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2488 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2490 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2491 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2494 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2495 vni = tunnel_id_to_key32(info->key.tun_id);
2497 dst_cache = &info->dst_cache;
2498 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2499 if (info->options_len < sizeof(*md))
2501 md = ip_tunnel_info_opts(info);
2503 ttl = info->key.ttl;
2504 tos = info->key.tos;
2505 label = info->key.label;
2506 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2508 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2509 vxlan->cfg.port_max, true);
2512 if (dst->sa.sa_family == AF_INET) {
2513 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2518 ifindex = sock4->sock->sk->sk_bound_dev_if;
2520 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2521 dst->sin.sin_addr.s_addr,
2522 &local_ip.sin.sin_addr.s_addr,
2531 /* Bypass encapsulation if the destination is local */
2532 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2533 dst_port, ifindex, vni,
2534 &rt->dst, rt->rt_flags);
2538 if (vxlan->cfg.df == VXLAN_DF_SET) {
2540 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2541 struct ethhdr *eth = eth_hdr(skb);
2543 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2544 (ntohs(eth->h_proto) == ETH_P_IP &&
2545 old_iph->frag_off & htons(IP_DF)))
2548 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2553 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
2555 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2556 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2557 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2558 vni, md, flags, udp_sum);
2562 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2563 dst->sin.sin_addr.s_addr, tos, ttl, df,
2564 src_port, dst_port, xnet, !udp_sum);
2565 #if IS_ENABLED(CONFIG_IPV6)
2567 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2570 ifindex = sock6->sock->sk->sk_bound_dev_if;
2572 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2573 label, &dst->sin6.sin6_addr,
2574 &local_ip.sin6.sin6_addr,
2578 err = PTR_ERR(ndst);
2584 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2586 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2587 dst_port, ifindex, vni,
2593 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
2595 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2596 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2597 skb_scrub_packet(skb, xnet);
2598 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2599 vni, md, flags, udp_sum);
2603 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2604 &local_ip.sin6.sin6_addr,
2605 &dst->sin6.sin6_addr, tos, ttl,
2606 label, src_port, dst_port, !udp_sum);
2614 dev->stats.tx_dropped++;
2621 dev->stats.collisions++;
2622 else if (err == -ENETUNREACH)
2623 dev->stats.tx_carrier_errors++;
2625 dev->stats.tx_errors++;
2629 /* Transmit local packets over Vxlan
2631 * Outer IP header inherits ECN and DF from inner header.
2632 * Outer UDP destination is the VXLAN assigned port.
2633 * source port is based on hash of flow
2635 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2637 struct vxlan_dev *vxlan = netdev_priv(dev);
2638 struct vxlan_rdst *rdst, *fdst = NULL;
2639 const struct ip_tunnel_info *info;
2640 bool did_rsc = false;
2641 struct vxlan_fdb *f;
2645 info = skb_tunnel_info(skb);
2647 skb_reset_mac_header(skb);
2649 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2650 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2651 info->mode & IP_TUNNEL_INFO_TX) {
2652 vni = tunnel_id_to_key32(info->key.tun_id);
2654 if (info && info->mode & IP_TUNNEL_INFO_TX)
2655 vxlan_xmit_one(skb, dev, vni, NULL, false);
2658 return NETDEV_TX_OK;
2662 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2664 if (ntohs(eth->h_proto) == ETH_P_ARP)
2665 return arp_reduce(dev, skb, vni);
2666 #if IS_ENABLED(CONFIG_IPV6)
2667 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2668 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2669 sizeof(struct nd_msg)) &&
2670 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2671 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2673 if (m->icmph.icmp6_code == 0 &&
2674 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2675 return neigh_reduce(dev, skb, vni);
2681 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2684 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2685 (ntohs(eth->h_proto) == ETH_P_IP ||
2686 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2687 did_rsc = route_shortcircuit(dev, skb);
2689 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2693 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2695 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2696 !is_multicast_ether_addr(eth->h_dest))
2697 vxlan_fdb_miss(vxlan, eth->h_dest);
2699 dev->stats.tx_dropped++;
2701 return NETDEV_TX_OK;
2705 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2706 struct sk_buff *skb1;
2712 skb1 = skb_clone(skb, GFP_ATOMIC);
2714 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2718 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2721 return NETDEV_TX_OK;
2724 /* Walk the forwarding table and purge stale entries */
2725 static void vxlan_cleanup(struct timer_list *t)
2727 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2728 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2731 if (!netif_running(vxlan->dev))
2734 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2735 struct hlist_node *p, *n;
2737 spin_lock(&vxlan->hash_lock[h]);
2738 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2740 = container_of(p, struct vxlan_fdb, hlist);
2741 unsigned long timeout;
2743 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2746 if (f->flags & NTF_EXT_LEARNED)
2749 timeout = f->used + vxlan->cfg.age_interval * HZ;
2750 if (time_before_eq(timeout, jiffies)) {
2751 netdev_dbg(vxlan->dev,
2752 "garbage collect %pM\n",
2754 f->state = NUD_STALE;
2755 vxlan_fdb_destroy(vxlan, f, true, true);
2756 } else if (time_before(timeout, next_timer))
2757 next_timer = timeout;
2759 spin_unlock(&vxlan->hash_lock[h]);
2762 mod_timer(&vxlan->age_timer, next_timer);
2765 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2767 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2769 spin_lock(&vn->sock_lock);
2770 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2771 #if IS_ENABLED(CONFIG_IPV6)
2772 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2774 spin_unlock(&vn->sock_lock);
2777 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2778 struct vxlan_dev_node *node)
2780 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2781 __be32 vni = vxlan->default_dst.remote_vni;
2783 node->vxlan = vxlan;
2784 spin_lock(&vn->sock_lock);
2785 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2786 spin_unlock(&vn->sock_lock);
2789 /* Setup stats when device is created */
2790 static int vxlan_init(struct net_device *dev)
2792 struct vxlan_dev *vxlan = netdev_priv(dev);
2795 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2799 err = gro_cells_init(&vxlan->gro_cells, dev);
2801 free_percpu(dev->tstats);
2808 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2810 struct vxlan_fdb *f;
2811 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2813 spin_lock_bh(&vxlan->hash_lock[hash_index]);
2814 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2816 vxlan_fdb_destroy(vxlan, f, true, true);
2817 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2820 static void vxlan_uninit(struct net_device *dev)
2822 struct vxlan_dev *vxlan = netdev_priv(dev);
2824 gro_cells_destroy(&vxlan->gro_cells);
2826 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2828 free_percpu(dev->tstats);
2831 /* Start ageing timer and join group when device is brought up */
2832 static int vxlan_open(struct net_device *dev)
2834 struct vxlan_dev *vxlan = netdev_priv(dev);
2837 ret = vxlan_sock_add(vxlan);
2841 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2842 ret = vxlan_igmp_join(vxlan);
2843 if (ret == -EADDRINUSE)
2846 vxlan_sock_release(vxlan);
2851 if (vxlan->cfg.age_interval)
2852 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2857 /* Purge the forwarding table */
2858 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2862 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2863 struct hlist_node *p, *n;
2865 spin_lock_bh(&vxlan->hash_lock[h]);
2866 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2868 = container_of(p, struct vxlan_fdb, hlist);
2869 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2871 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2872 if (is_zero_ether_addr(f->eth_addr) &&
2873 f->vni == vxlan->cfg.vni)
2875 vxlan_fdb_destroy(vxlan, f, true, true);
2877 spin_unlock_bh(&vxlan->hash_lock[h]);
2881 /* Cleanup timer and forwarding table on shutdown */
2882 static int vxlan_stop(struct net_device *dev)
2884 struct vxlan_dev *vxlan = netdev_priv(dev);
2885 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2888 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2889 !vxlan_group_used(vn, vxlan))
2890 ret = vxlan_igmp_leave(vxlan);
2892 del_timer_sync(&vxlan->age_timer);
2894 vxlan_flush(vxlan, false);
2895 vxlan_sock_release(vxlan);
2900 /* Stub, nothing needs to be done. */
2901 static void vxlan_set_multicast_list(struct net_device *dev)
2905 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2907 struct vxlan_dev *vxlan = netdev_priv(dev);
2908 struct vxlan_rdst *dst = &vxlan->default_dst;
2909 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2910 dst->remote_ifindex);
2911 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2913 /* This check is different than dev->max_mtu, because it looks at
2914 * the lowerdev->mtu, rather than the static dev->max_mtu
2917 int max_mtu = lowerdev->mtu -
2918 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2919 if (new_mtu > max_mtu)
2927 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2929 struct vxlan_dev *vxlan = netdev_priv(dev);
2930 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2931 __be16 sport, dport;
2933 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2934 vxlan->cfg.port_max, true);
2935 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2937 if (ip_tunnel_info_af(info) == AF_INET) {
2938 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2941 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2942 info->key.u.ipv4.dst,
2943 &info->key.u.ipv4.src, dport, sport,
2944 &info->dst_cache, info);
2949 #if IS_ENABLED(CONFIG_IPV6)
2950 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2951 struct dst_entry *ndst;
2953 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2954 info->key.label, &info->key.u.ipv6.dst,
2955 &info->key.u.ipv6.src, dport, sport,
2956 &info->dst_cache, info);
2958 return PTR_ERR(ndst);
2960 #else /* !CONFIG_IPV6 */
2961 return -EPFNOSUPPORT;
2964 info->key.tp_src = sport;
2965 info->key.tp_dst = dport;
2969 static const struct net_device_ops vxlan_netdev_ether_ops = {
2970 .ndo_init = vxlan_init,
2971 .ndo_uninit = vxlan_uninit,
2972 .ndo_open = vxlan_open,
2973 .ndo_stop = vxlan_stop,
2974 .ndo_start_xmit = vxlan_xmit,
2975 .ndo_get_stats64 = ip_tunnel_get_stats64,
2976 .ndo_set_rx_mode = vxlan_set_multicast_list,
2977 .ndo_change_mtu = vxlan_change_mtu,
2978 .ndo_validate_addr = eth_validate_addr,
2979 .ndo_set_mac_address = eth_mac_addr,
2980 .ndo_fdb_add = vxlan_fdb_add,
2981 .ndo_fdb_del = vxlan_fdb_delete,
2982 .ndo_fdb_dump = vxlan_fdb_dump,
2983 .ndo_fdb_get = vxlan_fdb_get,
2984 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2985 .ndo_change_proto_down = dev_change_proto_down_generic,
2988 static const struct net_device_ops vxlan_netdev_raw_ops = {
2989 .ndo_init = vxlan_init,
2990 .ndo_uninit = vxlan_uninit,
2991 .ndo_open = vxlan_open,
2992 .ndo_stop = vxlan_stop,
2993 .ndo_start_xmit = vxlan_xmit,
2994 .ndo_get_stats64 = ip_tunnel_get_stats64,
2995 .ndo_change_mtu = vxlan_change_mtu,
2996 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2999 /* Info for udev, that this is a virtual tunnel endpoint */
3000 static struct device_type vxlan_type = {
3004 /* Calls the ndo_udp_tunnel_add of the caller in order to
3005 * supply the listening VXLAN udp ports. Callers are expected
3006 * to implement the ndo_udp_tunnel_add.
3008 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3010 struct vxlan_sock *vs;
3011 struct net *net = dev_net(dev);
3012 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3015 spin_lock(&vn->sock_lock);
3016 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3017 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3018 unsigned short type;
3020 if (vs->flags & VXLAN_F_GPE)
3021 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3023 type = UDP_TUNNEL_TYPE_VXLAN;
3026 udp_tunnel_push_rx_port(dev, vs->sock, type);
3028 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3031 spin_unlock(&vn->sock_lock);
3034 /* Initialize the device structure. */
3035 static void vxlan_setup(struct net_device *dev)
3037 struct vxlan_dev *vxlan = netdev_priv(dev);
3040 eth_hw_addr_random(dev);
3043 dev->needs_free_netdev = true;
3044 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3046 dev->features |= NETIF_F_LLTX;
3047 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3048 dev->features |= NETIF_F_RXCSUM;
3049 dev->features |= NETIF_F_GSO_SOFTWARE;
3051 dev->vlan_features = dev->features;
3052 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3053 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3054 netif_keep_dst(dev);
3055 dev->priv_flags |= IFF_NO_QUEUE;
3057 /* MTU range: 68 - 65535 */
3058 dev->min_mtu = ETH_MIN_MTU;
3059 dev->max_mtu = ETH_MAX_MTU;
3061 INIT_LIST_HEAD(&vxlan->next);
3063 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3067 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3068 spin_lock_init(&vxlan->hash_lock[h]);
3069 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3073 static void vxlan_ether_setup(struct net_device *dev)
3075 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3076 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3077 dev->netdev_ops = &vxlan_netdev_ether_ops;
3080 static void vxlan_raw_setup(struct net_device *dev)
3082 dev->header_ops = NULL;
3083 dev->type = ARPHRD_NONE;
3084 dev->hard_header_len = 0;
3086 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3087 dev->netdev_ops = &vxlan_netdev_raw_ops;
3090 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3091 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3092 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
3093 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3094 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3095 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
3096 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3097 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3098 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3099 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3100 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3101 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3102 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3103 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3104 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3105 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3106 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3107 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3108 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3109 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3110 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3111 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3112 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3113 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3114 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3115 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3116 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3117 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3118 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3119 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3122 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3123 struct netlink_ext_ack *extack)
3125 if (tb[IFLA_ADDRESS]) {
3126 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3127 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3128 "Provided link layer address is not Ethernet");
3132 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3133 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3134 "Provided Ethernet address is not unicast");
3135 return -EADDRNOTAVAIL;
3140 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3142 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3143 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3144 "MTU must be between 68 and 65535");
3150 NL_SET_ERR_MSG(extack,
3151 "Required attributes not provided to perform the operation");
3155 if (data[IFLA_VXLAN_ID]) {
3156 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3158 if (id >= VXLAN_N_VID) {
3159 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3160 "VXLAN ID must be lower than 16777216");
3165 if (data[IFLA_VXLAN_PORT_RANGE]) {
3166 const struct ifla_vxlan_port_range *p
3167 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3169 if (ntohs(p->high) < ntohs(p->low)) {
3170 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3171 "Invalid source port range");
3176 if (data[IFLA_VXLAN_DF]) {
3177 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3179 if (df < 0 || df > VXLAN_DF_MAX) {
3180 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3181 "Invalid DF attribute");
3189 static void vxlan_get_drvinfo(struct net_device *netdev,
3190 struct ethtool_drvinfo *drvinfo)
3192 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3193 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3196 static const struct ethtool_ops vxlan_ethtool_ops = {
3197 .get_drvinfo = vxlan_get_drvinfo,
3198 .get_link = ethtool_op_get_link,
3201 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3202 __be16 port, u32 flags, int ifindex)
3204 struct socket *sock;
3205 struct udp_port_cfg udp_conf;
3208 memset(&udp_conf, 0, sizeof(udp_conf));
3211 udp_conf.family = AF_INET6;
3212 udp_conf.use_udp6_rx_checksums =
3213 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3214 udp_conf.ipv6_v6only = 1;
3216 udp_conf.family = AF_INET;
3219 udp_conf.local_udp_port = port;
3220 udp_conf.bind_ifindex = ifindex;
3222 /* Open UDP socket */
3223 err = udp_sock_create(net, &udp_conf, &sock);
3225 return ERR_PTR(err);
3230 /* Create new listen socket if needed */
3231 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3232 __be16 port, u32 flags,
3235 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3236 struct vxlan_sock *vs;
3237 struct socket *sock;
3239 struct udp_tunnel_sock_cfg tunnel_cfg;
3241 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3243 return ERR_PTR(-ENOMEM);
3245 for (h = 0; h < VNI_HASH_SIZE; ++h)
3246 INIT_HLIST_HEAD(&vs->vni_list[h]);
3248 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3251 return ERR_CAST(sock);
3255 refcount_set(&vs->refcnt, 1);
3256 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3258 spin_lock(&vn->sock_lock);
3259 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3260 udp_tunnel_notify_add_rx_port(sock,
3261 (vs->flags & VXLAN_F_GPE) ?
3262 UDP_TUNNEL_TYPE_VXLAN_GPE :
3263 UDP_TUNNEL_TYPE_VXLAN);
3264 spin_unlock(&vn->sock_lock);
3266 /* Mark socket as an encapsulation socket. */
3267 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3268 tunnel_cfg.sk_user_data = vs;
3269 tunnel_cfg.encap_type = 1;
3270 tunnel_cfg.encap_rcv = vxlan_rcv;
3271 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3272 tunnel_cfg.encap_destroy = NULL;
3273 tunnel_cfg.gro_receive = vxlan_gro_receive;
3274 tunnel_cfg.gro_complete = vxlan_gro_complete;
3276 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3281 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3283 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3284 struct vxlan_sock *vs = NULL;
3285 struct vxlan_dev_node *node;
3286 int l3mdev_index = 0;
3288 if (vxlan->cfg.remote_ifindex)
3289 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3290 vxlan->net, vxlan->cfg.remote_ifindex);
3292 if (!vxlan->cfg.no_share) {
3293 spin_lock(&vn->sock_lock);
3294 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3295 vxlan->cfg.dst_port, vxlan->cfg.flags,
3297 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3298 spin_unlock(&vn->sock_lock);
3301 spin_unlock(&vn->sock_lock);
3304 vs = vxlan_socket_create(vxlan->net, ipv6,
3305 vxlan->cfg.dst_port, vxlan->cfg.flags,
3309 #if IS_ENABLED(CONFIG_IPV6)
3311 rcu_assign_pointer(vxlan->vn6_sock, vs);
3312 node = &vxlan->hlist6;
3316 rcu_assign_pointer(vxlan->vn4_sock, vs);
3317 node = &vxlan->hlist4;
3319 vxlan_vs_add_dev(vs, vxlan, node);
3323 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3325 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3326 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3327 bool ipv4 = !ipv6 || metadata;
3330 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3331 #if IS_ENABLED(CONFIG_IPV6)
3332 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3334 ret = __vxlan_sock_add(vxlan, true);
3335 if (ret < 0 && ret != -EAFNOSUPPORT)
3340 ret = __vxlan_sock_add(vxlan, false);
3342 vxlan_sock_release(vxlan);
3346 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3347 struct net_device **lower,
3348 struct vxlan_dev *old,
3349 struct netlink_ext_ack *extack)
3351 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3352 struct vxlan_dev *tmp;
3353 bool use_ipv6 = false;
3355 if (conf->flags & VXLAN_F_GPE) {
3356 /* For now, allow GPE only together with
3357 * COLLECT_METADATA. This can be relaxed later; in such
3358 * case, the other side of the PtP link will have to be
3361 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3362 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3363 NL_SET_ERR_MSG(extack,
3364 "VXLAN GPE does not support this combination of attributes");
3369 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3370 /* Unless IPv6 is explicitly requested, assume IPv4 */
3371 conf->remote_ip.sa.sa_family = AF_INET;
3372 conf->saddr.sa.sa_family = AF_INET;
3373 } else if (!conf->remote_ip.sa.sa_family) {
3374 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3375 } else if (!conf->saddr.sa.sa_family) {
3376 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3379 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3380 NL_SET_ERR_MSG(extack,
3381 "Local and remote address must be from the same family");
3385 if (vxlan_addr_multicast(&conf->saddr)) {
3386 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3390 if (conf->saddr.sa.sa_family == AF_INET6) {
3391 if (!IS_ENABLED(CONFIG_IPV6)) {
3392 NL_SET_ERR_MSG(extack,
3393 "IPv6 support not enabled in the kernel");
3394 return -EPFNOSUPPORT;
3397 conf->flags |= VXLAN_F_IPV6;
3399 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3401 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3403 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3405 if (local_type & IPV6_ADDR_LINKLOCAL) {
3406 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3407 (remote_type != IPV6_ADDR_ANY)) {
3408 NL_SET_ERR_MSG(extack,
3409 "Invalid combination of local and remote address scopes");
3413 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3416 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3417 NL_SET_ERR_MSG(extack,
3418 "Invalid combination of local and remote address scopes");
3422 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3427 if (conf->label && !use_ipv6) {
3428 NL_SET_ERR_MSG(extack,
3429 "Label attribute only applies to IPv6 VXLAN devices");
3433 if (conf->remote_ifindex) {
3434 struct net_device *lowerdev;
3436 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3438 NL_SET_ERR_MSG(extack,
3439 "Invalid local interface, device not found");
3443 #if IS_ENABLED(CONFIG_IPV6)
3445 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3446 if (idev && idev->cnf.disable_ipv6) {
3447 NL_SET_ERR_MSG(extack,
3448 "IPv6 support disabled by administrator");
3456 if (vxlan_addr_multicast(&conf->remote_ip)) {
3457 NL_SET_ERR_MSG(extack,
3458 "Local interface required for multicast remote destination");
3463 #if IS_ENABLED(CONFIG_IPV6)
3464 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3465 NL_SET_ERR_MSG(extack,
3466 "Local interface required for link-local local/remote addresses");
3474 if (!conf->dst_port) {
3475 if (conf->flags & VXLAN_F_GPE)
3476 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3478 conf->dst_port = htons(vxlan_port);
3481 if (!conf->age_interval)
3482 conf->age_interval = FDB_AGE_DEFAULT;
3484 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3488 if (tmp->cfg.vni != conf->vni)
3490 if (tmp->cfg.dst_port != conf->dst_port)
3492 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3493 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3496 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3497 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3500 NL_SET_ERR_MSG(extack,
3501 "A VXLAN device with the specified VNI already exists");
3508 static void vxlan_config_apply(struct net_device *dev,
3509 struct vxlan_config *conf,
3510 struct net_device *lowerdev,
3511 struct net *src_net,
3514 struct vxlan_dev *vxlan = netdev_priv(dev);
3515 struct vxlan_rdst *dst = &vxlan->default_dst;
3516 unsigned short needed_headroom = ETH_HLEN;
3517 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3518 int max_mtu = ETH_MAX_MTU;
3521 if (conf->flags & VXLAN_F_GPE)
3522 vxlan_raw_setup(dev);
3524 vxlan_ether_setup(dev);
3527 dev->mtu = conf->mtu;
3529 vxlan->net = src_net;
3532 dst->remote_vni = conf->vni;
3534 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3537 dst->remote_ifindex = conf->remote_ifindex;
3539 dev->gso_max_size = lowerdev->gso_max_size;
3540 dev->gso_max_segs = lowerdev->gso_max_segs;
3542 needed_headroom = lowerdev->hard_header_len;
3543 needed_headroom += lowerdev->needed_headroom;
3545 dev->needed_tailroom = lowerdev->needed_tailroom;
3547 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3549 if (max_mtu < ETH_MIN_MTU)
3550 max_mtu = ETH_MIN_MTU;
3552 if (!changelink && !conf->mtu)
3556 if (dev->mtu > max_mtu)
3559 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3560 needed_headroom += VXLAN6_HEADROOM;
3562 needed_headroom += VXLAN_HEADROOM;
3563 dev->needed_headroom = needed_headroom;
3565 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3568 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3569 struct vxlan_config *conf, bool changelink,
3570 struct netlink_ext_ack *extack)
3572 struct vxlan_dev *vxlan = netdev_priv(dev);
3573 struct net_device *lowerdev;
3576 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3580 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3585 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3586 struct vxlan_config *conf,
3587 struct netlink_ext_ack *extack)
3589 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3590 struct vxlan_dev *vxlan = netdev_priv(dev);
3591 struct net_device *remote_dev = NULL;
3592 struct vxlan_fdb *f = NULL;
3593 bool unregister = false;
3594 struct vxlan_rdst *dst;
3597 dst = &vxlan->default_dst;
3598 err = vxlan_dev_configure(net, dev, conf, false, extack);
3602 dev->ethtool_ops = &vxlan_ethtool_ops;
3604 /* create an fdb entry for a valid default destination */
3605 if (!vxlan_addr_any(&dst->remote_ip)) {
3606 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3608 NUD_REACHABLE | NUD_PERMANENT,
3609 vxlan->cfg.dst_port,
3612 dst->remote_ifindex,
3618 err = register_netdevice(dev);
3623 if (dst->remote_ifindex) {
3624 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3630 err = netdev_upper_dev_link(remote_dev, dev, extack);
3635 err = rtnl_configure_link(dev, NULL);
3640 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3642 /* notify default fdb entry */
3643 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3644 RTM_NEWNEIGH, true, extack);
3646 vxlan_fdb_destroy(vxlan, f, false, false);
3648 netdev_upper_dev_unlink(remote_dev, dev);
3653 list_add(&vxlan->next, &vn->vxlan_list);
3655 dst->remote_dev = remote_dev;
3659 netdev_upper_dev_unlink(remote_dev, dev);
3661 /* unregister_netdevice() destroys the default FDB entry with deletion
3662 * notification. But the addition notification was not sent yet, so
3663 * destroy the entry by hand here.
3666 __vxlan_fdb_free(f);
3669 unregister_netdevice(dev);
3673 /* Set/clear flags based on attribute */
3674 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3675 int attrtype, unsigned long mask, bool changelink,
3676 bool changelink_supported,
3677 struct netlink_ext_ack *extack)
3679 unsigned long flags;
3684 if (changelink && !changelink_supported) {
3685 vxlan_flag_attr_error(attrtype, extack);
3689 if (vxlan_policy[attrtype].type == NLA_FLAG)
3690 flags = conf->flags | mask;
3691 else if (nla_get_u8(tb[attrtype]))
3692 flags = conf->flags | mask;
3694 flags = conf->flags & ~mask;
3696 conf->flags = flags;
3701 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3702 struct net_device *dev, struct vxlan_config *conf,
3703 bool changelink, struct netlink_ext_ack *extack)
3705 struct vxlan_dev *vxlan = netdev_priv(dev);
3708 memset(conf, 0, sizeof(*conf));
3710 /* if changelink operation, start with old existing cfg */
3712 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3714 if (data[IFLA_VXLAN_ID]) {
3715 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3717 if (changelink && (vni != conf->vni)) {
3718 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3721 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3724 if (data[IFLA_VXLAN_GROUP]) {
3725 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3726 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3730 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3731 conf->remote_ip.sa.sa_family = AF_INET;
3732 } else if (data[IFLA_VXLAN_GROUP6]) {
3733 if (!IS_ENABLED(CONFIG_IPV6)) {
3734 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3735 return -EPFNOSUPPORT;
3738 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3739 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3743 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3744 conf->remote_ip.sa.sa_family = AF_INET6;
3747 if (data[IFLA_VXLAN_LOCAL]) {
3748 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3749 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
3753 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3754 conf->saddr.sa.sa_family = AF_INET;
3755 } else if (data[IFLA_VXLAN_LOCAL6]) {
3756 if (!IS_ENABLED(CONFIG_IPV6)) {
3757 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
3758 return -EPFNOSUPPORT;
3761 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3762 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
3766 /* TODO: respect scope id */
3767 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3768 conf->saddr.sa.sa_family = AF_INET6;
3771 if (data[IFLA_VXLAN_LINK])
3772 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3774 if (data[IFLA_VXLAN_TOS])
3775 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3777 if (data[IFLA_VXLAN_TTL])
3778 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3780 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3781 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
3782 VXLAN_F_TTL_INHERIT, changelink, false,
3789 if (data[IFLA_VXLAN_LABEL])
3790 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3791 IPV6_FLOWLABEL_MASK;
3793 if (data[IFLA_VXLAN_LEARNING]) {
3794 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
3795 VXLAN_F_LEARN, changelink, true,
3799 } else if (!changelink) {
3800 /* default to learn on a new device */
3801 conf->flags |= VXLAN_F_LEARN;
3804 if (data[IFLA_VXLAN_AGEING])
3805 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3807 if (data[IFLA_VXLAN_PROXY]) {
3808 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
3809 VXLAN_F_PROXY, changelink, false,
3815 if (data[IFLA_VXLAN_RSC]) {
3816 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
3817 VXLAN_F_RSC, changelink, false,
3823 if (data[IFLA_VXLAN_L2MISS]) {
3824 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
3825 VXLAN_F_L2MISS, changelink, false,
3831 if (data[IFLA_VXLAN_L3MISS]) {
3832 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
3833 VXLAN_F_L3MISS, changelink, false,
3839 if (data[IFLA_VXLAN_LIMIT]) {
3841 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
3842 "Cannot change limit");
3845 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3848 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3849 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
3850 VXLAN_F_COLLECT_METADATA, changelink, false,
3856 if (data[IFLA_VXLAN_PORT_RANGE]) {
3858 const struct ifla_vxlan_port_range *p
3859 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3860 conf->port_min = ntohs(p->low);
3861 conf->port_max = ntohs(p->high);
3863 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3864 "Cannot change port range");
3869 if (data[IFLA_VXLAN_PORT]) {
3871 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
3872 "Cannot change port");
3875 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3878 if (data[IFLA_VXLAN_UDP_CSUM]) {
3880 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
3881 "Cannot change UDP_CSUM flag");
3884 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3885 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3888 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3889 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3890 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
3896 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3897 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3898 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
3904 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3905 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
3906 VXLAN_F_REMCSUM_TX, changelink, false,
3912 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3913 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
3914 VXLAN_F_REMCSUM_RX, changelink, false,
3920 if (data[IFLA_VXLAN_GBP]) {
3921 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
3922 VXLAN_F_GBP, changelink, false, extack);
3927 if (data[IFLA_VXLAN_GPE]) {
3928 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
3929 VXLAN_F_GPE, changelink, false,
3935 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3936 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
3937 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
3945 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3946 "Cannot change mtu");
3949 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3952 if (data[IFLA_VXLAN_DF])
3953 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
3958 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3959 struct nlattr *tb[], struct nlattr *data[],
3960 struct netlink_ext_ack *extack)
3962 struct vxlan_config conf;
3965 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
3969 return __vxlan_dev_create(src_net, dev, &conf, extack);
3972 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3973 struct nlattr *data[],
3974 struct netlink_ext_ack *extack)
3976 struct vxlan_dev *vxlan = netdev_priv(dev);
3977 struct net_device *lowerdev;
3978 struct vxlan_config conf;
3979 struct vxlan_rdst *dst;
3982 dst = &vxlan->default_dst;
3983 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
3987 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
3992 if (dst->remote_dev == lowerdev)
3995 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4000 /* handle default dst entry */
4001 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4002 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4004 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4005 if (!vxlan_addr_any(&conf.remote_ip)) {
4006 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4008 NUD_REACHABLE | NUD_PERMANENT,
4009 NLM_F_APPEND | NLM_F_CREATE,
4010 vxlan->cfg.dst_port,
4012 conf.remote_ifindex,
4013 NTF_SELF, true, extack);
4015 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4016 netdev_adjacent_change_abort(dst->remote_dev,
4021 if (!vxlan_addr_any(&dst->remote_ip))
4022 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4024 vxlan->cfg.dst_port,
4027 dst->remote_ifindex,
4029 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4032 if (conf.age_interval != vxlan->cfg.age_interval)
4033 mod_timer(&vxlan->age_timer, jiffies);
4035 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4036 if (lowerdev && lowerdev != dst->remote_dev) {
4037 dst->remote_dev = lowerdev;
4038 netdev_update_lockdep_key(lowerdev);
4040 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4044 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4046 struct vxlan_dev *vxlan = netdev_priv(dev);
4048 vxlan_flush(vxlan, true);
4050 list_del(&vxlan->next);
4051 unregister_netdevice_queue(dev, head);
4052 if (vxlan->default_dst.remote_dev)
4053 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4056 static size_t vxlan_get_size(const struct net_device *dev)
4059 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4060 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4061 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4062 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4063 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4064 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4065 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4066 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4067 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4068 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4069 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4070 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4071 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4072 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4073 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4074 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4075 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4076 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4077 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4078 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4079 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4080 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4081 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4082 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4086 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4088 const struct vxlan_dev *vxlan = netdev_priv(dev);
4089 const struct vxlan_rdst *dst = &vxlan->default_dst;
4090 struct ifla_vxlan_port_range ports = {
4091 .low = htons(vxlan->cfg.port_min),
4092 .high = htons(vxlan->cfg.port_max),
4095 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4096 goto nla_put_failure;
4098 if (!vxlan_addr_any(&dst->remote_ip)) {
4099 if (dst->remote_ip.sa.sa_family == AF_INET) {
4100 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4101 dst->remote_ip.sin.sin_addr.s_addr))
4102 goto nla_put_failure;
4103 #if IS_ENABLED(CONFIG_IPV6)
4105 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4106 &dst->remote_ip.sin6.sin6_addr))
4107 goto nla_put_failure;
4112 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4113 goto nla_put_failure;
4115 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4116 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4117 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4118 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4119 goto nla_put_failure;
4120 #if IS_ENABLED(CONFIG_IPV6)
4122 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4123 &vxlan->cfg.saddr.sin6.sin6_addr))
4124 goto nla_put_failure;
4129 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4130 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4131 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4132 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4133 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4134 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4135 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4136 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4137 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4138 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4139 nla_put_u8(skb, IFLA_VXLAN_RSC,
4140 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4141 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4142 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4143 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4144 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4145 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4146 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4147 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4148 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4149 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4150 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4151 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4152 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4153 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4154 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4155 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4156 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4157 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4158 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4159 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4160 goto nla_put_failure;
4162 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4163 goto nla_put_failure;
4165 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4166 nla_put_flag(skb, IFLA_VXLAN_GBP))
4167 goto nla_put_failure;
4169 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4170 nla_put_flag(skb, IFLA_VXLAN_GPE))
4171 goto nla_put_failure;
4173 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4174 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4175 goto nla_put_failure;
4183 static struct net *vxlan_get_link_net(const struct net_device *dev)
4185 struct vxlan_dev *vxlan = netdev_priv(dev);
4190 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4192 .maxtype = IFLA_VXLAN_MAX,
4193 .policy = vxlan_policy,
4194 .priv_size = sizeof(struct vxlan_dev),
4195 .setup = vxlan_setup,
4196 .validate = vxlan_validate,
4197 .newlink = vxlan_newlink,
4198 .changelink = vxlan_changelink,
4199 .dellink = vxlan_dellink,
4200 .get_size = vxlan_get_size,
4201 .fill_info = vxlan_fill_info,
4202 .get_link_net = vxlan_get_link_net,
4205 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4206 u8 name_assign_type,
4207 struct vxlan_config *conf)
4209 struct nlattr *tb[IFLA_MAX + 1];
4210 struct net_device *dev;
4213 memset(&tb, 0, sizeof(tb));
4215 dev = rtnl_create_link(net, name, name_assign_type,
4216 &vxlan_link_ops, tb, NULL);
4220 err = __vxlan_dev_create(net, dev, conf, NULL);
4223 return ERR_PTR(err);
4226 err = rtnl_configure_link(dev, NULL);
4228 LIST_HEAD(list_kill);
4230 vxlan_dellink(dev, &list_kill);
4231 unregister_netdevice_many(&list_kill);
4232 return ERR_PTR(err);
4237 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4239 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4240 struct net_device *dev)
4242 struct vxlan_dev *vxlan, *next;
4243 LIST_HEAD(list_kill);
4245 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4246 struct vxlan_rdst *dst = &vxlan->default_dst;
4248 /* In case we created vxlan device with carrier
4249 * and we loose the carrier due to module unload
4250 * we also need to remove vxlan device. In other
4251 * cases, it's not necessary and remote_ifindex
4252 * is 0 here, so no matches.
4254 if (dst->remote_ifindex == dev->ifindex)
4255 vxlan_dellink(vxlan->dev, &list_kill);
4258 unregister_netdevice_many(&list_kill);
4261 static int vxlan_netdevice_event(struct notifier_block *unused,
4262 unsigned long event, void *ptr)
4264 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4265 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4267 if (event == NETDEV_UNREGISTER) {
4268 vxlan_offload_rx_ports(dev, false);
4269 vxlan_handle_lowerdev_unregister(vn, dev);
4270 } else if (event == NETDEV_REGISTER) {
4271 vxlan_offload_rx_ports(dev, true);
4272 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4273 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4274 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4280 static struct notifier_block vxlan_notifier_block __read_mostly = {
4281 .notifier_call = vxlan_netdevice_event,
4285 vxlan_fdb_offloaded_set(struct net_device *dev,
4286 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4288 struct vxlan_dev *vxlan = netdev_priv(dev);
4289 struct vxlan_rdst *rdst;
4290 struct vxlan_fdb *f;
4293 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4295 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4297 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4301 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4302 fdb_info->remote_port,
4303 fdb_info->remote_vni,
4304 fdb_info->remote_ifindex);
4308 rdst->offloaded = fdb_info->offloaded;
4311 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4315 vxlan_fdb_external_learn_add(struct net_device *dev,
4316 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4318 struct vxlan_dev *vxlan = netdev_priv(dev);
4319 struct netlink_ext_ack *extack;
4323 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4324 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4326 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4327 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4329 NLM_F_CREATE | NLM_F_REPLACE,
4330 fdb_info->remote_port,
4332 fdb_info->remote_vni,
4333 fdb_info->remote_ifindex,
4334 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4336 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4342 vxlan_fdb_external_learn_del(struct net_device *dev,
4343 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4345 struct vxlan_dev *vxlan = netdev_priv(dev);
4346 struct vxlan_fdb *f;
4350 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4351 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4353 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4356 else if (f->flags & NTF_EXT_LEARNED)
4357 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4358 fdb_info->remote_ip,
4359 fdb_info->remote_port,
4361 fdb_info->remote_vni,
4362 fdb_info->remote_ifindex,
4365 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4370 static int vxlan_switchdev_event(struct notifier_block *unused,
4371 unsigned long event, void *ptr)
4373 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4374 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4378 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4379 vxlan_fdb_offloaded_set(dev, ptr);
4381 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4383 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4385 err = notifier_from_errno(err);
4388 fdb_info->offloaded = true;
4389 vxlan_fdb_offloaded_set(dev, fdb_info);
4391 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4393 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4395 err = notifier_from_errno(err);
4398 fdb_info->offloaded = false;
4399 vxlan_fdb_offloaded_set(dev, fdb_info);
4406 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4407 .notifier_call = vxlan_switchdev_event,
4410 static __net_init int vxlan_init_net(struct net *net)
4412 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4415 INIT_LIST_HEAD(&vn->vxlan_list);
4416 spin_lock_init(&vn->sock_lock);
4418 for (h = 0; h < PORT_HASH_SIZE; ++h)
4419 INIT_HLIST_HEAD(&vn->sock_list[h]);
4424 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4426 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4427 struct vxlan_dev *vxlan, *next;
4428 struct net_device *dev, *aux;
4430 for_each_netdev_safe(net, dev, aux)
4431 if (dev->rtnl_link_ops == &vxlan_link_ops)
4432 unregister_netdevice_queue(dev, head);
4434 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4435 /* If vxlan->dev is in the same netns, it has already been added
4436 * to the list by the previous loop.
4438 if (!net_eq(dev_net(vxlan->dev), net))
4439 unregister_netdevice_queue(vxlan->dev, head);
4444 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4451 list_for_each_entry(net, net_list, exit_list)
4452 vxlan_destroy_tunnels(net, &list);
4454 unregister_netdevice_many(&list);
4457 list_for_each_entry(net, net_list, exit_list) {
4458 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4460 for (h = 0; h < PORT_HASH_SIZE; ++h)
4461 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4465 static struct pernet_operations vxlan_net_ops = {
4466 .init = vxlan_init_net,
4467 .exit_batch = vxlan_exit_batch_net,
4468 .id = &vxlan_net_id,
4469 .size = sizeof(struct vxlan_net),
4472 static int __init vxlan_init_module(void)
4476 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4478 rc = register_pernet_subsys(&vxlan_net_ops);
4482 rc = register_netdevice_notifier(&vxlan_notifier_block);
4486 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4490 rc = rtnl_link_register(&vxlan_link_ops);
4496 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4498 unregister_netdevice_notifier(&vxlan_notifier_block);
4500 unregister_pernet_subsys(&vxlan_net_ops);
4504 late_initcall(vxlan_init_module);
4506 static void __exit vxlan_cleanup_module(void)
4508 rtnl_link_unregister(&vxlan_link_ops);
4509 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4510 unregister_netdevice_notifier(&vxlan_notifier_block);
4511 unregister_pernet_subsys(&vxlan_net_ops);
4512 /* rcu_barrier() is called by netns */
4514 module_exit(vxlan_cleanup_module);
4516 MODULE_LICENSE("GPL");
4517 MODULE_VERSION(VXLAN_VERSION);
4518 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4519 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4520 MODULE_ALIAS_RTNL_LINK("vxlan");