2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/kernel.h>
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
31 #include <net/ndisc.h>
33 #include <net/ip_tunnels.h>
36 #include <net/udp_tunnel.h>
37 #include <net/rtnetlink.h>
38 #include <net/route.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/vxlan.h>
44 #include <net/protocol.h>
45 #include <net/udp_tunnel.h>
46 #if IS_ENABLED(CONFIG_IPV6)
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
53 #define VXLAN_VERSION "0.1"
55 #define PORT_HASH_BITS 8
56 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
57 #define VNI_HASH_BITS 10
58 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
59 #define FDB_HASH_BITS 8
60 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
61 #define FDB_AGE_DEFAULT 300 /* 5 min */
62 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
64 /* UDP port for VXLAN traffic.
65 * The IANA assigned port is 4789, but the Linux default is 8472
66 * for compatibility with early adopters.
68 static unsigned short vxlan_port __read_mostly
= 8472;
69 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
70 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
72 static bool log_ecn_error
= true;
73 module_param(log_ecn_error
, bool, 0644);
74 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
76 static int vxlan_net_id
;
78 static const u8 all_zeros_mac
[ETH_ALEN
];
80 /* per-network namespace private data for this module */
82 struct list_head vxlan_list
;
83 struct hlist_head sock_list
[PORT_HASH_SIZE
];
88 struct sockaddr_in sin
;
89 struct sockaddr_in6 sin6
;
94 union vxlan_addr remote_ip
;
98 struct list_head list
;
102 /* Forwarding table entry */
104 struct hlist_node hlist
; /* linked list of entries */
106 unsigned long updated
; /* jiffies */
108 struct list_head remotes
;
109 u16 state
; /* see ndm_state */
110 u8 flags
; /* see ndm_flags */
111 u8 eth_addr
[ETH_ALEN
];
114 /* Pseudo network device */
116 struct hlist_node hlist
; /* vni hash table */
117 struct list_head next
; /* vxlan's per namespace list */
118 struct vxlan_sock
*vn_sock
; /* listening socket */
119 struct net_device
*dev
;
120 struct net
*net
; /* netns for packet i/o */
121 struct vxlan_rdst default_dst
; /* default destination */
122 union vxlan_addr saddr
; /* source address */
124 __u16 port_min
; /* source port range */
126 __u8 tos
; /* TOS override */
128 u32 flags
; /* VXLAN_F_* in vxlan.h */
130 struct work_struct sock_work
;
131 struct work_struct igmp_join
;
132 struct work_struct igmp_leave
;
134 unsigned long age_interval
;
135 struct timer_list age_timer
;
136 spinlock_t hash_lock
;
137 unsigned int addrcnt
;
138 unsigned int addrmax
;
140 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
143 /* salt for hash table */
144 static u32 vxlan_salt __read_mostly
;
145 static struct workqueue_struct
*vxlan_wq
;
147 static void vxlan_sock_work(struct work_struct
*work
);
149 #if IS_ENABLED(CONFIG_IPV6)
151 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
153 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
155 if (a
->sa
.sa_family
== AF_INET6
)
156 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
158 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
161 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
163 if (ipa
->sa
.sa_family
== AF_INET6
)
164 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
166 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
169 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
171 if (ipa
->sa
.sa_family
== AF_INET6
)
172 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
174 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
177 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
179 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
180 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
181 ip
->sa
.sa_family
= AF_INET6
;
183 } else if (nla_len(nla
) >= sizeof(__be32
)) {
184 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
185 ip
->sa
.sa_family
= AF_INET
;
188 return -EAFNOSUPPORT
;
192 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
193 const union vxlan_addr
*ip
)
195 if (ip
->sa
.sa_family
== AF_INET6
)
196 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
198 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
201 #else /* !CONFIG_IPV6 */
204 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
206 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
209 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
211 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
214 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
216 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
219 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
221 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
222 return -EAFNOSUPPORT
;
223 } else if (nla_len(nla
) >= sizeof(__be32
)) {
224 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
225 ip
->sa
.sa_family
= AF_INET
;
228 return -EAFNOSUPPORT
;
232 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
233 const union vxlan_addr
*ip
)
235 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
239 /* Virtual Network hash table head */
240 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
242 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
245 /* Socket hash table head */
246 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
248 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
250 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
253 /* First remote destination for a forwarding entry.
254 * Guaranteed to be non-NULL because remotes are never deleted.
256 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
258 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
261 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
263 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
266 /* Find VXLAN socket based on network namespace, address family and UDP port
267 * and enabled unshareable flags.
269 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
270 __be16 port
, u32 flags
)
272 struct vxlan_sock
*vs
;
274 flags
&= VXLAN_F_RCV_FLAGS
;
276 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
277 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
278 inet_sk(vs
->sock
->sk
)->sk
.sk_family
== family
&&
285 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
287 struct vxlan_dev
*vxlan
;
289 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
290 if (vxlan
->default_dst
.remote_vni
== id
)
297 /* Look up VNI in a per net namespace table */
298 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
,
299 sa_family_t family
, __be16 port
,
302 struct vxlan_sock
*vs
;
304 vs
= vxlan_find_sock(net
, family
, port
, flags
);
308 return vxlan_vs_find_vni(vs
, id
);
311 /* Fill in neighbour message in skbuff. */
312 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
313 const struct vxlan_fdb
*fdb
,
314 u32 portid
, u32 seq
, int type
, unsigned int flags
,
315 const struct vxlan_rdst
*rdst
)
317 unsigned long now
= jiffies
;
318 struct nda_cacheinfo ci
;
319 struct nlmsghdr
*nlh
;
321 bool send_ip
, send_eth
;
323 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
327 ndm
= nlmsg_data(nlh
);
328 memset(ndm
, 0, sizeof(*ndm
));
330 send_eth
= send_ip
= true;
332 if (type
== RTM_GETNEIGH
) {
333 ndm
->ndm_family
= AF_INET
;
334 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
335 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
337 ndm
->ndm_family
= AF_BRIDGE
;
338 ndm
->ndm_state
= fdb
->state
;
339 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
340 ndm
->ndm_flags
= fdb
->flags
;
341 ndm
->ndm_type
= RTN_UNICAST
;
343 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
344 nla_put_s32(skb
, NDA_NDM_IFINDEX_NETNSID
,
345 peernet2id(vxlan
->net
, dev_net(vxlan
->dev
))))
346 goto nla_put_failure
;
348 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
349 goto nla_put_failure
;
351 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
352 goto nla_put_failure
;
354 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
355 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
356 goto nla_put_failure
;
357 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
358 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
359 goto nla_put_failure
;
360 if (rdst
->remote_ifindex
&&
361 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
362 goto nla_put_failure
;
364 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
365 ci
.ndm_confirmed
= 0;
366 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
369 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
370 goto nla_put_failure
;
376 nlmsg_cancel(skb
, nlh
);
380 static inline size_t vxlan_nlmsg_size(void)
382 return NLMSG_ALIGN(sizeof(struct ndmsg
))
383 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
384 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
385 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
386 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
387 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
388 + nla_total_size(sizeof(struct nda_cacheinfo
));
391 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
392 struct vxlan_rdst
*rd
, int type
)
394 struct net
*net
= dev_net(vxlan
->dev
);
398 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
402 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
404 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
405 WARN_ON(err
== -EMSGSIZE
);
410 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
414 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
417 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
419 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
420 struct vxlan_fdb f
= {
423 struct vxlan_rdst remote
= {
424 .remote_ip
= *ipa
, /* goes to NDA_DST */
425 .remote_vni
= VXLAN_N_VID
,
428 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
431 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
433 struct vxlan_fdb f
= {
436 struct vxlan_rdst remote
= { };
438 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
440 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
443 /* Hash Ethernet address */
444 static u32
eth_hash(const unsigned char *addr
)
446 u64 value
= get_unaligned((u64
*)addr
);
448 /* only want 6 bytes */
454 return hash_64(value
, FDB_HASH_BITS
);
457 /* Hash chain to use given mac address */
458 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
461 return &vxlan
->fdb_head
[eth_hash(mac
)];
464 /* Look up Ethernet address in forwarding table */
465 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
468 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
471 hlist_for_each_entry_rcu(f
, head
, hlist
) {
472 if (ether_addr_equal(mac
, f
->eth_addr
))
479 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
484 f
= __vxlan_find_mac(vxlan
, mac
);
491 /* caller should hold vxlan->hash_lock */
492 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
493 union vxlan_addr
*ip
, __be16 port
,
494 __u32 vni
, __u32 ifindex
)
496 struct vxlan_rdst
*rd
;
498 list_for_each_entry(rd
, &f
->remotes
, list
) {
499 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
500 rd
->remote_port
== port
&&
501 rd
->remote_vni
== vni
&&
502 rd
->remote_ifindex
== ifindex
)
509 /* Replace destination of unicast mac */
510 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
511 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
513 struct vxlan_rdst
*rd
;
515 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
519 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
523 rd
->remote_port
= port
;
524 rd
->remote_vni
= vni
;
525 rd
->remote_ifindex
= ifindex
;
529 /* Add/update destinations for multicast */
530 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
531 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
532 __u32 ifindex
, struct vxlan_rdst
**rdp
)
534 struct vxlan_rdst
*rd
;
536 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
540 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
544 rd
->remote_port
= port
;
545 rd
->remote_vni
= vni
;
546 rd
->remote_ifindex
= ifindex
;
548 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
554 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
556 struct vxlanhdr
*vh
, size_t hdrlen
,
559 size_t start
, offset
, plen
;
562 if (skb
->remcsum_offload
)
565 if (!NAPI_GRO_CB(skb
)->csum_valid
)
568 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
569 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
570 offsetof(struct udphdr
, check
) :
571 offsetof(struct tcphdr
, check
));
573 plen
= hdrlen
+ offset
+ sizeof(u16
);
575 /* Pull checksum that will be written */
576 if (skb_gro_header_hard(skb
, off
+ plen
)) {
577 vh
= skb_gro_header_slow(skb
, off
+ plen
, off
);
582 delta
= remcsum_adjust((void *)vh
+ hdrlen
,
583 NAPI_GRO_CB(skb
)->csum
, start
, offset
);
585 /* Adjust skb->csum since we changed the packet */
586 skb
->csum
= csum_add(skb
->csum
, delta
);
587 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
589 skb
->remcsum_offload
= 1;
594 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
,
596 struct udp_offload
*uoff
)
598 struct sk_buff
*p
, **pp
= NULL
;
599 struct vxlanhdr
*vh
, *vh2
;
600 unsigned int hlen
, off_vx
;
602 struct vxlan_sock
*vs
= container_of(uoff
, struct vxlan_sock
,
606 off_vx
= skb_gro_offset(skb
);
607 hlen
= off_vx
+ sizeof(*vh
);
608 vh
= skb_gro_header_fast(skb
, off_vx
);
609 if (skb_gro_header_hard(skb
, hlen
)) {
610 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
615 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
616 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
618 flags
= ntohl(vh
->vx_flags
);
620 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
621 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
630 for (p
= *head
; p
; p
= p
->next
) {
631 if (!NAPI_GRO_CB(p
)->same_flow
)
634 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
635 if (vh
->vx_flags
!= vh2
->vx_flags
||
636 vh
->vx_vni
!= vh2
->vx_vni
) {
637 NAPI_GRO_CB(p
)->same_flow
= 0;
642 pp
= eth_gro_receive(head
, skb
);
645 NAPI_GRO_CB(skb
)->flush
|= flush
;
650 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
651 struct udp_offload
*uoff
)
653 udp_tunnel_gro_complete(skb
, nhoff
);
655 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
658 /* Notify netdevs that UDP port started listening */
659 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
661 struct net_device
*dev
;
662 struct sock
*sk
= vs
->sock
->sk
;
663 struct net
*net
= sock_net(sk
);
664 sa_family_t sa_family
= sk
->sk_family
;
665 __be16 port
= inet_sk(sk
)->inet_sport
;
668 if (sa_family
== AF_INET
) {
669 err
= udp_add_offload(&vs
->udp_offloads
);
671 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
675 for_each_netdev_rcu(net
, dev
) {
676 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
677 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
683 /* Notify netdevs that UDP port is no more listening */
684 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
686 struct net_device
*dev
;
687 struct sock
*sk
= vs
->sock
->sk
;
688 struct net
*net
= sock_net(sk
);
689 sa_family_t sa_family
= sk
->sk_family
;
690 __be16 port
= inet_sk(sk
)->inet_sport
;
693 for_each_netdev_rcu(net
, dev
) {
694 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
695 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
700 if (sa_family
== AF_INET
)
701 udp_del_offload(&vs
->udp_offloads
);
704 /* Add new entry to forwarding table -- assumes lock held */
705 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
706 const u8
*mac
, union vxlan_addr
*ip
,
707 __u16 state
, __u16 flags
,
708 __be16 port
, __u32 vni
, __u32 ifindex
,
711 struct vxlan_rdst
*rd
= NULL
;
715 f
= __vxlan_find_mac(vxlan
, mac
);
717 if (flags
& NLM_F_EXCL
) {
718 netdev_dbg(vxlan
->dev
,
719 "lost race to create %pM\n", mac
);
722 if (f
->state
!= state
) {
724 f
->updated
= jiffies
;
727 if (f
->flags
!= ndm_flags
) {
728 f
->flags
= ndm_flags
;
729 f
->updated
= jiffies
;
732 if ((flags
& NLM_F_REPLACE
)) {
733 /* Only change unicasts */
734 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
735 is_zero_ether_addr(f
->eth_addr
))) {
736 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
745 if ((flags
& NLM_F_APPEND
) &&
746 (is_multicast_ether_addr(f
->eth_addr
) ||
747 is_zero_ether_addr(f
->eth_addr
))) {
748 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
756 if (!(flags
& NLM_F_CREATE
))
759 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
762 /* Disallow replace to add a multicast entry */
763 if ((flags
& NLM_F_REPLACE
) &&
764 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
767 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
768 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
774 f
->flags
= ndm_flags
;
775 f
->updated
= f
->used
= jiffies
;
776 INIT_LIST_HEAD(&f
->remotes
);
777 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
779 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
782 hlist_add_head_rcu(&f
->hlist
,
783 vxlan_fdb_head(vxlan
, mac
));
788 rd
= first_remote_rtnl(f
);
789 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
795 static void vxlan_fdb_free(struct rcu_head
*head
)
797 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
798 struct vxlan_rdst
*rd
, *nd
;
800 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
805 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
807 netdev_dbg(vxlan
->dev
,
808 "delete %pM\n", f
->eth_addr
);
811 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
813 hlist_del_rcu(&f
->hlist
);
814 call_rcu(&f
->rcu
, vxlan_fdb_free
);
817 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
818 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
820 struct net
*net
= dev_net(vxlan
->dev
);
824 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
828 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
829 if (remote
->sa
.sa_family
== AF_INET
) {
830 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
831 ip
->sa
.sa_family
= AF_INET
;
832 #if IS_ENABLED(CONFIG_IPV6)
834 ip
->sin6
.sin6_addr
= in6addr_any
;
835 ip
->sa
.sa_family
= AF_INET6
;
841 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
843 *port
= nla_get_be16(tb
[NDA_PORT
]);
845 *port
= vxlan
->dst_port
;
849 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
851 *vni
= nla_get_u32(tb
[NDA_VNI
]);
853 *vni
= vxlan
->default_dst
.remote_vni
;
856 if (tb
[NDA_IFINDEX
]) {
857 struct net_device
*tdev
;
859 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
861 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
862 tdev
= __dev_get_by_index(net
, *ifindex
);
864 return -EADDRNOTAVAIL
;
872 /* Add static entry (via netlink) */
873 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
874 struct net_device
*dev
,
875 const unsigned char *addr
, u16 vid
, u16 flags
)
877 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
878 /* struct net *net = dev_net(vxlan->dev); */
884 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
885 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
890 if (tb
[NDA_DST
] == NULL
)
893 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
897 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
898 return -EAFNOSUPPORT
;
900 spin_lock_bh(&vxlan
->hash_lock
);
901 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
902 port
, vni
, ifindex
, ndm
->ndm_flags
);
903 spin_unlock_bh(&vxlan
->hash_lock
);
908 /* Delete entry (via netlink) */
909 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
910 struct net_device
*dev
,
911 const unsigned char *addr
, u16 vid
)
913 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
915 struct vxlan_rdst
*rd
= NULL
;
921 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
927 spin_lock_bh(&vxlan
->hash_lock
);
928 f
= vxlan_find_mac(vxlan
, addr
);
932 if (!vxlan_addr_any(&ip
)) {
933 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
940 /* remove a destination if it's not the only one on the list,
941 * otherwise destroy the fdb entry
943 if (rd
&& !list_is_singular(&f
->remotes
)) {
944 list_del_rcu(&rd
->list
);
945 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
950 vxlan_fdb_destroy(vxlan
, f
);
953 spin_unlock_bh(&vxlan
->hash_lock
);
958 /* Dump forwarding table */
959 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
960 struct net_device
*dev
,
961 struct net_device
*filter_dev
, int idx
)
963 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
966 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
970 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
971 struct vxlan_rdst
*rd
;
973 if (idx
< cb
->args
[0])
976 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
977 err
= vxlan_fdb_info(skb
, vxlan
, f
,
978 NETLINK_CB(cb
->skb
).portid
,
993 /* Watch incoming packets to learn mapping between Ethernet address
994 * and Tunnel endpoint.
995 * Return true if packet is bogus and should be droppped.
997 static bool vxlan_snoop(struct net_device
*dev
,
998 union vxlan_addr
*src_ip
, const u8
*src_mac
)
1000 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1001 struct vxlan_fdb
*f
;
1003 f
= vxlan_find_mac(vxlan
, src_mac
);
1005 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
1007 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
1010 /* Don't migrate static entries, drop packets */
1011 if (f
->state
& NUD_NOARP
)
1014 if (net_ratelimit())
1016 "%pM migrated from %pIS to %pIS\n",
1017 src_mac
, &rdst
->remote_ip
, &src_ip
);
1019 rdst
->remote_ip
= *src_ip
;
1020 f
->updated
= jiffies
;
1021 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1023 /* learned new entry */
1024 spin_lock(&vxlan
->hash_lock
);
1026 /* close off race between vxlan_flush and incoming packets */
1027 if (netif_running(dev
))
1028 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1030 NLM_F_EXCL
|NLM_F_CREATE
,
1032 vxlan
->default_dst
.remote_vni
,
1034 spin_unlock(&vxlan
->hash_lock
);
1040 /* See if multicast group is already in use by other ID */
1041 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1043 struct vxlan_dev
*vxlan
;
1045 /* The vxlan_sock is only used by dev, leaving group has
1046 * no effect on other vxlan devices.
1048 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1051 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1052 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1055 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1058 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1059 &dev
->default_dst
.remote_ip
))
1062 if (vxlan
->default_dst
.remote_ifindex
!=
1063 dev
->default_dst
.remote_ifindex
)
1072 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1074 atomic_inc(&vs
->refcnt
);
1077 void vxlan_sock_release(struct vxlan_sock
*vs
)
1079 struct sock
*sk
= vs
->sock
->sk
;
1080 struct net
*net
= sock_net(sk
);
1081 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1083 if (!atomic_dec_and_test(&vs
->refcnt
))
1086 spin_lock(&vn
->sock_lock
);
1087 hlist_del_rcu(&vs
->hlist
);
1088 vxlan_notify_del_rx_port(vs
);
1089 spin_unlock(&vn
->sock_lock
);
1091 queue_work(vxlan_wq
, &vs
->del_work
);
1093 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1095 /* Callback to update multicast group membership when first VNI on
1096 * multicast asddress is brought up
1097 * Done as workqueue because ip_mc_join_group acquires RTNL.
1099 static void vxlan_igmp_join(struct work_struct
*work
)
1101 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1102 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1103 struct sock
*sk
= vs
->sock
->sk
;
1104 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1105 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1108 if (ip
->sa
.sa_family
== AF_INET
) {
1109 struct ip_mreqn mreq
= {
1110 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1111 .imr_ifindex
= ifindex
,
1114 ip_mc_join_group(sk
, &mreq
);
1115 #if IS_ENABLED(CONFIG_IPV6)
1117 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1118 &ip
->sin6
.sin6_addr
);
1123 vxlan_sock_release(vs
);
1124 dev_put(vxlan
->dev
);
1127 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1128 static void vxlan_igmp_leave(struct work_struct
*work
)
1130 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1131 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1132 struct sock
*sk
= vs
->sock
->sk
;
1133 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1134 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1137 if (ip
->sa
.sa_family
== AF_INET
) {
1138 struct ip_mreqn mreq
= {
1139 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1140 .imr_ifindex
= ifindex
,
1143 ip_mc_leave_group(sk
, &mreq
);
1144 #if IS_ENABLED(CONFIG_IPV6)
1146 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1147 &ip
->sin6
.sin6_addr
);
1153 vxlan_sock_release(vs
);
1154 dev_put(vxlan
->dev
);
1157 static struct vxlanhdr
*vxlan_remcsum(struct sk_buff
*skb
, struct vxlanhdr
*vh
,
1158 size_t hdrlen
, u32 data
)
1160 size_t start
, offset
, plen
;
1163 if (skb
->remcsum_offload
) {
1164 /* Already processed in GRO path */
1165 skb
->remcsum_offload
= 0;
1169 start
= (data
& VXLAN_RCO_MASK
) << VXLAN_RCO_SHIFT
;
1170 offset
= start
+ ((data
& VXLAN_RCO_UDP
) ?
1171 offsetof(struct udphdr
, check
) :
1172 offsetof(struct tcphdr
, check
));
1174 plen
= hdrlen
+ offset
+ sizeof(u16
);
1176 if (!pskb_may_pull(skb
, plen
))
1179 vh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1181 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
))
1182 __skb_checksum_complete(skb
);
1184 delta
= remcsum_adjust((void *)vh
+ hdrlen
,
1185 skb
->csum
, start
, offset
);
1187 /* Adjust skb->csum since we changed the packet */
1188 skb
->csum
= csum_add(skb
->csum
, delta
);
1193 /* Callback from net/ipv4/udp.c to receive packets */
1194 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1196 struct vxlan_sock
*vs
;
1197 struct vxlanhdr
*vxh
;
1199 struct vxlan_metadata md
= {0};
1201 /* Need Vxlan and inner Ethernet header to be present */
1202 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1205 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1206 flags
= ntohl(vxh
->vx_flags
);
1207 vni
= ntohl(vxh
->vx_vni
);
1209 if (flags
& VXLAN_HF_VNI
) {
1210 flags
&= ~VXLAN_HF_VNI
;
1212 /* VNI flag always required to be set */
1216 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1218 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1220 vs
= rcu_dereference_sk_user_data(sk
);
1224 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
1225 vxh
= vxlan_remcsum(skb
, vxh
, sizeof(struct vxlanhdr
), vni
);
1229 flags
&= ~VXLAN_HF_RCO
;
1230 vni
&= VXLAN_VID_MASK
;
1233 /* For backwards compatibility, only allow reserved fields to be
1234 * used by VXLAN extensions if explicitly requested.
1236 if ((flags
& VXLAN_HF_GBP
) && (vs
->flags
& VXLAN_F_GBP
)) {
1237 struct vxlanhdr_gbp
*gbp
;
1239 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1240 md
.gbp
= ntohs(gbp
->policy_id
);
1242 if (gbp
->dont_learn
)
1243 md
.gbp
|= VXLAN_GBP_DONT_LEARN
;
1245 if (gbp
->policy_applied
)
1246 md
.gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1248 flags
&= ~VXLAN_GBP_USED_BITS
;
1251 if (flags
|| (vni
& ~VXLAN_VID_MASK
)) {
1252 /* If there are any unprocessed flags remaining treat
1253 * this as a malformed packet. This behavior diverges from
1254 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1255 * in reserved fields are to be ignored. The approach here
1256 * maintains compatbility with previous stack code, and also
1257 * is more robust and provides a little more security in
1258 * adding extensions to VXLAN.
1264 md
.vni
= vxh
->vx_vni
;
1265 vs
->rcv(vs
, skb
, &md
);
1269 /* Consume bad packet */
1274 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1275 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1278 /* Return non vxlan pkt */
1282 static void vxlan_rcv(struct vxlan_sock
*vs
, struct sk_buff
*skb
,
1283 struct vxlan_metadata
*md
)
1285 struct iphdr
*oip
= NULL
;
1286 struct ipv6hdr
*oip6
= NULL
;
1287 struct vxlan_dev
*vxlan
;
1288 struct pcpu_sw_netstats
*stats
;
1289 union vxlan_addr saddr
;
1292 union vxlan_addr
*remote_ip
;
1294 vni
= ntohl(md
->vni
) >> 8;
1295 /* Is this VNI defined? */
1296 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1300 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1301 skb_reset_mac_header(skb
);
1302 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1303 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1304 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1306 /* Ignore packet loops (and multicast echo) */
1307 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1310 /* Re-examine inner Ethernet packet */
1311 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1313 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1314 saddr
.sa
.sa_family
= AF_INET
;
1315 #if IS_ENABLED(CONFIG_IPV6)
1317 oip6
= ipv6_hdr(skb
);
1318 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1319 saddr
.sa
.sa_family
= AF_INET6
;
1323 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1324 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1327 skb_reset_network_header(skb
);
1328 skb
->mark
= md
->gbp
;
1331 err
= IP6_ECN_decapsulate(oip6
, skb
);
1333 err
= IP_ECN_decapsulate(oip
, skb
);
1335 if (unlikely(err
)) {
1336 if (log_ecn_error
) {
1338 net_info_ratelimited("non-ECT from %pI6\n",
1341 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1342 &oip
->saddr
, oip
->tos
);
1345 ++vxlan
->dev
->stats
.rx_frame_errors
;
1346 ++vxlan
->dev
->stats
.rx_errors
;
1351 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1352 u64_stats_update_begin(&stats
->syncp
);
1353 stats
->rx_packets
++;
1354 stats
->rx_bytes
+= skb
->len
;
1355 u64_stats_update_end(&stats
->syncp
);
1361 /* Consume bad packet */
1365 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1367 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1368 struct arphdr
*parp
;
1371 struct neighbour
*n
;
1373 if (dev
->flags
& IFF_NOARP
)
1376 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1377 dev
->stats
.tx_dropped
++;
1380 parp
= arp_hdr(skb
);
1382 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1383 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1384 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1385 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1386 parp
->ar_hln
!= dev
->addr_len
||
1389 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1391 arpptr
+= dev
->addr_len
; /* sha */
1392 memcpy(&sip
, arpptr
, sizeof(sip
));
1393 arpptr
+= sizeof(sip
);
1394 arpptr
+= dev
->addr_len
; /* tha */
1395 memcpy(&tip
, arpptr
, sizeof(tip
));
1397 if (ipv4_is_loopback(tip
) ||
1398 ipv4_is_multicast(tip
))
1401 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1404 struct vxlan_fdb
*f
;
1405 struct sk_buff
*reply
;
1407 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1412 f
= vxlan_find_mac(vxlan
, n
->ha
);
1413 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1414 /* bridge-local neighbor */
1419 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1427 skb_reset_mac_header(reply
);
1428 __skb_pull(reply
, skb_network_offset(reply
));
1429 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1430 reply
->pkt_type
= PACKET_HOST
;
1432 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1433 dev
->stats
.rx_dropped
++;
1434 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1435 union vxlan_addr ipa
= {
1436 .sin
.sin_addr
.s_addr
= tip
,
1437 .sin
.sin_family
= AF_INET
,
1440 vxlan_ip_miss(dev
, &ipa
);
1444 return NETDEV_TX_OK
;
1447 #if IS_ENABLED(CONFIG_IPV6)
1448 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1449 struct neighbour
*n
, bool isrouter
)
1451 struct net_device
*dev
= request
->dev
;
1452 struct sk_buff
*reply
;
1453 struct nd_msg
*ns
, *na
;
1454 struct ipv6hdr
*pip6
;
1456 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1463 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1464 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1465 reply
= alloc_skb(len
, GFP_ATOMIC
);
1469 reply
->protocol
= htons(ETH_P_IPV6
);
1471 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1472 skb_push(reply
, sizeof(struct ethhdr
));
1473 skb_set_mac_header(reply
, 0);
1475 ns
= (struct nd_msg
*)skb_transport_header(request
);
1477 daddr
= eth_hdr(request
)->h_source
;
1478 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1479 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1480 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1481 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1486 /* Ethernet header */
1487 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1488 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1489 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1490 reply
->protocol
= htons(ETH_P_IPV6
);
1492 skb_pull(reply
, sizeof(struct ethhdr
));
1493 skb_set_network_header(reply
, 0);
1494 skb_put(reply
, sizeof(struct ipv6hdr
));
1498 pip6
= ipv6_hdr(reply
);
1499 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1501 pip6
->priority
= ipv6_hdr(request
)->priority
;
1502 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1503 pip6
->hop_limit
= 255;
1504 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1505 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1507 skb_pull(reply
, sizeof(struct ipv6hdr
));
1508 skb_set_transport_header(reply
, 0);
1510 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1512 /* Neighbor Advertisement */
1513 memset(na
, 0, sizeof(*na
)+na_olen
);
1514 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1515 na
->icmph
.icmp6_router
= isrouter
;
1516 na
->icmph
.icmp6_override
= 1;
1517 na
->icmph
.icmp6_solicited
= 1;
1518 na
->target
= ns
->target
;
1519 ether_addr_copy(&na
->opt
[2], n
->ha
);
1520 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1521 na
->opt
[1] = na_olen
>> 3;
1523 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1524 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1525 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1527 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1529 skb_push(reply
, sizeof(struct ipv6hdr
));
1531 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1536 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1538 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1540 const struct ipv6hdr
*iphdr
;
1541 const struct in6_addr
*saddr
, *daddr
;
1542 struct neighbour
*n
;
1543 struct inet6_dev
*in6_dev
;
1545 in6_dev
= __in6_dev_get(dev
);
1549 iphdr
= ipv6_hdr(skb
);
1550 saddr
= &iphdr
->saddr
;
1551 daddr
= &iphdr
->daddr
;
1553 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1554 if (msg
->icmph
.icmp6_code
!= 0 ||
1555 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1558 if (ipv6_addr_loopback(daddr
) ||
1559 ipv6_addr_is_multicast(&msg
->target
))
1562 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1565 struct vxlan_fdb
*f
;
1566 struct sk_buff
*reply
;
1568 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1573 f
= vxlan_find_mac(vxlan
, n
->ha
);
1574 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1575 /* bridge-local neighbor */
1580 reply
= vxlan_na_create(skb
, n
,
1581 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1588 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1589 dev
->stats
.rx_dropped
++;
1591 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1592 union vxlan_addr ipa
= {
1593 .sin6
.sin6_addr
= msg
->target
,
1594 .sin6
.sin6_family
= AF_INET6
,
1597 vxlan_ip_miss(dev
, &ipa
);
1602 return NETDEV_TX_OK
;
1606 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1608 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1609 struct neighbour
*n
;
1611 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1615 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1620 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1623 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1624 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1625 union vxlan_addr ipa
= {
1626 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1627 .sin
.sin_family
= AF_INET
,
1630 vxlan_ip_miss(dev
, &ipa
);
1636 #if IS_ENABLED(CONFIG_IPV6)
1639 struct ipv6hdr
*pip6
;
1641 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1643 pip6
= ipv6_hdr(skb
);
1644 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1645 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1646 union vxlan_addr ipa
= {
1647 .sin6
.sin6_addr
= pip6
->daddr
,
1648 .sin6
.sin6_family
= AF_INET6
,
1651 vxlan_ip_miss(dev
, &ipa
);
1665 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1667 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1669 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1678 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1679 struct vxlan_metadata
*md
)
1681 struct vxlanhdr_gbp
*gbp
;
1683 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1684 vxh
->vx_flags
|= htonl(VXLAN_HF_GBP
);
1686 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1687 gbp
->dont_learn
= 1;
1689 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1690 gbp
->policy_applied
= 1;
1692 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1695 #if IS_ENABLED(CONFIG_IPV6)
1696 static int vxlan6_xmit_skb(struct dst_entry
*dst
, struct sk_buff
*skb
,
1697 struct net_device
*dev
, struct in6_addr
*saddr
,
1698 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1699 __be16 src_port
, __be16 dst_port
,
1700 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1702 struct vxlanhdr
*vxh
;
1705 bool udp_sum
= !(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
1706 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1707 u16 hdrlen
= sizeof(struct vxlanhdr
);
1709 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1710 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1711 int csum_start
= skb_checksum_start_offset(skb
);
1713 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1714 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1715 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1716 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1718 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1722 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1728 skb_scrub_packet(skb
, xnet
);
1730 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1731 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1732 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1734 /* Need space for new headers (invalidates iph ptr) */
1735 err
= skb_cow_head(skb
, min_headroom
);
1736 if (unlikely(err
)) {
1741 skb
= vlan_hwaccel_push_inside(skb
);
1742 if (WARN_ON(!skb
)) {
1747 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1748 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1749 vxh
->vx_vni
= md
->vni
;
1751 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1752 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1755 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1756 data
|= VXLAN_RCO_UDP
;
1758 vxh
->vx_vni
|= htonl(data
);
1759 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1761 if (!skb_is_gso(skb
)) {
1762 skb
->ip_summed
= CHECKSUM_NONE
;
1763 skb
->encapsulation
= 0;
1767 if (vxflags
& VXLAN_F_GBP
)
1768 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1770 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1772 udp_tunnel6_xmit_skb(dst
, skb
, dev
, saddr
, daddr
, prio
,
1773 ttl
, src_port
, dst_port
,
1774 !!(vxflags
& VXLAN_F_UDP_ZERO_CSUM6_TX
));
1782 int vxlan_xmit_skb(struct rtable
*rt
, struct sk_buff
*skb
,
1783 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1784 __be16 src_port
, __be16 dst_port
,
1785 struct vxlan_metadata
*md
, bool xnet
, u32 vxflags
)
1787 struct vxlanhdr
*vxh
;
1790 bool udp_sum
= !!(vxflags
& VXLAN_F_UDP_CSUM
);
1791 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1792 u16 hdrlen
= sizeof(struct vxlanhdr
);
1794 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1795 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1796 int csum_start
= skb_checksum_start_offset(skb
);
1798 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1799 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1800 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1801 skb
->csum_offset
== offsetof(struct tcphdr
, check
))) {
1803 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1807 skb
= iptunnel_handle_offloads(skb
, udp_sum
, type
);
1809 return PTR_ERR(skb
);
1811 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1812 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1813 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1815 /* Need space for new headers (invalidates iph ptr) */
1816 err
= skb_cow_head(skb
, min_headroom
);
1817 if (unlikely(err
)) {
1822 skb
= vlan_hwaccel_push_inside(skb
);
1826 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1827 vxh
->vx_flags
= htonl(VXLAN_HF_VNI
);
1828 vxh
->vx_vni
= md
->vni
;
1830 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1831 u32 data
= (skb_checksum_start_offset(skb
) - hdrlen
) >>
1834 if (skb
->csum_offset
== offsetof(struct udphdr
, check
))
1835 data
|= VXLAN_RCO_UDP
;
1837 vxh
->vx_vni
|= htonl(data
);
1838 vxh
->vx_flags
|= htonl(VXLAN_HF_RCO
);
1840 if (!skb_is_gso(skb
)) {
1841 skb
->ip_summed
= CHECKSUM_NONE
;
1842 skb
->encapsulation
= 0;
1846 if (vxflags
& VXLAN_F_GBP
)
1847 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1849 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1851 return udp_tunnel_xmit_skb(rt
, skb
, src
, dst
, tos
,
1852 ttl
, df
, src_port
, dst_port
, xnet
,
1853 !(vxflags
& VXLAN_F_UDP_CSUM
));
1855 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1857 /* Bypass encapsulation if the destination is local */
1858 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1859 struct vxlan_dev
*dst_vxlan
)
1861 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1862 union vxlan_addr loopback
;
1863 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1864 struct net_device
*dev
= skb
->dev
;
1867 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1868 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1869 skb
->pkt_type
= PACKET_HOST
;
1870 skb
->encapsulation
= 0;
1871 skb
->dev
= dst_vxlan
->dev
;
1872 __skb_pull(skb
, skb_network_offset(skb
));
1874 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1875 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1876 loopback
.sa
.sa_family
= AF_INET
;
1877 #if IS_ENABLED(CONFIG_IPV6)
1879 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1880 loopback
.sa
.sa_family
= AF_INET6
;
1884 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1885 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1887 u64_stats_update_begin(&tx_stats
->syncp
);
1888 tx_stats
->tx_packets
++;
1889 tx_stats
->tx_bytes
+= len
;
1890 u64_stats_update_end(&tx_stats
->syncp
);
1892 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1893 u64_stats_update_begin(&rx_stats
->syncp
);
1894 rx_stats
->rx_packets
++;
1895 rx_stats
->rx_bytes
+= len
;
1896 u64_stats_update_end(&rx_stats
->syncp
);
1898 dev
->stats
.rx_dropped
++;
1902 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1903 struct vxlan_rdst
*rdst
, bool did_rsc
)
1905 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1906 struct rtable
*rt
= NULL
;
1907 const struct iphdr
*old_iph
;
1909 union vxlan_addr
*dst
;
1910 struct vxlan_metadata md
;
1911 __be16 src_port
= 0, dst_port
;
1917 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1918 vni
= rdst
->remote_vni
;
1919 dst
= &rdst
->remote_ip
;
1921 if (vxlan_addr_any(dst
)) {
1923 /* short-circuited back to local bridge */
1924 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1930 old_iph
= ip_hdr(skb
);
1933 if (!ttl
&& vxlan_addr_multicast(dst
))
1938 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1940 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->port_min
,
1941 vxlan
->port_max
, true);
1943 if (dst
->sa
.sa_family
== AF_INET
) {
1944 memset(&fl4
, 0, sizeof(fl4
));
1945 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1946 fl4
.flowi4_tos
= RT_TOS(tos
);
1947 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1948 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1950 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1952 netdev_dbg(dev
, "no route to %pI4\n",
1953 &dst
->sin
.sin_addr
.s_addr
);
1954 dev
->stats
.tx_carrier_errors
++;
1958 if (rt
->dst
.dev
== dev
) {
1959 netdev_dbg(dev
, "circular route to %pI4\n",
1960 &dst
->sin
.sin_addr
.s_addr
);
1961 dev
->stats
.collisions
++;
1965 /* Bypass encapsulation if the destination is local */
1966 if (rt
->rt_flags
& RTCF_LOCAL
&&
1967 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1968 struct vxlan_dev
*dst_vxlan
;
1971 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
1972 dst
->sa
.sa_family
, dst_port
,
1976 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1980 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1981 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1982 md
.vni
= htonl(vni
<< 8);
1985 err
= vxlan_xmit_skb(rt
, skb
, fl4
.saddr
,
1986 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
1987 src_port
, dst_port
, &md
,
1988 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
1991 /* skb is already freed. */
1996 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1997 #if IS_ENABLED(CONFIG_IPV6)
1999 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
2000 struct dst_entry
*ndst
;
2004 memset(&fl6
, 0, sizeof(fl6
));
2005 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
2006 fl6
.daddr
= dst
->sin6
.sin6_addr
;
2007 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
2008 fl6
.flowi6_proto
= IPPROTO_UDP
;
2010 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
2011 netdev_dbg(dev
, "no route to %pI6\n",
2012 &dst
->sin6
.sin6_addr
);
2013 dev
->stats
.tx_carrier_errors
++;
2017 if (ndst
->dev
== dev
) {
2018 netdev_dbg(dev
, "circular route to %pI6\n",
2019 &dst
->sin6
.sin6_addr
);
2021 dev
->stats
.collisions
++;
2025 /* Bypass encapsulation if the destination is local */
2026 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2027 if (flags
& RTF_LOCAL
&&
2028 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2029 struct vxlan_dev
*dst_vxlan
;
2032 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2033 dst
->sa
.sa_family
, dst_port
,
2037 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2041 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2042 md
.vni
= htonl(vni
<< 8);
2045 err
= vxlan6_xmit_skb(ndst
, skb
, dev
, &fl6
.saddr
, &fl6
.daddr
,
2046 0, ttl
, src_port
, dst_port
, &md
,
2047 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)),
2055 dev
->stats
.tx_dropped
++;
2061 dev
->stats
.tx_errors
++;
2066 /* Transmit local packets over Vxlan
2068 * Outer IP header inherits ECN and DF from inner header.
2069 * Outer UDP destination is the VXLAN assigned port.
2070 * source port is based on hash of flow
2072 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2074 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2076 bool did_rsc
= false;
2077 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2078 struct vxlan_fdb
*f
;
2080 skb_reset_mac_header(skb
);
2083 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2084 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2085 return arp_reduce(dev
, skb
);
2086 #if IS_ENABLED(CONFIG_IPV6)
2087 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2088 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2089 + sizeof(struct nd_msg
)) &&
2090 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2093 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2094 if (msg
->icmph
.icmp6_code
== 0 &&
2095 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2096 return neigh_reduce(dev
, skb
);
2102 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2105 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2106 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2107 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2108 did_rsc
= route_shortcircuit(dev
, skb
);
2110 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2114 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2116 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2117 !is_multicast_ether_addr(eth
->h_dest
))
2118 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2120 dev
->stats
.tx_dropped
++;
2122 return NETDEV_TX_OK
;
2126 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2127 struct sk_buff
*skb1
;
2133 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2135 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2139 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2142 return NETDEV_TX_OK
;
2145 /* Walk the forwarding table and purge stale entries */
2146 static void vxlan_cleanup(unsigned long arg
)
2148 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2149 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2152 if (!netif_running(vxlan
->dev
))
2155 spin_lock_bh(&vxlan
->hash_lock
);
2156 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2157 struct hlist_node
*p
, *n
;
2158 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2160 = container_of(p
, struct vxlan_fdb
, hlist
);
2161 unsigned long timeout
;
2163 if (f
->state
& NUD_PERMANENT
)
2166 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2167 if (time_before_eq(timeout
, jiffies
)) {
2168 netdev_dbg(vxlan
->dev
,
2169 "garbage collect %pM\n",
2171 f
->state
= NUD_STALE
;
2172 vxlan_fdb_destroy(vxlan
, f
);
2173 } else if (time_before(timeout
, next_timer
))
2174 next_timer
= timeout
;
2177 spin_unlock_bh(&vxlan
->hash_lock
);
2179 mod_timer(&vxlan
->age_timer
, next_timer
);
2182 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2184 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2186 vxlan
->vn_sock
= vs
;
2187 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2190 /* Setup stats when device is created */
2191 static int vxlan_init(struct net_device
*dev
)
2193 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2194 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2195 struct vxlan_sock
*vs
;
2196 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2198 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2202 spin_lock(&vn
->sock_lock
);
2203 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2204 vxlan
->dst_port
, vxlan
->flags
);
2205 if (vs
&& atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2206 /* If we have a socket with same port already, reuse it */
2207 vxlan_vs_add_dev(vs
, vxlan
);
2209 /* otherwise make new socket outside of RTNL */
2211 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2213 spin_unlock(&vn
->sock_lock
);
2218 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2220 struct vxlan_fdb
*f
;
2222 spin_lock_bh(&vxlan
->hash_lock
);
2223 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2225 vxlan_fdb_destroy(vxlan
, f
);
2226 spin_unlock_bh(&vxlan
->hash_lock
);
2229 static void vxlan_uninit(struct net_device
*dev
)
2231 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2232 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2234 vxlan_fdb_delete_default(vxlan
);
2237 vxlan_sock_release(vs
);
2238 free_percpu(dev
->tstats
);
2241 /* Start ageing timer and join group when device is brought up */
2242 static int vxlan_open(struct net_device
*dev
)
2244 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2245 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2247 /* socket hasn't been created */
2251 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2252 vxlan_sock_hold(vs
);
2254 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2257 if (vxlan
->age_interval
)
2258 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2263 /* Purge the forwarding table */
2264 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2268 spin_lock_bh(&vxlan
->hash_lock
);
2269 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2270 struct hlist_node
*p
, *n
;
2271 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2273 = container_of(p
, struct vxlan_fdb
, hlist
);
2274 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2275 if (!is_zero_ether_addr(f
->eth_addr
))
2276 vxlan_fdb_destroy(vxlan
, f
);
2279 spin_unlock_bh(&vxlan
->hash_lock
);
2282 /* Cleanup timer and forwarding table on shutdown */
2283 static int vxlan_stop(struct net_device
*dev
)
2285 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2286 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2287 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2289 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2290 !vxlan_group_used(vn
, vxlan
)) {
2291 vxlan_sock_hold(vs
);
2293 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2296 del_timer_sync(&vxlan
->age_timer
);
2303 /* Stub, nothing needs to be done. */
2304 static void vxlan_set_multicast_list(struct net_device
*dev
)
2308 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2310 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2311 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2312 struct net_device
*lowerdev
;
2315 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2316 if (lowerdev
== NULL
)
2317 return eth_change_mtu(dev
, new_mtu
);
2319 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2320 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2322 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2324 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2331 static const struct net_device_ops vxlan_netdev_ops
= {
2332 .ndo_init
= vxlan_init
,
2333 .ndo_uninit
= vxlan_uninit
,
2334 .ndo_open
= vxlan_open
,
2335 .ndo_stop
= vxlan_stop
,
2336 .ndo_start_xmit
= vxlan_xmit
,
2337 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2338 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2339 .ndo_change_mtu
= vxlan_change_mtu
,
2340 .ndo_validate_addr
= eth_validate_addr
,
2341 .ndo_set_mac_address
= eth_mac_addr
,
2342 .ndo_fdb_add
= vxlan_fdb_add
,
2343 .ndo_fdb_del
= vxlan_fdb_delete
,
2344 .ndo_fdb_dump
= vxlan_fdb_dump
,
2347 /* Info for udev, that this is a virtual tunnel endpoint */
2348 static struct device_type vxlan_type
= {
2352 /* Calls the ndo_add_vxlan_port of the caller in order to
2353 * supply the listening VXLAN udp ports. Callers are expected
2354 * to implement the ndo_add_vxlan_port.
2356 void vxlan_get_rx_port(struct net_device
*dev
)
2358 struct vxlan_sock
*vs
;
2359 struct net
*net
= dev_net(dev
);
2360 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2361 sa_family_t sa_family
;
2365 spin_lock(&vn
->sock_lock
);
2366 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2367 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2368 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2369 sa_family
= vs
->sock
->sk
->sk_family
;
2370 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2374 spin_unlock(&vn
->sock_lock
);
2376 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2378 /* Initialize the device structure. */
2379 static void vxlan_setup(struct net_device
*dev
)
2381 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2384 eth_hw_addr_random(dev
);
2386 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2387 dev
->needed_headroom
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2389 dev
->needed_headroom
= ETH_HLEN
+ VXLAN_HEADROOM
;
2391 dev
->netdev_ops
= &vxlan_netdev_ops
;
2392 dev
->destructor
= free_netdev
;
2393 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2395 dev
->tx_queue_len
= 0;
2396 dev
->features
|= NETIF_F_LLTX
;
2397 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2398 dev
->features
|= NETIF_F_RXCSUM
;
2399 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2401 dev
->vlan_features
= dev
->features
;
2402 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2403 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2404 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2405 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2406 netif_keep_dst(dev
);
2407 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2409 INIT_LIST_HEAD(&vxlan
->next
);
2410 spin_lock_init(&vxlan
->hash_lock
);
2411 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2412 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2413 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2415 init_timer_deferrable(&vxlan
->age_timer
);
2416 vxlan
->age_timer
.function
= vxlan_cleanup
;
2417 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2419 vxlan
->dst_port
= htons(vxlan_port
);
2423 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2424 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2427 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2428 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2429 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2430 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2431 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2432 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2433 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2434 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2435 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2436 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2437 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2438 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2439 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2440 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2441 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2442 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2443 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2444 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2445 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2446 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2447 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2448 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2449 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2450 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2453 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2455 if (tb
[IFLA_ADDRESS
]) {
2456 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2457 pr_debug("invalid link address (not ethernet)\n");
2461 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2462 pr_debug("invalid all zero ethernet address\n");
2463 return -EADDRNOTAVAIL
;
2470 if (data
[IFLA_VXLAN_ID
]) {
2471 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2472 if (id
>= VXLAN_VID_MASK
)
2476 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2477 const struct ifla_vxlan_port_range
*p
2478 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2480 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2481 pr_debug("port range %u .. %u not valid\n",
2482 ntohs(p
->low
), ntohs(p
->high
));
2490 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2491 struct ethtool_drvinfo
*drvinfo
)
2493 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2494 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2497 static const struct ethtool_ops vxlan_ethtool_ops
= {
2498 .get_drvinfo
= vxlan_get_drvinfo
,
2499 .get_link
= ethtool_op_get_link
,
2502 static void vxlan_del_work(struct work_struct
*work
)
2504 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2505 udp_tunnel_sock_release(vs
->sock
);
2509 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2510 __be16 port
, u32 flags
)
2512 struct socket
*sock
;
2513 struct udp_port_cfg udp_conf
;
2516 memset(&udp_conf
, 0, sizeof(udp_conf
));
2519 udp_conf
.family
= AF_INET6
;
2520 udp_conf
.use_udp6_rx_checksums
=
2521 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2523 udp_conf
.family
= AF_INET
;
2524 udp_conf
.local_ip
.s_addr
= INADDR_ANY
;
2527 udp_conf
.local_udp_port
= port
;
2529 /* Open UDP socket */
2530 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2532 return ERR_PTR(err
);
2537 /* Create new listen socket if needed */
2538 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2539 vxlan_rcv_t
*rcv
, void *data
,
2542 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2543 struct vxlan_sock
*vs
;
2544 struct socket
*sock
;
2546 bool ipv6
= !!(flags
& VXLAN_F_IPV6
);
2547 struct udp_tunnel_sock_cfg tunnel_cfg
;
2549 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2551 return ERR_PTR(-ENOMEM
);
2553 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2554 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2556 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2558 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2561 return ERR_CAST(sock
);
2565 atomic_set(&vs
->refcnt
, 1);
2568 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2570 /* Initialize the vxlan udp offloads structure */
2571 vs
->udp_offloads
.port
= port
;
2572 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2573 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2575 spin_lock(&vn
->sock_lock
);
2576 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2577 vxlan_notify_add_rx_port(vs
);
2578 spin_unlock(&vn
->sock_lock
);
2580 /* Mark socket as an encapsulation socket. */
2581 tunnel_cfg
.sk_user_data
= vs
;
2582 tunnel_cfg
.encap_type
= 1;
2583 tunnel_cfg
.encap_rcv
= vxlan_udp_encap_recv
;
2584 tunnel_cfg
.encap_destroy
= NULL
;
2586 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2591 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2592 vxlan_rcv_t
*rcv
, void *data
,
2593 bool no_share
, u32 flags
)
2595 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2596 struct vxlan_sock
*vs
;
2597 bool ipv6
= flags
& VXLAN_F_IPV6
;
2599 vs
= vxlan_socket_create(net
, port
, rcv
, data
, flags
);
2603 if (no_share
) /* Return error if sharing is not allowed. */
2606 spin_lock(&vn
->sock_lock
);
2607 vs
= vxlan_find_sock(net
, ipv6
? AF_INET6
: AF_INET
, port
, flags
);
2608 if (vs
&& ((vs
->rcv
!= rcv
) ||
2609 !atomic_add_unless(&vs
->refcnt
, 1, 0)))
2610 vs
= ERR_PTR(-EBUSY
);
2611 spin_unlock(&vn
->sock_lock
);
2614 vs
= ERR_PTR(-EINVAL
);
2618 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2620 /* Scheduled at device creation to bind to a socket */
2621 static void vxlan_sock_work(struct work_struct
*work
)
2623 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2624 struct net
*net
= vxlan
->net
;
2625 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2626 __be16 port
= vxlan
->dst_port
;
2627 struct vxlan_sock
*nvs
;
2629 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
);
2630 spin_lock(&vn
->sock_lock
);
2632 vxlan_vs_add_dev(nvs
, vxlan
);
2633 spin_unlock(&vn
->sock_lock
);
2635 dev_put(vxlan
->dev
);
2638 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2639 struct nlattr
*tb
[], struct nlattr
*data
[])
2641 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2642 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2643 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2646 bool use_ipv6
= false;
2648 if (!data
[IFLA_VXLAN_ID
])
2651 vxlan
->net
= dev_net(dev
);
2653 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2654 dst
->remote_vni
= vni
;
2656 /* Unless IPv6 is explicitly requested, assume IPv4 */
2657 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2658 if (data
[IFLA_VXLAN_GROUP
]) {
2659 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2660 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2661 if (!IS_ENABLED(CONFIG_IPV6
))
2662 return -EPFNOSUPPORT
;
2664 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2665 sizeof(struct in6_addr
));
2666 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2670 if (data
[IFLA_VXLAN_LOCAL
]) {
2671 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2672 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2673 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2674 if (!IS_ENABLED(CONFIG_IPV6
))
2675 return -EPFNOSUPPORT
;
2677 /* TODO: respect scope id */
2678 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2679 sizeof(struct in6_addr
));
2680 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2684 if (data
[IFLA_VXLAN_LINK
] &&
2685 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2686 struct net_device
*lowerdev
2687 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2690 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2694 #if IS_ENABLED(CONFIG_IPV6)
2696 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2697 if (idev
&& idev
->cnf
.disable_ipv6
) {
2698 pr_info("IPv6 is disabled via sysctl\n");
2701 vxlan
->flags
|= VXLAN_F_IPV6
;
2706 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2708 dev
->needed_headroom
= lowerdev
->hard_header_len
+
2709 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2710 } else if (use_ipv6
)
2711 vxlan
->flags
|= VXLAN_F_IPV6
;
2713 if (data
[IFLA_VXLAN_TOS
])
2714 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2716 if (data
[IFLA_VXLAN_TTL
])
2717 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2719 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2720 vxlan
->flags
|= VXLAN_F_LEARN
;
2722 if (data
[IFLA_VXLAN_AGEING
])
2723 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2725 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2727 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2728 vxlan
->flags
|= VXLAN_F_PROXY
;
2730 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2731 vxlan
->flags
|= VXLAN_F_RSC
;
2733 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2734 vxlan
->flags
|= VXLAN_F_L2MISS
;
2736 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2737 vxlan
->flags
|= VXLAN_F_L3MISS
;
2739 if (data
[IFLA_VXLAN_LIMIT
])
2740 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2742 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2743 const struct ifla_vxlan_port_range
*p
2744 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2745 vxlan
->port_min
= ntohs(p
->low
);
2746 vxlan
->port_max
= ntohs(p
->high
);
2749 if (data
[IFLA_VXLAN_PORT
])
2750 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2752 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2753 vxlan
->flags
|= VXLAN_F_UDP_CSUM
;
2755 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2756 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2757 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2759 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2760 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2761 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2763 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
2764 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
2765 vxlan
->flags
|= VXLAN_F_REMCSUM_TX
;
2767 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
2768 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
2769 vxlan
->flags
|= VXLAN_F_REMCSUM_RX
;
2771 if (data
[IFLA_VXLAN_GBP
])
2772 vxlan
->flags
|= VXLAN_F_GBP
;
2774 if (vxlan_find_vni(net
, vni
, use_ipv6
? AF_INET6
: AF_INET
,
2775 vxlan
->dst_port
, vxlan
->flags
)) {
2776 pr_info("duplicate VNI %u\n", vni
);
2780 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2782 /* create an fdb entry for a valid default destination */
2783 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2784 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2785 &vxlan
->default_dst
.remote_ip
,
2786 NUD_REACHABLE
|NUD_PERMANENT
,
2787 NLM_F_EXCL
|NLM_F_CREATE
,
2789 vxlan
->default_dst
.remote_vni
,
2790 vxlan
->default_dst
.remote_ifindex
,
2796 err
= register_netdevice(dev
);
2798 vxlan_fdb_delete_default(vxlan
);
2802 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2807 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2809 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2810 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2812 spin_lock(&vn
->sock_lock
);
2813 if (!hlist_unhashed(&vxlan
->hlist
))
2814 hlist_del_rcu(&vxlan
->hlist
);
2815 spin_unlock(&vn
->sock_lock
);
2817 list_del(&vxlan
->next
);
2818 unregister_netdevice_queue(dev
, head
);
2821 static size_t vxlan_get_size(const struct net_device
*dev
)
2824 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2825 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2826 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2827 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2828 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2829 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2830 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2831 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2832 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2833 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2834 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2835 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2836 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2837 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2838 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2839 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
2840 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2841 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2842 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
2843 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
2847 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2849 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2850 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2851 struct ifla_vxlan_port_range ports
= {
2852 .low
= htons(vxlan
->port_min
),
2853 .high
= htons(vxlan
->port_max
),
2856 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2857 goto nla_put_failure
;
2859 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2860 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2861 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2862 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2863 goto nla_put_failure
;
2864 #if IS_ENABLED(CONFIG_IPV6)
2866 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2867 &dst
->remote_ip
.sin6
.sin6_addr
))
2868 goto nla_put_failure
;
2873 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2874 goto nla_put_failure
;
2876 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2877 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2878 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2879 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2880 goto nla_put_failure
;
2881 #if IS_ENABLED(CONFIG_IPV6)
2883 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2884 &vxlan
->saddr
.sin6
.sin6_addr
))
2885 goto nla_put_failure
;
2890 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2891 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2892 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2893 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2894 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2895 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2896 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2897 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2898 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2899 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2900 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2901 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2902 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2903 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
) ||
2904 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
2905 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
2906 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
2907 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
2908 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
2909 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
2910 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
2911 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
2912 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
2913 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
2914 goto nla_put_failure
;
2916 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2917 goto nla_put_failure
;
2919 if (vxlan
->flags
& VXLAN_F_GBP
&&
2920 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
2921 goto nla_put_failure
;
2929 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
2931 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2936 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2938 .maxtype
= IFLA_VXLAN_MAX
,
2939 .policy
= vxlan_policy
,
2940 .priv_size
= sizeof(struct vxlan_dev
),
2941 .setup
= vxlan_setup
,
2942 .validate
= vxlan_validate
,
2943 .newlink
= vxlan_newlink
,
2944 .dellink
= vxlan_dellink
,
2945 .get_size
= vxlan_get_size
,
2946 .fill_info
= vxlan_fill_info
,
2947 .get_link_net
= vxlan_get_link_net
,
2950 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2951 struct net_device
*dev
)
2953 struct vxlan_dev
*vxlan
, *next
;
2954 LIST_HEAD(list_kill
);
2956 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2957 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2959 /* In case we created vxlan device with carrier
2960 * and we loose the carrier due to module unload
2961 * we also need to remove vxlan device. In other
2962 * cases, it's not necessary and remote_ifindex
2963 * is 0 here, so no matches.
2965 if (dst
->remote_ifindex
== dev
->ifindex
)
2966 vxlan_dellink(vxlan
->dev
, &list_kill
);
2969 unregister_netdevice_many(&list_kill
);
2972 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2973 unsigned long event
, void *ptr
)
2975 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2976 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2978 if (event
== NETDEV_UNREGISTER
)
2979 vxlan_handle_lowerdev_unregister(vn
, dev
);
2984 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2985 .notifier_call
= vxlan_lowerdev_event
,
2988 static __net_init
int vxlan_init_net(struct net
*net
)
2990 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2993 INIT_LIST_HEAD(&vn
->vxlan_list
);
2994 spin_lock_init(&vn
->sock_lock
);
2996 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2997 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3002 static void __net_exit
vxlan_exit_net(struct net
*net
)
3004 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3005 struct vxlan_dev
*vxlan
, *next
;
3006 struct net_device
*dev
, *aux
;
3010 for_each_netdev_safe(net
, dev
, aux
)
3011 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3012 unregister_netdevice_queue(dev
, &list
);
3014 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3015 /* If vxlan->dev is in the same netns, it has already been added
3016 * to the list by the previous loop.
3018 if (!net_eq(dev_net(vxlan
->dev
), net
))
3019 unregister_netdevice_queue(dev
, &list
);
3022 unregister_netdevice_many(&list
);
3026 static struct pernet_operations vxlan_net_ops
= {
3027 .init
= vxlan_init_net
,
3028 .exit
= vxlan_exit_net
,
3029 .id
= &vxlan_net_id
,
3030 .size
= sizeof(struct vxlan_net
),
3033 static int __init
vxlan_init_module(void)
3037 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3041 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3043 rc
= register_pernet_subsys(&vxlan_net_ops
);
3047 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3051 rc
= rtnl_link_register(&vxlan_link_ops
);
3057 unregister_netdevice_notifier(&vxlan_notifier_block
);
3059 unregister_pernet_subsys(&vxlan_net_ops
);
3061 destroy_workqueue(vxlan_wq
);
3064 late_initcall(vxlan_init_module
);
3066 static void __exit
vxlan_cleanup_module(void)
3068 rtnl_link_unregister(&vxlan_link_ops
);
3069 unregister_netdevice_notifier(&vxlan_notifier_block
);
3070 destroy_workqueue(vxlan_wq
);
3071 unregister_pernet_subsys(&vxlan_net_ops
);
3072 /* rcu_barrier() is called by netns */
3074 module_exit(vxlan_cleanup_module
);
3076 MODULE_LICENSE("GPL");
3077 MODULE_VERSION(VXLAN_VERSION
);
3078 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3079 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3080 MODULE_ALIAS_RTNL_LINK("vxlan");