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/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/if_ether.h>
20 #include <linux/ethtool.h>
22 #include <net/ndisc.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/vxlan.h>
31 #if IS_ENABLED(CONFIG_IPV6)
32 #include <net/ip6_tunnel.h>
33 #include <net/ip6_checksum.h>
36 #define VXLAN_VERSION "0.1"
38 #define PORT_HASH_BITS 8
39 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
40 #define FDB_AGE_DEFAULT 300 /* 5 min */
41 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43 /* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
45 * for compatibility with early adopters.
47 static unsigned short vxlan_port __read_mostly
= 8472;
48 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
49 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
51 static bool log_ecn_error
= true;
52 module_param(log_ecn_error
, bool, 0644);
53 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
55 static unsigned int vxlan_net_id
;
56 static struct rtnl_link_ops vxlan_link_ops
;
58 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
60 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
62 /* per-network namespace private data for this module */
64 struct list_head vxlan_list
;
65 struct hlist_head sock_list
[PORT_HASH_SIZE
];
69 /* Forwarding table entry */
71 struct hlist_node hlist
; /* linked list of entries */
73 unsigned long updated
; /* jiffies */
75 struct list_head remotes
;
76 u8 eth_addr
[ETH_ALEN
];
77 u16 state
; /* see ndm_state */
79 u8 flags
; /* see ndm_flags */
82 /* salt for hash table */
83 static u32 vxlan_salt __read_mostly
;
85 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
87 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
88 ip_tunnel_collect_metadata();
91 #if IS_ENABLED(CONFIG_IPV6)
93 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
95 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
97 if (a
->sa
.sa_family
== AF_INET6
)
98 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
100 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
103 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
105 if (ipa
->sa
.sa_family
== AF_INET6
)
106 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
108 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
111 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
113 if (ipa
->sa
.sa_family
== AF_INET6
)
114 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
116 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
119 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
121 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
122 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
123 ip
->sa
.sa_family
= AF_INET6
;
125 } else if (nla_len(nla
) >= sizeof(__be32
)) {
126 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
127 ip
->sa
.sa_family
= AF_INET
;
130 return -EAFNOSUPPORT
;
134 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
135 const union vxlan_addr
*ip
)
137 if (ip
->sa
.sa_family
== AF_INET6
)
138 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
140 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
143 #else /* !CONFIG_IPV6 */
146 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
148 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
151 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
153 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
156 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
158 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
161 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
163 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
164 return -EAFNOSUPPORT
;
165 } else if (nla_len(nla
) >= sizeof(__be32
)) {
166 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
167 ip
->sa
.sa_family
= AF_INET
;
170 return -EAFNOSUPPORT
;
174 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
175 const union vxlan_addr
*ip
)
177 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
181 /* Virtual Network hash table head */
182 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
184 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
187 /* Socket hash table head */
188 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
190 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
192 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
195 /* First remote destination for a forwarding entry.
196 * Guaranteed to be non-NULL because remotes are never deleted.
198 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
200 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
203 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
205 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
208 /* Find VXLAN socket based on network namespace, address family and UDP port
209 * and enabled unshareable flags.
211 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
212 __be16 port
, u32 flags
)
214 struct vxlan_sock
*vs
;
216 flags
&= VXLAN_F_RCV_FLAGS
;
218 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
219 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
220 vxlan_get_sk_family(vs
) == family
&&
227 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, __be32 vni
)
229 struct vxlan_dev
*vxlan
;
231 /* For flow based devices, map all packets to VNI 0 */
232 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
235 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, vni
), hlist
) {
236 if (vxlan
->default_dst
.remote_vni
== vni
)
243 /* Look up VNI in a per net namespace table */
244 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, __be32 vni
,
245 sa_family_t family
, __be16 port
,
248 struct vxlan_sock
*vs
;
250 vs
= vxlan_find_sock(net
, family
, port
, flags
);
254 return vxlan_vs_find_vni(vs
, vni
);
257 /* Fill in neighbour message in skbuff. */
258 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
259 const struct vxlan_fdb
*fdb
,
260 u32 portid
, u32 seq
, int type
, unsigned int flags
,
261 const struct vxlan_rdst
*rdst
)
263 unsigned long now
= jiffies
;
264 struct nda_cacheinfo ci
;
265 struct nlmsghdr
*nlh
;
267 bool send_ip
, send_eth
;
269 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
273 ndm
= nlmsg_data(nlh
);
274 memset(ndm
, 0, sizeof(*ndm
));
276 send_eth
= send_ip
= true;
278 if (type
== RTM_GETNEIGH
) {
279 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
280 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
281 ndm
->ndm_family
= send_ip
? rdst
->remote_ip
.sa
.sa_family
: AF_INET
;
283 ndm
->ndm_family
= AF_BRIDGE
;
284 ndm
->ndm_state
= fdb
->state
;
285 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
286 ndm
->ndm_flags
= fdb
->flags
;
287 ndm
->ndm_type
= RTN_UNICAST
;
289 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
290 nla_put_s32(skb
, NDA_LINK_NETNSID
,
291 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
292 goto nla_put_failure
;
294 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
295 goto nla_put_failure
;
297 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
298 goto nla_put_failure
;
300 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
301 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
302 goto nla_put_failure
;
303 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
304 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
305 goto nla_put_failure
;
306 if ((vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) && fdb
->vni
&&
307 nla_put_u32(skb
, NDA_SRC_VNI
,
308 be32_to_cpu(fdb
->vni
)))
309 goto nla_put_failure
;
310 if (rdst
->remote_ifindex
&&
311 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
312 goto nla_put_failure
;
314 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
315 ci
.ndm_confirmed
= 0;
316 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
319 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
320 goto nla_put_failure
;
326 nlmsg_cancel(skb
, nlh
);
330 static inline size_t vxlan_nlmsg_size(void)
332 return NLMSG_ALIGN(sizeof(struct ndmsg
))
333 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
334 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
335 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
336 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
337 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
338 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
339 + nla_total_size(sizeof(struct nda_cacheinfo
));
342 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
343 struct vxlan_rdst
*rd
, int type
)
345 struct net
*net
= dev_net(vxlan
->dev
);
349 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
353 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
355 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
356 WARN_ON(err
== -EMSGSIZE
);
361 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
365 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
368 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
370 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
371 struct vxlan_fdb f
= {
374 struct vxlan_rdst remote
= {
375 .remote_ip
= *ipa
, /* goes to NDA_DST */
376 .remote_vni
= cpu_to_be32(VXLAN_N_VID
),
379 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
382 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
384 struct vxlan_fdb f
= {
387 struct vxlan_rdst remote
= { };
389 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
391 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
394 /* Hash Ethernet address */
395 static u32
eth_hash(const unsigned char *addr
)
397 u64 value
= get_unaligned((u64
*)addr
);
399 /* only want 6 bytes */
405 return hash_64(value
, FDB_HASH_BITS
);
408 static u32
eth_vni_hash(const unsigned char *addr
, __be32 vni
)
410 /* use 1 byte of OUI and 3 bytes of NIC */
411 u32 key
= get_unaligned((u32
*)(addr
+ 2));
413 return jhash_2words(key
, vni
, vxlan_salt
) & (FDB_HASH_SIZE
- 1);
416 /* Hash chain to use given mac address */
417 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
418 const u8
*mac
, __be32 vni
)
420 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)
421 return &vxlan
->fdb_head
[eth_vni_hash(mac
, vni
)];
423 return &vxlan
->fdb_head
[eth_hash(mac
)];
426 /* Look up Ethernet address in forwarding table */
427 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
428 const u8
*mac
, __be32 vni
)
430 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
, vni
);
433 hlist_for_each_entry_rcu(f
, head
, hlist
) {
434 if (ether_addr_equal(mac
, f
->eth_addr
)) {
435 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
447 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
448 const u8
*mac
, __be32 vni
)
452 f
= __vxlan_find_mac(vxlan
, mac
, vni
);
459 /* caller should hold vxlan->hash_lock */
460 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
461 union vxlan_addr
*ip
, __be16 port
,
462 __be32 vni
, __u32 ifindex
)
464 struct vxlan_rdst
*rd
;
466 list_for_each_entry(rd
, &f
->remotes
, list
) {
467 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
468 rd
->remote_port
== port
&&
469 rd
->remote_vni
== vni
&&
470 rd
->remote_ifindex
== ifindex
)
477 /* Replace destination of unicast mac */
478 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
479 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
482 struct vxlan_rdst
*rd
;
484 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
488 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
492 dst_cache_reset(&rd
->dst_cache
);
494 rd
->remote_port
= port
;
495 rd
->remote_vni
= vni
;
496 rd
->remote_ifindex
= ifindex
;
500 /* Add/update destinations for multicast */
501 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
502 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
503 __u32 ifindex
, struct vxlan_rdst
**rdp
)
505 struct vxlan_rdst
*rd
;
507 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
511 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
515 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
521 rd
->remote_port
= port
;
522 rd
->remote_vni
= vni
;
523 rd
->remote_ifindex
= ifindex
;
525 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
531 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
533 struct vxlanhdr
*vh
, size_t hdrlen
,
535 struct gro_remcsum
*grc
,
538 size_t start
, offset
;
540 if (skb
->remcsum_offload
)
543 if (!NAPI_GRO_CB(skb
)->csum_valid
)
546 start
= vxlan_rco_start(vni_field
);
547 offset
= start
+ vxlan_rco_offset(vni_field
);
549 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
550 start
, offset
, grc
, nopartial
);
552 skb
->remcsum_offload
= 1;
557 static struct sk_buff
**vxlan_gro_receive(struct sock
*sk
,
558 struct sk_buff
**head
,
561 struct sk_buff
*p
, **pp
= NULL
;
562 struct vxlanhdr
*vh
, *vh2
;
563 unsigned int hlen
, off_vx
;
565 struct vxlan_sock
*vs
= rcu_dereference_sk_user_data(sk
);
567 struct gro_remcsum grc
;
569 skb_gro_remcsum_init(&grc
);
571 off_vx
= skb_gro_offset(skb
);
572 hlen
= off_vx
+ sizeof(*vh
);
573 vh
= skb_gro_header_fast(skb
, off_vx
);
574 if (skb_gro_header_hard(skb
, hlen
)) {
575 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
580 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
582 flags
= vh
->vx_flags
;
584 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
585 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
588 VXLAN_F_REMCSUM_NOPARTIAL
));
594 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
596 for (p
= *head
; p
; p
= p
->next
) {
597 if (!NAPI_GRO_CB(p
)->same_flow
)
600 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
601 if (vh
->vx_flags
!= vh2
->vx_flags
||
602 vh
->vx_vni
!= vh2
->vx_vni
) {
603 NAPI_GRO_CB(p
)->same_flow
= 0;
608 pp
= call_gro_receive(eth_gro_receive
, head
, skb
);
612 skb_gro_remcsum_cleanup(skb
, &grc
);
613 NAPI_GRO_CB(skb
)->flush
|= flush
;
618 static int vxlan_gro_complete(struct sock
*sk
, struct sk_buff
*skb
, int nhoff
)
620 /* Sets 'skb->inner_mac_header' since we are always called with
621 * 'skb->encapsulation' set.
623 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
626 /* Add new entry to forwarding table -- assumes lock held */
627 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
628 const u8
*mac
, union vxlan_addr
*ip
,
629 __u16 state
, __u16 flags
,
630 __be16 port
, __be32 src_vni
, __be32 vni
,
631 __u32 ifindex
, __u8 ndm_flags
)
633 struct vxlan_rdst
*rd
= NULL
;
638 f
= __vxlan_find_mac(vxlan
, mac
, src_vni
);
640 if (flags
& NLM_F_EXCL
) {
641 netdev_dbg(vxlan
->dev
,
642 "lost race to create %pM\n", mac
);
645 if (f
->state
!= state
) {
647 f
->updated
= jiffies
;
650 if (f
->flags
!= ndm_flags
) {
651 f
->flags
= ndm_flags
;
652 f
->updated
= jiffies
;
655 if ((flags
& NLM_F_REPLACE
)) {
656 /* Only change unicasts */
657 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
658 is_zero_ether_addr(f
->eth_addr
))) {
659 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
664 if ((flags
& NLM_F_APPEND
) &&
665 (is_multicast_ether_addr(f
->eth_addr
) ||
666 is_zero_ether_addr(f
->eth_addr
))) {
667 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
674 if (!(flags
& NLM_F_CREATE
))
677 if (vxlan
->cfg
.addrmax
&&
678 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
681 /* Disallow replace to add a multicast entry */
682 if ((flags
& NLM_F_REPLACE
) &&
683 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
686 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
687 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
693 f
->flags
= ndm_flags
;
694 f
->updated
= f
->used
= jiffies
;
696 INIT_LIST_HEAD(&f
->remotes
);
697 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
699 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
706 hlist_add_head_rcu(&f
->hlist
,
707 vxlan_fdb_head(vxlan
, mac
, src_vni
));
712 rd
= first_remote_rtnl(f
);
713 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
719 static void vxlan_fdb_free(struct rcu_head
*head
)
721 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
722 struct vxlan_rdst
*rd
, *nd
;
724 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
725 dst_cache_destroy(&rd
->dst_cache
);
731 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
733 netdev_dbg(vxlan
->dev
,
734 "delete %pM\n", f
->eth_addr
);
737 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
739 hlist_del_rcu(&f
->hlist
);
740 call_rcu(&f
->rcu
, vxlan_fdb_free
);
743 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
744 union vxlan_addr
*ip
, __be16
*port
, __be32
*src_vni
,
745 __be32
*vni
, u32
*ifindex
)
747 struct net
*net
= dev_net(vxlan
->dev
);
751 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
755 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
756 if (remote
->sa
.sa_family
== AF_INET
) {
757 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
758 ip
->sa
.sa_family
= AF_INET
;
759 #if IS_ENABLED(CONFIG_IPV6)
761 ip
->sin6
.sin6_addr
= in6addr_any
;
762 ip
->sa
.sa_family
= AF_INET6
;
768 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
770 *port
= nla_get_be16(tb
[NDA_PORT
]);
772 *port
= vxlan
->cfg
.dst_port
;
776 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
778 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
780 *vni
= vxlan
->default_dst
.remote_vni
;
783 if (tb
[NDA_SRC_VNI
]) {
784 if (nla_len(tb
[NDA_SRC_VNI
]) != sizeof(u32
))
786 *src_vni
= cpu_to_be32(nla_get_u32(tb
[NDA_SRC_VNI
]));
788 *src_vni
= vxlan
->default_dst
.remote_vni
;
791 if (tb
[NDA_IFINDEX
]) {
792 struct net_device
*tdev
;
794 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
796 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
797 tdev
= __dev_get_by_index(net
, *ifindex
);
799 return -EADDRNOTAVAIL
;
807 /* Add static entry (via netlink) */
808 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
809 struct net_device
*dev
,
810 const unsigned char *addr
, u16 vid
, u16 flags
)
812 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
813 /* struct net *net = dev_net(vxlan->dev); */
820 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
821 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
826 if (tb
[NDA_DST
] == NULL
)
829 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
833 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
834 return -EAFNOSUPPORT
;
836 spin_lock_bh(&vxlan
->hash_lock
);
837 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
838 port
, src_vni
, vni
, ifindex
, ndm
->ndm_flags
);
839 spin_unlock_bh(&vxlan
->hash_lock
);
844 static int __vxlan_fdb_delete(struct vxlan_dev
*vxlan
,
845 const unsigned char *addr
, union vxlan_addr ip
,
846 __be16 port
, __be32 src_vni
, u32 vni
, u32 ifindex
,
850 struct vxlan_rdst
*rd
= NULL
;
853 f
= vxlan_find_mac(vxlan
, addr
, src_vni
);
857 if (!vxlan_addr_any(&ip
)) {
858 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
863 /* remove a destination if it's not the only one on the list,
864 * otherwise destroy the fdb entry
866 if (rd
&& !list_is_singular(&f
->remotes
)) {
867 list_del_rcu(&rd
->list
);
868 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
873 vxlan_fdb_destroy(vxlan
, f
);
879 /* Delete entry (via netlink) */
880 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
881 struct net_device
*dev
,
882 const unsigned char *addr
, u16 vid
)
884 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
891 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
895 spin_lock_bh(&vxlan
->hash_lock
);
896 err
= __vxlan_fdb_delete(vxlan
, addr
, ip
, port
, src_vni
, vni
, ifindex
,
898 spin_unlock_bh(&vxlan
->hash_lock
);
903 /* Dump forwarding table */
904 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
905 struct net_device
*dev
,
906 struct net_device
*filter_dev
, int *idx
)
908 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
912 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
915 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
916 struct vxlan_rdst
*rd
;
918 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
919 if (*idx
< cb
->args
[2])
922 err
= vxlan_fdb_info(skb
, vxlan
, f
,
923 NETLINK_CB(cb
->skb
).portid
,
938 /* Watch incoming packets to learn mapping between Ethernet address
939 * and Tunnel endpoint.
940 * Return true if packet is bogus and should be dropped.
942 static bool vxlan_snoop(struct net_device
*dev
,
943 union vxlan_addr
*src_ip
, const u8
*src_mac
,
946 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
949 f
= vxlan_find_mac(vxlan
, src_mac
, vni
);
951 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
953 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
956 /* Don't migrate static entries, drop packets */
957 if (f
->state
& NUD_NOARP
)
962 "%pM migrated from %pIS to %pIS\n",
963 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
965 rdst
->remote_ip
= *src_ip
;
966 f
->updated
= jiffies
;
967 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
969 /* learned new entry */
970 spin_lock(&vxlan
->hash_lock
);
972 /* close off race between vxlan_flush and incoming packets */
973 if (netif_running(dev
))
974 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
976 NLM_F_EXCL
|NLM_F_CREATE
,
979 vxlan
->default_dst
.remote_vni
,
981 spin_unlock(&vxlan
->hash_lock
);
987 /* See if multicast group is already in use by other ID */
988 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
990 struct vxlan_dev
*vxlan
;
991 struct vxlan_sock
*sock4
;
992 #if IS_ENABLED(CONFIG_IPV6)
993 struct vxlan_sock
*sock6
;
995 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
997 sock4
= rtnl_dereference(dev
->vn4_sock
);
999 /* The vxlan_sock is only used by dev, leaving group has
1000 * no effect on other vxlan devices.
1002 if (family
== AF_INET
&& sock4
&& atomic_read(&sock4
->refcnt
) == 1)
1004 #if IS_ENABLED(CONFIG_IPV6)
1005 sock6
= rtnl_dereference(dev
->vn6_sock
);
1006 if (family
== AF_INET6
&& sock6
&& atomic_read(&sock6
->refcnt
) == 1)
1010 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1011 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1014 if (family
== AF_INET
&&
1015 rtnl_dereference(vxlan
->vn4_sock
) != sock4
)
1017 #if IS_ENABLED(CONFIG_IPV6)
1018 if (family
== AF_INET6
&&
1019 rtnl_dereference(vxlan
->vn6_sock
) != sock6
)
1023 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1024 &dev
->default_dst
.remote_ip
))
1027 if (vxlan
->default_dst
.remote_ifindex
!=
1028 dev
->default_dst
.remote_ifindex
)
1037 static bool __vxlan_sock_release_prep(struct vxlan_sock
*vs
)
1039 struct vxlan_net
*vn
;
1043 if (!atomic_dec_and_test(&vs
->refcnt
))
1046 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1047 spin_lock(&vn
->sock_lock
);
1048 hlist_del_rcu(&vs
->hlist
);
1049 udp_tunnel_notify_del_rx_port(vs
->sock
,
1050 (vs
->flags
& VXLAN_F_GPE
) ?
1051 UDP_TUNNEL_TYPE_VXLAN_GPE
:
1052 UDP_TUNNEL_TYPE_VXLAN
);
1053 spin_unlock(&vn
->sock_lock
);
1058 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1060 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1061 #if IS_ENABLED(CONFIG_IPV6)
1062 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1064 rcu_assign_pointer(vxlan
->vn6_sock
, NULL
);
1067 rcu_assign_pointer(vxlan
->vn4_sock
, NULL
);
1070 if (__vxlan_sock_release_prep(sock4
)) {
1071 udp_tunnel_sock_release(sock4
->sock
);
1075 #if IS_ENABLED(CONFIG_IPV6)
1076 if (__vxlan_sock_release_prep(sock6
)) {
1077 udp_tunnel_sock_release(sock6
->sock
);
1083 /* Update multicast group membership when first VNI on
1084 * multicast address is brought up
1086 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1089 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1090 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1093 if (ip
->sa
.sa_family
== AF_INET
) {
1094 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1095 struct ip_mreqn mreq
= {
1096 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1097 .imr_ifindex
= ifindex
,
1100 sk
= sock4
->sock
->sk
;
1102 ret
= ip_mc_join_group(sk
, &mreq
);
1104 #if IS_ENABLED(CONFIG_IPV6)
1106 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1108 sk
= sock6
->sock
->sk
;
1110 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1111 &ip
->sin6
.sin6_addr
);
1119 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1120 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1123 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1124 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1127 if (ip
->sa
.sa_family
== AF_INET
) {
1128 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1129 struct ip_mreqn mreq
= {
1130 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1131 .imr_ifindex
= ifindex
,
1134 sk
= sock4
->sock
->sk
;
1136 ret
= ip_mc_leave_group(sk
, &mreq
);
1138 #if IS_ENABLED(CONFIG_IPV6)
1140 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1142 sk
= sock6
->sock
->sk
;
1144 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1145 &ip
->sin6
.sin6_addr
);
1153 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1154 struct sk_buff
*skb
, u32 vxflags
)
1156 size_t start
, offset
;
1158 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1161 start
= vxlan_rco_start(unparsed
->vx_vni
);
1162 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1164 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1167 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1168 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1170 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1171 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1175 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1176 struct sk_buff
*skb
, u32 vxflags
,
1177 struct vxlan_metadata
*md
)
1179 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1180 struct metadata_dst
*tun_dst
;
1182 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1185 md
->gbp
= ntohs(gbp
->policy_id
);
1187 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1189 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1190 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1192 if (gbp
->dont_learn
)
1193 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1195 if (gbp
->policy_applied
)
1196 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1198 /* In flow-based mode, GBP is carried in dst_metadata */
1199 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1200 skb
->mark
= md
->gbp
;
1202 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1205 static bool vxlan_parse_gpe_hdr(struct vxlanhdr
*unparsed
,
1207 struct sk_buff
*skb
, u32 vxflags
)
1209 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)unparsed
;
1211 /* Need to have Next Protocol set for interfaces in GPE mode. */
1212 if (!gpe
->np_applied
)
1214 /* "The initial version is 0. If a receiver does not support the
1215 * version indicated it MUST drop the packet.
1217 if (gpe
->version
!= 0)
1219 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1220 * processing MUST occur." However, we don't implement OAM
1221 * processing, thus drop the packet.
1226 switch (gpe
->next_protocol
) {
1227 case VXLAN_GPE_NP_IPV4
:
1228 *protocol
= htons(ETH_P_IP
);
1230 case VXLAN_GPE_NP_IPV6
:
1231 *protocol
= htons(ETH_P_IPV6
);
1233 case VXLAN_GPE_NP_ETHERNET
:
1234 *protocol
= htons(ETH_P_TEB
);
1240 unparsed
->vx_flags
&= ~VXLAN_GPE_USED_BITS
;
1244 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1245 struct vxlan_sock
*vs
,
1246 struct sk_buff
*skb
, __be32 vni
)
1248 union vxlan_addr saddr
;
1250 skb_reset_mac_header(skb
);
1251 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1252 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1254 /* Ignore packet loops (and multicast echo) */
1255 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1258 /* Get address from the outer IP header */
1259 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1260 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1261 saddr
.sa
.sa_family
= AF_INET
;
1262 #if IS_ENABLED(CONFIG_IPV6)
1264 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1265 saddr
.sa
.sa_family
= AF_INET6
;
1269 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1270 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
, vni
))
1276 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1277 struct sk_buff
*skb
)
1281 if (vxlan_get_sk_family(vs
) == AF_INET
)
1282 err
= IP_ECN_decapsulate(oiph
, skb
);
1283 #if IS_ENABLED(CONFIG_IPV6)
1285 err
= IP6_ECN_decapsulate(oiph
, skb
);
1288 if (unlikely(err
) && log_ecn_error
) {
1289 if (vxlan_get_sk_family(vs
) == AF_INET
)
1290 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1291 &((struct iphdr
*)oiph
)->saddr
,
1292 ((struct iphdr
*)oiph
)->tos
);
1294 net_info_ratelimited("non-ECT from %pI6\n",
1295 &((struct ipv6hdr
*)oiph
)->saddr
);
1300 /* Callback from net/ipv4/udp.c to receive packets */
1301 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1303 struct pcpu_sw_netstats
*stats
;
1304 struct vxlan_dev
*vxlan
;
1305 struct vxlan_sock
*vs
;
1306 struct vxlanhdr unparsed
;
1307 struct vxlan_metadata _md
;
1308 struct vxlan_metadata
*md
= &_md
;
1309 __be16 protocol
= htons(ETH_P_TEB
);
1310 bool raw_proto
= false;
1314 /* Need UDP and VXLAN header to be present */
1315 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1318 unparsed
= *vxlan_hdr(skb
);
1319 /* VNI flag always required to be set */
1320 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1321 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1322 ntohl(vxlan_hdr(skb
)->vx_flags
),
1323 ntohl(vxlan_hdr(skb
)->vx_vni
));
1324 /* Return non vxlan pkt */
1327 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1328 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1330 vs
= rcu_dereference_sk_user_data(sk
);
1334 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1336 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1340 /* For backwards compatibility, only allow reserved fields to be
1341 * used by VXLAN extensions if explicitly requested.
1343 if (vs
->flags
& VXLAN_F_GPE
) {
1344 if (!vxlan_parse_gpe_hdr(&unparsed
, &protocol
, skb
, vs
->flags
))
1349 if (__iptunnel_pull_header(skb
, VXLAN_HLEN
, protocol
, raw_proto
,
1350 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1353 if (vxlan_collect_metadata(vs
)) {
1354 struct metadata_dst
*tun_dst
;
1356 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1357 key32_to_tunnel_id(vni
), sizeof(*md
));
1362 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1364 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1366 memset(md
, 0, sizeof(*md
));
1369 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1370 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1372 if (vs
->flags
& VXLAN_F_GBP
)
1373 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1374 /* Note that GBP and GPE can never be active together. This is
1375 * ensured in vxlan_dev_configure.
1378 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1379 /* If there are any unprocessed flags remaining treat
1380 * this as a malformed packet. This behavior diverges from
1381 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1382 * in reserved fields are to be ignored. The approach here
1383 * maintains compatibility with previous stack code, and also
1384 * is more robust and provides a little more security in
1385 * adding extensions to VXLAN.
1391 if (!vxlan_set_mac(vxlan
, vs
, skb
, vni
))
1394 skb_reset_mac_header(skb
);
1395 skb
->dev
= vxlan
->dev
;
1396 skb
->pkt_type
= PACKET_HOST
;
1399 oiph
= skb_network_header(skb
);
1400 skb_reset_network_header(skb
);
1402 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1403 ++vxlan
->dev
->stats
.rx_frame_errors
;
1404 ++vxlan
->dev
->stats
.rx_errors
;
1408 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1409 u64_stats_update_begin(&stats
->syncp
);
1410 stats
->rx_packets
++;
1411 stats
->rx_bytes
+= skb
->len
;
1412 u64_stats_update_end(&stats
->syncp
);
1414 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1418 /* Consume bad packet */
1423 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1425 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1426 struct arphdr
*parp
;
1429 struct neighbour
*n
;
1431 if (dev
->flags
& IFF_NOARP
)
1434 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1435 dev
->stats
.tx_dropped
++;
1438 parp
= arp_hdr(skb
);
1440 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1441 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1442 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1443 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1444 parp
->ar_hln
!= dev
->addr_len
||
1447 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1449 arpptr
+= dev
->addr_len
; /* sha */
1450 memcpy(&sip
, arpptr
, sizeof(sip
));
1451 arpptr
+= sizeof(sip
);
1452 arpptr
+= dev
->addr_len
; /* tha */
1453 memcpy(&tip
, arpptr
, sizeof(tip
));
1455 if (ipv4_is_loopback(tip
) ||
1456 ipv4_is_multicast(tip
))
1459 n
= neigh_lookup(&arp_tbl
, &tip
, 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
, vni
);
1471 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1472 /* bridge-local neighbor */
1477 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1485 skb_reset_mac_header(reply
);
1486 __skb_pull(reply
, skb_network_offset(reply
));
1487 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1488 reply
->pkt_type
= PACKET_HOST
;
1490 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1491 dev
->stats
.rx_dropped
++;
1492 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1493 union vxlan_addr ipa
= {
1494 .sin
.sin_addr
.s_addr
= tip
,
1495 .sin
.sin_family
= AF_INET
,
1498 vxlan_ip_miss(dev
, &ipa
);
1502 return NETDEV_TX_OK
;
1505 #if IS_ENABLED(CONFIG_IPV6)
1506 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1507 struct neighbour
*n
, bool isrouter
)
1509 struct net_device
*dev
= request
->dev
;
1510 struct sk_buff
*reply
;
1511 struct nd_msg
*ns
, *na
;
1512 struct ipv6hdr
*pip6
;
1514 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1518 if (dev
== NULL
|| !pskb_may_pull(request
, request
->len
))
1521 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1522 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1523 reply
= alloc_skb(len
, GFP_ATOMIC
);
1527 reply
->protocol
= htons(ETH_P_IPV6
);
1529 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1530 skb_push(reply
, sizeof(struct ethhdr
));
1531 skb_reset_mac_header(reply
);
1533 ns
= (struct nd_msg
*)(ipv6_hdr(request
) + 1);
1535 daddr
= eth_hdr(request
)->h_source
;
1536 ns_olen
= request
->len
- skb_network_offset(request
) -
1537 sizeof(struct ipv6hdr
) - sizeof(*ns
);
1538 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1539 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1540 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1545 /* Ethernet header */
1546 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1547 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1548 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1549 reply
->protocol
= htons(ETH_P_IPV6
);
1551 skb_pull(reply
, sizeof(struct ethhdr
));
1552 skb_reset_network_header(reply
);
1553 skb_put(reply
, sizeof(struct ipv6hdr
));
1557 pip6
= ipv6_hdr(reply
);
1558 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1560 pip6
->priority
= ipv6_hdr(request
)->priority
;
1561 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1562 pip6
->hop_limit
= 255;
1563 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1564 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1566 skb_pull(reply
, sizeof(struct ipv6hdr
));
1567 skb_reset_transport_header(reply
);
1569 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1571 /* Neighbor Advertisement */
1572 memset(na
, 0, sizeof(*na
)+na_olen
);
1573 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1574 na
->icmph
.icmp6_router
= isrouter
;
1575 na
->icmph
.icmp6_override
= 1;
1576 na
->icmph
.icmp6_solicited
= 1;
1577 na
->target
= ns
->target
;
1578 ether_addr_copy(&na
->opt
[2], n
->ha
);
1579 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1580 na
->opt
[1] = na_olen
>> 3;
1582 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1583 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1584 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1586 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1588 skb_push(reply
, sizeof(struct ipv6hdr
));
1590 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1595 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1597 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1599 const struct ipv6hdr
*iphdr
;
1600 const struct in6_addr
*daddr
;
1601 struct neighbour
*n
;
1602 struct inet6_dev
*in6_dev
;
1604 in6_dev
= __in6_dev_get(dev
);
1608 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
) + sizeof(struct nd_msg
)))
1611 iphdr
= ipv6_hdr(skb
);
1612 daddr
= &iphdr
->daddr
;
1614 msg
= (struct nd_msg
*)(iphdr
+ 1);
1615 if (msg
->icmph
.icmp6_code
!= 0 ||
1616 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1619 if (ipv6_addr_loopback(daddr
) ||
1620 ipv6_addr_is_multicast(&msg
->target
))
1623 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1626 struct vxlan_fdb
*f
;
1627 struct sk_buff
*reply
;
1629 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1634 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1635 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1636 /* bridge-local neighbor */
1641 reply
= vxlan_na_create(skb
, n
,
1642 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1649 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1650 dev
->stats
.rx_dropped
++;
1652 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1653 union vxlan_addr ipa
= {
1654 .sin6
.sin6_addr
= msg
->target
,
1655 .sin6
.sin6_family
= AF_INET6
,
1658 vxlan_ip_miss(dev
, &ipa
);
1663 return NETDEV_TX_OK
;
1667 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1669 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1670 struct neighbour
*n
;
1672 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1676 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1681 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1684 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1685 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1686 union vxlan_addr ipa
= {
1687 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1688 .sin
.sin_family
= AF_INET
,
1691 vxlan_ip_miss(dev
, &ipa
);
1697 #if IS_ENABLED(CONFIG_IPV6)
1700 struct ipv6hdr
*pip6
;
1702 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1704 pip6
= ipv6_hdr(skb
);
1705 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1706 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1707 union vxlan_addr ipa
= {
1708 .sin6
.sin6_addr
= pip6
->daddr
,
1709 .sin6
.sin6_family
= AF_INET6
,
1712 vxlan_ip_miss(dev
, &ipa
);
1726 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1728 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1730 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1739 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1740 struct vxlan_metadata
*md
)
1742 struct vxlanhdr_gbp
*gbp
;
1747 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1748 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1750 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1751 gbp
->dont_learn
= 1;
1753 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1754 gbp
->policy_applied
= 1;
1756 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1759 static int vxlan_build_gpe_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1762 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)vxh
;
1764 gpe
->np_applied
= 1;
1767 case htons(ETH_P_IP
):
1768 gpe
->next_protocol
= VXLAN_GPE_NP_IPV4
;
1770 case htons(ETH_P_IPV6
):
1771 gpe
->next_protocol
= VXLAN_GPE_NP_IPV6
;
1773 case htons(ETH_P_TEB
):
1774 gpe
->next_protocol
= VXLAN_GPE_NP_ETHERNET
;
1777 return -EPFNOSUPPORT
;
1780 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1781 int iphdr_len
, __be32 vni
,
1782 struct vxlan_metadata
*md
, u32 vxflags
,
1785 struct vxlanhdr
*vxh
;
1788 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1789 __be16 inner_protocol
= htons(ETH_P_TEB
);
1791 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1792 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1793 int csum_start
= skb_checksum_start_offset(skb
);
1795 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1796 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1797 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1798 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
1799 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1802 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1803 + VXLAN_HLEN
+ iphdr_len
;
1805 /* Need space for new headers (invalidates iph ptr) */
1806 err
= skb_cow_head(skb
, min_headroom
);
1810 err
= iptunnel_handle_offloads(skb
, type
);
1814 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1815 vxh
->vx_flags
= VXLAN_HF_VNI
;
1816 vxh
->vx_vni
= vxlan_vni_field(vni
);
1818 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1821 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
1822 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
1823 vxh
->vx_flags
|= VXLAN_HF_RCO
;
1825 if (!skb_is_gso(skb
)) {
1826 skb
->ip_summed
= CHECKSUM_NONE
;
1827 skb
->encapsulation
= 0;
1831 if (vxflags
& VXLAN_F_GBP
)
1832 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1833 if (vxflags
& VXLAN_F_GPE
) {
1834 err
= vxlan_build_gpe_hdr(vxh
, vxflags
, skb
->protocol
);
1837 inner_protocol
= skb
->protocol
;
1840 skb_set_inner_protocol(skb
, inner_protocol
);
1844 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
, struct net_device
*dev
,
1845 struct vxlan_sock
*sock4
,
1846 struct sk_buff
*skb
, int oif
, u8 tos
,
1847 __be32 daddr
, __be32
*saddr
, __be16 dport
, __be16 sport
,
1848 struct dst_cache
*dst_cache
,
1849 const struct ip_tunnel_info
*info
)
1851 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1852 struct rtable
*rt
= NULL
;
1856 return ERR_PTR(-EIO
);
1861 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
1866 memset(&fl4
, 0, sizeof(fl4
));
1867 fl4
.flowi4_oif
= oif
;
1868 fl4
.flowi4_tos
= RT_TOS(tos
);
1869 fl4
.flowi4_mark
= skb
->mark
;
1870 fl4
.flowi4_proto
= IPPROTO_UDP
;
1873 fl4
.fl4_dport
= dport
;
1874 fl4
.fl4_sport
= sport
;
1876 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1877 if (likely(!IS_ERR(rt
))) {
1878 if (rt
->dst
.dev
== dev
) {
1879 netdev_dbg(dev
, "circular route to %pI4\n", &daddr
);
1881 return ERR_PTR(-ELOOP
);
1886 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
1888 netdev_dbg(dev
, "no route to %pI4\n", &daddr
);
1889 return ERR_PTR(-ENETUNREACH
);
1894 #if IS_ENABLED(CONFIG_IPV6)
1895 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1896 struct net_device
*dev
,
1897 struct vxlan_sock
*sock6
,
1898 struct sk_buff
*skb
, int oif
, u8 tos
,
1900 const struct in6_addr
*daddr
,
1901 struct in6_addr
*saddr
,
1902 __be16 dport
, __be16 sport
,
1903 struct dst_cache
*dst_cache
,
1904 const struct ip_tunnel_info
*info
)
1906 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1907 struct dst_entry
*ndst
;
1912 return ERR_PTR(-EIO
);
1917 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
1922 memset(&fl6
, 0, sizeof(fl6
));
1923 fl6
.flowi6_oif
= oif
;
1926 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
1927 fl6
.flowi6_mark
= skb
->mark
;
1928 fl6
.flowi6_proto
= IPPROTO_UDP
;
1929 fl6
.fl6_dport
= dport
;
1930 fl6
.fl6_sport
= sport
;
1932 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1935 if (unlikely(err
< 0)) {
1936 netdev_dbg(dev
, "no route to %pI6\n", daddr
);
1937 return ERR_PTR(-ENETUNREACH
);
1940 if (unlikely(ndst
->dev
== dev
)) {
1941 netdev_dbg(dev
, "circular route to %pI6\n", daddr
);
1943 return ERR_PTR(-ELOOP
);
1948 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
1953 /* Bypass encapsulation if the destination is local */
1954 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1955 struct vxlan_dev
*dst_vxlan
, __be32 vni
)
1957 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1958 union vxlan_addr loopback
;
1959 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1960 struct net_device
*dev
= skb
->dev
;
1963 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1964 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1965 skb
->pkt_type
= PACKET_HOST
;
1966 skb
->encapsulation
= 0;
1967 skb
->dev
= dst_vxlan
->dev
;
1968 __skb_pull(skb
, skb_network_offset(skb
));
1970 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1971 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1972 loopback
.sa
.sa_family
= AF_INET
;
1973 #if IS_ENABLED(CONFIG_IPV6)
1975 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1976 loopback
.sa
.sa_family
= AF_INET6
;
1980 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1981 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
, vni
);
1983 u64_stats_update_begin(&tx_stats
->syncp
);
1984 tx_stats
->tx_packets
++;
1985 tx_stats
->tx_bytes
+= len
;
1986 u64_stats_update_end(&tx_stats
->syncp
);
1988 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1989 u64_stats_update_begin(&rx_stats
->syncp
);
1990 rx_stats
->rx_packets
++;
1991 rx_stats
->rx_bytes
+= len
;
1992 u64_stats_update_end(&rx_stats
->syncp
);
1994 dev
->stats
.rx_dropped
++;
1998 static int encap_bypass_if_local(struct sk_buff
*skb
, struct net_device
*dev
,
1999 struct vxlan_dev
*vxlan
, union vxlan_addr
*daddr
,
2000 __be16 dst_port
, __be32 vni
, struct dst_entry
*dst
,
2003 #if IS_ENABLED(CONFIG_IPV6)
2004 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2005 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2006 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2008 BUILD_BUG_ON(RTCF_LOCAL
!= RTF_LOCAL
);
2010 /* Bypass encapsulation if the destination is local */
2011 if (rt_flags
& RTCF_LOCAL
&&
2012 !(rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2013 struct vxlan_dev
*dst_vxlan
;
2016 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2017 daddr
->sa
.sa_family
, dst_port
,
2020 dev
->stats
.tx_errors
++;
2025 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
, vni
);
2032 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
2033 __be32 default_vni
, struct vxlan_rdst
*rdst
,
2036 struct dst_cache
*dst_cache
;
2037 struct ip_tunnel_info
*info
;
2038 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2039 const struct iphdr
*old_iph
= ip_hdr(skb
);
2040 union vxlan_addr
*dst
;
2041 union vxlan_addr remote_ip
, local_ip
;
2042 struct vxlan_metadata _md
;
2043 struct vxlan_metadata
*md
= &_md
;
2044 __be16 src_port
= 0, dst_port
;
2045 struct dst_entry
*ndst
= NULL
;
2049 u32 flags
= vxlan
->flags
;
2050 bool udp_sum
= false;
2051 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
2053 info
= skb_tunnel_info(skb
);
2056 dst
= &rdst
->remote_ip
;
2057 if (vxlan_addr_any(dst
)) {
2059 /* short-circuited back to local bridge */
2060 vxlan_encap_bypass(skb
, vxlan
, vxlan
, default_vni
);
2066 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
2067 vni
= (rdst
->remote_vni
) ? : default_vni
;
2068 local_ip
= vxlan
->cfg
.saddr
;
2069 dst_cache
= &rdst
->dst_cache
;
2070 md
->gbp
= skb
->mark
;
2071 ttl
= vxlan
->cfg
.ttl
;
2072 if (!ttl
&& vxlan_addr_multicast(dst
))
2075 tos
= vxlan
->cfg
.tos
;
2077 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
2079 if (dst
->sa
.sa_family
== AF_INET
)
2080 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
2082 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2083 label
= vxlan
->cfg
.label
;
2086 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2090 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
2091 if (remote_ip
.sa
.sa_family
== AF_INET
) {
2092 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
2093 local_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.src
;
2095 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
2096 local_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.src
;
2099 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2100 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2101 dst_cache
= &info
->dst_cache
;
2102 if (info
->options_len
)
2103 md
= ip_tunnel_info_opts(info
);
2104 ttl
= info
->key
.ttl
;
2105 tos
= info
->key
.tos
;
2106 label
= info
->key
.label
;
2107 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
2109 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2110 vxlan
->cfg
.port_max
, true);
2113 if (dst
->sa
.sa_family
== AF_INET
) {
2114 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2118 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
,
2119 rdst
? rdst
->remote_ifindex
: 0, tos
,
2120 dst
->sin
.sin_addr
.s_addr
,
2121 &local_ip
.sin
.sin_addr
.s_addr
,
2129 /* Bypass encapsulation if the destination is local */
2131 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2132 dst_port
, vni
, &rt
->dst
,
2136 } else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
) {
2141 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2142 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2143 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct iphdr
),
2144 vni
, md
, flags
, udp_sum
);
2148 udp_tunnel_xmit_skb(rt
, sock4
->sock
->sk
, skb
, local_ip
.sin
.sin_addr
.s_addr
,
2149 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2150 src_port
, dst_port
, xnet
, !udp_sum
);
2151 #if IS_ENABLED(CONFIG_IPV6)
2153 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2155 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
,
2156 rdst
? rdst
->remote_ifindex
: 0, tos
,
2157 label
, &dst
->sin6
.sin6_addr
,
2158 &local_ip
.sin6
.sin6_addr
,
2162 err
= PTR_ERR(ndst
);
2168 u32 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2170 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2171 dst_port
, vni
, ndst
,
2177 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2178 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2179 skb_scrub_packet(skb
, xnet
);
2180 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2181 vni
, md
, flags
, udp_sum
);
2185 udp_tunnel6_xmit_skb(ndst
, sock6
->sock
->sk
, skb
, dev
,
2186 &local_ip
.sin6
.sin6_addr
,
2187 &dst
->sin6
.sin6_addr
, tos
, ttl
,
2188 label
, src_port
, dst_port
, !udp_sum
);
2196 dev
->stats
.tx_dropped
++;
2203 dev
->stats
.collisions
++;
2204 else if (err
== -ENETUNREACH
)
2205 dev
->stats
.tx_carrier_errors
++;
2207 dev
->stats
.tx_errors
++;
2211 /* Transmit local packets over Vxlan
2213 * Outer IP header inherits ECN and DF from inner header.
2214 * Outer UDP destination is the VXLAN assigned port.
2215 * source port is based on hash of flow
2217 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2219 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2220 const struct ip_tunnel_info
*info
;
2222 bool did_rsc
= false;
2223 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2224 struct vxlan_fdb
*f
;
2227 info
= skb_tunnel_info(skb
);
2229 skb_reset_mac_header(skb
);
2231 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
2232 if (info
&& info
->mode
& IP_TUNNEL_INFO_BRIDGE
&&
2233 info
->mode
& IP_TUNNEL_INFO_TX
) {
2234 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2236 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2237 vxlan_xmit_one(skb
, dev
, vni
, NULL
, false);
2240 return NETDEV_TX_OK
;
2244 if (vxlan
->flags
& VXLAN_F_PROXY
) {
2246 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2247 return arp_reduce(dev
, skb
, vni
);
2248 #if IS_ENABLED(CONFIG_IPV6)
2249 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
) {
2250 struct ipv6hdr
*hdr
, _hdr
;
2251 if ((hdr
= skb_header_pointer(skb
,
2252 skb_network_offset(skb
),
2253 sizeof(_hdr
), &_hdr
)) &&
2254 hdr
->nexthdr
== IPPROTO_ICMPV6
)
2255 return neigh_reduce(dev
, skb
, vni
);
2261 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2264 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2265 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2266 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2267 did_rsc
= route_shortcircuit(dev
, skb
);
2269 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2273 f
= vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2275 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2276 !is_multicast_ether_addr(eth
->h_dest
))
2277 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2279 dev
->stats
.tx_dropped
++;
2281 return NETDEV_TX_OK
;
2285 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2286 struct sk_buff
*skb1
;
2292 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2294 vxlan_xmit_one(skb1
, dev
, vni
, rdst
, did_rsc
);
2298 vxlan_xmit_one(skb
, dev
, vni
, fdst
, did_rsc
);
2301 return NETDEV_TX_OK
;
2304 /* Walk the forwarding table and purge stale entries */
2305 static void vxlan_cleanup(unsigned long arg
)
2307 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2308 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2311 if (!netif_running(vxlan
->dev
))
2314 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2315 struct hlist_node
*p
, *n
;
2317 spin_lock_bh(&vxlan
->hash_lock
);
2318 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2320 = container_of(p
, struct vxlan_fdb
, hlist
);
2321 unsigned long timeout
;
2323 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
2326 if (f
->flags
& NTF_EXT_LEARNED
)
2329 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2330 if (time_before_eq(timeout
, jiffies
)) {
2331 netdev_dbg(vxlan
->dev
,
2332 "garbage collect %pM\n",
2334 f
->state
= NUD_STALE
;
2335 vxlan_fdb_destroy(vxlan
, f
);
2336 } else if (time_before(timeout
, next_timer
))
2337 next_timer
= timeout
;
2339 spin_unlock_bh(&vxlan
->hash_lock
);
2342 mod_timer(&vxlan
->age_timer
, next_timer
);
2345 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2347 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2348 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2350 spin_lock(&vn
->sock_lock
);
2351 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2352 spin_unlock(&vn
->sock_lock
);
2355 /* Setup stats when device is created */
2356 static int vxlan_init(struct net_device
*dev
)
2358 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2365 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
, __be32 vni
)
2367 struct vxlan_fdb
*f
;
2369 spin_lock_bh(&vxlan
->hash_lock
);
2370 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2372 vxlan_fdb_destroy(vxlan
, f
);
2373 spin_unlock_bh(&vxlan
->hash_lock
);
2376 static void vxlan_uninit(struct net_device
*dev
)
2378 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2380 vxlan_fdb_delete_default(vxlan
, vxlan
->cfg
.vni
);
2382 free_percpu(dev
->tstats
);
2385 /* Start ageing timer and join group when device is brought up */
2386 static int vxlan_open(struct net_device
*dev
)
2388 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2391 ret
= vxlan_sock_add(vxlan
);
2395 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2396 ret
= vxlan_igmp_join(vxlan
);
2397 if (ret
== -EADDRINUSE
)
2400 vxlan_sock_release(vxlan
);
2405 if (vxlan
->cfg
.age_interval
)
2406 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2411 /* Purge the forwarding table */
2412 static void vxlan_flush(struct vxlan_dev
*vxlan
, bool do_all
)
2416 spin_lock_bh(&vxlan
->hash_lock
);
2417 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2418 struct hlist_node
*p
, *n
;
2419 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2421 = container_of(p
, struct vxlan_fdb
, hlist
);
2422 if (!do_all
&& (f
->state
& (NUD_PERMANENT
| NUD_NOARP
)))
2424 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2425 if (!is_zero_ether_addr(f
->eth_addr
))
2426 vxlan_fdb_destroy(vxlan
, f
);
2429 spin_unlock_bh(&vxlan
->hash_lock
);
2432 /* Cleanup timer and forwarding table on shutdown */
2433 static int vxlan_stop(struct net_device
*dev
)
2435 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2436 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2439 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2440 !vxlan_group_used(vn
, vxlan
))
2441 ret
= vxlan_igmp_leave(vxlan
);
2443 del_timer_sync(&vxlan
->age_timer
);
2445 vxlan_flush(vxlan
, false);
2446 vxlan_sock_release(vxlan
);
2451 /* Stub, nothing needs to be done. */
2452 static void vxlan_set_multicast_list(struct net_device
*dev
)
2456 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2458 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2459 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2460 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2461 dst
->remote_ifindex
);
2462 bool use_ipv6
= false;
2464 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2467 /* This check is different than dev->max_mtu, because it looks at
2468 * the lowerdev->mtu, rather than the static dev->max_mtu
2471 int max_mtu
= lowerdev
->mtu
-
2472 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2473 if (new_mtu
> max_mtu
)
2481 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2483 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2484 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2485 __be16 sport
, dport
;
2487 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2488 vxlan
->cfg
.port_max
, true);
2489 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2491 if (ip_tunnel_info_af(info
) == AF_INET
) {
2492 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2495 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, 0, info
->key
.tos
,
2496 info
->key
.u
.ipv4
.dst
,
2497 &info
->key
.u
.ipv4
.src
, dport
, sport
,
2498 &info
->dst_cache
, info
);
2503 #if IS_ENABLED(CONFIG_IPV6)
2504 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2505 struct dst_entry
*ndst
;
2507 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, 0, info
->key
.tos
,
2508 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2509 &info
->key
.u
.ipv6
.src
, dport
, sport
,
2510 &info
->dst_cache
, info
);
2512 return PTR_ERR(ndst
);
2514 #else /* !CONFIG_IPV6 */
2515 return -EPFNOSUPPORT
;
2518 info
->key
.tp_src
= sport
;
2519 info
->key
.tp_dst
= dport
;
2523 static const struct net_device_ops vxlan_netdev_ether_ops
= {
2524 .ndo_init
= vxlan_init
,
2525 .ndo_uninit
= vxlan_uninit
,
2526 .ndo_open
= vxlan_open
,
2527 .ndo_stop
= vxlan_stop
,
2528 .ndo_start_xmit
= vxlan_xmit
,
2529 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2530 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2531 .ndo_change_mtu
= vxlan_change_mtu
,
2532 .ndo_validate_addr
= eth_validate_addr
,
2533 .ndo_set_mac_address
= eth_mac_addr
,
2534 .ndo_fdb_add
= vxlan_fdb_add
,
2535 .ndo_fdb_del
= vxlan_fdb_delete
,
2536 .ndo_fdb_dump
= vxlan_fdb_dump
,
2537 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2540 static const struct net_device_ops vxlan_netdev_raw_ops
= {
2541 .ndo_init
= vxlan_init
,
2542 .ndo_uninit
= vxlan_uninit
,
2543 .ndo_open
= vxlan_open
,
2544 .ndo_stop
= vxlan_stop
,
2545 .ndo_start_xmit
= vxlan_xmit
,
2546 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2547 .ndo_change_mtu
= vxlan_change_mtu
,
2548 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2551 /* Info for udev, that this is a virtual tunnel endpoint */
2552 static struct device_type vxlan_type
= {
2556 /* Calls the ndo_udp_tunnel_add of the caller in order to
2557 * supply the listening VXLAN udp ports. Callers are expected
2558 * to implement the ndo_udp_tunnel_add.
2560 static void vxlan_push_rx_ports(struct net_device
*dev
)
2562 struct vxlan_sock
*vs
;
2563 struct net
*net
= dev_net(dev
);
2564 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2567 spin_lock(&vn
->sock_lock
);
2568 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2569 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
)
2570 udp_tunnel_push_rx_port(dev
, vs
->sock
,
2571 (vs
->flags
& VXLAN_F_GPE
) ?
2572 UDP_TUNNEL_TYPE_VXLAN_GPE
:
2573 UDP_TUNNEL_TYPE_VXLAN
);
2575 spin_unlock(&vn
->sock_lock
);
2578 /* Initialize the device structure. */
2579 static void vxlan_setup(struct net_device
*dev
)
2581 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2584 eth_hw_addr_random(dev
);
2587 dev
->destructor
= free_netdev
;
2588 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2590 dev
->features
|= NETIF_F_LLTX
;
2591 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2592 dev
->features
|= NETIF_F_RXCSUM
;
2593 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2595 dev
->vlan_features
= dev
->features
;
2596 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2597 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2598 netif_keep_dst(dev
);
2599 dev
->priv_flags
|= IFF_NO_QUEUE
;
2601 INIT_LIST_HEAD(&vxlan
->next
);
2602 spin_lock_init(&vxlan
->hash_lock
);
2604 init_timer_deferrable(&vxlan
->age_timer
);
2605 vxlan
->age_timer
.function
= vxlan_cleanup
;
2606 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2608 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2612 gro_cells_init(&vxlan
->gro_cells
, dev
);
2614 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2615 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2618 static void vxlan_ether_setup(struct net_device
*dev
)
2620 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2621 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2622 dev
->netdev_ops
= &vxlan_netdev_ether_ops
;
2625 static void vxlan_raw_setup(struct net_device
*dev
)
2627 dev
->header_ops
= NULL
;
2628 dev
->type
= ARPHRD_NONE
;
2629 dev
->hard_header_len
= 0;
2631 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
2632 dev
->netdev_ops
= &vxlan_netdev_raw_ops
;
2635 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2636 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2637 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2638 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2639 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2640 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2641 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2642 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2643 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2644 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
2645 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2646 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2647 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2648 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2649 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2650 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2651 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2652 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2653 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2654 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2655 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2656 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2657 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2658 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2659 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2660 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2661 [IFLA_VXLAN_GPE
] = { .type
= NLA_FLAG
, },
2662 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2665 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2667 if (tb
[IFLA_ADDRESS
]) {
2668 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2669 pr_debug("invalid link address (not ethernet)\n");
2673 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2674 pr_debug("invalid all zero ethernet address\n");
2675 return -EADDRNOTAVAIL
;
2682 if (data
[IFLA_VXLAN_ID
]) {
2683 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2684 if (id
>= VXLAN_N_VID
)
2688 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2689 const struct ifla_vxlan_port_range
*p
2690 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2692 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2693 pr_debug("port range %u .. %u not valid\n",
2694 ntohs(p
->low
), ntohs(p
->high
));
2702 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2703 struct ethtool_drvinfo
*drvinfo
)
2705 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2706 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2709 static const struct ethtool_ops vxlan_ethtool_ops
= {
2710 .get_drvinfo
= vxlan_get_drvinfo
,
2711 .get_link
= ethtool_op_get_link
,
2714 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2715 __be16 port
, u32 flags
)
2717 struct socket
*sock
;
2718 struct udp_port_cfg udp_conf
;
2721 memset(&udp_conf
, 0, sizeof(udp_conf
));
2724 udp_conf
.family
= AF_INET6
;
2725 udp_conf
.use_udp6_rx_checksums
=
2726 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2727 udp_conf
.ipv6_v6only
= 1;
2729 udp_conf
.family
= AF_INET
;
2732 udp_conf
.local_udp_port
= port
;
2734 /* Open UDP socket */
2735 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2737 return ERR_PTR(err
);
2742 /* Create new listen socket if needed */
2743 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2744 __be16 port
, u32 flags
)
2746 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2747 struct vxlan_sock
*vs
;
2748 struct socket
*sock
;
2750 struct udp_tunnel_sock_cfg tunnel_cfg
;
2752 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2754 return ERR_PTR(-ENOMEM
);
2756 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2757 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2759 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2762 return ERR_CAST(sock
);
2766 atomic_set(&vs
->refcnt
, 1);
2767 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2769 spin_lock(&vn
->sock_lock
);
2770 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2771 udp_tunnel_notify_add_rx_port(sock
,
2772 (vs
->flags
& VXLAN_F_GPE
) ?
2773 UDP_TUNNEL_TYPE_VXLAN_GPE
:
2774 UDP_TUNNEL_TYPE_VXLAN
);
2775 spin_unlock(&vn
->sock_lock
);
2777 /* Mark socket as an encapsulation socket. */
2778 memset(&tunnel_cfg
, 0, sizeof(tunnel_cfg
));
2779 tunnel_cfg
.sk_user_data
= vs
;
2780 tunnel_cfg
.encap_type
= 1;
2781 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
2782 tunnel_cfg
.encap_destroy
= NULL
;
2783 tunnel_cfg
.gro_receive
= vxlan_gro_receive
;
2784 tunnel_cfg
.gro_complete
= vxlan_gro_complete
;
2786 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2791 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2793 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2794 struct vxlan_sock
*vs
= NULL
;
2796 if (!vxlan
->cfg
.no_share
) {
2797 spin_lock(&vn
->sock_lock
);
2798 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2799 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2800 if (vs
&& !atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2801 spin_unlock(&vn
->sock_lock
);
2804 spin_unlock(&vn
->sock_lock
);
2807 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2808 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2811 #if IS_ENABLED(CONFIG_IPV6)
2813 rcu_assign_pointer(vxlan
->vn6_sock
, vs
);
2816 rcu_assign_pointer(vxlan
->vn4_sock
, vs
);
2817 vxlan_vs_add_dev(vs
, vxlan
);
2821 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2823 bool metadata
= vxlan
->flags
& VXLAN_F_COLLECT_METADATA
;
2824 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
|| metadata
;
2825 bool ipv4
= !ipv6
|| metadata
;
2828 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
2829 #if IS_ENABLED(CONFIG_IPV6)
2830 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
2832 ret
= __vxlan_sock_add(vxlan
, true);
2833 if (ret
< 0 && ret
!= -EAFNOSUPPORT
)
2838 ret
= __vxlan_sock_add(vxlan
, false);
2840 vxlan_sock_release(vxlan
);
2844 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2845 struct vxlan_config
*conf
,
2848 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2849 struct vxlan_dev
*vxlan
= netdev_priv(dev
), *tmp
;
2850 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2851 unsigned short needed_headroom
= ETH_HLEN
;
2852 bool use_ipv6
= false;
2853 __be16 default_port
= vxlan
->cfg
.dst_port
;
2854 struct net_device
*lowerdev
= NULL
;
2857 if (conf
->flags
& VXLAN_F_GPE
) {
2858 /* For now, allow GPE only together with
2859 * COLLECT_METADATA. This can be relaxed later; in such
2860 * case, the other side of the PtP link will have to be
2863 if ((conf
->flags
& ~VXLAN_F_ALLOWED_GPE
) ||
2864 !(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
2865 pr_info("unsupported combination of extensions\n");
2868 vxlan_raw_setup(dev
);
2870 vxlan_ether_setup(dev
);
2873 /* MTU range: 68 - 65535 */
2874 dev
->min_mtu
= ETH_MIN_MTU
;
2875 dev
->max_mtu
= ETH_MAX_MTU
;
2876 vxlan
->net
= src_net
;
2879 dst
->remote_vni
= conf
->vni
;
2881 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2883 /* Unless IPv6 is explicitly requested, assume IPv4 */
2884 if (!dst
->remote_ip
.sa
.sa_family
)
2885 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2887 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2888 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) {
2889 if (!IS_ENABLED(CONFIG_IPV6
))
2890 return -EPFNOSUPPORT
;
2892 vxlan
->flags
|= VXLAN_F_IPV6
;
2895 if (conf
->label
&& !use_ipv6
) {
2896 pr_info("label only supported in use with IPv6\n");
2900 if (conf
->remote_ifindex
&&
2901 conf
->remote_ifindex
!= vxlan
->cfg
.remote_ifindex
) {
2902 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2903 dst
->remote_ifindex
= conf
->remote_ifindex
;
2906 pr_info("ifindex %d does not exist\n",
2907 dst
->remote_ifindex
);
2911 #if IS_ENABLED(CONFIG_IPV6)
2913 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2914 if (idev
&& idev
->cnf
.disable_ipv6
) {
2915 pr_info("IPv6 is disabled via sysctl\n");
2922 dev
->mtu
= lowerdev
->mtu
-
2923 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2925 needed_headroom
= lowerdev
->hard_header_len
;
2926 } else if (!conf
->remote_ifindex
&&
2927 vxlan_addr_multicast(&dst
->remote_ip
)) {
2928 pr_info("multicast destination requires interface to be specified\n");
2933 dev
->gso_max_size
= lowerdev
->gso_max_size
;
2934 dev
->gso_max_segs
= lowerdev
->gso_max_segs
;
2938 int max_mtu
= ETH_MAX_MTU
;
2941 max_mtu
= lowerdev
->mtu
;
2943 max_mtu
-= (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2945 if (conf
->mtu
< dev
->min_mtu
|| conf
->mtu
> dev
->max_mtu
)
2948 dev
->mtu
= conf
->mtu
;
2950 if (conf
->mtu
> max_mtu
)
2954 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
2955 needed_headroom
+= VXLAN6_HEADROOM
;
2957 needed_headroom
+= VXLAN_HEADROOM
;
2958 dev
->needed_headroom
= needed_headroom
;
2960 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2961 if (!vxlan
->cfg
.dst_port
) {
2962 if (conf
->flags
& VXLAN_F_GPE
)
2963 vxlan
->cfg
.dst_port
= htons(4790); /* IANA VXLAN-GPE port */
2965 vxlan
->cfg
.dst_port
= default_port
;
2967 vxlan
->flags
|= conf
->flags
;
2969 if (!vxlan
->cfg
.age_interval
)
2970 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2975 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
2976 if (tmp
->cfg
.vni
== conf
->vni
&&
2977 (tmp
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
||
2978 tmp
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) == use_ipv6
&&
2979 tmp
->cfg
.dst_port
== vxlan
->cfg
.dst_port
&&
2980 (tmp
->flags
& VXLAN_F_RCV_FLAGS
) ==
2981 (vxlan
->flags
& VXLAN_F_RCV_FLAGS
)) {
2982 pr_info("duplicate VNI %u\n", be32_to_cpu(conf
->vni
));
2990 static int __vxlan_dev_create(struct net
*net
, struct net_device
*dev
,
2991 struct vxlan_config
*conf
)
2993 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2994 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2997 err
= vxlan_dev_configure(net
, dev
, conf
, false);
3001 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
3003 /* create an fdb entry for a valid default destination */
3004 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
3005 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3006 &vxlan
->default_dst
.remote_ip
,
3007 NUD_REACHABLE
| NUD_PERMANENT
,
3008 NLM_F_EXCL
| NLM_F_CREATE
,
3009 vxlan
->cfg
.dst_port
,
3010 vxlan
->default_dst
.remote_vni
,
3011 vxlan
->default_dst
.remote_vni
,
3012 vxlan
->default_dst
.remote_ifindex
,
3018 err
= register_netdevice(dev
);
3020 vxlan_fdb_delete_default(vxlan
, vxlan
->default_dst
.remote_vni
);
3024 list_add(&vxlan
->next
, &vn
->vxlan_list
);
3028 static int vxlan_nl2conf(struct nlattr
*tb
[], struct nlattr
*data
[],
3029 struct net_device
*dev
, struct vxlan_config
*conf
,
3032 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3034 memset(conf
, 0, sizeof(*conf
));
3036 /* if changelink operation, start with old existing cfg */
3038 memcpy(conf
, &vxlan
->cfg
, sizeof(*conf
));
3040 if (data
[IFLA_VXLAN_ID
]) {
3041 __be32 vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3043 if (changelink
&& (vni
!= conf
->vni
))
3045 conf
->vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3048 if (data
[IFLA_VXLAN_GROUP
]) {
3049 conf
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
3050 } else if (data
[IFLA_VXLAN_GROUP6
]) {
3051 if (!IS_ENABLED(CONFIG_IPV6
))
3052 return -EPFNOSUPPORT
;
3054 conf
->remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
3055 conf
->remote_ip
.sa
.sa_family
= AF_INET6
;
3058 if (data
[IFLA_VXLAN_LOCAL
]) {
3059 conf
->saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
3060 conf
->saddr
.sa
.sa_family
= AF_INET
;
3061 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
3062 if (!IS_ENABLED(CONFIG_IPV6
))
3063 return -EPFNOSUPPORT
;
3065 /* TODO: respect scope id */
3066 conf
->saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
3067 conf
->saddr
.sa
.sa_family
= AF_INET6
;
3070 if (data
[IFLA_VXLAN_LINK
])
3071 conf
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
3073 if (data
[IFLA_VXLAN_TOS
])
3074 conf
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
3076 if (data
[IFLA_VXLAN_TTL
])
3077 conf
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
3079 if (data
[IFLA_VXLAN_LABEL
])
3080 conf
->label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
3081 IPV6_FLOWLABEL_MASK
;
3083 if (data
[IFLA_VXLAN_LEARNING
]) {
3084 if (nla_get_u8(data
[IFLA_VXLAN_LEARNING
])) {
3085 conf
->flags
|= VXLAN_F_LEARN
;
3087 conf
->flags
&= ~VXLAN_F_LEARN
;
3088 vxlan
->flags
&= ~VXLAN_F_LEARN
;
3090 } else if (!changelink
) {
3091 /* default to learn on a new device */
3092 conf
->flags
|= VXLAN_F_LEARN
;
3095 if (data
[IFLA_VXLAN_AGEING
]) {
3098 conf
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
3101 if (data
[IFLA_VXLAN_PROXY
]) {
3104 if (nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
3105 conf
->flags
|= VXLAN_F_PROXY
;
3108 if (data
[IFLA_VXLAN_RSC
]) {
3111 if (nla_get_u8(data
[IFLA_VXLAN_RSC
]))
3112 conf
->flags
|= VXLAN_F_RSC
;
3115 if (data
[IFLA_VXLAN_L2MISS
]) {
3118 if (nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
3119 conf
->flags
|= VXLAN_F_L2MISS
;
3122 if (data
[IFLA_VXLAN_L3MISS
]) {
3125 if (nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
3126 conf
->flags
|= VXLAN_F_L3MISS
;
3129 if (data
[IFLA_VXLAN_LIMIT
]) {
3132 conf
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
3135 if (data
[IFLA_VXLAN_COLLECT_METADATA
]) {
3138 if (nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
3139 conf
->flags
|= VXLAN_F_COLLECT_METADATA
;
3142 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3144 const struct ifla_vxlan_port_range
*p
3145 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3146 conf
->port_min
= ntohs(p
->low
);
3147 conf
->port_max
= ntohs(p
->high
);
3153 if (data
[IFLA_VXLAN_PORT
]) {
3156 conf
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
3159 if (data
[IFLA_VXLAN_UDP_CSUM
]) {
3162 if (!nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
3163 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
3166 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]) {
3169 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
3170 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
3173 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]) {
3176 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
3177 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
3180 if (data
[IFLA_VXLAN_REMCSUM_TX
]) {
3183 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
3184 conf
->flags
|= VXLAN_F_REMCSUM_TX
;
3187 if (data
[IFLA_VXLAN_REMCSUM_RX
]) {
3190 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
3191 conf
->flags
|= VXLAN_F_REMCSUM_RX
;
3194 if (data
[IFLA_VXLAN_GBP
]) {
3197 conf
->flags
|= VXLAN_F_GBP
;
3200 if (data
[IFLA_VXLAN_GPE
]) {
3203 conf
->flags
|= VXLAN_F_GPE
;
3206 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
]) {
3209 conf
->flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
3215 conf
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3221 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
3222 struct nlattr
*tb
[], struct nlattr
*data
[])
3224 struct vxlan_config conf
;
3227 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, false);
3231 return __vxlan_dev_create(src_net
, dev
, &conf
);
3234 static int vxlan_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
3235 struct nlattr
*data
[])
3237 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3238 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3239 struct vxlan_rdst old_dst
;
3240 struct vxlan_config conf
;
3243 err
= vxlan_nl2conf(tb
, data
,
3248 memcpy(&old_dst
, dst
, sizeof(struct vxlan_rdst
));
3250 err
= vxlan_dev_configure(vxlan
->net
, dev
, &conf
, true);
3254 /* handle default dst entry */
3255 if (!vxlan_addr_equal(&dst
->remote_ip
, &old_dst
.remote_ip
)) {
3256 spin_lock_bh(&vxlan
->hash_lock
);
3257 if (!vxlan_addr_any(&old_dst
.remote_ip
))
3258 __vxlan_fdb_delete(vxlan
, all_zeros_mac
,
3260 vxlan
->cfg
.dst_port
,
3263 old_dst
.remote_ifindex
, 0);
3265 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3266 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3268 NUD_REACHABLE
| NUD_PERMANENT
,
3269 NLM_F_CREATE
| NLM_F_APPEND
,
3270 vxlan
->cfg
.dst_port
,
3273 dst
->remote_ifindex
,
3276 spin_unlock_bh(&vxlan
->hash_lock
);
3280 spin_unlock_bh(&vxlan
->hash_lock
);
3286 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
3288 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3289 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
3291 vxlan_flush(vxlan
, true);
3293 spin_lock(&vn
->sock_lock
);
3294 if (!hlist_unhashed(&vxlan
->hlist
))
3295 hlist_del_rcu(&vxlan
->hlist
);
3296 spin_unlock(&vn
->sock_lock
);
3298 gro_cells_destroy(&vxlan
->gro_cells
);
3299 list_del(&vxlan
->next
);
3300 unregister_netdevice_queue(dev
, head
);
3303 static size_t vxlan_get_size(const struct net_device
*dev
)
3306 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
3307 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
3308 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
3309 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
3310 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3311 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3312 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
3313 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3314 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3315 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3316 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3317 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3318 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3319 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3320 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3321 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3322 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3323 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3324 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3325 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3326 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3327 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3331 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3333 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3334 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3335 struct ifla_vxlan_port_range ports
= {
3336 .low
= htons(vxlan
->cfg
.port_min
),
3337 .high
= htons(vxlan
->cfg
.port_max
),
3340 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3341 goto nla_put_failure
;
3343 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3344 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3345 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3346 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3347 goto nla_put_failure
;
3348 #if IS_ENABLED(CONFIG_IPV6)
3350 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3351 &dst
->remote_ip
.sin6
.sin6_addr
))
3352 goto nla_put_failure
;
3357 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3358 goto nla_put_failure
;
3360 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3361 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3362 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3363 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3364 goto nla_put_failure
;
3365 #if IS_ENABLED(CONFIG_IPV6)
3367 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3368 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3369 goto nla_put_failure
;
3374 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3375 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3376 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
3377 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3378 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
3379 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3380 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
3381 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
3382 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3383 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
3384 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3385 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
3386 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3387 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
3388 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3389 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3390 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3391 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3392 !(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3393 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3394 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3395 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3396 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3397 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3398 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
3399 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3400 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
3401 goto nla_put_failure
;
3403 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3404 goto nla_put_failure
;
3406 if (vxlan
->flags
& VXLAN_F_GBP
&&
3407 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3408 goto nla_put_failure
;
3410 if (vxlan
->flags
& VXLAN_F_GPE
&&
3411 nla_put_flag(skb
, IFLA_VXLAN_GPE
))
3412 goto nla_put_failure
;
3414 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3415 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3416 goto nla_put_failure
;
3424 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3426 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3431 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3433 .maxtype
= IFLA_VXLAN_MAX
,
3434 .policy
= vxlan_policy
,
3435 .priv_size
= sizeof(struct vxlan_dev
),
3436 .setup
= vxlan_setup
,
3437 .validate
= vxlan_validate
,
3438 .newlink
= vxlan_newlink
,
3439 .changelink
= vxlan_changelink
,
3440 .dellink
= vxlan_dellink
,
3441 .get_size
= vxlan_get_size
,
3442 .fill_info
= vxlan_fill_info
,
3443 .get_link_net
= vxlan_get_link_net
,
3446 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
3447 u8 name_assign_type
,
3448 struct vxlan_config
*conf
)
3450 struct nlattr
*tb
[IFLA_MAX
+ 1];
3451 struct net_device
*dev
;
3454 memset(&tb
, 0, sizeof(tb
));
3456 dev
= rtnl_create_link(net
, name
, name_assign_type
,
3457 &vxlan_link_ops
, tb
);
3461 err
= __vxlan_dev_create(net
, dev
, conf
);
3464 return ERR_PTR(err
);
3467 err
= rtnl_configure_link(dev
, NULL
);
3469 LIST_HEAD(list_kill
);
3471 vxlan_dellink(dev
, &list_kill
);
3472 unregister_netdevice_many(&list_kill
);
3473 return ERR_PTR(err
);
3478 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
3480 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3481 struct net_device
*dev
)
3483 struct vxlan_dev
*vxlan
, *next
;
3484 LIST_HEAD(list_kill
);
3486 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3487 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3489 /* In case we created vxlan device with carrier
3490 * and we loose the carrier due to module unload
3491 * we also need to remove vxlan device. In other
3492 * cases, it's not necessary and remote_ifindex
3493 * is 0 here, so no matches.
3495 if (dst
->remote_ifindex
== dev
->ifindex
)
3496 vxlan_dellink(vxlan
->dev
, &list_kill
);
3499 unregister_netdevice_many(&list_kill
);
3502 static int vxlan_netdevice_event(struct notifier_block
*unused
,
3503 unsigned long event
, void *ptr
)
3505 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3506 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3508 if (event
== NETDEV_UNREGISTER
)
3509 vxlan_handle_lowerdev_unregister(vn
, dev
);
3510 else if (event
== NETDEV_UDP_TUNNEL_PUSH_INFO
)
3511 vxlan_push_rx_ports(dev
);
3516 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3517 .notifier_call
= vxlan_netdevice_event
,
3520 static __net_init
int vxlan_init_net(struct net
*net
)
3522 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3525 INIT_LIST_HEAD(&vn
->vxlan_list
);
3526 spin_lock_init(&vn
->sock_lock
);
3528 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3529 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3534 static void __net_exit
vxlan_exit_net(struct net
*net
)
3536 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3537 struct vxlan_dev
*vxlan
, *next
;
3538 struct net_device
*dev
, *aux
;
3542 for_each_netdev_safe(net
, dev
, aux
)
3543 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3544 unregister_netdevice_queue(dev
, &list
);
3546 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3547 /* If vxlan->dev is in the same netns, it has already been added
3548 * to the list by the previous loop.
3550 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3551 gro_cells_destroy(&vxlan
->gro_cells
);
3552 unregister_netdevice_queue(vxlan
->dev
, &list
);
3556 unregister_netdevice_many(&list
);
3560 static struct pernet_operations vxlan_net_ops
= {
3561 .init
= vxlan_init_net
,
3562 .exit
= vxlan_exit_net
,
3563 .id
= &vxlan_net_id
,
3564 .size
= sizeof(struct vxlan_net
),
3567 static int __init
vxlan_init_module(void)
3571 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3573 rc
= register_pernet_subsys(&vxlan_net_ops
);
3577 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3581 rc
= rtnl_link_register(&vxlan_link_ops
);
3587 unregister_netdevice_notifier(&vxlan_notifier_block
);
3589 unregister_pernet_subsys(&vxlan_net_ops
);
3593 late_initcall(vxlan_init_module
);
3595 static void __exit
vxlan_cleanup_module(void)
3597 rtnl_link_unregister(&vxlan_link_ops
);
3598 unregister_netdevice_notifier(&vxlan_notifier_block
);
3599 unregister_pernet_subsys(&vxlan_net_ops
);
3600 /* rcu_barrier() is called by netns */
3602 module_exit(vxlan_cleanup_module
);
3604 MODULE_LICENSE("GPL");
3605 MODULE_VERSION(VXLAN_VERSION
);
3606 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3607 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3608 MODULE_ALIAS_RTNL_LINK("vxlan");