2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/if_ether.h>
20 #include <linux/ethtool.h>
22 #include <net/ndisc.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/tun_proto.h>
30 #include <net/vxlan.h>
32 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_tunnel.h>
34 #include <net/ip6_checksum.h>
37 #define VXLAN_VERSION "0.1"
39 #define PORT_HASH_BITS 8
40 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
41 #define FDB_AGE_DEFAULT 300 /* 5 min */
42 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
44 /* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
46 * for compatibility with early adopters.
48 static unsigned short vxlan_port __read_mostly = 8472;
49 module_param_named(udp_port, vxlan_port, ushort, 0444);
50 MODULE_PARM_DESC(udp_port, "Destination UDP port");
52 static bool log_ecn_error = true;
53 module_param(log_ecn_error, bool, 0644);
54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
56 static unsigned int vxlan_net_id;
57 static struct rtnl_link_ops vxlan_link_ops;
59 static const u8 all_zeros_mac[ETH_ALEN + 2];
61 static int vxlan_sock_add(struct vxlan_dev *vxlan);
63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
65 /* per-network namespace private data for this module */
67 struct list_head vxlan_list;
68 struct hlist_head sock_list[PORT_HASH_SIZE];
72 /* Forwarding table entry */
74 struct hlist_node hlist; /* linked list of entries */
76 unsigned long updated; /* jiffies */
78 struct list_head remotes;
79 u8 eth_addr[ETH_ALEN];
80 u16 state; /* see ndm_state */
82 u8 flags; /* see ndm_flags */
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly;
88 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
94 #if IS_ENABLED(CONFIG_IPV6)
96 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
98 if (a->sa.sa_family != b->sa.sa_family)
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
106 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
108 if (nla_len(nla) >= sizeof(struct in6_addr)) {
109 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
110 ip->sa.sa_family = AF_INET6;
112 } else if (nla_len(nla) >= sizeof(__be32)) {
113 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
114 ip->sa.sa_family = AF_INET;
117 return -EAFNOSUPPORT;
121 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
122 const union vxlan_addr *ip)
124 if (ip->sa.sa_family == AF_INET6)
125 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
127 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
130 #else /* !CONFIG_IPV6 */
133 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
135 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
138 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
141 return -EAFNOSUPPORT;
142 } else if (nla_len(nla) >= sizeof(__be32)) {
143 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
144 ip->sa.sa_family = AF_INET;
147 return -EAFNOSUPPORT;
151 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
152 const union vxlan_addr *ip)
154 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
158 /* Virtual Network hash table head */
159 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
161 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
164 /* Socket hash table head */
165 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
167 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
169 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
172 /* First remote destination for a forwarding entry.
173 * Guaranteed to be non-NULL because remotes are never deleted.
175 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
177 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
180 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
182 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
185 /* Find VXLAN socket based on network namespace, address family and UDP port
186 * and enabled unshareable flags.
188 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
189 __be16 port, u32 flags)
191 struct vxlan_sock *vs;
193 flags &= VXLAN_F_RCV_FLAGS;
195 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
196 if (inet_sk(vs->sock->sk)->inet_sport == port &&
197 vxlan_get_sk_family(vs) == family &&
204 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
207 struct vxlan_dev_node *node;
209 /* For flow based devices, map all packets to VNI 0 */
210 if (vs->flags & VXLAN_F_COLLECT_METADATA)
213 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
214 if (node->vxlan->default_dst.remote_vni != vni)
217 if (IS_ENABLED(CONFIG_IPV6)) {
218 const struct vxlan_config *cfg = &node->vxlan->cfg;
220 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
221 cfg->remote_ifindex != ifindex)
231 /* Look up VNI in a per net namespace table */
232 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
233 __be32 vni, sa_family_t family,
234 __be16 port, u32 flags)
236 struct vxlan_sock *vs;
238 vs = vxlan_find_sock(net, family, port, flags);
242 return vxlan_vs_find_vni(vs, ifindex, vni);
245 /* Fill in neighbour message in skbuff. */
246 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
247 const struct vxlan_fdb *fdb,
248 u32 portid, u32 seq, int type, unsigned int flags,
249 const struct vxlan_rdst *rdst)
251 unsigned long now = jiffies;
252 struct nda_cacheinfo ci;
253 struct nlmsghdr *nlh;
255 bool send_ip, send_eth;
257 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
261 ndm = nlmsg_data(nlh);
262 memset(ndm, 0, sizeof(*ndm));
264 send_eth = send_ip = true;
266 if (type == RTM_GETNEIGH) {
267 send_ip = !vxlan_addr_any(&rdst->remote_ip);
268 send_eth = !is_zero_ether_addr(fdb->eth_addr);
269 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
271 ndm->ndm_family = AF_BRIDGE;
272 ndm->ndm_state = fdb->state;
273 ndm->ndm_ifindex = vxlan->dev->ifindex;
274 ndm->ndm_flags = fdb->flags;
276 ndm->ndm_flags |= NTF_OFFLOADED;
277 ndm->ndm_type = RTN_UNICAST;
279 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
280 nla_put_s32(skb, NDA_LINK_NETNSID,
281 peernet2id(dev_net(vxlan->dev), vxlan->net)))
282 goto nla_put_failure;
284 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
285 goto nla_put_failure;
287 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
288 goto nla_put_failure;
290 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
291 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
292 goto nla_put_failure;
293 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
294 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
295 goto nla_put_failure;
296 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
297 nla_put_u32(skb, NDA_SRC_VNI,
298 be32_to_cpu(fdb->vni)))
299 goto nla_put_failure;
300 if (rdst->remote_ifindex &&
301 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
302 goto nla_put_failure;
304 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
305 ci.ndm_confirmed = 0;
306 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
309 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
310 goto nla_put_failure;
316 nlmsg_cancel(skb, nlh);
320 static inline size_t vxlan_nlmsg_size(void)
322 return NLMSG_ALIGN(sizeof(struct ndmsg))
323 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
324 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
325 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
326 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
327 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
328 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
329 + nla_total_size(sizeof(struct nda_cacheinfo));
332 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
333 struct vxlan_rdst *rd, int type)
335 struct net *net = dev_net(vxlan->dev);
339 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
343 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
345 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
346 WARN_ON(err == -EMSGSIZE);
351 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
355 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
358 static void vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
359 struct vxlan_fdb *fdb,
360 struct vxlan_rdst *rd,
363 struct switchdev_notifier_vxlan_fdb_info info;
364 enum switchdev_notifier_type notifier_type;
369 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
370 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
372 info = (struct switchdev_notifier_vxlan_fdb_info){
373 .remote_ip = rd->remote_ip,
374 .remote_port = rd->remote_port,
375 .remote_vni = rd->remote_vni,
376 .remote_ifindex = rd->remote_ifindex,
378 .offloaded = rd->offloaded,
380 memcpy(info.eth_addr, fdb->eth_addr, ETH_ALEN);
382 call_switchdev_notifiers(notifier_type, vxlan->dev,
386 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
387 struct vxlan_rdst *rd, int type)
391 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, true);
394 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, false);
398 __vxlan_fdb_notify(vxlan, fdb, rd, type);
401 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
403 struct vxlan_dev *vxlan = netdev_priv(dev);
404 struct vxlan_fdb f = {
407 struct vxlan_rdst remote = {
408 .remote_ip = *ipa, /* goes to NDA_DST */
409 .remote_vni = cpu_to_be32(VXLAN_N_VID),
412 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
415 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
417 struct vxlan_fdb f = {
420 struct vxlan_rdst remote = { };
422 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
424 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
427 /* Hash Ethernet address */
428 static u32 eth_hash(const unsigned char *addr)
430 u64 value = get_unaligned((u64 *)addr);
432 /* only want 6 bytes */
438 return hash_64(value, FDB_HASH_BITS);
441 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
443 /* use 1 byte of OUI and 3 bytes of NIC */
444 u32 key = get_unaligned((u32 *)(addr + 2));
446 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
449 /* Hash chain to use given mac address */
450 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
451 const u8 *mac, __be32 vni)
453 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
454 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
456 return &vxlan->fdb_head[eth_hash(mac)];
459 /* Look up Ethernet address in forwarding table */
460 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
461 const u8 *mac, __be32 vni)
463 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
466 hlist_for_each_entry_rcu(f, head, hlist) {
467 if (ether_addr_equal(mac, f->eth_addr)) {
468 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
480 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
481 const u8 *mac, __be32 vni)
485 f = __vxlan_find_mac(vxlan, mac, vni);
486 if (f && f->used != jiffies)
492 /* caller should hold vxlan->hash_lock */
493 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
494 union vxlan_addr *ip, __be16 port,
495 __be32 vni, __u32 ifindex)
497 struct vxlan_rdst *rd;
499 list_for_each_entry(rd, &f->remotes, list) {
500 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
501 rd->remote_port == port &&
502 rd->remote_vni == vni &&
503 rd->remote_ifindex == ifindex)
510 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
511 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
513 struct vxlan_dev *vxlan = netdev_priv(dev);
514 u8 eth_addr[ETH_ALEN + 2] = { 0 };
515 struct vxlan_rdst *rdst;
519 if (is_multicast_ether_addr(mac) ||
520 is_zero_ether_addr(mac))
523 ether_addr_copy(eth_addr, mac);
527 f = __vxlan_find_mac(vxlan, eth_addr, vni);
533 rdst = first_remote_rcu(f);
535 memset(fdb_info, 0, sizeof(*fdb_info));
536 fdb_info->info.dev = dev;
537 fdb_info->remote_ip = rdst->remote_ip;
538 fdb_info->remote_port = rdst->remote_port;
539 fdb_info->remote_vni = rdst->remote_vni;
540 fdb_info->remote_ifindex = rdst->remote_ifindex;
542 fdb_info->offloaded = rdst->offloaded;
543 ether_addr_copy(fdb_info->eth_addr, mac);
549 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
551 /* Replace destination of unicast mac */
552 static int vxlan_fdb_replace(struct vxlan_fdb *f,
553 union vxlan_addr *ip, __be16 port, __be32 vni,
556 struct vxlan_rdst *rd;
558 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
562 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
566 dst_cache_reset(&rd->dst_cache);
568 rd->remote_port = port;
569 rd->remote_vni = vni;
570 rd->remote_ifindex = ifindex;
571 rd->offloaded = false;
575 /* Add/update destinations for multicast */
576 static int vxlan_fdb_append(struct vxlan_fdb *f,
577 union vxlan_addr *ip, __be16 port, __be32 vni,
578 __u32 ifindex, struct vxlan_rdst **rdp)
580 struct vxlan_rdst *rd;
582 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
586 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
590 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
596 rd->remote_port = port;
597 rd->offloaded = false;
598 rd->remote_vni = vni;
599 rd->remote_ifindex = ifindex;
601 list_add_tail_rcu(&rd->list, &f->remotes);
607 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
609 struct vxlanhdr *vh, size_t hdrlen,
611 struct gro_remcsum *grc,
614 size_t start, offset;
616 if (skb->remcsum_offload)
619 if (!NAPI_GRO_CB(skb)->csum_valid)
622 start = vxlan_rco_start(vni_field);
623 offset = start + vxlan_rco_offset(vni_field);
625 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
626 start, offset, grc, nopartial);
628 skb->remcsum_offload = 1;
633 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
634 struct list_head *head,
637 struct sk_buff *pp = NULL;
639 struct vxlanhdr *vh, *vh2;
640 unsigned int hlen, off_vx;
642 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
644 struct gro_remcsum grc;
646 skb_gro_remcsum_init(&grc);
648 off_vx = skb_gro_offset(skb);
649 hlen = off_vx + sizeof(*vh);
650 vh = skb_gro_header_fast(skb, off_vx);
651 if (skb_gro_header_hard(skb, hlen)) {
652 vh = skb_gro_header_slow(skb, hlen, off_vx);
657 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
659 flags = vh->vx_flags;
661 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
662 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
665 VXLAN_F_REMCSUM_NOPARTIAL));
671 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
673 list_for_each_entry(p, head, list) {
674 if (!NAPI_GRO_CB(p)->same_flow)
677 vh2 = (struct vxlanhdr *)(p->data + off_vx);
678 if (vh->vx_flags != vh2->vx_flags ||
679 vh->vx_vni != vh2->vx_vni) {
680 NAPI_GRO_CB(p)->same_flow = 0;
685 pp = call_gro_receive(eth_gro_receive, head, skb);
689 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
694 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
696 /* Sets 'skb->inner_mac_header' since we are always called with
697 * 'skb->encapsulation' set.
699 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
702 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
703 const u8 *mac, __u16 state,
704 __be32 src_vni, __u8 ndm_flags)
708 f = kmalloc(sizeof(*f), GFP_ATOMIC);
712 f->flags = ndm_flags;
713 f->updated = f->used = jiffies;
715 INIT_LIST_HEAD(&f->remotes);
716 memcpy(f->eth_addr, mac, ETH_ALEN);
721 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
722 const u8 *mac, union vxlan_addr *ip,
723 __u16 state, __be16 port, __be32 src_vni,
724 __be32 vni, __u32 ifindex, __u8 ndm_flags,
725 struct vxlan_fdb **fdb)
727 struct vxlan_rdst *rd = NULL;
731 if (vxlan->cfg.addrmax &&
732 vxlan->addrcnt >= vxlan->cfg.addrmax)
735 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
736 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
740 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
747 hlist_add_head_rcu(&f->hlist,
748 vxlan_fdb_head(vxlan, mac, src_vni));
755 /* Add new entry to forwarding table -- assumes lock held */
756 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
757 const u8 *mac, union vxlan_addr *ip,
758 __u16 state, __u16 flags,
759 __be16 port, __be32 src_vni, __be32 vni,
760 __u32 ifindex, __u8 ndm_flags)
762 __u8 fdb_flags = (ndm_flags & ~NTF_USE);
763 struct vxlan_rdst *rd = NULL;
768 f = __vxlan_find_mac(vxlan, mac, src_vni);
770 if (flags & NLM_F_EXCL) {
771 netdev_dbg(vxlan->dev,
772 "lost race to create %pM\n", mac);
775 if (f->state != state) {
777 f->updated = jiffies;
780 if (f->flags != fdb_flags) {
781 f->flags = fdb_flags;
782 f->updated = jiffies;
785 if ((flags & NLM_F_REPLACE)) {
786 /* Only change unicasts */
787 if (!(is_multicast_ether_addr(f->eth_addr) ||
788 is_zero_ether_addr(f->eth_addr))) {
789 notify |= vxlan_fdb_replace(f, ip, port, vni,
794 if ((flags & NLM_F_APPEND) &&
795 (is_multicast_ether_addr(f->eth_addr) ||
796 is_zero_ether_addr(f->eth_addr))) {
797 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
804 if (ndm_flags & NTF_USE)
807 if (!(flags & NLM_F_CREATE))
810 /* Disallow replace to add a multicast entry */
811 if ((flags & NLM_F_REPLACE) &&
812 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
815 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
816 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
817 vni, ifindex, fdb_flags, &f);
825 rd = first_remote_rtnl(f);
826 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
832 static void vxlan_fdb_free(struct rcu_head *head)
834 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
835 struct vxlan_rdst *rd, *nd;
837 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
838 dst_cache_destroy(&rd->dst_cache);
844 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
847 struct vxlan_rdst *rd;
849 netdev_dbg(vxlan->dev,
850 "delete %pM\n", f->eth_addr);
854 list_for_each_entry(rd, &f->remotes, list)
855 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
857 hlist_del_rcu(&f->hlist);
858 call_rcu(&f->rcu, vxlan_fdb_free);
861 static void vxlan_dst_free(struct rcu_head *head)
863 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
865 dst_cache_destroy(&rd->dst_cache);
869 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
870 struct vxlan_rdst *rd)
872 list_del_rcu(&rd->list);
873 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
874 call_rcu(&rd->rcu, vxlan_dst_free);
877 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
878 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
879 __be32 *vni, u32 *ifindex)
881 struct net *net = dev_net(vxlan->dev);
885 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
889 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
890 if (remote->sa.sa_family == AF_INET) {
891 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
892 ip->sa.sa_family = AF_INET;
893 #if IS_ENABLED(CONFIG_IPV6)
895 ip->sin6.sin6_addr = in6addr_any;
896 ip->sa.sa_family = AF_INET6;
902 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
904 *port = nla_get_be16(tb[NDA_PORT]);
906 *port = vxlan->cfg.dst_port;
910 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
912 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
914 *vni = vxlan->default_dst.remote_vni;
917 if (tb[NDA_SRC_VNI]) {
918 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
920 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
922 *src_vni = vxlan->default_dst.remote_vni;
925 if (tb[NDA_IFINDEX]) {
926 struct net_device *tdev;
928 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
930 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
931 tdev = __dev_get_by_index(net, *ifindex);
933 return -EADDRNOTAVAIL;
941 /* Add static entry (via netlink) */
942 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
943 struct net_device *dev,
944 const unsigned char *addr, u16 vid, u16 flags)
946 struct vxlan_dev *vxlan = netdev_priv(dev);
947 /* struct net *net = dev_net(vxlan->dev); */
954 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
955 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
960 if (tb[NDA_DST] == NULL)
963 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
967 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
968 return -EAFNOSUPPORT;
970 spin_lock_bh(&vxlan->hash_lock);
971 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
972 port, src_vni, vni, ifindex, ndm->ndm_flags);
973 spin_unlock_bh(&vxlan->hash_lock);
978 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
979 const unsigned char *addr, union vxlan_addr ip,
980 __be16 port, __be32 src_vni, __be32 vni,
981 u32 ifindex, u16 vid)
984 struct vxlan_rdst *rd = NULL;
987 f = vxlan_find_mac(vxlan, addr, src_vni);
991 if (!vxlan_addr_any(&ip)) {
992 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
997 /* remove a destination if it's not the only one on the list,
998 * otherwise destroy the fdb entry
1000 if (rd && !list_is_singular(&f->remotes)) {
1001 vxlan_fdb_dst_destroy(vxlan, f, rd);
1005 vxlan_fdb_destroy(vxlan, f, true);
1011 /* Delete entry (via netlink) */
1012 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1013 struct net_device *dev,
1014 const unsigned char *addr, u16 vid)
1016 struct vxlan_dev *vxlan = netdev_priv(dev);
1017 union vxlan_addr ip;
1018 __be32 src_vni, vni;
1023 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1027 spin_lock_bh(&vxlan->hash_lock);
1028 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1030 spin_unlock_bh(&vxlan->hash_lock);
1035 /* Dump forwarding table */
1036 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1037 struct net_device *dev,
1038 struct net_device *filter_dev, int *idx)
1040 struct vxlan_dev *vxlan = netdev_priv(dev);
1044 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1045 struct vxlan_fdb *f;
1047 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1048 struct vxlan_rdst *rd;
1050 list_for_each_entry_rcu(rd, &f->remotes, list) {
1051 if (*idx < cb->args[2])
1054 err = vxlan_fdb_info(skb, vxlan, f,
1055 NETLINK_CB(cb->skb).portid,
1070 /* Watch incoming packets to learn mapping between Ethernet address
1071 * and Tunnel endpoint.
1072 * Return true if packet is bogus and should be dropped.
1074 static bool vxlan_snoop(struct net_device *dev,
1075 union vxlan_addr *src_ip, const u8 *src_mac,
1076 u32 src_ifindex, __be32 vni)
1078 struct vxlan_dev *vxlan = netdev_priv(dev);
1079 struct vxlan_fdb *f;
1082 #if IS_ENABLED(CONFIG_IPV6)
1083 if (src_ip->sa.sa_family == AF_INET6 &&
1084 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1085 ifindex = src_ifindex;
1088 f = vxlan_find_mac(vxlan, src_mac, vni);
1090 struct vxlan_rdst *rdst = first_remote_rcu(f);
1092 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1093 rdst->remote_ifindex == ifindex))
1096 /* Don't migrate static entries, drop packets */
1097 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1100 if (net_ratelimit())
1102 "%pM migrated from %pIS to %pIS\n",
1103 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1105 rdst->remote_ip = *src_ip;
1106 f->updated = jiffies;
1107 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
1109 /* learned new entry */
1110 spin_lock(&vxlan->hash_lock);
1112 /* close off race between vxlan_flush and incoming packets */
1113 if (netif_running(dev))
1114 vxlan_fdb_update(vxlan, src_mac, src_ip,
1116 NLM_F_EXCL|NLM_F_CREATE,
1117 vxlan->cfg.dst_port,
1119 vxlan->default_dst.remote_vni,
1121 spin_unlock(&vxlan->hash_lock);
1127 /* See if multicast group is already in use by other ID */
1128 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1130 struct vxlan_dev *vxlan;
1131 struct vxlan_sock *sock4;
1132 #if IS_ENABLED(CONFIG_IPV6)
1133 struct vxlan_sock *sock6;
1135 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1137 sock4 = rtnl_dereference(dev->vn4_sock);
1139 /* The vxlan_sock is only used by dev, leaving group has
1140 * no effect on other vxlan devices.
1142 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1144 #if IS_ENABLED(CONFIG_IPV6)
1145 sock6 = rtnl_dereference(dev->vn6_sock);
1146 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1150 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1151 if (!netif_running(vxlan->dev) || vxlan == dev)
1154 if (family == AF_INET &&
1155 rtnl_dereference(vxlan->vn4_sock) != sock4)
1157 #if IS_ENABLED(CONFIG_IPV6)
1158 if (family == AF_INET6 &&
1159 rtnl_dereference(vxlan->vn6_sock) != sock6)
1163 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1164 &dev->default_dst.remote_ip))
1167 if (vxlan->default_dst.remote_ifindex !=
1168 dev->default_dst.remote_ifindex)
1177 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1179 struct vxlan_net *vn;
1183 if (!refcount_dec_and_test(&vs->refcnt))
1186 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1187 spin_lock(&vn->sock_lock);
1188 hlist_del_rcu(&vs->hlist);
1189 udp_tunnel_notify_del_rx_port(vs->sock,
1190 (vs->flags & VXLAN_F_GPE) ?
1191 UDP_TUNNEL_TYPE_VXLAN_GPE :
1192 UDP_TUNNEL_TYPE_VXLAN);
1193 spin_unlock(&vn->sock_lock);
1198 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1200 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1201 #if IS_ENABLED(CONFIG_IPV6)
1202 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1204 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1207 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1210 vxlan_vs_del_dev(vxlan);
1212 if (__vxlan_sock_release_prep(sock4)) {
1213 udp_tunnel_sock_release(sock4->sock);
1217 #if IS_ENABLED(CONFIG_IPV6)
1218 if (__vxlan_sock_release_prep(sock6)) {
1219 udp_tunnel_sock_release(sock6->sock);
1225 /* Update multicast group membership when first VNI on
1226 * multicast address is brought up
1228 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1231 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1232 int ifindex = vxlan->default_dst.remote_ifindex;
1235 if (ip->sa.sa_family == AF_INET) {
1236 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1237 struct ip_mreqn mreq = {
1238 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1239 .imr_ifindex = ifindex,
1242 sk = sock4->sock->sk;
1244 ret = ip_mc_join_group(sk, &mreq);
1246 #if IS_ENABLED(CONFIG_IPV6)
1248 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1250 sk = sock6->sock->sk;
1252 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1253 &ip->sin6.sin6_addr);
1261 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1262 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1265 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1266 int ifindex = vxlan->default_dst.remote_ifindex;
1269 if (ip->sa.sa_family == AF_INET) {
1270 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1271 struct ip_mreqn mreq = {
1272 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1273 .imr_ifindex = ifindex,
1276 sk = sock4->sock->sk;
1278 ret = ip_mc_leave_group(sk, &mreq);
1280 #if IS_ENABLED(CONFIG_IPV6)
1282 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1284 sk = sock6->sock->sk;
1286 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1287 &ip->sin6.sin6_addr);
1295 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1296 struct sk_buff *skb, u32 vxflags)
1298 size_t start, offset;
1300 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1303 start = vxlan_rco_start(unparsed->vx_vni);
1304 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1306 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1309 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1310 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1312 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1313 unparsed->vx_vni &= VXLAN_VNI_MASK;
1317 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1318 struct sk_buff *skb, u32 vxflags,
1319 struct vxlan_metadata *md)
1321 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1322 struct metadata_dst *tun_dst;
1324 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1327 md->gbp = ntohs(gbp->policy_id);
1329 tun_dst = (struct metadata_dst *)skb_dst(skb);
1331 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1332 tun_dst->u.tun_info.options_len = sizeof(*md);
1334 if (gbp->dont_learn)
1335 md->gbp |= VXLAN_GBP_DONT_LEARN;
1337 if (gbp->policy_applied)
1338 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1340 /* In flow-based mode, GBP is carried in dst_metadata */
1341 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1342 skb->mark = md->gbp;
1344 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1347 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1349 struct sk_buff *skb, u32 vxflags)
1351 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1353 /* Need to have Next Protocol set for interfaces in GPE mode. */
1354 if (!gpe->np_applied)
1356 /* "The initial version is 0. If a receiver does not support the
1357 * version indicated it MUST drop the packet.
1359 if (gpe->version != 0)
1361 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1362 * processing MUST occur." However, we don't implement OAM
1363 * processing, thus drop the packet.
1368 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1372 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1376 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1377 struct vxlan_sock *vs,
1378 struct sk_buff *skb, __be32 vni)
1380 union vxlan_addr saddr;
1381 u32 ifindex = skb->dev->ifindex;
1383 skb_reset_mac_header(skb);
1384 skb->protocol = eth_type_trans(skb, vxlan->dev);
1385 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1387 /* Ignore packet loops (and multicast echo) */
1388 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1391 /* Get address from the outer IP header */
1392 if (vxlan_get_sk_family(vs) == AF_INET) {
1393 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1394 saddr.sa.sa_family = AF_INET;
1395 #if IS_ENABLED(CONFIG_IPV6)
1397 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1398 saddr.sa.sa_family = AF_INET6;
1402 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1403 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1409 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1410 struct sk_buff *skb)
1414 if (vxlan_get_sk_family(vs) == AF_INET)
1415 err = IP_ECN_decapsulate(oiph, skb);
1416 #if IS_ENABLED(CONFIG_IPV6)
1418 err = IP6_ECN_decapsulate(oiph, skb);
1421 if (unlikely(err) && log_ecn_error) {
1422 if (vxlan_get_sk_family(vs) == AF_INET)
1423 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1424 &((struct iphdr *)oiph)->saddr,
1425 ((struct iphdr *)oiph)->tos);
1427 net_info_ratelimited("non-ECT from %pI6\n",
1428 &((struct ipv6hdr *)oiph)->saddr);
1433 /* Callback from net/ipv4/udp.c to receive packets */
1434 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1436 struct pcpu_sw_netstats *stats;
1437 struct vxlan_dev *vxlan;
1438 struct vxlan_sock *vs;
1439 struct vxlanhdr unparsed;
1440 struct vxlan_metadata _md;
1441 struct vxlan_metadata *md = &_md;
1442 __be16 protocol = htons(ETH_P_TEB);
1443 bool raw_proto = false;
1447 /* Need UDP and VXLAN header to be present */
1448 if (!pskb_may_pull(skb, VXLAN_HLEN))
1451 unparsed = *vxlan_hdr(skb);
1452 /* VNI flag always required to be set */
1453 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1454 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1455 ntohl(vxlan_hdr(skb)->vx_flags),
1456 ntohl(vxlan_hdr(skb)->vx_vni));
1457 /* Return non vxlan pkt */
1460 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1461 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1463 vs = rcu_dereference_sk_user_data(sk);
1467 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1469 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1473 /* For backwards compatibility, only allow reserved fields to be
1474 * used by VXLAN extensions if explicitly requested.
1476 if (vs->flags & VXLAN_F_GPE) {
1477 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1482 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1483 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1486 if (vxlan_collect_metadata(vs)) {
1487 struct metadata_dst *tun_dst;
1489 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1490 key32_to_tunnel_id(vni), sizeof(*md));
1495 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1497 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1499 memset(md, 0, sizeof(*md));
1502 if (vs->flags & VXLAN_F_REMCSUM_RX)
1503 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1505 if (vs->flags & VXLAN_F_GBP)
1506 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1507 /* Note that GBP and GPE can never be active together. This is
1508 * ensured in vxlan_dev_configure.
1511 if (unparsed.vx_flags || unparsed.vx_vni) {
1512 /* If there are any unprocessed flags remaining treat
1513 * this as a malformed packet. This behavior diverges from
1514 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1515 * in reserved fields are to be ignored. The approach here
1516 * maintains compatibility with previous stack code, and also
1517 * is more robust and provides a little more security in
1518 * adding extensions to VXLAN.
1524 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1527 skb_reset_mac_header(skb);
1528 skb->dev = vxlan->dev;
1529 skb->pkt_type = PACKET_HOST;
1532 oiph = skb_network_header(skb);
1533 skb_reset_network_header(skb);
1535 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1536 ++vxlan->dev->stats.rx_frame_errors;
1537 ++vxlan->dev->stats.rx_errors;
1541 stats = this_cpu_ptr(vxlan->dev->tstats);
1542 u64_stats_update_begin(&stats->syncp);
1543 stats->rx_packets++;
1544 stats->rx_bytes += skb->len;
1545 u64_stats_update_end(&stats->syncp);
1547 gro_cells_receive(&vxlan->gro_cells, skb);
1551 /* Consume bad packet */
1556 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1558 struct vxlan_dev *vxlan = netdev_priv(dev);
1559 struct arphdr *parp;
1562 struct neighbour *n;
1564 if (dev->flags & IFF_NOARP)
1567 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1568 dev->stats.tx_dropped++;
1571 parp = arp_hdr(skb);
1573 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1574 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1575 parp->ar_pro != htons(ETH_P_IP) ||
1576 parp->ar_op != htons(ARPOP_REQUEST) ||
1577 parp->ar_hln != dev->addr_len ||
1580 arpptr = (u8 *)parp + sizeof(struct arphdr);
1582 arpptr += dev->addr_len; /* sha */
1583 memcpy(&sip, arpptr, sizeof(sip));
1584 arpptr += sizeof(sip);
1585 arpptr += dev->addr_len; /* tha */
1586 memcpy(&tip, arpptr, sizeof(tip));
1588 if (ipv4_is_loopback(tip) ||
1589 ipv4_is_multicast(tip))
1592 n = neigh_lookup(&arp_tbl, &tip, dev);
1595 struct vxlan_fdb *f;
1596 struct sk_buff *reply;
1598 if (!(n->nud_state & NUD_CONNECTED)) {
1603 f = vxlan_find_mac(vxlan, n->ha, vni);
1604 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1605 /* bridge-local neighbor */
1610 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1618 skb_reset_mac_header(reply);
1619 __skb_pull(reply, skb_network_offset(reply));
1620 reply->ip_summed = CHECKSUM_UNNECESSARY;
1621 reply->pkt_type = PACKET_HOST;
1623 if (netif_rx_ni(reply) == NET_RX_DROP)
1624 dev->stats.rx_dropped++;
1625 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1626 union vxlan_addr ipa = {
1627 .sin.sin_addr.s_addr = tip,
1628 .sin.sin_family = AF_INET,
1631 vxlan_ip_miss(dev, &ipa);
1635 return NETDEV_TX_OK;
1638 #if IS_ENABLED(CONFIG_IPV6)
1639 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1640 struct neighbour *n, bool isrouter)
1642 struct net_device *dev = request->dev;
1643 struct sk_buff *reply;
1644 struct nd_msg *ns, *na;
1645 struct ipv6hdr *pip6;
1647 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1651 if (dev == NULL || !pskb_may_pull(request, request->len))
1654 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1655 sizeof(*na) + na_olen + dev->needed_tailroom;
1656 reply = alloc_skb(len, GFP_ATOMIC);
1660 reply->protocol = htons(ETH_P_IPV6);
1662 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1663 skb_push(reply, sizeof(struct ethhdr));
1664 skb_reset_mac_header(reply);
1666 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1668 daddr = eth_hdr(request)->h_source;
1669 ns_olen = request->len - skb_network_offset(request) -
1670 sizeof(struct ipv6hdr) - sizeof(*ns);
1671 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1672 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1673 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1678 /* Ethernet header */
1679 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1680 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1681 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1682 reply->protocol = htons(ETH_P_IPV6);
1684 skb_pull(reply, sizeof(struct ethhdr));
1685 skb_reset_network_header(reply);
1686 skb_put(reply, sizeof(struct ipv6hdr));
1690 pip6 = ipv6_hdr(reply);
1691 memset(pip6, 0, sizeof(struct ipv6hdr));
1693 pip6->priority = ipv6_hdr(request)->priority;
1694 pip6->nexthdr = IPPROTO_ICMPV6;
1695 pip6->hop_limit = 255;
1696 pip6->daddr = ipv6_hdr(request)->saddr;
1697 pip6->saddr = *(struct in6_addr *)n->primary_key;
1699 skb_pull(reply, sizeof(struct ipv6hdr));
1700 skb_reset_transport_header(reply);
1702 /* Neighbor Advertisement */
1703 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1704 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1705 na->icmph.icmp6_router = isrouter;
1706 na->icmph.icmp6_override = 1;
1707 na->icmph.icmp6_solicited = 1;
1708 na->target = ns->target;
1709 ether_addr_copy(&na->opt[2], n->ha);
1710 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1711 na->opt[1] = na_olen >> 3;
1713 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1714 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1715 csum_partial(na, sizeof(*na)+na_olen, 0));
1717 pip6->payload_len = htons(sizeof(*na)+na_olen);
1719 skb_push(reply, sizeof(struct ipv6hdr));
1721 reply->ip_summed = CHECKSUM_UNNECESSARY;
1726 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1728 struct vxlan_dev *vxlan = netdev_priv(dev);
1729 const struct in6_addr *daddr;
1730 const struct ipv6hdr *iphdr;
1731 struct inet6_dev *in6_dev;
1732 struct neighbour *n;
1735 in6_dev = __in6_dev_get(dev);
1739 iphdr = ipv6_hdr(skb);
1740 daddr = &iphdr->daddr;
1741 msg = (struct nd_msg *)(iphdr + 1);
1743 if (ipv6_addr_loopback(daddr) ||
1744 ipv6_addr_is_multicast(&msg->target))
1747 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1750 struct vxlan_fdb *f;
1751 struct sk_buff *reply;
1753 if (!(n->nud_state & NUD_CONNECTED)) {
1758 f = vxlan_find_mac(vxlan, n->ha, vni);
1759 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1760 /* bridge-local neighbor */
1765 reply = vxlan_na_create(skb, n,
1766 !!(f ? f->flags & NTF_ROUTER : 0));
1773 if (netif_rx_ni(reply) == NET_RX_DROP)
1774 dev->stats.rx_dropped++;
1776 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1777 union vxlan_addr ipa = {
1778 .sin6.sin6_addr = msg->target,
1779 .sin6.sin6_family = AF_INET6,
1782 vxlan_ip_miss(dev, &ipa);
1787 return NETDEV_TX_OK;
1791 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1793 struct vxlan_dev *vxlan = netdev_priv(dev);
1794 struct neighbour *n;
1796 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1800 switch (ntohs(eth_hdr(skb)->h_proto)) {
1805 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1808 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1809 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1810 union vxlan_addr ipa = {
1811 .sin.sin_addr.s_addr = pip->daddr,
1812 .sin.sin_family = AF_INET,
1815 vxlan_ip_miss(dev, &ipa);
1821 #if IS_ENABLED(CONFIG_IPV6)
1824 struct ipv6hdr *pip6;
1826 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1828 pip6 = ipv6_hdr(skb);
1829 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1830 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1831 union vxlan_addr ipa = {
1832 .sin6.sin6_addr = pip6->daddr,
1833 .sin6.sin6_family = AF_INET6,
1836 vxlan_ip_miss(dev, &ipa);
1850 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1852 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1854 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1863 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1864 struct vxlan_metadata *md)
1866 struct vxlanhdr_gbp *gbp;
1871 gbp = (struct vxlanhdr_gbp *)vxh;
1872 vxh->vx_flags |= VXLAN_HF_GBP;
1874 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1875 gbp->dont_learn = 1;
1877 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1878 gbp->policy_applied = 1;
1880 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1883 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1886 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1888 gpe->np_applied = 1;
1889 gpe->next_protocol = tun_p_from_eth_p(protocol);
1890 if (!gpe->next_protocol)
1891 return -EPFNOSUPPORT;
1895 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1896 int iphdr_len, __be32 vni,
1897 struct vxlan_metadata *md, u32 vxflags,
1900 struct vxlanhdr *vxh;
1903 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1904 __be16 inner_protocol = htons(ETH_P_TEB);
1906 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1907 skb->ip_summed == CHECKSUM_PARTIAL) {
1908 int csum_start = skb_checksum_start_offset(skb);
1910 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1911 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1912 (skb->csum_offset == offsetof(struct udphdr, check) ||
1913 skb->csum_offset == offsetof(struct tcphdr, check)))
1914 type |= SKB_GSO_TUNNEL_REMCSUM;
1917 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1918 + VXLAN_HLEN + iphdr_len;
1920 /* Need space for new headers (invalidates iph ptr) */
1921 err = skb_cow_head(skb, min_headroom);
1925 err = iptunnel_handle_offloads(skb, type);
1929 vxh = __skb_push(skb, sizeof(*vxh));
1930 vxh->vx_flags = VXLAN_HF_VNI;
1931 vxh->vx_vni = vxlan_vni_field(vni);
1933 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1936 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1937 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1938 vxh->vx_flags |= VXLAN_HF_RCO;
1940 if (!skb_is_gso(skb)) {
1941 skb->ip_summed = CHECKSUM_NONE;
1942 skb->encapsulation = 0;
1946 if (vxflags & VXLAN_F_GBP)
1947 vxlan_build_gbp_hdr(vxh, vxflags, md);
1948 if (vxflags & VXLAN_F_GPE) {
1949 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1952 inner_protocol = skb->protocol;
1955 skb_set_inner_protocol(skb, inner_protocol);
1959 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1960 struct vxlan_sock *sock4,
1961 struct sk_buff *skb, int oif, u8 tos,
1962 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
1963 struct dst_cache *dst_cache,
1964 const struct ip_tunnel_info *info)
1966 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1967 struct rtable *rt = NULL;
1971 return ERR_PTR(-EIO);
1976 rt = dst_cache_get_ip4(dst_cache, saddr);
1981 memset(&fl4, 0, sizeof(fl4));
1982 fl4.flowi4_oif = oif;
1983 fl4.flowi4_tos = RT_TOS(tos);
1984 fl4.flowi4_mark = skb->mark;
1985 fl4.flowi4_proto = IPPROTO_UDP;
1988 fl4.fl4_dport = dport;
1989 fl4.fl4_sport = sport;
1991 rt = ip_route_output_key(vxlan->net, &fl4);
1992 if (likely(!IS_ERR(rt))) {
1993 if (rt->dst.dev == dev) {
1994 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1996 return ERR_PTR(-ELOOP);
2001 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2003 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2004 return ERR_PTR(-ENETUNREACH);
2009 #if IS_ENABLED(CONFIG_IPV6)
2010 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2011 struct net_device *dev,
2012 struct vxlan_sock *sock6,
2013 struct sk_buff *skb, int oif, u8 tos,
2015 const struct in6_addr *daddr,
2016 struct in6_addr *saddr,
2017 __be16 dport, __be16 sport,
2018 struct dst_cache *dst_cache,
2019 const struct ip_tunnel_info *info)
2021 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2022 struct dst_entry *ndst;
2027 return ERR_PTR(-EIO);
2032 ndst = dst_cache_get_ip6(dst_cache, saddr);
2037 memset(&fl6, 0, sizeof(fl6));
2038 fl6.flowi6_oif = oif;
2041 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2042 fl6.flowi6_mark = skb->mark;
2043 fl6.flowi6_proto = IPPROTO_UDP;
2044 fl6.fl6_dport = dport;
2045 fl6.fl6_sport = sport;
2047 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
2050 if (unlikely(err < 0)) {
2051 netdev_dbg(dev, "no route to %pI6\n", daddr);
2052 return ERR_PTR(-ENETUNREACH);
2055 if (unlikely(ndst->dev == dev)) {
2056 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2058 return ERR_PTR(-ELOOP);
2063 dst_cache_set_ip6(dst_cache, ndst, saddr);
2068 /* Bypass encapsulation if the destination is local */
2069 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2070 struct vxlan_dev *dst_vxlan, __be32 vni)
2072 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2073 union vxlan_addr loopback;
2074 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2075 struct net_device *dev = skb->dev;
2078 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2079 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2080 skb->pkt_type = PACKET_HOST;
2081 skb->encapsulation = 0;
2082 skb->dev = dst_vxlan->dev;
2083 __skb_pull(skb, skb_network_offset(skb));
2085 if (remote_ip->sa.sa_family == AF_INET) {
2086 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2087 loopback.sa.sa_family = AF_INET;
2088 #if IS_ENABLED(CONFIG_IPV6)
2090 loopback.sin6.sin6_addr = in6addr_loopback;
2091 loopback.sa.sa_family = AF_INET6;
2095 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2096 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
2099 u64_stats_update_begin(&tx_stats->syncp);
2100 tx_stats->tx_packets++;
2101 tx_stats->tx_bytes += len;
2102 u64_stats_update_end(&tx_stats->syncp);
2104 if (netif_rx(skb) == NET_RX_SUCCESS) {
2105 u64_stats_update_begin(&rx_stats->syncp);
2106 rx_stats->rx_packets++;
2107 rx_stats->rx_bytes += len;
2108 u64_stats_update_end(&rx_stats->syncp);
2110 dev->stats.rx_dropped++;
2114 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2115 struct vxlan_dev *vxlan,
2116 union vxlan_addr *daddr,
2117 __be16 dst_port, int dst_ifindex, __be32 vni,
2118 struct dst_entry *dst,
2121 #if IS_ENABLED(CONFIG_IPV6)
2122 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2123 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2124 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2126 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2128 /* Bypass encapsulation if the destination is local */
2129 if (rt_flags & RTCF_LOCAL &&
2130 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2131 struct vxlan_dev *dst_vxlan;
2134 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2135 daddr->sa.sa_family, dst_port,
2138 dev->stats.tx_errors++;
2143 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2150 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2151 __be32 default_vni, struct vxlan_rdst *rdst,
2154 struct dst_cache *dst_cache;
2155 struct ip_tunnel_info *info;
2156 struct vxlan_dev *vxlan = netdev_priv(dev);
2157 const struct iphdr *old_iph = ip_hdr(skb);
2158 union vxlan_addr *dst;
2159 union vxlan_addr remote_ip, local_ip;
2160 struct vxlan_metadata _md;
2161 struct vxlan_metadata *md = &_md;
2162 __be16 src_port = 0, dst_port;
2163 struct dst_entry *ndst = NULL;
2168 u32 flags = vxlan->cfg.flags;
2169 bool udp_sum = false;
2170 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2172 info = skb_tunnel_info(skb);
2175 dst = &rdst->remote_ip;
2176 if (vxlan_addr_any(dst)) {
2178 /* short-circuited back to local bridge */
2179 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2185 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2186 vni = (rdst->remote_vni) ? : default_vni;
2187 ifindex = rdst->remote_ifindex;
2188 local_ip = vxlan->cfg.saddr;
2189 dst_cache = &rdst->dst_cache;
2190 md->gbp = skb->mark;
2191 if (flags & VXLAN_F_TTL_INHERIT) {
2192 ttl = ip_tunnel_get_ttl(old_iph, skb);
2194 ttl = vxlan->cfg.ttl;
2195 if (!ttl && vxlan_addr_multicast(dst))
2199 tos = vxlan->cfg.tos;
2201 tos = ip_tunnel_get_dsfield(old_iph, skb);
2203 if (dst->sa.sa_family == AF_INET)
2204 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2206 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2207 label = vxlan->cfg.label;
2210 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2214 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2215 if (remote_ip.sa.sa_family == AF_INET) {
2216 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2217 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2219 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2220 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2223 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2224 vni = tunnel_id_to_key32(info->key.tun_id);
2226 dst_cache = &info->dst_cache;
2227 if (info->options_len &&
2228 info->key.tun_flags & TUNNEL_VXLAN_OPT)
2229 md = ip_tunnel_info_opts(info);
2230 ttl = info->key.ttl;
2231 tos = info->key.tos;
2232 label = info->key.label;
2233 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2235 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2236 vxlan->cfg.port_max, true);
2239 if (dst->sa.sa_family == AF_INET) {
2240 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2244 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2245 dst->sin.sin_addr.s_addr,
2246 &local_ip.sin.sin_addr.s_addr,
2254 /* Bypass encapsulation if the destination is local */
2256 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2257 dst_port, ifindex, vni,
2258 &rt->dst, rt->rt_flags);
2261 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2266 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
2268 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2269 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2270 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2271 vni, md, flags, udp_sum);
2275 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2276 dst->sin.sin_addr.s_addr, tos, ttl, df,
2277 src_port, dst_port, xnet, !udp_sum);
2278 #if IS_ENABLED(CONFIG_IPV6)
2280 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2282 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2283 label, &dst->sin6.sin6_addr,
2284 &local_ip.sin6.sin6_addr,
2288 err = PTR_ERR(ndst);
2294 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2296 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2297 dst_port, ifindex, vni,
2303 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
2305 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2306 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2307 skb_scrub_packet(skb, xnet);
2308 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2309 vni, md, flags, udp_sum);
2313 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2314 &local_ip.sin6.sin6_addr,
2315 &dst->sin6.sin6_addr, tos, ttl,
2316 label, src_port, dst_port, !udp_sum);
2324 dev->stats.tx_dropped++;
2331 dev->stats.collisions++;
2332 else if (err == -ENETUNREACH)
2333 dev->stats.tx_carrier_errors++;
2335 dev->stats.tx_errors++;
2339 /* Transmit local packets over Vxlan
2341 * Outer IP header inherits ECN and DF from inner header.
2342 * Outer UDP destination is the VXLAN assigned port.
2343 * source port is based on hash of flow
2345 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2347 struct vxlan_dev *vxlan = netdev_priv(dev);
2348 struct vxlan_rdst *rdst, *fdst = NULL;
2349 const struct ip_tunnel_info *info;
2350 bool did_rsc = false;
2351 struct vxlan_fdb *f;
2355 info = skb_tunnel_info(skb);
2357 skb_reset_mac_header(skb);
2359 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2360 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2361 info->mode & IP_TUNNEL_INFO_TX) {
2362 vni = tunnel_id_to_key32(info->key.tun_id);
2364 if (info && info->mode & IP_TUNNEL_INFO_TX)
2365 vxlan_xmit_one(skb, dev, vni, NULL, false);
2368 return NETDEV_TX_OK;
2372 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2374 if (ntohs(eth->h_proto) == ETH_P_ARP)
2375 return arp_reduce(dev, skb, vni);
2376 #if IS_ENABLED(CONFIG_IPV6)
2377 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2378 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2379 sizeof(struct nd_msg)) &&
2380 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2381 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2383 if (m->icmph.icmp6_code == 0 &&
2384 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2385 return neigh_reduce(dev, skb, vni);
2391 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2394 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2395 (ntohs(eth->h_proto) == ETH_P_IP ||
2396 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2397 did_rsc = route_shortcircuit(dev, skb);
2399 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2403 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2405 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2406 !is_multicast_ether_addr(eth->h_dest))
2407 vxlan_fdb_miss(vxlan, eth->h_dest);
2409 dev->stats.tx_dropped++;
2411 return NETDEV_TX_OK;
2415 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2416 struct sk_buff *skb1;
2422 skb1 = skb_clone(skb, GFP_ATOMIC);
2424 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2428 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2431 return NETDEV_TX_OK;
2434 /* Walk the forwarding table and purge stale entries */
2435 static void vxlan_cleanup(struct timer_list *t)
2437 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2438 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2441 if (!netif_running(vxlan->dev))
2444 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2445 struct hlist_node *p, *n;
2447 spin_lock_bh(&vxlan->hash_lock);
2448 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2450 = container_of(p, struct vxlan_fdb, hlist);
2451 unsigned long timeout;
2453 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2456 if (f->flags & NTF_EXT_LEARNED)
2459 timeout = f->used + vxlan->cfg.age_interval * HZ;
2460 if (time_before_eq(timeout, jiffies)) {
2461 netdev_dbg(vxlan->dev,
2462 "garbage collect %pM\n",
2464 f->state = NUD_STALE;
2465 vxlan_fdb_destroy(vxlan, f, true);
2466 } else if (time_before(timeout, next_timer))
2467 next_timer = timeout;
2469 spin_unlock_bh(&vxlan->hash_lock);
2472 mod_timer(&vxlan->age_timer, next_timer);
2475 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2477 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2479 spin_lock(&vn->sock_lock);
2480 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2481 #if IS_ENABLED(CONFIG_IPV6)
2482 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2484 spin_unlock(&vn->sock_lock);
2487 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2488 struct vxlan_dev_node *node)
2490 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2491 __be32 vni = vxlan->default_dst.remote_vni;
2493 node->vxlan = vxlan;
2494 spin_lock(&vn->sock_lock);
2495 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2496 spin_unlock(&vn->sock_lock);
2499 /* Setup stats when device is created */
2500 static int vxlan_init(struct net_device *dev)
2502 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2509 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2511 struct vxlan_fdb *f;
2513 spin_lock_bh(&vxlan->hash_lock);
2514 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2516 vxlan_fdb_destroy(vxlan, f, true);
2517 spin_unlock_bh(&vxlan->hash_lock);
2520 static void vxlan_uninit(struct net_device *dev)
2522 struct vxlan_dev *vxlan = netdev_priv(dev);
2524 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2526 free_percpu(dev->tstats);
2529 /* Start ageing timer and join group when device is brought up */
2530 static int vxlan_open(struct net_device *dev)
2532 struct vxlan_dev *vxlan = netdev_priv(dev);
2535 ret = vxlan_sock_add(vxlan);
2539 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2540 ret = vxlan_igmp_join(vxlan);
2541 if (ret == -EADDRINUSE)
2544 vxlan_sock_release(vxlan);
2549 if (vxlan->cfg.age_interval)
2550 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2555 /* Purge the forwarding table */
2556 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2560 spin_lock_bh(&vxlan->hash_lock);
2561 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2562 struct hlist_node *p, *n;
2563 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2565 = container_of(p, struct vxlan_fdb, hlist);
2566 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2568 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2569 if (!is_zero_ether_addr(f->eth_addr))
2570 vxlan_fdb_destroy(vxlan, f, true);
2573 spin_unlock_bh(&vxlan->hash_lock);
2576 /* Cleanup timer and forwarding table on shutdown */
2577 static int vxlan_stop(struct net_device *dev)
2579 struct vxlan_dev *vxlan = netdev_priv(dev);
2580 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2583 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2584 !vxlan_group_used(vn, vxlan))
2585 ret = vxlan_igmp_leave(vxlan);
2587 del_timer_sync(&vxlan->age_timer);
2589 vxlan_flush(vxlan, false);
2590 vxlan_sock_release(vxlan);
2595 /* Stub, nothing needs to be done. */
2596 static void vxlan_set_multicast_list(struct net_device *dev)
2600 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2602 struct vxlan_dev *vxlan = netdev_priv(dev);
2603 struct vxlan_rdst *dst = &vxlan->default_dst;
2604 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2605 dst->remote_ifindex);
2606 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2608 /* This check is different than dev->max_mtu, because it looks at
2609 * the lowerdev->mtu, rather than the static dev->max_mtu
2612 int max_mtu = lowerdev->mtu -
2613 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2614 if (new_mtu > max_mtu)
2622 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2624 struct vxlan_dev *vxlan = netdev_priv(dev);
2625 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2626 __be16 sport, dport;
2628 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2629 vxlan->cfg.port_max, true);
2630 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2632 if (ip_tunnel_info_af(info) == AF_INET) {
2633 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2636 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2637 info->key.u.ipv4.dst,
2638 &info->key.u.ipv4.src, dport, sport,
2639 &info->dst_cache, info);
2644 #if IS_ENABLED(CONFIG_IPV6)
2645 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2646 struct dst_entry *ndst;
2648 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2649 info->key.label, &info->key.u.ipv6.dst,
2650 &info->key.u.ipv6.src, dport, sport,
2651 &info->dst_cache, info);
2653 return PTR_ERR(ndst);
2655 #else /* !CONFIG_IPV6 */
2656 return -EPFNOSUPPORT;
2659 info->key.tp_src = sport;
2660 info->key.tp_dst = dport;
2664 static const struct net_device_ops vxlan_netdev_ether_ops = {
2665 .ndo_init = vxlan_init,
2666 .ndo_uninit = vxlan_uninit,
2667 .ndo_open = vxlan_open,
2668 .ndo_stop = vxlan_stop,
2669 .ndo_start_xmit = vxlan_xmit,
2670 .ndo_get_stats64 = ip_tunnel_get_stats64,
2671 .ndo_set_rx_mode = vxlan_set_multicast_list,
2672 .ndo_change_mtu = vxlan_change_mtu,
2673 .ndo_validate_addr = eth_validate_addr,
2674 .ndo_set_mac_address = eth_mac_addr,
2675 .ndo_fdb_add = vxlan_fdb_add,
2676 .ndo_fdb_del = vxlan_fdb_delete,
2677 .ndo_fdb_dump = vxlan_fdb_dump,
2678 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2681 static const struct net_device_ops vxlan_netdev_raw_ops = {
2682 .ndo_init = vxlan_init,
2683 .ndo_uninit = vxlan_uninit,
2684 .ndo_open = vxlan_open,
2685 .ndo_stop = vxlan_stop,
2686 .ndo_start_xmit = vxlan_xmit,
2687 .ndo_get_stats64 = ip_tunnel_get_stats64,
2688 .ndo_change_mtu = vxlan_change_mtu,
2689 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2692 /* Info for udev, that this is a virtual tunnel endpoint */
2693 static struct device_type vxlan_type = {
2697 /* Calls the ndo_udp_tunnel_add of the caller in order to
2698 * supply the listening VXLAN udp ports. Callers are expected
2699 * to implement the ndo_udp_tunnel_add.
2701 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
2703 struct vxlan_sock *vs;
2704 struct net *net = dev_net(dev);
2705 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2708 spin_lock(&vn->sock_lock);
2709 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2710 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2711 unsigned short type;
2713 if (vs->flags & VXLAN_F_GPE)
2714 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2716 type = UDP_TUNNEL_TYPE_VXLAN;
2719 udp_tunnel_push_rx_port(dev, vs->sock, type);
2721 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2724 spin_unlock(&vn->sock_lock);
2727 /* Initialize the device structure. */
2728 static void vxlan_setup(struct net_device *dev)
2730 struct vxlan_dev *vxlan = netdev_priv(dev);
2733 eth_hw_addr_random(dev);
2736 dev->needs_free_netdev = true;
2737 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2739 dev->features |= NETIF_F_LLTX;
2740 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2741 dev->features |= NETIF_F_RXCSUM;
2742 dev->features |= NETIF_F_GSO_SOFTWARE;
2744 dev->vlan_features = dev->features;
2745 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2746 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2747 netif_keep_dst(dev);
2748 dev->priv_flags |= IFF_NO_QUEUE;
2750 /* MTU range: 68 - 65535 */
2751 dev->min_mtu = ETH_MIN_MTU;
2752 dev->max_mtu = ETH_MAX_MTU;
2754 INIT_LIST_HEAD(&vxlan->next);
2755 spin_lock_init(&vxlan->hash_lock);
2757 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
2761 gro_cells_init(&vxlan->gro_cells, dev);
2763 for (h = 0; h < FDB_HASH_SIZE; ++h)
2764 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2767 static void vxlan_ether_setup(struct net_device *dev)
2769 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2770 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2771 dev->netdev_ops = &vxlan_netdev_ether_ops;
2774 static void vxlan_raw_setup(struct net_device *dev)
2776 dev->header_ops = NULL;
2777 dev->type = ARPHRD_NONE;
2778 dev->hard_header_len = 0;
2780 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2781 dev->netdev_ops = &vxlan_netdev_raw_ops;
2784 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2785 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2786 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2787 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2788 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2789 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2790 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2791 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2792 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2793 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
2794 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2795 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2796 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2797 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2798 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2799 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2800 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2801 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2802 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2803 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2804 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2805 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2806 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2807 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2808 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2809 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2810 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
2811 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2812 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
2815 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2816 struct netlink_ext_ack *extack)
2818 if (tb[IFLA_ADDRESS]) {
2819 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2820 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2821 "Provided link layer address is not Ethernet");
2825 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2826 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2827 "Provided Ethernet address is not unicast");
2828 return -EADDRNOTAVAIL;
2833 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2835 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2836 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2837 "MTU must be between 68 and 65535");
2843 NL_SET_ERR_MSG(extack,
2844 "Required attributes not provided to perform the operation");
2848 if (data[IFLA_VXLAN_ID]) {
2849 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2851 if (id >= VXLAN_N_VID) {
2852 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2853 "VXLAN ID must be lower than 16777216");
2858 if (data[IFLA_VXLAN_PORT_RANGE]) {
2859 const struct ifla_vxlan_port_range *p
2860 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2862 if (ntohs(p->high) < ntohs(p->low)) {
2863 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2864 "Invalid source port range");
2872 static void vxlan_get_drvinfo(struct net_device *netdev,
2873 struct ethtool_drvinfo *drvinfo)
2875 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2876 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2879 static const struct ethtool_ops vxlan_ethtool_ops = {
2880 .get_drvinfo = vxlan_get_drvinfo,
2881 .get_link = ethtool_op_get_link,
2884 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2885 __be16 port, u32 flags)
2887 struct socket *sock;
2888 struct udp_port_cfg udp_conf;
2891 memset(&udp_conf, 0, sizeof(udp_conf));
2894 udp_conf.family = AF_INET6;
2895 udp_conf.use_udp6_rx_checksums =
2896 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2897 udp_conf.ipv6_v6only = 1;
2899 udp_conf.family = AF_INET;
2902 udp_conf.local_udp_port = port;
2904 /* Open UDP socket */
2905 err = udp_sock_create(net, &udp_conf, &sock);
2907 return ERR_PTR(err);
2912 /* Create new listen socket if needed */
2913 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2914 __be16 port, u32 flags)
2916 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2917 struct vxlan_sock *vs;
2918 struct socket *sock;
2920 struct udp_tunnel_sock_cfg tunnel_cfg;
2922 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2924 return ERR_PTR(-ENOMEM);
2926 for (h = 0; h < VNI_HASH_SIZE; ++h)
2927 INIT_HLIST_HEAD(&vs->vni_list[h]);
2929 sock = vxlan_create_sock(net, ipv6, port, flags);
2932 return ERR_CAST(sock);
2936 refcount_set(&vs->refcnt, 1);
2937 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2939 spin_lock(&vn->sock_lock);
2940 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2941 udp_tunnel_notify_add_rx_port(sock,
2942 (vs->flags & VXLAN_F_GPE) ?
2943 UDP_TUNNEL_TYPE_VXLAN_GPE :
2944 UDP_TUNNEL_TYPE_VXLAN);
2945 spin_unlock(&vn->sock_lock);
2947 /* Mark socket as an encapsulation socket. */
2948 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2949 tunnel_cfg.sk_user_data = vs;
2950 tunnel_cfg.encap_type = 1;
2951 tunnel_cfg.encap_rcv = vxlan_rcv;
2952 tunnel_cfg.encap_destroy = NULL;
2953 tunnel_cfg.gro_receive = vxlan_gro_receive;
2954 tunnel_cfg.gro_complete = vxlan_gro_complete;
2956 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2961 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2963 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2964 struct vxlan_sock *vs = NULL;
2965 struct vxlan_dev_node *node;
2967 if (!vxlan->cfg.no_share) {
2968 spin_lock(&vn->sock_lock);
2969 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2970 vxlan->cfg.dst_port, vxlan->cfg.flags);
2971 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
2972 spin_unlock(&vn->sock_lock);
2975 spin_unlock(&vn->sock_lock);
2978 vs = vxlan_socket_create(vxlan->net, ipv6,
2979 vxlan->cfg.dst_port, vxlan->cfg.flags);
2982 #if IS_ENABLED(CONFIG_IPV6)
2984 rcu_assign_pointer(vxlan->vn6_sock, vs);
2985 node = &vxlan->hlist6;
2989 rcu_assign_pointer(vxlan->vn4_sock, vs);
2990 node = &vxlan->hlist4;
2992 vxlan_vs_add_dev(vs, vxlan, node);
2996 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2998 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2999 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3000 bool ipv4 = !ipv6 || metadata;
3003 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3004 #if IS_ENABLED(CONFIG_IPV6)
3005 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3007 ret = __vxlan_sock_add(vxlan, true);
3008 if (ret < 0 && ret != -EAFNOSUPPORT)
3013 ret = __vxlan_sock_add(vxlan, false);
3015 vxlan_sock_release(vxlan);
3019 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3020 struct net_device **lower,
3021 struct vxlan_dev *old,
3022 struct netlink_ext_ack *extack)
3024 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3025 struct vxlan_dev *tmp;
3026 bool use_ipv6 = false;
3028 if (conf->flags & VXLAN_F_GPE) {
3029 /* For now, allow GPE only together with
3030 * COLLECT_METADATA. This can be relaxed later; in such
3031 * case, the other side of the PtP link will have to be
3034 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3035 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3036 NL_SET_ERR_MSG(extack,
3037 "VXLAN GPE does not support this combination of attributes");
3042 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3043 /* Unless IPv6 is explicitly requested, assume IPv4 */
3044 conf->remote_ip.sa.sa_family = AF_INET;
3045 conf->saddr.sa.sa_family = AF_INET;
3046 } else if (!conf->remote_ip.sa.sa_family) {
3047 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3048 } else if (!conf->saddr.sa.sa_family) {
3049 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3052 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3053 NL_SET_ERR_MSG(extack,
3054 "Local and remote address must be from the same family");
3058 if (vxlan_addr_multicast(&conf->saddr)) {
3059 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3063 if (conf->saddr.sa.sa_family == AF_INET6) {
3064 if (!IS_ENABLED(CONFIG_IPV6)) {
3065 NL_SET_ERR_MSG(extack,
3066 "IPv6 support not enabled in the kernel");
3067 return -EPFNOSUPPORT;
3070 conf->flags |= VXLAN_F_IPV6;
3072 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3074 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3076 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3078 if (local_type & IPV6_ADDR_LINKLOCAL) {
3079 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3080 (remote_type != IPV6_ADDR_ANY)) {
3081 NL_SET_ERR_MSG(extack,
3082 "Invalid combination of local and remote address scopes");
3086 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3089 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3090 NL_SET_ERR_MSG(extack,
3091 "Invalid combination of local and remote address scopes");
3095 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3100 if (conf->label && !use_ipv6) {
3101 NL_SET_ERR_MSG(extack,
3102 "Label attribute only applies to IPv6 VXLAN devices");
3106 if (conf->remote_ifindex) {
3107 struct net_device *lowerdev;
3109 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3111 NL_SET_ERR_MSG(extack,
3112 "Invalid local interface, device not found");
3116 #if IS_ENABLED(CONFIG_IPV6)
3118 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3119 if (idev && idev->cnf.disable_ipv6) {
3120 NL_SET_ERR_MSG(extack,
3121 "IPv6 support disabled by administrator");
3129 if (vxlan_addr_multicast(&conf->remote_ip)) {
3130 NL_SET_ERR_MSG(extack,
3131 "Local interface required for multicast remote destination");
3136 #if IS_ENABLED(CONFIG_IPV6)
3137 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3138 NL_SET_ERR_MSG(extack,
3139 "Local interface required for link-local local/remote addresses");
3147 if (!conf->dst_port) {
3148 if (conf->flags & VXLAN_F_GPE)
3149 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3151 conf->dst_port = htons(vxlan_port);
3154 if (!conf->age_interval)
3155 conf->age_interval = FDB_AGE_DEFAULT;
3157 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3161 if (tmp->cfg.vni != conf->vni)
3163 if (tmp->cfg.dst_port != conf->dst_port)
3165 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3166 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3169 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3170 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3173 NL_SET_ERR_MSG(extack,
3174 "A VXLAN device with the specified VNI already exists");
3181 static void vxlan_config_apply(struct net_device *dev,
3182 struct vxlan_config *conf,
3183 struct net_device *lowerdev,
3184 struct net *src_net,
3187 struct vxlan_dev *vxlan = netdev_priv(dev);
3188 struct vxlan_rdst *dst = &vxlan->default_dst;
3189 unsigned short needed_headroom = ETH_HLEN;
3190 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3191 int max_mtu = ETH_MAX_MTU;
3194 if (conf->flags & VXLAN_F_GPE)
3195 vxlan_raw_setup(dev);
3197 vxlan_ether_setup(dev);
3200 dev->mtu = conf->mtu;
3202 vxlan->net = src_net;
3205 dst->remote_vni = conf->vni;
3207 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3210 dst->remote_ifindex = conf->remote_ifindex;
3212 dev->gso_max_size = lowerdev->gso_max_size;
3213 dev->gso_max_segs = lowerdev->gso_max_segs;
3215 needed_headroom = lowerdev->hard_header_len;
3217 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3219 if (max_mtu < ETH_MIN_MTU)
3220 max_mtu = ETH_MIN_MTU;
3222 if (!changelink && !conf->mtu)
3226 if (dev->mtu > max_mtu)
3229 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3230 needed_headroom += VXLAN6_HEADROOM;
3232 needed_headroom += VXLAN_HEADROOM;
3233 dev->needed_headroom = needed_headroom;
3235 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3238 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3239 struct vxlan_config *conf, bool changelink,
3240 struct netlink_ext_ack *extack)
3242 struct vxlan_dev *vxlan = netdev_priv(dev);
3243 struct net_device *lowerdev;
3246 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3250 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3255 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3256 struct vxlan_config *conf,
3257 struct netlink_ext_ack *extack)
3259 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3260 struct vxlan_dev *vxlan = netdev_priv(dev);
3261 struct vxlan_fdb *f = NULL;
3262 bool unregister = false;
3265 err = vxlan_dev_configure(net, dev, conf, false, extack);
3269 dev->ethtool_ops = &vxlan_ethtool_ops;
3271 /* create an fdb entry for a valid default destination */
3272 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3273 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3274 &vxlan->default_dst.remote_ip,
3275 NUD_REACHABLE | NUD_PERMANENT,
3276 vxlan->cfg.dst_port,
3277 vxlan->default_dst.remote_vni,
3278 vxlan->default_dst.remote_vni,
3279 vxlan->default_dst.remote_ifindex,
3285 err = register_netdevice(dev);
3290 err = rtnl_configure_link(dev, NULL);
3294 /* notify default fdb entry */
3296 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH);
3298 list_add(&vxlan->next, &vn->vxlan_list);
3302 /* unregister_netdevice() destroys the default FDB entry with deletion
3303 * notification. But the addition notification was not sent yet, so
3304 * destroy the entry by hand here.
3307 vxlan_fdb_destroy(vxlan, f, false);
3309 unregister_netdevice(dev);
3313 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3314 struct net_device *dev, struct vxlan_config *conf,
3317 struct vxlan_dev *vxlan = netdev_priv(dev);
3319 memset(conf, 0, sizeof(*conf));
3321 /* if changelink operation, start with old existing cfg */
3323 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3325 if (data[IFLA_VXLAN_ID]) {
3326 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3328 if (changelink && (vni != conf->vni))
3330 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3333 if (data[IFLA_VXLAN_GROUP]) {
3334 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3337 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3338 conf->remote_ip.sa.sa_family = AF_INET;
3339 } else if (data[IFLA_VXLAN_GROUP6]) {
3340 if (!IS_ENABLED(CONFIG_IPV6))
3341 return -EPFNOSUPPORT;
3343 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3346 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3347 conf->remote_ip.sa.sa_family = AF_INET6;
3350 if (data[IFLA_VXLAN_LOCAL]) {
3351 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3354 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3355 conf->saddr.sa.sa_family = AF_INET;
3356 } else if (data[IFLA_VXLAN_LOCAL6]) {
3357 if (!IS_ENABLED(CONFIG_IPV6))
3358 return -EPFNOSUPPORT;
3360 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3363 /* TODO: respect scope id */
3364 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3365 conf->saddr.sa.sa_family = AF_INET6;
3368 if (data[IFLA_VXLAN_LINK])
3369 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3371 if (data[IFLA_VXLAN_TOS])
3372 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3374 if (data[IFLA_VXLAN_TTL])
3375 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3377 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3380 conf->flags |= VXLAN_F_TTL_INHERIT;
3383 if (data[IFLA_VXLAN_LABEL])
3384 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3385 IPV6_FLOWLABEL_MASK;
3387 if (data[IFLA_VXLAN_LEARNING]) {
3388 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
3389 conf->flags |= VXLAN_F_LEARN;
3391 conf->flags &= ~VXLAN_F_LEARN;
3392 } else if (!changelink) {
3393 /* default to learn on a new device */
3394 conf->flags |= VXLAN_F_LEARN;
3397 if (data[IFLA_VXLAN_AGEING]) {
3400 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3403 if (data[IFLA_VXLAN_PROXY]) {
3406 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3407 conf->flags |= VXLAN_F_PROXY;
3410 if (data[IFLA_VXLAN_RSC]) {
3413 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3414 conf->flags |= VXLAN_F_RSC;
3417 if (data[IFLA_VXLAN_L2MISS]) {
3420 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3421 conf->flags |= VXLAN_F_L2MISS;
3424 if (data[IFLA_VXLAN_L3MISS]) {
3427 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3428 conf->flags |= VXLAN_F_L3MISS;
3431 if (data[IFLA_VXLAN_LIMIT]) {
3434 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3437 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3440 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3441 conf->flags |= VXLAN_F_COLLECT_METADATA;
3444 if (data[IFLA_VXLAN_PORT_RANGE]) {
3446 const struct ifla_vxlan_port_range *p
3447 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3448 conf->port_min = ntohs(p->low);
3449 conf->port_max = ntohs(p->high);
3455 if (data[IFLA_VXLAN_PORT]) {
3458 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3461 if (data[IFLA_VXLAN_UDP_CSUM]) {
3464 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3465 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3468 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3471 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3472 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3475 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3478 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3479 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3482 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3485 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3486 conf->flags |= VXLAN_F_REMCSUM_TX;
3489 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3492 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3493 conf->flags |= VXLAN_F_REMCSUM_RX;
3496 if (data[IFLA_VXLAN_GBP]) {
3499 conf->flags |= VXLAN_F_GBP;
3502 if (data[IFLA_VXLAN_GPE]) {
3505 conf->flags |= VXLAN_F_GPE;
3508 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3511 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3517 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3523 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3524 struct nlattr *tb[], struct nlattr *data[],
3525 struct netlink_ext_ack *extack)
3527 struct vxlan_config conf;
3530 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3534 return __vxlan_dev_create(src_net, dev, &conf, extack);
3537 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3538 struct nlattr *data[],
3539 struct netlink_ext_ack *extack)
3541 struct vxlan_dev *vxlan = netdev_priv(dev);
3542 struct vxlan_rdst *dst = &vxlan->default_dst;
3543 struct vxlan_rdst old_dst;
3544 struct vxlan_config conf;
3547 err = vxlan_nl2conf(tb, data,
3552 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
3554 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
3558 /* handle default dst entry */
3559 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3560 spin_lock_bh(&vxlan->hash_lock);
3561 if (!vxlan_addr_any(&old_dst.remote_ip))
3562 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3564 vxlan->cfg.dst_port,
3567 old_dst.remote_ifindex, 0);
3569 if (!vxlan_addr_any(&dst->remote_ip)) {
3570 err = vxlan_fdb_update(vxlan, all_zeros_mac,
3572 NUD_REACHABLE | NUD_PERMANENT,
3573 NLM_F_APPEND | NLM_F_CREATE,
3574 vxlan->cfg.dst_port,
3577 dst->remote_ifindex,
3580 spin_unlock_bh(&vxlan->hash_lock);
3584 spin_unlock_bh(&vxlan->hash_lock);
3590 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3592 struct vxlan_dev *vxlan = netdev_priv(dev);
3594 vxlan_flush(vxlan, true);
3596 gro_cells_destroy(&vxlan->gro_cells);
3597 list_del(&vxlan->next);
3598 unregister_netdevice_queue(dev, head);
3601 static size_t vxlan_get_size(const struct net_device *dev)
3604 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
3605 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3606 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3607 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3608 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3609 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
3610 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3611 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3612 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
3613 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3614 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3615 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3616 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
3617 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
3618 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3619 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3620 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3621 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3622 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3623 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3624 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3625 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3626 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3630 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3632 const struct vxlan_dev *vxlan = netdev_priv(dev);
3633 const struct vxlan_rdst *dst = &vxlan->default_dst;
3634 struct ifla_vxlan_port_range ports = {
3635 .low = htons(vxlan->cfg.port_min),
3636 .high = htons(vxlan->cfg.port_max),
3639 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3640 goto nla_put_failure;
3642 if (!vxlan_addr_any(&dst->remote_ip)) {
3643 if (dst->remote_ip.sa.sa_family == AF_INET) {
3644 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3645 dst->remote_ip.sin.sin_addr.s_addr))
3646 goto nla_put_failure;
3647 #if IS_ENABLED(CONFIG_IPV6)
3649 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3650 &dst->remote_ip.sin6.sin6_addr))
3651 goto nla_put_failure;
3656 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3657 goto nla_put_failure;
3659 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3660 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3661 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3662 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3663 goto nla_put_failure;
3664 #if IS_ENABLED(CONFIG_IPV6)
3666 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3667 &vxlan->cfg.saddr.sin6.sin6_addr))
3668 goto nla_put_failure;
3673 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3674 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
3675 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
3676 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3677 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3678 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3679 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
3680 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3681 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3682 nla_put_u8(skb, IFLA_VXLAN_RSC,
3683 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
3684 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3685 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
3686 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3687 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
3688 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3689 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
3690 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3691 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3692 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3693 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3694 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3695 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3696 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3697 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3698 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3699 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3700 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
3701 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3702 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
3703 goto nla_put_failure;
3705 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3706 goto nla_put_failure;
3708 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3709 nla_put_flag(skb, IFLA_VXLAN_GBP))
3710 goto nla_put_failure;
3712 if (vxlan->cfg.flags & VXLAN_F_GPE &&
3713 nla_put_flag(skb, IFLA_VXLAN_GPE))
3714 goto nla_put_failure;
3716 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3717 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3718 goto nla_put_failure;
3726 static struct net *vxlan_get_link_net(const struct net_device *dev)
3728 struct vxlan_dev *vxlan = netdev_priv(dev);
3733 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3735 .maxtype = IFLA_VXLAN_MAX,
3736 .policy = vxlan_policy,
3737 .priv_size = sizeof(struct vxlan_dev),
3738 .setup = vxlan_setup,
3739 .validate = vxlan_validate,
3740 .newlink = vxlan_newlink,
3741 .changelink = vxlan_changelink,
3742 .dellink = vxlan_dellink,
3743 .get_size = vxlan_get_size,
3744 .fill_info = vxlan_fill_info,
3745 .get_link_net = vxlan_get_link_net,
3748 struct net_device *vxlan_dev_create(struct net *net, const char *name,
3749 u8 name_assign_type,
3750 struct vxlan_config *conf)
3752 struct nlattr *tb[IFLA_MAX + 1];
3753 struct net_device *dev;
3756 memset(&tb, 0, sizeof(tb));
3758 dev = rtnl_create_link(net, name, name_assign_type,
3759 &vxlan_link_ops, tb);
3763 err = __vxlan_dev_create(net, dev, conf, NULL);
3766 return ERR_PTR(err);
3769 err = rtnl_configure_link(dev, NULL);
3771 LIST_HEAD(list_kill);
3773 vxlan_dellink(dev, &list_kill);
3774 unregister_netdevice_many(&list_kill);
3775 return ERR_PTR(err);
3780 EXPORT_SYMBOL_GPL(vxlan_dev_create);
3782 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3783 struct net_device *dev)
3785 struct vxlan_dev *vxlan, *next;
3786 LIST_HEAD(list_kill);
3788 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3789 struct vxlan_rdst *dst = &vxlan->default_dst;
3791 /* In case we created vxlan device with carrier
3792 * and we loose the carrier due to module unload
3793 * we also need to remove vxlan device. In other
3794 * cases, it's not necessary and remote_ifindex
3795 * is 0 here, so no matches.
3797 if (dst->remote_ifindex == dev->ifindex)
3798 vxlan_dellink(vxlan->dev, &list_kill);
3801 unregister_netdevice_many(&list_kill);
3804 static int vxlan_netdevice_event(struct notifier_block *unused,
3805 unsigned long event, void *ptr)
3807 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3808 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3810 if (event == NETDEV_UNREGISTER) {
3811 vxlan_offload_rx_ports(dev, false);
3812 vxlan_handle_lowerdev_unregister(vn, dev);
3813 } else if (event == NETDEV_REGISTER) {
3814 vxlan_offload_rx_ports(dev, true);
3815 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3816 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
3817 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
3823 static struct notifier_block vxlan_notifier_block __read_mostly = {
3824 .notifier_call = vxlan_netdevice_event,
3828 vxlan_fdb_offloaded_set(struct net_device *dev,
3829 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
3831 struct vxlan_dev *vxlan = netdev_priv(dev);
3832 struct vxlan_rdst *rdst;
3833 struct vxlan_fdb *f;
3835 spin_lock_bh(&vxlan->hash_lock);
3837 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
3841 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
3842 fdb_info->remote_port,
3843 fdb_info->remote_vni,
3844 fdb_info->remote_ifindex);
3848 rdst->offloaded = fdb_info->offloaded;
3851 spin_unlock_bh(&vxlan->hash_lock);
3854 static int vxlan_switchdev_event(struct notifier_block *unused,
3855 unsigned long event, void *ptr)
3857 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
3860 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
3861 vxlan_fdb_offloaded_set(dev, ptr);
3868 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
3869 .notifier_call = vxlan_switchdev_event,
3872 static __net_init int vxlan_init_net(struct net *net)
3874 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3877 INIT_LIST_HEAD(&vn->vxlan_list);
3878 spin_lock_init(&vn->sock_lock);
3880 for (h = 0; h < PORT_HASH_SIZE; ++h)
3881 INIT_HLIST_HEAD(&vn->sock_list[h]);
3886 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
3888 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3889 struct vxlan_dev *vxlan, *next;
3890 struct net_device *dev, *aux;
3893 for_each_netdev_safe(net, dev, aux)
3894 if (dev->rtnl_link_ops == &vxlan_link_ops)
3895 unregister_netdevice_queue(dev, head);
3897 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3898 /* If vxlan->dev is in the same netns, it has already been added
3899 * to the list by the previous loop.
3901 if (!net_eq(dev_net(vxlan->dev), net)) {
3902 gro_cells_destroy(&vxlan->gro_cells);
3903 unregister_netdevice_queue(vxlan->dev, head);
3907 for (h = 0; h < PORT_HASH_SIZE; ++h)
3908 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
3911 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
3917 list_for_each_entry(net, net_list, exit_list)
3918 vxlan_destroy_tunnels(net, &list);
3920 unregister_netdevice_many(&list);
3924 static struct pernet_operations vxlan_net_ops = {
3925 .init = vxlan_init_net,
3926 .exit_batch = vxlan_exit_batch_net,
3927 .id = &vxlan_net_id,
3928 .size = sizeof(struct vxlan_net),
3931 static int __init vxlan_init_module(void)
3935 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3937 rc = register_pernet_subsys(&vxlan_net_ops);
3941 rc = register_netdevice_notifier(&vxlan_notifier_block);
3945 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
3949 rc = rtnl_link_register(&vxlan_link_ops);
3955 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
3957 unregister_netdevice_notifier(&vxlan_notifier_block);
3959 unregister_pernet_subsys(&vxlan_net_ops);
3963 late_initcall(vxlan_init_module);
3965 static void __exit vxlan_cleanup_module(void)
3967 rtnl_link_unregister(&vxlan_link_ops);
3968 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
3969 unregister_netdevice_notifier(&vxlan_notifier_block);
3970 unregister_pernet_subsys(&vxlan_net_ops);
3971 /* rcu_barrier() is called by netns */
3973 module_exit(vxlan_cleanup_module);
3975 MODULE_LICENSE("GPL");
3976 MODULE_VERSION(VXLAN_VERSION);
3977 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3978 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3979 MODULE_ALIAS_RTNL_LINK("vxlan");