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 */
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
= eth_gro_receive(head
, skb
);
611 skb_gro_remcsum_cleanup(skb
, &grc
);
612 NAPI_GRO_CB(skb
)->flush
|= flush
;
617 static int vxlan_gro_complete(struct sk_buff
*skb
, int nhoff
,
618 struct udp_offload
*uoff
)
620 udp_tunnel_gro_complete(skb
, nhoff
);
622 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
625 /* Notify netdevs that UDP port started listening */
626 static void vxlan_notify_add_rx_port(struct vxlan_sock
*vs
)
628 struct net_device
*dev
;
629 struct sock
*sk
= vs
->sock
->sk
;
630 struct net
*net
= sock_net(sk
);
631 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
632 __be16 port
= inet_sk(sk
)->inet_sport
;
635 if (sa_family
== AF_INET
) {
636 err
= udp_add_offload(net
, &vs
->udp_offloads
);
638 pr_warn("vxlan: udp_add_offload failed with status %d\n", err
);
642 for_each_netdev_rcu(net
, dev
) {
643 if (dev
->netdev_ops
->ndo_add_vxlan_port
)
644 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
650 /* Notify netdevs that UDP port is no more listening */
651 static void vxlan_notify_del_rx_port(struct vxlan_sock
*vs
)
653 struct net_device
*dev
;
654 struct sock
*sk
= vs
->sock
->sk
;
655 struct net
*net
= sock_net(sk
);
656 sa_family_t sa_family
= vxlan_get_sk_family(vs
);
657 __be16 port
= inet_sk(sk
)->inet_sport
;
660 for_each_netdev_rcu(net
, dev
) {
661 if (dev
->netdev_ops
->ndo_del_vxlan_port
)
662 dev
->netdev_ops
->ndo_del_vxlan_port(dev
, sa_family
,
667 if (sa_family
== AF_INET
)
668 udp_del_offload(&vs
->udp_offloads
);
671 /* Add new entry to forwarding table -- assumes lock held */
672 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
673 const u8
*mac
, union vxlan_addr
*ip
,
674 __u16 state
, __u16 flags
,
675 __be16 port
, __be32 vni
, __u32 ifindex
,
678 struct vxlan_rdst
*rd
= NULL
;
682 f
= __vxlan_find_mac(vxlan
, mac
);
684 if (flags
& NLM_F_EXCL
) {
685 netdev_dbg(vxlan
->dev
,
686 "lost race to create %pM\n", mac
);
689 if (f
->state
!= state
) {
691 f
->updated
= jiffies
;
694 if (f
->flags
!= ndm_flags
) {
695 f
->flags
= ndm_flags
;
696 f
->updated
= jiffies
;
699 if ((flags
& NLM_F_REPLACE
)) {
700 /* Only change unicasts */
701 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
702 is_zero_ether_addr(f
->eth_addr
))) {
703 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
708 if ((flags
& NLM_F_APPEND
) &&
709 (is_multicast_ether_addr(f
->eth_addr
) ||
710 is_zero_ether_addr(f
->eth_addr
))) {
711 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
,
719 if (!(flags
& NLM_F_CREATE
))
722 if (vxlan
->cfg
.addrmax
&&
723 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
726 /* Disallow replace to add a multicast entry */
727 if ((flags
& NLM_F_REPLACE
) &&
728 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
731 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
732 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
738 f
->flags
= ndm_flags
;
739 f
->updated
= f
->used
= jiffies
;
740 INIT_LIST_HEAD(&f
->remotes
);
741 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
743 vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
746 hlist_add_head_rcu(&f
->hlist
,
747 vxlan_fdb_head(vxlan
, mac
));
752 rd
= first_remote_rtnl(f
);
753 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
759 static void vxlan_fdb_free(struct rcu_head
*head
)
761 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
762 struct vxlan_rdst
*rd
, *nd
;
764 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
765 dst_cache_destroy(&rd
->dst_cache
);
771 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
773 netdev_dbg(vxlan
->dev
,
774 "delete %pM\n", f
->eth_addr
);
777 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
779 hlist_del_rcu(&f
->hlist
);
780 call_rcu(&f
->rcu
, vxlan_fdb_free
);
783 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
784 union vxlan_addr
*ip
, __be16
*port
, __be32
*vni
,
787 struct net
*net
= dev_net(vxlan
->dev
);
791 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
795 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
796 if (remote
->sa
.sa_family
== AF_INET
) {
797 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
798 ip
->sa
.sa_family
= AF_INET
;
799 #if IS_ENABLED(CONFIG_IPV6)
801 ip
->sin6
.sin6_addr
= in6addr_any
;
802 ip
->sa
.sa_family
= AF_INET6
;
808 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
810 *port
= nla_get_be16(tb
[NDA_PORT
]);
812 *port
= vxlan
->cfg
.dst_port
;
816 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
818 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
820 *vni
= vxlan
->default_dst
.remote_vni
;
823 if (tb
[NDA_IFINDEX
]) {
824 struct net_device
*tdev
;
826 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
828 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
829 tdev
= __dev_get_by_index(net
, *ifindex
);
831 return -EADDRNOTAVAIL
;
839 /* Add static entry (via netlink) */
840 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
841 struct net_device
*dev
,
842 const unsigned char *addr
, u16 vid
, u16 flags
)
844 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
845 /* struct net *net = dev_net(vxlan->dev); */
852 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
853 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
858 if (tb
[NDA_DST
] == NULL
)
861 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
865 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
866 return -EAFNOSUPPORT
;
868 spin_lock_bh(&vxlan
->hash_lock
);
869 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
870 port
, vni
, ifindex
, ndm
->ndm_flags
);
871 spin_unlock_bh(&vxlan
->hash_lock
);
876 /* Delete entry (via netlink) */
877 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
878 struct net_device
*dev
,
879 const unsigned char *addr
, u16 vid
)
881 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
883 struct vxlan_rdst
*rd
= NULL
;
890 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &vni
, &ifindex
);
896 spin_lock_bh(&vxlan
->hash_lock
);
897 f
= vxlan_find_mac(vxlan
, addr
);
901 if (!vxlan_addr_any(&ip
)) {
902 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
909 /* remove a destination if it's not the only one on the list,
910 * otherwise destroy the fdb entry
912 if (rd
&& !list_is_singular(&f
->remotes
)) {
913 list_del_rcu(&rd
->list
);
914 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
919 vxlan_fdb_destroy(vxlan
, f
);
922 spin_unlock_bh(&vxlan
->hash_lock
);
927 /* Dump forwarding table */
928 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
929 struct net_device
*dev
,
930 struct net_device
*filter_dev
, int idx
)
932 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
935 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
939 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
940 struct vxlan_rdst
*rd
;
942 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
943 if (idx
< cb
->args
[0])
946 err
= vxlan_fdb_info(skb
, vxlan
, f
,
947 NETLINK_CB(cb
->skb
).portid
,
964 /* Watch incoming packets to learn mapping between Ethernet address
965 * and Tunnel endpoint.
966 * Return true if packet is bogus and should be dropped.
968 static bool vxlan_snoop(struct net_device
*dev
,
969 union vxlan_addr
*src_ip
, const u8
*src_mac
)
971 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
974 f
= vxlan_find_mac(vxlan
, src_mac
);
976 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
978 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
981 /* Don't migrate static entries, drop packets */
982 if (f
->state
& NUD_NOARP
)
987 "%pM migrated from %pIS to %pIS\n",
988 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
990 rdst
->remote_ip
= *src_ip
;
991 f
->updated
= jiffies
;
992 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
994 /* learned new entry */
995 spin_lock(&vxlan
->hash_lock
);
997 /* close off race between vxlan_flush and incoming packets */
998 if (netif_running(dev
))
999 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1001 NLM_F_EXCL
|NLM_F_CREATE
,
1002 vxlan
->cfg
.dst_port
,
1003 vxlan
->default_dst
.remote_vni
,
1005 spin_unlock(&vxlan
->hash_lock
);
1011 /* See if multicast group is already in use by other ID */
1012 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1014 struct vxlan_dev
*vxlan
;
1015 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1017 /* The vxlan_sock is only used by dev, leaving group has
1018 * no effect on other vxlan devices.
1020 if (family
== AF_INET
&& dev
->vn4_sock
&&
1021 atomic_read(&dev
->vn4_sock
->refcnt
) == 1)
1023 #if IS_ENABLED(CONFIG_IPV6)
1024 if (family
== AF_INET6
&& dev
->vn6_sock
&&
1025 atomic_read(&dev
->vn6_sock
->refcnt
) == 1)
1029 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1030 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1033 if (family
== AF_INET
&& vxlan
->vn4_sock
!= dev
->vn4_sock
)
1035 #if IS_ENABLED(CONFIG_IPV6)
1036 if (family
== AF_INET6
&& vxlan
->vn6_sock
!= dev
->vn6_sock
)
1040 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1041 &dev
->default_dst
.remote_ip
))
1044 if (vxlan
->default_dst
.remote_ifindex
!=
1045 dev
->default_dst
.remote_ifindex
)
1054 static void __vxlan_sock_release(struct vxlan_sock
*vs
)
1056 struct vxlan_net
*vn
;
1060 if (!atomic_dec_and_test(&vs
->refcnt
))
1063 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1064 spin_lock(&vn
->sock_lock
);
1065 hlist_del_rcu(&vs
->hlist
);
1066 vxlan_notify_del_rx_port(vs
);
1067 spin_unlock(&vn
->sock_lock
);
1069 queue_work(vxlan_wq
, &vs
->del_work
);
1072 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1074 __vxlan_sock_release(vxlan
->vn4_sock
);
1075 #if IS_ENABLED(CONFIG_IPV6)
1076 __vxlan_sock_release(vxlan
->vn6_sock
);
1080 /* Update multicast group membership when first VNI on
1081 * multicast address is brought up
1083 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1086 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1087 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1090 if (ip
->sa
.sa_family
== AF_INET
) {
1091 struct ip_mreqn mreq
= {
1092 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1093 .imr_ifindex
= ifindex
,
1096 sk
= vxlan
->vn4_sock
->sock
->sk
;
1098 ret
= ip_mc_join_group(sk
, &mreq
);
1100 #if IS_ENABLED(CONFIG_IPV6)
1102 sk
= vxlan
->vn6_sock
->sock
->sk
;
1104 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1105 &ip
->sin6
.sin6_addr
);
1113 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1114 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1117 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1118 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1121 if (ip
->sa
.sa_family
== AF_INET
) {
1122 struct ip_mreqn mreq
= {
1123 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1124 .imr_ifindex
= ifindex
,
1127 sk
= vxlan
->vn4_sock
->sock
->sk
;
1129 ret
= ip_mc_leave_group(sk
, &mreq
);
1131 #if IS_ENABLED(CONFIG_IPV6)
1133 sk
= vxlan
->vn6_sock
->sock
->sk
;
1135 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1136 &ip
->sin6
.sin6_addr
);
1144 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1145 struct sk_buff
*skb
, u32 vxflags
)
1147 size_t start
, offset
, plen
;
1149 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1152 start
= vxlan_rco_start(unparsed
->vx_vni
);
1153 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1155 plen
= sizeof(struct vxlanhdr
) + offset
+ sizeof(u16
);
1157 if (!pskb_may_pull(skb
, plen
))
1160 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1161 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1163 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1164 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1168 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1169 struct sk_buff
*skb
, u32 vxflags
,
1170 struct vxlan_metadata
*md
)
1172 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1173 struct metadata_dst
*tun_dst
;
1175 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1178 md
->gbp
= ntohs(gbp
->policy_id
);
1180 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1182 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1183 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1185 if (gbp
->dont_learn
)
1186 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1188 if (gbp
->policy_applied
)
1189 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1191 /* In flow-based mode, GBP is carried in dst_metadata */
1192 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1193 skb
->mark
= md
->gbp
;
1195 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1198 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1199 struct vxlan_sock
*vs
,
1200 struct sk_buff
*skb
)
1202 union vxlan_addr saddr
;
1204 skb_reset_mac_header(skb
);
1205 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1206 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1208 /* Ignore packet loops (and multicast echo) */
1209 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1212 /* Get address from the outer IP header */
1213 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1214 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1215 saddr
.sa
.sa_family
= AF_INET
;
1216 #if IS_ENABLED(CONFIG_IPV6)
1218 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1219 saddr
.sa
.sa_family
= AF_INET6
;
1223 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1224 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
))
1230 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1231 struct sk_buff
*skb
)
1235 if (vxlan_get_sk_family(vs
) == AF_INET
)
1236 err
= IP_ECN_decapsulate(oiph
, skb
);
1237 #if IS_ENABLED(CONFIG_IPV6)
1239 err
= IP6_ECN_decapsulate(oiph
, skb
);
1242 if (unlikely(err
) && log_ecn_error
) {
1243 if (vxlan_get_sk_family(vs
) == AF_INET
)
1244 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1245 &((struct iphdr
*)oiph
)->saddr
,
1246 ((struct iphdr
*)oiph
)->tos
);
1248 net_info_ratelimited("non-ECT from %pI6\n",
1249 &((struct ipv6hdr
*)oiph
)->saddr
);
1254 /* Callback from net/ipv4/udp.c to receive packets */
1255 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1257 struct pcpu_sw_netstats
*stats
;
1258 struct vxlan_dev
*vxlan
;
1259 struct vxlan_sock
*vs
;
1260 struct vxlanhdr unparsed
;
1261 struct vxlan_metadata _md
;
1262 struct vxlan_metadata
*md
= &_md
;
1265 /* Need Vxlan and inner Ethernet header to be present */
1266 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1269 unparsed
= *vxlan_hdr(skb
);
1270 /* VNI flag always required to be set */
1271 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1272 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1273 ntohl(vxlan_hdr(skb
)->vx_flags
),
1274 ntohl(vxlan_hdr(skb
)->vx_vni
));
1275 /* Return non vxlan pkt */
1278 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1279 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1281 vs
= rcu_dereference_sk_user_data(sk
);
1285 vxlan
= vxlan_vs_find_vni(vs
, vxlan_vni(vxlan_hdr(skb
)->vx_vni
));
1289 if (iptunnel_pull_header(skb
, VXLAN_HLEN
, htons(ETH_P_TEB
),
1290 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1293 if (vxlan_collect_metadata(vs
)) {
1294 __be32 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1295 struct metadata_dst
*tun_dst
;
1297 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1298 vxlan_vni_to_tun_id(vni
), sizeof(*md
));
1303 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1305 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1307 memset(md
, 0, sizeof(*md
));
1310 /* For backwards compatibility, only allow reserved fields to be
1311 * used by VXLAN extensions if explicitly requested.
1313 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1314 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1316 if (vs
->flags
& VXLAN_F_GBP
)
1317 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1319 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1320 /* If there are any unprocessed flags remaining treat
1321 * this as a malformed packet. This behavior diverges from
1322 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1323 * in reserved fields are to be ignored. The approach here
1324 * maintains compatibility with previous stack code, and also
1325 * is more robust and provides a little more security in
1326 * adding extensions to VXLAN.
1331 if (!vxlan_set_mac(vxlan
, vs
, skb
))
1334 oiph
= skb_network_header(skb
);
1335 skb_reset_network_header(skb
);
1337 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1338 ++vxlan
->dev
->stats
.rx_frame_errors
;
1339 ++vxlan
->dev
->stats
.rx_errors
;
1343 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1344 u64_stats_update_begin(&stats
->syncp
);
1345 stats
->rx_packets
++;
1346 stats
->rx_bytes
+= skb
->len
;
1347 u64_stats_update_end(&stats
->syncp
);
1349 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1353 /* Consume bad packet */
1358 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1360 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1361 struct arphdr
*parp
;
1364 struct neighbour
*n
;
1366 if (dev
->flags
& IFF_NOARP
)
1369 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1370 dev
->stats
.tx_dropped
++;
1373 parp
= arp_hdr(skb
);
1375 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1376 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1377 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1378 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1379 parp
->ar_hln
!= dev
->addr_len
||
1382 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1384 arpptr
+= dev
->addr_len
; /* sha */
1385 memcpy(&sip
, arpptr
, sizeof(sip
));
1386 arpptr
+= sizeof(sip
);
1387 arpptr
+= dev
->addr_len
; /* tha */
1388 memcpy(&tip
, arpptr
, sizeof(tip
));
1390 if (ipv4_is_loopback(tip
) ||
1391 ipv4_is_multicast(tip
))
1394 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1397 struct vxlan_fdb
*f
;
1398 struct sk_buff
*reply
;
1400 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1405 f
= vxlan_find_mac(vxlan
, n
->ha
);
1406 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1407 /* bridge-local neighbor */
1412 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1420 skb_reset_mac_header(reply
);
1421 __skb_pull(reply
, skb_network_offset(reply
));
1422 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1423 reply
->pkt_type
= PACKET_HOST
;
1425 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1426 dev
->stats
.rx_dropped
++;
1427 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1428 union vxlan_addr ipa
= {
1429 .sin
.sin_addr
.s_addr
= tip
,
1430 .sin
.sin_family
= AF_INET
,
1433 vxlan_ip_miss(dev
, &ipa
);
1437 return NETDEV_TX_OK
;
1440 #if IS_ENABLED(CONFIG_IPV6)
1441 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1442 struct neighbour
*n
, bool isrouter
)
1444 struct net_device
*dev
= request
->dev
;
1445 struct sk_buff
*reply
;
1446 struct nd_msg
*ns
, *na
;
1447 struct ipv6hdr
*pip6
;
1449 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1456 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1457 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1458 reply
= alloc_skb(len
, GFP_ATOMIC
);
1462 reply
->protocol
= htons(ETH_P_IPV6
);
1464 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1465 skb_push(reply
, sizeof(struct ethhdr
));
1466 skb_reset_mac_header(reply
);
1468 ns
= (struct nd_msg
*)skb_transport_header(request
);
1470 daddr
= eth_hdr(request
)->h_source
;
1471 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1472 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1473 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1474 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1479 /* Ethernet header */
1480 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1481 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1482 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1483 reply
->protocol
= htons(ETH_P_IPV6
);
1485 skb_pull(reply
, sizeof(struct ethhdr
));
1486 skb_reset_network_header(reply
);
1487 skb_put(reply
, sizeof(struct ipv6hdr
));
1491 pip6
= ipv6_hdr(reply
);
1492 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1494 pip6
->priority
= ipv6_hdr(request
)->priority
;
1495 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1496 pip6
->hop_limit
= 255;
1497 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1498 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1500 skb_pull(reply
, sizeof(struct ipv6hdr
));
1501 skb_reset_transport_header(reply
);
1503 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1505 /* Neighbor Advertisement */
1506 memset(na
, 0, sizeof(*na
)+na_olen
);
1507 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1508 na
->icmph
.icmp6_router
= isrouter
;
1509 na
->icmph
.icmp6_override
= 1;
1510 na
->icmph
.icmp6_solicited
= 1;
1511 na
->target
= ns
->target
;
1512 ether_addr_copy(&na
->opt
[2], n
->ha
);
1513 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1514 na
->opt
[1] = na_olen
>> 3;
1516 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1517 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1518 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1520 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1522 skb_push(reply
, sizeof(struct ipv6hdr
));
1524 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1529 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
1531 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1533 const struct ipv6hdr
*iphdr
;
1534 const struct in6_addr
*saddr
, *daddr
;
1535 struct neighbour
*n
;
1536 struct inet6_dev
*in6_dev
;
1538 in6_dev
= __in6_dev_get(dev
);
1542 iphdr
= ipv6_hdr(skb
);
1543 saddr
= &iphdr
->saddr
;
1544 daddr
= &iphdr
->daddr
;
1546 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1547 if (msg
->icmph
.icmp6_code
!= 0 ||
1548 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1551 if (ipv6_addr_loopback(daddr
) ||
1552 ipv6_addr_is_multicast(&msg
->target
))
1555 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1558 struct vxlan_fdb
*f
;
1559 struct sk_buff
*reply
;
1561 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1566 f
= vxlan_find_mac(vxlan
, n
->ha
);
1567 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1568 /* bridge-local neighbor */
1573 reply
= vxlan_na_create(skb
, n
,
1574 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1581 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1582 dev
->stats
.rx_dropped
++;
1584 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1585 union vxlan_addr ipa
= {
1586 .sin6
.sin6_addr
= msg
->target
,
1587 .sin6
.sin6_family
= AF_INET6
,
1590 vxlan_ip_miss(dev
, &ipa
);
1595 return NETDEV_TX_OK
;
1599 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1601 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1602 struct neighbour
*n
;
1604 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1608 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1613 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1616 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1617 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1618 union vxlan_addr ipa
= {
1619 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1620 .sin
.sin_family
= AF_INET
,
1623 vxlan_ip_miss(dev
, &ipa
);
1629 #if IS_ENABLED(CONFIG_IPV6)
1632 struct ipv6hdr
*pip6
;
1634 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1636 pip6
= ipv6_hdr(skb
);
1637 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1638 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1639 union vxlan_addr ipa
= {
1640 .sin6
.sin6_addr
= pip6
->daddr
,
1641 .sin6
.sin6_family
= AF_INET6
,
1644 vxlan_ip_miss(dev
, &ipa
);
1658 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1660 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1662 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1671 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1672 struct vxlan_metadata
*md
)
1674 struct vxlanhdr_gbp
*gbp
;
1679 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1680 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1682 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1683 gbp
->dont_learn
= 1;
1685 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1686 gbp
->policy_applied
= 1;
1688 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1691 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1692 int iphdr_len
, __be32 vni
,
1693 struct vxlan_metadata
*md
, u32 vxflags
,
1696 struct vxlanhdr
*vxh
;
1699 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1701 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1702 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1703 int csum_start
= skb_checksum_start_offset(skb
);
1705 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1706 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1707 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1708 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
1709 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1712 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1713 + VXLAN_HLEN
+ iphdr_len
1714 + (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
1716 /* Need space for new headers (invalidates iph ptr) */
1717 err
= skb_cow_head(skb
, min_headroom
);
1718 if (unlikely(err
)) {
1723 skb
= vlan_hwaccel_push_inside(skb
);
1727 skb
= iptunnel_handle_offloads(skb
, type
);
1729 return PTR_ERR(skb
);
1731 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1732 vxh
->vx_flags
= VXLAN_HF_VNI
;
1733 vxh
->vx_vni
= vxlan_vni_field(vni
);
1735 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1738 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
1739 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
1740 vxh
->vx_flags
|= VXLAN_HF_RCO
;
1742 if (!skb_is_gso(skb
)) {
1743 skb
->ip_summed
= CHECKSUM_NONE
;
1744 skb
->encapsulation
= 0;
1748 if (vxflags
& VXLAN_F_GBP
)
1749 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1751 skb_set_inner_protocol(skb
, htons(ETH_P_TEB
));
1755 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
,
1756 struct sk_buff
*skb
, int oif
, u8 tos
,
1757 __be32 daddr
, __be32
*saddr
,
1758 struct dst_cache
*dst_cache
,
1759 struct ip_tunnel_info
*info
)
1761 struct rtable
*rt
= NULL
;
1762 bool use_cache
= false;
1765 /* when the ip_tunnel_info is availble, the tos used for lookup is
1766 * packet independent, so we can use the cache
1768 if (!skb
->mark
&& (!tos
|| info
)) {
1770 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
1775 memset(&fl4
, 0, sizeof(fl4
));
1776 fl4
.flowi4_oif
= oif
;
1777 fl4
.flowi4_tos
= RT_TOS(tos
);
1778 fl4
.flowi4_mark
= skb
->mark
;
1779 fl4
.flowi4_proto
= IPPROTO_UDP
;
1781 fl4
.saddr
= vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
;
1783 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1787 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
1792 #if IS_ENABLED(CONFIG_IPV6)
1793 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1794 struct sk_buff
*skb
, int oif
,
1795 const struct in6_addr
*daddr
,
1796 struct in6_addr
*saddr
,
1797 struct dst_cache
*dst_cache
)
1799 struct dst_entry
*ndst
;
1804 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
1809 memset(&fl6
, 0, sizeof(fl6
));
1810 fl6
.flowi6_oif
= oif
;
1812 fl6
.saddr
= vxlan
->cfg
.saddr
.sin6
.sin6_addr
;
1813 fl6
.flowi6_mark
= skb
->mark
;
1814 fl6
.flowi6_proto
= IPPROTO_UDP
;
1816 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1817 vxlan
->vn6_sock
->sock
->sk
,
1820 return ERR_PTR(err
);
1824 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
1829 /* Bypass encapsulation if the destination is local */
1830 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1831 struct vxlan_dev
*dst_vxlan
)
1833 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1834 union vxlan_addr loopback
;
1835 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1836 struct net_device
*dev
= skb
->dev
;
1839 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1840 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1841 skb
->pkt_type
= PACKET_HOST
;
1842 skb
->encapsulation
= 0;
1843 skb
->dev
= dst_vxlan
->dev
;
1844 __skb_pull(skb
, skb_network_offset(skb
));
1846 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1847 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1848 loopback
.sa
.sa_family
= AF_INET
;
1849 #if IS_ENABLED(CONFIG_IPV6)
1851 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1852 loopback
.sa
.sa_family
= AF_INET6
;
1856 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1857 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
);
1859 u64_stats_update_begin(&tx_stats
->syncp
);
1860 tx_stats
->tx_packets
++;
1861 tx_stats
->tx_bytes
+= len
;
1862 u64_stats_update_end(&tx_stats
->syncp
);
1864 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
1865 u64_stats_update_begin(&rx_stats
->syncp
);
1866 rx_stats
->rx_packets
++;
1867 rx_stats
->rx_bytes
+= len
;
1868 u64_stats_update_end(&rx_stats
->syncp
);
1870 dev
->stats
.rx_dropped
++;
1874 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1875 struct vxlan_rdst
*rdst
, bool did_rsc
)
1877 struct dst_cache
*dst_cache
;
1878 struct ip_tunnel_info
*info
;
1879 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1881 struct rtable
*rt
= NULL
;
1882 const struct iphdr
*old_iph
;
1883 union vxlan_addr
*dst
;
1884 union vxlan_addr remote_ip
;
1885 struct vxlan_metadata _md
;
1886 struct vxlan_metadata
*md
= &_md
;
1887 __be16 src_port
= 0, dst_port
;
1892 u32 flags
= vxlan
->flags
;
1893 bool udp_sum
= false;
1894 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
1896 info
= skb_tunnel_info(skb
);
1899 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
1900 vni
= rdst
->remote_vni
;
1901 dst
= &rdst
->remote_ip
;
1902 dst_cache
= &rdst
->dst_cache
;
1905 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1909 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
1910 vni
= vxlan_tun_id_to_vni(info
->key
.tun_id
);
1911 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
1912 if (remote_ip
.sa
.sa_family
== AF_INET
)
1913 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
1915 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
1917 dst_cache
= &info
->dst_cache
;
1920 if (vxlan_addr_any(dst
)) {
1922 /* short-circuited back to local bridge */
1923 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1929 old_iph
= ip_hdr(skb
);
1931 ttl
= vxlan
->cfg
.ttl
;
1932 if (!ttl
&& vxlan_addr_multicast(dst
))
1935 tos
= vxlan
->cfg
.tos
;
1937 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
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 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
1947 if (info
->options_len
)
1948 md
= ip_tunnel_info_opts(info
);
1950 md
->gbp
= skb
->mark
;
1953 if (dst
->sa
.sa_family
== AF_INET
) {
1956 if (!vxlan
->vn4_sock
)
1958 sk
= vxlan
->vn4_sock
->sock
->sk
;
1960 rt
= vxlan_get_route(vxlan
, skb
,
1961 rdst
? rdst
->remote_ifindex
: 0, tos
,
1962 dst
->sin
.sin_addr
.s_addr
, &saddr
,
1965 netdev_dbg(dev
, "no route to %pI4\n",
1966 &dst
->sin
.sin_addr
.s_addr
);
1967 dev
->stats
.tx_carrier_errors
++;
1971 if (rt
->dst
.dev
== dev
) {
1972 netdev_dbg(dev
, "circular route to %pI4\n",
1973 &dst
->sin
.sin_addr
.s_addr
);
1974 dev
->stats
.collisions
++;
1978 /* Bypass encapsulation if the destination is local */
1979 if (rt
->rt_flags
& RTCF_LOCAL
&&
1980 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1981 struct vxlan_dev
*dst_vxlan
;
1984 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
1985 dst
->sa
.sa_family
, dst_port
,
1989 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1994 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
1995 else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
)
1998 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1999 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2000 err
= vxlan_build_skb(skb
, &rt
->dst
, sizeof(struct iphdr
),
2001 vni
, md
, flags
, udp_sum
);
2005 udp_tunnel_xmit_skb(rt
, sk
, skb
, saddr
,
2006 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2007 src_port
, dst_port
, xnet
, !udp_sum
);
2008 #if IS_ENABLED(CONFIG_IPV6)
2010 struct dst_entry
*ndst
;
2011 struct in6_addr saddr
;
2014 if (!vxlan
->vn6_sock
)
2016 sk
= vxlan
->vn6_sock
->sock
->sk
;
2018 ndst
= vxlan6_get_route(vxlan
, skb
,
2019 rdst
? rdst
->remote_ifindex
: 0,
2020 &dst
->sin6
.sin6_addr
, &saddr
,
2023 netdev_dbg(dev
, "no route to %pI6\n",
2024 &dst
->sin6
.sin6_addr
);
2025 dev
->stats
.tx_carrier_errors
++;
2029 if (ndst
->dev
== dev
) {
2030 netdev_dbg(dev
, "circular route to %pI6\n",
2031 &dst
->sin6
.sin6_addr
);
2033 dev
->stats
.collisions
++;
2037 /* Bypass encapsulation if the destination is local */
2038 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2039 if (rt6i_flags
& RTF_LOCAL
&&
2040 !(rt6i_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2041 struct vxlan_dev
*dst_vxlan
;
2044 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2045 dst
->sa
.sa_family
, dst_port
,
2049 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
2054 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2056 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2057 skb_scrub_packet(skb
, xnet
);
2058 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2059 vni
, md
, flags
, udp_sum
);
2064 udp_tunnel6_xmit_skb(ndst
, sk
, skb
, dev
,
2065 &saddr
, &dst
->sin6
.sin6_addr
,
2066 0, ttl
, src_port
, dst_port
, !udp_sum
);
2073 dev
->stats
.tx_dropped
++;
2077 /* skb is already freed. */
2082 dev
->stats
.tx_errors
++;
2087 /* Transmit local packets over Vxlan
2089 * Outer IP header inherits ECN and DF from inner header.
2090 * Outer UDP destination is the VXLAN assigned port.
2091 * source port is based on hash of flow
2093 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2095 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2096 const struct ip_tunnel_info
*info
;
2098 bool did_rsc
= false;
2099 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2100 struct vxlan_fdb
*f
;
2102 info
= skb_tunnel_info(skb
);
2104 skb_reset_mac_header(skb
);
2107 if ((vxlan
->flags
& VXLAN_F_PROXY
)) {
2108 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2109 return arp_reduce(dev
, skb
);
2110 #if IS_ENABLED(CONFIG_IPV6)
2111 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2112 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2113 + sizeof(struct nd_msg
)) &&
2114 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2117 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2118 if (msg
->icmph
.icmp6_code
== 0 &&
2119 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2120 return neigh_reduce(dev
, skb
);
2126 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
2127 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2128 vxlan_xmit_one(skb
, dev
, NULL
, false);
2131 return NETDEV_TX_OK
;
2134 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2137 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2138 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2139 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2140 did_rsc
= route_shortcircuit(dev
, skb
);
2142 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
2146 f
= vxlan_find_mac(vxlan
, all_zeros_mac
);
2148 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2149 !is_multicast_ether_addr(eth
->h_dest
))
2150 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2152 dev
->stats
.tx_dropped
++;
2154 return NETDEV_TX_OK
;
2158 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2159 struct sk_buff
*skb1
;
2165 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2167 vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
2171 vxlan_xmit_one(skb
, dev
, fdst
, did_rsc
);
2174 return NETDEV_TX_OK
;
2177 /* Walk the forwarding table and purge stale entries */
2178 static void vxlan_cleanup(unsigned long arg
)
2180 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2181 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2184 if (!netif_running(vxlan
->dev
))
2187 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2188 struct hlist_node
*p
, *n
;
2190 spin_lock_bh(&vxlan
->hash_lock
);
2191 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2193 = container_of(p
, struct vxlan_fdb
, hlist
);
2194 unsigned long timeout
;
2196 if (f
->state
& NUD_PERMANENT
)
2199 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2200 if (time_before_eq(timeout
, jiffies
)) {
2201 netdev_dbg(vxlan
->dev
,
2202 "garbage collect %pM\n",
2204 f
->state
= NUD_STALE
;
2205 vxlan_fdb_destroy(vxlan
, f
);
2206 } else if (time_before(timeout
, next_timer
))
2207 next_timer
= timeout
;
2209 spin_unlock_bh(&vxlan
->hash_lock
);
2212 mod_timer(&vxlan
->age_timer
, next_timer
);
2215 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2217 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2218 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2220 spin_lock(&vn
->sock_lock
);
2221 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2222 spin_unlock(&vn
->sock_lock
);
2225 /* Setup stats when device is created */
2226 static int vxlan_init(struct net_device
*dev
)
2228 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2235 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
)
2237 struct vxlan_fdb
*f
;
2239 spin_lock_bh(&vxlan
->hash_lock
);
2240 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
);
2242 vxlan_fdb_destroy(vxlan
, f
);
2243 spin_unlock_bh(&vxlan
->hash_lock
);
2246 static void vxlan_uninit(struct net_device
*dev
)
2248 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2250 vxlan_fdb_delete_default(vxlan
);
2252 free_percpu(dev
->tstats
);
2255 /* Start ageing timer and join group when device is brought up */
2256 static int vxlan_open(struct net_device
*dev
)
2258 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2261 ret
= vxlan_sock_add(vxlan
);
2265 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2266 ret
= vxlan_igmp_join(vxlan
);
2267 if (ret
== -EADDRINUSE
)
2270 vxlan_sock_release(vxlan
);
2275 if (vxlan
->cfg
.age_interval
)
2276 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2281 /* Purge the forwarding table */
2282 static void vxlan_flush(struct vxlan_dev
*vxlan
)
2286 spin_lock_bh(&vxlan
->hash_lock
);
2287 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2288 struct hlist_node
*p
, *n
;
2289 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2291 = container_of(p
, struct vxlan_fdb
, hlist
);
2292 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2293 if (!is_zero_ether_addr(f
->eth_addr
))
2294 vxlan_fdb_destroy(vxlan
, f
);
2297 spin_unlock_bh(&vxlan
->hash_lock
);
2300 /* Cleanup timer and forwarding table on shutdown */
2301 static int vxlan_stop(struct net_device
*dev
)
2303 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2304 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2307 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2308 !vxlan_group_used(vn
, vxlan
))
2309 ret
= vxlan_igmp_leave(vxlan
);
2311 del_timer_sync(&vxlan
->age_timer
);
2314 vxlan_sock_release(vxlan
);
2319 /* Stub, nothing needs to be done. */
2320 static void vxlan_set_multicast_list(struct net_device
*dev
)
2324 static int __vxlan_change_mtu(struct net_device
*dev
,
2325 struct net_device
*lowerdev
,
2326 struct vxlan_rdst
*dst
, int new_mtu
, bool strict
)
2328 int max_mtu
= IP_MAX_MTU
;
2331 max_mtu
= lowerdev
->mtu
;
2333 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2334 max_mtu
-= VXLAN6_HEADROOM
;
2336 max_mtu
-= VXLAN_HEADROOM
;
2341 if (new_mtu
> max_mtu
) {
2352 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2354 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2355 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2356 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2357 dst
->remote_ifindex
);
2358 return __vxlan_change_mtu(dev
, lowerdev
, dst
, new_mtu
, true);
2361 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2363 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2364 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2365 __be16 sport
, dport
;
2367 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2368 vxlan
->cfg
.port_max
, true);
2369 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2371 if (ip_tunnel_info_af(info
) == AF_INET
) {
2374 if (!vxlan
->vn4_sock
)
2376 rt
= vxlan_get_route(vxlan
, skb
, 0, info
->key
.tos
,
2377 info
->key
.u
.ipv4
.dst
,
2378 &info
->key
.u
.ipv4
.src
, NULL
, info
);
2383 #if IS_ENABLED(CONFIG_IPV6)
2384 struct dst_entry
*ndst
;
2386 if (!vxlan
->vn6_sock
)
2388 ndst
= vxlan6_get_route(vxlan
, skb
, 0,
2389 &info
->key
.u
.ipv6
.dst
,
2390 &info
->key
.u
.ipv6
.src
, NULL
);
2392 return PTR_ERR(ndst
);
2394 #else /* !CONFIG_IPV6 */
2395 return -EPFNOSUPPORT
;
2398 info
->key
.tp_src
= sport
;
2399 info
->key
.tp_dst
= dport
;
2403 static const struct net_device_ops vxlan_netdev_ops
= {
2404 .ndo_init
= vxlan_init
,
2405 .ndo_uninit
= vxlan_uninit
,
2406 .ndo_open
= vxlan_open
,
2407 .ndo_stop
= vxlan_stop
,
2408 .ndo_start_xmit
= vxlan_xmit
,
2409 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2410 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2411 .ndo_change_mtu
= vxlan_change_mtu
,
2412 .ndo_validate_addr
= eth_validate_addr
,
2413 .ndo_set_mac_address
= eth_mac_addr
,
2414 .ndo_fdb_add
= vxlan_fdb_add
,
2415 .ndo_fdb_del
= vxlan_fdb_delete
,
2416 .ndo_fdb_dump
= vxlan_fdb_dump
,
2417 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2420 /* Info for udev, that this is a virtual tunnel endpoint */
2421 static struct device_type vxlan_type
= {
2425 /* Calls the ndo_add_vxlan_port of the caller in order to
2426 * supply the listening VXLAN udp ports. Callers are expected
2427 * to implement the ndo_add_vxlan_port.
2429 void vxlan_get_rx_port(struct net_device
*dev
)
2431 struct vxlan_sock
*vs
;
2432 struct net
*net
= dev_net(dev
);
2433 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2434 sa_family_t sa_family
;
2438 spin_lock(&vn
->sock_lock
);
2439 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2440 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2441 port
= inet_sk(vs
->sock
->sk
)->inet_sport
;
2442 sa_family
= vxlan_get_sk_family(vs
);
2443 dev
->netdev_ops
->ndo_add_vxlan_port(dev
, sa_family
,
2447 spin_unlock(&vn
->sock_lock
);
2449 EXPORT_SYMBOL_GPL(vxlan_get_rx_port
);
2451 /* Initialize the device structure. */
2452 static void vxlan_setup(struct net_device
*dev
)
2454 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2457 eth_hw_addr_random(dev
);
2460 dev
->netdev_ops
= &vxlan_netdev_ops
;
2461 dev
->destructor
= free_netdev
;
2462 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2464 dev
->features
|= NETIF_F_LLTX
;
2465 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2466 dev
->features
|= NETIF_F_RXCSUM
;
2467 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2469 dev
->vlan_features
= dev
->features
;
2470 dev
->features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2471 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2472 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2473 dev
->hw_features
|= NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_STAG_TX
;
2474 netif_keep_dst(dev
);
2475 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2476 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
| IFF_NO_QUEUE
;
2478 INIT_LIST_HEAD(&vxlan
->next
);
2479 spin_lock_init(&vxlan
->hash_lock
);
2481 init_timer_deferrable(&vxlan
->age_timer
);
2482 vxlan
->age_timer
.function
= vxlan_cleanup
;
2483 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2485 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2489 gro_cells_init(&vxlan
->gro_cells
, dev
);
2491 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2492 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2495 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2496 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2497 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2498 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2499 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2500 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2501 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2502 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2503 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2504 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2505 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2506 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2507 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2508 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2509 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2510 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2511 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2512 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2513 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2514 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2515 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2516 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2517 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2518 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2519 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2520 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2523 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2525 if (tb
[IFLA_ADDRESS
]) {
2526 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2527 pr_debug("invalid link address (not ethernet)\n");
2531 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2532 pr_debug("invalid all zero ethernet address\n");
2533 return -EADDRNOTAVAIL
;
2540 if (data
[IFLA_VXLAN_ID
]) {
2541 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2542 if (id
>= VXLAN_VID_MASK
)
2546 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2547 const struct ifla_vxlan_port_range
*p
2548 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2550 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2551 pr_debug("port range %u .. %u not valid\n",
2552 ntohs(p
->low
), ntohs(p
->high
));
2560 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2561 struct ethtool_drvinfo
*drvinfo
)
2563 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2564 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2567 static const struct ethtool_ops vxlan_ethtool_ops
= {
2568 .get_drvinfo
= vxlan_get_drvinfo
,
2569 .get_link
= ethtool_op_get_link
,
2572 static void vxlan_del_work(struct work_struct
*work
)
2574 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
2575 udp_tunnel_sock_release(vs
->sock
);
2579 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2580 __be16 port
, u32 flags
)
2582 struct socket
*sock
;
2583 struct udp_port_cfg udp_conf
;
2586 memset(&udp_conf
, 0, sizeof(udp_conf
));
2589 udp_conf
.family
= AF_INET6
;
2590 udp_conf
.use_udp6_rx_checksums
=
2591 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2592 udp_conf
.ipv6_v6only
= 1;
2594 udp_conf
.family
= AF_INET
;
2597 udp_conf
.local_udp_port
= port
;
2599 /* Open UDP socket */
2600 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2602 return ERR_PTR(err
);
2607 /* Create new listen socket if needed */
2608 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2609 __be16 port
, u32 flags
)
2611 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2612 struct vxlan_sock
*vs
;
2613 struct socket
*sock
;
2615 struct udp_tunnel_sock_cfg tunnel_cfg
;
2617 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2619 return ERR_PTR(-ENOMEM
);
2621 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2622 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2624 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
2626 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2628 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port
),
2631 return ERR_CAST(sock
);
2635 atomic_set(&vs
->refcnt
, 1);
2636 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2638 /* Initialize the vxlan udp offloads structure */
2639 vs
->udp_offloads
.port
= port
;
2640 vs
->udp_offloads
.callbacks
.gro_receive
= vxlan_gro_receive
;
2641 vs
->udp_offloads
.callbacks
.gro_complete
= vxlan_gro_complete
;
2643 spin_lock(&vn
->sock_lock
);
2644 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2645 vxlan_notify_add_rx_port(vs
);
2646 spin_unlock(&vn
->sock_lock
);
2648 /* Mark socket as an encapsulation socket. */
2649 tunnel_cfg
.sk_user_data
= vs
;
2650 tunnel_cfg
.encap_type
= 1;
2651 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
2652 tunnel_cfg
.encap_destroy
= NULL
;
2654 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2659 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2661 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2662 struct vxlan_sock
*vs
= NULL
;
2664 if (!vxlan
->cfg
.no_share
) {
2665 spin_lock(&vn
->sock_lock
);
2666 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2667 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2668 if (vs
&& !atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2669 spin_unlock(&vn
->sock_lock
);
2672 spin_unlock(&vn
->sock_lock
);
2675 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2676 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2679 #if IS_ENABLED(CONFIG_IPV6)
2681 vxlan
->vn6_sock
= vs
;
2684 vxlan
->vn4_sock
= vs
;
2685 vxlan_vs_add_dev(vs
, vxlan
);
2689 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2691 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2692 bool metadata
= vxlan
->flags
& VXLAN_F_COLLECT_METADATA
;
2695 vxlan
->vn4_sock
= NULL
;
2696 #if IS_ENABLED(CONFIG_IPV6)
2697 vxlan
->vn6_sock
= NULL
;
2698 if (ipv6
|| metadata
)
2699 ret
= __vxlan_sock_add(vxlan
, true);
2701 if (!ret
&& (!ipv6
|| metadata
))
2702 ret
= __vxlan_sock_add(vxlan
, false);
2704 vxlan_sock_release(vxlan
);
2708 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2709 struct vxlan_config
*conf
)
2711 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2712 struct vxlan_dev
*vxlan
= netdev_priv(dev
), *tmp
;
2713 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2714 unsigned short needed_headroom
= ETH_HLEN
;
2716 bool use_ipv6
= false;
2717 __be16 default_port
= vxlan
->cfg
.dst_port
;
2718 struct net_device
*lowerdev
= NULL
;
2720 vxlan
->net
= src_net
;
2722 dst
->remote_vni
= conf
->vni
;
2724 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2726 /* Unless IPv6 is explicitly requested, assume IPv4 */
2727 if (!dst
->remote_ip
.sa
.sa_family
)
2728 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2730 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2731 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) {
2732 if (!IS_ENABLED(CONFIG_IPV6
))
2733 return -EPFNOSUPPORT
;
2735 vxlan
->flags
|= VXLAN_F_IPV6
;
2738 if (conf
->remote_ifindex
) {
2739 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2740 dst
->remote_ifindex
= conf
->remote_ifindex
;
2743 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
2747 #if IS_ENABLED(CONFIG_IPV6)
2749 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2750 if (idev
&& idev
->cnf
.disable_ipv6
) {
2751 pr_info("IPv6 is disabled via sysctl\n");
2758 dev
->mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2760 needed_headroom
= lowerdev
->hard_header_len
;
2764 err
= __vxlan_change_mtu(dev
, lowerdev
, dst
, conf
->mtu
, false);
2769 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
2770 needed_headroom
+= VXLAN6_HEADROOM
;
2772 needed_headroom
+= VXLAN_HEADROOM
;
2773 dev
->needed_headroom
= needed_headroom
;
2775 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2776 if (!vxlan
->cfg
.dst_port
)
2777 vxlan
->cfg
.dst_port
= default_port
;
2778 vxlan
->flags
|= conf
->flags
;
2780 if (!vxlan
->cfg
.age_interval
)
2781 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2783 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
2784 if (tmp
->cfg
.vni
== conf
->vni
&&
2785 (tmp
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
||
2786 tmp
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) == use_ipv6
&&
2787 tmp
->cfg
.dst_port
== vxlan
->cfg
.dst_port
&&
2788 (tmp
->flags
& VXLAN_F_RCV_FLAGS
) ==
2789 (vxlan
->flags
& VXLAN_F_RCV_FLAGS
))
2793 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
2795 /* create an fdb entry for a valid default destination */
2796 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
2797 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
2798 &vxlan
->default_dst
.remote_ip
,
2799 NUD_REACHABLE
|NUD_PERMANENT
,
2800 NLM_F_EXCL
|NLM_F_CREATE
,
2801 vxlan
->cfg
.dst_port
,
2802 vxlan
->default_dst
.remote_vni
,
2803 vxlan
->default_dst
.remote_ifindex
,
2809 err
= register_netdevice(dev
);
2811 vxlan_fdb_delete_default(vxlan
);
2815 list_add(&vxlan
->next
, &vn
->vxlan_list
);
2820 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
2821 u8 name_assign_type
, struct vxlan_config
*conf
)
2823 struct nlattr
*tb
[IFLA_MAX
+1];
2824 struct net_device
*dev
;
2827 memset(&tb
, 0, sizeof(tb
));
2829 dev
= rtnl_create_link(net
, name
, name_assign_type
,
2830 &vxlan_link_ops
, tb
);
2834 err
= vxlan_dev_configure(net
, dev
, conf
);
2837 return ERR_PTR(err
);
2842 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
2844 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
2845 struct nlattr
*tb
[], struct nlattr
*data
[])
2847 struct vxlan_config conf
;
2850 memset(&conf
, 0, sizeof(conf
));
2852 if (data
[IFLA_VXLAN_ID
])
2853 conf
.vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
2855 if (data
[IFLA_VXLAN_GROUP
]) {
2856 conf
.remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
2857 } else if (data
[IFLA_VXLAN_GROUP6
]) {
2858 if (!IS_ENABLED(CONFIG_IPV6
))
2859 return -EPFNOSUPPORT
;
2861 conf
.remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
2862 conf
.remote_ip
.sa
.sa_family
= AF_INET6
;
2865 if (data
[IFLA_VXLAN_LOCAL
]) {
2866 conf
.saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
2867 conf
.saddr
.sa
.sa_family
= AF_INET
;
2868 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
2869 if (!IS_ENABLED(CONFIG_IPV6
))
2870 return -EPFNOSUPPORT
;
2872 /* TODO: respect scope id */
2873 conf
.saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
2874 conf
.saddr
.sa
.sa_family
= AF_INET6
;
2877 if (data
[IFLA_VXLAN_LINK
])
2878 conf
.remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
2880 if (data
[IFLA_VXLAN_TOS
])
2881 conf
.tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
2883 if (data
[IFLA_VXLAN_TTL
])
2884 conf
.ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
2886 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
2887 conf
.flags
|= VXLAN_F_LEARN
;
2889 if (data
[IFLA_VXLAN_AGEING
])
2890 conf
.age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
2892 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
2893 conf
.flags
|= VXLAN_F_PROXY
;
2895 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
2896 conf
.flags
|= VXLAN_F_RSC
;
2898 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
2899 conf
.flags
|= VXLAN_F_L2MISS
;
2901 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
2902 conf
.flags
|= VXLAN_F_L3MISS
;
2904 if (data
[IFLA_VXLAN_LIMIT
])
2905 conf
.addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
2907 if (data
[IFLA_VXLAN_COLLECT_METADATA
] &&
2908 nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
2909 conf
.flags
|= VXLAN_F_COLLECT_METADATA
;
2911 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2912 const struct ifla_vxlan_port_range
*p
2913 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2914 conf
.port_min
= ntohs(p
->low
);
2915 conf
.port_max
= ntohs(p
->high
);
2918 if (data
[IFLA_VXLAN_PORT
])
2919 conf
.dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
2921 if (data
[IFLA_VXLAN_UDP_CSUM
] &&
2922 !nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
2923 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
2925 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] &&
2926 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
2927 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
2929 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] &&
2930 nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
2931 conf
.flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
2933 if (data
[IFLA_VXLAN_REMCSUM_TX
] &&
2934 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
2935 conf
.flags
|= VXLAN_F_REMCSUM_TX
;
2937 if (data
[IFLA_VXLAN_REMCSUM_RX
] &&
2938 nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
2939 conf
.flags
|= VXLAN_F_REMCSUM_RX
;
2941 if (data
[IFLA_VXLAN_GBP
])
2942 conf
.flags
|= VXLAN_F_GBP
;
2944 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
])
2945 conf
.flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
2947 err
= vxlan_dev_configure(src_net
, dev
, &conf
);
2950 pr_info("ifindex %d does not exist\n", conf
.remote_ifindex
);
2954 pr_info("IPv6 is disabled via sysctl\n");
2958 pr_info("duplicate VNI %u\n", be32_to_cpu(conf
.vni
));
2965 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
2967 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2968 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2970 spin_lock(&vn
->sock_lock
);
2971 if (!hlist_unhashed(&vxlan
->hlist
))
2972 hlist_del_rcu(&vxlan
->hlist
);
2973 spin_unlock(&vn
->sock_lock
);
2975 gro_cells_destroy(&vxlan
->gro_cells
);
2976 list_del(&vxlan
->next
);
2977 unregister_netdevice_queue(dev
, head
);
2980 static size_t vxlan_get_size(const struct net_device
*dev
)
2983 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
2984 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
2985 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
2986 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
2987 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
2988 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
2989 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
2990 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
2991 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
2992 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
2993 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
2994 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
2995 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
2996 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
2997 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
2998 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
2999 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3000 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3001 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3002 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3003 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3007 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3009 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3010 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3011 struct ifla_vxlan_port_range ports
= {
3012 .low
= htons(vxlan
->cfg
.port_min
),
3013 .high
= htons(vxlan
->cfg
.port_max
),
3016 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3017 goto nla_put_failure
;
3019 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3020 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3021 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3022 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3023 goto nla_put_failure
;
3024 #if IS_ENABLED(CONFIG_IPV6)
3026 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3027 &dst
->remote_ip
.sin6
.sin6_addr
))
3028 goto nla_put_failure
;
3033 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3034 goto nla_put_failure
;
3036 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3037 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3038 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3039 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3040 goto nla_put_failure
;
3041 #if IS_ENABLED(CONFIG_IPV6)
3043 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3044 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3045 goto nla_put_failure
;
3050 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3051 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3052 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3053 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
3054 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3055 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
3056 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
3057 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3058 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
3059 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3060 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
3061 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3062 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
3063 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3064 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3065 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3066 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3067 !(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3068 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3069 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3070 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3071 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3072 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3073 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
3074 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3075 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
3076 goto nla_put_failure
;
3078 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3079 goto nla_put_failure
;
3081 if (vxlan
->flags
& VXLAN_F_GBP
&&
3082 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3083 goto nla_put_failure
;
3085 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3086 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3087 goto nla_put_failure
;
3095 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3097 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3102 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3104 .maxtype
= IFLA_VXLAN_MAX
,
3105 .policy
= vxlan_policy
,
3106 .priv_size
= sizeof(struct vxlan_dev
),
3107 .setup
= vxlan_setup
,
3108 .validate
= vxlan_validate
,
3109 .newlink
= vxlan_newlink
,
3110 .dellink
= vxlan_dellink
,
3111 .get_size
= vxlan_get_size
,
3112 .fill_info
= vxlan_fill_info
,
3113 .get_link_net
= vxlan_get_link_net
,
3116 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3117 struct net_device
*dev
)
3119 struct vxlan_dev
*vxlan
, *next
;
3120 LIST_HEAD(list_kill
);
3122 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3123 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3125 /* In case we created vxlan device with carrier
3126 * and we loose the carrier due to module unload
3127 * we also need to remove vxlan device. In other
3128 * cases, it's not necessary and remote_ifindex
3129 * is 0 here, so no matches.
3131 if (dst
->remote_ifindex
== dev
->ifindex
)
3132 vxlan_dellink(vxlan
->dev
, &list_kill
);
3135 unregister_netdevice_many(&list_kill
);
3138 static int vxlan_lowerdev_event(struct notifier_block
*unused
,
3139 unsigned long event
, void *ptr
)
3141 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3142 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3144 if (event
== NETDEV_UNREGISTER
)
3145 vxlan_handle_lowerdev_unregister(vn
, dev
);
3150 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3151 .notifier_call
= vxlan_lowerdev_event
,
3154 static __net_init
int vxlan_init_net(struct net
*net
)
3156 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3159 INIT_LIST_HEAD(&vn
->vxlan_list
);
3160 spin_lock_init(&vn
->sock_lock
);
3162 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3163 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3168 static void __net_exit
vxlan_exit_net(struct net
*net
)
3170 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3171 struct vxlan_dev
*vxlan
, *next
;
3172 struct net_device
*dev
, *aux
;
3176 for_each_netdev_safe(net
, dev
, aux
)
3177 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3178 unregister_netdevice_queue(dev
, &list
);
3180 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3181 /* If vxlan->dev is in the same netns, it has already been added
3182 * to the list by the previous loop.
3184 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3185 gro_cells_destroy(&vxlan
->gro_cells
);
3186 unregister_netdevice_queue(vxlan
->dev
, &list
);
3190 unregister_netdevice_many(&list
);
3194 static struct pernet_operations vxlan_net_ops
= {
3195 .init
= vxlan_init_net
,
3196 .exit
= vxlan_exit_net
,
3197 .id
= &vxlan_net_id
,
3198 .size
= sizeof(struct vxlan_net
),
3201 static int __init
vxlan_init_module(void)
3205 vxlan_wq
= alloc_workqueue("vxlan", 0, 0);
3209 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3211 rc
= register_pernet_subsys(&vxlan_net_ops
);
3215 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3219 rc
= rtnl_link_register(&vxlan_link_ops
);
3225 unregister_netdevice_notifier(&vxlan_notifier_block
);
3227 unregister_pernet_subsys(&vxlan_net_ops
);
3229 destroy_workqueue(vxlan_wq
);
3232 late_initcall(vxlan_init_module
);
3234 static void __exit
vxlan_cleanup_module(void)
3236 rtnl_link_unregister(&vxlan_link_ops
);
3237 unregister_netdevice_notifier(&vxlan_notifier_block
);
3238 destroy_workqueue(vxlan_wq
);
3239 unregister_pernet_subsys(&vxlan_net_ops
);
3240 /* rcu_barrier() is called by netns */
3242 module_exit(vxlan_cleanup_module
);
3244 MODULE_LICENSE("GPL");
3245 MODULE_VERSION(VXLAN_VERSION
);
3246 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3247 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3248 MODULE_ALIAS_RTNL_LINK("vxlan");