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
];
78 static struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
79 bool no_share
, u32 flags
);
81 /* per-network namespace private data for this module */
83 struct list_head vxlan_list
;
84 struct hlist_head sock_list
[PORT_HASH_SIZE
];
88 /* Forwarding table entry */
90 struct hlist_node hlist
; /* linked list of entries */
92 unsigned long updated
; /* jiffies */
94 struct list_head remotes
;
95 u8 eth_addr
[ETH_ALEN
];
96 u16 state
; /* see ndm_state */
97 u8 flags
; /* see ndm_flags */
100 /* salt for hash table */
101 static u32 vxlan_salt __read_mostly
;
102 static struct workqueue_struct
*vxlan_wq
;
104 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
106 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
107 ip_tunnel_collect_metadata();
110 #if IS_ENABLED(CONFIG_IPV6)
112 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
114 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
116 if (a
->sa
.sa_family
== AF_INET6
)
117 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
119 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
122 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
124 if (ipa
->sa
.sa_family
== AF_INET6
)
125 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
127 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
130 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
132 if (ipa
->sa
.sa_family
== AF_INET6
)
133 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
135 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
138 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
140 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
141 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
142 ip
->sa
.sa_family
= AF_INET6
;
144 } else if (nla_len(nla
) >= sizeof(__be32
)) {
145 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
146 ip
->sa
.sa_family
= AF_INET
;
149 return -EAFNOSUPPORT
;
153 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
154 const union vxlan_addr
*ip
)
156 if (ip
->sa
.sa_family
== AF_INET6
)
157 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
159 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
162 #else /* !CONFIG_IPV6 */
165 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
167 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
170 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
172 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
175 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
177 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
180 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
182 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
183 return -EAFNOSUPPORT
;
184 } else if (nla_len(nla
) >= sizeof(__be32
)) {
185 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
186 ip
->sa
.sa_family
= AF_INET
;
189 return -EAFNOSUPPORT
;
193 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
194 const union vxlan_addr
*ip
)
196 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
200 /* Virtual Network hash table head */
201 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
203 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
206 /* Socket hash table head */
207 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
209 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
211 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
214 /* First remote destination for a forwarding entry.
215 * Guaranteed to be non-NULL because remotes are never deleted.
217 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
219 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
222 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
224 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
227 /* Find VXLAN socket based on network namespace, address family and UDP port
228 * and enabled unshareable flags.
230 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
231 __be16 port
, u32 flags
)
233 struct vxlan_sock
*vs
;
235 flags
&= VXLAN_F_RCV_FLAGS
;
237 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
238 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
239 vxlan_get_sk_family(vs
) == family
&&
246 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
248 struct vxlan_dev
*vxlan
;
250 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
251 if (vxlan
->default_dst
.remote_vni
== id
)
258 /* Look up VNI in a per net namespace table */
259 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
,
260 sa_family_t family
, __be16 port
,
263 struct vxlan_sock
*vs
;
265 vs
= vxlan_find_sock(net
, family
, port
, flags
);
269 return vxlan_vs_find_vni(vs
, id
);
272 /* Fill in neighbour message in skbuff. */
273 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
274 const struct vxlan_fdb
*fdb
,
275 u32 portid
, u32 seq
, int type
, unsigned int flags
,
276 const struct vxlan_rdst
*rdst
)
278 unsigned long now
= jiffies
;
279 struct nda_cacheinfo ci
;
280 struct nlmsghdr
*nlh
;
282 bool send_ip
, send_eth
;
284 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
288 ndm
= nlmsg_data(nlh
);
289 memset(ndm
, 0, sizeof(*ndm
));
291 send_eth
= send_ip
= true;
293 if (type
== RTM_GETNEIGH
) {
294 ndm
->ndm_family
= AF_INET
;
295 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
296 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
298 ndm
->ndm_family
= AF_BRIDGE
;
299 ndm
->ndm_state
= fdb
->state
;
300 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
301 ndm
->ndm_flags
= fdb
->flags
;
302 ndm
->ndm_type
= RTN_UNICAST
;
304 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
305 nla_put_s32(skb
, NDA_LINK_NETNSID
,
306 peernet2id_alloc(dev_net(vxlan
->dev
), vxlan
->net
)))
307 goto nla_put_failure
;
309 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
310 goto nla_put_failure
;
312 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
313 goto nla_put_failure
;
315 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
316 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
317 goto nla_put_failure
;
318 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
319 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
320 goto nla_put_failure
;
321 if (rdst
->remote_ifindex
&&
322 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
323 goto nla_put_failure
;
325 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
326 ci
.ndm_confirmed
= 0;
327 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
330 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
331 goto nla_put_failure
;
337 nlmsg_cancel(skb
, nlh
);
341 static inline size_t vxlan_nlmsg_size(void)
343 return NLMSG_ALIGN(sizeof(struct ndmsg
))
344 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
345 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
346 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
347 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
348 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
349 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
350 + nla_total_size(sizeof(struct nda_cacheinfo
));
353 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
354 struct vxlan_rdst
*rd
, int type
)
356 struct net
*net
= dev_net(vxlan
->dev
);
360 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
364 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
366 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
367 WARN_ON(err
== -EMSGSIZE
);
372 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
376 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
379 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
381 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
382 struct vxlan_fdb f
= {
385 struct vxlan_rdst remote
= {
386 .remote_ip
= *ipa
, /* goes to NDA_DST */
387 .remote_vni
= VXLAN_N_VID
,
390 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
393 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
395 struct vxlan_fdb f
= {
398 struct vxlan_rdst remote
= { };
400 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
402 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
405 /* Hash Ethernet address */
406 static u32
eth_hash(const unsigned char *addr
)
408 u64 value
= get_unaligned((u64
*)addr
);
410 /* only want 6 bytes */
416 return hash_64(value
, FDB_HASH_BITS
);
419 /* Hash chain to use given mac address */
420 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
423 return &vxlan
->fdb_head
[eth_hash(mac
)];
426 /* Look up Ethernet address in forwarding table */
427 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
430 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
433 hlist_for_each_entry_rcu(f
, head
, hlist
) {
434 if (ether_addr_equal(mac
, f
->eth_addr
))
441 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
446 f
= __vxlan_find_mac(vxlan
, mac
);
453 /* caller should hold vxlan->hash_lock */
454 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
455 union vxlan_addr
*ip
, __be16 port
,
456 __u32 vni
, __u32 ifindex
)
458 struct vxlan_rdst
*rd
;
460 list_for_each_entry(rd
, &f
->remotes
, list
) {
461 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
462 rd
->remote_port
== port
&&
463 rd
->remote_vni
== vni
&&
464 rd
->remote_ifindex
== ifindex
)
471 /* Replace destination of unicast mac */
472 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
473 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
475 struct vxlan_rdst
*rd
;
477 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
481 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
485 rd
->remote_port
= port
;
486 rd
->remote_vni
= vni
;
487 rd
->remote_ifindex
= ifindex
;
491 /* Add/update destinations for multicast */
492 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
493 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
494 __u32 ifindex
, struct vxlan_rdst
**rdp
)
496 struct vxlan_rdst
*rd
;
498 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
502 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
506 rd
->remote_port
= port
;
507 rd
->remote_vni
= vni
;
508 rd
->remote_ifindex
= ifindex
;
510 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
516 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
518 struct vxlanhdr
*vh
, size_t hdrlen
,
519 u32 data
, struct gro_remcsum
*grc
,
522 size_t start
, offset
;
524 if (skb
->remcsum_offload
)
527 if (!NAPI_GRO_CB(skb
)->csum_valid
)
530 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
531 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
532 offsetof(struct udphdr
, check
) :
533 offsetof(struct tcphdr
, check
));
535 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
536 start
, offset
, grc
, nopartial
);
538 skb
->remcsum_offload
= 1;
543 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
,
545 struct udp_offload
*uoff
)
547 struct sk_buff
*p
, **pp
= NULL
;
548 struct vxlanhdr
*vh
, *vh2
;
549 unsigned int hlen
, off_vx
;
551 struct vxlan_sock
*vs
= container_of(uoff
, struct vxlan_sock
,
554 struct gro_remcsum grc
;
556 skb_gro_remcsum_init(&grc
);
558 off_vx
= skb_gro_offset(skb
);
559 hlen
= off_vx
+ sizeof(*vh
);
560 vh
= skb_gro_header_fast(skb
, off_vx
);
561 if (skb_gro_header_hard(skb
, hlen
)) {
562 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
567 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
569 flags
= ntohl(vh
->vx_flags
);
571 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
572 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
573 ntohl(vh
->vx_vni
), &grc
,
575 VXLAN_F_REMCSUM_NOPARTIAL
));
581 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
585 for (p
= *head
; p
; p
= p
->next
) {
586 if (!NAPI_GRO_CB(p
)->same_flow
)
589 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
590 if (vh
->vx_flags
!= vh2
->vx_flags
||
591 vh
->vx_vni
!= vh2
->vx_vni
) {
592 NAPI_GRO_CB(p
)->same_flow
= 0;
597 pp
= eth_gro_receive(head
, skb
);
600 skb_gro_remcsum_cleanup(skb
, &grc
);
601 NAPI_GRO_CB(skb
)->flush
|= flush
;
606 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
607 struct udp_offload
*uoff
)
609 udp_tunnel_gro_complete(skb
, nhoff
);
611 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
614 /* Notify netdevs that UDP port started listening */
615 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
617 struct net_device
*dev
;
618 struct sock
*sk
= vs
->sock
->sk
;
619 struct net
*net
= sock_net(sk
);
620 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
621 __be16 port
= inet_sk(sk
)->inet_sport
;
624 if (sa_family
== AF_INET
) {
625 err
= udp_add_offload(&vs
->udp_offloads
);
627 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
631 for_each_netdev_rcu(net
, dev
) {
632 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
633 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
639 /* Notify netdevs that UDP port is no more listening */
640 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
642 struct net_device
*dev
;
643 struct sock
*sk
= vs
->sock
->sk
;
644 struct net
*net
= sock_net(sk
);
645 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
646 __be16 port
= inet_sk(sk
)->inet_sport
;
649 for_each_netdev_rcu(net
, dev
) {
650 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
651 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
656 if (sa_family
== AF_INET
)
657 udp_del_offload(&vs
->udp_offloads
);
660 /* Add new entry to forwarding table -- assumes lock held */
661 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
662 const u8
*mac
, union vxlan_addr
*ip
,
663 __u16 state
, __u16 flags
,
664 __be16 port
, __u32 vni
, __u32 ifindex
,
667 struct vxlan_rdst
*rd
= NULL
;
671 f
= __vxlan_find_mac(vxlan
, mac
);
673 if (flags
& NLM_F_EXCL
) {
674 netdev_dbg(vxlan
->dev
,
675 "lost race to create %pM\n", mac
);
678 if (f
->state
!= state
) {
680 f
->updated
= jiffies
;
683 if (f
->flags
!= ndm_flags
) {
684 f
->flags
= ndm_flags
;
685 f
->updated
= jiffies
;
688 if ((flags
& NLM_F_REPLACE
)) {
689 /* Only change unicasts */
690 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
691 is_zero_ether_addr(f
->eth_addr
))) {
692 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
697 if ((flags
& NLM_F_APPEND
) &&
698 (is_multicast_ether_addr(f
->eth_addr
) ||
699 is_zero_ether_addr(f
->eth_addr
))) {
700 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
708 if (!(flags
& NLM_F_CREATE
))
711 if (vxlan
->cfg
.addrmax
&&
712 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
715 /* Disallow replace to add a multicast entry */
716 if ((flags
& NLM_F_REPLACE
) &&
717 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
720 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
721 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
727 f
->flags
= ndm_flags
;
728 f
->updated
= f
->used
= jiffies
;
729 INIT_LIST_HEAD(&f
->remotes
);
730 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
732 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
735 hlist_add_head_rcu(&f
->hlist
,
736 vxlan_fdb_head(vxlan
, mac
));
741 rd
= first_remote_rtnl(f
);
742 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
748 static void vxlan_fdb_free(struct rcu_head
*head
)
750 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
751 struct vxlan_rdst
*rd
, *nd
;
753 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
758 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
760 netdev_dbg(vxlan
->dev
,
761 "delete %pM\n", f
->eth_addr
);
764 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
766 hlist_del_rcu(&f
->hlist
);
767 call_rcu(&f
->rcu
, vxlan_fdb_free
);
770 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
771 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
773 struct net
*net
= dev_net(vxlan
->dev
);
777 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
781 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
782 if (remote
->sa
.sa_family
== AF_INET
) {
783 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
784 ip
->sa
.sa_family
= AF_INET
;
785 #if IS_ENABLED(CONFIG_IPV6)
787 ip
->sin6
.sin6_addr
= in6addr_any
;
788 ip
->sa
.sa_family
= AF_INET6
;
794 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
796 *port
= nla_get_be16(tb
[NDA_PORT
]);
798 *port
= vxlan
->cfg
.dst_port
;
802 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
804 *vni
= nla_get_u32(tb
[NDA_VNI
]);
806 *vni
= vxlan
->default_dst
.remote_vni
;
809 if (tb
[NDA_IFINDEX
]) {
810 struct net_device
*tdev
;
812 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
814 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
815 tdev
= __dev_get_by_index(net
, *ifindex
);
817 return -EADDRNOTAVAIL
;
825 /* Add static entry (via netlink) */
826 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
827 struct net_device
*dev
,
828 const unsigned char *addr
, u16 vid
, u16 flags
)
830 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
831 /* struct net *net = dev_net(vxlan->dev); */
837 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
838 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
843 if (tb
[NDA_DST
] == NULL
)
846 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
850 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
851 return -EAFNOSUPPORT
;
853 spin_lock_bh(&vxlan
->hash_lock
);
854 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
855 port
, vni
, ifindex
, ndm
->ndm_flags
);
856 spin_unlock_bh(&vxlan
->hash_lock
);
861 /* Delete entry (via netlink) */
862 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
863 struct net_device
*dev
,
864 const unsigned char *addr
, u16 vid
)
866 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
868 struct vxlan_rdst
*rd
= NULL
;
874 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
880 spin_lock_bh(&vxlan
->hash_lock
);
881 f
= vxlan_find_mac(vxlan
, addr
);
885 if (!vxlan_addr_any(&ip
)) {
886 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
893 /* remove a destination if it's not the only one on the list,
894 * otherwise destroy the fdb entry
896 if (rd
&& !list_is_singular(&f
->remotes
)) {
897 list_del_rcu(&rd
->list
);
898 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
903 vxlan_fdb_destroy(vxlan
, f
);
906 spin_unlock_bh(&vxlan
->hash_lock
);
911 /* Dump forwarding table */
912 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
913 struct net_device
*dev
,
914 struct net_device
*filter_dev
, int idx
)
916 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
919 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
923 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
924 struct vxlan_rdst
*rd
;
926 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
927 if (idx
< cb
->args
[0])
930 err
= vxlan_fdb_info(skb
, vxlan
, f
,
931 NETLINK_CB(cb
->skb
).portid
,
946 /* Watch incoming packets to learn mapping between Ethernet address
947 * and Tunnel endpoint.
948 * Return true if packet is bogus and should be dropped.
950 static bool vxlan_snoop(struct net_device
*dev
,
951 union vxlan_addr
*src_ip
, const u8
*src_mac
)
953 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
956 f
= vxlan_find_mac(vxlan
, src_mac
);
958 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
960 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
963 /* Don't migrate static entries, drop packets */
964 if (f
->state
& NUD_NOARP
)
969 "%pM migrated from %pIS to %pIS\n",
970 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
972 rdst
->remote_ip
= *src_ip
;
973 f
->updated
= jiffies
;
974 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
976 /* learned new entry */
977 spin_lock(&vxlan
->hash_lock
);
979 /* close off race between vxlan_flush and incoming packets */
980 if (netif_running(dev
))
981 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
983 NLM_F_EXCL
|NLM_F_CREATE
,
985 vxlan
->default_dst
.remote_vni
,
987 spin_unlock(&vxlan
->hash_lock
);
993 /* See if multicast group is already in use by other ID */
994 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
996 struct vxlan_dev
*vxlan
;
998 /* The vxlan_sock is only used by dev, leaving group has
999 * no effect on other vxlan devices.
1001 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1004 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1005 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1008 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1011 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1012 &dev
->default_dst
.remote_ip
))
1015 if (vxlan
->default_dst
.remote_ifindex
!=
1016 dev
->default_dst
.remote_ifindex
)
1025 static void vxlan_sock_release(struct vxlan_sock
*vs
)
1027 struct sock
*sk
= vs
->sock
->sk
;
1028 struct net
*net
= sock_net(sk
);
1029 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1031 if (!atomic_dec_and_test(&vs
->refcnt
))
1034 spin_lock(&vn
->sock_lock
);
1035 hlist_del_rcu(&vs
->hlist
);
1036 vxlan_notify_del_rx_port(vs
);
1037 spin_unlock(&vn
->sock_lock
);
1039 queue_work(vxlan_wq
, &vs
->del_work
);
1042 /* Update multicast group membership when first VNI on
1043 * multicast address is brought up
1045 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1047 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1048 struct sock
*sk
= vs
->sock
->sk
;
1049 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1050 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1054 if (ip
->sa
.sa_family
== AF_INET
) {
1055 struct ip_mreqn mreq
= {
1056 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1057 .imr_ifindex
= ifindex
,
1060 ret
= ip_mc_join_group(sk
, &mreq
);
1061 #if IS_ENABLED(CONFIG_IPV6)
1063 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1064 &ip
->sin6
.sin6_addr
);
1072 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1073 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1075 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1076 struct sock
*sk
= vs
->sock
->sk
;
1077 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1078 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1082 if (ip
->sa
.sa_family
== AF_INET
) {
1083 struct ip_mreqn mreq
= {
1084 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1085 .imr_ifindex
= ifindex
,
1088 ret
= ip_mc_leave_group(sk
, &mreq
);
1089 #if IS_ENABLED(CONFIG_IPV6)
1091 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1092 &ip
->sin6
.sin6_addr
);
1100 static struct vxlanhdr
*vxlan_remcsum(struct sk_buff
*skb
, struct vxlanhdr
*vh
,
1101 size_t hdrlen
, u32 data
, bool nopartial
)
1103 size_t start
, offset
, plen
;
1105 if (skb
->remcsum_offload
)
1108 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
1109 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
1110 offsetof(struct udphdr
, check
) :
1111 offsetof(struct tcphdr
, check
));
1113 plen
= hdrlen
+ offset
+ sizeof(u16
);
1115 if (!pskb_may_pull(skb
, plen
))
1118 vh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1120 skb_remcsum_process(skb
, (void *)vh
+ hdrlen
, start
, offset
,
1126 static void vxlan_rcv(struct vxlan_sock
*vs
, struct sk_buff
*skb
,
1127 struct vxlan_metadata
*md
, u32 vni
,
1128 struct metadata_dst
*tun_dst
)
1130 struct iphdr
*oip
= NULL
;
1131 struct ipv6hdr
*oip6
= NULL
;
1132 struct vxlan_dev
*vxlan
;
1133 struct pcpu_sw_netstats
*stats
;
1134 union vxlan_addr saddr
;
1136 union vxlan_addr
*remote_ip
;
1138 /* For flow based devices, map all packets to VNI 0 */
1139 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
1142 /* Is this VNI defined? */
1143 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1147 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1148 skb_reset_mac_header(skb
);
1149 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1150 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1151 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1153 /* Ignore packet loops (and multicast echo) */
1154 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1157 /* Re-examine inner Ethernet packet */
1158 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1160 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1161 saddr
.sa
.sa_family
= AF_INET
;
1162 #if IS_ENABLED(CONFIG_IPV6)
1164 oip6
= ipv6_hdr(skb
);
1165 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1166 saddr
.sa
.sa_family
= AF_INET6
;
1171 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1175 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1176 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1179 skb_reset_network_header(skb
);
1180 /* In flow-based mode, GBP is carried in dst_metadata */
1181 if (!(vs
->flags
& VXLAN_F_COLLECT_METADATA
))
1182 skb
->mark
= md
->gbp
;
1185 err
= IP6_ECN_decapsulate(oip6
, skb
);
1187 err
= IP_ECN_decapsulate(oip
, skb
);
1189 if (unlikely(err
)) {
1190 if (log_ecn_error
) {
1192 net_info_ratelimited("non-ECT from %pI6\n",
1195 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1196 &oip
->saddr
, oip
->tos
);
1199 ++vxlan
->dev
->stats
.rx_frame_errors
;
1200 ++vxlan
->dev
->stats
.rx_errors
;
1205 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1206 u64_stats_update_begin(&stats
->syncp
);
1207 stats
->rx_packets
++;
1208 stats
->rx_bytes
+= skb
->len
;
1209 u64_stats_update_end(&stats
->syncp
);
1211 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1216 dst_release((struct dst_entry
*)tun_dst
);
1218 /* Consume bad packet */
1222 /* Callback from net/ipv4/udp.c to receive packets */
1223 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1225 struct metadata_dst
*tun_dst
= NULL
;
1226 struct ip_tunnel_info
*info
;
1227 struct vxlan_sock
*vs
;
1228 struct vxlanhdr
*vxh
;
1230 struct vxlan_metadata _md
;
1231 struct vxlan_metadata
*md
= &_md
;
1233 /* Need Vxlan and inner Ethernet header to be present */
1234 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1237 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1238 flags
= ntohl(vxh
->vx_flags
);
1239 vni
= ntohl(vxh
->vx_vni
);
1241 if (flags
& VXLAN_HF_VNI
) {
1242 flags
&= ~VXLAN_HF_VNI
;
1244 /* VNI flag always required to be set */
1248 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1250 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1252 vs
= rcu_dereference_sk_user_data(sk
);
1256 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
1257 vxh
= vxlan_remcsum(skb
, vxh
, sizeof(struct vxlanhdr
), vni
,
1258 !!(vs
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1262 flags
&= ~VXLAN_HF_RCO
;
1263 vni
&= VXLAN_VNI_MASK
;
1266 if (vxlan_collect_metadata(vs
)) {
1267 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1268 cpu_to_be64(vni
>> 8), sizeof(*md
));
1273 info
= &tun_dst
->u
.tun_info
;
1274 md
= ip_tunnel_info_opts(info
, sizeof(*md
));
1276 memset(md
, 0, sizeof(*md
));
1279 /* For backwards compatibility, only allow reserved fields to be
1280 * used by VXLAN extensions if explicitly requested.
1282 if ((flags
& VXLAN_HF_GBP
) && (vs
->flags
& VXLAN_F_GBP
)) {
1283 struct vxlanhdr_gbp
*gbp
;
1285 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1286 md
->gbp
= ntohs(gbp
->policy_id
);
1289 info
->key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1291 if (gbp
->dont_learn
)
1292 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1294 if (gbp
->policy_applied
)
1295 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1297 flags
&= ~VXLAN_GBP_USED_BITS
;
1300 if (flags
|| vni
& ~VXLAN_VNI_MASK
) {
1301 /* If there are any unprocessed flags remaining treat
1302 * this as a malformed packet. This behavior diverges from
1303 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1304 * in reserved fields are to be ignored. The approach here
1305 * maintains compatibility with previous stack code, and also
1306 * is more robust and provides a little more security in
1307 * adding extensions to VXLAN.
1313 vxlan_rcv(vs
, skb
, md
, vni
>> 8, tun_dst
);
1317 /* Consume bad packet */
1322 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1323 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1327 dst_release((struct dst_entry
*)tun_dst
);
1329 /* Return non vxlan pkt */
1333 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1335 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1336 struct arphdr
*parp
;
1339 struct neighbour
*n
;
1341 if (dev
->flags
& IFF_NOARP
)
1344 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1345 dev
->stats
.tx_dropped
++;
1348 parp
= arp_hdr(skb
);
1350 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1351 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1352 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1353 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1354 parp
->ar_hln
!= dev
->addr_len
||
1357 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1359 arpptr
+= dev
->addr_len
; /* sha */
1360 memcpy(&sip
, arpptr
, sizeof(sip
));
1361 arpptr
+= sizeof(sip
);
1362 arpptr
+= dev
->addr_len
; /* tha */
1363 memcpy(&tip
, arpptr
, sizeof(tip
));
1365 if (ipv4_is_loopback(tip
) ||
1366 ipv4_is_multicast(tip
))
1369 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1372 struct vxlan_fdb
*f
;
1373 struct sk_buff
*reply
;
1375 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1380 f
= vxlan_find_mac(vxlan
, n
->ha
);
1381 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1382 /* bridge-local neighbor */
1387 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1395 skb_reset_mac_header(reply
);
1396 __skb_pull(reply
, skb_network_offset(reply
));
1397 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1398 reply
->pkt_type
= PACKET_HOST
;
1400 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1401 dev
->stats
.rx_dropped
++;
1402 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1403 union vxlan_addr ipa
= {
1404 .sin
.sin_addr
.s_addr
= tip
,
1405 .sin
.sin_family
= AF_INET
,
1408 vxlan_ip_miss(dev
, &ipa
);
1412 return NETDEV_TX_OK
;
1415 #if IS_ENABLED(CONFIG_IPV6)
1416 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1417 struct neighbour
*n
, bool isrouter
)
1419 struct net_device
*dev
= request
->dev
;
1420 struct sk_buff
*reply
;
1421 struct nd_msg
*ns
, *na
;
1422 struct ipv6hdr
*pip6
;
1424 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1431 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1432 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1433 reply
= alloc_skb(len
, GFP_ATOMIC
);
1437 reply
->protocol
= htons(ETH_P_IPV6
);
1439 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1440 skb_push(reply
, sizeof(struct ethhdr
));
1441 skb_set_mac_header(reply
, 0);
1443 ns
= (struct nd_msg
*)skb_transport_header(request
);
1445 daddr
= eth_hdr(request
)->h_source
;
1446 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1447 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1448 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1449 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1454 /* Ethernet header */
1455 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1456 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1457 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1458 reply
->protocol
= htons(ETH_P_IPV6
);
1460 skb_pull(reply
, sizeof(struct ethhdr
));
1461 skb_set_network_header(reply
, 0);
1462 skb_put(reply
, sizeof(struct ipv6hdr
));
1466 pip6
= ipv6_hdr(reply
);
1467 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1469 pip6
->priority
= ipv6_hdr(request
)->priority
;
1470 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1471 pip6
->hop_limit
= 255;
1472 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1473 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1475 skb_pull(reply
, sizeof(struct ipv6hdr
));
1476 skb_set_transport_header(reply
, 0);
1478 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1480 /* Neighbor Advertisement */
1481 memset(na
, 0, sizeof(*na
)+na_olen
);
1482 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1483 na
->icmph
.icmp6_router
= isrouter
;
1484 na
->icmph
.icmp6_override
= 1;
1485 na
->icmph
.icmp6_solicited
= 1;
1486 na
->target
= ns
->target
;
1487 ether_addr_copy(&na
->opt
[2], n
->ha
);
1488 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1489 na
->opt
[1] = na_olen
>> 3;
1491 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1492 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1493 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1495 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1497 skb_push(reply
, sizeof(struct ipv6hdr
));
1499 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1504 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1506 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1508 const struct ipv6hdr
*iphdr
;
1509 const struct in6_addr
*saddr
, *daddr
;
1510 struct neighbour
*n
;
1511 struct inet6_dev
*in6_dev
;
1513 in6_dev
= __in6_dev_get(dev
);
1517 iphdr
= ipv6_hdr(skb
);
1518 saddr
= &iphdr
->saddr
;
1519 daddr
= &iphdr
->daddr
;
1521 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1522 if (msg
->icmph
.icmp6_code
!= 0 ||
1523 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1526 if (ipv6_addr_loopback(daddr
) ||
1527 ipv6_addr_is_multicast(&msg
->target
))
1530 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1533 struct vxlan_fdb
*f
;
1534 struct sk_buff
*reply
;
1536 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1541 f
= vxlan_find_mac(vxlan
, n
->ha
);
1542 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1543 /* bridge-local neighbor */
1548 reply
= vxlan_na_create(skb
, n
,
1549 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1556 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1557 dev
->stats
.rx_dropped
++;
1559 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1560 union vxlan_addr ipa
= {
1561 .sin6
.sin6_addr
= msg
->target
,
1562 .sin6
.sin6_family
= AF_INET6
,
1565 vxlan_ip_miss(dev
, &ipa
);
1570 return NETDEV_TX_OK
;
1574 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1576 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1577 struct neighbour
*n
;
1579 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1583 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1588 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1591 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1592 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1593 union vxlan_addr ipa
= {
1594 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1595 .sin
.sin_family
= AF_INET
,
1598 vxlan_ip_miss(dev
, &ipa
);
1604 #if IS_ENABLED(CONFIG_IPV6)
1607 struct ipv6hdr
*pip6
;
1609 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1611 pip6
= ipv6_hdr(skb
);
1612 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1613 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1614 union vxlan_addr ipa
= {
1615 .sin6
.sin6_addr
= pip6
->daddr
,
1616 .sin6
.sin6_family
= AF_INET6
,
1619 vxlan_ip_miss(dev
, &ipa
);
1633 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1635 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1637 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1646 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1647 struct vxlan_metadata
*md
)
1649 struct vxlanhdr_gbp
*gbp
;
1654 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1655 vxh
->vx_flags
|= htonl(VXLAN_HF_GBP
);
1657 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1658 gbp
->dont_learn
= 1;
1660 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1661 gbp
->policy_applied
= 1;
1663 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1666 #if IS_ENABLED(CONFIG_IPV6)
1667 static int vxlan6_xmit_skb(struct dst_entry
*dst
, struct sock
*sk
,
1668 struct sk_buff
*skb
,
1669 struct net_device
*dev
, struct in6_addr
*saddr
,
1670 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1671 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1672 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1674 struct vxlanhdr
*vxh
;
1677 bool udp_sum
= !(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
1678 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1679 u16 hdrlen
= sizeof(struct vxlanhdr
);
1681 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1682 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1683 int csum_start
= skb_checksum_start_offset(skb
);
1685 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1686 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1687 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1688 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1690 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1694 skb_scrub_packet(skb
, xnet
);
1696 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1697 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1698 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1700 /* Need space for new headers (invalidates iph ptr) */
1701 err
= skb_cow_head(skb
, min_headroom
);
1702 if (unlikely(err
)) {
1707 skb
= vlan_hwaccel_push_inside(skb
);
1708 if (WARN_ON(!skb
)) {
1713 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1719 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1720 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1723 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1724 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1727 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1728 data
|= VXLAN_RCO_UDP
;
1730 vxh
->vx_vni
|= htonl(data
);
1731 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1733 if (!skb_is_gso(skb
)) {
1734 skb
->ip_summed
= CHECKSUM_NONE
;
1735 skb
->encapsulation
= 0;
1739 if (vxflags
& VXLAN_F_GBP
)
1740 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1742 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1744 udp_tunnel6_xmit_skb(dst
, sk
, skb
, dev
, saddr
, daddr
, prio
,
1745 ttl
, src_port
, dst_port
,
1746 !!(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
));
1754 static int vxlan_xmit_skb(struct rtable
*rt
, struct sock
*sk
, struct sk_buff
*skb
,
1755 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1756 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1757 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1759 struct vxlanhdr
*vxh
;
1762 bool udp_sum
= !!(vxflags
& VXLAN_F_UDP_CSUM
);
1763 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1764 u16 hdrlen
= sizeof(struct vxlanhdr
);
1766 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1767 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1768 int csum_start
= skb_checksum_start_offset(skb
);
1770 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1771 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1772 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1773 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1775 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1779 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1780 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1781 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1783 /* Need space for new headers (invalidates iph ptr) */
1784 err
= skb_cow_head(skb
, min_headroom
);
1785 if (unlikely(err
)) {
1790 skb
= vlan_hwaccel_push_inside(skb
);
1794 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1796 return PTR_ERR(skb
);
1798 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1799 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1802 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1803 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1806 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1807 data
|= VXLAN_RCO_UDP
;
1809 vxh
->vx_vni
|= htonl(data
);
1810 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1812 if (!skb_is_gso(skb
)) {
1813 skb
->ip_summed
= CHECKSUM_NONE
;
1814 skb
->encapsulation
= 0;
1818 if (vxflags
& VXLAN_F_GBP
)
1819 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1821 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1823 return udp_tunnel_xmit_skb(rt
, sk
, skb
, src
, dst
, tos
,
1824 ttl
, df
, src_port
, dst_port
, xnet
,
1825 !(vxflags
& VXLAN_F_UDP_CSUM
));
1828 /* Bypass encapsulation if the destination is local */
1829 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1830 struct vxlan_dev
*dst_vxlan
)
1832 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1833 union vxlan_addr loopback
;
1834 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1835 struct net_device
*dev
= skb
->dev
;
1838 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1839 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1840 skb
->pkt_type
= PACKET_HOST
;
1841 skb
->encapsulation
= 0;
1842 skb
->dev
= dst_vxlan
->dev
;
1843 __skb_pull(skb
, skb_network_offset(skb
));
1845 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1846 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1847 loopback
.sa
.sa_family
= AF_INET
;
1848 #if IS_ENABLED(CONFIG_IPV6)
1850 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1851 loopback
.sa
.sa_family
= AF_INET6
;
1855 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1856 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1858 u64_stats_update_begin(&tx_stats
->syncp
);
1859 tx_stats
->tx_packets
++;
1860 tx_stats
->tx_bytes
+= len
;
1861 u64_stats_update_end(&tx_stats
->syncp
);
1863 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1864 u64_stats_update_begin(&rx_stats
->syncp
);
1865 rx_stats
->rx_packets
++;
1866 rx_stats
->rx_bytes
+= len
;
1867 u64_stats_update_end(&rx_stats
->syncp
);
1869 dev
->stats
.rx_dropped
++;
1873 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1874 struct vxlan_rdst
*rdst
, bool did_rsc
)
1876 struct ip_tunnel_info
*info
;
1877 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1878 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1879 unsigned short family
= vxlan_get_sk_family(vxlan
->vn_sock
);
1880 struct rtable
*rt
= NULL
;
1881 const struct iphdr
*old_iph
;
1883 union vxlan_addr
*dst
;
1884 union vxlan_addr remote_ip
;
1885 struct vxlan_metadata _md
;
1886 struct vxlan_metadata
*md
= &_md
;
1887 __be16 src_port
= 0, dst_port
;
1892 u32 flags
= vxlan
->flags
;
1894 info
= skb_tunnel_info(skb
);
1897 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
1898 vni
= rdst
->remote_vni
;
1899 dst
= &rdst
->remote_ip
;
1902 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1907 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
1908 vni
= be64_to_cpu(info
->key
.tun_id
);
1909 remote_ip
.sa
.sa_family
= family
;
1910 if (family
== AF_INET
)
1911 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
1913 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
1917 if (vxlan_addr_any(dst
)) {
1919 /* short-circuited back to local bridge */
1920 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1926 old_iph
= ip_hdr(skb
);
1928 ttl
= vxlan
->cfg
.ttl
;
1929 if (!ttl
&& vxlan_addr_multicast(dst
))
1932 tos
= vxlan
->cfg
.tos
;
1934 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1936 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
1937 vxlan
->cfg
.port_max
, true);
1940 if (info
->key
.tun_flags
& TUNNEL_CSUM
)
1941 flags
|= VXLAN_F_UDP_CSUM
;
1943 flags
&= ~VXLAN_F_UDP_CSUM
;
1945 ttl
= info
->key
.ttl
;
1946 tos
= info
->key
.tos
;
1948 if (info
->options_len
)
1949 md
= ip_tunnel_info_opts(info
, sizeof(*md
));
1951 md
->gbp
= skb
->mark
;
1954 if (dst
->sa
.sa_family
== AF_INET
) {
1955 if (info
&& (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
))
1958 memset(&fl4
, 0, sizeof(fl4
));
1959 fl4
.flowi4_oif
= rdst
? rdst
->remote_ifindex
: 0;
1960 fl4
.flowi4_tos
= RT_TOS(tos
);
1961 fl4
.flowi4_mark
= skb
->mark
;
1962 fl4
.flowi4_proto
= IPPROTO_UDP
;
1963 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1964 fl4
.saddr
= vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
;
1966 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1968 netdev_dbg(dev
, "no route to %pI4\n",
1969 &dst
->sin
.sin_addr
.s_addr
);
1970 dev
->stats
.tx_carrier_errors
++;
1974 if (rt
->dst
.dev
== dev
) {
1975 netdev_dbg(dev
, "circular route to %pI4\n",
1976 &dst
->sin
.sin_addr
.s_addr
);
1977 dev
->stats
.collisions
++;
1981 /* Bypass encapsulation if the destination is local */
1982 if (rt
->rt_flags
& RTCF_LOCAL
&&
1983 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1984 struct vxlan_dev
*dst_vxlan
;
1987 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
1988 dst
->sa
.sa_family
, dst_port
,
1992 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1996 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1997 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1998 err
= vxlan_xmit_skb(rt
, sk
, skb
, fl4
.saddr
,
1999 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2000 src_port
, dst_port
, htonl(vni
<< 8), md
,
2001 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2004 /* skb is already freed. */
2009 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
2010 #if IS_ENABLED(CONFIG_IPV6)
2012 struct dst_entry
*ndst
;
2016 memset(&fl6
, 0, sizeof(fl6
));
2017 fl6
.flowi6_oif
= rdst
? rdst
->remote_ifindex
: 0;
2018 fl6
.daddr
= dst
->sin6
.sin6_addr
;
2019 fl6
.saddr
= vxlan
->cfg
.saddr
.sin6
.sin6_addr
;
2020 fl6
.flowi6_mark
= skb
->mark
;
2021 fl6
.flowi6_proto
= IPPROTO_UDP
;
2023 if (ipv6_stub
->ipv6_dst_lookup(vxlan
->net
, sk
, &ndst
, &fl6
)) {
2024 netdev_dbg(dev
, "no route to %pI6\n",
2025 &dst
->sin6
.sin6_addr
);
2026 dev
->stats
.tx_carrier_errors
++;
2030 if (ndst
->dev
== dev
) {
2031 netdev_dbg(dev
, "circular route to %pI6\n",
2032 &dst
->sin6
.sin6_addr
);
2034 dev
->stats
.collisions
++;
2038 /* Bypass encapsulation if the destination is local */
2039 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2040 if (rt6i_flags
& RTF_LOCAL
&&
2041 !(rt6i_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 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2055 err
= vxlan6_xmit_skb(ndst
, sk
, skb
, dev
, &fl6
.saddr
, &fl6
.daddr
,
2056 0, ttl
, src_port
, dst_port
, htonl(vni
<< 8), md
,
2057 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2065 dev
->stats
.tx_dropped
++;
2071 dev
->stats
.tx_errors
++;
2076 /* Transmit local packets over Vxlan
2078 * Outer IP header inherits ECN and DF from inner header.
2079 * Outer UDP destination is the VXLAN assigned port.
2080 * source port is based on hash of flow
2082 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2084 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2085 const struct ip_tunnel_info
*info
;
2087 bool did_rsc
= false;
2088 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2089 struct vxlan_fdb
*f
;
2091 info
= skb_tunnel_info(skb
);
2093 skb_reset_mac_header(skb
);
2096 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2097 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2098 return arp_reduce(dev
, skb
);
2099 #if IS_ENABLED(CONFIG_IPV6)
2100 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2101 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2102 + sizeof(struct nd_msg
)) &&
2103 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2106 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2107 if (msg
->icmph
.icmp6_code
== 0 &&
2108 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2109 return neigh_reduce(dev
, skb
);
2115 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
&&
2116 info
&& info
->mode
& IP_TUNNEL_INFO_TX
) {
2117 vxlan_xmit_one(skb
, dev
, NULL
, false);
2118 return NETDEV_TX_OK
;
2121 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2124 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2125 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2126 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2127 did_rsc
= route_shortcircuit(dev
, skb
);
2129 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2133 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2135 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2136 !is_multicast_ether_addr(eth
->h_dest
))
2137 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2139 dev
->stats
.tx_dropped
++;
2141 return NETDEV_TX_OK
;
2145 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2146 struct sk_buff
*skb1
;
2152 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2154 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2158 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2161 return NETDEV_TX_OK
;
2164 /* Walk the forwarding table and purge stale entries */
2165 static void vxlan_cleanup(unsigned long arg
)
2167 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2168 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2171 if (!netif_running(vxlan
->dev
))
2174 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2175 struct hlist_node
*p
, *n
;
2177 spin_lock_bh(&vxlan
->hash_lock
);
2178 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2180 = container_of(p
, struct vxlan_fdb
, hlist
);
2181 unsigned long timeout
;
2183 if (f
->state
& NUD_PERMANENT
)
2186 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2187 if (time_before_eq(timeout
, jiffies
)) {
2188 netdev_dbg(vxlan
->dev
,
2189 "garbage collect %pM\n",
2191 f
->state
= NUD_STALE
;
2192 vxlan_fdb_destroy(vxlan
, f
);
2193 } else if (time_before(timeout
, next_timer
))
2194 next_timer
= timeout
;
2196 spin_unlock_bh(&vxlan
->hash_lock
);
2199 mod_timer(&vxlan
->age_timer
, next_timer
);
2202 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2204 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2205 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2207 vxlan
->vn_sock
= vs
;
2208 spin_lock(&vn
->sock_lock
);
2209 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2210 spin_unlock(&vn
->sock_lock
);
2213 /* Setup stats when device is created */
2214 static int vxlan_init(struct net_device
*dev
)
2216 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2223 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2225 struct vxlan_fdb
*f
;
2227 spin_lock_bh(&vxlan
->hash_lock
);
2228 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2230 vxlan_fdb_destroy(vxlan
, f
);
2231 spin_unlock_bh(&vxlan
->hash_lock
);
2234 static void vxlan_uninit(struct net_device
*dev
)
2236 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2238 vxlan_fdb_delete_default(vxlan
);
2240 free_percpu(dev
->tstats
);
2243 /* Start ageing timer and join group when device is brought up */
2244 static int vxlan_open(struct net_device
*dev
)
2246 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2247 struct vxlan_sock
*vs
;
2250 vs
= vxlan_sock_add(vxlan
->net
, vxlan
->cfg
.dst_port
,
2251 vxlan
->cfg
.no_share
, vxlan
->flags
);
2255 vxlan_vs_add_dev(vs
, vxlan
);
2257 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2258 ret
= vxlan_igmp_join(vxlan
);
2259 if (ret
== -EADDRINUSE
)
2262 vxlan_sock_release(vs
);
2267 if (vxlan
->cfg
.age_interval
)
2268 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2273 /* Purge the forwarding table */
2274 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2278 spin_lock_bh(&vxlan
->hash_lock
);
2279 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2280 struct hlist_node
*p
, *n
;
2281 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2283 = container_of(p
, struct vxlan_fdb
, hlist
);
2284 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2285 if (!is_zero_ether_addr(f
->eth_addr
))
2286 vxlan_fdb_destroy(vxlan
, f
);
2289 spin_unlock_bh(&vxlan
->hash_lock
);
2292 /* Cleanup timer and forwarding table on shutdown */
2293 static int vxlan_stop(struct net_device
*dev
)
2295 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2296 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2297 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2300 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2301 !vxlan_group_used(vn
, vxlan
))
2302 ret
= vxlan_igmp_leave(vxlan
);
2304 del_timer_sync(&vxlan
->age_timer
);
2307 vxlan_sock_release(vs
);
2312 /* Stub, nothing needs to be done. */
2313 static void vxlan_set_multicast_list(struct net_device
*dev
)
2317 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2319 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2320 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2321 struct net_device
*lowerdev
;
2324 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2325 if (lowerdev
== NULL
)
2326 return eth_change_mtu(dev
, new_mtu
);
2328 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2329 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2331 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2333 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2340 static const struct net_device_ops vxlan_netdev_ops
= {
2341 .ndo_init
= vxlan_init
,
2342 .ndo_uninit
= vxlan_uninit
,
2343 .ndo_open
= vxlan_open
,
2344 .ndo_stop
= vxlan_stop
,
2345 .ndo_start_xmit
= vxlan_xmit
,
2346 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2347 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2348 .ndo_change_mtu
= vxlan_change_mtu
,
2349 .ndo_validate_addr
= eth_validate_addr
,
2350 .ndo_set_mac_address
= eth_mac_addr
,
2351 .ndo_fdb_add
= vxlan_fdb_add
,
2352 .ndo_fdb_del
= vxlan_fdb_delete
,
2353 .ndo_fdb_dump
= vxlan_fdb_dump
,
2356 /* Info for udev, that this is a virtual tunnel endpoint */
2357 static struct device_type vxlan_type
= {
2361 /* Calls the ndo_add_vxlan_port of the caller in order to
2362 * supply the listening VXLAN udp ports. Callers are expected
2363 * to implement the ndo_add_vxlan_port.
2365 void vxlan_get_rx_port(struct net_device
*dev
)
2367 struct vxlan_sock
*vs
;
2368 struct net
*net
= dev_net(dev
);
2369 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2370 sa_family_t sa_family
;
2374 spin_lock(&vn
->sock_lock
);
2375 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2376 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2377 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2378 sa_family
= vxlan_get_sk_family(vs
);
2379 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2383 spin_unlock(&vn
->sock_lock
);
2385 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2387 /* Initialize the device structure. */
2388 static void vxlan_setup(struct net_device
*dev
)
2390 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2393 eth_hw_addr_random(dev
);
2395 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2396 dev
->needed_headroom
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2398 dev
->needed_headroom
= ETH_HLEN
+ VXLAN_HEADROOM
;
2400 dev
->netdev_ops
= &vxlan_netdev_ops
;
2401 dev
->destructor
= free_netdev
;
2402 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2404 dev
->features
|= NETIF_F_LLTX
;
2405 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2406 dev
->features
|= NETIF_F_RXCSUM
;
2407 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2409 dev
->vlan_features
= dev
->features
;
2410 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2411 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2412 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2413 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2414 netif_keep_dst(dev
);
2415 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
| IFF_NO_QUEUE
;
2417 INIT_LIST_HEAD(&vxlan
->next
);
2418 spin_lock_init(&vxlan
->hash_lock
);
2420 init_timer_deferrable(&vxlan
->age_timer
);
2421 vxlan
->age_timer
.function
= vxlan_cleanup
;
2422 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2424 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2428 gro_cells_init(&vxlan
->gro_cells
, dev
);
2430 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2431 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2434 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2435 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2436 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2437 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2438 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2439 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2440 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2441 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2442 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2443 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2444 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2445 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2446 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2447 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2448 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2449 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2450 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2451 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2452 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2453 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2454 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2455 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2456 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2457 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2458 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2459 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2462 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2464 if (tb
[IFLA_ADDRESS
]) {
2465 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2466 pr_debug("invalid link address (not ethernet)\n");
2470 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2471 pr_debug("invalid all zero ethernet address\n");
2472 return -EADDRNOTAVAIL
;
2479 if (data
[IFLA_VXLAN_ID
]) {
2480 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2481 if (id
>= VXLAN_VID_MASK
)
2485 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2486 const struct ifla_vxlan_port_range
*p
2487 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2489 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2490 pr_debug("port range %u .. %u not valid\n",
2491 ntohs(p
->low
), ntohs(p
->high
));
2499 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2500 struct ethtool_drvinfo
*drvinfo
)
2502 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2503 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2506 static const struct ethtool_ops vxlan_ethtool_ops
= {
2507 .get_drvinfo
= vxlan_get_drvinfo
,
2508 .get_link
= ethtool_op_get_link
,
2511 static void vxlan_del_work(struct work_struct
*work
)
2513 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2514 udp_tunnel_sock_release(vs
->sock
);
2518 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2519 __be16 port
, u32 flags
)
2521 struct socket
*sock
;
2522 struct udp_port_cfg udp_conf
;
2525 memset(&udp_conf
, 0, sizeof(udp_conf
));
2528 udp_conf
.family
= AF_INET6
;
2529 udp_conf
.use_udp6_rx_checksums
=
2530 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2532 udp_conf
.family
= AF_INET
;
2535 udp_conf
.local_udp_port
= port
;
2537 /* Open UDP socket */
2538 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2540 return ERR_PTR(err
);
2545 /* Create new listen socket if needed */
2546 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2549 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2550 struct vxlan_sock
*vs
;
2551 struct socket
*sock
;
2553 bool ipv6
= !!(flags
& VXLAN_F_IPV6
);
2554 struct udp_tunnel_sock_cfg tunnel_cfg
;
2556 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2558 return ERR_PTR(-ENOMEM
);
2560 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2561 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2563 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2565 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2567 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port
),
2570 return ERR_CAST(sock
);
2574 atomic_set(&vs
->refcnt
, 1);
2575 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2577 /* Initialize the vxlan udp offloads structure */
2578 vs
->udp_offloads
.port
= port
;
2579 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2580 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2582 spin_lock(&vn
->sock_lock
);
2583 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2584 vxlan_notify_add_rx_port(vs
);
2585 spin_unlock(&vn
->sock_lock
);
2587 /* Mark socket as an encapsulation socket. */
2588 tunnel_cfg
.sk_user_data
= vs
;
2589 tunnel_cfg
.encap_type
= 1;
2590 tunnel_cfg
.encap_rcv
= vxlan_udp_encap_recv
;
2591 tunnel_cfg
.encap_destroy
= NULL
;
2593 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2598 static struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2599 bool no_share
, u32 flags
)
2601 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2602 struct vxlan_sock
*vs
;
2603 bool ipv6
= flags
& VXLAN_F_IPV6
;
2606 spin_lock(&vn
->sock_lock
);
2607 vs
= vxlan_find_sock(net
, ipv6
? AF_INET6
: AF_INET
, port
,
2610 if (!atomic_add_unless(&vs
->refcnt
, 1, 0))
2611 vs
= ERR_PTR(-EBUSY
);
2612 spin_unlock(&vn
->sock_lock
);
2615 spin_unlock(&vn
->sock_lock
);
2618 return vxlan_socket_create(net
, port
, flags
);
2621 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2622 struct vxlan_config
*conf
)
2624 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2625 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2626 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2628 bool use_ipv6
= false;
2629 __be16 default_port
= vxlan
->cfg
.dst_port
;
2631 vxlan
->net
= src_net
;
2633 dst
->remote_vni
= conf
->vni
;
2635 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2637 /* Unless IPv6 is explicitly requested, assume IPv4 */
2638 if (!dst
->remote_ip
.sa
.sa_family
)
2639 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2641 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2642 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
)
2645 if (conf
->remote_ifindex
) {
2646 struct net_device
*lowerdev
2647 = __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2649 dst
->remote_ifindex
= conf
->remote_ifindex
;
2652 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2656 #if IS_ENABLED(CONFIG_IPV6)
2658 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2659 if (idev
&& idev
->cnf
.disable_ipv6
) {
2660 pr_info("IPv6 is disabled via sysctl\n");
2663 vxlan
->flags
|= VXLAN_F_IPV6
;
2668 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2670 dev
->needed_headroom
= lowerdev
->hard_header_len
+
2671 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2672 } else if (use_ipv6
)
2673 vxlan
->flags
|= VXLAN_F_IPV6
;
2675 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2676 if (!vxlan
->cfg
.dst_port
)
2677 vxlan
->cfg
.dst_port
= default_port
;
2678 vxlan
->flags
|= conf
->flags
;
2680 if (!vxlan
->cfg
.age_interval
)
2681 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2683 if (vxlan_find_vni(src_net
, conf
->vni
, use_ipv6
? AF_INET6
: AF_INET
,
2684 vxlan
->cfg
.dst_port
, vxlan
->flags
))
2687 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2689 /* create an fdb entry for a valid default destination */
2690 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2691 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2692 &vxlan
->default_dst
.remote_ip
,
2693 NUD_REACHABLE
|NUD_PERMANENT
,
2694 NLM_F_EXCL
|NLM_F_CREATE
,
2695 vxlan
->cfg
.dst_port
,
2696 vxlan
->default_dst
.remote_vni
,
2697 vxlan
->default_dst
.remote_ifindex
,
2703 err
= register_netdevice(dev
);
2705 vxlan_fdb_delete_default(vxlan
);
2709 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2714 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
2715 u8 name_assign_type
, struct vxlan_config
*conf
)
2717 struct nlattr
*tb
[IFLA_MAX
+1];
2718 struct net_device
*dev
;
2721 memset(&tb
, 0, sizeof(tb
));
2723 dev
= rtnl_create_link(net
, name
, name_assign_type
,
2724 &vxlan_link_ops
, tb
);
2728 err
= vxlan_dev_configure(net
, dev
, conf
);
2731 return ERR_PTR(err
);
2736 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
2738 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
2739 struct nlattr
*tb
[], struct nlattr
*data
[])
2741 struct vxlan_config conf
;
2744 if (!data
[IFLA_VXLAN_ID
])
2747 memset(&conf
, 0, sizeof(conf
));
2748 conf
.vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2750 if (data
[IFLA_VXLAN_GROUP
]) {
2751 conf
.remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
2752 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2753 if (!IS_ENABLED(CONFIG_IPV6
))
2754 return -EPFNOSUPPORT
;
2756 conf
.remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
2757 conf
.remote_ip
.sa
.sa_family
= AF_INET6
;
2760 if (data
[IFLA_VXLAN_LOCAL
]) {
2761 conf
.saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
2762 conf
.saddr
.sa
.sa_family
= AF_INET
;
2763 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2764 if (!IS_ENABLED(CONFIG_IPV6
))
2765 return -EPFNOSUPPORT
;
2767 /* TODO: respect scope id */
2768 conf
.saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
2769 conf
.saddr
.sa
.sa_family
= AF_INET6
;
2772 if (data
[IFLA_VXLAN_LINK
])
2773 conf
.remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
2775 if (data
[IFLA_VXLAN_TOS
])
2776 conf
.tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2778 if (data
[IFLA_VXLAN_TTL
])
2779 conf
.ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2781 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2782 conf
.flags
|= VXLAN_F_LEARN
;
2784 if (data
[IFLA_VXLAN_AGEING
])
2785 conf
.age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2787 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2788 conf
.flags
|= VXLAN_F_PROXY
;
2790 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2791 conf
.flags
|= VXLAN_F_RSC
;
2793 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2794 conf
.flags
|= VXLAN_F_L2MISS
;
2796 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2797 conf
.flags
|= VXLAN_F_L3MISS
;
2799 if (data
[IFLA_VXLAN_LIMIT
])
2800 conf
.addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2802 if (data
[IFLA_VXLAN_COLLECT_METADATA
] &&
2803 nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
2804 conf
.flags
|= VXLAN_F_COLLECT_METADATA
;
2806 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2807 const struct ifla_vxlan_port_range
*p
2808 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2809 conf
.port_min
= ntohs(p
->low
);
2810 conf
.port_max
= ntohs(p
->high
);
2813 if (data
[IFLA_VXLAN_PORT
])
2814 conf
.dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2816 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2817 conf
.flags
|= VXLAN_F_UDP_CSUM
;
2819 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2820 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2821 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2823 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2824 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2825 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2827 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
2828 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
2829 conf
.flags
|= VXLAN_F_REMCSUM_TX
;
2831 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
2832 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
2833 conf
.flags
|= VXLAN_F_REMCSUM_RX
;
2835 if (data
[IFLA_VXLAN_GBP
])
2836 conf
.flags
|= VXLAN_F_GBP
;
2838 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
])
2839 conf
.flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
2841 err
= vxlan_dev_configure(src_net
, dev
, &conf
);
2844 pr_info("ifindex %d does not exist\n", conf
.remote_ifindex
);
2848 pr_info("IPv6 is disabled via sysctl\n");
2852 pr_info("duplicate VNI %u\n", conf
.vni
);
2859 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2861 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2862 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2864 spin_lock(&vn
->sock_lock
);
2865 if (!hlist_unhashed(&vxlan
->hlist
))
2866 hlist_del_rcu(&vxlan
->hlist
);
2867 spin_unlock(&vn
->sock_lock
);
2869 gro_cells_destroy(&vxlan
->gro_cells
);
2870 list_del(&vxlan
->next
);
2871 unregister_netdevice_queue(dev
, head
);
2874 static size_t vxlan_get_size(const struct net_device
*dev
)
2877 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2878 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2879 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2880 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2881 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2882 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2883 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2884 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2885 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2886 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2887 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2888 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
2889 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2890 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2891 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2892 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2893 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
2894 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2895 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2896 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
2897 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
2901 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2903 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2904 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2905 struct ifla_vxlan_port_range ports
= {
2906 .low
= htons(vxlan
->cfg
.port_min
),
2907 .high
= htons(vxlan
->cfg
.port_max
),
2910 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2911 goto nla_put_failure
;
2913 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2914 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2915 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
2916 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2917 goto nla_put_failure
;
2918 #if IS_ENABLED(CONFIG_IPV6)
2920 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
2921 &dst
->remote_ip
.sin6
.sin6_addr
))
2922 goto nla_put_failure
;
2927 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2928 goto nla_put_failure
;
2930 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
2931 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
2932 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
2933 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
2934 goto nla_put_failure
;
2935 #if IS_ENABLED(CONFIG_IPV6)
2937 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
2938 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
2939 goto nla_put_failure
;
2944 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
2945 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
2946 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2947 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2948 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2949 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2950 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2951 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2952 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2953 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2954 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2955 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
2956 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
2957 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
2958 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
2959 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
2960 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
2961 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
2962 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
2963 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
2964 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
2965 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
2966 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
2967 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
2968 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
2969 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
2970 goto nla_put_failure
;
2972 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2973 goto nla_put_failure
;
2975 if (vxlan
->flags
& VXLAN_F_GBP
&&
2976 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
2977 goto nla_put_failure
;
2979 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
2980 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
2981 goto nla_put_failure
;
2989 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
2991 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2996 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2998 .maxtype
= IFLA_VXLAN_MAX
,
2999 .policy
= vxlan_policy
,
3000 .priv_size
= sizeof(struct vxlan_dev
),
3001 .setup
= vxlan_setup
,
3002 .validate
= vxlan_validate
,
3003 .newlink
= vxlan_newlink
,
3004 .dellink
= vxlan_dellink
,
3005 .get_size
= vxlan_get_size
,
3006 .fill_info
= vxlan_fill_info
,
3007 .get_link_net
= vxlan_get_link_net
,
3010 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3011 struct net_device
*dev
)
3013 struct vxlan_dev
*vxlan
, *next
;
3014 LIST_HEAD(list_kill
);
3016 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3017 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3019 /* In case we created vxlan device with carrier
3020 * and we loose the carrier due to module unload
3021 * we also need to remove vxlan device. In other
3022 * cases, it's not necessary and remote_ifindex
3023 * is 0 here, so no matches.
3025 if (dst
->remote_ifindex
== dev
->ifindex
)
3026 vxlan_dellink(vxlan
->dev
, &list_kill
);
3029 unregister_netdevice_many(&list_kill
);
3032 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
3033 unsigned long event
, void *ptr
)
3035 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3036 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3038 if (event
== NETDEV_UNREGISTER
)
3039 vxlan_handle_lowerdev_unregister(vn
, dev
);
3044 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3045 .notifier_call
= vxlan_lowerdev_event
,
3048 static __net_init
int vxlan_init_net(struct net
*net
)
3050 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3053 INIT_LIST_HEAD(&vn
->vxlan_list
);
3054 spin_lock_init(&vn
->sock_lock
);
3056 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3057 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3062 static void __net_exit
vxlan_exit_net(struct net
*net
)
3064 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3065 struct vxlan_dev
*vxlan
, *next
;
3066 struct net_device
*dev
, *aux
;
3070 for_each_netdev_safe(net
, dev
, aux
)
3071 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3072 unregister_netdevice_queue(dev
, &list
);
3074 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3075 /* If vxlan->dev is in the same netns, it has already been added
3076 * to the list by the previous loop.
3078 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3079 gro_cells_destroy(&vxlan
->gro_cells
);
3080 unregister_netdevice_queue(vxlan
->dev
, &list
);
3084 unregister_netdevice_many(&list
);
3088 static struct pernet_operations vxlan_net_ops
= {
3089 .init
= vxlan_init_net
,
3090 .exit
= vxlan_exit_net
,
3091 .id
= &vxlan_net_id
,
3092 .size
= sizeof(struct vxlan_net
),
3095 static int __init
vxlan_init_module(void)
3099 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3103 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3105 rc
= register_pernet_subsys(&vxlan_net_ops
);
3109 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3113 rc
= rtnl_link_register(&vxlan_link_ops
);
3119 unregister_netdevice_notifier(&vxlan_notifier_block
);
3121 unregister_pernet_subsys(&vxlan_net_ops
);
3123 destroy_workqueue(vxlan_wq
);
3126 late_initcall(vxlan_init_module
);
3128 static void __exit
vxlan_cleanup_module(void)
3130 rtnl_link_unregister(&vxlan_link_ops
);
3131 unregister_netdevice_notifier(&vxlan_notifier_block
);
3132 destroy_workqueue(vxlan_wq
);
3133 unregister_pernet_subsys(&vxlan_net_ops
);
3134 /* rcu_barrier() is called by netns */
3136 module_exit(vxlan_cleanup_module
);
3138 MODULE_LICENSE("GPL");
3139 MODULE_VERSION(VXLAN_VERSION
);
3140 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3141 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3142 MODULE_ALIAS_RTNL_LINK("vxlan");