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 #if IS_ENABLED(CONFIG_IPV6)
45 #include <net/addrconf.h>
46 #include <net/ip6_tunnel.h>
47 #include <net/ip6_checksum.h>
50 #define VXLAN_VERSION "0.1"
52 #define PORT_HASH_BITS 8
53 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
54 #define VNI_HASH_BITS 10
55 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
56 #define FDB_HASH_BITS 8
57 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61 #define VXLAN_N_VID (1u << 24)
62 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
63 /* IP header + UDP + VXLAN + Ethernet header */
64 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
65 /* IPv6 header + UDP + VXLAN + Ethernet header */
66 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
67 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
69 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
71 /* VXLAN protocol header */
77 /* UDP port for VXLAN traffic.
78 * The IANA assigned port is 4789, but the Linux default is 8472
79 * for compatibility with early adopters.
81 static unsigned short vxlan_port __read_mostly
= 8472;
82 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
83 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
85 static bool log_ecn_error
= true;
86 module_param(log_ecn_error
, bool, 0644);
87 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
89 static int vxlan_net_id
;
91 static const u8 all_zeros_mac
[ETH_ALEN
];
93 /* per-network namespace private data for this module */
95 struct list_head vxlan_list
;
96 struct hlist_head sock_list
[PORT_HASH_SIZE
];
101 struct sockaddr_in sin
;
102 struct sockaddr_in6 sin6
;
107 union vxlan_addr remote_ip
;
111 struct list_head list
;
115 /* Forwarding table entry */
117 struct hlist_node hlist
; /* linked list of entries */
119 unsigned long updated
; /* jiffies */
121 struct list_head remotes
;
122 u16 state
; /* see ndm_state */
123 u8 flags
; /* see ndm_flags */
124 u8 eth_addr
[ETH_ALEN
];
127 /* Pseudo network device */
129 struct hlist_node hlist
; /* vni hash table */
130 struct list_head next
; /* vxlan's per namespace list */
131 struct vxlan_sock
*vn_sock
; /* listening socket */
132 struct net_device
*dev
;
133 struct vxlan_rdst default_dst
; /* default destination */
134 union vxlan_addr saddr
; /* source address */
136 __u16 port_min
; /* source port range */
138 __u8 tos
; /* TOS override */
140 u32 flags
; /* VXLAN_F_* below */
142 struct work_struct sock_work
;
143 struct work_struct igmp_join
;
144 struct work_struct igmp_leave
;
146 unsigned long age_interval
;
147 struct timer_list age_timer
;
148 spinlock_t hash_lock
;
149 unsigned int addrcnt
;
150 unsigned int addrmax
;
152 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
155 #define VXLAN_F_LEARN 0x01
156 #define VXLAN_F_PROXY 0x02
157 #define VXLAN_F_RSC 0x04
158 #define VXLAN_F_L2MISS 0x08
159 #define VXLAN_F_L3MISS 0x10
160 #define VXLAN_F_IPV6 0x20 /* internal flag */
162 /* salt for hash table */
163 static u32 vxlan_salt __read_mostly
;
164 static struct workqueue_struct
*vxlan_wq
;
166 static void vxlan_sock_work(struct work_struct
*work
);
168 #if IS_ENABLED(CONFIG_IPV6)
170 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
172 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
174 if (a
->sa
.sa_family
== AF_INET6
)
175 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
177 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
180 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
182 if (ipa
->sa
.sa_family
== AF_INET6
)
183 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
185 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
188 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
190 if (ipa
->sa
.sa_family
== AF_INET6
)
191 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
193 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
196 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
198 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
199 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
200 ip
->sa
.sa_family
= AF_INET6
;
202 } else if (nla_len(nla
) >= sizeof(__be32
)) {
203 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
204 ip
->sa
.sa_family
= AF_INET
;
207 return -EAFNOSUPPORT
;
211 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
212 const union vxlan_addr
*ip
)
214 if (ip
->sa
.sa_family
== AF_INET6
)
215 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
217 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
220 #else /* !CONFIG_IPV6 */
223 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
225 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
228 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
230 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
233 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
235 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
238 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
240 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
241 return -EAFNOSUPPORT
;
242 } else if (nla_len(nla
) >= sizeof(__be32
)) {
243 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
244 ip
->sa
.sa_family
= AF_INET
;
247 return -EAFNOSUPPORT
;
251 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
252 const union vxlan_addr
*ip
)
254 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
258 /* Virtual Network hash table head */
259 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
261 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
264 /* Socket hash table head */
265 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
267 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
269 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
272 /* First remote destination for a forwarding entry.
273 * Guaranteed to be non-NULL because remotes are never deleted.
275 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
277 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
280 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
282 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
285 /* Find VXLAN socket based on network namespace and UDP port */
286 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, __be16 port
)
288 struct vxlan_sock
*vs
;
290 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
291 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
297 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
299 struct vxlan_dev
*vxlan
;
301 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
302 if (vxlan
->default_dst
.remote_vni
== id
)
309 /* Look up VNI in a per net namespace table */
310 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
312 struct vxlan_sock
*vs
;
314 vs
= vxlan_find_sock(net
, port
);
318 return vxlan_vs_find_vni(vs
, id
);
321 /* Fill in neighbour message in skbuff. */
322 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
323 const struct vxlan_fdb
*fdb
,
324 u32 portid
, u32 seq
, int type
, unsigned int flags
,
325 const struct vxlan_rdst
*rdst
)
327 unsigned long now
= jiffies
;
328 struct nda_cacheinfo ci
;
329 struct nlmsghdr
*nlh
;
331 bool send_ip
, send_eth
;
333 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
337 ndm
= nlmsg_data(nlh
);
338 memset(ndm
, 0, sizeof(*ndm
));
340 send_eth
= send_ip
= true;
342 if (type
== RTM_GETNEIGH
) {
343 ndm
->ndm_family
= AF_INET
;
344 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
345 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
347 ndm
->ndm_family
= AF_BRIDGE
;
348 ndm
->ndm_state
= fdb
->state
;
349 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
350 ndm
->ndm_flags
= fdb
->flags
;
351 ndm
->ndm_type
= NDA_DST
;
353 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
354 goto nla_put_failure
;
356 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
357 goto nla_put_failure
;
359 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
360 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
361 goto nla_put_failure
;
362 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
363 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
364 goto nla_put_failure
;
365 if (rdst
->remote_ifindex
&&
366 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
367 goto nla_put_failure
;
369 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
370 ci
.ndm_confirmed
= 0;
371 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
374 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
375 goto nla_put_failure
;
377 return nlmsg_end(skb
, nlh
);
380 nlmsg_cancel(skb
, nlh
);
384 static inline size_t vxlan_nlmsg_size(void)
386 return NLMSG_ALIGN(sizeof(struct ndmsg
))
387 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
388 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
389 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
390 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
391 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
392 + nla_total_size(sizeof(struct nda_cacheinfo
));
395 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
,
396 struct vxlan_fdb
*fdb
, int type
)
398 struct net
*net
= dev_net(vxlan
->dev
);
402 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
406 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0,
407 first_remote_rtnl(fdb
));
409 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
410 WARN_ON(err
== -EMSGSIZE
);
415 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
419 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
422 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
424 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
425 struct vxlan_fdb f
= {
428 struct vxlan_rdst remote
= {
429 .remote_ip
= *ipa
, /* goes to NDA_DST */
430 .remote_vni
= VXLAN_N_VID
,
433 INIT_LIST_HEAD(&f
.remotes
);
434 list_add_rcu(&remote
.list
, &f
.remotes
);
436 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
439 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
441 struct vxlan_fdb f
= {
445 INIT_LIST_HEAD(&f
.remotes
);
446 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
448 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
451 /* Hash Ethernet address */
452 static u32
eth_hash(const unsigned char *addr
)
454 u64 value
= get_unaligned((u64
*)addr
);
456 /* only want 6 bytes */
462 return hash_64(value
, FDB_HASH_BITS
);
465 /* Hash chain to use given mac address */
466 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
469 return &vxlan
->fdb_head
[eth_hash(mac
)];
472 /* Look up Ethernet address in forwarding table */
473 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
477 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
480 hlist_for_each_entry_rcu(f
, head
, hlist
) {
481 if (ether_addr_equal(mac
, f
->eth_addr
))
488 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
493 f
= __vxlan_find_mac(vxlan
, mac
);
500 /* caller should hold vxlan->hash_lock */
501 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
502 union vxlan_addr
*ip
, __be16 port
,
503 __u32 vni
, __u32 ifindex
)
505 struct vxlan_rdst
*rd
;
507 list_for_each_entry(rd
, &f
->remotes
, list
) {
508 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
509 rd
->remote_port
== port
&&
510 rd
->remote_vni
== vni
&&
511 rd
->remote_ifindex
== ifindex
)
518 /* Replace destination of unicast mac */
519 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
520 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
522 struct vxlan_rdst
*rd
;
524 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
528 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
532 rd
->remote_port
= port
;
533 rd
->remote_vni
= vni
;
534 rd
->remote_ifindex
= ifindex
;
538 /* Add/update destinations for multicast */
539 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
540 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
542 struct vxlan_rdst
*rd
;
544 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
548 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
552 rd
->remote_port
= port
;
553 rd
->remote_vni
= vni
;
554 rd
->remote_ifindex
= ifindex
;
556 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
561 /* Notify netdevs that UDP port started listening */
562 static void vxlan_notify_add_rx_port(struct sock
*sk
)
564 struct net_device
*dev
;
565 struct net
*net
= sock_net(sk
);
566 sa_family_t sa_family
= sk
->sk_family
;
567 __be16 port
= inet_sk(sk
)->inet_sport
;
570 for_each_netdev_rcu(net
, dev
) {
571 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
572 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
578 /* Notify netdevs that UDP port is no more listening */
579 static void vxlan_notify_del_rx_port(struct sock
*sk
)
581 struct net_device
*dev
;
582 struct net
*net
= sock_net(sk
);
583 sa_family_t sa_family
= sk
->sk_family
;
584 __be16 port
= inet_sk(sk
)->inet_sport
;
587 for_each_netdev_rcu(net
, dev
) {
588 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
589 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
595 /* Add new entry to forwarding table -- assumes lock held */
596 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
597 const u8
*mac
, union vxlan_addr
*ip
,
598 __u16 state
, __u16 flags
,
599 __be16 port
, __u32 vni
, __u32 ifindex
,
605 f
= __vxlan_find_mac(vxlan
, mac
);
607 if (flags
& NLM_F_EXCL
) {
608 netdev_dbg(vxlan
->dev
,
609 "lost race to create %pM\n", mac
);
612 if (f
->state
!= state
) {
614 f
->updated
= jiffies
;
617 if (f
->flags
!= ndm_flags
) {
618 f
->flags
= ndm_flags
;
619 f
->updated
= jiffies
;
622 if ((flags
& NLM_F_REPLACE
)) {
623 /* Only change unicasts */
624 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
625 is_zero_ether_addr(f
->eth_addr
))) {
626 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
635 if ((flags
& NLM_F_APPEND
) &&
636 (is_multicast_ether_addr(f
->eth_addr
) ||
637 is_zero_ether_addr(f
->eth_addr
))) {
638 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
);
645 if (!(flags
& NLM_F_CREATE
))
648 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
651 /* Disallow replace to add a multicast entry */
652 if ((flags
& NLM_F_REPLACE
) &&
653 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
656 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
657 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
663 f
->flags
= ndm_flags
;
664 f
->updated
= f
->used
= jiffies
;
665 INIT_LIST_HEAD(&f
->remotes
);
666 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
668 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
);
671 hlist_add_head_rcu(&f
->hlist
,
672 vxlan_fdb_head(vxlan
, mac
));
676 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
681 static void vxlan_fdb_free(struct rcu_head
*head
)
683 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
684 struct vxlan_rdst
*rd
, *nd
;
686 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
691 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
693 netdev_dbg(vxlan
->dev
,
694 "delete %pM\n", f
->eth_addr
);
697 vxlan_fdb_notify(vxlan
, f
, RTM_DELNEIGH
);
699 hlist_del_rcu(&f
->hlist
);
700 call_rcu(&f
->rcu
, vxlan_fdb_free
);
703 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
704 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
706 struct net
*net
= dev_net(vxlan
->dev
);
710 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
714 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
715 if (remote
->sa
.sa_family
== AF_INET
) {
716 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
717 ip
->sa
.sa_family
= AF_INET
;
718 #if IS_ENABLED(CONFIG_IPV6)
720 ip
->sin6
.sin6_addr
= in6addr_any
;
721 ip
->sa
.sa_family
= AF_INET6
;
727 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
729 *port
= nla_get_be16(tb
[NDA_PORT
]);
731 *port
= vxlan
->dst_port
;
735 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
737 *vni
= nla_get_u32(tb
[NDA_VNI
]);
739 *vni
= vxlan
->default_dst
.remote_vni
;
742 if (tb
[NDA_IFINDEX
]) {
743 struct net_device
*tdev
;
745 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
747 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
748 tdev
= dev_get_by_index(net
, *ifindex
);
750 return -EADDRNOTAVAIL
;
759 /* Add static entry (via netlink) */
760 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
761 struct net_device
*dev
,
762 const unsigned char *addr
, u16 flags
)
764 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
765 /* struct net *net = dev_net(vxlan->dev); */
771 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
772 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
777 if (tb
[NDA_DST
] == NULL
)
780 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
784 spin_lock_bh(&vxlan
->hash_lock
);
785 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
786 port
, vni
, ifindex
, ndm
->ndm_flags
);
787 spin_unlock_bh(&vxlan
->hash_lock
);
792 /* Delete entry (via netlink) */
793 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
794 struct net_device
*dev
,
795 const unsigned char *addr
)
797 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
799 struct vxlan_rdst
*rd
= NULL
;
805 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
811 spin_lock_bh(&vxlan
->hash_lock
);
812 f
= vxlan_find_mac(vxlan
, addr
);
816 if (!vxlan_addr_any(&ip
)) {
817 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
824 /* remove a destination if it's not the only one on the list,
825 * otherwise destroy the fdb entry
827 if (rd
&& !list_is_singular(&f
->remotes
)) {
828 list_del_rcu(&rd
->list
);
833 vxlan_fdb_destroy(vxlan
, f
);
836 spin_unlock_bh(&vxlan
->hash_lock
);
841 /* Dump forwarding table */
842 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
843 struct net_device
*dev
, int idx
)
845 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
848 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
852 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
853 struct vxlan_rdst
*rd
;
855 if (idx
< cb
->args
[0])
858 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
859 err
= vxlan_fdb_info(skb
, vxlan
, f
,
860 NETLINK_CB(cb
->skb
).portid
,
875 /* Watch incoming packets to learn mapping between Ethernet address
876 * and Tunnel endpoint.
877 * Return true if packet is bogus and should be droppped.
879 static bool vxlan_snoop(struct net_device
*dev
,
880 union vxlan_addr
*src_ip
, const u8
*src_mac
)
882 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
885 f
= vxlan_find_mac(vxlan
, src_mac
);
887 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
889 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
892 /* Don't migrate static entries, drop packets */
893 if (f
->state
& NUD_NOARP
)
898 "%pM migrated from %pIS to %pIS\n",
899 src_mac
, &rdst
->remote_ip
, &src_ip
);
901 rdst
->remote_ip
= *src_ip
;
902 f
->updated
= jiffies
;
903 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
905 /* learned new entry */
906 spin_lock(&vxlan
->hash_lock
);
908 /* close off race between vxlan_flush and incoming packets */
909 if (netif_running(dev
))
910 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
912 NLM_F_EXCL
|NLM_F_CREATE
,
914 vxlan
->default_dst
.remote_vni
,
916 spin_unlock(&vxlan
->hash_lock
);
922 /* See if multicast group is already in use by other ID */
923 static bool vxlan_group_used(struct vxlan_net
*vn
, union vxlan_addr
*remote_ip
)
925 struct vxlan_dev
*vxlan
;
927 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
928 if (!netif_running(vxlan
->dev
))
931 if (vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
939 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
941 atomic_inc(&vs
->refcnt
);
944 void vxlan_sock_release(struct vxlan_sock
*vs
)
946 struct sock
*sk
= vs
->sock
->sk
;
947 struct net
*net
= sock_net(sk
);
948 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
950 if (!atomic_dec_and_test(&vs
->refcnt
))
953 spin_lock(&vn
->sock_lock
);
954 hlist_del_rcu(&vs
->hlist
);
955 rcu_assign_sk_user_data(vs
->sock
->sk
, NULL
);
956 vxlan_notify_del_rx_port(sk
);
957 spin_unlock(&vn
->sock_lock
);
959 queue_work(vxlan_wq
, &vs
->del_work
);
961 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
963 /* Callback to update multicast group membership when first VNI on
964 * multicast asddress is brought up
965 * Done as workqueue because ip_mc_join_group acquires RTNL.
967 static void vxlan_igmp_join(struct work_struct
*work
)
969 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
970 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
971 struct sock
*sk
= vs
->sock
->sk
;
972 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
973 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
976 if (ip
->sa
.sa_family
== AF_INET
) {
977 struct ip_mreqn mreq
= {
978 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
979 .imr_ifindex
= ifindex
,
982 ip_mc_join_group(sk
, &mreq
);
983 #if IS_ENABLED(CONFIG_IPV6)
985 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
986 &ip
->sin6
.sin6_addr
);
991 vxlan_sock_release(vs
);
995 /* Inverse of vxlan_igmp_join when last VNI is brought down */
996 static void vxlan_igmp_leave(struct work_struct
*work
)
998 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
999 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1000 struct sock
*sk
= vs
->sock
->sk
;
1001 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1002 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1005 if (ip
->sa
.sa_family
== AF_INET
) {
1006 struct ip_mreqn mreq
= {
1007 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1008 .imr_ifindex
= ifindex
,
1011 ip_mc_leave_group(sk
, &mreq
);
1012 #if IS_ENABLED(CONFIG_IPV6)
1014 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1015 &ip
->sin6
.sin6_addr
);
1021 vxlan_sock_release(vs
);
1022 dev_put(vxlan
->dev
);
1025 /* Callback from net/ipv4/udp.c to receive packets */
1026 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1028 struct vxlan_sock
*vs
;
1029 struct vxlanhdr
*vxh
;
1032 /* Need Vxlan and inner Ethernet header to be present */
1033 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1036 /* Return packets with reserved bits set */
1037 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1038 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
1039 (vxh
->vx_vni
& htonl(0xff))) {
1040 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1041 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1045 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1048 port
= inet_sk(sk
)->inet_sport
;
1050 vs
= rcu_dereference_sk_user_data(sk
);
1054 vs
->rcv(vs
, skb
, vxh
->vx_vni
);
1058 /* Consume bad packet */
1063 /* Return non vxlan pkt */
1067 static void vxlan_rcv(struct vxlan_sock
*vs
,
1068 struct sk_buff
*skb
, __be32 vx_vni
)
1070 struct iphdr
*oip
= NULL
;
1071 struct ipv6hdr
*oip6
= NULL
;
1072 struct vxlan_dev
*vxlan
;
1073 struct pcpu_tstats
*stats
;
1074 union vxlan_addr saddr
;
1077 union vxlan_addr
*remote_ip
;
1079 vni
= ntohl(vx_vni
) >> 8;
1080 /* Is this VNI defined? */
1081 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1085 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1086 skb_reset_mac_header(skb
);
1087 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1089 /* Ignore packet loops (and multicast echo) */
1090 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1093 /* Re-examine inner Ethernet packet */
1094 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1096 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1097 saddr
.sa
.sa_family
= AF_INET
;
1098 #if IS_ENABLED(CONFIG_IPV6)
1100 oip6
= ipv6_hdr(skb
);
1101 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1102 saddr
.sa
.sa_family
= AF_INET6
;
1106 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1107 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1110 skb_reset_network_header(skb
);
1112 /* If the NIC driver gave us an encapsulated packet with
1113 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1114 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1115 * for us. Otherwise force the upper layers to verify it.
1117 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
|| !skb
->encapsulation
||
1118 !(vxlan
->dev
->features
& NETIF_F_RXCSUM
))
1119 skb
->ip_summed
= CHECKSUM_NONE
;
1121 skb
->encapsulation
= 0;
1124 err
= IP6_ECN_decapsulate(oip6
, skb
);
1126 err
= IP_ECN_decapsulate(oip
, skb
);
1128 if (unlikely(err
)) {
1129 if (log_ecn_error
) {
1131 net_info_ratelimited("non-ECT from %pI6\n",
1134 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1135 &oip
->saddr
, oip
->tos
);
1138 ++vxlan
->dev
->stats
.rx_frame_errors
;
1139 ++vxlan
->dev
->stats
.rx_errors
;
1144 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1145 u64_stats_update_begin(&stats
->syncp
);
1146 stats
->rx_packets
++;
1147 stats
->rx_bytes
+= skb
->len
;
1148 u64_stats_update_end(&stats
->syncp
);
1154 /* Consume bad packet */
1158 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1160 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1161 struct arphdr
*parp
;
1164 struct neighbour
*n
;
1166 if (dev
->flags
& IFF_NOARP
)
1169 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1170 dev
->stats
.tx_dropped
++;
1173 parp
= arp_hdr(skb
);
1175 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1176 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1177 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1178 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1179 parp
->ar_hln
!= dev
->addr_len
||
1182 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1184 arpptr
+= dev
->addr_len
; /* sha */
1185 memcpy(&sip
, arpptr
, sizeof(sip
));
1186 arpptr
+= sizeof(sip
);
1187 arpptr
+= dev
->addr_len
; /* tha */
1188 memcpy(&tip
, arpptr
, sizeof(tip
));
1190 if (ipv4_is_loopback(tip
) ||
1191 ipv4_is_multicast(tip
))
1194 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1197 struct vxlan_fdb
*f
;
1198 struct sk_buff
*reply
;
1200 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1205 f
= vxlan_find_mac(vxlan
, n
->ha
);
1206 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1207 /* bridge-local neighbor */
1212 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1217 skb_reset_mac_header(reply
);
1218 __skb_pull(reply
, skb_network_offset(reply
));
1219 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1220 reply
->pkt_type
= PACKET_HOST
;
1222 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1223 dev
->stats
.rx_dropped
++;
1224 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1225 union vxlan_addr ipa
= {
1226 .sin
.sin_addr
.s_addr
= tip
,
1227 .sa
.sa_family
= AF_INET
,
1230 vxlan_ip_miss(dev
, &ipa
);
1234 return NETDEV_TX_OK
;
1237 #if IS_ENABLED(CONFIG_IPV6)
1238 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1240 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1241 struct neighbour
*n
;
1242 union vxlan_addr ipa
;
1243 const struct ipv6hdr
*iphdr
;
1244 const struct in6_addr
*saddr
, *daddr
;
1246 struct inet6_dev
*in6_dev
= NULL
;
1248 in6_dev
= __in6_dev_get(dev
);
1252 if (!pskb_may_pull(skb
, skb
->len
))
1255 iphdr
= ipv6_hdr(skb
);
1256 saddr
= &iphdr
->saddr
;
1257 daddr
= &iphdr
->daddr
;
1259 if (ipv6_addr_loopback(daddr
) ||
1260 ipv6_addr_is_multicast(daddr
))
1263 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1264 if (msg
->icmph
.icmp6_code
!= 0 ||
1265 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1268 n
= neigh_lookup(ipv6_stub
->nd_tbl
, daddr
, dev
);
1271 struct vxlan_fdb
*f
;
1273 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1278 f
= vxlan_find_mac(vxlan
, n
->ha
);
1279 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1280 /* bridge-local neighbor */
1285 ipv6_stub
->ndisc_send_na(dev
, n
, saddr
, &msg
->target
,
1286 !!in6_dev
->cnf
.forwarding
,
1287 true, false, false);
1289 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1290 ipa
.sin6
.sin6_addr
= *daddr
;
1291 ipa
.sa
.sa_family
= AF_INET6
;
1292 vxlan_ip_miss(dev
, &ipa
);
1297 return NETDEV_TX_OK
;
1301 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1303 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1304 struct neighbour
*n
;
1306 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1310 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1315 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1318 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1319 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1320 union vxlan_addr ipa
= {
1321 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1322 .sa
.sa_family
= AF_INET
,
1325 vxlan_ip_miss(dev
, &ipa
);
1331 #if IS_ENABLED(CONFIG_IPV6)
1334 struct ipv6hdr
*pip6
;
1336 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1338 pip6
= ipv6_hdr(skb
);
1339 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1340 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1341 union vxlan_addr ipa
= {
1342 .sin6
.sin6_addr
= pip6
->daddr
,
1343 .sa
.sa_family
= AF_INET6
,
1346 vxlan_ip_miss(dev
, &ipa
);
1360 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1362 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1364 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1373 static void vxlan_sock_put(struct sk_buff
*skb
)
1378 /* On transmit, associate with the tunnel socket */
1379 static void vxlan_set_owner(struct sock
*sk
, struct sk_buff
*skb
)
1384 skb
->destructor
= vxlan_sock_put
;
1387 /* Compute source port for outgoing packet
1388 * first choice to use L4 flow hash since it will spread
1389 * better and maybe available from hardware
1390 * secondary choice is to use jhash on the Ethernet header
1392 __be16
vxlan_src_port(__u16 port_min
, __u16 port_max
, struct sk_buff
*skb
)
1394 unsigned int range
= (port_max
- port_min
) + 1;
1397 hash
= skb_get_rxhash(skb
);
1399 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
1400 (__force u32
) skb
->protocol
);
1402 return htons((((u64
) hash
* range
) >> 32) + port_min
);
1404 EXPORT_SYMBOL_GPL(vxlan_src_port
);
1406 static int handle_offloads(struct sk_buff
*skb
)
1408 if (skb_is_gso(skb
)) {
1409 int err
= skb_unclone(skb
, GFP_ATOMIC
);
1413 skb_shinfo(skb
)->gso_type
|= SKB_GSO_UDP_TUNNEL
;
1414 } else if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
1415 skb
->ip_summed
= CHECKSUM_NONE
;
1420 #if IS_ENABLED(CONFIG_IPV6)
1421 static int vxlan6_xmit_skb(struct vxlan_sock
*vs
,
1422 struct dst_entry
*dst
, struct sk_buff
*skb
,
1423 struct net_device
*dev
, struct in6_addr
*saddr
,
1424 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1425 __be16 src_port
, __be16 dst_port
, __be32 vni
)
1427 struct ipv6hdr
*ip6h
;
1428 struct vxlanhdr
*vxh
;
1433 if (!skb
->encapsulation
) {
1434 skb_reset_inner_headers(skb
);
1435 skb
->encapsulation
= 1;
1438 skb_scrub_packet(skb
, false);
1440 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1441 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1442 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1444 /* Need space for new headers (invalidates iph ptr) */
1445 err
= skb_cow_head(skb
, min_headroom
);
1449 if (vlan_tx_tag_present(skb
)) {
1450 if (WARN_ON(!__vlan_put_tag(skb
,
1452 vlan_tx_tag_get(skb
))))
1458 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1459 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1462 __skb_push(skb
, sizeof(*uh
));
1463 skb_reset_transport_header(skb
);
1466 uh
->dest
= dst_port
;
1467 uh
->source
= src_port
;
1469 uh
->len
= htons(skb
->len
);
1472 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1473 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1475 skb_dst_set(skb
, dst
);
1477 if (!skb_is_gso(skb
) && !(dst
->dev
->features
& NETIF_F_IPV6_CSUM
)) {
1478 __wsum csum
= skb_checksum(skb
, 0, skb
->len
, 0);
1479 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1480 uh
->check
= csum_ipv6_magic(saddr
, daddr
, skb
->len
,
1483 uh
->check
= CSUM_MANGLED_0
;
1485 skb
->ip_summed
= CHECKSUM_PARTIAL
;
1486 skb
->csum_start
= skb_transport_header(skb
) - skb
->head
;
1487 skb
->csum_offset
= offsetof(struct udphdr
, check
);
1488 uh
->check
= ~csum_ipv6_magic(saddr
, daddr
,
1489 skb
->len
, IPPROTO_UDP
, 0);
1492 __skb_push(skb
, sizeof(*ip6h
));
1493 skb_reset_network_header(skb
);
1494 ip6h
= ipv6_hdr(skb
);
1496 ip6h
->priority
= prio
;
1497 ip6h
->flow_lbl
[0] = 0;
1498 ip6h
->flow_lbl
[1] = 0;
1499 ip6h
->flow_lbl
[2] = 0;
1500 ip6h
->payload_len
= htons(skb
->len
);
1501 ip6h
->nexthdr
= IPPROTO_UDP
;
1502 ip6h
->hop_limit
= ttl
;
1503 ip6h
->daddr
= *daddr
;
1504 ip6h
->saddr
= *saddr
;
1506 vxlan_set_owner(vs
->sock
->sk
, skb
);
1508 err
= handle_offloads(skb
);
1512 ip6tunnel_xmit(skb
, dev
);
1517 int vxlan_xmit_skb(struct vxlan_sock
*vs
,
1518 struct rtable
*rt
, struct sk_buff
*skb
,
1519 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1520 __be16 src_port
, __be16 dst_port
, __be32 vni
)
1522 struct vxlanhdr
*vxh
;
1527 if (!skb
->encapsulation
) {
1528 skb_reset_inner_headers(skb
);
1529 skb
->encapsulation
= 1;
1532 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1533 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1534 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1536 /* Need space for new headers (invalidates iph ptr) */
1537 err
= skb_cow_head(skb
, min_headroom
);
1541 if (vlan_tx_tag_present(skb
)) {
1542 if (WARN_ON(!__vlan_put_tag(skb
,
1544 vlan_tx_tag_get(skb
))))
1550 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1551 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1554 __skb_push(skb
, sizeof(*uh
));
1555 skb_reset_transport_header(skb
);
1558 uh
->dest
= dst_port
;
1559 uh
->source
= src_port
;
1561 uh
->len
= htons(skb
->len
);
1564 vxlan_set_owner(vs
->sock
->sk
, skb
);
1566 err
= handle_offloads(skb
);
1570 return iptunnel_xmit(rt
, skb
, src
, dst
, IPPROTO_UDP
, tos
, ttl
, df
,
1573 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1575 /* Bypass encapsulation if the destination is local */
1576 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1577 struct vxlan_dev
*dst_vxlan
)
1579 struct pcpu_tstats
*tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1580 struct pcpu_tstats
*rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1581 union vxlan_addr loopback
;
1582 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1584 skb
->pkt_type
= PACKET_HOST
;
1585 skb
->encapsulation
= 0;
1586 skb
->dev
= dst_vxlan
->dev
;
1587 __skb_pull(skb
, skb_network_offset(skb
));
1589 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1590 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1591 loopback
.sa
.sa_family
= AF_INET
;
1592 #if IS_ENABLED(CONFIG_IPV6)
1594 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1595 loopback
.sa
.sa_family
= AF_INET6
;
1599 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1600 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1602 u64_stats_update_begin(&tx_stats
->syncp
);
1603 tx_stats
->tx_packets
++;
1604 tx_stats
->tx_bytes
+= skb
->len
;
1605 u64_stats_update_end(&tx_stats
->syncp
);
1607 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1608 u64_stats_update_begin(&rx_stats
->syncp
);
1609 rx_stats
->rx_packets
++;
1610 rx_stats
->rx_bytes
+= skb
->len
;
1611 u64_stats_update_end(&rx_stats
->syncp
);
1613 skb
->dev
->stats
.rx_dropped
++;
1617 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1618 struct vxlan_rdst
*rdst
, bool did_rsc
)
1620 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1621 struct rtable
*rt
= NULL
;
1622 const struct iphdr
*old_iph
;
1624 union vxlan_addr
*dst
;
1625 __be16 src_port
= 0, dst_port
;
1631 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1632 vni
= rdst
->remote_vni
;
1633 dst
= &rdst
->remote_ip
;
1635 if (vxlan_addr_any(dst
)) {
1637 /* short-circuited back to local bridge */
1638 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1644 old_iph
= ip_hdr(skb
);
1647 if (!ttl
&& vxlan_addr_multicast(dst
))
1652 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1654 src_port
= vxlan_src_port(vxlan
->port_min
, vxlan
->port_max
, skb
);
1656 if (dst
->sa
.sa_family
== AF_INET
) {
1657 memset(&fl4
, 0, sizeof(fl4
));
1658 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1659 fl4
.flowi4_tos
= RT_TOS(tos
);
1660 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1661 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1663 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
1665 netdev_dbg(dev
, "no route to %pI4\n",
1666 &dst
->sin
.sin_addr
.s_addr
);
1667 dev
->stats
.tx_carrier_errors
++;
1671 if (rt
->dst
.dev
== dev
) {
1672 netdev_dbg(dev
, "circular route to %pI4\n",
1673 &dst
->sin
.sin_addr
.s_addr
);
1674 dev
->stats
.collisions
++;
1678 /* Bypass encapsulation if the destination is local */
1679 if (rt
->rt_flags
& RTCF_LOCAL
&&
1680 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1681 struct vxlan_dev
*dst_vxlan
;
1684 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
, dst_port
);
1687 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1691 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1692 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1694 err
= vxlan_xmit_skb(vxlan
->vn_sock
, rt
, skb
,
1695 fl4
.saddr
, dst
->sin
.sin_addr
.s_addr
,
1696 tos
, ttl
, df
, src_port
, dst_port
,
1701 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1702 #if IS_ENABLED(CONFIG_IPV6)
1704 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1705 struct dst_entry
*ndst
;
1709 memset(&fl6
, 0, sizeof(fl6
));
1710 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
1711 fl6
.daddr
= dst
->sin6
.sin6_addr
;
1712 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
1713 fl6
.flowi6_proto
= IPPROTO_UDP
;
1715 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
1716 netdev_dbg(dev
, "no route to %pI6\n",
1717 &dst
->sin6
.sin6_addr
);
1718 dev
->stats
.tx_carrier_errors
++;
1722 if (ndst
->dev
== dev
) {
1723 netdev_dbg(dev
, "circular route to %pI6\n",
1724 &dst
->sin6
.sin6_addr
);
1726 dev
->stats
.collisions
++;
1730 /* Bypass encapsulation if the destination is local */
1731 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
1732 if (flags
& RTF_LOCAL
&&
1733 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1734 struct vxlan_dev
*dst_vxlan
;
1737 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
, dst_port
);
1740 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1744 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
1746 err
= vxlan6_xmit_skb(vxlan
->vn_sock
, ndst
, skb
,
1747 dev
, &fl6
.saddr
, &fl6
.daddr
, 0, ttl
,
1748 src_port
, dst_port
, htonl(vni
<< 8));
1755 dev
->stats
.tx_dropped
++;
1761 dev
->stats
.tx_errors
++;
1766 /* Transmit local packets over Vxlan
1768 * Outer IP header inherits ECN and DF from inner header.
1769 * Outer UDP destination is the VXLAN assigned port.
1770 * source port is based on hash of flow
1772 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1774 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1776 bool did_rsc
= false;
1777 struct vxlan_rdst
*rdst
;
1778 struct vxlan_fdb
*f
;
1780 skb_reset_mac_header(skb
);
1783 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
1784 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
1785 return arp_reduce(dev
, skb
);
1786 #if IS_ENABLED(CONFIG_IPV6)
1787 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
1788 skb
->len
>= sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
) &&
1789 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
1792 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1793 if (msg
->icmph
.icmp6_code
== 0 &&
1794 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
1795 return neigh_reduce(dev
, skb
);
1800 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1803 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1804 (ntohs(eth
->h_proto
) == ETH_P_IP
||
1805 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
1806 did_rsc
= route_shortcircuit(dev
, skb
);
1808 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1812 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
1814 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
1815 !is_multicast_ether_addr(eth
->h_dest
))
1816 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
1818 dev
->stats
.tx_dropped
++;
1820 return NETDEV_TX_OK
;
1824 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
1825 struct sk_buff
*skb1
;
1827 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1829 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
1833 return NETDEV_TX_OK
;
1836 /* Walk the forwarding table and purge stale entries */
1837 static void vxlan_cleanup(unsigned long arg
)
1839 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1840 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1843 if (!netif_running(vxlan
->dev
))
1846 spin_lock_bh(&vxlan
->hash_lock
);
1847 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1848 struct hlist_node
*p
, *n
;
1849 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1851 = container_of(p
, struct vxlan_fdb
, hlist
);
1852 unsigned long timeout
;
1854 if (f
->state
& NUD_PERMANENT
)
1857 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
1858 if (time_before_eq(timeout
, jiffies
)) {
1859 netdev_dbg(vxlan
->dev
,
1860 "garbage collect %pM\n",
1862 f
->state
= NUD_STALE
;
1863 vxlan_fdb_destroy(vxlan
, f
);
1864 } else if (time_before(timeout
, next_timer
))
1865 next_timer
= timeout
;
1868 spin_unlock_bh(&vxlan
->hash_lock
);
1870 mod_timer(&vxlan
->age_timer
, next_timer
);
1873 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
1875 __u32 vni
= vxlan
->default_dst
.remote_vni
;
1877 vxlan
->vn_sock
= vs
;
1878 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
1881 /* Setup stats when device is created */
1882 static int vxlan_init(struct net_device
*dev
)
1884 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1885 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
1886 struct vxlan_sock
*vs
;
1888 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1892 spin_lock(&vn
->sock_lock
);
1893 vs
= vxlan_find_sock(dev_net(dev
), vxlan
->dst_port
);
1895 /* If we have a socket with same port already, reuse it */
1896 atomic_inc(&vs
->refcnt
);
1897 vxlan_vs_add_dev(vs
, vxlan
);
1899 /* otherwise make new socket outside of RTNL */
1901 queue_work(vxlan_wq
, &vxlan
->sock_work
);
1903 spin_unlock(&vn
->sock_lock
);
1908 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
1910 struct vxlan_fdb
*f
;
1912 spin_lock_bh(&vxlan
->hash_lock
);
1913 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
1915 vxlan_fdb_destroy(vxlan
, f
);
1916 spin_unlock_bh(&vxlan
->hash_lock
);
1919 static void vxlan_uninit(struct net_device
*dev
)
1921 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1922 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1924 vxlan_fdb_delete_default(vxlan
);
1927 vxlan_sock_release(vs
);
1928 free_percpu(dev
->tstats
);
1931 /* Start ageing timer and join group when device is brought up */
1932 static int vxlan_open(struct net_device
*dev
)
1934 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
1935 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1936 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1938 /* socket hasn't been created */
1942 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
1943 vxlan_group_used(vn
, &vxlan
->default_dst
.remote_ip
)) {
1944 vxlan_sock_hold(vs
);
1946 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
1949 if (vxlan
->age_interval
)
1950 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
1955 /* Purge the forwarding table */
1956 static void vxlan_flush(struct vxlan_dev
*vxlan
)
1960 spin_lock_bh(&vxlan
->hash_lock
);
1961 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1962 struct hlist_node
*p
, *n
;
1963 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1965 = container_of(p
, struct vxlan_fdb
, hlist
);
1966 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1967 if (!is_zero_ether_addr(f
->eth_addr
))
1968 vxlan_fdb_destroy(vxlan
, f
);
1971 spin_unlock_bh(&vxlan
->hash_lock
);
1974 /* Cleanup timer and forwarding table on shutdown */
1975 static int vxlan_stop(struct net_device
*dev
)
1977 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
1978 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1979 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1981 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
1982 ! vxlan_group_used(vn
, &vxlan
->default_dst
.remote_ip
)) {
1983 vxlan_sock_hold(vs
);
1985 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
1988 del_timer_sync(&vxlan
->age_timer
);
1995 /* Stub, nothing needs to be done. */
1996 static void vxlan_set_multicast_list(struct net_device
*dev
)
2000 static const struct net_device_ops vxlan_netdev_ops
= {
2001 .ndo_init
= vxlan_init
,
2002 .ndo_uninit
= vxlan_uninit
,
2003 .ndo_open
= vxlan_open
,
2004 .ndo_stop
= vxlan_stop
,
2005 .ndo_start_xmit
= vxlan_xmit
,
2006 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2007 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2008 .ndo_change_mtu
= eth_change_mtu
,
2009 .ndo_validate_addr
= eth_validate_addr
,
2010 .ndo_set_mac_address
= eth_mac_addr
,
2011 .ndo_fdb_add
= vxlan_fdb_add
,
2012 .ndo_fdb_del
= vxlan_fdb_delete
,
2013 .ndo_fdb_dump
= vxlan_fdb_dump
,
2016 /* Info for udev, that this is a virtual tunnel endpoint */
2017 static struct device_type vxlan_type
= {
2021 /* Calls the ndo_add_vxlan_port of the caller in order to
2022 * supply the listening VXLAN udp ports. Callers are expected
2023 * to implement the ndo_add_vxlan_port.
2025 void vxlan_get_rx_port(struct net_device
*dev
)
2027 struct vxlan_sock
*vs
;
2028 struct net
*net
= dev_net(dev
);
2029 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2030 sa_family_t sa_family
;
2034 spin_lock(&vn
->sock_lock
);
2035 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2036 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2037 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2038 sa_family
= vs
->sock
->sk
->sk_family
;
2039 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2043 spin_unlock(&vn
->sock_lock
);
2045 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2047 /* Initialize the device structure. */
2048 static void vxlan_setup(struct net_device
*dev
)
2050 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2054 eth_hw_addr_random(dev
);
2056 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2057 dev
->hard_header_len
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2059 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
2061 dev
->netdev_ops
= &vxlan_netdev_ops
;
2062 dev
->destructor
= free_netdev
;
2063 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2065 dev
->tx_queue_len
= 0;
2066 dev
->features
|= NETIF_F_LLTX
;
2067 dev
->features
|= NETIF_F_NETNS_LOCAL
;
2068 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2069 dev
->features
|= NETIF_F_RXCSUM
;
2070 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2072 dev
->vlan_features
= dev
->features
;
2073 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2074 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2075 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2076 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2077 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
2078 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2080 INIT_LIST_HEAD(&vxlan
->next
);
2081 spin_lock_init(&vxlan
->hash_lock
);
2082 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2083 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2084 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2086 init_timer_deferrable(&vxlan
->age_timer
);
2087 vxlan
->age_timer
.function
= vxlan_cleanup
;
2088 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2090 inet_get_local_port_range(&low
, &high
);
2091 vxlan
->port_min
= low
;
2092 vxlan
->port_max
= high
;
2093 vxlan
->dst_port
= htons(vxlan_port
);
2097 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2098 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2101 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2102 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2103 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2104 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2105 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2106 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2107 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2108 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2109 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2110 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2111 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2112 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2113 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2114 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2115 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2116 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2117 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2118 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2121 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2123 if (tb
[IFLA_ADDRESS
]) {
2124 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2125 pr_debug("invalid link address (not ethernet)\n");
2129 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2130 pr_debug("invalid all zero ethernet address\n");
2131 return -EADDRNOTAVAIL
;
2138 if (data
[IFLA_VXLAN_ID
]) {
2139 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2140 if (id
>= VXLAN_VID_MASK
)
2144 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2145 const struct ifla_vxlan_port_range
*p
2146 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2148 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2149 pr_debug("port range %u .. %u not valid\n",
2150 ntohs(p
->low
), ntohs(p
->high
));
2158 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2159 struct ethtool_drvinfo
*drvinfo
)
2161 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2162 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2165 static const struct ethtool_ops vxlan_ethtool_ops
= {
2166 .get_drvinfo
= vxlan_get_drvinfo
,
2167 .get_link
= ethtool_op_get_link
,
2170 static void vxlan_del_work(struct work_struct
*work
)
2172 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2174 sk_release_kernel(vs
->sock
->sk
);
2178 #if IS_ENABLED(CONFIG_IPV6)
2179 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2180 * could be used for both IPv4 and IPv6 communications, but
2181 * users may set bindv6only=1.
2183 static int create_v6_sock(struct net
*net
, __be16 port
, struct socket
**psock
)
2186 struct socket
*sock
;
2187 struct sockaddr_in6 vxlan_addr
= {
2188 .sin6_family
= AF_INET6
,
2193 rc
= sock_create_kern(AF_INET6
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2195 pr_debug("UDPv6 socket create failed\n");
2199 /* Put in proper namespace */
2201 sk_change_net(sk
, net
);
2203 kernel_setsockopt(sock
, SOL_IPV6
, IPV6_V6ONLY
,
2204 (char *)&val
, sizeof(val
));
2205 rc
= kernel_bind(sock
, (struct sockaddr
*)&vxlan_addr
,
2206 sizeof(struct sockaddr_in6
));
2208 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2209 &vxlan_addr
.sin6_addr
, ntohs(vxlan_addr
.sin6_port
), rc
);
2210 sk_release_kernel(sk
);
2213 /* At this point, IPv6 module should have been loaded in
2214 * sock_create_kern().
2219 /* Disable multicast loopback */
2220 inet_sk(sk
)->mc_loop
= 0;
2226 static int create_v6_sock(struct net
*net
, __be16 port
, struct socket
**psock
)
2228 return -EPFNOSUPPORT
;
2232 static int create_v4_sock(struct net
*net
, __be16 port
, struct socket
**psock
)
2235 struct socket
*sock
;
2236 struct sockaddr_in vxlan_addr
= {
2237 .sin_family
= AF_INET
,
2238 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
2243 /* Create UDP socket for encapsulation receive. */
2244 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &sock
);
2246 pr_debug("UDP socket create failed\n");
2250 /* Put in proper namespace */
2252 sk_change_net(sk
, net
);
2254 rc
= kernel_bind(sock
, (struct sockaddr
*) &vxlan_addr
,
2255 sizeof(vxlan_addr
));
2257 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2258 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
2259 sk_release_kernel(sk
);
2264 /* Disable multicast loopback */
2265 inet_sk(sk
)->mc_loop
= 0;
2269 /* Create new listen socket if needed */
2270 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2271 vxlan_rcv_t
*rcv
, void *data
, bool ipv6
)
2273 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2274 struct vxlan_sock
*vs
;
2275 struct socket
*sock
;
2280 vs
= kmalloc(sizeof(*vs
), GFP_KERNEL
);
2282 return ERR_PTR(-ENOMEM
);
2284 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2285 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2287 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2290 rc
= create_v6_sock(net
, port
, &sock
);
2292 rc
= create_v4_sock(net
, port
, &sock
);
2300 atomic_set(&vs
->refcnt
, 1);
2303 rcu_assign_sk_user_data(vs
->sock
->sk
, vs
);
2305 spin_lock(&vn
->sock_lock
);
2306 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2307 vxlan_notify_add_rx_port(sk
);
2308 spin_unlock(&vn
->sock_lock
);
2310 /* Mark socket as an encapsulation socket. */
2311 udp_sk(sk
)->encap_type
= 1;
2312 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
2313 #if IS_ENABLED(CONFIG_IPV6)
2315 ipv6_stub
->udpv6_encap_enable();
2323 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2324 vxlan_rcv_t
*rcv
, void *data
,
2325 bool no_share
, bool ipv6
)
2327 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2328 struct vxlan_sock
*vs
;
2330 vs
= vxlan_socket_create(net
, port
, rcv
, data
, ipv6
);
2334 if (no_share
) /* Return error if sharing is not allowed. */
2337 spin_lock(&vn
->sock_lock
);
2338 vs
= vxlan_find_sock(net
, port
);
2341 atomic_inc(&vs
->refcnt
);
2343 vs
= ERR_PTR(-EBUSY
);
2345 spin_unlock(&vn
->sock_lock
);
2348 vs
= ERR_PTR(-EINVAL
);
2352 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2354 /* Scheduled at device creation to bind to a socket */
2355 static void vxlan_sock_work(struct work_struct
*work
)
2357 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2358 struct net
*net
= dev_net(vxlan
->dev
);
2359 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2360 __be16 port
= vxlan
->dst_port
;
2361 struct vxlan_sock
*nvs
;
2363 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
& VXLAN_F_IPV6
);
2364 spin_lock(&vn
->sock_lock
);
2366 vxlan_vs_add_dev(nvs
, vxlan
);
2367 spin_unlock(&vn
->sock_lock
);
2369 dev_put(vxlan
->dev
);
2372 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2373 struct nlattr
*tb
[], struct nlattr
*data
[])
2375 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2376 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2377 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2380 bool use_ipv6
= false;
2382 if (!data
[IFLA_VXLAN_ID
])
2385 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2386 dst
->remote_vni
= vni
;
2388 if (data
[IFLA_VXLAN_GROUP
]) {
2389 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2390 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2391 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2392 if (!IS_ENABLED(CONFIG_IPV6
))
2393 return -EPFNOSUPPORT
;
2395 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2396 sizeof(struct in6_addr
));
2397 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2401 if (data
[IFLA_VXLAN_LOCAL
]) {
2402 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2403 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2404 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2405 if (!IS_ENABLED(CONFIG_IPV6
))
2406 return -EPFNOSUPPORT
;
2408 /* TODO: respect scope id */
2409 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2410 sizeof(struct in6_addr
));
2411 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2415 if (data
[IFLA_VXLAN_LINK
] &&
2416 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2417 struct net_device
*lowerdev
2418 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2421 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2425 #if IS_ENABLED(CONFIG_IPV6)
2427 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2428 if (idev
&& idev
->cnf
.disable_ipv6
) {
2429 pr_info("IPv6 is disabled via sysctl\n");
2432 vxlan
->flags
|= VXLAN_F_IPV6
;
2437 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2439 /* update header length based on lower device */
2440 dev
->hard_header_len
= lowerdev
->hard_header_len
+
2441 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2444 if (data
[IFLA_VXLAN_TOS
])
2445 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2447 if (data
[IFLA_VXLAN_TTL
])
2448 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2450 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2451 vxlan
->flags
|= VXLAN_F_LEARN
;
2453 if (data
[IFLA_VXLAN_AGEING
])
2454 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2456 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2458 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2459 vxlan
->flags
|= VXLAN_F_PROXY
;
2461 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2462 vxlan
->flags
|= VXLAN_F_RSC
;
2464 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2465 vxlan
->flags
|= VXLAN_F_L2MISS
;
2467 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2468 vxlan
->flags
|= VXLAN_F_L3MISS
;
2470 if (data
[IFLA_VXLAN_LIMIT
])
2471 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2473 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2474 const struct ifla_vxlan_port_range
*p
2475 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2476 vxlan
->port_min
= ntohs(p
->low
);
2477 vxlan
->port_max
= ntohs(p
->high
);
2480 if (data
[IFLA_VXLAN_PORT
])
2481 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2483 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
2484 pr_info("duplicate VNI %u\n", vni
);
2488 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
2490 /* create an fdb entry for a valid default destination */
2491 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2492 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2493 &vxlan
->default_dst
.remote_ip
,
2494 NUD_REACHABLE
|NUD_PERMANENT
,
2495 NLM_F_EXCL
|NLM_F_CREATE
,
2497 vxlan
->default_dst
.remote_vni
,
2498 vxlan
->default_dst
.remote_ifindex
,
2504 err
= register_netdevice(dev
);
2506 vxlan_fdb_delete_default(vxlan
);
2510 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2515 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2517 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2518 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2520 spin_lock(&vn
->sock_lock
);
2521 if (!hlist_unhashed(&vxlan
->hlist
))
2522 hlist_del_rcu(&vxlan
->hlist
);
2523 spin_unlock(&vn
->sock_lock
);
2525 list_del(&vxlan
->next
);
2526 unregister_netdevice_queue(dev
, head
);
2529 static size_t vxlan_get_size(const struct net_device
*dev
)
2532 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2533 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2534 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2535 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2536 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2537 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2538 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2539 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2540 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2541 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2542 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2543 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2544 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2545 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2546 nla_total_size(sizeof(__be16
))+ /* IFLA_VXLAN_PORT */
2550 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2552 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2553 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2554 struct ifla_vxlan_port_range ports
= {
2555 .low
= htons(vxlan
->port_min
),
2556 .high
= htons(vxlan
->port_max
),
2559 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2560 goto nla_put_failure
;
2562 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2563 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2564 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2565 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2566 goto nla_put_failure
;
2567 #if IS_ENABLED(CONFIG_IPV6)
2569 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2570 &dst
->remote_ip
.sin6
.sin6_addr
))
2571 goto nla_put_failure
;
2576 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2577 goto nla_put_failure
;
2579 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2580 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2581 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2582 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2583 goto nla_put_failure
;
2584 #if IS_ENABLED(CONFIG_IPV6)
2586 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2587 &vxlan
->saddr
.sin6
.sin6_addr
))
2588 goto nla_put_failure
;
2593 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2594 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2595 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2596 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2597 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2598 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2599 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2600 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2601 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2602 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2603 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2604 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2605 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2606 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
))
2607 goto nla_put_failure
;
2609 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2610 goto nla_put_failure
;
2618 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2620 .maxtype
= IFLA_VXLAN_MAX
,
2621 .policy
= vxlan_policy
,
2622 .priv_size
= sizeof(struct vxlan_dev
),
2623 .setup
= vxlan_setup
,
2624 .validate
= vxlan_validate
,
2625 .newlink
= vxlan_newlink
,
2626 .dellink
= vxlan_dellink
,
2627 .get_size
= vxlan_get_size
,
2628 .fill_info
= vxlan_fill_info
,
2631 static __net_init
int vxlan_init_net(struct net
*net
)
2633 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2636 INIT_LIST_HEAD(&vn
->vxlan_list
);
2637 spin_lock_init(&vn
->sock_lock
);
2639 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2640 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
2645 static __net_exit
void vxlan_exit_net(struct net
*net
)
2647 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2648 struct vxlan_dev
*vxlan
;
2652 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
)
2653 unregister_netdevice_queue(vxlan
->dev
, &list
);
2654 unregister_netdevice_many(&list
);
2658 static struct pernet_operations vxlan_net_ops
= {
2659 .init
= vxlan_init_net
,
2660 .exit
= vxlan_exit_net
,
2661 .id
= &vxlan_net_id
,
2662 .size
= sizeof(struct vxlan_net
),
2665 static int __init
vxlan_init_module(void)
2669 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
2673 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
2675 rc
= register_pernet_device(&vxlan_net_ops
);
2679 rc
= rtnl_link_register(&vxlan_link_ops
);
2686 unregister_pernet_device(&vxlan_net_ops
);
2688 destroy_workqueue(vxlan_wq
);
2691 late_initcall(vxlan_init_module
);
2693 static void __exit
vxlan_cleanup_module(void)
2695 rtnl_link_unregister(&vxlan_link_ops
);
2696 destroy_workqueue(vxlan_wq
);
2697 unregister_pernet_device(&vxlan_net_ops
);
2700 module_exit(vxlan_cleanup_module
);
2702 MODULE_LICENSE("GPL");
2703 MODULE_VERSION(VXLAN_VERSION
);
2704 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2705 MODULE_ALIAS_RTNL_LINK("vxlan");