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 /* Ignore packets from invalid src-address */
1782 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
1785 /* Get address from the outer IP header */
1786 if (vxlan_get_sk_family(vs) == AF_INET) {
1787 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1788 saddr.sa.sa_family = AF_INET;
1789 #if IS_ENABLED(CONFIG_IPV6)
1791 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1792 saddr.sa.sa_family = AF_INET6;
1796 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1797 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1803 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1804 struct sk_buff *skb)
1808 if (vxlan_get_sk_family(vs) == AF_INET)
1809 err = IP_ECN_decapsulate(oiph, skb);
1810 #if IS_ENABLED(CONFIG_IPV6)
1812 err = IP6_ECN_decapsulate(oiph, skb);
1815 if (unlikely(err) && log_ecn_error) {
1816 if (vxlan_get_sk_family(vs) == AF_INET)
1817 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1818 &((struct iphdr *)oiph)->saddr,
1819 ((struct iphdr *)oiph)->tos);
1821 net_info_ratelimited("non-ECT from %pI6\n",
1822 &((struct ipv6hdr *)oiph)->saddr);
1827 /* Callback from net/ipv4/udp.c to receive packets */
1828 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1830 struct vxlan_dev *vxlan;
1831 struct vxlan_sock *vs;
1832 struct vxlanhdr unparsed;
1833 struct vxlan_metadata _md;
1834 struct vxlan_metadata *md = &_md;
1835 __be16 protocol = htons(ETH_P_TEB);
1836 bool raw_proto = false;
1840 /* Need UDP and VXLAN header to be present */
1841 if (!pskb_may_pull(skb, VXLAN_HLEN))
1844 unparsed = *vxlan_hdr(skb);
1845 /* VNI flag always required to be set */
1846 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1847 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1848 ntohl(vxlan_hdr(skb)->vx_flags),
1849 ntohl(vxlan_hdr(skb)->vx_vni));
1850 /* Return non vxlan pkt */
1853 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1854 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1856 vs = rcu_dereference_sk_user_data(sk);
1860 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1862 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1866 /* For backwards compatibility, only allow reserved fields to be
1867 * used by VXLAN extensions if explicitly requested.
1869 if (vs->flags & VXLAN_F_GPE) {
1870 if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1872 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1876 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1877 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1880 if (vs->flags & VXLAN_F_REMCSUM_RX)
1881 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1884 if (vxlan_collect_metadata(vs)) {
1885 struct metadata_dst *tun_dst;
1887 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1888 key32_to_tunnel_id(vni), sizeof(*md));
1893 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1895 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1897 memset(md, 0, sizeof(*md));
1900 if (vs->flags & VXLAN_F_GBP)
1901 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1902 /* Note that GBP and GPE can never be active together. This is
1903 * ensured in vxlan_dev_configure.
1906 if (unparsed.vx_flags || unparsed.vx_vni) {
1907 /* If there are any unprocessed flags remaining treat
1908 * this as a malformed packet. This behavior diverges from
1909 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1910 * in reserved fields are to be ignored. The approach here
1911 * maintains compatibility with previous stack code, and also
1912 * is more robust and provides a little more security in
1913 * adding extensions to VXLAN.
1919 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1922 skb_reset_mac_header(skb);
1923 skb->dev = vxlan->dev;
1924 skb->pkt_type = PACKET_HOST;
1927 oiph = skb_network_header(skb);
1928 skb_reset_network_header(skb);
1930 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1931 ++vxlan->dev->stats.rx_frame_errors;
1932 ++vxlan->dev->stats.rx_errors;
1938 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1940 atomic_long_inc(&vxlan->dev->rx_dropped);
1944 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1945 gro_cells_receive(&vxlan->gro_cells, skb);
1952 /* Consume bad packet */
1957 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1958 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1960 struct vxlan_dev *vxlan;
1961 struct vxlan_sock *vs;
1962 struct vxlanhdr *hdr;
1965 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1968 hdr = vxlan_hdr(skb);
1970 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1973 vs = rcu_dereference_sk_user_data(sk);
1977 vni = vxlan_vni(hdr->vx_vni);
1978 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1985 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1987 struct vxlan_dev *vxlan = netdev_priv(dev);
1988 struct arphdr *parp;
1991 struct neighbour *n;
1993 if (dev->flags & IFF_NOARP)
1996 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1997 dev->stats.tx_dropped++;
2000 parp = arp_hdr(skb);
2002 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2003 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2004 parp->ar_pro != htons(ETH_P_IP) ||
2005 parp->ar_op != htons(ARPOP_REQUEST) ||
2006 parp->ar_hln != dev->addr_len ||
2009 arpptr = (u8 *)parp + sizeof(struct arphdr);
2011 arpptr += dev->addr_len; /* sha */
2012 memcpy(&sip, arpptr, sizeof(sip));
2013 arpptr += sizeof(sip);
2014 arpptr += dev->addr_len; /* tha */
2015 memcpy(&tip, arpptr, sizeof(tip));
2017 if (ipv4_is_loopback(tip) ||
2018 ipv4_is_multicast(tip))
2021 n = neigh_lookup(&arp_tbl, &tip, dev);
2024 struct vxlan_fdb *f;
2025 struct sk_buff *reply;
2027 if (!(n->nud_state & NUD_CONNECTED)) {
2032 f = vxlan_find_mac(vxlan, n->ha, vni);
2033 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2034 /* bridge-local neighbor */
2039 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2047 skb_reset_mac_header(reply);
2048 __skb_pull(reply, skb_network_offset(reply));
2049 reply->ip_summed = CHECKSUM_UNNECESSARY;
2050 reply->pkt_type = PACKET_HOST;
2052 if (netif_rx_ni(reply) == NET_RX_DROP)
2053 dev->stats.rx_dropped++;
2054 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2055 union vxlan_addr ipa = {
2056 .sin.sin_addr.s_addr = tip,
2057 .sin.sin_family = AF_INET,
2060 vxlan_ip_miss(dev, &ipa);
2064 return NETDEV_TX_OK;
2067 #if IS_ENABLED(CONFIG_IPV6)
2068 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2069 struct neighbour *n, bool isrouter)
2071 struct net_device *dev = request->dev;
2072 struct sk_buff *reply;
2073 struct nd_msg *ns, *na;
2074 struct ipv6hdr *pip6;
2076 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2080 if (dev == NULL || !pskb_may_pull(request, request->len))
2083 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2084 sizeof(*na) + na_olen + dev->needed_tailroom;
2085 reply = alloc_skb(len, GFP_ATOMIC);
2089 reply->protocol = htons(ETH_P_IPV6);
2091 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2092 skb_push(reply, sizeof(struct ethhdr));
2093 skb_reset_mac_header(reply);
2095 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2097 daddr = eth_hdr(request)->h_source;
2098 ns_olen = request->len - skb_network_offset(request) -
2099 sizeof(struct ipv6hdr) - sizeof(*ns);
2100 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2101 if (!ns->opt[i + 1]) {
2105 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2106 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2111 /* Ethernet header */
2112 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2113 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2114 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2115 reply->protocol = htons(ETH_P_IPV6);
2117 skb_pull(reply, sizeof(struct ethhdr));
2118 skb_reset_network_header(reply);
2119 skb_put(reply, sizeof(struct ipv6hdr));
2123 pip6 = ipv6_hdr(reply);
2124 memset(pip6, 0, sizeof(struct ipv6hdr));
2126 pip6->priority = ipv6_hdr(request)->priority;
2127 pip6->nexthdr = IPPROTO_ICMPV6;
2128 pip6->hop_limit = 255;
2129 pip6->daddr = ipv6_hdr(request)->saddr;
2130 pip6->saddr = *(struct in6_addr *)n->primary_key;
2132 skb_pull(reply, sizeof(struct ipv6hdr));
2133 skb_reset_transport_header(reply);
2135 /* Neighbor Advertisement */
2136 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2137 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2138 na->icmph.icmp6_router = isrouter;
2139 na->icmph.icmp6_override = 1;
2140 na->icmph.icmp6_solicited = 1;
2141 na->target = ns->target;
2142 ether_addr_copy(&na->opt[2], n->ha);
2143 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2144 na->opt[1] = na_olen >> 3;
2146 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2147 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2148 csum_partial(na, sizeof(*na)+na_olen, 0));
2150 pip6->payload_len = htons(sizeof(*na)+na_olen);
2152 skb_push(reply, sizeof(struct ipv6hdr));
2154 reply->ip_summed = CHECKSUM_UNNECESSARY;
2159 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2161 struct vxlan_dev *vxlan = netdev_priv(dev);
2162 const struct in6_addr *daddr;
2163 const struct ipv6hdr *iphdr;
2164 struct inet6_dev *in6_dev;
2165 struct neighbour *n;
2169 in6_dev = __in6_dev_get(dev);
2173 iphdr = ipv6_hdr(skb);
2174 daddr = &iphdr->daddr;
2175 msg = (struct nd_msg *)(iphdr + 1);
2177 if (ipv6_addr_loopback(daddr) ||
2178 ipv6_addr_is_multicast(&msg->target))
2181 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2184 struct vxlan_fdb *f;
2185 struct sk_buff *reply;
2187 if (!(n->nud_state & NUD_CONNECTED)) {
2192 f = vxlan_find_mac(vxlan, n->ha, vni);
2193 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2194 /* bridge-local neighbor */
2199 reply = vxlan_na_create(skb, n,
2200 !!(f ? f->flags & NTF_ROUTER : 0));
2207 if (netif_rx_ni(reply) == NET_RX_DROP)
2208 dev->stats.rx_dropped++;
2210 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2211 union vxlan_addr ipa = {
2212 .sin6.sin6_addr = msg->target,
2213 .sin6.sin6_family = AF_INET6,
2216 vxlan_ip_miss(dev, &ipa);
2222 return NETDEV_TX_OK;
2226 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2228 struct vxlan_dev *vxlan = netdev_priv(dev);
2229 struct neighbour *n;
2231 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2235 switch (ntohs(eth_hdr(skb)->h_proto)) {
2240 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2243 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2244 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2245 union vxlan_addr ipa = {
2246 .sin.sin_addr.s_addr = pip->daddr,
2247 .sin.sin_family = AF_INET,
2250 vxlan_ip_miss(dev, &ipa);
2256 #if IS_ENABLED(CONFIG_IPV6)
2259 struct ipv6hdr *pip6;
2261 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2263 pip6 = ipv6_hdr(skb);
2264 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2265 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2266 union vxlan_addr ipa = {
2267 .sin6.sin6_addr = pip6->daddr,
2268 .sin6.sin6_family = AF_INET6,
2271 vxlan_ip_miss(dev, &ipa);
2285 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2287 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2289 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2298 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2299 struct vxlan_metadata *md)
2301 struct vxlanhdr_gbp *gbp;
2306 gbp = (struct vxlanhdr_gbp *)vxh;
2307 vxh->vx_flags |= VXLAN_HF_GBP;
2309 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2310 gbp->dont_learn = 1;
2312 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2313 gbp->policy_applied = 1;
2315 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2318 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2321 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2323 gpe->np_applied = 1;
2324 gpe->next_protocol = tun_p_from_eth_p(protocol);
2325 if (!gpe->next_protocol)
2326 return -EPFNOSUPPORT;
2330 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2331 int iphdr_len, __be32 vni,
2332 struct vxlan_metadata *md, u32 vxflags,
2335 struct vxlanhdr *vxh;
2338 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2339 __be16 inner_protocol = htons(ETH_P_TEB);
2341 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2342 skb->ip_summed == CHECKSUM_PARTIAL) {
2343 int csum_start = skb_checksum_start_offset(skb);
2345 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2346 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2347 (skb->csum_offset == offsetof(struct udphdr, check) ||
2348 skb->csum_offset == offsetof(struct tcphdr, check)))
2349 type |= SKB_GSO_TUNNEL_REMCSUM;
2352 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2353 + VXLAN_HLEN + iphdr_len;
2355 /* Need space for new headers (invalidates iph ptr) */
2356 err = skb_cow_head(skb, min_headroom);
2360 err = iptunnel_handle_offloads(skb, type);
2364 vxh = __skb_push(skb, sizeof(*vxh));
2365 vxh->vx_flags = VXLAN_HF_VNI;
2366 vxh->vx_vni = vxlan_vni_field(vni);
2368 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2371 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2372 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2373 vxh->vx_flags |= VXLAN_HF_RCO;
2375 if (!skb_is_gso(skb)) {
2376 skb->ip_summed = CHECKSUM_NONE;
2377 skb->encapsulation = 0;
2381 if (vxflags & VXLAN_F_GBP)
2382 vxlan_build_gbp_hdr(vxh, vxflags, md);
2383 if (vxflags & VXLAN_F_GPE) {
2384 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2387 inner_protocol = skb->protocol;
2390 skb_set_inner_protocol(skb, inner_protocol);
2394 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2395 struct vxlan_sock *sock4,
2396 struct sk_buff *skb, int oif, u8 tos,
2397 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2398 struct dst_cache *dst_cache,
2399 const struct ip_tunnel_info *info)
2401 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2402 struct rtable *rt = NULL;
2406 return ERR_PTR(-EIO);
2411 rt = dst_cache_get_ip4(dst_cache, saddr);
2416 memset(&fl4, 0, sizeof(fl4));
2417 fl4.flowi4_oif = oif;
2418 fl4.flowi4_tos = RT_TOS(tos);
2419 fl4.flowi4_mark = skb->mark;
2420 fl4.flowi4_proto = IPPROTO_UDP;
2423 fl4.fl4_dport = dport;
2424 fl4.fl4_sport = sport;
2426 rt = ip_route_output_key(vxlan->net, &fl4);
2428 if (rt->dst.dev == dev) {
2429 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2431 return ERR_PTR(-ELOOP);
2436 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2438 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2439 return ERR_PTR(-ENETUNREACH);
2444 #if IS_ENABLED(CONFIG_IPV6)
2445 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2446 struct net_device *dev,
2447 struct vxlan_sock *sock6,
2448 struct sk_buff *skb, int oif, u8 tos,
2450 const struct in6_addr *daddr,
2451 struct in6_addr *saddr,
2452 __be16 dport, __be16 sport,
2453 struct dst_cache *dst_cache,
2454 const struct ip_tunnel_info *info)
2456 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2457 struct dst_entry *ndst;
2461 return ERR_PTR(-EIO);
2466 ndst = dst_cache_get_ip6(dst_cache, saddr);
2471 memset(&fl6, 0, sizeof(fl6));
2472 fl6.flowi6_oif = oif;
2475 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2476 fl6.flowi6_mark = skb->mark;
2477 fl6.flowi6_proto = IPPROTO_UDP;
2478 fl6.fl6_dport = dport;
2479 fl6.fl6_sport = sport;
2481 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2483 if (unlikely(IS_ERR(ndst))) {
2484 netdev_dbg(dev, "no route to %pI6\n", daddr);
2485 return ERR_PTR(-ENETUNREACH);
2488 if (unlikely(ndst->dev == dev)) {
2489 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2491 return ERR_PTR(-ELOOP);
2496 dst_cache_set_ip6(dst_cache, ndst, saddr);
2501 /* Bypass encapsulation if the destination is local */
2502 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2503 struct vxlan_dev *dst_vxlan, __be32 vni,
2506 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2507 union vxlan_addr loopback;
2508 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2509 struct net_device *dev;
2512 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2513 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2514 skb->pkt_type = PACKET_HOST;
2515 skb->encapsulation = 0;
2516 skb->dev = dst_vxlan->dev;
2517 __skb_pull(skb, skb_network_offset(skb));
2519 if (remote_ip->sa.sa_family == AF_INET) {
2520 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2521 loopback.sa.sa_family = AF_INET;
2522 #if IS_ENABLED(CONFIG_IPV6)
2524 loopback.sin6.sin6_addr = in6addr_loopback;
2525 loopback.sa.sa_family = AF_INET6;
2531 if (unlikely(!(dev->flags & IFF_UP))) {
2536 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2537 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2539 u64_stats_update_begin(&tx_stats->syncp);
2540 tx_stats->tx_packets++;
2541 tx_stats->tx_bytes += len;
2542 u64_stats_update_end(&tx_stats->syncp);
2544 if (netif_rx(skb) == NET_RX_SUCCESS) {
2545 u64_stats_update_begin(&rx_stats->syncp);
2546 rx_stats->rx_packets++;
2547 rx_stats->rx_bytes += len;
2548 u64_stats_update_end(&rx_stats->syncp);
2551 dev->stats.rx_dropped++;
2556 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2557 struct vxlan_dev *vxlan,
2558 union vxlan_addr *daddr,
2559 __be16 dst_port, int dst_ifindex, __be32 vni,
2560 struct dst_entry *dst,
2563 #if IS_ENABLED(CONFIG_IPV6)
2564 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2565 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2566 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2568 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2570 /* Bypass encapsulation if the destination is local */
2571 if (rt_flags & RTCF_LOCAL &&
2572 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2573 struct vxlan_dev *dst_vxlan;
2576 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2577 daddr->sa.sa_family, dst_port,
2580 dev->stats.tx_errors++;
2585 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2592 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2593 __be32 default_vni, struct vxlan_rdst *rdst,
2596 struct dst_cache *dst_cache;
2597 struct ip_tunnel_info *info;
2598 struct vxlan_dev *vxlan = netdev_priv(dev);
2599 const struct iphdr *old_iph = ip_hdr(skb);
2600 union vxlan_addr *dst;
2601 union vxlan_addr remote_ip, local_ip;
2602 struct vxlan_metadata _md;
2603 struct vxlan_metadata *md = &_md;
2604 __be16 src_port = 0, dst_port;
2605 struct dst_entry *ndst = NULL;
2610 u32 flags = vxlan->cfg.flags;
2611 bool udp_sum = false;
2612 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2614 info = skb_tunnel_info(skb);
2617 dst = &rdst->remote_ip;
2618 if (vxlan_addr_any(dst)) {
2620 /* short-circuited back to local bridge */
2621 vxlan_encap_bypass(skb, vxlan, vxlan,
2628 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2629 vni = (rdst->remote_vni) ? : default_vni;
2630 ifindex = rdst->remote_ifindex;
2631 local_ip = vxlan->cfg.saddr;
2632 dst_cache = &rdst->dst_cache;
2633 md->gbp = skb->mark;
2634 if (flags & VXLAN_F_TTL_INHERIT) {
2635 ttl = ip_tunnel_get_ttl(old_iph, skb);
2637 ttl = vxlan->cfg.ttl;
2638 if (!ttl && vxlan_addr_multicast(dst))
2642 tos = vxlan->cfg.tos;
2644 tos = ip_tunnel_get_dsfield(old_iph, skb);
2646 if (dst->sa.sa_family == AF_INET)
2647 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2649 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2650 label = vxlan->cfg.label;
2653 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2657 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2658 if (remote_ip.sa.sa_family == AF_INET) {
2659 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2660 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2662 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2663 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2666 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2667 vni = tunnel_id_to_key32(info->key.tun_id);
2669 dst_cache = &info->dst_cache;
2670 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2671 if (info->options_len < sizeof(*md))
2673 md = ip_tunnel_info_opts(info);
2675 ttl = info->key.ttl;
2676 tos = info->key.tos;
2677 label = info->key.label;
2678 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2680 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2681 vxlan->cfg.port_max, true);
2684 if (dst->sa.sa_family == AF_INET) {
2685 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2690 ifindex = sock4->sock->sk->sk_bound_dev_if;
2692 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2693 dst->sin.sin_addr.s_addr,
2694 &local_ip.sin.sin_addr.s_addr,
2703 /* Bypass encapsulation if the destination is local */
2704 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2705 dst_port, ifindex, vni,
2706 &rt->dst, rt->rt_flags);
2710 if (vxlan->cfg.df == VXLAN_DF_SET) {
2712 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2713 struct ethhdr *eth = eth_hdr(skb);
2715 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2716 (ntohs(eth->h_proto) == ETH_P_IP &&
2717 old_iph->frag_off & htons(IP_DF)))
2720 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2725 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2726 netif_is_any_bridge_port(dev));
2731 struct ip_tunnel_info *unclone;
2732 struct in_addr src, dst;
2734 unclone = skb_tunnel_info_unclone(skb);
2735 if (unlikely(!unclone))
2738 src = remote_ip.sin.sin_addr;
2739 dst = local_ip.sin.sin_addr;
2740 unclone->key.u.ipv4.src = src.s_addr;
2741 unclone->key.u.ipv4.dst = dst.s_addr;
2743 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2748 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2749 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2750 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2751 vni, md, flags, udp_sum);
2755 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2756 dst->sin.sin_addr.s_addr, tos, ttl, df,
2757 src_port, dst_port, xnet, !udp_sum);
2758 #if IS_ENABLED(CONFIG_IPV6)
2760 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2763 ifindex = sock6->sock->sk->sk_bound_dev_if;
2765 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2766 label, &dst->sin6.sin6_addr,
2767 &local_ip.sin6.sin6_addr,
2771 err = PTR_ERR(ndst);
2777 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2779 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2780 dst_port, ifindex, vni,
2786 err = skb_tunnel_check_pmtu(skb, ndst,
2787 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2788 netif_is_any_bridge_port(dev));
2793 struct ip_tunnel_info *unclone;
2794 struct in6_addr src, dst;
2796 unclone = skb_tunnel_info_unclone(skb);
2797 if (unlikely(!unclone))
2800 src = remote_ip.sin6.sin6_addr;
2801 dst = local_ip.sin6.sin6_addr;
2802 unclone->key.u.ipv6.src = src;
2803 unclone->key.u.ipv6.dst = dst;
2806 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2811 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2812 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2813 skb_scrub_packet(skb, xnet);
2814 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2815 vni, md, flags, udp_sum);
2819 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2820 &local_ip.sin6.sin6_addr,
2821 &dst->sin6.sin6_addr, tos, ttl,
2822 label, src_port, dst_port, !udp_sum);
2830 dev->stats.tx_dropped++;
2837 dev->stats.collisions++;
2838 else if (err == -ENETUNREACH)
2839 dev->stats.tx_carrier_errors++;
2841 dev->stats.tx_errors++;
2845 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2846 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2848 struct vxlan_rdst nh_rdst;
2853 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2854 hash = skb_get_hash(skb);
2857 nh = rcu_dereference(f->nh);
2862 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2865 if (likely(do_xmit))
2866 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2873 dev->stats.tx_dropped++;
2877 /* Transmit local packets over Vxlan
2879 * Outer IP header inherits ECN and DF from inner header.
2880 * Outer UDP destination is the VXLAN assigned port.
2881 * source port is based on hash of flow
2883 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2885 struct vxlan_dev *vxlan = netdev_priv(dev);
2886 struct vxlan_rdst *rdst, *fdst = NULL;
2887 const struct ip_tunnel_info *info;
2888 bool did_rsc = false;
2889 struct vxlan_fdb *f;
2893 info = skb_tunnel_info(skb);
2895 skb_reset_mac_header(skb);
2897 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2898 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2899 info->mode & IP_TUNNEL_INFO_TX) {
2900 vni = tunnel_id_to_key32(info->key.tun_id);
2902 if (info && info->mode & IP_TUNNEL_INFO_TX)
2903 vxlan_xmit_one(skb, dev, vni, NULL, false);
2906 return NETDEV_TX_OK;
2910 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2912 if (ntohs(eth->h_proto) == ETH_P_ARP)
2913 return arp_reduce(dev, skb, vni);
2914 #if IS_ENABLED(CONFIG_IPV6)
2915 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2916 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2917 sizeof(struct nd_msg)) &&
2918 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2919 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2921 if (m->icmph.icmp6_code == 0 &&
2922 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2923 return neigh_reduce(dev, skb, vni);
2929 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2932 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2933 (ntohs(eth->h_proto) == ETH_P_IP ||
2934 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2935 did_rsc = route_shortcircuit(dev, skb);
2937 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2941 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2943 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2944 !is_multicast_ether_addr(eth->h_dest))
2945 vxlan_fdb_miss(vxlan, eth->h_dest);
2947 dev->stats.tx_dropped++;
2949 return NETDEV_TX_OK;
2953 if (rcu_access_pointer(f->nh)) {
2954 vxlan_xmit_nh(skb, dev, f,
2955 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2957 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2958 struct sk_buff *skb1;
2964 skb1 = skb_clone(skb, GFP_ATOMIC);
2966 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2969 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2974 return NETDEV_TX_OK;
2977 /* Walk the forwarding table and purge stale entries */
2978 static void vxlan_cleanup(struct timer_list *t)
2980 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2981 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2984 if (!netif_running(vxlan->dev))
2987 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2988 struct hlist_node *p, *n;
2990 spin_lock(&vxlan->hash_lock[h]);
2991 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2993 = container_of(p, struct vxlan_fdb, hlist);
2994 unsigned long timeout;
2996 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2999 if (f->flags & NTF_EXT_LEARNED)
3002 timeout = f->used + vxlan->cfg.age_interval * HZ;
3003 if (time_before_eq(timeout, jiffies)) {
3004 netdev_dbg(vxlan->dev,
3005 "garbage collect %pM\n",
3007 f->state = NUD_STALE;
3008 vxlan_fdb_destroy(vxlan, f, true, true);
3009 } else if (time_before(timeout, next_timer))
3010 next_timer = timeout;
3012 spin_unlock(&vxlan->hash_lock[h]);
3015 mod_timer(&vxlan->age_timer, next_timer);
3018 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3020 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3022 spin_lock(&vn->sock_lock);
3023 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3024 #if IS_ENABLED(CONFIG_IPV6)
3025 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3027 spin_unlock(&vn->sock_lock);
3030 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3031 struct vxlan_dev_node *node)
3033 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3034 __be32 vni = vxlan->default_dst.remote_vni;
3036 node->vxlan = vxlan;
3037 spin_lock(&vn->sock_lock);
3038 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3039 spin_unlock(&vn->sock_lock);
3042 /* Setup stats when device is created */
3043 static int vxlan_init(struct net_device *dev)
3045 struct vxlan_dev *vxlan = netdev_priv(dev);
3048 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3052 err = gro_cells_init(&vxlan->gro_cells, dev);
3054 free_percpu(dev->tstats);
3061 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3063 struct vxlan_fdb *f;
3064 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3066 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3067 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3069 vxlan_fdb_destroy(vxlan, f, true, true);
3070 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3073 static void vxlan_uninit(struct net_device *dev)
3075 struct vxlan_dev *vxlan = netdev_priv(dev);
3077 gro_cells_destroy(&vxlan->gro_cells);
3079 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3081 free_percpu(dev->tstats);
3084 /* Start ageing timer and join group when device is brought up */
3085 static int vxlan_open(struct net_device *dev)
3087 struct vxlan_dev *vxlan = netdev_priv(dev);
3090 ret = vxlan_sock_add(vxlan);
3094 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3095 ret = vxlan_igmp_join(vxlan);
3096 if (ret == -EADDRINUSE)
3099 vxlan_sock_release(vxlan);
3104 if (vxlan->cfg.age_interval)
3105 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3110 /* Purge the forwarding table */
3111 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3115 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3116 struct hlist_node *p, *n;
3118 spin_lock_bh(&vxlan->hash_lock[h]);
3119 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3121 = container_of(p, struct vxlan_fdb, hlist);
3122 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3124 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3125 if (is_zero_ether_addr(f->eth_addr) &&
3126 f->vni == vxlan->cfg.vni)
3128 vxlan_fdb_destroy(vxlan, f, true, true);
3130 spin_unlock_bh(&vxlan->hash_lock[h]);
3134 /* Cleanup timer and forwarding table on shutdown */
3135 static int vxlan_stop(struct net_device *dev)
3137 struct vxlan_dev *vxlan = netdev_priv(dev);
3138 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3141 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3142 !vxlan_group_used(vn, vxlan))
3143 ret = vxlan_igmp_leave(vxlan);
3145 del_timer_sync(&vxlan->age_timer);
3147 vxlan_flush(vxlan, false);
3148 vxlan_sock_release(vxlan);
3153 /* Stub, nothing needs to be done. */
3154 static void vxlan_set_multicast_list(struct net_device *dev)
3158 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3160 struct vxlan_dev *vxlan = netdev_priv(dev);
3161 struct vxlan_rdst *dst = &vxlan->default_dst;
3162 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3163 dst->remote_ifindex);
3165 /* This check is different than dev->max_mtu, because it looks at
3166 * the lowerdev->mtu, rather than the static dev->max_mtu
3169 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3170 if (new_mtu > max_mtu)
3178 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3180 struct vxlan_dev *vxlan = netdev_priv(dev);
3181 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3182 __be16 sport, dport;
3184 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3185 vxlan->cfg.port_max, true);
3186 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3188 if (ip_tunnel_info_af(info) == AF_INET) {
3189 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3192 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3193 info->key.u.ipv4.dst,
3194 &info->key.u.ipv4.src, dport, sport,
3195 &info->dst_cache, info);
3200 #if IS_ENABLED(CONFIG_IPV6)
3201 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3202 struct dst_entry *ndst;
3204 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3205 info->key.label, &info->key.u.ipv6.dst,
3206 &info->key.u.ipv6.src, dport, sport,
3207 &info->dst_cache, info);
3209 return PTR_ERR(ndst);
3211 #else /* !CONFIG_IPV6 */
3212 return -EPFNOSUPPORT;
3215 info->key.tp_src = sport;
3216 info->key.tp_dst = dport;
3220 static const struct net_device_ops vxlan_netdev_ether_ops = {
3221 .ndo_init = vxlan_init,
3222 .ndo_uninit = vxlan_uninit,
3223 .ndo_open = vxlan_open,
3224 .ndo_stop = vxlan_stop,
3225 .ndo_start_xmit = vxlan_xmit,
3226 .ndo_get_stats64 = ip_tunnel_get_stats64,
3227 .ndo_set_rx_mode = vxlan_set_multicast_list,
3228 .ndo_change_mtu = vxlan_change_mtu,
3229 .ndo_validate_addr = eth_validate_addr,
3230 .ndo_set_mac_address = eth_mac_addr,
3231 .ndo_fdb_add = vxlan_fdb_add,
3232 .ndo_fdb_del = vxlan_fdb_delete,
3233 .ndo_fdb_dump = vxlan_fdb_dump,
3234 .ndo_fdb_get = vxlan_fdb_get,
3235 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3236 .ndo_change_proto_down = dev_change_proto_down_generic,
3239 static const struct net_device_ops vxlan_netdev_raw_ops = {
3240 .ndo_init = vxlan_init,
3241 .ndo_uninit = vxlan_uninit,
3242 .ndo_open = vxlan_open,
3243 .ndo_stop = vxlan_stop,
3244 .ndo_start_xmit = vxlan_xmit,
3245 .ndo_get_stats64 = ip_tunnel_get_stats64,
3246 .ndo_change_mtu = vxlan_change_mtu,
3247 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3250 /* Info for udev, that this is a virtual tunnel endpoint */
3251 static struct device_type vxlan_type = {
3255 /* Calls the ndo_udp_tunnel_add of the caller in order to
3256 * supply the listening VXLAN udp ports. Callers are expected
3257 * to implement the ndo_udp_tunnel_add.
3259 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3261 struct vxlan_sock *vs;
3262 struct net *net = dev_net(dev);
3263 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3266 spin_lock(&vn->sock_lock);
3267 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3268 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3269 unsigned short type;
3271 if (vs->flags & VXLAN_F_GPE)
3272 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3274 type = UDP_TUNNEL_TYPE_VXLAN;
3277 udp_tunnel_push_rx_port(dev, vs->sock, type);
3279 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3282 spin_unlock(&vn->sock_lock);
3285 /* Initialize the device structure. */
3286 static void vxlan_setup(struct net_device *dev)
3288 struct vxlan_dev *vxlan = netdev_priv(dev);
3291 eth_hw_addr_random(dev);
3294 dev->needs_free_netdev = true;
3295 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3297 dev->features |= NETIF_F_LLTX;
3298 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3299 dev->features |= NETIF_F_RXCSUM;
3300 dev->features |= NETIF_F_GSO_SOFTWARE;
3302 dev->vlan_features = dev->features;
3303 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3304 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3305 netif_keep_dst(dev);
3306 dev->priv_flags |= IFF_NO_QUEUE;
3308 /* MTU range: 68 - 65535 */
3309 dev->min_mtu = ETH_MIN_MTU;
3310 dev->max_mtu = ETH_MAX_MTU;
3312 INIT_LIST_HEAD(&vxlan->next);
3314 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3318 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3319 spin_lock_init(&vxlan->hash_lock[h]);
3320 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3324 static void vxlan_ether_setup(struct net_device *dev)
3326 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3327 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3328 dev->netdev_ops = &vxlan_netdev_ether_ops;
3331 static void vxlan_raw_setup(struct net_device *dev)
3333 dev->header_ops = NULL;
3334 dev->type = ARPHRD_NONE;
3335 dev->hard_header_len = 0;
3337 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3338 dev->netdev_ops = &vxlan_netdev_raw_ops;
3341 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3342 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3343 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3344 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3345 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3346 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3347 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3348 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3349 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3350 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3351 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3352 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3353 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3354 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3355 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3356 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3357 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3358 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3359 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3360 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3361 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3362 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3363 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3364 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3365 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3366 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3367 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3368 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3369 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3370 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3373 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3374 struct netlink_ext_ack *extack)
3376 if (tb[IFLA_ADDRESS]) {
3377 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3378 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3379 "Provided link layer address is not Ethernet");
3383 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3384 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3385 "Provided Ethernet address is not unicast");
3386 return -EADDRNOTAVAIL;
3391 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3393 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3394 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3395 "MTU must be between 68 and 65535");
3401 NL_SET_ERR_MSG(extack,
3402 "Required attributes not provided to perform the operation");
3406 if (data[IFLA_VXLAN_ID]) {
3407 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3409 if (id >= VXLAN_N_VID) {
3410 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3411 "VXLAN ID must be lower than 16777216");
3416 if (data[IFLA_VXLAN_PORT_RANGE]) {
3417 const struct ifla_vxlan_port_range *p
3418 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3420 if (ntohs(p->high) < ntohs(p->low)) {
3421 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3422 "Invalid source port range");
3427 if (data[IFLA_VXLAN_DF]) {
3428 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3430 if (df < 0 || df > VXLAN_DF_MAX) {
3431 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3432 "Invalid DF attribute");
3440 static void vxlan_get_drvinfo(struct net_device *netdev,
3441 struct ethtool_drvinfo *drvinfo)
3443 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3444 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3447 static int vxlan_get_link_ksettings(struct net_device *dev,
3448 struct ethtool_link_ksettings *cmd)
3450 struct vxlan_dev *vxlan = netdev_priv(dev);
3451 struct vxlan_rdst *dst = &vxlan->default_dst;
3452 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3453 dst->remote_ifindex);
3456 cmd->base.duplex = DUPLEX_UNKNOWN;
3457 cmd->base.port = PORT_OTHER;
3458 cmd->base.speed = SPEED_UNKNOWN;
3463 return __ethtool_get_link_ksettings(lowerdev, cmd);
3466 static const struct ethtool_ops vxlan_ethtool_ops = {
3467 .get_drvinfo = vxlan_get_drvinfo,
3468 .get_link = ethtool_op_get_link,
3469 .get_link_ksettings = vxlan_get_link_ksettings,
3472 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3473 __be16 port, u32 flags, int ifindex)
3475 struct socket *sock;
3476 struct udp_port_cfg udp_conf;
3479 memset(&udp_conf, 0, sizeof(udp_conf));
3482 udp_conf.family = AF_INET6;
3483 udp_conf.use_udp6_rx_checksums =
3484 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3485 udp_conf.ipv6_v6only = 1;
3487 udp_conf.family = AF_INET;
3490 udp_conf.local_udp_port = port;
3491 udp_conf.bind_ifindex = ifindex;
3493 /* Open UDP socket */
3494 err = udp_sock_create(net, &udp_conf, &sock);
3496 return ERR_PTR(err);
3501 /* Create new listen socket if needed */
3502 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3503 __be16 port, u32 flags,
3506 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3507 struct vxlan_sock *vs;
3508 struct socket *sock;
3510 struct udp_tunnel_sock_cfg tunnel_cfg;
3512 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3514 return ERR_PTR(-ENOMEM);
3516 for (h = 0; h < VNI_HASH_SIZE; ++h)
3517 INIT_HLIST_HEAD(&vs->vni_list[h]);
3519 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3522 return ERR_CAST(sock);
3526 refcount_set(&vs->refcnt, 1);
3527 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3529 spin_lock(&vn->sock_lock);
3530 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3531 udp_tunnel_notify_add_rx_port(sock,
3532 (vs->flags & VXLAN_F_GPE) ?
3533 UDP_TUNNEL_TYPE_VXLAN_GPE :
3534 UDP_TUNNEL_TYPE_VXLAN);
3535 spin_unlock(&vn->sock_lock);
3537 /* Mark socket as an encapsulation socket. */
3538 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3539 tunnel_cfg.sk_user_data = vs;
3540 tunnel_cfg.encap_type = 1;
3541 tunnel_cfg.encap_rcv = vxlan_rcv;
3542 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3543 tunnel_cfg.encap_destroy = NULL;
3544 tunnel_cfg.gro_receive = vxlan_gro_receive;
3545 tunnel_cfg.gro_complete = vxlan_gro_complete;
3547 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3552 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3554 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3555 struct vxlan_sock *vs = NULL;
3556 struct vxlan_dev_node *node;
3557 int l3mdev_index = 0;
3559 if (vxlan->cfg.remote_ifindex)
3560 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3561 vxlan->net, vxlan->cfg.remote_ifindex);
3563 if (!vxlan->cfg.no_share) {
3564 spin_lock(&vn->sock_lock);
3565 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3566 vxlan->cfg.dst_port, vxlan->cfg.flags,
3568 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3569 spin_unlock(&vn->sock_lock);
3572 spin_unlock(&vn->sock_lock);
3575 vs = vxlan_socket_create(vxlan->net, ipv6,
3576 vxlan->cfg.dst_port, vxlan->cfg.flags,
3580 #if IS_ENABLED(CONFIG_IPV6)
3582 rcu_assign_pointer(vxlan->vn6_sock, vs);
3583 node = &vxlan->hlist6;
3587 rcu_assign_pointer(vxlan->vn4_sock, vs);
3588 node = &vxlan->hlist4;
3590 vxlan_vs_add_dev(vs, vxlan, node);
3594 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3596 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3597 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3598 bool ipv4 = !ipv6 || metadata;
3601 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3602 #if IS_ENABLED(CONFIG_IPV6)
3603 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3605 ret = __vxlan_sock_add(vxlan, true);
3606 if (ret < 0 && ret != -EAFNOSUPPORT)
3611 ret = __vxlan_sock_add(vxlan, false);
3613 vxlan_sock_release(vxlan);
3617 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3618 struct net_device **lower,
3619 struct vxlan_dev *old,
3620 struct netlink_ext_ack *extack)
3622 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3623 struct vxlan_dev *tmp;
3624 bool use_ipv6 = false;
3626 if (conf->flags & VXLAN_F_GPE) {
3627 /* For now, allow GPE only together with
3628 * COLLECT_METADATA. This can be relaxed later; in such
3629 * case, the other side of the PtP link will have to be
3632 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3633 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3634 NL_SET_ERR_MSG(extack,
3635 "VXLAN GPE does not support this combination of attributes");
3640 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3641 /* Unless IPv6 is explicitly requested, assume IPv4 */
3642 conf->remote_ip.sa.sa_family = AF_INET;
3643 conf->saddr.sa.sa_family = AF_INET;
3644 } else if (!conf->remote_ip.sa.sa_family) {
3645 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3646 } else if (!conf->saddr.sa.sa_family) {
3647 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3650 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3651 NL_SET_ERR_MSG(extack,
3652 "Local and remote address must be from the same family");
3656 if (vxlan_addr_multicast(&conf->saddr)) {
3657 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3661 if (conf->saddr.sa.sa_family == AF_INET6) {
3662 if (!IS_ENABLED(CONFIG_IPV6)) {
3663 NL_SET_ERR_MSG(extack,
3664 "IPv6 support not enabled in the kernel");
3665 return -EPFNOSUPPORT;
3668 conf->flags |= VXLAN_F_IPV6;
3670 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3672 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3674 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3676 if (local_type & IPV6_ADDR_LINKLOCAL) {
3677 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3678 (remote_type != IPV6_ADDR_ANY)) {
3679 NL_SET_ERR_MSG(extack,
3680 "Invalid combination of local and remote address scopes");
3684 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3687 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3688 NL_SET_ERR_MSG(extack,
3689 "Invalid combination of local and remote address scopes");
3693 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3698 if (conf->label && !use_ipv6) {
3699 NL_SET_ERR_MSG(extack,
3700 "Label attribute only applies to IPv6 VXLAN devices");
3704 if (conf->remote_ifindex) {
3705 struct net_device *lowerdev;
3707 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3709 NL_SET_ERR_MSG(extack,
3710 "Invalid local interface, device not found");
3714 #if IS_ENABLED(CONFIG_IPV6)
3716 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3717 if (idev && idev->cnf.disable_ipv6) {
3718 NL_SET_ERR_MSG(extack,
3719 "IPv6 support disabled by administrator");
3727 if (vxlan_addr_multicast(&conf->remote_ip)) {
3728 NL_SET_ERR_MSG(extack,
3729 "Local interface required for multicast remote destination");
3734 #if IS_ENABLED(CONFIG_IPV6)
3735 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3736 NL_SET_ERR_MSG(extack,
3737 "Local interface required for link-local local/remote addresses");
3745 if (!conf->dst_port) {
3746 if (conf->flags & VXLAN_F_GPE)
3747 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3749 conf->dst_port = htons(vxlan_port);
3752 if (!conf->age_interval)
3753 conf->age_interval = FDB_AGE_DEFAULT;
3755 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3759 if (tmp->cfg.vni != conf->vni)
3761 if (tmp->cfg.dst_port != conf->dst_port)
3763 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3764 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3767 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3768 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3771 NL_SET_ERR_MSG(extack,
3772 "A VXLAN device with the specified VNI already exists");
3779 static void vxlan_config_apply(struct net_device *dev,
3780 struct vxlan_config *conf,
3781 struct net_device *lowerdev,
3782 struct net *src_net,
3785 struct vxlan_dev *vxlan = netdev_priv(dev);
3786 struct vxlan_rdst *dst = &vxlan->default_dst;
3787 unsigned short needed_headroom = ETH_HLEN;
3788 int max_mtu = ETH_MAX_MTU;
3789 u32 flags = conf->flags;
3792 if (flags & VXLAN_F_GPE)
3793 vxlan_raw_setup(dev);
3795 vxlan_ether_setup(dev);
3798 dev->mtu = conf->mtu;
3800 vxlan->net = src_net;
3803 dst->remote_vni = conf->vni;
3805 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3808 dst->remote_ifindex = conf->remote_ifindex;
3810 dev->gso_max_size = lowerdev->gso_max_size;
3811 dev->gso_max_segs = lowerdev->gso_max_segs;
3813 needed_headroom = lowerdev->hard_header_len;
3814 needed_headroom += lowerdev->needed_headroom;
3816 dev->needed_tailroom = lowerdev->needed_tailroom;
3818 max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3819 if (max_mtu < ETH_MIN_MTU)
3820 max_mtu = ETH_MIN_MTU;
3822 if (!changelink && !conf->mtu)
3826 if (dev->mtu > max_mtu)
3829 if (flags & VXLAN_F_COLLECT_METADATA)
3830 flags |= VXLAN_F_IPV6;
3831 needed_headroom += vxlan_headroom(flags);
3832 dev->needed_headroom = needed_headroom;
3834 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3837 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3838 struct vxlan_config *conf, bool changelink,
3839 struct netlink_ext_ack *extack)
3841 struct vxlan_dev *vxlan = netdev_priv(dev);
3842 struct net_device *lowerdev;
3845 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3849 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3854 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3855 struct vxlan_config *conf,
3856 struct netlink_ext_ack *extack)
3858 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3859 struct vxlan_dev *vxlan = netdev_priv(dev);
3860 struct net_device *remote_dev = NULL;
3861 struct vxlan_fdb *f = NULL;
3862 bool unregister = false;
3863 struct vxlan_rdst *dst;
3866 dst = &vxlan->default_dst;
3867 err = vxlan_dev_configure(net, dev, conf, false, extack);
3871 dev->ethtool_ops = &vxlan_ethtool_ops;
3873 /* create an fdb entry for a valid default destination */
3874 if (!vxlan_addr_any(&dst->remote_ip)) {
3875 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3877 NUD_REACHABLE | NUD_PERMANENT,
3878 vxlan->cfg.dst_port,
3881 dst->remote_ifindex,
3882 NTF_SELF, 0, &f, extack);
3887 err = register_netdevice(dev);
3892 if (dst->remote_ifindex) {
3893 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3899 err = netdev_upper_dev_link(remote_dev, dev, extack);
3904 err = rtnl_configure_link(dev, NULL);
3909 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3911 /* notify default fdb entry */
3912 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3913 RTM_NEWNEIGH, true, extack);
3915 vxlan_fdb_destroy(vxlan, f, false, false);
3917 netdev_upper_dev_unlink(remote_dev, dev);
3922 list_add(&vxlan->next, &vn->vxlan_list);
3924 dst->remote_dev = remote_dev;
3928 netdev_upper_dev_unlink(remote_dev, dev);
3930 /* unregister_netdevice() destroys the default FDB entry with deletion
3931 * notification. But the addition notification was not sent yet, so
3932 * destroy the entry by hand here.
3935 __vxlan_fdb_free(f);
3938 unregister_netdevice(dev);
3942 /* Set/clear flags based on attribute */
3943 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3944 int attrtype, unsigned long mask, bool changelink,
3945 bool changelink_supported,
3946 struct netlink_ext_ack *extack)
3948 unsigned long flags;
3953 if (changelink && !changelink_supported) {
3954 vxlan_flag_attr_error(attrtype, extack);
3958 if (vxlan_policy[attrtype].type == NLA_FLAG)
3959 flags = conf->flags | mask;
3960 else if (nla_get_u8(tb[attrtype]))
3961 flags = conf->flags | mask;
3963 flags = conf->flags & ~mask;
3965 conf->flags = flags;
3970 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3971 struct net_device *dev, struct vxlan_config *conf,
3972 bool changelink, struct netlink_ext_ack *extack)
3974 struct vxlan_dev *vxlan = netdev_priv(dev);
3977 memset(conf, 0, sizeof(*conf));
3979 /* if changelink operation, start with old existing cfg */
3981 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3983 if (data[IFLA_VXLAN_ID]) {
3984 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3986 if (changelink && (vni != conf->vni)) {
3987 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3990 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3993 if (data[IFLA_VXLAN_GROUP]) {
3994 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3995 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3999 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4000 conf->remote_ip.sa.sa_family = AF_INET;
4001 } else if (data[IFLA_VXLAN_GROUP6]) {
4002 if (!IS_ENABLED(CONFIG_IPV6)) {
4003 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4004 return -EPFNOSUPPORT;
4007 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4008 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4012 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4013 conf->remote_ip.sa.sa_family = AF_INET6;
4016 if (data[IFLA_VXLAN_LOCAL]) {
4017 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4018 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4022 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4023 conf->saddr.sa.sa_family = AF_INET;
4024 } else if (data[IFLA_VXLAN_LOCAL6]) {
4025 if (!IS_ENABLED(CONFIG_IPV6)) {
4026 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4027 return -EPFNOSUPPORT;
4030 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4031 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4035 /* TODO: respect scope id */
4036 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4037 conf->saddr.sa.sa_family = AF_INET6;
4040 if (data[IFLA_VXLAN_LINK])
4041 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4043 if (data[IFLA_VXLAN_TOS])
4044 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
4046 if (data[IFLA_VXLAN_TTL])
4047 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4049 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4050 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4051 VXLAN_F_TTL_INHERIT, changelink, false,
4058 if (data[IFLA_VXLAN_LABEL])
4059 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4060 IPV6_FLOWLABEL_MASK;
4062 if (data[IFLA_VXLAN_LEARNING]) {
4063 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4064 VXLAN_F_LEARN, changelink, true,
4068 } else if (!changelink) {
4069 /* default to learn on a new device */
4070 conf->flags |= VXLAN_F_LEARN;
4073 if (data[IFLA_VXLAN_AGEING])
4074 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4076 if (data[IFLA_VXLAN_PROXY]) {
4077 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4078 VXLAN_F_PROXY, changelink, false,
4084 if (data[IFLA_VXLAN_RSC]) {
4085 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4086 VXLAN_F_RSC, changelink, false,
4092 if (data[IFLA_VXLAN_L2MISS]) {
4093 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4094 VXLAN_F_L2MISS, changelink, false,
4100 if (data[IFLA_VXLAN_L3MISS]) {
4101 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4102 VXLAN_F_L3MISS, changelink, false,
4108 if (data[IFLA_VXLAN_LIMIT]) {
4110 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4111 "Cannot change limit");
4114 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4117 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4118 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4119 VXLAN_F_COLLECT_METADATA, changelink, false,
4125 if (data[IFLA_VXLAN_PORT_RANGE]) {
4127 const struct ifla_vxlan_port_range *p
4128 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4129 conf->port_min = ntohs(p->low);
4130 conf->port_max = ntohs(p->high);
4132 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4133 "Cannot change port range");
4138 if (data[IFLA_VXLAN_PORT]) {
4140 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4141 "Cannot change port");
4144 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4147 if (data[IFLA_VXLAN_UDP_CSUM]) {
4149 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4150 "Cannot change UDP_CSUM flag");
4153 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4154 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4157 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4158 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4159 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4165 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4166 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4167 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4173 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4174 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4175 VXLAN_F_REMCSUM_TX, changelink, false,
4181 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4182 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4183 VXLAN_F_REMCSUM_RX, changelink, false,
4189 if (data[IFLA_VXLAN_GBP]) {
4190 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4191 VXLAN_F_GBP, changelink, false, extack);
4196 if (data[IFLA_VXLAN_GPE]) {
4197 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4198 VXLAN_F_GPE, changelink, false,
4204 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4205 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4206 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4214 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4215 "Cannot change mtu");
4218 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4221 if (data[IFLA_VXLAN_DF])
4222 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4227 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4228 struct nlattr *tb[], struct nlattr *data[],
4229 struct netlink_ext_ack *extack)
4231 struct vxlan_config conf;
4234 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4238 return __vxlan_dev_create(src_net, dev, &conf, extack);
4241 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4242 struct nlattr *data[],
4243 struct netlink_ext_ack *extack)
4245 struct vxlan_dev *vxlan = netdev_priv(dev);
4246 struct net_device *lowerdev;
4247 struct vxlan_config conf;
4248 struct vxlan_rdst *dst;
4251 dst = &vxlan->default_dst;
4252 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4256 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4261 if (dst->remote_dev == lowerdev)
4264 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4269 /* handle default dst entry */
4270 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4271 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4273 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4274 if (!vxlan_addr_any(&conf.remote_ip)) {
4275 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4277 NUD_REACHABLE | NUD_PERMANENT,
4278 NLM_F_APPEND | NLM_F_CREATE,
4279 vxlan->cfg.dst_port,
4281 conf.remote_ifindex,
4282 NTF_SELF, 0, true, extack);
4284 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4285 netdev_adjacent_change_abort(dst->remote_dev,
4290 if (!vxlan_addr_any(&dst->remote_ip))
4291 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4293 vxlan->cfg.dst_port,
4296 dst->remote_ifindex,
4298 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4301 if (conf.age_interval != vxlan->cfg.age_interval)
4302 mod_timer(&vxlan->age_timer, jiffies);
4304 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4305 if (lowerdev && lowerdev != dst->remote_dev)
4306 dst->remote_dev = lowerdev;
4307 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4311 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4313 struct vxlan_dev *vxlan = netdev_priv(dev);
4315 vxlan_flush(vxlan, true);
4317 list_del(&vxlan->next);
4318 unregister_netdevice_queue(dev, head);
4319 if (vxlan->default_dst.remote_dev)
4320 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4323 static size_t vxlan_get_size(const struct net_device *dev)
4326 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4327 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4328 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4329 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4330 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4331 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4332 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4333 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4334 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4338 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4339 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4340 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4341 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4342 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4343 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4344 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4345 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4346 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4347 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4348 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4349 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4353 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4355 const struct vxlan_dev *vxlan = netdev_priv(dev);
4356 const struct vxlan_rdst *dst = &vxlan->default_dst;
4357 struct ifla_vxlan_port_range ports = {
4358 .low = htons(vxlan->cfg.port_min),
4359 .high = htons(vxlan->cfg.port_max),
4362 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4363 goto nla_put_failure;
4365 if (!vxlan_addr_any(&dst->remote_ip)) {
4366 if (dst->remote_ip.sa.sa_family == AF_INET) {
4367 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4368 dst->remote_ip.sin.sin_addr.s_addr))
4369 goto nla_put_failure;
4370 #if IS_ENABLED(CONFIG_IPV6)
4372 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4373 &dst->remote_ip.sin6.sin6_addr))
4374 goto nla_put_failure;
4379 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4380 goto nla_put_failure;
4382 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4383 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4384 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4385 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4386 goto nla_put_failure;
4387 #if IS_ENABLED(CONFIG_IPV6)
4389 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4390 &vxlan->cfg.saddr.sin6.sin6_addr))
4391 goto nla_put_failure;
4396 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4397 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4398 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4399 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4400 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4401 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4402 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4403 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4404 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4405 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4406 nla_put_u8(skb, IFLA_VXLAN_RSC,
4407 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4408 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4409 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4410 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4411 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4412 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4413 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4414 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4415 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4416 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4417 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4418 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4419 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4420 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4421 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4422 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4423 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4424 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4425 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4426 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4427 goto nla_put_failure;
4429 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4430 goto nla_put_failure;
4432 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4433 nla_put_flag(skb, IFLA_VXLAN_GBP))
4434 goto nla_put_failure;
4436 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4437 nla_put_flag(skb, IFLA_VXLAN_GPE))
4438 goto nla_put_failure;
4440 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4441 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4442 goto nla_put_failure;
4450 static struct net *vxlan_get_link_net(const struct net_device *dev)
4452 struct vxlan_dev *vxlan = netdev_priv(dev);
4457 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4459 .maxtype = IFLA_VXLAN_MAX,
4460 .policy = vxlan_policy,
4461 .priv_size = sizeof(struct vxlan_dev),
4462 .setup = vxlan_setup,
4463 .validate = vxlan_validate,
4464 .newlink = vxlan_newlink,
4465 .changelink = vxlan_changelink,
4466 .dellink = vxlan_dellink,
4467 .get_size = vxlan_get_size,
4468 .fill_info = vxlan_fill_info,
4469 .get_link_net = vxlan_get_link_net,
4472 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4473 u8 name_assign_type,
4474 struct vxlan_config *conf)
4476 struct nlattr *tb[IFLA_MAX + 1];
4477 struct net_device *dev;
4480 memset(&tb, 0, sizeof(tb));
4482 dev = rtnl_create_link(net, name, name_assign_type,
4483 &vxlan_link_ops, tb, NULL);
4487 err = __vxlan_dev_create(net, dev, conf, NULL);
4490 return ERR_PTR(err);
4493 err = rtnl_configure_link(dev, NULL);
4495 LIST_HEAD(list_kill);
4497 vxlan_dellink(dev, &list_kill);
4498 unregister_netdevice_many(&list_kill);
4499 return ERR_PTR(err);
4504 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4506 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4507 struct net_device *dev)
4509 struct vxlan_dev *vxlan, *next;
4510 LIST_HEAD(list_kill);
4512 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4513 struct vxlan_rdst *dst = &vxlan->default_dst;
4515 /* In case we created vxlan device with carrier
4516 * and we loose the carrier due to module unload
4517 * we also need to remove vxlan device. In other
4518 * cases, it's not necessary and remote_ifindex
4519 * is 0 here, so no matches.
4521 if (dst->remote_ifindex == dev->ifindex)
4522 vxlan_dellink(vxlan->dev, &list_kill);
4525 unregister_netdevice_many(&list_kill);
4528 static int vxlan_netdevice_event(struct notifier_block *unused,
4529 unsigned long event, void *ptr)
4531 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4532 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4534 if (event == NETDEV_UNREGISTER) {
4535 if (!dev->udp_tunnel_nic_info)
4536 vxlan_offload_rx_ports(dev, false);
4537 vxlan_handle_lowerdev_unregister(vn, dev);
4538 } else if (event == NETDEV_REGISTER) {
4539 if (!dev->udp_tunnel_nic_info)
4540 vxlan_offload_rx_ports(dev, true);
4541 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4542 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4543 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4549 static struct notifier_block vxlan_notifier_block __read_mostly = {
4550 .notifier_call = vxlan_netdevice_event,
4554 vxlan_fdb_offloaded_set(struct net_device *dev,
4555 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4557 struct vxlan_dev *vxlan = netdev_priv(dev);
4558 struct vxlan_rdst *rdst;
4559 struct vxlan_fdb *f;
4562 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4564 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4566 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4570 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4571 fdb_info->remote_port,
4572 fdb_info->remote_vni,
4573 fdb_info->remote_ifindex);
4577 rdst->offloaded = fdb_info->offloaded;
4580 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4584 vxlan_fdb_external_learn_add(struct net_device *dev,
4585 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4587 struct vxlan_dev *vxlan = netdev_priv(dev);
4588 struct netlink_ext_ack *extack;
4592 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4593 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4595 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4596 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4598 NLM_F_CREATE | NLM_F_REPLACE,
4599 fdb_info->remote_port,
4601 fdb_info->remote_vni,
4602 fdb_info->remote_ifindex,
4603 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4605 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4611 vxlan_fdb_external_learn_del(struct net_device *dev,
4612 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4614 struct vxlan_dev *vxlan = netdev_priv(dev);
4615 struct vxlan_fdb *f;
4619 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4620 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4622 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4625 else if (f->flags & NTF_EXT_LEARNED)
4626 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4627 fdb_info->remote_ip,
4628 fdb_info->remote_port,
4630 fdb_info->remote_vni,
4631 fdb_info->remote_ifindex,
4634 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4639 static int vxlan_switchdev_event(struct notifier_block *unused,
4640 unsigned long event, void *ptr)
4642 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4643 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4647 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4648 vxlan_fdb_offloaded_set(dev, ptr);
4650 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4652 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4654 err = notifier_from_errno(err);
4657 fdb_info->offloaded = true;
4658 vxlan_fdb_offloaded_set(dev, fdb_info);
4660 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4662 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4664 err = notifier_from_errno(err);
4667 fdb_info->offloaded = false;
4668 vxlan_fdb_offloaded_set(dev, fdb_info);
4675 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4676 .notifier_call = vxlan_switchdev_event,
4679 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4681 struct vxlan_fdb *fdb;
4682 struct vxlan_dev *vxlan;
4686 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4687 vxlan = rcu_dereference(fdb->vdev);
4689 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4690 vxlan->default_dst.remote_vni);
4691 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4692 if (!hlist_unhashed(&fdb->hlist))
4693 vxlan_fdb_destroy(vxlan, fdb, false, false);
4694 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4699 static int vxlan_nexthop_event(struct notifier_block *nb,
4700 unsigned long event, void *ptr)
4702 struct nexthop *nh = ptr;
4704 if (!nh || event != NEXTHOP_EVENT_DEL)
4707 vxlan_fdb_nh_flush(nh);
4712 static struct notifier_block vxlan_nexthop_notifier_block __read_mostly = {
4713 .notifier_call = vxlan_nexthop_event,
4716 static __net_init int vxlan_init_net(struct net *net)
4718 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4721 INIT_LIST_HEAD(&vn->vxlan_list);
4722 spin_lock_init(&vn->sock_lock);
4724 for (h = 0; h < PORT_HASH_SIZE; ++h)
4725 INIT_HLIST_HEAD(&vn->sock_list[h]);
4727 return register_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4730 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4732 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4733 struct vxlan_dev *vxlan, *next;
4734 struct net_device *dev, *aux;
4736 for_each_netdev_safe(net, dev, aux)
4737 if (dev->rtnl_link_ops == &vxlan_link_ops)
4738 unregister_netdevice_queue(dev, head);
4740 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4741 /* If vxlan->dev is in the same netns, it has already been added
4742 * to the list by the previous loop.
4744 if (!net_eq(dev_net(vxlan->dev), net))
4745 unregister_netdevice_queue(vxlan->dev, head);
4750 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4757 list_for_each_entry(net, net_list, exit_list)
4758 unregister_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4759 list_for_each_entry(net, net_list, exit_list)
4760 vxlan_destroy_tunnels(net, &list);
4762 unregister_netdevice_many(&list);
4765 list_for_each_entry(net, net_list, exit_list) {
4766 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4768 for (h = 0; h < PORT_HASH_SIZE; ++h)
4769 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4773 static struct pernet_operations vxlan_net_ops = {
4774 .init = vxlan_init_net,
4775 .exit_batch = vxlan_exit_batch_net,
4776 .id = &vxlan_net_id,
4777 .size = sizeof(struct vxlan_net),
4780 static int __init vxlan_init_module(void)
4784 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4786 rc = register_pernet_subsys(&vxlan_net_ops);
4790 rc = register_netdevice_notifier(&vxlan_notifier_block);
4794 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4798 rc = rtnl_link_register(&vxlan_link_ops);
4804 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4806 unregister_netdevice_notifier(&vxlan_notifier_block);
4808 unregister_pernet_subsys(&vxlan_net_ops);
4812 late_initcall(vxlan_init_module);
4814 static void __exit vxlan_cleanup_module(void)
4816 rtnl_link_unregister(&vxlan_link_ops);
4817 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4818 unregister_netdevice_notifier(&vxlan_notifier_block);
4819 unregister_pernet_subsys(&vxlan_net_ops);
4820 /* rcu_barrier() is called by netns */
4822 module_exit(vxlan_cleanup_module);
4824 MODULE_LICENSE("GPL");
4825 MODULE_VERSION(VXLAN_VERSION);
4826 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4827 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4828 MODULE_ALIAS_RTNL_LINK("vxlan");