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);
863 if (old_nh && old_nh->id == nhid)
866 nh = nexthop_find_by_id(vxlan->net, nhid);
868 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
873 if (!nexthop_get(nh)) {
874 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
878 if (!nexthop_is_fdb(nh)) {
879 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
883 if (!nexthop_is_multipath(nh)) {
884 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
888 /* check nexthop group family */
889 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
891 if (!nexthop_has_v4(nh)) {
893 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
898 if (nexthop_has_v4(nh)) {
900 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
907 list_del_rcu(&fdb->nh_list);
910 rcu_assign_pointer(fdb->nh, nh);
911 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
920 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
921 const u8 *mac, union vxlan_addr *ip,
922 __u16 state, __be16 port, __be32 src_vni,
923 __be32 vni, __u32 ifindex, __u16 ndm_flags,
924 u32 nhid, struct vxlan_fdb **fdb,
925 struct netlink_ext_ack *extack)
927 struct vxlan_rdst *rd = NULL;
931 if (vxlan->cfg.addrmax &&
932 vxlan->addrcnt >= vxlan->cfg.addrmax)
935 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
936 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
941 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
943 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
956 static void __vxlan_fdb_free(struct vxlan_fdb *f)
958 struct vxlan_rdst *rd, *nd;
961 nh = rcu_dereference_raw(f->nh);
963 rcu_assign_pointer(f->nh, NULL);
964 rcu_assign_pointer(f->vdev, NULL);
968 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
969 dst_cache_destroy(&rd->dst_cache);
975 static void vxlan_fdb_free(struct rcu_head *head)
977 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
982 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
983 bool do_notify, bool swdev_notify)
985 struct vxlan_rdst *rd;
987 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
991 if (rcu_access_pointer(f->nh))
992 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
995 list_for_each_entry(rd, &f->remotes, list)
996 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
1000 hlist_del_rcu(&f->hlist);
1001 list_del_rcu(&f->nh_list);
1002 call_rcu(&f->rcu, vxlan_fdb_free);
1005 static void vxlan_dst_free(struct rcu_head *head)
1007 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1009 dst_cache_destroy(&rd->dst_cache);
1013 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1014 union vxlan_addr *ip,
1015 __u16 state, __u16 flags,
1016 __be16 port, __be32 vni,
1017 __u32 ifindex, __u16 ndm_flags,
1018 struct vxlan_fdb *f, u32 nhid,
1020 struct netlink_ext_ack *extack)
1022 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1023 struct vxlan_rdst *rd = NULL;
1024 struct vxlan_rdst oldrd;
1029 if (nhid && !rcu_access_pointer(f->nh)) {
1030 NL_SET_ERR_MSG(extack,
1031 "Cannot replace an existing non nexthop fdb with a nexthop");
1035 if (nhid && (flags & NLM_F_APPEND)) {
1036 NL_SET_ERR_MSG(extack,
1037 "Cannot append to a nexthop fdb");
1041 /* Do not allow an externally learned entry to take over an entry added
1044 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1045 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1046 if (f->state != state) {
1048 f->updated = jiffies;
1051 if (f->flags != fdb_flags) {
1052 f->flags = fdb_flags;
1053 f->updated = jiffies;
1058 if ((flags & NLM_F_REPLACE)) {
1059 /* Only change unicasts */
1060 if (!(is_multicast_ether_addr(f->eth_addr) ||
1061 is_zero_ether_addr(f->eth_addr))) {
1063 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1067 rc = vxlan_fdb_replace(f, ip, port, vni,
1072 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1076 if ((flags & NLM_F_APPEND) &&
1077 (is_multicast_ether_addr(f->eth_addr) ||
1078 is_zero_ether_addr(f->eth_addr))) {
1079 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1086 if (ndm_flags & NTF_USE)
1091 rd = first_remote_rtnl(f);
1093 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1094 swdev_notify, extack);
1104 if ((flags & NLM_F_REPLACE) && rc)
1106 else if ((flags & NLM_F_APPEND) && rc) {
1107 list_del_rcu(&rd->list);
1108 call_rcu(&rd->rcu, vxlan_dst_free);
1113 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1114 const u8 *mac, union vxlan_addr *ip,
1115 __u16 state, __u16 flags,
1116 __be16 port, __be32 src_vni, __be32 vni,
1117 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1119 struct netlink_ext_ack *extack)
1121 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1122 struct vxlan_fdb *f;
1125 /* Disallow replace to add a multicast entry */
1126 if ((flags & NLM_F_REPLACE) &&
1127 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1130 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1131 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1132 vni, ifindex, fdb_flags, nhid, &f, extack);
1136 vxlan_fdb_insert(vxlan, mac, src_vni, f);
1137 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1138 swdev_notify, extack);
1145 vxlan_fdb_destroy(vxlan, f, false, false);
1149 /* Add new entry to forwarding table -- assumes lock held */
1150 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1151 const u8 *mac, union vxlan_addr *ip,
1152 __u16 state, __u16 flags,
1153 __be16 port, __be32 src_vni, __be32 vni,
1154 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1156 struct netlink_ext_ack *extack)
1158 struct vxlan_fdb *f;
1160 f = __vxlan_find_mac(vxlan, mac, src_vni);
1162 if (flags & NLM_F_EXCL) {
1163 netdev_dbg(vxlan->dev,
1164 "lost race to create %pM\n", mac);
1168 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1169 vni, ifindex, ndm_flags, f,
1170 nhid, swdev_notify, extack);
1172 if (!(flags & NLM_F_CREATE))
1175 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1176 port, src_vni, vni, ifindex,
1177 ndm_flags, nhid, swdev_notify,
1182 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1183 struct vxlan_rdst *rd, bool swdev_notify)
1185 list_del_rcu(&rd->list);
1186 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1187 call_rcu(&rd->rcu, vxlan_dst_free);
1190 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1191 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1192 __be32 *vni, u32 *ifindex, u32 *nhid)
1194 struct net *net = dev_net(vxlan->dev);
1197 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1202 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1206 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1208 if (remote->sa.sa_family == AF_INET) {
1209 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1210 ip->sa.sa_family = AF_INET;
1211 #if IS_ENABLED(CONFIG_IPV6)
1213 ip->sin6.sin6_addr = in6addr_any;
1214 ip->sa.sa_family = AF_INET6;
1220 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1222 *port = nla_get_be16(tb[NDA_PORT]);
1224 *port = vxlan->cfg.dst_port;
1228 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1230 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1232 *vni = vxlan->default_dst.remote_vni;
1235 if (tb[NDA_SRC_VNI]) {
1236 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1238 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1240 *src_vni = vxlan->default_dst.remote_vni;
1243 if (tb[NDA_IFINDEX]) {
1244 struct net_device *tdev;
1246 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1248 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1249 tdev = __dev_get_by_index(net, *ifindex);
1251 return -EADDRNOTAVAIL;
1257 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1264 /* Add static entry (via netlink) */
1265 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1266 struct net_device *dev,
1267 const unsigned char *addr, u16 vid, u16 flags,
1268 struct netlink_ext_ack *extack)
1270 struct vxlan_dev *vxlan = netdev_priv(dev);
1271 /* struct net *net = dev_net(vxlan->dev); */
1272 union vxlan_addr ip;
1274 __be32 src_vni, vni;
1279 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1280 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1285 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1288 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1293 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1294 return -EAFNOSUPPORT;
1296 hash_index = fdb_head_index(vxlan, addr, src_vni);
1297 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1298 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1299 port, src_vni, vni, ifindex,
1300 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1301 nhid, true, extack);
1302 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1307 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1308 const unsigned char *addr, union vxlan_addr ip,
1309 __be16 port, __be32 src_vni, __be32 vni,
1310 u32 ifindex, bool swdev_notify)
1312 struct vxlan_rdst *rd = NULL;
1313 struct vxlan_fdb *f;
1316 f = vxlan_find_mac(vxlan, addr, src_vni);
1320 if (!vxlan_addr_any(&ip)) {
1321 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1326 /* remove a destination if it's not the only one on the list,
1327 * otherwise destroy the fdb entry
1329 if (rd && !list_is_singular(&f->remotes)) {
1330 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1334 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1340 /* Delete entry (via netlink) */
1341 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1342 struct net_device *dev,
1343 const unsigned char *addr, u16 vid)
1345 struct vxlan_dev *vxlan = netdev_priv(dev);
1346 union vxlan_addr ip;
1347 __be32 src_vni, vni;
1353 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1358 hash_index = fdb_head_index(vxlan, addr, src_vni);
1359 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1360 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1362 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1367 /* Dump forwarding table */
1368 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1369 struct net_device *dev,
1370 struct net_device *filter_dev, int *idx)
1372 struct vxlan_dev *vxlan = netdev_priv(dev);
1376 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1377 struct vxlan_fdb *f;
1380 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1381 struct vxlan_rdst *rd;
1383 if (rcu_access_pointer(f->nh)) {
1384 if (*idx < cb->args[2])
1386 err = vxlan_fdb_info(skb, vxlan, f,
1387 NETLINK_CB(cb->skb).portid,
1400 list_for_each_entry_rcu(rd, &f->remotes, list) {
1401 if (*idx < cb->args[2])
1404 err = vxlan_fdb_info(skb, vxlan, f,
1405 NETLINK_CB(cb->skb).portid,
1423 static int vxlan_fdb_get(struct sk_buff *skb,
1424 struct nlattr *tb[],
1425 struct net_device *dev,
1426 const unsigned char *addr,
1427 u16 vid, u32 portid, u32 seq,
1428 struct netlink_ext_ack *extack)
1430 struct vxlan_dev *vxlan = netdev_priv(dev);
1431 struct vxlan_fdb *f;
1436 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1438 vni = vxlan->default_dst.remote_vni;
1442 f = __vxlan_find_mac(vxlan, addr, vni);
1444 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1449 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1450 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1456 /* Watch incoming packets to learn mapping between Ethernet address
1457 * and Tunnel endpoint.
1458 * Return true if packet is bogus and should be dropped.
1460 static bool vxlan_snoop(struct net_device *dev,
1461 union vxlan_addr *src_ip, const u8 *src_mac,
1462 u32 src_ifindex, __be32 vni)
1464 struct vxlan_dev *vxlan = netdev_priv(dev);
1465 struct vxlan_fdb *f;
1468 #if IS_ENABLED(CONFIG_IPV6)
1469 if (src_ip->sa.sa_family == AF_INET6 &&
1470 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1471 ifindex = src_ifindex;
1474 f = vxlan_find_mac(vxlan, src_mac, vni);
1476 struct vxlan_rdst *rdst = first_remote_rcu(f);
1478 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1479 rdst->remote_ifindex == ifindex))
1482 /* Don't migrate static entries, drop packets */
1483 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1486 /* Don't override an fdb with nexthop with a learnt entry */
1487 if (rcu_access_pointer(f->nh))
1490 if (net_ratelimit())
1492 "%pM migrated from %pIS to %pIS\n",
1493 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1495 rdst->remote_ip = *src_ip;
1496 f->updated = jiffies;
1497 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1499 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1501 /* learned new entry */
1502 spin_lock(&vxlan->hash_lock[hash_index]);
1504 /* close off race between vxlan_flush and incoming packets */
1505 if (netif_running(dev))
1506 vxlan_fdb_update(vxlan, src_mac, src_ip,
1508 NLM_F_EXCL|NLM_F_CREATE,
1509 vxlan->cfg.dst_port,
1511 vxlan->default_dst.remote_vni,
1512 ifindex, NTF_SELF, 0, true, NULL);
1513 spin_unlock(&vxlan->hash_lock[hash_index]);
1519 /* See if multicast group is already in use by other ID */
1520 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1522 struct vxlan_dev *vxlan;
1523 struct vxlan_sock *sock4;
1524 #if IS_ENABLED(CONFIG_IPV6)
1525 struct vxlan_sock *sock6;
1527 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1529 sock4 = rtnl_dereference(dev->vn4_sock);
1531 /* The vxlan_sock is only used by dev, leaving group has
1532 * no effect on other vxlan devices.
1534 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1536 #if IS_ENABLED(CONFIG_IPV6)
1537 sock6 = rtnl_dereference(dev->vn6_sock);
1538 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1542 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1543 if (!netif_running(vxlan->dev) || vxlan == dev)
1546 if (family == AF_INET &&
1547 rtnl_dereference(vxlan->vn4_sock) != sock4)
1549 #if IS_ENABLED(CONFIG_IPV6)
1550 if (family == AF_INET6 &&
1551 rtnl_dereference(vxlan->vn6_sock) != sock6)
1555 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1556 &dev->default_dst.remote_ip))
1559 if (vxlan->default_dst.remote_ifindex !=
1560 dev->default_dst.remote_ifindex)
1569 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1571 struct vxlan_net *vn;
1575 if (!refcount_dec_and_test(&vs->refcnt))
1578 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1579 spin_lock(&vn->sock_lock);
1580 hlist_del_rcu(&vs->hlist);
1581 udp_tunnel_notify_del_rx_port(vs->sock,
1582 (vs->flags & VXLAN_F_GPE) ?
1583 UDP_TUNNEL_TYPE_VXLAN_GPE :
1584 UDP_TUNNEL_TYPE_VXLAN);
1585 spin_unlock(&vn->sock_lock);
1590 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1592 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1593 #if IS_ENABLED(CONFIG_IPV6)
1594 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1596 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1599 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1602 vxlan_vs_del_dev(vxlan);
1604 if (__vxlan_sock_release_prep(sock4)) {
1605 udp_tunnel_sock_release(sock4->sock);
1609 #if IS_ENABLED(CONFIG_IPV6)
1610 if (__vxlan_sock_release_prep(sock6)) {
1611 udp_tunnel_sock_release(sock6->sock);
1617 /* Update multicast group membership when first VNI on
1618 * multicast address is brought up
1620 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1623 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1624 int ifindex = vxlan->default_dst.remote_ifindex;
1627 if (ip->sa.sa_family == AF_INET) {
1628 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1629 struct ip_mreqn mreq = {
1630 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1631 .imr_ifindex = ifindex,
1634 sk = sock4->sock->sk;
1636 ret = ip_mc_join_group(sk, &mreq);
1638 #if IS_ENABLED(CONFIG_IPV6)
1640 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1642 sk = sock6->sock->sk;
1644 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1645 &ip->sin6.sin6_addr);
1653 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1654 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1657 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1658 int ifindex = vxlan->default_dst.remote_ifindex;
1661 if (ip->sa.sa_family == AF_INET) {
1662 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1663 struct ip_mreqn mreq = {
1664 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1665 .imr_ifindex = ifindex,
1668 sk = sock4->sock->sk;
1670 ret = ip_mc_leave_group(sk, &mreq);
1672 #if IS_ENABLED(CONFIG_IPV6)
1674 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1676 sk = sock6->sock->sk;
1678 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1679 &ip->sin6.sin6_addr);
1687 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1688 struct sk_buff *skb, u32 vxflags)
1690 size_t start, offset;
1692 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1695 start = vxlan_rco_start(unparsed->vx_vni);
1696 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1698 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1701 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1702 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1704 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1705 unparsed->vx_vni &= VXLAN_VNI_MASK;
1709 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1710 struct sk_buff *skb, u32 vxflags,
1711 struct vxlan_metadata *md)
1713 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1714 struct metadata_dst *tun_dst;
1716 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1719 md->gbp = ntohs(gbp->policy_id);
1721 tun_dst = (struct metadata_dst *)skb_dst(skb);
1723 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1724 tun_dst->u.tun_info.options_len = sizeof(*md);
1726 if (gbp->dont_learn)
1727 md->gbp |= VXLAN_GBP_DONT_LEARN;
1729 if (gbp->policy_applied)
1730 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1732 /* In flow-based mode, GBP is carried in dst_metadata */
1733 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1734 skb->mark = md->gbp;
1736 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1739 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1741 struct sk_buff *skb, u32 vxflags)
1743 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1745 /* Need to have Next Protocol set for interfaces in GPE mode. */
1746 if (!gpe->np_applied)
1748 /* "The initial version is 0. If a receiver does not support the
1749 * version indicated it MUST drop the packet.
1751 if (gpe->version != 0)
1753 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1754 * processing MUST occur." However, we don't implement OAM
1755 * processing, thus drop the packet.
1760 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1764 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1768 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1769 struct vxlan_sock *vs,
1770 struct sk_buff *skb, __be32 vni)
1772 union vxlan_addr saddr;
1773 u32 ifindex = skb->dev->ifindex;
1775 skb_reset_mac_header(skb);
1776 skb->protocol = eth_type_trans(skb, vxlan->dev);
1777 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1779 /* Ignore packet loops (and multicast echo) */
1780 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1783 /* Get address from the outer IP header */
1784 if (vxlan_get_sk_family(vs) == AF_INET) {
1785 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1786 saddr.sa.sa_family = AF_INET;
1787 #if IS_ENABLED(CONFIG_IPV6)
1789 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1790 saddr.sa.sa_family = AF_INET6;
1794 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1795 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1801 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1802 struct sk_buff *skb)
1806 if (vxlan_get_sk_family(vs) == AF_INET)
1807 err = IP_ECN_decapsulate(oiph, skb);
1808 #if IS_ENABLED(CONFIG_IPV6)
1810 err = IP6_ECN_decapsulate(oiph, skb);
1813 if (unlikely(err) && log_ecn_error) {
1814 if (vxlan_get_sk_family(vs) == AF_INET)
1815 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1816 &((struct iphdr *)oiph)->saddr,
1817 ((struct iphdr *)oiph)->tos);
1819 net_info_ratelimited("non-ECT from %pI6\n",
1820 &((struct ipv6hdr *)oiph)->saddr);
1825 /* Callback from net/ipv4/udp.c to receive packets */
1826 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1828 struct pcpu_sw_netstats *stats;
1829 struct vxlan_dev *vxlan;
1830 struct vxlan_sock *vs;
1831 struct vxlanhdr unparsed;
1832 struct vxlan_metadata _md;
1833 struct vxlan_metadata *md = &_md;
1834 __be16 protocol = htons(ETH_P_TEB);
1835 bool raw_proto = false;
1839 /* Need UDP and VXLAN header to be present */
1840 if (!pskb_may_pull(skb, VXLAN_HLEN))
1843 unparsed = *vxlan_hdr(skb);
1844 /* VNI flag always required to be set */
1845 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1846 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1847 ntohl(vxlan_hdr(skb)->vx_flags),
1848 ntohl(vxlan_hdr(skb)->vx_vni));
1849 /* Return non vxlan pkt */
1852 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1853 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1855 vs = rcu_dereference_sk_user_data(sk);
1859 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1861 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1865 /* For backwards compatibility, only allow reserved fields to be
1866 * used by VXLAN extensions if explicitly requested.
1868 if (vs->flags & VXLAN_F_GPE) {
1869 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1874 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1875 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1878 if (vxlan_collect_metadata(vs)) {
1879 struct metadata_dst *tun_dst;
1881 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1882 key32_to_tunnel_id(vni), sizeof(*md));
1887 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1889 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1891 memset(md, 0, sizeof(*md));
1894 if (vs->flags & VXLAN_F_REMCSUM_RX)
1895 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1897 if (vs->flags & VXLAN_F_GBP)
1898 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1899 /* Note that GBP and GPE can never be active together. This is
1900 * ensured in vxlan_dev_configure.
1903 if (unparsed.vx_flags || unparsed.vx_vni) {
1904 /* If there are any unprocessed flags remaining treat
1905 * this as a malformed packet. This behavior diverges from
1906 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1907 * in reserved fields are to be ignored. The approach here
1908 * maintains compatibility with previous stack code, and also
1909 * is more robust and provides a little more security in
1910 * adding extensions to VXLAN.
1916 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1919 skb_reset_mac_header(skb);
1920 skb->dev = vxlan->dev;
1921 skb->pkt_type = PACKET_HOST;
1924 oiph = skb_network_header(skb);
1925 skb_reset_network_header(skb);
1927 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1928 ++vxlan->dev->stats.rx_frame_errors;
1929 ++vxlan->dev->stats.rx_errors;
1935 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1937 atomic_long_inc(&vxlan->dev->rx_dropped);
1941 stats = this_cpu_ptr(vxlan->dev->tstats);
1942 u64_stats_update_begin(&stats->syncp);
1943 stats->rx_packets++;
1944 stats->rx_bytes += skb->len;
1945 u64_stats_update_end(&stats->syncp);
1947 gro_cells_receive(&vxlan->gro_cells, skb);
1954 /* Consume bad packet */
1959 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1960 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1962 struct vxlan_dev *vxlan;
1963 struct vxlan_sock *vs;
1964 struct vxlanhdr *hdr;
1967 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1970 hdr = vxlan_hdr(skb);
1972 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1975 vs = rcu_dereference_sk_user_data(sk);
1979 vni = vxlan_vni(hdr->vx_vni);
1980 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1987 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1989 struct vxlan_dev *vxlan = netdev_priv(dev);
1990 struct arphdr *parp;
1993 struct neighbour *n;
1995 if (dev->flags & IFF_NOARP)
1998 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1999 dev->stats.tx_dropped++;
2002 parp = arp_hdr(skb);
2004 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2005 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2006 parp->ar_pro != htons(ETH_P_IP) ||
2007 parp->ar_op != htons(ARPOP_REQUEST) ||
2008 parp->ar_hln != dev->addr_len ||
2011 arpptr = (u8 *)parp + sizeof(struct arphdr);
2013 arpptr += dev->addr_len; /* sha */
2014 memcpy(&sip, arpptr, sizeof(sip));
2015 arpptr += sizeof(sip);
2016 arpptr += dev->addr_len; /* tha */
2017 memcpy(&tip, arpptr, sizeof(tip));
2019 if (ipv4_is_loopback(tip) ||
2020 ipv4_is_multicast(tip))
2023 n = neigh_lookup(&arp_tbl, &tip, dev);
2026 struct vxlan_fdb *f;
2027 struct sk_buff *reply;
2029 if (!(n->nud_state & NUD_CONNECTED)) {
2034 f = vxlan_find_mac(vxlan, n->ha, vni);
2035 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2036 /* bridge-local neighbor */
2041 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2049 skb_reset_mac_header(reply);
2050 __skb_pull(reply, skb_network_offset(reply));
2051 reply->ip_summed = CHECKSUM_UNNECESSARY;
2052 reply->pkt_type = PACKET_HOST;
2054 if (netif_rx_ni(reply) == NET_RX_DROP)
2055 dev->stats.rx_dropped++;
2056 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2057 union vxlan_addr ipa = {
2058 .sin.sin_addr.s_addr = tip,
2059 .sin.sin_family = AF_INET,
2062 vxlan_ip_miss(dev, &ipa);
2066 return NETDEV_TX_OK;
2069 #if IS_ENABLED(CONFIG_IPV6)
2070 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2071 struct neighbour *n, bool isrouter)
2073 struct net_device *dev = request->dev;
2074 struct sk_buff *reply;
2075 struct nd_msg *ns, *na;
2076 struct ipv6hdr *pip6;
2078 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2082 if (dev == NULL || !pskb_may_pull(request, request->len))
2085 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2086 sizeof(*na) + na_olen + dev->needed_tailroom;
2087 reply = alloc_skb(len, GFP_ATOMIC);
2091 reply->protocol = htons(ETH_P_IPV6);
2093 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2094 skb_push(reply, sizeof(struct ethhdr));
2095 skb_reset_mac_header(reply);
2097 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2099 daddr = eth_hdr(request)->h_source;
2100 ns_olen = request->len - skb_network_offset(request) -
2101 sizeof(struct ipv6hdr) - sizeof(*ns);
2102 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2103 if (!ns->opt[i + 1]) {
2107 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2108 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2113 /* Ethernet header */
2114 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2115 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2116 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2117 reply->protocol = htons(ETH_P_IPV6);
2119 skb_pull(reply, sizeof(struct ethhdr));
2120 skb_reset_network_header(reply);
2121 skb_put(reply, sizeof(struct ipv6hdr));
2125 pip6 = ipv6_hdr(reply);
2126 memset(pip6, 0, sizeof(struct ipv6hdr));
2128 pip6->priority = ipv6_hdr(request)->priority;
2129 pip6->nexthdr = IPPROTO_ICMPV6;
2130 pip6->hop_limit = 255;
2131 pip6->daddr = ipv6_hdr(request)->saddr;
2132 pip6->saddr = *(struct in6_addr *)n->primary_key;
2134 skb_pull(reply, sizeof(struct ipv6hdr));
2135 skb_reset_transport_header(reply);
2137 /* Neighbor Advertisement */
2138 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2139 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2140 na->icmph.icmp6_router = isrouter;
2141 na->icmph.icmp6_override = 1;
2142 na->icmph.icmp6_solicited = 1;
2143 na->target = ns->target;
2144 ether_addr_copy(&na->opt[2], n->ha);
2145 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2146 na->opt[1] = na_olen >> 3;
2148 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2149 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2150 csum_partial(na, sizeof(*na)+na_olen, 0));
2152 pip6->payload_len = htons(sizeof(*na)+na_olen);
2154 skb_push(reply, sizeof(struct ipv6hdr));
2156 reply->ip_summed = CHECKSUM_UNNECESSARY;
2161 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2163 struct vxlan_dev *vxlan = netdev_priv(dev);
2164 const struct in6_addr *daddr;
2165 const struct ipv6hdr *iphdr;
2166 struct inet6_dev *in6_dev;
2167 struct neighbour *n;
2170 in6_dev = __in6_dev_get(dev);
2174 iphdr = ipv6_hdr(skb);
2175 daddr = &iphdr->daddr;
2176 msg = (struct nd_msg *)(iphdr + 1);
2178 if (ipv6_addr_loopback(daddr) ||
2179 ipv6_addr_is_multicast(&msg->target))
2182 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2185 struct vxlan_fdb *f;
2186 struct sk_buff *reply;
2188 if (!(n->nud_state & NUD_CONNECTED)) {
2193 f = vxlan_find_mac(vxlan, n->ha, vni);
2194 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2195 /* bridge-local neighbor */
2200 reply = vxlan_na_create(skb, n,
2201 !!(f ? f->flags & NTF_ROUTER : 0));
2208 if (netif_rx_ni(reply) == NET_RX_DROP)
2209 dev->stats.rx_dropped++;
2211 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2212 union vxlan_addr ipa = {
2213 .sin6.sin6_addr = msg->target,
2214 .sin6.sin6_family = AF_INET6,
2217 vxlan_ip_miss(dev, &ipa);
2222 return NETDEV_TX_OK;
2226 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2228 struct vxlan_dev *vxlan = netdev_priv(dev);
2229 struct neighbour *n;
2231 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2235 switch (ntohs(eth_hdr(skb)->h_proto)) {
2240 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2243 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2244 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2245 union vxlan_addr ipa = {
2246 .sin.sin_addr.s_addr = pip->daddr,
2247 .sin.sin_family = AF_INET,
2250 vxlan_ip_miss(dev, &ipa);
2256 #if IS_ENABLED(CONFIG_IPV6)
2259 struct ipv6hdr *pip6;
2261 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2263 pip6 = ipv6_hdr(skb);
2264 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2265 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2266 union vxlan_addr ipa = {
2267 .sin6.sin6_addr = pip6->daddr,
2268 .sin6.sin6_family = AF_INET6,
2271 vxlan_ip_miss(dev, &ipa);
2285 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2287 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2289 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2298 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2299 struct vxlan_metadata *md)
2301 struct vxlanhdr_gbp *gbp;
2306 gbp = (struct vxlanhdr_gbp *)vxh;
2307 vxh->vx_flags |= VXLAN_HF_GBP;
2309 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2310 gbp->dont_learn = 1;
2312 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2313 gbp->policy_applied = 1;
2315 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2318 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2321 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2323 gpe->np_applied = 1;
2324 gpe->next_protocol = tun_p_from_eth_p(protocol);
2325 if (!gpe->next_protocol)
2326 return -EPFNOSUPPORT;
2330 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2331 int iphdr_len, __be32 vni,
2332 struct vxlan_metadata *md, u32 vxflags,
2335 struct vxlanhdr *vxh;
2338 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2339 __be16 inner_protocol = htons(ETH_P_TEB);
2341 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2342 skb->ip_summed == CHECKSUM_PARTIAL) {
2343 int csum_start = skb_checksum_start_offset(skb);
2345 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2346 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2347 (skb->csum_offset == offsetof(struct udphdr, check) ||
2348 skb->csum_offset == offsetof(struct tcphdr, check)))
2349 type |= SKB_GSO_TUNNEL_REMCSUM;
2352 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2353 + VXLAN_HLEN + iphdr_len;
2355 /* Need space for new headers (invalidates iph ptr) */
2356 err = skb_cow_head(skb, min_headroom);
2360 err = iptunnel_handle_offloads(skb, type);
2364 vxh = __skb_push(skb, sizeof(*vxh));
2365 vxh->vx_flags = VXLAN_HF_VNI;
2366 vxh->vx_vni = vxlan_vni_field(vni);
2368 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2371 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2372 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2373 vxh->vx_flags |= VXLAN_HF_RCO;
2375 if (!skb_is_gso(skb)) {
2376 skb->ip_summed = CHECKSUM_NONE;
2377 skb->encapsulation = 0;
2381 if (vxflags & VXLAN_F_GBP)
2382 vxlan_build_gbp_hdr(vxh, vxflags, md);
2383 if (vxflags & VXLAN_F_GPE) {
2384 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2387 inner_protocol = skb->protocol;
2390 skb_set_inner_protocol(skb, inner_protocol);
2394 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2395 struct vxlan_sock *sock4,
2396 struct sk_buff *skb, int oif, u8 tos,
2397 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2398 struct dst_cache *dst_cache,
2399 const struct ip_tunnel_info *info)
2401 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2402 struct rtable *rt = NULL;
2406 return ERR_PTR(-EIO);
2411 rt = dst_cache_get_ip4(dst_cache, saddr);
2416 memset(&fl4, 0, sizeof(fl4));
2417 fl4.flowi4_oif = oif;
2418 fl4.flowi4_tos = RT_TOS(tos);
2419 fl4.flowi4_mark = skb->mark;
2420 fl4.flowi4_proto = IPPROTO_UDP;
2423 fl4.fl4_dport = dport;
2424 fl4.fl4_sport = sport;
2426 rt = ip_route_output_key(vxlan->net, &fl4);
2428 if (rt->dst.dev == dev) {
2429 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2431 return ERR_PTR(-ELOOP);
2436 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2438 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2439 return ERR_PTR(-ENETUNREACH);
2444 #if IS_ENABLED(CONFIG_IPV6)
2445 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2446 struct net_device *dev,
2447 struct vxlan_sock *sock6,
2448 struct sk_buff *skb, int oif, u8 tos,
2450 const struct in6_addr *daddr,
2451 struct in6_addr *saddr,
2452 __be16 dport, __be16 sport,
2453 struct dst_cache *dst_cache,
2454 const struct ip_tunnel_info *info)
2456 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2457 struct dst_entry *ndst;
2461 return ERR_PTR(-EIO);
2466 ndst = dst_cache_get_ip6(dst_cache, saddr);
2471 memset(&fl6, 0, sizeof(fl6));
2472 fl6.flowi6_oif = oif;
2475 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2476 fl6.flowi6_mark = skb->mark;
2477 fl6.flowi6_proto = IPPROTO_UDP;
2478 fl6.fl6_dport = dport;
2479 fl6.fl6_sport = sport;
2481 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2483 if (unlikely(IS_ERR(ndst))) {
2484 netdev_dbg(dev, "no route to %pI6\n", daddr);
2485 return ERR_PTR(-ENETUNREACH);
2488 if (unlikely(ndst->dev == dev)) {
2489 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2491 return ERR_PTR(-ELOOP);
2496 dst_cache_set_ip6(dst_cache, ndst, saddr);
2501 /* Bypass encapsulation if the destination is local */
2502 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2503 struct vxlan_dev *dst_vxlan, __be32 vni,
2506 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2507 union vxlan_addr loopback;
2508 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2509 struct net_device *dev;
2512 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2513 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2514 skb->pkt_type = PACKET_HOST;
2515 skb->encapsulation = 0;
2516 skb->dev = dst_vxlan->dev;
2517 __skb_pull(skb, skb_network_offset(skb));
2519 if (remote_ip->sa.sa_family == AF_INET) {
2520 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2521 loopback.sa.sa_family = AF_INET;
2522 #if IS_ENABLED(CONFIG_IPV6)
2524 loopback.sin6.sin6_addr = in6addr_loopback;
2525 loopback.sa.sa_family = AF_INET6;
2531 if (unlikely(!(dev->flags & IFF_UP))) {
2536 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2537 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2539 u64_stats_update_begin(&tx_stats->syncp);
2540 tx_stats->tx_packets++;
2541 tx_stats->tx_bytes += len;
2542 u64_stats_update_end(&tx_stats->syncp);
2544 if (netif_rx(skb) == NET_RX_SUCCESS) {
2545 u64_stats_update_begin(&rx_stats->syncp);
2546 rx_stats->rx_packets++;
2547 rx_stats->rx_bytes += len;
2548 u64_stats_update_end(&rx_stats->syncp);
2551 dev->stats.rx_dropped++;
2556 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2557 struct vxlan_dev *vxlan,
2558 union vxlan_addr *daddr,
2559 __be16 dst_port, int dst_ifindex, __be32 vni,
2560 struct dst_entry *dst,
2563 #if IS_ENABLED(CONFIG_IPV6)
2564 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2565 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2566 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2568 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2570 /* Bypass encapsulation if the destination is local */
2571 if (rt_flags & RTCF_LOCAL &&
2572 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2573 struct vxlan_dev *dst_vxlan;
2576 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2577 daddr->sa.sa_family, dst_port,
2580 dev->stats.tx_errors++;
2585 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2592 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2593 __be32 default_vni, struct vxlan_rdst *rdst,
2596 struct dst_cache *dst_cache;
2597 struct ip_tunnel_info *info;
2598 struct vxlan_dev *vxlan = netdev_priv(dev);
2599 const struct iphdr *old_iph = ip_hdr(skb);
2600 union vxlan_addr *dst;
2601 union vxlan_addr remote_ip, local_ip;
2602 struct vxlan_metadata _md;
2603 struct vxlan_metadata *md = &_md;
2604 __be16 src_port = 0, dst_port;
2605 struct dst_entry *ndst = NULL;
2610 u32 flags = vxlan->cfg.flags;
2611 bool udp_sum = false;
2612 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2614 info = skb_tunnel_info(skb);
2617 dst = &rdst->remote_ip;
2618 if (vxlan_addr_any(dst)) {
2620 /* short-circuited back to local bridge */
2621 vxlan_encap_bypass(skb, vxlan, vxlan,
2628 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2629 vni = (rdst->remote_vni) ? : default_vni;
2630 ifindex = rdst->remote_ifindex;
2631 local_ip = vxlan->cfg.saddr;
2632 dst_cache = &rdst->dst_cache;
2633 md->gbp = skb->mark;
2634 if (flags & VXLAN_F_TTL_INHERIT) {
2635 ttl = ip_tunnel_get_ttl(old_iph, skb);
2637 ttl = vxlan->cfg.ttl;
2638 if (!ttl && vxlan_addr_multicast(dst))
2642 tos = vxlan->cfg.tos;
2644 tos = ip_tunnel_get_dsfield(old_iph, skb);
2646 if (dst->sa.sa_family == AF_INET)
2647 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2649 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2650 label = vxlan->cfg.label;
2653 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2657 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2658 if (remote_ip.sa.sa_family == AF_INET) {
2659 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2660 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2662 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2663 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2666 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2667 vni = tunnel_id_to_key32(info->key.tun_id);
2669 dst_cache = &info->dst_cache;
2670 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2671 if (info->options_len < sizeof(*md))
2673 md = ip_tunnel_info_opts(info);
2675 ttl = info->key.ttl;
2676 tos = info->key.tos;
2677 label = info->key.label;
2678 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2680 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2681 vxlan->cfg.port_max, true);
2684 if (dst->sa.sa_family == AF_INET) {
2685 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2690 ifindex = sock4->sock->sk->sk_bound_dev_if;
2692 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2693 dst->sin.sin_addr.s_addr,
2694 &local_ip.sin.sin_addr.s_addr,
2703 /* Bypass encapsulation if the destination is local */
2704 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2705 dst_port, ifindex, vni,
2706 &rt->dst, rt->rt_flags);
2710 if (vxlan->cfg.df == VXLAN_DF_SET) {
2712 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2713 struct ethhdr *eth = eth_hdr(skb);
2715 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2716 (ntohs(eth->h_proto) == ETH_P_IP &&
2717 old_iph->frag_off & htons(IP_DF)))
2720 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2725 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2726 netif_is_any_bridge_port(dev));
2731 struct in_addr src, dst;
2733 src = remote_ip.sin.sin_addr;
2734 dst = local_ip.sin.sin_addr;
2735 info->key.u.ipv4.src = src.s_addr;
2736 info->key.u.ipv4.dst = dst.s_addr;
2738 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2743 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2744 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2745 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2746 vni, md, flags, udp_sum);
2750 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2751 dst->sin.sin_addr.s_addr, tos, ttl, df,
2752 src_port, dst_port, xnet, !udp_sum);
2753 #if IS_ENABLED(CONFIG_IPV6)
2755 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2758 ifindex = sock6->sock->sk->sk_bound_dev_if;
2760 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2761 label, &dst->sin6.sin6_addr,
2762 &local_ip.sin6.sin6_addr,
2766 err = PTR_ERR(ndst);
2772 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2774 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2775 dst_port, ifindex, vni,
2781 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2782 netif_is_any_bridge_port(dev));
2787 struct in6_addr src, dst;
2789 src = remote_ip.sin6.sin6_addr;
2790 dst = local_ip.sin6.sin6_addr;
2791 info->key.u.ipv6.src = src;
2792 info->key.u.ipv6.dst = dst;
2795 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2800 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2801 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2802 skb_scrub_packet(skb, xnet);
2803 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2804 vni, md, flags, udp_sum);
2808 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2809 &local_ip.sin6.sin6_addr,
2810 &dst->sin6.sin6_addr, tos, ttl,
2811 label, src_port, dst_port, !udp_sum);
2819 dev->stats.tx_dropped++;
2826 dev->stats.collisions++;
2827 else if (err == -ENETUNREACH)
2828 dev->stats.tx_carrier_errors++;
2830 dev->stats.tx_errors++;
2834 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2835 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2837 struct vxlan_rdst nh_rdst;
2842 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2843 hash = skb_get_hash(skb);
2846 nh = rcu_dereference(f->nh);
2851 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2854 if (likely(do_xmit))
2855 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2862 dev->stats.tx_dropped++;
2866 /* Transmit local packets over Vxlan
2868 * Outer IP header inherits ECN and DF from inner header.
2869 * Outer UDP destination is the VXLAN assigned port.
2870 * source port is based on hash of flow
2872 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2874 struct vxlan_dev *vxlan = netdev_priv(dev);
2875 struct vxlan_rdst *rdst, *fdst = NULL;
2876 const struct ip_tunnel_info *info;
2877 bool did_rsc = false;
2878 struct vxlan_fdb *f;
2882 info = skb_tunnel_info(skb);
2884 skb_reset_mac_header(skb);
2886 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2887 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2888 info->mode & IP_TUNNEL_INFO_TX) {
2889 vni = tunnel_id_to_key32(info->key.tun_id);
2891 if (info && info->mode & IP_TUNNEL_INFO_TX)
2892 vxlan_xmit_one(skb, dev, vni, NULL, false);
2895 return NETDEV_TX_OK;
2899 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2901 if (ntohs(eth->h_proto) == ETH_P_ARP)
2902 return arp_reduce(dev, skb, vni);
2903 #if IS_ENABLED(CONFIG_IPV6)
2904 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2905 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2906 sizeof(struct nd_msg)) &&
2907 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2908 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2910 if (m->icmph.icmp6_code == 0 &&
2911 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2912 return neigh_reduce(dev, skb, vni);
2918 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2921 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2922 (ntohs(eth->h_proto) == ETH_P_IP ||
2923 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2924 did_rsc = route_shortcircuit(dev, skb);
2926 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2930 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2932 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2933 !is_multicast_ether_addr(eth->h_dest))
2934 vxlan_fdb_miss(vxlan, eth->h_dest);
2936 dev->stats.tx_dropped++;
2938 return NETDEV_TX_OK;
2942 if (rcu_access_pointer(f->nh)) {
2943 vxlan_xmit_nh(skb, dev, f,
2944 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2946 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2947 struct sk_buff *skb1;
2953 skb1 = skb_clone(skb, GFP_ATOMIC);
2955 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2958 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2963 return NETDEV_TX_OK;
2966 /* Walk the forwarding table and purge stale entries */
2967 static void vxlan_cleanup(struct timer_list *t)
2969 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2970 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2973 if (!netif_running(vxlan->dev))
2976 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2977 struct hlist_node *p, *n;
2979 spin_lock(&vxlan->hash_lock[h]);
2980 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2982 = container_of(p, struct vxlan_fdb, hlist);
2983 unsigned long timeout;
2985 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2988 if (f->flags & NTF_EXT_LEARNED)
2991 timeout = f->used + vxlan->cfg.age_interval * HZ;
2992 if (time_before_eq(timeout, jiffies)) {
2993 netdev_dbg(vxlan->dev,
2994 "garbage collect %pM\n",
2996 f->state = NUD_STALE;
2997 vxlan_fdb_destroy(vxlan, f, true, true);
2998 } else if (time_before(timeout, next_timer))
2999 next_timer = timeout;
3001 spin_unlock(&vxlan->hash_lock[h]);
3004 mod_timer(&vxlan->age_timer, next_timer);
3007 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3009 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3011 spin_lock(&vn->sock_lock);
3012 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3013 #if IS_ENABLED(CONFIG_IPV6)
3014 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3016 spin_unlock(&vn->sock_lock);
3019 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3020 struct vxlan_dev_node *node)
3022 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3023 __be32 vni = vxlan->default_dst.remote_vni;
3025 node->vxlan = vxlan;
3026 spin_lock(&vn->sock_lock);
3027 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3028 spin_unlock(&vn->sock_lock);
3031 /* Setup stats when device is created */
3032 static int vxlan_init(struct net_device *dev)
3034 struct vxlan_dev *vxlan = netdev_priv(dev);
3037 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3041 err = gro_cells_init(&vxlan->gro_cells, dev);
3043 free_percpu(dev->tstats);
3050 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3052 struct vxlan_fdb *f;
3053 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3055 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3056 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3058 vxlan_fdb_destroy(vxlan, f, true, true);
3059 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3062 static void vxlan_uninit(struct net_device *dev)
3064 struct vxlan_dev *vxlan = netdev_priv(dev);
3066 gro_cells_destroy(&vxlan->gro_cells);
3068 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3070 free_percpu(dev->tstats);
3073 /* Start ageing timer and join group when device is brought up */
3074 static int vxlan_open(struct net_device *dev)
3076 struct vxlan_dev *vxlan = netdev_priv(dev);
3079 ret = vxlan_sock_add(vxlan);
3083 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3084 ret = vxlan_igmp_join(vxlan);
3085 if (ret == -EADDRINUSE)
3088 vxlan_sock_release(vxlan);
3093 if (vxlan->cfg.age_interval)
3094 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3099 /* Purge the forwarding table */
3100 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3104 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3105 struct hlist_node *p, *n;
3107 spin_lock_bh(&vxlan->hash_lock[h]);
3108 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3110 = container_of(p, struct vxlan_fdb, hlist);
3111 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3113 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3114 if (is_zero_ether_addr(f->eth_addr) &&
3115 f->vni == vxlan->cfg.vni)
3117 vxlan_fdb_destroy(vxlan, f, true, true);
3119 spin_unlock_bh(&vxlan->hash_lock[h]);
3123 /* Cleanup timer and forwarding table on shutdown */
3124 static int vxlan_stop(struct net_device *dev)
3126 struct vxlan_dev *vxlan = netdev_priv(dev);
3127 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3130 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3131 !vxlan_group_used(vn, vxlan))
3132 ret = vxlan_igmp_leave(vxlan);
3134 del_timer_sync(&vxlan->age_timer);
3136 vxlan_flush(vxlan, false);
3137 vxlan_sock_release(vxlan);
3142 /* Stub, nothing needs to be done. */
3143 static void vxlan_set_multicast_list(struct net_device *dev)
3147 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3149 struct vxlan_dev *vxlan = netdev_priv(dev);
3150 struct vxlan_rdst *dst = &vxlan->default_dst;
3151 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3152 dst->remote_ifindex);
3153 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3155 /* This check is different than dev->max_mtu, because it looks at
3156 * the lowerdev->mtu, rather than the static dev->max_mtu
3159 int max_mtu = lowerdev->mtu -
3160 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3161 if (new_mtu > max_mtu)
3169 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3171 struct vxlan_dev *vxlan = netdev_priv(dev);
3172 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3173 __be16 sport, dport;
3175 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3176 vxlan->cfg.port_max, true);
3177 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3179 if (ip_tunnel_info_af(info) == AF_INET) {
3180 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3183 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3184 info->key.u.ipv4.dst,
3185 &info->key.u.ipv4.src, dport, sport,
3186 &info->dst_cache, info);
3191 #if IS_ENABLED(CONFIG_IPV6)
3192 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3193 struct dst_entry *ndst;
3195 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3196 info->key.label, &info->key.u.ipv6.dst,
3197 &info->key.u.ipv6.src, dport, sport,
3198 &info->dst_cache, info);
3200 return PTR_ERR(ndst);
3202 #else /* !CONFIG_IPV6 */
3203 return -EPFNOSUPPORT;
3206 info->key.tp_src = sport;
3207 info->key.tp_dst = dport;
3211 static const struct net_device_ops vxlan_netdev_ether_ops = {
3212 .ndo_init = vxlan_init,
3213 .ndo_uninit = vxlan_uninit,
3214 .ndo_open = vxlan_open,
3215 .ndo_stop = vxlan_stop,
3216 .ndo_start_xmit = vxlan_xmit,
3217 .ndo_get_stats64 = ip_tunnel_get_stats64,
3218 .ndo_set_rx_mode = vxlan_set_multicast_list,
3219 .ndo_change_mtu = vxlan_change_mtu,
3220 .ndo_validate_addr = eth_validate_addr,
3221 .ndo_set_mac_address = eth_mac_addr,
3222 .ndo_fdb_add = vxlan_fdb_add,
3223 .ndo_fdb_del = vxlan_fdb_delete,
3224 .ndo_fdb_dump = vxlan_fdb_dump,
3225 .ndo_fdb_get = vxlan_fdb_get,
3226 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3227 .ndo_change_proto_down = dev_change_proto_down_generic,
3230 static const struct net_device_ops vxlan_netdev_raw_ops = {
3231 .ndo_init = vxlan_init,
3232 .ndo_uninit = vxlan_uninit,
3233 .ndo_open = vxlan_open,
3234 .ndo_stop = vxlan_stop,
3235 .ndo_start_xmit = vxlan_xmit,
3236 .ndo_get_stats64 = ip_tunnel_get_stats64,
3237 .ndo_change_mtu = vxlan_change_mtu,
3238 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3241 /* Info for udev, that this is a virtual tunnel endpoint */
3242 static struct device_type vxlan_type = {
3246 /* Calls the ndo_udp_tunnel_add of the caller in order to
3247 * supply the listening VXLAN udp ports. Callers are expected
3248 * to implement the ndo_udp_tunnel_add.
3250 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3252 struct vxlan_sock *vs;
3253 struct net *net = dev_net(dev);
3254 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3257 spin_lock(&vn->sock_lock);
3258 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3259 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3260 unsigned short type;
3262 if (vs->flags & VXLAN_F_GPE)
3263 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3265 type = UDP_TUNNEL_TYPE_VXLAN;
3268 udp_tunnel_push_rx_port(dev, vs->sock, type);
3270 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3273 spin_unlock(&vn->sock_lock);
3276 /* Initialize the device structure. */
3277 static void vxlan_setup(struct net_device *dev)
3279 struct vxlan_dev *vxlan = netdev_priv(dev);
3282 eth_hw_addr_random(dev);
3285 dev->needs_free_netdev = true;
3286 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3288 dev->features |= NETIF_F_LLTX;
3289 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3290 dev->features |= NETIF_F_RXCSUM;
3291 dev->features |= NETIF_F_GSO_SOFTWARE;
3293 dev->vlan_features = dev->features;
3294 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3295 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3296 netif_keep_dst(dev);
3297 dev->priv_flags |= IFF_NO_QUEUE;
3299 /* MTU range: 68 - 65535 */
3300 dev->min_mtu = ETH_MIN_MTU;
3301 dev->max_mtu = ETH_MAX_MTU;
3303 INIT_LIST_HEAD(&vxlan->next);
3305 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3309 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3310 spin_lock_init(&vxlan->hash_lock[h]);
3311 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3315 static void vxlan_ether_setup(struct net_device *dev)
3317 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3318 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3319 dev->netdev_ops = &vxlan_netdev_ether_ops;
3322 static void vxlan_raw_setup(struct net_device *dev)
3324 dev->header_ops = NULL;
3325 dev->type = ARPHRD_NONE;
3326 dev->hard_header_len = 0;
3328 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3329 dev->netdev_ops = &vxlan_netdev_raw_ops;
3332 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3333 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3334 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3335 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3336 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3337 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3338 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3339 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3340 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3341 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3342 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3343 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3344 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3345 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3346 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3347 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3348 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3349 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3350 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3351 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3352 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3353 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3354 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3355 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3356 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3357 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3358 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3359 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3360 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3361 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3364 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3365 struct netlink_ext_ack *extack)
3367 if (tb[IFLA_ADDRESS]) {
3368 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3369 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3370 "Provided link layer address is not Ethernet");
3374 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3375 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3376 "Provided Ethernet address is not unicast");
3377 return -EADDRNOTAVAIL;
3382 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3384 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3385 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3386 "MTU must be between 68 and 65535");
3392 NL_SET_ERR_MSG(extack,
3393 "Required attributes not provided to perform the operation");
3397 if (data[IFLA_VXLAN_ID]) {
3398 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3400 if (id >= VXLAN_N_VID) {
3401 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3402 "VXLAN ID must be lower than 16777216");
3407 if (data[IFLA_VXLAN_PORT_RANGE]) {
3408 const struct ifla_vxlan_port_range *p
3409 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3411 if (ntohs(p->high) < ntohs(p->low)) {
3412 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3413 "Invalid source port range");
3418 if (data[IFLA_VXLAN_DF]) {
3419 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3421 if (df < 0 || df > VXLAN_DF_MAX) {
3422 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3423 "Invalid DF attribute");
3431 static void vxlan_get_drvinfo(struct net_device *netdev,
3432 struct ethtool_drvinfo *drvinfo)
3434 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3435 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3438 static int vxlan_get_link_ksettings(struct net_device *dev,
3439 struct ethtool_link_ksettings *cmd)
3441 struct vxlan_dev *vxlan = netdev_priv(dev);
3442 struct vxlan_rdst *dst = &vxlan->default_dst;
3443 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3444 dst->remote_ifindex);
3447 cmd->base.duplex = DUPLEX_UNKNOWN;
3448 cmd->base.port = PORT_OTHER;
3449 cmd->base.speed = SPEED_UNKNOWN;
3454 return __ethtool_get_link_ksettings(lowerdev, cmd);
3457 static const struct ethtool_ops vxlan_ethtool_ops = {
3458 .get_drvinfo = vxlan_get_drvinfo,
3459 .get_link = ethtool_op_get_link,
3460 .get_link_ksettings = vxlan_get_link_ksettings,
3463 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3464 __be16 port, u32 flags, int ifindex)
3466 struct socket *sock;
3467 struct udp_port_cfg udp_conf;
3470 memset(&udp_conf, 0, sizeof(udp_conf));
3473 udp_conf.family = AF_INET6;
3474 udp_conf.use_udp6_rx_checksums =
3475 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3476 udp_conf.ipv6_v6only = 1;
3478 udp_conf.family = AF_INET;
3481 udp_conf.local_udp_port = port;
3482 udp_conf.bind_ifindex = ifindex;
3484 /* Open UDP socket */
3485 err = udp_sock_create(net, &udp_conf, &sock);
3487 return ERR_PTR(err);
3492 /* Create new listen socket if needed */
3493 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3494 __be16 port, u32 flags,
3497 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3498 struct vxlan_sock *vs;
3499 struct socket *sock;
3501 struct udp_tunnel_sock_cfg tunnel_cfg;
3503 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3505 return ERR_PTR(-ENOMEM);
3507 for (h = 0; h < VNI_HASH_SIZE; ++h)
3508 INIT_HLIST_HEAD(&vs->vni_list[h]);
3510 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3513 return ERR_CAST(sock);
3517 refcount_set(&vs->refcnt, 1);
3518 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3520 spin_lock(&vn->sock_lock);
3521 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3522 udp_tunnel_notify_add_rx_port(sock,
3523 (vs->flags & VXLAN_F_GPE) ?
3524 UDP_TUNNEL_TYPE_VXLAN_GPE :
3525 UDP_TUNNEL_TYPE_VXLAN);
3526 spin_unlock(&vn->sock_lock);
3528 /* Mark socket as an encapsulation socket. */
3529 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3530 tunnel_cfg.sk_user_data = vs;
3531 tunnel_cfg.encap_type = 1;
3532 tunnel_cfg.encap_rcv = vxlan_rcv;
3533 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3534 tunnel_cfg.encap_destroy = NULL;
3535 tunnel_cfg.gro_receive = vxlan_gro_receive;
3536 tunnel_cfg.gro_complete = vxlan_gro_complete;
3538 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3543 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3545 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3546 struct vxlan_sock *vs = NULL;
3547 struct vxlan_dev_node *node;
3548 int l3mdev_index = 0;
3550 if (vxlan->cfg.remote_ifindex)
3551 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3552 vxlan->net, vxlan->cfg.remote_ifindex);
3554 if (!vxlan->cfg.no_share) {
3555 spin_lock(&vn->sock_lock);
3556 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3557 vxlan->cfg.dst_port, vxlan->cfg.flags,
3559 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3560 spin_unlock(&vn->sock_lock);
3563 spin_unlock(&vn->sock_lock);
3566 vs = vxlan_socket_create(vxlan->net, ipv6,
3567 vxlan->cfg.dst_port, vxlan->cfg.flags,
3571 #if IS_ENABLED(CONFIG_IPV6)
3573 rcu_assign_pointer(vxlan->vn6_sock, vs);
3574 node = &vxlan->hlist6;
3578 rcu_assign_pointer(vxlan->vn4_sock, vs);
3579 node = &vxlan->hlist4;
3581 vxlan_vs_add_dev(vs, vxlan, node);
3585 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3587 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3588 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3589 bool ipv4 = !ipv6 || metadata;
3592 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3593 #if IS_ENABLED(CONFIG_IPV6)
3594 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3596 ret = __vxlan_sock_add(vxlan, true);
3597 if (ret < 0 && ret != -EAFNOSUPPORT)
3602 ret = __vxlan_sock_add(vxlan, false);
3604 vxlan_sock_release(vxlan);
3608 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3609 struct net_device **lower,
3610 struct vxlan_dev *old,
3611 struct netlink_ext_ack *extack)
3613 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3614 struct vxlan_dev *tmp;
3615 bool use_ipv6 = false;
3617 if (conf->flags & VXLAN_F_GPE) {
3618 /* For now, allow GPE only together with
3619 * COLLECT_METADATA. This can be relaxed later; in such
3620 * case, the other side of the PtP link will have to be
3623 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3624 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3625 NL_SET_ERR_MSG(extack,
3626 "VXLAN GPE does not support this combination of attributes");
3631 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3632 /* Unless IPv6 is explicitly requested, assume IPv4 */
3633 conf->remote_ip.sa.sa_family = AF_INET;
3634 conf->saddr.sa.sa_family = AF_INET;
3635 } else if (!conf->remote_ip.sa.sa_family) {
3636 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3637 } else if (!conf->saddr.sa.sa_family) {
3638 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3641 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3642 NL_SET_ERR_MSG(extack,
3643 "Local and remote address must be from the same family");
3647 if (vxlan_addr_multicast(&conf->saddr)) {
3648 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3652 if (conf->saddr.sa.sa_family == AF_INET6) {
3653 if (!IS_ENABLED(CONFIG_IPV6)) {
3654 NL_SET_ERR_MSG(extack,
3655 "IPv6 support not enabled in the kernel");
3656 return -EPFNOSUPPORT;
3659 conf->flags |= VXLAN_F_IPV6;
3661 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3663 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3665 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3667 if (local_type & IPV6_ADDR_LINKLOCAL) {
3668 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3669 (remote_type != IPV6_ADDR_ANY)) {
3670 NL_SET_ERR_MSG(extack,
3671 "Invalid combination of local and remote address scopes");
3675 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3678 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3679 NL_SET_ERR_MSG(extack,
3680 "Invalid combination of local and remote address scopes");
3684 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3689 if (conf->label && !use_ipv6) {
3690 NL_SET_ERR_MSG(extack,
3691 "Label attribute only applies to IPv6 VXLAN devices");
3695 if (conf->remote_ifindex) {
3696 struct net_device *lowerdev;
3698 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3700 NL_SET_ERR_MSG(extack,
3701 "Invalid local interface, device not found");
3705 #if IS_ENABLED(CONFIG_IPV6)
3707 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3708 if (idev && idev->cnf.disable_ipv6) {
3709 NL_SET_ERR_MSG(extack,
3710 "IPv6 support disabled by administrator");
3718 if (vxlan_addr_multicast(&conf->remote_ip)) {
3719 NL_SET_ERR_MSG(extack,
3720 "Local interface required for multicast remote destination");
3725 #if IS_ENABLED(CONFIG_IPV6)
3726 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3727 NL_SET_ERR_MSG(extack,
3728 "Local interface required for link-local local/remote addresses");
3736 if (!conf->dst_port) {
3737 if (conf->flags & VXLAN_F_GPE)
3738 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3740 conf->dst_port = htons(vxlan_port);
3743 if (!conf->age_interval)
3744 conf->age_interval = FDB_AGE_DEFAULT;
3746 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3750 if (tmp->cfg.vni != conf->vni)
3752 if (tmp->cfg.dst_port != conf->dst_port)
3754 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3755 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3758 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3759 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3762 NL_SET_ERR_MSG(extack,
3763 "A VXLAN device with the specified VNI already exists");
3770 static void vxlan_config_apply(struct net_device *dev,
3771 struct vxlan_config *conf,
3772 struct net_device *lowerdev,
3773 struct net *src_net,
3776 struct vxlan_dev *vxlan = netdev_priv(dev);
3777 struct vxlan_rdst *dst = &vxlan->default_dst;
3778 unsigned short needed_headroom = ETH_HLEN;
3779 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3780 int max_mtu = ETH_MAX_MTU;
3783 if (conf->flags & VXLAN_F_GPE)
3784 vxlan_raw_setup(dev);
3786 vxlan_ether_setup(dev);
3789 dev->mtu = conf->mtu;
3791 vxlan->net = src_net;
3794 dst->remote_vni = conf->vni;
3796 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3799 dst->remote_ifindex = conf->remote_ifindex;
3801 dev->gso_max_size = lowerdev->gso_max_size;
3802 dev->gso_max_segs = lowerdev->gso_max_segs;
3804 needed_headroom = lowerdev->hard_header_len;
3806 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3808 if (max_mtu < ETH_MIN_MTU)
3809 max_mtu = ETH_MIN_MTU;
3811 if (!changelink && !conf->mtu)
3815 if (dev->mtu > max_mtu)
3818 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3819 needed_headroom += VXLAN6_HEADROOM;
3821 needed_headroom += VXLAN_HEADROOM;
3822 dev->needed_headroom = needed_headroom;
3824 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3827 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3828 struct vxlan_config *conf, bool changelink,
3829 struct netlink_ext_ack *extack)
3831 struct vxlan_dev *vxlan = netdev_priv(dev);
3832 struct net_device *lowerdev;
3835 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3839 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3844 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3845 struct vxlan_config *conf,
3846 struct netlink_ext_ack *extack)
3848 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3849 struct vxlan_dev *vxlan = netdev_priv(dev);
3850 struct net_device *remote_dev = NULL;
3851 struct vxlan_fdb *f = NULL;
3852 bool unregister = false;
3853 struct vxlan_rdst *dst;
3856 dst = &vxlan->default_dst;
3857 err = vxlan_dev_configure(net, dev, conf, false, extack);
3861 dev->ethtool_ops = &vxlan_ethtool_ops;
3863 /* create an fdb entry for a valid default destination */
3864 if (!vxlan_addr_any(&dst->remote_ip)) {
3865 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3867 NUD_REACHABLE | NUD_PERMANENT,
3868 vxlan->cfg.dst_port,
3871 dst->remote_ifindex,
3872 NTF_SELF, 0, &f, extack);
3877 err = register_netdevice(dev);
3882 if (dst->remote_ifindex) {
3883 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3887 err = netdev_upper_dev_link(remote_dev, dev, extack);
3892 err = rtnl_configure_link(dev, NULL);
3897 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3899 /* notify default fdb entry */
3900 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3901 RTM_NEWNEIGH, true, extack);
3903 vxlan_fdb_destroy(vxlan, f, false, false);
3905 netdev_upper_dev_unlink(remote_dev, dev);
3910 list_add(&vxlan->next, &vn->vxlan_list);
3912 dst->remote_dev = remote_dev;
3916 netdev_upper_dev_unlink(remote_dev, dev);
3918 /* unregister_netdevice() destroys the default FDB entry with deletion
3919 * notification. But the addition notification was not sent yet, so
3920 * destroy the entry by hand here.
3923 __vxlan_fdb_free(f);
3926 unregister_netdevice(dev);
3930 /* Set/clear flags based on attribute */
3931 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3932 int attrtype, unsigned long mask, bool changelink,
3933 bool changelink_supported,
3934 struct netlink_ext_ack *extack)
3936 unsigned long flags;
3941 if (changelink && !changelink_supported) {
3942 vxlan_flag_attr_error(attrtype, extack);
3946 if (vxlan_policy[attrtype].type == NLA_FLAG)
3947 flags = conf->flags | mask;
3948 else if (nla_get_u8(tb[attrtype]))
3949 flags = conf->flags | mask;
3951 flags = conf->flags & ~mask;
3953 conf->flags = flags;
3958 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3959 struct net_device *dev, struct vxlan_config *conf,
3960 bool changelink, struct netlink_ext_ack *extack)
3962 struct vxlan_dev *vxlan = netdev_priv(dev);
3965 memset(conf, 0, sizeof(*conf));
3967 /* if changelink operation, start with old existing cfg */
3969 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3971 if (data[IFLA_VXLAN_ID]) {
3972 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3974 if (changelink && (vni != conf->vni)) {
3975 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3978 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3981 if (data[IFLA_VXLAN_GROUP]) {
3982 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3983 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3987 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3988 conf->remote_ip.sa.sa_family = AF_INET;
3989 } else if (data[IFLA_VXLAN_GROUP6]) {
3990 if (!IS_ENABLED(CONFIG_IPV6)) {
3991 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3992 return -EPFNOSUPPORT;
3995 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3996 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4000 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4001 conf->remote_ip.sa.sa_family = AF_INET6;
4004 if (data[IFLA_VXLAN_LOCAL]) {
4005 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4006 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4010 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4011 conf->saddr.sa.sa_family = AF_INET;
4012 } else if (data[IFLA_VXLAN_LOCAL6]) {
4013 if (!IS_ENABLED(CONFIG_IPV6)) {
4014 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4015 return -EPFNOSUPPORT;
4018 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4019 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4023 /* TODO: respect scope id */
4024 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4025 conf->saddr.sa.sa_family = AF_INET6;
4028 if (data[IFLA_VXLAN_LINK])
4029 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4031 if (data[IFLA_VXLAN_TOS])
4032 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
4034 if (data[IFLA_VXLAN_TTL])
4035 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4037 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4038 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4039 VXLAN_F_TTL_INHERIT, changelink, false,
4046 if (data[IFLA_VXLAN_LABEL])
4047 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4048 IPV6_FLOWLABEL_MASK;
4050 if (data[IFLA_VXLAN_LEARNING]) {
4051 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4052 VXLAN_F_LEARN, changelink, true,
4056 } else if (!changelink) {
4057 /* default to learn on a new device */
4058 conf->flags |= VXLAN_F_LEARN;
4061 if (data[IFLA_VXLAN_AGEING])
4062 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4064 if (data[IFLA_VXLAN_PROXY]) {
4065 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4066 VXLAN_F_PROXY, changelink, false,
4072 if (data[IFLA_VXLAN_RSC]) {
4073 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4074 VXLAN_F_RSC, changelink, false,
4080 if (data[IFLA_VXLAN_L2MISS]) {
4081 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4082 VXLAN_F_L2MISS, changelink, false,
4088 if (data[IFLA_VXLAN_L3MISS]) {
4089 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4090 VXLAN_F_L3MISS, changelink, false,
4096 if (data[IFLA_VXLAN_LIMIT]) {
4098 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4099 "Cannot change limit");
4102 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4105 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4106 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4107 VXLAN_F_COLLECT_METADATA, changelink, false,
4113 if (data[IFLA_VXLAN_PORT_RANGE]) {
4115 const struct ifla_vxlan_port_range *p
4116 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4117 conf->port_min = ntohs(p->low);
4118 conf->port_max = ntohs(p->high);
4120 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4121 "Cannot change port range");
4126 if (data[IFLA_VXLAN_PORT]) {
4128 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4129 "Cannot change port");
4132 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4135 if (data[IFLA_VXLAN_UDP_CSUM]) {
4137 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4138 "Cannot change UDP_CSUM flag");
4141 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4142 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4145 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4146 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4147 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4153 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4154 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4155 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4161 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4162 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4163 VXLAN_F_REMCSUM_TX, changelink, false,
4169 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4170 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4171 VXLAN_F_REMCSUM_RX, changelink, false,
4177 if (data[IFLA_VXLAN_GBP]) {
4178 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4179 VXLAN_F_GBP, changelink, false, extack);
4184 if (data[IFLA_VXLAN_GPE]) {
4185 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4186 VXLAN_F_GPE, changelink, false,
4192 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4193 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4194 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4202 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4203 "Cannot change mtu");
4206 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4209 if (data[IFLA_VXLAN_DF])
4210 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4215 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4216 struct nlattr *tb[], struct nlattr *data[],
4217 struct netlink_ext_ack *extack)
4219 struct vxlan_config conf;
4222 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4226 return __vxlan_dev_create(src_net, dev, &conf, extack);
4229 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4230 struct nlattr *data[],
4231 struct netlink_ext_ack *extack)
4233 struct vxlan_dev *vxlan = netdev_priv(dev);
4234 struct net_device *lowerdev;
4235 struct vxlan_config conf;
4236 struct vxlan_rdst *dst;
4239 dst = &vxlan->default_dst;
4240 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4244 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4249 if (dst->remote_dev == lowerdev)
4252 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4257 /* handle default dst entry */
4258 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4259 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4261 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4262 if (!vxlan_addr_any(&conf.remote_ip)) {
4263 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4265 NUD_REACHABLE | NUD_PERMANENT,
4266 NLM_F_APPEND | NLM_F_CREATE,
4267 vxlan->cfg.dst_port,
4269 conf.remote_ifindex,
4270 NTF_SELF, 0, true, extack);
4272 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4273 netdev_adjacent_change_abort(dst->remote_dev,
4278 if (!vxlan_addr_any(&dst->remote_ip))
4279 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4281 vxlan->cfg.dst_port,
4284 dst->remote_ifindex,
4286 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4289 if (conf.age_interval != vxlan->cfg.age_interval)
4290 mod_timer(&vxlan->age_timer, jiffies);
4292 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4293 if (lowerdev && lowerdev != dst->remote_dev)
4294 dst->remote_dev = lowerdev;
4295 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4299 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4301 struct vxlan_dev *vxlan = netdev_priv(dev);
4303 vxlan_flush(vxlan, true);
4305 list_del(&vxlan->next);
4306 unregister_netdevice_queue(dev, head);
4307 if (vxlan->default_dst.remote_dev)
4308 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4311 static size_t vxlan_get_size(const struct net_device *dev)
4314 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4315 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4316 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4317 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4318 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4319 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4320 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4321 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4322 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4323 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4324 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4325 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4326 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4327 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4328 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4329 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4330 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4331 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4332 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4333 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4334 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4341 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4343 const struct vxlan_dev *vxlan = netdev_priv(dev);
4344 const struct vxlan_rdst *dst = &vxlan->default_dst;
4345 struct ifla_vxlan_port_range ports = {
4346 .low = htons(vxlan->cfg.port_min),
4347 .high = htons(vxlan->cfg.port_max),
4350 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4351 goto nla_put_failure;
4353 if (!vxlan_addr_any(&dst->remote_ip)) {
4354 if (dst->remote_ip.sa.sa_family == AF_INET) {
4355 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4356 dst->remote_ip.sin.sin_addr.s_addr))
4357 goto nla_put_failure;
4358 #if IS_ENABLED(CONFIG_IPV6)
4360 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4361 &dst->remote_ip.sin6.sin6_addr))
4362 goto nla_put_failure;
4367 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4368 goto nla_put_failure;
4370 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4371 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4372 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4373 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4374 goto nla_put_failure;
4375 #if IS_ENABLED(CONFIG_IPV6)
4377 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4378 &vxlan->cfg.saddr.sin6.sin6_addr))
4379 goto nla_put_failure;
4384 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4385 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4386 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4387 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4388 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4389 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4390 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4391 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4392 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4393 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4394 nla_put_u8(skb, IFLA_VXLAN_RSC,
4395 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4396 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4397 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4398 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4399 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4400 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4401 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4402 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4403 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4404 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4405 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4406 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4407 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4408 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4409 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4410 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4411 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4412 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4413 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4414 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4415 goto nla_put_failure;
4417 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4418 goto nla_put_failure;
4420 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4421 nla_put_flag(skb, IFLA_VXLAN_GBP))
4422 goto nla_put_failure;
4424 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4425 nla_put_flag(skb, IFLA_VXLAN_GPE))
4426 goto nla_put_failure;
4428 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4429 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4430 goto nla_put_failure;
4438 static struct net *vxlan_get_link_net(const struct net_device *dev)
4440 struct vxlan_dev *vxlan = netdev_priv(dev);
4445 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4447 .maxtype = IFLA_VXLAN_MAX,
4448 .policy = vxlan_policy,
4449 .priv_size = sizeof(struct vxlan_dev),
4450 .setup = vxlan_setup,
4451 .validate = vxlan_validate,
4452 .newlink = vxlan_newlink,
4453 .changelink = vxlan_changelink,
4454 .dellink = vxlan_dellink,
4455 .get_size = vxlan_get_size,
4456 .fill_info = vxlan_fill_info,
4457 .get_link_net = vxlan_get_link_net,
4460 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4461 u8 name_assign_type,
4462 struct vxlan_config *conf)
4464 struct nlattr *tb[IFLA_MAX + 1];
4465 struct net_device *dev;
4468 memset(&tb, 0, sizeof(tb));
4470 dev = rtnl_create_link(net, name, name_assign_type,
4471 &vxlan_link_ops, tb, NULL);
4475 err = __vxlan_dev_create(net, dev, conf, NULL);
4478 return ERR_PTR(err);
4481 err = rtnl_configure_link(dev, NULL);
4483 LIST_HEAD(list_kill);
4485 vxlan_dellink(dev, &list_kill);
4486 unregister_netdevice_many(&list_kill);
4487 return ERR_PTR(err);
4492 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4494 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4495 struct net_device *dev)
4497 struct vxlan_dev *vxlan, *next;
4498 LIST_HEAD(list_kill);
4500 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4501 struct vxlan_rdst *dst = &vxlan->default_dst;
4503 /* In case we created vxlan device with carrier
4504 * and we loose the carrier due to module unload
4505 * we also need to remove vxlan device. In other
4506 * cases, it's not necessary and remote_ifindex
4507 * is 0 here, so no matches.
4509 if (dst->remote_ifindex == dev->ifindex)
4510 vxlan_dellink(vxlan->dev, &list_kill);
4513 unregister_netdevice_many(&list_kill);
4516 static int vxlan_netdevice_event(struct notifier_block *unused,
4517 unsigned long event, void *ptr)
4519 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4520 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4522 if (event == NETDEV_UNREGISTER) {
4523 if (!dev->udp_tunnel_nic_info)
4524 vxlan_offload_rx_ports(dev, false);
4525 vxlan_handle_lowerdev_unregister(vn, dev);
4526 } else if (event == NETDEV_REGISTER) {
4527 if (!dev->udp_tunnel_nic_info)
4528 vxlan_offload_rx_ports(dev, true);
4529 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4530 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4531 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4537 static struct notifier_block vxlan_notifier_block __read_mostly = {
4538 .notifier_call = vxlan_netdevice_event,
4542 vxlan_fdb_offloaded_set(struct net_device *dev,
4543 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4545 struct vxlan_dev *vxlan = netdev_priv(dev);
4546 struct vxlan_rdst *rdst;
4547 struct vxlan_fdb *f;
4550 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4552 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4554 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4558 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4559 fdb_info->remote_port,
4560 fdb_info->remote_vni,
4561 fdb_info->remote_ifindex);
4565 rdst->offloaded = fdb_info->offloaded;
4568 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4572 vxlan_fdb_external_learn_add(struct net_device *dev,
4573 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4575 struct vxlan_dev *vxlan = netdev_priv(dev);
4576 struct netlink_ext_ack *extack;
4580 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4581 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4583 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4584 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4586 NLM_F_CREATE | NLM_F_REPLACE,
4587 fdb_info->remote_port,
4589 fdb_info->remote_vni,
4590 fdb_info->remote_ifindex,
4591 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4593 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4599 vxlan_fdb_external_learn_del(struct net_device *dev,
4600 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4602 struct vxlan_dev *vxlan = netdev_priv(dev);
4603 struct vxlan_fdb *f;
4607 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4608 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4610 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4613 else if (f->flags & NTF_EXT_LEARNED)
4614 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4615 fdb_info->remote_ip,
4616 fdb_info->remote_port,
4618 fdb_info->remote_vni,
4619 fdb_info->remote_ifindex,
4622 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4627 static int vxlan_switchdev_event(struct notifier_block *unused,
4628 unsigned long event, void *ptr)
4630 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4631 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4635 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4636 vxlan_fdb_offloaded_set(dev, ptr);
4638 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4640 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4642 err = notifier_from_errno(err);
4645 fdb_info->offloaded = true;
4646 vxlan_fdb_offloaded_set(dev, fdb_info);
4648 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4650 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4652 err = notifier_from_errno(err);
4655 fdb_info->offloaded = false;
4656 vxlan_fdb_offloaded_set(dev, fdb_info);
4663 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4664 .notifier_call = vxlan_switchdev_event,
4667 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4669 struct vxlan_fdb *fdb;
4670 struct vxlan_dev *vxlan;
4674 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4675 vxlan = rcu_dereference(fdb->vdev);
4677 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4678 vxlan->default_dst.remote_vni);
4679 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4680 if (!hlist_unhashed(&fdb->hlist))
4681 vxlan_fdb_destroy(vxlan, fdb, false, false);
4682 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4687 static int vxlan_nexthop_event(struct notifier_block *nb,
4688 unsigned long event, void *ptr)
4690 struct nexthop *nh = ptr;
4692 if (!nh || event != NEXTHOP_EVENT_DEL)
4695 vxlan_fdb_nh_flush(nh);
4700 static struct notifier_block vxlan_nexthop_notifier_block __read_mostly = {
4701 .notifier_call = vxlan_nexthop_event,
4704 static __net_init int vxlan_init_net(struct net *net)
4706 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4709 INIT_LIST_HEAD(&vn->vxlan_list);
4710 spin_lock_init(&vn->sock_lock);
4712 for (h = 0; h < PORT_HASH_SIZE; ++h)
4713 INIT_HLIST_HEAD(&vn->sock_list[h]);
4715 return register_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4718 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4720 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4721 struct vxlan_dev *vxlan, *next;
4722 struct net_device *dev, *aux;
4725 for_each_netdev_safe(net, dev, aux)
4726 if (dev->rtnl_link_ops == &vxlan_link_ops)
4727 unregister_netdevice_queue(dev, head);
4729 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4730 /* If vxlan->dev is in the same netns, it has already been added
4731 * to the list by the previous loop.
4733 if (!net_eq(dev_net(vxlan->dev), net))
4734 unregister_netdevice_queue(vxlan->dev, head);
4737 for (h = 0; h < PORT_HASH_SIZE; ++h)
4738 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4741 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4747 list_for_each_entry(net, net_list, exit_list)
4748 unregister_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4749 list_for_each_entry(net, net_list, exit_list)
4750 vxlan_destroy_tunnels(net, &list);
4752 unregister_netdevice_many(&list);
4756 static struct pernet_operations vxlan_net_ops = {
4757 .init = vxlan_init_net,
4758 .exit_batch = vxlan_exit_batch_net,
4759 .id = &vxlan_net_id,
4760 .size = sizeof(struct vxlan_net),
4763 static int __init vxlan_init_module(void)
4767 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4769 rc = register_pernet_subsys(&vxlan_net_ops);
4773 rc = register_netdevice_notifier(&vxlan_notifier_block);
4777 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4781 rc = rtnl_link_register(&vxlan_link_ops);
4787 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4789 unregister_netdevice_notifier(&vxlan_notifier_block);
4791 unregister_pernet_subsys(&vxlan_net_ops);
4795 late_initcall(vxlan_init_module);
4797 static void __exit vxlan_cleanup_module(void)
4799 rtnl_link_unregister(&vxlan_link_ops);
4800 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4801 unregister_netdevice_notifier(&vxlan_notifier_block);
4802 unregister_pernet_subsys(&vxlan_net_ops);
4803 /* rcu_barrier() is called by netns */
4805 module_exit(vxlan_cleanup_module);
4807 MODULE_LICENSE("GPL");
4808 MODULE_VERSION(VXLAN_VERSION);
4809 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4810 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4811 MODULE_ALIAS_RTNL_LINK("vxlan");