2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/errno.h>
20 #include <linux/slab.h>
21 #include <linux/skbuff.h>
22 #include <linux/rculist.h>
23 #include <linux/netdevice.h>
26 #include <linux/udp.h>
27 #include <linux/igmp.h>
28 #include <linux/etherdevice.h>
29 #include <linux/if_ether.h>
30 #include <linux/hash.h>
31 #include <linux/ethtool.h>
33 #include <net/ndisc.h>
35 #include <net/ip_tunnels.h>
38 #include <net/rtnetlink.h>
39 #include <net/route.h>
40 #include <net/dsfield.h>
41 #include <net/inet_ecn.h>
42 #include <net/net_namespace.h>
43 #include <net/netns/generic.h>
45 #define VXLAN_VERSION "0.1"
47 #define PORT_HASH_BITS 8
48 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
49 #define VNI_HASH_BITS 10
50 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
51 #define FDB_HASH_BITS 8
52 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
53 #define FDB_AGE_DEFAULT 300 /* 5 min */
54 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
56 #define VXLAN_N_VID (1u << 24)
57 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
58 /* IP header + UDP + VXLAN + Ethernet header */
59 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
61 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
63 /* VXLAN protocol header */
69 /* UDP port for VXLAN traffic.
70 * The IANA assigned port is 4789, but the Linux default is 8472
71 * for compatability with early adopters.
73 static unsigned int vxlan_port __read_mostly = 8472;
74 module_param_named(udp_port, vxlan_port, uint, 0444);
75 MODULE_PARM_DESC(udp_port, "Destination UDP port");
77 static bool log_ecn_error = true;
78 module_param(log_ecn_error, bool, 0644);
79 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
81 static unsigned int vxlan_net_id;
83 /* per UDP socket information */
85 struct hlist_node hlist;
87 struct work_struct del_work;
90 struct hlist_head vni_list[VNI_HASH_SIZE];
93 /* per-network namespace private data for this module */
95 struct list_head vxlan_list;
96 struct hlist_head sock_list[PORT_HASH_SIZE];
104 struct vxlan_rdst *remote_next;
107 /* Forwarding table entry */
109 struct hlist_node hlist; /* linked list of entries */
111 unsigned long updated; /* jiffies */
113 struct vxlan_rdst remote;
114 u16 state; /* see ndm_state */
115 u8 flags; /* see ndm_flags */
116 u8 eth_addr[ETH_ALEN];
119 /* Pseudo network device */
121 struct hlist_node hlist; /* vni hash table */
122 struct list_head next; /* vxlan's per namespace list */
123 struct vxlan_sock *vn_sock; /* listening socket */
124 struct net_device *dev;
125 struct vxlan_rdst default_dst; /* default destination */
126 __be32 saddr; /* source address */
128 __u16 port_min; /* source port range */
130 __u8 tos; /* TOS override */
132 u32 flags; /* VXLAN_F_* below */
134 struct work_struct igmp_work;
135 unsigned long age_interval;
136 struct timer_list age_timer;
137 spinlock_t hash_lock;
138 unsigned int addrcnt;
139 unsigned int addrmax;
141 struct hlist_head fdb_head[FDB_HASH_SIZE];
144 #define VXLAN_F_LEARN 0x01
145 #define VXLAN_F_PROXY 0x02
146 #define VXLAN_F_RSC 0x04
147 #define VXLAN_F_L2MISS 0x08
148 #define VXLAN_F_L3MISS 0x10
150 /* salt for hash table */
151 static u32 vxlan_salt __read_mostly;
152 static struct workqueue_struct *vxlan_wq;
154 /* Virtual Network hash table head */
155 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
157 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
160 /* Socket hash table head */
161 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
163 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
165 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
168 /* Find VXLAN socket based on network namespace and UDP port */
169 static struct vxlan_sock *vxlan_find_port(struct net *net, __be16 port)
171 struct vxlan_sock *vs;
173 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
174 if (inet_sk(vs->sock->sk)->inet_sport == port)
180 /* Look up VNI in a per net namespace table */
181 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
183 struct vxlan_sock *vs;
184 struct vxlan_dev *vxlan;
186 vs = vxlan_find_port(net, port);
190 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
191 if (vxlan->default_dst.remote_vni == id)
198 /* Fill in neighbour message in skbuff. */
199 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
200 const struct vxlan_fdb *fdb,
201 u32 portid, u32 seq, int type, unsigned int flags,
202 const struct vxlan_rdst *rdst)
204 unsigned long now = jiffies;
205 struct nda_cacheinfo ci;
206 struct nlmsghdr *nlh;
208 bool send_ip, send_eth;
210 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
214 ndm = nlmsg_data(nlh);
215 memset(ndm, 0, sizeof(*ndm));
217 send_eth = send_ip = true;
219 if (type == RTM_GETNEIGH) {
220 ndm->ndm_family = AF_INET;
221 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
222 send_eth = !is_zero_ether_addr(fdb->eth_addr);
224 ndm->ndm_family = AF_BRIDGE;
225 ndm->ndm_state = fdb->state;
226 ndm->ndm_ifindex = vxlan->dev->ifindex;
227 ndm->ndm_flags = fdb->flags;
228 ndm->ndm_type = NDA_DST;
230 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
231 goto nla_put_failure;
233 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
234 goto nla_put_failure;
236 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
237 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
238 goto nla_put_failure;
239 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
240 nla_put_be32(skb, NDA_VNI, rdst->remote_vni))
241 goto nla_put_failure;
242 if (rdst->remote_ifindex &&
243 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
244 goto nla_put_failure;
246 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
247 ci.ndm_confirmed = 0;
248 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
251 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
252 goto nla_put_failure;
254 return nlmsg_end(skb, nlh);
257 nlmsg_cancel(skb, nlh);
261 static inline size_t vxlan_nlmsg_size(void)
263 return NLMSG_ALIGN(sizeof(struct ndmsg))
264 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
265 + nla_total_size(sizeof(__be32)) /* NDA_DST */
266 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
267 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
268 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
269 + nla_total_size(sizeof(struct nda_cacheinfo));
272 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
273 const struct vxlan_fdb *fdb, int type)
275 struct net *net = dev_net(vxlan->dev);
279 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
283 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, &fdb->remote);
285 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
286 WARN_ON(err == -EMSGSIZE);
291 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
295 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
298 static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
300 struct vxlan_dev *vxlan = netdev_priv(dev);
303 memset(&f, 0, sizeof f);
305 f.remote.remote_ip = ipa; /* goes to NDA_DST */
306 f.remote.remote_vni = VXLAN_N_VID;
308 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
311 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
315 memset(&f, 0, sizeof f);
317 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
319 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
322 /* Hash Ethernet address */
323 static u32 eth_hash(const unsigned char *addr)
325 u64 value = get_unaligned((u64 *)addr);
327 /* only want 6 bytes */
333 return hash_64(value, FDB_HASH_BITS);
336 /* Hash chain to use given mac address */
337 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
340 return &vxlan->fdb_head[eth_hash(mac)];
343 /* Look up Ethernet address in forwarding table */
344 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
348 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
351 hlist_for_each_entry_rcu(f, head, hlist) {
352 if (compare_ether_addr(mac, f->eth_addr) == 0)
359 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
364 f = __vxlan_find_mac(vxlan, mac);
371 /* Add/update destinations for multicast */
372 static int vxlan_fdb_append(struct vxlan_fdb *f,
373 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
375 struct vxlan_rdst *rd_prev, *rd;
378 for (rd = &f->remote; rd; rd = rd->remote_next) {
379 if (rd->remote_ip == ip &&
380 rd->remote_port == port &&
381 rd->remote_vni == vni &&
382 rd->remote_ifindex == ifindex)
386 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
390 rd->remote_port = port;
391 rd->remote_vni = vni;
392 rd->remote_ifindex = ifindex;
393 rd->remote_next = NULL;
394 rd_prev->remote_next = rd;
398 /* Add new entry to forwarding table -- assumes lock held */
399 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
400 const u8 *mac, __be32 ip,
401 __u16 state, __u16 flags,
402 __be16 port, __u32 vni, __u32 ifindex,
408 f = __vxlan_find_mac(vxlan, mac);
410 if (flags & NLM_F_EXCL) {
411 netdev_dbg(vxlan->dev,
412 "lost race to create %pM\n", mac);
415 if (f->state != state) {
417 f->updated = jiffies;
420 if (f->flags != ndm_flags) {
421 f->flags = ndm_flags;
422 f->updated = jiffies;
425 if ((flags & NLM_F_APPEND) &&
426 is_multicast_ether_addr(f->eth_addr)) {
427 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
434 if (!(flags & NLM_F_CREATE))
437 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
440 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
441 f = kmalloc(sizeof(*f), GFP_ATOMIC);
446 f->remote.remote_ip = ip;
447 f->remote.remote_port = port;
448 f->remote.remote_vni = vni;
449 f->remote.remote_ifindex = ifindex;
450 f->remote.remote_next = NULL;
452 f->flags = ndm_flags;
453 f->updated = f->used = jiffies;
454 memcpy(f->eth_addr, mac, ETH_ALEN);
457 hlist_add_head_rcu(&f->hlist,
458 vxlan_fdb_head(vxlan, mac));
462 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
467 static void vxlan_fdb_free(struct rcu_head *head)
469 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
471 while (f->remote.remote_next) {
472 struct vxlan_rdst *rd = f->remote.remote_next;
474 f->remote.remote_next = rd->remote_next;
480 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
482 netdev_dbg(vxlan->dev,
483 "delete %pM\n", f->eth_addr);
486 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
488 hlist_del_rcu(&f->hlist);
489 call_rcu(&f->rcu, vxlan_fdb_free);
492 /* Add static entry (via netlink) */
493 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
494 struct net_device *dev,
495 const unsigned char *addr, u16 flags)
497 struct vxlan_dev *vxlan = netdev_priv(dev);
498 struct net *net = dev_net(vxlan->dev);
504 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
505 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
510 if (tb[NDA_DST] == NULL)
513 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
514 return -EAFNOSUPPORT;
516 ip = nla_get_be32(tb[NDA_DST]);
519 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
521 port = nla_get_be16(tb[NDA_PORT]);
523 port = vxlan->dst_port;
526 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
528 vni = nla_get_u32(tb[NDA_VNI]);
530 vni = vxlan->default_dst.remote_vni;
532 if (tb[NDA_IFINDEX]) {
533 struct net_device *tdev;
535 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
537 ifindex = nla_get_u32(tb[NDA_IFINDEX]);
538 tdev = dev_get_by_index(net, ifindex);
540 return -EADDRNOTAVAIL;
545 spin_lock_bh(&vxlan->hash_lock);
546 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
547 port, vni, ifindex, ndm->ndm_flags);
548 spin_unlock_bh(&vxlan->hash_lock);
553 /* Delete entry (via netlink) */
554 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
555 struct net_device *dev,
556 const unsigned char *addr)
558 struct vxlan_dev *vxlan = netdev_priv(dev);
562 spin_lock_bh(&vxlan->hash_lock);
563 f = vxlan_find_mac(vxlan, addr);
565 vxlan_fdb_destroy(vxlan, f);
568 spin_unlock_bh(&vxlan->hash_lock);
573 /* Dump forwarding table */
574 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
575 struct net_device *dev, int idx)
577 struct vxlan_dev *vxlan = netdev_priv(dev);
580 for (h = 0; h < FDB_HASH_SIZE; ++h) {
584 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
585 struct vxlan_rdst *rd;
586 for (rd = &f->remote; rd; rd = rd->remote_next) {
587 if (idx < cb->args[0])
590 err = vxlan_fdb_info(skb, vxlan, f,
591 NETLINK_CB(cb->skb).portid,
606 /* Watch incoming packets to learn mapping between Ethernet address
607 * and Tunnel endpoint.
608 * Return true if packet is bogus and should be droppped.
610 static bool vxlan_snoop(struct net_device *dev,
611 __be32 src_ip, const u8 *src_mac)
613 struct vxlan_dev *vxlan = netdev_priv(dev);
616 f = vxlan_find_mac(vxlan, src_mac);
618 if (likely(f->remote.remote_ip == src_ip))
621 /* Don't migrate static entries, drop packets */
622 if (f->state & NUD_NOARP)
627 "%pM migrated from %pI4 to %pI4\n",
628 src_mac, &f->remote.remote_ip, &src_ip);
630 f->remote.remote_ip = src_ip;
631 f->updated = jiffies;
632 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
634 /* learned new entry */
635 spin_lock(&vxlan->hash_lock);
637 /* close off race between vxlan_flush and incoming packets */
638 if (netif_running(dev))
639 vxlan_fdb_create(vxlan, src_mac, src_ip,
641 NLM_F_EXCL|NLM_F_CREATE,
643 vxlan->default_dst.remote_vni,
645 spin_unlock(&vxlan->hash_lock);
652 /* See if multicast group is already in use by other ID */
653 static bool vxlan_group_used(struct vxlan_net *vn, __be32 remote_ip)
655 struct vxlan_dev *vxlan;
657 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
658 if (!netif_running(vxlan->dev))
661 if (vxlan->default_dst.remote_ip == remote_ip)
668 static void vxlan_sock_hold(struct vxlan_sock *vs)
670 atomic_inc(&vs->refcnt);
673 static void vxlan_sock_release(struct vxlan_sock *vs)
675 if (!atomic_dec_and_test(&vs->refcnt))
678 hlist_del_rcu(&vs->hlist);
679 queue_work(vxlan_wq, &vs->del_work);
682 /* Callback to update multicast group membership.
683 * Scheduled when vxlan goes up/down.
685 static void vxlan_igmp_work(struct work_struct *work)
687 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_work);
688 struct vxlan_net *vn = net_generic(dev_net(vxlan->dev), vxlan_net_id);
689 struct vxlan_sock *vs = vxlan->vn_sock;
690 struct sock *sk = vs->sock->sk;
691 struct ip_mreqn mreq = {
692 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
693 .imr_ifindex = vxlan->default_dst.remote_ifindex,
697 if (vxlan_group_used(vn, vxlan->default_dst.remote_ip))
698 ip_mc_join_group(sk, &mreq);
700 ip_mc_leave_group(sk, &mreq);
703 vxlan_sock_release(vs);
707 /* Callback from net/ipv4/udp.c to receive packets */
708 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
711 struct vxlanhdr *vxh;
712 struct vxlan_dev *vxlan;
713 struct pcpu_tstats *stats;
718 /* pop off outer UDP header */
719 __skb_pull(skb, sizeof(struct udphdr));
721 /* Need Vxlan and inner Ethernet header to be present */
722 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
725 /* Drop packets with reserved bits set */
726 vxh = (struct vxlanhdr *) skb->data;
727 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
728 (vxh->vx_vni & htonl(0xff))) {
729 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
730 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
734 __skb_pull(skb, sizeof(struct vxlanhdr));
736 /* Is this VNI defined? */
737 vni = ntohl(vxh->vx_vni) >> 8;
738 port = inet_sk(sk)->inet_sport;
739 vxlan = vxlan_find_vni(sock_net(sk), vni, port);
741 netdev_dbg(skb->dev, "unknown vni %d port %u\n",
746 if (!pskb_may_pull(skb, ETH_HLEN)) {
747 vxlan->dev->stats.rx_length_errors++;
748 vxlan->dev->stats.rx_errors++;
752 skb_reset_mac_header(skb);
754 /* Re-examine inner Ethernet packet */
756 skb->protocol = eth_type_trans(skb, vxlan->dev);
758 /* Ignore packet loops (and multicast echo) */
759 if (compare_ether_addr(eth_hdr(skb)->h_source,
760 vxlan->dev->dev_addr) == 0)
763 if ((vxlan->flags & VXLAN_F_LEARN) &&
764 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source))
767 __skb_tunnel_rx(skb, vxlan->dev);
768 skb_reset_network_header(skb);
770 /* If the NIC driver gave us an encapsulated packet with
771 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
772 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
773 * for us. Otherwise force the upper layers to verify it.
775 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
776 !(vxlan->dev->features & NETIF_F_RXCSUM))
777 skb->ip_summed = CHECKSUM_NONE;
779 skb->encapsulation = 0;
781 err = IP_ECN_decapsulate(oip, skb);
784 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
785 &oip->saddr, oip->tos);
787 ++vxlan->dev->stats.rx_frame_errors;
788 ++vxlan->dev->stats.rx_errors;
793 stats = this_cpu_ptr(vxlan->dev->tstats);
794 u64_stats_update_begin(&stats->syncp);
796 stats->rx_bytes += skb->len;
797 u64_stats_update_end(&stats->syncp);
803 /* Put UDP header back */
804 __skb_push(skb, sizeof(struct udphdr));
808 /* Consume bad packet */
813 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
815 struct vxlan_dev *vxlan = netdev_priv(dev);
821 if (dev->flags & IFF_NOARP)
824 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
825 dev->stats.tx_dropped++;
830 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
831 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
832 parp->ar_pro != htons(ETH_P_IP) ||
833 parp->ar_op != htons(ARPOP_REQUEST) ||
834 parp->ar_hln != dev->addr_len ||
837 arpptr = (u8 *)parp + sizeof(struct arphdr);
839 arpptr += dev->addr_len; /* sha */
840 memcpy(&sip, arpptr, sizeof(sip));
841 arpptr += sizeof(sip);
842 arpptr += dev->addr_len; /* tha */
843 memcpy(&tip, arpptr, sizeof(tip));
845 if (ipv4_is_loopback(tip) ||
846 ipv4_is_multicast(tip))
849 n = neigh_lookup(&arp_tbl, &tip, dev);
853 struct sk_buff *reply;
855 if (!(n->nud_state & NUD_CONNECTED)) {
860 f = vxlan_find_mac(vxlan, n->ha);
861 if (f && f->remote.remote_ip == htonl(INADDR_ANY)) {
862 /* bridge-local neighbor */
867 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
872 skb_reset_mac_header(reply);
873 __skb_pull(reply, skb_network_offset(reply));
874 reply->ip_summed = CHECKSUM_UNNECESSARY;
875 reply->pkt_type = PACKET_HOST;
877 if (netif_rx_ni(reply) == NET_RX_DROP)
878 dev->stats.rx_dropped++;
879 } else if (vxlan->flags & VXLAN_F_L3MISS)
880 vxlan_ip_miss(dev, tip);
886 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
888 struct vxlan_dev *vxlan = netdev_priv(dev);
892 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
896 switch (ntohs(eth_hdr(skb)->h_proto)) {
898 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
901 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
910 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
912 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
914 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
918 } else if (vxlan->flags & VXLAN_F_L3MISS)
919 vxlan_ip_miss(dev, pip->daddr);
923 static void vxlan_sock_put(struct sk_buff *skb)
928 /* On transmit, associate with the tunnel socket */
929 static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
931 struct vxlan_dev *vxlan = netdev_priv(dev);
932 struct sock *sk = vxlan->vn_sock->sock->sk;
937 skb->destructor = vxlan_sock_put;
940 /* Compute source port for outgoing packet
941 * first choice to use L4 flow hash since it will spread
942 * better and maybe available from hardware
943 * secondary choice is to use jhash on the Ethernet header
945 static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
947 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
950 hash = skb_get_rxhash(skb);
952 hash = jhash(skb->data, 2 * ETH_ALEN,
953 (__force u32) skb->protocol);
955 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
958 static int handle_offloads(struct sk_buff *skb)
960 if (skb_is_gso(skb)) {
961 int err = skb_unclone(skb, GFP_ATOMIC);
965 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
966 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
967 skb->ip_summed = CHECKSUM_NONE;
972 /* Bypass encapsulation if the destination is local */
973 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
974 struct vxlan_dev *dst_vxlan)
976 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
977 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
979 skb->pkt_type = PACKET_HOST;
980 skb->encapsulation = 0;
981 skb->dev = dst_vxlan->dev;
982 __skb_pull(skb, skb_network_offset(skb));
984 if (dst_vxlan->flags & VXLAN_F_LEARN)
985 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
986 eth_hdr(skb)->h_source);
988 u64_stats_update_begin(&tx_stats->syncp);
989 tx_stats->tx_packets++;
990 tx_stats->tx_bytes += skb->len;
991 u64_stats_update_end(&tx_stats->syncp);
993 if (netif_rx(skb) == NET_RX_SUCCESS) {
994 u64_stats_update_begin(&rx_stats->syncp);
995 rx_stats->rx_packets++;
996 rx_stats->rx_bytes += skb->len;
997 u64_stats_update_end(&rx_stats->syncp);
999 skb->dev->stats.rx_dropped++;
1003 static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1004 struct vxlan_rdst *rdst, bool did_rsc)
1006 struct vxlan_dev *vxlan = netdev_priv(dev);
1008 const struct iphdr *old_iph;
1009 struct vxlanhdr *vxh;
1013 __be16 src_port, dst_port;
1019 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1020 vni = rdst->remote_vni;
1021 dst = rdst->remote_ip;
1025 /* short-circuited back to local bridge */
1026 vxlan_encap_bypass(skb, vxlan, vxlan);
1027 return NETDEV_TX_OK;
1032 if (!skb->encapsulation) {
1033 skb_reset_inner_headers(skb);
1034 skb->encapsulation = 1;
1037 /* Need space for new headers (invalidates iph ptr) */
1038 if (skb_cow_head(skb, VXLAN_HEADROOM))
1041 old_iph = ip_hdr(skb);
1044 if (!ttl && IN_MULTICAST(ntohl(dst)))
1049 tos = ip_tunnel_get_dsfield(old_iph, skb);
1051 src_port = vxlan_src_port(vxlan, skb);
1053 memset(&fl4, 0, sizeof(fl4));
1054 fl4.flowi4_oif = rdst->remote_ifindex;
1055 fl4.flowi4_tos = RT_TOS(tos);
1057 fl4.saddr = vxlan->saddr;
1059 rt = ip_route_output_key(dev_net(dev), &fl4);
1061 netdev_dbg(dev, "no route to %pI4\n", &dst);
1062 dev->stats.tx_carrier_errors++;
1066 if (rt->dst.dev == dev) {
1067 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1069 dev->stats.collisions++;
1073 /* Bypass encapsulation if the destination is local */
1074 if (rt->rt_flags & RTCF_LOCAL &&
1075 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1076 struct vxlan_dev *dst_vxlan;
1079 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1082 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1083 return NETDEV_TX_OK;
1085 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1086 vxh->vx_flags = htonl(VXLAN_FLAGS);
1087 vxh->vx_vni = htonl(vni << 8);
1089 __skb_push(skb, sizeof(*uh));
1090 skb_reset_transport_header(skb);
1093 uh->dest = dst_port;
1094 uh->source = src_port;
1096 uh->len = htons(skb->len);
1099 vxlan_set_owner(dev, skb);
1101 if (handle_offloads(skb))
1104 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1105 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1107 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, dst,
1108 IPPROTO_UDP, tos, ttl, df);
1109 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1111 return NETDEV_TX_OK;
1114 dev->stats.tx_dropped++;
1118 dev->stats.tx_errors++;
1121 return NETDEV_TX_OK;
1124 /* Transmit local packets over Vxlan
1126 * Outer IP header inherits ECN and DF from inner header.
1127 * Outer UDP destination is the VXLAN assigned port.
1128 * source port is based on hash of flow
1130 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1132 struct vxlan_dev *vxlan = netdev_priv(dev);
1134 bool did_rsc = false;
1135 struct vxlan_rdst *rdst0, *rdst;
1136 struct vxlan_fdb *f;
1139 skb_reset_mac_header(skb);
1142 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1143 return arp_reduce(dev, skb);
1145 f = vxlan_find_mac(vxlan, eth->h_dest);
1148 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1149 ntohs(eth->h_proto) == ETH_P_IP) {
1150 did_rsc = route_shortcircuit(dev, skb);
1152 f = vxlan_find_mac(vxlan, eth->h_dest);
1156 rdst0 = &vxlan->default_dst;
1158 if (rdst0->remote_ip == htonl(INADDR_ANY) &&
1159 (vxlan->flags & VXLAN_F_L2MISS) &&
1160 !is_multicast_ether_addr(eth->h_dest))
1161 vxlan_fdb_miss(vxlan, eth->h_dest);
1167 /* if there are multiple destinations, send copies */
1168 for (rdst = rdst0->remote_next; rdst; rdst = rdst->remote_next) {
1169 struct sk_buff *skb1;
1171 skb1 = skb_clone(skb, GFP_ATOMIC);
1173 rc1 = vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1174 if (rc == NETDEV_TX_OK)
1179 rc1 = vxlan_xmit_one(skb, dev, rdst0, did_rsc);
1180 if (rc == NETDEV_TX_OK)
1185 /* Walk the forwarding table and purge stale entries */
1186 static void vxlan_cleanup(unsigned long arg)
1188 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1189 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1192 if (!netif_running(vxlan->dev))
1195 spin_lock_bh(&vxlan->hash_lock);
1196 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1197 struct hlist_node *p, *n;
1198 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1200 = container_of(p, struct vxlan_fdb, hlist);
1201 unsigned long timeout;
1203 if (f->state & NUD_PERMANENT)
1206 timeout = f->used + vxlan->age_interval * HZ;
1207 if (time_before_eq(timeout, jiffies)) {
1208 netdev_dbg(vxlan->dev,
1209 "garbage collect %pM\n",
1211 f->state = NUD_STALE;
1212 vxlan_fdb_destroy(vxlan, f);
1213 } else if (time_before(timeout, next_timer))
1214 next_timer = timeout;
1217 spin_unlock_bh(&vxlan->hash_lock);
1219 mod_timer(&vxlan->age_timer, next_timer);
1222 /* Setup stats when device is created */
1223 static int vxlan_init(struct net_device *dev)
1225 dev->tstats = alloc_percpu(struct pcpu_tstats);
1232 /* Start ageing timer and join group when device is brought up */
1233 static int vxlan_open(struct net_device *dev)
1235 struct vxlan_dev *vxlan = netdev_priv(dev);
1237 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
1238 vxlan_sock_hold(vxlan->vn_sock);
1240 queue_work(vxlan_wq, &vxlan->igmp_work);
1243 if (vxlan->age_interval)
1244 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1249 /* Purge the forwarding table */
1250 static void vxlan_flush(struct vxlan_dev *vxlan)
1254 spin_lock_bh(&vxlan->hash_lock);
1255 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1256 struct hlist_node *p, *n;
1257 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1259 = container_of(p, struct vxlan_fdb, hlist);
1260 vxlan_fdb_destroy(vxlan, f);
1263 spin_unlock_bh(&vxlan->hash_lock);
1266 /* Cleanup timer and forwarding table on shutdown */
1267 static int vxlan_stop(struct net_device *dev)
1269 struct vxlan_dev *vxlan = netdev_priv(dev);
1271 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
1272 vxlan_sock_hold(vxlan->vn_sock);
1274 queue_work(vxlan_wq, &vxlan->igmp_work);
1277 del_timer_sync(&vxlan->age_timer);
1284 /* Stub, nothing needs to be done. */
1285 static void vxlan_set_multicast_list(struct net_device *dev)
1289 static const struct net_device_ops vxlan_netdev_ops = {
1290 .ndo_init = vxlan_init,
1291 .ndo_open = vxlan_open,
1292 .ndo_stop = vxlan_stop,
1293 .ndo_start_xmit = vxlan_xmit,
1294 .ndo_get_stats64 = ip_tunnel_get_stats64,
1295 .ndo_set_rx_mode = vxlan_set_multicast_list,
1296 .ndo_change_mtu = eth_change_mtu,
1297 .ndo_validate_addr = eth_validate_addr,
1298 .ndo_set_mac_address = eth_mac_addr,
1299 .ndo_fdb_add = vxlan_fdb_add,
1300 .ndo_fdb_del = vxlan_fdb_delete,
1301 .ndo_fdb_dump = vxlan_fdb_dump,
1304 /* Info for udev, that this is a virtual tunnel endpoint */
1305 static struct device_type vxlan_type = {
1309 static void vxlan_free(struct net_device *dev)
1311 free_percpu(dev->tstats);
1315 /* Initialize the device structure. */
1316 static void vxlan_setup(struct net_device *dev)
1318 struct vxlan_dev *vxlan = netdev_priv(dev);
1322 eth_hw_addr_random(dev);
1324 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
1326 dev->netdev_ops = &vxlan_netdev_ops;
1327 dev->destructor = vxlan_free;
1328 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1330 dev->tx_queue_len = 0;
1331 dev->features |= NETIF_F_LLTX;
1332 dev->features |= NETIF_F_NETNS_LOCAL;
1333 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1334 dev->features |= NETIF_F_RXCSUM;
1335 dev->features |= NETIF_F_GSO_SOFTWARE;
1337 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1338 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1339 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1340 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1342 INIT_LIST_HEAD(&vxlan->next);
1343 spin_lock_init(&vxlan->hash_lock);
1344 INIT_WORK(&vxlan->igmp_work, vxlan_igmp_work);
1346 init_timer_deferrable(&vxlan->age_timer);
1347 vxlan->age_timer.function = vxlan_cleanup;
1348 vxlan->age_timer.data = (unsigned long) vxlan;
1350 inet_get_local_port_range(&low, &high);
1351 vxlan->port_min = low;
1352 vxlan->port_max = high;
1353 vxlan->dst_port = htons(vxlan_port);
1357 for (h = 0; h < FDB_HASH_SIZE; ++h)
1358 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1361 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1362 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
1363 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1364 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1365 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1366 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1367 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1368 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1369 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1370 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
1371 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
1372 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1373 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1374 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1375 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
1376 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
1379 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1381 if (tb[IFLA_ADDRESS]) {
1382 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1383 pr_debug("invalid link address (not ethernet)\n");
1387 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1388 pr_debug("invalid all zero ethernet address\n");
1389 return -EADDRNOTAVAIL;
1396 if (data[IFLA_VXLAN_ID]) {
1397 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1398 if (id >= VXLAN_VID_MASK)
1402 if (data[IFLA_VXLAN_PORT_RANGE]) {
1403 const struct ifla_vxlan_port_range *p
1404 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1406 if (ntohs(p->high) < ntohs(p->low)) {
1407 pr_debug("port range %u .. %u not valid\n",
1408 ntohs(p->low), ntohs(p->high));
1416 static void vxlan_get_drvinfo(struct net_device *netdev,
1417 struct ethtool_drvinfo *drvinfo)
1419 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1420 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1423 static const struct ethtool_ops vxlan_ethtool_ops = {
1424 .get_drvinfo = vxlan_get_drvinfo,
1425 .get_link = ethtool_op_get_link,
1428 static void vxlan_del_work(struct work_struct *work)
1430 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
1432 sk_release_kernel(vs->sock->sk);
1436 /* Create new listen socket if needed */
1437 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port)
1439 struct vxlan_sock *vs;
1441 struct sockaddr_in vxlan_addr = {
1442 .sin_family = AF_INET,
1443 .sin_addr.s_addr = htonl(INADDR_ANY),
1448 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
1450 return ERR_PTR(-ENOMEM);
1452 for (h = 0; h < VNI_HASH_SIZE; ++h)
1453 INIT_HLIST_HEAD(&vs->vni_list[h]);
1455 INIT_WORK(&vs->del_work, vxlan_del_work);
1457 /* Create UDP socket for encapsulation receive. */
1458 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vs->sock);
1460 pr_debug("UDP socket create failed\n");
1465 /* Put in proper namespace */
1467 sk_change_net(sk, net);
1469 vxlan_addr.sin_port = port;
1471 rc = kernel_bind(vs->sock, (struct sockaddr *) &vxlan_addr,
1472 sizeof(vxlan_addr));
1474 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1475 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
1476 sk_release_kernel(sk);
1481 /* Disable multicast loopback */
1482 inet_sk(sk)->mc_loop = 0;
1484 /* Mark socket as an encapsulation socket. */
1485 udp_sk(sk)->encap_type = 1;
1486 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1488 atomic_set(&vs->refcnt, 1);
1493 static int vxlan_newlink(struct net *net, struct net_device *dev,
1494 struct nlattr *tb[], struct nlattr *data[])
1496 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1497 struct vxlan_dev *vxlan = netdev_priv(dev);
1498 struct vxlan_rdst *dst = &vxlan->default_dst;
1499 struct vxlan_sock *vs;
1503 if (!data[IFLA_VXLAN_ID])
1506 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1507 dst->remote_vni = vni;
1509 if (data[IFLA_VXLAN_GROUP])
1510 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
1512 if (data[IFLA_VXLAN_LOCAL])
1513 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1515 if (data[IFLA_VXLAN_LINK] &&
1516 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
1517 struct net_device *lowerdev
1518 = __dev_get_by_index(net, dst->remote_ifindex);
1521 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
1526 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1528 /* update header length based on lower device */
1529 dev->hard_header_len = lowerdev->hard_header_len +
1533 if (data[IFLA_VXLAN_TOS])
1534 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1536 if (data[IFLA_VXLAN_TTL])
1537 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1539 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
1540 vxlan->flags |= VXLAN_F_LEARN;
1542 if (data[IFLA_VXLAN_AGEING])
1543 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1545 vxlan->age_interval = FDB_AGE_DEFAULT;
1547 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1548 vxlan->flags |= VXLAN_F_PROXY;
1550 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1551 vxlan->flags |= VXLAN_F_RSC;
1553 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1554 vxlan->flags |= VXLAN_F_L2MISS;
1556 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1557 vxlan->flags |= VXLAN_F_L3MISS;
1559 if (data[IFLA_VXLAN_LIMIT])
1560 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1562 if (data[IFLA_VXLAN_PORT_RANGE]) {
1563 const struct ifla_vxlan_port_range *p
1564 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1565 vxlan->port_min = ntohs(p->low);
1566 vxlan->port_max = ntohs(p->high);
1569 if (data[IFLA_VXLAN_PORT])
1570 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1572 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
1573 pr_info("duplicate VNI %u\n", vni);
1577 vs = vxlan_find_port(net, vxlan->dst_port);
1579 atomic_inc(&vs->refcnt);
1581 /* Drop lock because socket create acquires RTNL lock */
1583 vs = vxlan_socket_create(net, vxlan->dst_port);
1588 hlist_add_head_rcu(&vs->hlist, vs_head(net, vxlan->dst_port));
1590 vxlan->vn_sock = vs;
1592 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1594 err = register_netdevice(dev);
1596 vxlan_sock_release(vs);
1600 list_add(&vxlan->next, &vn->vxlan_list);
1601 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1606 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1608 struct vxlan_dev *vxlan = netdev_priv(dev);
1609 struct vxlan_sock *vs = vxlan->vn_sock;
1611 hlist_del_rcu(&vxlan->hlist);
1612 list_del(&vxlan->next);
1613 unregister_netdevice_queue(dev, head);
1614 vxlan_sock_release(vs);
1617 static size_t vxlan_get_size(const struct net_device *dev)
1620 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
1621 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
1622 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1623 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1624 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1625 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1626 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
1627 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1628 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1629 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1630 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
1631 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1632 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
1633 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
1634 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
1638 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1640 const struct vxlan_dev *vxlan = netdev_priv(dev);
1641 const struct vxlan_rdst *dst = &vxlan->default_dst;
1642 struct ifla_vxlan_port_range ports = {
1643 .low = htons(vxlan->port_min),
1644 .high = htons(vxlan->port_max),
1647 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
1648 goto nla_put_failure;
1650 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
1651 goto nla_put_failure;
1653 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
1654 goto nla_put_failure;
1656 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
1657 goto nla_put_failure;
1659 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1660 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
1661 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1662 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1663 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1664 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1665 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1666 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1667 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1668 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1669 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
1670 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
1671 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1672 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
1673 goto nla_put_failure;
1675 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1676 goto nla_put_failure;
1684 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1686 .maxtype = IFLA_VXLAN_MAX,
1687 .policy = vxlan_policy,
1688 .priv_size = sizeof(struct vxlan_dev),
1689 .setup = vxlan_setup,
1690 .validate = vxlan_validate,
1691 .newlink = vxlan_newlink,
1692 .dellink = vxlan_dellink,
1693 .get_size = vxlan_get_size,
1694 .fill_info = vxlan_fill_info,
1697 static __net_init int vxlan_init_net(struct net *net)
1699 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1702 INIT_LIST_HEAD(&vn->vxlan_list);
1704 for (h = 0; h < PORT_HASH_SIZE; ++h)
1705 INIT_HLIST_HEAD(&vn->sock_list[h]);
1710 static __net_exit void vxlan_exit_net(struct net *net)
1712 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1713 struct vxlan_dev *vxlan;
1716 list_for_each_entry(vxlan, &vn->vxlan_list, next)
1717 dev_close(vxlan->dev);
1721 static struct pernet_operations vxlan_net_ops = {
1722 .init = vxlan_init_net,
1723 .exit = vxlan_exit_net,
1724 .id = &vxlan_net_id,
1725 .size = sizeof(struct vxlan_net),
1728 static int __init vxlan_init_module(void)
1732 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
1736 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1738 rc = register_pernet_device(&vxlan_net_ops);
1742 rc = rtnl_link_register(&vxlan_link_ops);
1749 unregister_pernet_device(&vxlan_net_ops);
1751 destroy_workqueue(vxlan_wq);
1754 late_initcall(vxlan_init_module);
1756 static void __exit vxlan_cleanup_module(void)
1758 unregister_pernet_device(&vxlan_net_ops);
1759 rtnl_link_unregister(&vxlan_link_ops);
1760 destroy_workqueue(vxlan_wq);
1763 module_exit(vxlan_cleanup_module);
1765 MODULE_LICENSE("GPL");
1766 MODULE_VERSION(VXLAN_VERSION);
1767 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
1768 MODULE_ALIAS_RTNL_LINK("vxlan");