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>
46 #if IS_ENABLED(CONFIG_IPV6)
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
52 #include <net/dst_metadata.h>
54 #define VXLAN_VERSION "0.1"
56 #define PORT_HASH_BITS 8
57 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61 /* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
63 * for compatibility with early adopters.
65 static unsigned short vxlan_port __read_mostly
= 8472;
66 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
67 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
69 static bool log_ecn_error
= true;
70 module_param(log_ecn_error
, bool, 0644);
71 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
73 static int vxlan_net_id
;
74 static struct rtnl_link_ops vxlan_link_ops
;
76 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
78 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
80 /* per-network namespace private data for this module */
82 struct list_head vxlan_list
;
83 struct hlist_head sock_list
[PORT_HASH_SIZE
];
87 /* Forwarding table entry */
89 struct hlist_node hlist
; /* linked list of entries */
91 unsigned long updated
; /* jiffies */
93 struct list_head remotes
;
94 u8 eth_addr
[ETH_ALEN
];
95 u16 state
; /* see ndm_state */
96 u8 flags
; /* see ndm_flags */
99 /* salt for hash table */
100 static u32 vxlan_salt __read_mostly
;
101 static struct workqueue_struct
*vxlan_wq
;
103 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
105 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
106 ip_tunnel_collect_metadata();
109 #if IS_ENABLED(CONFIG_IPV6)
111 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
113 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
115 if (a
->sa
.sa_family
== AF_INET6
)
116 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
118 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
121 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
123 if (ipa
->sa
.sa_family
== AF_INET6
)
124 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
126 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
129 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
131 if (ipa
->sa
.sa_family
== AF_INET6
)
132 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
134 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
137 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
139 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
140 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
141 ip
->sa
.sa_family
= AF_INET6
;
143 } else if (nla_len(nla
) >= sizeof(__be32
)) {
144 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
145 ip
->sa
.sa_family
= AF_INET
;
148 return -EAFNOSUPPORT
;
152 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
153 const union vxlan_addr
*ip
)
155 if (ip
->sa
.sa_family
== AF_INET6
)
156 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
158 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
161 #else /* !CONFIG_IPV6 */
164 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
166 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
169 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
171 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
174 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
176 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
179 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
181 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
182 return -EAFNOSUPPORT
;
183 } else if (nla_len(nla
) >= sizeof(__be32
)) {
184 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
185 ip
->sa
.sa_family
= AF_INET
;
188 return -EAFNOSUPPORT
;
192 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
193 const union vxlan_addr
*ip
)
195 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
199 /* Virtual Network hash table head */
200 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
202 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
205 /* Socket hash table head */
206 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
208 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
210 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
213 /* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
216 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
218 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
221 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
223 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
226 /* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
229 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
230 __be16 port
, u32 flags
)
232 struct vxlan_sock
*vs
;
234 flags
&= VXLAN_F_RCV_FLAGS
;
236 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
237 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
238 vxlan_get_sk_family(vs
) == family
&&
245 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, __be32 vni
)
247 struct vxlan_dev
*vxlan
;
249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
253 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, vni
), hlist
) {
254 if (vxlan
->default_dst
.remote_vni
== vni
)
261 /* Look up VNI in a per net namespace table */
262 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, __be32 vni
,
263 sa_family_t family
, __be16 port
,
266 struct vxlan_sock
*vs
;
268 vs
= vxlan_find_sock(net
, family
, port
, flags
);
272 return vxlan_vs_find_vni(vs
, vni
);
275 /* Fill in neighbour message in skbuff. */
276 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
277 const struct vxlan_fdb
*fdb
,
278 u32 portid
, u32 seq
, int type
, unsigned int flags
,
279 const struct vxlan_rdst
*rdst
)
281 unsigned long now
= jiffies
;
282 struct nda_cacheinfo ci
;
283 struct nlmsghdr
*nlh
;
285 bool send_ip
, send_eth
;
287 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
291 ndm
= nlmsg_data(nlh
);
292 memset(ndm
, 0, sizeof(*ndm
));
294 send_eth
= send_ip
= true;
296 if (type
== RTM_GETNEIGH
) {
297 ndm
->ndm_family
= AF_INET
;
298 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
299 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
301 ndm
->ndm_family
= AF_BRIDGE
;
302 ndm
->ndm_state
= fdb
->state
;
303 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
304 ndm
->ndm_flags
= fdb
->flags
;
305 ndm
->ndm_type
= RTN_UNICAST
;
307 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
308 nla_put_s32(skb
, NDA_LINK_NETNSID
,
309 peernet2id_alloc(dev_net(vxlan
->dev
), vxlan
->net
)))
310 goto nla_put_failure
;
312 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
313 goto nla_put_failure
;
315 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
316 goto nla_put_failure
;
318 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
319 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
320 goto nla_put_failure
;
321 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
322 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
323 goto nla_put_failure
;
324 if (rdst
->remote_ifindex
&&
325 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
326 goto nla_put_failure
;
328 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
329 ci
.ndm_confirmed
= 0;
330 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
333 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
334 goto nla_put_failure
;
340 nlmsg_cancel(skb
, nlh
);
344 static inline size_t vxlan_nlmsg_size(void)
346 return NLMSG_ALIGN(sizeof(struct ndmsg
))
347 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
348 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
349 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
350 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
352 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
353 + nla_total_size(sizeof(struct nda_cacheinfo
));
356 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
357 struct vxlan_rdst
*rd
, int type
)
359 struct net
*net
= dev_net(vxlan
->dev
);
363 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
367 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err
== -EMSGSIZE
);
375 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
379 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
382 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
384 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
385 struct vxlan_fdb f
= {
388 struct vxlan_rdst remote
= {
389 .remote_ip
= *ipa
, /* goes to NDA_DST */
390 .remote_vni
= cpu_to_be32(VXLAN_N_VID
),
393 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
396 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
398 struct vxlan_fdb f
= {
401 struct vxlan_rdst remote
= { };
403 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
405 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
408 /* Hash Ethernet address */
409 static u32
eth_hash(const unsigned char *addr
)
411 u64 value
= get_unaligned((u64
*)addr
);
413 /* only want 6 bytes */
419 return hash_64(value
, FDB_HASH_BITS
);
422 /* Hash chain to use given mac address */
423 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
426 return &vxlan
->fdb_head
[eth_hash(mac
)];
429 /* Look up Ethernet address in forwarding table */
430 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
433 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
436 hlist_for_each_entry_rcu(f
, head
, hlist
) {
437 if (ether_addr_equal(mac
, f
->eth_addr
))
444 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
449 f
= __vxlan_find_mac(vxlan
, mac
);
456 /* caller should hold vxlan->hash_lock */
457 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
458 union vxlan_addr
*ip
, __be16 port
,
459 __be32 vni
, __u32 ifindex
)
461 struct vxlan_rdst
*rd
;
463 list_for_each_entry(rd
, &f
->remotes
, list
) {
464 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
465 rd
->remote_port
== port
&&
466 rd
->remote_vni
== vni
&&
467 rd
->remote_ifindex
== ifindex
)
474 /* Replace destination of unicast mac */
475 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
476 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
479 struct vxlan_rdst
*rd
;
481 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
485 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
489 dst_cache_reset(&rd
->dst_cache
);
491 rd
->remote_port
= port
;
492 rd
->remote_vni
= vni
;
493 rd
->remote_ifindex
= ifindex
;
497 /* Add/update destinations for multicast */
498 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
499 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
500 __u32 ifindex
, struct vxlan_rdst
**rdp
)
502 struct vxlan_rdst
*rd
;
504 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
508 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
512 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
518 rd
->remote_port
= port
;
519 rd
->remote_vni
= vni
;
520 rd
->remote_ifindex
= ifindex
;
522 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
528 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
530 struct vxlanhdr
*vh
, size_t hdrlen
,
532 struct gro_remcsum
*grc
,
535 size_t start
, offset
;
537 if (skb
->remcsum_offload
)
540 if (!NAPI_GRO_CB(skb
)->csum_valid
)
543 start
= vxlan_rco_start(vni_field
);
544 offset
= start
+ vxlan_rco_offset(vni_field
);
546 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
547 start
, offset
, grc
, nopartial
);
549 skb
->remcsum_offload
= 1;
554 static struct sk_buff
**vxlan_gro_receive(struct sk_buff
**head
,
556 struct udp_offload
*uoff
)
558 struct sk_buff
*p
, **pp
= NULL
;
559 struct vxlanhdr
*vh
, *vh2
;
560 unsigned int hlen
, off_vx
;
562 struct vxlan_sock
*vs
= container_of(uoff
, struct vxlan_sock
,
565 struct gro_remcsum grc
;
567 skb_gro_remcsum_init(&grc
);
569 off_vx
= skb_gro_offset(skb
);
570 hlen
= off_vx
+ sizeof(*vh
);
571 vh
= skb_gro_header_fast(skb
, off_vx
);
572 if (skb_gro_header_hard(skb
, hlen
)) {
573 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
578 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
580 flags
= vh
->vx_flags
;
582 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
583 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
586 VXLAN_F_REMCSUM_NOPARTIAL
));
592 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
594 for (p
= *head
; p
; p
= p
->next
) {
595 if (!NAPI_GRO_CB(p
)->same_flow
)
598 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
599 if (vh
->vx_flags
!= vh2
->vx_flags
||
600 vh
->vx_vni
!= vh2
->vx_vni
) {
601 NAPI_GRO_CB(p
)->same_flow
= 0;
606 pp
= eth_gro_receive(head
, skb
);
610 skb_gro_remcsum_cleanup(skb
, &grc
);
611 NAPI_GRO_CB(skb
)->flush
|= flush
;
616 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
617 struct udp_offload
*uoff
)
619 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
622 /* Notify netdevs that UDP port started listening */
623 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
625 struct net_device
*dev
;
626 struct sock
*sk
= vs
->sock
->sk
;
627 struct net
*net
= sock_net(sk
);
628 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
629 __be16 port
= inet_sk(sk
)->inet_sport
;
632 if (sa_family
== AF_INET
) {
633 err
= udp_add_offload(net
, &vs
->udp_offloads
);
635 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
639 for_each_netdev_rcu(net
, dev
) {
640 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
641 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
647 /* Notify netdevs that UDP port is no more listening */
648 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
650 struct net_device
*dev
;
651 struct sock
*sk
= vs
->sock
->sk
;
652 struct net
*net
= sock_net(sk
);
653 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
654 __be16 port
= inet_sk(sk
)->inet_sport
;
657 for_each_netdev_rcu(net
, dev
) {
658 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
659 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
664 if (sa_family
== AF_INET
)
665 udp_del_offload(&vs
->udp_offloads
);
668 /* Add new entry to forwarding table -- assumes lock held */
669 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
670 const u8
*mac
, union vxlan_addr
*ip
,
671 __u16 state
, __u16 flags
,
672 __be16 port
, __be32 vni
, __u32 ifindex
,
675 struct vxlan_rdst
*rd
= NULL
;
679 f
= __vxlan_find_mac(vxlan
, mac
);
681 if (flags
& NLM_F_EXCL
) {
682 netdev_dbg(vxlan
->dev
,
683 "lost race to create %pM\n", mac
);
686 if (f
->state
!= state
) {
688 f
->updated
= jiffies
;
691 if (f
->flags
!= ndm_flags
) {
692 f
->flags
= ndm_flags
;
693 f
->updated
= jiffies
;
696 if ((flags
& NLM_F_REPLACE
)) {
697 /* Only change unicasts */
698 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
699 is_zero_ether_addr(f
->eth_addr
))) {
700 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
705 if ((flags
& NLM_F_APPEND
) &&
706 (is_multicast_ether_addr(f
->eth_addr
) ||
707 is_zero_ether_addr(f
->eth_addr
))) {
708 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
716 if (!(flags
& NLM_F_CREATE
))
719 if (vxlan
->cfg
.addrmax
&&
720 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
723 /* Disallow replace to add a multicast entry */
724 if ((flags
& NLM_F_REPLACE
) &&
725 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
728 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
729 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
735 f
->flags
= ndm_flags
;
736 f
->updated
= f
->used
= jiffies
;
737 INIT_LIST_HEAD(&f
->remotes
);
738 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
740 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
743 hlist_add_head_rcu(&f
->hlist
,
744 vxlan_fdb_head(vxlan
, mac
));
749 rd
= first_remote_rtnl(f
);
750 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
756 static void vxlan_fdb_free(struct rcu_head
*head
)
758 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
759 struct vxlan_rdst
*rd
, *nd
;
761 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
762 dst_cache_destroy(&rd
->dst_cache
);
768 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
770 netdev_dbg(vxlan
->dev
,
771 "delete %pM\n", f
->eth_addr
);
774 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
776 hlist_del_rcu(&f
->hlist
);
777 call_rcu(&f
->rcu
, vxlan_fdb_free
);
780 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
781 union vxlan_addr
*ip
, __be16
*port
, __be32
*vni
,
784 struct net
*net
= dev_net(vxlan
->dev
);
788 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
792 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
793 if (remote
->sa
.sa_family
== AF_INET
) {
794 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
795 ip
->sa
.sa_family
= AF_INET
;
796 #if IS_ENABLED(CONFIG_IPV6)
798 ip
->sin6
.sin6_addr
= in6addr_any
;
799 ip
->sa
.sa_family
= AF_INET6
;
805 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
807 *port
= nla_get_be16(tb
[NDA_PORT
]);
809 *port
= vxlan
->cfg
.dst_port
;
813 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
815 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
817 *vni
= vxlan
->default_dst
.remote_vni
;
820 if (tb
[NDA_IFINDEX
]) {
821 struct net_device
*tdev
;
823 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
825 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
826 tdev
= __dev_get_by_index(net
, *ifindex
);
828 return -EADDRNOTAVAIL
;
836 /* Add static entry (via netlink) */
837 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
838 struct net_device
*dev
,
839 const unsigned char *addr
, u16 vid
, u16 flags
)
841 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
842 /* struct net *net = dev_net(vxlan->dev); */
849 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
850 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
855 if (tb
[NDA_DST
] == NULL
)
858 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
862 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
863 return -EAFNOSUPPORT
;
865 spin_lock_bh(&vxlan
->hash_lock
);
866 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
867 port
, vni
, ifindex
, ndm
->ndm_flags
);
868 spin_unlock_bh(&vxlan
->hash_lock
);
873 /* Delete entry (via netlink) */
874 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
875 struct net_device
*dev
,
876 const unsigned char *addr
, u16 vid
)
878 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
880 struct vxlan_rdst
*rd
= NULL
;
887 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
893 spin_lock_bh(&vxlan
->hash_lock
);
894 f
= vxlan_find_mac(vxlan
, addr
);
898 if (!vxlan_addr_any(&ip
)) {
899 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
906 /* remove a destination if it's not the only one on the list,
907 * otherwise destroy the fdb entry
909 if (rd
&& !list_is_singular(&f
->remotes
)) {
910 list_del_rcu(&rd
->list
);
911 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
916 vxlan_fdb_destroy(vxlan
, f
);
919 spin_unlock_bh(&vxlan
->hash_lock
);
924 /* Dump forwarding table */
925 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
926 struct net_device
*dev
,
927 struct net_device
*filter_dev
, int idx
)
929 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
932 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
936 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
937 struct vxlan_rdst
*rd
;
939 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
940 if (idx
< cb
->args
[0])
943 err
= vxlan_fdb_info(skb
, vxlan
, f
,
944 NETLINK_CB(cb
->skb
).portid
,
961 /* Watch incoming packets to learn mapping between Ethernet address
962 * and Tunnel endpoint.
963 * Return true if packet is bogus and should be dropped.
965 static bool vxlan_snoop(struct net_device
*dev
,
966 union vxlan_addr
*src_ip
, const u8
*src_mac
)
968 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
971 f
= vxlan_find_mac(vxlan
, src_mac
);
973 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
975 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
978 /* Don't migrate static entries, drop packets */
979 if (f
->state
& NUD_NOARP
)
984 "%pM migrated from %pIS to %pIS\n",
985 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
987 rdst
->remote_ip
= *src_ip
;
988 f
->updated
= jiffies
;
989 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
991 /* learned new entry */
992 spin_lock(&vxlan
->hash_lock
);
994 /* close off race between vxlan_flush and incoming packets */
995 if (netif_running(dev
))
996 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
998 NLM_F_EXCL
|NLM_F_CREATE
,
1000 vxlan
->default_dst
.remote_vni
,
1002 spin_unlock(&vxlan
->hash_lock
);
1008 /* See if multicast group is already in use by other ID */
1009 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1011 struct vxlan_dev
*vxlan
;
1012 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1014 /* The vxlan_sock is only used by dev, leaving group has
1015 * no effect on other vxlan devices.
1017 if (family
== AF_INET
&& dev
->vn4_sock
&&
1018 atomic_read(&dev
->vn4_sock
->refcnt
) == 1)
1020 #if IS_ENABLED(CONFIG_IPV6)
1021 if (family
== AF_INET6
&& dev
->vn6_sock
&&
1022 atomic_read(&dev
->vn6_sock
->refcnt
) == 1)
1026 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1027 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1030 if (family
== AF_INET
&& vxlan
->vn4_sock
!= dev
->vn4_sock
)
1032 #if IS_ENABLED(CONFIG_IPV6)
1033 if (family
== AF_INET6
&& vxlan
->vn6_sock
!= dev
->vn6_sock
)
1037 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1038 &dev
->default_dst
.remote_ip
))
1041 if (vxlan
->default_dst
.remote_ifindex
!=
1042 dev
->default_dst
.remote_ifindex
)
1051 static void __vxlan_sock_release(struct vxlan_sock
*vs
)
1053 struct vxlan_net
*vn
;
1057 if (!atomic_dec_and_test(&vs
->refcnt
))
1060 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1061 spin_lock(&vn
->sock_lock
);
1062 hlist_del_rcu(&vs
->hlist
);
1063 vxlan_notify_del_rx_port(vs
);
1064 spin_unlock(&vn
->sock_lock
);
1066 queue_work(vxlan_wq
, &vs
->del_work
);
1069 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1071 __vxlan_sock_release(vxlan
->vn4_sock
);
1072 #if IS_ENABLED(CONFIG_IPV6)
1073 __vxlan_sock_release(vxlan
->vn6_sock
);
1077 /* Update multicast group membership when first VNI on
1078 * multicast address is brought up
1080 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1083 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1084 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1087 if (ip
->sa
.sa_family
== AF_INET
) {
1088 struct ip_mreqn mreq
= {
1089 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1090 .imr_ifindex
= ifindex
,
1093 sk
= vxlan
->vn4_sock
->sock
->sk
;
1095 ret
= ip_mc_join_group(sk
, &mreq
);
1097 #if IS_ENABLED(CONFIG_IPV6)
1099 sk
= vxlan
->vn6_sock
->sock
->sk
;
1101 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1102 &ip
->sin6
.sin6_addr
);
1110 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1111 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1114 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1115 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1118 if (ip
->sa
.sa_family
== AF_INET
) {
1119 struct ip_mreqn mreq
= {
1120 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1121 .imr_ifindex
= ifindex
,
1124 sk
= vxlan
->vn4_sock
->sock
->sk
;
1126 ret
= ip_mc_leave_group(sk
, &mreq
);
1128 #if IS_ENABLED(CONFIG_IPV6)
1130 sk
= vxlan
->vn6_sock
->sock
->sk
;
1132 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1133 &ip
->sin6
.sin6_addr
);
1141 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1142 struct sk_buff
*skb
, u32 vxflags
)
1144 size_t start
, offset
;
1146 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1149 start
= vxlan_rco_start(unparsed
->vx_vni
);
1150 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1152 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1155 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1156 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1158 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1159 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1163 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1164 struct sk_buff
*skb
, u32 vxflags
,
1165 struct vxlan_metadata
*md
)
1167 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1168 struct metadata_dst
*tun_dst
;
1170 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1173 md
->gbp
= ntohs(gbp
->policy_id
);
1175 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1177 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1178 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1180 if (gbp
->dont_learn
)
1181 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1183 if (gbp
->policy_applied
)
1184 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1186 /* In flow-based mode, GBP is carried in dst_metadata */
1187 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1188 skb
->mark
= md
->gbp
;
1190 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1193 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1194 struct vxlan_sock
*vs
,
1195 struct sk_buff
*skb
)
1197 union vxlan_addr saddr
;
1199 skb_reset_mac_header(skb
);
1200 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1201 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1203 /* Ignore packet loops (and multicast echo) */
1204 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1207 /* Get address from the outer IP header */
1208 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1209 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1210 saddr
.sa
.sa_family
= AF_INET
;
1211 #if IS_ENABLED(CONFIG_IPV6)
1213 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1214 saddr
.sa
.sa_family
= AF_INET6
;
1218 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1219 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1225 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1226 struct sk_buff
*skb
)
1230 if (vxlan_get_sk_family(vs
) == AF_INET
)
1231 err
= IP_ECN_decapsulate(oiph
, skb
);
1232 #if IS_ENABLED(CONFIG_IPV6)
1234 err
= IP6_ECN_decapsulate(oiph
, skb
);
1237 if (unlikely(err
) && log_ecn_error
) {
1238 if (vxlan_get_sk_family(vs
) == AF_INET
)
1239 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1240 &((struct iphdr
*)oiph
)->saddr
,
1241 ((struct iphdr
*)oiph
)->tos
);
1243 net_info_ratelimited("non-ECT from %pI6\n",
1244 &((struct ipv6hdr
*)oiph
)->saddr
);
1249 /* Callback from net/ipv4/udp.c to receive packets */
1250 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1252 struct pcpu_sw_netstats
*stats
;
1253 struct vxlan_dev
*vxlan
;
1254 struct vxlan_sock
*vs
;
1255 struct vxlanhdr unparsed
;
1256 struct vxlan_metadata _md
;
1257 struct vxlan_metadata
*md
= &_md
;
1260 /* Need Vxlan and inner Ethernet header to be present */
1261 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1264 unparsed
= *vxlan_hdr(skb
);
1265 /* VNI flag always required to be set */
1266 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1267 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1268 ntohl(vxlan_hdr(skb
)->vx_flags
),
1269 ntohl(vxlan_hdr(skb
)->vx_vni
));
1270 /* Return non vxlan pkt */
1273 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1274 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1276 vs
= rcu_dereference_sk_user_data(sk
);
1280 vxlan
= vxlan_vs_find_vni(vs
, vxlan_vni(vxlan_hdr(skb
)->vx_vni
));
1284 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
),
1285 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1288 if (vxlan_collect_metadata(vs
)) {
1289 __be32 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1290 struct metadata_dst
*tun_dst
;
1292 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1293 vxlan_vni_to_tun_id(vni
), sizeof(*md
));
1298 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1300 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1302 memset(md
, 0, sizeof(*md
));
1305 /* For backwards compatibility, only allow reserved fields to be
1306 * used by VXLAN extensions if explicitly requested.
1308 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1309 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1311 if (vs
->flags
& VXLAN_F_GBP
)
1312 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1314 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1315 /* If there are any unprocessed flags remaining treat
1316 * this as a malformed packet. This behavior diverges from
1317 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1318 * in reserved fields are to be ignored. The approach here
1319 * maintains compatibility with previous stack code, and also
1320 * is more robust and provides a little more security in
1321 * adding extensions to VXLAN.
1326 if (!vxlan_set_mac(vxlan
, vs
, skb
))
1329 oiph
= skb_network_header(skb
);
1330 skb_reset_network_header(skb
);
1332 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1333 ++vxlan
->dev
->stats
.rx_frame_errors
;
1334 ++vxlan
->dev
->stats
.rx_errors
;
1338 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1339 u64_stats_update_begin(&stats
->syncp
);
1340 stats
->rx_packets
++;
1341 stats
->rx_bytes
+= skb
->len
;
1342 u64_stats_update_end(&stats
->syncp
);
1344 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1348 /* Consume bad packet */
1353 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1355 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1356 struct arphdr
*parp
;
1359 struct neighbour
*n
;
1361 if (dev
->flags
& IFF_NOARP
)
1364 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1365 dev
->stats
.tx_dropped
++;
1368 parp
= arp_hdr(skb
);
1370 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1371 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1372 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1373 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1374 parp
->ar_hln
!= dev
->addr_len
||
1377 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1379 arpptr
+= dev
->addr_len
; /* sha */
1380 memcpy(&sip
, arpptr
, sizeof(sip
));
1381 arpptr
+= sizeof(sip
);
1382 arpptr
+= dev
->addr_len
; /* tha */
1383 memcpy(&tip
, arpptr
, sizeof(tip
));
1385 if (ipv4_is_loopback(tip
) ||
1386 ipv4_is_multicast(tip
))
1389 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1392 struct vxlan_fdb
*f
;
1393 struct sk_buff
*reply
;
1395 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1400 f
= vxlan_find_mac(vxlan
, n
->ha
);
1401 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1402 /* bridge-local neighbor */
1407 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1415 skb_reset_mac_header(reply
);
1416 __skb_pull(reply
, skb_network_offset(reply
));
1417 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1418 reply
->pkt_type
= PACKET_HOST
;
1420 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1421 dev
->stats
.rx_dropped
++;
1422 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1423 union vxlan_addr ipa
= {
1424 .sin
.sin_addr
.s_addr
= tip
,
1425 .sin
.sin_family
= AF_INET
,
1428 vxlan_ip_miss(dev
, &ipa
);
1432 return NETDEV_TX_OK
;
1435 #if IS_ENABLED(CONFIG_IPV6)
1436 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1437 struct neighbour
*n
, bool isrouter
)
1439 struct net_device
*dev
= request
->dev
;
1440 struct sk_buff
*reply
;
1441 struct nd_msg
*ns
, *na
;
1442 struct ipv6hdr
*pip6
;
1444 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1451 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1452 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1453 reply
= alloc_skb(len
, GFP_ATOMIC
);
1457 reply
->protocol
= htons(ETH_P_IPV6
);
1459 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1460 skb_push(reply
, sizeof(struct ethhdr
));
1461 skb_reset_mac_header(reply
);
1463 ns
= (struct nd_msg
*)skb_transport_header(request
);
1465 daddr
= eth_hdr(request
)->h_source
;
1466 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1467 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1468 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1469 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1474 /* Ethernet header */
1475 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1476 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1477 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1478 reply
->protocol
= htons(ETH_P_IPV6
);
1480 skb_pull(reply
, sizeof(struct ethhdr
));
1481 skb_reset_network_header(reply
);
1482 skb_put(reply
, sizeof(struct ipv6hdr
));
1486 pip6
= ipv6_hdr(reply
);
1487 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1489 pip6
->priority
= ipv6_hdr(request
)->priority
;
1490 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1491 pip6
->hop_limit
= 255;
1492 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1493 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1495 skb_pull(reply
, sizeof(struct ipv6hdr
));
1496 skb_reset_transport_header(reply
);
1498 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1500 /* Neighbor Advertisement */
1501 memset(na
, 0, sizeof(*na
)+na_olen
);
1502 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1503 na
->icmph
.icmp6_router
= isrouter
;
1504 na
->icmph
.icmp6_override
= 1;
1505 na
->icmph
.icmp6_solicited
= 1;
1506 na
->target
= ns
->target
;
1507 ether_addr_copy(&na
->opt
[2], n
->ha
);
1508 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1509 na
->opt
[1] = na_olen
>> 3;
1511 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1512 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1513 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1515 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1517 skb_push(reply
, sizeof(struct ipv6hdr
));
1519 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1524 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1526 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1528 const struct ipv6hdr
*iphdr
;
1529 const struct in6_addr
*saddr
, *daddr
;
1530 struct neighbour
*n
;
1531 struct inet6_dev
*in6_dev
;
1533 in6_dev
= __in6_dev_get(dev
);
1537 iphdr
= ipv6_hdr(skb
);
1538 saddr
= &iphdr
->saddr
;
1539 daddr
= &iphdr
->daddr
;
1541 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1542 if (msg
->icmph
.icmp6_code
!= 0 ||
1543 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1546 if (ipv6_addr_loopback(daddr
) ||
1547 ipv6_addr_is_multicast(&msg
->target
))
1550 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1553 struct vxlan_fdb
*f
;
1554 struct sk_buff
*reply
;
1556 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1561 f
= vxlan_find_mac(vxlan
, n
->ha
);
1562 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1563 /* bridge-local neighbor */
1568 reply
= vxlan_na_create(skb
, n
,
1569 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1576 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1577 dev
->stats
.rx_dropped
++;
1579 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1580 union vxlan_addr ipa
= {
1581 .sin6
.sin6_addr
= msg
->target
,
1582 .sin6
.sin6_family
= AF_INET6
,
1585 vxlan_ip_miss(dev
, &ipa
);
1590 return NETDEV_TX_OK
;
1594 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1596 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1597 struct neighbour
*n
;
1599 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1603 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1608 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1611 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1612 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1613 union vxlan_addr ipa
= {
1614 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1615 .sin
.sin_family
= AF_INET
,
1618 vxlan_ip_miss(dev
, &ipa
);
1624 #if IS_ENABLED(CONFIG_IPV6)
1627 struct ipv6hdr
*pip6
;
1629 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1631 pip6
= ipv6_hdr(skb
);
1632 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1633 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1634 union vxlan_addr ipa
= {
1635 .sin6
.sin6_addr
= pip6
->daddr
,
1636 .sin6
.sin6_family
= AF_INET6
,
1639 vxlan_ip_miss(dev
, &ipa
);
1653 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1655 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1657 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1666 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1667 struct vxlan_metadata
*md
)
1669 struct vxlanhdr_gbp
*gbp
;
1674 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1675 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1677 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1678 gbp
->dont_learn
= 1;
1680 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1681 gbp
->policy_applied
= 1;
1683 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1686 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1687 int iphdr_len
, __be32 vni
,
1688 struct vxlan_metadata
*md
, u32 vxflags
,
1691 struct vxlanhdr
*vxh
;
1694 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1696 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1697 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1698 int csum_start
= skb_checksum_start_offset(skb
);
1700 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1701 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1702 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1703 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
1704 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1707 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1708 + VXLAN_HLEN
+ iphdr_len
1709 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1711 /* Need space for new headers (invalidates iph ptr) */
1712 err
= skb_cow_head(skb
, min_headroom
);
1713 if (unlikely(err
)) {
1718 skb
= vlan_hwaccel_push_inside(skb
);
1722 skb
= iptunnel_handle_offloads(skb
, type
);
1724 return PTR_ERR(skb
);
1726 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1727 vxh
->vx_flags
= VXLAN_HF_VNI
;
1728 vxh
->vx_vni
= vxlan_vni_field(vni
);
1730 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1733 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
1734 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
1735 vxh
->vx_flags
|= VXLAN_HF_RCO
;
1737 if (!skb_is_gso(skb
)) {
1738 skb
->ip_summed
= CHECKSUM_NONE
;
1739 skb
->encapsulation
= 0;
1743 if (vxflags
& VXLAN_F_GBP
)
1744 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1746 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1750 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
,
1751 struct sk_buff
*skb
, int oif
, u8 tos
,
1752 __be32 daddr
, __be32
*saddr
,
1753 struct dst_cache
*dst_cache
,
1754 const struct ip_tunnel_info
*info
)
1756 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1757 struct rtable
*rt
= NULL
;
1763 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
1768 memset(&fl4
, 0, sizeof(fl4
));
1769 fl4
.flowi4_oif
= oif
;
1770 fl4
.flowi4_tos
= RT_TOS(tos
);
1771 fl4
.flowi4_mark
= skb
->mark
;
1772 fl4
.flowi4_proto
= IPPROTO_UDP
;
1774 fl4
.saddr
= vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
;
1776 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1780 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
1785 #if IS_ENABLED(CONFIG_IPV6)
1786 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1787 struct sk_buff
*skb
, int oif
, u8 tos
,
1789 const struct in6_addr
*daddr
,
1790 struct in6_addr
*saddr
,
1791 struct dst_cache
*dst_cache
,
1792 const struct ip_tunnel_info
*info
)
1794 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1795 struct dst_entry
*ndst
;
1802 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
1807 memset(&fl6
, 0, sizeof(fl6
));
1808 fl6
.flowi6_oif
= oif
;
1810 fl6
.saddr
= vxlan
->cfg
.saddr
.sin6
.sin6_addr
;
1811 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
1812 fl6
.flowi6_mark
= skb
->mark
;
1813 fl6
.flowi6_proto
= IPPROTO_UDP
;
1815 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1816 vxlan
->vn6_sock
->sock
->sk
,
1819 return ERR_PTR(err
);
1823 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
1828 /* Bypass encapsulation if the destination is local */
1829 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1830 struct vxlan_dev
*dst_vxlan
)
1832 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1833 union vxlan_addr loopback
;
1834 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1835 struct net_device
*dev
= skb
->dev
;
1838 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1839 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1840 skb
->pkt_type
= PACKET_HOST
;
1841 skb
->encapsulation
= 0;
1842 skb
->dev
= dst_vxlan
->dev
;
1843 __skb_pull(skb
, skb_network_offset(skb
));
1845 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1846 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1847 loopback
.sa
.sa_family
= AF_INET
;
1848 #if IS_ENABLED(CONFIG_IPV6)
1850 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1851 loopback
.sa
.sa_family
= AF_INET6
;
1855 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1856 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1858 u64_stats_update_begin(&tx_stats
->syncp
);
1859 tx_stats
->tx_packets
++;
1860 tx_stats
->tx_bytes
+= len
;
1861 u64_stats_update_end(&tx_stats
->syncp
);
1863 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1864 u64_stats_update_begin(&rx_stats
->syncp
);
1865 rx_stats
->rx_packets
++;
1866 rx_stats
->rx_bytes
+= len
;
1867 u64_stats_update_end(&rx_stats
->syncp
);
1869 dev
->stats
.rx_dropped
++;
1873 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1874 struct vxlan_rdst
*rdst
, bool did_rsc
)
1876 struct dst_cache
*dst_cache
;
1877 struct ip_tunnel_info
*info
;
1878 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1880 struct rtable
*rt
= NULL
;
1881 const struct iphdr
*old_iph
;
1882 union vxlan_addr
*dst
;
1883 union vxlan_addr remote_ip
;
1884 struct vxlan_metadata _md
;
1885 struct vxlan_metadata
*md
= &_md
;
1886 __be16 src_port
= 0, dst_port
;
1891 u32 flags
= vxlan
->flags
;
1892 bool udp_sum
= false;
1893 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
1895 info
= skb_tunnel_info(skb
);
1898 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
1899 vni
= rdst
->remote_vni
;
1900 dst
= &rdst
->remote_ip
;
1901 dst_cache
= &rdst
->dst_cache
;
1904 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1908 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
1909 vni
= vxlan_tun_id_to_vni(info
->key
.tun_id
);
1910 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
1911 if (remote_ip
.sa
.sa_family
== AF_INET
)
1912 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
1914 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
1916 dst_cache
= &info
->dst_cache
;
1919 if (vxlan_addr_any(dst
)) {
1921 /* short-circuited back to local bridge */
1922 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1928 old_iph
= ip_hdr(skb
);
1930 ttl
= vxlan
->cfg
.ttl
;
1931 if (!ttl
&& vxlan_addr_multicast(dst
))
1934 tos
= vxlan
->cfg
.tos
;
1936 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1938 label
= vxlan
->cfg
.label
;
1939 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
1940 vxlan
->cfg
.port_max
, true);
1943 ttl
= info
->key
.ttl
;
1944 tos
= info
->key
.tos
;
1945 label
= info
->key
.label
;
1946 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
1948 if (info
->options_len
)
1949 md
= ip_tunnel_info_opts(info
);
1951 md
->gbp
= skb
->mark
;
1954 if (dst
->sa
.sa_family
== AF_INET
) {
1957 if (!vxlan
->vn4_sock
)
1959 sk
= vxlan
->vn4_sock
->sock
->sk
;
1961 rt
= vxlan_get_route(vxlan
, skb
,
1962 rdst
? rdst
->remote_ifindex
: 0, tos
,
1963 dst
->sin
.sin_addr
.s_addr
, &saddr
,
1966 netdev_dbg(dev
, "no route to %pI4\n",
1967 &dst
->sin
.sin_addr
.s_addr
);
1968 dev
->stats
.tx_carrier_errors
++;
1972 if (rt
->dst
.dev
== dev
) {
1973 netdev_dbg(dev
, "circular route to %pI4\n",
1974 &dst
->sin
.sin_addr
.s_addr
);
1975 dev
->stats
.collisions
++;
1979 /* Bypass encapsulation if the destination is local */
1980 if (rt
->rt_flags
& RTCF_LOCAL
&&
1981 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1982 struct vxlan_dev
*dst_vxlan
;
1985 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
1986 dst
->sa
.sa_family
, dst_port
,
1990 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1995 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
1996 else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
)
1999 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2000 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2001 err
= vxlan_build_skb(skb
, &rt
->dst
, sizeof(struct iphdr
),
2002 vni
, md
, flags
, udp_sum
);
2006 udp_tunnel_xmit_skb(rt
, sk
, skb
, saddr
,
2007 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2008 src_port
, dst_port
, xnet
, !udp_sum
);
2009 #if IS_ENABLED(CONFIG_IPV6)
2011 struct dst_entry
*ndst
;
2012 struct in6_addr saddr
;
2015 if (!vxlan
->vn6_sock
)
2017 sk
= vxlan
->vn6_sock
->sock
->sk
;
2019 ndst
= vxlan6_get_route(vxlan
, skb
,
2020 rdst
? rdst
->remote_ifindex
: 0, tos
,
2021 label
, &dst
->sin6
.sin6_addr
, &saddr
,
2024 netdev_dbg(dev
, "no route to %pI6\n",
2025 &dst
->sin6
.sin6_addr
);
2026 dev
->stats
.tx_carrier_errors
++;
2030 if (ndst
->dev
== dev
) {
2031 netdev_dbg(dev
, "circular route to %pI6\n",
2032 &dst
->sin6
.sin6_addr
);
2034 dev
->stats
.collisions
++;
2038 /* Bypass encapsulation if the destination is local */
2039 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2040 if (rt6i_flags
& RTF_LOCAL
&&
2041 !(rt6i_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2042 struct vxlan_dev
*dst_vxlan
;
2045 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2046 dst
->sa
.sa_family
, dst_port
,
2050 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2055 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2057 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2058 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2059 skb_scrub_packet(skb
, xnet
);
2060 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2061 vni
, md
, flags
, udp_sum
);
2066 udp_tunnel6_xmit_skb(ndst
, sk
, skb
, dev
,
2067 &saddr
, &dst
->sin6
.sin6_addr
, tos
, ttl
,
2068 label
, src_port
, dst_port
, !udp_sum
);
2075 dev
->stats
.tx_dropped
++;
2079 /* skb is already freed. */
2084 dev
->stats
.tx_errors
++;
2089 /* Transmit local packets over Vxlan
2091 * Outer IP header inherits ECN and DF from inner header.
2092 * Outer UDP destination is the VXLAN assigned port.
2093 * source port is based on hash of flow
2095 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2097 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2098 const struct ip_tunnel_info
*info
;
2100 bool did_rsc
= false;
2101 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2102 struct vxlan_fdb
*f
;
2104 info
= skb_tunnel_info(skb
);
2106 skb_reset_mac_header(skb
);
2109 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2110 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2111 return arp_reduce(dev
, skb
);
2112 #if IS_ENABLED(CONFIG_IPV6)
2113 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2114 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2115 + sizeof(struct nd_msg
)) &&
2116 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2119 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2120 if (msg
->icmph
.icmp6_code
== 0 &&
2121 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2122 return neigh_reduce(dev
, skb
);
2128 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
2129 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2130 vxlan_xmit_one(skb
, dev
, NULL
, false);
2133 return NETDEV_TX_OK
;
2136 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2139 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2140 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2141 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2142 did_rsc
= route_shortcircuit(dev
, skb
);
2144 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2148 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2150 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2151 !is_multicast_ether_addr(eth
->h_dest
))
2152 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2154 dev
->stats
.tx_dropped
++;
2156 return NETDEV_TX_OK
;
2160 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2161 struct sk_buff
*skb1
;
2167 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2169 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2173 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2176 return NETDEV_TX_OK
;
2179 /* Walk the forwarding table and purge stale entries */
2180 static void vxlan_cleanup(unsigned long arg
)
2182 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2183 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2186 if (!netif_running(vxlan
->dev
))
2189 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2190 struct hlist_node
*p
, *n
;
2192 spin_lock_bh(&vxlan
->hash_lock
);
2193 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2195 = container_of(p
, struct vxlan_fdb
, hlist
);
2196 unsigned long timeout
;
2198 if (f
->state
& NUD_PERMANENT
)
2201 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2202 if (time_before_eq(timeout
, jiffies
)) {
2203 netdev_dbg(vxlan
->dev
,
2204 "garbage collect %pM\n",
2206 f
->state
= NUD_STALE
;
2207 vxlan_fdb_destroy(vxlan
, f
);
2208 } else if (time_before(timeout
, next_timer
))
2209 next_timer
= timeout
;
2211 spin_unlock_bh(&vxlan
->hash_lock
);
2214 mod_timer(&vxlan
->age_timer
, next_timer
);
2217 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2219 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2220 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2222 spin_lock(&vn
->sock_lock
);
2223 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2224 spin_unlock(&vn
->sock_lock
);
2227 /* Setup stats when device is created */
2228 static int vxlan_init(struct net_device
*dev
)
2230 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2237 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2239 struct vxlan_fdb
*f
;
2241 spin_lock_bh(&vxlan
->hash_lock
);
2242 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2244 vxlan_fdb_destroy(vxlan
, f
);
2245 spin_unlock_bh(&vxlan
->hash_lock
);
2248 static void vxlan_uninit(struct net_device
*dev
)
2250 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2252 vxlan_fdb_delete_default(vxlan
);
2254 free_percpu(dev
->tstats
);
2257 /* Start ageing timer and join group when device is brought up */
2258 static int vxlan_open(struct net_device
*dev
)
2260 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2263 ret
= vxlan_sock_add(vxlan
);
2267 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2268 ret
= vxlan_igmp_join(vxlan
);
2269 if (ret
== -EADDRINUSE
)
2272 vxlan_sock_release(vxlan
);
2277 if (vxlan
->cfg
.age_interval
)
2278 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2283 /* Purge the forwarding table */
2284 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2288 spin_lock_bh(&vxlan
->hash_lock
);
2289 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2290 struct hlist_node
*p
, *n
;
2291 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2293 = container_of(p
, struct vxlan_fdb
, hlist
);
2294 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2295 if (!is_zero_ether_addr(f
->eth_addr
))
2296 vxlan_fdb_destroy(vxlan
, f
);
2299 spin_unlock_bh(&vxlan
->hash_lock
);
2302 /* Cleanup timer and forwarding table on shutdown */
2303 static int vxlan_stop(struct net_device
*dev
)
2305 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2306 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2309 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2310 !vxlan_group_used(vn
, vxlan
))
2311 ret
= vxlan_igmp_leave(vxlan
);
2313 del_timer_sync(&vxlan
->age_timer
);
2316 vxlan_sock_release(vxlan
);
2321 /* Stub, nothing needs to be done. */
2322 static void vxlan_set_multicast_list(struct net_device
*dev
)
2326 static int __vxlan_change_mtu(struct net_device
*dev
,
2327 struct net_device
*lowerdev
,
2328 struct vxlan_rdst
*dst
, int new_mtu
, bool strict
)
2330 int max_mtu
= IP_MAX_MTU
;
2333 max_mtu
= lowerdev
->mtu
;
2335 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2336 max_mtu
-= VXLAN6_HEADROOM
;
2338 max_mtu
-= VXLAN_HEADROOM
;
2343 if (new_mtu
> max_mtu
) {
2354 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2356 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2357 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2358 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2359 dst
->remote_ifindex
);
2360 return __vxlan_change_mtu(dev
, lowerdev
, dst
, new_mtu
, true);
2363 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2365 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2366 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2367 __be16 sport
, dport
;
2369 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2370 vxlan
->cfg
.port_max
, true);
2371 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2373 if (ip_tunnel_info_af(info
) == AF_INET
) {
2376 if (!vxlan
->vn4_sock
)
2378 rt
= vxlan_get_route(vxlan
, skb
, 0, info
->key
.tos
,
2379 info
->key
.u
.ipv4
.dst
,
2380 &info
->key
.u
.ipv4
.src
, NULL
, info
);
2385 #if IS_ENABLED(CONFIG_IPV6)
2386 struct dst_entry
*ndst
;
2388 if (!vxlan
->vn6_sock
)
2390 ndst
= vxlan6_get_route(vxlan
, skb
, 0, info
->key
.tos
,
2391 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2392 &info
->key
.u
.ipv6
.src
, NULL
, info
);
2394 return PTR_ERR(ndst
);
2396 #else /* !CONFIG_IPV6 */
2397 return -EPFNOSUPPORT
;
2400 info
->key
.tp_src
= sport
;
2401 info
->key
.tp_dst
= dport
;
2405 static const struct net_device_ops vxlan_netdev_ops
= {
2406 .ndo_init
= vxlan_init
,
2407 .ndo_uninit
= vxlan_uninit
,
2408 .ndo_open
= vxlan_open
,
2409 .ndo_stop
= vxlan_stop
,
2410 .ndo_start_xmit
= vxlan_xmit
,
2411 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2412 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2413 .ndo_change_mtu
= vxlan_change_mtu
,
2414 .ndo_validate_addr
= eth_validate_addr
,
2415 .ndo_set_mac_address
= eth_mac_addr
,
2416 .ndo_fdb_add
= vxlan_fdb_add
,
2417 .ndo_fdb_del
= vxlan_fdb_delete
,
2418 .ndo_fdb_dump
= vxlan_fdb_dump
,
2419 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2422 /* Info for udev, that this is a virtual tunnel endpoint */
2423 static struct device_type vxlan_type
= {
2427 /* Calls the ndo_add_vxlan_port of the caller in order to
2428 * supply the listening VXLAN udp ports. Callers are expected
2429 * to implement the ndo_add_vxlan_port.
2431 void vxlan_get_rx_port(struct net_device
*dev
)
2433 struct vxlan_sock
*vs
;
2434 struct net
*net
= dev_net(dev
);
2435 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2436 sa_family_t sa_family
;
2440 spin_lock(&vn
->sock_lock
);
2441 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2442 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2443 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2444 sa_family
= vxlan_get_sk_family(vs
);
2445 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2449 spin_unlock(&vn
->sock_lock
);
2451 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2453 /* Initialize the device structure. */
2454 static void vxlan_setup(struct net_device
*dev
)
2456 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2459 eth_hw_addr_random(dev
);
2462 dev
->netdev_ops
= &vxlan_netdev_ops
;
2463 dev
->destructor
= free_netdev
;
2464 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2466 dev
->features
|= NETIF_F_LLTX
;
2467 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2468 dev
->features
|= NETIF_F_RXCSUM
;
2469 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2471 dev
->vlan_features
= dev
->features
;
2472 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2473 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2474 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2475 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2476 netif_keep_dst(dev
);
2477 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2478 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
| IFF_NO_QUEUE
;
2480 INIT_LIST_HEAD(&vxlan
->next
);
2481 spin_lock_init(&vxlan
->hash_lock
);
2483 init_timer_deferrable(&vxlan
->age_timer
);
2484 vxlan
->age_timer
.function
= vxlan_cleanup
;
2485 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2487 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2491 gro_cells_init(&vxlan
->gro_cells
, dev
);
2493 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2494 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2497 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2498 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2499 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2500 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2501 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2502 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2503 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2504 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2505 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2506 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
2507 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2508 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2509 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2510 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2511 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2512 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2513 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2514 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2515 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2516 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2517 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2518 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2519 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2520 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2521 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2522 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2523 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2526 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2528 if (tb
[IFLA_ADDRESS
]) {
2529 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2530 pr_debug("invalid link address (not ethernet)\n");
2534 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2535 pr_debug("invalid all zero ethernet address\n");
2536 return -EADDRNOTAVAIL
;
2543 if (data
[IFLA_VXLAN_ID
]) {
2544 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2545 if (id
>= VXLAN_VID_MASK
)
2549 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2550 const struct ifla_vxlan_port_range
*p
2551 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2553 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2554 pr_debug("port range %u .. %u not valid\n",
2555 ntohs(p
->low
), ntohs(p
->high
));
2563 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2564 struct ethtool_drvinfo
*drvinfo
)
2566 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2567 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2570 static const struct ethtool_ops vxlan_ethtool_ops
= {
2571 .get_drvinfo
= vxlan_get_drvinfo
,
2572 .get_link
= ethtool_op_get_link
,
2575 static void vxlan_del_work(struct work_struct
*work
)
2577 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2578 udp_tunnel_sock_release(vs
->sock
);
2582 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2583 __be16 port
, u32 flags
)
2585 struct socket
*sock
;
2586 struct udp_port_cfg udp_conf
;
2589 memset(&udp_conf
, 0, sizeof(udp_conf
));
2592 udp_conf
.family
= AF_INET6
;
2593 udp_conf
.use_udp6_rx_checksums
=
2594 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2595 udp_conf
.ipv6_v6only
= 1;
2597 udp_conf
.family
= AF_INET
;
2600 udp_conf
.local_udp_port
= port
;
2602 /* Open UDP socket */
2603 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2605 return ERR_PTR(err
);
2610 /* Create new listen socket if needed */
2611 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2612 __be16 port
, u32 flags
)
2614 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2615 struct vxlan_sock
*vs
;
2616 struct socket
*sock
;
2618 struct udp_tunnel_sock_cfg tunnel_cfg
;
2620 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2622 return ERR_PTR(-ENOMEM
);
2624 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2625 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2627 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2629 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2631 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port
),
2634 return ERR_CAST(sock
);
2638 atomic_set(&vs
->refcnt
, 1);
2639 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2641 /* Initialize the vxlan udp offloads structure */
2642 vs
->udp_offloads
.port
= port
;
2643 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2644 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2646 spin_lock(&vn
->sock_lock
);
2647 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2648 vxlan_notify_add_rx_port(vs
);
2649 spin_unlock(&vn
->sock_lock
);
2651 /* Mark socket as an encapsulation socket. */
2652 tunnel_cfg
.sk_user_data
= vs
;
2653 tunnel_cfg
.encap_type
= 1;
2654 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
2655 tunnel_cfg
.encap_destroy
= NULL
;
2657 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2662 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2664 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2665 struct vxlan_sock
*vs
= NULL
;
2667 if (!vxlan
->cfg
.no_share
) {
2668 spin_lock(&vn
->sock_lock
);
2669 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2670 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2671 if (vs
&& !atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2672 spin_unlock(&vn
->sock_lock
);
2675 spin_unlock(&vn
->sock_lock
);
2678 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2679 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2682 #if IS_ENABLED(CONFIG_IPV6)
2684 vxlan
->vn6_sock
= vs
;
2687 vxlan
->vn4_sock
= vs
;
2688 vxlan_vs_add_dev(vs
, vxlan
);
2692 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2694 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2695 bool metadata
= vxlan
->flags
& VXLAN_F_COLLECT_METADATA
;
2698 vxlan
->vn4_sock
= NULL
;
2699 #if IS_ENABLED(CONFIG_IPV6)
2700 vxlan
->vn6_sock
= NULL
;
2701 if (ipv6
|| metadata
)
2702 ret
= __vxlan_sock_add(vxlan
, true);
2704 if (!ret
&& (!ipv6
|| metadata
))
2705 ret
= __vxlan_sock_add(vxlan
, false);
2707 vxlan_sock_release(vxlan
);
2711 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2712 struct vxlan_config
*conf
)
2714 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2715 struct vxlan_dev
*vxlan
= netdev_priv(dev
), *tmp
;
2716 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2717 unsigned short needed_headroom
= ETH_HLEN
;
2719 bool use_ipv6
= false;
2720 __be16 default_port
= vxlan
->cfg
.dst_port
;
2721 struct net_device
*lowerdev
= NULL
;
2723 vxlan
->net
= src_net
;
2725 dst
->remote_vni
= conf
->vni
;
2727 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2729 /* Unless IPv6 is explicitly requested, assume IPv4 */
2730 if (!dst
->remote_ip
.sa
.sa_family
)
2731 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2733 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2734 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) {
2735 if (!IS_ENABLED(CONFIG_IPV6
))
2736 return -EPFNOSUPPORT
;
2738 vxlan
->flags
|= VXLAN_F_IPV6
;
2741 if (conf
->label
&& !use_ipv6
) {
2742 pr_info("label only supported in use with IPv6\n");
2746 if (conf
->remote_ifindex
) {
2747 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2748 dst
->remote_ifindex
= conf
->remote_ifindex
;
2751 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2755 #if IS_ENABLED(CONFIG_IPV6)
2757 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2758 if (idev
&& idev
->cnf
.disable_ipv6
) {
2759 pr_info("IPv6 is disabled via sysctl\n");
2766 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2768 needed_headroom
= lowerdev
->hard_header_len
;
2772 err
= __vxlan_change_mtu(dev
, lowerdev
, dst
, conf
->mtu
, false);
2777 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
2778 needed_headroom
+= VXLAN6_HEADROOM
;
2780 needed_headroom
+= VXLAN_HEADROOM
;
2781 dev
->needed_headroom
= needed_headroom
;
2783 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2784 if (!vxlan
->cfg
.dst_port
)
2785 vxlan
->cfg
.dst_port
= default_port
;
2786 vxlan
->flags
|= conf
->flags
;
2788 if (!vxlan
->cfg
.age_interval
)
2789 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2791 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
2792 if (tmp
->cfg
.vni
== conf
->vni
&&
2793 (tmp
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
||
2794 tmp
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) == use_ipv6
&&
2795 tmp
->cfg
.dst_port
== vxlan
->cfg
.dst_port
&&
2796 (tmp
->flags
& VXLAN_F_RCV_FLAGS
) ==
2797 (vxlan
->flags
& VXLAN_F_RCV_FLAGS
))
2801 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2803 /* create an fdb entry for a valid default destination */
2804 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2805 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2806 &vxlan
->default_dst
.remote_ip
,
2807 NUD_REACHABLE
|NUD_PERMANENT
,
2808 NLM_F_EXCL
|NLM_F_CREATE
,
2809 vxlan
->cfg
.dst_port
,
2810 vxlan
->default_dst
.remote_vni
,
2811 vxlan
->default_dst
.remote_ifindex
,
2817 err
= register_netdevice(dev
);
2819 vxlan_fdb_delete_default(vxlan
);
2823 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2828 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
2829 u8 name_assign_type
, struct vxlan_config
*conf
)
2831 struct nlattr
*tb
[IFLA_MAX
+1];
2832 struct net_device
*dev
;
2835 memset(&tb
, 0, sizeof(tb
));
2837 dev
= rtnl_create_link(net
, name
, name_assign_type
,
2838 &vxlan_link_ops
, tb
);
2842 err
= vxlan_dev_configure(net
, dev
, conf
);
2845 return ERR_PTR(err
);
2850 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
2852 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
2853 struct nlattr
*tb
[], struct nlattr
*data
[])
2855 struct vxlan_config conf
;
2858 memset(&conf
, 0, sizeof(conf
));
2860 if (data
[IFLA_VXLAN_ID
])
2861 conf
.vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
2863 if (data
[IFLA_VXLAN_GROUP
]) {
2864 conf
.remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
2865 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2866 if (!IS_ENABLED(CONFIG_IPV6
))
2867 return -EPFNOSUPPORT
;
2869 conf
.remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
2870 conf
.remote_ip
.sa
.sa_family
= AF_INET6
;
2873 if (data
[IFLA_VXLAN_LOCAL
]) {
2874 conf
.saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
2875 conf
.saddr
.sa
.sa_family
= AF_INET
;
2876 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2877 if (!IS_ENABLED(CONFIG_IPV6
))
2878 return -EPFNOSUPPORT
;
2880 /* TODO: respect scope id */
2881 conf
.saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
2882 conf
.saddr
.sa
.sa_family
= AF_INET6
;
2885 if (data
[IFLA_VXLAN_LINK
])
2886 conf
.remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
2888 if (data
[IFLA_VXLAN_TOS
])
2889 conf
.tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2891 if (data
[IFLA_VXLAN_TTL
])
2892 conf
.ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2894 if (data
[IFLA_VXLAN_LABEL
])
2895 conf
.label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
2896 IPV6_FLOWLABEL_MASK
;
2898 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2899 conf
.flags
|= VXLAN_F_LEARN
;
2901 if (data
[IFLA_VXLAN_AGEING
])
2902 conf
.age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2904 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2905 conf
.flags
|= VXLAN_F_PROXY
;
2907 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2908 conf
.flags
|= VXLAN_F_RSC
;
2910 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2911 conf
.flags
|= VXLAN_F_L2MISS
;
2913 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2914 conf
.flags
|= VXLAN_F_L3MISS
;
2916 if (data
[IFLA_VXLAN_LIMIT
])
2917 conf
.addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2919 if (data
[IFLA_VXLAN_COLLECT_METADATA
] &&
2920 nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
2921 conf
.flags
|= VXLAN_F_COLLECT_METADATA
;
2923 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2924 const struct ifla_vxlan_port_range
*p
2925 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2926 conf
.port_min
= ntohs(p
->low
);
2927 conf
.port_max
= ntohs(p
->high
);
2930 if (data
[IFLA_VXLAN_PORT
])
2931 conf
.dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2933 if (data
[IFLA_VXLAN_UDP_CSUM
] &&
2934 !nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2935 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
2937 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2938 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2939 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2941 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2942 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2943 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2945 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
2946 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
2947 conf
.flags
|= VXLAN_F_REMCSUM_TX
;
2949 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
2950 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
2951 conf
.flags
|= VXLAN_F_REMCSUM_RX
;
2953 if (data
[IFLA_VXLAN_GBP
])
2954 conf
.flags
|= VXLAN_F_GBP
;
2956 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
])
2957 conf
.flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
2959 err
= vxlan_dev_configure(src_net
, dev
, &conf
);
2962 pr_info("ifindex %d does not exist\n", conf
.remote_ifindex
);
2966 pr_info("IPv6 is disabled via sysctl\n");
2970 pr_info("duplicate VNI %u\n", be32_to_cpu(conf
.vni
));
2977 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2979 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2980 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2982 spin_lock(&vn
->sock_lock
);
2983 if (!hlist_unhashed(&vxlan
->hlist
))
2984 hlist_del_rcu(&vxlan
->hlist
);
2985 spin_unlock(&vn
->sock_lock
);
2987 gro_cells_destroy(&vxlan
->gro_cells
);
2988 list_del(&vxlan
->next
);
2989 unregister_netdevice_queue(dev
, head
);
2992 static size_t vxlan_get_size(const struct net_device
*dev
)
2995 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2996 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2997 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2998 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2999 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3000 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3001 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
3002 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3003 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3004 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3005 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3006 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3007 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3008 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3009 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3010 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3011 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3012 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3013 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3014 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3015 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3016 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3020 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3022 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3023 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3024 struct ifla_vxlan_port_range ports
= {
3025 .low
= htons(vxlan
->cfg
.port_min
),
3026 .high
= htons(vxlan
->cfg
.port_max
),
3029 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3030 goto nla_put_failure
;
3032 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3033 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3034 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3035 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3036 goto nla_put_failure
;
3037 #if IS_ENABLED(CONFIG_IPV6)
3039 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3040 &dst
->remote_ip
.sin6
.sin6_addr
))
3041 goto nla_put_failure
;
3046 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3047 goto nla_put_failure
;
3049 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3050 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3051 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3052 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3053 goto nla_put_failure
;
3054 #if IS_ENABLED(CONFIG_IPV6)
3056 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3057 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3058 goto nla_put_failure
;
3063 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3064 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3065 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
3066 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3067 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
3068 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3069 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
3070 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
3071 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3072 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
3073 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3074 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
3075 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3076 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
3077 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3078 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3079 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3080 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3081 !(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3082 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3083 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3084 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3085 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3086 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3087 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
3088 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3089 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
3090 goto nla_put_failure
;
3092 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3093 goto nla_put_failure
;
3095 if (vxlan
->flags
& VXLAN_F_GBP
&&
3096 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3097 goto nla_put_failure
;
3099 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3100 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3101 goto nla_put_failure
;
3109 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3111 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3116 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3118 .maxtype
= IFLA_VXLAN_MAX
,
3119 .policy
= vxlan_policy
,
3120 .priv_size
= sizeof(struct vxlan_dev
),
3121 .setup
= vxlan_setup
,
3122 .validate
= vxlan_validate
,
3123 .newlink
= vxlan_newlink
,
3124 .dellink
= vxlan_dellink
,
3125 .get_size
= vxlan_get_size
,
3126 .fill_info
= vxlan_fill_info
,
3127 .get_link_net
= vxlan_get_link_net
,
3130 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3131 struct net_device
*dev
)
3133 struct vxlan_dev
*vxlan
, *next
;
3134 LIST_HEAD(list_kill
);
3136 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3137 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3139 /* In case we created vxlan device with carrier
3140 * and we loose the carrier due to module unload
3141 * we also need to remove vxlan device. In other
3142 * cases, it's not necessary and remote_ifindex
3143 * is 0 here, so no matches.
3145 if (dst
->remote_ifindex
== dev
->ifindex
)
3146 vxlan_dellink(vxlan
->dev
, &list_kill
);
3149 unregister_netdevice_many(&list_kill
);
3152 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
3153 unsigned long event
, void *ptr
)
3155 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3156 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3158 if (event
== NETDEV_UNREGISTER
)
3159 vxlan_handle_lowerdev_unregister(vn
, dev
);
3164 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3165 .notifier_call
= vxlan_lowerdev_event
,
3168 static __net_init
int vxlan_init_net(struct net
*net
)
3170 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3173 INIT_LIST_HEAD(&vn
->vxlan_list
);
3174 spin_lock_init(&vn
->sock_lock
);
3176 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3177 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3182 static void __net_exit
vxlan_exit_net(struct net
*net
)
3184 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3185 struct vxlan_dev
*vxlan
, *next
;
3186 struct net_device
*dev
, *aux
;
3190 for_each_netdev_safe(net
, dev
, aux
)
3191 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3192 unregister_netdevice_queue(dev
, &list
);
3194 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3195 /* If vxlan->dev is in the same netns, it has already been added
3196 * to the list by the previous loop.
3198 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3199 gro_cells_destroy(&vxlan
->gro_cells
);
3200 unregister_netdevice_queue(vxlan
->dev
, &list
);
3204 unregister_netdevice_many(&list
);
3208 static struct pernet_operations vxlan_net_ops
= {
3209 .init
= vxlan_init_net
,
3210 .exit
= vxlan_exit_net
,
3211 .id
= &vxlan_net_id
,
3212 .size
= sizeof(struct vxlan_net
),
3215 static int __init
vxlan_init_module(void)
3219 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3223 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3225 rc
= register_pernet_subsys(&vxlan_net_ops
);
3229 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3233 rc
= rtnl_link_register(&vxlan_link_ops
);
3239 unregister_netdevice_notifier(&vxlan_notifier_block
);
3241 unregister_pernet_subsys(&vxlan_net_ops
);
3243 destroy_workqueue(vxlan_wq
);
3246 late_initcall(vxlan_init_module
);
3248 static void __exit
vxlan_cleanup_module(void)
3250 rtnl_link_unregister(&vxlan_link_ops
);
3251 unregister_netdevice_notifier(&vxlan_notifier_block
);
3252 destroy_workqueue(vxlan_wq
);
3253 unregister_pernet_subsys(&vxlan_net_ops
);
3254 /* rcu_barrier() is called by netns */
3256 module_exit(vxlan_cleanup_module
);
3258 MODULE_LICENSE("GPL");
3259 MODULE_VERSION(VXLAN_VERSION
);
3260 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3261 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3262 MODULE_ALIAS_RTNL_LINK("vxlan");