2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/if_ether.h>
20 #include <linux/ethtool.h>
22 #include <net/ndisc.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/vxlan.h>
31 #if IS_ENABLED(CONFIG_IPV6)
32 #include <net/ip6_tunnel.h>
33 #include <net/ip6_checksum.h>
36 #define VXLAN_VERSION "0.1"
38 #define PORT_HASH_BITS 8
39 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
40 #define FDB_AGE_DEFAULT 300 /* 5 min */
41 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43 /* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
45 * for compatibility with early adopters.
47 static unsigned short vxlan_port __read_mostly
= 8472;
48 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
49 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
51 static bool log_ecn_error
= true;
52 module_param(log_ecn_error
, bool, 0644);
53 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
55 static unsigned int vxlan_net_id
;
56 static struct rtnl_link_ops vxlan_link_ops
;
58 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
60 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
62 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
);
64 /* per-network namespace private data for this module */
66 struct list_head vxlan_list
;
67 struct hlist_head sock_list
[PORT_HASH_SIZE
];
71 /* Forwarding table entry */
73 struct hlist_node hlist
; /* linked list of entries */
75 unsigned long updated
; /* jiffies */
77 struct list_head remotes
;
78 u8 eth_addr
[ETH_ALEN
];
79 u16 state
; /* see ndm_state */
81 u8 flags
; /* see ndm_flags */
84 /* salt for hash table */
85 static u32 vxlan_salt __read_mostly
;
87 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
89 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
90 ip_tunnel_collect_metadata();
93 #if IS_ENABLED(CONFIG_IPV6)
95 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
97 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
99 if (a
->sa
.sa_family
== AF_INET6
)
100 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
102 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
105 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
107 if (ipa
->sa
.sa_family
== AF_INET6
)
108 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
110 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
113 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
115 if (ipa
->sa
.sa_family
== AF_INET6
)
116 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
118 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
121 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
123 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
124 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
125 ip
->sa
.sa_family
= AF_INET6
;
127 } else if (nla_len(nla
) >= sizeof(__be32
)) {
128 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
129 ip
->sa
.sa_family
= AF_INET
;
132 return -EAFNOSUPPORT
;
136 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
137 const union vxlan_addr
*ip
)
139 if (ip
->sa
.sa_family
== AF_INET6
)
140 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
142 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
145 #else /* !CONFIG_IPV6 */
148 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
150 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
153 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
155 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
158 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
160 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
163 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
165 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
166 return -EAFNOSUPPORT
;
167 } else if (nla_len(nla
) >= sizeof(__be32
)) {
168 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
169 ip
->sa
.sa_family
= AF_INET
;
172 return -EAFNOSUPPORT
;
176 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
177 const union vxlan_addr
*ip
)
179 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
183 /* Virtual Network hash table head */
184 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
186 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
189 /* Socket hash table head */
190 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
192 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
194 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
197 /* First remote destination for a forwarding entry.
198 * Guaranteed to be non-NULL because remotes are never deleted.
200 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
202 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
205 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
207 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
210 /* Find VXLAN socket based on network namespace, address family and UDP port
211 * and enabled unshareable flags.
213 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
214 __be16 port
, u32 flags
)
216 struct vxlan_sock
*vs
;
218 flags
&= VXLAN_F_RCV_FLAGS
;
220 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
221 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
222 vxlan_get_sk_family(vs
) == family
&&
229 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, __be32 vni
)
231 struct vxlan_dev
*vxlan
;
233 /* For flow based devices, map all packets to VNI 0 */
234 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
237 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, vni
), hlist
) {
238 if (vxlan
->default_dst
.remote_vni
== vni
)
245 /* Look up VNI in a per net namespace table */
246 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, __be32 vni
,
247 sa_family_t family
, __be16 port
,
250 struct vxlan_sock
*vs
;
252 vs
= vxlan_find_sock(net
, family
, port
, flags
);
256 return vxlan_vs_find_vni(vs
, vni
);
259 /* Fill in neighbour message in skbuff. */
260 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
261 const struct vxlan_fdb
*fdb
,
262 u32 portid
, u32 seq
, int type
, unsigned int flags
,
263 const struct vxlan_rdst
*rdst
)
265 unsigned long now
= jiffies
;
266 struct nda_cacheinfo ci
;
267 struct nlmsghdr
*nlh
;
269 bool send_ip
, send_eth
;
271 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
275 ndm
= nlmsg_data(nlh
);
276 memset(ndm
, 0, sizeof(*ndm
));
278 send_eth
= send_ip
= true;
280 if (type
== RTM_GETNEIGH
) {
281 ndm
->ndm_family
= AF_INET
;
282 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
283 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
285 ndm
->ndm_family
= AF_BRIDGE
;
286 ndm
->ndm_state
= fdb
->state
;
287 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
288 ndm
->ndm_flags
= fdb
->flags
;
289 ndm
->ndm_type
= RTN_UNICAST
;
291 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
292 nla_put_s32(skb
, NDA_LINK_NETNSID
,
293 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
294 goto nla_put_failure
;
296 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
297 goto nla_put_failure
;
299 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
300 goto nla_put_failure
;
302 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
303 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
304 goto nla_put_failure
;
305 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
306 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
307 goto nla_put_failure
;
308 if ((vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) && fdb
->vni
&&
309 nla_put_u32(skb
, NDA_SRC_VNI
,
310 be32_to_cpu(fdb
->vni
)))
311 goto nla_put_failure
;
312 if (rdst
->remote_ifindex
&&
313 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
314 goto nla_put_failure
;
316 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
317 ci
.ndm_confirmed
= 0;
318 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
321 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
322 goto nla_put_failure
;
328 nlmsg_cancel(skb
, nlh
);
332 static inline size_t vxlan_nlmsg_size(void)
334 return NLMSG_ALIGN(sizeof(struct ndmsg
))
335 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
336 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
337 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
338 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
339 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
340 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
341 + nla_total_size(sizeof(struct nda_cacheinfo
));
344 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
345 struct vxlan_rdst
*rd
, int type
)
347 struct net
*net
= dev_net(vxlan
->dev
);
351 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
355 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
357 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
358 WARN_ON(err
== -EMSGSIZE
);
363 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
367 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
370 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
372 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
373 struct vxlan_fdb f
= {
376 struct vxlan_rdst remote
= {
377 .remote_ip
= *ipa
, /* goes to NDA_DST */
378 .remote_vni
= cpu_to_be32(VXLAN_N_VID
),
381 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
384 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
386 struct vxlan_fdb f
= {
389 struct vxlan_rdst remote
= { };
391 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
393 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
396 /* Hash Ethernet address */
397 static u32
eth_hash(const unsigned char *addr
)
399 u64 value
= get_unaligned((u64
*)addr
);
401 /* only want 6 bytes */
407 return hash_64(value
, FDB_HASH_BITS
);
410 static u32
eth_vni_hash(const unsigned char *addr
, __be32 vni
)
412 /* use 1 byte of OUI and 3 bytes of NIC */
413 u32 key
= get_unaligned((u32
*)(addr
+ 2));
415 return jhash_2words(key
, vni
, vxlan_salt
) & (FDB_HASH_SIZE
- 1);
418 /* Hash chain to use given mac address */
419 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
420 const u8
*mac
, __be32 vni
)
422 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)
423 return &vxlan
->fdb_head
[eth_vni_hash(mac
, vni
)];
425 return &vxlan
->fdb_head
[eth_hash(mac
)];
428 /* Look up Ethernet address in forwarding table */
429 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
430 const u8
*mac
, __be32 vni
)
432 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
, vni
);
435 hlist_for_each_entry_rcu(f
, head
, hlist
) {
436 if (ether_addr_equal(mac
, f
->eth_addr
)) {
437 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
449 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
450 const u8
*mac
, __be32 vni
)
454 f
= __vxlan_find_mac(vxlan
, mac
, vni
);
461 /* caller should hold vxlan->hash_lock */
462 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
463 union vxlan_addr
*ip
, __be16 port
,
464 __be32 vni
, __u32 ifindex
)
466 struct vxlan_rdst
*rd
;
468 list_for_each_entry(rd
, &f
->remotes
, list
) {
469 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
470 rd
->remote_port
== port
&&
471 rd
->remote_vni
== vni
&&
472 rd
->remote_ifindex
== ifindex
)
479 /* Replace destination of unicast mac */
480 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
481 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
484 struct vxlan_rdst
*rd
;
486 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
490 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
494 dst_cache_reset(&rd
->dst_cache
);
496 rd
->remote_port
= port
;
497 rd
->remote_vni
= vni
;
498 rd
->remote_ifindex
= ifindex
;
502 /* Add/update destinations for multicast */
503 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
504 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
505 __u32 ifindex
, struct vxlan_rdst
**rdp
)
507 struct vxlan_rdst
*rd
;
509 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
513 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
517 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
523 rd
->remote_port
= port
;
524 rd
->remote_vni
= vni
;
525 rd
->remote_ifindex
= ifindex
;
527 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
533 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
535 struct vxlanhdr
*vh
, size_t hdrlen
,
537 struct gro_remcsum
*grc
,
540 size_t start
, offset
;
542 if (skb
->remcsum_offload
)
545 if (!NAPI_GRO_CB(skb
)->csum_valid
)
548 start
= vxlan_rco_start(vni_field
);
549 offset
= start
+ vxlan_rco_offset(vni_field
);
551 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
552 start
, offset
, grc
, nopartial
);
554 skb
->remcsum_offload
= 1;
559 static struct sk_buff
**vxlan_gro_receive(struct sock
*sk
,
560 struct sk_buff
**head
,
563 struct sk_buff
*p
, **pp
= NULL
;
564 struct vxlanhdr
*vh
, *vh2
;
565 unsigned int hlen
, off_vx
;
567 struct vxlan_sock
*vs
= rcu_dereference_sk_user_data(sk
);
569 struct gro_remcsum grc
;
571 skb_gro_remcsum_init(&grc
);
573 off_vx
= skb_gro_offset(skb
);
574 hlen
= off_vx
+ sizeof(*vh
);
575 vh
= skb_gro_header_fast(skb
, off_vx
);
576 if (skb_gro_header_hard(skb
, hlen
)) {
577 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
582 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
584 flags
= vh
->vx_flags
;
586 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
587 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
590 VXLAN_F_REMCSUM_NOPARTIAL
));
596 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
598 for (p
= *head
; p
; p
= p
->next
) {
599 if (!NAPI_GRO_CB(p
)->same_flow
)
602 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
603 if (vh
->vx_flags
!= vh2
->vx_flags
||
604 vh
->vx_vni
!= vh2
->vx_vni
) {
605 NAPI_GRO_CB(p
)->same_flow
= 0;
610 pp
= call_gro_receive(eth_gro_receive
, head
, skb
);
614 skb_gro_remcsum_cleanup(skb
, &grc
);
615 NAPI_GRO_CB(skb
)->flush
|= flush
;
620 static int vxlan_gro_complete(struct sock
*sk
, struct sk_buff
*skb
, int nhoff
)
622 /* Sets 'skb->inner_mac_header' since we are always called with
623 * 'skb->encapsulation' set.
625 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
628 /* Add new entry to forwarding table -- assumes lock held */
629 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
630 const u8
*mac
, union vxlan_addr
*ip
,
631 __u16 state
, __u16 flags
,
632 __be16 port
, __be32 src_vni
, __be32 vni
,
633 __u32 ifindex
, __u8 ndm_flags
)
635 struct vxlan_rdst
*rd
= NULL
;
640 f
= __vxlan_find_mac(vxlan
, mac
, src_vni
);
642 if (flags
& NLM_F_EXCL
) {
643 netdev_dbg(vxlan
->dev
,
644 "lost race to create %pM\n", mac
);
647 if (f
->state
!= state
) {
649 f
->updated
= jiffies
;
652 if (f
->flags
!= ndm_flags
) {
653 f
->flags
= ndm_flags
;
654 f
->updated
= jiffies
;
657 if ((flags
& NLM_F_REPLACE
)) {
658 /* Only change unicasts */
659 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
660 is_zero_ether_addr(f
->eth_addr
))) {
661 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
666 if ((flags
& NLM_F_APPEND
) &&
667 (is_multicast_ether_addr(f
->eth_addr
) ||
668 is_zero_ether_addr(f
->eth_addr
))) {
669 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
676 if (!(flags
& NLM_F_CREATE
))
679 if (vxlan
->cfg
.addrmax
&&
680 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
683 /* Disallow replace to add a multicast entry */
684 if ((flags
& NLM_F_REPLACE
) &&
685 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
688 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
689 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
695 f
->flags
= ndm_flags
;
696 f
->updated
= f
->used
= jiffies
;
698 INIT_LIST_HEAD(&f
->remotes
);
699 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
701 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
708 hlist_add_head_rcu(&f
->hlist
,
709 vxlan_fdb_head(vxlan
, mac
, src_vni
));
714 rd
= first_remote_rtnl(f
);
715 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
721 static void vxlan_fdb_free(struct rcu_head
*head
)
723 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
724 struct vxlan_rdst
*rd
, *nd
;
726 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
727 dst_cache_destroy(&rd
->dst_cache
);
733 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
735 netdev_dbg(vxlan
->dev
,
736 "delete %pM\n", f
->eth_addr
);
739 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
741 hlist_del_rcu(&f
->hlist
);
742 call_rcu(&f
->rcu
, vxlan_fdb_free
);
745 static void vxlan_dst_free(struct rcu_head
*head
)
747 struct vxlan_rdst
*rd
= container_of(head
, struct vxlan_rdst
, rcu
);
749 dst_cache_destroy(&rd
->dst_cache
);
753 static void vxlan_fdb_dst_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
754 struct vxlan_rdst
*rd
)
756 list_del_rcu(&rd
->list
);
757 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
758 call_rcu(&rd
->rcu
, vxlan_dst_free
);
761 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
762 union vxlan_addr
*ip
, __be16
*port
, __be32
*src_vni
,
763 __be32
*vni
, u32
*ifindex
)
765 struct net
*net
= dev_net(vxlan
->dev
);
769 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
773 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
774 if (remote
->sa
.sa_family
== AF_INET
) {
775 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
776 ip
->sa
.sa_family
= AF_INET
;
777 #if IS_ENABLED(CONFIG_IPV6)
779 ip
->sin6
.sin6_addr
= in6addr_any
;
780 ip
->sa
.sa_family
= AF_INET6
;
786 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
788 *port
= nla_get_be16(tb
[NDA_PORT
]);
790 *port
= vxlan
->cfg
.dst_port
;
794 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
796 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
798 *vni
= vxlan
->default_dst
.remote_vni
;
801 if (tb
[NDA_SRC_VNI
]) {
802 if (nla_len(tb
[NDA_SRC_VNI
]) != sizeof(u32
))
804 *src_vni
= cpu_to_be32(nla_get_u32(tb
[NDA_SRC_VNI
]));
806 *src_vni
= vxlan
->default_dst
.remote_vni
;
809 if (tb
[NDA_IFINDEX
]) {
810 struct net_device
*tdev
;
812 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
814 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
815 tdev
= __dev_get_by_index(net
, *ifindex
);
817 return -EADDRNOTAVAIL
;
825 /* Add static entry (via netlink) */
826 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
827 struct net_device
*dev
,
828 const unsigned char *addr
, u16 vid
, u16 flags
)
830 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
831 /* struct net *net = dev_net(vxlan->dev); */
838 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
839 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
844 if (tb
[NDA_DST
] == NULL
)
847 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
851 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
852 return -EAFNOSUPPORT
;
854 spin_lock_bh(&vxlan
->hash_lock
);
855 err
= vxlan_fdb_create(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
856 port
, src_vni
, vni
, ifindex
, ndm
->ndm_flags
);
857 spin_unlock_bh(&vxlan
->hash_lock
);
862 static int __vxlan_fdb_delete(struct vxlan_dev
*vxlan
,
863 const unsigned char *addr
, union vxlan_addr ip
,
864 __be16 port
, __be32 src_vni
, u32 vni
, u32 ifindex
,
868 struct vxlan_rdst
*rd
= NULL
;
871 f
= vxlan_find_mac(vxlan
, addr
, src_vni
);
875 if (!vxlan_addr_any(&ip
)) {
876 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
881 /* remove a destination if it's not the only one on the list,
882 * otherwise destroy the fdb entry
884 if (rd
&& !list_is_singular(&f
->remotes
)) {
885 vxlan_fdb_dst_destroy(vxlan
, f
, rd
);
889 vxlan_fdb_destroy(vxlan
, f
);
895 /* Delete entry (via netlink) */
896 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
897 struct net_device
*dev
,
898 const unsigned char *addr
, u16 vid
)
900 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
907 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
911 spin_lock_bh(&vxlan
->hash_lock
);
912 err
= __vxlan_fdb_delete(vxlan
, addr
, ip
, port
, src_vni
, vni
, ifindex
,
914 spin_unlock_bh(&vxlan
->hash_lock
);
919 /* Dump forwarding table */
920 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
921 struct net_device
*dev
,
922 struct net_device
*filter_dev
, int *idx
)
924 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
928 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
931 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
932 struct vxlan_rdst
*rd
;
934 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
935 if (*idx
< cb
->args
[2])
938 err
= vxlan_fdb_info(skb
, vxlan
, f
,
939 NETLINK_CB(cb
->skb
).portid
,
954 /* Watch incoming packets to learn mapping between Ethernet address
955 * and Tunnel endpoint.
956 * Return true if packet is bogus and should be dropped.
958 static bool vxlan_snoop(struct net_device
*dev
,
959 union vxlan_addr
*src_ip
, const u8
*src_mac
,
962 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
965 f
= vxlan_find_mac(vxlan
, src_mac
, vni
);
967 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
969 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
)))
972 /* Don't migrate static entries, drop packets */
973 if (f
->state
& NUD_NOARP
)
978 "%pM migrated from %pIS to %pIS\n",
979 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
981 rdst
->remote_ip
= *src_ip
;
982 f
->updated
= jiffies
;
983 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
985 /* learned new entry */
986 spin_lock(&vxlan
->hash_lock
);
988 /* close off race between vxlan_flush and incoming packets */
989 if (netif_running(dev
))
990 vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
992 NLM_F_EXCL
|NLM_F_CREATE
,
995 vxlan
->default_dst
.remote_vni
,
997 spin_unlock(&vxlan
->hash_lock
);
1003 /* See if multicast group is already in use by other ID */
1004 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1006 struct vxlan_dev
*vxlan
;
1007 struct vxlan_sock
*sock4
;
1008 #if IS_ENABLED(CONFIG_IPV6)
1009 struct vxlan_sock
*sock6
;
1011 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1013 sock4
= rtnl_dereference(dev
->vn4_sock
);
1015 /* The vxlan_sock is only used by dev, leaving group has
1016 * no effect on other vxlan devices.
1018 if (family
== AF_INET
&& sock4
&& atomic_read(&sock4
->refcnt
) == 1)
1020 #if IS_ENABLED(CONFIG_IPV6)
1021 sock6
= rtnl_dereference(dev
->vn6_sock
);
1022 if (family
== AF_INET6
&& sock6
&& atomic_read(&sock6
->refcnt
) == 1)
1026 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1027 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1030 if (family
== AF_INET
&&
1031 rtnl_dereference(vxlan
->vn4_sock
) != sock4
)
1033 #if IS_ENABLED(CONFIG_IPV6)
1034 if (family
== AF_INET6
&&
1035 rtnl_dereference(vxlan
->vn6_sock
) != sock6
)
1039 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1040 &dev
->default_dst
.remote_ip
))
1043 if (vxlan
->default_dst
.remote_ifindex
!=
1044 dev
->default_dst
.remote_ifindex
)
1053 static bool __vxlan_sock_release_prep(struct vxlan_sock
*vs
)
1055 struct vxlan_net
*vn
;
1059 if (!atomic_dec_and_test(&vs
->refcnt
))
1062 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1063 spin_lock(&vn
->sock_lock
);
1064 hlist_del_rcu(&vs
->hlist
);
1065 udp_tunnel_notify_del_rx_port(vs
->sock
,
1066 (vs
->flags
& VXLAN_F_GPE
) ?
1067 UDP_TUNNEL_TYPE_VXLAN_GPE
:
1068 UDP_TUNNEL_TYPE_VXLAN
);
1069 spin_unlock(&vn
->sock_lock
);
1074 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1076 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1077 #if IS_ENABLED(CONFIG_IPV6)
1078 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1080 rcu_assign_pointer(vxlan
->vn6_sock
, NULL
);
1083 rcu_assign_pointer(vxlan
->vn4_sock
, NULL
);
1086 vxlan_vs_del_dev(vxlan
);
1088 if (__vxlan_sock_release_prep(sock4
)) {
1089 udp_tunnel_sock_release(sock4
->sock
);
1093 #if IS_ENABLED(CONFIG_IPV6)
1094 if (__vxlan_sock_release_prep(sock6
)) {
1095 udp_tunnel_sock_release(sock6
->sock
);
1101 /* Update multicast group membership when first VNI on
1102 * multicast address is brought up
1104 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1107 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1108 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1111 if (ip
->sa
.sa_family
== AF_INET
) {
1112 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1113 struct ip_mreqn mreq
= {
1114 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1115 .imr_ifindex
= ifindex
,
1118 sk
= sock4
->sock
->sk
;
1120 ret
= ip_mc_join_group(sk
, &mreq
);
1122 #if IS_ENABLED(CONFIG_IPV6)
1124 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1126 sk
= sock6
->sock
->sk
;
1128 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1129 &ip
->sin6
.sin6_addr
);
1137 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1138 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1141 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1142 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1145 if (ip
->sa
.sa_family
== AF_INET
) {
1146 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1147 struct ip_mreqn mreq
= {
1148 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1149 .imr_ifindex
= ifindex
,
1152 sk
= sock4
->sock
->sk
;
1154 ret
= ip_mc_leave_group(sk
, &mreq
);
1156 #if IS_ENABLED(CONFIG_IPV6)
1158 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1160 sk
= sock6
->sock
->sk
;
1162 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1163 &ip
->sin6
.sin6_addr
);
1171 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1172 struct sk_buff
*skb
, u32 vxflags
)
1174 size_t start
, offset
;
1176 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1179 start
= vxlan_rco_start(unparsed
->vx_vni
);
1180 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1182 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1185 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1186 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1188 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1189 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1193 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1194 struct sk_buff
*skb
, u32 vxflags
,
1195 struct vxlan_metadata
*md
)
1197 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1198 struct metadata_dst
*tun_dst
;
1200 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1203 md
->gbp
= ntohs(gbp
->policy_id
);
1205 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1207 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1208 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1210 if (gbp
->dont_learn
)
1211 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1213 if (gbp
->policy_applied
)
1214 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1216 /* In flow-based mode, GBP is carried in dst_metadata */
1217 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1218 skb
->mark
= md
->gbp
;
1220 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1223 static bool vxlan_parse_gpe_hdr(struct vxlanhdr
*unparsed
,
1225 struct sk_buff
*skb
, u32 vxflags
)
1227 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)unparsed
;
1229 /* Need to have Next Protocol set for interfaces in GPE mode. */
1230 if (!gpe
->np_applied
)
1232 /* "The initial version is 0. If a receiver does not support the
1233 * version indicated it MUST drop the packet.
1235 if (gpe
->version
!= 0)
1237 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1238 * processing MUST occur." However, we don't implement OAM
1239 * processing, thus drop the packet.
1244 switch (gpe
->next_protocol
) {
1245 case VXLAN_GPE_NP_IPV4
:
1246 *protocol
= htons(ETH_P_IP
);
1248 case VXLAN_GPE_NP_IPV6
:
1249 *protocol
= htons(ETH_P_IPV6
);
1251 case VXLAN_GPE_NP_ETHERNET
:
1252 *protocol
= htons(ETH_P_TEB
);
1258 unparsed
->vx_flags
&= ~VXLAN_GPE_USED_BITS
;
1262 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1263 struct vxlan_sock
*vs
,
1264 struct sk_buff
*skb
, __be32 vni
)
1266 union vxlan_addr saddr
;
1268 skb_reset_mac_header(skb
);
1269 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1270 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1272 /* Ignore packet loops (and multicast echo) */
1273 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1276 /* Get address from the outer IP header */
1277 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1278 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1279 saddr
.sa
.sa_family
= AF_INET
;
1280 #if IS_ENABLED(CONFIG_IPV6)
1282 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1283 saddr
.sa
.sa_family
= AF_INET6
;
1287 if ((vxlan
->flags
& VXLAN_F_LEARN
) &&
1288 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
, vni
))
1294 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1295 struct sk_buff
*skb
)
1299 if (vxlan_get_sk_family(vs
) == AF_INET
)
1300 err
= IP_ECN_decapsulate(oiph
, skb
);
1301 #if IS_ENABLED(CONFIG_IPV6)
1303 err
= IP6_ECN_decapsulate(oiph
, skb
);
1306 if (unlikely(err
) && log_ecn_error
) {
1307 if (vxlan_get_sk_family(vs
) == AF_INET
)
1308 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1309 &((struct iphdr
*)oiph
)->saddr
,
1310 ((struct iphdr
*)oiph
)->tos
);
1312 net_info_ratelimited("non-ECT from %pI6\n",
1313 &((struct ipv6hdr
*)oiph
)->saddr
);
1318 /* Callback from net/ipv4/udp.c to receive packets */
1319 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1321 struct pcpu_sw_netstats
*stats
;
1322 struct vxlan_dev
*vxlan
;
1323 struct vxlan_sock
*vs
;
1324 struct vxlanhdr unparsed
;
1325 struct vxlan_metadata _md
;
1326 struct vxlan_metadata
*md
= &_md
;
1327 __be16 protocol
= htons(ETH_P_TEB
);
1328 bool raw_proto
= false;
1332 /* Need UDP and VXLAN header to be present */
1333 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1336 unparsed
= *vxlan_hdr(skb
);
1337 /* VNI flag always required to be set */
1338 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1339 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1340 ntohl(vxlan_hdr(skb
)->vx_flags
),
1341 ntohl(vxlan_hdr(skb
)->vx_vni
));
1342 /* Return non vxlan pkt */
1345 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1346 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1348 vs
= rcu_dereference_sk_user_data(sk
);
1352 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1354 vxlan
= vxlan_vs_find_vni(vs
, vni
);
1358 /* For backwards compatibility, only allow reserved fields to be
1359 * used by VXLAN extensions if explicitly requested.
1361 if (vs
->flags
& VXLAN_F_GPE
) {
1362 if (!vxlan_parse_gpe_hdr(&unparsed
, &protocol
, skb
, vs
->flags
))
1367 if (__iptunnel_pull_header(skb
, VXLAN_HLEN
, protocol
, raw_proto
,
1368 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1371 if (vxlan_collect_metadata(vs
)) {
1372 struct metadata_dst
*tun_dst
;
1374 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1375 key32_to_tunnel_id(vni
), sizeof(*md
));
1380 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1382 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1384 memset(md
, 0, sizeof(*md
));
1387 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1388 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1390 if (vs
->flags
& VXLAN_F_GBP
)
1391 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1392 /* Note that GBP and GPE can never be active together. This is
1393 * ensured in vxlan_dev_configure.
1396 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1397 /* If there are any unprocessed flags remaining treat
1398 * this as a malformed packet. This behavior diverges from
1399 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1400 * in reserved fields are to be ignored. The approach here
1401 * maintains compatibility with previous stack code, and also
1402 * is more robust and provides a little more security in
1403 * adding extensions to VXLAN.
1409 if (!vxlan_set_mac(vxlan
, vs
, skb
, vni
))
1412 skb_reset_mac_header(skb
);
1413 skb
->dev
= vxlan
->dev
;
1414 skb
->pkt_type
= PACKET_HOST
;
1417 oiph
= skb_network_header(skb
);
1418 skb_reset_network_header(skb
);
1420 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1421 ++vxlan
->dev
->stats
.rx_frame_errors
;
1422 ++vxlan
->dev
->stats
.rx_errors
;
1426 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1427 u64_stats_update_begin(&stats
->syncp
);
1428 stats
->rx_packets
++;
1429 stats
->rx_bytes
+= skb
->len
;
1430 u64_stats_update_end(&stats
->syncp
);
1432 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1436 /* Consume bad packet */
1441 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1443 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1444 struct arphdr
*parp
;
1447 struct neighbour
*n
;
1449 if (dev
->flags
& IFF_NOARP
)
1452 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1453 dev
->stats
.tx_dropped
++;
1456 parp
= arp_hdr(skb
);
1458 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1459 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1460 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1461 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1462 parp
->ar_hln
!= dev
->addr_len
||
1465 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1467 arpptr
+= dev
->addr_len
; /* sha */
1468 memcpy(&sip
, arpptr
, sizeof(sip
));
1469 arpptr
+= sizeof(sip
);
1470 arpptr
+= dev
->addr_len
; /* tha */
1471 memcpy(&tip
, arpptr
, sizeof(tip
));
1473 if (ipv4_is_loopback(tip
) ||
1474 ipv4_is_multicast(tip
))
1477 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1480 struct vxlan_fdb
*f
;
1481 struct sk_buff
*reply
;
1483 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1488 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1489 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1490 /* bridge-local neighbor */
1495 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1503 skb_reset_mac_header(reply
);
1504 __skb_pull(reply
, skb_network_offset(reply
));
1505 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1506 reply
->pkt_type
= PACKET_HOST
;
1508 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1509 dev
->stats
.rx_dropped
++;
1510 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1511 union vxlan_addr ipa
= {
1512 .sin
.sin_addr
.s_addr
= tip
,
1513 .sin
.sin_family
= AF_INET
,
1516 vxlan_ip_miss(dev
, &ipa
);
1520 return NETDEV_TX_OK
;
1523 #if IS_ENABLED(CONFIG_IPV6)
1524 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1525 struct neighbour
*n
, bool isrouter
)
1527 struct net_device
*dev
= request
->dev
;
1528 struct sk_buff
*reply
;
1529 struct nd_msg
*ns
, *na
;
1530 struct ipv6hdr
*pip6
;
1532 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1539 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1540 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1541 reply
= alloc_skb(len
, GFP_ATOMIC
);
1545 reply
->protocol
= htons(ETH_P_IPV6
);
1547 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1548 skb_push(reply
, sizeof(struct ethhdr
));
1549 skb_reset_mac_header(reply
);
1551 ns
= (struct nd_msg
*)skb_transport_header(request
);
1553 daddr
= eth_hdr(request
)->h_source
;
1554 ns_olen
= request
->len
- skb_transport_offset(request
) - sizeof(*ns
);
1555 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1556 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1557 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1562 /* Ethernet header */
1563 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1564 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1565 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1566 reply
->protocol
= htons(ETH_P_IPV6
);
1568 skb_pull(reply
, sizeof(struct ethhdr
));
1569 skb_reset_network_header(reply
);
1570 skb_put(reply
, sizeof(struct ipv6hdr
));
1574 pip6
= ipv6_hdr(reply
);
1575 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1577 pip6
->priority
= ipv6_hdr(request
)->priority
;
1578 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1579 pip6
->hop_limit
= 255;
1580 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1581 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1583 skb_pull(reply
, sizeof(struct ipv6hdr
));
1584 skb_reset_transport_header(reply
);
1586 na
= (struct nd_msg
*)skb_put(reply
, sizeof(*na
) + na_olen
);
1588 /* Neighbor Advertisement */
1589 memset(na
, 0, sizeof(*na
)+na_olen
);
1590 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1591 na
->icmph
.icmp6_router
= isrouter
;
1592 na
->icmph
.icmp6_override
= 1;
1593 na
->icmph
.icmp6_solicited
= 1;
1594 na
->target
= ns
->target
;
1595 ether_addr_copy(&na
->opt
[2], n
->ha
);
1596 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1597 na
->opt
[1] = na_olen
>> 3;
1599 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1600 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1601 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1603 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1605 skb_push(reply
, sizeof(struct ipv6hdr
));
1607 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1612 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1614 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1616 const struct ipv6hdr
*iphdr
;
1617 const struct in6_addr
*daddr
;
1618 struct neighbour
*n
;
1619 struct inet6_dev
*in6_dev
;
1621 in6_dev
= __in6_dev_get(dev
);
1625 iphdr
= ipv6_hdr(skb
);
1626 daddr
= &iphdr
->daddr
;
1628 msg
= (struct nd_msg
*)skb_transport_header(skb
);
1629 if (msg
->icmph
.icmp6_code
!= 0 ||
1630 msg
->icmph
.icmp6_type
!= NDISC_NEIGHBOUR_SOLICITATION
)
1633 if (ipv6_addr_loopback(daddr
) ||
1634 ipv6_addr_is_multicast(&msg
->target
))
1637 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1640 struct vxlan_fdb
*f
;
1641 struct sk_buff
*reply
;
1643 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1648 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1649 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1650 /* bridge-local neighbor */
1655 reply
= vxlan_na_create(skb
, n
,
1656 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1663 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1664 dev
->stats
.rx_dropped
++;
1666 } else if (vxlan
->flags
& VXLAN_F_L3MISS
) {
1667 union vxlan_addr ipa
= {
1668 .sin6
.sin6_addr
= msg
->target
,
1669 .sin6
.sin6_family
= AF_INET6
,
1672 vxlan_ip_miss(dev
, &ipa
);
1677 return NETDEV_TX_OK
;
1681 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1683 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1684 struct neighbour
*n
;
1686 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1690 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1695 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1698 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1699 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1700 union vxlan_addr ipa
= {
1701 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1702 .sin
.sin_family
= AF_INET
,
1705 vxlan_ip_miss(dev
, &ipa
);
1711 #if IS_ENABLED(CONFIG_IPV6)
1714 struct ipv6hdr
*pip6
;
1716 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1718 pip6
= ipv6_hdr(skb
);
1719 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1720 if (!n
&& (vxlan
->flags
& VXLAN_F_L3MISS
)) {
1721 union vxlan_addr ipa
= {
1722 .sin6
.sin6_addr
= pip6
->daddr
,
1723 .sin6
.sin6_family
= AF_INET6
,
1726 vxlan_ip_miss(dev
, &ipa
);
1740 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1742 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1744 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1753 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1754 struct vxlan_metadata
*md
)
1756 struct vxlanhdr_gbp
*gbp
;
1761 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1762 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1764 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1765 gbp
->dont_learn
= 1;
1767 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1768 gbp
->policy_applied
= 1;
1770 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1773 static int vxlan_build_gpe_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1776 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)vxh
;
1778 gpe
->np_applied
= 1;
1781 case htons(ETH_P_IP
):
1782 gpe
->next_protocol
= VXLAN_GPE_NP_IPV4
;
1784 case htons(ETH_P_IPV6
):
1785 gpe
->next_protocol
= VXLAN_GPE_NP_IPV6
;
1787 case htons(ETH_P_TEB
):
1788 gpe
->next_protocol
= VXLAN_GPE_NP_ETHERNET
;
1791 return -EPFNOSUPPORT
;
1794 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1795 int iphdr_len
, __be32 vni
,
1796 struct vxlan_metadata
*md
, u32 vxflags
,
1799 struct vxlanhdr
*vxh
;
1802 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
1803 __be16 inner_protocol
= htons(ETH_P_TEB
);
1805 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
1806 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1807 int csum_start
= skb_checksum_start_offset(skb
);
1809 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
1810 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
1811 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
1812 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
1813 type
|= SKB_GSO_TUNNEL_REMCSUM
;
1816 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
1817 + VXLAN_HLEN
+ iphdr_len
;
1819 /* Need space for new headers (invalidates iph ptr) */
1820 err
= skb_cow_head(skb
, min_headroom
);
1824 err
= iptunnel_handle_offloads(skb
, type
);
1828 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1829 vxh
->vx_flags
= VXLAN_HF_VNI
;
1830 vxh
->vx_vni
= vxlan_vni_field(vni
);
1832 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
1835 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
1836 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
1837 vxh
->vx_flags
|= VXLAN_HF_RCO
;
1839 if (!skb_is_gso(skb
)) {
1840 skb
->ip_summed
= CHECKSUM_NONE
;
1841 skb
->encapsulation
= 0;
1845 if (vxflags
& VXLAN_F_GBP
)
1846 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
1847 if (vxflags
& VXLAN_F_GPE
) {
1848 err
= vxlan_build_gpe_hdr(vxh
, vxflags
, skb
->protocol
);
1851 inner_protocol
= skb
->protocol
;
1854 skb_set_inner_protocol(skb
, inner_protocol
);
1858 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
, struct net_device
*dev
,
1859 struct vxlan_sock
*sock4
,
1860 struct sk_buff
*skb
, int oif
, u8 tos
,
1861 __be32 daddr
, __be32
*saddr
, __be16 dport
, __be16 sport
,
1862 struct dst_cache
*dst_cache
,
1863 const struct ip_tunnel_info
*info
)
1865 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1866 struct rtable
*rt
= NULL
;
1870 return ERR_PTR(-EIO
);
1875 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
1880 memset(&fl4
, 0, sizeof(fl4
));
1881 fl4
.flowi4_oif
= oif
;
1882 fl4
.flowi4_tos
= RT_TOS(tos
);
1883 fl4
.flowi4_mark
= skb
->mark
;
1884 fl4
.flowi4_proto
= IPPROTO_UDP
;
1887 fl4
.fl4_dport
= dport
;
1888 fl4
.fl4_sport
= sport
;
1890 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
1891 if (likely(!IS_ERR(rt
))) {
1892 if (rt
->dst
.dev
== dev
) {
1893 netdev_dbg(dev
, "circular route to %pI4\n", &daddr
);
1895 return ERR_PTR(-ELOOP
);
1900 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
1902 netdev_dbg(dev
, "no route to %pI4\n", &daddr
);
1903 return ERR_PTR(-ENETUNREACH
);
1908 #if IS_ENABLED(CONFIG_IPV6)
1909 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
1910 struct net_device
*dev
,
1911 struct vxlan_sock
*sock6
,
1912 struct sk_buff
*skb
, int oif
, u8 tos
,
1914 const struct in6_addr
*daddr
,
1915 struct in6_addr
*saddr
,
1916 __be16 dport
, __be16 sport
,
1917 struct dst_cache
*dst_cache
,
1918 const struct ip_tunnel_info
*info
)
1920 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
1921 struct dst_entry
*ndst
;
1926 return ERR_PTR(-EIO
);
1931 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
1936 memset(&fl6
, 0, sizeof(fl6
));
1937 fl6
.flowi6_oif
= oif
;
1940 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
1941 fl6
.flowi6_mark
= skb
->mark
;
1942 fl6
.flowi6_proto
= IPPROTO_UDP
;
1943 fl6
.fl6_dport
= dport
;
1944 fl6
.fl6_sport
= sport
;
1946 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
1949 if (unlikely(err
< 0)) {
1950 netdev_dbg(dev
, "no route to %pI6\n", daddr
);
1951 return ERR_PTR(-ENETUNREACH
);
1954 if (unlikely(ndst
->dev
== dev
)) {
1955 netdev_dbg(dev
, "circular route to %pI6\n", daddr
);
1957 return ERR_PTR(-ELOOP
);
1962 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
1967 /* Bypass encapsulation if the destination is local */
1968 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
1969 struct vxlan_dev
*dst_vxlan
, __be32 vni
)
1971 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
1972 union vxlan_addr loopback
;
1973 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
1974 struct net_device
*dev
= skb
->dev
;
1977 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
1978 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
1979 skb
->pkt_type
= PACKET_HOST
;
1980 skb
->encapsulation
= 0;
1981 skb
->dev
= dst_vxlan
->dev
;
1982 __skb_pull(skb
, skb_network_offset(skb
));
1984 if (remote_ip
->sa
.sa_family
== AF_INET
) {
1985 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
1986 loopback
.sa
.sa_family
= AF_INET
;
1987 #if IS_ENABLED(CONFIG_IPV6)
1989 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
1990 loopback
.sa
.sa_family
= AF_INET6
;
1994 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
1995 vxlan_snoop(skb
->dev
, &loopback
, eth_hdr(skb
)->h_source
, vni
);
1997 u64_stats_update_begin(&tx_stats
->syncp
);
1998 tx_stats
->tx_packets
++;
1999 tx_stats
->tx_bytes
+= len
;
2000 u64_stats_update_end(&tx_stats
->syncp
);
2002 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
2003 u64_stats_update_begin(&rx_stats
->syncp
);
2004 rx_stats
->rx_packets
++;
2005 rx_stats
->rx_bytes
+= len
;
2006 u64_stats_update_end(&rx_stats
->syncp
);
2008 dev
->stats
.rx_dropped
++;
2012 static int encap_bypass_if_local(struct sk_buff
*skb
, struct net_device
*dev
,
2013 struct vxlan_dev
*vxlan
, union vxlan_addr
*daddr
,
2014 __be16 dst_port
, __be32 vni
, struct dst_entry
*dst
,
2017 #if IS_ENABLED(CONFIG_IPV6)
2018 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2019 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2020 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2022 BUILD_BUG_ON(RTCF_LOCAL
!= RTF_LOCAL
);
2024 /* Bypass encapsulation if the destination is local */
2025 if (rt_flags
& RTCF_LOCAL
&&
2026 !(rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2027 struct vxlan_dev
*dst_vxlan
;
2030 dst_vxlan
= vxlan_find_vni(vxlan
->net
, vni
,
2031 daddr
->sa
.sa_family
, dst_port
,
2034 dev
->stats
.tx_errors
++;
2039 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
, vni
);
2046 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
2047 __be32 default_vni
, struct vxlan_rdst
*rdst
,
2050 struct dst_cache
*dst_cache
;
2051 struct ip_tunnel_info
*info
;
2052 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2053 const struct iphdr
*old_iph
= ip_hdr(skb
);
2054 union vxlan_addr
*dst
;
2055 union vxlan_addr remote_ip
, local_ip
;
2056 struct vxlan_metadata _md
;
2057 struct vxlan_metadata
*md
= &_md
;
2058 __be16 src_port
= 0, dst_port
;
2059 struct dst_entry
*ndst
= NULL
;
2063 u32 flags
= vxlan
->flags
;
2064 bool udp_sum
= false;
2065 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
2067 info
= skb_tunnel_info(skb
);
2070 dst
= &rdst
->remote_ip
;
2071 if (vxlan_addr_any(dst
)) {
2073 /* short-circuited back to local bridge */
2074 vxlan_encap_bypass(skb
, vxlan
, vxlan
, default_vni
);
2080 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
2081 vni
= (rdst
->remote_vni
) ? : default_vni
;
2082 local_ip
= vxlan
->cfg
.saddr
;
2083 dst_cache
= &rdst
->dst_cache
;
2084 md
->gbp
= skb
->mark
;
2085 ttl
= vxlan
->cfg
.ttl
;
2086 if (!ttl
&& vxlan_addr_multicast(dst
))
2089 tos
= vxlan
->cfg
.tos
;
2091 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
2093 if (dst
->sa
.sa_family
== AF_INET
)
2094 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
2096 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2097 label
= vxlan
->cfg
.label
;
2100 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2104 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
2105 if (remote_ip
.sa
.sa_family
== AF_INET
) {
2106 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
2107 local_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.src
;
2109 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
2110 local_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.src
;
2113 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2114 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2115 dst_cache
= &info
->dst_cache
;
2116 if (info
->options_len
)
2117 md
= ip_tunnel_info_opts(info
);
2118 ttl
= info
->key
.ttl
;
2119 tos
= info
->key
.tos
;
2120 label
= info
->key
.label
;
2121 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
2123 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2124 vxlan
->cfg
.port_max
, true);
2127 if (dst
->sa
.sa_family
== AF_INET
) {
2128 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2132 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
,
2133 rdst
? rdst
->remote_ifindex
: 0, tos
,
2134 dst
->sin
.sin_addr
.s_addr
,
2135 &local_ip
.sin
.sin_addr
.s_addr
,
2143 /* Bypass encapsulation if the destination is local */
2145 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2146 dst_port
, vni
, &rt
->dst
,
2150 } else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
) {
2155 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2156 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2157 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct iphdr
),
2158 vni
, md
, flags
, udp_sum
);
2162 udp_tunnel_xmit_skb(rt
, sock4
->sock
->sk
, skb
, local_ip
.sin
.sin_addr
.s_addr
,
2163 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2164 src_port
, dst_port
, xnet
, !udp_sum
);
2165 #if IS_ENABLED(CONFIG_IPV6)
2167 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2169 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
,
2170 rdst
? rdst
->remote_ifindex
: 0, tos
,
2171 label
, &dst
->sin6
.sin6_addr
,
2172 &local_ip
.sin6
.sin6_addr
,
2176 err
= PTR_ERR(ndst
);
2182 u32 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2184 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2185 dst_port
, vni
, ndst
,
2191 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2192 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2193 skb_scrub_packet(skb
, xnet
);
2194 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2195 vni
, md
, flags
, udp_sum
);
2199 udp_tunnel6_xmit_skb(ndst
, sock6
->sock
->sk
, skb
, dev
,
2200 &local_ip
.sin6
.sin6_addr
,
2201 &dst
->sin6
.sin6_addr
, tos
, ttl
,
2202 label
, src_port
, dst_port
, !udp_sum
);
2210 dev
->stats
.tx_dropped
++;
2217 dev
->stats
.collisions
++;
2218 else if (err
== -ENETUNREACH
)
2219 dev
->stats
.tx_carrier_errors
++;
2221 dev
->stats
.tx_errors
++;
2225 /* Transmit local packets over Vxlan
2227 * Outer IP header inherits ECN and DF from inner header.
2228 * Outer UDP destination is the VXLAN assigned port.
2229 * source port is based on hash of flow
2231 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2233 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2234 const struct ip_tunnel_info
*info
;
2236 bool did_rsc
= false;
2237 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2238 struct vxlan_fdb
*f
;
2241 info
= skb_tunnel_info(skb
);
2243 skb_reset_mac_header(skb
);
2245 if (vxlan
->flags
& VXLAN_F_COLLECT_METADATA
) {
2246 if (info
&& info
->mode
& IP_TUNNEL_INFO_BRIDGE
&&
2247 info
->mode
& IP_TUNNEL_INFO_TX
) {
2248 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2250 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2251 vxlan_xmit_one(skb
, dev
, vni
, NULL
, false);
2254 return NETDEV_TX_OK
;
2258 if (vxlan
->flags
& VXLAN_F_PROXY
) {
2260 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2261 return arp_reduce(dev
, skb
, vni
);
2262 #if IS_ENABLED(CONFIG_IPV6)
2263 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2264 pskb_may_pull(skb
, sizeof(struct ipv6hdr
)
2265 + sizeof(struct nd_msg
)) &&
2266 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2269 msg
= (struct nd_msg
*)skb_transport_header(skb
);
2270 if (msg
->icmph
.icmp6_code
== 0 &&
2271 msg
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2272 return neigh_reduce(dev
, skb
, vni
);
2278 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2281 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
2282 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2283 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2284 did_rsc
= route_shortcircuit(dev
, skb
);
2286 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2290 f
= vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2292 if ((vxlan
->flags
& VXLAN_F_L2MISS
) &&
2293 !is_multicast_ether_addr(eth
->h_dest
))
2294 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2296 dev
->stats
.tx_dropped
++;
2298 return NETDEV_TX_OK
;
2302 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2303 struct sk_buff
*skb1
;
2309 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2311 vxlan_xmit_one(skb1
, dev
, vni
, rdst
, did_rsc
);
2315 vxlan_xmit_one(skb
, dev
, vni
, fdst
, did_rsc
);
2318 return NETDEV_TX_OK
;
2321 /* Walk the forwarding table and purge stale entries */
2322 static void vxlan_cleanup(unsigned long arg
)
2324 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
2325 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2328 if (!netif_running(vxlan
->dev
))
2331 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2332 struct hlist_node
*p
, *n
;
2334 spin_lock_bh(&vxlan
->hash_lock
);
2335 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2337 = container_of(p
, struct vxlan_fdb
, hlist
);
2338 unsigned long timeout
;
2340 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
2343 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2344 if (time_before_eq(timeout
, jiffies
)) {
2345 netdev_dbg(vxlan
->dev
,
2346 "garbage collect %pM\n",
2348 f
->state
= NUD_STALE
;
2349 vxlan_fdb_destroy(vxlan
, f
);
2350 } else if (time_before(timeout
, next_timer
))
2351 next_timer
= timeout
;
2353 spin_unlock_bh(&vxlan
->hash_lock
);
2356 mod_timer(&vxlan
->age_timer
, next_timer
);
2359 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
)
2361 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2363 spin_lock(&vn
->sock_lock
);
2364 hlist_del_init_rcu(&vxlan
->hlist
);
2365 spin_unlock(&vn
->sock_lock
);
2368 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
)
2370 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2371 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2373 spin_lock(&vn
->sock_lock
);
2374 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
2375 spin_unlock(&vn
->sock_lock
);
2378 /* Setup stats when device is created */
2379 static int vxlan_init(struct net_device
*dev
)
2381 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2388 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
, __be32 vni
)
2390 struct vxlan_fdb
*f
;
2392 spin_lock_bh(&vxlan
->hash_lock
);
2393 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2395 vxlan_fdb_destroy(vxlan
, f
);
2396 spin_unlock_bh(&vxlan
->hash_lock
);
2399 static void vxlan_uninit(struct net_device
*dev
)
2401 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2403 vxlan_fdb_delete_default(vxlan
, vxlan
->cfg
.vni
);
2405 free_percpu(dev
->tstats
);
2408 /* Start ageing timer and join group when device is brought up */
2409 static int vxlan_open(struct net_device
*dev
)
2411 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2414 ret
= vxlan_sock_add(vxlan
);
2418 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2419 ret
= vxlan_igmp_join(vxlan
);
2420 if (ret
== -EADDRINUSE
)
2423 vxlan_sock_release(vxlan
);
2428 if (vxlan
->cfg
.age_interval
)
2429 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2434 /* Purge the forwarding table */
2435 static void vxlan_flush(struct vxlan_dev
*vxlan
, bool do_all
)
2439 spin_lock_bh(&vxlan
->hash_lock
);
2440 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2441 struct hlist_node
*p
, *n
;
2442 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2444 = container_of(p
, struct vxlan_fdb
, hlist
);
2445 if (!do_all
&& (f
->state
& (NUD_PERMANENT
| NUD_NOARP
)))
2447 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2448 if (!is_zero_ether_addr(f
->eth_addr
))
2449 vxlan_fdb_destroy(vxlan
, f
);
2452 spin_unlock_bh(&vxlan
->hash_lock
);
2455 /* Cleanup timer and forwarding table on shutdown */
2456 static int vxlan_stop(struct net_device
*dev
)
2458 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2459 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2462 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2463 !vxlan_group_used(vn
, vxlan
))
2464 ret
= vxlan_igmp_leave(vxlan
);
2466 del_timer_sync(&vxlan
->age_timer
);
2468 vxlan_flush(vxlan
, false);
2469 vxlan_sock_release(vxlan
);
2474 /* Stub, nothing needs to be done. */
2475 static void vxlan_set_multicast_list(struct net_device
*dev
)
2479 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2481 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2482 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2483 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2484 dst
->remote_ifindex
);
2485 bool use_ipv6
= false;
2487 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
)
2490 /* This check is different than dev->max_mtu, because it looks at
2491 * the lowerdev->mtu, rather than the static dev->max_mtu
2494 int max_mtu
= lowerdev
->mtu
-
2495 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2496 if (new_mtu
> max_mtu
)
2504 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2506 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2507 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2508 __be16 sport
, dport
;
2510 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2511 vxlan
->cfg
.port_max
, true);
2512 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2514 if (ip_tunnel_info_af(info
) == AF_INET
) {
2515 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2518 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, 0, info
->key
.tos
,
2519 info
->key
.u
.ipv4
.dst
,
2520 &info
->key
.u
.ipv4
.src
, dport
, sport
,
2521 &info
->dst_cache
, info
);
2526 #if IS_ENABLED(CONFIG_IPV6)
2527 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2528 struct dst_entry
*ndst
;
2530 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, 0, info
->key
.tos
,
2531 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2532 &info
->key
.u
.ipv6
.src
, dport
, sport
,
2533 &info
->dst_cache
, info
);
2535 return PTR_ERR(ndst
);
2537 #else /* !CONFIG_IPV6 */
2538 return -EPFNOSUPPORT
;
2541 info
->key
.tp_src
= sport
;
2542 info
->key
.tp_dst
= dport
;
2546 static const struct net_device_ops vxlan_netdev_ether_ops
= {
2547 .ndo_init
= vxlan_init
,
2548 .ndo_uninit
= vxlan_uninit
,
2549 .ndo_open
= vxlan_open
,
2550 .ndo_stop
= vxlan_stop
,
2551 .ndo_start_xmit
= vxlan_xmit
,
2552 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2553 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2554 .ndo_change_mtu
= vxlan_change_mtu
,
2555 .ndo_validate_addr
= eth_validate_addr
,
2556 .ndo_set_mac_address
= eth_mac_addr
,
2557 .ndo_fdb_add
= vxlan_fdb_add
,
2558 .ndo_fdb_del
= vxlan_fdb_delete
,
2559 .ndo_fdb_dump
= vxlan_fdb_dump
,
2560 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2563 static const struct net_device_ops vxlan_netdev_raw_ops
= {
2564 .ndo_init
= vxlan_init
,
2565 .ndo_uninit
= vxlan_uninit
,
2566 .ndo_open
= vxlan_open
,
2567 .ndo_stop
= vxlan_stop
,
2568 .ndo_start_xmit
= vxlan_xmit
,
2569 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2570 .ndo_change_mtu
= vxlan_change_mtu
,
2571 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2574 /* Info for udev, that this is a virtual tunnel endpoint */
2575 static struct device_type vxlan_type
= {
2579 /* Calls the ndo_udp_tunnel_add of the caller in order to
2580 * supply the listening VXLAN udp ports. Callers are expected
2581 * to implement the ndo_udp_tunnel_add.
2583 static void vxlan_push_rx_ports(struct net_device
*dev
)
2585 struct vxlan_sock
*vs
;
2586 struct net
*net
= dev_net(dev
);
2587 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2590 spin_lock(&vn
->sock_lock
);
2591 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2592 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
)
2593 udp_tunnel_push_rx_port(dev
, vs
->sock
,
2594 (vs
->flags
& VXLAN_F_GPE
) ?
2595 UDP_TUNNEL_TYPE_VXLAN_GPE
:
2596 UDP_TUNNEL_TYPE_VXLAN
);
2598 spin_unlock(&vn
->sock_lock
);
2601 /* Initialize the device structure. */
2602 static void vxlan_setup(struct net_device
*dev
)
2604 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2607 eth_hw_addr_random(dev
);
2610 dev
->needs_free_netdev
= true;
2611 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2613 dev
->features
|= NETIF_F_LLTX
;
2614 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2615 dev
->features
|= NETIF_F_RXCSUM
;
2616 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2618 dev
->vlan_features
= dev
->features
;
2619 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2620 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2621 netif_keep_dst(dev
);
2622 dev
->priv_flags
|= IFF_NO_QUEUE
;
2624 INIT_LIST_HEAD(&vxlan
->next
);
2625 spin_lock_init(&vxlan
->hash_lock
);
2627 init_timer_deferrable(&vxlan
->age_timer
);
2628 vxlan
->age_timer
.function
= vxlan_cleanup
;
2629 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
2631 vxlan
->cfg
.dst_port
= htons(vxlan_port
);
2635 gro_cells_init(&vxlan
->gro_cells
, dev
);
2637 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2638 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2641 static void vxlan_ether_setup(struct net_device
*dev
)
2643 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2644 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2645 dev
->netdev_ops
= &vxlan_netdev_ether_ops
;
2648 static void vxlan_raw_setup(struct net_device
*dev
)
2650 dev
->header_ops
= NULL
;
2651 dev
->type
= ARPHRD_NONE
;
2652 dev
->hard_header_len
= 0;
2654 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
2655 dev
->netdev_ops
= &vxlan_netdev_raw_ops
;
2658 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2659 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2660 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2661 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2662 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2663 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2664 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2665 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2666 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2667 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
2668 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2669 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2670 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2671 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2672 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2673 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2674 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2675 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2676 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2677 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2678 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2679 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2680 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2681 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2682 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2683 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2684 [IFLA_VXLAN_GPE
] = { .type
= NLA_FLAG
, },
2685 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2688 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
2690 if (tb
[IFLA_ADDRESS
]) {
2691 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2692 pr_debug("invalid link address (not ethernet)\n");
2696 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2697 pr_debug("invalid all zero ethernet address\n");
2698 return -EADDRNOTAVAIL
;
2705 if (data
[IFLA_VXLAN_ID
]) {
2706 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2707 if (id
>= VXLAN_N_VID
)
2711 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2712 const struct ifla_vxlan_port_range
*p
2713 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
2715 if (ntohs(p
->high
) < ntohs(p
->low
)) {
2716 pr_debug("port range %u .. %u not valid\n",
2717 ntohs(p
->low
), ntohs(p
->high
));
2725 static void vxlan_get_drvinfo(struct net_device
*netdev
,
2726 struct ethtool_drvinfo
*drvinfo
)
2728 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
2729 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
2732 static const struct ethtool_ops vxlan_ethtool_ops
= {
2733 .get_drvinfo
= vxlan_get_drvinfo
,
2734 .get_link
= ethtool_op_get_link
,
2737 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
2738 __be16 port
, u32 flags
)
2740 struct socket
*sock
;
2741 struct udp_port_cfg udp_conf
;
2744 memset(&udp_conf
, 0, sizeof(udp_conf
));
2747 udp_conf
.family
= AF_INET6
;
2748 udp_conf
.use_udp6_rx_checksums
=
2749 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
2750 udp_conf
.ipv6_v6only
= 1;
2752 udp_conf
.family
= AF_INET
;
2755 udp_conf
.local_udp_port
= port
;
2757 /* Open UDP socket */
2758 err
= udp_sock_create(net
, &udp_conf
, &sock
);
2760 return ERR_PTR(err
);
2765 /* Create new listen socket if needed */
2766 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
2767 __be16 port
, u32 flags
)
2769 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2770 struct vxlan_sock
*vs
;
2771 struct socket
*sock
;
2773 struct udp_tunnel_sock_cfg tunnel_cfg
;
2775 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
2777 return ERR_PTR(-ENOMEM
);
2779 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
2780 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
2782 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
2784 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port
),
2787 return ERR_CAST(sock
);
2791 atomic_set(&vs
->refcnt
, 1);
2792 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
2794 spin_lock(&vn
->sock_lock
);
2795 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
2796 udp_tunnel_notify_add_rx_port(sock
,
2797 (vs
->flags
& VXLAN_F_GPE
) ?
2798 UDP_TUNNEL_TYPE_VXLAN_GPE
:
2799 UDP_TUNNEL_TYPE_VXLAN
);
2800 spin_unlock(&vn
->sock_lock
);
2802 /* Mark socket as an encapsulation socket. */
2803 memset(&tunnel_cfg
, 0, sizeof(tunnel_cfg
));
2804 tunnel_cfg
.sk_user_data
= vs
;
2805 tunnel_cfg
.encap_type
= 1;
2806 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
2807 tunnel_cfg
.encap_destroy
= NULL
;
2808 tunnel_cfg
.gro_receive
= vxlan_gro_receive
;
2809 tunnel_cfg
.gro_complete
= vxlan_gro_complete
;
2811 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
2816 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
2818 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2819 struct vxlan_sock
*vs
= NULL
;
2821 if (!vxlan
->cfg
.no_share
) {
2822 spin_lock(&vn
->sock_lock
);
2823 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
2824 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2825 if (vs
&& !atomic_add_unless(&vs
->refcnt
, 1, 0)) {
2826 spin_unlock(&vn
->sock_lock
);
2829 spin_unlock(&vn
->sock_lock
);
2832 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
2833 vxlan
->cfg
.dst_port
, vxlan
->flags
);
2836 #if IS_ENABLED(CONFIG_IPV6)
2838 rcu_assign_pointer(vxlan
->vn6_sock
, vs
);
2841 rcu_assign_pointer(vxlan
->vn4_sock
, vs
);
2842 vxlan_vs_add_dev(vs
, vxlan
);
2846 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
2848 bool ipv6
= vxlan
->flags
& VXLAN_F_IPV6
;
2849 bool metadata
= vxlan
->flags
& VXLAN_F_COLLECT_METADATA
;
2852 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
2853 #if IS_ENABLED(CONFIG_IPV6)
2854 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
2855 if (ipv6
|| metadata
)
2856 ret
= __vxlan_sock_add(vxlan
, true);
2858 if (!ret
&& (!ipv6
|| metadata
))
2859 ret
= __vxlan_sock_add(vxlan
, false);
2861 vxlan_sock_release(vxlan
);
2865 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
2866 struct vxlan_config
*conf
,
2869 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
2870 struct vxlan_dev
*vxlan
= netdev_priv(dev
), *tmp
;
2871 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2872 unsigned short needed_headroom
= ETH_HLEN
;
2873 bool use_ipv6
= false;
2874 __be16 default_port
= vxlan
->cfg
.dst_port
;
2875 struct net_device
*lowerdev
= NULL
;
2878 if (conf
->flags
& VXLAN_F_GPE
) {
2879 /* For now, allow GPE only together with
2880 * COLLECT_METADATA. This can be relaxed later; in such
2881 * case, the other side of the PtP link will have to be
2884 if ((conf
->flags
& ~VXLAN_F_ALLOWED_GPE
) ||
2885 !(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
2886 pr_info("unsupported combination of extensions\n");
2889 vxlan_raw_setup(dev
);
2891 vxlan_ether_setup(dev
);
2894 /* MTU range: 68 - 65535 */
2895 dev
->min_mtu
= ETH_MIN_MTU
;
2896 dev
->max_mtu
= ETH_MAX_MTU
;
2897 vxlan
->net
= src_net
;
2900 dst
->remote_vni
= conf
->vni
;
2902 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
2904 /* Unless IPv6 is explicitly requested, assume IPv4 */
2905 if (!dst
->remote_ip
.sa
.sa_family
)
2906 dst
->remote_ip
.sa
.sa_family
= AF_INET
;
2908 if (dst
->remote_ip
.sa
.sa_family
== AF_INET6
||
2909 vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) {
2910 if (!IS_ENABLED(CONFIG_IPV6
))
2911 return -EPFNOSUPPORT
;
2913 vxlan
->flags
|= VXLAN_F_IPV6
;
2916 if (conf
->label
&& !use_ipv6
) {
2917 pr_info("label only supported in use with IPv6\n");
2921 if (conf
->remote_ifindex
&&
2922 conf
->remote_ifindex
!= vxlan
->cfg
.remote_ifindex
) {
2923 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
2924 dst
->remote_ifindex
= conf
->remote_ifindex
;
2927 pr_info("ifindex %d does not exist\n",
2928 dst
->remote_ifindex
);
2932 #if IS_ENABLED(CONFIG_IPV6)
2934 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
2935 if (idev
&& idev
->cnf
.disable_ipv6
) {
2936 pr_info("IPv6 is disabled via sysctl\n");
2943 dev
->mtu
= lowerdev
->mtu
-
2944 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2946 needed_headroom
= lowerdev
->hard_header_len
;
2947 } else if (!conf
->remote_ifindex
&&
2948 vxlan_addr_multicast(&dst
->remote_ip
)) {
2949 pr_info("multicast destination requires interface to be specified\n");
2954 int max_mtu
= ETH_MAX_MTU
;
2957 max_mtu
= lowerdev
->mtu
;
2959 max_mtu
-= (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2961 if (conf
->mtu
< dev
->min_mtu
|| conf
->mtu
> dev
->max_mtu
)
2964 dev
->mtu
= conf
->mtu
;
2966 if (conf
->mtu
> max_mtu
)
2970 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
2971 needed_headroom
+= VXLAN6_HEADROOM
;
2973 needed_headroom
+= VXLAN_HEADROOM
;
2974 dev
->needed_headroom
= needed_headroom
;
2976 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
2977 if (!vxlan
->cfg
.dst_port
) {
2978 if (conf
->flags
& VXLAN_F_GPE
)
2979 vxlan
->cfg
.dst_port
= htons(4790); /* IANA VXLAN-GPE port */
2981 vxlan
->cfg
.dst_port
= default_port
;
2983 vxlan
->flags
|= conf
->flags
;
2985 if (!vxlan
->cfg
.age_interval
)
2986 vxlan
->cfg
.age_interval
= FDB_AGE_DEFAULT
;
2991 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
2992 if (tmp
->cfg
.vni
== conf
->vni
&&
2993 (tmp
->default_dst
.remote_ip
.sa
.sa_family
== AF_INET6
||
2994 tmp
->cfg
.saddr
.sa
.sa_family
== AF_INET6
) == use_ipv6
&&
2995 tmp
->cfg
.dst_port
== vxlan
->cfg
.dst_port
&&
2996 (tmp
->flags
& VXLAN_F_RCV_FLAGS
) ==
2997 (vxlan
->flags
& VXLAN_F_RCV_FLAGS
)) {
2998 pr_info("duplicate VNI %u\n", be32_to_cpu(conf
->vni
));
3006 static int __vxlan_dev_create(struct net
*net
, struct net_device
*dev
,
3007 struct vxlan_config
*conf
)
3009 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3010 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3013 err
= vxlan_dev_configure(net
, dev
, conf
, false);
3017 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
3019 /* create an fdb entry for a valid default destination */
3020 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
3021 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3022 &vxlan
->default_dst
.remote_ip
,
3023 NUD_REACHABLE
| NUD_PERMANENT
,
3024 NLM_F_EXCL
| NLM_F_CREATE
,
3025 vxlan
->cfg
.dst_port
,
3026 vxlan
->default_dst
.remote_vni
,
3027 vxlan
->default_dst
.remote_vni
,
3028 vxlan
->default_dst
.remote_ifindex
,
3034 err
= register_netdevice(dev
);
3036 vxlan_fdb_delete_default(vxlan
, vxlan
->default_dst
.remote_vni
);
3040 list_add(&vxlan
->next
, &vn
->vxlan_list
);
3044 static int vxlan_nl2conf(struct nlattr
*tb
[], struct nlattr
*data
[],
3045 struct net_device
*dev
, struct vxlan_config
*conf
,
3048 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3050 memset(conf
, 0, sizeof(*conf
));
3052 /* if changelink operation, start with old existing cfg */
3054 memcpy(conf
, &vxlan
->cfg
, sizeof(*conf
));
3056 if (data
[IFLA_VXLAN_ID
]) {
3057 __be32 vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3059 if (changelink
&& (vni
!= conf
->vni
))
3061 conf
->vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3064 if (data
[IFLA_VXLAN_GROUP
]) {
3065 conf
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
3066 } else if (data
[IFLA_VXLAN_GROUP6
]) {
3067 if (!IS_ENABLED(CONFIG_IPV6
))
3068 return -EPFNOSUPPORT
;
3070 conf
->remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
3071 conf
->remote_ip
.sa
.sa_family
= AF_INET6
;
3074 if (data
[IFLA_VXLAN_LOCAL
]) {
3075 conf
->saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
3076 conf
->saddr
.sa
.sa_family
= AF_INET
;
3077 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
3078 if (!IS_ENABLED(CONFIG_IPV6
))
3079 return -EPFNOSUPPORT
;
3081 /* TODO: respect scope id */
3082 conf
->saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
3083 conf
->saddr
.sa
.sa_family
= AF_INET6
;
3086 if (data
[IFLA_VXLAN_LINK
])
3087 conf
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
3089 if (data
[IFLA_VXLAN_TOS
])
3090 conf
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
3092 if (data
[IFLA_VXLAN_TTL
])
3093 conf
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
3095 if (data
[IFLA_VXLAN_LABEL
])
3096 conf
->label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
3097 IPV6_FLOWLABEL_MASK
;
3099 if (data
[IFLA_VXLAN_LEARNING
]) {
3100 if (nla_get_u8(data
[IFLA_VXLAN_LEARNING
])) {
3101 conf
->flags
|= VXLAN_F_LEARN
;
3103 conf
->flags
&= ~VXLAN_F_LEARN
;
3104 vxlan
->flags
&= ~VXLAN_F_LEARN
;
3106 } else if (!changelink
) {
3107 /* default to learn on a new device */
3108 conf
->flags
|= VXLAN_F_LEARN
;
3111 if (data
[IFLA_VXLAN_AGEING
]) {
3114 conf
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
3117 if (data
[IFLA_VXLAN_PROXY
]) {
3120 if (nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
3121 conf
->flags
|= VXLAN_F_PROXY
;
3124 if (data
[IFLA_VXLAN_RSC
]) {
3127 if (nla_get_u8(data
[IFLA_VXLAN_RSC
]))
3128 conf
->flags
|= VXLAN_F_RSC
;
3131 if (data
[IFLA_VXLAN_L2MISS
]) {
3134 if (nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
3135 conf
->flags
|= VXLAN_F_L2MISS
;
3138 if (data
[IFLA_VXLAN_L3MISS
]) {
3141 if (nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
3142 conf
->flags
|= VXLAN_F_L3MISS
;
3145 if (data
[IFLA_VXLAN_LIMIT
]) {
3148 conf
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
3151 if (data
[IFLA_VXLAN_COLLECT_METADATA
]) {
3154 if (nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
3155 conf
->flags
|= VXLAN_F_COLLECT_METADATA
;
3158 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3160 const struct ifla_vxlan_port_range
*p
3161 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3162 conf
->port_min
= ntohs(p
->low
);
3163 conf
->port_max
= ntohs(p
->high
);
3169 if (data
[IFLA_VXLAN_PORT
]) {
3172 conf
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
3175 if (data
[IFLA_VXLAN_UDP_CSUM
]) {
3178 if (!nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
3179 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
3182 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]) {
3185 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
3186 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
3189 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]) {
3192 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
3193 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
3196 if (data
[IFLA_VXLAN_REMCSUM_TX
]) {
3199 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
3200 conf
->flags
|= VXLAN_F_REMCSUM_TX
;
3203 if (data
[IFLA_VXLAN_REMCSUM_RX
]) {
3206 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
3207 conf
->flags
|= VXLAN_F_REMCSUM_RX
;
3210 if (data
[IFLA_VXLAN_GBP
]) {
3213 conf
->flags
|= VXLAN_F_GBP
;
3216 if (data
[IFLA_VXLAN_GPE
]) {
3219 conf
->flags
|= VXLAN_F_GPE
;
3222 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
]) {
3225 conf
->flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
3231 conf
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3237 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
3238 struct nlattr
*tb
[], struct nlattr
*data
[])
3240 struct vxlan_config conf
;
3243 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, false);
3247 return __vxlan_dev_create(src_net
, dev
, &conf
);
3250 static int vxlan_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
3251 struct nlattr
*data
[])
3253 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3254 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3255 struct vxlan_rdst old_dst
;
3256 struct vxlan_config conf
;
3259 err
= vxlan_nl2conf(tb
, data
,
3264 memcpy(&old_dst
, dst
, sizeof(struct vxlan_rdst
));
3266 err
= vxlan_dev_configure(vxlan
->net
, dev
, &conf
, true);
3270 /* handle default dst entry */
3271 if (!vxlan_addr_equal(&dst
->remote_ip
, &old_dst
.remote_ip
)) {
3272 spin_lock_bh(&vxlan
->hash_lock
);
3273 if (!vxlan_addr_any(&old_dst
.remote_ip
))
3274 __vxlan_fdb_delete(vxlan
, all_zeros_mac
,
3276 vxlan
->cfg
.dst_port
,
3279 old_dst
.remote_ifindex
, 0);
3281 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3282 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3284 NUD_REACHABLE
| NUD_PERMANENT
,
3285 NLM_F_CREATE
| NLM_F_APPEND
,
3286 vxlan
->cfg
.dst_port
,
3289 dst
->remote_ifindex
,
3292 spin_unlock_bh(&vxlan
->hash_lock
);
3296 spin_unlock_bh(&vxlan
->hash_lock
);
3302 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
3304 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3306 vxlan_flush(vxlan
, true);
3308 gro_cells_destroy(&vxlan
->gro_cells
);
3309 list_del(&vxlan
->next
);
3310 unregister_netdevice_queue(dev
, head
);
3313 static size_t vxlan_get_size(const struct net_device
*dev
)
3316 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
3317 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
3318 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
3319 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
3320 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3321 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3322 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
3323 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3324 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3325 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3326 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3327 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3328 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3329 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3330 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3331 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3332 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3333 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3334 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3335 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3336 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3337 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3341 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3343 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3344 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3345 struct ifla_vxlan_port_range ports
= {
3346 .low
= htons(vxlan
->cfg
.port_min
),
3347 .high
= htons(vxlan
->cfg
.port_max
),
3350 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3351 goto nla_put_failure
;
3353 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3354 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3355 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3356 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3357 goto nla_put_failure
;
3358 #if IS_ENABLED(CONFIG_IPV6)
3360 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3361 &dst
->remote_ip
.sin6
.sin6_addr
))
3362 goto nla_put_failure
;
3367 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3368 goto nla_put_failure
;
3370 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3371 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3372 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3373 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3374 goto nla_put_failure
;
3375 #if IS_ENABLED(CONFIG_IPV6)
3377 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3378 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3379 goto nla_put_failure
;
3384 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3385 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3386 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
3387 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3388 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
3389 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3390 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
3391 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
3392 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3393 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
3394 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3395 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
3396 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3397 !!(vxlan
->flags
& VXLAN_F_COLLECT_METADATA
)) ||
3398 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3399 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3400 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3401 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3402 !(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3403 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3404 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3405 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3406 !!(vxlan
->flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3407 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3408 !!(vxlan
->flags
& VXLAN_F_REMCSUM_TX
)) ||
3409 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3410 !!(vxlan
->flags
& VXLAN_F_REMCSUM_RX
)))
3411 goto nla_put_failure
;
3413 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3414 goto nla_put_failure
;
3416 if (vxlan
->flags
& VXLAN_F_GBP
&&
3417 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3418 goto nla_put_failure
;
3420 if (vxlan
->flags
& VXLAN_F_GPE
&&
3421 nla_put_flag(skb
, IFLA_VXLAN_GPE
))
3422 goto nla_put_failure
;
3424 if (vxlan
->flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3425 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3426 goto nla_put_failure
;
3434 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3436 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3441 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3443 .maxtype
= IFLA_VXLAN_MAX
,
3444 .policy
= vxlan_policy
,
3445 .priv_size
= sizeof(struct vxlan_dev
),
3446 .setup
= vxlan_setup
,
3447 .validate
= vxlan_validate
,
3448 .newlink
= vxlan_newlink
,
3449 .changelink
= vxlan_changelink
,
3450 .dellink
= vxlan_dellink
,
3451 .get_size
= vxlan_get_size
,
3452 .fill_info
= vxlan_fill_info
,
3453 .get_link_net
= vxlan_get_link_net
,
3456 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
3457 u8 name_assign_type
,
3458 struct vxlan_config
*conf
)
3460 struct nlattr
*tb
[IFLA_MAX
+ 1];
3461 struct net_device
*dev
;
3464 memset(&tb
, 0, sizeof(tb
));
3466 dev
= rtnl_create_link(net
, name
, name_assign_type
,
3467 &vxlan_link_ops
, tb
);
3471 err
= __vxlan_dev_create(net
, dev
, conf
);
3474 return ERR_PTR(err
);
3477 err
= rtnl_configure_link(dev
, NULL
);
3479 LIST_HEAD(list_kill
);
3481 vxlan_dellink(dev
, &list_kill
);
3482 unregister_netdevice_many(&list_kill
);
3483 return ERR_PTR(err
);
3488 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
3490 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3491 struct net_device
*dev
)
3493 struct vxlan_dev
*vxlan
, *next
;
3494 LIST_HEAD(list_kill
);
3496 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3497 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3499 /* In case we created vxlan device with carrier
3500 * and we loose the carrier due to module unload
3501 * we also need to remove vxlan device. In other
3502 * cases, it's not necessary and remote_ifindex
3503 * is 0 here, so no matches.
3505 if (dst
->remote_ifindex
== dev
->ifindex
)
3506 vxlan_dellink(vxlan
->dev
, &list_kill
);
3509 unregister_netdevice_many(&list_kill
);
3512 static int vxlan_netdevice_event(struct notifier_block
*unused
,
3513 unsigned long event
, void *ptr
)
3515 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3516 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3518 if (event
== NETDEV_UNREGISTER
)
3519 vxlan_handle_lowerdev_unregister(vn
, dev
);
3520 else if (event
== NETDEV_UDP_TUNNEL_PUSH_INFO
)
3521 vxlan_push_rx_ports(dev
);
3526 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3527 .notifier_call
= vxlan_netdevice_event
,
3530 static __net_init
int vxlan_init_net(struct net
*net
)
3532 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3535 INIT_LIST_HEAD(&vn
->vxlan_list
);
3536 spin_lock_init(&vn
->sock_lock
);
3538 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3539 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3544 static void __net_exit
vxlan_exit_net(struct net
*net
)
3546 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3547 struct vxlan_dev
*vxlan
, *next
;
3548 struct net_device
*dev
, *aux
;
3552 for_each_netdev_safe(net
, dev
, aux
)
3553 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
3554 unregister_netdevice_queue(dev
, &list
);
3556 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3557 /* If vxlan->dev is in the same netns, it has already been added
3558 * to the list by the previous loop.
3560 if (!net_eq(dev_net(vxlan
->dev
), net
)) {
3561 gro_cells_destroy(&vxlan
->gro_cells
);
3562 unregister_netdevice_queue(vxlan
->dev
, &list
);
3566 unregister_netdevice_many(&list
);
3570 static struct pernet_operations vxlan_net_ops
= {
3571 .init
= vxlan_init_net
,
3572 .exit
= vxlan_exit_net
,
3573 .id
= &vxlan_net_id
,
3574 .size
= sizeof(struct vxlan_net
),
3577 static int __init
vxlan_init_module(void)
3581 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
3583 rc
= register_pernet_subsys(&vxlan_net_ops
);
3587 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
3591 rc
= rtnl_link_register(&vxlan_link_ops
);
3597 unregister_netdevice_notifier(&vxlan_notifier_block
);
3599 unregister_pernet_subsys(&vxlan_net_ops
);
3603 late_initcall(vxlan_init_module
);
3605 static void __exit
vxlan_cleanup_module(void)
3607 rtnl_link_unregister(&vxlan_link_ops
);
3608 unregister_netdevice_notifier(&vxlan_notifier_block
);
3609 unregister_pernet_subsys(&vxlan_net_ops
);
3610 /* rcu_barrier() is called by netns */
3612 module_exit(vxlan_cleanup_module
);
3614 MODULE_LICENSE("GPL");
3615 MODULE_VERSION(VXLAN_VERSION
);
3616 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3617 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3618 MODULE_ALIAS_RTNL_LINK("vxlan");