2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 * - use IANA UDP port number (when defined)
15 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/module.h>
20 #include <linux/errno.h>
21 #include <linux/slab.h>
22 #include <linux/skbuff.h>
23 #include <linux/rculist.h>
24 #include <linux/netdevice.h>
27 #include <linux/udp.h>
28 #include <linux/igmp.h>
29 #include <linux/etherdevice.h>
30 #include <linux/if_ether.h>
31 #include <linux/hash.h>
32 #include <linux/ethtool.h>
34 #include <net/ndisc.h>
36 #include <net/ip_tunnels.h>
39 #include <net/rtnetlink.h>
40 #include <net/route.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/net_namespace.h>
44 #include <net/netns/generic.h>
46 #define VXLAN_VERSION "0.1"
48 #define VNI_HASH_BITS 10
49 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
50 #define FDB_HASH_BITS 8
51 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
52 #define FDB_AGE_DEFAULT 300 /* 5 min */
53 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
55 #define VXLAN_N_VID (1u << 24)
56 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
57 /* IP header + UDP + VXLAN + Ethernet header */
58 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
60 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
62 /* VXLAN protocol header */
68 /* UDP port for VXLAN traffic. */
69 static unsigned int vxlan_port __read_mostly = 8472;
70 module_param_named(udp_port, vxlan_port, uint, 0444);
71 MODULE_PARM_DESC(udp_port, "Destination UDP port");
73 static bool log_ecn_error = true;
74 module_param(log_ecn_error, bool, 0644);
75 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
77 /* per-net private data for this module */
78 static unsigned int vxlan_net_id;
80 struct socket *sock; /* UDP encap socket */
81 struct hlist_head vni_list[VNI_HASH_SIZE];
90 struct vxlan_rdst *remote_next;
93 /* Forwarding table entry */
95 struct hlist_node hlist; /* linked list of entries */
97 unsigned long updated; /* jiffies */
99 struct vxlan_rdst remote;
100 u16 state; /* see ndm_state */
101 u8 eth_addr[ETH_ALEN];
104 /* Pseudo network device */
106 struct hlist_node hlist;
107 struct net_device *dev;
108 struct vxlan_rdst default_dst; /* default destination */
109 __be32 saddr; /* source address */
110 __u16 port_min; /* source port range */
112 __u8 tos; /* TOS override */
114 u32 flags; /* VXLAN_F_* below */
116 unsigned long age_interval;
117 struct timer_list age_timer;
118 spinlock_t hash_lock;
119 unsigned int addrcnt;
120 unsigned int addrmax;
122 struct hlist_head fdb_head[FDB_HASH_SIZE];
125 #define VXLAN_F_LEARN 0x01
126 #define VXLAN_F_PROXY 0x02
127 #define VXLAN_F_RSC 0x04
128 #define VXLAN_F_L2MISS 0x08
129 #define VXLAN_F_L3MISS 0x10
131 /* salt for hash table */
132 static u32 vxlan_salt __read_mostly;
134 static inline struct hlist_head *vni_head(struct net *net, u32 id)
136 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
138 return &vn->vni_list[hash_32(id, VNI_HASH_BITS)];
141 /* Look up VNI in a per net namespace table */
142 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id)
144 struct vxlan_dev *vxlan;
146 hlist_for_each_entry_rcu(vxlan, vni_head(net, id), hlist) {
147 if (vxlan->default_dst.remote_vni == id)
154 /* Fill in neighbour message in skbuff. */
155 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
156 const struct vxlan_fdb *fdb,
157 u32 portid, u32 seq, int type, unsigned int flags,
158 const struct vxlan_rdst *rdst)
160 unsigned long now = jiffies;
161 struct nda_cacheinfo ci;
162 struct nlmsghdr *nlh;
164 bool send_ip, send_eth;
166 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
170 ndm = nlmsg_data(nlh);
171 memset(ndm, 0, sizeof(*ndm));
173 send_eth = send_ip = true;
175 if (type == RTM_GETNEIGH) {
176 ndm->ndm_family = AF_INET;
177 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
178 send_eth = !is_zero_ether_addr(fdb->eth_addr);
180 ndm->ndm_family = AF_BRIDGE;
181 ndm->ndm_state = fdb->state;
182 ndm->ndm_ifindex = vxlan->dev->ifindex;
183 ndm->ndm_flags = NTF_SELF;
184 ndm->ndm_type = NDA_DST;
186 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
187 goto nla_put_failure;
189 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
190 goto nla_put_failure;
192 if (rdst->remote_port && rdst->remote_port != vxlan_port &&
193 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
194 goto nla_put_failure;
195 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
196 nla_put_be32(skb, NDA_VNI, rdst->remote_vni))
197 goto nla_put_failure;
198 if (rdst->remote_ifindex &&
199 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
200 goto nla_put_failure;
202 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
203 ci.ndm_confirmed = 0;
204 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
207 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
208 goto nla_put_failure;
210 return nlmsg_end(skb, nlh);
213 nlmsg_cancel(skb, nlh);
217 static inline size_t vxlan_nlmsg_size(void)
219 return NLMSG_ALIGN(sizeof(struct ndmsg))
220 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
221 + nla_total_size(sizeof(__be32)) /* NDA_DST */
222 + nla_total_size(sizeof(__be32)) /* NDA_PORT */
223 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
224 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
225 + nla_total_size(sizeof(struct nda_cacheinfo));
228 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
229 const struct vxlan_fdb *fdb, int type)
231 struct net *net = dev_net(vxlan->dev);
235 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
239 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, &fdb->remote);
241 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
242 WARN_ON(err == -EMSGSIZE);
247 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
251 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
254 static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
256 struct vxlan_dev *vxlan = netdev_priv(dev);
259 memset(&f, 0, sizeof f);
261 f.remote.remote_ip = ipa; /* goes to NDA_DST */
262 f.remote.remote_vni = VXLAN_N_VID;
264 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
267 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
271 memset(&f, 0, sizeof f);
273 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
275 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
278 /* Hash Ethernet address */
279 static u32 eth_hash(const unsigned char *addr)
281 u64 value = get_unaligned((u64 *)addr);
283 /* only want 6 bytes */
289 return hash_64(value, FDB_HASH_BITS);
292 /* Hash chain to use given mac address */
293 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
296 return &vxlan->fdb_head[eth_hash(mac)];
299 /* Look up Ethernet address in forwarding table */
300 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
304 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
307 hlist_for_each_entry_rcu(f, head, hlist) {
308 if (compare_ether_addr(mac, f->eth_addr) == 0)
315 /* Add/update destinations for multicast */
316 static int vxlan_fdb_append(struct vxlan_fdb *f,
317 __be32 ip, __u32 port, __u32 vni, __u32 ifindex)
319 struct vxlan_rdst *rd_prev, *rd;
322 for (rd = &f->remote; rd; rd = rd->remote_next) {
323 if (rd->remote_ip == ip &&
324 rd->remote_port == port &&
325 rd->remote_vni == vni &&
326 rd->remote_ifindex == ifindex)
330 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
334 rd->remote_port = port;
335 rd->remote_vni = vni;
336 rd->remote_ifindex = ifindex;
337 rd->remote_next = NULL;
338 rd_prev->remote_next = rd;
342 /* Add new entry to forwarding table -- assumes lock held */
343 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
344 const u8 *mac, __be32 ip,
345 __u16 state, __u16 flags,
346 __u32 port, __u32 vni, __u32 ifindex)
351 f = vxlan_find_mac(vxlan, mac);
353 if (flags & NLM_F_EXCL) {
354 netdev_dbg(vxlan->dev,
355 "lost race to create %pM\n", mac);
358 if (f->state != state) {
360 f->updated = jiffies;
363 if ((flags & NLM_F_APPEND) &&
364 is_multicast_ether_addr(f->eth_addr)) {
365 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
372 if (!(flags & NLM_F_CREATE))
375 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
378 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
379 f = kmalloc(sizeof(*f), GFP_ATOMIC);
384 f->remote.remote_ip = ip;
385 f->remote.remote_port = port;
386 f->remote.remote_vni = vni;
387 f->remote.remote_ifindex = ifindex;
388 f->remote.remote_next = NULL;
390 f->updated = f->used = jiffies;
391 memcpy(f->eth_addr, mac, ETH_ALEN);
394 hlist_add_head_rcu(&f->hlist,
395 vxlan_fdb_head(vxlan, mac));
399 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
404 static void vxlan_fdb_free(struct rcu_head *head)
406 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
408 while (f->remote.remote_next) {
409 struct vxlan_rdst *rd = f->remote.remote_next;
411 f->remote.remote_next = rd->remote_next;
417 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
419 netdev_dbg(vxlan->dev,
420 "delete %pM\n", f->eth_addr);
423 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
425 hlist_del_rcu(&f->hlist);
426 call_rcu(&f->rcu, vxlan_fdb_free);
429 /* Add static entry (via netlink) */
430 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
431 struct net_device *dev,
432 const unsigned char *addr, u16 flags)
434 struct vxlan_dev *vxlan = netdev_priv(dev);
435 struct net *net = dev_net(vxlan->dev);
437 u32 port, vni, ifindex;
440 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
441 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
446 if (tb[NDA_DST] == NULL)
449 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
450 return -EAFNOSUPPORT;
452 ip = nla_get_be32(tb[NDA_DST]);
455 if (nla_len(tb[NDA_PORT]) != sizeof(u32))
457 port = nla_get_u32(tb[NDA_PORT]);
462 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
464 vni = nla_get_u32(tb[NDA_VNI]);
466 vni = vxlan->default_dst.remote_vni;
468 if (tb[NDA_IFINDEX]) {
469 struct net_device *tdev;
471 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
473 ifindex = nla_get_u32(tb[NDA_IFINDEX]);
474 tdev = dev_get_by_index(net, ifindex);
476 return -EADDRNOTAVAIL;
481 spin_lock_bh(&vxlan->hash_lock);
482 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags, port,
484 spin_unlock_bh(&vxlan->hash_lock);
489 /* Delete entry (via netlink) */
490 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
491 struct net_device *dev,
492 const unsigned char *addr)
494 struct vxlan_dev *vxlan = netdev_priv(dev);
498 spin_lock_bh(&vxlan->hash_lock);
499 f = vxlan_find_mac(vxlan, addr);
501 vxlan_fdb_destroy(vxlan, f);
504 spin_unlock_bh(&vxlan->hash_lock);
509 /* Dump forwarding table */
510 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
511 struct net_device *dev, int idx)
513 struct vxlan_dev *vxlan = netdev_priv(dev);
516 for (h = 0; h < FDB_HASH_SIZE; ++h) {
520 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
521 struct vxlan_rdst *rd;
522 for (rd = &f->remote; rd; rd = rd->remote_next) {
523 if (idx < cb->args[0])
526 err = vxlan_fdb_info(skb, vxlan, f,
527 NETLINK_CB(cb->skb).portid,
542 /* Watch incoming packets to learn mapping between Ethernet address
543 * and Tunnel endpoint.
545 static void vxlan_snoop(struct net_device *dev,
546 __be32 src_ip, const u8 *src_mac)
548 struct vxlan_dev *vxlan = netdev_priv(dev);
552 f = vxlan_find_mac(vxlan, src_mac);
555 if (likely(f->remote.remote_ip == src_ip))
560 "%pM migrated from %pI4 to %pI4\n",
561 src_mac, &f->remote.remote_ip, &src_ip);
563 f->remote.remote_ip = src_ip;
564 f->updated = jiffies;
566 /* learned new entry */
567 spin_lock(&vxlan->hash_lock);
568 err = vxlan_fdb_create(vxlan, src_mac, src_ip,
570 NLM_F_EXCL|NLM_F_CREATE,
571 vxlan_port, vxlan->default_dst.remote_vni, 0);
572 spin_unlock(&vxlan->hash_lock);
577 /* See if multicast group is already in use by other ID */
578 static bool vxlan_group_used(struct vxlan_net *vn,
579 const struct vxlan_dev *this)
581 const struct vxlan_dev *vxlan;
584 for (h = 0; h < VNI_HASH_SIZE; ++h)
585 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist) {
589 if (!netif_running(vxlan->dev))
592 if (vxlan->default_dst.remote_ip == this->default_dst.remote_ip)
599 /* kernel equivalent to IP_ADD_MEMBERSHIP */
600 static int vxlan_join_group(struct net_device *dev)
602 struct vxlan_dev *vxlan = netdev_priv(dev);
603 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
604 struct sock *sk = vn->sock->sk;
605 struct ip_mreqn mreq = {
606 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
607 .imr_ifindex = vxlan->default_dst.remote_ifindex,
611 /* Already a member of group */
612 if (vxlan_group_used(vn, vxlan))
615 /* Need to drop RTNL to call multicast join */
618 err = ip_mc_join_group(sk, &mreq);
626 /* kernel equivalent to IP_DROP_MEMBERSHIP */
627 static int vxlan_leave_group(struct net_device *dev)
629 struct vxlan_dev *vxlan = netdev_priv(dev);
630 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
632 struct sock *sk = vn->sock->sk;
633 struct ip_mreqn mreq = {
634 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
635 .imr_ifindex = vxlan->default_dst.remote_ifindex,
638 /* Only leave group when last vxlan is done. */
639 if (vxlan_group_used(vn, vxlan))
642 /* Need to drop RTNL to call multicast leave */
645 err = ip_mc_leave_group(sk, &mreq);
652 /* Callback from net/ipv4/udp.c to receive packets */
653 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
656 struct vxlanhdr *vxh;
657 struct vxlan_dev *vxlan;
658 struct pcpu_tstats *stats;
662 /* pop off outer UDP header */
663 __skb_pull(skb, sizeof(struct udphdr));
665 /* Need Vxlan and inner Ethernet header to be present */
666 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
669 /* Drop packets with reserved bits set */
670 vxh = (struct vxlanhdr *) skb->data;
671 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
672 (vxh->vx_vni & htonl(0xff))) {
673 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
674 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
678 __skb_pull(skb, sizeof(struct vxlanhdr));
680 /* Is this VNI defined? */
681 vni = ntohl(vxh->vx_vni) >> 8;
682 vxlan = vxlan_find_vni(sock_net(sk), vni);
684 netdev_dbg(skb->dev, "unknown vni %d\n", vni);
688 if (!pskb_may_pull(skb, ETH_HLEN)) {
689 vxlan->dev->stats.rx_length_errors++;
690 vxlan->dev->stats.rx_errors++;
694 skb_reset_mac_header(skb);
696 /* Re-examine inner Ethernet packet */
698 skb->protocol = eth_type_trans(skb, vxlan->dev);
700 /* Ignore packet loops (and multicast echo) */
701 if (compare_ether_addr(eth_hdr(skb)->h_source,
702 vxlan->dev->dev_addr) == 0)
705 if (vxlan->flags & VXLAN_F_LEARN)
706 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source);
708 __skb_tunnel_rx(skb, vxlan->dev);
709 skb_reset_network_header(skb);
711 /* If the NIC driver gave us an encapsulated packet with
712 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
713 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
714 * for us. Otherwise force the upper layers to verify it.
716 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
717 !(vxlan->dev->features & NETIF_F_RXCSUM))
718 skb->ip_summed = CHECKSUM_NONE;
720 skb->encapsulation = 0;
722 err = IP_ECN_decapsulate(oip, skb);
725 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
726 &oip->saddr, oip->tos);
728 ++vxlan->dev->stats.rx_frame_errors;
729 ++vxlan->dev->stats.rx_errors;
734 stats = this_cpu_ptr(vxlan->dev->tstats);
735 u64_stats_update_begin(&stats->syncp);
737 stats->rx_bytes += skb->len;
738 u64_stats_update_end(&stats->syncp);
744 /* Put UDP header back */
745 __skb_push(skb, sizeof(struct udphdr));
749 /* Consume bad packet */
754 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
756 struct vxlan_dev *vxlan = netdev_priv(dev);
762 if (dev->flags & IFF_NOARP)
765 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
766 dev->stats.tx_dropped++;
771 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
772 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
773 parp->ar_pro != htons(ETH_P_IP) ||
774 parp->ar_op != htons(ARPOP_REQUEST) ||
775 parp->ar_hln != dev->addr_len ||
778 arpptr = (u8 *)parp + sizeof(struct arphdr);
780 arpptr += dev->addr_len; /* sha */
781 memcpy(&sip, arpptr, sizeof(sip));
782 arpptr += sizeof(sip);
783 arpptr += dev->addr_len; /* tha */
784 memcpy(&tip, arpptr, sizeof(tip));
786 if (ipv4_is_loopback(tip) ||
787 ipv4_is_multicast(tip))
790 n = neigh_lookup(&arp_tbl, &tip, dev);
794 struct sk_buff *reply;
796 if (!(n->nud_state & NUD_CONNECTED)) {
801 f = vxlan_find_mac(vxlan, n->ha);
802 if (f && f->remote.remote_ip == htonl(INADDR_ANY)) {
803 /* bridge-local neighbor */
808 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
813 skb_reset_mac_header(reply);
814 __skb_pull(reply, skb_network_offset(reply));
815 reply->ip_summed = CHECKSUM_UNNECESSARY;
816 reply->pkt_type = PACKET_HOST;
818 if (netif_rx_ni(reply) == NET_RX_DROP)
819 dev->stats.rx_dropped++;
820 } else if (vxlan->flags & VXLAN_F_L3MISS)
821 vxlan_ip_miss(dev, tip);
827 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
829 struct vxlan_dev *vxlan = netdev_priv(dev);
833 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
837 switch (ntohs(eth_hdr(skb)->h_proto)) {
839 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
842 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
851 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
853 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
855 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
859 } else if (vxlan->flags & VXLAN_F_L3MISS)
860 vxlan_ip_miss(dev, pip->daddr);
864 static void vxlan_sock_free(struct sk_buff *skb)
869 /* On transmit, associate with the tunnel socket */
870 static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
872 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
873 struct sock *sk = vn->sock->sk;
878 skb->destructor = vxlan_sock_free;
881 /* Compute source port for outgoing packet
882 * first choice to use L4 flow hash since it will spread
883 * better and maybe available from hardware
884 * secondary choice is to use jhash on the Ethernet header
886 static u16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
888 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
891 hash = skb_get_rxhash(skb);
893 hash = jhash(skb->data, 2 * ETH_ALEN,
894 (__force u32) skb->protocol);
896 return (((u64) hash * range) >> 32) + vxlan->port_min;
899 static int handle_offloads(struct sk_buff *skb)
901 if (skb_is_gso(skb)) {
902 int err = skb_unclone(skb, GFP_ATOMIC);
906 skb_shinfo(skb)->gso_type |= (SKB_GSO_UDP_TUNNEL | SKB_GSO_UDP);
907 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
908 skb->ip_summed = CHECKSUM_NONE;
913 /* Bypass encapsulation if the destination is local */
914 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
915 struct vxlan_dev *dst_vxlan)
917 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
918 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
920 skb->pkt_type = PACKET_HOST;
921 skb->encapsulation = 0;
922 skb->dev = dst_vxlan->dev;
923 __skb_pull(skb, skb_network_offset(skb));
925 if (dst_vxlan->flags & VXLAN_F_LEARN)
926 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
927 eth_hdr(skb)->h_source);
929 u64_stats_update_begin(&tx_stats->syncp);
930 tx_stats->tx_packets++;
931 tx_stats->tx_bytes += skb->len;
932 u64_stats_update_end(&tx_stats->syncp);
934 if (netif_rx(skb) == NET_RX_SUCCESS) {
935 u64_stats_update_begin(&rx_stats->syncp);
936 rx_stats->rx_packets++;
937 rx_stats->rx_bytes += skb->len;
938 u64_stats_update_end(&rx_stats->syncp);
940 skb->dev->stats.rx_dropped++;
944 static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
945 struct vxlan_rdst *rdst, bool did_rsc)
947 struct vxlan_dev *vxlan = netdev_priv(dev);
949 const struct iphdr *old_iph;
951 struct vxlanhdr *vxh;
955 __u16 src_port, dst_port;
960 dst_port = rdst->remote_port ? rdst->remote_port : vxlan_port;
961 vni = rdst->remote_vni;
962 dst = rdst->remote_ip;
966 /* short-circuited back to local bridge */
967 vxlan_encap_bypass(skb, vxlan, vxlan);
973 if (!skb->encapsulation) {
974 skb_reset_inner_headers(skb);
975 skb->encapsulation = 1;
978 /* Need space for new headers (invalidates iph ptr) */
979 if (skb_cow_head(skb, VXLAN_HEADROOM))
982 old_iph = ip_hdr(skb);
985 if (!ttl && IN_MULTICAST(ntohl(dst)))
990 tos = ip_tunnel_get_dsfield(old_iph, skb);
992 src_port = vxlan_src_port(vxlan, skb);
994 memset(&fl4, 0, sizeof(fl4));
995 fl4.flowi4_oif = rdst->remote_ifindex;
996 fl4.flowi4_tos = RT_TOS(tos);
998 fl4.saddr = vxlan->saddr;
1000 rt = ip_route_output_key(dev_net(dev), &fl4);
1002 netdev_dbg(dev, "no route to %pI4\n", &dst);
1003 dev->stats.tx_carrier_errors++;
1007 if (rt->dst.dev == dev) {
1008 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1010 dev->stats.collisions++;
1014 /* Bypass encapsulation if the destination is local */
1015 if (rt->rt_flags & RTCF_LOCAL &&
1016 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1017 struct vxlan_dev *dst_vxlan;
1020 dst_vxlan = vxlan_find_vni(dev_net(dev), vni);
1023 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1024 return NETDEV_TX_OK;
1027 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1028 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1031 skb_dst_set(skb, &rt->dst);
1033 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1034 vxh->vx_flags = htonl(VXLAN_FLAGS);
1035 vxh->vx_vni = htonl(vni << 8);
1037 __skb_push(skb, sizeof(*uh));
1038 skb_reset_transport_header(skb);
1041 uh->dest = htons(dst_port);
1042 uh->source = htons(src_port);
1044 uh->len = htons(skb->len);
1047 __skb_push(skb, sizeof(*iph));
1048 skb_reset_network_header(skb);
1051 iph->ihl = sizeof(struct iphdr) >> 2;
1053 iph->protocol = IPPROTO_UDP;
1054 iph->tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1056 iph->saddr = fl4.saddr;
1057 iph->ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1058 tunnel_ip_select_ident(skb, old_iph, &rt->dst);
1062 vxlan_set_owner(dev, skb);
1064 if (handle_offloads(skb))
1067 iptunnel_xmit(skb, dev);
1068 return NETDEV_TX_OK;
1071 dev->stats.tx_dropped++;
1075 dev->stats.tx_errors++;
1078 return NETDEV_TX_OK;
1081 /* Transmit local packets over Vxlan
1083 * Outer IP header inherits ECN and DF from inner header.
1084 * Outer UDP destination is the VXLAN assigned port.
1085 * source port is based on hash of flow
1087 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1089 struct vxlan_dev *vxlan = netdev_priv(dev);
1091 bool did_rsc = false;
1092 struct vxlan_rdst *rdst0, *rdst;
1093 struct vxlan_fdb *f;
1096 skb_reset_mac_header(skb);
1099 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1100 return arp_reduce(dev, skb);
1101 else if ((vxlan->flags&VXLAN_F_RSC) && ntohs(eth->h_proto) == ETH_P_IP)
1102 did_rsc = route_shortcircuit(dev, skb);
1104 f = vxlan_find_mac(vxlan, eth->h_dest);
1107 rdst0 = &vxlan->default_dst;
1109 if (rdst0->remote_ip == htonl(INADDR_ANY) &&
1110 (vxlan->flags & VXLAN_F_L2MISS) &&
1111 !is_multicast_ether_addr(eth->h_dest))
1112 vxlan_fdb_miss(vxlan, eth->h_dest);
1118 /* if there are multiple destinations, send copies */
1119 for (rdst = rdst0->remote_next; rdst; rdst = rdst->remote_next) {
1120 struct sk_buff *skb1;
1122 skb1 = skb_clone(skb, GFP_ATOMIC);
1123 rc1 = vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1124 if (rc == NETDEV_TX_OK)
1128 rc1 = vxlan_xmit_one(skb, dev, rdst0, did_rsc);
1129 if (rc == NETDEV_TX_OK)
1134 /* Walk the forwarding table and purge stale entries */
1135 static void vxlan_cleanup(unsigned long arg)
1137 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1138 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1141 if (!netif_running(vxlan->dev))
1144 spin_lock_bh(&vxlan->hash_lock);
1145 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1146 struct hlist_node *p, *n;
1147 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1149 = container_of(p, struct vxlan_fdb, hlist);
1150 unsigned long timeout;
1152 if (f->state & NUD_PERMANENT)
1155 timeout = f->used + vxlan->age_interval * HZ;
1156 if (time_before_eq(timeout, jiffies)) {
1157 netdev_dbg(vxlan->dev,
1158 "garbage collect %pM\n",
1160 f->state = NUD_STALE;
1161 vxlan_fdb_destroy(vxlan, f);
1162 } else if (time_before(timeout, next_timer))
1163 next_timer = timeout;
1166 spin_unlock_bh(&vxlan->hash_lock);
1168 mod_timer(&vxlan->age_timer, next_timer);
1171 /* Setup stats when device is created */
1172 static int vxlan_init(struct net_device *dev)
1174 dev->tstats = alloc_percpu(struct pcpu_tstats);
1181 /* Start ageing timer and join group when device is brought up */
1182 static int vxlan_open(struct net_device *dev)
1184 struct vxlan_dev *vxlan = netdev_priv(dev);
1187 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
1188 err = vxlan_join_group(dev);
1193 if (vxlan->age_interval)
1194 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1199 /* Purge the forwarding table */
1200 static void vxlan_flush(struct vxlan_dev *vxlan)
1204 spin_lock_bh(&vxlan->hash_lock);
1205 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1206 struct hlist_node *p, *n;
1207 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1209 = container_of(p, struct vxlan_fdb, hlist);
1210 vxlan_fdb_destroy(vxlan, f);
1213 spin_unlock_bh(&vxlan->hash_lock);
1216 /* Cleanup timer and forwarding table on shutdown */
1217 static int vxlan_stop(struct net_device *dev)
1219 struct vxlan_dev *vxlan = netdev_priv(dev);
1221 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)))
1222 vxlan_leave_group(dev);
1224 del_timer_sync(&vxlan->age_timer);
1231 /* Stub, nothing needs to be done. */
1232 static void vxlan_set_multicast_list(struct net_device *dev)
1236 static const struct net_device_ops vxlan_netdev_ops = {
1237 .ndo_init = vxlan_init,
1238 .ndo_open = vxlan_open,
1239 .ndo_stop = vxlan_stop,
1240 .ndo_start_xmit = vxlan_xmit,
1241 .ndo_get_stats64 = ip_tunnel_get_stats64,
1242 .ndo_set_rx_mode = vxlan_set_multicast_list,
1243 .ndo_change_mtu = eth_change_mtu,
1244 .ndo_validate_addr = eth_validate_addr,
1245 .ndo_set_mac_address = eth_mac_addr,
1246 .ndo_fdb_add = vxlan_fdb_add,
1247 .ndo_fdb_del = vxlan_fdb_delete,
1248 .ndo_fdb_dump = vxlan_fdb_dump,
1251 /* Info for udev, that this is a virtual tunnel endpoint */
1252 static struct device_type vxlan_type = {
1256 static void vxlan_free(struct net_device *dev)
1258 free_percpu(dev->tstats);
1262 /* Initialize the device structure. */
1263 static void vxlan_setup(struct net_device *dev)
1265 struct vxlan_dev *vxlan = netdev_priv(dev);
1269 eth_hw_addr_random(dev);
1271 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
1273 dev->netdev_ops = &vxlan_netdev_ops;
1274 dev->destructor = vxlan_free;
1275 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1277 dev->tx_queue_len = 0;
1278 dev->features |= NETIF_F_LLTX;
1279 dev->features |= NETIF_F_NETNS_LOCAL;
1280 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1281 dev->features |= NETIF_F_RXCSUM;
1282 dev->features |= NETIF_F_GSO_SOFTWARE;
1284 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1285 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1286 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1287 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1289 spin_lock_init(&vxlan->hash_lock);
1291 init_timer_deferrable(&vxlan->age_timer);
1292 vxlan->age_timer.function = vxlan_cleanup;
1293 vxlan->age_timer.data = (unsigned long) vxlan;
1295 inet_get_local_port_range(&low, &high);
1296 vxlan->port_min = low;
1297 vxlan->port_max = high;
1301 for (h = 0; h < FDB_HASH_SIZE; ++h)
1302 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1305 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1306 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
1307 [IFLA_VXLAN_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1308 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1309 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1310 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1311 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1312 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1313 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1314 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
1315 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
1316 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1317 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1318 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1319 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
1322 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1324 if (tb[IFLA_ADDRESS]) {
1325 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1326 pr_debug("invalid link address (not ethernet)\n");
1330 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1331 pr_debug("invalid all zero ethernet address\n");
1332 return -EADDRNOTAVAIL;
1339 if (data[IFLA_VXLAN_ID]) {
1340 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1341 if (id >= VXLAN_VID_MASK)
1345 if (data[IFLA_VXLAN_PORT_RANGE]) {
1346 const struct ifla_vxlan_port_range *p
1347 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1349 if (ntohs(p->high) < ntohs(p->low)) {
1350 pr_debug("port range %u .. %u not valid\n",
1351 ntohs(p->low), ntohs(p->high));
1359 static void vxlan_get_drvinfo(struct net_device *netdev,
1360 struct ethtool_drvinfo *drvinfo)
1362 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1363 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1366 static const struct ethtool_ops vxlan_ethtool_ops = {
1367 .get_drvinfo = vxlan_get_drvinfo,
1368 .get_link = ethtool_op_get_link,
1371 static int vxlan_newlink(struct net *net, struct net_device *dev,
1372 struct nlattr *tb[], struct nlattr *data[])
1374 struct vxlan_dev *vxlan = netdev_priv(dev);
1375 struct vxlan_rdst *dst = &vxlan->default_dst;
1379 if (!data[IFLA_VXLAN_ID])
1382 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1383 if (vxlan_find_vni(net, vni)) {
1384 pr_info("duplicate VNI %u\n", vni);
1387 dst->remote_vni = vni;
1389 if (data[IFLA_VXLAN_REMOTE])
1390 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_REMOTE]);
1392 if (data[IFLA_VXLAN_LOCAL])
1393 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1395 if (data[IFLA_VXLAN_LINK] &&
1396 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
1397 struct net_device *lowerdev
1398 = __dev_get_by_index(net, dst->remote_ifindex);
1401 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
1406 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1408 /* update header length based on lower device */
1409 dev->hard_header_len = lowerdev->hard_header_len +
1413 if (data[IFLA_VXLAN_TOS])
1414 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1416 if (data[IFLA_VXLAN_TTL])
1417 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1419 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
1420 vxlan->flags |= VXLAN_F_LEARN;
1422 if (data[IFLA_VXLAN_AGEING])
1423 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1425 vxlan->age_interval = FDB_AGE_DEFAULT;
1427 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1428 vxlan->flags |= VXLAN_F_PROXY;
1430 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1431 vxlan->flags |= VXLAN_F_RSC;
1433 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1434 vxlan->flags |= VXLAN_F_L2MISS;
1436 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1437 vxlan->flags |= VXLAN_F_L3MISS;
1439 if (data[IFLA_VXLAN_LIMIT])
1440 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1442 if (data[IFLA_VXLAN_PORT_RANGE]) {
1443 const struct ifla_vxlan_port_range *p
1444 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1445 vxlan->port_min = ntohs(p->low);
1446 vxlan->port_max = ntohs(p->high);
1449 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1451 err = register_netdevice(dev);
1453 hlist_add_head_rcu(&vxlan->hlist, vni_head(net, dst->remote_vni));
1458 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1460 struct vxlan_dev *vxlan = netdev_priv(dev);
1462 hlist_del_rcu(&vxlan->hlist);
1464 unregister_netdevice_queue(dev, head);
1467 static size_t vxlan_get_size(const struct net_device *dev)
1470 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
1471 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_REMOTE */
1472 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1473 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1474 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1475 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1476 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
1477 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1478 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1479 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1480 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
1481 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1482 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
1483 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
1487 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1489 const struct vxlan_dev *vxlan = netdev_priv(dev);
1490 const struct vxlan_rdst *dst = &vxlan->default_dst;
1491 struct ifla_vxlan_port_range ports = {
1492 .low = htons(vxlan->port_min),
1493 .high = htons(vxlan->port_max),
1496 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
1497 goto nla_put_failure;
1499 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_REMOTE, dst->remote_ip))
1500 goto nla_put_failure;
1502 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
1503 goto nla_put_failure;
1505 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
1506 goto nla_put_failure;
1508 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1509 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
1510 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1511 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1512 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1513 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1514 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1515 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1516 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1517 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1518 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
1519 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
1520 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax))
1521 goto nla_put_failure;
1523 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1524 goto nla_put_failure;
1532 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1534 .maxtype = IFLA_VXLAN_MAX,
1535 .policy = vxlan_policy,
1536 .priv_size = sizeof(struct vxlan_dev),
1537 .setup = vxlan_setup,
1538 .validate = vxlan_validate,
1539 .newlink = vxlan_newlink,
1540 .dellink = vxlan_dellink,
1541 .get_size = vxlan_get_size,
1542 .fill_info = vxlan_fill_info,
1545 static __net_init int vxlan_init_net(struct net *net)
1547 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1549 struct sockaddr_in vxlan_addr = {
1550 .sin_family = AF_INET,
1551 .sin_addr.s_addr = htonl(INADDR_ANY),
1556 /* Create UDP socket for encapsulation receive. */
1557 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vn->sock);
1559 pr_debug("UDP socket create failed\n");
1562 /* Put in proper namespace */
1564 sk_change_net(sk, net);
1566 vxlan_addr.sin_port = htons(vxlan_port);
1568 rc = kernel_bind(vn->sock, (struct sockaddr *) &vxlan_addr,
1569 sizeof(vxlan_addr));
1571 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1572 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
1573 sk_release_kernel(sk);
1578 /* Disable multicast loopback */
1579 inet_sk(sk)->mc_loop = 0;
1581 /* Mark socket as an encapsulation socket. */
1582 udp_sk(sk)->encap_type = 1;
1583 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1586 for (h = 0; h < VNI_HASH_SIZE; ++h)
1587 INIT_HLIST_HEAD(&vn->vni_list[h]);
1592 static __net_exit void vxlan_exit_net(struct net *net)
1594 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1595 struct vxlan_dev *vxlan;
1599 for (h = 0; h < VNI_HASH_SIZE; ++h)
1600 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist)
1601 dev_close(vxlan->dev);
1605 sk_release_kernel(vn->sock->sk);
1610 static struct pernet_operations vxlan_net_ops = {
1611 .init = vxlan_init_net,
1612 .exit = vxlan_exit_net,
1613 .id = &vxlan_net_id,
1614 .size = sizeof(struct vxlan_net),
1617 static int __init vxlan_init_module(void)
1621 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1623 rc = register_pernet_device(&vxlan_net_ops);
1627 rc = rtnl_link_register(&vxlan_link_ops);
1634 unregister_pernet_device(&vxlan_net_ops);
1638 module_init(vxlan_init_module);
1640 static void __exit vxlan_cleanup_module(void)
1642 rtnl_link_unregister(&vxlan_link_ops);
1643 unregister_pernet_device(&vxlan_net_ops);
1646 module_exit(vxlan_cleanup_module);
1648 MODULE_LICENSE("GPL");
1649 MODULE_VERSION(VXLAN_VERSION);
1650 MODULE_AUTHOR("Stephen Hemminger <shemminger@vyatta.com>");
1651 MODULE_ALIAS_RTNL_LINK("vxlan");