2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/if_ether.h>
20 #include <linux/ethtool.h>
22 #include <net/ndisc.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/tun_proto.h>
30 #include <net/vxlan.h>
32 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_tunnel.h>
34 #include <net/ip6_checksum.h>
37 #define VXLAN_VERSION "0.1"
39 #define PORT_HASH_BITS 8
40 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
41 #define FDB_AGE_DEFAULT 300 /* 5 min */
42 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
44 /* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
46 * for compatibility with early adopters.
48 static unsigned short vxlan_port __read_mostly
= 8472;
49 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
50 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
52 static bool log_ecn_error
= true;
53 module_param(log_ecn_error
, bool, 0644);
54 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
56 static unsigned int vxlan_net_id
;
57 static struct rtnl_link_ops vxlan_link_ops
;
59 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
61 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
63 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
);
65 /* per-network namespace private data for this module */
67 struct list_head vxlan_list
;
68 struct hlist_head sock_list
[PORT_HASH_SIZE
];
72 /* Forwarding table entry */
74 struct hlist_node hlist
; /* linked list of entries */
76 unsigned long updated
; /* jiffies */
78 struct list_head remotes
;
79 u8 eth_addr
[ETH_ALEN
];
80 u16 state
; /* see ndm_state */
82 u8 flags
; /* see ndm_flags */
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly
;
88 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
90 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
91 ip_tunnel_collect_metadata();
94 #if IS_ENABLED(CONFIG_IPV6)
96 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
98 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
100 if (a
->sa
.sa_family
== AF_INET6
)
101 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
103 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
106 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
108 if (ipa
->sa
.sa_family
== AF_INET6
)
109 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
111 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
114 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
116 if (ipa
->sa
.sa_family
== AF_INET6
)
117 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
119 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
122 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
124 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
125 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
126 ip
->sa
.sa_family
= AF_INET6
;
128 } else if (nla_len(nla
) >= sizeof(__be32
)) {
129 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
130 ip
->sa
.sa_family
= AF_INET
;
133 return -EAFNOSUPPORT
;
137 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
138 const union vxlan_addr
*ip
)
140 if (ip
->sa
.sa_family
== AF_INET6
)
141 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
143 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
146 #else /* !CONFIG_IPV6 */
149 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
151 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
154 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
156 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
159 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
161 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
164 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
166 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
167 return -EAFNOSUPPORT
;
168 } else if (nla_len(nla
) >= sizeof(__be32
)) {
169 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
170 ip
->sa
.sa_family
= AF_INET
;
173 return -EAFNOSUPPORT
;
177 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
178 const union vxlan_addr
*ip
)
180 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
184 /* Virtual Network hash table head */
185 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
187 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
190 /* Socket hash table head */
191 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
193 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
195 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
198 /* First remote destination for a forwarding entry.
199 * Guaranteed to be non-NULL because remotes are never deleted.
201 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
203 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
206 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
208 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
211 /* Find VXLAN socket based on network namespace, address family and UDP port
212 * and enabled unshareable flags.
214 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
215 __be16 port
, u32 flags
)
217 struct vxlan_sock
*vs
;
219 flags
&= VXLAN_F_RCV_FLAGS
;
221 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
222 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
223 vxlan_get_sk_family(vs
) == family
&&
230 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, int ifindex
,
233 struct vxlan_dev_node
*node
;
235 /* For flow based devices, map all packets to VNI 0 */
236 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
239 hlist_for_each_entry_rcu(node
, vni_head(vs
, vni
), hlist
) {
240 if (node
->vxlan
->default_dst
.remote_vni
!= vni
)
243 if (IS_ENABLED(CONFIG_IPV6
)) {
244 const struct vxlan_config
*cfg
= &node
->vxlan
->cfg
;
246 if ((cfg
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
247 cfg
->remote_ifindex
!= ifindex
)
257 /* Look up VNI in a per net namespace table */
258 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, int ifindex
,
259 __be32 vni
, sa_family_t family
,
260 __be16 port
, u32 flags
)
262 struct vxlan_sock
*vs
;
264 vs
= vxlan_find_sock(net
, family
, port
, flags
);
268 return vxlan_vs_find_vni(vs
, ifindex
, vni
);
271 /* Fill in neighbour message in skbuff. */
272 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
273 const struct vxlan_fdb
*fdb
,
274 u32 portid
, u32 seq
, int type
, unsigned int flags
,
275 const struct vxlan_rdst
*rdst
)
277 unsigned long now
= jiffies
;
278 struct nda_cacheinfo ci
;
279 struct nlmsghdr
*nlh
;
281 bool send_ip
, send_eth
;
283 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
287 ndm
= nlmsg_data(nlh
);
288 memset(ndm
, 0, sizeof(*ndm
));
290 send_eth
= send_ip
= true;
292 if (type
== RTM_GETNEIGH
) {
293 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
294 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
295 ndm
->ndm_family
= send_ip
? rdst
->remote_ip
.sa
.sa_family
: AF_INET
;
297 ndm
->ndm_family
= AF_BRIDGE
;
298 ndm
->ndm_state
= fdb
->state
;
299 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
300 ndm
->ndm_flags
= fdb
->flags
;
301 ndm
->ndm_type
= RTN_UNICAST
;
303 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
304 nla_put_s32(skb
, NDA_LINK_NETNSID
,
305 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
306 goto nla_put_failure
;
308 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
309 goto nla_put_failure
;
311 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
312 goto nla_put_failure
;
314 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
315 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
316 goto nla_put_failure
;
317 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
318 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
319 goto nla_put_failure
;
320 if ((vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) && fdb
->vni
&&
321 nla_put_u32(skb
, NDA_SRC_VNI
,
322 be32_to_cpu(fdb
->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 static u32
eth_vni_hash(const unsigned char *addr
, __be32 vni
)
424 /* use 1 byte of OUI and 3 bytes of NIC */
425 u32 key
= get_unaligned((u32
*)(addr
+ 2));
427 return jhash_2words(key
, vni
, vxlan_salt
) & (FDB_HASH_SIZE
- 1);
430 /* Hash chain to use given mac address */
431 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
432 const u8
*mac
, __be32 vni
)
434 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)
435 return &vxlan
->fdb_head
[eth_vni_hash(mac
, vni
)];
437 return &vxlan
->fdb_head
[eth_hash(mac
)];
440 /* Look up Ethernet address in forwarding table */
441 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
442 const u8
*mac
, __be32 vni
)
444 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
, vni
);
447 hlist_for_each_entry_rcu(f
, head
, hlist
) {
448 if (ether_addr_equal(mac
, f
->eth_addr
)) {
449 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
461 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
462 const u8
*mac
, __be32 vni
)
466 f
= __vxlan_find_mac(vxlan
, mac
, vni
);
473 /* caller should hold vxlan->hash_lock */
474 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
475 union vxlan_addr
*ip
, __be16 port
,
476 __be32 vni
, __u32 ifindex
)
478 struct vxlan_rdst
*rd
;
480 list_for_each_entry(rd
, &f
->remotes
, list
) {
481 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
482 rd
->remote_port
== port
&&
483 rd
->remote_vni
== vni
&&
484 rd
->remote_ifindex
== ifindex
)
491 /* Replace destination of unicast mac */
492 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
493 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
496 struct vxlan_rdst
*rd
;
498 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
502 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
506 dst_cache_reset(&rd
->dst_cache
);
508 rd
->remote_port
= port
;
509 rd
->remote_vni
= vni
;
510 rd
->remote_ifindex
= ifindex
;
514 /* Add/update destinations for multicast */
515 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
516 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
517 __u32 ifindex
, struct vxlan_rdst
**rdp
)
519 struct vxlan_rdst
*rd
;
521 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
525 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
529 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
535 rd
->remote_port
= port
;
536 rd
->remote_vni
= vni
;
537 rd
->remote_ifindex
= ifindex
;
539 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
545 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
547 struct vxlanhdr
*vh
, size_t hdrlen
,
549 struct gro_remcsum
*grc
,
552 size_t start
, offset
;
554 if (skb
->remcsum_offload
)
557 if (!NAPI_GRO_CB(skb
)->csum_valid
)
560 start
= vxlan_rco_start(vni_field
);
561 offset
= start
+ vxlan_rco_offset(vni_field
);
563 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
564 start
, offset
, grc
, nopartial
);
566 skb
->remcsum_offload
= 1;
571 static struct sk_buff
**vxlan_gro_receive(struct sock
*sk
,
572 struct sk_buff
**head
,
575 struct sk_buff
*p
, **pp
= NULL
;
576 struct vxlanhdr
*vh
, *vh2
;
577 unsigned int hlen
, off_vx
;
579 struct vxlan_sock
*vs
= rcu_dereference_sk_user_data(sk
);
581 struct gro_remcsum grc
;
583 skb_gro_remcsum_init(&grc
);
585 off_vx
= skb_gro_offset(skb
);
586 hlen
= off_vx
+ sizeof(*vh
);
587 vh
= skb_gro_header_fast(skb
, off_vx
);
588 if (skb_gro_header_hard(skb
, hlen
)) {
589 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
594 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
596 flags
= vh
->vx_flags
;
598 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
599 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
602 VXLAN_F_REMCSUM_NOPARTIAL
));
608 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
610 for (p
= *head
; p
; p
= p
->next
) {
611 if (!NAPI_GRO_CB(p
)->same_flow
)
614 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
615 if (vh
->vx_flags
!= vh2
->vx_flags
||
616 vh
->vx_vni
!= vh2
->vx_vni
) {
617 NAPI_GRO_CB(p
)->same_flow
= 0;
622 pp
= call_gro_receive(eth_gro_receive
, head
, skb
);
626 skb_gro_remcsum_cleanup(skb
, &grc
);
627 skb
->remcsum_offload
= 0;
628 NAPI_GRO_CB(skb
)->flush
|= flush
;
633 static int vxlan_gro_complete(struct sock
*sk
, struct sk_buff
*skb
, int nhoff
)
635 /* Sets 'skb->inner_mac_header' since we are always called with
636 * 'skb->encapsulation' set.
638 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
641 /* Add new entry to forwarding table -- assumes lock held */
642 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
643 const u8
*mac
, union vxlan_addr
*ip
,
644 __u16 state
, __u16 flags
,
645 __be16 port
, __be32 src_vni
, __be32 vni
,
646 __u32 ifindex
, __u8 ndm_flags
)
648 struct vxlan_rdst
*rd
= NULL
;
653 f
= __vxlan_find_mac(vxlan
, mac
, src_vni
);
655 if (flags
& NLM_F_EXCL
) {
656 netdev_dbg(vxlan
->dev
,
657 "lost race to create %pM\n", mac
);
660 if (f
->state
!= state
) {
662 f
->updated
= jiffies
;
665 if (f
->flags
!= ndm_flags
) {
666 f
->flags
= ndm_flags
;
667 f
->updated
= jiffies
;
670 if ((flags
& NLM_F_REPLACE
)) {
671 /* Only change unicasts */
672 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
673 is_zero_ether_addr(f
->eth_addr
))) {
674 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
679 if ((flags
& NLM_F_APPEND
) &&
680 (is_multicast_ether_addr(f
->eth_addr
) ||
681 is_zero_ether_addr(f
->eth_addr
))) {
682 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
689 if (!(flags
& NLM_F_CREATE
))
692 if (vxlan
->cfg
.addrmax
&&
693 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
696 /* Disallow replace to add a multicast entry */
697 if ((flags
& NLM_F_REPLACE
) &&
698 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
701 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
702 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
708 f
->flags
= ndm_flags
;
709 f
->updated
= f
->used
= jiffies
;
711 INIT_LIST_HEAD(&f
->remotes
);
712 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
714 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
721 hlist_add_head_rcu(&f
->hlist
,
722 vxlan_fdb_head(vxlan
, mac
, src_vni
));
727 rd
= first_remote_rtnl(f
);
728 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
734 static void vxlan_fdb_free(struct rcu_head
*head
)
736 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
737 struct vxlan_rdst
*rd
, *nd
;
739 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
740 dst_cache_destroy(&rd
->dst_cache
);
746 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
748 netdev_dbg(vxlan
->dev
,
749 "delete %pM\n", f
->eth_addr
);
752 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
754 hlist_del_rcu(&f
->hlist
);
755 call_rcu(&f
->rcu
, vxlan_fdb_free
);
758 static void vxlan_dst_free(struct rcu_head
*head
)
760 struct vxlan_rdst
*rd
= container_of(head
, struct vxlan_rdst
, rcu
);
762 dst_cache_destroy(&rd
->dst_cache
);
766 static void vxlan_fdb_dst_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
767 struct vxlan_rdst
*rd
)
769 list_del_rcu(&rd
->list
);
770 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
771 call_rcu(&rd
->rcu
, vxlan_dst_free
);
774 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
775 union vxlan_addr
*ip
, __be16
*port
, __be32
*src_vni
,
776 __be32
*vni
, u32
*ifindex
)
778 struct net
*net
= dev_net(vxlan
->dev
);
782 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
786 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
787 if (remote
->sa
.sa_family
== AF_INET
) {
788 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
789 ip
->sa
.sa_family
= AF_INET
;
790 #if IS_ENABLED(CONFIG_IPV6)
792 ip
->sin6
.sin6_addr
= in6addr_any
;
793 ip
->sa
.sa_family
= AF_INET6
;
799 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
801 *port
= nla_get_be16(tb
[NDA_PORT
]);
803 *port
= vxlan
->cfg
.dst_port
;
807 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
809 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
811 *vni
= vxlan
->default_dst
.remote_vni
;
814 if (tb
[NDA_SRC_VNI
]) {
815 if (nla_len(tb
[NDA_SRC_VNI
]) != sizeof(u32
))
817 *src_vni
= cpu_to_be32(nla_get_u32(tb
[NDA_SRC_VNI
]));
819 *src_vni
= vxlan
->default_dst
.remote_vni
;
822 if (tb
[NDA_IFINDEX
]) {
823 struct net_device
*tdev
;
825 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
827 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
828 tdev
= __dev_get_by_index(net
, *ifindex
);
830 return -EADDRNOTAVAIL
;
838 /* Add static entry (via netlink) */
839 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
840 struct net_device
*dev
,
841 const unsigned char *addr
, u16 vid
, u16 flags
)
843 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
844 /* struct net *net = dev_net(vxlan->dev); */
851 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
852 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
857 if (tb
[NDA_DST
] == NULL
)
860 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
864 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
865 return -EAFNOSUPPORT
;
867 spin_lock_bh(&vxlan
->hash_lock
);
868 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
869 port
, src_vni
, vni
, ifindex
, ndm
->ndm_flags
);
870 spin_unlock_bh(&vxlan
->hash_lock
);
875 static int __vxlan_fdb_delete(struct vxlan_dev
*vxlan
,
876 const unsigned char *addr
, union vxlan_addr ip
,
877 __be16 port
, __be32 src_vni
, __be32 vni
,
878 u32 ifindex
, u16 vid
)
881 struct vxlan_rdst
*rd
= NULL
;
884 f
= vxlan_find_mac(vxlan
, addr
, src_vni
);
888 if (!vxlan_addr_any(&ip
)) {
889 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
894 /* remove a destination if it's not the only one on the list,
895 * otherwise destroy the fdb entry
897 if (rd
&& !list_is_singular(&f
->remotes
)) {
898 vxlan_fdb_dst_destroy(vxlan
, f
, rd
);
902 vxlan_fdb_destroy(vxlan
, f
);
908 /* Delete entry (via netlink) */
909 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
910 struct net_device
*dev
,
911 const unsigned char *addr
, u16 vid
)
913 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
920 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
924 spin_lock_bh(&vxlan
->hash_lock
);
925 err
= __vxlan_fdb_delete(vxlan
, addr
, ip
, port
, src_vni
, vni
, ifindex
,
927 spin_unlock_bh(&vxlan
->hash_lock
);
932 /* Dump forwarding table */
933 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
934 struct net_device
*dev
,
935 struct net_device
*filter_dev
, int *idx
)
937 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
941 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
944 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
945 struct vxlan_rdst
*rd
;
947 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
948 if (*idx
< cb
->args
[2])
951 err
= vxlan_fdb_info(skb
, vxlan
, f
,
952 NETLINK_CB(cb
->skb
).portid
,
967 /* Watch incoming packets to learn mapping between Ethernet address
968 * and Tunnel endpoint.
969 * Return true if packet is bogus and should be dropped.
971 static bool vxlan_snoop(struct net_device
*dev
,
972 union vxlan_addr
*src_ip
, const u8
*src_mac
,
973 u32 src_ifindex
, __be32 vni
)
975 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
979 #if IS_ENABLED(CONFIG_IPV6)
980 if (src_ip
->sa
.sa_family
== AF_INET6
&&
981 (ipv6_addr_type(&src_ip
->sin6
.sin6_addr
) & IPV6_ADDR_LINKLOCAL
))
982 ifindex
= src_ifindex
;
985 f
= vxlan_find_mac(vxlan
, src_mac
, vni
);
987 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
989 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
) &&
990 rdst
->remote_ifindex
== ifindex
))
993 /* Don't migrate static entries, drop packets */
994 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
999 "%pM migrated from %pIS to %pIS\n",
1000 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
1002 rdst
->remote_ip
= *src_ip
;
1003 f
->updated
= jiffies
;
1004 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1006 /* learned new entry */
1007 spin_lock(&vxlan
->hash_lock
);
1009 /* close off race between vxlan_flush and incoming packets */
1010 if (netif_running(dev
))
1011 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
1013 NLM_F_EXCL
|NLM_F_CREATE
,
1014 vxlan
->cfg
.dst_port
,
1016 vxlan
->default_dst
.remote_vni
,
1018 spin_unlock(&vxlan
->hash_lock
);
1024 /* See if multicast group is already in use by other ID */
1025 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1027 struct vxlan_dev
*vxlan
;
1028 struct vxlan_sock
*sock4
;
1029 #if IS_ENABLED(CONFIG_IPV6)
1030 struct vxlan_sock
*sock6
;
1032 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1034 sock4
= rtnl_dereference(dev
->vn4_sock
);
1036 /* The vxlan_sock is only used by dev, leaving group has
1037 * no effect on other vxlan devices.
1039 if (family
== AF_INET
&& sock4
&& refcount_read(&sock4
->refcnt
) == 1)
1041 #if IS_ENABLED(CONFIG_IPV6)
1042 sock6
= rtnl_dereference(dev
->vn6_sock
);
1043 if (family
== AF_INET6
&& sock6
&& refcount_read(&sock6
->refcnt
) == 1)
1047 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1048 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1051 if (family
== AF_INET
&&
1052 rtnl_dereference(vxlan
->vn4_sock
) != sock4
)
1054 #if IS_ENABLED(CONFIG_IPV6)
1055 if (family
== AF_INET6
&&
1056 rtnl_dereference(vxlan
->vn6_sock
) != sock6
)
1060 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1061 &dev
->default_dst
.remote_ip
))
1064 if (vxlan
->default_dst
.remote_ifindex
!=
1065 dev
->default_dst
.remote_ifindex
)
1074 static bool __vxlan_sock_release_prep(struct vxlan_sock
*vs
)
1076 struct vxlan_net
*vn
;
1080 if (!refcount_dec_and_test(&vs
->refcnt
))
1083 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1084 spin_lock(&vn
->sock_lock
);
1085 hlist_del_rcu(&vs
->hlist
);
1086 udp_tunnel_notify_del_rx_port(vs
->sock
,
1087 (vs
->flags
& VXLAN_F_GPE
) ?
1088 UDP_TUNNEL_TYPE_VXLAN_GPE
:
1089 UDP_TUNNEL_TYPE_VXLAN
);
1090 spin_unlock(&vn
->sock_lock
);
1095 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1097 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1098 #if IS_ENABLED(CONFIG_IPV6)
1099 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1101 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
1104 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
1107 vxlan_vs_del_dev(vxlan
);
1109 if (__vxlan_sock_release_prep(sock4
)) {
1110 udp_tunnel_sock_release(sock4
->sock
);
1114 #if IS_ENABLED(CONFIG_IPV6)
1115 if (__vxlan_sock_release_prep(sock6
)) {
1116 udp_tunnel_sock_release(sock6
->sock
);
1122 /* Update multicast group membership when first VNI on
1123 * multicast address is brought up
1125 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1128 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1129 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1132 if (ip
->sa
.sa_family
== AF_INET
) {
1133 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1134 struct ip_mreqn mreq
= {
1135 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1136 .imr_ifindex
= ifindex
,
1139 sk
= sock4
->sock
->sk
;
1141 ret
= ip_mc_join_group(sk
, &mreq
);
1143 #if IS_ENABLED(CONFIG_IPV6)
1145 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1147 sk
= sock6
->sock
->sk
;
1149 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1150 &ip
->sin6
.sin6_addr
);
1158 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1159 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1162 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1163 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1166 if (ip
->sa
.sa_family
== AF_INET
) {
1167 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1168 struct ip_mreqn mreq
= {
1169 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1170 .imr_ifindex
= ifindex
,
1173 sk
= sock4
->sock
->sk
;
1175 ret
= ip_mc_leave_group(sk
, &mreq
);
1177 #if IS_ENABLED(CONFIG_IPV6)
1179 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1181 sk
= sock6
->sock
->sk
;
1183 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1184 &ip
->sin6
.sin6_addr
);
1192 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1193 struct sk_buff
*skb
, u32 vxflags
)
1195 size_t start
, offset
;
1197 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1200 start
= vxlan_rco_start(unparsed
->vx_vni
);
1201 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1203 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1206 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1207 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1209 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1210 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1214 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1215 struct sk_buff
*skb
, u32 vxflags
,
1216 struct vxlan_metadata
*md
)
1218 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1219 struct metadata_dst
*tun_dst
;
1221 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1224 md
->gbp
= ntohs(gbp
->policy_id
);
1226 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1228 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1229 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1231 if (gbp
->dont_learn
)
1232 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1234 if (gbp
->policy_applied
)
1235 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1237 /* In flow-based mode, GBP is carried in dst_metadata */
1238 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1239 skb
->mark
= md
->gbp
;
1241 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1244 static bool vxlan_parse_gpe_hdr(struct vxlanhdr
*unparsed
,
1246 struct sk_buff
*skb
, u32 vxflags
)
1248 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)unparsed
;
1250 /* Need to have Next Protocol set for interfaces in GPE mode. */
1251 if (!gpe
->np_applied
)
1253 /* "The initial version is 0. If a receiver does not support the
1254 * version indicated it MUST drop the packet.
1256 if (gpe
->version
!= 0)
1258 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1259 * processing MUST occur." However, we don't implement OAM
1260 * processing, thus drop the packet.
1265 *protocol
= tun_p_to_eth_p(gpe
->next_protocol
);
1269 unparsed
->vx_flags
&= ~VXLAN_GPE_USED_BITS
;
1273 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1274 struct vxlan_sock
*vs
,
1275 struct sk_buff
*skb
, __be32 vni
)
1277 union vxlan_addr saddr
;
1278 u32 ifindex
= skb
->dev
->ifindex
;
1280 skb_reset_mac_header(skb
);
1281 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1282 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1284 /* Ignore packet loops (and multicast echo) */
1285 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1288 /* Get address from the outer IP header */
1289 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1290 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1291 saddr
.sa
.sa_family
= AF_INET
;
1292 #if IS_ENABLED(CONFIG_IPV6)
1294 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1295 saddr
.sa
.sa_family
= AF_INET6
;
1299 if ((vxlan
->cfg
.flags
& VXLAN_F_LEARN
) &&
1300 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
, ifindex
, vni
))
1306 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1307 struct sk_buff
*skb
)
1311 if (vxlan_get_sk_family(vs
) == AF_INET
)
1312 err
= IP_ECN_decapsulate(oiph
, skb
);
1313 #if IS_ENABLED(CONFIG_IPV6)
1315 err
= IP6_ECN_decapsulate(oiph
, skb
);
1318 if (unlikely(err
) && log_ecn_error
) {
1319 if (vxlan_get_sk_family(vs
) == AF_INET
)
1320 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1321 &((struct iphdr
*)oiph
)->saddr
,
1322 ((struct iphdr
*)oiph
)->tos
);
1324 net_info_ratelimited("non-ECT from %pI6\n",
1325 &((struct ipv6hdr
*)oiph
)->saddr
);
1330 /* Callback from net/ipv4/udp.c to receive packets */
1331 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1333 struct pcpu_sw_netstats
*stats
;
1334 struct vxlan_dev
*vxlan
;
1335 struct vxlan_sock
*vs
;
1336 struct vxlanhdr unparsed
;
1337 struct vxlan_metadata _md
;
1338 struct vxlan_metadata
*md
= &_md
;
1339 __be16 protocol
= htons(ETH_P_TEB
);
1340 bool raw_proto
= false;
1344 /* Need UDP and VXLAN header to be present */
1345 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1348 unparsed
= *vxlan_hdr(skb
);
1349 /* VNI flag always required to be set */
1350 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1351 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1352 ntohl(vxlan_hdr(skb
)->vx_flags
),
1353 ntohl(vxlan_hdr(skb
)->vx_vni
));
1354 /* Return non vxlan pkt */
1357 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1358 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1360 vs
= rcu_dereference_sk_user_data(sk
);
1364 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1366 vxlan
= vxlan_vs_find_vni(vs
, skb
->dev
->ifindex
, vni
);
1370 /* For backwards compatibility, only allow reserved fields to be
1371 * used by VXLAN extensions if explicitly requested.
1373 if (vs
->flags
& VXLAN_F_GPE
) {
1374 if (!vxlan_parse_gpe_hdr(&unparsed
, &protocol
, skb
, vs
->flags
))
1379 if (__iptunnel_pull_header(skb
, VXLAN_HLEN
, protocol
, raw_proto
,
1380 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1383 if (vxlan_collect_metadata(vs
)) {
1384 struct metadata_dst
*tun_dst
;
1386 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1387 key32_to_tunnel_id(vni
), sizeof(*md
));
1392 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1394 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1396 memset(md
, 0, sizeof(*md
));
1399 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1400 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1402 if (vs
->flags
& VXLAN_F_GBP
)
1403 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1404 /* Note that GBP and GPE can never be active together. This is
1405 * ensured in vxlan_dev_configure.
1408 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1409 /* If there are any unprocessed flags remaining treat
1410 * this as a malformed packet. This behavior diverges from
1411 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1412 * in reserved fields are to be ignored. The approach here
1413 * maintains compatibility with previous stack code, and also
1414 * is more robust and provides a little more security in
1415 * adding extensions to VXLAN.
1421 if (!vxlan_set_mac(vxlan
, vs
, skb
, vni
))
1424 skb_reset_mac_header(skb
);
1425 skb
->dev
= vxlan
->dev
;
1426 skb
->pkt_type
= PACKET_HOST
;
1429 oiph
= skb_network_header(skb
);
1430 skb_reset_network_header(skb
);
1432 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1433 ++vxlan
->dev
->stats
.rx_frame_errors
;
1434 ++vxlan
->dev
->stats
.rx_errors
;
1438 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1439 u64_stats_update_begin(&stats
->syncp
);
1440 stats
->rx_packets
++;
1441 stats
->rx_bytes
+= skb
->len
;
1442 u64_stats_update_end(&stats
->syncp
);
1444 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1448 /* Consume bad packet */
1453 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1455 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1456 struct arphdr
*parp
;
1459 struct neighbour
*n
;
1461 if (dev
->flags
& IFF_NOARP
)
1464 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1465 dev
->stats
.tx_dropped
++;
1468 parp
= arp_hdr(skb
);
1470 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1471 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1472 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1473 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1474 parp
->ar_hln
!= dev
->addr_len
||
1477 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1479 arpptr
+= dev
->addr_len
; /* sha */
1480 memcpy(&sip
, arpptr
, sizeof(sip
));
1481 arpptr
+= sizeof(sip
);
1482 arpptr
+= dev
->addr_len
; /* tha */
1483 memcpy(&tip
, arpptr
, sizeof(tip
));
1485 if (ipv4_is_loopback(tip
) ||
1486 ipv4_is_multicast(tip
))
1489 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1492 struct vxlan_fdb
*f
;
1493 struct sk_buff
*reply
;
1495 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1500 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1501 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1502 /* bridge-local neighbor */
1507 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1515 skb_reset_mac_header(reply
);
1516 __skb_pull(reply
, skb_network_offset(reply
));
1517 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1518 reply
->pkt_type
= PACKET_HOST
;
1520 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1521 dev
->stats
.rx_dropped
++;
1522 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
1523 union vxlan_addr ipa
= {
1524 .sin
.sin_addr
.s_addr
= tip
,
1525 .sin
.sin_family
= AF_INET
,
1528 vxlan_ip_miss(dev
, &ipa
);
1532 return NETDEV_TX_OK
;
1535 #if IS_ENABLED(CONFIG_IPV6)
1536 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1537 struct neighbour
*n
, bool isrouter
)
1539 struct net_device
*dev
= request
->dev
;
1540 struct sk_buff
*reply
;
1541 struct nd_msg
*ns
, *na
;
1542 struct ipv6hdr
*pip6
;
1544 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1548 if (dev
== NULL
|| !pskb_may_pull(request
, request
->len
))
1551 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1552 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1553 reply
= alloc_skb(len
, GFP_ATOMIC
);
1557 reply
->protocol
= htons(ETH_P_IPV6
);
1559 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1560 skb_push(reply
, sizeof(struct ethhdr
));
1561 skb_reset_mac_header(reply
);
1563 ns
= (struct nd_msg
*)(ipv6_hdr(request
) + 1);
1565 daddr
= eth_hdr(request
)->h_source
;
1566 ns_olen
= request
->len
- skb_network_offset(request
) -
1567 sizeof(struct ipv6hdr
) - sizeof(*ns
);
1568 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1569 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1570 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1575 /* Ethernet header */
1576 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1577 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1578 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1579 reply
->protocol
= htons(ETH_P_IPV6
);
1581 skb_pull(reply
, sizeof(struct ethhdr
));
1582 skb_reset_network_header(reply
);
1583 skb_put(reply
, sizeof(struct ipv6hdr
));
1587 pip6
= ipv6_hdr(reply
);
1588 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1590 pip6
->priority
= ipv6_hdr(request
)->priority
;
1591 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1592 pip6
->hop_limit
= 255;
1593 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1594 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1596 skb_pull(reply
, sizeof(struct ipv6hdr
));
1597 skb_reset_transport_header(reply
);
1599 /* Neighbor Advertisement */
1600 na
= skb_put_zero(reply
, sizeof(*na
) + na_olen
);
1601 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1602 na
->icmph
.icmp6_router
= isrouter
;
1603 na
->icmph
.icmp6_override
= 1;
1604 na
->icmph
.icmp6_solicited
= 1;
1605 na
->target
= ns
->target
;
1606 ether_addr_copy(&na
->opt
[2], n
->ha
);
1607 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1608 na
->opt
[1] = na_olen
>> 3;
1610 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1611 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1612 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1614 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1616 skb_push(reply
, sizeof(struct ipv6hdr
));
1618 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1623 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1625 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1626 const struct in6_addr
*daddr
;
1627 const struct ipv6hdr
*iphdr
;
1628 struct inet6_dev
*in6_dev
;
1629 struct neighbour
*n
;
1632 in6_dev
= __in6_dev_get(dev
);
1636 iphdr
= ipv6_hdr(skb
);
1637 daddr
= &iphdr
->daddr
;
1638 msg
= (struct nd_msg
*)(iphdr
+ 1);
1640 if (ipv6_addr_loopback(daddr
) ||
1641 ipv6_addr_is_multicast(&msg
->target
))
1644 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1647 struct vxlan_fdb
*f
;
1648 struct sk_buff
*reply
;
1650 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1655 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1656 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1657 /* bridge-local neighbor */
1662 reply
= vxlan_na_create(skb
, n
,
1663 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1670 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1671 dev
->stats
.rx_dropped
++;
1673 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
1674 union vxlan_addr ipa
= {
1675 .sin6
.sin6_addr
= msg
->target
,
1676 .sin6
.sin6_family
= AF_INET6
,
1679 vxlan_ip_miss(dev
, &ipa
);
1684 return NETDEV_TX_OK
;
1688 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1690 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1691 struct neighbour
*n
;
1693 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1697 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1702 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1705 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1706 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
1707 union vxlan_addr ipa
= {
1708 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1709 .sin
.sin_family
= AF_INET
,
1712 vxlan_ip_miss(dev
, &ipa
);
1718 #if IS_ENABLED(CONFIG_IPV6)
1721 struct ipv6hdr
*pip6
;
1723 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1725 pip6
= ipv6_hdr(skb
);
1726 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1727 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
1728 union vxlan_addr ipa
= {
1729 .sin6
.sin6_addr
= pip6
->daddr
,
1730 .sin6
.sin6_family
= AF_INET6
,
1733 vxlan_ip_miss(dev
, &ipa
);
1747 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1749 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1751 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1760 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1761 struct vxlan_metadata
*md
)
1763 struct vxlanhdr_gbp
*gbp
;
1768 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1769 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1771 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1772 gbp
->dont_learn
= 1;
1774 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1775 gbp
->policy_applied
= 1;
1777 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1780 static int vxlan_build_gpe_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1783 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)vxh
;
1785 gpe
->np_applied
= 1;
1786 gpe
->next_protocol
= tun_p_from_eth_p(protocol
);
1787 if (!gpe
->next_protocol
)
1788 return -EPFNOSUPPORT
;
1792 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1793 int iphdr_len
, __be32 vni
,
1794 struct vxlan_metadata
*md
, u32 vxflags
,
1797 struct vxlanhdr
*vxh
;
1800 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1801 __be16 inner_protocol
= htons(ETH_P_TEB
);
1803 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1804 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1805 int csum_start
= skb_checksum_start_offset(skb
);
1807 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1808 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1809 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1810 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
1811 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1814 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1815 + VXLAN_HLEN
+ iphdr_len
;
1817 /* Need space for new headers (invalidates iph ptr) */
1818 err
= skb_cow_head(skb
, min_headroom
);
1822 err
= iptunnel_handle_offloads(skb
, type
);
1826 vxh
= __skb_push(skb
, sizeof(*vxh
));
1827 vxh
->vx_flags
= VXLAN_HF_VNI
;
1828 vxh
->vx_vni
= vxlan_vni_field(vni
);
1830 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1833 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
1834 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
1835 vxh
->vx_flags
|= VXLAN_HF_RCO
;
1837 if (!skb_is_gso(skb
)) {
1838 skb
->ip_summed
= CHECKSUM_NONE
;
1839 skb
->encapsulation
= 0;
1843 if (vxflags
& VXLAN_F_GBP
)
1844 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1845 if (vxflags
& VXLAN_F_GPE
) {
1846 err
= vxlan_build_gpe_hdr(vxh
, vxflags
, skb
->protocol
);
1849 inner_protocol
= skb
->protocol
;
1852 skb_set_inner_protocol(skb
, inner_protocol
);
1856 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
, struct net_device
*dev
,
1857 struct vxlan_sock
*sock4
,
1858 struct sk_buff
*skb
, int oif
, u8 tos
,
1859 __be32 daddr
, __be32
*saddr
, __be16 dport
, __be16 sport
,
1860 struct dst_cache
*dst_cache
,
1861 const struct ip_tunnel_info
*info
)
1863 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1864 struct rtable
*rt
= NULL
;
1868 return ERR_PTR(-EIO
);
1873 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
1878 memset(&fl4
, 0, sizeof(fl4
));
1879 fl4
.flowi4_oif
= oif
;
1880 fl4
.flowi4_tos
= RT_TOS(tos
);
1881 fl4
.flowi4_mark
= skb
->mark
;
1882 fl4
.flowi4_proto
= IPPROTO_UDP
;
1885 fl4
.fl4_dport
= dport
;
1886 fl4
.fl4_sport
= sport
;
1888 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1889 if (likely(!IS_ERR(rt
))) {
1890 if (rt
->dst
.dev
== dev
) {
1891 netdev_dbg(dev
, "circular route to %pI4\n", &daddr
);
1893 return ERR_PTR(-ELOOP
);
1898 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
1900 netdev_dbg(dev
, "no route to %pI4\n", &daddr
);
1901 return ERR_PTR(-ENETUNREACH
);
1906 #if IS_ENABLED(CONFIG_IPV6)
1907 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1908 struct net_device
*dev
,
1909 struct vxlan_sock
*sock6
,
1910 struct sk_buff
*skb
, int oif
, u8 tos
,
1912 const struct in6_addr
*daddr
,
1913 struct in6_addr
*saddr
,
1914 __be16 dport
, __be16 sport
,
1915 struct dst_cache
*dst_cache
,
1916 const struct ip_tunnel_info
*info
)
1918 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1919 struct dst_entry
*ndst
;
1924 return ERR_PTR(-EIO
);
1929 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
1934 memset(&fl6
, 0, sizeof(fl6
));
1935 fl6
.flowi6_oif
= oif
;
1938 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
1939 fl6
.flowi6_mark
= skb
->mark
;
1940 fl6
.flowi6_proto
= IPPROTO_UDP
;
1941 fl6
.fl6_dport
= dport
;
1942 fl6
.fl6_sport
= sport
;
1944 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1947 if (unlikely(err
< 0)) {
1948 netdev_dbg(dev
, "no route to %pI6\n", daddr
);
1949 return ERR_PTR(-ENETUNREACH
);
1952 if (unlikely(ndst
->dev
== dev
)) {
1953 netdev_dbg(dev
, "circular route to %pI6\n", daddr
);
1955 return ERR_PTR(-ELOOP
);
1960 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
1965 /* Bypass encapsulation if the destination is local */
1966 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1967 struct vxlan_dev
*dst_vxlan
, __be32 vni
)
1969 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1970 union vxlan_addr loopback
;
1971 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1972 struct net_device
*dev
= skb
->dev
;
1975 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1976 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1977 skb
->pkt_type
= PACKET_HOST
;
1978 skb
->encapsulation
= 0;
1979 skb
->dev
= dst_vxlan
->dev
;
1980 __skb_pull(skb
, skb_network_offset(skb
));
1982 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1983 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1984 loopback
.sa
.sa_family
= AF_INET
;
1985 #if IS_ENABLED(CONFIG_IPV6)
1987 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1988 loopback
.sa
.sa_family
= AF_INET6
;
1992 if (dst_vxlan
->cfg
.flags
& VXLAN_F_LEARN
)
1993 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
, 0,
1996 u64_stats_update_begin(&tx_stats
->syncp
);
1997 tx_stats
->tx_packets
++;
1998 tx_stats
->tx_bytes
+= len
;
1999 u64_stats_update_end(&tx_stats
->syncp
);
2001 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
2002 u64_stats_update_begin(&rx_stats
->syncp
);
2003 rx_stats
->rx_packets
++;
2004 rx_stats
->rx_bytes
+= len
;
2005 u64_stats_update_end(&rx_stats
->syncp
);
2007 dev
->stats
.rx_dropped
++;
2011 static int encap_bypass_if_local(struct sk_buff
*skb
, struct net_device
*dev
,
2012 struct vxlan_dev
*vxlan
,
2013 union vxlan_addr
*daddr
,
2014 __be16 dst_port
, int dst_ifindex
, __be32 vni
,
2015 struct dst_entry
*dst
,
2018 #if IS_ENABLED(CONFIG_IPV6)
2019 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2020 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2021 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2023 BUILD_BUG_ON(RTCF_LOCAL
!= RTF_LOCAL
);
2025 /* Bypass encapsulation if the destination is local */
2026 if (rt_flags
& RTCF_LOCAL
&&
2027 !(rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2028 struct vxlan_dev
*dst_vxlan
;
2031 dst_vxlan
= vxlan_find_vni(vxlan
->net
, dst_ifindex
, vni
,
2032 daddr
->sa
.sa_family
, dst_port
,
2035 dev
->stats
.tx_errors
++;
2040 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
, vni
);
2047 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
2048 __be32 default_vni
, struct vxlan_rdst
*rdst
,
2051 struct dst_cache
*dst_cache
;
2052 struct ip_tunnel_info
*info
;
2053 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2054 const struct iphdr
*old_iph
= ip_hdr(skb
);
2055 union vxlan_addr
*dst
;
2056 union vxlan_addr remote_ip
, local_ip
;
2057 struct vxlan_metadata _md
;
2058 struct vxlan_metadata
*md
= &_md
;
2059 __be16 src_port
= 0, dst_port
;
2060 struct dst_entry
*ndst
= NULL
;
2065 u32 flags
= vxlan
->cfg
.flags
;
2066 bool udp_sum
= false;
2067 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
2069 info
= skb_tunnel_info(skb
);
2072 dst
= &rdst
->remote_ip
;
2073 if (vxlan_addr_any(dst
)) {
2075 /* short-circuited back to local bridge */
2076 vxlan_encap_bypass(skb
, vxlan
, vxlan
, default_vni
);
2082 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
2083 vni
= (rdst
->remote_vni
) ? : default_vni
;
2084 ifindex
= rdst
->remote_ifindex
;
2085 local_ip
= vxlan
->cfg
.saddr
;
2086 dst_cache
= &rdst
->dst_cache
;
2087 md
->gbp
= skb
->mark
;
2088 ttl
= vxlan
->cfg
.ttl
;
2089 if (!ttl
&& vxlan_addr_multicast(dst
))
2092 tos
= vxlan
->cfg
.tos
;
2094 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
2096 if (dst
->sa
.sa_family
== AF_INET
)
2097 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
2099 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2100 label
= vxlan
->cfg
.label
;
2103 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2107 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
2108 if (remote_ip
.sa
.sa_family
== AF_INET
) {
2109 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
2110 local_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.src
;
2112 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
2113 local_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.src
;
2116 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2117 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2119 dst_cache
= &info
->dst_cache
;
2120 if (info
->options_len
)
2121 md
= ip_tunnel_info_opts(info
);
2122 ttl
= info
->key
.ttl
;
2123 tos
= info
->key
.tos
;
2124 label
= info
->key
.label
;
2125 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
2127 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2128 vxlan
->cfg
.port_max
, true);
2131 if (dst
->sa
.sa_family
== AF_INET
) {
2132 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2136 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, ifindex
, tos
,
2137 dst
->sin
.sin_addr
.s_addr
,
2138 &local_ip
.sin
.sin_addr
.s_addr
,
2146 /* Bypass encapsulation if the destination is local */
2148 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2149 dst_port
, ifindex
, vni
,
2150 &rt
->dst
, rt
->rt_flags
);
2153 } else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
) {
2159 int mtu
= dst_mtu(ndst
) - VXLAN_HEADROOM
;
2161 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
,
2165 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2166 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2167 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct iphdr
),
2168 vni
, md
, flags
, udp_sum
);
2172 udp_tunnel_xmit_skb(rt
, sock4
->sock
->sk
, skb
, local_ip
.sin
.sin_addr
.s_addr
,
2173 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2174 src_port
, dst_port
, xnet
, !udp_sum
);
2175 #if IS_ENABLED(CONFIG_IPV6)
2177 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2179 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, ifindex
, tos
,
2180 label
, &dst
->sin6
.sin6_addr
,
2181 &local_ip
.sin6
.sin6_addr
,
2185 err
= PTR_ERR(ndst
);
2191 u32 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2193 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2194 dst_port
, ifindex
, vni
,
2201 int mtu
= dst_mtu(ndst
) - VXLAN6_HEADROOM
;
2203 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
,
2207 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2208 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2209 skb_scrub_packet(skb
, xnet
);
2210 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2211 vni
, md
, flags
, udp_sum
);
2215 udp_tunnel6_xmit_skb(ndst
, sock6
->sock
->sk
, skb
, dev
,
2216 &local_ip
.sin6
.sin6_addr
,
2217 &dst
->sin6
.sin6_addr
, tos
, ttl
,
2218 label
, src_port
, dst_port
, !udp_sum
);
2226 dev
->stats
.tx_dropped
++;
2233 dev
->stats
.collisions
++;
2234 else if (err
== -ENETUNREACH
)
2235 dev
->stats
.tx_carrier_errors
++;
2237 dev
->stats
.tx_errors
++;
2241 /* Transmit local packets over Vxlan
2243 * Outer IP header inherits ECN and DF from inner header.
2244 * Outer UDP destination is the VXLAN assigned port.
2245 * source port is based on hash of flow
2247 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2249 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2250 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2251 const struct ip_tunnel_info
*info
;
2252 bool did_rsc
= false;
2253 struct vxlan_fdb
*f
;
2257 info
= skb_tunnel_info(skb
);
2259 skb_reset_mac_header(skb
);
2261 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
2262 if (info
&& info
->mode
& IP_TUNNEL_INFO_BRIDGE
&&
2263 info
->mode
& IP_TUNNEL_INFO_TX
) {
2264 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2266 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2267 vxlan_xmit_one(skb
, dev
, vni
, NULL
, false);
2270 return NETDEV_TX_OK
;
2274 if (vxlan
->cfg
.flags
& VXLAN_F_PROXY
) {
2276 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2277 return arp_reduce(dev
, skb
, vni
);
2278 #if IS_ENABLED(CONFIG_IPV6)
2279 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2280 pskb_may_pull(skb
, sizeof(struct ipv6hdr
) +
2281 sizeof(struct nd_msg
)) &&
2282 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2283 struct nd_msg
*m
= (struct nd_msg
*)(ipv6_hdr(skb
) + 1);
2285 if (m
->icmph
.icmp6_code
== 0 &&
2286 m
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2287 return neigh_reduce(dev
, skb
, vni
);
2293 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2296 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->cfg
.flags
& VXLAN_F_RSC
) &&
2297 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2298 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2299 did_rsc
= route_shortcircuit(dev
, skb
);
2301 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2305 f
= vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2307 if ((vxlan
->cfg
.flags
& VXLAN_F_L2MISS
) &&
2308 !is_multicast_ether_addr(eth
->h_dest
))
2309 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2311 dev
->stats
.tx_dropped
++;
2313 return NETDEV_TX_OK
;
2317 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2318 struct sk_buff
*skb1
;
2324 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2326 vxlan_xmit_one(skb1
, dev
, vni
, rdst
, did_rsc
);
2330 vxlan_xmit_one(skb
, dev
, vni
, fdst
, did_rsc
);
2333 return NETDEV_TX_OK
;
2336 /* Walk the forwarding table and purge stale entries */
2337 static void vxlan_cleanup(struct timer_list
*t
)
2339 struct vxlan_dev
*vxlan
= from_timer(vxlan
, t
, age_timer
);
2340 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2343 if (!netif_running(vxlan
->dev
))
2346 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2347 struct hlist_node
*p
, *n
;
2349 spin_lock_bh(&vxlan
->hash_lock
);
2350 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2352 = container_of(p
, struct vxlan_fdb
, hlist
);
2353 unsigned long timeout
;
2355 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
2358 if (f
->flags
& NTF_EXT_LEARNED
)
2361 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2362 if (time_before_eq(timeout
, jiffies
)) {
2363 netdev_dbg(vxlan
->dev
,
2364 "garbage collect %pM\n",
2366 f
->state
= NUD_STALE
;
2367 vxlan_fdb_destroy(vxlan
, f
);
2368 } else if (time_before(timeout
, next_timer
))
2369 next_timer
= timeout
;
2371 spin_unlock_bh(&vxlan
->hash_lock
);
2374 mod_timer(&vxlan
->age_timer
, next_timer
);
2377 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
)
2379 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2381 spin_lock(&vn
->sock_lock
);
2382 hlist_del_init_rcu(&vxlan
->hlist4
.hlist
);
2383 #if IS_ENABLED(CONFIG_IPV6)
2384 hlist_del_init_rcu(&vxlan
->hlist6
.hlist
);
2386 spin_unlock(&vn
->sock_lock
);
2389 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
,
2390 struct vxlan_dev_node
*node
)
2392 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2393 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2395 node
->vxlan
= vxlan
;
2396 spin_lock(&vn
->sock_lock
);
2397 hlist_add_head_rcu(&node
->hlist
, vni_head(vs
, vni
));
2398 spin_unlock(&vn
->sock_lock
);
2401 /* Setup stats when device is created */
2402 static int vxlan_init(struct net_device
*dev
)
2404 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2411 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
, __be32 vni
)
2413 struct vxlan_fdb
*f
;
2415 spin_lock_bh(&vxlan
->hash_lock
);
2416 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2418 vxlan_fdb_destroy(vxlan
, f
);
2419 spin_unlock_bh(&vxlan
->hash_lock
);
2422 static void vxlan_uninit(struct net_device
*dev
)
2424 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2426 vxlan_fdb_delete_default(vxlan
, vxlan
->cfg
.vni
);
2428 free_percpu(dev
->tstats
);
2431 /* Start ageing timer and join group when device is brought up */
2432 static int vxlan_open(struct net_device
*dev
)
2434 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2437 ret
= vxlan_sock_add(vxlan
);
2441 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2442 ret
= vxlan_igmp_join(vxlan
);
2443 if (ret
== -EADDRINUSE
)
2446 vxlan_sock_release(vxlan
);
2451 if (vxlan
->cfg
.age_interval
)
2452 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2457 /* Purge the forwarding table */
2458 static void vxlan_flush(struct vxlan_dev
*vxlan
, bool do_all
)
2462 spin_lock_bh(&vxlan
->hash_lock
);
2463 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2464 struct hlist_node
*p
, *n
;
2465 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2467 = container_of(p
, struct vxlan_fdb
, hlist
);
2468 if (!do_all
&& (f
->state
& (NUD_PERMANENT
| NUD_NOARP
)))
2470 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2471 if (!is_zero_ether_addr(f
->eth_addr
))
2472 vxlan_fdb_destroy(vxlan
, f
);
2475 spin_unlock_bh(&vxlan
->hash_lock
);
2478 /* Cleanup timer and forwarding table on shutdown */
2479 static int vxlan_stop(struct net_device
*dev
)
2481 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2482 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2485 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2486 !vxlan_group_used(vn
, vxlan
))
2487 ret
= vxlan_igmp_leave(vxlan
);
2489 del_timer_sync(&vxlan
->age_timer
);
2491 vxlan_flush(vxlan
, false);
2492 vxlan_sock_release(vxlan
);
2497 /* Stub, nothing needs to be done. */
2498 static void vxlan_set_multicast_list(struct net_device
*dev
)
2502 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2504 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2505 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2506 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2507 dst
->remote_ifindex
);
2508 bool use_ipv6
= !!(vxlan
->cfg
.flags
& VXLAN_F_IPV6
);
2510 /* This check is different than dev->max_mtu, because it looks at
2511 * the lowerdev->mtu, rather than the static dev->max_mtu
2514 int max_mtu
= lowerdev
->mtu
-
2515 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2516 if (new_mtu
> max_mtu
)
2524 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2526 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2527 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2528 __be16 sport
, dport
;
2530 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2531 vxlan
->cfg
.port_max
, true);
2532 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2534 if (ip_tunnel_info_af(info
) == AF_INET
) {
2535 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2538 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, 0, info
->key
.tos
,
2539 info
->key
.u
.ipv4
.dst
,
2540 &info
->key
.u
.ipv4
.src
, dport
, sport
,
2541 &info
->dst_cache
, info
);
2546 #if IS_ENABLED(CONFIG_IPV6)
2547 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2548 struct dst_entry
*ndst
;
2550 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, 0, info
->key
.tos
,
2551 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2552 &info
->key
.u
.ipv6
.src
, dport
, sport
,
2553 &info
->dst_cache
, info
);
2555 return PTR_ERR(ndst
);
2557 #else /* !CONFIG_IPV6 */
2558 return -EPFNOSUPPORT
;
2561 info
->key
.tp_src
= sport
;
2562 info
->key
.tp_dst
= dport
;
2566 static const struct net_device_ops vxlan_netdev_ether_ops
= {
2567 .ndo_init
= vxlan_init
,
2568 .ndo_uninit
= vxlan_uninit
,
2569 .ndo_open
= vxlan_open
,
2570 .ndo_stop
= vxlan_stop
,
2571 .ndo_start_xmit
= vxlan_xmit
,
2572 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2573 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2574 .ndo_change_mtu
= vxlan_change_mtu
,
2575 .ndo_validate_addr
= eth_validate_addr
,
2576 .ndo_set_mac_address
= eth_mac_addr
,
2577 .ndo_fdb_add
= vxlan_fdb_add
,
2578 .ndo_fdb_del
= vxlan_fdb_delete
,
2579 .ndo_fdb_dump
= vxlan_fdb_dump
,
2580 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2583 static const struct net_device_ops vxlan_netdev_raw_ops
= {
2584 .ndo_init
= vxlan_init
,
2585 .ndo_uninit
= vxlan_uninit
,
2586 .ndo_open
= vxlan_open
,
2587 .ndo_stop
= vxlan_stop
,
2588 .ndo_start_xmit
= vxlan_xmit
,
2589 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2590 .ndo_change_mtu
= vxlan_change_mtu
,
2591 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2594 /* Info for udev, that this is a virtual tunnel endpoint */
2595 static struct device_type vxlan_type
= {
2599 /* Calls the ndo_udp_tunnel_add of the caller in order to
2600 * supply the listening VXLAN udp ports. Callers are expected
2601 * to implement the ndo_udp_tunnel_add.
2603 static void vxlan_offload_rx_ports(struct net_device
*dev
, bool push
)
2605 struct vxlan_sock
*vs
;
2606 struct net
*net
= dev_net(dev
);
2607 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2610 spin_lock(&vn
->sock_lock
);
2611 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2612 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2613 unsigned short type
;
2615 if (vs
->flags
& VXLAN_F_GPE
)
2616 type
= UDP_TUNNEL_TYPE_VXLAN_GPE
;
2618 type
= UDP_TUNNEL_TYPE_VXLAN
;
2621 udp_tunnel_push_rx_port(dev
, vs
->sock
, type
);
2623 udp_tunnel_drop_rx_port(dev
, vs
->sock
, type
);
2626 spin_unlock(&vn
->sock_lock
);
2629 /* Initialize the device structure. */
2630 static void vxlan_setup(struct net_device
*dev
)
2632 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2635 eth_hw_addr_random(dev
);
2638 dev
->needs_free_netdev
= true;
2639 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2641 dev
->features
|= NETIF_F_LLTX
;
2642 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2643 dev
->features
|= NETIF_F_RXCSUM
;
2644 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2646 dev
->vlan_features
= dev
->features
;
2647 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2648 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2649 netif_keep_dst(dev
);
2650 dev
->priv_flags
|= IFF_NO_QUEUE
;
2652 /* MTU range: 68 - 65535 */
2653 dev
->min_mtu
= ETH_MIN_MTU
;
2654 dev
->max_mtu
= ETH_MAX_MTU
;
2656 INIT_LIST_HEAD(&vxlan
->next
);
2657 spin_lock_init(&vxlan
->hash_lock
);
2659 timer_setup(&vxlan
->age_timer
, vxlan_cleanup
, TIMER_DEFERRABLE
);
2663 gro_cells_init(&vxlan
->gro_cells
, dev
);
2665 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2666 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2669 static void vxlan_ether_setup(struct net_device
*dev
)
2671 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2672 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2673 dev
->netdev_ops
= &vxlan_netdev_ether_ops
;
2676 static void vxlan_raw_setup(struct net_device
*dev
)
2678 dev
->header_ops
= NULL
;
2679 dev
->type
= ARPHRD_NONE
;
2680 dev
->hard_header_len
= 0;
2682 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
2683 dev
->netdev_ops
= &vxlan_netdev_raw_ops
;
2686 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2687 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2688 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2689 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2690 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2691 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2692 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2693 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2694 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2695 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
2696 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2697 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2698 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2699 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2700 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2701 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2702 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2703 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2704 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2705 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2706 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2707 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2708 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2709 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2710 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2711 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2712 [IFLA_VXLAN_GPE
] = { .type
= NLA_FLAG
, },
2713 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2716 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
2717 struct netlink_ext_ack
*extack
)
2719 if (tb
[IFLA_ADDRESS
]) {
2720 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2721 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
2722 "Provided link layer address is not Ethernet");
2726 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2727 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
2728 "Provided Ethernet address is not unicast");
2729 return -EADDRNOTAVAIL
;
2734 u32 mtu
= nla_get_u32(tb
[IFLA_MTU
]);
2736 if (mtu
< ETH_MIN_MTU
|| mtu
> ETH_MAX_MTU
) {
2737 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_MTU
],
2738 "MTU must be between 68 and 65535");
2744 NL_SET_ERR_MSG(extack
,
2745 "Required attributes not provided to perform the operation");
2749 if (data
[IFLA_VXLAN_ID
]) {
2750 u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2752 if (id
>= VXLAN_N_VID
) {
2753 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_ID
],
2754 "VXLAN ID must be lower than 16777216");
2759 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2760 const struct ifla_vxlan_port_range
*p
2761 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2763 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2764 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_PORT_RANGE
],
2765 "Invalid source port range");
2773 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2774 struct ethtool_drvinfo
*drvinfo
)
2776 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2777 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2780 static const struct ethtool_ops vxlan_ethtool_ops
= {
2781 .get_drvinfo
= vxlan_get_drvinfo
,
2782 .get_link
= ethtool_op_get_link
,
2785 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2786 __be16 port
, u32 flags
)
2788 struct socket
*sock
;
2789 struct udp_port_cfg udp_conf
;
2792 memset(&udp_conf
, 0, sizeof(udp_conf
));
2795 udp_conf
.family
= AF_INET6
;
2796 udp_conf
.use_udp6_rx_checksums
=
2797 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2798 udp_conf
.ipv6_v6only
= 1;
2800 udp_conf
.family
= AF_INET
;
2803 udp_conf
.local_udp_port
= port
;
2805 /* Open UDP socket */
2806 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2808 return ERR_PTR(err
);
2813 /* Create new listen socket if needed */
2814 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2815 __be16 port
, u32 flags
)
2817 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2818 struct vxlan_sock
*vs
;
2819 struct socket
*sock
;
2821 struct udp_tunnel_sock_cfg tunnel_cfg
;
2823 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2825 return ERR_PTR(-ENOMEM
);
2827 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2828 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2830 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2833 return ERR_CAST(sock
);
2837 refcount_set(&vs
->refcnt
, 1);
2838 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2840 spin_lock(&vn
->sock_lock
);
2841 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2842 udp_tunnel_notify_add_rx_port(sock
,
2843 (vs
->flags
& VXLAN_F_GPE
) ?
2844 UDP_TUNNEL_TYPE_VXLAN_GPE
:
2845 UDP_TUNNEL_TYPE_VXLAN
);
2846 spin_unlock(&vn
->sock_lock
);
2848 /* Mark socket as an encapsulation socket. */
2849 memset(&tunnel_cfg
, 0, sizeof(tunnel_cfg
));
2850 tunnel_cfg
.sk_user_data
= vs
;
2851 tunnel_cfg
.encap_type
= 1;
2852 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
2853 tunnel_cfg
.encap_destroy
= NULL
;
2854 tunnel_cfg
.gro_receive
= vxlan_gro_receive
;
2855 tunnel_cfg
.gro_complete
= vxlan_gro_complete
;
2857 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2862 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2864 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2865 struct vxlan_sock
*vs
= NULL
;
2866 struct vxlan_dev_node
*node
;
2868 if (!vxlan
->cfg
.no_share
) {
2869 spin_lock(&vn
->sock_lock
);
2870 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2871 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
);
2872 if (vs
&& !refcount_inc_not_zero(&vs
->refcnt
)) {
2873 spin_unlock(&vn
->sock_lock
);
2876 spin_unlock(&vn
->sock_lock
);
2879 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2880 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
);
2883 #if IS_ENABLED(CONFIG_IPV6)
2885 rcu_assign_pointer(vxlan
->vn6_sock
, vs
);
2886 node
= &vxlan
->hlist6
;
2890 rcu_assign_pointer(vxlan
->vn4_sock
, vs
);
2891 node
= &vxlan
->hlist4
;
2893 vxlan_vs_add_dev(vs
, vxlan
, node
);
2897 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2899 bool metadata
= vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
;
2900 bool ipv6
= vxlan
->cfg
.flags
& VXLAN_F_IPV6
|| metadata
;
2901 bool ipv4
= !ipv6
|| metadata
;
2904 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
2905 #if IS_ENABLED(CONFIG_IPV6)
2906 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
2908 ret
= __vxlan_sock_add(vxlan
, true);
2909 if (ret
< 0 && ret
!= -EAFNOSUPPORT
)
2914 ret
= __vxlan_sock_add(vxlan
, false);
2916 vxlan_sock_release(vxlan
);
2920 static int vxlan_config_validate(struct net
*src_net
, struct vxlan_config
*conf
,
2921 struct net_device
**lower
,
2922 struct vxlan_dev
*old
,
2923 struct netlink_ext_ack
*extack
)
2925 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2926 struct vxlan_dev
*tmp
;
2927 bool use_ipv6
= false;
2929 if (conf
->flags
& VXLAN_F_GPE
) {
2930 /* For now, allow GPE only together with
2931 * COLLECT_METADATA. This can be relaxed later; in such
2932 * case, the other side of the PtP link will have to be
2935 if ((conf
->flags
& ~VXLAN_F_ALLOWED_GPE
) ||
2936 !(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
2937 NL_SET_ERR_MSG(extack
,
2938 "VXLAN GPE does not support this combination of attributes");
2943 if (!conf
->remote_ip
.sa
.sa_family
&& !conf
->saddr
.sa
.sa_family
) {
2944 /* Unless IPv6 is explicitly requested, assume IPv4 */
2945 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
2946 conf
->saddr
.sa
.sa_family
= AF_INET
;
2947 } else if (!conf
->remote_ip
.sa
.sa_family
) {
2948 conf
->remote_ip
.sa
.sa_family
= conf
->saddr
.sa
.sa_family
;
2949 } else if (!conf
->saddr
.sa
.sa_family
) {
2950 conf
->saddr
.sa
.sa_family
= conf
->remote_ip
.sa
.sa_family
;
2953 if (conf
->saddr
.sa
.sa_family
!= conf
->remote_ip
.sa
.sa_family
) {
2954 NL_SET_ERR_MSG(extack
,
2955 "Local and remote address must be from the same family");
2959 if (vxlan_addr_multicast(&conf
->saddr
)) {
2960 NL_SET_ERR_MSG(extack
, "Local address cannot be multicast");
2964 if (conf
->saddr
.sa
.sa_family
== AF_INET6
) {
2965 if (!IS_ENABLED(CONFIG_IPV6
)) {
2966 NL_SET_ERR_MSG(extack
,
2967 "IPv6 support not enabled in the kernel");
2968 return -EPFNOSUPPORT
;
2971 conf
->flags
|= VXLAN_F_IPV6
;
2973 if (!(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
2975 ipv6_addr_type(&conf
->saddr
.sin6
.sin6_addr
);
2977 ipv6_addr_type(&conf
->remote_ip
.sin6
.sin6_addr
);
2979 if (local_type
& IPV6_ADDR_LINKLOCAL
) {
2980 if (!(remote_type
& IPV6_ADDR_LINKLOCAL
) &&
2981 (remote_type
!= IPV6_ADDR_ANY
)) {
2982 NL_SET_ERR_MSG(extack
,
2983 "Invalid combination of local and remote address scopes");
2987 conf
->flags
|= VXLAN_F_IPV6_LINKLOCAL
;
2990 (IPV6_ADDR_UNICAST
| IPV6_ADDR_LINKLOCAL
)) {
2991 NL_SET_ERR_MSG(extack
,
2992 "Invalid combination of local and remote address scopes");
2996 conf
->flags
&= ~VXLAN_F_IPV6_LINKLOCAL
;
3001 if (conf
->label
&& !use_ipv6
) {
3002 NL_SET_ERR_MSG(extack
,
3003 "Label attribute only applies to IPv6 VXLAN devices");
3007 if (conf
->remote_ifindex
) {
3008 struct net_device
*lowerdev
;
3010 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
3012 NL_SET_ERR_MSG(extack
,
3013 "Invalid local interface, device not found");
3017 #if IS_ENABLED(CONFIG_IPV6)
3019 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
3020 if (idev
&& idev
->cnf
.disable_ipv6
) {
3021 NL_SET_ERR_MSG(extack
,
3022 "IPv6 support disabled by administrator");
3030 if (vxlan_addr_multicast(&conf
->remote_ip
)) {
3031 NL_SET_ERR_MSG(extack
,
3032 "Local interface required for multicast remote destination");
3037 #if IS_ENABLED(CONFIG_IPV6)
3038 if (conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) {
3039 NL_SET_ERR_MSG(extack
,
3040 "Local interface required for link-local local/remote addresses");
3048 if (!conf
->dst_port
) {
3049 if (conf
->flags
& VXLAN_F_GPE
)
3050 conf
->dst_port
= htons(4790); /* IANA VXLAN-GPE port */
3052 conf
->dst_port
= htons(vxlan_port
);
3055 if (!conf
->age_interval
)
3056 conf
->age_interval
= FDB_AGE_DEFAULT
;
3058 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
3062 if (tmp
->cfg
.vni
!= conf
->vni
)
3064 if (tmp
->cfg
.dst_port
!= conf
->dst_port
)
3066 if ((tmp
->cfg
.flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)) !=
3067 (conf
->flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)))
3070 if ((conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
3071 tmp
->cfg
.remote_ifindex
!= conf
->remote_ifindex
)
3074 NL_SET_ERR_MSG(extack
,
3075 "A VXLAN device with the specified VNI already exists");
3082 static void vxlan_config_apply(struct net_device
*dev
,
3083 struct vxlan_config
*conf
,
3084 struct net_device
*lowerdev
,
3085 struct net
*src_net
,
3088 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3089 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3090 unsigned short needed_headroom
= ETH_HLEN
;
3091 bool use_ipv6
= !!(conf
->flags
& VXLAN_F_IPV6
);
3092 int max_mtu
= ETH_MAX_MTU
;
3095 if (conf
->flags
& VXLAN_F_GPE
)
3096 vxlan_raw_setup(dev
);
3098 vxlan_ether_setup(dev
);
3101 dev
->mtu
= conf
->mtu
;
3103 vxlan
->net
= src_net
;
3106 dst
->remote_vni
= conf
->vni
;
3108 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
3111 dst
->remote_ifindex
= conf
->remote_ifindex
;
3113 dev
->gso_max_size
= lowerdev
->gso_max_size
;
3114 dev
->gso_max_segs
= lowerdev
->gso_max_segs
;
3116 needed_headroom
= lowerdev
->hard_header_len
;
3118 max_mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
:
3120 if (max_mtu
< ETH_MIN_MTU
)
3121 max_mtu
= ETH_MIN_MTU
;
3123 if (!changelink
&& !conf
->mtu
)
3127 if (dev
->mtu
> max_mtu
)
3130 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
3131 needed_headroom
+= VXLAN6_HEADROOM
;
3133 needed_headroom
+= VXLAN_HEADROOM
;
3134 dev
->needed_headroom
= needed_headroom
;
3136 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
3139 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
3140 struct vxlan_config
*conf
, bool changelink
,
3141 struct netlink_ext_ack
*extack
)
3143 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3144 struct net_device
*lowerdev
;
3147 ret
= vxlan_config_validate(src_net
, conf
, &lowerdev
, vxlan
, extack
);
3151 vxlan_config_apply(dev
, conf
, lowerdev
, src_net
, changelink
);
3156 static int __vxlan_dev_create(struct net
*net
, struct net_device
*dev
,
3157 struct vxlan_config
*conf
,
3158 struct netlink_ext_ack
*extack
)
3160 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3161 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3164 err
= vxlan_dev_configure(net
, dev
, conf
, false, extack
);
3168 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
3170 /* create an fdb entry for a valid default destination */
3171 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
3172 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3173 &vxlan
->default_dst
.remote_ip
,
3174 NUD_REACHABLE
| NUD_PERMANENT
,
3175 NLM_F_EXCL
| NLM_F_CREATE
,
3176 vxlan
->cfg
.dst_port
,
3177 vxlan
->default_dst
.remote_vni
,
3178 vxlan
->default_dst
.remote_vni
,
3179 vxlan
->default_dst
.remote_ifindex
,
3185 err
= register_netdevice(dev
);
3187 vxlan_fdb_delete_default(vxlan
, vxlan
->default_dst
.remote_vni
);
3191 list_add(&vxlan
->next
, &vn
->vxlan_list
);
3195 static int vxlan_nl2conf(struct nlattr
*tb
[], struct nlattr
*data
[],
3196 struct net_device
*dev
, struct vxlan_config
*conf
,
3199 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3201 memset(conf
, 0, sizeof(*conf
));
3203 /* if changelink operation, start with old existing cfg */
3205 memcpy(conf
, &vxlan
->cfg
, sizeof(*conf
));
3207 if (data
[IFLA_VXLAN_ID
]) {
3208 __be32 vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3210 if (changelink
&& (vni
!= conf
->vni
))
3212 conf
->vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3215 if (data
[IFLA_VXLAN_GROUP
]) {
3216 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET
))
3219 conf
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
3220 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
3221 } else if (data
[IFLA_VXLAN_GROUP6
]) {
3222 if (!IS_ENABLED(CONFIG_IPV6
))
3223 return -EPFNOSUPPORT
;
3225 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET6
))
3228 conf
->remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
3229 conf
->remote_ip
.sa
.sa_family
= AF_INET6
;
3232 if (data
[IFLA_VXLAN_LOCAL
]) {
3233 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET
))
3236 conf
->saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
3237 conf
->saddr
.sa
.sa_family
= AF_INET
;
3238 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
3239 if (!IS_ENABLED(CONFIG_IPV6
))
3240 return -EPFNOSUPPORT
;
3242 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET6
))
3245 /* TODO: respect scope id */
3246 conf
->saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
3247 conf
->saddr
.sa
.sa_family
= AF_INET6
;
3250 if (data
[IFLA_VXLAN_LINK
])
3251 conf
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
3253 if (data
[IFLA_VXLAN_TOS
])
3254 conf
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
3256 if (data
[IFLA_VXLAN_TTL
])
3257 conf
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
3259 if (data
[IFLA_VXLAN_LABEL
])
3260 conf
->label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
3261 IPV6_FLOWLABEL_MASK
;
3263 if (data
[IFLA_VXLAN_LEARNING
]) {
3264 if (nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
3265 conf
->flags
|= VXLAN_F_LEARN
;
3267 conf
->flags
&= ~VXLAN_F_LEARN
;
3268 } else if (!changelink
) {
3269 /* default to learn on a new device */
3270 conf
->flags
|= VXLAN_F_LEARN
;
3273 if (data
[IFLA_VXLAN_AGEING
]) {
3276 conf
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
3279 if (data
[IFLA_VXLAN_PROXY
]) {
3282 if (nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
3283 conf
->flags
|= VXLAN_F_PROXY
;
3286 if (data
[IFLA_VXLAN_RSC
]) {
3289 if (nla_get_u8(data
[IFLA_VXLAN_RSC
]))
3290 conf
->flags
|= VXLAN_F_RSC
;
3293 if (data
[IFLA_VXLAN_L2MISS
]) {
3296 if (nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
3297 conf
->flags
|= VXLAN_F_L2MISS
;
3300 if (data
[IFLA_VXLAN_L3MISS
]) {
3303 if (nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
3304 conf
->flags
|= VXLAN_F_L3MISS
;
3307 if (data
[IFLA_VXLAN_LIMIT
]) {
3310 conf
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
3313 if (data
[IFLA_VXLAN_COLLECT_METADATA
]) {
3316 if (nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
3317 conf
->flags
|= VXLAN_F_COLLECT_METADATA
;
3320 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3322 const struct ifla_vxlan_port_range
*p
3323 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3324 conf
->port_min
= ntohs(p
->low
);
3325 conf
->port_max
= ntohs(p
->high
);
3331 if (data
[IFLA_VXLAN_PORT
]) {
3334 conf
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
3337 if (data
[IFLA_VXLAN_UDP_CSUM
]) {
3340 if (!nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
3341 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
3344 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]) {
3347 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
3348 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
3351 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]) {
3354 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
3355 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
3358 if (data
[IFLA_VXLAN_REMCSUM_TX
]) {
3361 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
3362 conf
->flags
|= VXLAN_F_REMCSUM_TX
;
3365 if (data
[IFLA_VXLAN_REMCSUM_RX
]) {
3368 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
3369 conf
->flags
|= VXLAN_F_REMCSUM_RX
;
3372 if (data
[IFLA_VXLAN_GBP
]) {
3375 conf
->flags
|= VXLAN_F_GBP
;
3378 if (data
[IFLA_VXLAN_GPE
]) {
3381 conf
->flags
|= VXLAN_F_GPE
;
3384 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
]) {
3387 conf
->flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
3393 conf
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3399 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
3400 struct nlattr
*tb
[], struct nlattr
*data
[],
3401 struct netlink_ext_ack
*extack
)
3403 struct vxlan_config conf
;
3406 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, false);
3410 return __vxlan_dev_create(src_net
, dev
, &conf
, extack
);
3413 static int vxlan_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
3414 struct nlattr
*data
[],
3415 struct netlink_ext_ack
*extack
)
3417 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3418 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3419 struct vxlan_rdst old_dst
;
3420 struct vxlan_config conf
;
3423 err
= vxlan_nl2conf(tb
, data
,
3428 memcpy(&old_dst
, dst
, sizeof(struct vxlan_rdst
));
3430 err
= vxlan_dev_configure(vxlan
->net
, dev
, &conf
, true, extack
);
3434 /* handle default dst entry */
3435 if (!vxlan_addr_equal(&dst
->remote_ip
, &old_dst
.remote_ip
)) {
3436 spin_lock_bh(&vxlan
->hash_lock
);
3437 if (!vxlan_addr_any(&old_dst
.remote_ip
))
3438 __vxlan_fdb_delete(vxlan
, all_zeros_mac
,
3440 vxlan
->cfg
.dst_port
,
3443 old_dst
.remote_ifindex
, 0);
3445 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3446 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3448 NUD_REACHABLE
| NUD_PERMANENT
,
3449 NLM_F_CREATE
| NLM_F_APPEND
,
3450 vxlan
->cfg
.dst_port
,
3453 dst
->remote_ifindex
,
3456 spin_unlock_bh(&vxlan
->hash_lock
);
3460 spin_unlock_bh(&vxlan
->hash_lock
);
3466 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
3468 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3470 vxlan_flush(vxlan
, true);
3472 gro_cells_destroy(&vxlan
->gro_cells
);
3473 list_del(&vxlan
->next
);
3474 unregister_netdevice_queue(dev
, head
);
3477 static size_t vxlan_get_size(const struct net_device
*dev
)
3480 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
3481 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
3482 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
3483 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
3484 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3485 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3486 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
3487 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3488 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3489 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3490 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3491 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3492 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3493 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3494 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3495 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3496 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3497 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3498 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3499 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3500 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3501 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3505 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3507 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3508 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3509 struct ifla_vxlan_port_range ports
= {
3510 .low
= htons(vxlan
->cfg
.port_min
),
3511 .high
= htons(vxlan
->cfg
.port_max
),
3514 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3515 goto nla_put_failure
;
3517 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3518 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3519 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3520 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3521 goto nla_put_failure
;
3522 #if IS_ENABLED(CONFIG_IPV6)
3524 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3525 &dst
->remote_ip
.sin6
.sin6_addr
))
3526 goto nla_put_failure
;
3531 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3532 goto nla_put_failure
;
3534 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3535 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3536 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3537 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3538 goto nla_put_failure
;
3539 #if IS_ENABLED(CONFIG_IPV6)
3541 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3542 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3543 goto nla_put_failure
;
3548 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3549 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3550 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
3551 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3552 !!(vxlan
->cfg
.flags
& VXLAN_F_LEARN
)) ||
3553 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3554 !!(vxlan
->cfg
.flags
& VXLAN_F_PROXY
)) ||
3555 nla_put_u8(skb
, IFLA_VXLAN_RSC
,
3556 !!(vxlan
->cfg
.flags
& VXLAN_F_RSC
)) ||
3557 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3558 !!(vxlan
->cfg
.flags
& VXLAN_F_L2MISS
)) ||
3559 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3560 !!(vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) ||
3561 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3562 !!(vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)) ||
3563 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3564 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3565 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3566 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3567 !(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3568 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3569 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3570 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3571 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3572 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3573 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_TX
)) ||
3574 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3575 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_RX
)))
3576 goto nla_put_failure
;
3578 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3579 goto nla_put_failure
;
3581 if (vxlan
->cfg
.flags
& VXLAN_F_GBP
&&
3582 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3583 goto nla_put_failure
;
3585 if (vxlan
->cfg
.flags
& VXLAN_F_GPE
&&
3586 nla_put_flag(skb
, IFLA_VXLAN_GPE
))
3587 goto nla_put_failure
;
3589 if (vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3590 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3591 goto nla_put_failure
;
3599 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3601 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3606 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3608 .maxtype
= IFLA_VXLAN_MAX
,
3609 .policy
= vxlan_policy
,
3610 .priv_size
= sizeof(struct vxlan_dev
),
3611 .setup
= vxlan_setup
,
3612 .validate
= vxlan_validate
,
3613 .newlink
= vxlan_newlink
,
3614 .changelink
= vxlan_changelink
,
3615 .dellink
= vxlan_dellink
,
3616 .get_size
= vxlan_get_size
,
3617 .fill_info
= vxlan_fill_info
,
3618 .get_link_net
= vxlan_get_link_net
,
3621 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
3622 u8 name_assign_type
,
3623 struct vxlan_config
*conf
)
3625 struct nlattr
*tb
[IFLA_MAX
+ 1];
3626 struct net_device
*dev
;
3629 memset(&tb
, 0, sizeof(tb
));
3631 dev
= rtnl_create_link(net
, name
, name_assign_type
,
3632 &vxlan_link_ops
, tb
);
3636 err
= __vxlan_dev_create(net
, dev
, conf
, NULL
);
3639 return ERR_PTR(err
);
3642 err
= rtnl_configure_link(dev
, NULL
);
3644 LIST_HEAD(list_kill
);
3646 vxlan_dellink(dev
, &list_kill
);
3647 unregister_netdevice_many(&list_kill
);
3648 return ERR_PTR(err
);
3653 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
3655 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3656 struct net_device
*dev
)
3658 struct vxlan_dev
*vxlan
, *next
;
3659 LIST_HEAD(list_kill
);
3661 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3662 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3664 /* In case we created vxlan device with carrier
3665 * and we loose the carrier due to module unload
3666 * we also need to remove vxlan device. In other
3667 * cases, it's not necessary and remote_ifindex
3668 * is 0 here, so no matches.
3670 if (dst
->remote_ifindex
== dev
->ifindex
)
3671 vxlan_dellink(vxlan
->dev
, &list_kill
);
3674 unregister_netdevice_many(&list_kill
);
3677 static int vxlan_netdevice_event(struct notifier_block
*unused
,
3678 unsigned long event
, void *ptr
)
3680 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3681 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3683 if (event
== NETDEV_UNREGISTER
) {
3684 vxlan_offload_rx_ports(dev
, false);
3685 vxlan_handle_lowerdev_unregister(vn
, dev
);
3686 } else if (event
== NETDEV_REGISTER
) {
3687 vxlan_offload_rx_ports(dev
, true);
3688 } else if (event
== NETDEV_UDP_TUNNEL_PUSH_INFO
||
3689 event
== NETDEV_UDP_TUNNEL_DROP_INFO
) {
3690 vxlan_offload_rx_ports(dev
, event
== NETDEV_UDP_TUNNEL_PUSH_INFO
);
3696 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3697 .notifier_call
= vxlan_netdevice_event
,
3700 static __net_init
int vxlan_init_net(struct net
*net
)
3702 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3705 INIT_LIST_HEAD(&vn
->vxlan_list
);
3706 spin_lock_init(&vn
->sock_lock
);
3708 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3709 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3714 static void __net_exit
vxlan_exit_net(struct net
*net
)
3716 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3717 struct vxlan_dev
*vxlan
, *next
;
3718 struct net_device
*dev
, *aux
;
3723 for_each_netdev_safe(net
, dev
, aux
)
3724 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3725 unregister_netdevice_queue(dev
, &list
);
3727 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3728 /* If vxlan->dev is in the same netns, it has already been added
3729 * to the list by the previous loop.
3731 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3732 gro_cells_destroy(&vxlan
->gro_cells
);
3733 unregister_netdevice_queue(vxlan
->dev
, &list
);
3737 unregister_netdevice_many(&list
);
3740 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3741 WARN_ON_ONCE(!hlist_empty(&vn
->sock_list
[h
]));
3744 static struct pernet_operations vxlan_net_ops
= {
3745 .init
= vxlan_init_net
,
3746 .exit
= vxlan_exit_net
,
3747 .id
= &vxlan_net_id
,
3748 .size
= sizeof(struct vxlan_net
),
3751 static int __init
vxlan_init_module(void)
3755 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3757 rc
= register_pernet_subsys(&vxlan_net_ops
);
3761 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3765 rc
= rtnl_link_register(&vxlan_link_ops
);
3771 unregister_netdevice_notifier(&vxlan_notifier_block
);
3773 unregister_pernet_subsys(&vxlan_net_ops
);
3777 late_initcall(vxlan_init_module
);
3779 static void __exit
vxlan_cleanup_module(void)
3781 rtnl_link_unregister(&vxlan_link_ops
);
3782 unregister_netdevice_notifier(&vxlan_notifier_block
);
3783 unregister_pernet_subsys(&vxlan_net_ops
);
3784 /* rcu_barrier() is called by netns */
3786 module_exit(vxlan_cleanup_module
);
3788 MODULE_LICENSE("GPL");
3789 MODULE_VERSION(VXLAN_VERSION
);
3790 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3791 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3792 MODULE_ALIAS_RTNL_LINK("vxlan");