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/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
31 #include <net/ndisc.h>
33 #include <net/ip_tunnels.h>
36 #include <net/udp_tunnel.h>
37 #include <net/rtnetlink.h>
38 #include <net/route.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/vxlan.h>
44 #include <net/protocol.h>
46 #if IS_ENABLED(CONFIG_IPV6)
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
52 #include <net/dst_metadata.h>
54 #define VXLAN_VERSION "0.1"
56 #define PORT_HASH_BITS 8
57 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61 /* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
63 * for compatibility with early adopters.
65 static unsigned short vxlan_port __read_mostly = 8472;
66 module_param_named(udp_port, vxlan_port, ushort, 0444);
67 MODULE_PARM_DESC(udp_port, "Destination UDP port");
69 static bool log_ecn_error = true;
70 module_param(log_ecn_error, bool, 0644);
71 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
73 static int vxlan_net_id;
74 static struct rtnl_link_ops vxlan_link_ops;
76 static const u8 all_zeros_mac[ETH_ALEN + 2];
78 static int vxlan_sock_add(struct vxlan_dev *vxlan);
80 /* per-network namespace private data for this module */
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
87 /* Forwarding table entry */
89 struct hlist_node hlist; /* linked list of entries */
91 unsigned long updated; /* jiffies */
93 struct list_head remotes;
94 u8 eth_addr[ETH_ALEN];
95 u16 state; /* see ndm_state */
96 u8 flags; /* see ndm_flags */
99 /* salt for hash table */
100 static u32 vxlan_salt __read_mostly;
101 static struct workqueue_struct *vxlan_wq;
103 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
109 #if IS_ENABLED(CONFIG_IPV6)
111 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
113 if (a->sa.sa_family != b->sa.sa_family)
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
121 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
129 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
137 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
141 ip->sa.sa_family = AF_INET6;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
145 ip->sa.sa_family = AF_INET;
148 return -EAFNOSUPPORT;
152 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
153 const union vxlan_addr *ip)
155 if (ip->sa.sa_family == AF_INET6)
156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
161 #else /* !CONFIG_IPV6 */
164 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
169 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
174 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
179 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
185 ip->sa.sa_family = AF_INET;
188 return -EAFNOSUPPORT;
192 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
199 /* Virtual Network hash table head */
200 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
202 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
205 /* Socket hash table head */
206 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
213 /* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
216 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
221 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
226 /* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
229 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
232 struct vxlan_sock *vs;
234 flags &= VXLAN_F_RCV_FLAGS;
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
238 vxlan_get_sk_family(vs) == family &&
245 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
247 struct vxlan_dev *vxlan;
249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs->flags & VXLAN_F_COLLECT_METADATA)
253 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
254 if (vxlan->default_dst.remote_vni == vni)
261 /* Look up VNI in a per net namespace table */
262 static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
263 sa_family_t family, __be16 port,
266 struct vxlan_sock *vs;
268 vs = vxlan_find_sock(net, family, port, flags);
272 return vxlan_vs_find_vni(vs, vni);
275 /* Fill in neighbour message in skbuff. */
276 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
277 const struct vxlan_fdb *fdb,
278 u32 portid, u32 seq, int type, unsigned int flags,
279 const struct vxlan_rdst *rdst)
281 unsigned long now = jiffies;
282 struct nda_cacheinfo ci;
283 struct nlmsghdr *nlh;
285 bool send_ip, send_eth;
287 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
291 ndm = nlmsg_data(nlh);
292 memset(ndm, 0, sizeof(*ndm));
294 send_eth = send_ip = true;
296 if (type == RTM_GETNEIGH) {
297 ndm->ndm_family = AF_INET;
298 send_ip = !vxlan_addr_any(&rdst->remote_ip);
299 send_eth = !is_zero_ether_addr(fdb->eth_addr);
301 ndm->ndm_family = AF_BRIDGE;
302 ndm->ndm_state = fdb->state;
303 ndm->ndm_ifindex = vxlan->dev->ifindex;
304 ndm->ndm_flags = fdb->flags;
305 ndm->ndm_type = RTN_UNICAST;
307 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
308 nla_put_s32(skb, NDA_LINK_NETNSID,
309 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
310 goto nla_put_failure;
312 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
313 goto nla_put_failure;
315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
316 goto nla_put_failure;
318 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
319 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
320 goto nla_put_failure;
321 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
322 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
326 goto nla_put_failure;
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
340 nlmsg_cancel(skb, nlh);
344 static inline size_t vxlan_nlmsg_size(void)
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
353 + nla_total_size(sizeof(struct nda_cacheinfo));
356 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
359 struct net *net = dev_net(vxlan->dev);
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
382 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
384 struct vxlan_dev *vxlan = netdev_priv(dev);
385 struct vxlan_fdb f = {
388 struct vxlan_rdst remote = {
389 .remote_ip = *ipa, /* goes to NDA_DST */
390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
396 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
398 struct vxlan_fdb f = {
401 struct vxlan_rdst remote = { };
403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
408 /* Hash Ethernet address */
409 static u32 eth_hash(const unsigned char *addr)
411 u64 value = get_unaligned((u64 *)addr);
413 /* only want 6 bytes */
419 return hash_64(value, FDB_HASH_BITS);
422 /* Hash chain to use given mac address */
423 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
426 return &vxlan->fdb_head[eth_hash(mac)];
429 /* Look up Ethernet address in forwarding table */
430 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
433 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
436 hlist_for_each_entry_rcu(f, head, hlist) {
437 if (ether_addr_equal(mac, f->eth_addr))
444 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
449 f = __vxlan_find_mac(vxlan, mac);
456 /* caller should hold vxlan->hash_lock */
457 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
458 union vxlan_addr *ip, __be16 port,
459 __be32 vni, __u32 ifindex)
461 struct vxlan_rdst *rd;
463 list_for_each_entry(rd, &f->remotes, list) {
464 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
465 rd->remote_port == port &&
466 rd->remote_vni == vni &&
467 rd->remote_ifindex == ifindex)
474 /* Replace destination of unicast mac */
475 static int vxlan_fdb_replace(struct vxlan_fdb *f,
476 union vxlan_addr *ip, __be16 port, __be32 vni,
479 struct vxlan_rdst *rd;
481 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
485 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
489 dst_cache_reset(&rd->dst_cache);
491 rd->remote_port = port;
492 rd->remote_vni = vni;
493 rd->remote_ifindex = ifindex;
497 /* Add/update destinations for multicast */
498 static int vxlan_fdb_append(struct vxlan_fdb *f,
499 union vxlan_addr *ip, __be16 port, __be32 vni,
500 __u32 ifindex, struct vxlan_rdst **rdp)
502 struct vxlan_rdst *rd;
504 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
508 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
512 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
522 list_add_tail_rcu(&rd->list, &f->remotes);
528 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
530 struct vxlanhdr *vh, size_t hdrlen,
532 struct gro_remcsum *grc,
535 size_t start, offset;
537 if (skb->remcsum_offload)
540 if (!NAPI_GRO_CB(skb)->csum_valid)
543 start = vxlan_rco_start(vni_field);
544 offset = start + vxlan_rco_offset(vni_field);
546 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
547 start, offset, grc, nopartial);
549 skb->remcsum_offload = 1;
554 static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
556 struct udp_offload *uoff)
558 struct sk_buff *p, **pp = NULL;
559 struct vxlanhdr *vh, *vh2;
560 unsigned int hlen, off_vx;
562 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
565 struct gro_remcsum grc;
567 skb_gro_remcsum_init(&grc);
569 off_vx = skb_gro_offset(skb);
570 hlen = off_vx + sizeof(*vh);
571 vh = skb_gro_header_fast(skb, off_vx);
572 if (skb_gro_header_hard(skb, hlen)) {
573 vh = skb_gro_header_slow(skb, hlen, off_vx);
578 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
580 flags = vh->vx_flags;
582 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
583 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
586 VXLAN_F_REMCSUM_NOPARTIAL));
592 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
596 for (p = *head; p; p = p->next) {
597 if (!NAPI_GRO_CB(p)->same_flow)
600 vh2 = (struct vxlanhdr *)(p->data + off_vx);
601 if (vh->vx_flags != vh2->vx_flags ||
602 vh->vx_vni != vh2->vx_vni) {
603 NAPI_GRO_CB(p)->same_flow = 0;
608 pp = eth_gro_receive(head, skb);
611 skb_gro_remcsum_cleanup(skb, &grc);
612 NAPI_GRO_CB(skb)->flush |= flush;
617 static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
618 struct udp_offload *uoff)
620 udp_tunnel_gro_complete(skb, nhoff);
622 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
625 /* Notify netdevs that UDP port started listening */
626 static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
628 struct net_device *dev;
629 struct sock *sk = vs->sock->sk;
630 struct net *net = sock_net(sk);
631 sa_family_t sa_family = vxlan_get_sk_family(vs);
632 __be16 port = inet_sk(sk)->inet_sport;
635 if (sa_family == AF_INET) {
636 err = udp_add_offload(net, &vs->udp_offloads);
638 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
642 for_each_netdev_rcu(net, dev) {
643 if (dev->netdev_ops->ndo_add_vxlan_port)
644 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
650 /* Notify netdevs that UDP port is no more listening */
651 static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
653 struct net_device *dev;
654 struct sock *sk = vs->sock->sk;
655 struct net *net = sock_net(sk);
656 sa_family_t sa_family = vxlan_get_sk_family(vs);
657 __be16 port = inet_sk(sk)->inet_sport;
660 for_each_netdev_rcu(net, dev) {
661 if (dev->netdev_ops->ndo_del_vxlan_port)
662 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
667 if (sa_family == AF_INET)
668 udp_del_offload(&vs->udp_offloads);
671 /* Add new entry to forwarding table -- assumes lock held */
672 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
673 const u8 *mac, union vxlan_addr *ip,
674 __u16 state, __u16 flags,
675 __be16 port, __be32 vni, __u32 ifindex,
678 struct vxlan_rdst *rd = NULL;
682 f = __vxlan_find_mac(vxlan, mac);
684 if (flags & NLM_F_EXCL) {
685 netdev_dbg(vxlan->dev,
686 "lost race to create %pM\n", mac);
689 if (f->state != state) {
691 f->updated = jiffies;
694 if (f->flags != ndm_flags) {
695 f->flags = ndm_flags;
696 f->updated = jiffies;
699 if ((flags & NLM_F_REPLACE)) {
700 /* Only change unicasts */
701 if (!(is_multicast_ether_addr(f->eth_addr) ||
702 is_zero_ether_addr(f->eth_addr))) {
703 notify |= vxlan_fdb_replace(f, ip, port, vni,
708 if ((flags & NLM_F_APPEND) &&
709 (is_multicast_ether_addr(f->eth_addr) ||
710 is_zero_ether_addr(f->eth_addr))) {
711 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
719 if (!(flags & NLM_F_CREATE))
722 if (vxlan->cfg.addrmax &&
723 vxlan->addrcnt >= vxlan->cfg.addrmax)
726 /* Disallow replace to add a multicast entry */
727 if ((flags & NLM_F_REPLACE) &&
728 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
731 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
732 f = kmalloc(sizeof(*f), GFP_ATOMIC);
738 f->flags = ndm_flags;
739 f->updated = f->used = jiffies;
740 INIT_LIST_HEAD(&f->remotes);
741 memcpy(f->eth_addr, mac, ETH_ALEN);
743 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
746 hlist_add_head_rcu(&f->hlist,
747 vxlan_fdb_head(vxlan, mac));
752 rd = first_remote_rtnl(f);
753 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
759 static void vxlan_fdb_free(struct rcu_head *head)
761 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
762 struct vxlan_rdst *rd, *nd;
764 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
765 dst_cache_destroy(&rd->dst_cache);
771 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
773 netdev_dbg(vxlan->dev,
774 "delete %pM\n", f->eth_addr);
777 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
779 hlist_del_rcu(&f->hlist);
780 call_rcu(&f->rcu, vxlan_fdb_free);
783 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
784 union vxlan_addr *ip, __be16 *port, __be32 *vni,
787 struct net *net = dev_net(vxlan->dev);
791 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
795 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
796 if (remote->sa.sa_family == AF_INET) {
797 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
798 ip->sa.sa_family = AF_INET;
799 #if IS_ENABLED(CONFIG_IPV6)
801 ip->sin6.sin6_addr = in6addr_any;
802 ip->sa.sa_family = AF_INET6;
808 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
810 *port = nla_get_be16(tb[NDA_PORT]);
812 *port = vxlan->cfg.dst_port;
816 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
818 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
820 *vni = vxlan->default_dst.remote_vni;
823 if (tb[NDA_IFINDEX]) {
824 struct net_device *tdev;
826 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
828 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
829 tdev = __dev_get_by_index(net, *ifindex);
831 return -EADDRNOTAVAIL;
839 /* Add static entry (via netlink) */
840 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
841 struct net_device *dev,
842 const unsigned char *addr, u16 vid, u16 flags)
844 struct vxlan_dev *vxlan = netdev_priv(dev);
845 /* struct net *net = dev_net(vxlan->dev); */
852 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
853 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
858 if (tb[NDA_DST] == NULL)
861 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
865 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
866 return -EAFNOSUPPORT;
868 spin_lock_bh(&vxlan->hash_lock);
869 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
870 port, vni, ifindex, ndm->ndm_flags);
871 spin_unlock_bh(&vxlan->hash_lock);
876 /* Delete entry (via netlink) */
877 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
878 struct net_device *dev,
879 const unsigned char *addr, u16 vid)
881 struct vxlan_dev *vxlan = netdev_priv(dev);
883 struct vxlan_rdst *rd = NULL;
890 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
896 spin_lock_bh(&vxlan->hash_lock);
897 f = vxlan_find_mac(vxlan, addr);
901 if (!vxlan_addr_any(&ip)) {
902 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
909 /* remove a destination if it's not the only one on the list,
910 * otherwise destroy the fdb entry
912 if (rd && !list_is_singular(&f->remotes)) {
913 list_del_rcu(&rd->list);
914 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
919 vxlan_fdb_destroy(vxlan, f);
922 spin_unlock_bh(&vxlan->hash_lock);
927 /* Dump forwarding table */
928 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
929 struct net_device *dev,
930 struct net_device *filter_dev, int idx)
932 struct vxlan_dev *vxlan = netdev_priv(dev);
935 for (h = 0; h < FDB_HASH_SIZE; ++h) {
939 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
940 struct vxlan_rdst *rd;
942 list_for_each_entry_rcu(rd, &f->remotes, list) {
943 if (idx < cb->args[0])
946 err = vxlan_fdb_info(skb, vxlan, f,
947 NETLINK_CB(cb->skb).portid,
962 /* Watch incoming packets to learn mapping between Ethernet address
963 * and Tunnel endpoint.
964 * Return true if packet is bogus and should be dropped.
966 static bool vxlan_snoop(struct net_device *dev,
967 union vxlan_addr *src_ip, const u8 *src_mac)
969 struct vxlan_dev *vxlan = netdev_priv(dev);
972 f = vxlan_find_mac(vxlan, src_mac);
974 struct vxlan_rdst *rdst = first_remote_rcu(f);
976 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
979 /* Don't migrate static entries, drop packets */
980 if (f->state & NUD_NOARP)
985 "%pM migrated from %pIS to %pIS\n",
986 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
988 rdst->remote_ip = *src_ip;
989 f->updated = jiffies;
990 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
992 /* learned new entry */
993 spin_lock(&vxlan->hash_lock);
995 /* close off race between vxlan_flush and incoming packets */
996 if (netif_running(dev))
997 vxlan_fdb_create(vxlan, src_mac, src_ip,
999 NLM_F_EXCL|NLM_F_CREATE,
1000 vxlan->cfg.dst_port,
1001 vxlan->default_dst.remote_vni,
1003 spin_unlock(&vxlan->hash_lock);
1009 /* See if multicast group is already in use by other ID */
1010 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1012 struct vxlan_dev *vxlan;
1013 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1015 /* The vxlan_sock is only used by dev, leaving group has
1016 * no effect on other vxlan devices.
1018 if (family == AF_INET && dev->vn4_sock &&
1019 atomic_read(&dev->vn4_sock->refcnt) == 1)
1021 #if IS_ENABLED(CONFIG_IPV6)
1022 if (family == AF_INET6 && dev->vn6_sock &&
1023 atomic_read(&dev->vn6_sock->refcnt) == 1)
1027 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1028 if (!netif_running(vxlan->dev) || vxlan == dev)
1031 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
1033 #if IS_ENABLED(CONFIG_IPV6)
1034 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1038 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1039 &dev->default_dst.remote_ip))
1042 if (vxlan->default_dst.remote_ifindex !=
1043 dev->default_dst.remote_ifindex)
1052 static void __vxlan_sock_release(struct vxlan_sock *vs)
1054 struct vxlan_net *vn;
1058 if (!atomic_dec_and_test(&vs->refcnt))
1061 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1062 spin_lock(&vn->sock_lock);
1063 hlist_del_rcu(&vs->hlist);
1064 vxlan_notify_del_rx_port(vs);
1065 spin_unlock(&vn->sock_lock);
1067 queue_work(vxlan_wq, &vs->del_work);
1070 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1072 __vxlan_sock_release(vxlan->vn4_sock);
1073 #if IS_ENABLED(CONFIG_IPV6)
1074 __vxlan_sock_release(vxlan->vn6_sock);
1078 /* Update multicast group membership when first VNI on
1079 * multicast address is brought up
1081 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1084 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1085 int ifindex = vxlan->default_dst.remote_ifindex;
1088 if (ip->sa.sa_family == AF_INET) {
1089 struct ip_mreqn mreq = {
1090 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1091 .imr_ifindex = ifindex,
1094 sk = vxlan->vn4_sock->sock->sk;
1096 ret = ip_mc_join_group(sk, &mreq);
1098 #if IS_ENABLED(CONFIG_IPV6)
1100 sk = vxlan->vn6_sock->sock->sk;
1102 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1103 &ip->sin6.sin6_addr);
1111 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1112 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1115 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1116 int ifindex = vxlan->default_dst.remote_ifindex;
1119 if (ip->sa.sa_family == AF_INET) {
1120 struct ip_mreqn mreq = {
1121 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1122 .imr_ifindex = ifindex,
1125 sk = vxlan->vn4_sock->sock->sk;
1127 ret = ip_mc_leave_group(sk, &mreq);
1129 #if IS_ENABLED(CONFIG_IPV6)
1131 sk = vxlan->vn6_sock->sock->sk;
1133 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1134 &ip->sin6.sin6_addr);
1142 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1143 struct sk_buff *skb, u32 vxflags)
1145 size_t start, offset, plen;
1147 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1150 start = vxlan_rco_start(unparsed->vx_vni);
1151 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1153 plen = sizeof(struct vxlanhdr) + offset + sizeof(u16);
1155 if (!pskb_may_pull(skb, plen))
1158 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1159 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1161 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1162 unparsed->vx_vni &= VXLAN_VNI_MASK;
1166 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1167 struct vxlan_metadata *md,
1168 struct metadata_dst *tun_dst)
1170 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1172 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1175 md->gbp = ntohs(gbp->policy_id);
1178 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1180 if (gbp->dont_learn)
1181 md->gbp |= VXLAN_GBP_DONT_LEARN;
1183 if (gbp->policy_applied)
1184 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1187 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1190 static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1191 struct vxlan_metadata *md, __be32 vni,
1192 struct metadata_dst *tun_dst)
1194 struct iphdr *oip = NULL;
1195 struct ipv6hdr *oip6 = NULL;
1196 struct vxlan_dev *vxlan;
1197 struct pcpu_sw_netstats *stats;
1198 union vxlan_addr saddr;
1201 /* Is this VNI defined? */
1202 vxlan = vxlan_vs_find_vni(vs, vni);
1206 skb_reset_mac_header(skb);
1207 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1208 skb->protocol = eth_type_trans(skb, vxlan->dev);
1209 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1211 /* Ignore packet loops (and multicast echo) */
1212 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1215 /* Get data from the outer IP header */
1216 if (vxlan_get_sk_family(vs) == AF_INET) {
1218 saddr.sin.sin_addr.s_addr = oip->saddr;
1219 saddr.sa.sa_family = AF_INET;
1220 #if IS_ENABLED(CONFIG_IPV6)
1222 oip6 = ipv6_hdr(skb);
1223 saddr.sin6.sin6_addr = oip6->saddr;
1224 saddr.sa.sa_family = AF_INET6;
1229 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1233 if ((vxlan->flags & VXLAN_F_LEARN) &&
1234 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1237 skb_reset_network_header(skb);
1238 /* In flow-based mode, GBP is carried in dst_metadata */
1239 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
1240 skb->mark = md->gbp;
1243 err = IP6_ECN_decapsulate(oip6, skb);
1245 err = IP_ECN_decapsulate(oip, skb);
1247 if (unlikely(err)) {
1248 if (log_ecn_error) {
1250 net_info_ratelimited("non-ECT from %pI6\n",
1253 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1254 &oip->saddr, oip->tos);
1257 ++vxlan->dev->stats.rx_frame_errors;
1258 ++vxlan->dev->stats.rx_errors;
1263 stats = this_cpu_ptr(vxlan->dev->tstats);
1264 u64_stats_update_begin(&stats->syncp);
1265 stats->rx_packets++;
1266 stats->rx_bytes += skb->len;
1267 u64_stats_update_end(&stats->syncp);
1269 gro_cells_receive(&vxlan->gro_cells, skb);
1274 dst_release((struct dst_entry *)tun_dst);
1276 /* Consume bad packet */
1280 /* Callback from net/ipv4/udp.c to receive packets */
1281 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1283 struct metadata_dst *tun_dst = NULL;
1284 struct vxlan_sock *vs;
1285 struct vxlanhdr unparsed;
1286 struct vxlan_metadata _md;
1287 struct vxlan_metadata *md = &_md;
1289 /* Need Vxlan and inner Ethernet header to be present */
1290 if (!pskb_may_pull(skb, VXLAN_HLEN))
1293 unparsed = *vxlan_hdr(skb);
1294 /* VNI flag always required to be set */
1295 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1296 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1297 ntohl(vxlan_hdr(skb)->vx_flags),
1298 ntohl(vxlan_hdr(skb)->vx_vni));
1299 /* Return non vxlan pkt */
1302 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1303 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1305 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1308 vs = rcu_dereference_sk_user_data(sk);
1312 if (vxlan_collect_metadata(vs)) {
1313 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1315 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1316 vxlan_vni_to_tun_id(vni), sizeof(*md));
1321 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1323 memset(md, 0, sizeof(*md));
1326 /* For backwards compatibility, only allow reserved fields to be
1327 * used by VXLAN extensions if explicitly requested.
1329 if (vs->flags & VXLAN_F_REMCSUM_RX)
1330 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1332 if (vs->flags & VXLAN_F_GBP)
1333 vxlan_parse_gbp_hdr(&unparsed, md, tun_dst);
1335 if (unparsed.vx_flags || unparsed.vx_vni) {
1336 /* If there are any unprocessed flags remaining treat
1337 * this as a malformed packet. This behavior diverges from
1338 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1339 * in reserved fields are to be ignored. The approach here
1340 * maintains compatibility with previous stack code, and also
1341 * is more robust and provides a little more security in
1342 * adding extensions to VXLAN.
1347 vxlan_rcv(vs, skb, md, vxlan_vni(vxlan_hdr(skb)->vx_vni), tun_dst);
1352 dst_release((struct dst_entry *)tun_dst);
1354 /* Consume bad packet */
1359 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1361 struct vxlan_dev *vxlan = netdev_priv(dev);
1362 struct arphdr *parp;
1365 struct neighbour *n;
1367 if (dev->flags & IFF_NOARP)
1370 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1371 dev->stats.tx_dropped++;
1374 parp = arp_hdr(skb);
1376 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1377 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1378 parp->ar_pro != htons(ETH_P_IP) ||
1379 parp->ar_op != htons(ARPOP_REQUEST) ||
1380 parp->ar_hln != dev->addr_len ||
1383 arpptr = (u8 *)parp + sizeof(struct arphdr);
1385 arpptr += dev->addr_len; /* sha */
1386 memcpy(&sip, arpptr, sizeof(sip));
1387 arpptr += sizeof(sip);
1388 arpptr += dev->addr_len; /* tha */
1389 memcpy(&tip, arpptr, sizeof(tip));
1391 if (ipv4_is_loopback(tip) ||
1392 ipv4_is_multicast(tip))
1395 n = neigh_lookup(&arp_tbl, &tip, dev);
1398 struct vxlan_fdb *f;
1399 struct sk_buff *reply;
1401 if (!(n->nud_state & NUD_CONNECTED)) {
1406 f = vxlan_find_mac(vxlan, n->ha);
1407 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1408 /* bridge-local neighbor */
1413 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1421 skb_reset_mac_header(reply);
1422 __skb_pull(reply, skb_network_offset(reply));
1423 reply->ip_summed = CHECKSUM_UNNECESSARY;
1424 reply->pkt_type = PACKET_HOST;
1426 if (netif_rx_ni(reply) == NET_RX_DROP)
1427 dev->stats.rx_dropped++;
1428 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1429 union vxlan_addr ipa = {
1430 .sin.sin_addr.s_addr = tip,
1431 .sin.sin_family = AF_INET,
1434 vxlan_ip_miss(dev, &ipa);
1438 return NETDEV_TX_OK;
1441 #if IS_ENABLED(CONFIG_IPV6)
1442 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1443 struct neighbour *n, bool isrouter)
1445 struct net_device *dev = request->dev;
1446 struct sk_buff *reply;
1447 struct nd_msg *ns, *na;
1448 struct ipv6hdr *pip6;
1450 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1457 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1458 sizeof(*na) + na_olen + dev->needed_tailroom;
1459 reply = alloc_skb(len, GFP_ATOMIC);
1463 reply->protocol = htons(ETH_P_IPV6);
1465 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1466 skb_push(reply, sizeof(struct ethhdr));
1467 skb_set_mac_header(reply, 0);
1469 ns = (struct nd_msg *)skb_transport_header(request);
1471 daddr = eth_hdr(request)->h_source;
1472 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1473 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1474 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1475 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1480 /* Ethernet header */
1481 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1482 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1483 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1484 reply->protocol = htons(ETH_P_IPV6);
1486 skb_pull(reply, sizeof(struct ethhdr));
1487 skb_set_network_header(reply, 0);
1488 skb_put(reply, sizeof(struct ipv6hdr));
1492 pip6 = ipv6_hdr(reply);
1493 memset(pip6, 0, sizeof(struct ipv6hdr));
1495 pip6->priority = ipv6_hdr(request)->priority;
1496 pip6->nexthdr = IPPROTO_ICMPV6;
1497 pip6->hop_limit = 255;
1498 pip6->daddr = ipv6_hdr(request)->saddr;
1499 pip6->saddr = *(struct in6_addr *)n->primary_key;
1501 skb_pull(reply, sizeof(struct ipv6hdr));
1502 skb_set_transport_header(reply, 0);
1504 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1506 /* Neighbor Advertisement */
1507 memset(na, 0, sizeof(*na)+na_olen);
1508 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1509 na->icmph.icmp6_router = isrouter;
1510 na->icmph.icmp6_override = 1;
1511 na->icmph.icmp6_solicited = 1;
1512 na->target = ns->target;
1513 ether_addr_copy(&na->opt[2], n->ha);
1514 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1515 na->opt[1] = na_olen >> 3;
1517 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1518 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1519 csum_partial(na, sizeof(*na)+na_olen, 0));
1521 pip6->payload_len = htons(sizeof(*na)+na_olen);
1523 skb_push(reply, sizeof(struct ipv6hdr));
1525 reply->ip_summed = CHECKSUM_UNNECESSARY;
1530 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1532 struct vxlan_dev *vxlan = netdev_priv(dev);
1534 const struct ipv6hdr *iphdr;
1535 const struct in6_addr *saddr, *daddr;
1536 struct neighbour *n;
1537 struct inet6_dev *in6_dev;
1539 in6_dev = __in6_dev_get(dev);
1543 iphdr = ipv6_hdr(skb);
1544 saddr = &iphdr->saddr;
1545 daddr = &iphdr->daddr;
1547 msg = (struct nd_msg *)skb_transport_header(skb);
1548 if (msg->icmph.icmp6_code != 0 ||
1549 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1552 if (ipv6_addr_loopback(daddr) ||
1553 ipv6_addr_is_multicast(&msg->target))
1556 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1559 struct vxlan_fdb *f;
1560 struct sk_buff *reply;
1562 if (!(n->nud_state & NUD_CONNECTED)) {
1567 f = vxlan_find_mac(vxlan, n->ha);
1568 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1569 /* bridge-local neighbor */
1574 reply = vxlan_na_create(skb, n,
1575 !!(f ? f->flags & NTF_ROUTER : 0));
1582 if (netif_rx_ni(reply) == NET_RX_DROP)
1583 dev->stats.rx_dropped++;
1585 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1586 union vxlan_addr ipa = {
1587 .sin6.sin6_addr = msg->target,
1588 .sin6.sin6_family = AF_INET6,
1591 vxlan_ip_miss(dev, &ipa);
1596 return NETDEV_TX_OK;
1600 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1602 struct vxlan_dev *vxlan = netdev_priv(dev);
1603 struct neighbour *n;
1605 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1609 switch (ntohs(eth_hdr(skb)->h_proto)) {
1614 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1617 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1618 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1619 union vxlan_addr ipa = {
1620 .sin.sin_addr.s_addr = pip->daddr,
1621 .sin.sin_family = AF_INET,
1624 vxlan_ip_miss(dev, &ipa);
1630 #if IS_ENABLED(CONFIG_IPV6)
1633 struct ipv6hdr *pip6;
1635 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1637 pip6 = ipv6_hdr(skb);
1638 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1639 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1640 union vxlan_addr ipa = {
1641 .sin6.sin6_addr = pip6->daddr,
1642 .sin6.sin6_family = AF_INET6,
1645 vxlan_ip_miss(dev, &ipa);
1659 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1661 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1663 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1672 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1673 struct vxlan_metadata *md)
1675 struct vxlanhdr_gbp *gbp;
1680 gbp = (struct vxlanhdr_gbp *)vxh;
1681 vxh->vx_flags |= VXLAN_HF_GBP;
1683 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1684 gbp->dont_learn = 1;
1686 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1687 gbp->policy_applied = 1;
1689 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1692 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1693 int iphdr_len, __be32 vni,
1694 struct vxlan_metadata *md, u32 vxflags,
1697 struct vxlanhdr *vxh;
1700 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1702 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1703 skb->ip_summed == CHECKSUM_PARTIAL) {
1704 int csum_start = skb_checksum_start_offset(skb);
1706 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1707 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1708 (skb->csum_offset == offsetof(struct udphdr, check) ||
1709 skb->csum_offset == offsetof(struct tcphdr, check)))
1710 type |= SKB_GSO_TUNNEL_REMCSUM;
1713 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1714 + VXLAN_HLEN + iphdr_len
1715 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
1717 /* Need space for new headers (invalidates iph ptr) */
1718 err = skb_cow_head(skb, min_headroom);
1719 if (unlikely(err)) {
1724 skb = vlan_hwaccel_push_inside(skb);
1728 skb = iptunnel_handle_offloads(skb, type);
1730 return PTR_ERR(skb);
1732 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1733 vxh->vx_flags = VXLAN_HF_VNI;
1734 vxh->vx_vni = vxlan_vni_field(vni);
1736 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1739 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1740 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1741 vxh->vx_flags |= VXLAN_HF_RCO;
1743 if (!skb_is_gso(skb)) {
1744 skb->ip_summed = CHECKSUM_NONE;
1745 skb->encapsulation = 0;
1749 if (vxflags & VXLAN_F_GBP)
1750 vxlan_build_gbp_hdr(vxh, vxflags, md);
1752 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1756 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1757 struct sk_buff *skb, int oif, u8 tos,
1758 __be32 daddr, __be32 *saddr,
1759 struct dst_cache *dst_cache,
1760 struct ip_tunnel_info *info)
1762 struct rtable *rt = NULL;
1763 bool use_cache = false;
1766 /* when the ip_tunnel_info is availble, the tos used for lookup is
1767 * packet independent, so we can use the cache
1769 if (!skb->mark && (!tos || info)) {
1771 rt = dst_cache_get_ip4(dst_cache, saddr);
1776 memset(&fl4, 0, sizeof(fl4));
1777 fl4.flowi4_oif = oif;
1778 fl4.flowi4_tos = RT_TOS(tos);
1779 fl4.flowi4_mark = skb->mark;
1780 fl4.flowi4_proto = IPPROTO_UDP;
1782 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1784 rt = ip_route_output_key(vxlan->net, &fl4);
1788 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1793 #if IS_ENABLED(CONFIG_IPV6)
1794 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1795 struct sk_buff *skb, int oif,
1796 const struct in6_addr *daddr,
1797 struct in6_addr *saddr,
1798 struct dst_cache *dst_cache)
1800 struct dst_entry *ndst;
1805 ndst = dst_cache_get_ip6(dst_cache, saddr);
1810 memset(&fl6, 0, sizeof(fl6));
1811 fl6.flowi6_oif = oif;
1813 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1814 fl6.flowi6_mark = skb->mark;
1815 fl6.flowi6_proto = IPPROTO_UDP;
1817 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1818 vxlan->vn6_sock->sock->sk,
1821 return ERR_PTR(err);
1825 dst_cache_set_ip6(dst_cache, ndst, saddr);
1830 /* Bypass encapsulation if the destination is local */
1831 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1832 struct vxlan_dev *dst_vxlan)
1834 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1835 union vxlan_addr loopback;
1836 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1837 struct net_device *dev = skb->dev;
1840 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1841 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1842 skb->pkt_type = PACKET_HOST;
1843 skb->encapsulation = 0;
1844 skb->dev = dst_vxlan->dev;
1845 __skb_pull(skb, skb_network_offset(skb));
1847 if (remote_ip->sa.sa_family == AF_INET) {
1848 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1849 loopback.sa.sa_family = AF_INET;
1850 #if IS_ENABLED(CONFIG_IPV6)
1852 loopback.sin6.sin6_addr = in6addr_loopback;
1853 loopback.sa.sa_family = AF_INET6;
1857 if (dst_vxlan->flags & VXLAN_F_LEARN)
1858 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1860 u64_stats_update_begin(&tx_stats->syncp);
1861 tx_stats->tx_packets++;
1862 tx_stats->tx_bytes += len;
1863 u64_stats_update_end(&tx_stats->syncp);
1865 if (netif_rx(skb) == NET_RX_SUCCESS) {
1866 u64_stats_update_begin(&rx_stats->syncp);
1867 rx_stats->rx_packets++;
1868 rx_stats->rx_bytes += len;
1869 u64_stats_update_end(&rx_stats->syncp);
1871 dev->stats.rx_dropped++;
1875 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1876 struct vxlan_rdst *rdst, bool did_rsc)
1878 struct dst_cache *dst_cache;
1879 struct ip_tunnel_info *info;
1880 struct vxlan_dev *vxlan = netdev_priv(dev);
1882 struct rtable *rt = NULL;
1883 const struct iphdr *old_iph;
1884 union vxlan_addr *dst;
1885 union vxlan_addr remote_ip;
1886 struct vxlan_metadata _md;
1887 struct vxlan_metadata *md = &_md;
1888 __be16 src_port = 0, dst_port;
1893 u32 flags = vxlan->flags;
1894 bool udp_sum = false;
1895 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
1897 info = skb_tunnel_info(skb);
1900 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
1901 vni = rdst->remote_vni;
1902 dst = &rdst->remote_ip;
1903 dst_cache = &rdst->dst_cache;
1906 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1910 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
1911 vni = vxlan_tun_id_to_vni(info->key.tun_id);
1912 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1913 if (remote_ip.sa.sa_family == AF_INET)
1914 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1916 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
1918 dst_cache = &info->dst_cache;
1921 if (vxlan_addr_any(dst)) {
1923 /* short-circuited back to local bridge */
1924 vxlan_encap_bypass(skb, vxlan, vxlan);
1930 old_iph = ip_hdr(skb);
1932 ttl = vxlan->cfg.ttl;
1933 if (!ttl && vxlan_addr_multicast(dst))
1936 tos = vxlan->cfg.tos;
1938 tos = ip_tunnel_get_dsfield(old_iph, skb);
1940 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1941 vxlan->cfg.port_max, true);
1944 ttl = info->key.ttl;
1945 tos = info->key.tos;
1946 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
1948 if (info->options_len)
1949 md = ip_tunnel_info_opts(info);
1951 md->gbp = skb->mark;
1954 if (dst->sa.sa_family == AF_INET) {
1957 if (!vxlan->vn4_sock)
1959 sk = vxlan->vn4_sock->sock->sk;
1962 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1965 udp_sum = !!(flags & VXLAN_F_UDP_CSUM);
1968 rt = vxlan_get_route(vxlan, skb,
1969 rdst ? rdst->remote_ifindex : 0, tos,
1970 dst->sin.sin_addr.s_addr, &saddr,
1973 netdev_dbg(dev, "no route to %pI4\n",
1974 &dst->sin.sin_addr.s_addr);
1975 dev->stats.tx_carrier_errors++;
1979 if (rt->dst.dev == dev) {
1980 netdev_dbg(dev, "circular route to %pI4\n",
1981 &dst->sin.sin_addr.s_addr);
1982 dev->stats.collisions++;
1986 /* Bypass encapsulation if the destination is local */
1987 if (rt->rt_flags & RTCF_LOCAL &&
1988 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1989 struct vxlan_dev *dst_vxlan;
1992 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1993 dst->sa.sa_family, dst_port,
1997 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2001 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2002 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2003 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
2004 vni, md, flags, udp_sum);
2008 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2009 dst->sin.sin_addr.s_addr, tos, ttl, df,
2010 src_port, dst_port, xnet, !udp_sum);
2011 #if IS_ENABLED(CONFIG_IPV6)
2013 struct dst_entry *ndst;
2014 struct in6_addr saddr;
2017 if (!vxlan->vn6_sock)
2019 sk = vxlan->vn6_sock->sock->sk;
2021 ndst = vxlan6_get_route(vxlan, skb,
2022 rdst ? rdst->remote_ifindex : 0,
2023 &dst->sin6.sin6_addr, &saddr,
2026 netdev_dbg(dev, "no route to %pI6\n",
2027 &dst->sin6.sin6_addr);
2028 dev->stats.tx_carrier_errors++;
2032 if (ndst->dev == dev) {
2033 netdev_dbg(dev, "circular route to %pI6\n",
2034 &dst->sin6.sin6_addr);
2036 dev->stats.collisions++;
2040 /* Bypass encapsulation if the destination is local */
2041 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2042 if (rt6i_flags & RTF_LOCAL &&
2043 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2044 struct vxlan_dev *dst_vxlan;
2047 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2048 dst->sa.sa_family, dst_port,
2052 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2057 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2059 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2060 skb_scrub_packet(skb, xnet);
2061 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2062 vni, md, flags, udp_sum);
2067 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
2068 &saddr, &dst->sin6.sin6_addr,
2069 0, ttl, src_port, dst_port, !udp_sum);
2076 dev->stats.tx_dropped++;
2080 /* skb is already freed. */
2085 dev->stats.tx_errors++;
2090 /* Transmit local packets over Vxlan
2092 * Outer IP header inherits ECN and DF from inner header.
2093 * Outer UDP destination is the VXLAN assigned port.
2094 * source port is based on hash of flow
2096 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2098 struct vxlan_dev *vxlan = netdev_priv(dev);
2099 const struct ip_tunnel_info *info;
2101 bool did_rsc = false;
2102 struct vxlan_rdst *rdst, *fdst = NULL;
2103 struct vxlan_fdb *f;
2105 info = skb_tunnel_info(skb);
2107 skb_reset_mac_header(skb);
2110 if ((vxlan->flags & VXLAN_F_PROXY)) {
2111 if (ntohs(eth->h_proto) == ETH_P_ARP)
2112 return arp_reduce(dev, skb);
2113 #if IS_ENABLED(CONFIG_IPV6)
2114 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2115 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2116 + sizeof(struct nd_msg)) &&
2117 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2120 msg = (struct nd_msg *)skb_transport_header(skb);
2121 if (msg->icmph.icmp6_code == 0 &&
2122 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2123 return neigh_reduce(dev, skb);
2129 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
2130 info && info->mode & IP_TUNNEL_INFO_TX) {
2131 vxlan_xmit_one(skb, dev, NULL, false);
2132 return NETDEV_TX_OK;
2135 f = vxlan_find_mac(vxlan, eth->h_dest);
2138 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
2139 (ntohs(eth->h_proto) == ETH_P_IP ||
2140 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2141 did_rsc = route_shortcircuit(dev, skb);
2143 f = vxlan_find_mac(vxlan, eth->h_dest);
2147 f = vxlan_find_mac(vxlan, all_zeros_mac);
2149 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2150 !is_multicast_ether_addr(eth->h_dest))
2151 vxlan_fdb_miss(vxlan, eth->h_dest);
2153 dev->stats.tx_dropped++;
2155 return NETDEV_TX_OK;
2159 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2160 struct sk_buff *skb1;
2166 skb1 = skb_clone(skb, GFP_ATOMIC);
2168 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2172 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2175 return NETDEV_TX_OK;
2178 /* Walk the forwarding table and purge stale entries */
2179 static void vxlan_cleanup(unsigned long arg)
2181 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2182 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2185 if (!netif_running(vxlan->dev))
2188 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2189 struct hlist_node *p, *n;
2191 spin_lock_bh(&vxlan->hash_lock);
2192 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2194 = container_of(p, struct vxlan_fdb, hlist);
2195 unsigned long timeout;
2197 if (f->state & NUD_PERMANENT)
2200 timeout = f->used + vxlan->cfg.age_interval * HZ;
2201 if (time_before_eq(timeout, jiffies)) {
2202 netdev_dbg(vxlan->dev,
2203 "garbage collect %pM\n",
2205 f->state = NUD_STALE;
2206 vxlan_fdb_destroy(vxlan, f);
2207 } else if (time_before(timeout, next_timer))
2208 next_timer = timeout;
2210 spin_unlock_bh(&vxlan->hash_lock);
2213 mod_timer(&vxlan->age_timer, next_timer);
2216 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2218 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2219 __be32 vni = vxlan->default_dst.remote_vni;
2221 spin_lock(&vn->sock_lock);
2222 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2223 spin_unlock(&vn->sock_lock);
2226 /* Setup stats when device is created */
2227 static int vxlan_init(struct net_device *dev)
2229 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2236 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2238 struct vxlan_fdb *f;
2240 spin_lock_bh(&vxlan->hash_lock);
2241 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2243 vxlan_fdb_destroy(vxlan, f);
2244 spin_unlock_bh(&vxlan->hash_lock);
2247 static void vxlan_uninit(struct net_device *dev)
2249 struct vxlan_dev *vxlan = netdev_priv(dev);
2251 vxlan_fdb_delete_default(vxlan);
2253 free_percpu(dev->tstats);
2256 /* Start ageing timer and join group when device is brought up */
2257 static int vxlan_open(struct net_device *dev)
2259 struct vxlan_dev *vxlan = netdev_priv(dev);
2262 ret = vxlan_sock_add(vxlan);
2266 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2267 ret = vxlan_igmp_join(vxlan);
2268 if (ret == -EADDRINUSE)
2271 vxlan_sock_release(vxlan);
2276 if (vxlan->cfg.age_interval)
2277 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2282 /* Purge the forwarding table */
2283 static void vxlan_flush(struct vxlan_dev *vxlan)
2287 spin_lock_bh(&vxlan->hash_lock);
2288 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2289 struct hlist_node *p, *n;
2290 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2292 = container_of(p, struct vxlan_fdb, hlist);
2293 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2294 if (!is_zero_ether_addr(f->eth_addr))
2295 vxlan_fdb_destroy(vxlan, f);
2298 spin_unlock_bh(&vxlan->hash_lock);
2301 /* Cleanup timer and forwarding table on shutdown */
2302 static int vxlan_stop(struct net_device *dev)
2304 struct vxlan_dev *vxlan = netdev_priv(dev);
2305 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2308 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2309 !vxlan_group_used(vn, vxlan))
2310 ret = vxlan_igmp_leave(vxlan);
2312 del_timer_sync(&vxlan->age_timer);
2315 vxlan_sock_release(vxlan);
2320 /* Stub, nothing needs to be done. */
2321 static void vxlan_set_multicast_list(struct net_device *dev)
2325 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2327 struct vxlan_dev *vxlan = netdev_priv(dev);
2328 struct vxlan_rdst *dst = &vxlan->default_dst;
2329 struct net_device *lowerdev;
2332 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
2333 if (lowerdev == NULL)
2334 return eth_change_mtu(dev, new_mtu);
2336 if (dst->remote_ip.sa.sa_family == AF_INET6)
2337 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2339 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2341 if (new_mtu < 68 || new_mtu > max_mtu)
2348 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2350 struct vxlan_dev *vxlan = netdev_priv(dev);
2351 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2352 __be16 sport, dport;
2354 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2355 vxlan->cfg.port_max, true);
2356 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2358 if (ip_tunnel_info_af(info) == AF_INET) {
2361 if (!vxlan->vn4_sock)
2363 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2364 info->key.u.ipv4.dst,
2365 &info->key.u.ipv4.src, NULL, info);
2370 #if IS_ENABLED(CONFIG_IPV6)
2371 struct dst_entry *ndst;
2373 if (!vxlan->vn6_sock)
2375 ndst = vxlan6_get_route(vxlan, skb, 0,
2376 &info->key.u.ipv6.dst,
2377 &info->key.u.ipv6.src, NULL);
2379 return PTR_ERR(ndst);
2381 #else /* !CONFIG_IPV6 */
2382 return -EPFNOSUPPORT;
2385 info->key.tp_src = sport;
2386 info->key.tp_dst = dport;
2390 static const struct net_device_ops vxlan_netdev_ops = {
2391 .ndo_init = vxlan_init,
2392 .ndo_uninit = vxlan_uninit,
2393 .ndo_open = vxlan_open,
2394 .ndo_stop = vxlan_stop,
2395 .ndo_start_xmit = vxlan_xmit,
2396 .ndo_get_stats64 = ip_tunnel_get_stats64,
2397 .ndo_set_rx_mode = vxlan_set_multicast_list,
2398 .ndo_change_mtu = vxlan_change_mtu,
2399 .ndo_validate_addr = eth_validate_addr,
2400 .ndo_set_mac_address = eth_mac_addr,
2401 .ndo_fdb_add = vxlan_fdb_add,
2402 .ndo_fdb_del = vxlan_fdb_delete,
2403 .ndo_fdb_dump = vxlan_fdb_dump,
2404 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2407 /* Info for udev, that this is a virtual tunnel endpoint */
2408 static struct device_type vxlan_type = {
2412 /* Calls the ndo_add_vxlan_port of the caller in order to
2413 * supply the listening VXLAN udp ports. Callers are expected
2414 * to implement the ndo_add_vxlan_port.
2416 void vxlan_get_rx_port(struct net_device *dev)
2418 struct vxlan_sock *vs;
2419 struct net *net = dev_net(dev);
2420 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2421 sa_family_t sa_family;
2425 spin_lock(&vn->sock_lock);
2426 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2427 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2428 port = inet_sk(vs->sock->sk)->inet_sport;
2429 sa_family = vxlan_get_sk_family(vs);
2430 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2434 spin_unlock(&vn->sock_lock);
2436 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2438 /* Initialize the device structure. */
2439 static void vxlan_setup(struct net_device *dev)
2441 struct vxlan_dev *vxlan = netdev_priv(dev);
2444 eth_hw_addr_random(dev);
2447 dev->netdev_ops = &vxlan_netdev_ops;
2448 dev->destructor = free_netdev;
2449 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2451 dev->features |= NETIF_F_LLTX;
2452 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2453 dev->features |= NETIF_F_RXCSUM;
2454 dev->features |= NETIF_F_GSO_SOFTWARE;
2456 dev->vlan_features = dev->features;
2457 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2458 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2459 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2460 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2461 netif_keep_dst(dev);
2462 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
2464 INIT_LIST_HEAD(&vxlan->next);
2465 spin_lock_init(&vxlan->hash_lock);
2467 init_timer_deferrable(&vxlan->age_timer);
2468 vxlan->age_timer.function = vxlan_cleanup;
2469 vxlan->age_timer.data = (unsigned long) vxlan;
2471 vxlan->cfg.dst_port = htons(vxlan_port);
2475 gro_cells_init(&vxlan->gro_cells, dev);
2477 for (h = 0; h < FDB_HASH_SIZE; ++h)
2478 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2481 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2482 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2483 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2484 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2485 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2486 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2487 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2488 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2489 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2490 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2491 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2492 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2493 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2494 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2495 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2496 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2497 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2498 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2499 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2500 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2501 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2502 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2503 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2504 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2505 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2506 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2509 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2511 if (tb[IFLA_ADDRESS]) {
2512 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2513 pr_debug("invalid link address (not ethernet)\n");
2517 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2518 pr_debug("invalid all zero ethernet address\n");
2519 return -EADDRNOTAVAIL;
2526 if (data[IFLA_VXLAN_ID]) {
2527 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2528 if (id >= VXLAN_VID_MASK)
2532 if (data[IFLA_VXLAN_PORT_RANGE]) {
2533 const struct ifla_vxlan_port_range *p
2534 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2536 if (ntohs(p->high) < ntohs(p->low)) {
2537 pr_debug("port range %u .. %u not valid\n",
2538 ntohs(p->low), ntohs(p->high));
2546 static void vxlan_get_drvinfo(struct net_device *netdev,
2547 struct ethtool_drvinfo *drvinfo)
2549 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2550 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2553 static const struct ethtool_ops vxlan_ethtool_ops = {
2554 .get_drvinfo = vxlan_get_drvinfo,
2555 .get_link = ethtool_op_get_link,
2558 static void vxlan_del_work(struct work_struct *work)
2560 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2561 udp_tunnel_sock_release(vs->sock);
2565 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2566 __be16 port, u32 flags)
2568 struct socket *sock;
2569 struct udp_port_cfg udp_conf;
2572 memset(&udp_conf, 0, sizeof(udp_conf));
2575 udp_conf.family = AF_INET6;
2576 udp_conf.use_udp6_rx_checksums =
2577 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2578 udp_conf.ipv6_v6only = 1;
2580 udp_conf.family = AF_INET;
2583 udp_conf.local_udp_port = port;
2585 /* Open UDP socket */
2586 err = udp_sock_create(net, &udp_conf, &sock);
2588 return ERR_PTR(err);
2593 /* Create new listen socket if needed */
2594 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2595 __be16 port, u32 flags)
2597 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2598 struct vxlan_sock *vs;
2599 struct socket *sock;
2601 struct udp_tunnel_sock_cfg tunnel_cfg;
2603 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2605 return ERR_PTR(-ENOMEM);
2607 for (h = 0; h < VNI_HASH_SIZE; ++h)
2608 INIT_HLIST_HEAD(&vs->vni_list[h]);
2610 INIT_WORK(&vs->del_work, vxlan_del_work);
2612 sock = vxlan_create_sock(net, ipv6, port, flags);
2614 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2617 return ERR_CAST(sock);
2621 atomic_set(&vs->refcnt, 1);
2622 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2624 /* Initialize the vxlan udp offloads structure */
2625 vs->udp_offloads.port = port;
2626 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2627 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2629 spin_lock(&vn->sock_lock);
2630 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2631 vxlan_notify_add_rx_port(vs);
2632 spin_unlock(&vn->sock_lock);
2634 /* Mark socket as an encapsulation socket. */
2635 tunnel_cfg.sk_user_data = vs;
2636 tunnel_cfg.encap_type = 1;
2637 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2638 tunnel_cfg.encap_destroy = NULL;
2640 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2645 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2647 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2648 struct vxlan_sock *vs = NULL;
2650 if (!vxlan->cfg.no_share) {
2651 spin_lock(&vn->sock_lock);
2652 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2653 vxlan->cfg.dst_port, vxlan->flags);
2654 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
2655 spin_unlock(&vn->sock_lock);
2658 spin_unlock(&vn->sock_lock);
2661 vs = vxlan_socket_create(vxlan->net, ipv6,
2662 vxlan->cfg.dst_port, vxlan->flags);
2665 #if IS_ENABLED(CONFIG_IPV6)
2667 vxlan->vn6_sock = vs;
2670 vxlan->vn4_sock = vs;
2671 vxlan_vs_add_dev(vs, vxlan);
2675 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2677 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2678 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2681 vxlan->vn4_sock = NULL;
2682 #if IS_ENABLED(CONFIG_IPV6)
2683 vxlan->vn6_sock = NULL;
2684 if (ipv6 || metadata)
2685 ret = __vxlan_sock_add(vxlan, true);
2687 if (!ret && (!ipv6 || metadata))
2688 ret = __vxlan_sock_add(vxlan, false);
2690 vxlan_sock_release(vxlan);
2694 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2695 struct vxlan_config *conf)
2697 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2698 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
2699 struct vxlan_rdst *dst = &vxlan->default_dst;
2700 unsigned short needed_headroom = ETH_HLEN;
2702 bool use_ipv6 = false;
2703 __be16 default_port = vxlan->cfg.dst_port;
2705 vxlan->net = src_net;
2707 dst->remote_vni = conf->vni;
2709 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
2711 /* Unless IPv6 is explicitly requested, assume IPv4 */
2712 if (!dst->remote_ip.sa.sa_family)
2713 dst->remote_ip.sa.sa_family = AF_INET;
2715 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
2716 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2717 if (!IS_ENABLED(CONFIG_IPV6))
2718 return -EPFNOSUPPORT;
2720 vxlan->flags |= VXLAN_F_IPV6;
2723 if (conf->remote_ifindex) {
2724 struct net_device *lowerdev
2725 = __dev_get_by_index(src_net, conf->remote_ifindex);
2727 dst->remote_ifindex = conf->remote_ifindex;
2730 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2734 #if IS_ENABLED(CONFIG_IPV6)
2736 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2737 if (idev && idev->cnf.disable_ipv6) {
2738 pr_info("IPv6 is disabled via sysctl\n");
2745 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2747 needed_headroom = lowerdev->hard_header_len;
2750 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2751 needed_headroom += VXLAN6_HEADROOM;
2753 needed_headroom += VXLAN_HEADROOM;
2754 dev->needed_headroom = needed_headroom;
2756 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2757 if (!vxlan->cfg.dst_port)
2758 vxlan->cfg.dst_port = default_port;
2759 vxlan->flags |= conf->flags;
2761 if (!vxlan->cfg.age_interval)
2762 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2764 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2765 if (tmp->cfg.vni == conf->vni &&
2766 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2767 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2768 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2769 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2770 (vxlan->flags & VXLAN_F_RCV_FLAGS))
2774 dev->ethtool_ops = &vxlan_ethtool_ops;
2776 /* create an fdb entry for a valid default destination */
2777 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2778 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2779 &vxlan->default_dst.remote_ip,
2780 NUD_REACHABLE|NUD_PERMANENT,
2781 NLM_F_EXCL|NLM_F_CREATE,
2782 vxlan->cfg.dst_port,
2783 vxlan->default_dst.remote_vni,
2784 vxlan->default_dst.remote_ifindex,
2790 err = register_netdevice(dev);
2792 vxlan_fdb_delete_default(vxlan);
2796 list_add(&vxlan->next, &vn->vxlan_list);
2801 struct net_device *vxlan_dev_create(struct net *net, const char *name,
2802 u8 name_assign_type, struct vxlan_config *conf)
2804 struct nlattr *tb[IFLA_MAX+1];
2805 struct net_device *dev;
2808 memset(&tb, 0, sizeof(tb));
2810 dev = rtnl_create_link(net, name, name_assign_type,
2811 &vxlan_link_ops, tb);
2815 err = vxlan_dev_configure(net, dev, conf);
2818 return ERR_PTR(err);
2823 EXPORT_SYMBOL_GPL(vxlan_dev_create);
2825 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2826 struct nlattr *tb[], struct nlattr *data[])
2828 struct vxlan_config conf;
2831 memset(&conf, 0, sizeof(conf));
2833 if (data[IFLA_VXLAN_ID])
2834 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
2836 if (data[IFLA_VXLAN_GROUP]) {
2837 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2838 } else if (data[IFLA_VXLAN_GROUP6]) {
2839 if (!IS_ENABLED(CONFIG_IPV6))
2840 return -EPFNOSUPPORT;
2842 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2843 conf.remote_ip.sa.sa_family = AF_INET6;
2846 if (data[IFLA_VXLAN_LOCAL]) {
2847 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2848 conf.saddr.sa.sa_family = AF_INET;
2849 } else if (data[IFLA_VXLAN_LOCAL6]) {
2850 if (!IS_ENABLED(CONFIG_IPV6))
2851 return -EPFNOSUPPORT;
2853 /* TODO: respect scope id */
2854 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2855 conf.saddr.sa.sa_family = AF_INET6;
2858 if (data[IFLA_VXLAN_LINK])
2859 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2861 if (data[IFLA_VXLAN_TOS])
2862 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2864 if (data[IFLA_VXLAN_TTL])
2865 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2867 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2868 conf.flags |= VXLAN_F_LEARN;
2870 if (data[IFLA_VXLAN_AGEING])
2871 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2873 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2874 conf.flags |= VXLAN_F_PROXY;
2876 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2877 conf.flags |= VXLAN_F_RSC;
2879 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2880 conf.flags |= VXLAN_F_L2MISS;
2882 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2883 conf.flags |= VXLAN_F_L3MISS;
2885 if (data[IFLA_VXLAN_LIMIT])
2886 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2888 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2889 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2890 conf.flags |= VXLAN_F_COLLECT_METADATA;
2892 if (data[IFLA_VXLAN_PORT_RANGE]) {
2893 const struct ifla_vxlan_port_range *p
2894 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2895 conf.port_min = ntohs(p->low);
2896 conf.port_max = ntohs(p->high);
2899 if (data[IFLA_VXLAN_PORT])
2900 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2902 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2903 conf.flags |= VXLAN_F_UDP_CSUM;
2905 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2906 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2907 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2909 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2910 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2911 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2913 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2914 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2915 conf.flags |= VXLAN_F_REMCSUM_TX;
2917 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2918 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2919 conf.flags |= VXLAN_F_REMCSUM_RX;
2921 if (data[IFLA_VXLAN_GBP])
2922 conf.flags |= VXLAN_F_GBP;
2924 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
2925 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
2927 err = vxlan_dev_configure(src_net, dev, &conf);
2930 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2934 pr_info("IPv6 is disabled via sysctl\n");
2938 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
2945 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2947 struct vxlan_dev *vxlan = netdev_priv(dev);
2948 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2950 spin_lock(&vn->sock_lock);
2951 if (!hlist_unhashed(&vxlan->hlist))
2952 hlist_del_rcu(&vxlan->hlist);
2953 spin_unlock(&vn->sock_lock);
2955 gro_cells_destroy(&vxlan->gro_cells);
2956 list_del(&vxlan->next);
2957 unregister_netdevice_queue(dev, head);
2960 static size_t vxlan_get_size(const struct net_device *dev)
2963 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2964 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2965 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2966 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2967 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2968 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2969 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2970 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2971 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2972 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2973 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2974 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
2975 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2976 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2977 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2978 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2979 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2980 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2981 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2982 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2983 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
2987 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2989 const struct vxlan_dev *vxlan = netdev_priv(dev);
2990 const struct vxlan_rdst *dst = &vxlan->default_dst;
2991 struct ifla_vxlan_port_range ports = {
2992 .low = htons(vxlan->cfg.port_min),
2993 .high = htons(vxlan->cfg.port_max),
2996 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
2997 goto nla_put_failure;
2999 if (!vxlan_addr_any(&dst->remote_ip)) {
3000 if (dst->remote_ip.sa.sa_family == AF_INET) {
3001 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3002 dst->remote_ip.sin.sin_addr.s_addr))
3003 goto nla_put_failure;
3004 #if IS_ENABLED(CONFIG_IPV6)
3006 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3007 &dst->remote_ip.sin6.sin6_addr))
3008 goto nla_put_failure;
3013 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3014 goto nla_put_failure;
3016 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3017 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3018 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3019 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3020 goto nla_put_failure;
3021 #if IS_ENABLED(CONFIG_IPV6)
3023 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3024 &vxlan->cfg.saddr.sin6.sin6_addr))
3025 goto nla_put_failure;
3030 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3031 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3032 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3033 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3034 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3035 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3036 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3037 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3038 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3039 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3040 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
3041 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3042 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
3043 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3044 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3045 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3046 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3047 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3048 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3049 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3050 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3051 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3052 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3053 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3054 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3055 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
3056 goto nla_put_failure;
3058 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3059 goto nla_put_failure;
3061 if (vxlan->flags & VXLAN_F_GBP &&
3062 nla_put_flag(skb, IFLA_VXLAN_GBP))
3063 goto nla_put_failure;
3065 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3066 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3067 goto nla_put_failure;
3075 static struct net *vxlan_get_link_net(const struct net_device *dev)
3077 struct vxlan_dev *vxlan = netdev_priv(dev);
3082 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3084 .maxtype = IFLA_VXLAN_MAX,
3085 .policy = vxlan_policy,
3086 .priv_size = sizeof(struct vxlan_dev),
3087 .setup = vxlan_setup,
3088 .validate = vxlan_validate,
3089 .newlink = vxlan_newlink,
3090 .dellink = vxlan_dellink,
3091 .get_size = vxlan_get_size,
3092 .fill_info = vxlan_fill_info,
3093 .get_link_net = vxlan_get_link_net,
3096 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3097 struct net_device *dev)
3099 struct vxlan_dev *vxlan, *next;
3100 LIST_HEAD(list_kill);
3102 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3103 struct vxlan_rdst *dst = &vxlan->default_dst;
3105 /* In case we created vxlan device with carrier
3106 * and we loose the carrier due to module unload
3107 * we also need to remove vxlan device. In other
3108 * cases, it's not necessary and remote_ifindex
3109 * is 0 here, so no matches.
3111 if (dst->remote_ifindex == dev->ifindex)
3112 vxlan_dellink(vxlan->dev, &list_kill);
3115 unregister_netdevice_many(&list_kill);
3118 static int vxlan_lowerdev_event(struct notifier_block *unused,
3119 unsigned long event, void *ptr)
3121 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3122 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3124 if (event == NETDEV_UNREGISTER)
3125 vxlan_handle_lowerdev_unregister(vn, dev);
3130 static struct notifier_block vxlan_notifier_block __read_mostly = {
3131 .notifier_call = vxlan_lowerdev_event,
3134 static __net_init int vxlan_init_net(struct net *net)
3136 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3139 INIT_LIST_HEAD(&vn->vxlan_list);
3140 spin_lock_init(&vn->sock_lock);
3142 for (h = 0; h < PORT_HASH_SIZE; ++h)
3143 INIT_HLIST_HEAD(&vn->sock_list[h]);
3148 static void __net_exit vxlan_exit_net(struct net *net)
3150 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3151 struct vxlan_dev *vxlan, *next;
3152 struct net_device *dev, *aux;
3156 for_each_netdev_safe(net, dev, aux)
3157 if (dev->rtnl_link_ops == &vxlan_link_ops)
3158 unregister_netdevice_queue(dev, &list);
3160 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3161 /* If vxlan->dev is in the same netns, it has already been added
3162 * to the list by the previous loop.
3164 if (!net_eq(dev_net(vxlan->dev), net)) {
3165 gro_cells_destroy(&vxlan->gro_cells);
3166 unregister_netdevice_queue(vxlan->dev, &list);
3170 unregister_netdevice_many(&list);
3174 static struct pernet_operations vxlan_net_ops = {
3175 .init = vxlan_init_net,
3176 .exit = vxlan_exit_net,
3177 .id = &vxlan_net_id,
3178 .size = sizeof(struct vxlan_net),
3181 static int __init vxlan_init_module(void)
3185 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3189 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3191 rc = register_pernet_subsys(&vxlan_net_ops);
3195 rc = register_netdevice_notifier(&vxlan_notifier_block);
3199 rc = rtnl_link_register(&vxlan_link_ops);
3205 unregister_netdevice_notifier(&vxlan_notifier_block);
3207 unregister_pernet_subsys(&vxlan_net_ops);
3209 destroy_workqueue(vxlan_wq);
3212 late_initcall(vxlan_init_module);
3214 static void __exit vxlan_cleanup_module(void)
3216 rtnl_link_unregister(&vxlan_link_ops);
3217 unregister_netdevice_notifier(&vxlan_notifier_block);
3218 destroy_workqueue(vxlan_wq);
3219 unregister_pernet_subsys(&vxlan_net_ops);
3220 /* rcu_barrier() is called by netns */
3222 module_exit(vxlan_cleanup_module);
3224 MODULE_LICENSE("GPL");
3225 MODULE_VERSION(VXLAN_VERSION);
3226 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3227 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3228 MODULE_ALIAS_RTNL_LINK("vxlan");