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>
52 #include <net/dst_metadata.h>
54 #define VXLAN_VERSION "0.1"
56 #define PORT_HASH_BITS 8
57 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61 /* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
63 * for compatibility with early adopters.
65 static unsigned short vxlan_port __read_mostly
= 8472;
66 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
67 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
69 static bool log_ecn_error
= true;
70 module_param(log_ecn_error
, bool, 0644);
71 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
73 static int vxlan_net_id
;
74 static struct rtnl_link_ops vxlan_link_ops
;
76 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
78 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
80 /* per-network namespace private data for this module */
82 struct list_head vxlan_list
;
83 struct hlist_head sock_list
[PORT_HASH_SIZE
];
87 /* Forwarding table entry */
89 struct hlist_node hlist
; /* linked list of entries */
91 unsigned long updated
; /* jiffies */
93 struct list_head remotes
;
94 u8 eth_addr
[ETH_ALEN
];
95 u16 state
; /* see ndm_state */
96 u8 flags
; /* see ndm_flags */
99 /* salt for hash table */
100 static u32 vxlan_salt __read_mostly
;
101 static struct workqueue_struct
*vxlan_wq
;
103 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
105 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
106 ip_tunnel_collect_metadata();
109 #if IS_ENABLED(CONFIG_IPV6)
111 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
113 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
115 if (a
->sa
.sa_family
== AF_INET6
)
116 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
118 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
121 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
123 if (ipa
->sa
.sa_family
== AF_INET6
)
124 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
126 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
129 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
131 if (ipa
->sa
.sa_family
== AF_INET6
)
132 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
134 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
137 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
139 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
140 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
141 ip
->sa
.sa_family
= AF_INET6
;
143 } else if (nla_len(nla
) >= sizeof(__be32
)) {
144 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
145 ip
->sa
.sa_family
= AF_INET
;
148 return -EAFNOSUPPORT
;
152 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
153 const union vxlan_addr
*ip
)
155 if (ip
->sa
.sa_family
== AF_INET6
)
156 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
158 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
161 #else /* !CONFIG_IPV6 */
164 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
166 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
169 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
171 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
174 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
176 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
179 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
181 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
182 return -EAFNOSUPPORT
;
183 } else if (nla_len(nla
) >= sizeof(__be32
)) {
184 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(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 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
199 /* Virtual Network hash table head */
200 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
202 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
205 /* Socket hash table head */
206 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
208 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
210 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
213 /* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
216 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
218 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
221 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
223 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
226 /* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
229 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
230 __be16 port
, u32 flags
)
232 struct vxlan_sock
*vs
;
234 flags
&= VXLAN_F_RCV_FLAGS
;
236 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
237 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
238 vxlan_get_sk_family(vs
) == family
&&
245 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
247 struct vxlan_dev
*vxlan
;
249 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
250 if (vxlan
->default_dst
.remote_vni
== id
)
257 /* Look up VNI in a per net namespace table */
258 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
,
259 sa_family_t family
, __be16 port
,
262 struct vxlan_sock
*vs
;
264 vs
= vxlan_find_sock(net
, family
, port
, flags
);
268 return vxlan_vs_find_vni(vs
, id
);
271 /* Fill in neighbour message in skbuff. */
272 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
273 const struct vxlan_fdb
*fdb
,
274 u32 portid
, u32 seq
, int type
, unsigned int flags
,
275 const struct vxlan_rdst
*rdst
)
277 unsigned long now
= jiffies
;
278 struct nda_cacheinfo ci
;
279 struct nlmsghdr
*nlh
;
281 bool send_ip
, send_eth
;
283 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
287 ndm
= nlmsg_data(nlh
);
288 memset(ndm
, 0, sizeof(*ndm
));
290 send_eth
= send_ip
= true;
292 if (type
== RTM_GETNEIGH
) {
293 ndm
->ndm_family
= AF_INET
;
294 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
295 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
297 ndm
->ndm_family
= AF_BRIDGE
;
298 ndm
->ndm_state
= fdb
->state
;
299 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
300 ndm
->ndm_flags
= fdb
->flags
;
301 ndm
->ndm_type
= RTN_UNICAST
;
303 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
304 nla_put_s32(skb
, NDA_LINK_NETNSID
,
305 peernet2id_alloc(dev_net(vxlan
->dev
), vxlan
->net
)))
306 goto nla_put_failure
;
308 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
309 goto nla_put_failure
;
311 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
312 goto nla_put_failure
;
314 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
315 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
316 goto nla_put_failure
;
317 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
318 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
319 goto nla_put_failure
;
320 if (rdst
->remote_ifindex
&&
321 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
322 goto nla_put_failure
;
324 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
325 ci
.ndm_confirmed
= 0;
326 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
329 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
330 goto nla_put_failure
;
336 nlmsg_cancel(skb
, nlh
);
340 static inline size_t vxlan_nlmsg_size(void)
342 return NLMSG_ALIGN(sizeof(struct ndmsg
))
343 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
344 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
345 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
346 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
347 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
348 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
349 + nla_total_size(sizeof(struct nda_cacheinfo
));
352 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
353 struct vxlan_rdst
*rd
, int type
)
355 struct net
*net
= dev_net(vxlan
->dev
);
359 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
363 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
365 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
366 WARN_ON(err
== -EMSGSIZE
);
371 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
375 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
378 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
380 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
381 struct vxlan_fdb f
= {
384 struct vxlan_rdst remote
= {
385 .remote_ip
= *ipa
, /* goes to NDA_DST */
386 .remote_vni
= VXLAN_N_VID
,
389 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
392 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
394 struct vxlan_fdb f
= {
397 struct vxlan_rdst remote
= { };
399 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
401 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
404 /* Hash Ethernet address */
405 static u32
eth_hash(const unsigned char *addr
)
407 u64 value
= get_unaligned((u64
*)addr
);
409 /* only want 6 bytes */
415 return hash_64(value
, FDB_HASH_BITS
);
418 /* Hash chain to use given mac address */
419 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
422 return &vxlan
->fdb_head
[eth_hash(mac
)];
425 /* Look up Ethernet address in forwarding table */
426 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
429 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
432 hlist_for_each_entry_rcu(f
, head
, hlist
) {
433 if (ether_addr_equal(mac
, f
->eth_addr
))
440 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
445 f
= __vxlan_find_mac(vxlan
, mac
);
452 /* caller should hold vxlan->hash_lock */
453 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
454 union vxlan_addr
*ip
, __be16 port
,
455 __u32 vni
, __u32 ifindex
)
457 struct vxlan_rdst
*rd
;
459 list_for_each_entry(rd
, &f
->remotes
, list
) {
460 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
461 rd
->remote_port
== port
&&
462 rd
->remote_vni
== vni
&&
463 rd
->remote_ifindex
== ifindex
)
470 /* Replace destination of unicast mac */
471 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
472 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
474 struct vxlan_rdst
*rd
;
476 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
480 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
484 rd
->remote_port
= port
;
485 rd
->remote_vni
= vni
;
486 rd
->remote_ifindex
= ifindex
;
490 /* Add/update destinations for multicast */
491 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
492 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
493 __u32 ifindex
, struct vxlan_rdst
**rdp
)
495 struct vxlan_rdst
*rd
;
497 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
501 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
505 rd
->remote_port
= port
;
506 rd
->remote_vni
= vni
;
507 rd
->remote_ifindex
= ifindex
;
509 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
515 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
517 struct vxlanhdr
*vh
, size_t hdrlen
,
518 u32 data
, struct gro_remcsum
*grc
,
521 size_t start
, offset
;
523 if (skb
->remcsum_offload
)
526 if (!NAPI_GRO_CB(skb
)->csum_valid
)
529 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
530 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
531 offsetof(struct udphdr
, check
) :
532 offsetof(struct tcphdr
, check
));
534 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
535 start
, offset
, grc
, nopartial
);
537 skb
->remcsum_offload
= 1;
542 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
,
544 struct udp_offload
*uoff
)
546 struct sk_buff
*p
, **pp
= NULL
;
547 struct vxlanhdr
*vh
, *vh2
;
548 unsigned int hlen
, off_vx
;
550 struct vxlan_sock
*vs
= container_of(uoff
, struct vxlan_sock
,
553 struct gro_remcsum grc
;
555 skb_gro_remcsum_init(&grc
);
557 off_vx
= skb_gro_offset(skb
);
558 hlen
= off_vx
+ sizeof(*vh
);
559 vh
= skb_gro_header_fast(skb
, off_vx
);
560 if (skb_gro_header_hard(skb
, hlen
)) {
561 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
566 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
568 flags
= ntohl(vh
->vx_flags
);
570 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
571 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
572 ntohl(vh
->vx_vni
), &grc
,
574 VXLAN_F_REMCSUM_NOPARTIAL
));
580 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
584 for (p
= *head
; p
; p
= p
->next
) {
585 if (!NAPI_GRO_CB(p
)->same_flow
)
588 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
589 if (vh
->vx_flags
!= vh2
->vx_flags
||
590 vh
->vx_vni
!= vh2
->vx_vni
) {
591 NAPI_GRO_CB(p
)->same_flow
= 0;
596 pp
= eth_gro_receive(head
, skb
);
599 skb_gro_remcsum_cleanup(skb
, &grc
);
600 NAPI_GRO_CB(skb
)->flush
|= flush
;
605 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
606 struct udp_offload
*uoff
)
608 udp_tunnel_gro_complete(skb
, nhoff
);
610 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
613 /* Notify netdevs that UDP port started listening */
614 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
616 struct net_device
*dev
;
617 struct sock
*sk
= vs
->sock
->sk
;
618 struct net
*net
= sock_net(sk
);
619 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
620 __be16 port
= inet_sk(sk
)->inet_sport
;
623 if (sa_family
== AF_INET
) {
624 err
= udp_add_offload(net
, &vs
->udp_offloads
);
626 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
630 for_each_netdev_rcu(net
, dev
) {
631 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
632 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
638 /* Notify netdevs that UDP port is no more listening */
639 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
641 struct net_device
*dev
;
642 struct sock
*sk
= vs
->sock
->sk
;
643 struct net
*net
= sock_net(sk
);
644 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
645 __be16 port
= inet_sk(sk
)->inet_sport
;
648 for_each_netdev_rcu(net
, dev
) {
649 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
650 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
655 if (sa_family
== AF_INET
)
656 udp_del_offload(&vs
->udp_offloads
);
659 /* Add new entry to forwarding table -- assumes lock held */
660 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
661 const u8
*mac
, union vxlan_addr
*ip
,
662 __u16 state
, __u16 flags
,
663 __be16 port
, __u32 vni
, __u32 ifindex
,
666 struct vxlan_rdst
*rd
= NULL
;
670 f
= __vxlan_find_mac(vxlan
, mac
);
672 if (flags
& NLM_F_EXCL
) {
673 netdev_dbg(vxlan
->dev
,
674 "lost race to create %pM\n", mac
);
677 if (f
->state
!= state
) {
679 f
->updated
= jiffies
;
682 if (f
->flags
!= ndm_flags
) {
683 f
->flags
= ndm_flags
;
684 f
->updated
= jiffies
;
687 if ((flags
& NLM_F_REPLACE
)) {
688 /* Only change unicasts */
689 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
690 is_zero_ether_addr(f
->eth_addr
))) {
691 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
696 if ((flags
& NLM_F_APPEND
) &&
697 (is_multicast_ether_addr(f
->eth_addr
) ||
698 is_zero_ether_addr(f
->eth_addr
))) {
699 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
707 if (!(flags
& NLM_F_CREATE
))
710 if (vxlan
->cfg
.addrmax
&&
711 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
714 /* Disallow replace to add a multicast entry */
715 if ((flags
& NLM_F_REPLACE
) &&
716 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
719 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
720 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
726 f
->flags
= ndm_flags
;
727 f
->updated
= f
->used
= jiffies
;
728 INIT_LIST_HEAD(&f
->remotes
);
729 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
731 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
734 hlist_add_head_rcu(&f
->hlist
,
735 vxlan_fdb_head(vxlan
, mac
));
740 rd
= first_remote_rtnl(f
);
741 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
747 static void vxlan_fdb_free(struct rcu_head
*head
)
749 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
750 struct vxlan_rdst
*rd
, *nd
;
752 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
757 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
759 netdev_dbg(vxlan
->dev
,
760 "delete %pM\n", f
->eth_addr
);
763 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
765 hlist_del_rcu(&f
->hlist
);
766 call_rcu(&f
->rcu
, vxlan_fdb_free
);
769 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
770 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
772 struct net
*net
= dev_net(vxlan
->dev
);
776 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
780 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
781 if (remote
->sa
.sa_family
== AF_INET
) {
782 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
783 ip
->sa
.sa_family
= AF_INET
;
784 #if IS_ENABLED(CONFIG_IPV6)
786 ip
->sin6
.sin6_addr
= in6addr_any
;
787 ip
->sa
.sa_family
= AF_INET6
;
793 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
795 *port
= nla_get_be16(tb
[NDA_PORT
]);
797 *port
= vxlan
->cfg
.dst_port
;
801 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
803 *vni
= nla_get_u32(tb
[NDA_VNI
]);
805 *vni
= vxlan
->default_dst
.remote_vni
;
808 if (tb
[NDA_IFINDEX
]) {
809 struct net_device
*tdev
;
811 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
813 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
814 tdev
= __dev_get_by_index(net
, *ifindex
);
816 return -EADDRNOTAVAIL
;
824 /* Add static entry (via netlink) */
825 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
826 struct net_device
*dev
,
827 const unsigned char *addr
, u16 vid
, u16 flags
)
829 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
830 /* struct net *net = dev_net(vxlan->dev); */
836 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
837 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
842 if (tb
[NDA_DST
] == NULL
)
845 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
849 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
850 return -EAFNOSUPPORT
;
852 spin_lock_bh(&vxlan
->hash_lock
);
853 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
854 port
, vni
, ifindex
, ndm
->ndm_flags
);
855 spin_unlock_bh(&vxlan
->hash_lock
);
860 /* Delete entry (via netlink) */
861 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
862 struct net_device
*dev
,
863 const unsigned char *addr
, u16 vid
)
865 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
867 struct vxlan_rdst
*rd
= NULL
;
873 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
879 spin_lock_bh(&vxlan
->hash_lock
);
880 f
= vxlan_find_mac(vxlan
, addr
);
884 if (!vxlan_addr_any(&ip
)) {
885 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
895 if (rd
&& !list_is_singular(&f
->remotes
)) {
896 list_del_rcu(&rd
->list
);
897 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
902 vxlan_fdb_destroy(vxlan
, f
);
905 spin_unlock_bh(&vxlan
->hash_lock
);
910 /* Dump forwarding table */
911 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
912 struct net_device
*dev
,
913 struct net_device
*filter_dev
, int idx
)
915 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
918 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
922 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
923 struct vxlan_rdst
*rd
;
925 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
926 if (idx
< cb
->args
[0])
929 err
= vxlan_fdb_info(skb
, vxlan
, f
,
930 NETLINK_CB(cb
->skb
).portid
,
947 /* Watch incoming packets to learn mapping between Ethernet address
948 * and Tunnel endpoint.
949 * Return true if packet is bogus and should be dropped.
951 static bool vxlan_snoop(struct net_device
*dev
,
952 union vxlan_addr
*src_ip
, const u8
*src_mac
)
954 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
957 f
= vxlan_find_mac(vxlan
, src_mac
);
959 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
961 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
964 /* Don't migrate static entries, drop packets */
965 if (f
->state
& NUD_NOARP
)
970 "%pM migrated from %pIS to %pIS\n",
971 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
973 rdst
->remote_ip
= *src_ip
;
974 f
->updated
= jiffies
;
975 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
977 /* learned new entry */
978 spin_lock(&vxlan
->hash_lock
);
980 /* close off race between vxlan_flush and incoming packets */
981 if (netif_running(dev
))
982 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
984 NLM_F_EXCL
|NLM_F_CREATE
,
986 vxlan
->default_dst
.remote_vni
,
988 spin_unlock(&vxlan
->hash_lock
);
994 /* See if multicast group is already in use by other ID */
995 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
997 struct vxlan_dev
*vxlan
;
998 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1000 /* The vxlan_sock is only used by dev, leaving group has
1001 * no effect on other vxlan devices.
1003 if (family
== AF_INET
&& dev
->vn4_sock
&&
1004 atomic_read(&dev
->vn4_sock
->refcnt
) == 1)
1006 #if IS_ENABLED(CONFIG_IPV6)
1007 if (family
== AF_INET6
&& dev
->vn6_sock
&&
1008 atomic_read(&dev
->vn6_sock
->refcnt
) == 1)
1012 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1013 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1016 if (family
== AF_INET
&& vxlan
->vn4_sock
!= dev
->vn4_sock
)
1018 #if IS_ENABLED(CONFIG_IPV6)
1019 if (family
== AF_INET6
&& vxlan
->vn6_sock
!= dev
->vn6_sock
)
1023 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1024 &dev
->default_dst
.remote_ip
))
1027 if (vxlan
->default_dst
.remote_ifindex
!=
1028 dev
->default_dst
.remote_ifindex
)
1037 static void __vxlan_sock_release(struct vxlan_sock
*vs
)
1039 struct vxlan_net
*vn
;
1043 if (!atomic_dec_and_test(&vs
->refcnt
))
1046 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1047 spin_lock(&vn
->sock_lock
);
1048 hlist_del_rcu(&vs
->hlist
);
1049 vxlan_notify_del_rx_port(vs
);
1050 spin_unlock(&vn
->sock_lock
);
1052 queue_work(vxlan_wq
, &vs
->del_work
);
1055 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1057 __vxlan_sock_release(vxlan
->vn4_sock
);
1058 #if IS_ENABLED(CONFIG_IPV6)
1059 __vxlan_sock_release(vxlan
->vn6_sock
);
1063 /* Update multicast group membership when first VNI on
1064 * multicast address is brought up
1066 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1069 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1070 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1073 if (ip
->sa
.sa_family
== AF_INET
) {
1074 struct ip_mreqn mreq
= {
1075 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1076 .imr_ifindex
= ifindex
,
1079 sk
= vxlan
->vn4_sock
->sock
->sk
;
1081 ret
= ip_mc_join_group(sk
, &mreq
);
1083 #if IS_ENABLED(CONFIG_IPV6)
1085 sk
= vxlan
->vn6_sock
->sock
->sk
;
1087 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1088 &ip
->sin6
.sin6_addr
);
1096 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1097 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1100 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1101 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1104 if (ip
->sa
.sa_family
== AF_INET
) {
1105 struct ip_mreqn mreq
= {
1106 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1107 .imr_ifindex
= ifindex
,
1110 sk
= vxlan
->vn4_sock
->sock
->sk
;
1112 ret
= ip_mc_leave_group(sk
, &mreq
);
1114 #if IS_ENABLED(CONFIG_IPV6)
1116 sk
= vxlan
->vn6_sock
->sock
->sk
;
1118 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1119 &ip
->sin6
.sin6_addr
);
1127 static struct vxlanhdr
*vxlan_remcsum(struct sk_buff
*skb
, struct vxlanhdr
*vh
,
1128 size_t hdrlen
, u32 data
, bool nopartial
)
1130 size_t start
, offset
, plen
;
1132 if (skb
->remcsum_offload
)
1135 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
1136 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
1137 offsetof(struct udphdr
, check
) :
1138 offsetof(struct tcphdr
, check
));
1140 plen
= hdrlen
+ offset
+ sizeof(u16
);
1142 if (!pskb_may_pull(skb
, plen
))
1145 vh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1147 skb_remcsum_process(skb
, (void *)vh
+ hdrlen
, start
, offset
,
1153 static void vxlan_rcv(struct vxlan_sock
*vs
, struct sk_buff
*skb
,
1154 struct vxlan_metadata
*md
, u32 vni
,
1155 struct metadata_dst
*tun_dst
)
1157 struct iphdr
*oip
= NULL
;
1158 struct ipv6hdr
*oip6
= NULL
;
1159 struct vxlan_dev
*vxlan
;
1160 struct pcpu_sw_netstats
*stats
;
1161 union vxlan_addr saddr
;
1164 /* For flow based devices, map all packets to VNI 0 */
1165 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
1168 /* Is this VNI defined? */
1169 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1173 skb_reset_mac_header(skb
);
1174 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1175 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1176 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1178 /* Ignore packet loops (and multicast echo) */
1179 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1182 /* Get data from the outer IP header */
1183 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1185 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1186 saddr
.sa
.sa_family
= AF_INET
;
1187 #if IS_ENABLED(CONFIG_IPV6)
1189 oip6
= ipv6_hdr(skb
);
1190 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1191 saddr
.sa
.sa_family
= AF_INET6
;
1196 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1200 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1201 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1204 skb_reset_network_header(skb
);
1205 /* In flow-based mode, GBP is carried in dst_metadata */
1206 if (!(vs
->flags
& VXLAN_F_COLLECT_METADATA
))
1207 skb
->mark
= md
->gbp
;
1210 err
= IP6_ECN_decapsulate(oip6
, skb
);
1212 err
= IP_ECN_decapsulate(oip
, skb
);
1214 if (unlikely(err
)) {
1215 if (log_ecn_error
) {
1217 net_info_ratelimited("non-ECT from %pI6\n",
1220 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1221 &oip
->saddr
, oip
->tos
);
1224 ++vxlan
->dev
->stats
.rx_frame_errors
;
1225 ++vxlan
->dev
->stats
.rx_errors
;
1230 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1231 u64_stats_update_begin(&stats
->syncp
);
1232 stats
->rx_packets
++;
1233 stats
->rx_bytes
+= skb
->len
;
1234 u64_stats_update_end(&stats
->syncp
);
1236 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1241 dst_release((struct dst_entry
*)tun_dst
);
1243 /* Consume bad packet */
1247 /* Callback from net/ipv4/udp.c to receive packets */
1248 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1250 struct metadata_dst
*tun_dst
= NULL
;
1251 struct vxlan_sock
*vs
;
1252 struct vxlanhdr
*vxh
;
1254 struct vxlan_metadata _md
;
1255 struct vxlan_metadata
*md
= &_md
;
1257 /* Need Vxlan and inner Ethernet header to be present */
1258 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1261 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1262 flags
= ntohl(vxh
->vx_flags
);
1263 vni
= ntohl(vxh
->vx_vni
);
1265 if (flags
& VXLAN_HF_VNI
) {
1266 flags
&= ~VXLAN_HF_VNI
;
1268 /* VNI flag always required to be set */
1272 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1274 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1276 vs
= rcu_dereference_sk_user_data(sk
);
1280 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
1281 vxh
= vxlan_remcsum(skb
, vxh
, sizeof(struct vxlanhdr
), vni
,
1282 !!(vs
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1286 flags
&= ~VXLAN_HF_RCO
;
1287 vni
&= VXLAN_VNI_MASK
;
1290 if (vxlan_collect_metadata(vs
)) {
1291 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1292 cpu_to_be64(vni
>> 8), sizeof(*md
));
1297 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1299 memset(md
, 0, sizeof(*md
));
1302 /* For backwards compatibility, only allow reserved fields to be
1303 * used by VXLAN extensions if explicitly requested.
1305 if ((flags
& VXLAN_HF_GBP
) && (vs
->flags
& VXLAN_F_GBP
)) {
1306 struct vxlanhdr_gbp
*gbp
;
1308 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1309 md
->gbp
= ntohs(gbp
->policy_id
);
1312 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1313 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1316 if (gbp
->dont_learn
)
1317 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1319 if (gbp
->policy_applied
)
1320 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1322 flags
&= ~VXLAN_GBP_USED_BITS
;
1325 if (flags
|| vni
& ~VXLAN_VNI_MASK
) {
1326 /* If there are any unprocessed flags remaining treat
1327 * this as a malformed packet. This behavior diverges from
1328 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1329 * in reserved fields are to be ignored. The approach here
1330 * maintains compatibility with previous stack code, and also
1331 * is more robust and provides a little more security in
1332 * adding extensions to VXLAN.
1338 vxlan_rcv(vs
, skb
, md
, vni
>> 8, tun_dst
);
1342 /* Consume bad packet */
1347 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1348 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1352 dst_release((struct dst_entry
*)tun_dst
);
1354 /* Return non vxlan pkt */
1358 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1360 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1361 struct arphdr
*parp
;
1364 struct neighbour
*n
;
1366 if (dev
->flags
& IFF_NOARP
)
1369 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1370 dev
->stats
.tx_dropped
++;
1373 parp
= arp_hdr(skb
);
1375 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1376 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1377 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1378 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1379 parp
->ar_hln
!= dev
->addr_len
||
1382 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1384 arpptr
+= dev
->addr_len
; /* sha */
1385 memcpy(&sip
, arpptr
, sizeof(sip
));
1386 arpptr
+= sizeof(sip
);
1387 arpptr
+= dev
->addr_len
; /* tha */
1388 memcpy(&tip
, arpptr
, sizeof(tip
));
1390 if (ipv4_is_loopback(tip
) ||
1391 ipv4_is_multicast(tip
))
1394 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1397 struct vxlan_fdb
*f
;
1398 struct sk_buff
*reply
;
1400 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1405 f
= vxlan_find_mac(vxlan
, n
->ha
);
1406 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1407 /* bridge-local neighbor */
1412 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1420 skb_reset_mac_header(reply
);
1421 __skb_pull(reply
, skb_network_offset(reply
));
1422 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1423 reply
->pkt_type
= PACKET_HOST
;
1425 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1426 dev
->stats
.rx_dropped
++;
1427 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1428 union vxlan_addr ipa
= {
1429 .sin
.sin_addr
.s_addr
= tip
,
1430 .sin
.sin_family
= AF_INET
,
1433 vxlan_ip_miss(dev
, &ipa
);
1437 return NETDEV_TX_OK
;
1440 #if IS_ENABLED(CONFIG_IPV6)
1441 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1442 struct neighbour
*n
, bool isrouter
)
1444 struct net_device
*dev
= request
->dev
;
1445 struct sk_buff
*reply
;
1446 struct nd_msg
*ns
, *na
;
1447 struct ipv6hdr
*pip6
;
1449 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1456 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1457 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1458 reply
= alloc_skb(len
, GFP_ATOMIC
);
1462 reply
->protocol
= htons(ETH_P_IPV6
);
1464 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1465 skb_push(reply
, sizeof(struct ethhdr
));
1466 skb_set_mac_header(reply
, 0);
1468 ns
= (struct nd_msg
*)skb_transport_header(request
);
1470 daddr
= eth_hdr(request
)->h_source
;
1471 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1472 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1473 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1474 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1479 /* Ethernet header */
1480 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1481 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1482 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1483 reply
->protocol
= htons(ETH_P_IPV6
);
1485 skb_pull(reply
, sizeof(struct ethhdr
));
1486 skb_set_network_header(reply
, 0);
1487 skb_put(reply
, sizeof(struct ipv6hdr
));
1491 pip6
= ipv6_hdr(reply
);
1492 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1494 pip6
->priority
= ipv6_hdr(request
)->priority
;
1495 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1496 pip6
->hop_limit
= 255;
1497 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1498 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1500 skb_pull(reply
, sizeof(struct ipv6hdr
));
1501 skb_set_transport_header(reply
, 0);
1503 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1505 /* Neighbor Advertisement */
1506 memset(na
, 0, sizeof(*na
)+na_olen
);
1507 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1508 na
->icmph
.icmp6_router
= isrouter
;
1509 na
->icmph
.icmp6_override
= 1;
1510 na
->icmph
.icmp6_solicited
= 1;
1511 na
->target
= ns
->target
;
1512 ether_addr_copy(&na
->opt
[2], n
->ha
);
1513 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1514 na
->opt
[1] = na_olen
>> 3;
1516 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1517 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1518 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1520 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1522 skb_push(reply
, sizeof(struct ipv6hdr
));
1524 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1529 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1531 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1533 const struct ipv6hdr
*iphdr
;
1534 const struct in6_addr
*saddr
, *daddr
;
1535 struct neighbour
*n
;
1536 struct inet6_dev
*in6_dev
;
1538 in6_dev
= __in6_dev_get(dev
);
1542 iphdr
= ipv6_hdr(skb
);
1543 saddr
= &iphdr
->saddr
;
1544 daddr
= &iphdr
->daddr
;
1546 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1547 if (msg
->icmph
.icmp6_code
!= 0 ||
1548 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1551 if (ipv6_addr_loopback(daddr
) ||
1552 ipv6_addr_is_multicast(&msg
->target
))
1555 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1558 struct vxlan_fdb
*f
;
1559 struct sk_buff
*reply
;
1561 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1566 f
= vxlan_find_mac(vxlan
, n
->ha
);
1567 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1568 /* bridge-local neighbor */
1573 reply
= vxlan_na_create(skb
, n
,
1574 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1581 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1582 dev
->stats
.rx_dropped
++;
1584 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1585 union vxlan_addr ipa
= {
1586 .sin6
.sin6_addr
= msg
->target
,
1587 .sin6
.sin6_family
= AF_INET6
,
1590 vxlan_ip_miss(dev
, &ipa
);
1595 return NETDEV_TX_OK
;
1599 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1601 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1602 struct neighbour
*n
;
1604 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1608 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1613 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1616 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1617 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1618 union vxlan_addr ipa
= {
1619 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1620 .sin
.sin_family
= AF_INET
,
1623 vxlan_ip_miss(dev
, &ipa
);
1629 #if IS_ENABLED(CONFIG_IPV6)
1632 struct ipv6hdr
*pip6
;
1634 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1636 pip6
= ipv6_hdr(skb
);
1637 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1638 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1639 union vxlan_addr ipa
= {
1640 .sin6
.sin6_addr
= pip6
->daddr
,
1641 .sin6
.sin6_family
= AF_INET6
,
1644 vxlan_ip_miss(dev
, &ipa
);
1658 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1660 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1662 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1671 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1672 struct vxlan_metadata
*md
)
1674 struct vxlanhdr_gbp
*gbp
;
1679 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1680 vxh
->vx_flags
|= htonl(VXLAN_HF_GBP
);
1682 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1683 gbp
->dont_learn
= 1;
1685 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1686 gbp
->policy_applied
= 1;
1688 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1691 #if IS_ENABLED(CONFIG_IPV6)
1692 static int vxlan6_xmit_skb(struct dst_entry
*dst
, struct sock
*sk
,
1693 struct sk_buff
*skb
,
1694 struct net_device
*dev
, struct in6_addr
*saddr
,
1695 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1696 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1697 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1699 struct vxlanhdr
*vxh
;
1702 bool udp_sum
= !(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
1703 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1704 u16 hdrlen
= sizeof(struct vxlanhdr
);
1706 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1707 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1708 int csum_start
= skb_checksum_start_offset(skb
);
1710 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1711 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1712 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1713 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1715 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1719 skb_scrub_packet(skb
, xnet
);
1721 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1722 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1723 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1725 /* Need space for new headers (invalidates iph ptr) */
1726 err
= skb_cow_head(skb
, min_headroom
);
1727 if (unlikely(err
)) {
1732 skb
= vlan_hwaccel_push_inside(skb
);
1733 if (WARN_ON(!skb
)) {
1738 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1744 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1745 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1748 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1749 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1752 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1753 data
|= VXLAN_RCO_UDP
;
1755 vxh
->vx_vni
|= htonl(data
);
1756 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1758 if (!skb_is_gso(skb
)) {
1759 skb
->ip_summed
= CHECKSUM_NONE
;
1760 skb
->encapsulation
= 0;
1764 if (vxflags
& VXLAN_F_GBP
)
1765 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1767 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1769 udp_tunnel6_xmit_skb(dst
, sk
, skb
, dev
, saddr
, daddr
, prio
,
1770 ttl
, src_port
, dst_port
,
1771 !!(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
));
1779 static int vxlan_xmit_skb(struct rtable
*rt
, struct sock
*sk
, struct sk_buff
*skb
,
1780 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1781 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1782 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1784 struct vxlanhdr
*vxh
;
1787 bool udp_sum
= !!(vxflags
& VXLAN_F_UDP_CSUM
);
1788 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1789 u16 hdrlen
= sizeof(struct vxlanhdr
);
1791 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1792 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1793 int csum_start
= skb_checksum_start_offset(skb
);
1795 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1796 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1797 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1798 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1800 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1804 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1805 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1806 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1808 /* Need space for new headers (invalidates iph ptr) */
1809 err
= skb_cow_head(skb
, min_headroom
);
1810 if (unlikely(err
)) {
1815 skb
= vlan_hwaccel_push_inside(skb
);
1819 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1821 return PTR_ERR(skb
);
1823 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1824 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1827 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1828 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1831 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1832 data
|= VXLAN_RCO_UDP
;
1834 vxh
->vx_vni
|= htonl(data
);
1835 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1837 if (!skb_is_gso(skb
)) {
1838 skb
->ip_summed
= CHECKSUM_NONE
;
1839 skb
->encapsulation
= 0;
1843 if (vxflags
& VXLAN_F_GBP
)
1844 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1846 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1848 udp_tunnel_xmit_skb(rt
, sk
, skb
, src
, dst
, tos
, ttl
, df
,
1849 src_port
, dst_port
, xnet
,
1850 !(vxflags
& VXLAN_F_UDP_CSUM
));
1854 #if IS_ENABLED(CONFIG_IPV6)
1855 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1856 struct sk_buff
*skb
, int oif
,
1857 const struct in6_addr
*daddr
,
1858 struct in6_addr
*saddr
)
1860 struct dst_entry
*ndst
;
1864 memset(&fl6
, 0, sizeof(fl6
));
1865 fl6
.flowi6_oif
= oif
;
1867 fl6
.saddr
= vxlan
->cfg
.saddr
.sin6
.sin6_addr
;
1868 fl6
.flowi6_mark
= skb
->mark
;
1869 fl6
.flowi6_proto
= IPPROTO_UDP
;
1871 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1872 vxlan
->vn6_sock
->sock
->sk
,
1875 return ERR_PTR(err
);
1882 /* Bypass encapsulation if the destination is local */
1883 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1884 struct vxlan_dev
*dst_vxlan
)
1886 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1887 union vxlan_addr loopback
;
1888 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1889 struct net_device
*dev
= skb
->dev
;
1892 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1893 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1894 skb
->pkt_type
= PACKET_HOST
;
1895 skb
->encapsulation
= 0;
1896 skb
->dev
= dst_vxlan
->dev
;
1897 __skb_pull(skb
, skb_network_offset(skb
));
1899 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1900 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1901 loopback
.sa
.sa_family
= AF_INET
;
1902 #if IS_ENABLED(CONFIG_IPV6)
1904 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1905 loopback
.sa
.sa_family
= AF_INET6
;
1909 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1910 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1912 u64_stats_update_begin(&tx_stats
->syncp
);
1913 tx_stats
->tx_packets
++;
1914 tx_stats
->tx_bytes
+= len
;
1915 u64_stats_update_end(&tx_stats
->syncp
);
1917 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1918 u64_stats_update_begin(&rx_stats
->syncp
);
1919 rx_stats
->rx_packets
++;
1920 rx_stats
->rx_bytes
+= len
;
1921 u64_stats_update_end(&rx_stats
->syncp
);
1923 dev
->stats
.rx_dropped
++;
1927 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1928 struct vxlan_rdst
*rdst
, bool did_rsc
)
1930 struct ip_tunnel_info
*info
;
1931 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1933 struct rtable
*rt
= NULL
;
1934 const struct iphdr
*old_iph
;
1936 union vxlan_addr
*dst
;
1937 union vxlan_addr remote_ip
;
1938 struct vxlan_metadata _md
;
1939 struct vxlan_metadata
*md
= &_md
;
1940 __be16 src_port
= 0, dst_port
;
1945 u32 flags
= vxlan
->flags
;
1947 info
= skb_tunnel_info(skb
);
1950 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
1951 vni
= rdst
->remote_vni
;
1952 dst
= &rdst
->remote_ip
;
1955 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1959 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
1960 vni
= be64_to_cpu(info
->key
.tun_id
);
1961 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
1962 if (remote_ip
.sa
.sa_family
== AF_INET
)
1963 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
1965 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
1969 if (vxlan_addr_any(dst
)) {
1971 /* short-circuited back to local bridge */
1972 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1978 old_iph
= ip_hdr(skb
);
1980 ttl
= vxlan
->cfg
.ttl
;
1981 if (!ttl
&& vxlan_addr_multicast(dst
))
1984 tos
= vxlan
->cfg
.tos
;
1986 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1988 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
1989 vxlan
->cfg
.port_max
, true);
1992 ttl
= info
->key
.ttl
;
1993 tos
= info
->key
.tos
;
1995 if (info
->options_len
)
1996 md
= ip_tunnel_info_opts(info
);
1998 md
->gbp
= skb
->mark
;
2001 if (dst
->sa
.sa_family
== AF_INET
) {
2002 if (!vxlan
->vn4_sock
)
2004 sk
= vxlan
->vn4_sock
->sock
->sk
;
2007 if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
)
2010 if (info
->key
.tun_flags
& TUNNEL_CSUM
)
2011 flags
|= VXLAN_F_UDP_CSUM
;
2013 flags
&= ~VXLAN_F_UDP_CSUM
;
2016 memset(&fl4
, 0, sizeof(fl4
));
2017 fl4
.flowi4_oif
= rdst
? rdst
->remote_ifindex
: 0;
2018 fl4
.flowi4_tos
= RT_TOS(tos
);
2019 fl4
.flowi4_mark
= skb
->mark
;
2020 fl4
.flowi4_proto
= IPPROTO_UDP
;
2021 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
2022 fl4
.saddr
= vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
;
2024 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
2026 netdev_dbg(dev
, "no route to %pI4\n",
2027 &dst
->sin
.sin_addr
.s_addr
);
2028 dev
->stats
.tx_carrier_errors
++;
2032 if (rt
->dst
.dev
== dev
) {
2033 netdev_dbg(dev
, "circular route to %pI4\n",
2034 &dst
->sin
.sin_addr
.s_addr
);
2035 dev
->stats
.collisions
++;
2039 /* Bypass encapsulation if the destination is local */
2040 if (rt
->rt_flags
& RTCF_LOCAL
&&
2041 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2042 struct vxlan_dev
*dst_vxlan
;
2045 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2046 dst
->sa
.sa_family
, dst_port
,
2050 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2054 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2055 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2056 err
= vxlan_xmit_skb(rt
, sk
, skb
, fl4
.saddr
,
2057 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2058 src_port
, dst_port
, htonl(vni
<< 8), md
,
2059 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2062 /* skb is already freed. */
2066 #if IS_ENABLED(CONFIG_IPV6)
2068 struct dst_entry
*ndst
;
2069 struct in6_addr saddr
;
2072 if (!vxlan
->vn6_sock
)
2074 sk
= vxlan
->vn6_sock
->sock
->sk
;
2076 ndst
= vxlan6_get_route(vxlan
, skb
,
2077 rdst
? rdst
->remote_ifindex
: 0,
2078 &dst
->sin6
.sin6_addr
, &saddr
);
2080 netdev_dbg(dev
, "no route to %pI6\n",
2081 &dst
->sin6
.sin6_addr
);
2082 dev
->stats
.tx_carrier_errors
++;
2086 if (ndst
->dev
== dev
) {
2087 netdev_dbg(dev
, "circular route to %pI6\n",
2088 &dst
->sin6
.sin6_addr
);
2090 dev
->stats
.collisions
++;
2094 /* Bypass encapsulation if the destination is local */
2095 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2096 if (rt6i_flags
& RTF_LOCAL
&&
2097 !(rt6i_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2098 struct vxlan_dev
*dst_vxlan
;
2101 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2102 dst
->sa
.sa_family
, dst_port
,
2106 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2111 if (info
->key
.tun_flags
& TUNNEL_CSUM
)
2112 flags
&= ~VXLAN_F_UDP_ZERO_CSUM6_TX
;
2114 flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2117 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2118 err
= vxlan6_xmit_skb(ndst
, sk
, skb
, dev
, &saddr
, &dst
->sin6
.sin6_addr
,
2119 0, ttl
, src_port
, dst_port
, htonl(vni
<< 8), md
,
2120 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2128 dev
->stats
.tx_dropped
++;
2134 dev
->stats
.tx_errors
++;
2139 /* Transmit local packets over Vxlan
2141 * Outer IP header inherits ECN and DF from inner header.
2142 * Outer UDP destination is the VXLAN assigned port.
2143 * source port is based on hash of flow
2145 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2147 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2148 const struct ip_tunnel_info
*info
;
2150 bool did_rsc
= false;
2151 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2152 struct vxlan_fdb
*f
;
2154 info
= skb_tunnel_info(skb
);
2156 skb_reset_mac_header(skb
);
2159 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2160 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2161 return arp_reduce(dev
, skb
);
2162 #if IS_ENABLED(CONFIG_IPV6)
2163 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2164 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2165 + sizeof(struct nd_msg
)) &&
2166 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2169 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2170 if (msg
->icmph
.icmp6_code
== 0 &&
2171 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2172 return neigh_reduce(dev
, skb
);
2178 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
2179 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2180 vxlan_xmit_one(skb
, dev
, NULL
, false);
2183 return NETDEV_TX_OK
;
2186 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2189 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2190 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2191 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2192 did_rsc
= route_shortcircuit(dev
, skb
);
2194 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2198 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2200 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2201 !is_multicast_ether_addr(eth
->h_dest
))
2202 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2204 dev
->stats
.tx_dropped
++;
2206 return NETDEV_TX_OK
;
2210 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2211 struct sk_buff
*skb1
;
2217 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2219 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2223 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2226 return NETDEV_TX_OK
;
2229 /* Walk the forwarding table and purge stale entries */
2230 static void vxlan_cleanup(unsigned long arg
)
2232 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2233 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2236 if (!netif_running(vxlan
->dev
))
2239 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2240 struct hlist_node
*p
, *n
;
2242 spin_lock_bh(&vxlan
->hash_lock
);
2243 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2245 = container_of(p
, struct vxlan_fdb
, hlist
);
2246 unsigned long timeout
;
2248 if (f
->state
& NUD_PERMANENT
)
2251 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2252 if (time_before_eq(timeout
, jiffies
)) {
2253 netdev_dbg(vxlan
->dev
,
2254 "garbage collect %pM\n",
2256 f
->state
= NUD_STALE
;
2257 vxlan_fdb_destroy(vxlan
, f
);
2258 } else if (time_before(timeout
, next_timer
))
2259 next_timer
= timeout
;
2261 spin_unlock_bh(&vxlan
->hash_lock
);
2264 mod_timer(&vxlan
->age_timer
, next_timer
);
2267 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2269 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2270 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2272 spin_lock(&vn
->sock_lock
);
2273 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2274 spin_unlock(&vn
->sock_lock
);
2277 /* Setup stats when device is created */
2278 static int vxlan_init(struct net_device
*dev
)
2280 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2287 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2289 struct vxlan_fdb
*f
;
2291 spin_lock_bh(&vxlan
->hash_lock
);
2292 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2294 vxlan_fdb_destroy(vxlan
, f
);
2295 spin_unlock_bh(&vxlan
->hash_lock
);
2298 static void vxlan_uninit(struct net_device
*dev
)
2300 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2302 vxlan_fdb_delete_default(vxlan
);
2304 free_percpu(dev
->tstats
);
2307 /* Start ageing timer and join group when device is brought up */
2308 static int vxlan_open(struct net_device
*dev
)
2310 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2313 ret
= vxlan_sock_add(vxlan
);
2317 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2318 ret
= vxlan_igmp_join(vxlan
);
2319 if (ret
== -EADDRINUSE
)
2322 vxlan_sock_release(vxlan
);
2327 if (vxlan
->cfg
.age_interval
)
2328 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2333 /* Purge the forwarding table */
2334 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2338 spin_lock_bh(&vxlan
->hash_lock
);
2339 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2340 struct hlist_node
*p
, *n
;
2341 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2343 = container_of(p
, struct vxlan_fdb
, hlist
);
2344 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2345 if (!is_zero_ether_addr(f
->eth_addr
))
2346 vxlan_fdb_destroy(vxlan
, f
);
2349 spin_unlock_bh(&vxlan
->hash_lock
);
2352 /* Cleanup timer and forwarding table on shutdown */
2353 static int vxlan_stop(struct net_device
*dev
)
2355 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2356 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2359 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2360 !vxlan_group_used(vn
, vxlan
))
2361 ret
= vxlan_igmp_leave(vxlan
);
2363 del_timer_sync(&vxlan
->age_timer
);
2366 vxlan_sock_release(vxlan
);
2371 /* Stub, nothing needs to be done. */
2372 static void vxlan_set_multicast_list(struct net_device
*dev
)
2376 static int __vxlan_change_mtu(struct net_device
*dev
,
2377 struct net_device
*lowerdev
,
2378 struct vxlan_rdst
*dst
, int new_mtu
, bool strict
)
2380 int max_mtu
= IP_MAX_MTU
;
2383 max_mtu
= lowerdev
->mtu
;
2385 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2386 max_mtu
-= VXLAN6_HEADROOM
;
2388 max_mtu
-= VXLAN_HEADROOM
;
2393 if (new_mtu
> max_mtu
) {
2404 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2406 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2407 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2408 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2409 dst
->remote_ifindex
);
2410 return __vxlan_change_mtu(dev
, lowerdev
, dst
, new_mtu
, true);
2413 static int egress_ipv4_tun_info(struct net_device
*dev
, struct sk_buff
*skb
,
2414 struct ip_tunnel_info
*info
,
2415 __be16 sport
, __be16 dport
)
2417 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2421 memset(&fl4
, 0, sizeof(fl4
));
2422 fl4
.flowi4_tos
= RT_TOS(info
->key
.tos
);
2423 fl4
.flowi4_mark
= skb
->mark
;
2424 fl4
.flowi4_proto
= IPPROTO_UDP
;
2425 fl4
.daddr
= info
->key
.u
.ipv4
.dst
;
2427 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
2432 info
->key
.u
.ipv4
.src
= fl4
.saddr
;
2433 info
->key
.tp_src
= sport
;
2434 info
->key
.tp_dst
= dport
;
2438 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2440 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2441 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2442 __be16 sport
, dport
;
2444 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2445 vxlan
->cfg
.port_max
, true);
2446 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2448 if (ip_tunnel_info_af(info
) == AF_INET
) {
2449 if (!vxlan
->vn4_sock
)
2451 return egress_ipv4_tun_info(dev
, skb
, info
, sport
, dport
);
2453 #if IS_ENABLED(CONFIG_IPV6)
2454 struct dst_entry
*ndst
;
2456 if (!vxlan
->vn6_sock
)
2458 ndst
= vxlan6_get_route(vxlan
, skb
, 0,
2459 &info
->key
.u
.ipv6
.dst
,
2460 &info
->key
.u
.ipv6
.src
);
2462 return PTR_ERR(ndst
);
2465 info
->key
.tp_src
= sport
;
2466 info
->key
.tp_dst
= dport
;
2467 #else /* !CONFIG_IPV6 */
2468 return -EPFNOSUPPORT
;
2474 static const struct net_device_ops vxlan_netdev_ops
= {
2475 .ndo_init
= vxlan_init
,
2476 .ndo_uninit
= vxlan_uninit
,
2477 .ndo_open
= vxlan_open
,
2478 .ndo_stop
= vxlan_stop
,
2479 .ndo_start_xmit
= vxlan_xmit
,
2480 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2481 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2482 .ndo_change_mtu
= vxlan_change_mtu
,
2483 .ndo_validate_addr
= eth_validate_addr
,
2484 .ndo_set_mac_address
= eth_mac_addr
,
2485 .ndo_fdb_add
= vxlan_fdb_add
,
2486 .ndo_fdb_del
= vxlan_fdb_delete
,
2487 .ndo_fdb_dump
= vxlan_fdb_dump
,
2488 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2491 /* Info for udev, that this is a virtual tunnel endpoint */
2492 static struct device_type vxlan_type
= {
2496 /* Calls the ndo_add_vxlan_port of the caller in order to
2497 * supply the listening VXLAN udp ports. Callers are expected
2498 * to implement the ndo_add_vxlan_port.
2500 void vxlan_get_rx_port(struct net_device
*dev
)
2502 struct vxlan_sock
*vs
;
2503 struct net
*net
= dev_net(dev
);
2504 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2505 sa_family_t sa_family
;
2509 spin_lock(&vn
->sock_lock
);
2510 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2511 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2512 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2513 sa_family
= vxlan_get_sk_family(vs
);
2514 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2518 spin_unlock(&vn
->sock_lock
);
2520 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2522 /* Initialize the device structure. */
2523 static void vxlan_setup(struct net_device
*dev
)
2525 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2528 eth_hw_addr_random(dev
);
2531 dev
->netdev_ops
= &vxlan_netdev_ops
;
2532 dev
->destructor
= free_netdev
;
2533 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2535 dev
->features
|= NETIF_F_LLTX
;
2536 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2537 dev
->features
|= NETIF_F_RXCSUM
;
2538 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2540 dev
->vlan_features
= dev
->features
;
2541 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2542 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2543 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2544 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2545 netif_keep_dst(dev
);
2546 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2547 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
| IFF_NO_QUEUE
;
2549 INIT_LIST_HEAD(&vxlan
->next
);
2550 spin_lock_init(&vxlan
->hash_lock
);
2552 init_timer_deferrable(&vxlan
->age_timer
);
2553 vxlan
->age_timer
.function
= vxlan_cleanup
;
2554 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2556 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2560 gro_cells_init(&vxlan
->gro_cells
, dev
);
2562 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2563 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2566 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2567 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2568 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2569 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2570 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2571 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2572 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2573 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2574 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2575 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2576 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2577 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2578 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2579 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2580 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2581 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2582 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2583 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2584 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2585 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2586 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2587 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2588 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2589 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2590 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2591 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2594 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2596 if (tb
[IFLA_ADDRESS
]) {
2597 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2598 pr_debug("invalid link address (not ethernet)\n");
2602 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2603 pr_debug("invalid all zero ethernet address\n");
2604 return -EADDRNOTAVAIL
;
2611 if (data
[IFLA_VXLAN_ID
]) {
2612 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2613 if (id
>= VXLAN_VID_MASK
)
2617 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2618 const struct ifla_vxlan_port_range
*p
2619 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2621 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2622 pr_debug("port range %u .. %u not valid\n",
2623 ntohs(p
->low
), ntohs(p
->high
));
2631 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2632 struct ethtool_drvinfo
*drvinfo
)
2634 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2635 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2638 static const struct ethtool_ops vxlan_ethtool_ops
= {
2639 .get_drvinfo
= vxlan_get_drvinfo
,
2640 .get_link
= ethtool_op_get_link
,
2643 static void vxlan_del_work(struct work_struct
*work
)
2645 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2646 udp_tunnel_sock_release(vs
->sock
);
2650 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2651 __be16 port
, u32 flags
)
2653 struct socket
*sock
;
2654 struct udp_port_cfg udp_conf
;
2657 memset(&udp_conf
, 0, sizeof(udp_conf
));
2660 udp_conf
.family
= AF_INET6
;
2661 udp_conf
.use_udp6_rx_checksums
=
2662 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2663 udp_conf
.ipv6_v6only
= 1;
2665 udp_conf
.family
= AF_INET
;
2668 udp_conf
.local_udp_port
= port
;
2670 /* Open UDP socket */
2671 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2673 return ERR_PTR(err
);
2678 /* Create new listen socket if needed */
2679 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2680 __be16 port
, u32 flags
)
2682 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2683 struct vxlan_sock
*vs
;
2684 struct socket
*sock
;
2686 struct udp_tunnel_sock_cfg tunnel_cfg
;
2688 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2690 return ERR_PTR(-ENOMEM
);
2692 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2693 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2695 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2697 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2699 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port
),
2702 return ERR_CAST(sock
);
2706 atomic_set(&vs
->refcnt
, 1);
2707 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2709 /* Initialize the vxlan udp offloads structure */
2710 vs
->udp_offloads
.port
= port
;
2711 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2712 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2714 spin_lock(&vn
->sock_lock
);
2715 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2716 vxlan_notify_add_rx_port(vs
);
2717 spin_unlock(&vn
->sock_lock
);
2719 /* Mark socket as an encapsulation socket. */
2720 tunnel_cfg
.sk_user_data
= vs
;
2721 tunnel_cfg
.encap_type
= 1;
2722 tunnel_cfg
.encap_rcv
= vxlan_udp_encap_recv
;
2723 tunnel_cfg
.encap_destroy
= NULL
;
2725 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2730 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2732 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2733 struct vxlan_sock
*vs
= NULL
;
2735 if (!vxlan
->cfg
.no_share
) {
2736 spin_lock(&vn
->sock_lock
);
2737 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2738 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2739 if (vs
&& !atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2740 spin_unlock(&vn
->sock_lock
);
2743 spin_unlock(&vn
->sock_lock
);
2746 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2747 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2750 #if IS_ENABLED(CONFIG_IPV6)
2752 vxlan
->vn6_sock
= vs
;
2755 vxlan
->vn4_sock
= vs
;
2756 vxlan_vs_add_dev(vs
, vxlan
);
2760 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2762 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2763 bool metadata
= vxlan
->flags
& VXLAN_F_COLLECT_METADATA
;
2766 vxlan
->vn4_sock
= NULL
;
2767 #if IS_ENABLED(CONFIG_IPV6)
2768 vxlan
->vn6_sock
= NULL
;
2769 if (ipv6
|| metadata
)
2770 ret
= __vxlan_sock_add(vxlan
, true);
2772 if (!ret
&& (!ipv6
|| metadata
))
2773 ret
= __vxlan_sock_add(vxlan
, false);
2775 vxlan_sock_release(vxlan
);
2779 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2780 struct vxlan_config
*conf
)
2782 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2783 struct vxlan_dev
*vxlan
= netdev_priv(dev
), *tmp
;
2784 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2785 unsigned short needed_headroom
= ETH_HLEN
;
2787 bool use_ipv6
= false;
2788 __be16 default_port
= vxlan
->cfg
.dst_port
;
2789 struct net_device
*lowerdev
= NULL
;
2791 vxlan
->net
= src_net
;
2793 dst
->remote_vni
= conf
->vni
;
2795 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2797 /* Unless IPv6 is explicitly requested, assume IPv4 */
2798 if (!dst
->remote_ip
.sa
.sa_family
)
2799 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2801 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2802 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) {
2803 if (!IS_ENABLED(CONFIG_IPV6
))
2804 return -EPFNOSUPPORT
;
2806 vxlan
->flags
|= VXLAN_F_IPV6
;
2809 if (conf
->remote_ifindex
) {
2810 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2811 dst
->remote_ifindex
= conf
->remote_ifindex
;
2814 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2818 #if IS_ENABLED(CONFIG_IPV6)
2820 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2821 if (idev
&& idev
->cnf
.disable_ipv6
) {
2822 pr_info("IPv6 is disabled via sysctl\n");
2829 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2831 needed_headroom
= lowerdev
->hard_header_len
;
2835 err
= __vxlan_change_mtu(dev
, lowerdev
, dst
, conf
->mtu
, false);
2840 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
2841 needed_headroom
+= VXLAN6_HEADROOM
;
2843 needed_headroom
+= VXLAN_HEADROOM
;
2844 dev
->needed_headroom
= needed_headroom
;
2846 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2847 if (!vxlan
->cfg
.dst_port
)
2848 vxlan
->cfg
.dst_port
= default_port
;
2849 vxlan
->flags
|= conf
->flags
;
2851 if (!vxlan
->cfg
.age_interval
)
2852 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2854 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
2855 if (tmp
->cfg
.vni
== conf
->vni
&&
2856 (tmp
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
||
2857 tmp
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) == use_ipv6
&&
2858 tmp
->cfg
.dst_port
== vxlan
->cfg
.dst_port
&&
2859 (tmp
->flags
& VXLAN_F_RCV_FLAGS
) ==
2860 (vxlan
->flags
& VXLAN_F_RCV_FLAGS
))
2864 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2866 /* create an fdb entry for a valid default destination */
2867 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2868 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2869 &vxlan
->default_dst
.remote_ip
,
2870 NUD_REACHABLE
|NUD_PERMANENT
,
2871 NLM_F_EXCL
|NLM_F_CREATE
,
2872 vxlan
->cfg
.dst_port
,
2873 vxlan
->default_dst
.remote_vni
,
2874 vxlan
->default_dst
.remote_ifindex
,
2880 err
= register_netdevice(dev
);
2882 vxlan_fdb_delete_default(vxlan
);
2886 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2891 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
2892 u8 name_assign_type
, struct vxlan_config
*conf
)
2894 struct nlattr
*tb
[IFLA_MAX
+1];
2895 struct net_device
*dev
;
2898 memset(&tb
, 0, sizeof(tb
));
2900 dev
= rtnl_create_link(net
, name
, name_assign_type
,
2901 &vxlan_link_ops
, tb
);
2905 err
= vxlan_dev_configure(net
, dev
, conf
);
2908 return ERR_PTR(err
);
2913 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
2915 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
2916 struct nlattr
*tb
[], struct nlattr
*data
[])
2918 struct vxlan_config conf
;
2921 memset(&conf
, 0, sizeof(conf
));
2923 if (data
[IFLA_VXLAN_ID
])
2924 conf
.vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2926 if (data
[IFLA_VXLAN_GROUP
]) {
2927 conf
.remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
2928 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2929 if (!IS_ENABLED(CONFIG_IPV6
))
2930 return -EPFNOSUPPORT
;
2932 conf
.remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
2933 conf
.remote_ip
.sa
.sa_family
= AF_INET6
;
2936 if (data
[IFLA_VXLAN_LOCAL
]) {
2937 conf
.saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
2938 conf
.saddr
.sa
.sa_family
= AF_INET
;
2939 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2940 if (!IS_ENABLED(CONFIG_IPV6
))
2941 return -EPFNOSUPPORT
;
2943 /* TODO: respect scope id */
2944 conf
.saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
2945 conf
.saddr
.sa
.sa_family
= AF_INET6
;
2948 if (data
[IFLA_VXLAN_LINK
])
2949 conf
.remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
2951 if (data
[IFLA_VXLAN_TOS
])
2952 conf
.tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2954 if (data
[IFLA_VXLAN_TTL
])
2955 conf
.ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2957 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2958 conf
.flags
|= VXLAN_F_LEARN
;
2960 if (data
[IFLA_VXLAN_AGEING
])
2961 conf
.age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2963 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2964 conf
.flags
|= VXLAN_F_PROXY
;
2966 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2967 conf
.flags
|= VXLAN_F_RSC
;
2969 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2970 conf
.flags
|= VXLAN_F_L2MISS
;
2972 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2973 conf
.flags
|= VXLAN_F_L3MISS
;
2975 if (data
[IFLA_VXLAN_LIMIT
])
2976 conf
.addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2978 if (data
[IFLA_VXLAN_COLLECT_METADATA
] &&
2979 nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
2980 conf
.flags
|= VXLAN_F_COLLECT_METADATA
;
2982 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2983 const struct ifla_vxlan_port_range
*p
2984 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2985 conf
.port_min
= ntohs(p
->low
);
2986 conf
.port_max
= ntohs(p
->high
);
2989 if (data
[IFLA_VXLAN_PORT
])
2990 conf
.dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2992 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2993 conf
.flags
|= VXLAN_F_UDP_CSUM
;
2995 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2996 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2997 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2999 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
3000 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
3001 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
3003 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
3004 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
3005 conf
.flags
|= VXLAN_F_REMCSUM_TX
;
3007 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
3008 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
3009 conf
.flags
|= VXLAN_F_REMCSUM_RX
;
3011 if (data
[IFLA_VXLAN_GBP
])
3012 conf
.flags
|= VXLAN_F_GBP
;
3014 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
])
3015 conf
.flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
3017 err
= vxlan_dev_configure(src_net
, dev
, &conf
);
3020 pr_info("ifindex %d does not exist\n", conf
.remote_ifindex
);
3024 pr_info("IPv6 is disabled via sysctl\n");
3028 pr_info("duplicate VNI %u\n", conf
.vni
);
3035 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
3037 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3038 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
3040 spin_lock(&vn
->sock_lock
);
3041 if (!hlist_unhashed(&vxlan
->hlist
))
3042 hlist_del_rcu(&vxlan
->hlist
);
3043 spin_unlock(&vn
->sock_lock
);
3045 gro_cells_destroy(&vxlan
->gro_cells
);
3046 list_del(&vxlan
->next
);
3047 unregister_netdevice_queue(dev
, head
);
3050 static size_t vxlan_get_size(const struct net_device
*dev
)
3053 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
3054 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
3055 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
3056 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
3057 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3058 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3059 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3060 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3061 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3062 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3063 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3064 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3065 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3066 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3067 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3068 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3069 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3070 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3071 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3072 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3073 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3077 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3079 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3080 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3081 struct ifla_vxlan_port_range ports
= {
3082 .low
= htons(vxlan
->cfg
.port_min
),
3083 .high
= htons(vxlan
->cfg
.port_max
),
3086 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
3087 goto nla_put_failure
;
3089 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3090 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3091 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3092 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3093 goto nla_put_failure
;
3094 #if IS_ENABLED(CONFIG_IPV6)
3096 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3097 &dst
->remote_ip
.sin6
.sin6_addr
))
3098 goto nla_put_failure
;
3103 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3104 goto nla_put_failure
;
3106 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3107 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3108 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3109 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3110 goto nla_put_failure
;
3111 #if IS_ENABLED(CONFIG_IPV6)
3113 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3114 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3115 goto nla_put_failure
;
3120 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3121 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3122 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3123 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
3124 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3125 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
3126 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
3127 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3128 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
3129 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3130 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
3131 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3132 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
3133 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3134 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3135 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3136 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3137 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
3138 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3139 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3140 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3141 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3142 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3143 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
3144 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3145 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
3146 goto nla_put_failure
;
3148 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3149 goto nla_put_failure
;
3151 if (vxlan
->flags
& VXLAN_F_GBP
&&
3152 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3153 goto nla_put_failure
;
3155 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3156 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3157 goto nla_put_failure
;
3165 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3167 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3172 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3174 .maxtype
= IFLA_VXLAN_MAX
,
3175 .policy
= vxlan_policy
,
3176 .priv_size
= sizeof(struct vxlan_dev
),
3177 .setup
= vxlan_setup
,
3178 .validate
= vxlan_validate
,
3179 .newlink
= vxlan_newlink
,
3180 .dellink
= vxlan_dellink
,
3181 .get_size
= vxlan_get_size
,
3182 .fill_info
= vxlan_fill_info
,
3183 .get_link_net
= vxlan_get_link_net
,
3186 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3187 struct net_device
*dev
)
3189 struct vxlan_dev
*vxlan
, *next
;
3190 LIST_HEAD(list_kill
);
3192 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3193 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3195 /* In case we created vxlan device with carrier
3196 * and we loose the carrier due to module unload
3197 * we also need to remove vxlan device. In other
3198 * cases, it's not necessary and remote_ifindex
3199 * is 0 here, so no matches.
3201 if (dst
->remote_ifindex
== dev
->ifindex
)
3202 vxlan_dellink(vxlan
->dev
, &list_kill
);
3205 unregister_netdevice_many(&list_kill
);
3208 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
3209 unsigned long event
, void *ptr
)
3211 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3212 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3214 if (event
== NETDEV_UNREGISTER
)
3215 vxlan_handle_lowerdev_unregister(vn
, dev
);
3220 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3221 .notifier_call
= vxlan_lowerdev_event
,
3224 static __net_init
int vxlan_init_net(struct net
*net
)
3226 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3229 INIT_LIST_HEAD(&vn
->vxlan_list
);
3230 spin_lock_init(&vn
->sock_lock
);
3232 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3233 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3238 static void __net_exit
vxlan_exit_net(struct net
*net
)
3240 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3241 struct vxlan_dev
*vxlan
, *next
;
3242 struct net_device
*dev
, *aux
;
3246 for_each_netdev_safe(net
, dev
, aux
)
3247 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3248 unregister_netdevice_queue(dev
, &list
);
3250 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3251 /* If vxlan->dev is in the same netns, it has already been added
3252 * to the list by the previous loop.
3254 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3255 gro_cells_destroy(&vxlan
->gro_cells
);
3256 unregister_netdevice_queue(vxlan
->dev
, &list
);
3260 unregister_netdevice_many(&list
);
3264 static struct pernet_operations vxlan_net_ops
= {
3265 .init
= vxlan_init_net
,
3266 .exit
= vxlan_exit_net
,
3267 .id
= &vxlan_net_id
,
3268 .size
= sizeof(struct vxlan_net
),
3271 static int __init
vxlan_init_module(void)
3275 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3279 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3281 rc
= register_pernet_subsys(&vxlan_net_ops
);
3285 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3289 rc
= rtnl_link_register(&vxlan_link_ops
);
3295 unregister_netdevice_notifier(&vxlan_notifier_block
);
3297 unregister_pernet_subsys(&vxlan_net_ops
);
3299 destroy_workqueue(vxlan_wq
);
3302 late_initcall(vxlan_init_module
);
3304 static void __exit
vxlan_cleanup_module(void)
3306 rtnl_link_unregister(&vxlan_link_ops
);
3307 unregister_netdevice_notifier(&vxlan_notifier_block
);
3308 destroy_workqueue(vxlan_wq
);
3309 unregister_pernet_subsys(&vxlan_net_ops
);
3310 /* rcu_barrier() is called by netns */
3312 module_exit(vxlan_cleanup_module
);
3314 MODULE_LICENSE("GPL");
3315 MODULE_VERSION(VXLAN_VERSION
);
3316 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3317 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3318 MODULE_ALIAS_RTNL_LINK("vxlan");