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 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 * The IANA assigned port is 4789, but the Linux default is 8472
69 * for compatability with early adopters.
71 static unsigned int vxlan_port __read_mostly
= 8472;
72 module_param_named(udp_port
, vxlan_port
, uint
, 0444);
73 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
75 static bool log_ecn_error
= true;
76 module_param(log_ecn_error
, bool, 0644);
77 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
79 /* per-net private data for this module */
80 static unsigned int vxlan_net_id
;
82 struct socket
*sock
; /* UDP encap socket */
83 struct hlist_head vni_list
[VNI_HASH_SIZE
];
92 struct vxlan_rdst
*remote_next
;
95 /* Forwarding table entry */
97 struct hlist_node hlist
; /* linked list of entries */
99 unsigned long updated
; /* jiffies */
101 struct vxlan_rdst remote
;
102 u16 state
; /* see ndm_state */
103 u8 flags
; /* see ndm_flags */
104 u8 eth_addr
[ETH_ALEN
];
107 /* Pseudo network device */
109 struct hlist_node hlist
;
110 struct net_device
*dev
;
111 struct vxlan_rdst default_dst
; /* default destination */
112 __be32 saddr
; /* source address */
114 __u16 port_min
; /* source port range */
116 __u8 tos
; /* TOS override */
118 u32 flags
; /* VXLAN_F_* below */
120 unsigned long age_interval
;
121 struct timer_list age_timer
;
122 spinlock_t hash_lock
;
123 unsigned int addrcnt
;
124 unsigned int addrmax
;
126 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
129 #define VXLAN_F_LEARN 0x01
130 #define VXLAN_F_PROXY 0x02
131 #define VXLAN_F_RSC 0x04
132 #define VXLAN_F_L2MISS 0x08
133 #define VXLAN_F_L3MISS 0x10
135 /* salt for hash table */
136 static u32 vxlan_salt __read_mostly
;
138 static inline struct hlist_head
*vni_head(struct net
*net
, u32 id
)
140 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
142 return &vn
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
145 /* Look up VNI in a per net namespace table */
146 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
)
148 struct vxlan_dev
*vxlan
;
150 hlist_for_each_entry_rcu(vxlan
, vni_head(net
, id
), hlist
) {
151 if (vxlan
->default_dst
.remote_vni
== id
)
158 /* Fill in neighbour message in skbuff. */
159 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
160 const struct vxlan_fdb
*fdb
,
161 u32 portid
, u32 seq
, int type
, unsigned int flags
,
162 const struct vxlan_rdst
*rdst
)
164 unsigned long now
= jiffies
;
165 struct nda_cacheinfo ci
;
166 struct nlmsghdr
*nlh
;
168 bool send_ip
, send_eth
;
170 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
174 ndm
= nlmsg_data(nlh
);
175 memset(ndm
, 0, sizeof(*ndm
));
177 send_eth
= send_ip
= true;
179 if (type
== RTM_GETNEIGH
) {
180 ndm
->ndm_family
= AF_INET
;
181 send_ip
= rdst
->remote_ip
!= htonl(INADDR_ANY
);
182 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
184 ndm
->ndm_family
= AF_BRIDGE
;
185 ndm
->ndm_state
= fdb
->state
;
186 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
187 ndm
->ndm_flags
= fdb
->flags
;
188 ndm
->ndm_type
= NDA_DST
;
190 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
191 goto nla_put_failure
;
193 if (send_ip
&& nla_put_be32(skb
, NDA_DST
, rdst
->remote_ip
))
194 goto nla_put_failure
;
196 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
197 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
198 goto nla_put_failure
;
199 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
200 nla_put_be32(skb
, NDA_VNI
, rdst
->remote_vni
))
201 goto nla_put_failure
;
202 if (rdst
->remote_ifindex
&&
203 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
204 goto nla_put_failure
;
206 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
207 ci
.ndm_confirmed
= 0;
208 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
211 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
212 goto nla_put_failure
;
214 return nlmsg_end(skb
, nlh
);
217 nlmsg_cancel(skb
, nlh
);
221 static inline size_t vxlan_nlmsg_size(void)
223 return NLMSG_ALIGN(sizeof(struct ndmsg
))
224 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
225 + nla_total_size(sizeof(__be32
)) /* NDA_DST */
226 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
227 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
228 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
229 + nla_total_size(sizeof(struct nda_cacheinfo
));
232 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
,
233 const struct vxlan_fdb
*fdb
, int type
)
235 struct net
*net
= dev_net(vxlan
->dev
);
239 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
243 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, &fdb
->remote
);
245 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
246 WARN_ON(err
== -EMSGSIZE
);
251 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
255 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
258 static void vxlan_ip_miss(struct net_device
*dev
, __be32 ipa
)
260 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
263 memset(&f
, 0, sizeof f
);
265 f
.remote
.remote_ip
= ipa
; /* goes to NDA_DST */
266 f
.remote
.remote_vni
= VXLAN_N_VID
;
268 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
271 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
275 memset(&f
, 0, sizeof f
);
277 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
279 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
282 /* Hash Ethernet address */
283 static u32
eth_hash(const unsigned char *addr
)
285 u64 value
= get_unaligned((u64
*)addr
);
287 /* only want 6 bytes */
293 return hash_64(value
, FDB_HASH_BITS
);
296 /* Hash chain to use given mac address */
297 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
300 return &vxlan
->fdb_head
[eth_hash(mac
)];
303 /* Look up Ethernet address in forwarding table */
304 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
308 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
311 hlist_for_each_entry_rcu(f
, head
, hlist
) {
312 if (compare_ether_addr(mac
, f
->eth_addr
) == 0)
319 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
324 f
= __vxlan_find_mac(vxlan
, mac
);
331 /* Add/update destinations for multicast */
332 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
333 __be32 ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
335 struct vxlan_rdst
*rd_prev
, *rd
;
338 for (rd
= &f
->remote
; rd
; rd
= rd
->remote_next
) {
339 if (rd
->remote_ip
== ip
&&
340 rd
->remote_port
== port
&&
341 rd
->remote_vni
== vni
&&
342 rd
->remote_ifindex
== ifindex
)
346 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
350 rd
->remote_port
= port
;
351 rd
->remote_vni
= vni
;
352 rd
->remote_ifindex
= ifindex
;
353 rd
->remote_next
= NULL
;
354 rd_prev
->remote_next
= rd
;
358 /* Add new entry to forwarding table -- assumes lock held */
359 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
360 const u8
*mac
, __be32 ip
,
361 __u16 state
, __u16 flags
,
362 __be16 port
, __u32 vni
, __u32 ifindex
,
368 f
= __vxlan_find_mac(vxlan
, mac
);
370 if (flags
& NLM_F_EXCL
) {
371 netdev_dbg(vxlan
->dev
,
372 "lost race to create %pM\n", mac
);
375 if (f
->state
!= state
) {
377 f
->updated
= jiffies
;
380 if (f
->flags
!= ndm_flags
) {
381 f
->flags
= ndm_flags
;
382 f
->updated
= jiffies
;
385 if ((flags
& NLM_F_APPEND
) &&
386 is_multicast_ether_addr(f
->eth_addr
)) {
387 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
);
394 if (!(flags
& NLM_F_CREATE
))
397 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
400 netdev_dbg(vxlan
->dev
, "add %pM -> %pI4\n", mac
, &ip
);
401 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
406 f
->remote
.remote_ip
= ip
;
407 f
->remote
.remote_port
= port
;
408 f
->remote
.remote_vni
= vni
;
409 f
->remote
.remote_ifindex
= ifindex
;
410 f
->remote
.remote_next
= NULL
;
412 f
->flags
= ndm_flags
;
413 f
->updated
= f
->used
= jiffies
;
414 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
417 hlist_add_head_rcu(&f
->hlist
,
418 vxlan_fdb_head(vxlan
, mac
));
422 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
427 static void vxlan_fdb_free(struct rcu_head
*head
)
429 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
431 while (f
->remote
.remote_next
) {
432 struct vxlan_rdst
*rd
= f
->remote
.remote_next
;
434 f
->remote
.remote_next
= rd
->remote_next
;
440 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
442 netdev_dbg(vxlan
->dev
,
443 "delete %pM\n", f
->eth_addr
);
446 vxlan_fdb_notify(vxlan
, f
, RTM_DELNEIGH
);
448 hlist_del_rcu(&f
->hlist
);
449 call_rcu(&f
->rcu
, vxlan_fdb_free
);
452 /* Add static entry (via netlink) */
453 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
454 struct net_device
*dev
,
455 const unsigned char *addr
, u16 flags
)
457 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
458 struct net
*net
= dev_net(vxlan
->dev
);
464 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
465 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
470 if (tb
[NDA_DST
] == NULL
)
473 if (nla_len(tb
[NDA_DST
]) != sizeof(__be32
))
474 return -EAFNOSUPPORT
;
476 ip
= nla_get_be32(tb
[NDA_DST
]);
479 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
481 port
= nla_get_be16(tb
[NDA_PORT
]);
483 port
= vxlan
->dst_port
;
486 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
488 vni
= nla_get_u32(tb
[NDA_VNI
]);
490 vni
= vxlan
->default_dst
.remote_vni
;
492 if (tb
[NDA_IFINDEX
]) {
493 struct net_device
*tdev
;
495 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
497 ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
498 tdev
= dev_get_by_index(net
, ifindex
);
500 return -EADDRNOTAVAIL
;
505 spin_lock_bh(&vxlan
->hash_lock
);
506 err
= vxlan_fdb_create(vxlan
, addr
, ip
, ndm
->ndm_state
, flags
,
507 port
, vni
, ifindex
, ndm
->ndm_flags
);
508 spin_unlock_bh(&vxlan
->hash_lock
);
513 /* Delete entry (via netlink) */
514 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
515 struct net_device
*dev
,
516 const unsigned char *addr
)
518 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
522 spin_lock_bh(&vxlan
->hash_lock
);
523 f
= vxlan_find_mac(vxlan
, addr
);
525 vxlan_fdb_destroy(vxlan
, f
);
528 spin_unlock_bh(&vxlan
->hash_lock
);
533 /* Dump forwarding table */
534 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
535 struct net_device
*dev
, int idx
)
537 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
540 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
544 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
545 struct vxlan_rdst
*rd
;
546 for (rd
= &f
->remote
; rd
; rd
= rd
->remote_next
) {
547 if (idx
< cb
->args
[0])
550 err
= vxlan_fdb_info(skb
, vxlan
, f
,
551 NETLINK_CB(cb
->skb
).portid
,
566 /* Watch incoming packets to learn mapping between Ethernet address
567 * and Tunnel endpoint.
569 static void vxlan_snoop(struct net_device
*dev
,
570 __be32 src_ip
, const u8
*src_mac
)
572 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
576 f
= vxlan_find_mac(vxlan
, src_mac
);
578 if (likely(f
->remote
.remote_ip
== src_ip
))
583 "%pM migrated from %pI4 to %pI4\n",
584 src_mac
, &f
->remote
.remote_ip
, &src_ip
);
586 f
->remote
.remote_ip
= src_ip
;
587 f
->updated
= jiffies
;
589 /* learned new entry */
590 spin_lock(&vxlan
->hash_lock
);
591 err
= vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
593 NLM_F_EXCL
|NLM_F_CREATE
,
595 vxlan
->default_dst
.remote_vni
,
597 spin_unlock(&vxlan
->hash_lock
);
602 /* See if multicast group is already in use by other ID */
603 static bool vxlan_group_used(struct vxlan_net
*vn
,
604 const struct vxlan_dev
*this)
606 const struct vxlan_dev
*vxlan
;
609 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
610 hlist_for_each_entry(vxlan
, &vn
->vni_list
[h
], hlist
) {
614 if (!netif_running(vxlan
->dev
))
617 if (vxlan
->default_dst
.remote_ip
== this->default_dst
.remote_ip
)
624 /* kernel equivalent to IP_ADD_MEMBERSHIP */
625 static int vxlan_join_group(struct net_device
*dev
)
627 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
628 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
629 struct sock
*sk
= vn
->sock
->sk
;
630 struct ip_mreqn mreq
= {
631 .imr_multiaddr
.s_addr
= vxlan
->default_dst
.remote_ip
,
632 .imr_ifindex
= vxlan
->default_dst
.remote_ifindex
,
636 /* Already a member of group */
637 if (vxlan_group_used(vn
, vxlan
))
640 /* Need to drop RTNL to call multicast join */
643 err
= ip_mc_join_group(sk
, &mreq
);
651 /* kernel equivalent to IP_DROP_MEMBERSHIP */
652 static int vxlan_leave_group(struct net_device
*dev
)
654 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
655 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
657 struct sock
*sk
= vn
->sock
->sk
;
658 struct ip_mreqn mreq
= {
659 .imr_multiaddr
.s_addr
= vxlan
->default_dst
.remote_ip
,
660 .imr_ifindex
= vxlan
->default_dst
.remote_ifindex
,
663 /* Only leave group when last vxlan is done. */
664 if (vxlan_group_used(vn
, vxlan
))
667 /* Need to drop RTNL to call multicast leave */
670 err
= ip_mc_leave_group(sk
, &mreq
);
677 /* Callback from net/ipv4/udp.c to receive packets */
678 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
681 struct vxlanhdr
*vxh
;
682 struct vxlan_dev
*vxlan
;
683 struct pcpu_tstats
*stats
;
687 /* pop off outer UDP header */
688 __skb_pull(skb
, sizeof(struct udphdr
));
690 /* Need Vxlan and inner Ethernet header to be present */
691 if (!pskb_may_pull(skb
, sizeof(struct vxlanhdr
)))
694 /* Drop packets with reserved bits set */
695 vxh
= (struct vxlanhdr
*) skb
->data
;
696 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
697 (vxh
->vx_vni
& htonl(0xff))) {
698 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
699 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
703 __skb_pull(skb
, sizeof(struct vxlanhdr
));
705 /* Is this VNI defined? */
706 vni
= ntohl(vxh
->vx_vni
) >> 8;
707 vxlan
= vxlan_find_vni(sock_net(sk
), vni
);
709 netdev_dbg(skb
->dev
, "unknown vni %d\n", vni
);
713 if (!pskb_may_pull(skb
, ETH_HLEN
)) {
714 vxlan
->dev
->stats
.rx_length_errors
++;
715 vxlan
->dev
->stats
.rx_errors
++;
719 skb_reset_mac_header(skb
);
721 /* Re-examine inner Ethernet packet */
723 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
725 /* Ignore packet loops (and multicast echo) */
726 if (compare_ether_addr(eth_hdr(skb
)->h_source
,
727 vxlan
->dev
->dev_addr
) == 0)
730 if (vxlan
->flags
& VXLAN_F_LEARN
)
731 vxlan_snoop(skb
->dev
, oip
->saddr
, eth_hdr(skb
)->h_source
);
733 __skb_tunnel_rx(skb
, vxlan
->dev
);
734 skb_reset_network_header(skb
);
736 /* If the NIC driver gave us an encapsulated packet with
737 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
738 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
739 * for us. Otherwise force the upper layers to verify it.
741 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
|| !skb
->encapsulation
||
742 !(vxlan
->dev
->features
& NETIF_F_RXCSUM
))
743 skb
->ip_summed
= CHECKSUM_NONE
;
745 skb
->encapsulation
= 0;
747 err
= IP_ECN_decapsulate(oip
, skb
);
750 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
751 &oip
->saddr
, oip
->tos
);
753 ++vxlan
->dev
->stats
.rx_frame_errors
;
754 ++vxlan
->dev
->stats
.rx_errors
;
759 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
760 u64_stats_update_begin(&stats
->syncp
);
762 stats
->rx_bytes
+= skb
->len
;
763 u64_stats_update_end(&stats
->syncp
);
769 /* Put UDP header back */
770 __skb_push(skb
, sizeof(struct udphdr
));
774 /* Consume bad packet */
779 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
781 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
787 if (dev
->flags
& IFF_NOARP
)
790 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
791 dev
->stats
.tx_dropped
++;
796 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
797 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
798 parp
->ar_pro
!= htons(ETH_P_IP
) ||
799 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
800 parp
->ar_hln
!= dev
->addr_len
||
803 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
805 arpptr
+= dev
->addr_len
; /* sha */
806 memcpy(&sip
, arpptr
, sizeof(sip
));
807 arpptr
+= sizeof(sip
);
808 arpptr
+= dev
->addr_len
; /* tha */
809 memcpy(&tip
, arpptr
, sizeof(tip
));
811 if (ipv4_is_loopback(tip
) ||
812 ipv4_is_multicast(tip
))
815 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
819 struct sk_buff
*reply
;
821 if (!(n
->nud_state
& NUD_CONNECTED
)) {
826 f
= vxlan_find_mac(vxlan
, n
->ha
);
827 if (f
&& f
->remote
.remote_ip
== htonl(INADDR_ANY
)) {
828 /* bridge-local neighbor */
833 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
838 skb_reset_mac_header(reply
);
839 __skb_pull(reply
, skb_network_offset(reply
));
840 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
841 reply
->pkt_type
= PACKET_HOST
;
843 if (netif_rx_ni(reply
) == NET_RX_DROP
)
844 dev
->stats
.rx_dropped
++;
845 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
846 vxlan_ip_miss(dev
, tip
);
852 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
854 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
858 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
862 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
864 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
867 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
876 diff
= compare_ether_addr(eth_hdr(skb
)->h_dest
, n
->ha
) != 0;
878 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
880 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
884 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
885 vxlan_ip_miss(dev
, pip
->daddr
);
889 static void vxlan_sock_free(struct sk_buff
*skb
)
894 /* On transmit, associate with the tunnel socket */
895 static void vxlan_set_owner(struct net_device
*dev
, struct sk_buff
*skb
)
897 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
898 struct sock
*sk
= vn
->sock
->sk
;
903 skb
->destructor
= vxlan_sock_free
;
906 /* Compute source port for outgoing packet
907 * first choice to use L4 flow hash since it will spread
908 * better and maybe available from hardware
909 * secondary choice is to use jhash on the Ethernet header
911 static __be16
vxlan_src_port(const struct vxlan_dev
*vxlan
, struct sk_buff
*skb
)
913 unsigned int range
= (vxlan
->port_max
- vxlan
->port_min
) + 1;
916 hash
= skb_get_rxhash(skb
);
918 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
919 (__force u32
) skb
->protocol
);
921 return htons((((u64
) hash
* range
) >> 32) + vxlan
->port_min
);
924 static int handle_offloads(struct sk_buff
*skb
)
926 if (skb_is_gso(skb
)) {
927 int err
= skb_unclone(skb
, GFP_ATOMIC
);
931 skb_shinfo(skb
)->gso_type
|= SKB_GSO_UDP_TUNNEL
;
932 } else if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
933 skb
->ip_summed
= CHECKSUM_NONE
;
938 /* Bypass encapsulation if the destination is local */
939 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
940 struct vxlan_dev
*dst_vxlan
)
942 struct pcpu_tstats
*tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
943 struct pcpu_tstats
*rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
945 skb
->pkt_type
= PACKET_HOST
;
946 skb
->encapsulation
= 0;
947 skb
->dev
= dst_vxlan
->dev
;
948 __skb_pull(skb
, skb_network_offset(skb
));
950 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
951 vxlan_snoop(skb
->dev
, htonl(INADDR_LOOPBACK
),
952 eth_hdr(skb
)->h_source
);
954 u64_stats_update_begin(&tx_stats
->syncp
);
955 tx_stats
->tx_packets
++;
956 tx_stats
->tx_bytes
+= skb
->len
;
957 u64_stats_update_end(&tx_stats
->syncp
);
959 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
960 u64_stats_update_begin(&rx_stats
->syncp
);
961 rx_stats
->rx_packets
++;
962 rx_stats
->rx_bytes
+= skb
->len
;
963 u64_stats_update_end(&rx_stats
->syncp
);
965 skb
->dev
->stats
.rx_dropped
++;
969 static netdev_tx_t
vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
970 struct vxlan_rdst
*rdst
, bool did_rsc
)
972 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
974 const struct iphdr
*old_iph
;
976 struct vxlanhdr
*vxh
;
980 __be16 src_port
, dst_port
;
985 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
986 vni
= rdst
->remote_vni
;
987 dst
= rdst
->remote_ip
;
991 /* short-circuited back to local bridge */
992 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
998 if (!skb
->encapsulation
) {
999 skb_reset_inner_headers(skb
);
1000 skb
->encapsulation
= 1;
1003 /* Need space for new headers (invalidates iph ptr) */
1004 if (skb_cow_head(skb
, VXLAN_HEADROOM
))
1007 old_iph
= ip_hdr(skb
);
1010 if (!ttl
&& IN_MULTICAST(ntohl(dst
)))
1015 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1017 src_port
= vxlan_src_port(vxlan
, skb
);
1019 memset(&fl4
, 0, sizeof(fl4
));
1020 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1021 fl4
.flowi4_tos
= RT_TOS(tos
);
1023 fl4
.saddr
= vxlan
->saddr
;
1025 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
1027 netdev_dbg(dev
, "no route to %pI4\n", &dst
);
1028 dev
->stats
.tx_carrier_errors
++;
1032 if (rt
->dst
.dev
== dev
) {
1033 netdev_dbg(dev
, "circular route to %pI4\n", &dst
);
1035 dev
->stats
.collisions
++;
1039 /* Bypass encapsulation if the destination is local */
1040 if (rt
->rt_flags
& RTCF_LOCAL
&&
1041 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1042 struct vxlan_dev
*dst_vxlan
;
1045 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
);
1048 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1049 return NETDEV_TX_OK
;
1052 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1053 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1056 skb_dst_set(skb
, &rt
->dst
);
1058 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1059 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1060 vxh
->vx_vni
= htonl(vni
<< 8);
1062 __skb_push(skb
, sizeof(*uh
));
1063 skb_reset_transport_header(skb
);
1066 uh
->dest
= dst_port
;
1067 uh
->source
= src_port
;
1069 uh
->len
= htons(skb
->len
);
1072 __skb_push(skb
, sizeof(*iph
));
1073 skb_reset_network_header(skb
);
1076 iph
->ihl
= sizeof(struct iphdr
) >> 2;
1078 iph
->protocol
= IPPROTO_UDP
;
1079 iph
->tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1081 iph
->saddr
= fl4
.saddr
;
1082 iph
->ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1083 tunnel_ip_select_ident(skb
, old_iph
, &rt
->dst
);
1087 vxlan_set_owner(dev
, skb
);
1089 if (handle_offloads(skb
))
1092 iptunnel_xmit(skb
, dev
);
1093 return NETDEV_TX_OK
;
1096 dev
->stats
.tx_dropped
++;
1100 dev
->stats
.tx_errors
++;
1103 return NETDEV_TX_OK
;
1106 /* Transmit local packets over Vxlan
1108 * Outer IP header inherits ECN and DF from inner header.
1109 * Outer UDP destination is the VXLAN assigned port.
1110 * source port is based on hash of flow
1112 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1114 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1116 bool did_rsc
= false;
1117 struct vxlan_rdst
*rdst0
, *rdst
;
1118 struct vxlan_fdb
*f
;
1121 skb_reset_mac_header(skb
);
1124 if ((vxlan
->flags
& VXLAN_F_PROXY
) && ntohs(eth
->h_proto
) == ETH_P_ARP
)
1125 return arp_reduce(dev
, skb
);
1127 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1130 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1131 ntohs(eth
->h_proto
) == ETH_P_IP
) {
1132 did_rsc
= route_shortcircuit(dev
, skb
);
1134 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1138 rdst0
= &vxlan
->default_dst
;
1140 if (rdst0
->remote_ip
== htonl(INADDR_ANY
) &&
1141 (vxlan
->flags
& VXLAN_F_L2MISS
) &&
1142 !is_multicast_ether_addr(eth
->h_dest
))
1143 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
1149 /* if there are multiple destinations, send copies */
1150 for (rdst
= rdst0
->remote_next
; rdst
; rdst
= rdst
->remote_next
) {
1151 struct sk_buff
*skb1
;
1153 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1154 rc1
= vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
1155 if (rc
== NETDEV_TX_OK
)
1159 rc1
= vxlan_xmit_one(skb
, dev
, rdst0
, did_rsc
);
1160 if (rc
== NETDEV_TX_OK
)
1165 /* Walk the forwarding table and purge stale entries */
1166 static void vxlan_cleanup(unsigned long arg
)
1168 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1169 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1172 if (!netif_running(vxlan
->dev
))
1175 spin_lock_bh(&vxlan
->hash_lock
);
1176 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1177 struct hlist_node
*p
, *n
;
1178 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1180 = container_of(p
, struct vxlan_fdb
, hlist
);
1181 unsigned long timeout
;
1183 if (f
->state
& NUD_PERMANENT
)
1186 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
1187 if (time_before_eq(timeout
, jiffies
)) {
1188 netdev_dbg(vxlan
->dev
,
1189 "garbage collect %pM\n",
1191 f
->state
= NUD_STALE
;
1192 vxlan_fdb_destroy(vxlan
, f
);
1193 } else if (time_before(timeout
, next_timer
))
1194 next_timer
= timeout
;
1197 spin_unlock_bh(&vxlan
->hash_lock
);
1199 mod_timer(&vxlan
->age_timer
, next_timer
);
1202 /* Setup stats when device is created */
1203 static int vxlan_init(struct net_device
*dev
)
1205 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1212 /* Start ageing timer and join group when device is brought up */
1213 static int vxlan_open(struct net_device
*dev
)
1215 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1218 if (IN_MULTICAST(ntohl(vxlan
->default_dst
.remote_ip
))) {
1219 err
= vxlan_join_group(dev
);
1224 if (vxlan
->age_interval
)
1225 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
1230 /* Purge the forwarding table */
1231 static void vxlan_flush(struct vxlan_dev
*vxlan
)
1235 spin_lock_bh(&vxlan
->hash_lock
);
1236 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1237 struct hlist_node
*p
, *n
;
1238 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1240 = container_of(p
, struct vxlan_fdb
, hlist
);
1241 vxlan_fdb_destroy(vxlan
, f
);
1244 spin_unlock_bh(&vxlan
->hash_lock
);
1247 /* Cleanup timer and forwarding table on shutdown */
1248 static int vxlan_stop(struct net_device
*dev
)
1250 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1252 if (IN_MULTICAST(ntohl(vxlan
->default_dst
.remote_ip
)))
1253 vxlan_leave_group(dev
);
1255 del_timer_sync(&vxlan
->age_timer
);
1262 /* Stub, nothing needs to be done. */
1263 static void vxlan_set_multicast_list(struct net_device
*dev
)
1267 static const struct net_device_ops vxlan_netdev_ops
= {
1268 .ndo_init
= vxlan_init
,
1269 .ndo_open
= vxlan_open
,
1270 .ndo_stop
= vxlan_stop
,
1271 .ndo_start_xmit
= vxlan_xmit
,
1272 .ndo_get_stats64
= ip_tunnel_get_stats64
,
1273 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
1274 .ndo_change_mtu
= eth_change_mtu
,
1275 .ndo_validate_addr
= eth_validate_addr
,
1276 .ndo_set_mac_address
= eth_mac_addr
,
1277 .ndo_fdb_add
= vxlan_fdb_add
,
1278 .ndo_fdb_del
= vxlan_fdb_delete
,
1279 .ndo_fdb_dump
= vxlan_fdb_dump
,
1282 /* Info for udev, that this is a virtual tunnel endpoint */
1283 static struct device_type vxlan_type
= {
1287 static void vxlan_free(struct net_device
*dev
)
1289 free_percpu(dev
->tstats
);
1293 /* Initialize the device structure. */
1294 static void vxlan_setup(struct net_device
*dev
)
1296 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1300 eth_hw_addr_random(dev
);
1302 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
1304 dev
->netdev_ops
= &vxlan_netdev_ops
;
1305 dev
->destructor
= vxlan_free
;
1306 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
1308 dev
->tx_queue_len
= 0;
1309 dev
->features
|= NETIF_F_LLTX
;
1310 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1311 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1312 dev
->features
|= NETIF_F_RXCSUM
;
1313 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
1315 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
1316 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
1317 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1318 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1320 spin_lock_init(&vxlan
->hash_lock
);
1322 init_timer_deferrable(&vxlan
->age_timer
);
1323 vxlan
->age_timer
.function
= vxlan_cleanup
;
1324 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
1326 inet_get_local_port_range(&low
, &high
);
1327 vxlan
->port_min
= low
;
1328 vxlan
->port_max
= high
;
1329 vxlan
->dst_port
= htons(vxlan_port
);
1333 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
1334 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
1337 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
1338 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
1339 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
1340 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
1341 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
1342 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
1343 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
1344 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
1345 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
1346 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
1347 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
1348 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
1349 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
1350 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
1351 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
1352 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
1355 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
1357 if (tb
[IFLA_ADDRESS
]) {
1358 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
1359 pr_debug("invalid link address (not ethernet)\n");
1363 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
1364 pr_debug("invalid all zero ethernet address\n");
1365 return -EADDRNOTAVAIL
;
1372 if (data
[IFLA_VXLAN_ID
]) {
1373 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1374 if (id
>= VXLAN_VID_MASK
)
1378 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1379 const struct ifla_vxlan_port_range
*p
1380 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1382 if (ntohs(p
->high
) < ntohs(p
->low
)) {
1383 pr_debug("port range %u .. %u not valid\n",
1384 ntohs(p
->low
), ntohs(p
->high
));
1392 static void vxlan_get_drvinfo(struct net_device
*netdev
,
1393 struct ethtool_drvinfo
*drvinfo
)
1395 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
1396 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
1399 static const struct ethtool_ops vxlan_ethtool_ops
= {
1400 .get_drvinfo
= vxlan_get_drvinfo
,
1401 .get_link
= ethtool_op_get_link
,
1404 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
1405 struct nlattr
*tb
[], struct nlattr
*data
[])
1407 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1408 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
1412 if (!data
[IFLA_VXLAN_ID
])
1415 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1416 if (vxlan_find_vni(net
, vni
)) {
1417 pr_info("duplicate VNI %u\n", vni
);
1420 dst
->remote_vni
= vni
;
1422 if (data
[IFLA_VXLAN_GROUP
])
1423 dst
->remote_ip
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1425 if (data
[IFLA_VXLAN_LOCAL
])
1426 vxlan
->saddr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
1428 if (data
[IFLA_VXLAN_LINK
] &&
1429 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
1430 struct net_device
*lowerdev
1431 = __dev_get_by_index(net
, dst
->remote_ifindex
);
1434 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
1439 dev
->mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
1441 /* update header length based on lower device */
1442 dev
->hard_header_len
= lowerdev
->hard_header_len
+
1446 if (data
[IFLA_VXLAN_TOS
])
1447 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
1449 if (data
[IFLA_VXLAN_TTL
])
1450 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
1452 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
1453 vxlan
->flags
|= VXLAN_F_LEARN
;
1455 if (data
[IFLA_VXLAN_AGEING
])
1456 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
1458 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
1460 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
1461 vxlan
->flags
|= VXLAN_F_PROXY
;
1463 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
1464 vxlan
->flags
|= VXLAN_F_RSC
;
1466 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
1467 vxlan
->flags
|= VXLAN_F_L2MISS
;
1469 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
1470 vxlan
->flags
|= VXLAN_F_L3MISS
;
1472 if (data
[IFLA_VXLAN_LIMIT
])
1473 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
1475 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1476 const struct ifla_vxlan_port_range
*p
1477 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1478 vxlan
->port_min
= ntohs(p
->low
);
1479 vxlan
->port_max
= ntohs(p
->high
);
1482 if (data
[IFLA_VXLAN_PORT
])
1483 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
1485 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
1487 err
= register_netdevice(dev
);
1489 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(net
, dst
->remote_vni
));
1494 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
1496 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1498 hlist_del_rcu(&vxlan
->hlist
);
1500 unregister_netdevice_queue(dev
, head
);
1503 static size_t vxlan_get_size(const struct net_device
*dev
)
1506 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
1507 nla_total_size(sizeof(__be32
)) +/* IFLA_VXLAN_GROUP */
1508 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
1509 nla_total_size(sizeof(__be32
))+ /* IFLA_VXLAN_LOCAL */
1510 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
1511 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
1512 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
1513 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
1514 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
1515 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
1516 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
1517 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
1518 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
1519 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
1520 nla_total_size(sizeof(__be16
))+ /* IFLA_VXLAN_PORT */
1524 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1526 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1527 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
1528 struct ifla_vxlan_port_range ports
= {
1529 .low
= htons(vxlan
->port_min
),
1530 .high
= htons(vxlan
->port_max
),
1533 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
1534 goto nla_put_failure
;
1536 if (dst
->remote_ip
&& nla_put_be32(skb
, IFLA_VXLAN_GROUP
, dst
->remote_ip
))
1537 goto nla_put_failure
;
1539 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
1540 goto nla_put_failure
;
1542 if (vxlan
->saddr
&& nla_put_be32(skb
, IFLA_VXLAN_LOCAL
, vxlan
->saddr
))
1543 goto nla_put_failure
;
1545 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
1546 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
1547 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
1548 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
1549 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
1550 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
1551 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
1552 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
1553 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
1554 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
1555 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
1556 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
1557 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
1558 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
))
1559 goto nla_put_failure
;
1561 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
1562 goto nla_put_failure
;
1570 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
1572 .maxtype
= IFLA_VXLAN_MAX
,
1573 .policy
= vxlan_policy
,
1574 .priv_size
= sizeof(struct vxlan_dev
),
1575 .setup
= vxlan_setup
,
1576 .validate
= vxlan_validate
,
1577 .newlink
= vxlan_newlink
,
1578 .dellink
= vxlan_dellink
,
1579 .get_size
= vxlan_get_size
,
1580 .fill_info
= vxlan_fill_info
,
1583 static __net_init
int vxlan_init_net(struct net
*net
)
1585 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1587 struct sockaddr_in vxlan_addr
= {
1588 .sin_family
= AF_INET
,
1589 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
1594 /* Create UDP socket for encapsulation receive. */
1595 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &vn
->sock
);
1597 pr_debug("UDP socket create failed\n");
1600 /* Put in proper namespace */
1602 sk_change_net(sk
, net
);
1604 vxlan_addr
.sin_port
= htons(vxlan_port
);
1606 rc
= kernel_bind(vn
->sock
, (struct sockaddr
*) &vxlan_addr
,
1607 sizeof(vxlan_addr
));
1609 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1610 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
1611 sk_release_kernel(sk
);
1616 /* Disable multicast loopback */
1617 inet_sk(sk
)->mc_loop
= 0;
1619 /* Mark socket as an encapsulation socket. */
1620 udp_sk(sk
)->encap_type
= 1;
1621 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
1624 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1625 INIT_HLIST_HEAD(&vn
->vni_list
[h
]);
1630 static __net_exit
void vxlan_exit_net(struct net
*net
)
1632 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1633 struct vxlan_dev
*vxlan
;
1637 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1638 hlist_for_each_entry(vxlan
, &vn
->vni_list
[h
], hlist
)
1639 dev_close(vxlan
->dev
);
1643 sk_release_kernel(vn
->sock
->sk
);
1648 static struct pernet_operations vxlan_net_ops
= {
1649 .init
= vxlan_init_net
,
1650 .exit
= vxlan_exit_net
,
1651 .id
= &vxlan_net_id
,
1652 .size
= sizeof(struct vxlan_net
),
1655 static int __init
vxlan_init_module(void)
1659 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
1661 rc
= register_pernet_device(&vxlan_net_ops
);
1665 rc
= rtnl_link_register(&vxlan_link_ops
);
1672 unregister_pernet_device(&vxlan_net_ops
);
1676 module_init(vxlan_init_module
);
1678 static void __exit
vxlan_cleanup_module(void)
1680 rtnl_link_unregister(&vxlan_link_ops
);
1681 unregister_pernet_device(&vxlan_net_ops
);
1684 module_exit(vxlan_cleanup_module
);
1686 MODULE_LICENSE("GPL");
1687 MODULE_VERSION(VXLAN_VERSION
);
1688 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
1689 MODULE_ALIAS_RTNL_LINK("vxlan");