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>
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 VNI_HASH_BITS 10
48 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
49 #define FDB_HASH_BITS 8
50 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
51 #define FDB_AGE_DEFAULT 300 /* 5 min */
52 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
54 #define VXLAN_N_VID (1u << 24)
55 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
56 /* IP header + UDP + VXLAN + Ethernet header */
57 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
59 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
61 /* VXLAN protocol header */
67 /* UDP port for VXLAN traffic. */
68 static unsigned int vxlan_port __read_mostly
= 8472;
69 module_param_named(udp_port
, vxlan_port
, uint
, 0444);
70 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
72 static bool log_ecn_error
= true;
73 module_param(log_ecn_error
, bool, 0644);
74 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
76 /* per-net private data for this module */
77 static unsigned int vxlan_net_id
;
79 struct socket
*sock
; /* UDP encap socket */
80 struct hlist_head vni_list
[VNI_HASH_SIZE
];
83 /* Forwarding table entry */
85 struct hlist_node hlist
; /* linked list of entries */
87 unsigned long updated
; /* jiffies */
90 u16 state
; /* see ndm_state */
91 u8 eth_addr
[ETH_ALEN
];
94 /* Per-cpu network traffic stats */
100 struct u64_stats_sync syncp
;
103 /* Pseudo network device */
105 struct hlist_node hlist
;
106 struct net_device
*dev
;
107 struct vxlan_stats __percpu
*stats
;
108 __u32 vni
; /* virtual network id */
109 __be32 gaddr
; /* multicast group */
110 __be32 saddr
; /* source address */
111 unsigned int link
; /* link to multicast over */
112 __u16 port_min
; /* source port range */
114 __u8 tos
; /* TOS override */
116 u32 flags
; /* VXLAN_F_* below */
118 unsigned long age_interval
;
119 struct timer_list age_timer
;
120 spinlock_t hash_lock
;
121 unsigned int addrcnt
;
122 unsigned int addrmax
;
124 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
127 #define VXLAN_F_LEARN 0x01
128 #define VXLAN_F_PROXY 0x02
129 #define VXLAN_F_RSC 0x04
130 #define VXLAN_F_L2MISS 0x08
131 #define VXLAN_F_L3MISS 0x10
133 /* salt for hash table */
134 static u32 vxlan_salt __read_mostly
;
136 static inline struct hlist_head
*vni_head(struct net
*net
, u32 id
)
138 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
140 return &vn
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
143 /* Look up VNI in a per net namespace table */
144 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
)
146 struct vxlan_dev
*vxlan
;
147 struct hlist_node
*node
;
149 hlist_for_each_entry_rcu(vxlan
, node
, vni_head(net
, id
), hlist
) {
150 if (vxlan
->vni
== id
)
157 /* Fill in neighbour message in skbuff. */
158 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
159 const struct vxlan_fdb
*fdb
,
160 u32 portid
, u32 seq
, int type
, unsigned int flags
)
162 unsigned long now
= jiffies
;
163 struct nda_cacheinfo ci
;
164 struct nlmsghdr
*nlh
;
166 bool send_ip
, send_eth
;
168 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
172 ndm
= nlmsg_data(nlh
);
173 memset(ndm
, 0, sizeof(*ndm
));
175 send_eth
= send_ip
= true;
177 if (type
== RTM_GETNEIGH
) {
178 ndm
->ndm_family
= AF_INET
;
179 send_ip
= fdb
->remote_ip
!= 0;
180 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
182 ndm
->ndm_family
= AF_BRIDGE
;
183 ndm
->ndm_state
= fdb
->state
;
184 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
185 ndm
->ndm_flags
= NTF_SELF
;
186 ndm
->ndm_type
= NDA_DST
;
188 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
189 goto nla_put_failure
;
191 if (send_ip
&& nla_put_be32(skb
, NDA_DST
, fdb
->remote_ip
))
192 goto nla_put_failure
;
194 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
195 ci
.ndm_confirmed
= 0;
196 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
199 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
200 goto nla_put_failure
;
202 return nlmsg_end(skb
, nlh
);
205 nlmsg_cancel(skb
, nlh
);
209 static inline size_t vxlan_nlmsg_size(void)
211 return NLMSG_ALIGN(sizeof(struct ndmsg
))
212 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
213 + nla_total_size(sizeof(__be32
)) /* NDA_DST */
214 + nla_total_size(sizeof(struct nda_cacheinfo
));
217 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
,
218 const struct vxlan_fdb
*fdb
, int type
)
220 struct net
*net
= dev_net(vxlan
->dev
);
224 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
228 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0);
230 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
231 WARN_ON(err
== -EMSGSIZE
);
236 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
240 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
243 static void vxlan_ip_miss(struct net_device
*dev
, __be32 ipa
)
245 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
248 memset(&f
, 0, sizeof f
);
250 f
.remote_ip
= ipa
; /* goes to NDA_DST */
252 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
255 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
259 memset(&f
, 0, sizeof f
);
261 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
263 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
266 /* Hash Ethernet address */
267 static u32
eth_hash(const unsigned char *addr
)
269 u64 value
= get_unaligned((u64
*)addr
);
271 /* only want 6 bytes */
277 return hash_64(value
, FDB_HASH_BITS
);
280 /* Hash chain to use given mac address */
281 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
284 return &vxlan
->fdb_head
[eth_hash(mac
)];
287 /* Look up Ethernet address in forwarding table */
288 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
292 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
294 struct hlist_node
*node
;
296 hlist_for_each_entry_rcu(f
, node
, head
, hlist
) {
297 if (compare_ether_addr(mac
, f
->eth_addr
) == 0)
304 /* Add new entry to forwarding table -- assumes lock held */
305 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
306 const u8
*mac
, __be32 ip
,
307 __u16 state
, __u16 flags
)
312 f
= vxlan_find_mac(vxlan
, mac
);
314 if (flags
& NLM_F_EXCL
) {
315 netdev_dbg(vxlan
->dev
,
316 "lost race to create %pM\n", mac
);
319 if (f
->state
!= state
) {
321 f
->updated
= jiffies
;
325 if (!(flags
& NLM_F_CREATE
))
328 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
331 netdev_dbg(vxlan
->dev
, "add %pM -> %pI4\n", mac
, &ip
);
332 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
339 f
->updated
= f
->used
= jiffies
;
340 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
343 hlist_add_head_rcu(&f
->hlist
,
344 vxlan_fdb_head(vxlan
, mac
));
348 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
353 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
355 netdev_dbg(vxlan
->dev
,
356 "delete %pM\n", f
->eth_addr
);
359 vxlan_fdb_notify(vxlan
, f
, RTM_DELNEIGH
);
361 hlist_del_rcu(&f
->hlist
);
365 /* Add static entry (via netlink) */
366 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
367 struct net_device
*dev
,
368 const unsigned char *addr
, u16 flags
)
370 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
374 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
375 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
380 if (tb
[NDA_DST
] == NULL
)
383 if (nla_len(tb
[NDA_DST
]) != sizeof(__be32
))
384 return -EAFNOSUPPORT
;
386 ip
= nla_get_be32(tb
[NDA_DST
]);
388 spin_lock_bh(&vxlan
->hash_lock
);
389 err
= vxlan_fdb_create(vxlan
, addr
, ip
, ndm
->ndm_state
, flags
);
390 spin_unlock_bh(&vxlan
->hash_lock
);
395 /* Delete entry (via netlink) */
396 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
397 struct net_device
*dev
,
398 const unsigned char *addr
)
400 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
404 spin_lock_bh(&vxlan
->hash_lock
);
405 f
= vxlan_find_mac(vxlan
, addr
);
407 vxlan_fdb_destroy(vxlan
, f
);
410 spin_unlock_bh(&vxlan
->hash_lock
);
415 /* Dump forwarding table */
416 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
417 struct net_device
*dev
, int idx
)
419 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
422 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
424 struct hlist_node
*n
;
427 hlist_for_each_entry_rcu(f
, n
, &vxlan
->fdb_head
[h
], hlist
) {
428 if (idx
< cb
->args
[0])
431 err
= vxlan_fdb_info(skb
, vxlan
, f
,
432 NETLINK_CB(cb
->skb
).portid
,
446 /* Watch incoming packets to learn mapping between Ethernet address
447 * and Tunnel endpoint.
449 static void vxlan_snoop(struct net_device
*dev
,
450 __be32 src_ip
, const u8
*src_mac
)
452 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
456 f
= vxlan_find_mac(vxlan
, src_mac
);
459 if (likely(f
->remote_ip
== src_ip
))
464 "%pM migrated from %pI4 to %pI4\n",
465 src_mac
, &f
->remote_ip
, &src_ip
);
467 f
->remote_ip
= src_ip
;
468 f
->updated
= jiffies
;
470 /* learned new entry */
471 spin_lock(&vxlan
->hash_lock
);
472 err
= vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
474 NLM_F_EXCL
|NLM_F_CREATE
);
475 spin_unlock(&vxlan
->hash_lock
);
480 /* See if multicast group is already in use by other ID */
481 static bool vxlan_group_used(struct vxlan_net
*vn
,
482 const struct vxlan_dev
*this)
484 const struct vxlan_dev
*vxlan
;
485 struct hlist_node
*node
;
488 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
489 hlist_for_each_entry(vxlan
, node
, &vn
->vni_list
[h
], hlist
) {
493 if (!netif_running(vxlan
->dev
))
496 if (vxlan
->gaddr
== this->gaddr
)
503 /* kernel equivalent to IP_ADD_MEMBERSHIP */
504 static int vxlan_join_group(struct net_device
*dev
)
506 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
507 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
508 struct sock
*sk
= vn
->sock
->sk
;
509 struct ip_mreqn mreq
= {
510 .imr_multiaddr
.s_addr
= vxlan
->gaddr
,
511 .imr_ifindex
= vxlan
->link
,
515 /* Already a member of group */
516 if (vxlan_group_used(vn
, vxlan
))
519 /* Need to drop RTNL to call multicast join */
522 err
= ip_mc_join_group(sk
, &mreq
);
530 /* kernel equivalent to IP_DROP_MEMBERSHIP */
531 static int vxlan_leave_group(struct net_device
*dev
)
533 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
534 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
536 struct sock
*sk
= vn
->sock
->sk
;
537 struct ip_mreqn mreq
= {
538 .imr_multiaddr
.s_addr
= vxlan
->gaddr
,
539 .imr_ifindex
= vxlan
->link
,
542 /* Only leave group when last vxlan is done. */
543 if (vxlan_group_used(vn
, vxlan
))
546 /* Need to drop RTNL to call multicast leave */
549 err
= ip_mc_leave_group(sk
, &mreq
);
556 /* Callback from net/ipv4/udp.c to receive packets */
557 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
560 struct vxlanhdr
*vxh
;
561 struct vxlan_dev
*vxlan
;
562 struct vxlan_stats
*stats
;
566 /* pop off outer UDP header */
567 __skb_pull(skb
, sizeof(struct udphdr
));
569 /* Need Vxlan and inner Ethernet header to be present */
570 if (!pskb_may_pull(skb
, sizeof(struct vxlanhdr
)))
573 /* Drop packets with reserved bits set */
574 vxh
= (struct vxlanhdr
*) skb
->data
;
575 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
576 (vxh
->vx_vni
& htonl(0xff))) {
577 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
578 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
582 __skb_pull(skb
, sizeof(struct vxlanhdr
));
584 /* Is this VNI defined? */
585 vni
= ntohl(vxh
->vx_vni
) >> 8;
586 vxlan
= vxlan_find_vni(sock_net(sk
), vni
);
588 netdev_dbg(skb
->dev
, "unknown vni %d\n", vni
);
592 if (!pskb_may_pull(skb
, ETH_HLEN
)) {
593 vxlan
->dev
->stats
.rx_length_errors
++;
594 vxlan
->dev
->stats
.rx_errors
++;
598 skb_reset_mac_header(skb
);
600 /* Re-examine inner Ethernet packet */
602 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
604 /* Ignore packet loops (and multicast echo) */
605 if (compare_ether_addr(eth_hdr(skb
)->h_source
,
606 vxlan
->dev
->dev_addr
) == 0)
609 if (vxlan
->flags
& VXLAN_F_LEARN
)
610 vxlan_snoop(skb
->dev
, oip
->saddr
, eth_hdr(skb
)->h_source
);
612 __skb_tunnel_rx(skb
, vxlan
->dev
);
613 skb_reset_network_header(skb
);
615 /* If the NIC driver gave us an encapsulated packet with
616 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
617 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
618 * for us. Otherwise force the upper layers to verify it.
620 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
|| !skb
->encapsulation
||
621 !(vxlan
->dev
->features
& NETIF_F_RXCSUM
))
622 skb
->ip_summed
= CHECKSUM_NONE
;
624 skb
->encapsulation
= 0;
626 err
= IP_ECN_decapsulate(oip
, skb
);
629 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
630 &oip
->saddr
, oip
->tos
);
632 ++vxlan
->dev
->stats
.rx_frame_errors
;
633 ++vxlan
->dev
->stats
.rx_errors
;
638 stats
= this_cpu_ptr(vxlan
->stats
);
639 u64_stats_update_begin(&stats
->syncp
);
641 stats
->rx_bytes
+= skb
->len
;
642 u64_stats_update_end(&stats
->syncp
);
648 /* Put UDP header back */
649 __skb_push(skb
, sizeof(struct udphdr
));
653 /* Consume bad packet */
658 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
660 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
666 if (dev
->flags
& IFF_NOARP
)
669 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
670 dev
->stats
.tx_dropped
++;
675 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
676 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
677 parp
->ar_pro
!= htons(ETH_P_IP
) ||
678 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
679 parp
->ar_hln
!= dev
->addr_len
||
682 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
684 arpptr
+= dev
->addr_len
; /* sha */
685 memcpy(&sip
, arpptr
, sizeof(sip
));
686 arpptr
+= sizeof(sip
);
687 arpptr
+= dev
->addr_len
; /* tha */
688 memcpy(&tip
, arpptr
, sizeof(tip
));
690 if (ipv4_is_loopback(tip
) ||
691 ipv4_is_multicast(tip
))
694 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
697 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
699 struct sk_buff
*reply
;
701 if (!(n
->nud_state
& NUD_CONNECTED
)) {
706 f
= vxlan_find_mac(vxlan
, n
->ha
);
707 if (f
&& f
->remote_ip
== 0) {
708 /* bridge-local neighbor */
713 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
718 skb_reset_mac_header(reply
);
719 __skb_pull(reply
, skb_network_offset(reply
));
720 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
721 reply
->pkt_type
= PACKET_HOST
;
723 if (netif_rx_ni(reply
) == NET_RX_DROP
)
724 dev
->stats
.rx_dropped
++;
725 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
726 vxlan_ip_miss(dev
, tip
);
732 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
734 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
738 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
742 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
744 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
747 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
756 diff
= compare_ether_addr(eth_hdr(skb
)->h_dest
, n
->ha
) != 0;
758 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
760 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
764 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
765 vxlan_ip_miss(dev
, pip
->daddr
);
769 /* Extract dsfield from inner protocol */
770 static inline u8
vxlan_get_dsfield(const struct iphdr
*iph
,
771 const struct sk_buff
*skb
)
773 if (skb
->protocol
== htons(ETH_P_IP
))
775 else if (skb
->protocol
== htons(ETH_P_IPV6
))
776 return ipv6_get_dsfield((const struct ipv6hdr
*)iph
);
781 /* Propogate ECN bits out */
782 static inline u8
vxlan_ecn_encap(u8 tos
,
783 const struct iphdr
*iph
,
784 const struct sk_buff
*skb
)
786 u8 inner
= vxlan_get_dsfield(iph
, skb
);
788 return INET_ECN_encapsulate(tos
, inner
);
791 static void vxlan_sock_free(struct sk_buff
*skb
)
796 /* On transmit, associate with the tunnel socket */
797 static void vxlan_set_owner(struct net_device
*dev
, struct sk_buff
*skb
)
799 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
800 struct sock
*sk
= vn
->sock
->sk
;
805 skb
->destructor
= vxlan_sock_free
;
808 /* Compute source port for outgoing packet
809 * first choice to use L4 flow hash since it will spread
810 * better and maybe available from hardware
811 * secondary choice is to use jhash on the Ethernet header
813 static u16
vxlan_src_port(const struct vxlan_dev
*vxlan
, struct sk_buff
*skb
)
815 unsigned int range
= (vxlan
->port_max
- vxlan
->port_min
) + 1;
818 hash
= skb_get_rxhash(skb
);
820 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
821 (__force u32
) skb
->protocol
);
823 return (((u64
) hash
* range
) >> 32) + vxlan
->port_min
;
826 /* Transmit local packets over Vxlan
828 * Outer IP header inherits ECN and DF from inner header.
829 * Outer UDP destination is the VXLAN assigned port.
830 * source port is based on hash of flow
832 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
834 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
836 const struct iphdr
*old_iph
;
839 struct vxlanhdr
*vxh
;
842 unsigned int pkt_len
= skb
->len
;
848 bool did_rsc
= false;
849 const struct vxlan_fdb
*f
;
851 skb_reset_mac_header(skb
);
854 if ((vxlan
->flags
& VXLAN_F_PROXY
) && ntohs(eth
->h_proto
) == ETH_P_ARP
)
855 return arp_reduce(dev
, skb
);
856 else if ((vxlan
->flags
&VXLAN_F_RSC
) && ntohs(eth
->h_proto
) == ETH_P_IP
)
857 did_rsc
= route_shortcircuit(dev
, skb
);
859 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
863 if (!dst
&& (vxlan
->flags
& VXLAN_F_L2MISS
) &&
864 !is_multicast_ether_addr(eth
->h_dest
))
865 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
871 __skb_pull(skb
, skb_network_offset(skb
));
872 skb
->ip_summed
= CHECKSUM_NONE
;
873 skb
->pkt_type
= PACKET_HOST
;
875 /* short-circuited back to local bridge */
876 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
877 struct vxlan_stats
*stats
=
878 this_cpu_ptr(vxlan
->stats
);
880 u64_stats_update_begin(&stats
->syncp
);
882 stats
->tx_bytes
+= pkt_len
;
883 u64_stats_update_end(&stats
->syncp
);
885 dev
->stats
.tx_errors
++;
886 dev
->stats
.tx_aborted_errors
++;
893 if (!skb
->encapsulation
) {
894 skb_reset_inner_headers(skb
);
895 skb
->encapsulation
= 1;
898 /* Need space for new headers (invalidates iph ptr) */
899 if (skb_cow_head(skb
, VXLAN_HEADROOM
))
902 old_iph
= ip_hdr(skb
);
905 if (!ttl
&& IN_MULTICAST(ntohl(dst
)))
910 tos
= vxlan_get_dsfield(old_iph
, skb
);
912 src_port
= vxlan_src_port(vxlan
, skb
);
914 memset(&fl4
, 0, sizeof(fl4
));
915 fl4
.flowi4_oif
= vxlan
->link
;
916 fl4
.flowi4_tos
= RT_TOS(tos
);
918 fl4
.saddr
= vxlan
->saddr
;
920 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
922 netdev_dbg(dev
, "no route to %pI4\n", &dst
);
923 dev
->stats
.tx_carrier_errors
++;
927 if (rt
->dst
.dev
== dev
) {
928 netdev_dbg(dev
, "circular route to %pI4\n", &dst
);
930 dev
->stats
.collisions
++;
934 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
935 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
938 skb_dst_set(skb
, &rt
->dst
);
940 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
941 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
942 vxh
->vx_vni
= htonl(vxlan
->vni
<< 8);
944 __skb_push(skb
, sizeof(*uh
));
945 skb_reset_transport_header(skb
);
948 uh
->dest
= htons(vxlan_port
);
949 uh
->source
= htons(src_port
);
951 uh
->len
= htons(skb
->len
);
954 __skb_push(skb
, sizeof(*iph
));
955 skb_reset_network_header(skb
);
958 iph
->ihl
= sizeof(struct iphdr
) >> 2;
960 iph
->protocol
= IPPROTO_UDP
;
961 iph
->tos
= vxlan_ecn_encap(tos
, old_iph
, skb
);
963 iph
->saddr
= fl4
.saddr
;
964 iph
->ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
966 vxlan_set_owner(dev
, skb
);
968 /* See iptunnel_xmit() */
969 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
970 skb
->ip_summed
= CHECKSUM_NONE
;
971 ip_select_ident(iph
, &rt
->dst
, NULL
);
973 err
= ip_local_out(skb
);
974 if (likely(net_xmit_eval(err
) == 0)) {
975 struct vxlan_stats
*stats
= this_cpu_ptr(vxlan
->stats
);
977 u64_stats_update_begin(&stats
->syncp
);
979 stats
->tx_bytes
+= pkt_len
;
980 u64_stats_update_end(&stats
->syncp
);
982 dev
->stats
.tx_errors
++;
983 dev
->stats
.tx_aborted_errors
++;
988 dev
->stats
.tx_dropped
++;
992 dev
->stats
.tx_errors
++;
998 /* Walk the forwarding table and purge stale entries */
999 static void vxlan_cleanup(unsigned long arg
)
1001 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1002 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1005 if (!netif_running(vxlan
->dev
))
1008 spin_lock_bh(&vxlan
->hash_lock
);
1009 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1010 struct hlist_node
*p
, *n
;
1011 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1013 = container_of(p
, struct vxlan_fdb
, hlist
);
1014 unsigned long timeout
;
1016 if (f
->state
& NUD_PERMANENT
)
1019 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
1020 if (time_before_eq(timeout
, jiffies
)) {
1021 netdev_dbg(vxlan
->dev
,
1022 "garbage collect %pM\n",
1024 f
->state
= NUD_STALE
;
1025 vxlan_fdb_destroy(vxlan
, f
);
1026 } else if (time_before(timeout
, next_timer
))
1027 next_timer
= timeout
;
1030 spin_unlock_bh(&vxlan
->hash_lock
);
1032 mod_timer(&vxlan
->age_timer
, next_timer
);
1035 /* Setup stats when device is created */
1036 static int vxlan_init(struct net_device
*dev
)
1038 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1040 vxlan
->stats
= alloc_percpu(struct vxlan_stats
);
1047 /* Start ageing timer and join group when device is brought up */
1048 static int vxlan_open(struct net_device
*dev
)
1050 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1054 err
= vxlan_join_group(dev
);
1059 if (vxlan
->age_interval
)
1060 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
1065 /* Purge the forwarding table */
1066 static void vxlan_flush(struct vxlan_dev
*vxlan
)
1070 spin_lock_bh(&vxlan
->hash_lock
);
1071 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1072 struct hlist_node
*p
, *n
;
1073 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1075 = container_of(p
, struct vxlan_fdb
, hlist
);
1076 vxlan_fdb_destroy(vxlan
, f
);
1079 spin_unlock_bh(&vxlan
->hash_lock
);
1082 /* Cleanup timer and forwarding table on shutdown */
1083 static int vxlan_stop(struct net_device
*dev
)
1085 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1088 vxlan_leave_group(dev
);
1090 del_timer_sync(&vxlan
->age_timer
);
1097 /* Merge per-cpu statistics */
1098 static struct rtnl_link_stats64
*vxlan_stats64(struct net_device
*dev
,
1099 struct rtnl_link_stats64
*stats
)
1101 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1102 struct vxlan_stats tmp
, sum
= { 0 };
1105 for_each_possible_cpu(cpu
) {
1107 const struct vxlan_stats
*stats
1108 = per_cpu_ptr(vxlan
->stats
, cpu
);
1111 start
= u64_stats_fetch_begin_bh(&stats
->syncp
);
1112 memcpy(&tmp
, stats
, sizeof(tmp
));
1113 } while (u64_stats_fetch_retry_bh(&stats
->syncp
, start
));
1115 sum
.tx_bytes
+= tmp
.tx_bytes
;
1116 sum
.tx_packets
+= tmp
.tx_packets
;
1117 sum
.rx_bytes
+= tmp
.rx_bytes
;
1118 sum
.rx_packets
+= tmp
.rx_packets
;
1121 stats
->tx_bytes
= sum
.tx_bytes
;
1122 stats
->tx_packets
= sum
.tx_packets
;
1123 stats
->rx_bytes
= sum
.rx_bytes
;
1124 stats
->rx_packets
= sum
.rx_packets
;
1126 stats
->multicast
= dev
->stats
.multicast
;
1127 stats
->rx_length_errors
= dev
->stats
.rx_length_errors
;
1128 stats
->rx_frame_errors
= dev
->stats
.rx_frame_errors
;
1129 stats
->rx_errors
= dev
->stats
.rx_errors
;
1131 stats
->tx_dropped
= dev
->stats
.tx_dropped
;
1132 stats
->tx_carrier_errors
= dev
->stats
.tx_carrier_errors
;
1133 stats
->tx_aborted_errors
= dev
->stats
.tx_aborted_errors
;
1134 stats
->collisions
= dev
->stats
.collisions
;
1135 stats
->tx_errors
= dev
->stats
.tx_errors
;
1140 /* Stub, nothing needs to be done. */
1141 static void vxlan_set_multicast_list(struct net_device
*dev
)
1145 static const struct net_device_ops vxlan_netdev_ops
= {
1146 .ndo_init
= vxlan_init
,
1147 .ndo_open
= vxlan_open
,
1148 .ndo_stop
= vxlan_stop
,
1149 .ndo_start_xmit
= vxlan_xmit
,
1150 .ndo_get_stats64
= vxlan_stats64
,
1151 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
1152 .ndo_change_mtu
= eth_change_mtu
,
1153 .ndo_validate_addr
= eth_validate_addr
,
1154 .ndo_set_mac_address
= eth_mac_addr
,
1155 .ndo_fdb_add
= vxlan_fdb_add
,
1156 .ndo_fdb_del
= vxlan_fdb_delete
,
1157 .ndo_fdb_dump
= vxlan_fdb_dump
,
1160 /* Info for udev, that this is a virtual tunnel endpoint */
1161 static struct device_type vxlan_type
= {
1165 static void vxlan_free(struct net_device
*dev
)
1167 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1169 free_percpu(vxlan
->stats
);
1173 /* Initialize the device structure. */
1174 static void vxlan_setup(struct net_device
*dev
)
1176 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1180 eth_hw_addr_random(dev
);
1182 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
1184 dev
->netdev_ops
= &vxlan_netdev_ops
;
1185 dev
->destructor
= vxlan_free
;
1186 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
1188 dev
->tx_queue_len
= 0;
1189 dev
->features
|= NETIF_F_LLTX
;
1190 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1191 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1192 dev
->features
|= NETIF_F_RXCSUM
;
1194 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
1195 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1196 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1198 spin_lock_init(&vxlan
->hash_lock
);
1200 init_timer_deferrable(&vxlan
->age_timer
);
1201 vxlan
->age_timer
.function
= vxlan_cleanup
;
1202 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
1204 inet_get_local_port_range(&low
, &high
);
1205 vxlan
->port_min
= low
;
1206 vxlan
->port_max
= high
;
1210 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
1211 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
1214 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
1215 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
1216 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
1217 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
1218 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
1219 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
1220 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
1221 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
1222 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
1223 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
1224 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
1225 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
1226 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
1227 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
1228 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
1231 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
1233 if (tb
[IFLA_ADDRESS
]) {
1234 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
1235 pr_debug("invalid link address (not ethernet)\n");
1239 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
1240 pr_debug("invalid all zero ethernet address\n");
1241 return -EADDRNOTAVAIL
;
1248 if (data
[IFLA_VXLAN_ID
]) {
1249 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1250 if (id
>= VXLAN_VID_MASK
)
1254 if (data
[IFLA_VXLAN_GROUP
]) {
1255 __be32 gaddr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1256 if (!IN_MULTICAST(ntohl(gaddr
))) {
1257 pr_debug("group address is not IPv4 multicast\n");
1258 return -EADDRNOTAVAIL
;
1262 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1263 const struct ifla_vxlan_port_range
*p
1264 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1266 if (ntohs(p
->high
) < ntohs(p
->low
)) {
1267 pr_debug("port range %u .. %u not valid\n",
1268 ntohs(p
->low
), ntohs(p
->high
));
1276 static void vxlan_get_drvinfo(struct net_device
*netdev
,
1277 struct ethtool_drvinfo
*drvinfo
)
1279 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
1280 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
1283 static const struct ethtool_ops vxlan_ethtool_ops
= {
1284 .get_drvinfo
= vxlan_get_drvinfo
,
1285 .get_link
= ethtool_op_get_link
,
1288 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
1289 struct nlattr
*tb
[], struct nlattr
*data
[])
1291 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1295 if (!data
[IFLA_VXLAN_ID
])
1298 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1299 if (vxlan_find_vni(net
, vni
)) {
1300 pr_info("duplicate VNI %u\n", vni
);
1305 if (data
[IFLA_VXLAN_GROUP
])
1306 vxlan
->gaddr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1308 if (data
[IFLA_VXLAN_LOCAL
])
1309 vxlan
->saddr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
1311 if (data
[IFLA_VXLAN_LINK
] &&
1312 (vxlan
->link
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
1313 struct net_device
*lowerdev
1314 = __dev_get_by_index(net
, vxlan
->link
);
1317 pr_info("ifindex %d does not exist\n", vxlan
->link
);
1322 dev
->mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
1324 /* update header length based on lower device */
1325 dev
->hard_header_len
= lowerdev
->hard_header_len
+
1329 if (data
[IFLA_VXLAN_TOS
])
1330 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
1332 if (data
[IFLA_VXLAN_TTL
])
1333 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
1335 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
1336 vxlan
->flags
|= VXLAN_F_LEARN
;
1338 if (data
[IFLA_VXLAN_AGEING
])
1339 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
1341 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
1343 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
1344 vxlan
->flags
|= VXLAN_F_PROXY
;
1346 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
1347 vxlan
->flags
|= VXLAN_F_RSC
;
1349 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
1350 vxlan
->flags
|= VXLAN_F_L2MISS
;
1352 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
1353 vxlan
->flags
|= VXLAN_F_L3MISS
;
1355 if (data
[IFLA_VXLAN_LIMIT
])
1356 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
1358 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1359 const struct ifla_vxlan_port_range
*p
1360 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1361 vxlan
->port_min
= ntohs(p
->low
);
1362 vxlan
->port_max
= ntohs(p
->high
);
1365 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
1367 err
= register_netdevice(dev
);
1369 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(net
, vxlan
->vni
));
1374 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
1376 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1378 hlist_del_rcu(&vxlan
->hlist
);
1380 unregister_netdevice_queue(dev
, head
);
1383 static size_t vxlan_get_size(const struct net_device
*dev
)
1386 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
1387 nla_total_size(sizeof(__be32
)) +/* IFLA_VXLAN_GROUP */
1388 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
1389 nla_total_size(sizeof(__be32
))+ /* IFLA_VXLAN_LOCAL */
1390 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
1391 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
1392 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
1393 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
1394 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
1395 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
1396 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
1397 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
1398 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
1399 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
1403 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1405 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1406 struct ifla_vxlan_port_range ports
= {
1407 .low
= htons(vxlan
->port_min
),
1408 .high
= htons(vxlan
->port_max
),
1411 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, vxlan
->vni
))
1412 goto nla_put_failure
;
1414 if (vxlan
->gaddr
&& nla_put_be32(skb
, IFLA_VXLAN_GROUP
, vxlan
->gaddr
))
1415 goto nla_put_failure
;
1417 if (vxlan
->link
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, vxlan
->link
))
1418 goto nla_put_failure
;
1420 if (vxlan
->saddr
&& nla_put_be32(skb
, IFLA_VXLAN_LOCAL
, vxlan
->saddr
))
1421 goto nla_put_failure
;
1423 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
1424 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
1425 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
1426 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
1427 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
1428 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
1429 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
1430 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
1431 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
1432 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
1433 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
1434 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
1435 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
))
1436 goto nla_put_failure
;
1438 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
1439 goto nla_put_failure
;
1447 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
1449 .maxtype
= IFLA_VXLAN_MAX
,
1450 .policy
= vxlan_policy
,
1451 .priv_size
= sizeof(struct vxlan_dev
),
1452 .setup
= vxlan_setup
,
1453 .validate
= vxlan_validate
,
1454 .newlink
= vxlan_newlink
,
1455 .dellink
= vxlan_dellink
,
1456 .get_size
= vxlan_get_size
,
1457 .fill_info
= vxlan_fill_info
,
1460 static __net_init
int vxlan_init_net(struct net
*net
)
1462 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1464 struct sockaddr_in vxlan_addr
= {
1465 .sin_family
= AF_INET
,
1466 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
1471 /* Create UDP socket for encapsulation receive. */
1472 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &vn
->sock
);
1474 pr_debug("UDP socket create failed\n");
1477 /* Put in proper namespace */
1479 sk_change_net(sk
, net
);
1481 vxlan_addr
.sin_port
= htons(vxlan_port
);
1483 rc
= kernel_bind(vn
->sock
, (struct sockaddr
*) &vxlan_addr
,
1484 sizeof(vxlan_addr
));
1486 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1487 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
1488 sk_release_kernel(sk
);
1493 /* Disable multicast loopback */
1494 inet_sk(sk
)->mc_loop
= 0;
1496 /* Mark socket as an encapsulation socket. */
1497 udp_sk(sk
)->encap_type
= 1;
1498 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
1501 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1502 INIT_HLIST_HEAD(&vn
->vni_list
[h
]);
1507 static __net_exit
void vxlan_exit_net(struct net
*net
)
1509 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1512 sk_release_kernel(vn
->sock
->sk
);
1517 static struct pernet_operations vxlan_net_ops
= {
1518 .init
= vxlan_init_net
,
1519 .exit
= vxlan_exit_net
,
1520 .id
= &vxlan_net_id
,
1521 .size
= sizeof(struct vxlan_net
),
1524 static int __init
vxlan_init_module(void)
1528 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
1530 rc
= register_pernet_device(&vxlan_net_ops
);
1534 rc
= rtnl_link_register(&vxlan_link_ops
);
1541 unregister_pernet_device(&vxlan_net_ops
);
1545 module_init(vxlan_init_module
);
1547 static void __exit
vxlan_cleanup_module(void)
1549 rtnl_link_unregister(&vxlan_link_ops
);
1550 unregister_pernet_device(&vxlan_net_ops
);
1552 module_exit(vxlan_cleanup_module
);
1554 MODULE_LICENSE("GPL");
1555 MODULE_VERSION(VXLAN_VERSION
);
1556 MODULE_AUTHOR("Stephen Hemminger <shemminger@vyatta.com>");
1557 MODULE_ALIAS_RTNL_LINK("vxlan");