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/udp_tunnel.h>
37 #include <net/rtnetlink.h>
38 #include <net/route.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/vxlan.h>
44 #include <net/protocol.h>
45 #include <net/udp_tunnel.h>
46 #if IS_ENABLED(CONFIG_IPV6)
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
53 #define VXLAN_VERSION "0.1"
55 #define PORT_HASH_BITS 8
56 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
57 #define VNI_HASH_BITS 10
58 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
59 #define FDB_HASH_BITS 8
60 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
61 #define FDB_AGE_DEFAULT 300 /* 5 min */
62 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
64 /* UDP port for VXLAN traffic.
65 * The IANA assigned port is 4789, but the Linux default is 8472
66 * for compatibility with early adopters.
68 static unsigned short vxlan_port __read_mostly
= 8472;
69 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
70 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
72 static bool log_ecn_error
= true;
73 module_param(log_ecn_error
, bool, 0644);
74 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
76 static int vxlan_net_id
;
78 static const u8 all_zeros_mac
[ETH_ALEN
];
80 /* per-network namespace private data for this module */
82 struct list_head vxlan_list
;
83 struct hlist_head sock_list
[PORT_HASH_SIZE
];
88 struct sockaddr_in sin
;
89 struct sockaddr_in6 sin6
;
94 union vxlan_addr remote_ip
;
98 struct list_head list
;
102 /* Forwarding table entry */
104 struct hlist_node hlist
; /* linked list of entries */
106 unsigned long updated
; /* jiffies */
108 struct list_head remotes
;
109 u16 state
; /* see ndm_state */
110 u8 flags
; /* see ndm_flags */
111 u8 eth_addr
[ETH_ALEN
];
114 /* Pseudo network device */
116 struct hlist_node hlist
; /* vni hash table */
117 struct list_head next
; /* vxlan's per namespace list */
118 struct vxlan_sock
*vn_sock
; /* listening socket */
119 struct net_device
*dev
;
120 struct net
*net
; /* netns for packet i/o */
121 struct vxlan_rdst default_dst
; /* default destination */
122 union vxlan_addr saddr
; /* source address */
124 __u16 port_min
; /* source port range */
126 __u8 tos
; /* TOS override */
128 u32 flags
; /* VXLAN_F_* in vxlan.h */
130 struct work_struct sock_work
;
131 struct work_struct igmp_join
;
132 struct work_struct igmp_leave
;
134 unsigned long age_interval
;
135 struct timer_list age_timer
;
136 spinlock_t hash_lock
;
137 unsigned int addrcnt
;
138 unsigned int addrmax
;
140 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
143 /* salt for hash table */
144 static u32 vxlan_salt __read_mostly
;
145 static struct workqueue_struct
*vxlan_wq
;
147 static void vxlan_sock_work(struct work_struct
*work
);
149 #if IS_ENABLED(CONFIG_IPV6)
151 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
153 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
155 if (a
->sa
.sa_family
== AF_INET6
)
156 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
158 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
161 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
163 if (ipa
->sa
.sa_family
== AF_INET6
)
164 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
166 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
169 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
171 if (ipa
->sa
.sa_family
== AF_INET6
)
172 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
174 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
177 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
179 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
180 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
181 ip
->sa
.sa_family
= AF_INET6
;
183 } else if (nla_len(nla
) >= sizeof(__be32
)) {
184 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
185 ip
->sa
.sa_family
= AF_INET
;
188 return -EAFNOSUPPORT
;
192 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
193 const union vxlan_addr
*ip
)
195 if (ip
->sa
.sa_family
== AF_INET6
)
196 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
198 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
201 #else /* !CONFIG_IPV6 */
204 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
206 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
209 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
211 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
214 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
216 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
219 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
221 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
222 return -EAFNOSUPPORT
;
223 } else if (nla_len(nla
) >= sizeof(__be32
)) {
224 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
225 ip
->sa
.sa_family
= AF_INET
;
228 return -EAFNOSUPPORT
;
232 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
233 const union vxlan_addr
*ip
)
235 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
239 /* Virtual Network hash table head */
240 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
242 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
245 /* Socket hash table head */
246 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
248 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
250 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
253 /* First remote destination for a forwarding entry.
254 * Guaranteed to be non-NULL because remotes are never deleted.
256 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
258 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
261 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
263 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
266 /* Find VXLAN socket based on network namespace, address family and UDP port
267 * and enabled unshareable flags.
269 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
270 __be16 port
, u32 flags
)
272 struct vxlan_sock
*vs
;
274 flags
&= VXLAN_F_RCV_FLAGS
;
276 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
277 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
278 inet_sk(vs
->sock
->sk
)->sk
.sk_family
== family
&&
285 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
287 struct vxlan_dev
*vxlan
;
289 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
290 if (vxlan
->default_dst
.remote_vni
== id
)
297 /* Look up VNI in a per net namespace table */
298 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
,
299 sa_family_t family
, __be16 port
,
302 struct vxlan_sock
*vs
;
304 vs
= vxlan_find_sock(net
, family
, port
, flags
);
308 return vxlan_vs_find_vni(vs
, id
);
311 /* Fill in neighbour message in skbuff. */
312 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
313 const struct vxlan_fdb
*fdb
,
314 u32 portid
, u32 seq
, int type
, unsigned int flags
,
315 const struct vxlan_rdst
*rdst
)
317 unsigned long now
= jiffies
;
318 struct nda_cacheinfo ci
;
319 struct nlmsghdr
*nlh
;
321 bool send_ip
, send_eth
;
323 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
327 ndm
= nlmsg_data(nlh
);
328 memset(ndm
, 0, sizeof(*ndm
));
330 send_eth
= send_ip
= true;
332 if (type
== RTM_GETNEIGH
) {
333 ndm
->ndm_family
= AF_INET
;
334 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
335 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
337 ndm
->ndm_family
= AF_BRIDGE
;
338 ndm
->ndm_state
= fdb
->state
;
339 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
340 ndm
->ndm_flags
= fdb
->flags
;
341 ndm
->ndm_type
= RTN_UNICAST
;
343 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
344 nla_put_s32(skb
, NDA_LINK_NETNSID
,
345 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
346 goto nla_put_failure
;
348 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
349 goto nla_put_failure
;
351 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
352 goto nla_put_failure
;
354 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
355 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
356 goto nla_put_failure
;
357 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
358 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
359 goto nla_put_failure
;
360 if (rdst
->remote_ifindex
&&
361 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
362 goto nla_put_failure
;
364 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
365 ci
.ndm_confirmed
= 0;
366 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
369 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
370 goto nla_put_failure
;
376 nlmsg_cancel(skb
, nlh
);
380 static inline size_t vxlan_nlmsg_size(void)
382 return NLMSG_ALIGN(sizeof(struct ndmsg
))
383 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
384 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
385 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
386 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
387 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
388 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
389 + nla_total_size(sizeof(struct nda_cacheinfo
));
392 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
393 struct vxlan_rdst
*rd
, int type
)
395 struct net
*net
= dev_net(vxlan
->dev
);
399 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
403 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
405 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
406 WARN_ON(err
== -EMSGSIZE
);
411 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
415 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
418 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
420 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
421 struct vxlan_fdb f
= {
424 struct vxlan_rdst remote
= {
425 .remote_ip
= *ipa
, /* goes to NDA_DST */
426 .remote_vni
= VXLAN_N_VID
,
429 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
432 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
434 struct vxlan_fdb f
= {
437 struct vxlan_rdst remote
= { };
439 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
441 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
444 /* Hash Ethernet address */
445 static u32
eth_hash(const unsigned char *addr
)
447 u64 value
= get_unaligned((u64
*)addr
);
449 /* only want 6 bytes */
455 return hash_64(value
, FDB_HASH_BITS
);
458 /* Hash chain to use given mac address */
459 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
462 return &vxlan
->fdb_head
[eth_hash(mac
)];
465 /* Look up Ethernet address in forwarding table */
466 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
469 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
472 hlist_for_each_entry_rcu(f
, head
, hlist
) {
473 if (ether_addr_equal(mac
, f
->eth_addr
))
480 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
485 f
= __vxlan_find_mac(vxlan
, mac
);
492 /* caller should hold vxlan->hash_lock */
493 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
494 union vxlan_addr
*ip
, __be16 port
,
495 __u32 vni
, __u32 ifindex
)
497 struct vxlan_rdst
*rd
;
499 list_for_each_entry(rd
, &f
->remotes
, list
) {
500 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
501 rd
->remote_port
== port
&&
502 rd
->remote_vni
== vni
&&
503 rd
->remote_ifindex
== ifindex
)
510 /* Replace destination of unicast mac */
511 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
512 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
514 struct vxlan_rdst
*rd
;
516 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
520 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
524 rd
->remote_port
= port
;
525 rd
->remote_vni
= vni
;
526 rd
->remote_ifindex
= ifindex
;
530 /* Add/update destinations for multicast */
531 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
532 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
533 __u32 ifindex
, struct vxlan_rdst
**rdp
)
535 struct vxlan_rdst
*rd
;
537 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
541 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
545 rd
->remote_port
= port
;
546 rd
->remote_vni
= vni
;
547 rd
->remote_ifindex
= ifindex
;
549 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
555 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
557 struct vxlanhdr
*vh
, size_t hdrlen
,
558 u32 data
, struct gro_remcsum
*grc
,
561 size_t start
, offset
, plen
;
563 if (skb
->remcsum_offload
)
566 if (!NAPI_GRO_CB(skb
)->csum_valid
)
569 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
570 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
571 offsetof(struct udphdr
, check
) :
572 offsetof(struct tcphdr
, check
));
574 plen
= hdrlen
+ offset
+ sizeof(u16
);
576 /* Pull checksum that will be written */
577 if (skb_gro_header_hard(skb
, off
+ plen
)) {
578 vh
= skb_gro_header_slow(skb
, off
+ plen
, off
);
583 skb_gro_remcsum_process(skb
, (void *)vh
+ hdrlen
,
584 start
, offset
, grc
, nopartial
);
586 skb
->remcsum_offload
= 1;
591 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
,
593 struct udp_offload
*uoff
)
595 struct sk_buff
*p
, **pp
= NULL
;
596 struct vxlanhdr
*vh
, *vh2
;
597 unsigned int hlen
, off_vx
;
599 struct vxlan_sock
*vs
= container_of(uoff
, struct vxlan_sock
,
602 struct gro_remcsum grc
;
604 skb_gro_remcsum_init(&grc
);
606 off_vx
= skb_gro_offset(skb
);
607 hlen
= off_vx
+ sizeof(*vh
);
608 vh
= skb_gro_header_fast(skb
, off_vx
);
609 if (skb_gro_header_hard(skb
, hlen
)) {
610 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
615 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
616 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
618 flags
= ntohl(vh
->vx_flags
);
620 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
621 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
622 ntohl(vh
->vx_vni
), &grc
,
624 VXLAN_F_REMCSUM_NOPARTIAL
));
632 for (p
= *head
; p
; p
= p
->next
) {
633 if (!NAPI_GRO_CB(p
)->same_flow
)
636 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
637 if (vh
->vx_flags
!= vh2
->vx_flags
||
638 vh
->vx_vni
!= vh2
->vx_vni
) {
639 NAPI_GRO_CB(p
)->same_flow
= 0;
644 pp
= eth_gro_receive(head
, skb
);
647 skb_gro_remcsum_cleanup(skb
, &grc
);
648 NAPI_GRO_CB(skb
)->flush
|= flush
;
653 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
654 struct udp_offload
*uoff
)
656 udp_tunnel_gro_complete(skb
, nhoff
);
658 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
661 /* Notify netdevs that UDP port started listening */
662 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
664 struct net_device
*dev
;
665 struct sock
*sk
= vs
->sock
->sk
;
666 struct net
*net
= sock_net(sk
);
667 sa_family_t sa_family
= sk
->sk_family
;
668 __be16 port
= inet_sk(sk
)->inet_sport
;
671 if (sa_family
== AF_INET
) {
672 err
= udp_add_offload(&vs
->udp_offloads
);
674 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
678 for_each_netdev_rcu(net
, dev
) {
679 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
680 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
686 /* Notify netdevs that UDP port is no more listening */
687 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
689 struct net_device
*dev
;
690 struct sock
*sk
= vs
->sock
->sk
;
691 struct net
*net
= sock_net(sk
);
692 sa_family_t sa_family
= sk
->sk_family
;
693 __be16 port
= inet_sk(sk
)->inet_sport
;
696 for_each_netdev_rcu(net
, dev
) {
697 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
698 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
703 if (sa_family
== AF_INET
)
704 udp_del_offload(&vs
->udp_offloads
);
707 /* Add new entry to forwarding table -- assumes lock held */
708 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
709 const u8
*mac
, union vxlan_addr
*ip
,
710 __u16 state
, __u16 flags
,
711 __be16 port
, __u32 vni
, __u32 ifindex
,
714 struct vxlan_rdst
*rd
= NULL
;
718 f
= __vxlan_find_mac(vxlan
, mac
);
720 if (flags
& NLM_F_EXCL
) {
721 netdev_dbg(vxlan
->dev
,
722 "lost race to create %pM\n", mac
);
725 if (f
->state
!= state
) {
727 f
->updated
= jiffies
;
730 if (f
->flags
!= ndm_flags
) {
731 f
->flags
= ndm_flags
;
732 f
->updated
= jiffies
;
735 if ((flags
& NLM_F_REPLACE
)) {
736 /* Only change unicasts */
737 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
738 is_zero_ether_addr(f
->eth_addr
))) {
739 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
748 if ((flags
& NLM_F_APPEND
) &&
749 (is_multicast_ether_addr(f
->eth_addr
) ||
750 is_zero_ether_addr(f
->eth_addr
))) {
751 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
759 if (!(flags
& NLM_F_CREATE
))
762 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
765 /* Disallow replace to add a multicast entry */
766 if ((flags
& NLM_F_REPLACE
) &&
767 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
770 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
771 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
777 f
->flags
= ndm_flags
;
778 f
->updated
= f
->used
= jiffies
;
779 INIT_LIST_HEAD(&f
->remotes
);
780 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
782 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
785 hlist_add_head_rcu(&f
->hlist
,
786 vxlan_fdb_head(vxlan
, mac
));
791 rd
= first_remote_rtnl(f
);
792 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
798 static void vxlan_fdb_free(struct rcu_head
*head
)
800 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
801 struct vxlan_rdst
*rd
, *nd
;
803 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
808 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
810 netdev_dbg(vxlan
->dev
,
811 "delete %pM\n", f
->eth_addr
);
814 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
816 hlist_del_rcu(&f
->hlist
);
817 call_rcu(&f
->rcu
, vxlan_fdb_free
);
820 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
821 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
823 struct net
*net
= dev_net(vxlan
->dev
);
827 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
831 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
832 if (remote
->sa
.sa_family
== AF_INET
) {
833 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
834 ip
->sa
.sa_family
= AF_INET
;
835 #if IS_ENABLED(CONFIG_IPV6)
837 ip
->sin6
.sin6_addr
= in6addr_any
;
838 ip
->sa
.sa_family
= AF_INET6
;
844 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
846 *port
= nla_get_be16(tb
[NDA_PORT
]);
848 *port
= vxlan
->dst_port
;
852 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
854 *vni
= nla_get_u32(tb
[NDA_VNI
]);
856 *vni
= vxlan
->default_dst
.remote_vni
;
859 if (tb
[NDA_IFINDEX
]) {
860 struct net_device
*tdev
;
862 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
864 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
865 tdev
= __dev_get_by_index(net
, *ifindex
);
867 return -EADDRNOTAVAIL
;
875 /* Add static entry (via netlink) */
876 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
877 struct net_device
*dev
,
878 const unsigned char *addr
, u16 vid
, u16 flags
)
880 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
881 /* struct net *net = dev_net(vxlan->dev); */
887 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
888 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
893 if (tb
[NDA_DST
] == NULL
)
896 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
900 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
901 return -EAFNOSUPPORT
;
903 spin_lock_bh(&vxlan
->hash_lock
);
904 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
905 port
, vni
, ifindex
, ndm
->ndm_flags
);
906 spin_unlock_bh(&vxlan
->hash_lock
);
911 /* Delete entry (via netlink) */
912 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
913 struct net_device
*dev
,
914 const unsigned char *addr
, u16 vid
)
916 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
918 struct vxlan_rdst
*rd
= NULL
;
924 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
930 spin_lock_bh(&vxlan
->hash_lock
);
931 f
= vxlan_find_mac(vxlan
, addr
);
935 if (!vxlan_addr_any(&ip
)) {
936 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
943 /* remove a destination if it's not the only one on the list,
944 * otherwise destroy the fdb entry
946 if (rd
&& !list_is_singular(&f
->remotes
)) {
947 list_del_rcu(&rd
->list
);
948 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
953 vxlan_fdb_destroy(vxlan
, f
);
956 spin_unlock_bh(&vxlan
->hash_lock
);
961 /* Dump forwarding table */
962 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
963 struct net_device
*dev
,
964 struct net_device
*filter_dev
, int idx
)
966 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
969 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
973 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
974 struct vxlan_rdst
*rd
;
976 if (idx
< cb
->args
[0])
979 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
980 err
= vxlan_fdb_info(skb
, vxlan
, f
,
981 NETLINK_CB(cb
->skb
).portid
,
996 /* Watch incoming packets to learn mapping between Ethernet address
997 * and Tunnel endpoint.
998 * Return true if packet is bogus and should be droppped.
1000 static bool vxlan_snoop(struct net_device
*dev
,
1001 union vxlan_addr
*src_ip
, const u8
*src_mac
)
1003 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1004 struct vxlan_fdb
*f
;
1006 f
= vxlan_find_mac(vxlan
, src_mac
);
1008 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
1010 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
1013 /* Don't migrate static entries, drop packets */
1014 if (f
->state
& NUD_NOARP
)
1017 if (net_ratelimit())
1019 "%pM migrated from %pIS to %pIS\n",
1020 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
1022 rdst
->remote_ip
= *src_ip
;
1023 f
->updated
= jiffies
;
1024 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1026 /* learned new entry */
1027 spin_lock(&vxlan
->hash_lock
);
1029 /* close off race between vxlan_flush and incoming packets */
1030 if (netif_running(dev
))
1031 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1033 NLM_F_EXCL
|NLM_F_CREATE
,
1035 vxlan
->default_dst
.remote_vni
,
1037 spin_unlock(&vxlan
->hash_lock
);
1043 /* See if multicast group is already in use by other ID */
1044 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1046 struct vxlan_dev
*vxlan
;
1048 /* The vxlan_sock is only used by dev, leaving group has
1049 * no effect on other vxlan devices.
1051 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1054 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1055 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1058 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1061 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1062 &dev
->default_dst
.remote_ip
))
1065 if (vxlan
->default_dst
.remote_ifindex
!=
1066 dev
->default_dst
.remote_ifindex
)
1075 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1077 atomic_inc(&vs
->refcnt
);
1080 void vxlan_sock_release(struct vxlan_sock
*vs
)
1082 struct sock
*sk
= vs
->sock
->sk
;
1083 struct net
*net
= sock_net(sk
);
1084 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1086 if (!atomic_dec_and_test(&vs
->refcnt
))
1089 spin_lock(&vn
->sock_lock
);
1090 hlist_del_rcu(&vs
->hlist
);
1091 vxlan_notify_del_rx_port(vs
);
1092 spin_unlock(&vn
->sock_lock
);
1094 queue_work(vxlan_wq
, &vs
->del_work
);
1096 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1098 /* Callback to update multicast group membership when first VNI on
1099 * multicast asddress is brought up
1100 * Done as workqueue because ip_mc_join_group acquires RTNL.
1102 static void vxlan_igmp_join(struct work_struct
*work
)
1104 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1105 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1106 struct sock
*sk
= vs
->sock
->sk
;
1107 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1108 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1111 if (ip
->sa
.sa_family
== AF_INET
) {
1112 struct ip_mreqn mreq
= {
1113 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1114 .imr_ifindex
= ifindex
,
1117 ip_mc_join_group(sk
, &mreq
);
1118 #if IS_ENABLED(CONFIG_IPV6)
1120 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1121 &ip
->sin6
.sin6_addr
);
1126 vxlan_sock_release(vs
);
1127 dev_put(vxlan
->dev
);
1130 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1131 static void vxlan_igmp_leave(struct work_struct
*work
)
1133 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1134 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1135 struct sock
*sk
= vs
->sock
->sk
;
1136 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1137 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1140 if (ip
->sa
.sa_family
== AF_INET
) {
1141 struct ip_mreqn mreq
= {
1142 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1143 .imr_ifindex
= ifindex
,
1146 ip_mc_leave_group(sk
, &mreq
);
1147 #if IS_ENABLED(CONFIG_IPV6)
1149 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1150 &ip
->sin6
.sin6_addr
);
1156 vxlan_sock_release(vs
);
1157 dev_put(vxlan
->dev
);
1160 static struct vxlanhdr
*vxlan_remcsum(struct sk_buff
*skb
, struct vxlanhdr
*vh
,
1161 size_t hdrlen
, u32 data
, bool nopartial
)
1163 size_t start
, offset
, plen
;
1165 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
1166 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
1167 offsetof(struct udphdr
, check
) :
1168 offsetof(struct tcphdr
, check
));
1170 plen
= hdrlen
+ offset
+ sizeof(u16
);
1172 if (!pskb_may_pull(skb
, plen
))
1175 vh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1177 skb_remcsum_process(skb
, (void *)vh
+ hdrlen
, start
, offset
,
1183 /* Callback from net/ipv4/udp.c to receive packets */
1184 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1186 struct vxlan_sock
*vs
;
1187 struct vxlanhdr
*vxh
;
1189 struct vxlan_metadata md
= {0};
1191 /* Need Vxlan and inner Ethernet header to be present */
1192 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1195 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1196 flags
= ntohl(vxh
->vx_flags
);
1197 vni
= ntohl(vxh
->vx_vni
);
1199 if (flags
& VXLAN_HF_VNI
) {
1200 flags
&= ~VXLAN_HF_VNI
;
1202 /* VNI flag always required to be set */
1206 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1208 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1210 vs
= rcu_dereference_sk_user_data(sk
);
1214 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
1215 vxh
= vxlan_remcsum(skb
, vxh
, sizeof(struct vxlanhdr
), vni
,
1216 !!(vs
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1220 flags
&= ~VXLAN_HF_RCO
;
1221 vni
&= VXLAN_VNI_MASK
;
1224 /* For backwards compatibility, only allow reserved fields to be
1225 * used by VXLAN extensions if explicitly requested.
1227 if ((flags
& VXLAN_HF_GBP
) && (vs
->flags
& VXLAN_F_GBP
)) {
1228 struct vxlanhdr_gbp
*gbp
;
1230 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1231 md
.gbp
= ntohs(gbp
->policy_id
);
1233 if (gbp
->dont_learn
)
1234 md
.gbp
|= VXLAN_GBP_DONT_LEARN
;
1236 if (gbp
->policy_applied
)
1237 md
.gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1239 flags
&= ~VXLAN_GBP_USED_BITS
;
1242 if (flags
|| vni
& ~VXLAN_VNI_MASK
) {
1243 /* If there are any unprocessed flags remaining treat
1244 * this as a malformed packet. This behavior diverges from
1245 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1246 * in reserved fields are to be ignored. The approach here
1247 * maintains compatbility with previous stack code, and also
1248 * is more robust and provides a little more security in
1249 * adding extensions to VXLAN.
1255 md
.vni
= vxh
->vx_vni
;
1256 vs
->rcv(vs
, skb
, &md
);
1260 /* Consume bad packet */
1265 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1266 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1269 /* Return non vxlan pkt */
1273 static void vxlan_rcv(struct vxlan_sock
*vs
, struct sk_buff
*skb
,
1274 struct vxlan_metadata
*md
)
1276 struct iphdr
*oip
= NULL
;
1277 struct ipv6hdr
*oip6
= NULL
;
1278 struct vxlan_dev
*vxlan
;
1279 struct pcpu_sw_netstats
*stats
;
1280 union vxlan_addr saddr
;
1283 union vxlan_addr
*remote_ip
;
1285 vni
= ntohl(md
->vni
) >> 8;
1286 /* Is this VNI defined? */
1287 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1291 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1292 skb_reset_mac_header(skb
);
1293 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1294 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1295 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1297 /* Ignore packet loops (and multicast echo) */
1298 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1301 /* Re-examine inner Ethernet packet */
1302 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1304 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1305 saddr
.sa
.sa_family
= AF_INET
;
1306 #if IS_ENABLED(CONFIG_IPV6)
1308 oip6
= ipv6_hdr(skb
);
1309 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1310 saddr
.sa
.sa_family
= AF_INET6
;
1314 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1315 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1318 skb_reset_network_header(skb
);
1319 skb
->mark
= md
->gbp
;
1322 err
= IP6_ECN_decapsulate(oip6
, skb
);
1324 err
= IP_ECN_decapsulate(oip
, skb
);
1326 if (unlikely(err
)) {
1327 if (log_ecn_error
) {
1329 net_info_ratelimited("non-ECT from %pI6\n",
1332 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1333 &oip
->saddr
, oip
->tos
);
1336 ++vxlan
->dev
->stats
.rx_frame_errors
;
1337 ++vxlan
->dev
->stats
.rx_errors
;
1342 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1343 u64_stats_update_begin(&stats
->syncp
);
1344 stats
->rx_packets
++;
1345 stats
->rx_bytes
+= skb
->len
;
1346 u64_stats_update_end(&stats
->syncp
);
1352 /* Consume bad packet */
1356 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1358 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1359 struct arphdr
*parp
;
1362 struct neighbour
*n
;
1364 if (dev
->flags
& IFF_NOARP
)
1367 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1368 dev
->stats
.tx_dropped
++;
1371 parp
= arp_hdr(skb
);
1373 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1374 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1375 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1376 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1377 parp
->ar_hln
!= dev
->addr_len
||
1380 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1382 arpptr
+= dev
->addr_len
; /* sha */
1383 memcpy(&sip
, arpptr
, sizeof(sip
));
1384 arpptr
+= sizeof(sip
);
1385 arpptr
+= dev
->addr_len
; /* tha */
1386 memcpy(&tip
, arpptr
, sizeof(tip
));
1388 if (ipv4_is_loopback(tip
) ||
1389 ipv4_is_multicast(tip
))
1392 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1395 struct vxlan_fdb
*f
;
1396 struct sk_buff
*reply
;
1398 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1403 f
= vxlan_find_mac(vxlan
, n
->ha
);
1404 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1405 /* bridge-local neighbor */
1410 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1418 skb_reset_mac_header(reply
);
1419 __skb_pull(reply
, skb_network_offset(reply
));
1420 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1421 reply
->pkt_type
= PACKET_HOST
;
1423 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1424 dev
->stats
.rx_dropped
++;
1425 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1426 union vxlan_addr ipa
= {
1427 .sin
.sin_addr
.s_addr
= tip
,
1428 .sin
.sin_family
= AF_INET
,
1431 vxlan_ip_miss(dev
, &ipa
);
1435 return NETDEV_TX_OK
;
1438 #if IS_ENABLED(CONFIG_IPV6)
1439 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1440 struct neighbour
*n
, bool isrouter
)
1442 struct net_device
*dev
= request
->dev
;
1443 struct sk_buff
*reply
;
1444 struct nd_msg
*ns
, *na
;
1445 struct ipv6hdr
*pip6
;
1447 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1454 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1455 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1456 reply
= alloc_skb(len
, GFP_ATOMIC
);
1460 reply
->protocol
= htons(ETH_P_IPV6
);
1462 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1463 skb_push(reply
, sizeof(struct ethhdr
));
1464 skb_set_mac_header(reply
, 0);
1466 ns
= (struct nd_msg
*)skb_transport_header(request
);
1468 daddr
= eth_hdr(request
)->h_source
;
1469 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1470 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1471 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1472 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1477 /* Ethernet header */
1478 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1479 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1480 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1481 reply
->protocol
= htons(ETH_P_IPV6
);
1483 skb_pull(reply
, sizeof(struct ethhdr
));
1484 skb_set_network_header(reply
, 0);
1485 skb_put(reply
, sizeof(struct ipv6hdr
));
1489 pip6
= ipv6_hdr(reply
);
1490 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1492 pip6
->priority
= ipv6_hdr(request
)->priority
;
1493 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1494 pip6
->hop_limit
= 255;
1495 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1496 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1498 skb_pull(reply
, sizeof(struct ipv6hdr
));
1499 skb_set_transport_header(reply
, 0);
1501 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1503 /* Neighbor Advertisement */
1504 memset(na
, 0, sizeof(*na
)+na_olen
);
1505 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1506 na
->icmph
.icmp6_router
= isrouter
;
1507 na
->icmph
.icmp6_override
= 1;
1508 na
->icmph
.icmp6_solicited
= 1;
1509 na
->target
= ns
->target
;
1510 ether_addr_copy(&na
->opt
[2], n
->ha
);
1511 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1512 na
->opt
[1] = na_olen
>> 3;
1514 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1515 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1516 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1518 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1520 skb_push(reply
, sizeof(struct ipv6hdr
));
1522 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1527 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1529 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1531 const struct ipv6hdr
*iphdr
;
1532 const struct in6_addr
*saddr
, *daddr
;
1533 struct neighbour
*n
;
1534 struct inet6_dev
*in6_dev
;
1536 in6_dev
= __in6_dev_get(dev
);
1540 iphdr
= ipv6_hdr(skb
);
1541 saddr
= &iphdr
->saddr
;
1542 daddr
= &iphdr
->daddr
;
1544 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1545 if (msg
->icmph
.icmp6_code
!= 0 ||
1546 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1549 if (ipv6_addr_loopback(daddr
) ||
1550 ipv6_addr_is_multicast(&msg
->target
))
1553 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1556 struct vxlan_fdb
*f
;
1557 struct sk_buff
*reply
;
1559 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1564 f
= vxlan_find_mac(vxlan
, n
->ha
);
1565 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1566 /* bridge-local neighbor */
1571 reply
= vxlan_na_create(skb
, n
,
1572 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1579 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1580 dev
->stats
.rx_dropped
++;
1582 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1583 union vxlan_addr ipa
= {
1584 .sin6
.sin6_addr
= msg
->target
,
1585 .sin6
.sin6_family
= AF_INET6
,
1588 vxlan_ip_miss(dev
, &ipa
);
1593 return NETDEV_TX_OK
;
1597 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1599 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1600 struct neighbour
*n
;
1602 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1606 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1611 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1614 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1615 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1616 union vxlan_addr ipa
= {
1617 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1618 .sin
.sin_family
= AF_INET
,
1621 vxlan_ip_miss(dev
, &ipa
);
1627 #if IS_ENABLED(CONFIG_IPV6)
1630 struct ipv6hdr
*pip6
;
1632 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1634 pip6
= ipv6_hdr(skb
);
1635 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1636 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1637 union vxlan_addr ipa
= {
1638 .sin6
.sin6_addr
= pip6
->daddr
,
1639 .sin6
.sin6_family
= AF_INET6
,
1642 vxlan_ip_miss(dev
, &ipa
);
1656 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1658 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1660 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1669 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1670 struct vxlan_metadata
*md
)
1672 struct vxlanhdr_gbp
*gbp
;
1677 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1678 vxh
->vx_flags
|= htonl(VXLAN_HF_GBP
);
1680 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1681 gbp
->dont_learn
= 1;
1683 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1684 gbp
->policy_applied
= 1;
1686 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1689 #if IS_ENABLED(CONFIG_IPV6)
1690 static int vxlan6_xmit_skb(struct dst_entry
*dst
, struct sk_buff
*skb
,
1691 struct net_device
*dev
, struct in6_addr
*saddr
,
1692 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1693 __be16 src_port
, __be16 dst_port
,
1694 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1696 struct vxlanhdr
*vxh
;
1699 bool udp_sum
= !(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
1700 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1701 u16 hdrlen
= sizeof(struct vxlanhdr
);
1703 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1704 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1705 int csum_start
= skb_checksum_start_offset(skb
);
1707 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1708 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1709 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1710 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1712 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1716 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1722 skb_scrub_packet(skb
, xnet
);
1724 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1725 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1726 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1728 /* Need space for new headers (invalidates iph ptr) */
1729 err
= skb_cow_head(skb
, min_headroom
);
1730 if (unlikely(err
)) {
1735 skb
= vlan_hwaccel_push_inside(skb
);
1736 if (WARN_ON(!skb
)) {
1741 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1742 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1743 vxh
->vx_vni
= md
->vni
;
1745 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1746 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1749 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1750 data
|= VXLAN_RCO_UDP
;
1752 vxh
->vx_vni
|= htonl(data
);
1753 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1755 if (!skb_is_gso(skb
)) {
1756 skb
->ip_summed
= CHECKSUM_NONE
;
1757 skb
->encapsulation
= 0;
1761 if (vxflags
& VXLAN_F_GBP
)
1762 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1764 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1766 udp_tunnel6_xmit_skb(dst
, skb
, dev
, saddr
, daddr
, prio
,
1767 ttl
, src_port
, dst_port
,
1768 !!(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
));
1776 int vxlan_xmit_skb(struct rtable
*rt
, struct sk_buff
*skb
,
1777 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1778 __be16 src_port
, __be16 dst_port
,
1779 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1781 struct vxlanhdr
*vxh
;
1784 bool udp_sum
= !!(vxflags
& VXLAN_F_UDP_CSUM
);
1785 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1786 u16 hdrlen
= sizeof(struct vxlanhdr
);
1788 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1789 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1790 int csum_start
= skb_checksum_start_offset(skb
);
1792 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1793 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1794 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1795 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1797 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1801 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1803 return PTR_ERR(skb
);
1805 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1806 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1807 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1809 /* Need space for new headers (invalidates iph ptr) */
1810 err
= skb_cow_head(skb
, min_headroom
);
1811 if (unlikely(err
)) {
1816 skb
= vlan_hwaccel_push_inside(skb
);
1820 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1821 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1822 vxh
->vx_vni
= md
->vni
;
1824 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1825 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1828 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1829 data
|= VXLAN_RCO_UDP
;
1831 vxh
->vx_vni
|= htonl(data
);
1832 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1834 if (!skb_is_gso(skb
)) {
1835 skb
->ip_summed
= CHECKSUM_NONE
;
1836 skb
->encapsulation
= 0;
1840 if (vxflags
& VXLAN_F_GBP
)
1841 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1843 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1845 return udp_tunnel_xmit_skb(rt
, skb
, src
, dst
, tos
,
1846 ttl
, df
, src_port
, dst_port
, xnet
,
1847 !(vxflags
& VXLAN_F_UDP_CSUM
));
1849 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1851 /* Bypass encapsulation if the destination is local */
1852 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1853 struct vxlan_dev
*dst_vxlan
)
1855 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1856 union vxlan_addr loopback
;
1857 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1858 struct net_device
*dev
= skb
->dev
;
1861 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1862 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1863 skb
->pkt_type
= PACKET_HOST
;
1864 skb
->encapsulation
= 0;
1865 skb
->dev
= dst_vxlan
->dev
;
1866 __skb_pull(skb
, skb_network_offset(skb
));
1868 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1869 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1870 loopback
.sa
.sa_family
= AF_INET
;
1871 #if IS_ENABLED(CONFIG_IPV6)
1873 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1874 loopback
.sa
.sa_family
= AF_INET6
;
1878 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1879 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1881 u64_stats_update_begin(&tx_stats
->syncp
);
1882 tx_stats
->tx_packets
++;
1883 tx_stats
->tx_bytes
+= len
;
1884 u64_stats_update_end(&tx_stats
->syncp
);
1886 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1887 u64_stats_update_begin(&rx_stats
->syncp
);
1888 rx_stats
->rx_packets
++;
1889 rx_stats
->rx_bytes
+= len
;
1890 u64_stats_update_end(&rx_stats
->syncp
);
1892 dev
->stats
.rx_dropped
++;
1896 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1897 struct vxlan_rdst
*rdst
, bool did_rsc
)
1899 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1900 struct rtable
*rt
= NULL
;
1901 const struct iphdr
*old_iph
;
1903 union vxlan_addr
*dst
;
1904 struct vxlan_metadata md
;
1905 __be16 src_port
= 0, dst_port
;
1911 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1912 vni
= rdst
->remote_vni
;
1913 dst
= &rdst
->remote_ip
;
1915 if (vxlan_addr_any(dst
)) {
1917 /* short-circuited back to local bridge */
1918 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1924 old_iph
= ip_hdr(skb
);
1927 if (!ttl
&& vxlan_addr_multicast(dst
))
1932 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1934 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->port_min
,
1935 vxlan
->port_max
, true);
1937 if (dst
->sa
.sa_family
== AF_INET
) {
1938 memset(&fl4
, 0, sizeof(fl4
));
1939 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1940 fl4
.flowi4_tos
= RT_TOS(tos
);
1941 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1942 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1944 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1946 netdev_dbg(dev
, "no route to %pI4\n",
1947 &dst
->sin
.sin_addr
.s_addr
);
1948 dev
->stats
.tx_carrier_errors
++;
1952 if (rt
->dst
.dev
== dev
) {
1953 netdev_dbg(dev
, "circular route to %pI4\n",
1954 &dst
->sin
.sin_addr
.s_addr
);
1955 dev
->stats
.collisions
++;
1959 /* Bypass encapsulation if the destination is local */
1960 if (rt
->rt_flags
& RTCF_LOCAL
&&
1961 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1962 struct vxlan_dev
*dst_vxlan
;
1965 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
1966 dst
->sa
.sa_family
, dst_port
,
1970 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1974 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1975 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1976 md
.vni
= htonl(vni
<< 8);
1979 err
= vxlan_xmit_skb(rt
, skb
, fl4
.saddr
,
1980 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
1981 src_port
, dst_port
, &md
,
1982 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
1985 /* skb is already freed. */
1990 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1991 #if IS_ENABLED(CONFIG_IPV6)
1993 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1994 struct dst_entry
*ndst
;
1998 memset(&fl6
, 0, sizeof(fl6
));
1999 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
2000 fl6
.daddr
= dst
->sin6
.sin6_addr
;
2001 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
2002 fl6
.flowi6_proto
= IPPROTO_UDP
;
2004 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
2005 netdev_dbg(dev
, "no route to %pI6\n",
2006 &dst
->sin6
.sin6_addr
);
2007 dev
->stats
.tx_carrier_errors
++;
2011 if (ndst
->dev
== dev
) {
2012 netdev_dbg(dev
, "circular route to %pI6\n",
2013 &dst
->sin6
.sin6_addr
);
2015 dev
->stats
.collisions
++;
2019 /* Bypass encapsulation if the destination is local */
2020 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2021 if (flags
& RTF_LOCAL
&&
2022 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2023 struct vxlan_dev
*dst_vxlan
;
2026 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2027 dst
->sa
.sa_family
, dst_port
,
2031 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2035 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2036 md
.vni
= htonl(vni
<< 8);
2039 err
= vxlan6_xmit_skb(ndst
, skb
, dev
, &fl6
.saddr
, &fl6
.daddr
,
2040 0, ttl
, src_port
, dst_port
, &md
,
2041 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2049 dev
->stats
.tx_dropped
++;
2055 dev
->stats
.tx_errors
++;
2060 /* Transmit local packets over Vxlan
2062 * Outer IP header inherits ECN and DF from inner header.
2063 * Outer UDP destination is the VXLAN assigned port.
2064 * source port is based on hash of flow
2066 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2068 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2070 bool did_rsc
= false;
2071 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2072 struct vxlan_fdb
*f
;
2074 skb_reset_mac_header(skb
);
2077 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2078 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2079 return arp_reduce(dev
, skb
);
2080 #if IS_ENABLED(CONFIG_IPV6)
2081 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2082 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2083 + sizeof(struct nd_msg
)) &&
2084 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2087 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2088 if (msg
->icmph
.icmp6_code
== 0 &&
2089 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2090 return neigh_reduce(dev
, skb
);
2096 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2099 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2100 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2101 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2102 did_rsc
= route_shortcircuit(dev
, skb
);
2104 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2108 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2110 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2111 !is_multicast_ether_addr(eth
->h_dest
))
2112 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2114 dev
->stats
.tx_dropped
++;
2116 return NETDEV_TX_OK
;
2120 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2121 struct sk_buff
*skb1
;
2127 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2129 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2133 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2136 return NETDEV_TX_OK
;
2139 /* Walk the forwarding table and purge stale entries */
2140 static void vxlan_cleanup(unsigned long arg
)
2142 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2143 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2146 if (!netif_running(vxlan
->dev
))
2149 spin_lock_bh(&vxlan
->hash_lock
);
2150 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2151 struct hlist_node
*p
, *n
;
2152 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2154 = container_of(p
, struct vxlan_fdb
, hlist
);
2155 unsigned long timeout
;
2157 if (f
->state
& NUD_PERMANENT
)
2160 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2161 if (time_before_eq(timeout
, jiffies
)) {
2162 netdev_dbg(vxlan
->dev
,
2163 "garbage collect %pM\n",
2165 f
->state
= NUD_STALE
;
2166 vxlan_fdb_destroy(vxlan
, f
);
2167 } else if (time_before(timeout
, next_timer
))
2168 next_timer
= timeout
;
2171 spin_unlock_bh(&vxlan
->hash_lock
);
2173 mod_timer(&vxlan
->age_timer
, next_timer
);
2176 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2178 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2180 vxlan
->vn_sock
= vs
;
2181 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2184 /* Setup stats when device is created */
2185 static int vxlan_init(struct net_device
*dev
)
2187 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2188 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2189 struct vxlan_sock
*vs
;
2190 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2192 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2196 spin_lock(&vn
->sock_lock
);
2197 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2198 vxlan
->dst_port
, vxlan
->flags
);
2199 if (vs
&& atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2200 /* If we have a socket with same port already, reuse it */
2201 vxlan_vs_add_dev(vs
, vxlan
);
2203 /* otherwise make new socket outside of RTNL */
2205 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2207 spin_unlock(&vn
->sock_lock
);
2212 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2214 struct vxlan_fdb
*f
;
2216 spin_lock_bh(&vxlan
->hash_lock
);
2217 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2219 vxlan_fdb_destroy(vxlan
, f
);
2220 spin_unlock_bh(&vxlan
->hash_lock
);
2223 static void vxlan_uninit(struct net_device
*dev
)
2225 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2226 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2228 vxlan_fdb_delete_default(vxlan
);
2231 vxlan_sock_release(vs
);
2232 free_percpu(dev
->tstats
);
2235 /* Start ageing timer and join group when device is brought up */
2236 static int vxlan_open(struct net_device
*dev
)
2238 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2239 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2241 /* socket hasn't been created */
2245 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2246 vxlan_sock_hold(vs
);
2248 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2251 if (vxlan
->age_interval
)
2252 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2257 /* Purge the forwarding table */
2258 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2262 spin_lock_bh(&vxlan
->hash_lock
);
2263 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2264 struct hlist_node
*p
, *n
;
2265 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2267 = container_of(p
, struct vxlan_fdb
, hlist
);
2268 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2269 if (!is_zero_ether_addr(f
->eth_addr
))
2270 vxlan_fdb_destroy(vxlan
, f
);
2273 spin_unlock_bh(&vxlan
->hash_lock
);
2276 /* Cleanup timer and forwarding table on shutdown */
2277 static int vxlan_stop(struct net_device
*dev
)
2279 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2280 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2281 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2283 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2284 !vxlan_group_used(vn
, vxlan
)) {
2285 vxlan_sock_hold(vs
);
2287 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2290 del_timer_sync(&vxlan
->age_timer
);
2297 /* Stub, nothing needs to be done. */
2298 static void vxlan_set_multicast_list(struct net_device
*dev
)
2302 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2304 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2305 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2306 struct net_device
*lowerdev
;
2309 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2310 if (lowerdev
== NULL
)
2311 return eth_change_mtu(dev
, new_mtu
);
2313 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2314 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2316 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2318 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2325 static const struct net_device_ops vxlan_netdev_ops
= {
2326 .ndo_init
= vxlan_init
,
2327 .ndo_uninit
= vxlan_uninit
,
2328 .ndo_open
= vxlan_open
,
2329 .ndo_stop
= vxlan_stop
,
2330 .ndo_start_xmit
= vxlan_xmit
,
2331 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2332 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2333 .ndo_change_mtu
= vxlan_change_mtu
,
2334 .ndo_validate_addr
= eth_validate_addr
,
2335 .ndo_set_mac_address
= eth_mac_addr
,
2336 .ndo_fdb_add
= vxlan_fdb_add
,
2337 .ndo_fdb_del
= vxlan_fdb_delete
,
2338 .ndo_fdb_dump
= vxlan_fdb_dump
,
2341 /* Info for udev, that this is a virtual tunnel endpoint */
2342 static struct device_type vxlan_type
= {
2346 /* Calls the ndo_add_vxlan_port of the caller in order to
2347 * supply the listening VXLAN udp ports. Callers are expected
2348 * to implement the ndo_add_vxlan_port.
2350 void vxlan_get_rx_port(struct net_device
*dev
)
2352 struct vxlan_sock
*vs
;
2353 struct net
*net
= dev_net(dev
);
2354 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2355 sa_family_t sa_family
;
2359 spin_lock(&vn
->sock_lock
);
2360 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2361 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2362 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2363 sa_family
= vs
->sock
->sk
->sk_family
;
2364 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2368 spin_unlock(&vn
->sock_lock
);
2370 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2372 /* Initialize the device structure. */
2373 static void vxlan_setup(struct net_device
*dev
)
2375 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2378 eth_hw_addr_random(dev
);
2380 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2381 dev
->needed_headroom
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2383 dev
->needed_headroom
= ETH_HLEN
+ VXLAN_HEADROOM
;
2385 dev
->netdev_ops
= &vxlan_netdev_ops
;
2386 dev
->destructor
= free_netdev
;
2387 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2389 dev
->tx_queue_len
= 0;
2390 dev
->features
|= NETIF_F_LLTX
;
2391 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2392 dev
->features
|= NETIF_F_RXCSUM
;
2393 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2395 dev
->vlan_features
= dev
->features
;
2396 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2397 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2398 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2399 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2400 netif_keep_dst(dev
);
2401 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2403 INIT_LIST_HEAD(&vxlan
->next
);
2404 spin_lock_init(&vxlan
->hash_lock
);
2405 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2406 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2407 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2409 init_timer_deferrable(&vxlan
->age_timer
);
2410 vxlan
->age_timer
.function
= vxlan_cleanup
;
2411 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2413 vxlan
->dst_port
= htons(vxlan_port
);
2417 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2418 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2421 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2422 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2423 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2424 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2425 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2426 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2427 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2428 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2429 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2430 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2431 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2432 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2433 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2434 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2435 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2436 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2437 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2438 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2439 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2440 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2441 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2442 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2443 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2444 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2445 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2448 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2450 if (tb
[IFLA_ADDRESS
]) {
2451 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2452 pr_debug("invalid link address (not ethernet)\n");
2456 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2457 pr_debug("invalid all zero ethernet address\n");
2458 return -EADDRNOTAVAIL
;
2465 if (data
[IFLA_VXLAN_ID
]) {
2466 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2467 if (id
>= VXLAN_VID_MASK
)
2471 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2472 const struct ifla_vxlan_port_range
*p
2473 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2475 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2476 pr_debug("port range %u .. %u not valid\n",
2477 ntohs(p
->low
), ntohs(p
->high
));
2485 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2486 struct ethtool_drvinfo
*drvinfo
)
2488 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2489 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2492 static const struct ethtool_ops vxlan_ethtool_ops
= {
2493 .get_drvinfo
= vxlan_get_drvinfo
,
2494 .get_link
= ethtool_op_get_link
,
2497 static void vxlan_del_work(struct work_struct
*work
)
2499 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2500 udp_tunnel_sock_release(vs
->sock
);
2504 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2505 __be16 port
, u32 flags
)
2507 struct socket
*sock
;
2508 struct udp_port_cfg udp_conf
;
2511 memset(&udp_conf
, 0, sizeof(udp_conf
));
2514 udp_conf
.family
= AF_INET6
;
2515 udp_conf
.use_udp6_rx_checksums
=
2516 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2518 udp_conf
.family
= AF_INET
;
2519 udp_conf
.local_ip
.s_addr
= INADDR_ANY
;
2522 udp_conf
.local_udp_port
= port
;
2524 /* Open UDP socket */
2525 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2527 return ERR_PTR(err
);
2532 /* Create new listen socket if needed */
2533 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2534 vxlan_rcv_t
*rcv
, void *data
,
2537 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2538 struct vxlan_sock
*vs
;
2539 struct socket
*sock
;
2541 bool ipv6
= !!(flags
& VXLAN_F_IPV6
);
2542 struct udp_tunnel_sock_cfg tunnel_cfg
;
2544 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2546 return ERR_PTR(-ENOMEM
);
2548 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2549 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2551 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2553 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2556 return ERR_CAST(sock
);
2560 atomic_set(&vs
->refcnt
, 1);
2563 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2565 /* Initialize the vxlan udp offloads structure */
2566 vs
->udp_offloads
.port
= port
;
2567 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2568 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2570 spin_lock(&vn
->sock_lock
);
2571 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2572 vxlan_notify_add_rx_port(vs
);
2573 spin_unlock(&vn
->sock_lock
);
2575 /* Mark socket as an encapsulation socket. */
2576 tunnel_cfg
.sk_user_data
= vs
;
2577 tunnel_cfg
.encap_type
= 1;
2578 tunnel_cfg
.encap_rcv
= vxlan_udp_encap_recv
;
2579 tunnel_cfg
.encap_destroy
= NULL
;
2581 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2586 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2587 vxlan_rcv_t
*rcv
, void *data
,
2588 bool no_share
, u32 flags
)
2590 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2591 struct vxlan_sock
*vs
;
2592 bool ipv6
= flags
& VXLAN_F_IPV6
;
2594 vs
= vxlan_socket_create(net
, port
, rcv
, data
, flags
);
2598 if (no_share
) /* Return error if sharing is not allowed. */
2601 spin_lock(&vn
->sock_lock
);
2602 vs
= vxlan_find_sock(net
, ipv6
? AF_INET6
: AF_INET
, port
, flags
);
2603 if (vs
&& ((vs
->rcv
!= rcv
) ||
2604 !atomic_add_unless(&vs
->refcnt
, 1, 0)))
2605 vs
= ERR_PTR(-EBUSY
);
2606 spin_unlock(&vn
->sock_lock
);
2609 vs
= ERR_PTR(-EINVAL
);
2613 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2615 /* Scheduled at device creation to bind to a socket */
2616 static void vxlan_sock_work(struct work_struct
*work
)
2618 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2619 struct net
*net
= vxlan
->net
;
2620 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2621 __be16 port
= vxlan
->dst_port
;
2622 struct vxlan_sock
*nvs
;
2624 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
);
2625 spin_lock(&vn
->sock_lock
);
2627 vxlan_vs_add_dev(nvs
, vxlan
);
2628 spin_unlock(&vn
->sock_lock
);
2630 dev_put(vxlan
->dev
);
2633 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
2634 struct nlattr
*tb
[], struct nlattr
*data
[])
2636 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2637 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2638 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2641 bool use_ipv6
= false;
2643 if (!data
[IFLA_VXLAN_ID
])
2646 vxlan
->net
= src_net
;
2648 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2649 dst
->remote_vni
= vni
;
2651 /* Unless IPv6 is explicitly requested, assume IPv4 */
2652 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2653 if (data
[IFLA_VXLAN_GROUP
]) {
2654 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2655 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2656 if (!IS_ENABLED(CONFIG_IPV6
))
2657 return -EPFNOSUPPORT
;
2659 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2660 sizeof(struct in6_addr
));
2661 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2665 if (data
[IFLA_VXLAN_LOCAL
]) {
2666 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2667 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2668 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2669 if (!IS_ENABLED(CONFIG_IPV6
))
2670 return -EPFNOSUPPORT
;
2672 /* TODO: respect scope id */
2673 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2674 sizeof(struct in6_addr
));
2675 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2679 if (data
[IFLA_VXLAN_LINK
] &&
2680 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2681 struct net_device
*lowerdev
2682 = __dev_get_by_index(src_net
, dst
->remote_ifindex
);
2685 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2689 #if IS_ENABLED(CONFIG_IPV6)
2691 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2692 if (idev
&& idev
->cnf
.disable_ipv6
) {
2693 pr_info("IPv6 is disabled via sysctl\n");
2696 vxlan
->flags
|= VXLAN_F_IPV6
;
2701 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2703 dev
->needed_headroom
= lowerdev
->hard_header_len
+
2704 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2705 } else if (use_ipv6
)
2706 vxlan
->flags
|= VXLAN_F_IPV6
;
2708 if (data
[IFLA_VXLAN_TOS
])
2709 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2711 if (data
[IFLA_VXLAN_TTL
])
2712 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2714 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2715 vxlan
->flags
|= VXLAN_F_LEARN
;
2717 if (data
[IFLA_VXLAN_AGEING
])
2718 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2720 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2722 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2723 vxlan
->flags
|= VXLAN_F_PROXY
;
2725 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2726 vxlan
->flags
|= VXLAN_F_RSC
;
2728 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2729 vxlan
->flags
|= VXLAN_F_L2MISS
;
2731 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2732 vxlan
->flags
|= VXLAN_F_L3MISS
;
2734 if (data
[IFLA_VXLAN_LIMIT
])
2735 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2737 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2738 const struct ifla_vxlan_port_range
*p
2739 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2740 vxlan
->port_min
= ntohs(p
->low
);
2741 vxlan
->port_max
= ntohs(p
->high
);
2744 if (data
[IFLA_VXLAN_PORT
])
2745 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2747 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2748 vxlan
->flags
|= VXLAN_F_UDP_CSUM
;
2750 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2751 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2752 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2754 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2755 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2756 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2758 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
2759 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
2760 vxlan
->flags
|= VXLAN_F_REMCSUM_TX
;
2762 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
2763 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
2764 vxlan
->flags
|= VXLAN_F_REMCSUM_RX
;
2766 if (data
[IFLA_VXLAN_GBP
])
2767 vxlan
->flags
|= VXLAN_F_GBP
;
2769 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
])
2770 vxlan
->flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
2772 if (vxlan_find_vni(src_net
, vni
, use_ipv6
? AF_INET6
: AF_INET
,
2773 vxlan
->dst_port
, vxlan
->flags
)) {
2774 pr_info("duplicate VNI %u\n", vni
);
2778 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2780 /* create an fdb entry for a valid default destination */
2781 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2782 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2783 &vxlan
->default_dst
.remote_ip
,
2784 NUD_REACHABLE
|NUD_PERMANENT
,
2785 NLM_F_EXCL
|NLM_F_CREATE
,
2787 vxlan
->default_dst
.remote_vni
,
2788 vxlan
->default_dst
.remote_ifindex
,
2794 err
= register_netdevice(dev
);
2796 vxlan_fdb_delete_default(vxlan
);
2800 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2805 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2807 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2808 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2810 spin_lock(&vn
->sock_lock
);
2811 if (!hlist_unhashed(&vxlan
->hlist
))
2812 hlist_del_rcu(&vxlan
->hlist
);
2813 spin_unlock(&vn
->sock_lock
);
2815 list_del(&vxlan
->next
);
2816 unregister_netdevice_queue(dev
, head
);
2819 static size_t vxlan_get_size(const struct net_device
*dev
)
2822 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2823 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2824 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2825 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2826 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2827 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2828 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2829 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2830 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2831 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2832 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2833 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2834 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2835 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2836 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2837 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
2838 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2839 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2840 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
2841 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
2845 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2847 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2848 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2849 struct ifla_vxlan_port_range ports
= {
2850 .low
= htons(vxlan
->port_min
),
2851 .high
= htons(vxlan
->port_max
),
2854 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2855 goto nla_put_failure
;
2857 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2858 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2859 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2860 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2861 goto nla_put_failure
;
2862 #if IS_ENABLED(CONFIG_IPV6)
2864 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2865 &dst
->remote_ip
.sin6
.sin6_addr
))
2866 goto nla_put_failure
;
2871 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2872 goto nla_put_failure
;
2874 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2875 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2876 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2877 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2878 goto nla_put_failure
;
2879 #if IS_ENABLED(CONFIG_IPV6)
2881 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2882 &vxlan
->saddr
.sin6
.sin6_addr
))
2883 goto nla_put_failure
;
2888 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2889 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2890 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2891 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2892 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2893 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2894 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2895 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2896 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2897 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2898 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2899 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2900 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2901 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
) ||
2902 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
2903 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
2904 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
2905 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
2906 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
2907 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
2908 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
2909 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
2910 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
2911 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
2912 goto nla_put_failure
;
2914 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2915 goto nla_put_failure
;
2917 if (vxlan
->flags
& VXLAN_F_GBP
&&
2918 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
2919 goto nla_put_failure
;
2921 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
2922 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
2923 goto nla_put_failure
;
2931 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
2933 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2938 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2940 .maxtype
= IFLA_VXLAN_MAX
,
2941 .policy
= vxlan_policy
,
2942 .priv_size
= sizeof(struct vxlan_dev
),
2943 .setup
= vxlan_setup
,
2944 .validate
= vxlan_validate
,
2945 .newlink
= vxlan_newlink
,
2946 .dellink
= vxlan_dellink
,
2947 .get_size
= vxlan_get_size
,
2948 .fill_info
= vxlan_fill_info
,
2949 .get_link_net
= vxlan_get_link_net
,
2952 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2953 struct net_device
*dev
)
2955 struct vxlan_dev
*vxlan
, *next
;
2956 LIST_HEAD(list_kill
);
2958 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2959 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2961 /* In case we created vxlan device with carrier
2962 * and we loose the carrier due to module unload
2963 * we also need to remove vxlan device. In other
2964 * cases, it's not necessary and remote_ifindex
2965 * is 0 here, so no matches.
2967 if (dst
->remote_ifindex
== dev
->ifindex
)
2968 vxlan_dellink(vxlan
->dev
, &list_kill
);
2971 unregister_netdevice_many(&list_kill
);
2974 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2975 unsigned long event
, void *ptr
)
2977 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2978 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2980 if (event
== NETDEV_UNREGISTER
)
2981 vxlan_handle_lowerdev_unregister(vn
, dev
);
2986 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2987 .notifier_call
= vxlan_lowerdev_event
,
2990 static __net_init
int vxlan_init_net(struct net
*net
)
2992 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2995 INIT_LIST_HEAD(&vn
->vxlan_list
);
2996 spin_lock_init(&vn
->sock_lock
);
2998 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2999 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3004 static void __net_exit
vxlan_exit_net(struct net
*net
)
3006 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3007 struct vxlan_dev
*vxlan
, *next
;
3008 struct net_device
*dev
, *aux
;
3012 for_each_netdev_safe(net
, dev
, aux
)
3013 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3014 unregister_netdevice_queue(dev
, &list
);
3016 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3017 /* If vxlan->dev is in the same netns, it has already been added
3018 * to the list by the previous loop.
3020 if (!net_eq(dev_net(vxlan
->dev
), net
))
3021 unregister_netdevice_queue(dev
, &list
);
3024 unregister_netdevice_many(&list
);
3028 static struct pernet_operations vxlan_net_ops
= {
3029 .init
= vxlan_init_net
,
3030 .exit
= vxlan_exit_net
,
3031 .id
= &vxlan_net_id
,
3032 .size
= sizeof(struct vxlan_net
),
3035 static int __init
vxlan_init_module(void)
3039 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3043 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3045 rc
= register_pernet_subsys(&vxlan_net_ops
);
3049 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3053 rc
= rtnl_link_register(&vxlan_link_ops
);
3059 unregister_netdevice_notifier(&vxlan_notifier_block
);
3061 unregister_pernet_subsys(&vxlan_net_ops
);
3063 destroy_workqueue(vxlan_wq
);
3066 late_initcall(vxlan_init_module
);
3068 static void __exit
vxlan_cleanup_module(void)
3070 rtnl_link_unregister(&vxlan_link_ops
);
3071 unregister_netdevice_notifier(&vxlan_notifier_block
);
3072 destroy_workqueue(vxlan_wq
);
3073 unregister_pernet_subsys(&vxlan_net_ops
);
3074 /* rcu_barrier() is called by netns */
3076 module_exit(vxlan_cleanup_module
);
3078 MODULE_LICENSE("GPL");
3079 MODULE_VERSION(VXLAN_VERSION
);
3080 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3081 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3082 MODULE_ALIAS_RTNL_LINK("vxlan");