1 // SPDX-License-Identifier: GPL-2.0-only
3 * VXLAN: Virtual eXtensible Local Area Network
5 * Copyright (c) 2012-2013 Vyatta Inc.
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
19 #include <net/ndisc.h>
21 #include <net/ipv6_stubs.h>
24 #include <net/rtnetlink.h>
25 #include <net/inet_ecn.h>
26 #include <net/net_namespace.h>
27 #include <net/netns/generic.h>
28 #include <net/tun_proto.h>
29 #include <net/vxlan.h>
30 #include <net/nexthop.h>
32 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_tunnel.h>
34 #include <net/ip6_checksum.h>
37 #include "vxlan_private.h"
39 #define VXLAN_VERSION "0.1"
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 unsigned int vxlan_net_id;
58 const u8 all_zeros_mac[ETH_ALEN + 2];
59 static struct rtnl_link_ops vxlan_link_ops;
61 static int vxlan_sock_add(struct vxlan_dev *vxlan);
63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
65 /* salt for hash table */
66 static u32 vxlan_salt __read_mostly;
68 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
70 return vs->flags & VXLAN_F_COLLECT_METADATA ||
71 ip_tunnel_collect_metadata();
74 /* Find VXLAN socket based on network namespace, address family, UDP port,
75 * enabled unshareable flags and socket device binding (see l3mdev with
78 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
79 __be16 port, u32 flags, int ifindex)
81 struct vxlan_sock *vs;
83 flags &= VXLAN_F_RCV_FLAGS;
85 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
86 if (inet_sk(vs->sock->sk)->inet_sport == port &&
87 vxlan_get_sk_family(vs) == family &&
89 vs->sock->sk->sk_bound_dev_if == ifindex)
95 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
96 int ifindex, __be32 vni,
97 struct vxlan_vni_node **vninode)
99 struct vxlan_vni_node *vnode;
100 struct vxlan_dev_node *node;
102 /* For flow based devices, map all packets to VNI 0 */
103 if (vs->flags & VXLAN_F_COLLECT_METADATA &&
104 !(vs->flags & VXLAN_F_VNIFILTER))
107 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
111 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
112 vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
115 } else if (node->vxlan->default_dst.remote_vni != vni) {
119 if (IS_ENABLED(CONFIG_IPV6)) {
120 const struct vxlan_config *cfg = &node->vxlan->cfg;
122 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
123 cfg->remote_ifindex != ifindex)
135 /* Look up VNI in a per net namespace table */
136 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
137 __be32 vni, sa_family_t family,
138 __be16 port, u32 flags)
140 struct vxlan_sock *vs;
142 vs = vxlan_find_sock(net, family, port, flags, ifindex);
146 return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
149 /* Fill in neighbour message in skbuff. */
150 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
151 const struct vxlan_fdb *fdb,
152 u32 portid, u32 seq, int type, unsigned int flags,
153 const struct vxlan_rdst *rdst)
155 unsigned long now = jiffies;
156 struct nda_cacheinfo ci;
157 bool send_ip, send_eth;
158 struct nlmsghdr *nlh;
164 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
168 ndm = nlmsg_data(nlh);
169 memset(ndm, 0, sizeof(*ndm));
171 send_eth = send_ip = true;
174 nh = rcu_dereference(fdb->nh);
176 nh_family = nexthop_get_family(nh);
181 if (type == RTM_GETNEIGH) {
183 send_ip = !vxlan_addr_any(&rdst->remote_ip);
184 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
186 ndm->ndm_family = nh_family;
188 send_eth = !is_zero_ether_addr(fdb->eth_addr);
190 ndm->ndm_family = AF_BRIDGE;
191 ndm->ndm_state = fdb->state;
192 ndm->ndm_ifindex = vxlan->dev->ifindex;
193 ndm->ndm_flags = fdb->flags;
194 if (rdst && rdst->offloaded)
195 ndm->ndm_flags |= NTF_OFFLOADED;
196 ndm->ndm_type = RTN_UNICAST;
198 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
199 nla_put_s32(skb, NDA_LINK_NETNSID,
200 peernet2id(dev_net(vxlan->dev), vxlan->net)))
201 goto nla_put_failure;
203 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
204 goto nla_put_failure;
206 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
207 goto nla_put_failure;
209 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
211 goto nla_put_failure;
213 if (rdst->remote_port &&
214 rdst->remote_port != vxlan->cfg.dst_port &&
215 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
216 goto nla_put_failure;
217 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
218 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
219 goto nla_put_failure;
220 if (rdst->remote_ifindex &&
221 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
222 goto nla_put_failure;
225 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
226 nla_put_u32(skb, NDA_SRC_VNI,
227 be32_to_cpu(fdb->vni)))
228 goto nla_put_failure;
230 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
231 ci.ndm_confirmed = 0;
232 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
235 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
236 goto nla_put_failure;
242 nlmsg_cancel(skb, nlh);
246 static inline size_t vxlan_nlmsg_size(void)
248 return NLMSG_ALIGN(sizeof(struct ndmsg))
249 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
250 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
251 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
252 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
253 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
254 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
255 + nla_total_size(sizeof(struct nda_cacheinfo));
258 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
259 struct vxlan_rdst *rd, int type)
261 struct net *net = dev_net(vxlan->dev);
265 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
269 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
271 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
272 WARN_ON(err == -EMSGSIZE);
277 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
281 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
284 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
285 const struct vxlan_fdb *fdb,
286 const struct vxlan_rdst *rd,
287 struct netlink_ext_ack *extack,
288 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
290 fdb_info->info.dev = vxlan->dev;
291 fdb_info->info.extack = extack;
292 fdb_info->remote_ip = rd->remote_ip;
293 fdb_info->remote_port = rd->remote_port;
294 fdb_info->remote_vni = rd->remote_vni;
295 fdb_info->remote_ifindex = rd->remote_ifindex;
296 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
297 fdb_info->vni = fdb->vni;
298 fdb_info->offloaded = rd->offloaded;
299 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
302 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
303 struct vxlan_fdb *fdb,
304 struct vxlan_rdst *rd,
306 struct netlink_ext_ack *extack)
308 struct switchdev_notifier_vxlan_fdb_info info;
309 enum switchdev_notifier_type notifier_type;
315 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
316 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
317 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
318 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
320 return notifier_to_errno(ret);
323 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
324 struct vxlan_rdst *rd, int type, bool swdev_notify,
325 struct netlink_ext_ack *extack)
329 if (swdev_notify && rd) {
332 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
338 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
344 __vxlan_fdb_notify(vxlan, fdb, rd, type);
348 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
350 struct vxlan_dev *vxlan = netdev_priv(dev);
351 struct vxlan_fdb f = {
354 struct vxlan_rdst remote = {
355 .remote_ip = *ipa, /* goes to NDA_DST */
356 .remote_vni = cpu_to_be32(VXLAN_N_VID),
359 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
362 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
364 struct vxlan_fdb f = {
367 struct vxlan_rdst remote = { };
369 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
371 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
374 /* Hash Ethernet address */
375 static u32 eth_hash(const unsigned char *addr)
377 u64 value = get_unaligned((u64 *)addr);
379 /* only want 6 bytes */
385 return hash_64(value, FDB_HASH_BITS);
388 u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
390 /* use 1 byte of OUI and 3 bytes of NIC */
391 u32 key = get_unaligned((u32 *)(addr + 2));
393 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
396 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
398 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
399 return eth_vni_hash(mac, vni);
401 return eth_hash(mac);
404 /* Hash chain to use given mac address */
405 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
406 const u8 *mac, __be32 vni)
408 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
411 /* Look up Ethernet address in forwarding table */
412 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
413 const u8 *mac, __be32 vni)
415 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
418 hlist_for_each_entry_rcu(f, head, hlist) {
419 if (ether_addr_equal(mac, f->eth_addr)) {
420 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
432 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
433 const u8 *mac, __be32 vni)
437 f = __vxlan_find_mac(vxlan, mac, vni);
438 if (f && f->used != jiffies)
444 /* caller should hold vxlan->hash_lock */
445 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
446 union vxlan_addr *ip, __be16 port,
447 __be32 vni, __u32 ifindex)
449 struct vxlan_rdst *rd;
451 list_for_each_entry(rd, &f->remotes, list) {
452 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
453 rd->remote_port == port &&
454 rd->remote_vni == vni &&
455 rd->remote_ifindex == ifindex)
462 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
463 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
465 struct vxlan_dev *vxlan = netdev_priv(dev);
466 u8 eth_addr[ETH_ALEN + 2] = { 0 };
467 struct vxlan_rdst *rdst;
471 if (is_multicast_ether_addr(mac) ||
472 is_zero_ether_addr(mac))
475 ether_addr_copy(eth_addr, mac);
479 f = __vxlan_find_mac(vxlan, eth_addr, vni);
485 rdst = first_remote_rcu(f);
486 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
492 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
494 static int vxlan_fdb_notify_one(struct notifier_block *nb,
495 const struct vxlan_dev *vxlan,
496 const struct vxlan_fdb *f,
497 const struct vxlan_rdst *rdst,
498 struct netlink_ext_ack *extack)
500 struct switchdev_notifier_vxlan_fdb_info fdb_info;
503 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
504 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
506 return notifier_to_errno(rc);
509 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
510 struct notifier_block *nb,
511 struct netlink_ext_ack *extack)
513 struct vxlan_dev *vxlan;
514 struct vxlan_rdst *rdst;
519 if (!netif_is_vxlan(dev))
521 vxlan = netdev_priv(dev);
523 for (h = 0; h < FDB_HASH_SIZE; ++h) {
524 spin_lock_bh(&vxlan->hash_lock[h]);
525 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
527 list_for_each_entry(rdst, &f->remotes, list) {
528 rc = vxlan_fdb_notify_one(nb, vxlan,
536 spin_unlock_bh(&vxlan->hash_lock[h]);
541 spin_unlock_bh(&vxlan->hash_lock[h]);
544 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
546 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
548 struct vxlan_dev *vxlan;
549 struct vxlan_rdst *rdst;
553 if (!netif_is_vxlan(dev))
555 vxlan = netdev_priv(dev);
557 for (h = 0; h < FDB_HASH_SIZE; ++h) {
558 spin_lock_bh(&vxlan->hash_lock[h]);
559 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
561 list_for_each_entry(rdst, &f->remotes, list)
562 rdst->offloaded = false;
563 spin_unlock_bh(&vxlan->hash_lock[h]);
567 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
569 /* Replace destination of unicast mac */
570 static int vxlan_fdb_replace(struct vxlan_fdb *f,
571 union vxlan_addr *ip, __be16 port, __be32 vni,
572 __u32 ifindex, struct vxlan_rdst *oldrd)
574 struct vxlan_rdst *rd;
576 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
580 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
585 dst_cache_reset(&rd->dst_cache);
587 rd->remote_port = port;
588 rd->remote_vni = vni;
589 rd->remote_ifindex = ifindex;
590 rd->offloaded = false;
594 /* Add/update destinations for multicast */
595 static int vxlan_fdb_append(struct vxlan_fdb *f,
596 union vxlan_addr *ip, __be16 port, __be32 vni,
597 __u32 ifindex, struct vxlan_rdst **rdp)
599 struct vxlan_rdst *rd;
601 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
605 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
609 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
615 rd->remote_port = port;
616 rd->offloaded = false;
617 rd->remote_vni = vni;
618 rd->remote_ifindex = ifindex;
620 list_add_tail_rcu(&rd->list, &f->remotes);
626 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol)
628 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr;
630 /* Need to have Next Protocol set for interfaces in GPE mode. */
631 if (!gpe->np_applied)
633 /* "The initial version is 0. If a receiver does not support the
634 * version indicated it MUST drop the packet.
636 if (gpe->version != 0)
638 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
639 * processing MUST occur." However, we don't implement OAM
640 * processing, thus drop the packet.
645 *protocol = tun_p_to_eth_p(gpe->next_protocol);
652 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
654 struct vxlanhdr *vh, size_t hdrlen,
656 struct gro_remcsum *grc,
659 size_t start, offset;
661 if (skb->remcsum_offload)
664 if (!NAPI_GRO_CB(skb)->csum_valid)
667 start = vxlan_rco_start(vni_field);
668 offset = start + vxlan_rco_offset(vni_field);
670 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
671 start, offset, grc, nopartial);
673 skb->remcsum_offload = 1;
678 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
679 struct list_head *head,
681 struct gro_remcsum *grc)
684 struct vxlanhdr *vh, *vh2;
685 unsigned int hlen, off_vx;
686 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
689 skb_gro_remcsum_init(grc);
691 off_vx = skb_gro_offset(skb);
692 hlen = off_vx + sizeof(*vh);
693 vh = skb_gro_header(skb, hlen, off_vx);
697 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
699 flags = vh->vx_flags;
701 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
702 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
705 VXLAN_F_REMCSUM_NOPARTIAL));
711 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
713 list_for_each_entry(p, head, list) {
714 if (!NAPI_GRO_CB(p)->same_flow)
717 vh2 = (struct vxlanhdr *)(p->data + off_vx);
718 if (vh->vx_flags != vh2->vx_flags ||
719 vh->vx_vni != vh2->vx_vni) {
720 NAPI_GRO_CB(p)->same_flow = 0;
728 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
729 struct list_head *head,
732 struct sk_buff *pp = NULL;
733 struct gro_remcsum grc;
736 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
737 pp = call_gro_receive(eth_gro_receive, head, skb);
740 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
744 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
745 struct list_head *head,
748 const struct packet_offload *ptype;
749 struct sk_buff *pp = NULL;
750 struct gro_remcsum grc;
755 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
757 if (!vxlan_parse_gpe_proto(vh, &protocol))
759 ptype = gro_find_receive_by_type(protocol);
762 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
766 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
770 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
772 /* Sets 'skb->inner_mac_header' since we are always called with
773 * 'skb->encapsulation' set.
775 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
778 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
780 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
781 const struct packet_offload *ptype;
785 if (!vxlan_parse_gpe_proto(vh, &protocol))
787 ptype = gro_find_complete_by_type(protocol);
789 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
793 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
794 __u16 state, __be32 src_vni,
799 f = kmalloc(sizeof(*f), GFP_ATOMIC);
803 f->flags = ndm_flags;
804 f->updated = f->used = jiffies;
807 RCU_INIT_POINTER(f->vdev, vxlan);
808 INIT_LIST_HEAD(&f->nh_list);
809 INIT_LIST_HEAD(&f->remotes);
810 memcpy(f->eth_addr, mac, ETH_ALEN);
815 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
816 __be32 src_vni, struct vxlan_fdb *f)
819 hlist_add_head_rcu(&f->hlist,
820 vxlan_fdb_head(vxlan, mac, src_vni));
823 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
824 u32 nhid, struct netlink_ext_ack *extack)
826 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
830 if (old_nh && old_nh->id == nhid)
833 nh = nexthop_find_by_id(vxlan->net, nhid);
835 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
839 if (!nexthop_get(nh)) {
840 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
844 if (!nexthop_is_fdb(nh)) {
845 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
849 if (!nexthop_is_multipath(nh)) {
850 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
854 /* check nexthop group family */
855 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
857 if (!nexthop_has_v4(nh)) {
859 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
864 if (nexthop_has_v4(nh)) {
866 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
872 list_del_rcu(&fdb->nh_list);
875 rcu_assign_pointer(fdb->nh, nh);
876 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
885 int vxlan_fdb_create(struct vxlan_dev *vxlan,
886 const u8 *mac, union vxlan_addr *ip,
887 __u16 state, __be16 port, __be32 src_vni,
888 __be32 vni, __u32 ifindex, __u16 ndm_flags,
889 u32 nhid, struct vxlan_fdb **fdb,
890 struct netlink_ext_ack *extack)
892 struct vxlan_rdst *rd = NULL;
896 if (vxlan->cfg.addrmax &&
897 vxlan->addrcnt >= vxlan->cfg.addrmax)
900 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
901 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
906 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
908 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
921 static void __vxlan_fdb_free(struct vxlan_fdb *f)
923 struct vxlan_rdst *rd, *nd;
926 nh = rcu_dereference_raw(f->nh);
928 rcu_assign_pointer(f->nh, NULL);
929 rcu_assign_pointer(f->vdev, NULL);
933 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
934 dst_cache_destroy(&rd->dst_cache);
940 static void vxlan_fdb_free(struct rcu_head *head)
942 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
947 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
948 bool do_notify, bool swdev_notify)
950 struct vxlan_rdst *rd;
952 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
956 if (rcu_access_pointer(f->nh))
957 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
960 list_for_each_entry(rd, &f->remotes, list)
961 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
965 hlist_del_rcu(&f->hlist);
966 list_del_rcu(&f->nh_list);
967 call_rcu(&f->rcu, vxlan_fdb_free);
970 static void vxlan_dst_free(struct rcu_head *head)
972 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
974 dst_cache_destroy(&rd->dst_cache);
978 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
979 union vxlan_addr *ip,
980 __u16 state, __u16 flags,
981 __be16 port, __be32 vni,
982 __u32 ifindex, __u16 ndm_flags,
983 struct vxlan_fdb *f, u32 nhid,
985 struct netlink_ext_ack *extack)
987 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
988 struct vxlan_rdst *rd = NULL;
989 struct vxlan_rdst oldrd;
994 if (nhid && !rcu_access_pointer(f->nh)) {
995 NL_SET_ERR_MSG(extack,
996 "Cannot replace an existing non nexthop fdb with a nexthop");
1000 if (nhid && (flags & NLM_F_APPEND)) {
1001 NL_SET_ERR_MSG(extack,
1002 "Cannot append to a nexthop fdb");
1006 /* Do not allow an externally learned entry to take over an entry added
1009 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1010 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1011 if (f->state != state) {
1013 f->updated = jiffies;
1016 if (f->flags != fdb_flags) {
1017 f->flags = fdb_flags;
1018 f->updated = jiffies;
1023 if ((flags & NLM_F_REPLACE)) {
1024 /* Only change unicasts */
1025 if (!(is_multicast_ether_addr(f->eth_addr) ||
1026 is_zero_ether_addr(f->eth_addr))) {
1028 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1032 rc = vxlan_fdb_replace(f, ip, port, vni,
1037 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1041 if ((flags & NLM_F_APPEND) &&
1042 (is_multicast_ether_addr(f->eth_addr) ||
1043 is_zero_ether_addr(f->eth_addr))) {
1044 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1051 if (ndm_flags & NTF_USE)
1056 rd = first_remote_rtnl(f);
1058 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1059 swdev_notify, extack);
1069 if ((flags & NLM_F_REPLACE) && rc)
1071 else if ((flags & NLM_F_APPEND) && rc) {
1072 list_del_rcu(&rd->list);
1073 call_rcu(&rd->rcu, vxlan_dst_free);
1078 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1079 const u8 *mac, union vxlan_addr *ip,
1080 __u16 state, __u16 flags,
1081 __be16 port, __be32 src_vni, __be32 vni,
1082 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1084 struct netlink_ext_ack *extack)
1086 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1087 struct vxlan_fdb *f;
1090 /* Disallow replace to add a multicast entry */
1091 if ((flags & NLM_F_REPLACE) &&
1092 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1095 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1096 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1097 vni, ifindex, fdb_flags, nhid, &f, extack);
1101 vxlan_fdb_insert(vxlan, mac, src_vni, f);
1102 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1103 swdev_notify, extack);
1110 vxlan_fdb_destroy(vxlan, f, false, false);
1114 /* Add new entry to forwarding table -- assumes lock held */
1115 int vxlan_fdb_update(struct vxlan_dev *vxlan,
1116 const u8 *mac, union vxlan_addr *ip,
1117 __u16 state, __u16 flags,
1118 __be16 port, __be32 src_vni, __be32 vni,
1119 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1121 struct netlink_ext_ack *extack)
1123 struct vxlan_fdb *f;
1125 f = __vxlan_find_mac(vxlan, mac, src_vni);
1127 if (flags & NLM_F_EXCL) {
1128 netdev_dbg(vxlan->dev,
1129 "lost race to create %pM\n", mac);
1133 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1134 vni, ifindex, ndm_flags, f,
1135 nhid, swdev_notify, extack);
1137 if (!(flags & NLM_F_CREATE))
1140 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1141 port, src_vni, vni, ifindex,
1142 ndm_flags, nhid, swdev_notify,
1147 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1148 struct vxlan_rdst *rd, bool swdev_notify)
1150 list_del_rcu(&rd->list);
1151 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1152 call_rcu(&rd->rcu, vxlan_dst_free);
1155 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1156 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1157 __be32 *vni, u32 *ifindex, u32 *nhid,
1158 struct netlink_ext_ack *extack)
1160 struct net *net = dev_net(vxlan->dev);
1163 if (tb[NDA_NH_ID] &&
1164 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1165 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1170 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1172 NL_SET_ERR_MSG(extack, "Unsupported address family");
1176 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1178 if (remote->sa.sa_family == AF_INET) {
1179 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1180 ip->sa.sa_family = AF_INET;
1181 #if IS_ENABLED(CONFIG_IPV6)
1183 ip->sin6.sin6_addr = in6addr_any;
1184 ip->sa.sa_family = AF_INET6;
1190 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1191 NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1194 *port = nla_get_be16(tb[NDA_PORT]);
1196 *port = vxlan->cfg.dst_port;
1200 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1201 NL_SET_ERR_MSG(extack, "Invalid vni");
1204 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1206 *vni = vxlan->default_dst.remote_vni;
1209 if (tb[NDA_SRC_VNI]) {
1210 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1211 NL_SET_ERR_MSG(extack, "Invalid src vni");
1214 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1216 *src_vni = vxlan->default_dst.remote_vni;
1219 if (tb[NDA_IFINDEX]) {
1220 struct net_device *tdev;
1222 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1223 NL_SET_ERR_MSG(extack, "Invalid ifindex");
1226 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1227 tdev = __dev_get_by_index(net, *ifindex);
1229 NL_SET_ERR_MSG(extack, "Device not found");
1230 return -EADDRNOTAVAIL;
1237 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1244 /* Add static entry (via netlink) */
1245 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1246 struct net_device *dev,
1247 const unsigned char *addr, u16 vid, u16 flags,
1248 struct netlink_ext_ack *extack)
1250 struct vxlan_dev *vxlan = netdev_priv(dev);
1251 /* struct net *net = dev_net(vxlan->dev); */
1252 union vxlan_addr ip;
1254 __be32 src_vni, vni;
1259 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1260 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1265 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1268 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1273 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1274 return -EAFNOSUPPORT;
1276 hash_index = fdb_head_index(vxlan, addr, src_vni);
1277 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1278 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1279 port, src_vni, vni, ifindex,
1280 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1281 nhid, true, extack);
1282 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1287 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1288 const unsigned char *addr, union vxlan_addr ip,
1289 __be16 port, __be32 src_vni, __be32 vni,
1290 u32 ifindex, bool swdev_notify)
1292 struct vxlan_rdst *rd = NULL;
1293 struct vxlan_fdb *f;
1296 f = vxlan_find_mac(vxlan, addr, src_vni);
1300 if (!vxlan_addr_any(&ip)) {
1301 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1306 /* remove a destination if it's not the only one on the list,
1307 * otherwise destroy the fdb entry
1309 if (rd && !list_is_singular(&f->remotes)) {
1310 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1314 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1320 /* Delete entry (via netlink) */
1321 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1322 struct net_device *dev,
1323 const unsigned char *addr, u16 vid,
1324 struct netlink_ext_ack *extack)
1326 struct vxlan_dev *vxlan = netdev_priv(dev);
1327 union vxlan_addr ip;
1328 __be32 src_vni, vni;
1334 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1339 hash_index = fdb_head_index(vxlan, addr, src_vni);
1340 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1341 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1343 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1348 /* Dump forwarding table */
1349 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1350 struct net_device *dev,
1351 struct net_device *filter_dev, int *idx)
1353 struct vxlan_dev *vxlan = netdev_priv(dev);
1357 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1358 struct vxlan_fdb *f;
1361 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1362 struct vxlan_rdst *rd;
1364 if (rcu_access_pointer(f->nh)) {
1365 if (*idx < cb->args[2])
1367 err = vxlan_fdb_info(skb, vxlan, f,
1368 NETLINK_CB(cb->skb).portid,
1381 list_for_each_entry_rcu(rd, &f->remotes, list) {
1382 if (*idx < cb->args[2])
1385 err = vxlan_fdb_info(skb, vxlan, f,
1386 NETLINK_CB(cb->skb).portid,
1404 static int vxlan_fdb_get(struct sk_buff *skb,
1405 struct nlattr *tb[],
1406 struct net_device *dev,
1407 const unsigned char *addr,
1408 u16 vid, u32 portid, u32 seq,
1409 struct netlink_ext_ack *extack)
1411 struct vxlan_dev *vxlan = netdev_priv(dev);
1412 struct vxlan_fdb *f;
1417 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1419 vni = vxlan->default_dst.remote_vni;
1423 f = __vxlan_find_mac(vxlan, addr, vni);
1425 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1430 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1431 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1437 /* Watch incoming packets to learn mapping between Ethernet address
1438 * and Tunnel endpoint.
1439 * Return true if packet is bogus and should be dropped.
1441 static bool vxlan_snoop(struct net_device *dev,
1442 union vxlan_addr *src_ip, const u8 *src_mac,
1443 u32 src_ifindex, __be32 vni)
1445 struct vxlan_dev *vxlan = netdev_priv(dev);
1446 struct vxlan_fdb *f;
1449 #if IS_ENABLED(CONFIG_IPV6)
1450 if (src_ip->sa.sa_family == AF_INET6 &&
1451 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1452 ifindex = src_ifindex;
1455 f = vxlan_find_mac(vxlan, src_mac, vni);
1457 struct vxlan_rdst *rdst = first_remote_rcu(f);
1459 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1460 rdst->remote_ifindex == ifindex))
1463 /* Don't migrate static entries, drop packets */
1464 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1467 /* Don't override an fdb with nexthop with a learnt entry */
1468 if (rcu_access_pointer(f->nh))
1471 if (net_ratelimit())
1473 "%pM migrated from %pIS to %pIS\n",
1474 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1476 rdst->remote_ip = *src_ip;
1477 f->updated = jiffies;
1478 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1480 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1482 /* learned new entry */
1483 spin_lock(&vxlan->hash_lock[hash_index]);
1485 /* close off race between vxlan_flush and incoming packets */
1486 if (netif_running(dev))
1487 vxlan_fdb_update(vxlan, src_mac, src_ip,
1489 NLM_F_EXCL|NLM_F_CREATE,
1490 vxlan->cfg.dst_port,
1492 vxlan->default_dst.remote_vni,
1493 ifindex, NTF_SELF, 0, true, NULL);
1494 spin_unlock(&vxlan->hash_lock[hash_index]);
1500 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1502 struct vxlan_net *vn;
1506 if (!refcount_dec_and_test(&vs->refcnt))
1509 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1510 spin_lock(&vn->sock_lock);
1511 hlist_del_rcu(&vs->hlist);
1512 udp_tunnel_notify_del_rx_port(vs->sock,
1513 (vs->flags & VXLAN_F_GPE) ?
1514 UDP_TUNNEL_TYPE_VXLAN_GPE :
1515 UDP_TUNNEL_TYPE_VXLAN);
1516 spin_unlock(&vn->sock_lock);
1521 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1523 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1524 #if IS_ENABLED(CONFIG_IPV6)
1525 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1527 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1530 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1533 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1534 vxlan_vs_del_vnigrp(vxlan);
1536 vxlan_vs_del_dev(vxlan);
1538 if (__vxlan_sock_release_prep(sock4)) {
1539 udp_tunnel_sock_release(sock4->sock);
1543 #if IS_ENABLED(CONFIG_IPV6)
1544 if (__vxlan_sock_release_prep(sock6)) {
1545 udp_tunnel_sock_release(sock6->sock);
1551 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1552 struct sk_buff *skb, u32 vxflags)
1554 size_t start, offset;
1556 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1559 start = vxlan_rco_start(unparsed->vx_vni);
1560 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1562 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1565 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1566 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1568 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1569 unparsed->vx_vni &= VXLAN_VNI_MASK;
1573 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1574 struct sk_buff *skb, u32 vxflags,
1575 struct vxlan_metadata *md)
1577 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1578 struct metadata_dst *tun_dst;
1580 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1583 md->gbp = ntohs(gbp->policy_id);
1585 tun_dst = (struct metadata_dst *)skb_dst(skb);
1587 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1588 tun_dst->u.tun_info.options_len = sizeof(*md);
1590 if (gbp->dont_learn)
1591 md->gbp |= VXLAN_GBP_DONT_LEARN;
1593 if (gbp->policy_applied)
1594 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1596 /* In flow-based mode, GBP is carried in dst_metadata */
1597 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1598 skb->mark = md->gbp;
1600 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1603 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1604 struct vxlan_sock *vs,
1605 struct sk_buff *skb, __be32 vni)
1607 union vxlan_addr saddr;
1608 u32 ifindex = skb->dev->ifindex;
1610 skb_reset_mac_header(skb);
1611 skb->protocol = eth_type_trans(skb, vxlan->dev);
1612 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1614 /* Ignore packet loops (and multicast echo) */
1615 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1618 /* Get address from the outer IP header */
1619 if (vxlan_get_sk_family(vs) == AF_INET) {
1620 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1621 saddr.sa.sa_family = AF_INET;
1622 #if IS_ENABLED(CONFIG_IPV6)
1624 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1625 saddr.sa.sa_family = AF_INET6;
1629 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1630 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1636 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1637 struct sk_buff *skb)
1641 if (vxlan_get_sk_family(vs) == AF_INET)
1642 err = IP_ECN_decapsulate(oiph, skb);
1643 #if IS_ENABLED(CONFIG_IPV6)
1645 err = IP6_ECN_decapsulate(oiph, skb);
1648 if (unlikely(err) && log_ecn_error) {
1649 if (vxlan_get_sk_family(vs) == AF_INET)
1650 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1651 &((struct iphdr *)oiph)->saddr,
1652 ((struct iphdr *)oiph)->tos);
1654 net_info_ratelimited("non-ECT from %pI6\n",
1655 &((struct ipv6hdr *)oiph)->saddr);
1660 /* Callback from net/ipv4/udp.c to receive packets */
1661 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1663 struct vxlan_vni_node *vninode = NULL;
1664 struct vxlan_dev *vxlan;
1665 struct vxlan_sock *vs;
1666 struct vxlanhdr unparsed;
1667 struct vxlan_metadata _md;
1668 struct vxlan_metadata *md = &_md;
1669 __be16 protocol = htons(ETH_P_TEB);
1670 bool raw_proto = false;
1674 /* Need UDP and VXLAN header to be present */
1675 if (!pskb_may_pull(skb, VXLAN_HLEN))
1678 unparsed = *vxlan_hdr(skb);
1679 /* VNI flag always required to be set */
1680 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1681 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1682 ntohl(vxlan_hdr(skb)->vx_flags),
1683 ntohl(vxlan_hdr(skb)->vx_vni));
1684 /* Return non vxlan pkt */
1687 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1688 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1690 vs = rcu_dereference_sk_user_data(sk);
1694 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1696 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1700 /* For backwards compatibility, only allow reserved fields to be
1701 * used by VXLAN extensions if explicitly requested.
1703 if (vs->flags & VXLAN_F_GPE) {
1704 if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1706 unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1710 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1711 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1714 if (vs->flags & VXLAN_F_REMCSUM_RX)
1715 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1718 if (vxlan_collect_metadata(vs)) {
1719 struct metadata_dst *tun_dst;
1721 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1722 key32_to_tunnel_id(vni), sizeof(*md));
1727 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1729 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1731 memset(md, 0, sizeof(*md));
1734 if (vs->flags & VXLAN_F_GBP)
1735 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1736 /* Note that GBP and GPE can never be active together. This is
1737 * ensured in vxlan_dev_configure.
1740 if (unparsed.vx_flags || unparsed.vx_vni) {
1741 /* If there are any unprocessed flags remaining treat
1742 * this as a malformed packet. This behavior diverges from
1743 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1744 * in reserved fields are to be ignored. The approach here
1745 * maintains compatibility with previous stack code, and also
1746 * is more robust and provides a little more security in
1747 * adding extensions to VXLAN.
1753 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1756 skb_reset_mac_header(skb);
1757 skb->dev = vxlan->dev;
1758 skb->pkt_type = PACKET_HOST;
1761 oiph = skb_network_header(skb);
1762 skb_reset_network_header(skb);
1764 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1765 ++vxlan->dev->stats.rx_frame_errors;
1766 ++vxlan->dev->stats.rx_errors;
1767 vxlan_vnifilter_count(vxlan, vni, vninode,
1768 VXLAN_VNI_STATS_RX_ERRORS, 0);
1774 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1776 dev_core_stats_rx_dropped_inc(vxlan->dev);
1777 vxlan_vnifilter_count(vxlan, vni, vninode,
1778 VXLAN_VNI_STATS_RX_DROPS, 0);
1782 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1783 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1784 gro_cells_receive(&vxlan->gro_cells, skb);
1791 /* Consume bad packet */
1796 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1797 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1799 struct vxlan_dev *vxlan;
1800 struct vxlan_sock *vs;
1801 struct vxlanhdr *hdr;
1804 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1807 hdr = vxlan_hdr(skb);
1809 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1812 vs = rcu_dereference_sk_user_data(sk);
1816 vni = vxlan_vni(hdr->vx_vni);
1817 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1824 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1826 struct vxlan_dev *vxlan = netdev_priv(dev);
1827 struct arphdr *parp;
1830 struct neighbour *n;
1832 if (dev->flags & IFF_NOARP)
1835 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1836 dev->stats.tx_dropped++;
1839 parp = arp_hdr(skb);
1841 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1842 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1843 parp->ar_pro != htons(ETH_P_IP) ||
1844 parp->ar_op != htons(ARPOP_REQUEST) ||
1845 parp->ar_hln != dev->addr_len ||
1848 arpptr = (u8 *)parp + sizeof(struct arphdr);
1850 arpptr += dev->addr_len; /* sha */
1851 memcpy(&sip, arpptr, sizeof(sip));
1852 arpptr += sizeof(sip);
1853 arpptr += dev->addr_len; /* tha */
1854 memcpy(&tip, arpptr, sizeof(tip));
1856 if (ipv4_is_loopback(tip) ||
1857 ipv4_is_multicast(tip))
1860 n = neigh_lookup(&arp_tbl, &tip, dev);
1863 struct vxlan_fdb *f;
1864 struct sk_buff *reply;
1866 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1871 f = vxlan_find_mac(vxlan, n->ha, vni);
1872 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1873 /* bridge-local neighbor */
1878 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1886 skb_reset_mac_header(reply);
1887 __skb_pull(reply, skb_network_offset(reply));
1888 reply->ip_summed = CHECKSUM_UNNECESSARY;
1889 reply->pkt_type = PACKET_HOST;
1891 if (netif_rx(reply) == NET_RX_DROP) {
1892 dev->stats.rx_dropped++;
1893 vxlan_vnifilter_count(vxlan, vni, NULL,
1894 VXLAN_VNI_STATS_RX_DROPS, 0);
1897 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1898 union vxlan_addr ipa = {
1899 .sin.sin_addr.s_addr = tip,
1900 .sin.sin_family = AF_INET,
1903 vxlan_ip_miss(dev, &ipa);
1907 return NETDEV_TX_OK;
1910 #if IS_ENABLED(CONFIG_IPV6)
1911 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1912 struct neighbour *n, bool isrouter)
1914 struct net_device *dev = request->dev;
1915 struct sk_buff *reply;
1916 struct nd_msg *ns, *na;
1917 struct ipv6hdr *pip6;
1919 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1923 if (dev == NULL || !pskb_may_pull(request, request->len))
1926 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1927 sizeof(*na) + na_olen + dev->needed_tailroom;
1928 reply = alloc_skb(len, GFP_ATOMIC);
1932 reply->protocol = htons(ETH_P_IPV6);
1934 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1935 skb_push(reply, sizeof(struct ethhdr));
1936 skb_reset_mac_header(reply);
1938 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1940 daddr = eth_hdr(request)->h_source;
1941 ns_olen = request->len - skb_network_offset(request) -
1942 sizeof(struct ipv6hdr) - sizeof(*ns);
1943 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1944 if (!ns->opt[i + 1]) {
1948 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1949 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1954 /* Ethernet header */
1955 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1956 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1957 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1958 reply->protocol = htons(ETH_P_IPV6);
1960 skb_pull(reply, sizeof(struct ethhdr));
1961 skb_reset_network_header(reply);
1962 skb_put(reply, sizeof(struct ipv6hdr));
1966 pip6 = ipv6_hdr(reply);
1967 memset(pip6, 0, sizeof(struct ipv6hdr));
1969 pip6->priority = ipv6_hdr(request)->priority;
1970 pip6->nexthdr = IPPROTO_ICMPV6;
1971 pip6->hop_limit = 255;
1972 pip6->daddr = ipv6_hdr(request)->saddr;
1973 pip6->saddr = *(struct in6_addr *)n->primary_key;
1975 skb_pull(reply, sizeof(struct ipv6hdr));
1976 skb_reset_transport_header(reply);
1978 /* Neighbor Advertisement */
1979 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1980 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1981 na->icmph.icmp6_router = isrouter;
1982 na->icmph.icmp6_override = 1;
1983 na->icmph.icmp6_solicited = 1;
1984 na->target = ns->target;
1985 ether_addr_copy(&na->opt[2], n->ha);
1986 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1987 na->opt[1] = na_olen >> 3;
1989 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1990 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1991 csum_partial(na, sizeof(*na)+na_olen, 0));
1993 pip6->payload_len = htons(sizeof(*na)+na_olen);
1995 skb_push(reply, sizeof(struct ipv6hdr));
1997 reply->ip_summed = CHECKSUM_UNNECESSARY;
2002 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2004 struct vxlan_dev *vxlan = netdev_priv(dev);
2005 const struct in6_addr *daddr;
2006 const struct ipv6hdr *iphdr;
2007 struct inet6_dev *in6_dev;
2008 struct neighbour *n;
2012 in6_dev = __in6_dev_get(dev);
2016 iphdr = ipv6_hdr(skb);
2017 daddr = &iphdr->daddr;
2018 msg = (struct nd_msg *)(iphdr + 1);
2020 if (ipv6_addr_loopback(daddr) ||
2021 ipv6_addr_is_multicast(&msg->target))
2024 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2027 struct vxlan_fdb *f;
2028 struct sk_buff *reply;
2030 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2035 f = vxlan_find_mac(vxlan, n->ha, vni);
2036 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2037 /* bridge-local neighbor */
2042 reply = vxlan_na_create(skb, n,
2043 !!(f ? f->flags & NTF_ROUTER : 0));
2050 if (netif_rx(reply) == NET_RX_DROP) {
2051 dev->stats.rx_dropped++;
2052 vxlan_vnifilter_count(vxlan, vni, NULL,
2053 VXLAN_VNI_STATS_RX_DROPS, 0);
2055 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2056 union vxlan_addr ipa = {
2057 .sin6.sin6_addr = msg->target,
2058 .sin6.sin6_family = AF_INET6,
2061 vxlan_ip_miss(dev, &ipa);
2067 return NETDEV_TX_OK;
2071 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2073 struct vxlan_dev *vxlan = netdev_priv(dev);
2074 struct neighbour *n;
2076 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2080 switch (ntohs(eth_hdr(skb)->h_proto)) {
2085 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2088 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2089 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2090 union vxlan_addr ipa = {
2091 .sin.sin_addr.s_addr = pip->daddr,
2092 .sin.sin_family = AF_INET,
2095 vxlan_ip_miss(dev, &ipa);
2101 #if IS_ENABLED(CONFIG_IPV6)
2104 struct ipv6hdr *pip6;
2106 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2108 pip6 = ipv6_hdr(skb);
2109 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2110 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2111 union vxlan_addr ipa = {
2112 .sin6.sin6_addr = pip6->daddr,
2113 .sin6.sin6_family = AF_INET6,
2116 vxlan_ip_miss(dev, &ipa);
2130 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2132 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2134 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2143 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2145 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2147 gpe->np_applied = 1;
2148 gpe->next_protocol = tun_p_from_eth_p(protocol);
2149 if (!gpe->next_protocol)
2150 return -EPFNOSUPPORT;
2154 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2155 int iphdr_len, __be32 vni,
2156 struct vxlan_metadata *md, u32 vxflags,
2159 struct vxlanhdr *vxh;
2162 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2163 __be16 inner_protocol = htons(ETH_P_TEB);
2165 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2166 skb->ip_summed == CHECKSUM_PARTIAL) {
2167 int csum_start = skb_checksum_start_offset(skb);
2169 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2170 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2171 (skb->csum_offset == offsetof(struct udphdr, check) ||
2172 skb->csum_offset == offsetof(struct tcphdr, check)))
2173 type |= SKB_GSO_TUNNEL_REMCSUM;
2176 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2177 + VXLAN_HLEN + iphdr_len;
2179 /* Need space for new headers (invalidates iph ptr) */
2180 err = skb_cow_head(skb, min_headroom);
2184 err = iptunnel_handle_offloads(skb, type);
2188 vxh = __skb_push(skb, sizeof(*vxh));
2189 vxh->vx_flags = VXLAN_HF_VNI;
2190 vxh->vx_vni = vxlan_vni_field(vni);
2192 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2195 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2196 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2197 vxh->vx_flags |= VXLAN_HF_RCO;
2199 if (!skb_is_gso(skb)) {
2200 skb->ip_summed = CHECKSUM_NONE;
2201 skb->encapsulation = 0;
2205 if (vxflags & VXLAN_F_GBP)
2206 vxlan_build_gbp_hdr(vxh, md);
2207 if (vxflags & VXLAN_F_GPE) {
2208 err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2211 inner_protocol = skb->protocol;
2214 skb_set_inner_protocol(skb, inner_protocol);
2218 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2219 struct vxlan_sock *sock4,
2220 struct sk_buff *skb, int oif, u8 tos,
2221 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2222 __u8 flow_flags, struct dst_cache *dst_cache,
2223 const struct ip_tunnel_info *info)
2225 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2226 struct rtable *rt = NULL;
2230 return ERR_PTR(-EIO);
2235 rt = dst_cache_get_ip4(dst_cache, saddr);
2240 memset(&fl4, 0, sizeof(fl4));
2241 fl4.flowi4_oif = oif;
2242 fl4.flowi4_tos = RT_TOS(tos);
2243 fl4.flowi4_mark = skb->mark;
2244 fl4.flowi4_proto = IPPROTO_UDP;
2247 fl4.fl4_dport = dport;
2248 fl4.fl4_sport = sport;
2249 fl4.flowi4_flags = flow_flags;
2251 rt = ip_route_output_key(vxlan->net, &fl4);
2253 if (rt->dst.dev == dev) {
2254 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2256 return ERR_PTR(-ELOOP);
2261 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2263 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2264 return ERR_PTR(-ENETUNREACH);
2269 #if IS_ENABLED(CONFIG_IPV6)
2270 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2271 struct net_device *dev,
2272 struct vxlan_sock *sock6,
2273 struct sk_buff *skb, int oif, u8 tos,
2275 const struct in6_addr *daddr,
2276 struct in6_addr *saddr,
2277 __be16 dport, __be16 sport,
2278 struct dst_cache *dst_cache,
2279 const struct ip_tunnel_info *info)
2281 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2282 struct dst_entry *ndst;
2286 return ERR_PTR(-EIO);
2291 ndst = dst_cache_get_ip6(dst_cache, saddr);
2296 memset(&fl6, 0, sizeof(fl6));
2297 fl6.flowi6_oif = oif;
2300 fl6.flowlabel = ip6_make_flowinfo(tos, label);
2301 fl6.flowi6_mark = skb->mark;
2302 fl6.flowi6_proto = IPPROTO_UDP;
2303 fl6.fl6_dport = dport;
2304 fl6.fl6_sport = sport;
2306 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2309 netdev_dbg(dev, "no route to %pI6\n", daddr);
2310 return ERR_PTR(-ENETUNREACH);
2313 if (unlikely(ndst->dev == dev)) {
2314 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2316 return ERR_PTR(-ELOOP);
2321 dst_cache_set_ip6(dst_cache, ndst, saddr);
2326 /* Bypass encapsulation if the destination is local */
2327 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2328 struct vxlan_dev *dst_vxlan, __be32 vni,
2331 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2332 union vxlan_addr loopback;
2333 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2334 struct net_device *dev;
2337 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2338 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2339 skb->pkt_type = PACKET_HOST;
2340 skb->encapsulation = 0;
2341 skb->dev = dst_vxlan->dev;
2342 __skb_pull(skb, skb_network_offset(skb));
2344 if (remote_ip->sa.sa_family == AF_INET) {
2345 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2346 loopback.sa.sa_family = AF_INET;
2347 #if IS_ENABLED(CONFIG_IPV6)
2349 loopback.sin6.sin6_addr = in6addr_loopback;
2350 loopback.sa.sa_family = AF_INET6;
2356 if (unlikely(!(dev->flags & IFF_UP))) {
2361 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2362 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2364 u64_stats_update_begin(&tx_stats->syncp);
2365 u64_stats_inc(&tx_stats->tx_packets);
2366 u64_stats_add(&tx_stats->tx_bytes, len);
2367 u64_stats_update_end(&tx_stats->syncp);
2368 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2370 if (__netif_rx(skb) == NET_RX_SUCCESS) {
2371 u64_stats_update_begin(&rx_stats->syncp);
2372 u64_stats_inc(&rx_stats->rx_packets);
2373 u64_stats_add(&rx_stats->rx_bytes, len);
2374 u64_stats_update_end(&rx_stats->syncp);
2375 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2379 dev->stats.rx_dropped++;
2380 vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2381 VXLAN_VNI_STATS_RX_DROPS, 0);
2386 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2387 struct vxlan_dev *vxlan,
2388 union vxlan_addr *daddr,
2389 __be16 dst_port, int dst_ifindex, __be32 vni,
2390 struct dst_entry *dst,
2393 #if IS_ENABLED(CONFIG_IPV6)
2394 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2395 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2396 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2398 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2400 /* Bypass encapsulation if the destination is local */
2401 if (rt_flags & RTCF_LOCAL &&
2402 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2403 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2404 struct vxlan_dev *dst_vxlan;
2407 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2408 daddr->sa.sa_family, dst_port,
2411 dev->stats.tx_errors++;
2412 vxlan_vnifilter_count(vxlan, vni, NULL,
2413 VXLAN_VNI_STATS_TX_ERRORS, 0);
2418 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2425 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2426 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2428 struct dst_cache *dst_cache;
2429 struct ip_tunnel_info *info;
2430 struct vxlan_dev *vxlan = netdev_priv(dev);
2431 const struct iphdr *old_iph = ip_hdr(skb);
2432 union vxlan_addr *dst;
2433 union vxlan_addr remote_ip, local_ip;
2434 struct vxlan_metadata _md;
2435 struct vxlan_metadata *md = &_md;
2436 unsigned int pkt_len = skb->len;
2437 __be16 src_port = 0, dst_port;
2438 struct dst_entry *ndst = NULL;
2439 __u8 tos, ttl, flow_flags = 0;
2442 u32 flags = vxlan->cfg.flags;
2443 bool udp_sum = false;
2444 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2446 #if IS_ENABLED(CONFIG_IPV6)
2450 info = skb_tunnel_info(skb);
2453 dst = &rdst->remote_ip;
2454 if (vxlan_addr_any(dst)) {
2456 /* short-circuited back to local bridge */
2457 vxlan_encap_bypass(skb, vxlan, vxlan,
2464 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2465 vni = (rdst->remote_vni) ? : default_vni;
2466 ifindex = rdst->remote_ifindex;
2467 local_ip = vxlan->cfg.saddr;
2468 dst_cache = &rdst->dst_cache;
2469 md->gbp = skb->mark;
2470 if (flags & VXLAN_F_TTL_INHERIT) {
2471 ttl = ip_tunnel_get_ttl(old_iph, skb);
2473 ttl = vxlan->cfg.ttl;
2474 if (!ttl && vxlan_addr_multicast(dst))
2478 tos = vxlan->cfg.tos;
2480 tos = ip_tunnel_get_dsfield(old_iph, skb);
2482 if (dst->sa.sa_family == AF_INET)
2483 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2485 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2486 #if IS_ENABLED(CONFIG_IPV6)
2487 label = vxlan->cfg.label;
2491 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2495 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2496 if (remote_ip.sa.sa_family == AF_INET) {
2497 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2498 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2500 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2501 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2504 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2505 flow_flags = info->key.flow_flags;
2506 vni = tunnel_id_to_key32(info->key.tun_id);
2508 dst_cache = &info->dst_cache;
2509 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2510 if (info->options_len < sizeof(*md))
2512 md = ip_tunnel_info_opts(info);
2514 ttl = info->key.ttl;
2515 tos = info->key.tos;
2516 #if IS_ENABLED(CONFIG_IPV6)
2517 label = info->key.label;
2519 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2521 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2522 vxlan->cfg.port_max, true);
2525 if (dst->sa.sa_family == AF_INET) {
2526 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2531 ifindex = sock4->sock->sk->sk_bound_dev_if;
2533 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2534 dst->sin.sin_addr.s_addr,
2535 &local_ip.sin.sin_addr.s_addr,
2536 dst_port, src_port, flow_flags,
2544 /* Bypass encapsulation if the destination is local */
2545 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2546 dst_port, ifindex, vni,
2547 &rt->dst, rt->rt_flags);
2551 if (vxlan->cfg.df == VXLAN_DF_SET) {
2553 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2554 struct ethhdr *eth = eth_hdr(skb);
2556 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2557 (ntohs(eth->h_proto) == ETH_P_IP &&
2558 old_iph->frag_off & htons(IP_DF)))
2561 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2566 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2567 netif_is_any_bridge_port(dev));
2572 struct ip_tunnel_info *unclone;
2573 struct in_addr src, dst;
2575 unclone = skb_tunnel_info_unclone(skb);
2576 if (unlikely(!unclone))
2579 src = remote_ip.sin.sin_addr;
2580 dst = local_ip.sin.sin_addr;
2581 unclone->key.u.ipv4.src = src.s_addr;
2582 unclone->key.u.ipv4.dst = dst.s_addr;
2584 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2589 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2590 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2591 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2592 vni, md, flags, udp_sum);
2596 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2597 dst->sin.sin_addr.s_addr, tos, ttl, df,
2598 src_port, dst_port, xnet, !udp_sum);
2599 #if IS_ENABLED(CONFIG_IPV6)
2601 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2604 ifindex = sock6->sock->sk->sk_bound_dev_if;
2606 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2607 label, &dst->sin6.sin6_addr,
2608 &local_ip.sin6.sin6_addr,
2612 err = PTR_ERR(ndst);
2618 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2620 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2621 dst_port, ifindex, vni,
2627 err = skb_tunnel_check_pmtu(skb, ndst,
2628 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2629 netif_is_any_bridge_port(dev));
2634 struct ip_tunnel_info *unclone;
2635 struct in6_addr src, dst;
2637 unclone = skb_tunnel_info_unclone(skb);
2638 if (unlikely(!unclone))
2641 src = remote_ip.sin6.sin6_addr;
2642 dst = local_ip.sin6.sin6_addr;
2643 unclone->key.u.ipv6.src = src;
2644 unclone->key.u.ipv6.dst = dst;
2647 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2652 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2653 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2654 skb_scrub_packet(skb, xnet);
2655 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2656 vni, md, flags, udp_sum);
2660 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2661 &local_ip.sin6.sin6_addr,
2662 &dst->sin6.sin6_addr, tos, ttl,
2663 label, src_port, dst_port, !udp_sum);
2666 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2672 dev->stats.tx_dropped++;
2673 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2680 dev->stats.collisions++;
2681 else if (err == -ENETUNREACH)
2682 dev->stats.tx_carrier_errors++;
2684 dev->stats.tx_errors++;
2685 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2689 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2690 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2692 struct vxlan_rdst nh_rdst;
2697 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2698 hash = skb_get_hash(skb);
2701 nh = rcu_dereference(f->nh);
2706 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2709 if (likely(do_xmit))
2710 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2717 dev->stats.tx_dropped++;
2718 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2719 VXLAN_VNI_STATS_TX_DROPS, 0);
2723 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2724 u32 nhid, __be32 vni)
2726 struct vxlan_dev *vxlan = netdev_priv(dev);
2727 struct vxlan_rdst nh_rdst;
2732 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2733 hash = skb_get_hash(skb);
2736 nh = nexthop_find_by_id(dev_net(dev), nhid);
2737 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2741 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2744 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2747 if (likely(do_xmit))
2748 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2752 return NETDEV_TX_OK;
2755 dev->stats.tx_dropped++;
2756 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2757 VXLAN_VNI_STATS_TX_DROPS, 0);
2759 return NETDEV_TX_OK;
2762 /* Transmit local packets over Vxlan
2764 * Outer IP header inherits ECN and DF from inner header.
2765 * Outer UDP destination is the VXLAN assigned port.
2766 * source port is based on hash of flow
2768 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2770 struct vxlan_dev *vxlan = netdev_priv(dev);
2771 struct vxlan_rdst *rdst, *fdst = NULL;
2772 const struct ip_tunnel_info *info;
2773 bool did_rsc = false;
2774 struct vxlan_fdb *f;
2779 info = skb_tunnel_info(skb);
2781 skb_reset_mac_header(skb);
2783 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2784 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2785 info->mode & IP_TUNNEL_INFO_TX) {
2786 vni = tunnel_id_to_key32(info->key.tun_id);
2787 nhid = info->key.nhid;
2789 if (info && info->mode & IP_TUNNEL_INFO_TX)
2790 vxlan_xmit_one(skb, dev, vni, NULL, false);
2793 return NETDEV_TX_OK;
2797 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2799 if (ntohs(eth->h_proto) == ETH_P_ARP)
2800 return arp_reduce(dev, skb, vni);
2801 #if IS_ENABLED(CONFIG_IPV6)
2802 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2803 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2804 sizeof(struct nd_msg)) &&
2805 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2806 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2808 if (m->icmph.icmp6_code == 0 &&
2809 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2810 return neigh_reduce(dev, skb, vni);
2816 return vxlan_xmit_nhid(skb, dev, nhid, vni);
2818 if (vxlan->cfg.flags & VXLAN_F_MDB) {
2819 struct vxlan_mdb_entry *mdb_entry;
2822 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2826 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2834 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2837 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2838 (ntohs(eth->h_proto) == ETH_P_IP ||
2839 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2840 did_rsc = route_shortcircuit(dev, skb);
2842 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2846 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2848 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2849 !is_multicast_ether_addr(eth->h_dest))
2850 vxlan_fdb_miss(vxlan, eth->h_dest);
2852 dev->stats.tx_dropped++;
2853 vxlan_vnifilter_count(vxlan, vni, NULL,
2854 VXLAN_VNI_STATS_TX_DROPS, 0);
2856 return NETDEV_TX_OK;
2860 if (rcu_access_pointer(f->nh)) {
2861 vxlan_xmit_nh(skb, dev, f,
2862 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2864 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2865 struct sk_buff *skb1;
2871 skb1 = skb_clone(skb, GFP_ATOMIC);
2873 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2876 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2881 return NETDEV_TX_OK;
2884 /* Walk the forwarding table and purge stale entries */
2885 static void vxlan_cleanup(struct timer_list *t)
2887 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2888 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2891 if (!netif_running(vxlan->dev))
2894 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2895 struct hlist_node *p, *n;
2897 spin_lock(&vxlan->hash_lock[h]);
2898 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2900 = container_of(p, struct vxlan_fdb, hlist);
2901 unsigned long timeout;
2903 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2906 if (f->flags & NTF_EXT_LEARNED)
2909 timeout = f->used + vxlan->cfg.age_interval * HZ;
2910 if (time_before_eq(timeout, jiffies)) {
2911 netdev_dbg(vxlan->dev,
2912 "garbage collect %pM\n",
2914 f->state = NUD_STALE;
2915 vxlan_fdb_destroy(vxlan, f, true, true);
2916 } else if (time_before(timeout, next_timer))
2917 next_timer = timeout;
2919 spin_unlock(&vxlan->hash_lock[h]);
2922 mod_timer(&vxlan->age_timer, next_timer);
2925 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2927 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2929 spin_lock(&vn->sock_lock);
2930 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2931 #if IS_ENABLED(CONFIG_IPV6)
2932 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2934 spin_unlock(&vn->sock_lock);
2937 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2938 struct vxlan_dev_node *node)
2940 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2941 __be32 vni = vxlan->default_dst.remote_vni;
2943 node->vxlan = vxlan;
2944 spin_lock(&vn->sock_lock);
2945 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2946 spin_unlock(&vn->sock_lock);
2949 /* Setup stats when device is created */
2950 static int vxlan_init(struct net_device *dev)
2952 struct vxlan_dev *vxlan = netdev_priv(dev);
2955 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2956 vxlan_vnigroup_init(vxlan);
2958 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2961 goto err_vnigroup_uninit;
2964 err = gro_cells_init(&vxlan->gro_cells, dev);
2966 goto err_free_percpu;
2968 err = vxlan_mdb_init(vxlan);
2970 goto err_gro_cells_destroy;
2974 err_gro_cells_destroy:
2975 gro_cells_destroy(&vxlan->gro_cells);
2977 free_percpu(dev->tstats);
2978 err_vnigroup_uninit:
2979 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2980 vxlan_vnigroup_uninit(vxlan);
2984 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2986 struct vxlan_fdb *f;
2987 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2989 spin_lock_bh(&vxlan->hash_lock[hash_index]);
2990 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2992 vxlan_fdb_destroy(vxlan, f, true, true);
2993 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2996 static void vxlan_uninit(struct net_device *dev)
2998 struct vxlan_dev *vxlan = netdev_priv(dev);
3000 vxlan_mdb_fini(vxlan);
3002 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
3003 vxlan_vnigroup_uninit(vxlan);
3005 gro_cells_destroy(&vxlan->gro_cells);
3007 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3009 free_percpu(dev->tstats);
3012 /* Start ageing timer and join group when device is brought up */
3013 static int vxlan_open(struct net_device *dev)
3015 struct vxlan_dev *vxlan = netdev_priv(dev);
3018 ret = vxlan_sock_add(vxlan);
3022 ret = vxlan_multicast_join(vxlan);
3024 vxlan_sock_release(vxlan);
3028 if (vxlan->cfg.age_interval)
3029 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3034 /* Purge the forwarding table */
3035 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3039 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3040 struct hlist_node *p, *n;
3042 spin_lock_bh(&vxlan->hash_lock[h]);
3043 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3045 = container_of(p, struct vxlan_fdb, hlist);
3046 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3048 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3049 if (is_zero_ether_addr(f->eth_addr) &&
3050 f->vni == vxlan->cfg.vni)
3052 vxlan_fdb_destroy(vxlan, f, true, true);
3054 spin_unlock_bh(&vxlan->hash_lock[h]);
3058 /* Cleanup timer and forwarding table on shutdown */
3059 static int vxlan_stop(struct net_device *dev)
3061 struct vxlan_dev *vxlan = netdev_priv(dev);
3063 vxlan_multicast_leave(vxlan);
3065 del_timer_sync(&vxlan->age_timer);
3067 vxlan_flush(vxlan, false);
3068 vxlan_sock_release(vxlan);
3073 /* Stub, nothing needs to be done. */
3074 static void vxlan_set_multicast_list(struct net_device *dev)
3078 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3080 struct vxlan_dev *vxlan = netdev_priv(dev);
3081 struct vxlan_rdst *dst = &vxlan->default_dst;
3082 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3083 dst->remote_ifindex);
3085 /* This check is different than dev->max_mtu, because it looks at
3086 * the lowerdev->mtu, rather than the static dev->max_mtu
3089 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3090 if (new_mtu > max_mtu)
3098 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3100 struct vxlan_dev *vxlan = netdev_priv(dev);
3101 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3102 __be16 sport, dport;
3104 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3105 vxlan->cfg.port_max, true);
3106 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3108 if (ip_tunnel_info_af(info) == AF_INET) {
3109 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3112 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3113 info->key.u.ipv4.dst,
3114 &info->key.u.ipv4.src, dport, sport,
3115 info->key.flow_flags, &info->dst_cache,
3121 #if IS_ENABLED(CONFIG_IPV6)
3122 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3123 struct dst_entry *ndst;
3125 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3126 info->key.label, &info->key.u.ipv6.dst,
3127 &info->key.u.ipv6.src, dport, sport,
3128 &info->dst_cache, info);
3130 return PTR_ERR(ndst);
3132 #else /* !CONFIG_IPV6 */
3133 return -EPFNOSUPPORT;
3136 info->key.tp_src = sport;
3137 info->key.tp_dst = dport;
3141 static const struct net_device_ops vxlan_netdev_ether_ops = {
3142 .ndo_init = vxlan_init,
3143 .ndo_uninit = vxlan_uninit,
3144 .ndo_open = vxlan_open,
3145 .ndo_stop = vxlan_stop,
3146 .ndo_start_xmit = vxlan_xmit,
3147 .ndo_get_stats64 = dev_get_tstats64,
3148 .ndo_set_rx_mode = vxlan_set_multicast_list,
3149 .ndo_change_mtu = vxlan_change_mtu,
3150 .ndo_validate_addr = eth_validate_addr,
3151 .ndo_set_mac_address = eth_mac_addr,
3152 .ndo_fdb_add = vxlan_fdb_add,
3153 .ndo_fdb_del = vxlan_fdb_delete,
3154 .ndo_fdb_dump = vxlan_fdb_dump,
3155 .ndo_fdb_get = vxlan_fdb_get,
3156 .ndo_mdb_add = vxlan_mdb_add,
3157 .ndo_mdb_del = vxlan_mdb_del,
3158 .ndo_mdb_dump = vxlan_mdb_dump,
3159 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3162 static const struct net_device_ops vxlan_netdev_raw_ops = {
3163 .ndo_init = vxlan_init,
3164 .ndo_uninit = vxlan_uninit,
3165 .ndo_open = vxlan_open,
3166 .ndo_stop = vxlan_stop,
3167 .ndo_start_xmit = vxlan_xmit,
3168 .ndo_get_stats64 = dev_get_tstats64,
3169 .ndo_change_mtu = vxlan_change_mtu,
3170 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3173 /* Info for udev, that this is a virtual tunnel endpoint */
3174 static struct device_type vxlan_type = {
3178 /* Calls the ndo_udp_tunnel_add of the caller in order to
3179 * supply the listening VXLAN udp ports. Callers are expected
3180 * to implement the ndo_udp_tunnel_add.
3182 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3184 struct vxlan_sock *vs;
3185 struct net *net = dev_net(dev);
3186 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3189 spin_lock(&vn->sock_lock);
3190 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3191 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3192 unsigned short type;
3194 if (vs->flags & VXLAN_F_GPE)
3195 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3197 type = UDP_TUNNEL_TYPE_VXLAN;
3200 udp_tunnel_push_rx_port(dev, vs->sock, type);
3202 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3205 spin_unlock(&vn->sock_lock);
3208 /* Initialize the device structure. */
3209 static void vxlan_setup(struct net_device *dev)
3211 struct vxlan_dev *vxlan = netdev_priv(dev);
3214 eth_hw_addr_random(dev);
3217 dev->needs_free_netdev = true;
3218 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3220 dev->features |= NETIF_F_LLTX;
3221 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3222 dev->features |= NETIF_F_RXCSUM;
3223 dev->features |= NETIF_F_GSO_SOFTWARE;
3225 dev->vlan_features = dev->features;
3226 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3227 dev->hw_features |= NETIF_F_RXCSUM;
3228 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3229 netif_keep_dst(dev);
3230 dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN;
3232 /* MTU range: 68 - 65535 */
3233 dev->min_mtu = ETH_MIN_MTU;
3234 dev->max_mtu = ETH_MAX_MTU;
3236 INIT_LIST_HEAD(&vxlan->next);
3238 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3242 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3243 spin_lock_init(&vxlan->hash_lock[h]);
3244 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3248 static void vxlan_ether_setup(struct net_device *dev)
3250 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3251 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3252 dev->netdev_ops = &vxlan_netdev_ether_ops;
3255 static void vxlan_raw_setup(struct net_device *dev)
3257 dev->header_ops = NULL;
3258 dev->type = ARPHRD_NONE;
3259 dev->hard_header_len = 0;
3261 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3262 dev->netdev_ops = &vxlan_netdev_raw_ops;
3265 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3266 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3267 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3268 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3269 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3270 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3271 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3272 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3273 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3274 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3275 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3276 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3277 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3278 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3279 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3280 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3281 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3282 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3283 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3284 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3285 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3286 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3287 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3288 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3289 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3290 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3291 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3292 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3293 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3294 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3295 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3296 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 },
3297 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1),
3300 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3301 struct netlink_ext_ack *extack)
3303 if (tb[IFLA_ADDRESS]) {
3304 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3305 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3306 "Provided link layer address is not Ethernet");
3310 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3311 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3312 "Provided Ethernet address is not unicast");
3313 return -EADDRNOTAVAIL;
3318 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3320 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3321 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3322 "MTU must be between 68 and 65535");
3328 NL_SET_ERR_MSG(extack,
3329 "Required attributes not provided to perform the operation");
3333 if (data[IFLA_VXLAN_ID]) {
3334 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3336 if (id >= VXLAN_N_VID) {
3337 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3338 "VXLAN ID must be lower than 16777216");
3343 if (data[IFLA_VXLAN_PORT_RANGE]) {
3344 const struct ifla_vxlan_port_range *p
3345 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3347 if (ntohs(p->high) < ntohs(p->low)) {
3348 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3349 "Invalid source port range");
3354 if (data[IFLA_VXLAN_DF]) {
3355 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3357 if (df < 0 || df > VXLAN_DF_MAX) {
3358 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3359 "Invalid DF attribute");
3367 static void vxlan_get_drvinfo(struct net_device *netdev,
3368 struct ethtool_drvinfo *drvinfo)
3370 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3371 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3374 static int vxlan_get_link_ksettings(struct net_device *dev,
3375 struct ethtool_link_ksettings *cmd)
3377 struct vxlan_dev *vxlan = netdev_priv(dev);
3378 struct vxlan_rdst *dst = &vxlan->default_dst;
3379 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3380 dst->remote_ifindex);
3383 cmd->base.duplex = DUPLEX_UNKNOWN;
3384 cmd->base.port = PORT_OTHER;
3385 cmd->base.speed = SPEED_UNKNOWN;
3390 return __ethtool_get_link_ksettings(lowerdev, cmd);
3393 static const struct ethtool_ops vxlan_ethtool_ops = {
3394 .get_drvinfo = vxlan_get_drvinfo,
3395 .get_link = ethtool_op_get_link,
3396 .get_link_ksettings = vxlan_get_link_ksettings,
3399 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3400 __be16 port, u32 flags, int ifindex)
3402 struct socket *sock;
3403 struct udp_port_cfg udp_conf;
3406 memset(&udp_conf, 0, sizeof(udp_conf));
3409 udp_conf.family = AF_INET6;
3410 udp_conf.use_udp6_rx_checksums =
3411 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3412 udp_conf.ipv6_v6only = 1;
3414 udp_conf.family = AF_INET;
3417 udp_conf.local_udp_port = port;
3418 udp_conf.bind_ifindex = ifindex;
3420 /* Open UDP socket */
3421 err = udp_sock_create(net, &udp_conf, &sock);
3423 return ERR_PTR(err);
3425 udp_allow_gso(sock->sk);
3429 /* Create new listen socket if needed */
3430 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3431 __be16 port, u32 flags,
3434 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3435 struct vxlan_sock *vs;
3436 struct socket *sock;
3438 struct udp_tunnel_sock_cfg tunnel_cfg;
3440 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3442 return ERR_PTR(-ENOMEM);
3444 for (h = 0; h < VNI_HASH_SIZE; ++h)
3445 INIT_HLIST_HEAD(&vs->vni_list[h]);
3447 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3450 return ERR_CAST(sock);
3454 refcount_set(&vs->refcnt, 1);
3455 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3457 spin_lock(&vn->sock_lock);
3458 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3459 udp_tunnel_notify_add_rx_port(sock,
3460 (vs->flags & VXLAN_F_GPE) ?
3461 UDP_TUNNEL_TYPE_VXLAN_GPE :
3462 UDP_TUNNEL_TYPE_VXLAN);
3463 spin_unlock(&vn->sock_lock);
3465 /* Mark socket as an encapsulation socket. */
3466 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3467 tunnel_cfg.sk_user_data = vs;
3468 tunnel_cfg.encap_type = 1;
3469 tunnel_cfg.encap_rcv = vxlan_rcv;
3470 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3471 tunnel_cfg.encap_destroy = NULL;
3472 if (vs->flags & VXLAN_F_GPE) {
3473 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3474 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3476 tunnel_cfg.gro_receive = vxlan_gro_receive;
3477 tunnel_cfg.gro_complete = vxlan_gro_complete;
3480 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3485 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3487 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3488 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3489 struct vxlan_sock *vs = NULL;
3490 struct vxlan_dev_node *node;
3491 int l3mdev_index = 0;
3493 if (vxlan->cfg.remote_ifindex)
3494 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3495 vxlan->net, vxlan->cfg.remote_ifindex);
3497 if (!vxlan->cfg.no_share) {
3498 spin_lock(&vn->sock_lock);
3499 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3500 vxlan->cfg.dst_port, vxlan->cfg.flags,
3502 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3503 spin_unlock(&vn->sock_lock);
3506 spin_unlock(&vn->sock_lock);
3509 vs = vxlan_socket_create(vxlan->net, ipv6,
3510 vxlan->cfg.dst_port, vxlan->cfg.flags,
3514 #if IS_ENABLED(CONFIG_IPV6)
3516 rcu_assign_pointer(vxlan->vn6_sock, vs);
3517 node = &vxlan->hlist6;
3521 rcu_assign_pointer(vxlan->vn4_sock, vs);
3522 node = &vxlan->hlist4;
3525 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3526 vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3528 vxlan_vs_add_dev(vs, vxlan, node);
3533 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3535 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3536 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3537 bool ipv4 = !ipv6 || metadata;
3540 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3541 #if IS_ENABLED(CONFIG_IPV6)
3542 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3544 ret = __vxlan_sock_add(vxlan, true);
3545 if (ret < 0 && ret != -EAFNOSUPPORT)
3550 ret = __vxlan_sock_add(vxlan, false);
3552 vxlan_sock_release(vxlan);
3556 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3557 struct vxlan_config *conf, __be32 vni)
3559 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3560 struct vxlan_dev *tmp;
3562 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3565 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3566 if (!vxlan_vnifilter_lookup(tmp, vni))
3568 } else if (tmp->cfg.vni != vni) {
3571 if (tmp->cfg.dst_port != conf->dst_port)
3573 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3574 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3577 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3578 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3587 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3588 struct net_device **lower,
3589 struct vxlan_dev *old,
3590 struct netlink_ext_ack *extack)
3592 bool use_ipv6 = false;
3594 if (conf->flags & VXLAN_F_GPE) {
3595 /* For now, allow GPE only together with
3596 * COLLECT_METADATA. This can be relaxed later; in such
3597 * case, the other side of the PtP link will have to be
3600 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3601 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3602 NL_SET_ERR_MSG(extack,
3603 "VXLAN GPE does not support this combination of attributes");
3608 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3609 /* Unless IPv6 is explicitly requested, assume IPv4 */
3610 conf->remote_ip.sa.sa_family = AF_INET;
3611 conf->saddr.sa.sa_family = AF_INET;
3612 } else if (!conf->remote_ip.sa.sa_family) {
3613 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3614 } else if (!conf->saddr.sa.sa_family) {
3615 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3618 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3619 NL_SET_ERR_MSG(extack,
3620 "Local and remote address must be from the same family");
3624 if (vxlan_addr_multicast(&conf->saddr)) {
3625 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3629 if (conf->saddr.sa.sa_family == AF_INET6) {
3630 if (!IS_ENABLED(CONFIG_IPV6)) {
3631 NL_SET_ERR_MSG(extack,
3632 "IPv6 support not enabled in the kernel");
3633 return -EPFNOSUPPORT;
3636 conf->flags |= VXLAN_F_IPV6;
3638 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3640 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3642 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3644 if (local_type & IPV6_ADDR_LINKLOCAL) {
3645 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3646 (remote_type != IPV6_ADDR_ANY)) {
3647 NL_SET_ERR_MSG(extack,
3648 "Invalid combination of local and remote address scopes");
3652 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3655 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3656 NL_SET_ERR_MSG(extack,
3657 "Invalid combination of local and remote address scopes");
3661 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3666 if (conf->label && !use_ipv6) {
3667 NL_SET_ERR_MSG(extack,
3668 "Label attribute only applies to IPv6 VXLAN devices");
3672 if (conf->remote_ifindex) {
3673 struct net_device *lowerdev;
3675 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3677 NL_SET_ERR_MSG(extack,
3678 "Invalid local interface, device not found");
3682 #if IS_ENABLED(CONFIG_IPV6)
3684 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3686 if (idev && idev->cnf.disable_ipv6) {
3687 NL_SET_ERR_MSG(extack,
3688 "IPv6 support disabled by administrator");
3696 if (vxlan_addr_multicast(&conf->remote_ip)) {
3697 NL_SET_ERR_MSG(extack,
3698 "Local interface required for multicast remote destination");
3703 #if IS_ENABLED(CONFIG_IPV6)
3704 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3705 NL_SET_ERR_MSG(extack,
3706 "Local interface required for link-local local/remote addresses");
3714 if (!conf->dst_port) {
3715 if (conf->flags & VXLAN_F_GPE)
3716 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3718 conf->dst_port = htons(vxlan_port);
3721 if (!conf->age_interval)
3722 conf->age_interval = FDB_AGE_DEFAULT;
3724 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3725 NL_SET_ERR_MSG(extack,
3726 "A VXLAN device with the specified VNI already exists");
3733 static void vxlan_config_apply(struct net_device *dev,
3734 struct vxlan_config *conf,
3735 struct net_device *lowerdev,
3736 struct net *src_net,
3739 struct vxlan_dev *vxlan = netdev_priv(dev);
3740 struct vxlan_rdst *dst = &vxlan->default_dst;
3741 unsigned short needed_headroom = ETH_HLEN;
3742 int max_mtu = ETH_MAX_MTU;
3743 u32 flags = conf->flags;
3746 if (flags & VXLAN_F_GPE)
3747 vxlan_raw_setup(dev);
3749 vxlan_ether_setup(dev);
3752 dev->mtu = conf->mtu;
3754 vxlan->net = src_net;
3757 dst->remote_vni = conf->vni;
3759 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3762 dst->remote_ifindex = conf->remote_ifindex;
3764 netif_inherit_tso_max(dev, lowerdev);
3766 needed_headroom = lowerdev->hard_header_len;
3767 needed_headroom += lowerdev->needed_headroom;
3769 dev->needed_tailroom = lowerdev->needed_tailroom;
3771 max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3772 if (max_mtu < ETH_MIN_MTU)
3773 max_mtu = ETH_MIN_MTU;
3775 if (!changelink && !conf->mtu)
3779 if (dev->mtu > max_mtu)
3782 if (flags & VXLAN_F_COLLECT_METADATA)
3783 flags |= VXLAN_F_IPV6;
3784 needed_headroom += vxlan_headroom(flags);
3785 dev->needed_headroom = needed_headroom;
3787 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3790 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3791 struct vxlan_config *conf, bool changelink,
3792 struct netlink_ext_ack *extack)
3794 struct vxlan_dev *vxlan = netdev_priv(dev);
3795 struct net_device *lowerdev;
3798 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3802 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3807 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3808 struct vxlan_config *conf,
3809 struct netlink_ext_ack *extack)
3811 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3812 struct vxlan_dev *vxlan = netdev_priv(dev);
3813 struct net_device *remote_dev = NULL;
3814 struct vxlan_fdb *f = NULL;
3815 bool unregister = false;
3816 struct vxlan_rdst *dst;
3819 dst = &vxlan->default_dst;
3820 err = vxlan_dev_configure(net, dev, conf, false, extack);
3824 dev->ethtool_ops = &vxlan_ethtool_ops;
3826 /* create an fdb entry for a valid default destination */
3827 if (!vxlan_addr_any(&dst->remote_ip)) {
3828 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3830 NUD_REACHABLE | NUD_PERMANENT,
3831 vxlan->cfg.dst_port,
3834 dst->remote_ifindex,
3835 NTF_SELF, 0, &f, extack);
3840 err = register_netdevice(dev);
3845 if (dst->remote_ifindex) {
3846 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3852 err = netdev_upper_dev_link(remote_dev, dev, extack);
3857 err = rtnl_configure_link(dev, NULL, 0, NULL);
3862 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3864 /* notify default fdb entry */
3865 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3866 RTM_NEWNEIGH, true, extack);
3868 vxlan_fdb_destroy(vxlan, f, false, false);
3870 netdev_upper_dev_unlink(remote_dev, dev);
3875 list_add(&vxlan->next, &vn->vxlan_list);
3877 dst->remote_dev = remote_dev;
3881 netdev_upper_dev_unlink(remote_dev, dev);
3883 /* unregister_netdevice() destroys the default FDB entry with deletion
3884 * notification. But the addition notification was not sent yet, so
3885 * destroy the entry by hand here.
3888 __vxlan_fdb_free(f);
3891 unregister_netdevice(dev);
3895 /* Set/clear flags based on attribute */
3896 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3897 int attrtype, unsigned long mask, bool changelink,
3898 bool changelink_supported,
3899 struct netlink_ext_ack *extack)
3901 unsigned long flags;
3906 if (changelink && !changelink_supported) {
3907 vxlan_flag_attr_error(attrtype, extack);
3911 if (vxlan_policy[attrtype].type == NLA_FLAG)
3912 flags = conf->flags | mask;
3913 else if (nla_get_u8(tb[attrtype]))
3914 flags = conf->flags | mask;
3916 flags = conf->flags & ~mask;
3918 conf->flags = flags;
3923 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3924 struct net_device *dev, struct vxlan_config *conf,
3925 bool changelink, struct netlink_ext_ack *extack)
3927 struct vxlan_dev *vxlan = netdev_priv(dev);
3930 memset(conf, 0, sizeof(*conf));
3932 /* if changelink operation, start with old existing cfg */
3934 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3936 if (data[IFLA_VXLAN_ID]) {
3937 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3939 if (changelink && (vni != conf->vni)) {
3940 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3943 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3946 if (data[IFLA_VXLAN_GROUP]) {
3947 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3948 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3952 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3953 conf->remote_ip.sa.sa_family = AF_INET;
3954 } else if (data[IFLA_VXLAN_GROUP6]) {
3955 if (!IS_ENABLED(CONFIG_IPV6)) {
3956 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3957 return -EPFNOSUPPORT;
3960 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3961 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3965 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3966 conf->remote_ip.sa.sa_family = AF_INET6;
3969 if (data[IFLA_VXLAN_LOCAL]) {
3970 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3971 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
3975 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3976 conf->saddr.sa.sa_family = AF_INET;
3977 } else if (data[IFLA_VXLAN_LOCAL6]) {
3978 if (!IS_ENABLED(CONFIG_IPV6)) {
3979 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
3980 return -EPFNOSUPPORT;
3983 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3984 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
3988 /* TODO: respect scope id */
3989 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3990 conf->saddr.sa.sa_family = AF_INET6;
3993 if (data[IFLA_VXLAN_LINK])
3994 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3996 if (data[IFLA_VXLAN_TOS])
3997 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3999 if (data[IFLA_VXLAN_TTL])
4000 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4002 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4003 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4004 VXLAN_F_TTL_INHERIT, changelink, false,
4011 if (data[IFLA_VXLAN_LABEL])
4012 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4013 IPV6_FLOWLABEL_MASK;
4015 if (data[IFLA_VXLAN_LEARNING]) {
4016 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4017 VXLAN_F_LEARN, changelink, true,
4021 } else if (!changelink) {
4022 /* default to learn on a new device */
4023 conf->flags |= VXLAN_F_LEARN;
4026 if (data[IFLA_VXLAN_AGEING])
4027 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4029 if (data[IFLA_VXLAN_PROXY]) {
4030 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4031 VXLAN_F_PROXY, changelink, false,
4037 if (data[IFLA_VXLAN_RSC]) {
4038 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4039 VXLAN_F_RSC, changelink, false,
4045 if (data[IFLA_VXLAN_L2MISS]) {
4046 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4047 VXLAN_F_L2MISS, changelink, false,
4053 if (data[IFLA_VXLAN_L3MISS]) {
4054 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4055 VXLAN_F_L3MISS, changelink, false,
4061 if (data[IFLA_VXLAN_LIMIT]) {
4063 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4064 "Cannot change limit");
4067 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4070 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4071 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4072 VXLAN_F_COLLECT_METADATA, changelink, false,
4078 if (data[IFLA_VXLAN_PORT_RANGE]) {
4080 const struct ifla_vxlan_port_range *p
4081 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4082 conf->port_min = ntohs(p->low);
4083 conf->port_max = ntohs(p->high);
4085 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4086 "Cannot change port range");
4091 if (data[IFLA_VXLAN_PORT]) {
4093 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4094 "Cannot change port");
4097 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4100 if (data[IFLA_VXLAN_UDP_CSUM]) {
4102 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4103 "Cannot change UDP_CSUM flag");
4106 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4107 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4110 if (data[IFLA_VXLAN_LOCALBYPASS]) {
4111 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4112 VXLAN_F_LOCALBYPASS, changelink,
4116 } else if (!changelink) {
4117 /* default to local bypass on a new device */
4118 conf->flags |= VXLAN_F_LOCALBYPASS;
4121 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4122 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4123 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4129 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4130 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4131 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4137 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4138 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4139 VXLAN_F_REMCSUM_TX, changelink, false,
4145 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4146 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4147 VXLAN_F_REMCSUM_RX, changelink, false,
4153 if (data[IFLA_VXLAN_GBP]) {
4154 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4155 VXLAN_F_GBP, changelink, false, extack);
4160 if (data[IFLA_VXLAN_GPE]) {
4161 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4162 VXLAN_F_GPE, changelink, false,
4168 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4169 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4170 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4178 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4179 "Cannot change mtu");
4182 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4185 if (data[IFLA_VXLAN_DF])
4186 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4188 if (data[IFLA_VXLAN_VNIFILTER]) {
4189 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4190 VXLAN_F_VNIFILTER, changelink, false,
4195 if ((conf->flags & VXLAN_F_VNIFILTER) &&
4196 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4197 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4198 "vxlan vnifilter only valid in collect metadata mode");
4206 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4207 struct nlattr *tb[], struct nlattr *data[],
4208 struct netlink_ext_ack *extack)
4210 struct vxlan_config conf;
4213 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4217 return __vxlan_dev_create(src_net, dev, &conf, extack);
4220 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4221 struct nlattr *data[],
4222 struct netlink_ext_ack *extack)
4224 struct vxlan_dev *vxlan = netdev_priv(dev);
4225 struct net_device *lowerdev;
4226 struct vxlan_config conf;
4227 struct vxlan_rdst *dst;
4230 dst = &vxlan->default_dst;
4231 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4235 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4240 if (dst->remote_dev == lowerdev)
4243 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4248 /* handle default dst entry */
4249 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4250 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4252 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4253 if (!vxlan_addr_any(&conf.remote_ip)) {
4254 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4256 NUD_REACHABLE | NUD_PERMANENT,
4257 NLM_F_APPEND | NLM_F_CREATE,
4258 vxlan->cfg.dst_port,
4260 conf.remote_ifindex,
4261 NTF_SELF, 0, true, extack);
4263 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4264 netdev_adjacent_change_abort(dst->remote_dev,
4269 if (!vxlan_addr_any(&dst->remote_ip))
4270 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4272 vxlan->cfg.dst_port,
4275 dst->remote_ifindex,
4277 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4279 /* If vni filtering device, also update fdb entries of
4280 * all vnis that were using default remote ip
4282 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4283 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4284 &conf.remote_ip, extack);
4286 netdev_adjacent_change_abort(dst->remote_dev,
4293 if (conf.age_interval != vxlan->cfg.age_interval)
4294 mod_timer(&vxlan->age_timer, jiffies);
4296 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4297 if (lowerdev && lowerdev != dst->remote_dev)
4298 dst->remote_dev = lowerdev;
4299 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4303 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4305 struct vxlan_dev *vxlan = netdev_priv(dev);
4307 vxlan_flush(vxlan, true);
4309 list_del(&vxlan->next);
4310 unregister_netdevice_queue(dev, head);
4311 if (vxlan->default_dst.remote_dev)
4312 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4315 static size_t vxlan_get_size(const struct net_device *dev)
4318 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4319 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4320 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4321 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4322 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4323 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4324 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4325 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4326 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4327 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4328 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4329 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4330 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4331 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4332 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4333 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4334 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4335 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4336 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4338 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4339 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4340 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4341 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4342 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4346 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4348 const struct vxlan_dev *vxlan = netdev_priv(dev);
4349 const struct vxlan_rdst *dst = &vxlan->default_dst;
4350 struct ifla_vxlan_port_range ports = {
4351 .low = htons(vxlan->cfg.port_min),
4352 .high = htons(vxlan->cfg.port_max),
4355 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4356 goto nla_put_failure;
4358 if (!vxlan_addr_any(&dst->remote_ip)) {
4359 if (dst->remote_ip.sa.sa_family == AF_INET) {
4360 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4361 dst->remote_ip.sin.sin_addr.s_addr))
4362 goto nla_put_failure;
4363 #if IS_ENABLED(CONFIG_IPV6)
4365 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4366 &dst->remote_ip.sin6.sin6_addr))
4367 goto nla_put_failure;
4372 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4373 goto nla_put_failure;
4375 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4376 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4377 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4378 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4379 goto nla_put_failure;
4380 #if IS_ENABLED(CONFIG_IPV6)
4382 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4383 &vxlan->cfg.saddr.sin6.sin6_addr))
4384 goto nla_put_failure;
4389 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4390 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4391 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4392 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4393 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4394 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4395 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4396 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4397 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4398 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4399 nla_put_u8(skb, IFLA_VXLAN_RSC,
4400 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4401 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4402 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4403 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4404 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4405 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4406 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4407 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4408 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4409 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4410 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4411 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4412 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4413 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4414 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4415 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4416 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4417 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4418 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4419 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4420 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4421 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4422 goto nla_put_failure;
4424 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4425 goto nla_put_failure;
4427 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4428 nla_put_flag(skb, IFLA_VXLAN_GBP))
4429 goto nla_put_failure;
4431 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4432 nla_put_flag(skb, IFLA_VXLAN_GPE))
4433 goto nla_put_failure;
4435 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4436 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4437 goto nla_put_failure;
4439 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4440 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4441 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4442 goto nla_put_failure;
4450 static struct net *vxlan_get_link_net(const struct net_device *dev)
4452 struct vxlan_dev *vxlan = netdev_priv(dev);
4457 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4459 .maxtype = IFLA_VXLAN_MAX,
4460 .policy = vxlan_policy,
4461 .priv_size = sizeof(struct vxlan_dev),
4462 .setup = vxlan_setup,
4463 .validate = vxlan_validate,
4464 .newlink = vxlan_newlink,
4465 .changelink = vxlan_changelink,
4466 .dellink = vxlan_dellink,
4467 .get_size = vxlan_get_size,
4468 .fill_info = vxlan_fill_info,
4469 .get_link_net = vxlan_get_link_net,
4472 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4473 u8 name_assign_type,
4474 struct vxlan_config *conf)
4476 struct nlattr *tb[IFLA_MAX + 1];
4477 struct net_device *dev;
4480 memset(&tb, 0, sizeof(tb));
4482 dev = rtnl_create_link(net, name, name_assign_type,
4483 &vxlan_link_ops, tb, NULL);
4487 err = __vxlan_dev_create(net, dev, conf, NULL);
4490 return ERR_PTR(err);
4493 err = rtnl_configure_link(dev, NULL, 0, NULL);
4495 LIST_HEAD(list_kill);
4497 vxlan_dellink(dev, &list_kill);
4498 unregister_netdevice_many(&list_kill);
4499 return ERR_PTR(err);
4504 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4506 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4507 struct net_device *dev)
4509 struct vxlan_dev *vxlan, *next;
4510 LIST_HEAD(list_kill);
4512 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4513 struct vxlan_rdst *dst = &vxlan->default_dst;
4515 /* In case we created vxlan device with carrier
4516 * and we loose the carrier due to module unload
4517 * we also need to remove vxlan device. In other
4518 * cases, it's not necessary and remote_ifindex
4519 * is 0 here, so no matches.
4521 if (dst->remote_ifindex == dev->ifindex)
4522 vxlan_dellink(vxlan->dev, &list_kill);
4525 unregister_netdevice_many(&list_kill);
4528 static int vxlan_netdevice_event(struct notifier_block *unused,
4529 unsigned long event, void *ptr)
4531 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4532 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4534 if (event == NETDEV_UNREGISTER)
4535 vxlan_handle_lowerdev_unregister(vn, dev);
4536 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4537 vxlan_offload_rx_ports(dev, true);
4538 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4539 vxlan_offload_rx_ports(dev, false);
4544 static struct notifier_block vxlan_notifier_block __read_mostly = {
4545 .notifier_call = vxlan_netdevice_event,
4549 vxlan_fdb_offloaded_set(struct net_device *dev,
4550 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4552 struct vxlan_dev *vxlan = netdev_priv(dev);
4553 struct vxlan_rdst *rdst;
4554 struct vxlan_fdb *f;
4557 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4559 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4561 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4565 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4566 fdb_info->remote_port,
4567 fdb_info->remote_vni,
4568 fdb_info->remote_ifindex);
4572 rdst->offloaded = fdb_info->offloaded;
4575 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4579 vxlan_fdb_external_learn_add(struct net_device *dev,
4580 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4582 struct vxlan_dev *vxlan = netdev_priv(dev);
4583 struct netlink_ext_ack *extack;
4587 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4588 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4590 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4591 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4593 NLM_F_CREATE | NLM_F_REPLACE,
4594 fdb_info->remote_port,
4596 fdb_info->remote_vni,
4597 fdb_info->remote_ifindex,
4598 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4600 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4606 vxlan_fdb_external_learn_del(struct net_device *dev,
4607 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4609 struct vxlan_dev *vxlan = netdev_priv(dev);
4610 struct vxlan_fdb *f;
4614 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4615 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4617 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4620 else if (f->flags & NTF_EXT_LEARNED)
4621 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4622 fdb_info->remote_ip,
4623 fdb_info->remote_port,
4625 fdb_info->remote_vni,
4626 fdb_info->remote_ifindex,
4629 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4634 static int vxlan_switchdev_event(struct notifier_block *unused,
4635 unsigned long event, void *ptr)
4637 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4638 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4642 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4643 vxlan_fdb_offloaded_set(dev, ptr);
4645 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4647 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4649 err = notifier_from_errno(err);
4652 fdb_info->offloaded = true;
4653 vxlan_fdb_offloaded_set(dev, fdb_info);
4655 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4657 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4659 err = notifier_from_errno(err);
4662 fdb_info->offloaded = false;
4663 vxlan_fdb_offloaded_set(dev, fdb_info);
4670 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4671 .notifier_call = vxlan_switchdev_event,
4674 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4676 struct vxlan_fdb *fdb;
4677 struct vxlan_dev *vxlan;
4681 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4682 vxlan = rcu_dereference(fdb->vdev);
4684 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4685 vxlan->default_dst.remote_vni);
4686 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4687 if (!hlist_unhashed(&fdb->hlist))
4688 vxlan_fdb_destroy(vxlan, fdb, false, false);
4689 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4694 static int vxlan_nexthop_event(struct notifier_block *nb,
4695 unsigned long event, void *ptr)
4697 struct nh_notifier_info *info = ptr;
4700 if (event != NEXTHOP_EVENT_DEL)
4703 nh = nexthop_find_by_id(info->net, info->id);
4707 vxlan_fdb_nh_flush(nh);
4712 static __net_init int vxlan_init_net(struct net *net)
4714 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4717 INIT_LIST_HEAD(&vn->vxlan_list);
4718 spin_lock_init(&vn->sock_lock);
4719 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4721 for (h = 0; h < PORT_HASH_SIZE; ++h)
4722 INIT_HLIST_HEAD(&vn->sock_list[h]);
4724 return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4728 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4730 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4731 struct vxlan_dev *vxlan, *next;
4732 struct net_device *dev, *aux;
4734 for_each_netdev_safe(net, dev, aux)
4735 if (dev->rtnl_link_ops == &vxlan_link_ops)
4736 unregister_netdevice_queue(dev, head);
4738 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4739 /* If vxlan->dev is in the same netns, it has already been added
4740 * to the list by the previous loop.
4742 if (!net_eq(dev_net(vxlan->dev), net))
4743 unregister_netdevice_queue(vxlan->dev, head);
4748 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4754 list_for_each_entry(net, net_list, exit_list) {
4755 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4757 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4760 list_for_each_entry(net, net_list, exit_list)
4761 vxlan_destroy_tunnels(net, &list);
4763 unregister_netdevice_many(&list);
4766 list_for_each_entry(net, net_list, exit_list) {
4767 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4769 for (h = 0; h < PORT_HASH_SIZE; ++h)
4770 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4774 static struct pernet_operations vxlan_net_ops = {
4775 .init = vxlan_init_net,
4776 .exit_batch = vxlan_exit_batch_net,
4777 .id = &vxlan_net_id,
4778 .size = sizeof(struct vxlan_net),
4781 static int __init vxlan_init_module(void)
4785 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4787 rc = register_pernet_subsys(&vxlan_net_ops);
4791 rc = register_netdevice_notifier(&vxlan_notifier_block);
4795 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4799 rc = rtnl_link_register(&vxlan_link_ops);
4803 vxlan_vnifilter_init();
4807 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4809 unregister_netdevice_notifier(&vxlan_notifier_block);
4811 unregister_pernet_subsys(&vxlan_net_ops);
4815 late_initcall(vxlan_init_module);
4817 static void __exit vxlan_cleanup_module(void)
4819 vxlan_vnifilter_uninit();
4820 rtnl_link_unregister(&vxlan_link_ops);
4821 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4822 unregister_netdevice_notifier(&vxlan_notifier_block);
4823 unregister_pernet_subsys(&vxlan_net_ops);
4824 /* rcu_barrier() is called by netns */
4826 module_exit(vxlan_cleanup_module);
4828 MODULE_LICENSE("GPL");
4829 MODULE_VERSION(VXLAN_VERSION);
4830 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4831 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4832 MODULE_ALIAS_RTNL_LINK("vxlan");