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.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/kernel.h>
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
31 #include <net/ndisc.h>
33 #include <net/ip_tunnels.h>
36 #include <net/rtnetlink.h>
37 #include <net/route.h>
38 #include <net/dsfield.h>
39 #include <net/inet_ecn.h>
40 #include <net/net_namespace.h>
41 #include <net/netns/generic.h>
42 #include <net/vxlan.h>
43 #include <net/protocol.h>
44 #if IS_ENABLED(CONFIG_IPV6)
46 #include <net/addrconf.h>
47 #include <net/ip6_tunnel.h>
48 #include <net/ip6_checksum.h>
51 #define VXLAN_VERSION "0.1"
53 #define PORT_HASH_BITS 8
54 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
55 #define VNI_HASH_BITS 10
56 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
57 #define FDB_HASH_BITS 8
58 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
59 #define FDB_AGE_DEFAULT 300 /* 5 min */
60 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
62 #define VXLAN_N_VID (1u << 24)
63 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
64 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
66 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
68 /* VXLAN protocol header */
74 /* UDP port for VXLAN traffic.
75 * The IANA assigned port is 4789, but the Linux default is 8472
76 * for compatibility with early adopters.
78 static unsigned short vxlan_port __read_mostly
= 8472;
79 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
80 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
82 static bool log_ecn_error
= true;
83 module_param(log_ecn_error
, bool, 0644);
84 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
86 static int vxlan_net_id
;
88 static const u8 all_zeros_mac
[ETH_ALEN
];
90 /* per-network namespace private data for this module */
92 struct list_head vxlan_list
;
93 struct hlist_head sock_list
[PORT_HASH_SIZE
];
98 struct sockaddr_in sin
;
99 struct sockaddr_in6 sin6
;
104 union vxlan_addr remote_ip
;
108 struct list_head list
;
112 /* Forwarding table entry */
114 struct hlist_node hlist
; /* linked list of entries */
116 unsigned long updated
; /* jiffies */
118 struct list_head remotes
;
119 u16 state
; /* see ndm_state */
120 u8 flags
; /* see ndm_flags */
121 u8 eth_addr
[ETH_ALEN
];
124 /* Pseudo network device */
126 struct hlist_node hlist
; /* vni hash table */
127 struct list_head next
; /* vxlan's per namespace list */
128 struct vxlan_sock
*vn_sock
; /* listening socket */
129 struct net_device
*dev
;
130 struct vxlan_rdst default_dst
; /* default destination */
131 union vxlan_addr saddr
; /* source address */
133 __u16 port_min
; /* source port range */
135 __u8 tos
; /* TOS override */
137 u32 flags
; /* VXLAN_F_* below */
139 struct work_struct sock_work
;
140 struct work_struct igmp_join
;
141 struct work_struct igmp_leave
;
143 unsigned long age_interval
;
144 struct timer_list age_timer
;
145 spinlock_t hash_lock
;
146 unsigned int addrcnt
;
147 unsigned int addrmax
;
149 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
152 #define VXLAN_F_LEARN 0x01
153 #define VXLAN_F_PROXY 0x02
154 #define VXLAN_F_RSC 0x04
155 #define VXLAN_F_L2MISS 0x08
156 #define VXLAN_F_L3MISS 0x10
157 #define VXLAN_F_IPV6 0x20 /* internal flag */
159 /* salt for hash table */
160 static u32 vxlan_salt __read_mostly
;
161 static struct workqueue_struct
*vxlan_wq
;
163 static void vxlan_sock_work(struct work_struct
*work
);
165 #if IS_ENABLED(CONFIG_IPV6)
167 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
169 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
171 if (a
->sa
.sa_family
== AF_INET6
)
172 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
174 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
177 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
179 if (ipa
->sa
.sa_family
== AF_INET6
)
180 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
182 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
185 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
187 if (ipa
->sa
.sa_family
== AF_INET6
)
188 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
190 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
193 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
195 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
196 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
197 ip
->sa
.sa_family
= AF_INET6
;
199 } else if (nla_len(nla
) >= sizeof(__be32
)) {
200 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
201 ip
->sa
.sa_family
= AF_INET
;
204 return -EAFNOSUPPORT
;
208 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
209 const union vxlan_addr
*ip
)
211 if (ip
->sa
.sa_family
== AF_INET6
)
212 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
214 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
217 #else /* !CONFIG_IPV6 */
220 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
222 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
225 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
227 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
230 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
232 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
235 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
237 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
238 return -EAFNOSUPPORT
;
239 } else if (nla_len(nla
) >= sizeof(__be32
)) {
240 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
241 ip
->sa
.sa_family
= AF_INET
;
244 return -EAFNOSUPPORT
;
248 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
249 const union vxlan_addr
*ip
)
251 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
255 /* Virtual Network hash table head */
256 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
258 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
261 /* Socket hash table head */
262 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
264 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
266 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
269 /* First remote destination for a forwarding entry.
270 * Guaranteed to be non-NULL because remotes are never deleted.
272 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
274 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
277 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
279 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
282 /* Find VXLAN socket based on network namespace and UDP port */
283 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, __be16 port
)
285 struct vxlan_sock
*vs
;
287 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
288 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
294 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
296 struct vxlan_dev
*vxlan
;
298 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
299 if (vxlan
->default_dst
.remote_vni
== id
)
306 /* Look up VNI in a per net namespace table */
307 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
309 struct vxlan_sock
*vs
;
311 vs
= vxlan_find_sock(net
, port
);
315 return vxlan_vs_find_vni(vs
, id
);
318 /* Fill in neighbour message in skbuff. */
319 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
320 const struct vxlan_fdb
*fdb
,
321 u32 portid
, u32 seq
, int type
, unsigned int flags
,
322 const struct vxlan_rdst
*rdst
)
324 unsigned long now
= jiffies
;
325 struct nda_cacheinfo ci
;
326 struct nlmsghdr
*nlh
;
328 bool send_ip
, send_eth
;
330 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
334 ndm
= nlmsg_data(nlh
);
335 memset(ndm
, 0, sizeof(*ndm
));
337 send_eth
= send_ip
= true;
339 if (type
== RTM_GETNEIGH
) {
340 ndm
->ndm_family
= AF_INET
;
341 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
342 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
344 ndm
->ndm_family
= AF_BRIDGE
;
345 ndm
->ndm_state
= fdb
->state
;
346 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
347 ndm
->ndm_flags
= fdb
->flags
;
348 ndm
->ndm_type
= NDA_DST
;
350 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
351 goto nla_put_failure
;
353 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
354 goto nla_put_failure
;
356 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
357 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
358 goto nla_put_failure
;
359 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
360 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
361 goto nla_put_failure
;
362 if (rdst
->remote_ifindex
&&
363 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
364 goto nla_put_failure
;
366 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
367 ci
.ndm_confirmed
= 0;
368 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
371 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
372 goto nla_put_failure
;
374 return nlmsg_end(skb
, nlh
);
377 nlmsg_cancel(skb
, nlh
);
381 static inline size_t vxlan_nlmsg_size(void)
383 return NLMSG_ALIGN(sizeof(struct ndmsg
))
384 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
385 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
386 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
387 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
388 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
389 + nla_total_size(sizeof(struct nda_cacheinfo
));
392 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
393 struct vxlan_rdst
*rd
, int type
)
395 struct net
*net
= dev_net(vxlan
->dev
);
399 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
403 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
405 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
406 WARN_ON(err
== -EMSGSIZE
);
411 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
415 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
418 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
420 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
421 struct vxlan_fdb f
= {
424 struct vxlan_rdst remote
= {
425 .remote_ip
= *ipa
, /* goes to NDA_DST */
426 .remote_vni
= VXLAN_N_VID
,
429 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
432 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
434 struct vxlan_fdb f
= {
437 struct vxlan_rdst remote
= { };
439 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
441 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
444 /* Hash Ethernet address */
445 static u32
eth_hash(const unsigned char *addr
)
447 u64 value
= get_unaligned((u64
*)addr
);
449 /* only want 6 bytes */
455 return hash_64(value
, FDB_HASH_BITS
);
458 /* Hash chain to use given mac address */
459 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
462 return &vxlan
->fdb_head
[eth_hash(mac
)];
465 /* Look up Ethernet address in forwarding table */
466 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
469 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
472 hlist_for_each_entry_rcu(f
, head
, hlist
) {
473 if (ether_addr_equal(mac
, f
->eth_addr
))
480 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
485 f
= __vxlan_find_mac(vxlan
, mac
);
492 /* caller should hold vxlan->hash_lock */
493 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
494 union vxlan_addr
*ip
, __be16 port
,
495 __u32 vni
, __u32 ifindex
)
497 struct vxlan_rdst
*rd
;
499 list_for_each_entry(rd
, &f
->remotes
, list
) {
500 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
501 rd
->remote_port
== port
&&
502 rd
->remote_vni
== vni
&&
503 rd
->remote_ifindex
== ifindex
)
510 /* Replace destination of unicast mac */
511 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
512 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
514 struct vxlan_rdst
*rd
;
516 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
520 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
524 rd
->remote_port
= port
;
525 rd
->remote_vni
= vni
;
526 rd
->remote_ifindex
= ifindex
;
530 /* Add/update destinations for multicast */
531 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
532 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
533 __u32 ifindex
, struct vxlan_rdst
**rdp
)
535 struct vxlan_rdst
*rd
;
537 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
541 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
545 rd
->remote_port
= port
;
546 rd
->remote_vni
= vni
;
547 rd
->remote_ifindex
= ifindex
;
549 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
555 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
)
557 struct sk_buff
*p
, **pp
= NULL
;
558 struct vxlanhdr
*vh
, *vh2
;
559 struct ethhdr
*eh
, *eh2
;
560 unsigned int hlen
, off_vx
, off_eth
;
561 const struct packet_offload
*ptype
;
565 off_vx
= skb_gro_offset(skb
);
566 hlen
= off_vx
+ sizeof(*vh
);
567 vh
= skb_gro_header_fast(skb
, off_vx
);
568 if (skb_gro_header_hard(skb
, hlen
)) {
569 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
573 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
575 off_eth
= skb_gro_offset(skb
);
576 hlen
= off_eth
+ sizeof(*eh
);
577 eh
= skb_gro_header_fast(skb
, off_eth
);
578 if (skb_gro_header_hard(skb
, hlen
)) {
579 eh
= skb_gro_header_slow(skb
, hlen
, off_eth
);
586 for (p
= *head
; p
; p
= p
->next
) {
587 if (!NAPI_GRO_CB(p
)->same_flow
)
590 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
591 eh2
= (struct ethhdr
*)(p
->data
+ off_eth
);
592 if (vh
->vx_vni
!= vh2
->vx_vni
|| compare_ether_header(eh
, eh2
)) {
593 NAPI_GRO_CB(p
)->same_flow
= 0;
601 ptype
= gro_find_receive_by_type(type
);
607 skb_gro_pull(skb
, sizeof(*eh
)); /* pull inner eth header */
608 pp
= ptype
->callbacks
.gro_receive(head
, skb
);
613 NAPI_GRO_CB(skb
)->flush
|= flush
;
618 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
)
621 struct packet_offload
*ptype
;
623 int vxlan_len
= sizeof(struct vxlanhdr
) + sizeof(struct ethhdr
);
626 eh
= (struct ethhdr
*)(skb
->data
+ nhoff
+ sizeof(struct vxlanhdr
));
630 ptype
= gro_find_complete_by_type(type
);
632 err
= ptype
->callbacks
.gro_complete(skb
, nhoff
+ vxlan_len
);
638 /* Notify netdevs that UDP port started listening */
639 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
641 struct net_device
*dev
;
642 struct sock
*sk
= vs
->sock
->sk
;
643 struct net
*net
= sock_net(sk
);
644 sa_family_t sa_family
= sk
->sk_family
;
645 __be16 port
= inet_sk(sk
)->inet_sport
;
648 if (sa_family
== AF_INET
) {
649 err
= udp_add_offload(&vs
->udp_offloads
);
651 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
655 for_each_netdev_rcu(net
, dev
) {
656 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
657 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
663 /* Notify netdevs that UDP port is no more listening */
664 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
666 struct net_device
*dev
;
667 struct sock
*sk
= vs
->sock
->sk
;
668 struct net
*net
= sock_net(sk
);
669 sa_family_t sa_family
= sk
->sk_family
;
670 __be16 port
= inet_sk(sk
)->inet_sport
;
673 for_each_netdev_rcu(net
, dev
) {
674 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
675 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
680 if (sa_family
== AF_INET
)
681 udp_del_offload(&vs
->udp_offloads
);
684 /* Add new entry to forwarding table -- assumes lock held */
685 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
686 const u8
*mac
, union vxlan_addr
*ip
,
687 __u16 state
, __u16 flags
,
688 __be16 port
, __u32 vni
, __u32 ifindex
,
691 struct vxlan_rdst
*rd
= NULL
;
695 f
= __vxlan_find_mac(vxlan
, mac
);
697 if (flags
& NLM_F_EXCL
) {
698 netdev_dbg(vxlan
->dev
,
699 "lost race to create %pM\n", mac
);
702 if (f
->state
!= state
) {
704 f
->updated
= jiffies
;
707 if (f
->flags
!= ndm_flags
) {
708 f
->flags
= ndm_flags
;
709 f
->updated
= jiffies
;
712 if ((flags
& NLM_F_REPLACE
)) {
713 /* Only change unicasts */
714 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
715 is_zero_ether_addr(f
->eth_addr
))) {
716 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
725 if ((flags
& NLM_F_APPEND
) &&
726 (is_multicast_ether_addr(f
->eth_addr
) ||
727 is_zero_ether_addr(f
->eth_addr
))) {
728 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
736 if (!(flags
& NLM_F_CREATE
))
739 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
742 /* Disallow replace to add a multicast entry */
743 if ((flags
& NLM_F_REPLACE
) &&
744 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
747 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
748 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
754 f
->flags
= ndm_flags
;
755 f
->updated
= f
->used
= jiffies
;
756 INIT_LIST_HEAD(&f
->remotes
);
757 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
759 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
762 hlist_add_head_rcu(&f
->hlist
,
763 vxlan_fdb_head(vxlan
, mac
));
768 rd
= first_remote_rtnl(f
);
769 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
775 static void vxlan_fdb_free(struct rcu_head
*head
)
777 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
778 struct vxlan_rdst
*rd
, *nd
;
780 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
785 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
787 netdev_dbg(vxlan
->dev
,
788 "delete %pM\n", f
->eth_addr
);
791 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
793 hlist_del_rcu(&f
->hlist
);
794 call_rcu(&f
->rcu
, vxlan_fdb_free
);
797 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
798 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
800 struct net
*net
= dev_net(vxlan
->dev
);
804 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
808 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
809 if (remote
->sa
.sa_family
== AF_INET
) {
810 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
811 ip
->sa
.sa_family
= AF_INET
;
812 #if IS_ENABLED(CONFIG_IPV6)
814 ip
->sin6
.sin6_addr
= in6addr_any
;
815 ip
->sa
.sa_family
= AF_INET6
;
821 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
823 *port
= nla_get_be16(tb
[NDA_PORT
]);
825 *port
= vxlan
->dst_port
;
829 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
831 *vni
= nla_get_u32(tb
[NDA_VNI
]);
833 *vni
= vxlan
->default_dst
.remote_vni
;
836 if (tb
[NDA_IFINDEX
]) {
837 struct net_device
*tdev
;
839 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
841 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
842 tdev
= __dev_get_by_index(net
, *ifindex
);
844 return -EADDRNOTAVAIL
;
852 /* Add static entry (via netlink) */
853 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
854 struct net_device
*dev
,
855 const unsigned char *addr
, u16 flags
)
857 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
858 /* struct net *net = dev_net(vxlan->dev); */
864 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
865 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
870 if (tb
[NDA_DST
] == NULL
)
873 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
877 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
878 return -EAFNOSUPPORT
;
880 spin_lock_bh(&vxlan
->hash_lock
);
881 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
882 port
, vni
, ifindex
, ndm
->ndm_flags
);
883 spin_unlock_bh(&vxlan
->hash_lock
);
888 /* Delete entry (via netlink) */
889 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
890 struct net_device
*dev
,
891 const unsigned char *addr
)
893 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
895 struct vxlan_rdst
*rd
= NULL
;
901 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
907 spin_lock_bh(&vxlan
->hash_lock
);
908 f
= vxlan_find_mac(vxlan
, addr
);
912 if (!vxlan_addr_any(&ip
)) {
913 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
920 /* remove a destination if it's not the only one on the list,
921 * otherwise destroy the fdb entry
923 if (rd
&& !list_is_singular(&f
->remotes
)) {
924 list_del_rcu(&rd
->list
);
925 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
930 vxlan_fdb_destroy(vxlan
, f
);
933 spin_unlock_bh(&vxlan
->hash_lock
);
938 /* Dump forwarding table */
939 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
940 struct net_device
*dev
, int idx
)
942 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
945 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
949 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
950 struct vxlan_rdst
*rd
;
952 if (idx
< cb
->args
[0])
955 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
956 err
= vxlan_fdb_info(skb
, vxlan
, f
,
957 NETLINK_CB(cb
->skb
).portid
,
972 /* Watch incoming packets to learn mapping between Ethernet address
973 * and Tunnel endpoint.
974 * Return true if packet is bogus and should be droppped.
976 static bool vxlan_snoop(struct net_device
*dev
,
977 union vxlan_addr
*src_ip
, const u8
*src_mac
)
979 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
982 f
= vxlan_find_mac(vxlan
, src_mac
);
984 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
986 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
989 /* Don't migrate static entries, drop packets */
990 if (f
->state
& NUD_NOARP
)
995 "%pM migrated from %pIS to %pIS\n",
996 src_mac
, &rdst
->remote_ip
, &src_ip
);
998 rdst
->remote_ip
= *src_ip
;
999 f
->updated
= jiffies
;
1000 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1002 /* learned new entry */
1003 spin_lock(&vxlan
->hash_lock
);
1005 /* close off race between vxlan_flush and incoming packets */
1006 if (netif_running(dev
))
1007 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1009 NLM_F_EXCL
|NLM_F_CREATE
,
1011 vxlan
->default_dst
.remote_vni
,
1013 spin_unlock(&vxlan
->hash_lock
);
1019 /* See if multicast group is already in use by other ID */
1020 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1022 struct vxlan_dev
*vxlan
;
1024 /* The vxlan_sock is only used by dev, leaving group has
1025 * no effect on other vxlan devices.
1027 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1030 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1031 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1034 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1037 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1038 &dev
->default_dst
.remote_ip
))
1041 if (vxlan
->default_dst
.remote_ifindex
!=
1042 dev
->default_dst
.remote_ifindex
)
1051 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1053 atomic_inc(&vs
->refcnt
);
1056 void vxlan_sock_release(struct vxlan_sock
*vs
)
1058 struct sock
*sk
= vs
->sock
->sk
;
1059 struct net
*net
= sock_net(sk
);
1060 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1062 if (!atomic_dec_and_test(&vs
->refcnt
))
1065 spin_lock(&vn
->sock_lock
);
1066 hlist_del_rcu(&vs
->hlist
);
1067 rcu_assign_sk_user_data(vs
->sock
->sk
, NULL
);
1068 vxlan_notify_del_rx_port(vs
);
1069 spin_unlock(&vn
->sock_lock
);
1071 queue_work(vxlan_wq
, &vs
->del_work
);
1073 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1075 /* Callback to update multicast group membership when first VNI on
1076 * multicast asddress is brought up
1077 * Done as workqueue because ip_mc_join_group acquires RTNL.
1079 static void vxlan_igmp_join(struct work_struct
*work
)
1081 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1082 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1083 struct sock
*sk
= vs
->sock
->sk
;
1084 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1085 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1088 if (ip
->sa
.sa_family
== AF_INET
) {
1089 struct ip_mreqn mreq
= {
1090 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1091 .imr_ifindex
= ifindex
,
1094 ip_mc_join_group(sk
, &mreq
);
1095 #if IS_ENABLED(CONFIG_IPV6)
1097 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1098 &ip
->sin6
.sin6_addr
);
1103 vxlan_sock_release(vs
);
1104 dev_put(vxlan
->dev
);
1107 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1108 static void vxlan_igmp_leave(struct work_struct
*work
)
1110 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1111 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1112 struct sock
*sk
= vs
->sock
->sk
;
1113 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1114 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1117 if (ip
->sa
.sa_family
== AF_INET
) {
1118 struct ip_mreqn mreq
= {
1119 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1120 .imr_ifindex
= ifindex
,
1123 ip_mc_leave_group(sk
, &mreq
);
1124 #if IS_ENABLED(CONFIG_IPV6)
1126 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1127 &ip
->sin6
.sin6_addr
);
1133 vxlan_sock_release(vs
);
1134 dev_put(vxlan
->dev
);
1137 /* Callback from net/ipv4/udp.c to receive packets */
1138 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1140 struct vxlan_sock
*vs
;
1141 struct vxlanhdr
*vxh
;
1143 /* Need Vxlan and inner Ethernet header to be present */
1144 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1147 /* Return packets with reserved bits set */
1148 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1149 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
1150 (vxh
->vx_vni
& htonl(0xff))) {
1151 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1152 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1156 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1159 vs
= rcu_dereference_sk_user_data(sk
);
1163 /* If the NIC driver gave us an encapsulated packet
1164 * with the encapsulation mark, the device checksummed it
1165 * for us. Otherwise force the upper layers to verify it.
1167 if ((skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
) ||
1168 !skb
->encapsulation
)
1169 skb
->ip_summed
= CHECKSUM_NONE
;
1171 skb
->encapsulation
= 0;
1173 vs
->rcv(vs
, skb
, vxh
->vx_vni
);
1177 /* Consume bad packet */
1182 /* Return non vxlan pkt */
1186 static void vxlan_rcv(struct vxlan_sock
*vs
,
1187 struct sk_buff
*skb
, __be32 vx_vni
)
1189 struct iphdr
*oip
= NULL
;
1190 struct ipv6hdr
*oip6
= NULL
;
1191 struct vxlan_dev
*vxlan
;
1192 struct pcpu_sw_netstats
*stats
;
1193 union vxlan_addr saddr
;
1196 union vxlan_addr
*remote_ip
;
1198 vni
= ntohl(vx_vni
) >> 8;
1199 /* Is this VNI defined? */
1200 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1204 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1205 skb_reset_mac_header(skb
);
1206 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1208 /* Ignore packet loops (and multicast echo) */
1209 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1212 /* Re-examine inner Ethernet packet */
1213 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1215 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1216 saddr
.sa
.sa_family
= AF_INET
;
1217 #if IS_ENABLED(CONFIG_IPV6)
1219 oip6
= ipv6_hdr(skb
);
1220 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1221 saddr
.sa
.sa_family
= AF_INET6
;
1225 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1226 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1229 skb_reset_network_header(skb
);
1232 err
= IP6_ECN_decapsulate(oip6
, skb
);
1234 err
= IP_ECN_decapsulate(oip
, skb
);
1236 if (unlikely(err
)) {
1237 if (log_ecn_error
) {
1239 net_info_ratelimited("non-ECT from %pI6\n",
1242 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1243 &oip
->saddr
, oip
->tos
);
1246 ++vxlan
->dev
->stats
.rx_frame_errors
;
1247 ++vxlan
->dev
->stats
.rx_errors
;
1252 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1253 u64_stats_update_begin(&stats
->syncp
);
1254 stats
->rx_packets
++;
1255 stats
->rx_bytes
+= skb
->len
;
1256 u64_stats_update_end(&stats
->syncp
);
1262 /* Consume bad packet */
1266 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1268 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1269 struct arphdr
*parp
;
1272 struct neighbour
*n
;
1274 if (dev
->flags
& IFF_NOARP
)
1277 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1278 dev
->stats
.tx_dropped
++;
1281 parp
= arp_hdr(skb
);
1283 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1284 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1285 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1286 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1287 parp
->ar_hln
!= dev
->addr_len
||
1290 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1292 arpptr
+= dev
->addr_len
; /* sha */
1293 memcpy(&sip
, arpptr
, sizeof(sip
));
1294 arpptr
+= sizeof(sip
);
1295 arpptr
+= dev
->addr_len
; /* tha */
1296 memcpy(&tip
, arpptr
, sizeof(tip
));
1298 if (ipv4_is_loopback(tip
) ||
1299 ipv4_is_multicast(tip
))
1302 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1305 struct vxlan_fdb
*f
;
1306 struct sk_buff
*reply
;
1308 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1313 f
= vxlan_find_mac(vxlan
, n
->ha
);
1314 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1315 /* bridge-local neighbor */
1320 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1328 skb_reset_mac_header(reply
);
1329 __skb_pull(reply
, skb_network_offset(reply
));
1330 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1331 reply
->pkt_type
= PACKET_HOST
;
1333 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1334 dev
->stats
.rx_dropped
++;
1335 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1336 union vxlan_addr ipa
= {
1337 .sin
.sin_addr
.s_addr
= tip
,
1338 .sa
.sa_family
= AF_INET
,
1341 vxlan_ip_miss(dev
, &ipa
);
1345 return NETDEV_TX_OK
;
1348 #if IS_ENABLED(CONFIG_IPV6)
1350 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1351 struct neighbour
*n
, bool isrouter
)
1353 struct net_device
*dev
= request
->dev
;
1354 struct sk_buff
*reply
;
1355 struct nd_msg
*ns
, *na
;
1356 struct ipv6hdr
*pip6
;
1358 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1365 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1366 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1367 reply
= alloc_skb(len
, GFP_ATOMIC
);
1371 reply
->protocol
= htons(ETH_P_IPV6
);
1373 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1374 skb_push(reply
, sizeof(struct ethhdr
));
1375 skb_set_mac_header(reply
, 0);
1377 ns
= (struct nd_msg
*)skb_transport_header(request
);
1379 daddr
= eth_hdr(request
)->h_source
;
1380 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1381 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1382 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1383 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1388 /* Ethernet header */
1389 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1390 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1391 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1392 reply
->protocol
= htons(ETH_P_IPV6
);
1394 skb_pull(reply
, sizeof(struct ethhdr
));
1395 skb_set_network_header(reply
, 0);
1396 skb_put(reply
, sizeof(struct ipv6hdr
));
1400 pip6
= ipv6_hdr(reply
);
1401 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1403 pip6
->priority
= ipv6_hdr(request
)->priority
;
1404 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1405 pip6
->hop_limit
= 255;
1406 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1407 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1409 skb_pull(reply
, sizeof(struct ipv6hdr
));
1410 skb_set_transport_header(reply
, 0);
1412 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1414 /* Neighbor Advertisement */
1415 memset(na
, 0, sizeof(*na
)+na_olen
);
1416 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1417 na
->icmph
.icmp6_router
= isrouter
;
1418 na
->icmph
.icmp6_override
= 1;
1419 na
->icmph
.icmp6_solicited
= 1;
1420 na
->target
= ns
->target
;
1421 ether_addr_copy(&na
->opt
[2], n
->ha
);
1422 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1423 na
->opt
[1] = na_olen
>> 3;
1425 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1426 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1427 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1429 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1431 skb_push(reply
, sizeof(struct ipv6hdr
));
1433 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1438 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1440 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1442 const struct ipv6hdr
*iphdr
;
1443 const struct in6_addr
*saddr
, *daddr
;
1444 struct neighbour
*n
;
1445 struct inet6_dev
*in6_dev
;
1447 in6_dev
= __in6_dev_get(dev
);
1451 if (!pskb_may_pull(skb
, skb
->len
))
1454 iphdr
= ipv6_hdr(skb
);
1455 saddr
= &iphdr
->saddr
;
1456 daddr
= &iphdr
->daddr
;
1458 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1459 if (msg
->icmph
.icmp6_code
!= 0 ||
1460 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1463 if (ipv6_addr_loopback(daddr
) ||
1464 ipv6_addr_is_multicast(&msg
->target
))
1467 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1470 struct vxlan_fdb
*f
;
1471 struct sk_buff
*reply
;
1473 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1478 f
= vxlan_find_mac(vxlan
, n
->ha
);
1479 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1480 /* bridge-local neighbor */
1485 reply
= vxlan_na_create(skb
, n
,
1486 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1493 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1494 dev
->stats
.rx_dropped
++;
1496 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1497 union vxlan_addr ipa
= {
1498 .sin6
.sin6_addr
= msg
->target
,
1499 .sa
.sa_family
= AF_INET6
,
1502 vxlan_ip_miss(dev
, &ipa
);
1507 return NETDEV_TX_OK
;
1511 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1513 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1514 struct neighbour
*n
;
1516 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1520 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1525 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1528 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1529 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1530 union vxlan_addr ipa
= {
1531 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1532 .sa
.sa_family
= AF_INET
,
1535 vxlan_ip_miss(dev
, &ipa
);
1541 #if IS_ENABLED(CONFIG_IPV6)
1544 struct ipv6hdr
*pip6
;
1546 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1548 pip6
= ipv6_hdr(skb
);
1549 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1550 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1551 union vxlan_addr ipa
= {
1552 .sin6
.sin6_addr
= pip6
->daddr
,
1553 .sa
.sa_family
= AF_INET6
,
1556 vxlan_ip_miss(dev
, &ipa
);
1570 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1572 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1574 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1583 /* Compute source port for outgoing packet
1584 * first choice to use L4 flow hash since it will spread
1585 * better and maybe available from hardware
1586 * secondary choice is to use jhash on the Ethernet header
1588 __be16
vxlan_src_port(__u16 port_min
, __u16 port_max
, struct sk_buff
*skb
)
1590 unsigned int range
= (port_max
- port_min
) + 1;
1593 hash
= skb_get_hash(skb
);
1595 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
1596 (__force u32
) skb
->protocol
);
1598 return htons((((u64
) hash
* range
) >> 32) + port_min
);
1600 EXPORT_SYMBOL_GPL(vxlan_src_port
);
1602 static int handle_offloads(struct sk_buff
*skb
)
1604 if (skb_is_gso(skb
)) {
1605 int err
= skb_unclone(skb
, GFP_ATOMIC
);
1609 skb_shinfo(skb
)->gso_type
|= SKB_GSO_UDP_TUNNEL
;
1610 } else if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
1611 skb
->ip_summed
= CHECKSUM_NONE
;
1616 #if IS_ENABLED(CONFIG_IPV6)
1617 static int vxlan6_xmit_skb(struct vxlan_sock
*vs
,
1618 struct dst_entry
*dst
, struct sk_buff
*skb
,
1619 struct net_device
*dev
, struct in6_addr
*saddr
,
1620 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1621 __be16 src_port
, __be16 dst_port
, __be32 vni
)
1623 struct ipv6hdr
*ip6h
;
1624 struct vxlanhdr
*vxh
;
1629 if (!skb
->encapsulation
) {
1630 skb_reset_inner_headers(skb
);
1631 skb
->encapsulation
= 1;
1634 skb_scrub_packet(skb
, false);
1636 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1637 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1638 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1640 /* Need space for new headers (invalidates iph ptr) */
1641 err
= skb_cow_head(skb
, min_headroom
);
1645 if (vlan_tx_tag_present(skb
)) {
1646 if (WARN_ON(!__vlan_put_tag(skb
,
1648 vlan_tx_tag_get(skb
))))
1654 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1655 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1658 __skb_push(skb
, sizeof(*uh
));
1659 skb_reset_transport_header(skb
);
1662 uh
->dest
= dst_port
;
1663 uh
->source
= src_port
;
1665 uh
->len
= htons(skb
->len
);
1668 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1669 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1671 skb_dst_set(skb
, dst
);
1673 if (!skb_is_gso(skb
) && !(dst
->dev
->features
& NETIF_F_IPV6_CSUM
)) {
1674 __wsum csum
= skb_checksum(skb
, 0, skb
->len
, 0);
1675 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1676 uh
->check
= csum_ipv6_magic(saddr
, daddr
, skb
->len
,
1679 uh
->check
= CSUM_MANGLED_0
;
1681 skb
->ip_summed
= CHECKSUM_PARTIAL
;
1682 skb
->csum_start
= skb_transport_header(skb
) - skb
->head
;
1683 skb
->csum_offset
= offsetof(struct udphdr
, check
);
1684 uh
->check
= ~csum_ipv6_magic(saddr
, daddr
,
1685 skb
->len
, IPPROTO_UDP
, 0);
1688 __skb_push(skb
, sizeof(*ip6h
));
1689 skb_reset_network_header(skb
);
1690 ip6h
= ipv6_hdr(skb
);
1692 ip6h
->priority
= prio
;
1693 ip6h
->flow_lbl
[0] = 0;
1694 ip6h
->flow_lbl
[1] = 0;
1695 ip6h
->flow_lbl
[2] = 0;
1696 ip6h
->payload_len
= htons(skb
->len
);
1697 ip6h
->nexthdr
= IPPROTO_UDP
;
1698 ip6h
->hop_limit
= ttl
;
1699 ip6h
->daddr
= *daddr
;
1700 ip6h
->saddr
= *saddr
;
1702 err
= handle_offloads(skb
);
1706 ip6tunnel_xmit(skb
, dev
);
1711 int vxlan_xmit_skb(struct vxlan_sock
*vs
,
1712 struct rtable
*rt
, struct sk_buff
*skb
,
1713 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1714 __be16 src_port
, __be16 dst_port
, __be32 vni
)
1716 struct vxlanhdr
*vxh
;
1721 if (!skb
->encapsulation
) {
1722 skb_reset_inner_headers(skb
);
1723 skb
->encapsulation
= 1;
1726 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1727 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1728 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1730 /* Need space for new headers (invalidates iph ptr) */
1731 err
= skb_cow_head(skb
, min_headroom
);
1735 if (vlan_tx_tag_present(skb
)) {
1736 if (WARN_ON(!__vlan_put_tag(skb
,
1738 vlan_tx_tag_get(skb
))))
1744 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1745 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1748 __skb_push(skb
, sizeof(*uh
));
1749 skb_reset_transport_header(skb
);
1752 uh
->dest
= dst_port
;
1753 uh
->source
= src_port
;
1755 uh
->len
= htons(skb
->len
);
1758 err
= handle_offloads(skb
);
1762 return iptunnel_xmit(vs
->sock
->sk
, rt
, skb
, src
, dst
, IPPROTO_UDP
,
1763 tos
, ttl
, df
, false);
1765 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1767 /* Bypass encapsulation if the destination is local */
1768 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1769 struct vxlan_dev
*dst_vxlan
)
1771 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1772 union vxlan_addr loopback
;
1773 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1775 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1776 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1777 skb
->pkt_type
= PACKET_HOST
;
1778 skb
->encapsulation
= 0;
1779 skb
->dev
= dst_vxlan
->dev
;
1780 __skb_pull(skb
, skb_network_offset(skb
));
1782 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1783 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1784 loopback
.sa
.sa_family
= AF_INET
;
1785 #if IS_ENABLED(CONFIG_IPV6)
1787 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1788 loopback
.sa
.sa_family
= AF_INET6
;
1792 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1793 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1795 u64_stats_update_begin(&tx_stats
->syncp
);
1796 tx_stats
->tx_packets
++;
1797 tx_stats
->tx_bytes
+= skb
->len
;
1798 u64_stats_update_end(&tx_stats
->syncp
);
1800 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1801 u64_stats_update_begin(&rx_stats
->syncp
);
1802 rx_stats
->rx_packets
++;
1803 rx_stats
->rx_bytes
+= skb
->len
;
1804 u64_stats_update_end(&rx_stats
->syncp
);
1806 skb
->dev
->stats
.rx_dropped
++;
1810 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1811 struct vxlan_rdst
*rdst
, bool did_rsc
)
1813 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1814 struct rtable
*rt
= NULL
;
1815 const struct iphdr
*old_iph
;
1817 union vxlan_addr
*dst
;
1818 __be16 src_port
= 0, dst_port
;
1824 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1825 vni
= rdst
->remote_vni
;
1826 dst
= &rdst
->remote_ip
;
1828 if (vxlan_addr_any(dst
)) {
1830 /* short-circuited back to local bridge */
1831 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1837 old_iph
= ip_hdr(skb
);
1840 if (!ttl
&& vxlan_addr_multicast(dst
))
1845 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1847 src_port
= vxlan_src_port(vxlan
->port_min
, vxlan
->port_max
, skb
);
1849 if (dst
->sa
.sa_family
== AF_INET
) {
1850 memset(&fl4
, 0, sizeof(fl4
));
1851 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1852 fl4
.flowi4_tos
= RT_TOS(tos
);
1853 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1854 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1856 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
1858 netdev_dbg(dev
, "no route to %pI4\n",
1859 &dst
->sin
.sin_addr
.s_addr
);
1860 dev
->stats
.tx_carrier_errors
++;
1864 if (rt
->dst
.dev
== dev
) {
1865 netdev_dbg(dev
, "circular route to %pI4\n",
1866 &dst
->sin
.sin_addr
.s_addr
);
1867 dev
->stats
.collisions
++;
1871 /* Bypass encapsulation if the destination is local */
1872 if (rt
->rt_flags
& RTCF_LOCAL
&&
1873 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1874 struct vxlan_dev
*dst_vxlan
;
1877 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
, dst_port
);
1880 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1884 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1885 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1887 err
= vxlan_xmit_skb(vxlan
->vn_sock
, rt
, skb
,
1888 fl4
.saddr
, dst
->sin
.sin_addr
.s_addr
,
1889 tos
, ttl
, df
, src_port
, dst_port
,
1894 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1895 #if IS_ENABLED(CONFIG_IPV6)
1897 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1898 struct dst_entry
*ndst
;
1902 memset(&fl6
, 0, sizeof(fl6
));
1903 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
1904 fl6
.daddr
= dst
->sin6
.sin6_addr
;
1905 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
1906 fl6
.flowi6_proto
= IPPROTO_UDP
;
1908 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
1909 netdev_dbg(dev
, "no route to %pI6\n",
1910 &dst
->sin6
.sin6_addr
);
1911 dev
->stats
.tx_carrier_errors
++;
1915 if (ndst
->dev
== dev
) {
1916 netdev_dbg(dev
, "circular route to %pI6\n",
1917 &dst
->sin6
.sin6_addr
);
1919 dev
->stats
.collisions
++;
1923 /* Bypass encapsulation if the destination is local */
1924 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
1925 if (flags
& RTF_LOCAL
&&
1926 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1927 struct vxlan_dev
*dst_vxlan
;
1930 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
, dst_port
);
1933 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1937 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
1939 err
= vxlan6_xmit_skb(vxlan
->vn_sock
, ndst
, skb
,
1940 dev
, &fl6
.saddr
, &fl6
.daddr
, 0, ttl
,
1941 src_port
, dst_port
, htonl(vni
<< 8));
1948 dev
->stats
.tx_dropped
++;
1954 dev
->stats
.tx_errors
++;
1959 /* Transmit local packets over Vxlan
1961 * Outer IP header inherits ECN and DF from inner header.
1962 * Outer UDP destination is the VXLAN assigned port.
1963 * source port is based on hash of flow
1965 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1967 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1969 bool did_rsc
= false;
1970 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
1971 struct vxlan_fdb
*f
;
1973 skb_reset_mac_header(skb
);
1976 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
1977 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
1978 return arp_reduce(dev
, skb
);
1979 #if IS_ENABLED(CONFIG_IPV6)
1980 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
1981 skb
->len
>= sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
) &&
1982 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
1985 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1986 if (msg
->icmph
.icmp6_code
== 0 &&
1987 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
1988 return neigh_reduce(dev
, skb
);
1993 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1996 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1997 (ntohs(eth
->h_proto
) == ETH_P_IP
||
1998 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
1999 did_rsc
= route_shortcircuit(dev
, skb
);
2001 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2005 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2007 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2008 !is_multicast_ether_addr(eth
->h_dest
))
2009 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2011 dev
->stats
.tx_dropped
++;
2013 return NETDEV_TX_OK
;
2017 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2018 struct sk_buff
*skb1
;
2024 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2026 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2030 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2033 return NETDEV_TX_OK
;
2036 /* Walk the forwarding table and purge stale entries */
2037 static void vxlan_cleanup(unsigned long arg
)
2039 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2040 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2043 if (!netif_running(vxlan
->dev
))
2046 spin_lock_bh(&vxlan
->hash_lock
);
2047 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2048 struct hlist_node
*p
, *n
;
2049 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2051 = container_of(p
, struct vxlan_fdb
, hlist
);
2052 unsigned long timeout
;
2054 if (f
->state
& NUD_PERMANENT
)
2057 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2058 if (time_before_eq(timeout
, jiffies
)) {
2059 netdev_dbg(vxlan
->dev
,
2060 "garbage collect %pM\n",
2062 f
->state
= NUD_STALE
;
2063 vxlan_fdb_destroy(vxlan
, f
);
2064 } else if (time_before(timeout
, next_timer
))
2065 next_timer
= timeout
;
2068 spin_unlock_bh(&vxlan
->hash_lock
);
2070 mod_timer(&vxlan
->age_timer
, next_timer
);
2073 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2075 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2077 vxlan
->vn_sock
= vs
;
2078 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2081 /* Setup stats when device is created */
2082 static int vxlan_init(struct net_device
*dev
)
2084 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2085 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2086 struct vxlan_sock
*vs
;
2088 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2092 spin_lock(&vn
->sock_lock
);
2093 vs
= vxlan_find_sock(dev_net(dev
), vxlan
->dst_port
);
2095 /* If we have a socket with same port already, reuse it */
2096 atomic_inc(&vs
->refcnt
);
2097 vxlan_vs_add_dev(vs
, vxlan
);
2099 /* otherwise make new socket outside of RTNL */
2101 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2103 spin_unlock(&vn
->sock_lock
);
2108 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2110 struct vxlan_fdb
*f
;
2112 spin_lock_bh(&vxlan
->hash_lock
);
2113 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2115 vxlan_fdb_destroy(vxlan
, f
);
2116 spin_unlock_bh(&vxlan
->hash_lock
);
2119 static void vxlan_uninit(struct net_device
*dev
)
2121 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2122 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2124 vxlan_fdb_delete_default(vxlan
);
2127 vxlan_sock_release(vs
);
2128 free_percpu(dev
->tstats
);
2131 /* Start ageing timer and join group when device is brought up */
2132 static int vxlan_open(struct net_device
*dev
)
2134 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2135 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2137 /* socket hasn't been created */
2141 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2142 vxlan_sock_hold(vs
);
2144 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2147 if (vxlan
->age_interval
)
2148 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2153 /* Purge the forwarding table */
2154 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2158 spin_lock_bh(&vxlan
->hash_lock
);
2159 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2160 struct hlist_node
*p
, *n
;
2161 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2163 = container_of(p
, struct vxlan_fdb
, hlist
);
2164 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2165 if (!is_zero_ether_addr(f
->eth_addr
))
2166 vxlan_fdb_destroy(vxlan
, f
);
2169 spin_unlock_bh(&vxlan
->hash_lock
);
2172 /* Cleanup timer and forwarding table on shutdown */
2173 static int vxlan_stop(struct net_device
*dev
)
2175 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2176 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2177 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2179 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2180 !vxlan_group_used(vn
, vxlan
)) {
2181 vxlan_sock_hold(vs
);
2183 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2186 del_timer_sync(&vxlan
->age_timer
);
2193 /* Stub, nothing needs to be done. */
2194 static void vxlan_set_multicast_list(struct net_device
*dev
)
2198 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2200 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2201 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2202 struct net_device
*lowerdev
;
2205 lowerdev
= __dev_get_by_index(dev_net(dev
), dst
->remote_ifindex
);
2206 if (lowerdev
== NULL
)
2207 return eth_change_mtu(dev
, new_mtu
);
2209 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2210 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2212 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2214 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2221 static const struct net_device_ops vxlan_netdev_ops
= {
2222 .ndo_init
= vxlan_init
,
2223 .ndo_uninit
= vxlan_uninit
,
2224 .ndo_open
= vxlan_open
,
2225 .ndo_stop
= vxlan_stop
,
2226 .ndo_start_xmit
= vxlan_xmit
,
2227 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2228 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2229 .ndo_change_mtu
= vxlan_change_mtu
,
2230 .ndo_validate_addr
= eth_validate_addr
,
2231 .ndo_set_mac_address
= eth_mac_addr
,
2232 .ndo_fdb_add
= vxlan_fdb_add
,
2233 .ndo_fdb_del
= vxlan_fdb_delete
,
2234 .ndo_fdb_dump
= vxlan_fdb_dump
,
2237 /* Info for udev, that this is a virtual tunnel endpoint */
2238 static struct device_type vxlan_type
= {
2242 /* Calls the ndo_add_vxlan_port of the caller in order to
2243 * supply the listening VXLAN udp ports. Callers are expected
2244 * to implement the ndo_add_vxlan_port.
2246 void vxlan_get_rx_port(struct net_device
*dev
)
2248 struct vxlan_sock
*vs
;
2249 struct net
*net
= dev_net(dev
);
2250 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2251 sa_family_t sa_family
;
2255 spin_lock(&vn
->sock_lock
);
2256 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2257 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2258 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2259 sa_family
= vs
->sock
->sk
->sk_family
;
2260 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2264 spin_unlock(&vn
->sock_lock
);
2266 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2268 /* Initialize the device structure. */
2269 static void vxlan_setup(struct net_device
*dev
)
2271 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2275 eth_hw_addr_random(dev
);
2277 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2278 dev
->hard_header_len
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2280 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
2282 dev
->netdev_ops
= &vxlan_netdev_ops
;
2283 dev
->destructor
= free_netdev
;
2284 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2286 dev
->tx_queue_len
= 0;
2287 dev
->features
|= NETIF_F_LLTX
;
2288 dev
->features
|= NETIF_F_NETNS_LOCAL
;
2289 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2290 dev
->features
|= NETIF_F_RXCSUM
;
2291 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2293 dev
->vlan_features
= dev
->features
;
2294 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2295 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2296 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2297 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2298 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
2299 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2301 INIT_LIST_HEAD(&vxlan
->next
);
2302 spin_lock_init(&vxlan
->hash_lock
);
2303 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2304 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2305 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2307 init_timer_deferrable(&vxlan
->age_timer
);
2308 vxlan
->age_timer
.function
= vxlan_cleanup
;
2309 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2311 inet_get_local_port_range(dev_net(dev
), &low
, &high
);
2312 vxlan
->port_min
= low
;
2313 vxlan
->port_max
= high
;
2314 vxlan
->dst_port
= htons(vxlan_port
);
2318 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2319 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2322 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2323 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2324 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2325 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2326 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2327 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2328 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2329 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2330 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2331 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2332 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2333 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2334 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2335 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2336 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2337 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2338 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2339 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2342 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2344 if (tb
[IFLA_ADDRESS
]) {
2345 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2346 pr_debug("invalid link address (not ethernet)\n");
2350 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2351 pr_debug("invalid all zero ethernet address\n");
2352 return -EADDRNOTAVAIL
;
2359 if (data
[IFLA_VXLAN_ID
]) {
2360 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2361 if (id
>= VXLAN_VID_MASK
)
2365 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2366 const struct ifla_vxlan_port_range
*p
2367 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2369 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2370 pr_debug("port range %u .. %u not valid\n",
2371 ntohs(p
->low
), ntohs(p
->high
));
2379 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2380 struct ethtool_drvinfo
*drvinfo
)
2382 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2383 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2386 static const struct ethtool_ops vxlan_ethtool_ops
= {
2387 .get_drvinfo
= vxlan_get_drvinfo
,
2388 .get_link
= ethtool_op_get_link
,
2391 static void vxlan_del_work(struct work_struct
*work
)
2393 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2395 sk_release_kernel(vs
->sock
->sk
);
2399 #if IS_ENABLED(CONFIG_IPV6)
2400 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2401 * could be used for both IPv4 and IPv6 communications, but
2402 * users may set bindv6only=1.
2404 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
)
2407 struct socket
*sock
;
2408 struct sockaddr_in6 vxlan_addr
= {
2409 .sin6_family
= AF_INET6
,
2414 rc
= sock_create_kern(AF_INET6
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2416 pr_debug("UDPv6 socket create failed\n");
2420 /* Put in proper namespace */
2422 sk_change_net(sk
, net
);
2424 kernel_setsockopt(sock
, SOL_IPV6
, IPV6_V6ONLY
,
2425 (char *)&val
, sizeof(val
));
2426 rc
= kernel_bind(sock
, (struct sockaddr
*)&vxlan_addr
,
2427 sizeof(struct sockaddr_in6
));
2429 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2430 &vxlan_addr
.sin6_addr
, ntohs(vxlan_addr
.sin6_port
), rc
);
2431 sk_release_kernel(sk
);
2434 /* At this point, IPv6 module should have been loaded in
2435 * sock_create_kern().
2439 /* Disable multicast loopback */
2440 inet_sk(sk
)->mc_loop
= 0;
2446 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
)
2448 return ERR_PTR(-EPFNOSUPPORT
);
2452 static struct socket
*create_v4_sock(struct net
*net
, __be16 port
)
2455 struct socket
*sock
;
2456 struct sockaddr_in vxlan_addr
= {
2457 .sin_family
= AF_INET
,
2458 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
2463 /* Create UDP socket for encapsulation receive. */
2464 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2466 pr_debug("UDP socket create failed\n");
2470 /* Put in proper namespace */
2472 sk_change_net(sk
, net
);
2474 rc
= kernel_bind(sock
, (struct sockaddr
*) &vxlan_addr
,
2475 sizeof(vxlan_addr
));
2477 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2478 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
2479 sk_release_kernel(sk
);
2483 /* Disable multicast loopback */
2484 inet_sk(sk
)->mc_loop
= 0;
2488 /* Create new listen socket if needed */
2489 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2490 vxlan_rcv_t
*rcv
, void *data
, bool ipv6
)
2492 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2493 struct vxlan_sock
*vs
;
2494 struct socket
*sock
;
2498 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2500 return ERR_PTR(-ENOMEM
);
2502 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2503 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2505 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2508 sock
= create_v6_sock(net
, port
);
2510 sock
= create_v4_sock(net
, port
);
2513 return ERR_CAST(sock
);
2518 atomic_set(&vs
->refcnt
, 1);
2521 rcu_assign_sk_user_data(vs
->sock
->sk
, vs
);
2523 /* Initialize the vxlan udp offloads structure */
2524 vs
->udp_offloads
.port
= port
;
2525 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2526 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2528 spin_lock(&vn
->sock_lock
);
2529 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2530 vxlan_notify_add_rx_port(vs
);
2531 spin_unlock(&vn
->sock_lock
);
2533 /* Mark socket as an encapsulation socket. */
2534 udp_sk(sk
)->encap_type
= 1;
2535 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
2536 #if IS_ENABLED(CONFIG_IPV6)
2538 ipv6_stub
->udpv6_encap_enable();
2546 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2547 vxlan_rcv_t
*rcv
, void *data
,
2548 bool no_share
, bool ipv6
)
2550 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2551 struct vxlan_sock
*vs
;
2553 vs
= vxlan_socket_create(net
, port
, rcv
, data
, ipv6
);
2557 if (no_share
) /* Return error if sharing is not allowed. */
2560 spin_lock(&vn
->sock_lock
);
2561 vs
= vxlan_find_sock(net
, port
);
2564 atomic_inc(&vs
->refcnt
);
2566 vs
= ERR_PTR(-EBUSY
);
2568 spin_unlock(&vn
->sock_lock
);
2571 vs
= ERR_PTR(-EINVAL
);
2575 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2577 /* Scheduled at device creation to bind to a socket */
2578 static void vxlan_sock_work(struct work_struct
*work
)
2580 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2581 struct net
*net
= dev_net(vxlan
->dev
);
2582 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2583 __be16 port
= vxlan
->dst_port
;
2584 struct vxlan_sock
*nvs
;
2586 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
& VXLAN_F_IPV6
);
2587 spin_lock(&vn
->sock_lock
);
2589 vxlan_vs_add_dev(nvs
, vxlan
);
2590 spin_unlock(&vn
->sock_lock
);
2592 dev_put(vxlan
->dev
);
2595 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2596 struct nlattr
*tb
[], struct nlattr
*data
[])
2598 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2599 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2600 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2603 bool use_ipv6
= false;
2605 if (!data
[IFLA_VXLAN_ID
])
2608 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2609 dst
->remote_vni
= vni
;
2611 /* Unless IPv6 is explicitly requested, assume IPv4 */
2612 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2613 if (data
[IFLA_VXLAN_GROUP
]) {
2614 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2615 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2616 if (!IS_ENABLED(CONFIG_IPV6
))
2617 return -EPFNOSUPPORT
;
2619 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2620 sizeof(struct in6_addr
));
2621 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2625 if (data
[IFLA_VXLAN_LOCAL
]) {
2626 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2627 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2628 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2629 if (!IS_ENABLED(CONFIG_IPV6
))
2630 return -EPFNOSUPPORT
;
2632 /* TODO: respect scope id */
2633 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2634 sizeof(struct in6_addr
));
2635 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2639 if (data
[IFLA_VXLAN_LINK
] &&
2640 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2641 struct net_device
*lowerdev
2642 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2645 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2649 #if IS_ENABLED(CONFIG_IPV6)
2651 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2652 if (idev
&& idev
->cnf
.disable_ipv6
) {
2653 pr_info("IPv6 is disabled via sysctl\n");
2656 vxlan
->flags
|= VXLAN_F_IPV6
;
2661 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2663 /* update header length based on lower device */
2664 dev
->hard_header_len
= lowerdev
->hard_header_len
+
2665 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2666 } else if (use_ipv6
)
2667 vxlan
->flags
|= VXLAN_F_IPV6
;
2669 if (data
[IFLA_VXLAN_TOS
])
2670 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2672 if (data
[IFLA_VXLAN_TTL
])
2673 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2675 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2676 vxlan
->flags
|= VXLAN_F_LEARN
;
2678 if (data
[IFLA_VXLAN_AGEING
])
2679 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2681 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2683 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2684 vxlan
->flags
|= VXLAN_F_PROXY
;
2686 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2687 vxlan
->flags
|= VXLAN_F_RSC
;
2689 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2690 vxlan
->flags
|= VXLAN_F_L2MISS
;
2692 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2693 vxlan
->flags
|= VXLAN_F_L3MISS
;
2695 if (data
[IFLA_VXLAN_LIMIT
])
2696 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2698 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2699 const struct ifla_vxlan_port_range
*p
2700 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2701 vxlan
->port_min
= ntohs(p
->low
);
2702 vxlan
->port_max
= ntohs(p
->high
);
2705 if (data
[IFLA_VXLAN_PORT
])
2706 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2708 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
2709 pr_info("duplicate VNI %u\n", vni
);
2713 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
2715 /* create an fdb entry for a valid default destination */
2716 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2717 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2718 &vxlan
->default_dst
.remote_ip
,
2719 NUD_REACHABLE
|NUD_PERMANENT
,
2720 NLM_F_EXCL
|NLM_F_CREATE
,
2722 vxlan
->default_dst
.remote_vni
,
2723 vxlan
->default_dst
.remote_ifindex
,
2729 err
= register_netdevice(dev
);
2731 vxlan_fdb_delete_default(vxlan
);
2735 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2740 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2742 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2743 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2745 spin_lock(&vn
->sock_lock
);
2746 if (!hlist_unhashed(&vxlan
->hlist
))
2747 hlist_del_rcu(&vxlan
->hlist
);
2748 spin_unlock(&vn
->sock_lock
);
2750 list_del(&vxlan
->next
);
2751 unregister_netdevice_queue(dev
, head
);
2754 static size_t vxlan_get_size(const struct net_device
*dev
)
2757 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2758 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2759 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2760 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2761 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2762 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2763 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2764 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2765 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2766 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2767 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2768 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2769 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2770 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2771 nla_total_size(sizeof(__be16
))+ /* IFLA_VXLAN_PORT */
2775 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2777 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2778 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2779 struct ifla_vxlan_port_range ports
= {
2780 .low
= htons(vxlan
->port_min
),
2781 .high
= htons(vxlan
->port_max
),
2784 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2785 goto nla_put_failure
;
2787 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2788 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2789 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2790 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2791 goto nla_put_failure
;
2792 #if IS_ENABLED(CONFIG_IPV6)
2794 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2795 &dst
->remote_ip
.sin6
.sin6_addr
))
2796 goto nla_put_failure
;
2801 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2802 goto nla_put_failure
;
2804 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2805 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2806 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2807 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2808 goto nla_put_failure
;
2809 #if IS_ENABLED(CONFIG_IPV6)
2811 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2812 &vxlan
->saddr
.sin6
.sin6_addr
))
2813 goto nla_put_failure
;
2818 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2819 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2820 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2821 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2822 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2823 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2824 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2825 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2826 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2827 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2828 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2829 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2830 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2831 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
))
2832 goto nla_put_failure
;
2834 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2835 goto nla_put_failure
;
2843 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2845 .maxtype
= IFLA_VXLAN_MAX
,
2846 .policy
= vxlan_policy
,
2847 .priv_size
= sizeof(struct vxlan_dev
),
2848 .setup
= vxlan_setup
,
2849 .validate
= vxlan_validate
,
2850 .newlink
= vxlan_newlink
,
2851 .dellink
= vxlan_dellink
,
2852 .get_size
= vxlan_get_size
,
2853 .fill_info
= vxlan_fill_info
,
2856 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2857 struct net_device
*dev
)
2859 struct vxlan_dev
*vxlan
, *next
;
2860 LIST_HEAD(list_kill
);
2862 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2863 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2865 /* In case we created vxlan device with carrier
2866 * and we loose the carrier due to module unload
2867 * we also need to remove vxlan device. In other
2868 * cases, it's not necessary and remote_ifindex
2869 * is 0 here, so no matches.
2871 if (dst
->remote_ifindex
== dev
->ifindex
)
2872 vxlan_dellink(vxlan
->dev
, &list_kill
);
2875 unregister_netdevice_many(&list_kill
);
2878 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2879 unsigned long event
, void *ptr
)
2881 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2882 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2884 if (event
== NETDEV_UNREGISTER
)
2885 vxlan_handle_lowerdev_unregister(vn
, dev
);
2890 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2891 .notifier_call
= vxlan_lowerdev_event
,
2894 static __net_init
int vxlan_init_net(struct net
*net
)
2896 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2899 INIT_LIST_HEAD(&vn
->vxlan_list
);
2900 spin_lock_init(&vn
->sock_lock
);
2902 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2903 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
2908 static struct pernet_operations vxlan_net_ops
= {
2909 .init
= vxlan_init_net
,
2910 .id
= &vxlan_net_id
,
2911 .size
= sizeof(struct vxlan_net
),
2914 static int __init
vxlan_init_module(void)
2918 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
2922 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
2924 rc
= register_pernet_subsys(&vxlan_net_ops
);
2928 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
2932 rc
= rtnl_link_register(&vxlan_link_ops
);
2938 unregister_netdevice_notifier(&vxlan_notifier_block
);
2940 unregister_pernet_subsys(&vxlan_net_ops
);
2942 destroy_workqueue(vxlan_wq
);
2945 late_initcall(vxlan_init_module
);
2947 static void __exit
vxlan_cleanup_module(void)
2949 rtnl_link_unregister(&vxlan_link_ops
);
2950 unregister_netdevice_notifier(&vxlan_notifier_block
);
2951 destroy_workqueue(vxlan_wq
);
2952 unregister_pernet_subsys(&vxlan_net_ops
);
2953 /* rcu_barrier() is called by netns */
2955 module_exit(vxlan_cleanup_module
);
2957 MODULE_LICENSE("GPL");
2958 MODULE_VERSION(VXLAN_VERSION
);
2959 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2960 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2961 MODULE_ALIAS_RTNL_LINK("vxlan");