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 net
*net
; /* netns for packet i/o */
131 struct vxlan_rdst default_dst
; /* default destination */
132 union vxlan_addr saddr
; /* source address */
134 __u16 port_min
; /* source port range */
136 __u8 tos
; /* TOS override */
138 u32 flags
; /* VXLAN_F_* in vxlan.h */
140 struct work_struct sock_work
;
141 struct work_struct igmp_join
;
142 struct work_struct igmp_leave
;
144 unsigned long age_interval
;
145 struct timer_list age_timer
;
146 spinlock_t hash_lock
;
147 unsigned int addrcnt
;
148 unsigned int addrmax
;
150 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
153 /* salt for hash table */
154 static u32 vxlan_salt __read_mostly
;
155 static struct workqueue_struct
*vxlan_wq
;
157 static void vxlan_sock_work(struct work_struct
*work
);
159 #if IS_ENABLED(CONFIG_IPV6)
161 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
163 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
165 if (a
->sa
.sa_family
== AF_INET6
)
166 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
168 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
171 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
173 if (ipa
->sa
.sa_family
== AF_INET6
)
174 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
176 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
179 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
181 if (ipa
->sa
.sa_family
== AF_INET6
)
182 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
184 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
187 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
189 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
190 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
191 ip
->sa
.sa_family
= AF_INET6
;
193 } else if (nla_len(nla
) >= sizeof(__be32
)) {
194 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
195 ip
->sa
.sa_family
= AF_INET
;
198 return -EAFNOSUPPORT
;
202 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
203 const union vxlan_addr
*ip
)
205 if (ip
->sa
.sa_family
== AF_INET6
)
206 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
208 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
211 #else /* !CONFIG_IPV6 */
214 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
216 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
219 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
221 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
224 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
226 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
229 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
231 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
232 return -EAFNOSUPPORT
;
233 } else if (nla_len(nla
) >= sizeof(__be32
)) {
234 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
235 ip
->sa
.sa_family
= AF_INET
;
238 return -EAFNOSUPPORT
;
242 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
243 const union vxlan_addr
*ip
)
245 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
249 /* Virtual Network hash table head */
250 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
252 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
255 /* Socket hash table head */
256 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
258 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
260 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
263 /* First remote destination for a forwarding entry.
264 * Guaranteed to be non-NULL because remotes are never deleted.
266 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
268 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
271 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
273 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
276 /* Find VXLAN socket based on network namespace and UDP port */
277 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, __be16 port
)
279 struct vxlan_sock
*vs
;
281 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
282 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
288 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
290 struct vxlan_dev
*vxlan
;
292 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
293 if (vxlan
->default_dst
.remote_vni
== id
)
300 /* Look up VNI in a per net namespace table */
301 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
303 struct vxlan_sock
*vs
;
305 vs
= vxlan_find_sock(net
, port
);
309 return vxlan_vs_find_vni(vs
, id
);
312 /* Fill in neighbour message in skbuff. */
313 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
314 const struct vxlan_fdb
*fdb
,
315 u32 portid
, u32 seq
, int type
, unsigned int flags
,
316 const struct vxlan_rdst
*rdst
)
318 unsigned long now
= jiffies
;
319 struct nda_cacheinfo ci
;
320 struct nlmsghdr
*nlh
;
322 bool send_ip
, send_eth
;
324 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
328 ndm
= nlmsg_data(nlh
);
329 memset(ndm
, 0, sizeof(*ndm
));
331 send_eth
= send_ip
= true;
333 if (type
== RTM_GETNEIGH
) {
334 ndm
->ndm_family
= AF_INET
;
335 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
336 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
338 ndm
->ndm_family
= AF_BRIDGE
;
339 ndm
->ndm_state
= fdb
->state
;
340 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
341 ndm
->ndm_flags
= fdb
->flags
;
342 ndm
->ndm_type
= NDA_DST
;
344 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
345 goto nla_put_failure
;
347 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
348 goto nla_put_failure
;
350 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
351 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
352 goto nla_put_failure
;
353 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
354 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
355 goto nla_put_failure
;
356 if (rdst
->remote_ifindex
&&
357 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
358 goto nla_put_failure
;
360 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
361 ci
.ndm_confirmed
= 0;
362 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
365 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
366 goto nla_put_failure
;
368 return nlmsg_end(skb
, nlh
);
371 nlmsg_cancel(skb
, nlh
);
375 static inline size_t vxlan_nlmsg_size(void)
377 return NLMSG_ALIGN(sizeof(struct ndmsg
))
378 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
379 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
380 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
381 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
382 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
383 + nla_total_size(sizeof(struct nda_cacheinfo
));
386 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
387 struct vxlan_rdst
*rd
, int type
)
389 struct net
*net
= dev_net(vxlan
->dev
);
393 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
397 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
399 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
400 WARN_ON(err
== -EMSGSIZE
);
405 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
409 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
412 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
414 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
415 struct vxlan_fdb f
= {
418 struct vxlan_rdst remote
= {
419 .remote_ip
= *ipa
, /* goes to NDA_DST */
420 .remote_vni
= VXLAN_N_VID
,
423 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
426 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
428 struct vxlan_fdb f
= {
431 struct vxlan_rdst remote
= { };
433 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
435 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
438 /* Hash Ethernet address */
439 static u32
eth_hash(const unsigned char *addr
)
441 u64 value
= get_unaligned((u64
*)addr
);
443 /* only want 6 bytes */
449 return hash_64(value
, FDB_HASH_BITS
);
452 /* Hash chain to use given mac address */
453 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
456 return &vxlan
->fdb_head
[eth_hash(mac
)];
459 /* Look up Ethernet address in forwarding table */
460 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
463 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
466 hlist_for_each_entry_rcu(f
, head
, hlist
) {
467 if (ether_addr_equal(mac
, f
->eth_addr
))
474 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
479 f
= __vxlan_find_mac(vxlan
, mac
);
486 /* caller should hold vxlan->hash_lock */
487 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
488 union vxlan_addr
*ip
, __be16 port
,
489 __u32 vni
, __u32 ifindex
)
491 struct vxlan_rdst
*rd
;
493 list_for_each_entry(rd
, &f
->remotes
, list
) {
494 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
495 rd
->remote_port
== port
&&
496 rd
->remote_vni
== vni
&&
497 rd
->remote_ifindex
== ifindex
)
504 /* Replace destination of unicast mac */
505 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
506 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
508 struct vxlan_rdst
*rd
;
510 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
514 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
518 rd
->remote_port
= port
;
519 rd
->remote_vni
= vni
;
520 rd
->remote_ifindex
= ifindex
;
524 /* Add/update destinations for multicast */
525 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
526 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
527 __u32 ifindex
, struct vxlan_rdst
**rdp
)
529 struct vxlan_rdst
*rd
;
531 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
535 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
539 rd
->remote_port
= port
;
540 rd
->remote_vni
= vni
;
541 rd
->remote_ifindex
= ifindex
;
543 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
549 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
)
551 struct sk_buff
*p
, **pp
= NULL
;
552 struct vxlanhdr
*vh
, *vh2
;
553 struct ethhdr
*eh
, *eh2
;
554 unsigned int hlen
, off_vx
, off_eth
;
555 const struct packet_offload
*ptype
;
559 off_vx
= skb_gro_offset(skb
);
560 hlen
= off_vx
+ sizeof(*vh
);
561 vh
= skb_gro_header_fast(skb
, off_vx
);
562 if (skb_gro_header_hard(skb
, hlen
)) {
563 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
567 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
568 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
570 off_eth
= skb_gro_offset(skb
);
571 hlen
= off_eth
+ sizeof(*eh
);
572 eh
= skb_gro_header_fast(skb
, off_eth
);
573 if (skb_gro_header_hard(skb
, hlen
)) {
574 eh
= skb_gro_header_slow(skb
, hlen
, off_eth
);
581 for (p
= *head
; p
; p
= p
->next
) {
582 if (!NAPI_GRO_CB(p
)->same_flow
)
585 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
586 eh2
= (struct ethhdr
*)(p
->data
+ off_eth
);
587 if (vh
->vx_vni
!= vh2
->vx_vni
|| compare_ether_header(eh
, eh2
)) {
588 NAPI_GRO_CB(p
)->same_flow
= 0;
596 ptype
= gro_find_receive_by_type(type
);
602 skb_gro_pull(skb
, sizeof(*eh
)); /* pull inner eth header */
603 skb_gro_postpull_rcsum(skb
, eh
, sizeof(*eh
));
604 pp
= ptype
->callbacks
.gro_receive(head
, skb
);
609 NAPI_GRO_CB(skb
)->flush
|= flush
;
614 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
)
617 struct packet_offload
*ptype
;
619 int vxlan_len
= sizeof(struct vxlanhdr
) + sizeof(struct ethhdr
);
622 eh
= (struct ethhdr
*)(skb
->data
+ nhoff
+ sizeof(struct vxlanhdr
));
626 ptype
= gro_find_complete_by_type(type
);
628 err
= ptype
->callbacks
.gro_complete(skb
, nhoff
+ vxlan_len
);
634 /* Notify netdevs that UDP port started listening */
635 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
637 struct net_device
*dev
;
638 struct sock
*sk
= vs
->sock
->sk
;
639 struct net
*net
= sock_net(sk
);
640 sa_family_t sa_family
= sk
->sk_family
;
641 __be16 port
= inet_sk(sk
)->inet_sport
;
644 if (sa_family
== AF_INET
) {
645 err
= udp_add_offload(&vs
->udp_offloads
);
647 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
651 for_each_netdev_rcu(net
, dev
) {
652 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
653 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
659 /* Notify netdevs that UDP port is no more listening */
660 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
662 struct net_device
*dev
;
663 struct sock
*sk
= vs
->sock
->sk
;
664 struct net
*net
= sock_net(sk
);
665 sa_family_t sa_family
= sk
->sk_family
;
666 __be16 port
= inet_sk(sk
)->inet_sport
;
669 for_each_netdev_rcu(net
, dev
) {
670 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
671 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
676 if (sa_family
== AF_INET
)
677 udp_del_offload(&vs
->udp_offloads
);
680 /* Add new entry to forwarding table -- assumes lock held */
681 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
682 const u8
*mac
, union vxlan_addr
*ip
,
683 __u16 state
, __u16 flags
,
684 __be16 port
, __u32 vni
, __u32 ifindex
,
687 struct vxlan_rdst
*rd
= NULL
;
691 f
= __vxlan_find_mac(vxlan
, mac
);
693 if (flags
& NLM_F_EXCL
) {
694 netdev_dbg(vxlan
->dev
,
695 "lost race to create %pM\n", mac
);
698 if (f
->state
!= state
) {
700 f
->updated
= jiffies
;
703 if (f
->flags
!= ndm_flags
) {
704 f
->flags
= ndm_flags
;
705 f
->updated
= jiffies
;
708 if ((flags
& NLM_F_REPLACE
)) {
709 /* Only change unicasts */
710 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
711 is_zero_ether_addr(f
->eth_addr
))) {
712 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
721 if ((flags
& NLM_F_APPEND
) &&
722 (is_multicast_ether_addr(f
->eth_addr
) ||
723 is_zero_ether_addr(f
->eth_addr
))) {
724 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
732 if (!(flags
& NLM_F_CREATE
))
735 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
738 /* Disallow replace to add a multicast entry */
739 if ((flags
& NLM_F_REPLACE
) &&
740 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
743 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
744 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
750 f
->flags
= ndm_flags
;
751 f
->updated
= f
->used
= jiffies
;
752 INIT_LIST_HEAD(&f
->remotes
);
753 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
755 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
758 hlist_add_head_rcu(&f
->hlist
,
759 vxlan_fdb_head(vxlan
, mac
));
764 rd
= first_remote_rtnl(f
);
765 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
771 static void vxlan_fdb_free(struct rcu_head
*head
)
773 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
774 struct vxlan_rdst
*rd
, *nd
;
776 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
781 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
783 netdev_dbg(vxlan
->dev
,
784 "delete %pM\n", f
->eth_addr
);
787 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
789 hlist_del_rcu(&f
->hlist
);
790 call_rcu(&f
->rcu
, vxlan_fdb_free
);
793 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
794 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
796 struct net
*net
= dev_net(vxlan
->dev
);
800 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
804 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
805 if (remote
->sa
.sa_family
== AF_INET
) {
806 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
807 ip
->sa
.sa_family
= AF_INET
;
808 #if IS_ENABLED(CONFIG_IPV6)
810 ip
->sin6
.sin6_addr
= in6addr_any
;
811 ip
->sa
.sa_family
= AF_INET6
;
817 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
819 *port
= nla_get_be16(tb
[NDA_PORT
]);
821 *port
= vxlan
->dst_port
;
825 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
827 *vni
= nla_get_u32(tb
[NDA_VNI
]);
829 *vni
= vxlan
->default_dst
.remote_vni
;
832 if (tb
[NDA_IFINDEX
]) {
833 struct net_device
*tdev
;
835 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
837 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
838 tdev
= __dev_get_by_index(net
, *ifindex
);
840 return -EADDRNOTAVAIL
;
848 /* Add static entry (via netlink) */
849 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
850 struct net_device
*dev
,
851 const unsigned char *addr
, u16 flags
)
853 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
854 /* struct net *net = dev_net(vxlan->dev); */
860 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
861 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
866 if (tb
[NDA_DST
] == NULL
)
869 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
873 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
874 return -EAFNOSUPPORT
;
876 spin_lock_bh(&vxlan
->hash_lock
);
877 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
878 port
, vni
, ifindex
, ndm
->ndm_flags
);
879 spin_unlock_bh(&vxlan
->hash_lock
);
884 /* Delete entry (via netlink) */
885 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
886 struct net_device
*dev
,
887 const unsigned char *addr
)
889 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
891 struct vxlan_rdst
*rd
= NULL
;
897 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
903 spin_lock_bh(&vxlan
->hash_lock
);
904 f
= vxlan_find_mac(vxlan
, addr
);
908 if (!vxlan_addr_any(&ip
)) {
909 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
916 /* remove a destination if it's not the only one on the list,
917 * otherwise destroy the fdb entry
919 if (rd
&& !list_is_singular(&f
->remotes
)) {
920 list_del_rcu(&rd
->list
);
921 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
926 vxlan_fdb_destroy(vxlan
, f
);
929 spin_unlock_bh(&vxlan
->hash_lock
);
934 /* Dump forwarding table */
935 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
936 struct net_device
*dev
, int idx
)
938 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
941 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
945 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
946 struct vxlan_rdst
*rd
;
948 if (idx
< cb
->args
[0])
951 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
952 err
= vxlan_fdb_info(skb
, vxlan
, f
,
953 NETLINK_CB(cb
->skb
).portid
,
968 /* Watch incoming packets to learn mapping between Ethernet address
969 * and Tunnel endpoint.
970 * Return true if packet is bogus and should be droppped.
972 static bool vxlan_snoop(struct net_device
*dev
,
973 union vxlan_addr
*src_ip
, const u8
*src_mac
)
975 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
978 f
= vxlan_find_mac(vxlan
, src_mac
);
980 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
982 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
985 /* Don't migrate static entries, drop packets */
986 if (f
->state
& NUD_NOARP
)
991 "%pM migrated from %pIS to %pIS\n",
992 src_mac
, &rdst
->remote_ip
, &src_ip
);
994 rdst
->remote_ip
= *src_ip
;
995 f
->updated
= jiffies
;
996 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
998 /* learned new entry */
999 spin_lock(&vxlan
->hash_lock
);
1001 /* close off race between vxlan_flush and incoming packets */
1002 if (netif_running(dev
))
1003 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1005 NLM_F_EXCL
|NLM_F_CREATE
,
1007 vxlan
->default_dst
.remote_vni
,
1009 spin_unlock(&vxlan
->hash_lock
);
1015 /* See if multicast group is already in use by other ID */
1016 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1018 struct vxlan_dev
*vxlan
;
1020 /* The vxlan_sock is only used by dev, leaving group has
1021 * no effect on other vxlan devices.
1023 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1026 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1027 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1030 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1033 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1034 &dev
->default_dst
.remote_ip
))
1037 if (vxlan
->default_dst
.remote_ifindex
!=
1038 dev
->default_dst
.remote_ifindex
)
1047 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1049 atomic_inc(&vs
->refcnt
);
1052 void vxlan_sock_release(struct vxlan_sock
*vs
)
1054 struct sock
*sk
= vs
->sock
->sk
;
1055 struct net
*net
= sock_net(sk
);
1056 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1058 if (!atomic_dec_and_test(&vs
->refcnt
))
1061 spin_lock(&vn
->sock_lock
);
1062 hlist_del_rcu(&vs
->hlist
);
1063 rcu_assign_sk_user_data(vs
->sock
->sk
, NULL
);
1064 vxlan_notify_del_rx_port(vs
);
1065 spin_unlock(&vn
->sock_lock
);
1067 queue_work(vxlan_wq
, &vs
->del_work
);
1069 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1071 /* Callback to update multicast group membership when first VNI on
1072 * multicast asddress is brought up
1073 * Done as workqueue because ip_mc_join_group acquires RTNL.
1075 static void vxlan_igmp_join(struct work_struct
*work
)
1077 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1078 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1079 struct sock
*sk
= vs
->sock
->sk
;
1080 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1081 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1084 if (ip
->sa
.sa_family
== AF_INET
) {
1085 struct ip_mreqn mreq
= {
1086 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1087 .imr_ifindex
= ifindex
,
1090 ip_mc_join_group(sk
, &mreq
);
1091 #if IS_ENABLED(CONFIG_IPV6)
1093 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1094 &ip
->sin6
.sin6_addr
);
1099 vxlan_sock_release(vs
);
1100 dev_put(vxlan
->dev
);
1103 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1104 static void vxlan_igmp_leave(struct work_struct
*work
)
1106 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1107 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1108 struct sock
*sk
= vs
->sock
->sk
;
1109 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1110 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1113 if (ip
->sa
.sa_family
== AF_INET
) {
1114 struct ip_mreqn mreq
= {
1115 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1116 .imr_ifindex
= ifindex
,
1119 ip_mc_leave_group(sk
, &mreq
);
1120 #if IS_ENABLED(CONFIG_IPV6)
1122 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1123 &ip
->sin6
.sin6_addr
);
1129 vxlan_sock_release(vs
);
1130 dev_put(vxlan
->dev
);
1133 /* Callback from net/ipv4/udp.c to receive packets */
1134 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1136 struct vxlan_sock
*vs
;
1137 struct vxlanhdr
*vxh
;
1139 /* Need Vxlan and inner Ethernet header to be present */
1140 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1143 /* Return packets with reserved bits set */
1144 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1145 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
1146 (vxh
->vx_vni
& htonl(0xff))) {
1147 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1148 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1152 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1155 vs
= rcu_dereference_sk_user_data(sk
);
1159 skb_pop_rcv_encapsulation(skb
);
1161 vs
->rcv(vs
, skb
, vxh
->vx_vni
);
1165 /* Consume bad packet */
1170 /* Return non vxlan pkt */
1174 static void vxlan_rcv(struct vxlan_sock
*vs
,
1175 struct sk_buff
*skb
, __be32 vx_vni
)
1177 struct iphdr
*oip
= NULL
;
1178 struct ipv6hdr
*oip6
= NULL
;
1179 struct vxlan_dev
*vxlan
;
1180 struct pcpu_sw_netstats
*stats
;
1181 union vxlan_addr saddr
;
1184 union vxlan_addr
*remote_ip
;
1186 vni
= ntohl(vx_vni
) >> 8;
1187 /* Is this VNI defined? */
1188 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1192 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1193 skb_reset_mac_header(skb
);
1194 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1195 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1196 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1198 /* Ignore packet loops (and multicast echo) */
1199 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1202 /* Re-examine inner Ethernet packet */
1203 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1205 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1206 saddr
.sa
.sa_family
= AF_INET
;
1207 #if IS_ENABLED(CONFIG_IPV6)
1209 oip6
= ipv6_hdr(skb
);
1210 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1211 saddr
.sa
.sa_family
= AF_INET6
;
1215 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1216 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1219 skb_reset_network_header(skb
);
1222 err
= IP6_ECN_decapsulate(oip6
, skb
);
1224 err
= IP_ECN_decapsulate(oip
, skb
);
1226 if (unlikely(err
)) {
1227 if (log_ecn_error
) {
1229 net_info_ratelimited("non-ECT from %pI6\n",
1232 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1233 &oip
->saddr
, oip
->tos
);
1236 ++vxlan
->dev
->stats
.rx_frame_errors
;
1237 ++vxlan
->dev
->stats
.rx_errors
;
1242 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1243 u64_stats_update_begin(&stats
->syncp
);
1244 stats
->rx_packets
++;
1245 stats
->rx_bytes
+= skb
->len
;
1246 u64_stats_update_end(&stats
->syncp
);
1252 /* Consume bad packet */
1256 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1258 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1259 struct arphdr
*parp
;
1262 struct neighbour
*n
;
1264 if (dev
->flags
& IFF_NOARP
)
1267 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1268 dev
->stats
.tx_dropped
++;
1271 parp
= arp_hdr(skb
);
1273 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1274 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1275 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1276 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1277 parp
->ar_hln
!= dev
->addr_len
||
1280 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1282 arpptr
+= dev
->addr_len
; /* sha */
1283 memcpy(&sip
, arpptr
, sizeof(sip
));
1284 arpptr
+= sizeof(sip
);
1285 arpptr
+= dev
->addr_len
; /* tha */
1286 memcpy(&tip
, arpptr
, sizeof(tip
));
1288 if (ipv4_is_loopback(tip
) ||
1289 ipv4_is_multicast(tip
))
1292 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1295 struct vxlan_fdb
*f
;
1296 struct sk_buff
*reply
;
1298 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1303 f
= vxlan_find_mac(vxlan
, n
->ha
);
1304 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1305 /* bridge-local neighbor */
1310 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1318 skb_reset_mac_header(reply
);
1319 __skb_pull(reply
, skb_network_offset(reply
));
1320 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1321 reply
->pkt_type
= PACKET_HOST
;
1323 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1324 dev
->stats
.rx_dropped
++;
1325 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1326 union vxlan_addr ipa
= {
1327 .sin
.sin_addr
.s_addr
= tip
,
1328 .sa
.sa_family
= AF_INET
,
1331 vxlan_ip_miss(dev
, &ipa
);
1335 return NETDEV_TX_OK
;
1338 #if IS_ENABLED(CONFIG_IPV6)
1340 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1341 struct neighbour
*n
, bool isrouter
)
1343 struct net_device
*dev
= request
->dev
;
1344 struct sk_buff
*reply
;
1345 struct nd_msg
*ns
, *na
;
1346 struct ipv6hdr
*pip6
;
1348 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1355 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1356 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1357 reply
= alloc_skb(len
, GFP_ATOMIC
);
1361 reply
->protocol
= htons(ETH_P_IPV6
);
1363 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1364 skb_push(reply
, sizeof(struct ethhdr
));
1365 skb_set_mac_header(reply
, 0);
1367 ns
= (struct nd_msg
*)skb_transport_header(request
);
1369 daddr
= eth_hdr(request
)->h_source
;
1370 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1371 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1372 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1373 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1378 /* Ethernet header */
1379 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1380 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1381 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1382 reply
->protocol
= htons(ETH_P_IPV6
);
1384 skb_pull(reply
, sizeof(struct ethhdr
));
1385 skb_set_network_header(reply
, 0);
1386 skb_put(reply
, sizeof(struct ipv6hdr
));
1390 pip6
= ipv6_hdr(reply
);
1391 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1393 pip6
->priority
= ipv6_hdr(request
)->priority
;
1394 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1395 pip6
->hop_limit
= 255;
1396 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1397 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1399 skb_pull(reply
, sizeof(struct ipv6hdr
));
1400 skb_set_transport_header(reply
, 0);
1402 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1404 /* Neighbor Advertisement */
1405 memset(na
, 0, sizeof(*na
)+na_olen
);
1406 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1407 na
->icmph
.icmp6_router
= isrouter
;
1408 na
->icmph
.icmp6_override
= 1;
1409 na
->icmph
.icmp6_solicited
= 1;
1410 na
->target
= ns
->target
;
1411 ether_addr_copy(&na
->opt
[2], n
->ha
);
1412 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1413 na
->opt
[1] = na_olen
>> 3;
1415 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1416 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1417 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1419 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1421 skb_push(reply
, sizeof(struct ipv6hdr
));
1423 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1428 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1430 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1432 const struct ipv6hdr
*iphdr
;
1433 const struct in6_addr
*saddr
, *daddr
;
1434 struct neighbour
*n
;
1435 struct inet6_dev
*in6_dev
;
1437 in6_dev
= __in6_dev_get(dev
);
1441 if (!pskb_may_pull(skb
, skb
->len
))
1444 iphdr
= ipv6_hdr(skb
);
1445 saddr
= &iphdr
->saddr
;
1446 daddr
= &iphdr
->daddr
;
1448 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1449 if (msg
->icmph
.icmp6_code
!= 0 ||
1450 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1453 if (ipv6_addr_loopback(daddr
) ||
1454 ipv6_addr_is_multicast(&msg
->target
))
1457 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1460 struct vxlan_fdb
*f
;
1461 struct sk_buff
*reply
;
1463 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1468 f
= vxlan_find_mac(vxlan
, n
->ha
);
1469 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1470 /* bridge-local neighbor */
1475 reply
= vxlan_na_create(skb
, n
,
1476 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1483 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1484 dev
->stats
.rx_dropped
++;
1486 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1487 union vxlan_addr ipa
= {
1488 .sin6
.sin6_addr
= msg
->target
,
1489 .sa
.sa_family
= AF_INET6
,
1492 vxlan_ip_miss(dev
, &ipa
);
1497 return NETDEV_TX_OK
;
1501 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1503 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1504 struct neighbour
*n
;
1506 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1510 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1515 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1518 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1519 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1520 union vxlan_addr ipa
= {
1521 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1522 .sa
.sa_family
= AF_INET
,
1525 vxlan_ip_miss(dev
, &ipa
);
1531 #if IS_ENABLED(CONFIG_IPV6)
1534 struct ipv6hdr
*pip6
;
1536 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1538 pip6
= ipv6_hdr(skb
);
1539 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1540 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1541 union vxlan_addr ipa
= {
1542 .sin6
.sin6_addr
= pip6
->daddr
,
1543 .sa
.sa_family
= AF_INET6
,
1546 vxlan_ip_miss(dev
, &ipa
);
1560 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1562 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1564 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1573 /* Compute source port for outgoing packet
1574 * first choice to use L4 flow hash since it will spread
1575 * better and maybe available from hardware
1576 * secondary choice is to use jhash on the Ethernet header
1578 __be16
vxlan_src_port(__u16 port_min
, __u16 port_max
, struct sk_buff
*skb
)
1580 unsigned int range
= (port_max
- port_min
) + 1;
1583 hash
= skb_get_hash(skb
);
1585 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
1586 (__force u32
) skb
->protocol
);
1588 return htons((((u64
) hash
* range
) >> 32) + port_min
);
1590 EXPORT_SYMBOL_GPL(vxlan_src_port
);
1592 static inline struct sk_buff
*vxlan_handle_offloads(struct sk_buff
*skb
,
1595 int type
= udp_csum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1596 return iptunnel_handle_offloads(skb
, udp_csum
, type
);
1599 #if IS_ENABLED(CONFIG_IPV6)
1600 static int vxlan6_xmit_skb(struct vxlan_sock
*vs
,
1601 struct dst_entry
*dst
, struct sk_buff
*skb
,
1602 struct net_device
*dev
, struct in6_addr
*saddr
,
1603 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1604 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1607 struct ipv6hdr
*ip6h
;
1608 struct vxlanhdr
*vxh
;
1613 skb
= vxlan_handle_offloads(skb
, !udp_get_no_check6_tx(vs
->sock
->sk
));
1617 skb_scrub_packet(skb
, xnet
);
1619 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1620 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1621 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1623 /* Need space for new headers (invalidates iph ptr) */
1624 err
= skb_cow_head(skb
, min_headroom
);
1628 if (vlan_tx_tag_present(skb
)) {
1629 if (WARN_ON(!__vlan_put_tag(skb
,
1631 vlan_tx_tag_get(skb
))))
1637 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1638 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1641 __skb_push(skb
, sizeof(*uh
));
1642 skb_reset_transport_header(skb
);
1645 uh
->dest
= dst_port
;
1646 uh
->source
= src_port
;
1648 uh
->len
= htons(skb
->len
);
1650 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1651 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1653 skb_dst_set(skb
, dst
);
1655 udp6_set_csum(udp_get_no_check6_tx(vs
->sock
->sk
), skb
,
1656 saddr
, daddr
, skb
->len
);
1658 __skb_push(skb
, sizeof(*ip6h
));
1659 skb_reset_network_header(skb
);
1660 ip6h
= ipv6_hdr(skb
);
1662 ip6h
->priority
= prio
;
1663 ip6h
->flow_lbl
[0] = 0;
1664 ip6h
->flow_lbl
[1] = 0;
1665 ip6h
->flow_lbl
[2] = 0;
1666 ip6h
->payload_len
= htons(skb
->len
);
1667 ip6h
->nexthdr
= IPPROTO_UDP
;
1668 ip6h
->hop_limit
= ttl
;
1669 ip6h
->daddr
= *daddr
;
1670 ip6h
->saddr
= *saddr
;
1672 ip6tunnel_xmit(skb
, dev
);
1677 int vxlan_xmit_skb(struct vxlan_sock
*vs
,
1678 struct rtable
*rt
, struct sk_buff
*skb
,
1679 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1680 __be16 src_port
, __be16 dst_port
, __be32 vni
, bool xnet
)
1682 struct vxlanhdr
*vxh
;
1687 skb
= vxlan_handle_offloads(skb
, !vs
->sock
->sk
->sk_no_check_tx
);
1691 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1692 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1693 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1695 /* Need space for new headers (invalidates iph ptr) */
1696 err
= skb_cow_head(skb
, min_headroom
);
1700 if (vlan_tx_tag_present(skb
)) {
1701 if (WARN_ON(!__vlan_put_tag(skb
,
1703 vlan_tx_tag_get(skb
))))
1709 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1710 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1713 __skb_push(skb
, sizeof(*uh
));
1714 skb_reset_transport_header(skb
);
1717 uh
->dest
= dst_port
;
1718 uh
->source
= src_port
;
1720 uh
->len
= htons(skb
->len
);
1722 udp_set_csum(vs
->sock
->sk
->sk_no_check_tx
, skb
,
1723 src
, dst
, skb
->len
);
1725 return iptunnel_xmit(vs
->sock
->sk
, rt
, skb
, src
, dst
, IPPROTO_UDP
,
1726 tos
, ttl
, df
, xnet
);
1728 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1730 /* Bypass encapsulation if the destination is local */
1731 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1732 struct vxlan_dev
*dst_vxlan
)
1734 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1735 union vxlan_addr loopback
;
1736 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1738 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1739 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1740 skb
->pkt_type
= PACKET_HOST
;
1741 skb
->encapsulation
= 0;
1742 skb
->dev
= dst_vxlan
->dev
;
1743 __skb_pull(skb
, skb_network_offset(skb
));
1745 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1746 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1747 loopback
.sa
.sa_family
= AF_INET
;
1748 #if IS_ENABLED(CONFIG_IPV6)
1750 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1751 loopback
.sa
.sa_family
= AF_INET6
;
1755 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1756 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1758 u64_stats_update_begin(&tx_stats
->syncp
);
1759 tx_stats
->tx_packets
++;
1760 tx_stats
->tx_bytes
+= skb
->len
;
1761 u64_stats_update_end(&tx_stats
->syncp
);
1763 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1764 u64_stats_update_begin(&rx_stats
->syncp
);
1765 rx_stats
->rx_packets
++;
1766 rx_stats
->rx_bytes
+= skb
->len
;
1767 u64_stats_update_end(&rx_stats
->syncp
);
1769 skb
->dev
->stats
.rx_dropped
++;
1773 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1774 struct vxlan_rdst
*rdst
, bool did_rsc
)
1776 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1777 struct rtable
*rt
= NULL
;
1778 const struct iphdr
*old_iph
;
1780 union vxlan_addr
*dst
;
1781 __be16 src_port
= 0, dst_port
;
1787 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1788 vni
= rdst
->remote_vni
;
1789 dst
= &rdst
->remote_ip
;
1791 if (vxlan_addr_any(dst
)) {
1793 /* short-circuited back to local bridge */
1794 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1800 old_iph
= ip_hdr(skb
);
1803 if (!ttl
&& vxlan_addr_multicast(dst
))
1808 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1810 src_port
= vxlan_src_port(vxlan
->port_min
, vxlan
->port_max
, skb
);
1812 if (dst
->sa
.sa_family
== AF_INET
) {
1813 memset(&fl4
, 0, sizeof(fl4
));
1814 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1815 fl4
.flowi4_tos
= RT_TOS(tos
);
1816 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1817 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1819 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1821 netdev_dbg(dev
, "no route to %pI4\n",
1822 &dst
->sin
.sin_addr
.s_addr
);
1823 dev
->stats
.tx_carrier_errors
++;
1827 if (rt
->dst
.dev
== dev
) {
1828 netdev_dbg(dev
, "circular route to %pI4\n",
1829 &dst
->sin
.sin_addr
.s_addr
);
1830 dev
->stats
.collisions
++;
1834 /* Bypass encapsulation if the destination is local */
1835 if (rt
->rt_flags
& RTCF_LOCAL
&&
1836 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1837 struct vxlan_dev
*dst_vxlan
;
1840 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1843 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1847 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1848 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1850 err
= vxlan_xmit_skb(vxlan
->vn_sock
, rt
, skb
,
1851 fl4
.saddr
, dst
->sin
.sin_addr
.s_addr
,
1852 tos
, ttl
, df
, src_port
, dst_port
,
1854 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1858 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1859 #if IS_ENABLED(CONFIG_IPV6)
1861 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1862 struct dst_entry
*ndst
;
1866 memset(&fl6
, 0, sizeof(fl6
));
1867 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
1868 fl6
.daddr
= dst
->sin6
.sin6_addr
;
1869 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
1870 fl6
.flowi6_proto
= IPPROTO_UDP
;
1872 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
1873 netdev_dbg(dev
, "no route to %pI6\n",
1874 &dst
->sin6
.sin6_addr
);
1875 dev
->stats
.tx_carrier_errors
++;
1879 if (ndst
->dev
== dev
) {
1880 netdev_dbg(dev
, "circular route to %pI6\n",
1881 &dst
->sin6
.sin6_addr
);
1883 dev
->stats
.collisions
++;
1887 /* Bypass encapsulation if the destination is local */
1888 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
1889 if (flags
& RTF_LOCAL
&&
1890 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1891 struct vxlan_dev
*dst_vxlan
;
1894 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1897 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1901 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
1903 err
= vxlan6_xmit_skb(vxlan
->vn_sock
, ndst
, skb
,
1904 dev
, &fl6
.saddr
, &fl6
.daddr
, 0, ttl
,
1905 src_port
, dst_port
, htonl(vni
<< 8),
1906 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1913 dev
->stats
.tx_dropped
++;
1919 dev
->stats
.tx_errors
++;
1924 /* Transmit local packets over Vxlan
1926 * Outer IP header inherits ECN and DF from inner header.
1927 * Outer UDP destination is the VXLAN assigned port.
1928 * source port is based on hash of flow
1930 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1932 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1934 bool did_rsc
= false;
1935 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
1936 struct vxlan_fdb
*f
;
1938 skb_reset_mac_header(skb
);
1941 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
1942 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
1943 return arp_reduce(dev
, skb
);
1944 #if IS_ENABLED(CONFIG_IPV6)
1945 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
1946 skb
->len
>= sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
) &&
1947 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
1950 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1951 if (msg
->icmph
.icmp6_code
== 0 &&
1952 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
1953 return neigh_reduce(dev
, skb
);
1958 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1961 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1962 (ntohs(eth
->h_proto
) == ETH_P_IP
||
1963 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
1964 did_rsc
= route_shortcircuit(dev
, skb
);
1966 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1970 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
1972 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
1973 !is_multicast_ether_addr(eth
->h_dest
))
1974 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
1976 dev
->stats
.tx_dropped
++;
1978 return NETDEV_TX_OK
;
1982 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
1983 struct sk_buff
*skb1
;
1989 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1991 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
1995 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
1998 return NETDEV_TX_OK
;
2001 /* Walk the forwarding table and purge stale entries */
2002 static void vxlan_cleanup(unsigned long arg
)
2004 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2005 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2008 if (!netif_running(vxlan
->dev
))
2011 spin_lock_bh(&vxlan
->hash_lock
);
2012 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2013 struct hlist_node
*p
, *n
;
2014 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2016 = container_of(p
, struct vxlan_fdb
, hlist
);
2017 unsigned long timeout
;
2019 if (f
->state
& NUD_PERMANENT
)
2022 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2023 if (time_before_eq(timeout
, jiffies
)) {
2024 netdev_dbg(vxlan
->dev
,
2025 "garbage collect %pM\n",
2027 f
->state
= NUD_STALE
;
2028 vxlan_fdb_destroy(vxlan
, f
);
2029 } else if (time_before(timeout
, next_timer
))
2030 next_timer
= timeout
;
2033 spin_unlock_bh(&vxlan
->hash_lock
);
2035 mod_timer(&vxlan
->age_timer
, next_timer
);
2038 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2040 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2042 vxlan
->vn_sock
= vs
;
2043 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2046 /* Setup stats when device is created */
2047 static int vxlan_init(struct net_device
*dev
)
2049 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2050 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2051 struct vxlan_sock
*vs
;
2053 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2057 spin_lock(&vn
->sock_lock
);
2058 vs
= vxlan_find_sock(vxlan
->net
, vxlan
->dst_port
);
2060 /* If we have a socket with same port already, reuse it */
2061 atomic_inc(&vs
->refcnt
);
2062 vxlan_vs_add_dev(vs
, vxlan
);
2064 /* otherwise make new socket outside of RTNL */
2066 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2068 spin_unlock(&vn
->sock_lock
);
2073 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2075 struct vxlan_fdb
*f
;
2077 spin_lock_bh(&vxlan
->hash_lock
);
2078 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2080 vxlan_fdb_destroy(vxlan
, f
);
2081 spin_unlock_bh(&vxlan
->hash_lock
);
2084 static void vxlan_uninit(struct net_device
*dev
)
2086 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2087 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2089 vxlan_fdb_delete_default(vxlan
);
2092 vxlan_sock_release(vs
);
2093 free_percpu(dev
->tstats
);
2096 /* Start ageing timer and join group when device is brought up */
2097 static int vxlan_open(struct net_device
*dev
)
2099 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2100 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2102 /* socket hasn't been created */
2106 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2107 vxlan_sock_hold(vs
);
2109 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2112 if (vxlan
->age_interval
)
2113 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2118 /* Purge the forwarding table */
2119 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2123 spin_lock_bh(&vxlan
->hash_lock
);
2124 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2125 struct hlist_node
*p
, *n
;
2126 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2128 = container_of(p
, struct vxlan_fdb
, hlist
);
2129 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2130 if (!is_zero_ether_addr(f
->eth_addr
))
2131 vxlan_fdb_destroy(vxlan
, f
);
2134 spin_unlock_bh(&vxlan
->hash_lock
);
2137 /* Cleanup timer and forwarding table on shutdown */
2138 static int vxlan_stop(struct net_device
*dev
)
2140 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2141 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2142 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2144 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2145 !vxlan_group_used(vn
, vxlan
)) {
2146 vxlan_sock_hold(vs
);
2148 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2151 del_timer_sync(&vxlan
->age_timer
);
2158 /* Stub, nothing needs to be done. */
2159 static void vxlan_set_multicast_list(struct net_device
*dev
)
2163 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2165 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2166 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2167 struct net_device
*lowerdev
;
2170 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2171 if (lowerdev
== NULL
)
2172 return eth_change_mtu(dev
, new_mtu
);
2174 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2175 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2177 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2179 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2186 static const struct net_device_ops vxlan_netdev_ops
= {
2187 .ndo_init
= vxlan_init
,
2188 .ndo_uninit
= vxlan_uninit
,
2189 .ndo_open
= vxlan_open
,
2190 .ndo_stop
= vxlan_stop
,
2191 .ndo_start_xmit
= vxlan_xmit
,
2192 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2193 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2194 .ndo_change_mtu
= vxlan_change_mtu
,
2195 .ndo_validate_addr
= eth_validate_addr
,
2196 .ndo_set_mac_address
= eth_mac_addr
,
2197 .ndo_fdb_add
= vxlan_fdb_add
,
2198 .ndo_fdb_del
= vxlan_fdb_delete
,
2199 .ndo_fdb_dump
= vxlan_fdb_dump
,
2202 /* Info for udev, that this is a virtual tunnel endpoint */
2203 static struct device_type vxlan_type
= {
2207 /* Calls the ndo_add_vxlan_port of the caller in order to
2208 * supply the listening VXLAN udp ports. Callers are expected
2209 * to implement the ndo_add_vxlan_port.
2211 void vxlan_get_rx_port(struct net_device
*dev
)
2213 struct vxlan_sock
*vs
;
2214 struct net
*net
= dev_net(dev
);
2215 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2216 sa_family_t sa_family
;
2220 spin_lock(&vn
->sock_lock
);
2221 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2222 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2223 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2224 sa_family
= vs
->sock
->sk
->sk_family
;
2225 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2229 spin_unlock(&vn
->sock_lock
);
2231 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2233 /* Initialize the device structure. */
2234 static void vxlan_setup(struct net_device
*dev
)
2236 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2240 eth_hw_addr_random(dev
);
2242 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2243 dev
->needed_headroom
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2245 dev
->needed_headroom
= ETH_HLEN
+ VXLAN_HEADROOM
;
2247 dev
->netdev_ops
= &vxlan_netdev_ops
;
2248 dev
->destructor
= free_netdev
;
2249 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2251 dev
->tx_queue_len
= 0;
2252 dev
->features
|= NETIF_F_LLTX
;
2253 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2254 dev
->features
|= NETIF_F_RXCSUM
;
2255 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2257 dev
->vlan_features
= dev
->features
;
2258 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2259 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2260 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2261 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2262 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
2263 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2265 INIT_LIST_HEAD(&vxlan
->next
);
2266 spin_lock_init(&vxlan
->hash_lock
);
2267 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2268 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2269 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2271 init_timer_deferrable(&vxlan
->age_timer
);
2272 vxlan
->age_timer
.function
= vxlan_cleanup
;
2273 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2275 inet_get_local_port_range(dev_net(dev
), &low
, &high
);
2276 vxlan
->port_min
= low
;
2277 vxlan
->port_max
= high
;
2278 vxlan
->dst_port
= htons(vxlan_port
);
2282 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2283 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2286 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2287 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2288 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2289 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2290 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2291 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2292 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2293 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2294 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2295 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2296 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2297 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2298 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2299 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2300 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2301 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2302 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2303 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2306 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2308 if (tb
[IFLA_ADDRESS
]) {
2309 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2310 pr_debug("invalid link address (not ethernet)\n");
2314 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2315 pr_debug("invalid all zero ethernet address\n");
2316 return -EADDRNOTAVAIL
;
2323 if (data
[IFLA_VXLAN_ID
]) {
2324 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2325 if (id
>= VXLAN_VID_MASK
)
2329 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2330 const struct ifla_vxlan_port_range
*p
2331 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2333 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2334 pr_debug("port range %u .. %u not valid\n",
2335 ntohs(p
->low
), ntohs(p
->high
));
2343 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2344 struct ethtool_drvinfo
*drvinfo
)
2346 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2347 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2350 static const struct ethtool_ops vxlan_ethtool_ops
= {
2351 .get_drvinfo
= vxlan_get_drvinfo
,
2352 .get_link
= ethtool_op_get_link
,
2355 static void vxlan_del_work(struct work_struct
*work
)
2357 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2359 sk_release_kernel(vs
->sock
->sk
);
2363 #if IS_ENABLED(CONFIG_IPV6)
2364 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2365 * could be used for both IPv4 and IPv6 communications, but
2366 * users may set bindv6only=1.
2368 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
, u32 flags
)
2371 struct socket
*sock
;
2372 struct sockaddr_in6 vxlan_addr
= {
2373 .sin6_family
= AF_INET6
,
2378 rc
= sock_create_kern(AF_INET6
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2380 pr_debug("UDPv6 socket create failed\n");
2384 /* Put in proper namespace */
2386 sk_change_net(sk
, net
);
2388 kernel_setsockopt(sock
, SOL_IPV6
, IPV6_V6ONLY
,
2389 (char *)&val
, sizeof(val
));
2390 rc
= kernel_bind(sock
, (struct sockaddr
*)&vxlan_addr
,
2391 sizeof(struct sockaddr_in6
));
2393 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2394 &vxlan_addr
.sin6_addr
, ntohs(vxlan_addr
.sin6_port
), rc
);
2395 sk_release_kernel(sk
);
2398 /* At this point, IPv6 module should have been loaded in
2399 * sock_create_kern().
2403 /* Disable multicast loopback */
2404 inet_sk(sk
)->mc_loop
= 0;
2406 if (flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)
2407 udp_set_no_check6_tx(sk
, true);
2409 if (flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)
2410 udp_set_no_check6_rx(sk
, true);
2417 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
, u32 flags
)
2419 return ERR_PTR(-EPFNOSUPPORT
);
2423 static struct socket
*create_v4_sock(struct net
*net
, __be16 port
, u32 flags
)
2426 struct socket
*sock
;
2427 struct sockaddr_in vxlan_addr
= {
2428 .sin_family
= AF_INET
,
2429 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
2434 /* Create UDP socket for encapsulation receive. */
2435 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2437 pr_debug("UDP socket create failed\n");
2441 /* Put in proper namespace */
2443 sk_change_net(sk
, net
);
2445 rc
= kernel_bind(sock
, (struct sockaddr
*) &vxlan_addr
,
2446 sizeof(vxlan_addr
));
2448 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2449 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
2450 sk_release_kernel(sk
);
2454 /* Disable multicast loopback */
2455 inet_sk(sk
)->mc_loop
= 0;
2457 if (!(flags
& VXLAN_F_UDP_CSUM
))
2458 sock
->sk
->sk_no_check_tx
= 1;
2463 /* Create new listen socket if needed */
2464 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2465 vxlan_rcv_t
*rcv
, void *data
,
2468 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2469 struct vxlan_sock
*vs
;
2470 struct socket
*sock
;
2473 bool ipv6
= !!(flags
& VXLAN_F_IPV6
);
2475 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2477 return ERR_PTR(-ENOMEM
);
2479 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2480 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2482 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2485 sock
= create_v6_sock(net
, port
, flags
);
2487 sock
= create_v4_sock(net
, port
, flags
);
2490 return ERR_CAST(sock
);
2495 atomic_set(&vs
->refcnt
, 1);
2498 rcu_assign_sk_user_data(vs
->sock
->sk
, vs
);
2500 /* Initialize the vxlan udp offloads structure */
2501 vs
->udp_offloads
.port
= port
;
2502 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2503 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2505 spin_lock(&vn
->sock_lock
);
2506 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2507 vxlan_notify_add_rx_port(vs
);
2508 spin_unlock(&vn
->sock_lock
);
2510 /* Mark socket as an encapsulation socket. */
2511 udp_sk(sk
)->encap_type
= 1;
2512 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
2513 #if IS_ENABLED(CONFIG_IPV6)
2515 ipv6_stub
->udpv6_encap_enable();
2523 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2524 vxlan_rcv_t
*rcv
, void *data
,
2525 bool no_share
, u32 flags
)
2527 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2528 struct vxlan_sock
*vs
;
2530 vs
= vxlan_socket_create(net
, port
, rcv
, data
, flags
);
2534 if (no_share
) /* Return error if sharing is not allowed. */
2537 spin_lock(&vn
->sock_lock
);
2538 vs
= vxlan_find_sock(net
, port
);
2541 atomic_inc(&vs
->refcnt
);
2543 vs
= ERR_PTR(-EBUSY
);
2545 spin_unlock(&vn
->sock_lock
);
2548 vs
= ERR_PTR(-EINVAL
);
2552 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2554 /* Scheduled at device creation to bind to a socket */
2555 static void vxlan_sock_work(struct work_struct
*work
)
2557 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2558 struct net
*net
= vxlan
->net
;
2559 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2560 __be16 port
= vxlan
->dst_port
;
2561 struct vxlan_sock
*nvs
;
2563 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
);
2564 spin_lock(&vn
->sock_lock
);
2566 vxlan_vs_add_dev(nvs
, vxlan
);
2567 spin_unlock(&vn
->sock_lock
);
2569 dev_put(vxlan
->dev
);
2572 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2573 struct nlattr
*tb
[], struct nlattr
*data
[])
2575 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2576 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2577 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2580 bool use_ipv6
= false;
2582 if (!data
[IFLA_VXLAN_ID
])
2585 vxlan
->net
= dev_net(dev
);
2587 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2588 dst
->remote_vni
= vni
;
2590 /* Unless IPv6 is explicitly requested, assume IPv4 */
2591 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2592 if (data
[IFLA_VXLAN_GROUP
]) {
2593 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2594 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2595 if (!IS_ENABLED(CONFIG_IPV6
))
2596 return -EPFNOSUPPORT
;
2598 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2599 sizeof(struct in6_addr
));
2600 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2604 if (data
[IFLA_VXLAN_LOCAL
]) {
2605 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2606 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2607 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2608 if (!IS_ENABLED(CONFIG_IPV6
))
2609 return -EPFNOSUPPORT
;
2611 /* TODO: respect scope id */
2612 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2613 sizeof(struct in6_addr
));
2614 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2618 if (data
[IFLA_VXLAN_LINK
] &&
2619 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2620 struct net_device
*lowerdev
2621 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2624 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2628 #if IS_ENABLED(CONFIG_IPV6)
2630 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2631 if (idev
&& idev
->cnf
.disable_ipv6
) {
2632 pr_info("IPv6 is disabled via sysctl\n");
2635 vxlan
->flags
|= VXLAN_F_IPV6
;
2640 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2642 dev
->needed_headroom
= lowerdev
->hard_header_len
+
2643 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2644 } else if (use_ipv6
)
2645 vxlan
->flags
|= VXLAN_F_IPV6
;
2647 if (data
[IFLA_VXLAN_TOS
])
2648 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2650 if (data
[IFLA_VXLAN_TTL
])
2651 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2653 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2654 vxlan
->flags
|= VXLAN_F_LEARN
;
2656 if (data
[IFLA_VXLAN_AGEING
])
2657 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2659 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2661 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2662 vxlan
->flags
|= VXLAN_F_PROXY
;
2664 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2665 vxlan
->flags
|= VXLAN_F_RSC
;
2667 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2668 vxlan
->flags
|= VXLAN_F_L2MISS
;
2670 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2671 vxlan
->flags
|= VXLAN_F_L3MISS
;
2673 if (data
[IFLA_VXLAN_LIMIT
])
2674 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2676 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2677 const struct ifla_vxlan_port_range
*p
2678 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2679 vxlan
->port_min
= ntohs(p
->low
);
2680 vxlan
->port_max
= ntohs(p
->high
);
2683 if (data
[IFLA_VXLAN_PORT
])
2684 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2686 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2687 vxlan
->flags
|= VXLAN_F_UDP_CSUM
;
2689 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2690 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2691 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2693 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2694 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2695 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2697 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
2698 pr_info("duplicate VNI %u\n", vni
);
2702 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2704 /* create an fdb entry for a valid default destination */
2705 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2706 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2707 &vxlan
->default_dst
.remote_ip
,
2708 NUD_REACHABLE
|NUD_PERMANENT
,
2709 NLM_F_EXCL
|NLM_F_CREATE
,
2711 vxlan
->default_dst
.remote_vni
,
2712 vxlan
->default_dst
.remote_ifindex
,
2718 err
= register_netdevice(dev
);
2720 vxlan_fdb_delete_default(vxlan
);
2724 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2729 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2731 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2732 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2734 spin_lock(&vn
->sock_lock
);
2735 if (!hlist_unhashed(&vxlan
->hlist
))
2736 hlist_del_rcu(&vxlan
->hlist
);
2737 spin_unlock(&vn
->sock_lock
);
2739 list_del(&vxlan
->next
);
2740 unregister_netdevice_queue(dev
, head
);
2743 static size_t vxlan_get_size(const struct net_device
*dev
)
2746 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2747 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2748 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2749 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2750 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2751 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2752 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2753 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2754 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2755 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2756 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2757 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2758 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2759 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2760 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2761 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
2762 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2763 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2767 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2769 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2770 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2771 struct ifla_vxlan_port_range ports
= {
2772 .low
= htons(vxlan
->port_min
),
2773 .high
= htons(vxlan
->port_max
),
2776 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2777 goto nla_put_failure
;
2779 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2780 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2781 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2782 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2783 goto nla_put_failure
;
2784 #if IS_ENABLED(CONFIG_IPV6)
2786 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2787 &dst
->remote_ip
.sin6
.sin6_addr
))
2788 goto nla_put_failure
;
2793 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2794 goto nla_put_failure
;
2796 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2797 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2798 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2799 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2800 goto nla_put_failure
;
2801 #if IS_ENABLED(CONFIG_IPV6)
2803 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2804 &vxlan
->saddr
.sin6
.sin6_addr
))
2805 goto nla_put_failure
;
2810 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2811 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2812 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2813 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2814 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2815 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2816 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2817 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2818 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2819 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2820 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2821 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2822 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2823 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
) ||
2824 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
2825 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
2826 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
2827 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
2828 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
2829 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)))
2830 goto nla_put_failure
;
2832 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2833 goto nla_put_failure
;
2841 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2843 .maxtype
= IFLA_VXLAN_MAX
,
2844 .policy
= vxlan_policy
,
2845 .priv_size
= sizeof(struct vxlan_dev
),
2846 .setup
= vxlan_setup
,
2847 .validate
= vxlan_validate
,
2848 .newlink
= vxlan_newlink
,
2849 .dellink
= vxlan_dellink
,
2850 .get_size
= vxlan_get_size
,
2851 .fill_info
= vxlan_fill_info
,
2854 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2855 struct net_device
*dev
)
2857 struct vxlan_dev
*vxlan
, *next
;
2858 LIST_HEAD(list_kill
);
2860 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2861 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2863 /* In case we created vxlan device with carrier
2864 * and we loose the carrier due to module unload
2865 * we also need to remove vxlan device. In other
2866 * cases, it's not necessary and remote_ifindex
2867 * is 0 here, so no matches.
2869 if (dst
->remote_ifindex
== dev
->ifindex
)
2870 vxlan_dellink(vxlan
->dev
, &list_kill
);
2873 unregister_netdevice_many(&list_kill
);
2876 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2877 unsigned long event
, void *ptr
)
2879 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2880 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2882 if (event
== NETDEV_UNREGISTER
)
2883 vxlan_handle_lowerdev_unregister(vn
, dev
);
2888 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2889 .notifier_call
= vxlan_lowerdev_event
,
2892 static __net_init
int vxlan_init_net(struct net
*net
)
2894 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2897 INIT_LIST_HEAD(&vn
->vxlan_list
);
2898 spin_lock_init(&vn
->sock_lock
);
2900 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2901 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
2906 static void __net_exit
vxlan_exit_net(struct net
*net
)
2908 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2909 struct vxlan_dev
*vxlan
, *next
;
2910 struct net_device
*dev
, *aux
;
2914 for_each_netdev_safe(net
, dev
, aux
)
2915 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
2916 unregister_netdevice_queue(dev
, &list
);
2918 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2919 /* If vxlan->dev is in the same netns, it has already been added
2920 * to the list by the previous loop.
2922 if (!net_eq(dev_net(vxlan
->dev
), net
))
2923 unregister_netdevice_queue(dev
, &list
);
2926 unregister_netdevice_many(&list
);
2930 static struct pernet_operations vxlan_net_ops
= {
2931 .init
= vxlan_init_net
,
2932 .exit
= vxlan_exit_net
,
2933 .id
= &vxlan_net_id
,
2934 .size
= sizeof(struct vxlan_net
),
2937 static int __init
vxlan_init_module(void)
2941 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
2945 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
2947 rc
= register_pernet_subsys(&vxlan_net_ops
);
2951 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
2955 rc
= rtnl_link_register(&vxlan_link_ops
);
2961 unregister_netdevice_notifier(&vxlan_notifier_block
);
2963 unregister_pernet_subsys(&vxlan_net_ops
);
2965 destroy_workqueue(vxlan_wq
);
2968 late_initcall(vxlan_init_module
);
2970 static void __exit
vxlan_cleanup_module(void)
2972 rtnl_link_unregister(&vxlan_link_ops
);
2973 unregister_netdevice_notifier(&vxlan_notifier_block
);
2974 destroy_workqueue(vxlan_wq
);
2975 unregister_pernet_subsys(&vxlan_net_ops
);
2976 /* rcu_barrier() is called by netns */
2978 module_exit(vxlan_cleanup_module
);
2980 MODULE_LICENSE("GPL");
2981 MODULE_VERSION(VXLAN_VERSION
);
2982 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2983 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2984 MODULE_ALIAS_RTNL_LINK("vxlan");