1 // SPDX-License-Identifier: GPL-2.0-only
3 * VXLAN: Virtual eXtensible Local Area Network
5 * Copyright (c) 2012-2013 Vyatta Inc.
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
19 #include <net/ndisc.h>
20 #include <net/ipv6_stubs.h>
23 #include <net/rtnetlink.h>
24 #include <net/inet_ecn.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27 #include <net/tun_proto.h>
28 #include <net/vxlan.h>
29 #include <net/nexthop.h>
31 #if IS_ENABLED(CONFIG_IPV6)
32 #include <net/ip6_tunnel.h>
33 #include <net/ip6_checksum.h>
36 #define VXLAN_VERSION "0.1"
38 #define PORT_HASH_BITS 8
39 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
40 #define FDB_AGE_DEFAULT 300 /* 5 min */
41 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43 /* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
45 * for compatibility with early adopters.
47 static unsigned short vxlan_port __read_mostly = 8472;
48 module_param_named(udp_port, vxlan_port, ushort, 0444);
49 MODULE_PARM_DESC(udp_port, "Destination UDP port");
51 static bool log_ecn_error = true;
52 module_param(log_ecn_error, bool, 0644);
53 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55 static unsigned int vxlan_net_id;
56 static struct rtnl_link_ops vxlan_link_ops;
58 static const u8 all_zeros_mac[ETH_ALEN + 2];
60 static int vxlan_sock_add(struct vxlan_dev *vxlan);
62 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64 /* per-network namespace private data for this module */
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
71 /* Forwarding table entry */
73 struct hlist_node hlist; /* linked list of entries */
75 unsigned long updated; /* jiffies */
77 struct list_head remotes;
78 u8 eth_addr[ETH_ALEN];
79 u16 state; /* see ndm_state */
81 u16 flags; /* see ndm_flags and below */
82 struct list_head nh_list;
83 struct nexthop __rcu *nh;
84 struct vxlan_dev __rcu *vdev;
87 #define NTF_VXLAN_ADDED_BY_USER 0x100
89 /* salt for hash table */
90 static u32 vxlan_salt __read_mostly;
92 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
94 return vs->flags & VXLAN_F_COLLECT_METADATA ||
95 ip_tunnel_collect_metadata();
98 #if IS_ENABLED(CONFIG_IPV6)
100 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
102 if (a->sa.sa_family != b->sa.sa_family)
104 if (a->sa.sa_family == AF_INET6)
105 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
107 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
110 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
112 if (nla_len(nla) >= sizeof(struct in6_addr)) {
113 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
114 ip->sa.sa_family = AF_INET6;
116 } else if (nla_len(nla) >= sizeof(__be32)) {
117 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
118 ip->sa.sa_family = AF_INET;
121 return -EAFNOSUPPORT;
125 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
126 const union vxlan_addr *ip)
128 if (ip->sa.sa_family == AF_INET6)
129 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
131 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
134 #else /* !CONFIG_IPV6 */
137 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
139 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
142 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
144 if (nla_len(nla) >= sizeof(struct in6_addr)) {
145 return -EAFNOSUPPORT;
146 } else if (nla_len(nla) >= sizeof(__be32)) {
147 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
148 ip->sa.sa_family = AF_INET;
151 return -EAFNOSUPPORT;
155 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
156 const union vxlan_addr *ip)
158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
162 /* Virtual Network hash table head */
163 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
165 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
168 /* Socket hash table head */
169 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
171 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
176 /* First remote destination for a forwarding entry.
177 * Guaranteed to be non-NULL because remotes are never deleted.
179 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
181 if (rcu_access_pointer(fdb->nh))
183 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
186 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
188 if (rcu_access_pointer(fdb->nh))
190 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
193 /* Find VXLAN socket based on network namespace, address family and UDP port
194 * and enabled unshareable flags.
196 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
197 __be16 port, u32 flags, int ifindex)
199 struct vxlan_sock *vs;
201 flags &= VXLAN_F_RCV_FLAGS;
203 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
204 if (inet_sk(vs->sock->sk)->inet_sport == port &&
205 vxlan_get_sk_family(vs) == family &&
206 vs->flags == flags &&
207 vs->sock->sk->sk_bound_dev_if == ifindex)
213 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
216 struct vxlan_dev_node *node;
218 /* For flow based devices, map all packets to VNI 0 */
219 if (vs->flags & VXLAN_F_COLLECT_METADATA)
222 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
223 if (node->vxlan->default_dst.remote_vni != vni)
226 if (IS_ENABLED(CONFIG_IPV6)) {
227 const struct vxlan_config *cfg = &node->vxlan->cfg;
229 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
230 cfg->remote_ifindex != ifindex)
240 /* Look up VNI in a per net namespace table */
241 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
242 __be32 vni, sa_family_t family,
243 __be16 port, u32 flags)
245 struct vxlan_sock *vs;
247 vs = vxlan_find_sock(net, family, port, flags, ifindex);
251 return vxlan_vs_find_vni(vs, ifindex, vni);
254 /* Fill in neighbour message in skbuff. */
255 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
256 const struct vxlan_fdb *fdb,
257 u32 portid, u32 seq, int type, unsigned int flags,
258 const struct vxlan_rdst *rdst)
260 unsigned long now = jiffies;
261 struct nda_cacheinfo ci;
262 bool send_ip, send_eth;
263 struct nlmsghdr *nlh;
269 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
273 ndm = nlmsg_data(nlh);
274 memset(ndm, 0, sizeof(*ndm));
276 send_eth = send_ip = true;
279 nh = rcu_dereference(fdb->nh);
281 nh_family = nexthop_get_family(nh);
286 if (type == RTM_GETNEIGH) {
288 send_ip = !vxlan_addr_any(&rdst->remote_ip);
289 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
291 ndm->ndm_family = nh_family;
293 send_eth = !is_zero_ether_addr(fdb->eth_addr);
295 ndm->ndm_family = AF_BRIDGE;
296 ndm->ndm_state = fdb->state;
297 ndm->ndm_ifindex = vxlan->dev->ifindex;
298 ndm->ndm_flags = fdb->flags;
299 if (rdst && rdst->offloaded)
300 ndm->ndm_flags |= NTF_OFFLOADED;
301 ndm->ndm_type = RTN_UNICAST;
303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
304 nla_put_s32(skb, NDA_LINK_NETNSID,
305 peernet2id(dev_net(vxlan->dev), vxlan->net)))
306 goto nla_put_failure;
308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
309 goto nla_put_failure;
311 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
312 goto nla_put_failure;
314 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
316 goto nla_put_failure;
318 if (rdst->remote_port &&
319 rdst->remote_port != vxlan->cfg.dst_port &&
320 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
321 goto nla_put_failure;
322 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
323 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
324 goto nla_put_failure;
325 if (rdst->remote_ifindex &&
326 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
327 goto nla_put_failure;
330 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
331 nla_put_u32(skb, NDA_SRC_VNI,
332 be32_to_cpu(fdb->vni)))
333 goto nla_put_failure;
335 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
336 ci.ndm_confirmed = 0;
337 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
340 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
341 goto nla_put_failure;
347 nlmsg_cancel(skb, nlh);
351 static inline size_t vxlan_nlmsg_size(void)
353 return NLMSG_ALIGN(sizeof(struct ndmsg))
354 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
355 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
356 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
357 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
358 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
359 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
360 + nla_total_size(sizeof(struct nda_cacheinfo));
363 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
364 struct vxlan_rdst *rd, int type)
366 struct net *net = dev_net(vxlan->dev);
370 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
374 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
376 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
377 WARN_ON(err == -EMSGSIZE);
382 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
386 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
389 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
390 const struct vxlan_fdb *fdb,
391 const struct vxlan_rdst *rd,
392 struct netlink_ext_ack *extack,
393 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
395 fdb_info->info.dev = vxlan->dev;
396 fdb_info->info.extack = extack;
397 fdb_info->remote_ip = rd->remote_ip;
398 fdb_info->remote_port = rd->remote_port;
399 fdb_info->remote_vni = rd->remote_vni;
400 fdb_info->remote_ifindex = rd->remote_ifindex;
401 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
402 fdb_info->vni = fdb->vni;
403 fdb_info->offloaded = rd->offloaded;
404 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
407 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
408 struct vxlan_fdb *fdb,
409 struct vxlan_rdst *rd,
411 struct netlink_ext_ack *extack)
413 struct switchdev_notifier_vxlan_fdb_info info;
414 enum switchdev_notifier_type notifier_type;
420 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
421 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
422 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
423 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
425 return notifier_to_errno(ret);
428 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
429 struct vxlan_rdst *rd, int type, bool swdev_notify,
430 struct netlink_ext_ack *extack)
434 if (swdev_notify && rd) {
437 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
443 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
449 __vxlan_fdb_notify(vxlan, fdb, rd, type);
453 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
455 struct vxlan_dev *vxlan = netdev_priv(dev);
456 struct vxlan_fdb f = {
459 struct vxlan_rdst remote = {
460 .remote_ip = *ipa, /* goes to NDA_DST */
461 .remote_vni = cpu_to_be32(VXLAN_N_VID),
464 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
467 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
469 struct vxlan_fdb f = {
472 struct vxlan_rdst remote = { };
474 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
476 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
479 /* Hash Ethernet address */
480 static u32 eth_hash(const unsigned char *addr)
482 u64 value = get_unaligned((u64 *)addr);
484 /* only want 6 bytes */
490 return hash_64(value, FDB_HASH_BITS);
493 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
495 /* use 1 byte of OUI and 3 bytes of NIC */
496 u32 key = get_unaligned((u32 *)(addr + 2));
498 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
501 static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
503 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
504 return eth_vni_hash(mac, vni);
506 return eth_hash(mac);
509 /* Hash chain to use given mac address */
510 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
511 const u8 *mac, __be32 vni)
513 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
516 /* Look up Ethernet address in forwarding table */
517 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
518 const u8 *mac, __be32 vni)
520 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
523 hlist_for_each_entry_rcu(f, head, hlist) {
524 if (ether_addr_equal(mac, f->eth_addr)) {
525 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
537 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
538 const u8 *mac, __be32 vni)
542 f = __vxlan_find_mac(vxlan, mac, vni);
543 if (f && f->used != jiffies)
549 /* caller should hold vxlan->hash_lock */
550 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
551 union vxlan_addr *ip, __be16 port,
552 __be32 vni, __u32 ifindex)
554 struct vxlan_rdst *rd;
556 list_for_each_entry(rd, &f->remotes, list) {
557 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
558 rd->remote_port == port &&
559 rd->remote_vni == vni &&
560 rd->remote_ifindex == ifindex)
567 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
568 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
570 struct vxlan_dev *vxlan = netdev_priv(dev);
571 u8 eth_addr[ETH_ALEN + 2] = { 0 };
572 struct vxlan_rdst *rdst;
576 if (is_multicast_ether_addr(mac) ||
577 is_zero_ether_addr(mac))
580 ether_addr_copy(eth_addr, mac);
584 f = __vxlan_find_mac(vxlan, eth_addr, vni);
590 rdst = first_remote_rcu(f);
591 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
597 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
599 static int vxlan_fdb_notify_one(struct notifier_block *nb,
600 const struct vxlan_dev *vxlan,
601 const struct vxlan_fdb *f,
602 const struct vxlan_rdst *rdst,
603 struct netlink_ext_ack *extack)
605 struct switchdev_notifier_vxlan_fdb_info fdb_info;
608 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
609 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
611 return notifier_to_errno(rc);
614 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
615 struct notifier_block *nb,
616 struct netlink_ext_ack *extack)
618 struct vxlan_dev *vxlan;
619 struct vxlan_rdst *rdst;
624 if (!netif_is_vxlan(dev))
626 vxlan = netdev_priv(dev);
628 for (h = 0; h < FDB_HASH_SIZE; ++h) {
629 spin_lock_bh(&vxlan->hash_lock[h]);
630 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
632 list_for_each_entry(rdst, &f->remotes, list) {
633 rc = vxlan_fdb_notify_one(nb, vxlan,
641 spin_unlock_bh(&vxlan->hash_lock[h]);
646 spin_unlock_bh(&vxlan->hash_lock[h]);
649 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
651 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
653 struct vxlan_dev *vxlan;
654 struct vxlan_rdst *rdst;
658 if (!netif_is_vxlan(dev))
660 vxlan = netdev_priv(dev);
662 for (h = 0; h < FDB_HASH_SIZE; ++h) {
663 spin_lock_bh(&vxlan->hash_lock[h]);
664 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
666 list_for_each_entry(rdst, &f->remotes, list)
667 rdst->offloaded = false;
668 spin_unlock_bh(&vxlan->hash_lock[h]);
672 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
674 /* Replace destination of unicast mac */
675 static int vxlan_fdb_replace(struct vxlan_fdb *f,
676 union vxlan_addr *ip, __be16 port, __be32 vni,
677 __u32 ifindex, struct vxlan_rdst *oldrd)
679 struct vxlan_rdst *rd;
681 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
685 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
690 dst_cache_reset(&rd->dst_cache);
692 rd->remote_port = port;
693 rd->remote_vni = vni;
694 rd->remote_ifindex = ifindex;
695 rd->offloaded = false;
699 /* Add/update destinations for multicast */
700 static int vxlan_fdb_append(struct vxlan_fdb *f,
701 union vxlan_addr *ip, __be16 port, __be32 vni,
702 __u32 ifindex, struct vxlan_rdst **rdp)
704 struct vxlan_rdst *rd;
706 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
710 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
714 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
720 rd->remote_port = port;
721 rd->offloaded = false;
722 rd->remote_vni = vni;
723 rd->remote_ifindex = ifindex;
725 list_add_tail_rcu(&rd->list, &f->remotes);
731 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
733 struct vxlanhdr *vh, size_t hdrlen,
735 struct gro_remcsum *grc,
738 size_t start, offset;
740 if (skb->remcsum_offload)
743 if (!NAPI_GRO_CB(skb)->csum_valid)
746 start = vxlan_rco_start(vni_field);
747 offset = start + vxlan_rco_offset(vni_field);
749 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
750 start, offset, grc, nopartial);
752 skb->remcsum_offload = 1;
757 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
758 struct list_head *head,
761 struct sk_buff *pp = NULL;
763 struct vxlanhdr *vh, *vh2;
764 unsigned int hlen, off_vx;
766 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
768 struct gro_remcsum grc;
770 skb_gro_remcsum_init(&grc);
772 off_vx = skb_gro_offset(skb);
773 hlen = off_vx + sizeof(*vh);
774 vh = skb_gro_header_fast(skb, off_vx);
775 if (skb_gro_header_hard(skb, hlen)) {
776 vh = skb_gro_header_slow(skb, hlen, off_vx);
781 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
783 flags = vh->vx_flags;
785 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
786 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
789 VXLAN_F_REMCSUM_NOPARTIAL));
795 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
797 list_for_each_entry(p, head, list) {
798 if (!NAPI_GRO_CB(p)->same_flow)
801 vh2 = (struct vxlanhdr *)(p->data + off_vx);
802 if (vh->vx_flags != vh2->vx_flags ||
803 vh->vx_vni != vh2->vx_vni) {
804 NAPI_GRO_CB(p)->same_flow = 0;
809 pp = call_gro_receive(eth_gro_receive, head, skb);
813 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
818 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
820 /* Sets 'skb->inner_mac_header' since we are always called with
821 * 'skb->encapsulation' set.
823 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
826 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
827 __u16 state, __be32 src_vni,
832 f = kmalloc(sizeof(*f), GFP_ATOMIC);
836 f->flags = ndm_flags;
837 f->updated = f->used = jiffies;
840 RCU_INIT_POINTER(f->vdev, vxlan);
841 INIT_LIST_HEAD(&f->nh_list);
842 INIT_LIST_HEAD(&f->remotes);
843 memcpy(f->eth_addr, mac, ETH_ALEN);
848 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
849 __be32 src_vni, struct vxlan_fdb *f)
852 hlist_add_head_rcu(&f->hlist,
853 vxlan_fdb_head(vxlan, mac, src_vni));
856 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
857 u32 nhid, struct netlink_ext_ack *extack)
859 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
860 struct nh_group *nhg;
864 if (old_nh && old_nh->id == nhid)
867 nh = nexthop_find_by_id(vxlan->net, nhid);
869 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
874 if (!nexthop_get(nh)) {
875 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
879 if (!nh->is_fdb_nh) {
880 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
884 nhg = rtnl_dereference(nh->nh_grp);
885 if (!nh->is_group || !nhg->mpath) {
886 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
890 /* check nexthop group family */
891 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
895 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
902 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
909 list_del_rcu(&fdb->nh_list);
912 rcu_assign_pointer(fdb->nh, nh);
913 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
922 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
923 const u8 *mac, union vxlan_addr *ip,
924 __u16 state, __be16 port, __be32 src_vni,
925 __be32 vni, __u32 ifindex, __u16 ndm_flags,
926 u32 nhid, struct vxlan_fdb **fdb,
927 struct netlink_ext_ack *extack)
929 struct vxlan_rdst *rd = NULL;
933 if (vxlan->cfg.addrmax &&
934 vxlan->addrcnt >= vxlan->cfg.addrmax)
937 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
938 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
943 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
945 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
958 static void __vxlan_fdb_free(struct vxlan_fdb *f)
960 struct vxlan_rdst *rd, *nd;
963 nh = rcu_dereference_raw(f->nh);
965 rcu_assign_pointer(f->nh, NULL);
966 rcu_assign_pointer(f->vdev, NULL);
970 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
971 dst_cache_destroy(&rd->dst_cache);
977 static void vxlan_fdb_free(struct rcu_head *head)
979 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
984 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
985 bool do_notify, bool swdev_notify)
987 struct vxlan_rdst *rd;
989 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
993 if (rcu_access_pointer(f->nh))
994 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
997 list_for_each_entry(rd, &f->remotes, list)
998 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
1002 hlist_del_rcu(&f->hlist);
1003 list_del_rcu(&f->nh_list);
1004 call_rcu(&f->rcu, vxlan_fdb_free);
1007 static void vxlan_dst_free(struct rcu_head *head)
1009 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1011 dst_cache_destroy(&rd->dst_cache);
1015 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1016 union vxlan_addr *ip,
1017 __u16 state, __u16 flags,
1018 __be16 port, __be32 vni,
1019 __u32 ifindex, __u16 ndm_flags,
1020 struct vxlan_fdb *f, u32 nhid,
1022 struct netlink_ext_ack *extack)
1024 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1025 struct vxlan_rdst *rd = NULL;
1026 struct vxlan_rdst oldrd;
1031 if (nhid && !rcu_access_pointer(f->nh)) {
1032 NL_SET_ERR_MSG(extack,
1033 "Cannot replace an existing non nexthop fdb with a nexthop");
1037 if (nhid && (flags & NLM_F_APPEND)) {
1038 NL_SET_ERR_MSG(extack,
1039 "Cannot append to a nexthop fdb");
1043 /* Do not allow an externally learned entry to take over an entry added
1046 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1047 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1048 if (f->state != state) {
1050 f->updated = jiffies;
1053 if (f->flags != fdb_flags) {
1054 f->flags = fdb_flags;
1055 f->updated = jiffies;
1060 if ((flags & NLM_F_REPLACE)) {
1061 /* Only change unicasts */
1062 if (!(is_multicast_ether_addr(f->eth_addr) ||
1063 is_zero_ether_addr(f->eth_addr))) {
1065 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1069 rc = vxlan_fdb_replace(f, ip, port, vni,
1074 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1078 if ((flags & NLM_F_APPEND) &&
1079 (is_multicast_ether_addr(f->eth_addr) ||
1080 is_zero_ether_addr(f->eth_addr))) {
1081 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1088 if (ndm_flags & NTF_USE)
1093 rd = first_remote_rtnl(f);
1095 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1096 swdev_notify, extack);
1106 if ((flags & NLM_F_REPLACE) && rc)
1108 else if ((flags & NLM_F_APPEND) && rc) {
1109 list_del_rcu(&rd->list);
1110 call_rcu(&rd->rcu, vxlan_dst_free);
1115 static int vxlan_fdb_update_create(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 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1124 struct vxlan_fdb *f;
1127 /* Disallow replace to add a multicast entry */
1128 if ((flags & NLM_F_REPLACE) &&
1129 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1132 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1133 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1134 vni, ifindex, fdb_flags, nhid, &f, extack);
1138 vxlan_fdb_insert(vxlan, mac, src_vni, f);
1139 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1140 swdev_notify, extack);
1147 vxlan_fdb_destroy(vxlan, f, false, false);
1151 /* Add new entry to forwarding table -- assumes lock held */
1152 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1153 const u8 *mac, union vxlan_addr *ip,
1154 __u16 state, __u16 flags,
1155 __be16 port, __be32 src_vni, __be32 vni,
1156 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1158 struct netlink_ext_ack *extack)
1160 struct vxlan_fdb *f;
1162 f = __vxlan_find_mac(vxlan, mac, src_vni);
1164 if (flags & NLM_F_EXCL) {
1165 netdev_dbg(vxlan->dev,
1166 "lost race to create %pM\n", mac);
1170 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1171 vni, ifindex, ndm_flags, f,
1172 nhid, swdev_notify, extack);
1174 if (!(flags & NLM_F_CREATE))
1177 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1178 port, src_vni, vni, ifindex,
1179 ndm_flags, nhid, swdev_notify,
1184 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1185 struct vxlan_rdst *rd, bool swdev_notify)
1187 list_del_rcu(&rd->list);
1188 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1189 call_rcu(&rd->rcu, vxlan_dst_free);
1192 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1193 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1194 __be32 *vni, u32 *ifindex, u32 *nhid)
1196 struct net *net = dev_net(vxlan->dev);
1199 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1204 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1208 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1210 if (remote->sa.sa_family == AF_INET) {
1211 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1212 ip->sa.sa_family = AF_INET;
1213 #if IS_ENABLED(CONFIG_IPV6)
1215 ip->sin6.sin6_addr = in6addr_any;
1216 ip->sa.sa_family = AF_INET6;
1222 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1224 *port = nla_get_be16(tb[NDA_PORT]);
1226 *port = vxlan->cfg.dst_port;
1230 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1232 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1234 *vni = vxlan->default_dst.remote_vni;
1237 if (tb[NDA_SRC_VNI]) {
1238 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1240 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1242 *src_vni = vxlan->default_dst.remote_vni;
1245 if (tb[NDA_IFINDEX]) {
1246 struct net_device *tdev;
1248 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1250 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1251 tdev = __dev_get_by_index(net, *ifindex);
1253 return -EADDRNOTAVAIL;
1259 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1266 /* Add static entry (via netlink) */
1267 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1268 struct net_device *dev,
1269 const unsigned char *addr, u16 vid, u16 flags,
1270 struct netlink_ext_ack *extack)
1272 struct vxlan_dev *vxlan = netdev_priv(dev);
1273 /* struct net *net = dev_net(vxlan->dev); */
1274 union vxlan_addr ip;
1276 __be32 src_vni, vni;
1281 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1282 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1287 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1290 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1295 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1296 return -EAFNOSUPPORT;
1298 hash_index = fdb_head_index(vxlan, addr, src_vni);
1299 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1300 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1301 port, src_vni, vni, ifindex,
1302 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1303 nhid, true, extack);
1304 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1309 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1310 const unsigned char *addr, union vxlan_addr ip,
1311 __be16 port, __be32 src_vni, __be32 vni,
1312 u32 ifindex, bool swdev_notify)
1314 struct vxlan_rdst *rd = NULL;
1315 struct vxlan_fdb *f;
1318 f = vxlan_find_mac(vxlan, addr, src_vni);
1322 if (!vxlan_addr_any(&ip)) {
1323 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1328 /* remove a destination if it's not the only one on the list,
1329 * otherwise destroy the fdb entry
1331 if (rd && !list_is_singular(&f->remotes)) {
1332 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1336 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1342 /* Delete entry (via netlink) */
1343 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1344 struct net_device *dev,
1345 const unsigned char *addr, u16 vid)
1347 struct vxlan_dev *vxlan = netdev_priv(dev);
1348 union vxlan_addr ip;
1349 __be32 src_vni, vni;
1355 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1360 hash_index = fdb_head_index(vxlan, addr, src_vni);
1361 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1362 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1364 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1369 /* Dump forwarding table */
1370 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1371 struct net_device *dev,
1372 struct net_device *filter_dev, int *idx)
1374 struct vxlan_dev *vxlan = netdev_priv(dev);
1378 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1379 struct vxlan_fdb *f;
1381 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1382 struct vxlan_rdst *rd;
1384 if (rcu_access_pointer(f->nh)) {
1385 err = vxlan_fdb_info(skb, vxlan, f,
1386 NETLINK_CB(cb->skb).portid,
1395 list_for_each_entry_rcu(rd, &f->remotes, list) {
1396 if (*idx < cb->args[2])
1399 err = vxlan_fdb_info(skb, vxlan, f,
1400 NETLINK_CB(cb->skb).portid,
1415 static int vxlan_fdb_get(struct sk_buff *skb,
1416 struct nlattr *tb[],
1417 struct net_device *dev,
1418 const unsigned char *addr,
1419 u16 vid, u32 portid, u32 seq,
1420 struct netlink_ext_ack *extack)
1422 struct vxlan_dev *vxlan = netdev_priv(dev);
1423 struct vxlan_fdb *f;
1428 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1430 vni = vxlan->default_dst.remote_vni;
1434 f = __vxlan_find_mac(vxlan, addr, vni);
1436 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1441 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1442 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1448 /* Watch incoming packets to learn mapping between Ethernet address
1449 * and Tunnel endpoint.
1450 * Return true if packet is bogus and should be dropped.
1452 static bool vxlan_snoop(struct net_device *dev,
1453 union vxlan_addr *src_ip, const u8 *src_mac,
1454 u32 src_ifindex, __be32 vni)
1456 struct vxlan_dev *vxlan = netdev_priv(dev);
1457 struct vxlan_fdb *f;
1460 #if IS_ENABLED(CONFIG_IPV6)
1461 if (src_ip->sa.sa_family == AF_INET6 &&
1462 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1463 ifindex = src_ifindex;
1466 f = vxlan_find_mac(vxlan, src_mac, vni);
1468 struct vxlan_rdst *rdst = first_remote_rcu(f);
1470 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1471 rdst->remote_ifindex == ifindex))
1474 /* Don't migrate static entries, drop packets */
1475 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1478 /* Don't override an fdb with nexthop with a learnt entry */
1479 if (rcu_access_pointer(f->nh))
1482 if (net_ratelimit())
1484 "%pM migrated from %pIS to %pIS\n",
1485 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1487 rdst->remote_ip = *src_ip;
1488 f->updated = jiffies;
1489 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1491 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1493 /* learned new entry */
1494 spin_lock(&vxlan->hash_lock[hash_index]);
1496 /* close off race between vxlan_flush and incoming packets */
1497 if (netif_running(dev))
1498 vxlan_fdb_update(vxlan, src_mac, src_ip,
1500 NLM_F_EXCL|NLM_F_CREATE,
1501 vxlan->cfg.dst_port,
1503 vxlan->default_dst.remote_vni,
1504 ifindex, NTF_SELF, 0, true, NULL);
1505 spin_unlock(&vxlan->hash_lock[hash_index]);
1511 /* See if multicast group is already in use by other ID */
1512 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1514 struct vxlan_dev *vxlan;
1515 struct vxlan_sock *sock4;
1516 #if IS_ENABLED(CONFIG_IPV6)
1517 struct vxlan_sock *sock6;
1519 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1521 sock4 = rtnl_dereference(dev->vn4_sock);
1523 /* The vxlan_sock is only used by dev, leaving group has
1524 * no effect on other vxlan devices.
1526 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1528 #if IS_ENABLED(CONFIG_IPV6)
1529 sock6 = rtnl_dereference(dev->vn6_sock);
1530 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1534 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1535 if (!netif_running(vxlan->dev) || vxlan == dev)
1538 if (family == AF_INET &&
1539 rtnl_dereference(vxlan->vn4_sock) != sock4)
1541 #if IS_ENABLED(CONFIG_IPV6)
1542 if (family == AF_INET6 &&
1543 rtnl_dereference(vxlan->vn6_sock) != sock6)
1547 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1548 &dev->default_dst.remote_ip))
1551 if (vxlan->default_dst.remote_ifindex !=
1552 dev->default_dst.remote_ifindex)
1561 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1563 struct vxlan_net *vn;
1567 if (!refcount_dec_and_test(&vs->refcnt))
1570 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1571 spin_lock(&vn->sock_lock);
1572 hlist_del_rcu(&vs->hlist);
1573 udp_tunnel_notify_del_rx_port(vs->sock,
1574 (vs->flags & VXLAN_F_GPE) ?
1575 UDP_TUNNEL_TYPE_VXLAN_GPE :
1576 UDP_TUNNEL_TYPE_VXLAN);
1577 spin_unlock(&vn->sock_lock);
1582 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1584 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1585 #if IS_ENABLED(CONFIG_IPV6)
1586 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1588 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1591 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1594 vxlan_vs_del_dev(vxlan);
1596 if (__vxlan_sock_release_prep(sock4)) {
1597 udp_tunnel_sock_release(sock4->sock);
1601 #if IS_ENABLED(CONFIG_IPV6)
1602 if (__vxlan_sock_release_prep(sock6)) {
1603 udp_tunnel_sock_release(sock6->sock);
1609 /* Update multicast group membership when first VNI on
1610 * multicast address is brought up
1612 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1615 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1616 int ifindex = vxlan->default_dst.remote_ifindex;
1619 if (ip->sa.sa_family == AF_INET) {
1620 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1621 struct ip_mreqn mreq = {
1622 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1623 .imr_ifindex = ifindex,
1626 sk = sock4->sock->sk;
1628 ret = ip_mc_join_group(sk, &mreq);
1630 #if IS_ENABLED(CONFIG_IPV6)
1632 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1634 sk = sock6->sock->sk;
1636 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1637 &ip->sin6.sin6_addr);
1645 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1646 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1649 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1650 int ifindex = vxlan->default_dst.remote_ifindex;
1653 if (ip->sa.sa_family == AF_INET) {
1654 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1655 struct ip_mreqn mreq = {
1656 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1657 .imr_ifindex = ifindex,
1660 sk = sock4->sock->sk;
1662 ret = ip_mc_leave_group(sk, &mreq);
1664 #if IS_ENABLED(CONFIG_IPV6)
1666 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1668 sk = sock6->sock->sk;
1670 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1671 &ip->sin6.sin6_addr);
1679 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1680 struct sk_buff *skb, u32 vxflags)
1682 size_t start, offset;
1684 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1687 start = vxlan_rco_start(unparsed->vx_vni);
1688 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1690 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1693 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1694 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1696 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1697 unparsed->vx_vni &= VXLAN_VNI_MASK;
1701 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1702 struct sk_buff *skb, u32 vxflags,
1703 struct vxlan_metadata *md)
1705 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1706 struct metadata_dst *tun_dst;
1708 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1711 md->gbp = ntohs(gbp->policy_id);
1713 tun_dst = (struct metadata_dst *)skb_dst(skb);
1715 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1716 tun_dst->u.tun_info.options_len = sizeof(*md);
1718 if (gbp->dont_learn)
1719 md->gbp |= VXLAN_GBP_DONT_LEARN;
1721 if (gbp->policy_applied)
1722 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1724 /* In flow-based mode, GBP is carried in dst_metadata */
1725 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1726 skb->mark = md->gbp;
1728 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1731 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1733 struct sk_buff *skb, u32 vxflags)
1735 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1737 /* Need to have Next Protocol set for interfaces in GPE mode. */
1738 if (!gpe->np_applied)
1740 /* "The initial version is 0. If a receiver does not support the
1741 * version indicated it MUST drop the packet.
1743 if (gpe->version != 0)
1745 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1746 * processing MUST occur." However, we don't implement OAM
1747 * processing, thus drop the packet.
1752 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1756 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1760 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1761 struct vxlan_sock *vs,
1762 struct sk_buff *skb, __be32 vni)
1764 union vxlan_addr saddr;
1765 u32 ifindex = skb->dev->ifindex;
1767 skb_reset_mac_header(skb);
1768 skb->protocol = eth_type_trans(skb, vxlan->dev);
1769 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1771 /* Ignore packet loops (and multicast echo) */
1772 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1775 /* Get address from the outer IP header */
1776 if (vxlan_get_sk_family(vs) == AF_INET) {
1777 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1778 saddr.sa.sa_family = AF_INET;
1779 #if IS_ENABLED(CONFIG_IPV6)
1781 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1782 saddr.sa.sa_family = AF_INET6;
1786 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1787 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1793 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1794 struct sk_buff *skb)
1798 if (vxlan_get_sk_family(vs) == AF_INET)
1799 err = IP_ECN_decapsulate(oiph, skb);
1800 #if IS_ENABLED(CONFIG_IPV6)
1802 err = IP6_ECN_decapsulate(oiph, skb);
1805 if (unlikely(err) && log_ecn_error) {
1806 if (vxlan_get_sk_family(vs) == AF_INET)
1807 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1808 &((struct iphdr *)oiph)->saddr,
1809 ((struct iphdr *)oiph)->tos);
1811 net_info_ratelimited("non-ECT from %pI6\n",
1812 &((struct ipv6hdr *)oiph)->saddr);
1817 /* Callback from net/ipv4/udp.c to receive packets */
1818 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1820 struct pcpu_sw_netstats *stats;
1821 struct vxlan_dev *vxlan;
1822 struct vxlan_sock *vs;
1823 struct vxlanhdr unparsed;
1824 struct vxlan_metadata _md;
1825 struct vxlan_metadata *md = &_md;
1826 __be16 protocol = htons(ETH_P_TEB);
1827 bool raw_proto = false;
1831 /* Need UDP and VXLAN header to be present */
1832 if (!pskb_may_pull(skb, VXLAN_HLEN))
1835 unparsed = *vxlan_hdr(skb);
1836 /* VNI flag always required to be set */
1837 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1838 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1839 ntohl(vxlan_hdr(skb)->vx_flags),
1840 ntohl(vxlan_hdr(skb)->vx_vni));
1841 /* Return non vxlan pkt */
1844 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1845 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1847 vs = rcu_dereference_sk_user_data(sk);
1851 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1853 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1857 /* For backwards compatibility, only allow reserved fields to be
1858 * used by VXLAN extensions if explicitly requested.
1860 if (vs->flags & VXLAN_F_GPE) {
1861 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1866 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1867 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1870 if (vxlan_collect_metadata(vs)) {
1871 struct metadata_dst *tun_dst;
1873 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1874 key32_to_tunnel_id(vni), sizeof(*md));
1879 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1881 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1883 memset(md, 0, sizeof(*md));
1886 if (vs->flags & VXLAN_F_REMCSUM_RX)
1887 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1889 if (vs->flags & VXLAN_F_GBP)
1890 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1891 /* Note that GBP and GPE can never be active together. This is
1892 * ensured in vxlan_dev_configure.
1895 if (unparsed.vx_flags || unparsed.vx_vni) {
1896 /* If there are any unprocessed flags remaining treat
1897 * this as a malformed packet. This behavior diverges from
1898 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1899 * in reserved fields are to be ignored. The approach here
1900 * maintains compatibility with previous stack code, and also
1901 * is more robust and provides a little more security in
1902 * adding extensions to VXLAN.
1908 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1911 skb_reset_mac_header(skb);
1912 skb->dev = vxlan->dev;
1913 skb->pkt_type = PACKET_HOST;
1916 oiph = skb_network_header(skb);
1917 skb_reset_network_header(skb);
1919 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1920 ++vxlan->dev->stats.rx_frame_errors;
1921 ++vxlan->dev->stats.rx_errors;
1927 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1929 atomic_long_inc(&vxlan->dev->rx_dropped);
1933 stats = this_cpu_ptr(vxlan->dev->tstats);
1934 u64_stats_update_begin(&stats->syncp);
1935 stats->rx_packets++;
1936 stats->rx_bytes += skb->len;
1937 u64_stats_update_end(&stats->syncp);
1939 gro_cells_receive(&vxlan->gro_cells, skb);
1946 /* Consume bad packet */
1951 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1952 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1954 struct vxlan_dev *vxlan;
1955 struct vxlan_sock *vs;
1956 struct vxlanhdr *hdr;
1959 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1962 hdr = vxlan_hdr(skb);
1964 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1967 vs = rcu_dereference_sk_user_data(sk);
1971 vni = vxlan_vni(hdr->vx_vni);
1972 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1979 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1981 struct vxlan_dev *vxlan = netdev_priv(dev);
1982 struct arphdr *parp;
1985 struct neighbour *n;
1987 if (dev->flags & IFF_NOARP)
1990 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1991 dev->stats.tx_dropped++;
1994 parp = arp_hdr(skb);
1996 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1997 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1998 parp->ar_pro != htons(ETH_P_IP) ||
1999 parp->ar_op != htons(ARPOP_REQUEST) ||
2000 parp->ar_hln != dev->addr_len ||
2003 arpptr = (u8 *)parp + sizeof(struct arphdr);
2005 arpptr += dev->addr_len; /* sha */
2006 memcpy(&sip, arpptr, sizeof(sip));
2007 arpptr += sizeof(sip);
2008 arpptr += dev->addr_len; /* tha */
2009 memcpy(&tip, arpptr, sizeof(tip));
2011 if (ipv4_is_loopback(tip) ||
2012 ipv4_is_multicast(tip))
2015 n = neigh_lookup(&arp_tbl, &tip, dev);
2018 struct vxlan_fdb *f;
2019 struct sk_buff *reply;
2021 if (!(n->nud_state & NUD_CONNECTED)) {
2026 f = vxlan_find_mac(vxlan, n->ha, vni);
2027 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2028 /* bridge-local neighbor */
2033 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2041 skb_reset_mac_header(reply);
2042 __skb_pull(reply, skb_network_offset(reply));
2043 reply->ip_summed = CHECKSUM_UNNECESSARY;
2044 reply->pkt_type = PACKET_HOST;
2046 if (netif_rx_ni(reply) == NET_RX_DROP)
2047 dev->stats.rx_dropped++;
2048 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2049 union vxlan_addr ipa = {
2050 .sin.sin_addr.s_addr = tip,
2051 .sin.sin_family = AF_INET,
2054 vxlan_ip_miss(dev, &ipa);
2058 return NETDEV_TX_OK;
2061 #if IS_ENABLED(CONFIG_IPV6)
2062 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2063 struct neighbour *n, bool isrouter)
2065 struct net_device *dev = request->dev;
2066 struct sk_buff *reply;
2067 struct nd_msg *ns, *na;
2068 struct ipv6hdr *pip6;
2070 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2074 if (dev == NULL || !pskb_may_pull(request, request->len))
2077 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2078 sizeof(*na) + na_olen + dev->needed_tailroom;
2079 reply = alloc_skb(len, GFP_ATOMIC);
2083 reply->protocol = htons(ETH_P_IPV6);
2085 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2086 skb_push(reply, sizeof(struct ethhdr));
2087 skb_reset_mac_header(reply);
2089 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2091 daddr = eth_hdr(request)->h_source;
2092 ns_olen = request->len - skb_network_offset(request) -
2093 sizeof(struct ipv6hdr) - sizeof(*ns);
2094 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2095 if (!ns->opt[i + 1]) {
2099 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2100 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2105 /* Ethernet header */
2106 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2107 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2108 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2109 reply->protocol = htons(ETH_P_IPV6);
2111 skb_pull(reply, sizeof(struct ethhdr));
2112 skb_reset_network_header(reply);
2113 skb_put(reply, sizeof(struct ipv6hdr));
2117 pip6 = ipv6_hdr(reply);
2118 memset(pip6, 0, sizeof(struct ipv6hdr));
2120 pip6->priority = ipv6_hdr(request)->priority;
2121 pip6->nexthdr = IPPROTO_ICMPV6;
2122 pip6->hop_limit = 255;
2123 pip6->daddr = ipv6_hdr(request)->saddr;
2124 pip6->saddr = *(struct in6_addr *)n->primary_key;
2126 skb_pull(reply, sizeof(struct ipv6hdr));
2127 skb_reset_transport_header(reply);
2129 /* Neighbor Advertisement */
2130 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2131 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2132 na->icmph.icmp6_router = isrouter;
2133 na->icmph.icmp6_override = 1;
2134 na->icmph.icmp6_solicited = 1;
2135 na->target = ns->target;
2136 ether_addr_copy(&na->opt[2], n->ha);
2137 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2138 na->opt[1] = na_olen >> 3;
2140 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2141 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2142 csum_partial(na, sizeof(*na)+na_olen, 0));
2144 pip6->payload_len = htons(sizeof(*na)+na_olen);
2146 skb_push(reply, sizeof(struct ipv6hdr));
2148 reply->ip_summed = CHECKSUM_UNNECESSARY;
2153 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2155 struct vxlan_dev *vxlan = netdev_priv(dev);
2156 const struct in6_addr *daddr;
2157 const struct ipv6hdr *iphdr;
2158 struct inet6_dev *in6_dev;
2159 struct neighbour *n;
2162 in6_dev = __in6_dev_get(dev);
2166 iphdr = ipv6_hdr(skb);
2167 daddr = &iphdr->daddr;
2168 msg = (struct nd_msg *)(iphdr + 1);
2170 if (ipv6_addr_loopback(daddr) ||
2171 ipv6_addr_is_multicast(&msg->target))
2174 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2177 struct vxlan_fdb *f;
2178 struct sk_buff *reply;
2180 if (!(n->nud_state & NUD_CONNECTED)) {
2185 f = vxlan_find_mac(vxlan, n->ha, vni);
2186 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2187 /* bridge-local neighbor */
2192 reply = vxlan_na_create(skb, n,
2193 !!(f ? f->flags & NTF_ROUTER : 0));
2200 if (netif_rx_ni(reply) == NET_RX_DROP)
2201 dev->stats.rx_dropped++;
2203 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2204 union vxlan_addr ipa = {
2205 .sin6.sin6_addr = msg->target,
2206 .sin6.sin6_family = AF_INET6,
2209 vxlan_ip_miss(dev, &ipa);
2214 return NETDEV_TX_OK;
2218 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2220 struct vxlan_dev *vxlan = netdev_priv(dev);
2221 struct neighbour *n;
2223 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2227 switch (ntohs(eth_hdr(skb)->h_proto)) {
2232 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2235 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2236 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2237 union vxlan_addr ipa = {
2238 .sin.sin_addr.s_addr = pip->daddr,
2239 .sin.sin_family = AF_INET,
2242 vxlan_ip_miss(dev, &ipa);
2248 #if IS_ENABLED(CONFIG_IPV6)
2251 struct ipv6hdr *pip6;
2253 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2255 pip6 = ipv6_hdr(skb);
2256 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2257 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2258 union vxlan_addr ipa = {
2259 .sin6.sin6_addr = pip6->daddr,
2260 .sin6.sin6_family = AF_INET6,
2263 vxlan_ip_miss(dev, &ipa);
2277 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2279 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2281 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2290 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2291 struct vxlan_metadata *md)
2293 struct vxlanhdr_gbp *gbp;
2298 gbp = (struct vxlanhdr_gbp *)vxh;
2299 vxh->vx_flags |= VXLAN_HF_GBP;
2301 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2302 gbp->dont_learn = 1;
2304 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2305 gbp->policy_applied = 1;
2307 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2310 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2313 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2315 gpe->np_applied = 1;
2316 gpe->next_protocol = tun_p_from_eth_p(protocol);
2317 if (!gpe->next_protocol)
2318 return -EPFNOSUPPORT;
2322 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2323 int iphdr_len, __be32 vni,
2324 struct vxlan_metadata *md, u32 vxflags,
2327 struct vxlanhdr *vxh;
2330 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2331 __be16 inner_protocol = htons(ETH_P_TEB);
2333 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2334 skb->ip_summed == CHECKSUM_PARTIAL) {
2335 int csum_start = skb_checksum_start_offset(skb);
2337 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2338 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2339 (skb->csum_offset == offsetof(struct udphdr, check) ||
2340 skb->csum_offset == offsetof(struct tcphdr, check)))
2341 type |= SKB_GSO_TUNNEL_REMCSUM;
2344 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2345 + VXLAN_HLEN + iphdr_len;
2347 /* Need space for new headers (invalidates iph ptr) */
2348 err = skb_cow_head(skb, min_headroom);
2352 err = iptunnel_handle_offloads(skb, type);
2356 vxh = __skb_push(skb, sizeof(*vxh));
2357 vxh->vx_flags = VXLAN_HF_VNI;
2358 vxh->vx_vni = vxlan_vni_field(vni);
2360 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2363 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2364 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2365 vxh->vx_flags |= VXLAN_HF_RCO;
2367 if (!skb_is_gso(skb)) {
2368 skb->ip_summed = CHECKSUM_NONE;
2369 skb->encapsulation = 0;
2373 if (vxflags & VXLAN_F_GBP)
2374 vxlan_build_gbp_hdr(vxh, vxflags, md);
2375 if (vxflags & VXLAN_F_GPE) {
2376 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2379 inner_protocol = skb->protocol;
2382 skb_set_inner_protocol(skb, inner_protocol);
2386 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2387 struct vxlan_sock *sock4,
2388 struct sk_buff *skb, int oif, u8 tos,
2389 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2390 struct dst_cache *dst_cache,
2391 const struct ip_tunnel_info *info)
2393 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2394 struct rtable *rt = NULL;
2398 return ERR_PTR(-EIO);
2403 rt = dst_cache_get_ip4(dst_cache, saddr);
2408 memset(&fl4, 0, sizeof(fl4));
2409 fl4.flowi4_oif = oif;
2410 fl4.flowi4_tos = RT_TOS(tos);
2411 fl4.flowi4_mark = skb->mark;
2412 fl4.flowi4_proto = IPPROTO_UDP;
2415 fl4.fl4_dport = dport;
2416 fl4.fl4_sport = sport;
2418 rt = ip_route_output_key(vxlan->net, &fl4);
2420 if (rt->dst.dev == dev) {
2421 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2423 return ERR_PTR(-ELOOP);
2428 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2430 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2431 return ERR_PTR(-ENETUNREACH);
2436 #if IS_ENABLED(CONFIG_IPV6)
2437 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2438 struct net_device *dev,
2439 struct vxlan_sock *sock6,
2440 struct sk_buff *skb, int oif, u8 tos,
2442 const struct in6_addr *daddr,
2443 struct in6_addr *saddr,
2444 __be16 dport, __be16 sport,
2445 struct dst_cache *dst_cache,
2446 const struct ip_tunnel_info *info)
2448 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2449 struct dst_entry *ndst;
2453 return ERR_PTR(-EIO);
2458 ndst = dst_cache_get_ip6(dst_cache, saddr);
2463 memset(&fl6, 0, sizeof(fl6));
2464 fl6.flowi6_oif = oif;
2467 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2468 fl6.flowi6_mark = skb->mark;
2469 fl6.flowi6_proto = IPPROTO_UDP;
2470 fl6.fl6_dport = dport;
2471 fl6.fl6_sport = sport;
2473 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2475 if (unlikely(IS_ERR(ndst))) {
2476 netdev_dbg(dev, "no route to %pI6\n", daddr);
2477 return ERR_PTR(-ENETUNREACH);
2480 if (unlikely(ndst->dev == dev)) {
2481 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2483 return ERR_PTR(-ELOOP);
2488 dst_cache_set_ip6(dst_cache, ndst, saddr);
2493 /* Bypass encapsulation if the destination is local */
2494 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2495 struct vxlan_dev *dst_vxlan, __be32 vni)
2497 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2498 union vxlan_addr loopback;
2499 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2500 struct net_device *dev;
2503 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2504 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2505 skb->pkt_type = PACKET_HOST;
2506 skb->encapsulation = 0;
2507 skb->dev = dst_vxlan->dev;
2508 __skb_pull(skb, skb_network_offset(skb));
2510 if (remote_ip->sa.sa_family == AF_INET) {
2511 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2512 loopback.sa.sa_family = AF_INET;
2513 #if IS_ENABLED(CONFIG_IPV6)
2515 loopback.sin6.sin6_addr = in6addr_loopback;
2516 loopback.sa.sa_family = AF_INET6;
2522 if (unlikely(!(dev->flags & IFF_UP))) {
2527 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2528 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2530 u64_stats_update_begin(&tx_stats->syncp);
2531 tx_stats->tx_packets++;
2532 tx_stats->tx_bytes += len;
2533 u64_stats_update_end(&tx_stats->syncp);
2535 if (netif_rx(skb) == NET_RX_SUCCESS) {
2536 u64_stats_update_begin(&rx_stats->syncp);
2537 rx_stats->rx_packets++;
2538 rx_stats->rx_bytes += len;
2539 u64_stats_update_end(&rx_stats->syncp);
2542 dev->stats.rx_dropped++;
2547 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2548 struct vxlan_dev *vxlan,
2549 union vxlan_addr *daddr,
2550 __be16 dst_port, int dst_ifindex, __be32 vni,
2551 struct dst_entry *dst,
2554 #if IS_ENABLED(CONFIG_IPV6)
2555 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2556 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2557 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2559 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2561 /* Bypass encapsulation if the destination is local */
2562 if (rt_flags & RTCF_LOCAL &&
2563 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2564 struct vxlan_dev *dst_vxlan;
2567 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2568 daddr->sa.sa_family, dst_port,
2571 dev->stats.tx_errors++;
2576 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2583 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2584 __be32 default_vni, struct vxlan_rdst *rdst,
2587 struct dst_cache *dst_cache;
2588 struct ip_tunnel_info *info;
2589 struct vxlan_dev *vxlan = netdev_priv(dev);
2590 const struct iphdr *old_iph = ip_hdr(skb);
2591 union vxlan_addr *dst;
2592 union vxlan_addr remote_ip, local_ip;
2593 struct vxlan_metadata _md;
2594 struct vxlan_metadata *md = &_md;
2595 __be16 src_port = 0, dst_port;
2596 struct dst_entry *ndst = NULL;
2601 u32 flags = vxlan->cfg.flags;
2602 bool udp_sum = false;
2603 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2605 info = skb_tunnel_info(skb);
2608 dst = &rdst->remote_ip;
2609 if (vxlan_addr_any(dst)) {
2611 /* short-circuited back to local bridge */
2612 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2618 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2619 vni = (rdst->remote_vni) ? : default_vni;
2620 ifindex = rdst->remote_ifindex;
2621 local_ip = vxlan->cfg.saddr;
2622 dst_cache = &rdst->dst_cache;
2623 md->gbp = skb->mark;
2624 if (flags & VXLAN_F_TTL_INHERIT) {
2625 ttl = ip_tunnel_get_ttl(old_iph, skb);
2627 ttl = vxlan->cfg.ttl;
2628 if (!ttl && vxlan_addr_multicast(dst))
2632 tos = vxlan->cfg.tos;
2634 tos = ip_tunnel_get_dsfield(old_iph, skb);
2636 if (dst->sa.sa_family == AF_INET)
2637 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2639 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2640 label = vxlan->cfg.label;
2643 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2647 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2648 if (remote_ip.sa.sa_family == AF_INET) {
2649 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2650 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2652 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2653 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2656 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2657 vni = tunnel_id_to_key32(info->key.tun_id);
2659 dst_cache = &info->dst_cache;
2660 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2661 if (info->options_len < sizeof(*md))
2663 md = ip_tunnel_info_opts(info);
2665 ttl = info->key.ttl;
2666 tos = info->key.tos;
2667 label = info->key.label;
2668 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2670 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2671 vxlan->cfg.port_max, true);
2674 if (dst->sa.sa_family == AF_INET) {
2675 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2680 ifindex = sock4->sock->sk->sk_bound_dev_if;
2682 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2683 dst->sin.sin_addr.s_addr,
2684 &local_ip.sin.sin_addr.s_addr,
2693 /* Bypass encapsulation if the destination is local */
2694 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2695 dst_port, ifindex, vni,
2696 &rt->dst, rt->rt_flags);
2700 if (vxlan->cfg.df == VXLAN_DF_SET) {
2702 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2703 struct ethhdr *eth = eth_hdr(skb);
2705 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2706 (ntohs(eth->h_proto) == ETH_P_IP &&
2707 old_iph->frag_off & htons(IP_DF)))
2710 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2715 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
2717 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
2718 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2719 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2720 vni, md, flags, udp_sum);
2724 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2725 dst->sin.sin_addr.s_addr, tos, ttl, df,
2726 src_port, dst_port, xnet, !udp_sum);
2727 #if IS_ENABLED(CONFIG_IPV6)
2729 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2732 ifindex = sock6->sock->sk->sk_bound_dev_if;
2734 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2735 label, &dst->sin6.sin6_addr,
2736 &local_ip.sin6.sin6_addr,
2740 err = PTR_ERR(ndst);
2746 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2748 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2749 dst_port, ifindex, vni,
2755 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
2757 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
2758 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2759 skb_scrub_packet(skb, xnet);
2760 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2761 vni, md, flags, udp_sum);
2765 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2766 &local_ip.sin6.sin6_addr,
2767 &dst->sin6.sin6_addr, tos, ttl,
2768 label, src_port, dst_port, !udp_sum);
2776 dev->stats.tx_dropped++;
2783 dev->stats.collisions++;
2784 else if (err == -ENETUNREACH)
2785 dev->stats.tx_carrier_errors++;
2787 dev->stats.tx_errors++;
2791 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2792 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2794 struct vxlan_rdst nh_rdst;
2799 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2800 hash = skb_get_hash(skb);
2803 nh = rcu_dereference(f->nh);
2808 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2811 if (likely(do_xmit))
2812 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2819 dev->stats.tx_dropped++;
2823 /* Transmit local packets over Vxlan
2825 * Outer IP header inherits ECN and DF from inner header.
2826 * Outer UDP destination is the VXLAN assigned port.
2827 * source port is based on hash of flow
2829 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2831 struct vxlan_dev *vxlan = netdev_priv(dev);
2832 struct vxlan_rdst *rdst, *fdst = NULL;
2833 const struct ip_tunnel_info *info;
2834 bool did_rsc = false;
2835 struct vxlan_fdb *f;
2839 info = skb_tunnel_info(skb);
2841 skb_reset_mac_header(skb);
2843 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2844 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2845 info->mode & IP_TUNNEL_INFO_TX) {
2846 vni = tunnel_id_to_key32(info->key.tun_id);
2848 if (info && info->mode & IP_TUNNEL_INFO_TX)
2849 vxlan_xmit_one(skb, dev, vni, NULL, false);
2852 return NETDEV_TX_OK;
2856 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2858 if (ntohs(eth->h_proto) == ETH_P_ARP)
2859 return arp_reduce(dev, skb, vni);
2860 #if IS_ENABLED(CONFIG_IPV6)
2861 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2862 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2863 sizeof(struct nd_msg)) &&
2864 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2865 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2867 if (m->icmph.icmp6_code == 0 &&
2868 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2869 return neigh_reduce(dev, skb, vni);
2875 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2878 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2879 (ntohs(eth->h_proto) == ETH_P_IP ||
2880 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2881 did_rsc = route_shortcircuit(dev, skb);
2883 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2887 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2889 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2890 !is_multicast_ether_addr(eth->h_dest))
2891 vxlan_fdb_miss(vxlan, eth->h_dest);
2893 dev->stats.tx_dropped++;
2895 return NETDEV_TX_OK;
2899 if (rcu_access_pointer(f->nh)) {
2900 vxlan_xmit_nh(skb, dev, f,
2901 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2903 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2904 struct sk_buff *skb1;
2910 skb1 = skb_clone(skb, GFP_ATOMIC);
2912 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2915 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2920 return NETDEV_TX_OK;
2923 /* Walk the forwarding table and purge stale entries */
2924 static void vxlan_cleanup(struct timer_list *t)
2926 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2927 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2930 if (!netif_running(vxlan->dev))
2933 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2934 struct hlist_node *p, *n;
2936 spin_lock(&vxlan->hash_lock[h]);
2937 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2939 = container_of(p, struct vxlan_fdb, hlist);
2940 unsigned long timeout;
2942 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2945 if (f->flags & NTF_EXT_LEARNED)
2948 timeout = f->used + vxlan->cfg.age_interval * HZ;
2949 if (time_before_eq(timeout, jiffies)) {
2950 netdev_dbg(vxlan->dev,
2951 "garbage collect %pM\n",
2953 f->state = NUD_STALE;
2954 vxlan_fdb_destroy(vxlan, f, true, true);
2955 } else if (time_before(timeout, next_timer))
2956 next_timer = timeout;
2958 spin_unlock(&vxlan->hash_lock[h]);
2961 mod_timer(&vxlan->age_timer, next_timer);
2964 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2966 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2968 spin_lock(&vn->sock_lock);
2969 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2970 #if IS_ENABLED(CONFIG_IPV6)
2971 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2973 spin_unlock(&vn->sock_lock);
2976 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2977 struct vxlan_dev_node *node)
2979 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2980 __be32 vni = vxlan->default_dst.remote_vni;
2982 node->vxlan = vxlan;
2983 spin_lock(&vn->sock_lock);
2984 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2985 spin_unlock(&vn->sock_lock);
2988 /* Setup stats when device is created */
2989 static int vxlan_init(struct net_device *dev)
2991 struct vxlan_dev *vxlan = netdev_priv(dev);
2994 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2998 err = gro_cells_init(&vxlan->gro_cells, dev);
3000 free_percpu(dev->tstats);
3007 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3009 struct vxlan_fdb *f;
3010 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3012 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3013 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3015 vxlan_fdb_destroy(vxlan, f, true, true);
3016 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3019 static void vxlan_uninit(struct net_device *dev)
3021 struct vxlan_dev *vxlan = netdev_priv(dev);
3023 gro_cells_destroy(&vxlan->gro_cells);
3025 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3027 free_percpu(dev->tstats);
3030 /* Start ageing timer and join group when device is brought up */
3031 static int vxlan_open(struct net_device *dev)
3033 struct vxlan_dev *vxlan = netdev_priv(dev);
3036 ret = vxlan_sock_add(vxlan);
3040 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3041 ret = vxlan_igmp_join(vxlan);
3042 if (ret == -EADDRINUSE)
3045 vxlan_sock_release(vxlan);
3050 if (vxlan->cfg.age_interval)
3051 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3056 /* Purge the forwarding table */
3057 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3061 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3062 struct hlist_node *p, *n;
3064 spin_lock_bh(&vxlan->hash_lock[h]);
3065 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3067 = container_of(p, struct vxlan_fdb, hlist);
3068 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3070 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3071 if (!is_zero_ether_addr(f->eth_addr))
3072 vxlan_fdb_destroy(vxlan, f, true, true);
3074 spin_unlock_bh(&vxlan->hash_lock[h]);
3078 /* Cleanup timer and forwarding table on shutdown */
3079 static int vxlan_stop(struct net_device *dev)
3081 struct vxlan_dev *vxlan = netdev_priv(dev);
3082 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3085 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3086 !vxlan_group_used(vn, vxlan))
3087 ret = vxlan_igmp_leave(vxlan);
3089 del_timer_sync(&vxlan->age_timer);
3091 vxlan_flush(vxlan, false);
3092 vxlan_sock_release(vxlan);
3097 /* Stub, nothing needs to be done. */
3098 static void vxlan_set_multicast_list(struct net_device *dev)
3102 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3104 struct vxlan_dev *vxlan = netdev_priv(dev);
3105 struct vxlan_rdst *dst = &vxlan->default_dst;
3106 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3107 dst->remote_ifindex);
3108 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3110 /* This check is different than dev->max_mtu, because it looks at
3111 * the lowerdev->mtu, rather than the static dev->max_mtu
3114 int max_mtu = lowerdev->mtu -
3115 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3116 if (new_mtu > max_mtu)
3124 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3126 struct vxlan_dev *vxlan = netdev_priv(dev);
3127 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3128 __be16 sport, dport;
3130 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3131 vxlan->cfg.port_max, true);
3132 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3134 if (ip_tunnel_info_af(info) == AF_INET) {
3135 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3138 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3139 info->key.u.ipv4.dst,
3140 &info->key.u.ipv4.src, dport, sport,
3141 &info->dst_cache, info);
3146 #if IS_ENABLED(CONFIG_IPV6)
3147 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3148 struct dst_entry *ndst;
3150 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3151 info->key.label, &info->key.u.ipv6.dst,
3152 &info->key.u.ipv6.src, dport, sport,
3153 &info->dst_cache, info);
3155 return PTR_ERR(ndst);
3157 #else /* !CONFIG_IPV6 */
3158 return -EPFNOSUPPORT;
3161 info->key.tp_src = sport;
3162 info->key.tp_dst = dport;
3166 static const struct net_device_ops vxlan_netdev_ether_ops = {
3167 .ndo_init = vxlan_init,
3168 .ndo_uninit = vxlan_uninit,
3169 .ndo_open = vxlan_open,
3170 .ndo_stop = vxlan_stop,
3171 .ndo_start_xmit = vxlan_xmit,
3172 .ndo_get_stats64 = ip_tunnel_get_stats64,
3173 .ndo_set_rx_mode = vxlan_set_multicast_list,
3174 .ndo_change_mtu = vxlan_change_mtu,
3175 .ndo_validate_addr = eth_validate_addr,
3176 .ndo_set_mac_address = eth_mac_addr,
3177 .ndo_fdb_add = vxlan_fdb_add,
3178 .ndo_fdb_del = vxlan_fdb_delete,
3179 .ndo_fdb_dump = vxlan_fdb_dump,
3180 .ndo_fdb_get = vxlan_fdb_get,
3181 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3182 .ndo_change_proto_down = dev_change_proto_down_generic,
3185 static const struct net_device_ops vxlan_netdev_raw_ops = {
3186 .ndo_init = vxlan_init,
3187 .ndo_uninit = vxlan_uninit,
3188 .ndo_open = vxlan_open,
3189 .ndo_stop = vxlan_stop,
3190 .ndo_start_xmit = vxlan_xmit,
3191 .ndo_get_stats64 = ip_tunnel_get_stats64,
3192 .ndo_change_mtu = vxlan_change_mtu,
3193 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3196 /* Info for udev, that this is a virtual tunnel endpoint */
3197 static struct device_type vxlan_type = {
3201 /* Calls the ndo_udp_tunnel_add of the caller in order to
3202 * supply the listening VXLAN udp ports. Callers are expected
3203 * to implement the ndo_udp_tunnel_add.
3205 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3207 struct vxlan_sock *vs;
3208 struct net *net = dev_net(dev);
3209 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3212 spin_lock(&vn->sock_lock);
3213 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3214 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3215 unsigned short type;
3217 if (vs->flags & VXLAN_F_GPE)
3218 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3220 type = UDP_TUNNEL_TYPE_VXLAN;
3223 udp_tunnel_push_rx_port(dev, vs->sock, type);
3225 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3228 spin_unlock(&vn->sock_lock);
3231 /* Initialize the device structure. */
3232 static void vxlan_setup(struct net_device *dev)
3234 struct vxlan_dev *vxlan = netdev_priv(dev);
3237 eth_hw_addr_random(dev);
3240 dev->needs_free_netdev = true;
3241 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3243 dev->features |= NETIF_F_LLTX;
3244 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3245 dev->features |= NETIF_F_RXCSUM;
3246 dev->features |= NETIF_F_GSO_SOFTWARE;
3248 dev->vlan_features = dev->features;
3249 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3250 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3251 netif_keep_dst(dev);
3252 dev->priv_flags |= IFF_NO_QUEUE;
3254 /* MTU range: 68 - 65535 */
3255 dev->min_mtu = ETH_MIN_MTU;
3256 dev->max_mtu = ETH_MAX_MTU;
3258 INIT_LIST_HEAD(&vxlan->next);
3260 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3264 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3265 spin_lock_init(&vxlan->hash_lock[h]);
3266 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3270 static void vxlan_ether_setup(struct net_device *dev)
3272 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3273 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3274 dev->netdev_ops = &vxlan_netdev_ether_ops;
3277 static void vxlan_raw_setup(struct net_device *dev)
3279 dev->header_ops = NULL;
3280 dev->type = ARPHRD_NONE;
3281 dev->hard_header_len = 0;
3283 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3284 dev->netdev_ops = &vxlan_netdev_raw_ops;
3287 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3288 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3289 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3290 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3291 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3292 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3293 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3294 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3295 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3296 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3297 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3298 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3299 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3300 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3301 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3302 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3303 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3304 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3305 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3306 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3307 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3308 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3309 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3310 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3311 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3312 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3313 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3314 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3315 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3316 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3319 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3320 struct netlink_ext_ack *extack)
3322 if (tb[IFLA_ADDRESS]) {
3323 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3324 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3325 "Provided link layer address is not Ethernet");
3329 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3330 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3331 "Provided Ethernet address is not unicast");
3332 return -EADDRNOTAVAIL;
3337 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3339 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3340 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3341 "MTU must be between 68 and 65535");
3347 NL_SET_ERR_MSG(extack,
3348 "Required attributes not provided to perform the operation");
3352 if (data[IFLA_VXLAN_ID]) {
3353 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3355 if (id >= VXLAN_N_VID) {
3356 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3357 "VXLAN ID must be lower than 16777216");
3362 if (data[IFLA_VXLAN_PORT_RANGE]) {
3363 const struct ifla_vxlan_port_range *p
3364 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3366 if (ntohs(p->high) < ntohs(p->low)) {
3367 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3368 "Invalid source port range");
3373 if (data[IFLA_VXLAN_DF]) {
3374 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3376 if (df < 0 || df > VXLAN_DF_MAX) {
3377 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3378 "Invalid DF attribute");
3386 static void vxlan_get_drvinfo(struct net_device *netdev,
3387 struct ethtool_drvinfo *drvinfo)
3389 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3390 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3393 static int vxlan_get_link_ksettings(struct net_device *dev,
3394 struct ethtool_link_ksettings *cmd)
3396 struct vxlan_dev *vxlan = netdev_priv(dev);
3397 struct vxlan_rdst *dst = &vxlan->default_dst;
3398 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3399 dst->remote_ifindex);
3402 cmd->base.duplex = DUPLEX_UNKNOWN;
3403 cmd->base.port = PORT_OTHER;
3404 cmd->base.speed = SPEED_UNKNOWN;
3409 return __ethtool_get_link_ksettings(lowerdev, cmd);
3412 static const struct ethtool_ops vxlan_ethtool_ops = {
3413 .get_drvinfo = vxlan_get_drvinfo,
3414 .get_link = ethtool_op_get_link,
3415 .get_link_ksettings = vxlan_get_link_ksettings,
3418 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3419 __be16 port, u32 flags, int ifindex)
3421 struct socket *sock;
3422 struct udp_port_cfg udp_conf;
3425 memset(&udp_conf, 0, sizeof(udp_conf));
3428 udp_conf.family = AF_INET6;
3429 udp_conf.use_udp6_rx_checksums =
3430 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3431 udp_conf.ipv6_v6only = 1;
3433 udp_conf.family = AF_INET;
3436 udp_conf.local_udp_port = port;
3437 udp_conf.bind_ifindex = ifindex;
3439 /* Open UDP socket */
3440 err = udp_sock_create(net, &udp_conf, &sock);
3442 return ERR_PTR(err);
3447 /* Create new listen socket if needed */
3448 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3449 __be16 port, u32 flags,
3452 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3453 struct vxlan_sock *vs;
3454 struct socket *sock;
3456 struct udp_tunnel_sock_cfg tunnel_cfg;
3458 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3460 return ERR_PTR(-ENOMEM);
3462 for (h = 0; h < VNI_HASH_SIZE; ++h)
3463 INIT_HLIST_HEAD(&vs->vni_list[h]);
3465 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3468 return ERR_CAST(sock);
3472 refcount_set(&vs->refcnt, 1);
3473 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3475 spin_lock(&vn->sock_lock);
3476 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3477 udp_tunnel_notify_add_rx_port(sock,
3478 (vs->flags & VXLAN_F_GPE) ?
3479 UDP_TUNNEL_TYPE_VXLAN_GPE :
3480 UDP_TUNNEL_TYPE_VXLAN);
3481 spin_unlock(&vn->sock_lock);
3483 /* Mark socket as an encapsulation socket. */
3484 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3485 tunnel_cfg.sk_user_data = vs;
3486 tunnel_cfg.encap_type = 1;
3487 tunnel_cfg.encap_rcv = vxlan_rcv;
3488 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3489 tunnel_cfg.encap_destroy = NULL;
3490 tunnel_cfg.gro_receive = vxlan_gro_receive;
3491 tunnel_cfg.gro_complete = vxlan_gro_complete;
3493 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3498 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3500 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3501 struct vxlan_sock *vs = NULL;
3502 struct vxlan_dev_node *node;
3503 int l3mdev_index = 0;
3505 if (vxlan->cfg.remote_ifindex)
3506 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3507 vxlan->net, vxlan->cfg.remote_ifindex);
3509 if (!vxlan->cfg.no_share) {
3510 spin_lock(&vn->sock_lock);
3511 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3512 vxlan->cfg.dst_port, vxlan->cfg.flags,
3514 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3515 spin_unlock(&vn->sock_lock);
3518 spin_unlock(&vn->sock_lock);
3521 vs = vxlan_socket_create(vxlan->net, ipv6,
3522 vxlan->cfg.dst_port, vxlan->cfg.flags,
3526 #if IS_ENABLED(CONFIG_IPV6)
3528 rcu_assign_pointer(vxlan->vn6_sock, vs);
3529 node = &vxlan->hlist6;
3533 rcu_assign_pointer(vxlan->vn4_sock, vs);
3534 node = &vxlan->hlist4;
3536 vxlan_vs_add_dev(vs, vxlan, node);
3540 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3542 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3543 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3544 bool ipv4 = !ipv6 || metadata;
3547 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3548 #if IS_ENABLED(CONFIG_IPV6)
3549 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3551 ret = __vxlan_sock_add(vxlan, true);
3552 if (ret < 0 && ret != -EAFNOSUPPORT)
3557 ret = __vxlan_sock_add(vxlan, false);
3559 vxlan_sock_release(vxlan);
3563 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3564 struct net_device **lower,
3565 struct vxlan_dev *old,
3566 struct netlink_ext_ack *extack)
3568 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3569 struct vxlan_dev *tmp;
3570 bool use_ipv6 = false;
3572 if (conf->flags & VXLAN_F_GPE) {
3573 /* For now, allow GPE only together with
3574 * COLLECT_METADATA. This can be relaxed later; in such
3575 * case, the other side of the PtP link will have to be
3578 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3579 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3580 NL_SET_ERR_MSG(extack,
3581 "VXLAN GPE does not support this combination of attributes");
3586 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3587 /* Unless IPv6 is explicitly requested, assume IPv4 */
3588 conf->remote_ip.sa.sa_family = AF_INET;
3589 conf->saddr.sa.sa_family = AF_INET;
3590 } else if (!conf->remote_ip.sa.sa_family) {
3591 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3592 } else if (!conf->saddr.sa.sa_family) {
3593 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3596 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3597 NL_SET_ERR_MSG(extack,
3598 "Local and remote address must be from the same family");
3602 if (vxlan_addr_multicast(&conf->saddr)) {
3603 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3607 if (conf->saddr.sa.sa_family == AF_INET6) {
3608 if (!IS_ENABLED(CONFIG_IPV6)) {
3609 NL_SET_ERR_MSG(extack,
3610 "IPv6 support not enabled in the kernel");
3611 return -EPFNOSUPPORT;
3614 conf->flags |= VXLAN_F_IPV6;
3616 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3618 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3620 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3622 if (local_type & IPV6_ADDR_LINKLOCAL) {
3623 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3624 (remote_type != IPV6_ADDR_ANY)) {
3625 NL_SET_ERR_MSG(extack,
3626 "Invalid combination of local and remote address scopes");
3630 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3633 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3634 NL_SET_ERR_MSG(extack,
3635 "Invalid combination of local and remote address scopes");
3639 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3644 if (conf->label && !use_ipv6) {
3645 NL_SET_ERR_MSG(extack,
3646 "Label attribute only applies to IPv6 VXLAN devices");
3650 if (conf->remote_ifindex) {
3651 struct net_device *lowerdev;
3653 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3655 NL_SET_ERR_MSG(extack,
3656 "Invalid local interface, device not found");
3660 #if IS_ENABLED(CONFIG_IPV6)
3662 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3663 if (idev && idev->cnf.disable_ipv6) {
3664 NL_SET_ERR_MSG(extack,
3665 "IPv6 support disabled by administrator");
3673 if (vxlan_addr_multicast(&conf->remote_ip)) {
3674 NL_SET_ERR_MSG(extack,
3675 "Local interface required for multicast remote destination");
3680 #if IS_ENABLED(CONFIG_IPV6)
3681 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3682 NL_SET_ERR_MSG(extack,
3683 "Local interface required for link-local local/remote addresses");
3691 if (!conf->dst_port) {
3692 if (conf->flags & VXLAN_F_GPE)
3693 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3695 conf->dst_port = htons(vxlan_port);
3698 if (!conf->age_interval)
3699 conf->age_interval = FDB_AGE_DEFAULT;
3701 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3705 if (tmp->cfg.vni != conf->vni)
3707 if (tmp->cfg.dst_port != conf->dst_port)
3709 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3710 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3713 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3714 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3717 NL_SET_ERR_MSG(extack,
3718 "A VXLAN device with the specified VNI already exists");
3725 static void vxlan_config_apply(struct net_device *dev,
3726 struct vxlan_config *conf,
3727 struct net_device *lowerdev,
3728 struct net *src_net,
3731 struct vxlan_dev *vxlan = netdev_priv(dev);
3732 struct vxlan_rdst *dst = &vxlan->default_dst;
3733 unsigned short needed_headroom = ETH_HLEN;
3734 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3735 int max_mtu = ETH_MAX_MTU;
3738 if (conf->flags & VXLAN_F_GPE)
3739 vxlan_raw_setup(dev);
3741 vxlan_ether_setup(dev);
3744 dev->mtu = conf->mtu;
3746 vxlan->net = src_net;
3749 dst->remote_vni = conf->vni;
3751 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3754 dst->remote_ifindex = conf->remote_ifindex;
3756 dev->gso_max_size = lowerdev->gso_max_size;
3757 dev->gso_max_segs = lowerdev->gso_max_segs;
3759 needed_headroom = lowerdev->hard_header_len;
3761 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3763 if (max_mtu < ETH_MIN_MTU)
3764 max_mtu = ETH_MIN_MTU;
3766 if (!changelink && !conf->mtu)
3770 if (dev->mtu > max_mtu)
3773 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3774 needed_headroom += VXLAN6_HEADROOM;
3776 needed_headroom += VXLAN_HEADROOM;
3777 dev->needed_headroom = needed_headroom;
3779 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3782 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3783 struct vxlan_config *conf, bool changelink,
3784 struct netlink_ext_ack *extack)
3786 struct vxlan_dev *vxlan = netdev_priv(dev);
3787 struct net_device *lowerdev;
3790 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3794 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3799 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3800 struct vxlan_config *conf,
3801 struct netlink_ext_ack *extack)
3803 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3804 struct vxlan_dev *vxlan = netdev_priv(dev);
3805 struct net_device *remote_dev = NULL;
3806 struct vxlan_fdb *f = NULL;
3807 bool unregister = false;
3808 struct vxlan_rdst *dst;
3811 dst = &vxlan->default_dst;
3812 err = vxlan_dev_configure(net, dev, conf, false, extack);
3816 dev->ethtool_ops = &vxlan_ethtool_ops;
3818 /* create an fdb entry for a valid default destination */
3819 if (!vxlan_addr_any(&dst->remote_ip)) {
3820 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3822 NUD_REACHABLE | NUD_PERMANENT,
3823 vxlan->cfg.dst_port,
3826 dst->remote_ifindex,
3827 NTF_SELF, 0, &f, extack);
3832 err = register_netdevice(dev);
3837 if (dst->remote_ifindex) {
3838 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3842 err = netdev_upper_dev_link(remote_dev, dev, extack);
3847 err = rtnl_configure_link(dev, NULL);
3852 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3854 /* notify default fdb entry */
3855 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3856 RTM_NEWNEIGH, true, extack);
3858 vxlan_fdb_destroy(vxlan, f, false, false);
3860 netdev_upper_dev_unlink(remote_dev, dev);
3865 list_add(&vxlan->next, &vn->vxlan_list);
3867 dst->remote_dev = remote_dev;
3871 netdev_upper_dev_unlink(remote_dev, dev);
3873 /* unregister_netdevice() destroys the default FDB entry with deletion
3874 * notification. But the addition notification was not sent yet, so
3875 * destroy the entry by hand here.
3878 __vxlan_fdb_free(f);
3881 unregister_netdevice(dev);
3885 /* Set/clear flags based on attribute */
3886 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3887 int attrtype, unsigned long mask, bool changelink,
3888 bool changelink_supported,
3889 struct netlink_ext_ack *extack)
3891 unsigned long flags;
3896 if (changelink && !changelink_supported) {
3897 vxlan_flag_attr_error(attrtype, extack);
3901 if (vxlan_policy[attrtype].type == NLA_FLAG)
3902 flags = conf->flags | mask;
3903 else if (nla_get_u8(tb[attrtype]))
3904 flags = conf->flags | mask;
3906 flags = conf->flags & ~mask;
3908 conf->flags = flags;
3913 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3914 struct net_device *dev, struct vxlan_config *conf,
3915 bool changelink, struct netlink_ext_ack *extack)
3917 struct vxlan_dev *vxlan = netdev_priv(dev);
3920 memset(conf, 0, sizeof(*conf));
3922 /* if changelink operation, start with old existing cfg */
3924 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3926 if (data[IFLA_VXLAN_ID]) {
3927 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3929 if (changelink && (vni != conf->vni)) {
3930 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3933 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3936 if (data[IFLA_VXLAN_GROUP]) {
3937 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3938 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3942 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3943 conf->remote_ip.sa.sa_family = AF_INET;
3944 } else if (data[IFLA_VXLAN_GROUP6]) {
3945 if (!IS_ENABLED(CONFIG_IPV6)) {
3946 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3947 return -EPFNOSUPPORT;
3950 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3951 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3955 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3956 conf->remote_ip.sa.sa_family = AF_INET6;
3959 if (data[IFLA_VXLAN_LOCAL]) {
3960 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3961 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
3965 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3966 conf->saddr.sa.sa_family = AF_INET;
3967 } else if (data[IFLA_VXLAN_LOCAL6]) {
3968 if (!IS_ENABLED(CONFIG_IPV6)) {
3969 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
3970 return -EPFNOSUPPORT;
3973 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3974 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
3978 /* TODO: respect scope id */
3979 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3980 conf->saddr.sa.sa_family = AF_INET6;
3983 if (data[IFLA_VXLAN_LINK])
3984 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3986 if (data[IFLA_VXLAN_TOS])
3987 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3989 if (data[IFLA_VXLAN_TTL])
3990 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3992 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3993 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
3994 VXLAN_F_TTL_INHERIT, changelink, false,
4001 if (data[IFLA_VXLAN_LABEL])
4002 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4003 IPV6_FLOWLABEL_MASK;
4005 if (data[IFLA_VXLAN_LEARNING]) {
4006 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4007 VXLAN_F_LEARN, changelink, true,
4011 } else if (!changelink) {
4012 /* default to learn on a new device */
4013 conf->flags |= VXLAN_F_LEARN;
4016 if (data[IFLA_VXLAN_AGEING])
4017 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4019 if (data[IFLA_VXLAN_PROXY]) {
4020 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4021 VXLAN_F_PROXY, changelink, false,
4027 if (data[IFLA_VXLAN_RSC]) {
4028 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4029 VXLAN_F_RSC, changelink, false,
4035 if (data[IFLA_VXLAN_L2MISS]) {
4036 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4037 VXLAN_F_L2MISS, changelink, false,
4043 if (data[IFLA_VXLAN_L3MISS]) {
4044 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4045 VXLAN_F_L3MISS, changelink, false,
4051 if (data[IFLA_VXLAN_LIMIT]) {
4053 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4054 "Cannot change limit");
4057 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4060 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4061 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4062 VXLAN_F_COLLECT_METADATA, changelink, false,
4068 if (data[IFLA_VXLAN_PORT_RANGE]) {
4070 const struct ifla_vxlan_port_range *p
4071 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4072 conf->port_min = ntohs(p->low);
4073 conf->port_max = ntohs(p->high);
4075 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4076 "Cannot change port range");
4081 if (data[IFLA_VXLAN_PORT]) {
4083 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4084 "Cannot change port");
4087 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4090 if (data[IFLA_VXLAN_UDP_CSUM]) {
4092 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4093 "Cannot change UDP_CSUM flag");
4096 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4097 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4100 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4101 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4102 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4108 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4109 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4110 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4116 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4117 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4118 VXLAN_F_REMCSUM_TX, changelink, false,
4124 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4125 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4126 VXLAN_F_REMCSUM_RX, changelink, false,
4132 if (data[IFLA_VXLAN_GBP]) {
4133 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4134 VXLAN_F_GBP, changelink, false, extack);
4139 if (data[IFLA_VXLAN_GPE]) {
4140 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4141 VXLAN_F_GPE, changelink, false,
4147 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4148 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4149 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4157 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4158 "Cannot change mtu");
4161 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4164 if (data[IFLA_VXLAN_DF])
4165 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4170 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4171 struct nlattr *tb[], struct nlattr *data[],
4172 struct netlink_ext_ack *extack)
4174 struct vxlan_config conf;
4177 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4181 return __vxlan_dev_create(src_net, dev, &conf, extack);
4184 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4185 struct nlattr *data[],
4186 struct netlink_ext_ack *extack)
4188 struct vxlan_dev *vxlan = netdev_priv(dev);
4189 struct net_device *lowerdev;
4190 struct vxlan_config conf;
4191 struct vxlan_rdst *dst;
4194 dst = &vxlan->default_dst;
4195 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4199 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4204 if (dst->remote_dev == lowerdev)
4207 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4212 /* handle default dst entry */
4213 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4214 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4216 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4217 if (!vxlan_addr_any(&conf.remote_ip)) {
4218 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4220 NUD_REACHABLE | NUD_PERMANENT,
4221 NLM_F_APPEND | NLM_F_CREATE,
4222 vxlan->cfg.dst_port,
4224 conf.remote_ifindex,
4225 NTF_SELF, 0, true, extack);
4227 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4228 netdev_adjacent_change_abort(dst->remote_dev,
4233 if (!vxlan_addr_any(&dst->remote_ip))
4234 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4236 vxlan->cfg.dst_port,
4239 dst->remote_ifindex,
4241 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4244 if (conf.age_interval != vxlan->cfg.age_interval)
4245 mod_timer(&vxlan->age_timer, jiffies);
4247 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4248 if (lowerdev && lowerdev != dst->remote_dev) {
4249 dst->remote_dev = lowerdev;
4250 netdev_update_lockdep_key(lowerdev);
4252 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4256 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4258 struct vxlan_dev *vxlan = netdev_priv(dev);
4260 vxlan_flush(vxlan, true);
4262 list_del(&vxlan->next);
4263 unregister_netdevice_queue(dev, head);
4264 if (vxlan->default_dst.remote_dev)
4265 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4268 static size_t vxlan_get_size(const struct net_device *dev)
4271 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4272 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4273 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4274 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4275 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4276 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4277 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4278 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4279 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4280 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4281 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4282 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4283 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4284 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4285 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4286 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4287 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4288 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4289 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4290 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4291 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4292 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4293 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4294 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4298 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4300 const struct vxlan_dev *vxlan = netdev_priv(dev);
4301 const struct vxlan_rdst *dst = &vxlan->default_dst;
4302 struct ifla_vxlan_port_range ports = {
4303 .low = htons(vxlan->cfg.port_min),
4304 .high = htons(vxlan->cfg.port_max),
4307 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4308 goto nla_put_failure;
4310 if (!vxlan_addr_any(&dst->remote_ip)) {
4311 if (dst->remote_ip.sa.sa_family == AF_INET) {
4312 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4313 dst->remote_ip.sin.sin_addr.s_addr))
4314 goto nla_put_failure;
4315 #if IS_ENABLED(CONFIG_IPV6)
4317 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4318 &dst->remote_ip.sin6.sin6_addr))
4319 goto nla_put_failure;
4324 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4325 goto nla_put_failure;
4327 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4328 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4329 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4330 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4331 goto nla_put_failure;
4332 #if IS_ENABLED(CONFIG_IPV6)
4334 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4335 &vxlan->cfg.saddr.sin6.sin6_addr))
4336 goto nla_put_failure;
4341 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4342 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4343 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4344 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4345 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4346 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4347 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4348 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4349 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4350 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4351 nla_put_u8(skb, IFLA_VXLAN_RSC,
4352 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4353 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4354 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4355 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4356 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4357 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4358 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4359 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4360 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4361 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4362 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4363 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4364 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4365 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4366 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4367 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4368 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4369 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4370 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4371 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4372 goto nla_put_failure;
4374 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4375 goto nla_put_failure;
4377 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4378 nla_put_flag(skb, IFLA_VXLAN_GBP))
4379 goto nla_put_failure;
4381 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4382 nla_put_flag(skb, IFLA_VXLAN_GPE))
4383 goto nla_put_failure;
4385 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4386 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4387 goto nla_put_failure;
4395 static struct net *vxlan_get_link_net(const struct net_device *dev)
4397 struct vxlan_dev *vxlan = netdev_priv(dev);
4402 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4404 .maxtype = IFLA_VXLAN_MAX,
4405 .policy = vxlan_policy,
4406 .priv_size = sizeof(struct vxlan_dev),
4407 .setup = vxlan_setup,
4408 .validate = vxlan_validate,
4409 .newlink = vxlan_newlink,
4410 .changelink = vxlan_changelink,
4411 .dellink = vxlan_dellink,
4412 .get_size = vxlan_get_size,
4413 .fill_info = vxlan_fill_info,
4414 .get_link_net = vxlan_get_link_net,
4417 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4418 u8 name_assign_type,
4419 struct vxlan_config *conf)
4421 struct nlattr *tb[IFLA_MAX + 1];
4422 struct net_device *dev;
4425 memset(&tb, 0, sizeof(tb));
4427 dev = rtnl_create_link(net, name, name_assign_type,
4428 &vxlan_link_ops, tb, NULL);
4432 err = __vxlan_dev_create(net, dev, conf, NULL);
4435 return ERR_PTR(err);
4438 err = rtnl_configure_link(dev, NULL);
4440 LIST_HEAD(list_kill);
4442 vxlan_dellink(dev, &list_kill);
4443 unregister_netdevice_many(&list_kill);
4444 return ERR_PTR(err);
4449 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4451 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4452 struct net_device *dev)
4454 struct vxlan_dev *vxlan, *next;
4455 LIST_HEAD(list_kill);
4457 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4458 struct vxlan_rdst *dst = &vxlan->default_dst;
4460 /* In case we created vxlan device with carrier
4461 * and we loose the carrier due to module unload
4462 * we also need to remove vxlan device. In other
4463 * cases, it's not necessary and remote_ifindex
4464 * is 0 here, so no matches.
4466 if (dst->remote_ifindex == dev->ifindex)
4467 vxlan_dellink(vxlan->dev, &list_kill);
4470 unregister_netdevice_many(&list_kill);
4473 static int vxlan_netdevice_event(struct notifier_block *unused,
4474 unsigned long event, void *ptr)
4476 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4477 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4479 if (event == NETDEV_UNREGISTER) {
4480 vxlan_offload_rx_ports(dev, false);
4481 vxlan_handle_lowerdev_unregister(vn, dev);
4482 } else if (event == NETDEV_REGISTER) {
4483 vxlan_offload_rx_ports(dev, true);
4484 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4485 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4486 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4492 static struct notifier_block vxlan_notifier_block __read_mostly = {
4493 .notifier_call = vxlan_netdevice_event,
4497 vxlan_fdb_offloaded_set(struct net_device *dev,
4498 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4500 struct vxlan_dev *vxlan = netdev_priv(dev);
4501 struct vxlan_rdst *rdst;
4502 struct vxlan_fdb *f;
4505 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4507 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4509 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4513 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4514 fdb_info->remote_port,
4515 fdb_info->remote_vni,
4516 fdb_info->remote_ifindex);
4520 rdst->offloaded = fdb_info->offloaded;
4523 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4527 vxlan_fdb_external_learn_add(struct net_device *dev,
4528 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4530 struct vxlan_dev *vxlan = netdev_priv(dev);
4531 struct netlink_ext_ack *extack;
4535 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4536 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4538 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4539 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4541 NLM_F_CREATE | NLM_F_REPLACE,
4542 fdb_info->remote_port,
4544 fdb_info->remote_vni,
4545 fdb_info->remote_ifindex,
4546 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4548 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4554 vxlan_fdb_external_learn_del(struct net_device *dev,
4555 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4557 struct vxlan_dev *vxlan = netdev_priv(dev);
4558 struct vxlan_fdb *f;
4562 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4563 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4565 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4568 else if (f->flags & NTF_EXT_LEARNED)
4569 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4570 fdb_info->remote_ip,
4571 fdb_info->remote_port,
4573 fdb_info->remote_vni,
4574 fdb_info->remote_ifindex,
4577 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4582 static int vxlan_switchdev_event(struct notifier_block *unused,
4583 unsigned long event, void *ptr)
4585 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4586 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4590 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4591 vxlan_fdb_offloaded_set(dev, ptr);
4593 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4595 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4597 err = notifier_from_errno(err);
4600 fdb_info->offloaded = true;
4601 vxlan_fdb_offloaded_set(dev, fdb_info);
4603 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4605 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4607 err = notifier_from_errno(err);
4610 fdb_info->offloaded = false;
4611 vxlan_fdb_offloaded_set(dev, fdb_info);
4618 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4619 .notifier_call = vxlan_switchdev_event,
4622 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4624 struct vxlan_fdb *fdb;
4625 struct vxlan_dev *vxlan;
4629 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4630 vxlan = rcu_dereference(fdb->vdev);
4632 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4633 vxlan->default_dst.remote_vni);
4634 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4635 if (!hlist_unhashed(&fdb->hlist))
4636 vxlan_fdb_destroy(vxlan, fdb, false, false);
4637 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4642 static int vxlan_nexthop_event(struct notifier_block *nb,
4643 unsigned long event, void *ptr)
4645 struct nexthop *nh = ptr;
4647 if (!nh || event != NEXTHOP_EVENT_DEL)
4650 vxlan_fdb_nh_flush(nh);
4655 static struct notifier_block vxlan_nexthop_notifier_block __read_mostly = {
4656 .notifier_call = vxlan_nexthop_event,
4659 static __net_init int vxlan_init_net(struct net *net)
4661 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4664 INIT_LIST_HEAD(&vn->vxlan_list);
4665 spin_lock_init(&vn->sock_lock);
4667 for (h = 0; h < PORT_HASH_SIZE; ++h)
4668 INIT_HLIST_HEAD(&vn->sock_list[h]);
4670 return register_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4673 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4675 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4676 struct vxlan_dev *vxlan, *next;
4677 struct net_device *dev, *aux;
4680 for_each_netdev_safe(net, dev, aux)
4681 if (dev->rtnl_link_ops == &vxlan_link_ops)
4682 unregister_netdevice_queue(dev, head);
4684 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4685 /* If vxlan->dev is in the same netns, it has already been added
4686 * to the list by the previous loop.
4688 if (!net_eq(dev_net(vxlan->dev), net))
4689 unregister_netdevice_queue(vxlan->dev, head);
4692 for (h = 0; h < PORT_HASH_SIZE; ++h)
4693 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4696 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4702 list_for_each_entry(net, net_list, exit_list)
4703 unregister_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4704 list_for_each_entry(net, net_list, exit_list)
4705 vxlan_destroy_tunnels(net, &list);
4707 unregister_netdevice_many(&list);
4711 static struct pernet_operations vxlan_net_ops = {
4712 .init = vxlan_init_net,
4713 .exit_batch = vxlan_exit_batch_net,
4714 .id = &vxlan_net_id,
4715 .size = sizeof(struct vxlan_net),
4718 static int __init vxlan_init_module(void)
4722 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4724 rc = register_pernet_subsys(&vxlan_net_ops);
4728 rc = register_netdevice_notifier(&vxlan_notifier_block);
4732 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4736 rc = rtnl_link_register(&vxlan_link_ops);
4742 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4744 unregister_netdevice_notifier(&vxlan_notifier_block);
4746 unregister_pernet_subsys(&vxlan_net_ops);
4750 late_initcall(vxlan_init_module);
4752 static void __exit vxlan_cleanup_module(void)
4754 rtnl_link_unregister(&vxlan_link_ops);
4755 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4756 unregister_netdevice_notifier(&vxlan_notifier_block);
4757 unregister_pernet_subsys(&vxlan_net_ops);
4758 /* rcu_barrier() is called by netns */
4760 module_exit(vxlan_cleanup_module);
4762 MODULE_LICENSE("GPL");
4763 MODULE_VERSION(VXLAN_VERSION);
4764 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4765 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4766 MODULE_ALIAS_RTNL_LINK("vxlan");