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>
29 #include <net/nexthop.h>
31 #if IS_ENABLED(CONFIG_IPV6)
32 #include <net/ip6_tunnel.h>
33 #include <net/ip6_checksum.h>
36 #define VXLAN_VERSION "0.1"
38 #define PORT_HASH_BITS 8
39 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
40 #define FDB_AGE_DEFAULT 300 /* 5 min */
41 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43 /* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
45 * for compatibility with early adopters.
47 static unsigned short vxlan_port __read_mostly = 8472;
48 module_param_named(udp_port, vxlan_port, ushort, 0444);
49 MODULE_PARM_DESC(udp_port, "Destination UDP port");
51 static bool log_ecn_error = true;
52 module_param(log_ecn_error, bool, 0644);
53 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55 static unsigned int vxlan_net_id;
56 static struct rtnl_link_ops vxlan_link_ops;
58 static const u8 all_zeros_mac[ETH_ALEN + 2];
60 static int vxlan_sock_add(struct vxlan_dev *vxlan);
62 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64 /* per-network namespace private data for this module */
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
71 /* Forwarding table entry */
73 struct hlist_node hlist; /* linked list of entries */
75 unsigned long updated; /* jiffies */
77 struct list_head remotes;
78 u8 eth_addr[ETH_ALEN];
79 u16 state; /* see ndm_state */
81 u16 flags; /* see ndm_flags and below */
82 struct list_head nh_list;
83 struct nexthop __rcu *nh;
84 struct vxlan_dev __rcu *vdev;
87 #define NTF_VXLAN_ADDED_BY_USER 0x100
89 /* salt for hash table */
90 static u32 vxlan_salt __read_mostly;
92 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
94 return vs->flags & VXLAN_F_COLLECT_METADATA ||
95 ip_tunnel_collect_metadata();
98 #if IS_ENABLED(CONFIG_IPV6)
100 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
102 if (a->sa.sa_family != b->sa.sa_family)
104 if (a->sa.sa_family == AF_INET6)
105 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
107 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
110 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
112 if (nla_len(nla) >= sizeof(struct in6_addr)) {
113 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
114 ip->sa.sa_family = AF_INET6;
116 } else if (nla_len(nla) >= sizeof(__be32)) {
117 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
118 ip->sa.sa_family = AF_INET;
121 return -EAFNOSUPPORT;
125 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
126 const union vxlan_addr *ip)
128 if (ip->sa.sa_family == AF_INET6)
129 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
131 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
134 #else /* !CONFIG_IPV6 */
137 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
139 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
142 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
144 if (nla_len(nla) >= sizeof(struct in6_addr)) {
145 return -EAFNOSUPPORT;
146 } else if (nla_len(nla) >= sizeof(__be32)) {
147 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
148 ip->sa.sa_family = AF_INET;
151 return -EAFNOSUPPORT;
155 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
156 const union vxlan_addr *ip)
158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
162 /* Virtual Network hash table head */
163 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
165 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
168 /* Socket hash table head */
169 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
171 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
176 /* First remote destination for a forwarding entry.
177 * Guaranteed to be non-NULL because remotes are never deleted.
179 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
181 if (rcu_access_pointer(fdb->nh))
183 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
186 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
188 if (rcu_access_pointer(fdb->nh))
190 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
193 /* Find VXLAN socket based on network namespace, address family, UDP port,
194 * enabled unshareable flags and socket device binding (see l3mdev with
197 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
198 __be16 port, u32 flags, int ifindex)
200 struct vxlan_sock *vs;
202 flags &= VXLAN_F_RCV_FLAGS;
204 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
205 if (inet_sk(vs->sock->sk)->inet_sport == port &&
206 vxlan_get_sk_family(vs) == family &&
207 vs->flags == flags &&
208 vs->sock->sk->sk_bound_dev_if == ifindex)
214 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
217 struct vxlan_dev_node *node;
219 /* For flow based devices, map all packets to VNI 0 */
220 if (vs->flags & VXLAN_F_COLLECT_METADATA)
223 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
224 if (node->vxlan->default_dst.remote_vni != vni)
227 if (IS_ENABLED(CONFIG_IPV6)) {
228 const struct vxlan_config *cfg = &node->vxlan->cfg;
230 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
231 cfg->remote_ifindex != ifindex)
241 /* Look up VNI in a per net namespace table */
242 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
243 __be32 vni, sa_family_t family,
244 __be16 port, u32 flags)
246 struct vxlan_sock *vs;
248 vs = vxlan_find_sock(net, family, port, flags, ifindex);
252 return vxlan_vs_find_vni(vs, ifindex, vni);
255 /* Fill in neighbour message in skbuff. */
256 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
257 const struct vxlan_fdb *fdb,
258 u32 portid, u32 seq, int type, unsigned int flags,
259 const struct vxlan_rdst *rdst)
261 unsigned long now = jiffies;
262 struct nda_cacheinfo ci;
263 bool send_ip, send_eth;
264 struct nlmsghdr *nlh;
270 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
274 ndm = nlmsg_data(nlh);
275 memset(ndm, 0, sizeof(*ndm));
277 send_eth = send_ip = true;
280 nh = rcu_dereference(fdb->nh);
282 nh_family = nexthop_get_family(nh);
287 if (type == RTM_GETNEIGH) {
289 send_ip = !vxlan_addr_any(&rdst->remote_ip);
290 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
292 ndm->ndm_family = nh_family;
294 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 ndm->ndm_family = AF_BRIDGE;
297 ndm->ndm_state = fdb->state;
298 ndm->ndm_ifindex = vxlan->dev->ifindex;
299 ndm->ndm_flags = fdb->flags;
300 if (rdst && rdst->offloaded)
301 ndm->ndm_flags |= NTF_OFFLOADED;
302 ndm->ndm_type = RTN_UNICAST;
304 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
305 nla_put_s32(skb, NDA_LINK_NETNSID,
306 peernet2id(dev_net(vxlan->dev), vxlan->net)))
307 goto nla_put_failure;
309 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
310 goto nla_put_failure;
312 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
313 goto nla_put_failure;
315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
317 goto nla_put_failure;
319 if (rdst->remote_port &&
320 rdst->remote_port != vxlan->cfg.dst_port &&
321 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
322 goto nla_put_failure;
323 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
324 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
325 goto nla_put_failure;
326 if (rdst->remote_ifindex &&
327 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
328 goto nla_put_failure;
331 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
332 nla_put_u32(skb, NDA_SRC_VNI,
333 be32_to_cpu(fdb->vni)))
334 goto nla_put_failure;
336 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
337 ci.ndm_confirmed = 0;
338 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
341 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
342 goto nla_put_failure;
348 nlmsg_cancel(skb, nlh);
352 static inline size_t vxlan_nlmsg_size(void)
354 return NLMSG_ALIGN(sizeof(struct ndmsg))
355 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
356 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
357 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
358 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
359 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
360 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
361 + nla_total_size(sizeof(struct nda_cacheinfo));
364 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
365 struct vxlan_rdst *rd, int type)
367 struct net *net = dev_net(vxlan->dev);
371 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
375 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
377 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
378 WARN_ON(err == -EMSGSIZE);
383 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
387 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
390 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
391 const struct vxlan_fdb *fdb,
392 const struct vxlan_rdst *rd,
393 struct netlink_ext_ack *extack,
394 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
396 fdb_info->info.dev = vxlan->dev;
397 fdb_info->info.extack = extack;
398 fdb_info->remote_ip = rd->remote_ip;
399 fdb_info->remote_port = rd->remote_port;
400 fdb_info->remote_vni = rd->remote_vni;
401 fdb_info->remote_ifindex = rd->remote_ifindex;
402 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
403 fdb_info->vni = fdb->vni;
404 fdb_info->offloaded = rd->offloaded;
405 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
408 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
409 struct vxlan_fdb *fdb,
410 struct vxlan_rdst *rd,
412 struct netlink_ext_ack *extack)
414 struct switchdev_notifier_vxlan_fdb_info info;
415 enum switchdev_notifier_type notifier_type;
421 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
422 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
423 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
424 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
426 return notifier_to_errno(ret);
429 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
430 struct vxlan_rdst *rd, int type, bool swdev_notify,
431 struct netlink_ext_ack *extack)
435 if (swdev_notify && rd) {
438 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
444 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
450 __vxlan_fdb_notify(vxlan, fdb, rd, type);
454 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
456 struct vxlan_dev *vxlan = netdev_priv(dev);
457 struct vxlan_fdb f = {
460 struct vxlan_rdst remote = {
461 .remote_ip = *ipa, /* goes to NDA_DST */
462 .remote_vni = cpu_to_be32(VXLAN_N_VID),
465 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
468 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
470 struct vxlan_fdb f = {
473 struct vxlan_rdst remote = { };
475 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
477 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
480 /* Hash Ethernet address */
481 static u32 eth_hash(const unsigned char *addr)
483 u64 value = get_unaligned((u64 *)addr);
485 /* only want 6 bytes */
491 return hash_64(value, FDB_HASH_BITS);
494 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
496 /* use 1 byte of OUI and 3 bytes of NIC */
497 u32 key = get_unaligned((u32 *)(addr + 2));
499 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
502 static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
504 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
505 return eth_vni_hash(mac, vni);
507 return eth_hash(mac);
510 /* Hash chain to use given mac address */
511 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
512 const u8 *mac, __be32 vni)
514 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
517 /* Look up Ethernet address in forwarding table */
518 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
519 const u8 *mac, __be32 vni)
521 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
524 hlist_for_each_entry_rcu(f, head, hlist) {
525 if (ether_addr_equal(mac, f->eth_addr)) {
526 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
538 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
539 const u8 *mac, __be32 vni)
543 f = __vxlan_find_mac(vxlan, mac, vni);
544 if (f && f->used != jiffies)
550 /* caller should hold vxlan->hash_lock */
551 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
552 union vxlan_addr *ip, __be16 port,
553 __be32 vni, __u32 ifindex)
555 struct vxlan_rdst *rd;
557 list_for_each_entry(rd, &f->remotes, list) {
558 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
559 rd->remote_port == port &&
560 rd->remote_vni == vni &&
561 rd->remote_ifindex == ifindex)
568 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
569 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
571 struct vxlan_dev *vxlan = netdev_priv(dev);
572 u8 eth_addr[ETH_ALEN + 2] = { 0 };
573 struct vxlan_rdst *rdst;
577 if (is_multicast_ether_addr(mac) ||
578 is_zero_ether_addr(mac))
581 ether_addr_copy(eth_addr, mac);
585 f = __vxlan_find_mac(vxlan, eth_addr, vni);
591 rdst = first_remote_rcu(f);
592 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
598 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
600 static int vxlan_fdb_notify_one(struct notifier_block *nb,
601 const struct vxlan_dev *vxlan,
602 const struct vxlan_fdb *f,
603 const struct vxlan_rdst *rdst,
604 struct netlink_ext_ack *extack)
606 struct switchdev_notifier_vxlan_fdb_info fdb_info;
609 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
610 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
612 return notifier_to_errno(rc);
615 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
616 struct notifier_block *nb,
617 struct netlink_ext_ack *extack)
619 struct vxlan_dev *vxlan;
620 struct vxlan_rdst *rdst;
625 if (!netif_is_vxlan(dev))
627 vxlan = netdev_priv(dev);
629 for (h = 0; h < FDB_HASH_SIZE; ++h) {
630 spin_lock_bh(&vxlan->hash_lock[h]);
631 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
633 list_for_each_entry(rdst, &f->remotes, list) {
634 rc = vxlan_fdb_notify_one(nb, vxlan,
642 spin_unlock_bh(&vxlan->hash_lock[h]);
647 spin_unlock_bh(&vxlan->hash_lock[h]);
650 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
652 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
654 struct vxlan_dev *vxlan;
655 struct vxlan_rdst *rdst;
659 if (!netif_is_vxlan(dev))
661 vxlan = netdev_priv(dev);
663 for (h = 0; h < FDB_HASH_SIZE; ++h) {
664 spin_lock_bh(&vxlan->hash_lock[h]);
665 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
667 list_for_each_entry(rdst, &f->remotes, list)
668 rdst->offloaded = false;
669 spin_unlock_bh(&vxlan->hash_lock[h]);
673 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
675 /* Replace destination of unicast mac */
676 static int vxlan_fdb_replace(struct vxlan_fdb *f,
677 union vxlan_addr *ip, __be16 port, __be32 vni,
678 __u32 ifindex, struct vxlan_rdst *oldrd)
680 struct vxlan_rdst *rd;
682 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
686 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
691 dst_cache_reset(&rd->dst_cache);
693 rd->remote_port = port;
694 rd->remote_vni = vni;
695 rd->remote_ifindex = ifindex;
696 rd->offloaded = false;
700 /* Add/update destinations for multicast */
701 static int vxlan_fdb_append(struct vxlan_fdb *f,
702 union vxlan_addr *ip, __be16 port, __be32 vni,
703 __u32 ifindex, struct vxlan_rdst **rdp)
705 struct vxlan_rdst *rd;
707 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
711 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
715 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
721 rd->remote_port = port;
722 rd->offloaded = false;
723 rd->remote_vni = vni;
724 rd->remote_ifindex = ifindex;
726 list_add_tail_rcu(&rd->list, &f->remotes);
732 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol)
734 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr;
736 /* Need to have Next Protocol set for interfaces in GPE mode. */
737 if (!gpe->np_applied)
739 /* "The initial version is 0. If a receiver does not support the
740 * version indicated it MUST drop the packet.
742 if (gpe->version != 0)
744 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
745 * processing MUST occur." However, we don't implement OAM
746 * processing, thus drop the packet.
751 *protocol = tun_p_to_eth_p(gpe->next_protocol);
758 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
760 struct vxlanhdr *vh, size_t hdrlen,
762 struct gro_remcsum *grc,
765 size_t start, offset;
767 if (skb->remcsum_offload)
770 if (!NAPI_GRO_CB(skb)->csum_valid)
773 start = vxlan_rco_start(vni_field);
774 offset = start + vxlan_rco_offset(vni_field);
776 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
777 start, offset, grc, nopartial);
779 skb->remcsum_offload = 1;
784 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
785 struct list_head *head,
788 struct sk_buff *pp = NULL;
790 struct vxlanhdr *vh, *vh2;
791 unsigned int hlen, off_vx;
793 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
795 struct gro_remcsum grc;
797 skb_gro_remcsum_init(&grc);
799 off_vx = skb_gro_offset(skb);
800 hlen = off_vx + sizeof(*vh);
801 vh = skb_gro_header_fast(skb, off_vx);
802 if (skb_gro_header_hard(skb, hlen)) {
803 vh = skb_gro_header_slow(skb, hlen, off_vx);
808 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
810 flags = vh->vx_flags;
812 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
813 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
816 VXLAN_F_REMCSUM_NOPARTIAL));
822 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
824 list_for_each_entry(p, head, list) {
825 if (!NAPI_GRO_CB(p)->same_flow)
828 vh2 = (struct vxlanhdr *)(p->data + off_vx);
829 if (vh->vx_flags != vh2->vx_flags ||
830 vh->vx_vni != vh2->vx_vni) {
831 NAPI_GRO_CB(p)->same_flow = 0;
836 pp = call_gro_receive(eth_gro_receive, head, skb);
840 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
845 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
847 /* Sets 'skb->inner_mac_header' since we are always called with
848 * 'skb->encapsulation' set.
850 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
853 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
854 __u16 state, __be32 src_vni,
859 f = kmalloc(sizeof(*f), GFP_ATOMIC);
863 f->flags = ndm_flags;
864 f->updated = f->used = jiffies;
867 RCU_INIT_POINTER(f->vdev, vxlan);
868 INIT_LIST_HEAD(&f->nh_list);
869 INIT_LIST_HEAD(&f->remotes);
870 memcpy(f->eth_addr, mac, ETH_ALEN);
875 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
876 __be32 src_vni, struct vxlan_fdb *f)
879 hlist_add_head_rcu(&f->hlist,
880 vxlan_fdb_head(vxlan, mac, src_vni));
883 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
884 u32 nhid, struct netlink_ext_ack *extack)
886 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
890 if (old_nh && old_nh->id == nhid)
893 nh = nexthop_find_by_id(vxlan->net, nhid);
895 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
900 if (!nexthop_get(nh)) {
901 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
905 if (!nexthop_is_fdb(nh)) {
906 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
910 if (!nexthop_is_multipath(nh)) {
911 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
915 /* check nexthop group family */
916 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
918 if (!nexthop_has_v4(nh)) {
920 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
925 if (nexthop_has_v4(nh)) {
927 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
934 list_del_rcu(&fdb->nh_list);
937 rcu_assign_pointer(fdb->nh, nh);
938 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
947 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
948 const u8 *mac, union vxlan_addr *ip,
949 __u16 state, __be16 port, __be32 src_vni,
950 __be32 vni, __u32 ifindex, __u16 ndm_flags,
951 u32 nhid, struct vxlan_fdb **fdb,
952 struct netlink_ext_ack *extack)
954 struct vxlan_rdst *rd = NULL;
958 if (vxlan->cfg.addrmax &&
959 vxlan->addrcnt >= vxlan->cfg.addrmax)
962 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
963 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
968 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
970 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
983 static void __vxlan_fdb_free(struct vxlan_fdb *f)
985 struct vxlan_rdst *rd, *nd;
988 nh = rcu_dereference_raw(f->nh);
990 rcu_assign_pointer(f->nh, NULL);
991 rcu_assign_pointer(f->vdev, NULL);
995 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
996 dst_cache_destroy(&rd->dst_cache);
1002 static void vxlan_fdb_free(struct rcu_head *head)
1004 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
1006 __vxlan_fdb_free(f);
1009 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1010 bool do_notify, bool swdev_notify)
1012 struct vxlan_rdst *rd;
1014 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
1018 if (rcu_access_pointer(f->nh))
1019 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
1020 swdev_notify, NULL);
1022 list_for_each_entry(rd, &f->remotes, list)
1023 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
1024 swdev_notify, NULL);
1027 hlist_del_rcu(&f->hlist);
1028 list_del_rcu(&f->nh_list);
1029 call_rcu(&f->rcu, vxlan_fdb_free);
1032 static void vxlan_dst_free(struct rcu_head *head)
1034 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1036 dst_cache_destroy(&rd->dst_cache);
1040 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1041 union vxlan_addr *ip,
1042 __u16 state, __u16 flags,
1043 __be16 port, __be32 vni,
1044 __u32 ifindex, __u16 ndm_flags,
1045 struct vxlan_fdb *f, u32 nhid,
1047 struct netlink_ext_ack *extack)
1049 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1050 struct vxlan_rdst *rd = NULL;
1051 struct vxlan_rdst oldrd;
1056 if (nhid && !rcu_access_pointer(f->nh)) {
1057 NL_SET_ERR_MSG(extack,
1058 "Cannot replace an existing non nexthop fdb with a nexthop");
1062 if (nhid && (flags & NLM_F_APPEND)) {
1063 NL_SET_ERR_MSG(extack,
1064 "Cannot append to a nexthop fdb");
1068 /* Do not allow an externally learned entry to take over an entry added
1071 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1072 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1073 if (f->state != state) {
1075 f->updated = jiffies;
1078 if (f->flags != fdb_flags) {
1079 f->flags = fdb_flags;
1080 f->updated = jiffies;
1085 if ((flags & NLM_F_REPLACE)) {
1086 /* Only change unicasts */
1087 if (!(is_multicast_ether_addr(f->eth_addr) ||
1088 is_zero_ether_addr(f->eth_addr))) {
1090 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1094 rc = vxlan_fdb_replace(f, ip, port, vni,
1099 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1103 if ((flags & NLM_F_APPEND) &&
1104 (is_multicast_ether_addr(f->eth_addr) ||
1105 is_zero_ether_addr(f->eth_addr))) {
1106 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1113 if (ndm_flags & NTF_USE)
1118 rd = first_remote_rtnl(f);
1120 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1121 swdev_notify, extack);
1131 if ((flags & NLM_F_REPLACE) && rc)
1133 else if ((flags & NLM_F_APPEND) && rc) {
1134 list_del_rcu(&rd->list);
1135 call_rcu(&rd->rcu, vxlan_dst_free);
1140 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1141 const u8 *mac, union vxlan_addr *ip,
1142 __u16 state, __u16 flags,
1143 __be16 port, __be32 src_vni, __be32 vni,
1144 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1146 struct netlink_ext_ack *extack)
1148 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1149 struct vxlan_fdb *f;
1152 /* Disallow replace to add a multicast entry */
1153 if ((flags & NLM_F_REPLACE) &&
1154 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1157 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1158 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1159 vni, ifindex, fdb_flags, nhid, &f, extack);
1163 vxlan_fdb_insert(vxlan, mac, src_vni, f);
1164 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1165 swdev_notify, extack);
1172 vxlan_fdb_destroy(vxlan, f, false, false);
1176 /* Add new entry to forwarding table -- assumes lock held */
1177 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1178 const u8 *mac, union vxlan_addr *ip,
1179 __u16 state, __u16 flags,
1180 __be16 port, __be32 src_vni, __be32 vni,
1181 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1183 struct netlink_ext_ack *extack)
1185 struct vxlan_fdb *f;
1187 f = __vxlan_find_mac(vxlan, mac, src_vni);
1189 if (flags & NLM_F_EXCL) {
1190 netdev_dbg(vxlan->dev,
1191 "lost race to create %pM\n", mac);
1195 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1196 vni, ifindex, ndm_flags, f,
1197 nhid, swdev_notify, extack);
1199 if (!(flags & NLM_F_CREATE))
1202 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1203 port, src_vni, vni, ifindex,
1204 ndm_flags, nhid, swdev_notify,
1209 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1210 struct vxlan_rdst *rd, bool swdev_notify)
1212 list_del_rcu(&rd->list);
1213 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1214 call_rcu(&rd->rcu, vxlan_dst_free);
1217 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1218 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1219 __be32 *vni, u32 *ifindex, u32 *nhid)
1221 struct net *net = dev_net(vxlan->dev);
1224 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1229 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1233 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1235 if (remote->sa.sa_family == AF_INET) {
1236 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1237 ip->sa.sa_family = AF_INET;
1238 #if IS_ENABLED(CONFIG_IPV6)
1240 ip->sin6.sin6_addr = in6addr_any;
1241 ip->sa.sa_family = AF_INET6;
1247 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1249 *port = nla_get_be16(tb[NDA_PORT]);
1251 *port = vxlan->cfg.dst_port;
1255 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1257 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1259 *vni = vxlan->default_dst.remote_vni;
1262 if (tb[NDA_SRC_VNI]) {
1263 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1265 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1267 *src_vni = vxlan->default_dst.remote_vni;
1270 if (tb[NDA_IFINDEX]) {
1271 struct net_device *tdev;
1273 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1275 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1276 tdev = __dev_get_by_index(net, *ifindex);
1278 return -EADDRNOTAVAIL;
1284 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1291 /* Add static entry (via netlink) */
1292 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1293 struct net_device *dev,
1294 const unsigned char *addr, u16 vid, u16 flags,
1295 struct netlink_ext_ack *extack)
1297 struct vxlan_dev *vxlan = netdev_priv(dev);
1298 /* struct net *net = dev_net(vxlan->dev); */
1299 union vxlan_addr ip;
1301 __be32 src_vni, vni;
1306 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1307 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1312 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1315 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1320 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1321 return -EAFNOSUPPORT;
1323 hash_index = fdb_head_index(vxlan, addr, src_vni);
1324 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1325 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1326 port, src_vni, vni, ifindex,
1327 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1328 nhid, true, extack);
1329 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1334 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1335 const unsigned char *addr, union vxlan_addr ip,
1336 __be16 port, __be32 src_vni, __be32 vni,
1337 u32 ifindex, bool swdev_notify)
1339 struct vxlan_rdst *rd = NULL;
1340 struct vxlan_fdb *f;
1343 f = vxlan_find_mac(vxlan, addr, src_vni);
1347 if (!vxlan_addr_any(&ip)) {
1348 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1353 /* remove a destination if it's not the only one on the list,
1354 * otherwise destroy the fdb entry
1356 if (rd && !list_is_singular(&f->remotes)) {
1357 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1361 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1367 /* Delete entry (via netlink) */
1368 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1369 struct net_device *dev,
1370 const unsigned char *addr, u16 vid)
1372 struct vxlan_dev *vxlan = netdev_priv(dev);
1373 union vxlan_addr ip;
1374 __be32 src_vni, vni;
1380 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1385 hash_index = fdb_head_index(vxlan, addr, src_vni);
1386 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1387 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1389 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1394 /* Dump forwarding table */
1395 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1396 struct net_device *dev,
1397 struct net_device *filter_dev, int *idx)
1399 struct vxlan_dev *vxlan = netdev_priv(dev);
1403 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1404 struct vxlan_fdb *f;
1407 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1408 struct vxlan_rdst *rd;
1410 if (rcu_access_pointer(f->nh)) {
1411 if (*idx < cb->args[2])
1413 err = vxlan_fdb_info(skb, vxlan, f,
1414 NETLINK_CB(cb->skb).portid,
1427 list_for_each_entry_rcu(rd, &f->remotes, list) {
1428 if (*idx < cb->args[2])
1431 err = vxlan_fdb_info(skb, vxlan, f,
1432 NETLINK_CB(cb->skb).portid,
1450 static int vxlan_fdb_get(struct sk_buff *skb,
1451 struct nlattr *tb[],
1452 struct net_device *dev,
1453 const unsigned char *addr,
1454 u16 vid, u32 portid, u32 seq,
1455 struct netlink_ext_ack *extack)
1457 struct vxlan_dev *vxlan = netdev_priv(dev);
1458 struct vxlan_fdb *f;
1463 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1465 vni = vxlan->default_dst.remote_vni;
1469 f = __vxlan_find_mac(vxlan, addr, vni);
1471 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1476 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1477 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1483 /* Watch incoming packets to learn mapping between Ethernet address
1484 * and Tunnel endpoint.
1485 * Return true if packet is bogus and should be dropped.
1487 static bool vxlan_snoop(struct net_device *dev,
1488 union vxlan_addr *src_ip, const u8 *src_mac,
1489 u32 src_ifindex, __be32 vni)
1491 struct vxlan_dev *vxlan = netdev_priv(dev);
1492 struct vxlan_fdb *f;
1495 #if IS_ENABLED(CONFIG_IPV6)
1496 if (src_ip->sa.sa_family == AF_INET6 &&
1497 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1498 ifindex = src_ifindex;
1501 f = vxlan_find_mac(vxlan, src_mac, vni);
1503 struct vxlan_rdst *rdst = first_remote_rcu(f);
1505 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1506 rdst->remote_ifindex == ifindex))
1509 /* Don't migrate static entries, drop packets */
1510 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1513 /* Don't override an fdb with nexthop with a learnt entry */
1514 if (rcu_access_pointer(f->nh))
1517 if (net_ratelimit())
1519 "%pM migrated from %pIS to %pIS\n",
1520 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1522 rdst->remote_ip = *src_ip;
1523 f->updated = jiffies;
1524 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1526 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1528 /* learned new entry */
1529 spin_lock(&vxlan->hash_lock[hash_index]);
1531 /* close off race between vxlan_flush and incoming packets */
1532 if (netif_running(dev))
1533 vxlan_fdb_update(vxlan, src_mac, src_ip,
1535 NLM_F_EXCL|NLM_F_CREATE,
1536 vxlan->cfg.dst_port,
1538 vxlan->default_dst.remote_vni,
1539 ifindex, NTF_SELF, 0, true, NULL);
1540 spin_unlock(&vxlan->hash_lock[hash_index]);
1546 /* See if multicast group is already in use by other ID */
1547 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1549 struct vxlan_dev *vxlan;
1550 struct vxlan_sock *sock4;
1551 #if IS_ENABLED(CONFIG_IPV6)
1552 struct vxlan_sock *sock6;
1554 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1556 sock4 = rtnl_dereference(dev->vn4_sock);
1558 /* The vxlan_sock is only used by dev, leaving group has
1559 * no effect on other vxlan devices.
1561 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1563 #if IS_ENABLED(CONFIG_IPV6)
1564 sock6 = rtnl_dereference(dev->vn6_sock);
1565 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1569 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1570 if (!netif_running(vxlan->dev) || vxlan == dev)
1573 if (family == AF_INET &&
1574 rtnl_dereference(vxlan->vn4_sock) != sock4)
1576 #if IS_ENABLED(CONFIG_IPV6)
1577 if (family == AF_INET6 &&
1578 rtnl_dereference(vxlan->vn6_sock) != sock6)
1582 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1583 &dev->default_dst.remote_ip))
1586 if (vxlan->default_dst.remote_ifindex !=
1587 dev->default_dst.remote_ifindex)
1596 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1598 struct vxlan_net *vn;
1602 if (!refcount_dec_and_test(&vs->refcnt))
1605 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1606 spin_lock(&vn->sock_lock);
1607 hlist_del_rcu(&vs->hlist);
1608 udp_tunnel_notify_del_rx_port(vs->sock,
1609 (vs->flags & VXLAN_F_GPE) ?
1610 UDP_TUNNEL_TYPE_VXLAN_GPE :
1611 UDP_TUNNEL_TYPE_VXLAN);
1612 spin_unlock(&vn->sock_lock);
1617 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1619 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1620 #if IS_ENABLED(CONFIG_IPV6)
1621 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1623 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1626 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1629 vxlan_vs_del_dev(vxlan);
1631 if (__vxlan_sock_release_prep(sock4)) {
1632 udp_tunnel_sock_release(sock4->sock);
1636 #if IS_ENABLED(CONFIG_IPV6)
1637 if (__vxlan_sock_release_prep(sock6)) {
1638 udp_tunnel_sock_release(sock6->sock);
1644 /* Update multicast group membership when first VNI on
1645 * multicast address is brought up
1647 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1650 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1651 int ifindex = vxlan->default_dst.remote_ifindex;
1654 if (ip->sa.sa_family == AF_INET) {
1655 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1656 struct ip_mreqn mreq = {
1657 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1658 .imr_ifindex = ifindex,
1661 sk = sock4->sock->sk;
1663 ret = ip_mc_join_group(sk, &mreq);
1665 #if IS_ENABLED(CONFIG_IPV6)
1667 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1669 sk = sock6->sock->sk;
1671 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1672 &ip->sin6.sin6_addr);
1680 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1681 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1684 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1685 int ifindex = vxlan->default_dst.remote_ifindex;
1688 if (ip->sa.sa_family == AF_INET) {
1689 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1690 struct ip_mreqn mreq = {
1691 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1692 .imr_ifindex = ifindex,
1695 sk = sock4->sock->sk;
1697 ret = ip_mc_leave_group(sk, &mreq);
1699 #if IS_ENABLED(CONFIG_IPV6)
1701 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1703 sk = sock6->sock->sk;
1705 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1706 &ip->sin6.sin6_addr);
1714 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1715 struct sk_buff *skb, u32 vxflags)
1717 size_t start, offset;
1719 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1722 start = vxlan_rco_start(unparsed->vx_vni);
1723 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1725 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1728 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1729 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1731 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1732 unparsed->vx_vni &= VXLAN_VNI_MASK;
1736 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1737 struct sk_buff *skb, u32 vxflags,
1738 struct vxlan_metadata *md)
1740 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1741 struct metadata_dst *tun_dst;
1743 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1746 md->gbp = ntohs(gbp->policy_id);
1748 tun_dst = (struct metadata_dst *)skb_dst(skb);
1750 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1751 tun_dst->u.tun_info.options_len = sizeof(*md);
1753 if (gbp->dont_learn)
1754 md->gbp |= VXLAN_GBP_DONT_LEARN;
1756 if (gbp->policy_applied)
1757 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1759 /* In flow-based mode, GBP is carried in dst_metadata */
1760 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1761 skb->mark = md->gbp;
1763 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1766 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1767 struct vxlan_sock *vs,
1768 struct sk_buff *skb, __be32 vni)
1770 union vxlan_addr saddr;
1771 u32 ifindex = skb->dev->ifindex;
1773 skb_reset_mac_header(skb);
1774 skb->protocol = eth_type_trans(skb, vxlan->dev);
1775 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1777 /* Ignore packet loops (and multicast echo) */
1778 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1781 /* Get address from the outer IP header */
1782 if (vxlan_get_sk_family(vs) == AF_INET) {
1783 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1784 saddr.sa.sa_family = AF_INET;
1785 #if IS_ENABLED(CONFIG_IPV6)
1787 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1788 saddr.sa.sa_family = AF_INET6;
1792 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1793 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1799 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1800 struct sk_buff *skb)
1804 if (vxlan_get_sk_family(vs) == AF_INET)
1805 err = IP_ECN_decapsulate(oiph, skb);
1806 #if IS_ENABLED(CONFIG_IPV6)
1808 err = IP6_ECN_decapsulate(oiph, skb);
1811 if (unlikely(err) && log_ecn_error) {
1812 if (vxlan_get_sk_family(vs) == AF_INET)
1813 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1814 &((struct iphdr *)oiph)->saddr,
1815 ((struct iphdr *)oiph)->tos);
1817 net_info_ratelimited("non-ECT from %pI6\n",
1818 &((struct ipv6hdr *)oiph)->saddr);
1823 /* Callback from net/ipv4/udp.c to receive packets */
1824 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1826 struct vxlan_dev *vxlan;
1827 struct vxlan_sock *vs;
1828 struct vxlanhdr unparsed;
1829 struct vxlan_metadata _md;
1830 struct vxlan_metadata *md = &_md;
1831 __be16 protocol = htons(ETH_P_TEB);
1832 bool raw_proto = false;
1836 /* Need UDP and VXLAN header to be present */
1837 if (!pskb_may_pull(skb, VXLAN_HLEN))
1840 unparsed = *vxlan_hdr(skb);
1841 /* VNI flag always required to be set */
1842 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1843 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1844 ntohl(vxlan_hdr(skb)->vx_flags),
1845 ntohl(vxlan_hdr(skb)->vx_vni));
1846 /* Return non vxlan pkt */
1849 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1850 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1852 vs = rcu_dereference_sk_user_data(sk);
1856 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1858 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1862 /* For backwards compatibility, only allow reserved fields to be
1863 * used by VXLAN extensions if explicitly requested.
1865 if (vs->flags & VXLAN_F_GPE) {
1866 if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1868 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1872 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1873 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1876 if (vs->flags & VXLAN_F_REMCSUM_RX)
1877 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1880 if (vxlan_collect_metadata(vs)) {
1881 struct metadata_dst *tun_dst;
1883 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1884 key32_to_tunnel_id(vni), sizeof(*md));
1889 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1891 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1893 memset(md, 0, sizeof(*md));
1896 if (vs->flags & VXLAN_F_GBP)
1897 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1898 /* Note that GBP and GPE can never be active together. This is
1899 * ensured in vxlan_dev_configure.
1902 if (unparsed.vx_flags || unparsed.vx_vni) {
1903 /* If there are any unprocessed flags remaining treat
1904 * this as a malformed packet. This behavior diverges from
1905 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1906 * in reserved fields are to be ignored. The approach here
1907 * maintains compatibility with previous stack code, and also
1908 * is more robust and provides a little more security in
1909 * adding extensions to VXLAN.
1915 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1918 skb_reset_mac_header(skb);
1919 skb->dev = vxlan->dev;
1920 skb->pkt_type = PACKET_HOST;
1923 oiph = skb_network_header(skb);
1924 skb_reset_network_header(skb);
1926 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1927 ++vxlan->dev->stats.rx_frame_errors;
1928 ++vxlan->dev->stats.rx_errors;
1934 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1936 atomic_long_inc(&vxlan->dev->rx_dropped);
1940 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1941 gro_cells_receive(&vxlan->gro_cells, skb);
1948 /* Consume bad packet */
1953 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1954 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1956 struct vxlan_dev *vxlan;
1957 struct vxlan_sock *vs;
1958 struct vxlanhdr *hdr;
1961 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1964 hdr = vxlan_hdr(skb);
1966 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1969 vs = rcu_dereference_sk_user_data(sk);
1973 vni = vxlan_vni(hdr->vx_vni);
1974 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1981 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1983 struct vxlan_dev *vxlan = netdev_priv(dev);
1984 struct arphdr *parp;
1987 struct neighbour *n;
1989 if (dev->flags & IFF_NOARP)
1992 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1993 dev->stats.tx_dropped++;
1996 parp = arp_hdr(skb);
1998 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1999 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2000 parp->ar_pro != htons(ETH_P_IP) ||
2001 parp->ar_op != htons(ARPOP_REQUEST) ||
2002 parp->ar_hln != dev->addr_len ||
2005 arpptr = (u8 *)parp + sizeof(struct arphdr);
2007 arpptr += dev->addr_len; /* sha */
2008 memcpy(&sip, arpptr, sizeof(sip));
2009 arpptr += sizeof(sip);
2010 arpptr += dev->addr_len; /* tha */
2011 memcpy(&tip, arpptr, sizeof(tip));
2013 if (ipv4_is_loopback(tip) ||
2014 ipv4_is_multicast(tip))
2017 n = neigh_lookup(&arp_tbl, &tip, dev);
2020 struct vxlan_fdb *f;
2021 struct sk_buff *reply;
2023 if (!(n->nud_state & NUD_CONNECTED)) {
2028 f = vxlan_find_mac(vxlan, n->ha, vni);
2029 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2030 /* bridge-local neighbor */
2035 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2043 skb_reset_mac_header(reply);
2044 __skb_pull(reply, skb_network_offset(reply));
2045 reply->ip_summed = CHECKSUM_UNNECESSARY;
2046 reply->pkt_type = PACKET_HOST;
2048 if (netif_rx_ni(reply) == NET_RX_DROP)
2049 dev->stats.rx_dropped++;
2050 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2051 union vxlan_addr ipa = {
2052 .sin.sin_addr.s_addr = tip,
2053 .sin.sin_family = AF_INET,
2056 vxlan_ip_miss(dev, &ipa);
2060 return NETDEV_TX_OK;
2063 #if IS_ENABLED(CONFIG_IPV6)
2064 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2065 struct neighbour *n, bool isrouter)
2067 struct net_device *dev = request->dev;
2068 struct sk_buff *reply;
2069 struct nd_msg *ns, *na;
2070 struct ipv6hdr *pip6;
2072 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2076 if (dev == NULL || !pskb_may_pull(request, request->len))
2079 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2080 sizeof(*na) + na_olen + dev->needed_tailroom;
2081 reply = alloc_skb(len, GFP_ATOMIC);
2085 reply->protocol = htons(ETH_P_IPV6);
2087 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2088 skb_push(reply, sizeof(struct ethhdr));
2089 skb_reset_mac_header(reply);
2091 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2093 daddr = eth_hdr(request)->h_source;
2094 ns_olen = request->len - skb_network_offset(request) -
2095 sizeof(struct ipv6hdr) - sizeof(*ns);
2096 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2097 if (!ns->opt[i + 1]) {
2101 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2102 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2107 /* Ethernet header */
2108 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2109 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2110 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2111 reply->protocol = htons(ETH_P_IPV6);
2113 skb_pull(reply, sizeof(struct ethhdr));
2114 skb_reset_network_header(reply);
2115 skb_put(reply, sizeof(struct ipv6hdr));
2119 pip6 = ipv6_hdr(reply);
2120 memset(pip6, 0, sizeof(struct ipv6hdr));
2122 pip6->priority = ipv6_hdr(request)->priority;
2123 pip6->nexthdr = IPPROTO_ICMPV6;
2124 pip6->hop_limit = 255;
2125 pip6->daddr = ipv6_hdr(request)->saddr;
2126 pip6->saddr = *(struct in6_addr *)n->primary_key;
2128 skb_pull(reply, sizeof(struct ipv6hdr));
2129 skb_reset_transport_header(reply);
2131 /* Neighbor Advertisement */
2132 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2133 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2134 na->icmph.icmp6_router = isrouter;
2135 na->icmph.icmp6_override = 1;
2136 na->icmph.icmp6_solicited = 1;
2137 na->target = ns->target;
2138 ether_addr_copy(&na->opt[2], n->ha);
2139 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2140 na->opt[1] = na_olen >> 3;
2142 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2143 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2144 csum_partial(na, sizeof(*na)+na_olen, 0));
2146 pip6->payload_len = htons(sizeof(*na)+na_olen);
2148 skb_push(reply, sizeof(struct ipv6hdr));
2150 reply->ip_summed = CHECKSUM_UNNECESSARY;
2155 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2157 struct vxlan_dev *vxlan = netdev_priv(dev);
2158 const struct in6_addr *daddr;
2159 const struct ipv6hdr *iphdr;
2160 struct inet6_dev *in6_dev;
2161 struct neighbour *n;
2165 in6_dev = __in6_dev_get(dev);
2169 iphdr = ipv6_hdr(skb);
2170 daddr = &iphdr->daddr;
2171 msg = (struct nd_msg *)(iphdr + 1);
2173 if (ipv6_addr_loopback(daddr) ||
2174 ipv6_addr_is_multicast(&msg->target))
2177 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2180 struct vxlan_fdb *f;
2181 struct sk_buff *reply;
2183 if (!(n->nud_state & NUD_CONNECTED)) {
2188 f = vxlan_find_mac(vxlan, n->ha, vni);
2189 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2190 /* bridge-local neighbor */
2195 reply = vxlan_na_create(skb, n,
2196 !!(f ? f->flags & NTF_ROUTER : 0));
2203 if (netif_rx_ni(reply) == NET_RX_DROP)
2204 dev->stats.rx_dropped++;
2206 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2207 union vxlan_addr ipa = {
2208 .sin6.sin6_addr = msg->target,
2209 .sin6.sin6_family = AF_INET6,
2212 vxlan_ip_miss(dev, &ipa);
2218 return NETDEV_TX_OK;
2222 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2224 struct vxlan_dev *vxlan = netdev_priv(dev);
2225 struct neighbour *n;
2227 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2231 switch (ntohs(eth_hdr(skb)->h_proto)) {
2236 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2239 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2240 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2241 union vxlan_addr ipa = {
2242 .sin.sin_addr.s_addr = pip->daddr,
2243 .sin.sin_family = AF_INET,
2246 vxlan_ip_miss(dev, &ipa);
2252 #if IS_ENABLED(CONFIG_IPV6)
2255 struct ipv6hdr *pip6;
2257 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2259 pip6 = ipv6_hdr(skb);
2260 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2261 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2262 union vxlan_addr ipa = {
2263 .sin6.sin6_addr = pip6->daddr,
2264 .sin6.sin6_family = AF_INET6,
2267 vxlan_ip_miss(dev, &ipa);
2281 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2283 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2285 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2294 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2295 struct vxlan_metadata *md)
2297 struct vxlanhdr_gbp *gbp;
2302 gbp = (struct vxlanhdr_gbp *)vxh;
2303 vxh->vx_flags |= VXLAN_HF_GBP;
2305 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2306 gbp->dont_learn = 1;
2308 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2309 gbp->policy_applied = 1;
2311 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2314 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2317 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2319 gpe->np_applied = 1;
2320 gpe->next_protocol = tun_p_from_eth_p(protocol);
2321 if (!gpe->next_protocol)
2322 return -EPFNOSUPPORT;
2326 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2327 int iphdr_len, __be32 vni,
2328 struct vxlan_metadata *md, u32 vxflags,
2331 struct vxlanhdr *vxh;
2334 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2335 __be16 inner_protocol = htons(ETH_P_TEB);
2337 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2338 skb->ip_summed == CHECKSUM_PARTIAL) {
2339 int csum_start = skb_checksum_start_offset(skb);
2341 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2342 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2343 (skb->csum_offset == offsetof(struct udphdr, check) ||
2344 skb->csum_offset == offsetof(struct tcphdr, check)))
2345 type |= SKB_GSO_TUNNEL_REMCSUM;
2348 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2349 + VXLAN_HLEN + iphdr_len;
2351 /* Need space for new headers (invalidates iph ptr) */
2352 err = skb_cow_head(skb, min_headroom);
2356 err = iptunnel_handle_offloads(skb, type);
2360 vxh = __skb_push(skb, sizeof(*vxh));
2361 vxh->vx_flags = VXLAN_HF_VNI;
2362 vxh->vx_vni = vxlan_vni_field(vni);
2364 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2367 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2368 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2369 vxh->vx_flags |= VXLAN_HF_RCO;
2371 if (!skb_is_gso(skb)) {
2372 skb->ip_summed = CHECKSUM_NONE;
2373 skb->encapsulation = 0;
2377 if (vxflags & VXLAN_F_GBP)
2378 vxlan_build_gbp_hdr(vxh, vxflags, md);
2379 if (vxflags & VXLAN_F_GPE) {
2380 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2383 inner_protocol = skb->protocol;
2386 skb_set_inner_protocol(skb, inner_protocol);
2390 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2391 struct vxlan_sock *sock4,
2392 struct sk_buff *skb, int oif, u8 tos,
2393 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2394 struct dst_cache *dst_cache,
2395 const struct ip_tunnel_info *info)
2397 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2398 struct rtable *rt = NULL;
2402 return ERR_PTR(-EIO);
2407 rt = dst_cache_get_ip4(dst_cache, saddr);
2412 memset(&fl4, 0, sizeof(fl4));
2413 fl4.flowi4_oif = oif;
2414 fl4.flowi4_tos = RT_TOS(tos);
2415 fl4.flowi4_mark = skb->mark;
2416 fl4.flowi4_proto = IPPROTO_UDP;
2419 fl4.fl4_dport = dport;
2420 fl4.fl4_sport = sport;
2422 rt = ip_route_output_key(vxlan->net, &fl4);
2424 if (rt->dst.dev == dev) {
2425 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2427 return ERR_PTR(-ELOOP);
2432 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2434 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2435 return ERR_PTR(-ENETUNREACH);
2440 #if IS_ENABLED(CONFIG_IPV6)
2441 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2442 struct net_device *dev,
2443 struct vxlan_sock *sock6,
2444 struct sk_buff *skb, int oif, u8 tos,
2446 const struct in6_addr *daddr,
2447 struct in6_addr *saddr,
2448 __be16 dport, __be16 sport,
2449 struct dst_cache *dst_cache,
2450 const struct ip_tunnel_info *info)
2452 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2453 struct dst_entry *ndst;
2457 return ERR_PTR(-EIO);
2462 ndst = dst_cache_get_ip6(dst_cache, saddr);
2467 memset(&fl6, 0, sizeof(fl6));
2468 fl6.flowi6_oif = oif;
2471 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2472 fl6.flowi6_mark = skb->mark;
2473 fl6.flowi6_proto = IPPROTO_UDP;
2474 fl6.fl6_dport = dport;
2475 fl6.fl6_sport = sport;
2477 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2479 if (unlikely(IS_ERR(ndst))) {
2480 netdev_dbg(dev, "no route to %pI6\n", daddr);
2481 return ERR_PTR(-ENETUNREACH);
2484 if (unlikely(ndst->dev == dev)) {
2485 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2487 return ERR_PTR(-ELOOP);
2492 dst_cache_set_ip6(dst_cache, ndst, saddr);
2497 /* Bypass encapsulation if the destination is local */
2498 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2499 struct vxlan_dev *dst_vxlan, __be32 vni,
2502 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2503 union vxlan_addr loopback;
2504 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2505 struct net_device *dev;
2508 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2509 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2510 skb->pkt_type = PACKET_HOST;
2511 skb->encapsulation = 0;
2512 skb->dev = dst_vxlan->dev;
2513 __skb_pull(skb, skb_network_offset(skb));
2515 if (remote_ip->sa.sa_family == AF_INET) {
2516 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2517 loopback.sa.sa_family = AF_INET;
2518 #if IS_ENABLED(CONFIG_IPV6)
2520 loopback.sin6.sin6_addr = in6addr_loopback;
2521 loopback.sa.sa_family = AF_INET6;
2527 if (unlikely(!(dev->flags & IFF_UP))) {
2532 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2533 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2535 u64_stats_update_begin(&tx_stats->syncp);
2536 tx_stats->tx_packets++;
2537 tx_stats->tx_bytes += len;
2538 u64_stats_update_end(&tx_stats->syncp);
2540 if (netif_rx(skb) == NET_RX_SUCCESS) {
2541 u64_stats_update_begin(&rx_stats->syncp);
2542 rx_stats->rx_packets++;
2543 rx_stats->rx_bytes += len;
2544 u64_stats_update_end(&rx_stats->syncp);
2547 dev->stats.rx_dropped++;
2552 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2553 struct vxlan_dev *vxlan,
2554 union vxlan_addr *daddr,
2555 __be16 dst_port, int dst_ifindex, __be32 vni,
2556 struct dst_entry *dst,
2559 #if IS_ENABLED(CONFIG_IPV6)
2560 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2561 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2562 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2564 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2566 /* Bypass encapsulation if the destination is local */
2567 if (rt_flags & RTCF_LOCAL &&
2568 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2569 struct vxlan_dev *dst_vxlan;
2572 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2573 daddr->sa.sa_family, dst_port,
2576 dev->stats.tx_errors++;
2581 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2588 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2589 __be32 default_vni, struct vxlan_rdst *rdst,
2592 struct dst_cache *dst_cache;
2593 struct ip_tunnel_info *info;
2594 struct vxlan_dev *vxlan = netdev_priv(dev);
2595 const struct iphdr *old_iph = ip_hdr(skb);
2596 union vxlan_addr *dst;
2597 union vxlan_addr remote_ip, local_ip;
2598 struct vxlan_metadata _md;
2599 struct vxlan_metadata *md = &_md;
2600 __be16 src_port = 0, dst_port;
2601 struct dst_entry *ndst = NULL;
2606 u32 flags = vxlan->cfg.flags;
2607 bool udp_sum = false;
2608 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2610 info = skb_tunnel_info(skb);
2613 dst = &rdst->remote_ip;
2614 if (vxlan_addr_any(dst)) {
2616 /* short-circuited back to local bridge */
2617 vxlan_encap_bypass(skb, vxlan, vxlan,
2624 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2625 vni = (rdst->remote_vni) ? : default_vni;
2626 ifindex = rdst->remote_ifindex;
2627 local_ip = vxlan->cfg.saddr;
2628 dst_cache = &rdst->dst_cache;
2629 md->gbp = skb->mark;
2630 if (flags & VXLAN_F_TTL_INHERIT) {
2631 ttl = ip_tunnel_get_ttl(old_iph, skb);
2633 ttl = vxlan->cfg.ttl;
2634 if (!ttl && vxlan_addr_multicast(dst))
2638 tos = vxlan->cfg.tos;
2640 tos = ip_tunnel_get_dsfield(old_iph, skb);
2642 if (dst->sa.sa_family == AF_INET)
2643 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2645 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2646 label = vxlan->cfg.label;
2649 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2653 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2654 if (remote_ip.sa.sa_family == AF_INET) {
2655 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2656 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2658 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2659 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2662 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2663 vni = tunnel_id_to_key32(info->key.tun_id);
2665 dst_cache = &info->dst_cache;
2666 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2667 if (info->options_len < sizeof(*md))
2669 md = ip_tunnel_info_opts(info);
2671 ttl = info->key.ttl;
2672 tos = info->key.tos;
2673 label = info->key.label;
2674 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2676 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2677 vxlan->cfg.port_max, true);
2680 if (dst->sa.sa_family == AF_INET) {
2681 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2686 ifindex = sock4->sock->sk->sk_bound_dev_if;
2688 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2689 dst->sin.sin_addr.s_addr,
2690 &local_ip.sin.sin_addr.s_addr,
2699 /* Bypass encapsulation if the destination is local */
2700 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2701 dst_port, ifindex, vni,
2702 &rt->dst, rt->rt_flags);
2706 if (vxlan->cfg.df == VXLAN_DF_SET) {
2708 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2709 struct ethhdr *eth = eth_hdr(skb);
2711 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2712 (ntohs(eth->h_proto) == ETH_P_IP &&
2713 old_iph->frag_off & htons(IP_DF)))
2716 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2721 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2722 netif_is_any_bridge_port(dev));
2727 struct ip_tunnel_info *unclone;
2728 struct in_addr src, dst;
2730 unclone = skb_tunnel_info_unclone(skb);
2731 if (unlikely(!unclone))
2734 src = remote_ip.sin.sin_addr;
2735 dst = local_ip.sin.sin_addr;
2736 unclone->key.u.ipv4.src = src.s_addr;
2737 unclone->key.u.ipv4.dst = dst.s_addr;
2739 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2744 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2745 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2746 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2747 vni, md, flags, udp_sum);
2751 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2752 dst->sin.sin_addr.s_addr, tos, ttl, df,
2753 src_port, dst_port, xnet, !udp_sum);
2754 #if IS_ENABLED(CONFIG_IPV6)
2756 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2759 ifindex = sock6->sock->sk->sk_bound_dev_if;
2761 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2762 label, &dst->sin6.sin6_addr,
2763 &local_ip.sin6.sin6_addr,
2767 err = PTR_ERR(ndst);
2773 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2775 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2776 dst_port, ifindex, vni,
2782 err = skb_tunnel_check_pmtu(skb, ndst,
2783 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2784 netif_is_any_bridge_port(dev));
2789 struct ip_tunnel_info *unclone;
2790 struct in6_addr src, dst;
2792 unclone = skb_tunnel_info_unclone(skb);
2793 if (unlikely(!unclone))
2796 src = remote_ip.sin6.sin6_addr;
2797 dst = local_ip.sin6.sin6_addr;
2798 unclone->key.u.ipv6.src = src;
2799 unclone->key.u.ipv6.dst = dst;
2802 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2807 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2808 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2809 skb_scrub_packet(skb, xnet);
2810 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2811 vni, md, flags, udp_sum);
2815 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2816 &local_ip.sin6.sin6_addr,
2817 &dst->sin6.sin6_addr, tos, ttl,
2818 label, src_port, dst_port, !udp_sum);
2826 dev->stats.tx_dropped++;
2833 dev->stats.collisions++;
2834 else if (err == -ENETUNREACH)
2835 dev->stats.tx_carrier_errors++;
2837 dev->stats.tx_errors++;
2841 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2842 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2844 struct vxlan_rdst nh_rdst;
2849 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2850 hash = skb_get_hash(skb);
2853 nh = rcu_dereference(f->nh);
2858 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2861 if (likely(do_xmit))
2862 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2869 dev->stats.tx_dropped++;
2873 /* Transmit local packets over Vxlan
2875 * Outer IP header inherits ECN and DF from inner header.
2876 * Outer UDP destination is the VXLAN assigned port.
2877 * source port is based on hash of flow
2879 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2881 struct vxlan_dev *vxlan = netdev_priv(dev);
2882 struct vxlan_rdst *rdst, *fdst = NULL;
2883 const struct ip_tunnel_info *info;
2884 bool did_rsc = false;
2885 struct vxlan_fdb *f;
2889 info = skb_tunnel_info(skb);
2891 skb_reset_mac_header(skb);
2893 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2894 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2895 info->mode & IP_TUNNEL_INFO_TX) {
2896 vni = tunnel_id_to_key32(info->key.tun_id);
2898 if (info && info->mode & IP_TUNNEL_INFO_TX)
2899 vxlan_xmit_one(skb, dev, vni, NULL, false);
2902 return NETDEV_TX_OK;
2906 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2908 if (ntohs(eth->h_proto) == ETH_P_ARP)
2909 return arp_reduce(dev, skb, vni);
2910 #if IS_ENABLED(CONFIG_IPV6)
2911 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2912 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2913 sizeof(struct nd_msg)) &&
2914 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2915 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2917 if (m->icmph.icmp6_code == 0 &&
2918 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2919 return neigh_reduce(dev, skb, vni);
2925 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2928 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2929 (ntohs(eth->h_proto) == ETH_P_IP ||
2930 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2931 did_rsc = route_shortcircuit(dev, skb);
2933 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2937 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2939 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2940 !is_multicast_ether_addr(eth->h_dest))
2941 vxlan_fdb_miss(vxlan, eth->h_dest);
2943 dev->stats.tx_dropped++;
2945 return NETDEV_TX_OK;
2949 if (rcu_access_pointer(f->nh)) {
2950 vxlan_xmit_nh(skb, dev, f,
2951 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2953 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2954 struct sk_buff *skb1;
2960 skb1 = skb_clone(skb, GFP_ATOMIC);
2962 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2965 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2970 return NETDEV_TX_OK;
2973 /* Walk the forwarding table and purge stale entries */
2974 static void vxlan_cleanup(struct timer_list *t)
2976 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2977 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2980 if (!netif_running(vxlan->dev))
2983 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2984 struct hlist_node *p, *n;
2986 spin_lock(&vxlan->hash_lock[h]);
2987 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2989 = container_of(p, struct vxlan_fdb, hlist);
2990 unsigned long timeout;
2992 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2995 if (f->flags & NTF_EXT_LEARNED)
2998 timeout = f->used + vxlan->cfg.age_interval * HZ;
2999 if (time_before_eq(timeout, jiffies)) {
3000 netdev_dbg(vxlan->dev,
3001 "garbage collect %pM\n",
3003 f->state = NUD_STALE;
3004 vxlan_fdb_destroy(vxlan, f, true, true);
3005 } else if (time_before(timeout, next_timer))
3006 next_timer = timeout;
3008 spin_unlock(&vxlan->hash_lock[h]);
3011 mod_timer(&vxlan->age_timer, next_timer);
3014 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3016 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3018 spin_lock(&vn->sock_lock);
3019 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3020 #if IS_ENABLED(CONFIG_IPV6)
3021 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3023 spin_unlock(&vn->sock_lock);
3026 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3027 struct vxlan_dev_node *node)
3029 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3030 __be32 vni = vxlan->default_dst.remote_vni;
3032 node->vxlan = vxlan;
3033 spin_lock(&vn->sock_lock);
3034 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3035 spin_unlock(&vn->sock_lock);
3038 /* Setup stats when device is created */
3039 static int vxlan_init(struct net_device *dev)
3041 struct vxlan_dev *vxlan = netdev_priv(dev);
3044 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3048 err = gro_cells_init(&vxlan->gro_cells, dev);
3050 free_percpu(dev->tstats);
3057 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3059 struct vxlan_fdb *f;
3060 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3062 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3063 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3065 vxlan_fdb_destroy(vxlan, f, true, true);
3066 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3069 static void vxlan_uninit(struct net_device *dev)
3071 struct vxlan_dev *vxlan = netdev_priv(dev);
3073 gro_cells_destroy(&vxlan->gro_cells);
3075 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3077 free_percpu(dev->tstats);
3080 /* Start ageing timer and join group when device is brought up */
3081 static int vxlan_open(struct net_device *dev)
3083 struct vxlan_dev *vxlan = netdev_priv(dev);
3086 ret = vxlan_sock_add(vxlan);
3090 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3091 ret = vxlan_igmp_join(vxlan);
3092 if (ret == -EADDRINUSE)
3095 vxlan_sock_release(vxlan);
3100 if (vxlan->cfg.age_interval)
3101 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3106 /* Purge the forwarding table */
3107 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3111 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3112 struct hlist_node *p, *n;
3114 spin_lock_bh(&vxlan->hash_lock[h]);
3115 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3117 = container_of(p, struct vxlan_fdb, hlist);
3118 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3120 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3121 if (is_zero_ether_addr(f->eth_addr) &&
3122 f->vni == vxlan->cfg.vni)
3124 vxlan_fdb_destroy(vxlan, f, true, true);
3126 spin_unlock_bh(&vxlan->hash_lock[h]);
3130 /* Cleanup timer and forwarding table on shutdown */
3131 static int vxlan_stop(struct net_device *dev)
3133 struct vxlan_dev *vxlan = netdev_priv(dev);
3134 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3137 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3138 !vxlan_group_used(vn, vxlan))
3139 ret = vxlan_igmp_leave(vxlan);
3141 del_timer_sync(&vxlan->age_timer);
3143 vxlan_flush(vxlan, false);
3144 vxlan_sock_release(vxlan);
3149 /* Stub, nothing needs to be done. */
3150 static void vxlan_set_multicast_list(struct net_device *dev)
3154 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3156 struct vxlan_dev *vxlan = netdev_priv(dev);
3157 struct vxlan_rdst *dst = &vxlan->default_dst;
3158 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3159 dst->remote_ifindex);
3161 /* This check is different than dev->max_mtu, because it looks at
3162 * the lowerdev->mtu, rather than the static dev->max_mtu
3165 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3166 if (new_mtu > max_mtu)
3174 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3176 struct vxlan_dev *vxlan = netdev_priv(dev);
3177 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3178 __be16 sport, dport;
3180 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3181 vxlan->cfg.port_max, true);
3182 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3184 if (ip_tunnel_info_af(info) == AF_INET) {
3185 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3188 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3189 info->key.u.ipv4.dst,
3190 &info->key.u.ipv4.src, dport, sport,
3191 &info->dst_cache, info);
3196 #if IS_ENABLED(CONFIG_IPV6)
3197 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3198 struct dst_entry *ndst;
3200 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3201 info->key.label, &info->key.u.ipv6.dst,
3202 &info->key.u.ipv6.src, dport, sport,
3203 &info->dst_cache, info);
3205 return PTR_ERR(ndst);
3207 #else /* !CONFIG_IPV6 */
3208 return -EPFNOSUPPORT;
3211 info->key.tp_src = sport;
3212 info->key.tp_dst = dport;
3216 static const struct net_device_ops vxlan_netdev_ether_ops = {
3217 .ndo_init = vxlan_init,
3218 .ndo_uninit = vxlan_uninit,
3219 .ndo_open = vxlan_open,
3220 .ndo_stop = vxlan_stop,
3221 .ndo_start_xmit = vxlan_xmit,
3222 .ndo_get_stats64 = ip_tunnel_get_stats64,
3223 .ndo_set_rx_mode = vxlan_set_multicast_list,
3224 .ndo_change_mtu = vxlan_change_mtu,
3225 .ndo_validate_addr = eth_validate_addr,
3226 .ndo_set_mac_address = eth_mac_addr,
3227 .ndo_fdb_add = vxlan_fdb_add,
3228 .ndo_fdb_del = vxlan_fdb_delete,
3229 .ndo_fdb_dump = vxlan_fdb_dump,
3230 .ndo_fdb_get = vxlan_fdb_get,
3231 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3232 .ndo_change_proto_down = dev_change_proto_down_generic,
3235 static const struct net_device_ops vxlan_netdev_raw_ops = {
3236 .ndo_init = vxlan_init,
3237 .ndo_uninit = vxlan_uninit,
3238 .ndo_open = vxlan_open,
3239 .ndo_stop = vxlan_stop,
3240 .ndo_start_xmit = vxlan_xmit,
3241 .ndo_get_stats64 = ip_tunnel_get_stats64,
3242 .ndo_change_mtu = vxlan_change_mtu,
3243 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3246 /* Info for udev, that this is a virtual tunnel endpoint */
3247 static struct device_type vxlan_type = {
3251 /* Calls the ndo_udp_tunnel_add of the caller in order to
3252 * supply the listening VXLAN udp ports. Callers are expected
3253 * to implement the ndo_udp_tunnel_add.
3255 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3257 struct vxlan_sock *vs;
3258 struct net *net = dev_net(dev);
3259 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3262 spin_lock(&vn->sock_lock);
3263 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3264 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3265 unsigned short type;
3267 if (vs->flags & VXLAN_F_GPE)
3268 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3270 type = UDP_TUNNEL_TYPE_VXLAN;
3273 udp_tunnel_push_rx_port(dev, vs->sock, type);
3275 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3278 spin_unlock(&vn->sock_lock);
3281 /* Initialize the device structure. */
3282 static void vxlan_setup(struct net_device *dev)
3284 struct vxlan_dev *vxlan = netdev_priv(dev);
3287 eth_hw_addr_random(dev);
3290 dev->needs_free_netdev = true;
3291 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3293 dev->features |= NETIF_F_LLTX;
3294 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3295 dev->features |= NETIF_F_RXCSUM;
3296 dev->features |= NETIF_F_GSO_SOFTWARE;
3298 dev->vlan_features = dev->features;
3299 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3300 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3301 netif_keep_dst(dev);
3302 dev->priv_flags |= IFF_NO_QUEUE;
3304 /* MTU range: 68 - 65535 */
3305 dev->min_mtu = ETH_MIN_MTU;
3306 dev->max_mtu = ETH_MAX_MTU;
3308 INIT_LIST_HEAD(&vxlan->next);
3310 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3314 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3315 spin_lock_init(&vxlan->hash_lock[h]);
3316 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3320 static void vxlan_ether_setup(struct net_device *dev)
3322 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3323 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3324 dev->netdev_ops = &vxlan_netdev_ether_ops;
3327 static void vxlan_raw_setup(struct net_device *dev)
3329 dev->header_ops = NULL;
3330 dev->type = ARPHRD_NONE;
3331 dev->hard_header_len = 0;
3333 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3334 dev->netdev_ops = &vxlan_netdev_raw_ops;
3337 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3338 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3339 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3340 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3341 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3342 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3343 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3344 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3345 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3346 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3347 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3348 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3349 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3350 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3351 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3352 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3353 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3354 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3355 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3356 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3357 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3358 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3359 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3360 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3361 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3362 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3363 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3364 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3365 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3366 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3369 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3370 struct netlink_ext_ack *extack)
3372 if (tb[IFLA_ADDRESS]) {
3373 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3374 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3375 "Provided link layer address is not Ethernet");
3379 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3380 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3381 "Provided Ethernet address is not unicast");
3382 return -EADDRNOTAVAIL;
3387 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3389 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3390 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3391 "MTU must be between 68 and 65535");
3397 NL_SET_ERR_MSG(extack,
3398 "Required attributes not provided to perform the operation");
3402 if (data[IFLA_VXLAN_ID]) {
3403 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3405 if (id >= VXLAN_N_VID) {
3406 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3407 "VXLAN ID must be lower than 16777216");
3412 if (data[IFLA_VXLAN_PORT_RANGE]) {
3413 const struct ifla_vxlan_port_range *p
3414 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3416 if (ntohs(p->high) < ntohs(p->low)) {
3417 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3418 "Invalid source port range");
3423 if (data[IFLA_VXLAN_DF]) {
3424 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3426 if (df < 0 || df > VXLAN_DF_MAX) {
3427 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3428 "Invalid DF attribute");
3436 static void vxlan_get_drvinfo(struct net_device *netdev,
3437 struct ethtool_drvinfo *drvinfo)
3439 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3440 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3443 static int vxlan_get_link_ksettings(struct net_device *dev,
3444 struct ethtool_link_ksettings *cmd)
3446 struct vxlan_dev *vxlan = netdev_priv(dev);
3447 struct vxlan_rdst *dst = &vxlan->default_dst;
3448 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3449 dst->remote_ifindex);
3452 cmd->base.duplex = DUPLEX_UNKNOWN;
3453 cmd->base.port = PORT_OTHER;
3454 cmd->base.speed = SPEED_UNKNOWN;
3459 return __ethtool_get_link_ksettings(lowerdev, cmd);
3462 static const struct ethtool_ops vxlan_ethtool_ops = {
3463 .get_drvinfo = vxlan_get_drvinfo,
3464 .get_link = ethtool_op_get_link,
3465 .get_link_ksettings = vxlan_get_link_ksettings,
3468 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3469 __be16 port, u32 flags, int ifindex)
3471 struct socket *sock;
3472 struct udp_port_cfg udp_conf;
3475 memset(&udp_conf, 0, sizeof(udp_conf));
3478 udp_conf.family = AF_INET6;
3479 udp_conf.use_udp6_rx_checksums =
3480 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3481 udp_conf.ipv6_v6only = 1;
3483 udp_conf.family = AF_INET;
3486 udp_conf.local_udp_port = port;
3487 udp_conf.bind_ifindex = ifindex;
3489 /* Open UDP socket */
3490 err = udp_sock_create(net, &udp_conf, &sock);
3492 return ERR_PTR(err);
3497 /* Create new listen socket if needed */
3498 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3499 __be16 port, u32 flags,
3502 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3503 struct vxlan_sock *vs;
3504 struct socket *sock;
3506 struct udp_tunnel_sock_cfg tunnel_cfg;
3508 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3510 return ERR_PTR(-ENOMEM);
3512 for (h = 0; h < VNI_HASH_SIZE; ++h)
3513 INIT_HLIST_HEAD(&vs->vni_list[h]);
3515 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3518 return ERR_CAST(sock);
3522 refcount_set(&vs->refcnt, 1);
3523 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3525 spin_lock(&vn->sock_lock);
3526 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3527 udp_tunnel_notify_add_rx_port(sock,
3528 (vs->flags & VXLAN_F_GPE) ?
3529 UDP_TUNNEL_TYPE_VXLAN_GPE :
3530 UDP_TUNNEL_TYPE_VXLAN);
3531 spin_unlock(&vn->sock_lock);
3533 /* Mark socket as an encapsulation socket. */
3534 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3535 tunnel_cfg.sk_user_data = vs;
3536 tunnel_cfg.encap_type = 1;
3537 tunnel_cfg.encap_rcv = vxlan_rcv;
3538 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3539 tunnel_cfg.encap_destroy = NULL;
3540 tunnel_cfg.gro_receive = vxlan_gro_receive;
3541 tunnel_cfg.gro_complete = vxlan_gro_complete;
3543 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3548 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3550 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3551 struct vxlan_sock *vs = NULL;
3552 struct vxlan_dev_node *node;
3553 int l3mdev_index = 0;
3555 if (vxlan->cfg.remote_ifindex)
3556 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3557 vxlan->net, vxlan->cfg.remote_ifindex);
3559 if (!vxlan->cfg.no_share) {
3560 spin_lock(&vn->sock_lock);
3561 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3562 vxlan->cfg.dst_port, vxlan->cfg.flags,
3564 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3565 spin_unlock(&vn->sock_lock);
3568 spin_unlock(&vn->sock_lock);
3571 vs = vxlan_socket_create(vxlan->net, ipv6,
3572 vxlan->cfg.dst_port, vxlan->cfg.flags,
3576 #if IS_ENABLED(CONFIG_IPV6)
3578 rcu_assign_pointer(vxlan->vn6_sock, vs);
3579 node = &vxlan->hlist6;
3583 rcu_assign_pointer(vxlan->vn4_sock, vs);
3584 node = &vxlan->hlist4;
3586 vxlan_vs_add_dev(vs, vxlan, node);
3590 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3592 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3593 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3594 bool ipv4 = !ipv6 || metadata;
3597 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3598 #if IS_ENABLED(CONFIG_IPV6)
3599 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3601 ret = __vxlan_sock_add(vxlan, true);
3602 if (ret < 0 && ret != -EAFNOSUPPORT)
3607 ret = __vxlan_sock_add(vxlan, false);
3609 vxlan_sock_release(vxlan);
3613 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3614 struct net_device **lower,
3615 struct vxlan_dev *old,
3616 struct netlink_ext_ack *extack)
3618 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3619 struct vxlan_dev *tmp;
3620 bool use_ipv6 = false;
3622 if (conf->flags & VXLAN_F_GPE) {
3623 /* For now, allow GPE only together with
3624 * COLLECT_METADATA. This can be relaxed later; in such
3625 * case, the other side of the PtP link will have to be
3628 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3629 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3630 NL_SET_ERR_MSG(extack,
3631 "VXLAN GPE does not support this combination of attributes");
3636 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3637 /* Unless IPv6 is explicitly requested, assume IPv4 */
3638 conf->remote_ip.sa.sa_family = AF_INET;
3639 conf->saddr.sa.sa_family = AF_INET;
3640 } else if (!conf->remote_ip.sa.sa_family) {
3641 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3642 } else if (!conf->saddr.sa.sa_family) {
3643 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3646 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3647 NL_SET_ERR_MSG(extack,
3648 "Local and remote address must be from the same family");
3652 if (vxlan_addr_multicast(&conf->saddr)) {
3653 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3657 if (conf->saddr.sa.sa_family == AF_INET6) {
3658 if (!IS_ENABLED(CONFIG_IPV6)) {
3659 NL_SET_ERR_MSG(extack,
3660 "IPv6 support not enabled in the kernel");
3661 return -EPFNOSUPPORT;
3664 conf->flags |= VXLAN_F_IPV6;
3666 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3668 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3670 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3672 if (local_type & IPV6_ADDR_LINKLOCAL) {
3673 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3674 (remote_type != IPV6_ADDR_ANY)) {
3675 NL_SET_ERR_MSG(extack,
3676 "Invalid combination of local and remote address scopes");
3680 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3683 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3684 NL_SET_ERR_MSG(extack,
3685 "Invalid combination of local and remote address scopes");
3689 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3694 if (conf->label && !use_ipv6) {
3695 NL_SET_ERR_MSG(extack,
3696 "Label attribute only applies to IPv6 VXLAN devices");
3700 if (conf->remote_ifindex) {
3701 struct net_device *lowerdev;
3703 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3705 NL_SET_ERR_MSG(extack,
3706 "Invalid local interface, device not found");
3710 #if IS_ENABLED(CONFIG_IPV6)
3712 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3713 if (idev && idev->cnf.disable_ipv6) {
3714 NL_SET_ERR_MSG(extack,
3715 "IPv6 support disabled by administrator");
3723 if (vxlan_addr_multicast(&conf->remote_ip)) {
3724 NL_SET_ERR_MSG(extack,
3725 "Local interface required for multicast remote destination");
3730 #if IS_ENABLED(CONFIG_IPV6)
3731 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3732 NL_SET_ERR_MSG(extack,
3733 "Local interface required for link-local local/remote addresses");
3741 if (!conf->dst_port) {
3742 if (conf->flags & VXLAN_F_GPE)
3743 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3745 conf->dst_port = htons(vxlan_port);
3748 if (!conf->age_interval)
3749 conf->age_interval = FDB_AGE_DEFAULT;
3751 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3755 if (tmp->cfg.vni != conf->vni)
3757 if (tmp->cfg.dst_port != conf->dst_port)
3759 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3760 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3763 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3764 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3767 NL_SET_ERR_MSG(extack,
3768 "A VXLAN device with the specified VNI already exists");
3775 static void vxlan_config_apply(struct net_device *dev,
3776 struct vxlan_config *conf,
3777 struct net_device *lowerdev,
3778 struct net *src_net,
3781 struct vxlan_dev *vxlan = netdev_priv(dev);
3782 struct vxlan_rdst *dst = &vxlan->default_dst;
3783 unsigned short needed_headroom = ETH_HLEN;
3784 int max_mtu = ETH_MAX_MTU;
3785 u32 flags = conf->flags;
3788 if (flags & VXLAN_F_GPE)
3789 vxlan_raw_setup(dev);
3791 vxlan_ether_setup(dev);
3794 dev->mtu = conf->mtu;
3796 vxlan->net = src_net;
3799 dst->remote_vni = conf->vni;
3801 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3804 dst->remote_ifindex = conf->remote_ifindex;
3806 dev->gso_max_size = lowerdev->gso_max_size;
3807 dev->gso_max_segs = lowerdev->gso_max_segs;
3809 needed_headroom = lowerdev->hard_header_len;
3810 needed_headroom += lowerdev->needed_headroom;
3812 dev->needed_tailroom = lowerdev->needed_tailroom;
3814 max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3815 if (max_mtu < ETH_MIN_MTU)
3816 max_mtu = ETH_MIN_MTU;
3818 if (!changelink && !conf->mtu)
3822 if (dev->mtu > max_mtu)
3825 if (flags & VXLAN_F_COLLECT_METADATA)
3826 flags |= VXLAN_F_IPV6;
3827 needed_headroom += vxlan_headroom(flags);
3828 dev->needed_headroom = needed_headroom;
3830 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3833 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3834 struct vxlan_config *conf, bool changelink,
3835 struct netlink_ext_ack *extack)
3837 struct vxlan_dev *vxlan = netdev_priv(dev);
3838 struct net_device *lowerdev;
3841 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3845 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3850 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3851 struct vxlan_config *conf,
3852 struct netlink_ext_ack *extack)
3854 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3855 struct vxlan_dev *vxlan = netdev_priv(dev);
3856 struct net_device *remote_dev = NULL;
3857 struct vxlan_fdb *f = NULL;
3858 bool unregister = false;
3859 struct vxlan_rdst *dst;
3862 dst = &vxlan->default_dst;
3863 err = vxlan_dev_configure(net, dev, conf, false, extack);
3867 dev->ethtool_ops = &vxlan_ethtool_ops;
3869 /* create an fdb entry for a valid default destination */
3870 if (!vxlan_addr_any(&dst->remote_ip)) {
3871 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3873 NUD_REACHABLE | NUD_PERMANENT,
3874 vxlan->cfg.dst_port,
3877 dst->remote_ifindex,
3878 NTF_SELF, 0, &f, extack);
3883 err = register_netdevice(dev);
3888 if (dst->remote_ifindex) {
3889 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3895 err = netdev_upper_dev_link(remote_dev, dev, extack);
3900 err = rtnl_configure_link(dev, NULL);
3905 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3907 /* notify default fdb entry */
3908 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3909 RTM_NEWNEIGH, true, extack);
3911 vxlan_fdb_destroy(vxlan, f, false, false);
3913 netdev_upper_dev_unlink(remote_dev, dev);
3918 list_add(&vxlan->next, &vn->vxlan_list);
3920 dst->remote_dev = remote_dev;
3924 netdev_upper_dev_unlink(remote_dev, dev);
3926 /* unregister_netdevice() destroys the default FDB entry with deletion
3927 * notification. But the addition notification was not sent yet, so
3928 * destroy the entry by hand here.
3931 __vxlan_fdb_free(f);
3934 unregister_netdevice(dev);
3938 /* Set/clear flags based on attribute */
3939 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3940 int attrtype, unsigned long mask, bool changelink,
3941 bool changelink_supported,
3942 struct netlink_ext_ack *extack)
3944 unsigned long flags;
3949 if (changelink && !changelink_supported) {
3950 vxlan_flag_attr_error(attrtype, extack);
3954 if (vxlan_policy[attrtype].type == NLA_FLAG)
3955 flags = conf->flags | mask;
3956 else if (nla_get_u8(tb[attrtype]))
3957 flags = conf->flags | mask;
3959 flags = conf->flags & ~mask;
3961 conf->flags = flags;
3966 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3967 struct net_device *dev, struct vxlan_config *conf,
3968 bool changelink, struct netlink_ext_ack *extack)
3970 struct vxlan_dev *vxlan = netdev_priv(dev);
3973 memset(conf, 0, sizeof(*conf));
3975 /* if changelink operation, start with old existing cfg */
3977 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3979 if (data[IFLA_VXLAN_ID]) {
3980 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3982 if (changelink && (vni != conf->vni)) {
3983 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3986 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3989 if (data[IFLA_VXLAN_GROUP]) {
3990 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3991 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3995 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3996 conf->remote_ip.sa.sa_family = AF_INET;
3997 } else if (data[IFLA_VXLAN_GROUP6]) {
3998 if (!IS_ENABLED(CONFIG_IPV6)) {
3999 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4000 return -EPFNOSUPPORT;
4003 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4004 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4008 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4009 conf->remote_ip.sa.sa_family = AF_INET6;
4012 if (data[IFLA_VXLAN_LOCAL]) {
4013 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4014 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4018 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4019 conf->saddr.sa.sa_family = AF_INET;
4020 } else if (data[IFLA_VXLAN_LOCAL6]) {
4021 if (!IS_ENABLED(CONFIG_IPV6)) {
4022 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4023 return -EPFNOSUPPORT;
4026 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4027 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4031 /* TODO: respect scope id */
4032 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4033 conf->saddr.sa.sa_family = AF_INET6;
4036 if (data[IFLA_VXLAN_LINK])
4037 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4039 if (data[IFLA_VXLAN_TOS])
4040 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
4042 if (data[IFLA_VXLAN_TTL])
4043 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4045 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4046 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4047 VXLAN_F_TTL_INHERIT, changelink, false,
4054 if (data[IFLA_VXLAN_LABEL])
4055 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4056 IPV6_FLOWLABEL_MASK;
4058 if (data[IFLA_VXLAN_LEARNING]) {
4059 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4060 VXLAN_F_LEARN, changelink, true,
4064 } else if (!changelink) {
4065 /* default to learn on a new device */
4066 conf->flags |= VXLAN_F_LEARN;
4069 if (data[IFLA_VXLAN_AGEING])
4070 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4072 if (data[IFLA_VXLAN_PROXY]) {
4073 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4074 VXLAN_F_PROXY, changelink, false,
4080 if (data[IFLA_VXLAN_RSC]) {
4081 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4082 VXLAN_F_RSC, changelink, false,
4088 if (data[IFLA_VXLAN_L2MISS]) {
4089 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4090 VXLAN_F_L2MISS, changelink, false,
4096 if (data[IFLA_VXLAN_L3MISS]) {
4097 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4098 VXLAN_F_L3MISS, changelink, false,
4104 if (data[IFLA_VXLAN_LIMIT]) {
4106 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4107 "Cannot change limit");
4110 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4113 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4114 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4115 VXLAN_F_COLLECT_METADATA, changelink, false,
4121 if (data[IFLA_VXLAN_PORT_RANGE]) {
4123 const struct ifla_vxlan_port_range *p
4124 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4125 conf->port_min = ntohs(p->low);
4126 conf->port_max = ntohs(p->high);
4128 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4129 "Cannot change port range");
4134 if (data[IFLA_VXLAN_PORT]) {
4136 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4137 "Cannot change port");
4140 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4143 if (data[IFLA_VXLAN_UDP_CSUM]) {
4145 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4146 "Cannot change UDP_CSUM flag");
4149 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4150 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4153 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4154 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4155 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4161 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4162 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4163 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4169 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4170 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4171 VXLAN_F_REMCSUM_TX, changelink, false,
4177 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4178 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4179 VXLAN_F_REMCSUM_RX, changelink, false,
4185 if (data[IFLA_VXLAN_GBP]) {
4186 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4187 VXLAN_F_GBP, changelink, false, extack);
4192 if (data[IFLA_VXLAN_GPE]) {
4193 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4194 VXLAN_F_GPE, changelink, false,
4200 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4201 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4202 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4210 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4211 "Cannot change mtu");
4214 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4217 if (data[IFLA_VXLAN_DF])
4218 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4223 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4224 struct nlattr *tb[], struct nlattr *data[],
4225 struct netlink_ext_ack *extack)
4227 struct vxlan_config conf;
4230 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4234 return __vxlan_dev_create(src_net, dev, &conf, extack);
4237 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4238 struct nlattr *data[],
4239 struct netlink_ext_ack *extack)
4241 struct vxlan_dev *vxlan = netdev_priv(dev);
4242 struct net_device *lowerdev;
4243 struct vxlan_config conf;
4244 struct vxlan_rdst *dst;
4247 dst = &vxlan->default_dst;
4248 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4252 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4257 if (dst->remote_dev == lowerdev)
4260 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4265 /* handle default dst entry */
4266 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4267 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4269 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4270 if (!vxlan_addr_any(&conf.remote_ip)) {
4271 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4273 NUD_REACHABLE | NUD_PERMANENT,
4274 NLM_F_APPEND | NLM_F_CREATE,
4275 vxlan->cfg.dst_port,
4277 conf.remote_ifindex,
4278 NTF_SELF, 0, true, extack);
4280 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4281 netdev_adjacent_change_abort(dst->remote_dev,
4286 if (!vxlan_addr_any(&dst->remote_ip))
4287 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4289 vxlan->cfg.dst_port,
4292 dst->remote_ifindex,
4294 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4297 if (conf.age_interval != vxlan->cfg.age_interval)
4298 mod_timer(&vxlan->age_timer, jiffies);
4300 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4301 if (lowerdev && lowerdev != dst->remote_dev)
4302 dst->remote_dev = lowerdev;
4303 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4307 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4309 struct vxlan_dev *vxlan = netdev_priv(dev);
4311 vxlan_flush(vxlan, true);
4313 list_del(&vxlan->next);
4314 unregister_netdevice_queue(dev, head);
4315 if (vxlan->default_dst.remote_dev)
4316 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4319 static size_t vxlan_get_size(const struct net_device *dev)
4322 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4323 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4324 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4325 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4326 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4327 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4328 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4329 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4330 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4331 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4332 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4333 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4334 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4337 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4338 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4339 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4340 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4341 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4342 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4343 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4344 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4345 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4349 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4351 const struct vxlan_dev *vxlan = netdev_priv(dev);
4352 const struct vxlan_rdst *dst = &vxlan->default_dst;
4353 struct ifla_vxlan_port_range ports = {
4354 .low = htons(vxlan->cfg.port_min),
4355 .high = htons(vxlan->cfg.port_max),
4358 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4359 goto nla_put_failure;
4361 if (!vxlan_addr_any(&dst->remote_ip)) {
4362 if (dst->remote_ip.sa.sa_family == AF_INET) {
4363 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4364 dst->remote_ip.sin.sin_addr.s_addr))
4365 goto nla_put_failure;
4366 #if IS_ENABLED(CONFIG_IPV6)
4368 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4369 &dst->remote_ip.sin6.sin6_addr))
4370 goto nla_put_failure;
4375 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4376 goto nla_put_failure;
4378 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4379 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4380 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4381 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4382 goto nla_put_failure;
4383 #if IS_ENABLED(CONFIG_IPV6)
4385 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4386 &vxlan->cfg.saddr.sin6.sin6_addr))
4387 goto nla_put_failure;
4392 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4393 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4394 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4395 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4396 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4397 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4398 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4399 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4400 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4401 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4402 nla_put_u8(skb, IFLA_VXLAN_RSC,
4403 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4404 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4405 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4406 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4407 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4408 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4409 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4410 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4411 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4412 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4413 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4414 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4415 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4416 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4417 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4418 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4419 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4420 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4421 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4422 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4423 goto nla_put_failure;
4425 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4426 goto nla_put_failure;
4428 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4429 nla_put_flag(skb, IFLA_VXLAN_GBP))
4430 goto nla_put_failure;
4432 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4433 nla_put_flag(skb, IFLA_VXLAN_GPE))
4434 goto nla_put_failure;
4436 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4437 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4438 goto nla_put_failure;
4446 static struct net *vxlan_get_link_net(const struct net_device *dev)
4448 struct vxlan_dev *vxlan = netdev_priv(dev);
4453 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4455 .maxtype = IFLA_VXLAN_MAX,
4456 .policy = vxlan_policy,
4457 .priv_size = sizeof(struct vxlan_dev),
4458 .setup = vxlan_setup,
4459 .validate = vxlan_validate,
4460 .newlink = vxlan_newlink,
4461 .changelink = vxlan_changelink,
4462 .dellink = vxlan_dellink,
4463 .get_size = vxlan_get_size,
4464 .fill_info = vxlan_fill_info,
4465 .get_link_net = vxlan_get_link_net,
4468 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4469 u8 name_assign_type,
4470 struct vxlan_config *conf)
4472 struct nlattr *tb[IFLA_MAX + 1];
4473 struct net_device *dev;
4476 memset(&tb, 0, sizeof(tb));
4478 dev = rtnl_create_link(net, name, name_assign_type,
4479 &vxlan_link_ops, tb, NULL);
4483 err = __vxlan_dev_create(net, dev, conf, NULL);
4486 return ERR_PTR(err);
4489 err = rtnl_configure_link(dev, NULL);
4491 LIST_HEAD(list_kill);
4493 vxlan_dellink(dev, &list_kill);
4494 unregister_netdevice_many(&list_kill);
4495 return ERR_PTR(err);
4500 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4502 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4503 struct net_device *dev)
4505 struct vxlan_dev *vxlan, *next;
4506 LIST_HEAD(list_kill);
4508 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4509 struct vxlan_rdst *dst = &vxlan->default_dst;
4511 /* In case we created vxlan device with carrier
4512 * and we loose the carrier due to module unload
4513 * we also need to remove vxlan device. In other
4514 * cases, it's not necessary and remote_ifindex
4515 * is 0 here, so no matches.
4517 if (dst->remote_ifindex == dev->ifindex)
4518 vxlan_dellink(vxlan->dev, &list_kill);
4521 unregister_netdevice_many(&list_kill);
4524 static int vxlan_netdevice_event(struct notifier_block *unused,
4525 unsigned long event, void *ptr)
4527 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4528 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4530 if (event == NETDEV_UNREGISTER) {
4531 if (!dev->udp_tunnel_nic_info)
4532 vxlan_offload_rx_ports(dev, false);
4533 vxlan_handle_lowerdev_unregister(vn, dev);
4534 } else if (event == NETDEV_REGISTER) {
4535 if (!dev->udp_tunnel_nic_info)
4536 vxlan_offload_rx_ports(dev, true);
4537 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4538 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4539 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4545 static struct notifier_block vxlan_notifier_block __read_mostly = {
4546 .notifier_call = vxlan_netdevice_event,
4550 vxlan_fdb_offloaded_set(struct net_device *dev,
4551 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4553 struct vxlan_dev *vxlan = netdev_priv(dev);
4554 struct vxlan_rdst *rdst;
4555 struct vxlan_fdb *f;
4558 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4560 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4562 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4566 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4567 fdb_info->remote_port,
4568 fdb_info->remote_vni,
4569 fdb_info->remote_ifindex);
4573 rdst->offloaded = fdb_info->offloaded;
4576 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4580 vxlan_fdb_external_learn_add(struct net_device *dev,
4581 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4583 struct vxlan_dev *vxlan = netdev_priv(dev);
4584 struct netlink_ext_ack *extack;
4588 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4589 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4591 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4592 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4594 NLM_F_CREATE | NLM_F_REPLACE,
4595 fdb_info->remote_port,
4597 fdb_info->remote_vni,
4598 fdb_info->remote_ifindex,
4599 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4601 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4607 vxlan_fdb_external_learn_del(struct net_device *dev,
4608 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4610 struct vxlan_dev *vxlan = netdev_priv(dev);
4611 struct vxlan_fdb *f;
4615 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4616 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4618 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4621 else if (f->flags & NTF_EXT_LEARNED)
4622 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4623 fdb_info->remote_ip,
4624 fdb_info->remote_port,
4626 fdb_info->remote_vni,
4627 fdb_info->remote_ifindex,
4630 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4635 static int vxlan_switchdev_event(struct notifier_block *unused,
4636 unsigned long event, void *ptr)
4638 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4639 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4643 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4644 vxlan_fdb_offloaded_set(dev, ptr);
4646 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4648 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4650 err = notifier_from_errno(err);
4653 fdb_info->offloaded = true;
4654 vxlan_fdb_offloaded_set(dev, fdb_info);
4656 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4658 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4660 err = notifier_from_errno(err);
4663 fdb_info->offloaded = false;
4664 vxlan_fdb_offloaded_set(dev, fdb_info);
4671 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4672 .notifier_call = vxlan_switchdev_event,
4675 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4677 struct vxlan_fdb *fdb;
4678 struct vxlan_dev *vxlan;
4682 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4683 vxlan = rcu_dereference(fdb->vdev);
4685 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4686 vxlan->default_dst.remote_vni);
4687 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4688 if (!hlist_unhashed(&fdb->hlist))
4689 vxlan_fdb_destroy(vxlan, fdb, false, false);
4690 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4695 static int vxlan_nexthop_event(struct notifier_block *nb,
4696 unsigned long event, void *ptr)
4698 struct nexthop *nh = ptr;
4700 if (!nh || event != NEXTHOP_EVENT_DEL)
4703 vxlan_fdb_nh_flush(nh);
4708 static struct notifier_block vxlan_nexthop_notifier_block __read_mostly = {
4709 .notifier_call = vxlan_nexthop_event,
4712 static __net_init int vxlan_init_net(struct net *net)
4714 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4717 INIT_LIST_HEAD(&vn->vxlan_list);
4718 spin_lock_init(&vn->sock_lock);
4720 for (h = 0; h < PORT_HASH_SIZE; ++h)
4721 INIT_HLIST_HEAD(&vn->sock_list[h]);
4723 return register_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4726 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4728 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4729 struct vxlan_dev *vxlan, *next;
4730 struct net_device *dev, *aux;
4732 for_each_netdev_safe(net, dev, aux)
4733 if (dev->rtnl_link_ops == &vxlan_link_ops)
4734 unregister_netdevice_queue(dev, head);
4736 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4737 /* If vxlan->dev is in the same netns, it has already been added
4738 * to the list by the previous loop.
4740 if (!net_eq(dev_net(vxlan->dev), net))
4741 unregister_netdevice_queue(vxlan->dev, head);
4746 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4753 list_for_each_entry(net, net_list, exit_list)
4754 unregister_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4755 list_for_each_entry(net, net_list, exit_list)
4756 vxlan_destroy_tunnels(net, &list);
4758 unregister_netdevice_many(&list);
4761 list_for_each_entry(net, net_list, exit_list) {
4762 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4764 for (h = 0; h < PORT_HASH_SIZE; ++h)
4765 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4769 static struct pernet_operations vxlan_net_ops = {
4770 .init = vxlan_init_net,
4771 .exit_batch = vxlan_exit_batch_net,
4772 .id = &vxlan_net_id,
4773 .size = sizeof(struct vxlan_net),
4776 static int __init vxlan_init_module(void)
4780 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4782 rc = register_pernet_subsys(&vxlan_net_ops);
4786 rc = register_netdevice_notifier(&vxlan_notifier_block);
4790 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4794 rc = rtnl_link_register(&vxlan_link_ops);
4800 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4802 unregister_netdevice_notifier(&vxlan_notifier_block);
4804 unregister_pernet_subsys(&vxlan_net_ops);
4808 late_initcall(vxlan_init_module);
4810 static void __exit vxlan_cleanup_module(void)
4812 rtnl_link_unregister(&vxlan_link_ops);
4813 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4814 unregister_netdevice_notifier(&vxlan_notifier_block);
4815 unregister_pernet_subsys(&vxlan_net_ops);
4816 /* rcu_barrier() is called by netns */
4818 module_exit(vxlan_cleanup_module);
4820 MODULE_LICENSE("GPL");
4821 MODULE_VERSION(VXLAN_VERSION);
4822 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4823 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4824 MODULE_ALIAS_RTNL_LINK("vxlan");