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 #if IS_ENABLED(CONFIG_IPV6)
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/ip6_checksum.h>
52 #define VXLAN_VERSION "0.1"
54 #define PORT_HASH_BITS 8
55 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
56 #define VNI_HASH_BITS 10
57 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
58 #define FDB_HASH_BITS 8
59 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
60 #define FDB_AGE_DEFAULT 300 /* 5 min */
61 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
63 #define VXLAN_N_VID (1u << 24)
64 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
65 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
67 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
69 /* VXLAN protocol header */
75 /* UDP port for VXLAN traffic.
76 * The IANA assigned port is 4789, but the Linux default is 8472
77 * for compatibility with early adopters.
79 static unsigned short vxlan_port __read_mostly
= 8472;
80 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
81 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
83 static bool log_ecn_error
= true;
84 module_param(log_ecn_error
, bool, 0644);
85 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
87 static int vxlan_net_id
;
89 static const u8 all_zeros_mac
[ETH_ALEN
];
91 /* per-network namespace private data for this module */
93 struct list_head vxlan_list
;
94 struct hlist_head sock_list
[PORT_HASH_SIZE
];
99 struct sockaddr_in sin
;
100 struct sockaddr_in6 sin6
;
105 union vxlan_addr remote_ip
;
109 struct list_head list
;
113 /* Forwarding table entry */
115 struct hlist_node hlist
; /* linked list of entries */
117 unsigned long updated
; /* jiffies */
119 struct list_head remotes
;
120 u16 state
; /* see ndm_state */
121 u8 flags
; /* see ndm_flags */
122 u8 eth_addr
[ETH_ALEN
];
125 /* Pseudo network device */
127 struct hlist_node hlist
; /* vni hash table */
128 struct list_head next
; /* vxlan's per namespace list */
129 struct vxlan_sock
*vn_sock
; /* listening socket */
130 struct net_device
*dev
;
131 struct net
*net
; /* netns for packet i/o */
132 struct vxlan_rdst default_dst
; /* default destination */
133 union vxlan_addr saddr
; /* source address */
135 __u16 port_min
; /* source port range */
137 __u8 tos
; /* TOS override */
139 u32 flags
; /* VXLAN_F_* in vxlan.h */
141 struct work_struct sock_work
;
142 struct work_struct igmp_join
;
143 struct work_struct igmp_leave
;
145 unsigned long age_interval
;
146 struct timer_list age_timer
;
147 spinlock_t hash_lock
;
148 unsigned int addrcnt
;
149 unsigned int addrmax
;
151 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
154 /* salt for hash table */
155 static u32 vxlan_salt __read_mostly
;
156 static struct workqueue_struct
*vxlan_wq
;
158 static void vxlan_sock_work(struct work_struct
*work
);
160 #if IS_ENABLED(CONFIG_IPV6)
162 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
164 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
166 if (a
->sa
.sa_family
== AF_INET6
)
167 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
169 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
172 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
174 if (ipa
->sa
.sa_family
== AF_INET6
)
175 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
177 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
180 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
182 if (ipa
->sa
.sa_family
== AF_INET6
)
183 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
185 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
188 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
190 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
191 nla_memcpy(&ip
->sin6
.sin6_addr
, nla
, sizeof(struct in6_addr
));
192 ip
->sa
.sa_family
= AF_INET6
;
194 } else if (nla_len(nla
) >= sizeof(__be32
)) {
195 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
196 ip
->sa
.sa_family
= AF_INET
;
199 return -EAFNOSUPPORT
;
203 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
204 const union vxlan_addr
*ip
)
206 if (ip
->sa
.sa_family
== AF_INET6
)
207 return nla_put(skb
, attr
, sizeof(struct in6_addr
), &ip
->sin6
.sin6_addr
);
209 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
212 #else /* !CONFIG_IPV6 */
215 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
217 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
220 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
222 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
225 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
227 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
230 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
232 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
233 return -EAFNOSUPPORT
;
234 } else if (nla_len(nla
) >= sizeof(__be32
)) {
235 ip
->sin
.sin_addr
.s_addr
= nla_get_be32(nla
);
236 ip
->sa
.sa_family
= AF_INET
;
239 return -EAFNOSUPPORT
;
243 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
244 const union vxlan_addr
*ip
)
246 return nla_put_be32(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
250 /* Virtual Network hash table head */
251 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
253 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
256 /* Socket hash table head */
257 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
259 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
261 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
264 /* First remote destination for a forwarding entry.
265 * Guaranteed to be non-NULL because remotes are never deleted.
267 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
269 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
272 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
274 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
277 /* Find VXLAN socket based on network namespace and UDP port */
278 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, __be16 port
)
280 struct vxlan_sock
*vs
;
282 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
283 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
289 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, u32 id
)
291 struct vxlan_dev
*vxlan
;
293 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
294 if (vxlan
->default_dst
.remote_vni
== id
)
301 /* Look up VNI in a per net namespace table */
302 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
304 struct vxlan_sock
*vs
;
306 vs
= vxlan_find_sock(net
, port
);
310 return vxlan_vs_find_vni(vs
, id
);
313 /* Fill in neighbour message in skbuff. */
314 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
315 const struct vxlan_fdb
*fdb
,
316 u32 portid
, u32 seq
, int type
, unsigned int flags
,
317 const struct vxlan_rdst
*rdst
)
319 unsigned long now
= jiffies
;
320 struct nda_cacheinfo ci
;
321 struct nlmsghdr
*nlh
;
323 bool send_ip
, send_eth
;
325 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
329 ndm
= nlmsg_data(nlh
);
330 memset(ndm
, 0, sizeof(*ndm
));
332 send_eth
= send_ip
= true;
334 if (type
== RTM_GETNEIGH
) {
335 ndm
->ndm_family
= AF_INET
;
336 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
337 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
339 ndm
->ndm_family
= AF_BRIDGE
;
340 ndm
->ndm_state
= fdb
->state
;
341 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
342 ndm
->ndm_flags
= fdb
->flags
;
343 ndm
->ndm_type
= RTN_UNICAST
;
345 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
346 goto nla_put_failure
;
348 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
349 goto nla_put_failure
;
351 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
352 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
353 goto nla_put_failure
;
354 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
355 nla_put_u32(skb
, NDA_VNI
, rdst
->remote_vni
))
356 goto nla_put_failure
;
357 if (rdst
->remote_ifindex
&&
358 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
359 goto nla_put_failure
;
361 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
362 ci
.ndm_confirmed
= 0;
363 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
366 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
367 goto nla_put_failure
;
369 return nlmsg_end(skb
, nlh
);
372 nlmsg_cancel(skb
, nlh
);
376 static inline size_t vxlan_nlmsg_size(void)
378 return NLMSG_ALIGN(sizeof(struct ndmsg
))
379 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
380 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
381 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
382 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
383 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
384 + nla_total_size(sizeof(struct nda_cacheinfo
));
387 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
388 struct vxlan_rdst
*rd
, int type
)
390 struct net
*net
= dev_net(vxlan
->dev
);
394 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
398 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
400 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
401 WARN_ON(err
== -EMSGSIZE
);
406 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
410 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
413 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
415 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
416 struct vxlan_fdb f
= {
419 struct vxlan_rdst remote
= {
420 .remote_ip
= *ipa
, /* goes to NDA_DST */
421 .remote_vni
= VXLAN_N_VID
,
424 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
427 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
429 struct vxlan_fdb f
= {
432 struct vxlan_rdst remote
= { };
434 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
436 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
439 /* Hash Ethernet address */
440 static u32
eth_hash(const unsigned char *addr
)
442 u64 value
= get_unaligned((u64
*)addr
);
444 /* only want 6 bytes */
450 return hash_64(value
, FDB_HASH_BITS
);
453 /* Hash chain to use given mac address */
454 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
457 return &vxlan
->fdb_head
[eth_hash(mac
)];
460 /* Look up Ethernet address in forwarding table */
461 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
464 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
467 hlist_for_each_entry_rcu(f
, head
, hlist
) {
468 if (ether_addr_equal(mac
, f
->eth_addr
))
475 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
480 f
= __vxlan_find_mac(vxlan
, mac
);
487 /* caller should hold vxlan->hash_lock */
488 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
489 union vxlan_addr
*ip
, __be16 port
,
490 __u32 vni
, __u32 ifindex
)
492 struct vxlan_rdst
*rd
;
494 list_for_each_entry(rd
, &f
->remotes
, list
) {
495 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
496 rd
->remote_port
== port
&&
497 rd
->remote_vni
== vni
&&
498 rd
->remote_ifindex
== ifindex
)
505 /* Replace destination of unicast mac */
506 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
507 union vxlan_addr
*ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
509 struct vxlan_rdst
*rd
;
511 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
515 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
519 rd
->remote_port
= port
;
520 rd
->remote_vni
= vni
;
521 rd
->remote_ifindex
= ifindex
;
525 /* Add/update destinations for multicast */
526 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
527 union vxlan_addr
*ip
, __be16 port
, __u32 vni
,
528 __u32 ifindex
, struct vxlan_rdst
**rdp
)
530 struct vxlan_rdst
*rd
;
532 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
536 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
540 rd
->remote_port
= port
;
541 rd
->remote_vni
= vni
;
542 rd
->remote_ifindex
= ifindex
;
544 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
550 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
)
552 struct sk_buff
*p
, **pp
= NULL
;
553 struct vxlanhdr
*vh
, *vh2
;
554 struct ethhdr
*eh
, *eh2
;
555 unsigned int hlen
, off_vx
, off_eth
;
556 const struct packet_offload
*ptype
;
560 off_vx
= skb_gro_offset(skb
);
561 hlen
= off_vx
+ sizeof(*vh
);
562 vh
= skb_gro_header_fast(skb
, off_vx
);
563 if (skb_gro_header_hard(skb
, hlen
)) {
564 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
568 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
569 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
571 off_eth
= skb_gro_offset(skb
);
572 hlen
= off_eth
+ sizeof(*eh
);
573 eh
= skb_gro_header_fast(skb
, off_eth
);
574 if (skb_gro_header_hard(skb
, hlen
)) {
575 eh
= skb_gro_header_slow(skb
, hlen
, off_eth
);
582 for (p
= *head
; p
; p
= p
->next
) {
583 if (!NAPI_GRO_CB(p
)->same_flow
)
586 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
587 eh2
= (struct ethhdr
*)(p
->data
+ off_eth
);
588 if (vh
->vx_vni
!= vh2
->vx_vni
|| compare_ether_header(eh
, eh2
)) {
589 NAPI_GRO_CB(p
)->same_flow
= 0;
597 ptype
= gro_find_receive_by_type(type
);
603 skb_gro_pull(skb
, sizeof(*eh
)); /* pull inner eth header */
604 skb_gro_postpull_rcsum(skb
, eh
, sizeof(*eh
));
605 pp
= ptype
->callbacks
.gro_receive(head
, skb
);
610 NAPI_GRO_CB(skb
)->flush
|= flush
;
615 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
)
618 struct packet_offload
*ptype
;
620 int vxlan_len
= sizeof(struct vxlanhdr
) + sizeof(struct ethhdr
);
623 eh
= (struct ethhdr
*)(skb
->data
+ nhoff
+ sizeof(struct vxlanhdr
));
627 ptype
= gro_find_complete_by_type(type
);
629 err
= ptype
->callbacks
.gro_complete(skb
, nhoff
+ vxlan_len
);
635 /* Notify netdevs that UDP port started listening */
636 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
638 struct net_device
*dev
;
639 struct sock
*sk
= vs
->sock
->sk
;
640 struct net
*net
= sock_net(sk
);
641 sa_family_t sa_family
= sk
->sk_family
;
642 __be16 port
= inet_sk(sk
)->inet_sport
;
645 if (sa_family
== AF_INET
) {
646 err
= udp_add_offload(&vs
->udp_offloads
);
648 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
652 for_each_netdev_rcu(net
, dev
) {
653 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
654 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
660 /* Notify netdevs that UDP port is no more listening */
661 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
663 struct net_device
*dev
;
664 struct sock
*sk
= vs
->sock
->sk
;
665 struct net
*net
= sock_net(sk
);
666 sa_family_t sa_family
= sk
->sk_family
;
667 __be16 port
= inet_sk(sk
)->inet_sport
;
670 for_each_netdev_rcu(net
, dev
) {
671 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
672 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
677 if (sa_family
== AF_INET
)
678 udp_del_offload(&vs
->udp_offloads
);
681 /* Add new entry to forwarding table -- assumes lock held */
682 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
683 const u8
*mac
, union vxlan_addr
*ip
,
684 __u16 state
, __u16 flags
,
685 __be16 port
, __u32 vni
, __u32 ifindex
,
688 struct vxlan_rdst
*rd
= NULL
;
692 f
= __vxlan_find_mac(vxlan
, mac
);
694 if (flags
& NLM_F_EXCL
) {
695 netdev_dbg(vxlan
->dev
,
696 "lost race to create %pM\n", mac
);
699 if (f
->state
!= state
) {
701 f
->updated
= jiffies
;
704 if (f
->flags
!= ndm_flags
) {
705 f
->flags
= ndm_flags
;
706 f
->updated
= jiffies
;
709 if ((flags
& NLM_F_REPLACE
)) {
710 /* Only change unicasts */
711 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
712 is_zero_ether_addr(f
->eth_addr
))) {
713 int rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
722 if ((flags
& NLM_F_APPEND
) &&
723 (is_multicast_ether_addr(f
->eth_addr
) ||
724 is_zero_ether_addr(f
->eth_addr
))) {
725 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
733 if (!(flags
& NLM_F_CREATE
))
736 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
739 /* Disallow replace to add a multicast entry */
740 if ((flags
& NLM_F_REPLACE
) &&
741 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
744 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
745 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
751 f
->flags
= ndm_flags
;
752 f
->updated
= f
->used
= jiffies
;
753 INIT_LIST_HEAD(&f
->remotes
);
754 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
756 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
759 hlist_add_head_rcu(&f
->hlist
,
760 vxlan_fdb_head(vxlan
, mac
));
765 rd
= first_remote_rtnl(f
);
766 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
772 static void vxlan_fdb_free(struct rcu_head
*head
)
774 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
775 struct vxlan_rdst
*rd
, *nd
;
777 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
)
782 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
784 netdev_dbg(vxlan
->dev
,
785 "delete %pM\n", f
->eth_addr
);
788 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
790 hlist_del_rcu(&f
->hlist
);
791 call_rcu(&f
->rcu
, vxlan_fdb_free
);
794 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
795 union vxlan_addr
*ip
, __be16
*port
, u32
*vni
, u32
*ifindex
)
797 struct net
*net
= dev_net(vxlan
->dev
);
801 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
805 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
806 if (remote
->sa
.sa_family
== AF_INET
) {
807 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
808 ip
->sa
.sa_family
= AF_INET
;
809 #if IS_ENABLED(CONFIG_IPV6)
811 ip
->sin6
.sin6_addr
= in6addr_any
;
812 ip
->sa
.sa_family
= AF_INET6
;
818 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
820 *port
= nla_get_be16(tb
[NDA_PORT
]);
822 *port
= vxlan
->dst_port
;
826 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
828 *vni
= nla_get_u32(tb
[NDA_VNI
]);
830 *vni
= vxlan
->default_dst
.remote_vni
;
833 if (tb
[NDA_IFINDEX
]) {
834 struct net_device
*tdev
;
836 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
838 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
839 tdev
= __dev_get_by_index(net
, *ifindex
);
841 return -EADDRNOTAVAIL
;
849 /* Add static entry (via netlink) */
850 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
851 struct net_device
*dev
,
852 const unsigned char *addr
, u16 flags
)
854 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
855 /* struct net *net = dev_net(vxlan->dev); */
861 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
862 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
867 if (tb
[NDA_DST
] == NULL
)
870 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
874 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
875 return -EAFNOSUPPORT
;
877 spin_lock_bh(&vxlan
->hash_lock
);
878 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
879 port
, vni
, ifindex
, ndm
->ndm_flags
);
880 spin_unlock_bh(&vxlan
->hash_lock
);
885 /* Delete entry (via netlink) */
886 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
887 struct net_device
*dev
,
888 const unsigned char *addr
)
890 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
892 struct vxlan_rdst
*rd
= NULL
;
898 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
904 spin_lock_bh(&vxlan
->hash_lock
);
905 f
= vxlan_find_mac(vxlan
, addr
);
909 if (!vxlan_addr_any(&ip
)) {
910 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
917 /* remove a destination if it's not the only one on the list,
918 * otherwise destroy the fdb entry
920 if (rd
&& !list_is_singular(&f
->remotes
)) {
921 list_del_rcu(&rd
->list
);
922 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
927 vxlan_fdb_destroy(vxlan
, f
);
930 spin_unlock_bh(&vxlan
->hash_lock
);
935 /* Dump forwarding table */
936 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
937 struct net_device
*dev
,
938 struct net_device
*filter_dev
, int idx
)
940 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
943 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
947 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
948 struct vxlan_rdst
*rd
;
950 if (idx
< cb
->args
[0])
953 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
954 err
= vxlan_fdb_info(skb
, vxlan
, f
,
955 NETLINK_CB(cb
->skb
).portid
,
970 /* Watch incoming packets to learn mapping between Ethernet address
971 * and Tunnel endpoint.
972 * Return true if packet is bogus and should be droppped.
974 static bool vxlan_snoop(struct net_device
*dev
,
975 union vxlan_addr
*src_ip
, const u8
*src_mac
)
977 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
980 f
= vxlan_find_mac(vxlan
, src_mac
);
982 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
984 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
987 /* Don't migrate static entries, drop packets */
988 if (f
->state
& NUD_NOARP
)
993 "%pM migrated from %pIS to %pIS\n",
994 src_mac
, &rdst
->remote_ip
, &src_ip
);
996 rdst
->remote_ip
= *src_ip
;
997 f
->updated
= jiffies
;
998 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1000 /* learned new entry */
1001 spin_lock(&vxlan
->hash_lock
);
1003 /* close off race between vxlan_flush and incoming packets */
1004 if (netif_running(dev
))
1005 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1007 NLM_F_EXCL
|NLM_F_CREATE
,
1009 vxlan
->default_dst
.remote_vni
,
1011 spin_unlock(&vxlan
->hash_lock
);
1017 /* See if multicast group is already in use by other ID */
1018 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1020 struct vxlan_dev
*vxlan
;
1022 /* The vxlan_sock is only used by dev, leaving group has
1023 * no effect on other vxlan devices.
1025 if (atomic_read(&dev
->vn_sock
->refcnt
) == 1)
1028 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1029 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1032 if (vxlan
->vn_sock
!= dev
->vn_sock
)
1035 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1036 &dev
->default_dst
.remote_ip
))
1039 if (vxlan
->default_dst
.remote_ifindex
!=
1040 dev
->default_dst
.remote_ifindex
)
1049 static void vxlan_sock_hold(struct vxlan_sock
*vs
)
1051 atomic_inc(&vs
->refcnt
);
1054 void vxlan_sock_release(struct vxlan_sock
*vs
)
1056 struct sock
*sk
= vs
->sock
->sk
;
1057 struct net
*net
= sock_net(sk
);
1058 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1060 if (!atomic_dec_and_test(&vs
->refcnt
))
1063 spin_lock(&vn
->sock_lock
);
1064 hlist_del_rcu(&vs
->hlist
);
1065 rcu_assign_sk_user_data(vs
->sock
->sk
, NULL
);
1066 vxlan_notify_del_rx_port(vs
);
1067 spin_unlock(&vn
->sock_lock
);
1069 queue_work(vxlan_wq
, &vs
->del_work
);
1071 EXPORT_SYMBOL_GPL(vxlan_sock_release
);
1073 /* Callback to update multicast group membership when first VNI on
1074 * multicast asddress is brought up
1075 * Done as workqueue because ip_mc_join_group acquires RTNL.
1077 static void vxlan_igmp_join(struct work_struct
*work
)
1079 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_join
);
1080 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1081 struct sock
*sk
= vs
->sock
->sk
;
1082 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1083 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1086 if (ip
->sa
.sa_family
== AF_INET
) {
1087 struct ip_mreqn mreq
= {
1088 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1089 .imr_ifindex
= ifindex
,
1092 ip_mc_join_group(sk
, &mreq
);
1093 #if IS_ENABLED(CONFIG_IPV6)
1095 ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1096 &ip
->sin6
.sin6_addr
);
1101 vxlan_sock_release(vs
);
1102 dev_put(vxlan
->dev
);
1105 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1106 static void vxlan_igmp_leave(struct work_struct
*work
)
1108 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, igmp_leave
);
1109 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1110 struct sock
*sk
= vs
->sock
->sk
;
1111 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1112 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1115 if (ip
->sa
.sa_family
== AF_INET
) {
1116 struct ip_mreqn mreq
= {
1117 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1118 .imr_ifindex
= ifindex
,
1121 ip_mc_leave_group(sk
, &mreq
);
1122 #if IS_ENABLED(CONFIG_IPV6)
1124 ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1125 &ip
->sin6
.sin6_addr
);
1131 vxlan_sock_release(vs
);
1132 dev_put(vxlan
->dev
);
1135 /* Callback from net/ipv4/udp.c to receive packets */
1136 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
1138 struct vxlan_sock
*vs
;
1139 struct vxlanhdr
*vxh
;
1141 /* Need Vxlan and inner Ethernet header to be present */
1142 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1145 /* Return packets with reserved bits set */
1146 vxh
= (struct vxlanhdr
*)(udp_hdr(skb
) + 1);
1147 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
1148 (vxh
->vx_vni
& htonl(0xff))) {
1149 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1150 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
1154 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
)))
1157 vs
= rcu_dereference_sk_user_data(sk
);
1161 skb_pop_rcv_encapsulation(skb
);
1163 vs
->rcv(vs
, skb
, vxh
->vx_vni
);
1167 /* Consume bad packet */
1172 /* Return non vxlan pkt */
1176 static void vxlan_rcv(struct vxlan_sock
*vs
,
1177 struct sk_buff
*skb
, __be32 vx_vni
)
1179 struct iphdr
*oip
= NULL
;
1180 struct ipv6hdr
*oip6
= NULL
;
1181 struct vxlan_dev
*vxlan
;
1182 struct pcpu_sw_netstats
*stats
;
1183 union vxlan_addr saddr
;
1186 union vxlan_addr
*remote_ip
;
1188 vni
= ntohl(vx_vni
) >> 8;
1189 /* Is this VNI defined? */
1190 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1194 remote_ip
= &vxlan
->default_dst
.remote_ip
;
1195 skb_reset_mac_header(skb
);
1196 skb_scrub_packet(skb
, !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1197 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1198 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1200 /* Ignore packet loops (and multicast echo) */
1201 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1204 /* Re-examine inner Ethernet packet */
1205 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1207 saddr
.sin
.sin_addr
.s_addr
= oip
->saddr
;
1208 saddr
.sa
.sa_family
= AF_INET
;
1209 #if IS_ENABLED(CONFIG_IPV6)
1211 oip6
= ipv6_hdr(skb
);
1212 saddr
.sin6
.sin6_addr
= oip6
->saddr
;
1213 saddr
.sa
.sa_family
= AF_INET6
;
1217 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1218 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1221 skb_reset_network_header(skb
);
1224 err
= IP6_ECN_decapsulate(oip6
, skb
);
1226 err
= IP_ECN_decapsulate(oip
, skb
);
1228 if (unlikely(err
)) {
1229 if (log_ecn_error
) {
1231 net_info_ratelimited("non-ECT from %pI6\n",
1234 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1235 &oip
->saddr
, oip
->tos
);
1238 ++vxlan
->dev
->stats
.rx_frame_errors
;
1239 ++vxlan
->dev
->stats
.rx_errors
;
1244 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1245 u64_stats_update_begin(&stats
->syncp
);
1246 stats
->rx_packets
++;
1247 stats
->rx_bytes
+= skb
->len
;
1248 u64_stats_update_end(&stats
->syncp
);
1254 /* Consume bad packet */
1258 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1260 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1261 struct arphdr
*parp
;
1264 struct neighbour
*n
;
1266 if (dev
->flags
& IFF_NOARP
)
1269 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1270 dev
->stats
.tx_dropped
++;
1273 parp
= arp_hdr(skb
);
1275 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1276 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1277 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1278 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1279 parp
->ar_hln
!= dev
->addr_len
||
1282 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1284 arpptr
+= dev
->addr_len
; /* sha */
1285 memcpy(&sip
, arpptr
, sizeof(sip
));
1286 arpptr
+= sizeof(sip
);
1287 arpptr
+= dev
->addr_len
; /* tha */
1288 memcpy(&tip
, arpptr
, sizeof(tip
));
1290 if (ipv4_is_loopback(tip
) ||
1291 ipv4_is_multicast(tip
))
1294 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1297 struct vxlan_fdb
*f
;
1298 struct sk_buff
*reply
;
1300 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1305 f
= vxlan_find_mac(vxlan
, n
->ha
);
1306 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1307 /* bridge-local neighbor */
1312 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1320 skb_reset_mac_header(reply
);
1321 __skb_pull(reply
, skb_network_offset(reply
));
1322 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1323 reply
->pkt_type
= PACKET_HOST
;
1325 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1326 dev
->stats
.rx_dropped
++;
1327 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1328 union vxlan_addr ipa
= {
1329 .sin
.sin_addr
.s_addr
= tip
,
1330 .sa
.sa_family
= AF_INET
,
1333 vxlan_ip_miss(dev
, &ipa
);
1337 return NETDEV_TX_OK
;
1340 #if IS_ENABLED(CONFIG_IPV6)
1342 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1343 struct neighbour
*n
, bool isrouter
)
1345 struct net_device
*dev
= request
->dev
;
1346 struct sk_buff
*reply
;
1347 struct nd_msg
*ns
, *na
;
1348 struct ipv6hdr
*pip6
;
1350 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1357 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1358 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1359 reply
= alloc_skb(len
, GFP_ATOMIC
);
1363 reply
->protocol
= htons(ETH_P_IPV6
);
1365 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1366 skb_push(reply
, sizeof(struct ethhdr
));
1367 skb_set_mac_header(reply
, 0);
1369 ns
= (struct nd_msg
*)skb_transport_header(request
);
1371 daddr
= eth_hdr(request
)->h_source
;
1372 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1373 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1374 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1375 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1380 /* Ethernet header */
1381 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1382 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1383 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1384 reply
->protocol
= htons(ETH_P_IPV6
);
1386 skb_pull(reply
, sizeof(struct ethhdr
));
1387 skb_set_network_header(reply
, 0);
1388 skb_put(reply
, sizeof(struct ipv6hdr
));
1392 pip6
= ipv6_hdr(reply
);
1393 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1395 pip6
->priority
= ipv6_hdr(request
)->priority
;
1396 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1397 pip6
->hop_limit
= 255;
1398 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1399 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1401 skb_pull(reply
, sizeof(struct ipv6hdr
));
1402 skb_set_transport_header(reply
, 0);
1404 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1406 /* Neighbor Advertisement */
1407 memset(na
, 0, sizeof(*na
)+na_olen
);
1408 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1409 na
->icmph
.icmp6_router
= isrouter
;
1410 na
->icmph
.icmp6_override
= 1;
1411 na
->icmph
.icmp6_solicited
= 1;
1412 na
->target
= ns
->target
;
1413 ether_addr_copy(&na
->opt
[2], n
->ha
);
1414 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1415 na
->opt
[1] = na_olen
>> 3;
1417 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1418 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1419 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1421 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1423 skb_push(reply
, sizeof(struct ipv6hdr
));
1425 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1430 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1432 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1434 const struct ipv6hdr
*iphdr
;
1435 const struct in6_addr
*saddr
, *daddr
;
1436 struct neighbour
*n
;
1437 struct inet6_dev
*in6_dev
;
1439 in6_dev
= __in6_dev_get(dev
);
1443 if (!pskb_may_pull(skb
, skb
->len
))
1446 iphdr
= ipv6_hdr(skb
);
1447 saddr
= &iphdr
->saddr
;
1448 daddr
= &iphdr
->daddr
;
1450 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1451 if (msg
->icmph
.icmp6_code
!= 0 ||
1452 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1455 if (ipv6_addr_loopback(daddr
) ||
1456 ipv6_addr_is_multicast(&msg
->target
))
1459 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1462 struct vxlan_fdb
*f
;
1463 struct sk_buff
*reply
;
1465 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1470 f
= vxlan_find_mac(vxlan
, n
->ha
);
1471 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1472 /* bridge-local neighbor */
1477 reply
= vxlan_na_create(skb
, n
,
1478 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1485 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1486 dev
->stats
.rx_dropped
++;
1488 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1489 union vxlan_addr ipa
= {
1490 .sin6
.sin6_addr
= msg
->target
,
1491 .sa
.sa_family
= AF_INET6
,
1494 vxlan_ip_miss(dev
, &ipa
);
1499 return NETDEV_TX_OK
;
1503 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1505 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1506 struct neighbour
*n
;
1508 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1512 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1517 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1520 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1521 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1522 union vxlan_addr ipa
= {
1523 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1524 .sa
.sa_family
= AF_INET
,
1527 vxlan_ip_miss(dev
, &ipa
);
1533 #if IS_ENABLED(CONFIG_IPV6)
1536 struct ipv6hdr
*pip6
;
1538 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1540 pip6
= ipv6_hdr(skb
);
1541 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1542 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1543 union vxlan_addr ipa
= {
1544 .sin6
.sin6_addr
= pip6
->daddr
,
1545 .sa
.sa_family
= AF_INET6
,
1548 vxlan_ip_miss(dev
, &ipa
);
1562 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1564 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1566 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1575 static inline struct sk_buff
*vxlan_handle_offloads(struct sk_buff
*skb
,
1578 int type
= udp_csum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1579 return iptunnel_handle_offloads(skb
, udp_csum
, type
);
1582 #if IS_ENABLED(CONFIG_IPV6)
1583 static int vxlan6_xmit_skb(struct vxlan_sock
*vs
,
1584 struct dst_entry
*dst
, struct sk_buff
*skb
,
1585 struct net_device
*dev
, struct in6_addr
*saddr
,
1586 struct in6_addr
*daddr
, __u8 prio
, __u8 ttl
,
1587 __be16 src_port
, __be16 dst_port
, __be32 vni
,
1590 struct ipv6hdr
*ip6h
;
1591 struct vxlanhdr
*vxh
;
1596 skb
= vxlan_handle_offloads(skb
, !udp_get_no_check6_tx(vs
->sock
->sk
));
1600 skb_scrub_packet(skb
, xnet
);
1602 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1603 + VXLAN_HLEN
+ sizeof(struct ipv6hdr
)
1604 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1606 /* Need space for new headers (invalidates iph ptr) */
1607 err
= skb_cow_head(skb
, min_headroom
);
1611 if (vlan_tx_tag_present(skb
)) {
1612 if (WARN_ON(!__vlan_put_tag(skb
,
1614 vlan_tx_tag_get(skb
))))
1620 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1621 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1624 __skb_push(skb
, sizeof(*uh
));
1625 skb_reset_transport_header(skb
);
1628 uh
->dest
= dst_port
;
1629 uh
->source
= src_port
;
1631 uh
->len
= htons(skb
->len
);
1633 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1634 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1636 skb_dst_set(skb
, dst
);
1638 udp6_set_csum(udp_get_no_check6_tx(vs
->sock
->sk
), skb
,
1639 saddr
, daddr
, skb
->len
);
1641 __skb_push(skb
, sizeof(*ip6h
));
1642 skb_reset_network_header(skb
);
1643 ip6h
= ipv6_hdr(skb
);
1645 ip6h
->priority
= prio
;
1646 ip6h
->flow_lbl
[0] = 0;
1647 ip6h
->flow_lbl
[1] = 0;
1648 ip6h
->flow_lbl
[2] = 0;
1649 ip6h
->payload_len
= htons(skb
->len
);
1650 ip6h
->nexthdr
= IPPROTO_UDP
;
1651 ip6h
->hop_limit
= ttl
;
1652 ip6h
->daddr
= *daddr
;
1653 ip6h
->saddr
= *saddr
;
1655 ip6tunnel_xmit(skb
, dev
);
1660 int vxlan_xmit_skb(struct vxlan_sock
*vs
,
1661 struct rtable
*rt
, struct sk_buff
*skb
,
1662 __be32 src
, __be32 dst
, __u8 tos
, __u8 ttl
, __be16 df
,
1663 __be16 src_port
, __be16 dst_port
, __be32 vni
, bool xnet
)
1665 struct vxlanhdr
*vxh
;
1670 skb
= vxlan_handle_offloads(skb
, !vs
->sock
->sk
->sk_no_check_tx
);
1674 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
1675 + VXLAN_HLEN
+ sizeof(struct iphdr
)
1676 + (vlan_tx_tag_present(skb
) ? VLAN_HLEN
: 0);
1678 /* Need space for new headers (invalidates iph ptr) */
1679 err
= skb_cow_head(skb
, min_headroom
);
1683 if (vlan_tx_tag_present(skb
)) {
1684 if (WARN_ON(!__vlan_put_tag(skb
,
1686 vlan_tx_tag_get(skb
))))
1692 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1693 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1696 __skb_push(skb
, sizeof(*uh
));
1697 skb_reset_transport_header(skb
);
1700 uh
->dest
= dst_port
;
1701 uh
->source
= src_port
;
1703 uh
->len
= htons(skb
->len
);
1705 udp_set_csum(vs
->sock
->sk
->sk_no_check_tx
, skb
,
1706 src
, dst
, skb
->len
);
1708 return iptunnel_xmit(vs
->sock
->sk
, rt
, skb
, src
, dst
, IPPROTO_UDP
,
1709 tos
, ttl
, df
, xnet
);
1711 EXPORT_SYMBOL_GPL(vxlan_xmit_skb
);
1713 /* Bypass encapsulation if the destination is local */
1714 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1715 struct vxlan_dev
*dst_vxlan
)
1717 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1718 union vxlan_addr loopback
;
1719 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1721 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1722 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1723 skb
->pkt_type
= PACKET_HOST
;
1724 skb
->encapsulation
= 0;
1725 skb
->dev
= dst_vxlan
->dev
;
1726 __skb_pull(skb
, skb_network_offset(skb
));
1728 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1729 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1730 loopback
.sa
.sa_family
= AF_INET
;
1731 #if IS_ENABLED(CONFIG_IPV6)
1733 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1734 loopback
.sa
.sa_family
= AF_INET6
;
1738 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1739 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1741 u64_stats_update_begin(&tx_stats
->syncp
);
1742 tx_stats
->tx_packets
++;
1743 tx_stats
->tx_bytes
+= skb
->len
;
1744 u64_stats_update_end(&tx_stats
->syncp
);
1746 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1747 u64_stats_update_begin(&rx_stats
->syncp
);
1748 rx_stats
->rx_packets
++;
1749 rx_stats
->rx_bytes
+= skb
->len
;
1750 u64_stats_update_end(&rx_stats
->syncp
);
1752 skb
->dev
->stats
.rx_dropped
++;
1756 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1757 struct vxlan_rdst
*rdst
, bool did_rsc
)
1759 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1760 struct rtable
*rt
= NULL
;
1761 const struct iphdr
*old_iph
;
1763 union vxlan_addr
*dst
;
1764 __be16 src_port
= 0, dst_port
;
1770 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1771 vni
= rdst
->remote_vni
;
1772 dst
= &rdst
->remote_ip
;
1774 if (vxlan_addr_any(dst
)) {
1776 /* short-circuited back to local bridge */
1777 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1783 old_iph
= ip_hdr(skb
);
1786 if (!ttl
&& vxlan_addr_multicast(dst
))
1791 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1793 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->port_min
,
1794 vxlan
->port_max
, true);
1796 if (dst
->sa
.sa_family
== AF_INET
) {
1797 memset(&fl4
, 0, sizeof(fl4
));
1798 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1799 fl4
.flowi4_tos
= RT_TOS(tos
);
1800 fl4
.daddr
= dst
->sin
.sin_addr
.s_addr
;
1801 fl4
.saddr
= vxlan
->saddr
.sin
.sin_addr
.s_addr
;
1803 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1805 netdev_dbg(dev
, "no route to %pI4\n",
1806 &dst
->sin
.sin_addr
.s_addr
);
1807 dev
->stats
.tx_carrier_errors
++;
1811 if (rt
->dst
.dev
== dev
) {
1812 netdev_dbg(dev
, "circular route to %pI4\n",
1813 &dst
->sin
.sin_addr
.s_addr
);
1814 dev
->stats
.collisions
++;
1818 /* Bypass encapsulation if the destination is local */
1819 if (rt
->rt_flags
& RTCF_LOCAL
&&
1820 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1821 struct vxlan_dev
*dst_vxlan
;
1824 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1827 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1831 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1832 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1834 err
= vxlan_xmit_skb(vxlan
->vn_sock
, rt
, skb
,
1835 fl4
.saddr
, dst
->sin
.sin_addr
.s_addr
,
1836 tos
, ttl
, df
, src_port
, dst_port
,
1838 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1842 iptunnel_xmit_stats(err
, &dev
->stats
, dev
->tstats
);
1843 #if IS_ENABLED(CONFIG_IPV6)
1845 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
1846 struct dst_entry
*ndst
;
1850 memset(&fl6
, 0, sizeof(fl6
));
1851 fl6
.flowi6_oif
= rdst
->remote_ifindex
;
1852 fl6
.daddr
= dst
->sin6
.sin6_addr
;
1853 fl6
.saddr
= vxlan
->saddr
.sin6
.sin6_addr
;
1854 fl6
.flowi6_proto
= IPPROTO_UDP
;
1856 if (ipv6_stub
->ipv6_dst_lookup(sk
, &ndst
, &fl6
)) {
1857 netdev_dbg(dev
, "no route to %pI6\n",
1858 &dst
->sin6
.sin6_addr
);
1859 dev
->stats
.tx_carrier_errors
++;
1863 if (ndst
->dev
== dev
) {
1864 netdev_dbg(dev
, "circular route to %pI6\n",
1865 &dst
->sin6
.sin6_addr
);
1867 dev
->stats
.collisions
++;
1871 /* Bypass encapsulation if the destination is local */
1872 flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
1873 if (flags
& RTF_LOCAL
&&
1874 !(flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1875 struct vxlan_dev
*dst_vxlan
;
1878 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
, dst_port
);
1881 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1885 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
1887 err
= vxlan6_xmit_skb(vxlan
->vn_sock
, ndst
, skb
,
1888 dev
, &fl6
.saddr
, &fl6
.daddr
, 0, ttl
,
1889 src_port
, dst_port
, htonl(vni
<< 8),
1890 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
)));
1897 dev
->stats
.tx_dropped
++;
1903 dev
->stats
.tx_errors
++;
1908 /* Transmit local packets over Vxlan
1910 * Outer IP header inherits ECN and DF from inner header.
1911 * Outer UDP destination is the VXLAN assigned port.
1912 * source port is based on hash of flow
1914 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1916 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1918 bool did_rsc
= false;
1919 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
1920 struct vxlan_fdb
*f
;
1922 skb_reset_mac_header(skb
);
1925 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
1926 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
1927 return arp_reduce(dev
, skb
);
1928 #if IS_ENABLED(CONFIG_IPV6)
1929 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
1930 skb
->len
>= sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
) &&
1931 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
1934 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1935 if (msg
->icmph
.icmp6_code
== 0 &&
1936 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
1937 return neigh_reduce(dev
, skb
);
1942 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1945 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1946 (ntohs(eth
->h_proto
) == ETH_P_IP
||
1947 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
1948 did_rsc
= route_shortcircuit(dev
, skb
);
1950 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1954 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
1956 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
1957 !is_multicast_ether_addr(eth
->h_dest
))
1958 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
1960 dev
->stats
.tx_dropped
++;
1962 return NETDEV_TX_OK
;
1966 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
1967 struct sk_buff
*skb1
;
1973 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1975 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
1979 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
1982 return NETDEV_TX_OK
;
1985 /* Walk the forwarding table and purge stale entries */
1986 static void vxlan_cleanup(unsigned long arg
)
1988 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1989 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1992 if (!netif_running(vxlan
->dev
))
1995 spin_lock_bh(&vxlan
->hash_lock
);
1996 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1997 struct hlist_node
*p
, *n
;
1998 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2000 = container_of(p
, struct vxlan_fdb
, hlist
);
2001 unsigned long timeout
;
2003 if (f
->state
& NUD_PERMANENT
)
2006 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
2007 if (time_before_eq(timeout
, jiffies
)) {
2008 netdev_dbg(vxlan
->dev
,
2009 "garbage collect %pM\n",
2011 f
->state
= NUD_STALE
;
2012 vxlan_fdb_destroy(vxlan
, f
);
2013 } else if (time_before(timeout
, next_timer
))
2014 next_timer
= timeout
;
2017 spin_unlock_bh(&vxlan
->hash_lock
);
2019 mod_timer(&vxlan
->age_timer
, next_timer
);
2022 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2024 __u32 vni
= vxlan
->default_dst
.remote_vni
;
2026 vxlan
->vn_sock
= vs
;
2027 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2030 /* Setup stats when device is created */
2031 static int vxlan_init(struct net_device
*dev
)
2033 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2034 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2035 struct vxlan_sock
*vs
;
2037 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2041 spin_lock(&vn
->sock_lock
);
2042 vs
= vxlan_find_sock(vxlan
->net
, vxlan
->dst_port
);
2044 /* If we have a socket with same port already, reuse it */
2045 atomic_inc(&vs
->refcnt
);
2046 vxlan_vs_add_dev(vs
, vxlan
);
2048 /* otherwise make new socket outside of RTNL */
2050 queue_work(vxlan_wq
, &vxlan
->sock_work
);
2052 spin_unlock(&vn
->sock_lock
);
2057 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2059 struct vxlan_fdb
*f
;
2061 spin_lock_bh(&vxlan
->hash_lock
);
2062 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2064 vxlan_fdb_destroy(vxlan
, f
);
2065 spin_unlock_bh(&vxlan
->hash_lock
);
2068 static void vxlan_uninit(struct net_device
*dev
)
2070 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2071 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2073 vxlan_fdb_delete_default(vxlan
);
2076 vxlan_sock_release(vs
);
2077 free_percpu(dev
->tstats
);
2080 /* Start ageing timer and join group when device is brought up */
2081 static int vxlan_open(struct net_device
*dev
)
2083 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2084 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2086 /* socket hasn't been created */
2090 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2091 vxlan_sock_hold(vs
);
2093 queue_work(vxlan_wq
, &vxlan
->igmp_join
);
2096 if (vxlan
->age_interval
)
2097 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2102 /* Purge the forwarding table */
2103 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2107 spin_lock_bh(&vxlan
->hash_lock
);
2108 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2109 struct hlist_node
*p
, *n
;
2110 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2112 = container_of(p
, struct vxlan_fdb
, hlist
);
2113 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2114 if (!is_zero_ether_addr(f
->eth_addr
))
2115 vxlan_fdb_destroy(vxlan
, f
);
2118 spin_unlock_bh(&vxlan
->hash_lock
);
2121 /* Cleanup timer and forwarding table on shutdown */
2122 static int vxlan_stop(struct net_device
*dev
)
2124 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2125 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2126 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
2128 if (vs
&& vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2129 !vxlan_group_used(vn
, vxlan
)) {
2130 vxlan_sock_hold(vs
);
2132 queue_work(vxlan_wq
, &vxlan
->igmp_leave
);
2135 del_timer_sync(&vxlan
->age_timer
);
2142 /* Stub, nothing needs to be done. */
2143 static void vxlan_set_multicast_list(struct net_device
*dev
)
2147 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2149 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2150 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2151 struct net_device
*lowerdev
;
2154 lowerdev
= __dev_get_by_index(vxlan
->net
, dst
->remote_ifindex
);
2155 if (lowerdev
== NULL
)
2156 return eth_change_mtu(dev
, new_mtu
);
2158 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2159 max_mtu
= lowerdev
->mtu
- VXLAN6_HEADROOM
;
2161 max_mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
2163 if (new_mtu
< 68 || new_mtu
> max_mtu
)
2170 static const struct net_device_ops vxlan_netdev_ops
= {
2171 .ndo_init
= vxlan_init
,
2172 .ndo_uninit
= vxlan_uninit
,
2173 .ndo_open
= vxlan_open
,
2174 .ndo_stop
= vxlan_stop
,
2175 .ndo_start_xmit
= vxlan_xmit
,
2176 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2177 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2178 .ndo_change_mtu
= vxlan_change_mtu
,
2179 .ndo_validate_addr
= eth_validate_addr
,
2180 .ndo_set_mac_address
= eth_mac_addr
,
2181 .ndo_fdb_add
= vxlan_fdb_add
,
2182 .ndo_fdb_del
= vxlan_fdb_delete
,
2183 .ndo_fdb_dump
= vxlan_fdb_dump
,
2186 /* Info for udev, that this is a virtual tunnel endpoint */
2187 static struct device_type vxlan_type
= {
2191 /* Calls the ndo_add_vxlan_port of the caller in order to
2192 * supply the listening VXLAN udp ports. Callers are expected
2193 * to implement the ndo_add_vxlan_port.
2195 void vxlan_get_rx_port(struct net_device
*dev
)
2197 struct vxlan_sock
*vs
;
2198 struct net
*net
= dev_net(dev
);
2199 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2200 sa_family_t sa_family
;
2204 spin_lock(&vn
->sock_lock
);
2205 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2206 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2207 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2208 sa_family
= vs
->sock
->sk
->sk_family
;
2209 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2213 spin_unlock(&vn
->sock_lock
);
2215 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2217 /* Initialize the device structure. */
2218 static void vxlan_setup(struct net_device
*dev
)
2220 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2223 eth_hw_addr_random(dev
);
2225 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
)
2226 dev
->needed_headroom
= ETH_HLEN
+ VXLAN6_HEADROOM
;
2228 dev
->needed_headroom
= ETH_HLEN
+ VXLAN_HEADROOM
;
2230 dev
->netdev_ops
= &vxlan_netdev_ops
;
2231 dev
->destructor
= free_netdev
;
2232 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2234 dev
->tx_queue_len
= 0;
2235 dev
->features
|= NETIF_F_LLTX
;
2236 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2237 dev
->features
|= NETIF_F_RXCSUM
;
2238 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2240 dev
->vlan_features
= dev
->features
;
2241 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2242 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2243 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2244 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2245 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
2246 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2248 INIT_LIST_HEAD(&vxlan
->next
);
2249 spin_lock_init(&vxlan
->hash_lock
);
2250 INIT_WORK(&vxlan
->igmp_join
, vxlan_igmp_join
);
2251 INIT_WORK(&vxlan
->igmp_leave
, vxlan_igmp_leave
);
2252 INIT_WORK(&vxlan
->sock_work
, vxlan_sock_work
);
2254 init_timer_deferrable(&vxlan
->age_timer
);
2255 vxlan
->age_timer
.function
= vxlan_cleanup
;
2256 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2258 vxlan
->dst_port
= htons(vxlan_port
);
2262 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2263 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2266 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2267 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2268 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2269 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2270 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2271 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2272 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2273 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2274 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2275 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2276 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2277 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2278 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2279 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2280 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2281 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2282 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2283 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2286 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2288 if (tb
[IFLA_ADDRESS
]) {
2289 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2290 pr_debug("invalid link address (not ethernet)\n");
2294 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2295 pr_debug("invalid all zero ethernet address\n");
2296 return -EADDRNOTAVAIL
;
2303 if (data
[IFLA_VXLAN_ID
]) {
2304 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2305 if (id
>= VXLAN_VID_MASK
)
2309 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2310 const struct ifla_vxlan_port_range
*p
2311 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2313 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2314 pr_debug("port range %u .. %u not valid\n",
2315 ntohs(p
->low
), ntohs(p
->high
));
2323 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2324 struct ethtool_drvinfo
*drvinfo
)
2326 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2327 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2330 static const struct ethtool_ops vxlan_ethtool_ops
= {
2331 .get_drvinfo
= vxlan_get_drvinfo
,
2332 .get_link
= ethtool_op_get_link
,
2335 static void vxlan_del_work(struct work_struct
*work
)
2337 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2339 sk_release_kernel(vs
->sock
->sk
);
2343 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2344 __be16 port
, u32 flags
)
2346 struct socket
*sock
;
2347 struct udp_port_cfg udp_conf
;
2350 memset(&udp_conf
, 0, sizeof(udp_conf
));
2353 udp_conf
.family
= AF_INET6
;
2354 udp_conf
.use_udp6_tx_checksums
=
2355 !!(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2356 udp_conf
.use_udp6_rx_checksums
=
2357 !!(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2359 udp_conf
.family
= AF_INET
;
2360 udp_conf
.local_ip
.s_addr
= INADDR_ANY
;
2361 udp_conf
.use_udp_checksums
=
2362 !!(flags
& VXLAN_F_UDP_CSUM
);
2365 udp_conf
.local_udp_port
= port
;
2367 /* Open UDP socket */
2368 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2370 return ERR_PTR(err
);
2372 /* Disable multicast loopback */
2373 inet_sk(sock
->sk
)->mc_loop
= 0;
2378 /* Create new listen socket if needed */
2379 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
,
2380 vxlan_rcv_t
*rcv
, void *data
,
2383 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2384 struct vxlan_sock
*vs
;
2385 struct socket
*sock
;
2388 bool ipv6
= !!(flags
& VXLAN_F_IPV6
);
2390 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2392 return ERR_PTR(-ENOMEM
);
2394 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2395 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2397 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2399 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2402 return ERR_CAST(sock
);
2407 atomic_set(&vs
->refcnt
, 1);
2410 rcu_assign_sk_user_data(vs
->sock
->sk
, vs
);
2412 /* Initialize the vxlan udp offloads structure */
2413 vs
->udp_offloads
.port
= port
;
2414 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2415 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2417 spin_lock(&vn
->sock_lock
);
2418 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2419 vxlan_notify_add_rx_port(vs
);
2420 spin_unlock(&vn
->sock_lock
);
2422 /* Mark socket as an encapsulation socket. */
2423 udp_sk(sk
)->encap_type
= 1;
2424 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
2425 #if IS_ENABLED(CONFIG_IPV6)
2427 ipv6_stub
->udpv6_encap_enable();
2435 struct vxlan_sock
*vxlan_sock_add(struct net
*net
, __be16 port
,
2436 vxlan_rcv_t
*rcv
, void *data
,
2437 bool no_share
, u32 flags
)
2439 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2440 struct vxlan_sock
*vs
;
2442 vs
= vxlan_socket_create(net
, port
, rcv
, data
, flags
);
2446 if (no_share
) /* Return error if sharing is not allowed. */
2449 spin_lock(&vn
->sock_lock
);
2450 vs
= vxlan_find_sock(net
, port
);
2453 atomic_inc(&vs
->refcnt
);
2455 vs
= ERR_PTR(-EBUSY
);
2457 spin_unlock(&vn
->sock_lock
);
2460 vs
= ERR_PTR(-EINVAL
);
2464 EXPORT_SYMBOL_GPL(vxlan_sock_add
);
2466 /* Scheduled at device creation to bind to a socket */
2467 static void vxlan_sock_work(struct work_struct
*work
)
2469 struct vxlan_dev
*vxlan
= container_of(work
, struct vxlan_dev
, sock_work
);
2470 struct net
*net
= vxlan
->net
;
2471 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2472 __be16 port
= vxlan
->dst_port
;
2473 struct vxlan_sock
*nvs
;
2475 nvs
= vxlan_sock_add(net
, port
, vxlan_rcv
, NULL
, false, vxlan
->flags
);
2476 spin_lock(&vn
->sock_lock
);
2478 vxlan_vs_add_dev(nvs
, vxlan
);
2479 spin_unlock(&vn
->sock_lock
);
2481 dev_put(vxlan
->dev
);
2484 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
2485 struct nlattr
*tb
[], struct nlattr
*data
[])
2487 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2488 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2489 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2492 bool use_ipv6
= false;
2494 if (!data
[IFLA_VXLAN_ID
])
2497 vxlan
->net
= dev_net(dev
);
2499 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2500 dst
->remote_vni
= vni
;
2502 /* Unless IPv6 is explicitly requested, assume IPv4 */
2503 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2504 if (data
[IFLA_VXLAN_GROUP
]) {
2505 dst
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
2506 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2507 if (!IS_ENABLED(CONFIG_IPV6
))
2508 return -EPFNOSUPPORT
;
2510 nla_memcpy(&dst
->remote_ip
.sin6
.sin6_addr
, data
[IFLA_VXLAN_GROUP6
],
2511 sizeof(struct in6_addr
));
2512 dst
->remote_ip
.sa
.sa_family
= AF_INET6
;
2516 if (data
[IFLA_VXLAN_LOCAL
]) {
2517 vxlan
->saddr
.sin
.sin_addr
.s_addr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
2518 vxlan
->saddr
.sa
.sa_family
= AF_INET
;
2519 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2520 if (!IS_ENABLED(CONFIG_IPV6
))
2521 return -EPFNOSUPPORT
;
2523 /* TODO: respect scope id */
2524 nla_memcpy(&vxlan
->saddr
.sin6
.sin6_addr
, data
[IFLA_VXLAN_LOCAL6
],
2525 sizeof(struct in6_addr
));
2526 vxlan
->saddr
.sa
.sa_family
= AF_INET6
;
2530 if (data
[IFLA_VXLAN_LINK
] &&
2531 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
2532 struct net_device
*lowerdev
2533 = __dev_get_by_index(net
, dst
->remote_ifindex
);
2536 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2540 #if IS_ENABLED(CONFIG_IPV6)
2542 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2543 if (idev
&& idev
->cnf
.disable_ipv6
) {
2544 pr_info("IPv6 is disabled via sysctl\n");
2547 vxlan
->flags
|= VXLAN_F_IPV6
;
2552 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2554 dev
->needed_headroom
= lowerdev
->hard_header_len
+
2555 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2556 } else if (use_ipv6
)
2557 vxlan
->flags
|= VXLAN_F_IPV6
;
2559 if (data
[IFLA_VXLAN_TOS
])
2560 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2562 if (data
[IFLA_VXLAN_TTL
])
2563 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2565 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2566 vxlan
->flags
|= VXLAN_F_LEARN
;
2568 if (data
[IFLA_VXLAN_AGEING
])
2569 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2571 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
2573 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2574 vxlan
->flags
|= VXLAN_F_PROXY
;
2576 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2577 vxlan
->flags
|= VXLAN_F_RSC
;
2579 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2580 vxlan
->flags
|= VXLAN_F_L2MISS
;
2582 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2583 vxlan
->flags
|= VXLAN_F_L3MISS
;
2585 if (data
[IFLA_VXLAN_LIMIT
])
2586 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2588 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2589 const struct ifla_vxlan_port_range
*p
2590 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2591 vxlan
->port_min
= ntohs(p
->low
);
2592 vxlan
->port_max
= ntohs(p
->high
);
2595 if (data
[IFLA_VXLAN_PORT
])
2596 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2598 if (data
[IFLA_VXLAN_UDP_CSUM
] && nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2599 vxlan
->flags
|= VXLAN_F_UDP_CSUM
;
2601 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2602 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2603 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2605 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2606 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2607 vxlan
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2609 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
2610 pr_info("duplicate VNI %u\n", vni
);
2614 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2616 /* create an fdb entry for a valid default destination */
2617 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2618 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2619 &vxlan
->default_dst
.remote_ip
,
2620 NUD_REACHABLE
|NUD_PERMANENT
,
2621 NLM_F_EXCL
|NLM_F_CREATE
,
2623 vxlan
->default_dst
.remote_vni
,
2624 vxlan
->default_dst
.remote_ifindex
,
2630 err
= register_netdevice(dev
);
2632 vxlan_fdb_delete_default(vxlan
);
2636 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2641 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2643 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2644 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2646 spin_lock(&vn
->sock_lock
);
2647 if (!hlist_unhashed(&vxlan
->hlist
))
2648 hlist_del_rcu(&vxlan
->hlist
);
2649 spin_unlock(&vn
->sock_lock
);
2651 list_del(&vxlan
->next
);
2652 unregister_netdevice_queue(dev
, head
);
2655 static size_t vxlan_get_size(const struct net_device
*dev
)
2658 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2659 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2660 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2661 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2662 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2663 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2664 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2665 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2666 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2667 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2668 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2669 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2670 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2671 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2672 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2673 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
2674 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2675 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2679 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
2681 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2682 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2683 struct ifla_vxlan_port_range ports
= {
2684 .low
= htons(vxlan
->port_min
),
2685 .high
= htons(vxlan
->port_max
),
2688 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
2689 goto nla_put_failure
;
2691 if (!vxlan_addr_any(&dst
->remote_ip
)) {
2692 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
2693 if (nla_put_be32(skb
, IFLA_VXLAN_GROUP
,
2694 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
2695 goto nla_put_failure
;
2696 #if IS_ENABLED(CONFIG_IPV6)
2698 if (nla_put(skb
, IFLA_VXLAN_GROUP6
, sizeof(struct in6_addr
),
2699 &dst
->remote_ip
.sin6
.sin6_addr
))
2700 goto nla_put_failure
;
2705 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
2706 goto nla_put_failure
;
2708 if (!vxlan_addr_any(&vxlan
->saddr
)) {
2709 if (vxlan
->saddr
.sa
.sa_family
== AF_INET
) {
2710 if (nla_put_be32(skb
, IFLA_VXLAN_LOCAL
,
2711 vxlan
->saddr
.sin
.sin_addr
.s_addr
))
2712 goto nla_put_failure
;
2713 #if IS_ENABLED(CONFIG_IPV6)
2715 if (nla_put(skb
, IFLA_VXLAN_LOCAL6
, sizeof(struct in6_addr
),
2716 &vxlan
->saddr
.sin6
.sin6_addr
))
2717 goto nla_put_failure
;
2722 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
2723 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
2724 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
2725 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
2726 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
2727 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
2728 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
2729 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
2730 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
2731 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
2732 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
2733 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
2734 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
2735 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
) ||
2736 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
2737 !!(vxlan
->flags
& VXLAN_F_UDP_CSUM
)) ||
2738 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
2739 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
2740 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
2741 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)))
2742 goto nla_put_failure
;
2744 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
2745 goto nla_put_failure
;
2753 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
2755 .maxtype
= IFLA_VXLAN_MAX
,
2756 .policy
= vxlan_policy
,
2757 .priv_size
= sizeof(struct vxlan_dev
),
2758 .setup
= vxlan_setup
,
2759 .validate
= vxlan_validate
,
2760 .newlink
= vxlan_newlink
,
2761 .dellink
= vxlan_dellink
,
2762 .get_size
= vxlan_get_size
,
2763 .fill_info
= vxlan_fill_info
,
2766 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
2767 struct net_device
*dev
)
2769 struct vxlan_dev
*vxlan
, *next
;
2770 LIST_HEAD(list_kill
);
2772 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2773 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2775 /* In case we created vxlan device with carrier
2776 * and we loose the carrier due to module unload
2777 * we also need to remove vxlan device. In other
2778 * cases, it's not necessary and remote_ifindex
2779 * is 0 here, so no matches.
2781 if (dst
->remote_ifindex
== dev
->ifindex
)
2782 vxlan_dellink(vxlan
->dev
, &list_kill
);
2785 unregister_netdevice_many(&list_kill
);
2788 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
2789 unsigned long event
, void *ptr
)
2791 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2792 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
2794 if (event
== NETDEV_UNREGISTER
)
2795 vxlan_handle_lowerdev_unregister(vn
, dev
);
2800 static struct notifier_block vxlan_notifier_block __read_mostly
= {
2801 .notifier_call
= vxlan_lowerdev_event
,
2804 static __net_init
int vxlan_init_net(struct net
*net
)
2806 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2809 INIT_LIST_HEAD(&vn
->vxlan_list
);
2810 spin_lock_init(&vn
->sock_lock
);
2812 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
2813 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
2818 static void __net_exit
vxlan_exit_net(struct net
*net
)
2820 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2821 struct vxlan_dev
*vxlan
, *next
;
2822 struct net_device
*dev
, *aux
;
2826 for_each_netdev_safe(net
, dev
, aux
)
2827 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
2828 unregister_netdevice_queue(dev
, &list
);
2830 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
2831 /* If vxlan->dev is in the same netns, it has already been added
2832 * to the list by the previous loop.
2834 if (!net_eq(dev_net(vxlan
->dev
), net
))
2835 unregister_netdevice_queue(dev
, &list
);
2838 unregister_netdevice_many(&list
);
2842 static struct pernet_operations vxlan_net_ops
= {
2843 .init
= vxlan_init_net
,
2844 .exit
= vxlan_exit_net
,
2845 .id
= &vxlan_net_id
,
2846 .size
= sizeof(struct vxlan_net
),
2849 static int __init
vxlan_init_module(void)
2853 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
2857 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
2859 rc
= register_pernet_subsys(&vxlan_net_ops
);
2863 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
2867 rc
= rtnl_link_register(&vxlan_link_ops
);
2873 unregister_netdevice_notifier(&vxlan_notifier_block
);
2875 unregister_pernet_subsys(&vxlan_net_ops
);
2877 destroy_workqueue(vxlan_wq
);
2880 late_initcall(vxlan_init_module
);
2882 static void __exit
vxlan_cleanup_module(void)
2884 rtnl_link_unregister(&vxlan_link_ops
);
2885 unregister_netdevice_notifier(&vxlan_notifier_block
);
2886 destroy_workqueue(vxlan_wq
);
2887 unregister_pernet_subsys(&vxlan_net_ops
);
2888 /* rcu_barrier() is called by netns */
2890 module_exit(vxlan_cleanup_module
);
2892 MODULE_LICENSE("GPL");
2893 MODULE_VERSION(VXLAN_VERSION
);
2894 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2895 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2896 MODULE_ALIAS_RTNL_LINK("vxlan");