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_* below */
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 #define VXLAN_F_LEARN 0x01
154 #define VXLAN_F_PROXY 0x02
155 #define VXLAN_F_RSC 0x04
156 #define VXLAN_F_L2MISS 0x08
157 #define VXLAN_F_L3MISS 0x10
158 #define VXLAN_F_IPV6 0x20 /* internal flag */
160 /* salt for hash table */
161 static u32 vxlan_salt __read_mostly
;
162 static struct workqueue_struct
*vxlan_wq
;
164 static void vxlan_sock_work(struct work_struct
*work
);
166 #if IS_ENABLED(CONFIG_IPV6)
168 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
170 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
172 if (a
->sa
.sa_family
== AF_INET6
)
173 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
175 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
178 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
180 if (ipa
->sa
.sa_family
== AF_INET6
)
181 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
183 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
186 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
188 if (ipa
->sa
.sa_family
== AF_INET6
)
189 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
191 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
194 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
196 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
197 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
198 ip
->sa
.sa_family
= AF_INET6
;
200 } else if (nla_len(nla
) >= sizeof(__be32
)) {
201 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
202 ip
->sa
.sa_family
= AF_INET
;
205 return -EAFNOSUPPORT
;
209 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
210 const union vxlan_addr
*ip
)
212 if (ip
->sa
.sa_family
== AF_INET6
)
213 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
215 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
218 #else /* !CONFIG_IPV6 */
221 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
223 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
226 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
228 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
231 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
233 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
236 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
238 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
239 return -EAFNOSUPPORT
;
240 } else if (nla_len(nla
) >= sizeof(__be32
)) {
241 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
242 ip
->sa
.sa_family
= AF_INET
;
245 return -EAFNOSUPPORT
;
249 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
250 const union vxlan_addr
*ip
)
252 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
256 /* Virtual Network hash table head */
257 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
259 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
262 /* Socket hash table head */
263 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
265 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
267 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
270 /* First remote destination for a forwarding entry.
271 * Guaranteed to be non-NULL because remotes are never deleted.
273 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
275 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
278 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
280 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
283 /* Find VXLAN socket based on network namespace and UDP port */
284 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, __be16 port
)
286 struct vxlan_sock
*vs
;
288 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
289 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
295 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
297 struct vxlan_dev
*vxlan
;
299 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
300 if (vxlan
->default_dst
.remote_vni
== id
)
307 /* Look up VNI in a per net namespace table */
308 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
310 struct vxlan_sock
*vs
;
312 vs
= vxlan_find_sock(net
, port
);
316 return vxlan_vs_find_vni(vs
, id
);
319 /* Fill in neighbour message in skbuff. */
320 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
321 const struct vxlan_fdb
*fdb
,
322 u32 portid
, u32 seq
, int type
, unsigned int flags
,
323 const struct vxlan_rdst
*rdst
)
325 unsigned long now
= jiffies
;
326 struct nda_cacheinfo ci
;
327 struct nlmsghdr
*nlh
;
329 bool send_ip
, send_eth
;
331 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
335 ndm
= nlmsg_data(nlh
);
336 memset(ndm
, 0, sizeof(*ndm
));
338 send_eth
= send_ip
= true;
340 if (type
== RTM_GETNEIGH
) {
341 ndm
->ndm_family
= AF_INET
;
342 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
343 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
345 ndm
->ndm_family
= AF_BRIDGE
;
346 ndm
->ndm_state
= fdb
->state
;
347 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
348 ndm
->ndm_flags
= fdb
->flags
;
349 ndm
->ndm_type
= NDA_DST
;
351 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
352 goto nla_put_failure
;
354 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
355 goto nla_put_failure
;
357 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
358 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
359 goto nla_put_failure
;
360 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
361 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
362 goto nla_put_failure
;
363 if (rdst
->remote_ifindex
&&
364 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
365 goto nla_put_failure
;
367 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
368 ci
.ndm_confirmed
= 0;
369 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
372 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
373 goto nla_put_failure
;
375 return nlmsg_end(skb
, nlh
);
378 nlmsg_cancel(skb
, nlh
);
382 static inline size_t vxlan_nlmsg_size(void)
384 return NLMSG_ALIGN(sizeof(struct ndmsg
))
385 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
386 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
387 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
388 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
389 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
390 + nla_total_size(sizeof(struct nda_cacheinfo
));
393 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
394 struct vxlan_rdst
*rd
, int type
)
396 struct net
*net
= dev_net(vxlan
->dev
);
400 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
404 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
406 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
407 WARN_ON(err
== -EMSGSIZE
);
412 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
416 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
419 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
421 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
422 struct vxlan_fdb f
= {
425 struct vxlan_rdst remote
= {
426 .remote_ip
= *ipa
, /* goes to NDA_DST */
427 .remote_vni
= VXLAN_N_VID
,
430 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
433 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
435 struct vxlan_fdb f
= {
438 struct vxlan_rdst remote
= { };
440 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
442 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
445 /* Hash Ethernet address */
446 static u32
eth_hash(const unsigned char *addr
)
448 u64 value
= get_unaligned((u64
*)addr
);
450 /* only want 6 bytes */
456 return hash_64(value
, FDB_HASH_BITS
);
459 /* Hash chain to use given mac address */
460 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
463 return &vxlan
->fdb_head
[eth_hash(mac
)];
466 /* Look up Ethernet address in forwarding table */
467 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
470 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
473 hlist_for_each_entry_rcu(f
, head
, hlist
) {
474 if (ether_addr_equal(mac
, f
->eth_addr
))
481 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
486 f
= __vxlan_find_mac(vxlan
, mac
);
493 /* caller should hold vxlan->hash_lock */
494 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
495 union vxlan_addr
*ip
, __be16 port
,
496 __u32 vni
, __u32 ifindex
)
498 struct vxlan_rdst
*rd
;
500 list_for_each_entry(rd
, &f
->remotes
, list
) {
501 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
502 rd
->remote_port
== port
&&
503 rd
->remote_vni
== vni
&&
504 rd
->remote_ifindex
== ifindex
)
511 /* Replace destination of unicast mac */
512 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
513 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
515 struct vxlan_rdst
*rd
;
517 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
521 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
525 rd
->remote_port
= port
;
526 rd
->remote_vni
= vni
;
527 rd
->remote_ifindex
= ifindex
;
531 /* Add/update destinations for multicast */
532 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
533 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
534 __u32 ifindex
, struct vxlan_rdst
**rdp
)
536 struct vxlan_rdst
*rd
;
538 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
542 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
546 rd
->remote_port
= port
;
547 rd
->remote_vni
= vni
;
548 rd
->remote_ifindex
= ifindex
;
550 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
556 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
)
558 struct sk_buff
*p
, **pp
= NULL
;
559 struct vxlanhdr
*vh
, *vh2
;
560 struct ethhdr
*eh
, *eh2
;
561 unsigned int hlen
, off_vx
, off_eth
;
562 const struct packet_offload
*ptype
;
566 off_vx
= skb_gro_offset(skb
);
567 hlen
= off_vx
+ sizeof(*vh
);
568 vh
= skb_gro_header_fast(skb
, off_vx
);
569 if (skb_gro_header_hard(skb
, hlen
)) {
570 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
574 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
576 off_eth
= skb_gro_offset(skb
);
577 hlen
= off_eth
+ sizeof(*eh
);
578 eh
= skb_gro_header_fast(skb
, off_eth
);
579 if (skb_gro_header_hard(skb
, hlen
)) {
580 eh
= skb_gro_header_slow(skb
, hlen
, off_eth
);
587 for (p
= *head
; p
; p
= p
->next
) {
588 if (!NAPI_GRO_CB(p
)->same_flow
)
591 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
592 eh2
= (struct ethhdr
*)(p
->data
+ off_eth
);
593 if (vh
->vx_vni
!= vh2
->vx_vni
|| compare_ether_header(eh
, eh2
)) {
594 NAPI_GRO_CB(p
)->same_flow
= 0;
602 ptype
= gro_find_receive_by_type(type
);
608 skb_gro_pull(skb
, sizeof(*eh
)); /* pull inner eth header */
609 pp
= ptype
->callbacks
.gro_receive(head
, skb
);
614 NAPI_GRO_CB(skb
)->flush
|= flush
;
619 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
)
622 struct packet_offload
*ptype
;
624 int vxlan_len
= sizeof(struct vxlanhdr
) + sizeof(struct ethhdr
);
627 eh
= (struct ethhdr
*)(skb
->data
+ nhoff
+ sizeof(struct vxlanhdr
));
631 ptype
= gro_find_complete_by_type(type
);
633 err
= ptype
->callbacks
.gro_complete(skb
, nhoff
+ vxlan_len
);
639 /* Notify netdevs that UDP port started listening */
640 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
642 struct net_device
*dev
;
643 struct sock
*sk
= vs
->sock
->sk
;
644 struct net
*net
= sock_net(sk
);
645 sa_family_t sa_family
= sk
->sk_family
;
646 __be16 port
= inet_sk(sk
)->inet_sport
;
649 if (sa_family
== AF_INET
) {
650 err
= udp_add_offload(&vs
->udp_offloads
);
652 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
656 for_each_netdev_rcu(net
, dev
) {
657 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
658 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
664 /* Notify netdevs that UDP port is no more listening */
665 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
667 struct net_device
*dev
;
668 struct sock
*sk
= vs
->sock
->sk
;
669 struct net
*net
= sock_net(sk
);
670 sa_family_t sa_family
= sk
->sk_family
;
671 __be16 port
= inet_sk(sk
)->inet_sport
;
674 for_each_netdev_rcu(net
, dev
) {
675 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
676 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
681 if (sa_family
== AF_INET
)
682 udp_del_offload(&vs
->udp_offloads
);
685 /* Add new entry to forwarding table -- assumes lock held */
686 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
687 const u8
*mac
, union vxlan_addr
*ip
,
688 __u16 state
, __u16 flags
,
689 __be16 port
, __u32 vni
, __u32 ifindex
,
692 struct vxlan_rdst
*rd
= NULL
;
696 f
= __vxlan_find_mac(vxlan
, mac
);
698 if (flags
& NLM_F_EXCL
) {
699 netdev_dbg(vxlan
->dev
,
700 "lost race to create %pM\n", mac
);
703 if (f
->state
!= state
) {
705 f
->updated
= jiffies
;
708 if (f
->flags
!= ndm_flags
) {
709 f
->flags
= ndm_flags
;
710 f
->updated
= jiffies
;
713 if ((flags
& NLM_F_REPLACE
)) {
714 /* Only change unicasts */
715 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
716 is_zero_ether_addr(f
->eth_addr
))) {
717 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
726 if ((flags
& NLM_F_APPEND
) &&
727 (is_multicast_ether_addr(f
->eth_addr
) ||
728 is_zero_ether_addr(f
->eth_addr
))) {
729 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
737 if (!(flags
& NLM_F_CREATE
))
740 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
743 /* Disallow replace to add a multicast entry */
744 if ((flags
& NLM_F_REPLACE
) &&
745 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
748 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
749 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
755 f
->flags
= ndm_flags
;
756 f
->updated
= f
->used
= jiffies
;
757 INIT_LIST_HEAD(&f
->remotes
);
758 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
760 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
763 hlist_add_head_rcu(&f
->hlist
,
764 vxlan_fdb_head(vxlan
, mac
));
769 rd
= first_remote_rtnl(f
);
770 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
776 static void vxlan_fdb_free(struct rcu_head
*head
)
778 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
779 struct vxlan_rdst
*rd
, *nd
;
781 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
786 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
788 netdev_dbg(vxlan
->dev
,
789 "delete %pM\n", f
->eth_addr
);
792 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
794 hlist_del_rcu(&f
->hlist
);
795 call_rcu(&f
->rcu
, vxlan_fdb_free
);
798 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
799 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
801 struct net
*net
= dev_net(vxlan
->dev
);
805 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
809 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
810 if (remote
->sa
.sa_family
== AF_INET
) {
811 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
812 ip
->sa
.sa_family
= AF_INET
;
813 #if IS_ENABLED(CONFIG_IPV6)
815 ip
->sin6
.sin6_addr
= in6addr_any
;
816 ip
->sa
.sa_family
= AF_INET6
;
822 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
824 *port
= nla_get_be16(tb
[NDA_PORT
]);
826 *port
= vxlan
->dst_port
;
830 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
832 *vni
= nla_get_u32(tb
[NDA_VNI
]);
834 *vni
= vxlan
->default_dst
.remote_vni
;
837 if (tb
[NDA_IFINDEX
]) {
838 struct net_device
*tdev
;
840 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
842 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
843 tdev
= __dev_get_by_index(net
, *ifindex
);
845 return -EADDRNOTAVAIL
;
853 /* Add static entry (via netlink) */
854 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
855 struct net_device
*dev
,
856 const unsigned char *addr
, u16 flags
)
858 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
859 /* struct net *net = dev_net(vxlan->dev); */
865 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
866 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
871 if (tb
[NDA_DST
] == NULL
)
874 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
878 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
879 return -EAFNOSUPPORT
;
881 spin_lock_bh(&vxlan
->hash_lock
);
882 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
883 port
, vni
, ifindex
, ndm
->ndm_flags
);
884 spin_unlock_bh(&vxlan
->hash_lock
);
889 /* Delete entry (via netlink) */
890 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
891 struct net_device
*dev
,
892 const unsigned char *addr
)
894 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
896 struct vxlan_rdst
*rd
= NULL
;
902 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
908 spin_lock_bh(&vxlan
->hash_lock
);
909 f
= vxlan_find_mac(vxlan
, addr
);
913 if (!vxlan_addr_any(&ip
)) {
914 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
921 /* remove a destination if it's not the only one on the list,
922 * otherwise destroy the fdb entry
924 if (rd
&& !list_is_singular(&f
->remotes
)) {
925 list_del_rcu(&rd
->list
);
926 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
931 vxlan_fdb_destroy(vxlan
, f
);
934 spin_unlock_bh(&vxlan
->hash_lock
);
939 /* Dump forwarding table */
940 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
941 struct net_device
*dev
, int idx
)
943 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
946 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
950 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
951 struct vxlan_rdst
*rd
;
953 if (idx
< cb
->args
[0])
956 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
957 err
= vxlan_fdb_info(skb
, vxlan
, f
,
958 NETLINK_CB(cb
->skb
).portid
,
973 /* Watch incoming packets to learn mapping between Ethernet address
974 * and Tunnel endpoint.
975 * Return true if packet is bogus and should be droppped.
977 static bool vxlan_snoop(struct net_device
*dev
,
978 union vxlan_addr
*src_ip
, const u8
*src_mac
)
980 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
983 f
= vxlan_find_mac(vxlan
, src_mac
);
985 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
987 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
990 /* Don't migrate static entries, drop packets */
991 if (f
->state
& NUD_NOARP
)
996 "%pM migrated from %pIS to %pIS\n",
997 src_mac
, &rdst
->remote_ip
, &src_ip
);
999 rdst
->remote_ip
= *src_ip
;
1000 f
->updated
= jiffies
;
1001 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1003 /* learned new entry */
1004 spin_lock(&vxlan
->hash_lock
);
1006 /* close off race between vxlan_flush and incoming packets */
1007 if (netif_running(dev
))
1008 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1010 NLM_F_EXCL
|NLM_F_CREATE
,
1012 vxlan
->default_dst
.remote_vni
,
1014 spin_unlock(&vxlan
->hash_lock
);
1020 /* See if multicast group is already in use by other ID */
1021 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1023 struct vxlan_dev
*vxlan
;
1025 /* The vxlan_sock is only used by dev, leaving group has
1026 * no effect on other vxlan devices.
1028 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1031 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1032 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1035 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1038 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1039 &dev
->default_dst
.remote_ip
))
1042 if (vxlan
->default_dst
.remote_ifindex
!=
1043 dev
->default_dst
.remote_ifindex
)
1052 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1054 atomic_inc(&vs
->refcnt
);
1057 void vxlan_sock_release(struct vxlan_sock
*vs
)
1059 struct sock
*sk
= vs
->sock
->sk
;
1060 struct net
*net
= sock_net(sk
);
1061 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1063 if (!atomic_dec_and_test(&vs
->refcnt
))
1066 spin_lock(&vn
->sock_lock
);
1067 hlist_del_rcu(&vs
->hlist
);
1068 rcu_assign_sk_user_data(vs
->sock
->sk
, NULL
);
1069 vxlan_notify_del_rx_port(vs
);
1070 spin_unlock(&vn
->sock_lock
);
1072 queue_work(vxlan_wq
, &vs
->del_work
);
1074 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1076 /* Callback to update multicast group membership when first VNI on
1077 * multicast asddress is brought up
1078 * Done as workqueue because ip_mc_join_group acquires RTNL.
1080 static void vxlan_igmp_join(struct work_struct
*work
)
1082 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1083 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1084 struct sock
*sk
= vs
->sock
->sk
;
1085 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1086 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1089 if (ip
->sa
.sa_family
== AF_INET
) {
1090 struct ip_mreqn mreq
= {
1091 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1092 .imr_ifindex
= ifindex
,
1095 ip_mc_join_group(sk
, &mreq
);
1096 #if IS_ENABLED(CONFIG_IPV6)
1098 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1099 &ip
->sin6
.sin6_addr
);
1104 vxlan_sock_release(vs
);
1105 dev_put(vxlan
->dev
);
1108 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1109 static void vxlan_igmp_leave(struct work_struct
*work
)
1111 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1112 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1113 struct sock
*sk
= vs
->sock
->sk
;
1114 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1115 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1118 if (ip
->sa
.sa_family
== AF_INET
) {
1119 struct ip_mreqn mreq
= {
1120 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1121 .imr_ifindex
= ifindex
,
1124 ip_mc_leave_group(sk
, &mreq
);
1125 #if IS_ENABLED(CONFIG_IPV6)
1127 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1128 &ip
->sin6
.sin6_addr
);
1134 vxlan_sock_release(vs
);
1135 dev_put(vxlan
->dev
);
1138 /* Callback from net/ipv4/udp.c to receive packets */
1139 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1141 struct vxlan_sock
*vs
;
1142 struct vxlanhdr
*vxh
;
1144 /* Need Vxlan and inner Ethernet header to be present */
1145 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1148 /* Return packets with reserved bits set */
1149 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1150 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
1151 (vxh
->vx_vni
& htonl(0xff))) {
1152 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1153 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1157 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1160 vs
= rcu_dereference_sk_user_data(sk
);
1164 /* If the NIC driver gave us an encapsulated packet
1165 * with the encapsulation mark, the device checksummed it
1166 * for us. Otherwise force the upper layers to verify it.
1168 if ((skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
) ||
1169 !skb
->encapsulation
)
1170 skb
->ip_summed
= CHECKSUM_NONE
;
1172 skb
->encapsulation
= 0;
1174 vs
->rcv(vs
, skb
, vxh
->vx_vni
);
1178 /* Consume bad packet */
1183 /* Return non vxlan pkt */
1187 static void vxlan_rcv(struct vxlan_sock
*vs
,
1188 struct sk_buff
*skb
, __be32 vx_vni
)
1190 struct iphdr
*oip
= NULL
;
1191 struct ipv6hdr
*oip6
= NULL
;
1192 struct vxlan_dev
*vxlan
;
1193 struct pcpu_sw_netstats
*stats
;
1194 union vxlan_addr saddr
;
1197 union vxlan_addr
*remote_ip
;
1199 vni
= ntohl(vx_vni
) >> 8;
1200 /* Is this VNI defined? */
1201 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1205 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1206 skb_reset_mac_header(skb
);
1207 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1208 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1210 /* Ignore packet loops (and multicast echo) */
1211 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1214 /* Re-examine inner Ethernet packet */
1215 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1217 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1218 saddr
.sa
.sa_family
= AF_INET
;
1219 #if IS_ENABLED(CONFIG_IPV6)
1221 oip6
= ipv6_hdr(skb
);
1222 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1223 saddr
.sa
.sa_family
= AF_INET6
;
1227 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1228 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1231 skb_reset_network_header(skb
);
1234 err
= IP6_ECN_decapsulate(oip6
, skb
);
1236 err
= IP_ECN_decapsulate(oip
, skb
);
1238 if (unlikely(err
)) {
1239 if (log_ecn_error
) {
1241 net_info_ratelimited("non-ECT from %pI6\n",
1244 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1245 &oip
->saddr
, oip
->tos
);
1248 ++vxlan
->dev
->stats
.rx_frame_errors
;
1249 ++vxlan
->dev
->stats
.rx_errors
;
1254 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1255 u64_stats_update_begin(&stats
->syncp
);
1256 stats
->rx_packets
++;
1257 stats
->rx_bytes
+= skb
->len
;
1258 u64_stats_update_end(&stats
->syncp
);
1264 /* Consume bad packet */
1268 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1270 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1271 struct arphdr
*parp
;
1274 struct neighbour
*n
;
1276 if (dev
->flags
& IFF_NOARP
)
1279 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1280 dev
->stats
.tx_dropped
++;
1283 parp
= arp_hdr(skb
);
1285 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1286 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1287 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1288 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1289 parp
->ar_hln
!= dev
->addr_len
||
1292 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1294 arpptr
+= dev
->addr_len
; /* sha */
1295 memcpy(&sip
, arpptr
, sizeof(sip
));
1296 arpptr
+= sizeof(sip
);
1297 arpptr
+= dev
->addr_len
; /* tha */
1298 memcpy(&tip
, arpptr
, sizeof(tip
));
1300 if (ipv4_is_loopback(tip
) ||
1301 ipv4_is_multicast(tip
))
1304 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1307 struct vxlan_fdb
*f
;
1308 struct sk_buff
*reply
;
1310 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1315 f
= vxlan_find_mac(vxlan
, n
->ha
);
1316 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1317 /* bridge-local neighbor */
1322 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1330 skb_reset_mac_header(reply
);
1331 __skb_pull(reply
, skb_network_offset(reply
));
1332 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1333 reply
->pkt_type
= PACKET_HOST
;
1335 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1336 dev
->stats
.rx_dropped
++;
1337 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1338 union vxlan_addr ipa
= {
1339 .sin
.sin_addr
.s_addr
= tip
,
1340 .sa
.sa_family
= AF_INET
,
1343 vxlan_ip_miss(dev
, &ipa
);
1347 return NETDEV_TX_OK
;
1350 #if IS_ENABLED(CONFIG_IPV6)
1352 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1353 struct neighbour
*n
, bool isrouter
)
1355 struct net_device
*dev
= request
->dev
;
1356 struct sk_buff
*reply
;
1357 struct nd_msg
*ns
, *na
;
1358 struct ipv6hdr
*pip6
;
1360 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1367 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1368 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1369 reply
= alloc_skb(len
, GFP_ATOMIC
);
1373 reply
->protocol
= htons(ETH_P_IPV6
);
1375 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1376 skb_push(reply
, sizeof(struct ethhdr
));
1377 skb_set_mac_header(reply
, 0);
1379 ns
= (struct nd_msg
*)skb_transport_header(request
);
1381 daddr
= eth_hdr(request
)->h_source
;
1382 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1383 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1384 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1385 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1390 /* Ethernet header */
1391 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1392 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1393 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1394 reply
->protocol
= htons(ETH_P_IPV6
);
1396 skb_pull(reply
, sizeof(struct ethhdr
));
1397 skb_set_network_header(reply
, 0);
1398 skb_put(reply
, sizeof(struct ipv6hdr
));
1402 pip6
= ipv6_hdr(reply
);
1403 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1405 pip6
->priority
= ipv6_hdr(request
)->priority
;
1406 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1407 pip6
->hop_limit
= 255;
1408 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1409 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1411 skb_pull(reply
, sizeof(struct ipv6hdr
));
1412 skb_set_transport_header(reply
, 0);
1414 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1416 /* Neighbor Advertisement */
1417 memset(na
, 0, sizeof(*na
)+na_olen
);
1418 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1419 na
->icmph
.icmp6_router
= isrouter
;
1420 na
->icmph
.icmp6_override
= 1;
1421 na
->icmph
.icmp6_solicited
= 1;
1422 na
->target
= ns
->target
;
1423 ether_addr_copy(&na
->opt
[2], n
->ha
);
1424 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1425 na
->opt
[1] = na_olen
>> 3;
1427 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1428 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1429 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1431 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1433 skb_push(reply
, sizeof(struct ipv6hdr
));
1435 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1440 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1442 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1444 const struct ipv6hdr
*iphdr
;
1445 const struct in6_addr
*saddr
, *daddr
;
1446 struct neighbour
*n
;
1447 struct inet6_dev
*in6_dev
;
1449 in6_dev
= __in6_dev_get(dev
);
1453 if (!pskb_may_pull(skb
, skb
->len
))
1456 iphdr
= ipv6_hdr(skb
);
1457 saddr
= &iphdr
->saddr
;
1458 daddr
= &iphdr
->daddr
;
1460 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1461 if (msg
->icmph
.icmp6_code
!= 0 ||
1462 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1465 if (ipv6_addr_loopback(daddr
) ||
1466 ipv6_addr_is_multicast(&msg
->target
))
1469 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1472 struct vxlan_fdb
*f
;
1473 struct sk_buff
*reply
;
1475 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1480 f
= vxlan_find_mac(vxlan
, n
->ha
);
1481 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1482 /* bridge-local neighbor */
1487 reply
= vxlan_na_create(skb
, n
,
1488 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1495 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1496 dev
->stats
.rx_dropped
++;
1498 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1499 union vxlan_addr ipa
= {
1500 .sin6
.sin6_addr
= msg
->target
,
1501 .sa
.sa_family
= AF_INET6
,
1504 vxlan_ip_miss(dev
, &ipa
);
1509 return NETDEV_TX_OK
;
1513 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1515 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1516 struct neighbour
*n
;
1518 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1522 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1527 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1530 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1531 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1532 union vxlan_addr ipa
= {
1533 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1534 .sa
.sa_family
= AF_INET
,
1537 vxlan_ip_miss(dev
, &ipa
);
1543 #if IS_ENABLED(CONFIG_IPV6)
1546 struct ipv6hdr
*pip6
;
1548 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1550 pip6
= ipv6_hdr(skb
);
1551 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1552 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1553 union vxlan_addr ipa
= {
1554 .sin6
.sin6_addr
= pip6
->daddr
,
1555 .sa
.sa_family
= AF_INET6
,
1558 vxlan_ip_miss(dev
, &ipa
);
1572 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1574 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1576 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1585 /* Compute source port for outgoing packet
1586 * first choice to use L4 flow hash since it will spread
1587 * better and maybe available from hardware
1588 * secondary choice is to use jhash on the Ethernet header
1590 __be16
vxlan_src_port(__u16 port_min
, __u16 port_max
, struct sk_buff
*skb
)
1592 unsigned int range
= (port_max
- port_min
) + 1;
1595 hash
= skb_get_hash(skb
);
1597 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
1598 (__force u32
) skb
->protocol
);
1600 return htons((((u64
) hash
* range
) >> 32) + port_min
);
1602 EXPORT_SYMBOL_GPL(vxlan_src_port
);
1604 static int handle_offloads(struct sk_buff
*skb
)
1606 if (skb_is_gso(skb
)) {
1607 int err
= skb_unclone(skb
, GFP_ATOMIC
);
1611 skb_shinfo(skb
)->gso_type
|= SKB_GSO_UDP_TUNNEL
;
1612 } else if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
1613 skb
->ip_summed
= CHECKSUM_NONE
;
1618 #if IS_ENABLED(CONFIG_IPV6)
1619 static int vxlan6_xmit_skb(struct vxlan_sock
*vs
,
1620 struct dst_entry
*dst
, struct sk_buff
*skb
,
1621 struct net_device
*dev
, struct in6_addr
*saddr
,
1622 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1623 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1626 struct ipv6hdr
*ip6h
;
1627 struct vxlanhdr
*vxh
;
1632 if (!skb
->encapsulation
) {
1633 skb_reset_inner_headers(skb
);
1634 skb
->encapsulation
= 1;
1637 skb_scrub_packet(skb
, xnet
);
1639 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1640 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1641 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1643 /* Need space for new headers (invalidates iph ptr) */
1644 err
= skb_cow_head(skb
, min_headroom
);
1648 if (vlan_tx_tag_present(skb
)) {
1649 if (WARN_ON(!__vlan_put_tag(skb
,
1651 vlan_tx_tag_get(skb
))))
1657 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1658 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1661 __skb_push(skb
, sizeof(*uh
));
1662 skb_reset_transport_header(skb
);
1665 uh
->dest
= dst_port
;
1666 uh
->source
= src_port
;
1668 uh
->len
= htons(skb
->len
);
1671 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1672 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1674 skb_dst_set(skb
, dst
);
1676 if (!skb_is_gso(skb
) && !(dst
->dev
->features
& NETIF_F_IPV6_CSUM
)) {
1677 __wsum csum
= skb_checksum(skb
, 0, skb
->len
, 0);
1678 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1679 uh
->check
= csum_ipv6_magic(saddr
, daddr
, skb
->len
,
1682 uh
->check
= CSUM_MANGLED_0
;
1684 skb
->ip_summed
= CHECKSUM_PARTIAL
;
1685 skb
->csum_start
= skb_transport_header(skb
) - skb
->head
;
1686 skb
->csum_offset
= offsetof(struct udphdr
, check
);
1687 uh
->check
= ~csum_ipv6_magic(saddr
, daddr
,
1688 skb
->len
, IPPROTO_UDP
, 0);
1691 __skb_push(skb
, sizeof(*ip6h
));
1692 skb_reset_network_header(skb
);
1693 ip6h
= ipv6_hdr(skb
);
1695 ip6h
->priority
= prio
;
1696 ip6h
->flow_lbl
[0] = 0;
1697 ip6h
->flow_lbl
[1] = 0;
1698 ip6h
->flow_lbl
[2] = 0;
1699 ip6h
->payload_len
= htons(skb
->len
);
1700 ip6h
->nexthdr
= IPPROTO_UDP
;
1701 ip6h
->hop_limit
= ttl
;
1702 ip6h
->daddr
= *daddr
;
1703 ip6h
->saddr
= *saddr
;
1705 err
= handle_offloads(skb
);
1709 ip6tunnel_xmit(skb
, dev
);
1714 int vxlan_xmit_skb(struct vxlan_sock
*vs
,
1715 struct rtable
*rt
, struct sk_buff
*skb
,
1716 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1717 __be16 src_port
, __be16 dst_port
, __be32 vni
, bool xnet
)
1719 struct vxlanhdr
*vxh
;
1724 if (!skb
->encapsulation
) {
1725 skb_reset_inner_headers(skb
);
1726 skb
->encapsulation
= 1;
1729 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1730 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1731 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1733 /* Need space for new headers (invalidates iph ptr) */
1734 err
= skb_cow_head(skb
, min_headroom
);
1738 if (vlan_tx_tag_present(skb
)) {
1739 if (WARN_ON(!__vlan_put_tag(skb
,
1741 vlan_tx_tag_get(skb
))))
1747 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1748 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1751 __skb_push(skb
, sizeof(*uh
));
1752 skb_reset_transport_header(skb
);
1755 uh
->dest
= dst_port
;
1756 uh
->source
= src_port
;
1758 uh
->len
= htons(skb
->len
);
1761 err
= handle_offloads(skb
);
1765 return iptunnel_xmit(vs
->sock
->sk
, rt
, skb
, src
, dst
, IPPROTO_UDP
,
1766 tos
, ttl
, df
, xnet
);
1768 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1770 /* Bypass encapsulation if the destination is local */
1771 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1772 struct vxlan_dev
*dst_vxlan
)
1774 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1775 union vxlan_addr loopback
;
1776 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1778 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1779 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1780 skb
->pkt_type
= PACKET_HOST
;
1781 skb
->encapsulation
= 0;
1782 skb
->dev
= dst_vxlan
->dev
;
1783 __skb_pull(skb
, skb_network_offset(skb
));
1785 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1786 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1787 loopback
.sa
.sa_family
= AF_INET
;
1788 #if IS_ENABLED(CONFIG_IPV6)
1790 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1791 loopback
.sa
.sa_family
= AF_INET6
;
1795 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1796 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1798 u64_stats_update_begin(&tx_stats
->syncp
);
1799 tx_stats
->tx_packets
++;
1800 tx_stats
->tx_bytes
+= skb
->len
;
1801 u64_stats_update_end(&tx_stats
->syncp
);
1803 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1804 u64_stats_update_begin(&rx_stats
->syncp
);
1805 rx_stats
->rx_packets
++;
1806 rx_stats
->rx_bytes
+= skb
->len
;
1807 u64_stats_update_end(&rx_stats
->syncp
);
1809 skb
->dev
->stats
.rx_dropped
++;
1813 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1814 struct vxlan_rdst
*rdst
, bool did_rsc
)
1816 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1817 struct rtable
*rt
= NULL
;
1818 const struct iphdr
*old_iph
;
1820 union vxlan_addr
*dst
;
1821 __be16 src_port
= 0, dst_port
;
1827 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1828 vni
= rdst
->remote_vni
;
1829 dst
= &rdst
->remote_ip
;
1831 if (vxlan_addr_any(dst
)) {
1833 /* short-circuited back to local bridge */
1834 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1840 old_iph
= ip_hdr(skb
);
1843 if (!ttl
&& vxlan_addr_multicast(dst
))
1848 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1850 src_port
= vxlan_src_port(vxlan
->port_min
, vxlan
->port_max
, skb
);
1852 if (dst
->sa
.sa_family
== AF_INET
) {
1853 memset(&fl4
, 0, sizeof(fl4
));
1854 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1855 fl4
.flowi4_tos
= RT_TOS(tos
);
1856 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1857 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1859 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1861 netdev_dbg(dev
, "no route to %pI4\n",
1862 &dst
->sin
.sin_addr
.s_addr
);
1863 dev
->stats
.tx_carrier_errors
++;
1867 if (rt
->dst
.dev
== dev
) {
1868 netdev_dbg(dev
, "circular route to %pI4\n",
1869 &dst
->sin
.sin_addr
.s_addr
);
1870 dev
->stats
.collisions
++;
1874 /* Bypass encapsulation if the destination is local */
1875 if (rt
->rt_flags
& RTCF_LOCAL
&&
1876 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1877 struct vxlan_dev
*dst_vxlan
;
1880 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1883 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1887 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1888 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1890 err
= vxlan_xmit_skb(vxlan
->vn_sock
, rt
, skb
,
1891 fl4
.saddr
, dst
->sin
.sin_addr
.s_addr
,
1892 tos
, ttl
, df
, src_port
, dst_port
,
1894 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1898 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1899 #if IS_ENABLED(CONFIG_IPV6)
1901 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1902 struct dst_entry
*ndst
;
1906 memset(&fl6
, 0, sizeof(fl6
));
1907 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
1908 fl6
.daddr
= dst
->sin6
.sin6_addr
;
1909 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
1910 fl6
.flowi6_proto
= IPPROTO_UDP
;
1912 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
1913 netdev_dbg(dev
, "no route to %pI6\n",
1914 &dst
->sin6
.sin6_addr
);
1915 dev
->stats
.tx_carrier_errors
++;
1919 if (ndst
->dev
== dev
) {
1920 netdev_dbg(dev
, "circular route to %pI6\n",
1921 &dst
->sin6
.sin6_addr
);
1923 dev
->stats
.collisions
++;
1927 /* Bypass encapsulation if the destination is local */
1928 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
1929 if (flags
& RTF_LOCAL
&&
1930 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1931 struct vxlan_dev
*dst_vxlan
;
1934 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1937 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1941 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
1943 err
= vxlan6_xmit_skb(vxlan
->vn_sock
, ndst
, skb
,
1944 dev
, &fl6
.saddr
, &fl6
.daddr
, 0, ttl
,
1945 src_port
, dst_port
, htonl(vni
<< 8),
1946 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1953 dev
->stats
.tx_dropped
++;
1959 dev
->stats
.tx_errors
++;
1964 /* Transmit local packets over Vxlan
1966 * Outer IP header inherits ECN and DF from inner header.
1967 * Outer UDP destination is the VXLAN assigned port.
1968 * source port is based on hash of flow
1970 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1972 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1974 bool did_rsc
= false;
1975 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
1976 struct vxlan_fdb
*f
;
1978 skb_reset_mac_header(skb
);
1981 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
1982 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
1983 return arp_reduce(dev
, skb
);
1984 #if IS_ENABLED(CONFIG_IPV6)
1985 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
1986 skb
->len
>= sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
) &&
1987 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
1990 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1991 if (msg
->icmph
.icmp6_code
== 0 &&
1992 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
1993 return neigh_reduce(dev
, skb
);
1998 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2001 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2002 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2003 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2004 did_rsc
= route_shortcircuit(dev
, skb
);
2006 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2010 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2012 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2013 !is_multicast_ether_addr(eth
->h_dest
))
2014 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2016 dev
->stats
.tx_dropped
++;
2018 return NETDEV_TX_OK
;
2022 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2023 struct sk_buff
*skb1
;
2029 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2031 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2035 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2038 return NETDEV_TX_OK
;
2041 /* Walk the forwarding table and purge stale entries */
2042 static void vxlan_cleanup(unsigned long arg
)
2044 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2045 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2048 if (!netif_running(vxlan
->dev
))
2051 spin_lock_bh(&vxlan
->hash_lock
);
2052 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2053 struct hlist_node
*p
, *n
;
2054 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2056 = container_of(p
, struct vxlan_fdb
, hlist
);
2057 unsigned long timeout
;
2059 if (f
->state
& NUD_PERMANENT
)
2062 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2063 if (time_before_eq(timeout
, jiffies
)) {
2064 netdev_dbg(vxlan
->dev
,
2065 "garbage collect %pM\n",
2067 f
->state
= NUD_STALE
;
2068 vxlan_fdb_destroy(vxlan
, f
);
2069 } else if (time_before(timeout
, next_timer
))
2070 next_timer
= timeout
;
2073 spin_unlock_bh(&vxlan
->hash_lock
);
2075 mod_timer(&vxlan
->age_timer
, next_timer
);
2078 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2080 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2082 vxlan
->vn_sock
= vs
;
2083 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2086 /* Setup stats when device is created */
2087 static int vxlan_init(struct net_device
*dev
)
2089 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2090 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2091 struct vxlan_sock
*vs
;
2093 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2097 spin_lock(&vn
->sock_lock
);
2098 vs
= vxlan_find_sock(vxlan
->net
, vxlan
->dst_port
);
2100 /* If we have a socket with same port already, reuse it */
2101 atomic_inc(&vs
->refcnt
);
2102 vxlan_vs_add_dev(vs
, vxlan
);
2104 /* otherwise make new socket outside of RTNL */
2106 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2108 spin_unlock(&vn
->sock_lock
);
2113 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2115 struct vxlan_fdb
*f
;
2117 spin_lock_bh(&vxlan
->hash_lock
);
2118 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2120 vxlan_fdb_destroy(vxlan
, f
);
2121 spin_unlock_bh(&vxlan
->hash_lock
);
2124 static void vxlan_uninit(struct net_device
*dev
)
2126 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2127 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2129 vxlan_fdb_delete_default(vxlan
);
2132 vxlan_sock_release(vs
);
2133 free_percpu(dev
->tstats
);
2136 /* Start ageing timer and join group when device is brought up */
2137 static int vxlan_open(struct net_device
*dev
)
2139 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2140 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2142 /* socket hasn't been created */
2146 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2147 vxlan_sock_hold(vs
);
2149 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2152 if (vxlan
->age_interval
)
2153 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2158 /* Purge the forwarding table */
2159 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2163 spin_lock_bh(&vxlan
->hash_lock
);
2164 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2165 struct hlist_node
*p
, *n
;
2166 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2168 = container_of(p
, struct vxlan_fdb
, hlist
);
2169 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2170 if (!is_zero_ether_addr(f
->eth_addr
))
2171 vxlan_fdb_destroy(vxlan
, f
);
2174 spin_unlock_bh(&vxlan
->hash_lock
);
2177 /* Cleanup timer and forwarding table on shutdown */
2178 static int vxlan_stop(struct net_device
*dev
)
2180 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2181 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2182 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2184 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2185 !vxlan_group_used(vn
, vxlan
)) {
2186 vxlan_sock_hold(vs
);
2188 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2191 del_timer_sync(&vxlan
->age_timer
);
2198 /* Stub, nothing needs to be done. */
2199 static void vxlan_set_multicast_list(struct net_device
*dev
)
2203 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2205 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2206 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2207 struct net_device
*lowerdev
;
2210 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2211 if (lowerdev
== NULL
)
2212 return eth_change_mtu(dev
, new_mtu
);
2214 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2215 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2217 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2219 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2226 static const struct net_device_ops vxlan_netdev_ops
= {
2227 .ndo_init
= vxlan_init
,
2228 .ndo_uninit
= vxlan_uninit
,
2229 .ndo_open
= vxlan_open
,
2230 .ndo_stop
= vxlan_stop
,
2231 .ndo_start_xmit
= vxlan_xmit
,
2232 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2233 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2234 .ndo_change_mtu
= vxlan_change_mtu
,
2235 .ndo_validate_addr
= eth_validate_addr
,
2236 .ndo_set_mac_address
= eth_mac_addr
,
2237 .ndo_fdb_add
= vxlan_fdb_add
,
2238 .ndo_fdb_del
= vxlan_fdb_delete
,
2239 .ndo_fdb_dump
= vxlan_fdb_dump
,
2242 /* Info for udev, that this is a virtual tunnel endpoint */
2243 static struct device_type vxlan_type
= {
2247 /* Calls the ndo_add_vxlan_port of the caller in order to
2248 * supply the listening VXLAN udp ports. Callers are expected
2249 * to implement the ndo_add_vxlan_port.
2251 void vxlan_get_rx_port(struct net_device
*dev
)
2253 struct vxlan_sock
*vs
;
2254 struct net
*net
= dev_net(dev
);
2255 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2256 sa_family_t sa_family
;
2260 spin_lock(&vn
->sock_lock
);
2261 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2262 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2263 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2264 sa_family
= vs
->sock
->sk
->sk_family
;
2265 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2269 spin_unlock(&vn
->sock_lock
);
2271 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2273 /* Initialize the device structure. */
2274 static void vxlan_setup(struct net_device
*dev
)
2276 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2280 eth_hw_addr_random(dev
);
2282 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2283 dev
->hard_header_len
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2285 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
2287 dev
->netdev_ops
= &vxlan_netdev_ops
;
2288 dev
->destructor
= free_netdev
;
2289 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2291 dev
->tx_queue_len
= 0;
2292 dev
->features
|= NETIF_F_LLTX
;
2293 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2294 dev
->features
|= NETIF_F_RXCSUM
;
2295 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2297 dev
->vlan_features
= dev
->features
;
2298 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2299 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2300 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2301 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2302 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
2303 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2305 INIT_LIST_HEAD(&vxlan
->next
);
2306 spin_lock_init(&vxlan
->hash_lock
);
2307 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2308 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2309 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2311 init_timer_deferrable(&vxlan
->age_timer
);
2312 vxlan
->age_timer
.function
= vxlan_cleanup
;
2313 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2315 inet_get_local_port_range(dev_net(dev
), &low
, &high
);
2316 vxlan
->port_min
= low
;
2317 vxlan
->port_max
= high
;
2318 vxlan
->dst_port
= htons(vxlan_port
);
2322 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2323 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2326 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2327 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2328 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2329 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2330 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2331 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2332 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2333 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2334 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2335 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2336 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2337 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2338 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2339 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2340 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2341 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2342 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2343 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2346 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2348 if (tb
[IFLA_ADDRESS
]) {
2349 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2350 pr_debug("invalid link address (not ethernet)\n");
2354 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2355 pr_debug("invalid all zero ethernet address\n");
2356 return -EADDRNOTAVAIL
;
2363 if (data
[IFLA_VXLAN_ID
]) {
2364 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2365 if (id
>= VXLAN_VID_MASK
)
2369 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2370 const struct ifla_vxlan_port_range
*p
2371 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2373 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2374 pr_debug("port range %u .. %u not valid\n",
2375 ntohs(p
->low
), ntohs(p
->high
));
2383 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2384 struct ethtool_drvinfo
*drvinfo
)
2386 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2387 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2390 static const struct ethtool_ops vxlan_ethtool_ops
= {
2391 .get_drvinfo
= vxlan_get_drvinfo
,
2392 .get_link
= ethtool_op_get_link
,
2395 static void vxlan_del_work(struct work_struct
*work
)
2397 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2399 sk_release_kernel(vs
->sock
->sk
);
2403 #if IS_ENABLED(CONFIG_IPV6)
2404 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2405 * could be used for both IPv4 and IPv6 communications, but
2406 * users may set bindv6only=1.
2408 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
)
2411 struct socket
*sock
;
2412 struct sockaddr_in6 vxlan_addr
= {
2413 .sin6_family
= AF_INET6
,
2418 rc
= sock_create_kern(AF_INET6
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2420 pr_debug("UDPv6 socket create failed\n");
2424 /* Put in proper namespace */
2426 sk_change_net(sk
, net
);
2428 kernel_setsockopt(sock
, SOL_IPV6
, IPV6_V6ONLY
,
2429 (char *)&val
, sizeof(val
));
2430 rc
= kernel_bind(sock
, (struct sockaddr
*)&vxlan_addr
,
2431 sizeof(struct sockaddr_in6
));
2433 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2434 &vxlan_addr
.sin6_addr
, ntohs(vxlan_addr
.sin6_port
), rc
);
2435 sk_release_kernel(sk
);
2438 /* At this point, IPv6 module should have been loaded in
2439 * sock_create_kern().
2443 /* Disable multicast loopback */
2444 inet_sk(sk
)->mc_loop
= 0;
2450 static struct socket
*create_v6_sock(struct net
*net
, __be16 port
)
2452 return ERR_PTR(-EPFNOSUPPORT
);
2456 static struct socket
*create_v4_sock(struct net
*net
, __be16 port
)
2459 struct socket
*sock
;
2460 struct sockaddr_in vxlan_addr
= {
2461 .sin_family
= AF_INET
,
2462 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
2467 /* Create UDP socket for encapsulation receive. */
2468 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2470 pr_debug("UDP socket create failed\n");
2474 /* Put in proper namespace */
2476 sk_change_net(sk
, net
);
2478 rc
= kernel_bind(sock
, (struct sockaddr
*) &vxlan_addr
,
2479 sizeof(vxlan_addr
));
2481 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2482 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
2483 sk_release_kernel(sk
);
2487 /* Disable multicast loopback */
2488 inet_sk(sk
)->mc_loop
= 0;
2492 /* Create new listen socket if needed */
2493 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2494 vxlan_rcv_t
*rcv
, void *data
, bool ipv6
)
2496 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2497 struct vxlan_sock
*vs
;
2498 struct socket
*sock
;
2502 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2504 return ERR_PTR(-ENOMEM
);
2506 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2507 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2509 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2512 sock
= create_v6_sock(net
, port
);
2514 sock
= create_v4_sock(net
, port
);
2517 return ERR_CAST(sock
);
2522 atomic_set(&vs
->refcnt
, 1);
2525 rcu_assign_sk_user_data(vs
->sock
->sk
, vs
);
2527 /* Initialize the vxlan udp offloads structure */
2528 vs
->udp_offloads
.port
= port
;
2529 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2530 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2532 spin_lock(&vn
->sock_lock
);
2533 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2534 vxlan_notify_add_rx_port(vs
);
2535 spin_unlock(&vn
->sock_lock
);
2537 /* Mark socket as an encapsulation socket. */
2538 udp_sk(sk
)->encap_type
= 1;
2539 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
2540 #if IS_ENABLED(CONFIG_IPV6)
2542 ipv6_stub
->udpv6_encap_enable();
2550 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2551 vxlan_rcv_t
*rcv
, void *data
,
2552 bool no_share
, bool ipv6
)
2554 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2555 struct vxlan_sock
*vs
;
2557 vs
= vxlan_socket_create(net
, port
, rcv
, data
, ipv6
);
2561 if (no_share
) /* Return error if sharing is not allowed. */
2564 spin_lock(&vn
->sock_lock
);
2565 vs
= vxlan_find_sock(net
, port
);
2568 atomic_inc(&vs
->refcnt
);
2570 vs
= ERR_PTR(-EBUSY
);
2572 spin_unlock(&vn
->sock_lock
);
2575 vs
= ERR_PTR(-EINVAL
);
2579 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2581 /* Scheduled at device creation to bind to a socket */
2582 static void vxlan_sock_work(struct work_struct
*work
)
2584 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2585 struct net
*net
= vxlan
->net
;
2586 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2587 __be16 port
= vxlan
->dst_port
;
2588 struct vxlan_sock
*nvs
;
2590 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
& VXLAN_F_IPV6
);
2591 spin_lock(&vn
->sock_lock
);
2593 vxlan_vs_add_dev(nvs
, vxlan
);
2594 spin_unlock(&vn
->sock_lock
);
2596 dev_put(vxlan
->dev
);
2599 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2600 struct nlattr
*tb
[], struct nlattr
*data
[])
2602 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2603 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2604 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2607 bool use_ipv6
= false;
2609 if (!data
[IFLA_VXLAN_ID
])
2612 vxlan
->net
= dev_net(dev
);
2614 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2615 dst
->remote_vni
= vni
;
2617 /* Unless IPv6 is explicitly requested, assume IPv4 */
2618 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2619 if (data
[IFLA_VXLAN_GROUP
]) {
2620 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2621 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2622 if (!IS_ENABLED(CONFIG_IPV6
))
2623 return -EPFNOSUPPORT
;
2625 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2626 sizeof(struct in6_addr
));
2627 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2631 if (data
[IFLA_VXLAN_LOCAL
]) {
2632 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2633 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2634 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2635 if (!IS_ENABLED(CONFIG_IPV6
))
2636 return -EPFNOSUPPORT
;
2638 /* TODO: respect scope id */
2639 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2640 sizeof(struct in6_addr
));
2641 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2645 if (data
[IFLA_VXLAN_LINK
] &&
2646 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2647 struct net_device
*lowerdev
2648 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2651 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2655 #if IS_ENABLED(CONFIG_IPV6)
2657 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2658 if (idev
&& idev
->cnf
.disable_ipv6
) {
2659 pr_info("IPv6 is disabled via sysctl\n");
2662 vxlan
->flags
|= VXLAN_F_IPV6
;
2667 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2669 /* update header length based on lower device */
2670 dev
->hard_header_len
= lowerdev
->hard_header_len
+
2671 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2672 } else if (use_ipv6
)
2673 vxlan
->flags
|= VXLAN_F_IPV6
;
2675 if (data
[IFLA_VXLAN_TOS
])
2676 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2678 if (data
[IFLA_VXLAN_TTL
])
2679 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2681 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2682 vxlan
->flags
|= VXLAN_F_LEARN
;
2684 if (data
[IFLA_VXLAN_AGEING
])
2685 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2687 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2689 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2690 vxlan
->flags
|= VXLAN_F_PROXY
;
2692 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2693 vxlan
->flags
|= VXLAN_F_RSC
;
2695 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2696 vxlan
->flags
|= VXLAN_F_L2MISS
;
2698 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2699 vxlan
->flags
|= VXLAN_F_L3MISS
;
2701 if (data
[IFLA_VXLAN_LIMIT
])
2702 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2704 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2705 const struct ifla_vxlan_port_range
*p
2706 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2707 vxlan
->port_min
= ntohs(p
->low
);
2708 vxlan
->port_max
= ntohs(p
->high
);
2711 if (data
[IFLA_VXLAN_PORT
])
2712 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2714 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
2715 pr_info("duplicate VNI %u\n", vni
);
2719 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2721 /* create an fdb entry for a valid default destination */
2722 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2723 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2724 &vxlan
->default_dst
.remote_ip
,
2725 NUD_REACHABLE
|NUD_PERMANENT
,
2726 NLM_F_EXCL
|NLM_F_CREATE
,
2728 vxlan
->default_dst
.remote_vni
,
2729 vxlan
->default_dst
.remote_ifindex
,
2735 err
= register_netdevice(dev
);
2737 vxlan_fdb_delete_default(vxlan
);
2741 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2746 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2748 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2749 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2751 spin_lock(&vn
->sock_lock
);
2752 if (!hlist_unhashed(&vxlan
->hlist
))
2753 hlist_del_rcu(&vxlan
->hlist
);
2754 spin_unlock(&vn
->sock_lock
);
2756 list_del(&vxlan
->next
);
2757 unregister_netdevice_queue(dev
, head
);
2760 static size_t vxlan_get_size(const struct net_device
*dev
)
2763 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2764 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2765 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2766 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2767 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2768 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2769 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2770 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2771 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2772 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2773 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2774 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2775 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2776 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2777 nla_total_size(sizeof(__be16
))+ /* IFLA_VXLAN_PORT */
2781 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2783 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2784 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2785 struct ifla_vxlan_port_range ports
= {
2786 .low
= htons(vxlan
->port_min
),
2787 .high
= htons(vxlan
->port_max
),
2790 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2791 goto nla_put_failure
;
2793 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2794 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2795 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2796 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2797 goto nla_put_failure
;
2798 #if IS_ENABLED(CONFIG_IPV6)
2800 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2801 &dst
->remote_ip
.sin6
.sin6_addr
))
2802 goto nla_put_failure
;
2807 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2808 goto nla_put_failure
;
2810 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2811 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2812 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2813 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2814 goto nla_put_failure
;
2815 #if IS_ENABLED(CONFIG_IPV6)
2817 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2818 &vxlan
->saddr
.sin6
.sin6_addr
))
2819 goto nla_put_failure
;
2824 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2825 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2826 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2827 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2828 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2829 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2830 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2831 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2832 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2833 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2834 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2835 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2836 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2837 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
))
2838 goto nla_put_failure
;
2840 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2841 goto nla_put_failure
;
2849 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2851 .maxtype
= IFLA_VXLAN_MAX
,
2852 .policy
= vxlan_policy
,
2853 .priv_size
= sizeof(struct vxlan_dev
),
2854 .setup
= vxlan_setup
,
2855 .validate
= vxlan_validate
,
2856 .newlink
= vxlan_newlink
,
2857 .dellink
= vxlan_dellink
,
2858 .get_size
= vxlan_get_size
,
2859 .fill_info
= vxlan_fill_info
,
2862 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2863 struct net_device
*dev
)
2865 struct vxlan_dev
*vxlan
, *next
;
2866 LIST_HEAD(list_kill
);
2868 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2869 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2871 /* In case we created vxlan device with carrier
2872 * and we loose the carrier due to module unload
2873 * we also need to remove vxlan device. In other
2874 * cases, it's not necessary and remote_ifindex
2875 * is 0 here, so no matches.
2877 if (dst
->remote_ifindex
== dev
->ifindex
)
2878 vxlan_dellink(vxlan
->dev
, &list_kill
);
2881 unregister_netdevice_many(&list_kill
);
2884 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2885 unsigned long event
, void *ptr
)
2887 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2888 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2890 if (event
== NETDEV_UNREGISTER
)
2891 vxlan_handle_lowerdev_unregister(vn
, dev
);
2896 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2897 .notifier_call
= vxlan_lowerdev_event
,
2900 static __net_init
int vxlan_init_net(struct net
*net
)
2902 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2905 INIT_LIST_HEAD(&vn
->vxlan_list
);
2906 spin_lock_init(&vn
->sock_lock
);
2908 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2909 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
2914 static void __net_exit
vxlan_exit_net(struct net
*net
)
2916 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2917 struct vxlan_dev
*vxlan
, *next
;
2918 struct net_device
*dev
, *aux
;
2922 for_each_netdev_safe(net
, dev
, aux
)
2923 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
2924 unregister_netdevice_queue(dev
, &list
);
2926 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2927 /* If vxlan->dev is in the same netns, it has already been added
2928 * to the list by the previous loop.
2930 if (!net_eq(dev_net(vxlan
->dev
), net
))
2931 unregister_netdevice_queue(dev
, &list
);
2934 unregister_netdevice_many(&list
);
2938 static struct pernet_operations vxlan_net_ops
= {
2939 .init
= vxlan_init_net
,
2940 .exit
= vxlan_exit_net
,
2941 .id
= &vxlan_net_id
,
2942 .size
= sizeof(struct vxlan_net
),
2945 static int __init
vxlan_init_module(void)
2949 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
2953 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
2955 rc
= register_pernet_subsys(&vxlan_net_ops
);
2959 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
2963 rc
= rtnl_link_register(&vxlan_link_ops
);
2969 unregister_netdevice_notifier(&vxlan_notifier_block
);
2971 unregister_pernet_subsys(&vxlan_net_ops
);
2973 destroy_workqueue(vxlan_wq
);
2976 late_initcall(vxlan_init_module
);
2978 static void __exit
vxlan_cleanup_module(void)
2980 rtnl_link_unregister(&vxlan_link_ops
);
2981 unregister_netdevice_notifier(&vxlan_notifier_block
);
2982 destroy_workqueue(vxlan_wq
);
2983 unregister_pernet_subsys(&vxlan_net_ops
);
2984 /* rcu_barrier() is called by netns */
2986 module_exit(vxlan_cleanup_module
);
2988 MODULE_LICENSE("GPL");
2989 MODULE_VERSION(VXLAN_VERSION
);
2990 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2991 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2992 MODULE_ALIAS_RTNL_LINK("vxlan");