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/inetdevice.h>
20 #include <linux/if_ether.h>
21 #include <linux/ethtool.h>
23 #include <net/ndisc.h>
26 #include <net/rtnetlink.h>
27 #include <net/inet_ecn.h>
28 #include <net/net_namespace.h>
29 #include <net/netns/generic.h>
30 #include <net/tun_proto.h>
31 #include <net/vxlan.h>
33 #if IS_ENABLED(CONFIG_IPV6)
34 #include <net/ip6_tunnel.h>
35 #include <net/ip6_checksum.h>
38 #define VXLAN_VERSION "0.1"
40 #define PORT_HASH_BITS 8
41 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
42 #define FDB_AGE_DEFAULT 300 /* 5 min */
43 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
45 /* UDP port for VXLAN traffic.
46 * The IANA assigned port is 4789, but the Linux default is 8472
47 * for compatibility with early adopters.
49 static unsigned short vxlan_port __read_mostly
= 8472;
50 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
51 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
53 static bool log_ecn_error
= true;
54 module_param(log_ecn_error
, bool, 0644);
55 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
57 static unsigned int vxlan_net_id
;
58 static struct rtnl_link_ops vxlan_link_ops
;
60 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
62 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
64 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
);
66 /* per-network namespace private data for this module */
68 struct list_head vxlan_list
;
69 struct hlist_head sock_list
[PORT_HASH_SIZE
];
73 /* Forwarding table entry */
75 struct hlist_node hlist
; /* linked list of entries */
77 unsigned long updated
; /* jiffies */
79 struct list_head remotes
;
80 u8 eth_addr
[ETH_ALEN
];
81 u16 state
; /* see ndm_state */
83 u8 flags
; /* see ndm_flags */
86 /* salt for hash table */
87 static u32 vxlan_salt __read_mostly
;
89 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
91 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
92 ip_tunnel_collect_metadata();
95 static struct ip_fan_map
*vxlan_fan_find_map(struct vxlan_dev
*vxlan
, __be32 daddr
)
97 struct ip_fan_map
*fan_map
;
100 list_for_each_entry_rcu(fan_map
, &vxlan
->fan
.fan_maps
, list
) {
101 if (fan_map
->overlay
==
102 (daddr
& inet_make_mask(fan_map
->overlay_prefix
))) {
112 static void vxlan_fan_flush_map(struct vxlan_dev
*vxlan
)
114 struct ip_fan_map
*fan_map
;
116 list_for_each_entry_rcu(fan_map
, &vxlan
->fan
.fan_maps
, list
) {
117 list_del_rcu(&fan_map
->list
);
118 kfree_rcu(fan_map
, rcu
);
122 static int vxlan_fan_del_map(struct vxlan_dev
*vxlan
, __be32 overlay
)
124 struct ip_fan_map
*fan_map
;
126 fan_map
= vxlan_fan_find_map(vxlan
, overlay
);
130 list_del_rcu(&fan_map
->list
);
131 kfree_rcu(fan_map
, rcu
);
136 static int vxlan_fan_add_map(struct vxlan_dev
*vxlan
, struct ifla_fan_map
*map
)
138 __be32 overlay_mask
, underlay_mask
;
139 struct ip_fan_map
*fan_map
;
141 overlay_mask
= inet_make_mask(map
->overlay_prefix
);
142 underlay_mask
= inet_make_mask(map
->underlay_prefix
);
144 netdev_dbg(vxlan
->dev
, "vfam: map: o %x/%d u %x/%d om %x um %x\n",
145 map
->overlay
, map
->overlay_prefix
,
146 map
->underlay
, map
->underlay_prefix
,
147 overlay_mask
, underlay_mask
);
149 if ((map
->overlay
& ~overlay_mask
) || (map
->underlay
& ~underlay_mask
))
152 if (!(map
->overlay
& overlay_mask
) && (map
->underlay
& underlay_mask
))
155 /* Special case: overlay 0 and underlay 0: flush all mappings */
156 if (!map
->overlay
&& !map
->underlay
) {
157 vxlan_fan_flush_map(vxlan
);
161 /* Special case: overlay set and underlay 0: clear map for overlay */
163 return vxlan_fan_del_map(vxlan
, map
->overlay
);
165 if (vxlan_fan_find_map(vxlan
, map
->overlay
))
168 fan_map
= kmalloc(sizeof(*fan_map
), GFP_KERNEL
);
169 fan_map
->underlay
= map
->underlay
;
170 fan_map
->overlay
= map
->overlay
;
171 fan_map
->underlay_prefix
= map
->underlay_prefix
;
172 fan_map
->overlay_mask
= ntohl(overlay_mask
);
173 fan_map
->overlay_prefix
= map
->overlay_prefix
;
175 list_add_tail_rcu(&fan_map
->list
, &vxlan
->fan
.fan_maps
);
180 static int vxlan_parse_fan_map(struct nlattr
*data
[], struct vxlan_dev
*vxlan
)
182 struct ifla_fan_map
*map
;
186 nla_for_each_nested(attr
, data
[IFLA_IPTUN_FAN_MAP
], rem
) {
187 map
= nla_data(attr
);
188 rv
= vxlan_fan_add_map(vxlan
, map
);
196 static int vxlan_fan_build_rdst(struct vxlan_dev
*vxlan
, struct sk_buff
*skb
,
197 struct vxlan_rdst
*fan_rdst
)
199 struct ip_fan_map
*f_map
;
200 union vxlan_addr
*va
;
208 switch (eth
->h_proto
) {
209 case htons(ETH_P_IP
):
215 case htons(ETH_P_ARP
):
220 netdev_dbg(vxlan
->dev
,
221 "vfbr: arp sha %pM sip %pI4 tha %pM tip %pI4\n",
222 arp_ptr
, arp_ptr
+ skb
->dev
->addr_len
,
223 arp_ptr
+ skb
->dev
->addr_len
+ 4,
224 arp_ptr
+ (skb
->dev
->addr_len
* 2) + 4);
225 arp_ptr
+= (skb
->dev
->addr_len
* 2) + 4;
226 memcpy(&daddr
, arp_ptr
, 4);
229 netdev_dbg(vxlan
->dev
, "vfbr: unknown eth p %x\n", eth
->h_proto
);
233 f_map
= vxlan_fan_find_map(vxlan
, daddr
);
237 daddr
= ntohl(daddr
);
238 underlay
= ntohl(f_map
->underlay
);
242 memset(fan_rdst
, 0, sizeof(*fan_rdst
));
243 va
= &fan_rdst
->remote_ip
;
244 va
->sa
.sa_family
= AF_INET
;
245 fan_rdst
->remote_vni
= vxlan
->default_dst
.remote_vni
;
246 va
->sin
.sin_addr
.s_addr
= htonl(underlay
|
247 ((daddr
& ~f_map
->overlay_mask
) >>
248 (32 - f_map
->overlay_prefix
-
249 (32 - f_map
->underlay_prefix
))));
250 netdev_dbg(vxlan
->dev
, "vfbr: daddr %x ul %x dst %x\n",
251 daddr
, underlay
, va
->sin
.sin_addr
.s_addr
);
256 #if IS_ENABLED(CONFIG_IPV6)
258 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
260 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
262 if (a
->sa
.sa_family
== AF_INET6
)
263 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
265 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
268 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
270 if (ipa
->sa
.sa_family
== AF_INET6
)
271 return ipv6_addr_any(&ipa
->sin6
.sin6_addr
);
273 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
276 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
278 if (ipa
->sa
.sa_family
== AF_INET6
)
279 return ipv6_addr_is_multicast(&ipa
->sin6
.sin6_addr
);
281 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
284 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
286 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
287 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
288 ip
->sa
.sa_family
= AF_INET6
;
290 } else if (nla_len(nla
) >= sizeof(__be32
)) {
291 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
292 ip
->sa
.sa_family
= AF_INET
;
295 return -EAFNOSUPPORT
;
299 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
300 const union vxlan_addr
*ip
)
302 if (ip
->sa
.sa_family
== AF_INET6
)
303 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
305 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
308 #else /* !CONFIG_IPV6 */
311 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
313 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
316 static inline bool vxlan_addr_any(const union vxlan_addr
*ipa
)
318 return ipa
->sin
.sin_addr
.s_addr
== htonl(INADDR_ANY
);
321 static inline bool vxlan_addr_multicast(const union vxlan_addr
*ipa
)
323 return IN_MULTICAST(ntohl(ipa
->sin
.sin_addr
.s_addr
));
326 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
328 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
329 return -EAFNOSUPPORT
;
330 } else if (nla_len(nla
) >= sizeof(__be32
)) {
331 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
332 ip
->sa
.sa_family
= AF_INET
;
335 return -EAFNOSUPPORT
;
339 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
340 const union vxlan_addr
*ip
)
342 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
346 /* Virtual Network hash table head */
347 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
349 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
352 /* Socket hash table head */
353 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
355 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
357 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
360 /* First remote destination for a forwarding entry.
361 * Guaranteed to be non-NULL because remotes are never deleted.
363 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
365 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
368 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
370 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
373 /* Find VXLAN socket based on network namespace, address family and UDP port
374 * and enabled unshareable flags.
376 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
377 __be16 port
, u32 flags
)
379 struct vxlan_sock
*vs
;
381 flags
&= VXLAN_F_RCV_FLAGS
;
383 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
384 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
385 vxlan_get_sk_family(vs
) == family
&&
392 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, int ifindex
,
395 struct vxlan_dev_node
*node
;
397 /* For flow based devices, map all packets to VNI 0 */
398 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
401 hlist_for_each_entry_rcu(node
, vni_head(vs
, vni
), hlist
) {
402 if (node
->vxlan
->default_dst
.remote_vni
!= vni
)
405 if (IS_ENABLED(CONFIG_IPV6
)) {
406 const struct vxlan_config
*cfg
= &node
->vxlan
->cfg
;
408 if ((cfg
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
409 cfg
->remote_ifindex
!= ifindex
)
419 /* Look up VNI in a per net namespace table */
420 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, int ifindex
,
421 __be32 vni
, sa_family_t family
,
422 __be16 port
, u32 flags
)
424 struct vxlan_sock
*vs
;
426 vs
= vxlan_find_sock(net
, family
, port
, flags
);
430 return vxlan_vs_find_vni(vs
, ifindex
, vni
);
433 /* Fill in neighbour message in skbuff. */
434 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
435 const struct vxlan_fdb
*fdb
,
436 u32 portid
, u32 seq
, int type
, unsigned int flags
,
437 const struct vxlan_rdst
*rdst
)
439 unsigned long now
= jiffies
;
440 struct nda_cacheinfo ci
;
441 struct nlmsghdr
*nlh
;
443 bool send_ip
, send_eth
;
445 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
449 ndm
= nlmsg_data(nlh
);
450 memset(ndm
, 0, sizeof(*ndm
));
452 send_eth
= send_ip
= true;
454 if (type
== RTM_GETNEIGH
) {
455 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
456 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
457 ndm
->ndm_family
= send_ip
? rdst
->remote_ip
.sa
.sa_family
: AF_INET
;
459 ndm
->ndm_family
= AF_BRIDGE
;
460 ndm
->ndm_state
= fdb
->state
;
461 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
462 ndm
->ndm_flags
= fdb
->flags
;
463 ndm
->ndm_type
= RTN_UNICAST
;
465 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
466 nla_put_s32(skb
, NDA_LINK_NETNSID
,
467 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
468 goto nla_put_failure
;
470 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
471 goto nla_put_failure
;
473 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
474 goto nla_put_failure
;
476 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
477 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
478 goto nla_put_failure
;
479 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
480 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
481 goto nla_put_failure
;
482 if ((vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) && fdb
->vni
&&
483 nla_put_u32(skb
, NDA_SRC_VNI
,
484 be32_to_cpu(fdb
->vni
)))
485 goto nla_put_failure
;
486 if (rdst
->remote_ifindex
&&
487 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
488 goto nla_put_failure
;
490 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
491 ci
.ndm_confirmed
= 0;
492 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
495 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
496 goto nla_put_failure
;
502 nlmsg_cancel(skb
, nlh
);
506 static inline size_t vxlan_nlmsg_size(void)
508 return NLMSG_ALIGN(sizeof(struct ndmsg
))
509 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
510 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
511 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
512 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
513 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
514 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
515 + nla_total_size(sizeof(struct nda_cacheinfo
));
518 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
519 struct vxlan_rdst
*rd
, int type
)
521 struct net
*net
= dev_net(vxlan
->dev
);
525 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
529 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
531 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
532 WARN_ON(err
== -EMSGSIZE
);
537 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
541 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
544 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
546 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
547 struct vxlan_fdb f
= {
550 struct vxlan_rdst remote
= {
551 .remote_ip
= *ipa
, /* goes to NDA_DST */
552 .remote_vni
= cpu_to_be32(VXLAN_N_VID
),
555 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
558 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
560 struct vxlan_fdb f
= {
563 struct vxlan_rdst remote
= { };
565 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
567 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
);
570 /* Hash Ethernet address */
571 static u32
eth_hash(const unsigned char *addr
)
573 u64 value
= get_unaligned((u64
*)addr
);
575 /* only want 6 bytes */
581 return hash_64(value
, FDB_HASH_BITS
);
584 static u32
eth_vni_hash(const unsigned char *addr
, __be32 vni
)
586 /* use 1 byte of OUI and 3 bytes of NIC */
587 u32 key
= get_unaligned((u32
*)(addr
+ 2));
589 return jhash_2words(key
, vni
, vxlan_salt
) & (FDB_HASH_SIZE
- 1);
592 /* Hash chain to use given mac address */
593 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
594 const u8
*mac
, __be32 vni
)
596 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)
597 return &vxlan
->fdb_head
[eth_vni_hash(mac
, vni
)];
599 return &vxlan
->fdb_head
[eth_hash(mac
)];
602 /* Look up Ethernet address in forwarding table */
603 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
604 const u8
*mac
, __be32 vni
)
606 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
, vni
);
609 hlist_for_each_entry_rcu(f
, head
, hlist
) {
610 if (ether_addr_equal(mac
, f
->eth_addr
)) {
611 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
623 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
624 const u8
*mac
, __be32 vni
)
628 f
= __vxlan_find_mac(vxlan
, mac
, vni
);
635 /* caller should hold vxlan->hash_lock */
636 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
637 union vxlan_addr
*ip
, __be16 port
,
638 __be32 vni
, __u32 ifindex
)
640 struct vxlan_rdst
*rd
;
642 list_for_each_entry(rd
, &f
->remotes
, list
) {
643 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
644 rd
->remote_port
== port
&&
645 rd
->remote_vni
== vni
&&
646 rd
->remote_ifindex
== ifindex
)
653 /* Replace destination of unicast mac */
654 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
655 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
658 struct vxlan_rdst
*rd
;
660 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
664 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
668 dst_cache_reset(&rd
->dst_cache
);
670 rd
->remote_port
= port
;
671 rd
->remote_vni
= vni
;
672 rd
->remote_ifindex
= ifindex
;
676 /* Add/update destinations for multicast */
677 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
678 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
679 __u32 ifindex
, struct vxlan_rdst
**rdp
)
681 struct vxlan_rdst
*rd
;
683 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
687 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
691 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
697 rd
->remote_port
= port
;
698 rd
->remote_vni
= vni
;
699 rd
->remote_ifindex
= ifindex
;
701 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
707 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
709 struct vxlanhdr
*vh
, size_t hdrlen
,
711 struct gro_remcsum
*grc
,
714 size_t start
, offset
;
716 if (skb
->remcsum_offload
)
719 if (!NAPI_GRO_CB(skb
)->csum_valid
)
722 start
= vxlan_rco_start(vni_field
);
723 offset
= start
+ vxlan_rco_offset(vni_field
);
725 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
726 start
, offset
, grc
, nopartial
);
728 skb
->remcsum_offload
= 1;
733 static struct sk_buff
**vxlan_gro_receive(struct sock
*sk
,
734 struct sk_buff
**head
,
737 struct sk_buff
*p
, **pp
= NULL
;
738 struct vxlanhdr
*vh
, *vh2
;
739 unsigned int hlen
, off_vx
;
741 struct vxlan_sock
*vs
= rcu_dereference_sk_user_data(sk
);
743 struct gro_remcsum grc
;
745 skb_gro_remcsum_init(&grc
);
747 off_vx
= skb_gro_offset(skb
);
748 hlen
= off_vx
+ sizeof(*vh
);
749 vh
= skb_gro_header_fast(skb
, off_vx
);
750 if (skb_gro_header_hard(skb
, hlen
)) {
751 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
756 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
758 flags
= vh
->vx_flags
;
760 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
761 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
764 VXLAN_F_REMCSUM_NOPARTIAL
));
770 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
772 for (p
= *head
; p
; p
= p
->next
) {
773 if (!NAPI_GRO_CB(p
)->same_flow
)
776 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
777 if (vh
->vx_flags
!= vh2
->vx_flags
||
778 vh
->vx_vni
!= vh2
->vx_vni
) {
779 NAPI_GRO_CB(p
)->same_flow
= 0;
784 pp
= call_gro_receive(eth_gro_receive
, head
, skb
);
788 skb_gro_flush_final_remcsum(skb
, pp
, flush
, &grc
);
793 static int vxlan_gro_complete(struct sock
*sk
, struct sk_buff
*skb
, int nhoff
)
795 /* Sets 'skb->inner_mac_header' since we are always called with
796 * 'skb->encapsulation' set.
798 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
801 static struct vxlan_fdb
*vxlan_fdb_alloc(struct vxlan_dev
*vxlan
,
802 const u8
*mac
, __u16 state
,
803 __be32 src_vni
, __u8 ndm_flags
)
807 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
811 f
->flags
= ndm_flags
;
812 f
->updated
= f
->used
= jiffies
;
814 INIT_LIST_HEAD(&f
->remotes
);
815 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
820 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
821 const u8
*mac
, union vxlan_addr
*ip
,
822 __u16 state
, __be16 port
, __be32 src_vni
,
823 __be32 vni
, __u32 ifindex
, __u8 ndm_flags
,
824 struct vxlan_fdb
**fdb
)
826 struct vxlan_rdst
*rd
= NULL
;
830 if (vxlan
->cfg
.addrmax
&&
831 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
834 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
835 f
= vxlan_fdb_alloc(vxlan
, mac
, state
, src_vni
, ndm_flags
);
839 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
846 hlist_add_head_rcu(&f
->hlist
,
847 vxlan_fdb_head(vxlan
, mac
, src_vni
));
854 /* Add new entry to forwarding table -- assumes lock held */
855 static int vxlan_fdb_update(struct vxlan_dev
*vxlan
,
856 const u8
*mac
, union vxlan_addr
*ip
,
857 __u16 state
, __u16 flags
,
858 __be16 port
, __be32 src_vni
, __be32 vni
,
859 __u32 ifindex
, __u8 ndm_flags
)
861 struct vxlan_rdst
*rd
= NULL
;
866 f
= __vxlan_find_mac(vxlan
, mac
, src_vni
);
868 if (flags
& NLM_F_EXCL
) {
869 netdev_dbg(vxlan
->dev
,
870 "lost race to create %pM\n", mac
);
873 if (f
->state
!= state
) {
875 f
->updated
= jiffies
;
878 if (f
->flags
!= ndm_flags
) {
879 f
->flags
= ndm_flags
;
880 f
->updated
= jiffies
;
883 if ((flags
& NLM_F_REPLACE
)) {
884 /* Only change unicasts */
885 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
886 is_zero_ether_addr(f
->eth_addr
))) {
887 notify
|= vxlan_fdb_replace(f
, ip
, port
, vni
,
892 if ((flags
& NLM_F_APPEND
) &&
893 (is_multicast_ether_addr(f
->eth_addr
) ||
894 is_zero_ether_addr(f
->eth_addr
))) {
895 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
902 if (!(flags
& NLM_F_CREATE
))
905 /* Disallow replace to add a multicast entry */
906 if ((flags
& NLM_F_REPLACE
) &&
907 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
910 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
911 rc
= vxlan_fdb_create(vxlan
, mac
, ip
, state
, port
, src_vni
,
912 vni
, ifindex
, ndm_flags
, &f
);
920 rd
= first_remote_rtnl(f
);
921 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
);
927 static void vxlan_fdb_free(struct rcu_head
*head
)
929 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
930 struct vxlan_rdst
*rd
, *nd
;
932 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
933 dst_cache_destroy(&rd
->dst_cache
);
939 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
942 netdev_dbg(vxlan
->dev
,
943 "delete %pM\n", f
->eth_addr
);
947 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_DELNEIGH
);
949 hlist_del_rcu(&f
->hlist
);
950 call_rcu(&f
->rcu
, vxlan_fdb_free
);
953 static void vxlan_dst_free(struct rcu_head
*head
)
955 struct vxlan_rdst
*rd
= container_of(head
, struct vxlan_rdst
, rcu
);
957 dst_cache_destroy(&rd
->dst_cache
);
961 static void vxlan_fdb_dst_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
962 struct vxlan_rdst
*rd
)
964 list_del_rcu(&rd
->list
);
965 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
);
966 call_rcu(&rd
->rcu
, vxlan_dst_free
);
969 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
970 union vxlan_addr
*ip
, __be16
*port
, __be32
*src_vni
,
971 __be32
*vni
, u32
*ifindex
)
973 struct net
*net
= dev_net(vxlan
->dev
);
977 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
981 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
982 if (remote
->sa
.sa_family
== AF_INET
) {
983 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
984 ip
->sa
.sa_family
= AF_INET
;
985 #if IS_ENABLED(CONFIG_IPV6)
987 ip
->sin6
.sin6_addr
= in6addr_any
;
988 ip
->sa
.sa_family
= AF_INET6
;
994 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
996 *port
= nla_get_be16(tb
[NDA_PORT
]);
998 *port
= vxlan
->cfg
.dst_port
;
1002 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
1004 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
1006 *vni
= vxlan
->default_dst
.remote_vni
;
1009 if (tb
[NDA_SRC_VNI
]) {
1010 if (nla_len(tb
[NDA_SRC_VNI
]) != sizeof(u32
))
1012 *src_vni
= cpu_to_be32(nla_get_u32(tb
[NDA_SRC_VNI
]));
1014 *src_vni
= vxlan
->default_dst
.remote_vni
;
1017 if (tb
[NDA_IFINDEX
]) {
1018 struct net_device
*tdev
;
1020 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
1022 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
1023 tdev
= __dev_get_by_index(net
, *ifindex
);
1025 return -EADDRNOTAVAIL
;
1033 /* Add static entry (via netlink) */
1034 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
1035 struct net_device
*dev
,
1036 const unsigned char *addr
, u16 vid
, u16 flags
)
1038 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1039 /* struct net *net = dev_net(vxlan->dev); */
1040 union vxlan_addr ip
;
1042 __be32 src_vni
, vni
;
1046 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
1047 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1052 if (tb
[NDA_DST
] == NULL
)
1055 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
1059 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
1060 return -EAFNOSUPPORT
;
1062 spin_lock_bh(&vxlan
->hash_lock
);
1063 err
= vxlan_fdb_update(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
1064 port
, src_vni
, vni
, ifindex
, ndm
->ndm_flags
);
1065 spin_unlock_bh(&vxlan
->hash_lock
);
1070 static int __vxlan_fdb_delete(struct vxlan_dev
*vxlan
,
1071 const unsigned char *addr
, union vxlan_addr ip
,
1072 __be16 port
, __be32 src_vni
, __be32 vni
,
1073 u32 ifindex
, u16 vid
)
1075 struct vxlan_fdb
*f
;
1076 struct vxlan_rdst
*rd
= NULL
;
1079 f
= vxlan_find_mac(vxlan
, addr
, src_vni
);
1083 if (!vxlan_addr_any(&ip
)) {
1084 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
1089 /* remove a destination if it's not the only one on the list,
1090 * otherwise destroy the fdb entry
1092 if (rd
&& !list_is_singular(&f
->remotes
)) {
1093 vxlan_fdb_dst_destroy(vxlan
, f
, rd
);
1097 vxlan_fdb_destroy(vxlan
, f
, true);
1103 /* Delete entry (via netlink) */
1104 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
1105 struct net_device
*dev
,
1106 const unsigned char *addr
, u16 vid
)
1108 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1109 union vxlan_addr ip
;
1110 __be32 src_vni
, vni
;
1115 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
1119 spin_lock_bh(&vxlan
->hash_lock
);
1120 err
= __vxlan_fdb_delete(vxlan
, addr
, ip
, port
, src_vni
, vni
, ifindex
,
1122 spin_unlock_bh(&vxlan
->hash_lock
);
1127 /* Dump forwarding table */
1128 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
1129 struct net_device
*dev
,
1130 struct net_device
*filter_dev
, int *idx
)
1132 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1136 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1137 struct vxlan_fdb
*f
;
1139 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
1140 struct vxlan_rdst
*rd
;
1142 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
1143 if (*idx
< cb
->args
[2])
1146 err
= vxlan_fdb_info(skb
, vxlan
, f
,
1147 NETLINK_CB(cb
->skb
).portid
,
1162 /* Watch incoming packets to learn mapping between Ethernet address
1163 * and Tunnel endpoint.
1164 * Return true if packet is bogus and should be dropped.
1166 static bool vxlan_snoop(struct net_device
*dev
,
1167 union vxlan_addr
*src_ip
, const u8
*src_mac
,
1168 u32 src_ifindex
, __be32 vni
)
1170 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1171 struct vxlan_fdb
*f
;
1174 #if IS_ENABLED(CONFIG_IPV6)
1175 if (src_ip
->sa
.sa_family
== AF_INET6
&&
1176 (ipv6_addr_type(&src_ip
->sin6
.sin6_addr
) & IPV6_ADDR_LINKLOCAL
))
1177 ifindex
= src_ifindex
;
1180 f
= vxlan_find_mac(vxlan
, src_mac
, vni
);
1182 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
1184 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
) &&
1185 rdst
->remote_ifindex
== ifindex
))
1188 /* Don't migrate static entries, drop packets */
1189 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
1192 if (net_ratelimit())
1194 "%pM migrated from %pIS to %pIS\n",
1195 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
1197 rdst
->remote_ip
= *src_ip
;
1198 f
->updated
= jiffies
;
1199 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
);
1201 /* learned new entry */
1202 spin_lock(&vxlan
->hash_lock
);
1204 /* close off race between vxlan_flush and incoming packets */
1205 if (netif_running(dev
))
1206 vxlan_fdb_update(vxlan
, src_mac
, src_ip
,
1208 NLM_F_EXCL
|NLM_F_CREATE
,
1209 vxlan
->cfg
.dst_port
,
1211 vxlan
->default_dst
.remote_vni
,
1213 spin_unlock(&vxlan
->hash_lock
);
1219 /* See if multicast group is already in use by other ID */
1220 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1222 struct vxlan_dev
*vxlan
;
1223 struct vxlan_sock
*sock4
;
1224 #if IS_ENABLED(CONFIG_IPV6)
1225 struct vxlan_sock
*sock6
;
1227 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1229 sock4
= rtnl_dereference(dev
->vn4_sock
);
1231 /* The vxlan_sock is only used by dev, leaving group has
1232 * no effect on other vxlan devices.
1234 if (family
== AF_INET
&& sock4
&& refcount_read(&sock4
->refcnt
) == 1)
1236 #if IS_ENABLED(CONFIG_IPV6)
1237 sock6
= rtnl_dereference(dev
->vn6_sock
);
1238 if (family
== AF_INET6
&& sock6
&& refcount_read(&sock6
->refcnt
) == 1)
1242 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1243 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1246 if (family
== AF_INET
&&
1247 rtnl_dereference(vxlan
->vn4_sock
) != sock4
)
1249 #if IS_ENABLED(CONFIG_IPV6)
1250 if (family
== AF_INET6
&&
1251 rtnl_dereference(vxlan
->vn6_sock
) != sock6
)
1255 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1256 &dev
->default_dst
.remote_ip
))
1259 if (vxlan
->default_dst
.remote_ifindex
!=
1260 dev
->default_dst
.remote_ifindex
)
1269 static bool __vxlan_sock_release_prep(struct vxlan_sock
*vs
)
1271 struct vxlan_net
*vn
;
1275 if (!refcount_dec_and_test(&vs
->refcnt
))
1278 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1279 spin_lock(&vn
->sock_lock
);
1280 hlist_del_rcu(&vs
->hlist
);
1281 udp_tunnel_notify_del_rx_port(vs
->sock
,
1282 (vs
->flags
& VXLAN_F_GPE
) ?
1283 UDP_TUNNEL_TYPE_VXLAN_GPE
:
1284 UDP_TUNNEL_TYPE_VXLAN
);
1285 spin_unlock(&vn
->sock_lock
);
1290 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1292 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1293 #if IS_ENABLED(CONFIG_IPV6)
1294 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1296 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
1299 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
1302 vxlan_vs_del_dev(vxlan
);
1304 if (__vxlan_sock_release_prep(sock4
)) {
1305 udp_tunnel_sock_release(sock4
->sock
);
1309 #if IS_ENABLED(CONFIG_IPV6)
1310 if (__vxlan_sock_release_prep(sock6
)) {
1311 udp_tunnel_sock_release(sock6
->sock
);
1317 /* Update multicast group membership when first VNI on
1318 * multicast address is brought up
1320 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1323 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1324 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1327 if (ip
->sa
.sa_family
== AF_INET
) {
1328 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1329 struct ip_mreqn mreq
= {
1330 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1331 .imr_ifindex
= ifindex
,
1334 sk
= sock4
->sock
->sk
;
1336 ret
= ip_mc_join_group(sk
, &mreq
);
1338 #if IS_ENABLED(CONFIG_IPV6)
1340 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1342 sk
= sock6
->sock
->sk
;
1344 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1345 &ip
->sin6
.sin6_addr
);
1353 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1354 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1357 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1358 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1361 if (ip
->sa
.sa_family
== AF_INET
) {
1362 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1363 struct ip_mreqn mreq
= {
1364 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1365 .imr_ifindex
= ifindex
,
1368 sk
= sock4
->sock
->sk
;
1370 ret
= ip_mc_leave_group(sk
, &mreq
);
1372 #if IS_ENABLED(CONFIG_IPV6)
1374 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1376 sk
= sock6
->sock
->sk
;
1378 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1379 &ip
->sin6
.sin6_addr
);
1387 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1388 struct sk_buff
*skb
, u32 vxflags
)
1390 size_t start
, offset
;
1392 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1395 start
= vxlan_rco_start(unparsed
->vx_vni
);
1396 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1398 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1401 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1402 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1404 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1405 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1409 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1410 struct sk_buff
*skb
, u32 vxflags
,
1411 struct vxlan_metadata
*md
)
1413 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1414 struct metadata_dst
*tun_dst
;
1416 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1419 md
->gbp
= ntohs(gbp
->policy_id
);
1421 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1423 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1424 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1426 if (gbp
->dont_learn
)
1427 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1429 if (gbp
->policy_applied
)
1430 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1432 /* In flow-based mode, GBP is carried in dst_metadata */
1433 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1434 skb
->mark
= md
->gbp
;
1436 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1439 static bool vxlan_parse_gpe_hdr(struct vxlanhdr
*unparsed
,
1441 struct sk_buff
*skb
, u32 vxflags
)
1443 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)unparsed
;
1445 /* Need to have Next Protocol set for interfaces in GPE mode. */
1446 if (!gpe
->np_applied
)
1448 /* "The initial version is 0. If a receiver does not support the
1449 * version indicated it MUST drop the packet.
1451 if (gpe
->version
!= 0)
1453 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1454 * processing MUST occur." However, we don't implement OAM
1455 * processing, thus drop the packet.
1460 *protocol
= tun_p_to_eth_p(gpe
->next_protocol
);
1464 unparsed
->vx_flags
&= ~VXLAN_GPE_USED_BITS
;
1468 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1469 struct vxlan_sock
*vs
,
1470 struct sk_buff
*skb
, __be32 vni
)
1472 union vxlan_addr saddr
;
1473 u32 ifindex
= skb
->dev
->ifindex
;
1475 skb_reset_mac_header(skb
);
1476 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1477 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1479 /* Ignore packet loops (and multicast echo) */
1480 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1483 /* Get address from the outer IP header */
1484 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1485 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1486 saddr
.sa
.sa_family
= AF_INET
;
1487 #if IS_ENABLED(CONFIG_IPV6)
1489 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1490 saddr
.sa
.sa_family
= AF_INET6
;
1494 if ((vxlan
->cfg
.flags
& VXLAN_F_LEARN
) &&
1495 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
, ifindex
, vni
))
1501 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1502 struct sk_buff
*skb
)
1506 if (vxlan_get_sk_family(vs
) == AF_INET
)
1507 err
= IP_ECN_decapsulate(oiph
, skb
);
1508 #if IS_ENABLED(CONFIG_IPV6)
1510 err
= IP6_ECN_decapsulate(oiph
, skb
);
1513 if (unlikely(err
) && log_ecn_error
) {
1514 if (vxlan_get_sk_family(vs
) == AF_INET
)
1515 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1516 &((struct iphdr
*)oiph
)->saddr
,
1517 ((struct iphdr
*)oiph
)->tos
);
1519 net_info_ratelimited("non-ECT from %pI6\n",
1520 &((struct ipv6hdr
*)oiph
)->saddr
);
1525 /* Callback from net/ipv4/udp.c to receive packets */
1526 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1528 struct pcpu_sw_netstats
*stats
;
1529 struct vxlan_dev
*vxlan
;
1530 struct vxlan_sock
*vs
;
1531 struct vxlanhdr unparsed
;
1532 struct vxlan_metadata _md
;
1533 struct vxlan_metadata
*md
= &_md
;
1534 __be16 protocol
= htons(ETH_P_TEB
);
1535 bool raw_proto
= false;
1539 /* Need UDP and VXLAN header to be present */
1540 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1543 unparsed
= *vxlan_hdr(skb
);
1544 /* VNI flag always required to be set */
1545 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1546 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1547 ntohl(vxlan_hdr(skb
)->vx_flags
),
1548 ntohl(vxlan_hdr(skb
)->vx_vni
));
1549 /* Return non vxlan pkt */
1552 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1553 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1555 vs
= rcu_dereference_sk_user_data(sk
);
1559 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1561 vxlan
= vxlan_vs_find_vni(vs
, skb
->dev
->ifindex
, vni
);
1565 /* For backwards compatibility, only allow reserved fields to be
1566 * used by VXLAN extensions if explicitly requested.
1568 if (vs
->flags
& VXLAN_F_GPE
) {
1569 if (!vxlan_parse_gpe_hdr(&unparsed
, &protocol
, skb
, vs
->flags
))
1574 if (__iptunnel_pull_header(skb
, VXLAN_HLEN
, protocol
, raw_proto
,
1575 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1578 if (vxlan_collect_metadata(vs
)) {
1579 struct metadata_dst
*tun_dst
;
1581 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1582 key32_to_tunnel_id(vni
), sizeof(*md
));
1587 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1589 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1591 memset(md
, 0, sizeof(*md
));
1594 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1595 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1597 if (vs
->flags
& VXLAN_F_GBP
)
1598 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1599 /* Note that GBP and GPE can never be active together. This is
1600 * ensured in vxlan_dev_configure.
1603 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1604 /* If there are any unprocessed flags remaining treat
1605 * this as a malformed packet. This behavior diverges from
1606 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1607 * in reserved fields are to be ignored. The approach here
1608 * maintains compatibility with previous stack code, and also
1609 * is more robust and provides a little more security in
1610 * adding extensions to VXLAN.
1616 if (!vxlan_set_mac(vxlan
, vs
, skb
, vni
))
1619 skb_reset_mac_header(skb
);
1620 skb
->dev
= vxlan
->dev
;
1621 skb
->pkt_type
= PACKET_HOST
;
1624 oiph
= skb_network_header(skb
);
1625 skb_reset_network_header(skb
);
1627 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1628 ++vxlan
->dev
->stats
.rx_frame_errors
;
1629 ++vxlan
->dev
->stats
.rx_errors
;
1635 if (unlikely(!(vxlan
->dev
->flags
& IFF_UP
))) {
1637 atomic_long_inc(&vxlan
->dev
->rx_dropped
);
1641 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1642 u64_stats_update_begin(&stats
->syncp
);
1643 stats
->rx_packets
++;
1644 stats
->rx_bytes
+= skb
->len
;
1645 u64_stats_update_end(&stats
->syncp
);
1647 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1654 /* Consume bad packet */
1659 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1661 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1662 struct arphdr
*parp
;
1665 struct neighbour
*n
;
1667 if (dev
->flags
& IFF_NOARP
)
1670 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1671 dev
->stats
.tx_dropped
++;
1674 parp
= arp_hdr(skb
);
1676 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1677 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1678 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1679 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1680 parp
->ar_hln
!= dev
->addr_len
||
1683 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1685 arpptr
+= dev
->addr_len
; /* sha */
1686 memcpy(&sip
, arpptr
, sizeof(sip
));
1687 arpptr
+= sizeof(sip
);
1688 arpptr
+= dev
->addr_len
; /* tha */
1689 memcpy(&tip
, arpptr
, sizeof(tip
));
1691 if (ipv4_is_loopback(tip
) ||
1692 ipv4_is_multicast(tip
))
1695 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1698 struct vxlan_fdb
*f
;
1699 struct sk_buff
*reply
;
1701 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1706 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1707 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1708 /* bridge-local neighbor */
1713 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1721 skb_reset_mac_header(reply
);
1722 __skb_pull(reply
, skb_network_offset(reply
));
1723 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1724 reply
->pkt_type
= PACKET_HOST
;
1726 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1727 dev
->stats
.rx_dropped
++;
1728 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
1729 union vxlan_addr ipa
= {
1730 .sin
.sin_addr
.s_addr
= tip
,
1731 .sin
.sin_family
= AF_INET
,
1734 vxlan_ip_miss(dev
, &ipa
);
1738 return NETDEV_TX_OK
;
1741 #if IS_ENABLED(CONFIG_IPV6)
1742 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1743 struct neighbour
*n
, bool isrouter
)
1745 struct net_device
*dev
= request
->dev
;
1746 struct sk_buff
*reply
;
1747 struct nd_msg
*ns
, *na
;
1748 struct ipv6hdr
*pip6
;
1750 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1754 if (dev
== NULL
|| !pskb_may_pull(request
, request
->len
))
1757 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1758 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1759 reply
= alloc_skb(len
, GFP_ATOMIC
);
1763 reply
->protocol
= htons(ETH_P_IPV6
);
1765 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1766 skb_push(reply
, sizeof(struct ethhdr
));
1767 skb_reset_mac_header(reply
);
1769 ns
= (struct nd_msg
*)(ipv6_hdr(request
) + 1);
1771 daddr
= eth_hdr(request
)->h_source
;
1772 ns_olen
= request
->len
- skb_network_offset(request
) -
1773 sizeof(struct ipv6hdr
) - sizeof(*ns
);
1774 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1775 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1776 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1781 /* Ethernet header */
1782 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1783 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1784 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1785 reply
->protocol
= htons(ETH_P_IPV6
);
1787 skb_pull(reply
, sizeof(struct ethhdr
));
1788 skb_reset_network_header(reply
);
1789 skb_put(reply
, sizeof(struct ipv6hdr
));
1793 pip6
= ipv6_hdr(reply
);
1794 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1796 pip6
->priority
= ipv6_hdr(request
)->priority
;
1797 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1798 pip6
->hop_limit
= 255;
1799 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1800 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1802 skb_pull(reply
, sizeof(struct ipv6hdr
));
1803 skb_reset_transport_header(reply
);
1805 /* Neighbor Advertisement */
1806 na
= skb_put_zero(reply
, sizeof(*na
) + na_olen
);
1807 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1808 na
->icmph
.icmp6_router
= isrouter
;
1809 na
->icmph
.icmp6_override
= 1;
1810 na
->icmph
.icmp6_solicited
= 1;
1811 na
->target
= ns
->target
;
1812 ether_addr_copy(&na
->opt
[2], n
->ha
);
1813 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1814 na
->opt
[1] = na_olen
>> 3;
1816 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1817 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1818 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1820 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1822 skb_push(reply
, sizeof(struct ipv6hdr
));
1824 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1829 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1831 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1832 const struct in6_addr
*daddr
;
1833 const struct ipv6hdr
*iphdr
;
1834 struct inet6_dev
*in6_dev
;
1835 struct neighbour
*n
;
1838 in6_dev
= __in6_dev_get(dev
);
1842 iphdr
= ipv6_hdr(skb
);
1843 daddr
= &iphdr
->daddr
;
1844 msg
= (struct nd_msg
*)(iphdr
+ 1);
1846 if (ipv6_addr_loopback(daddr
) ||
1847 ipv6_addr_is_multicast(&msg
->target
))
1850 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
1853 struct vxlan_fdb
*f
;
1854 struct sk_buff
*reply
;
1856 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1861 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1862 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1863 /* bridge-local neighbor */
1868 reply
= vxlan_na_create(skb
, n
,
1869 !!(f
? f
->flags
& NTF_ROUTER
: 0));
1876 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1877 dev
->stats
.rx_dropped
++;
1879 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
1880 union vxlan_addr ipa
= {
1881 .sin6
.sin6_addr
= msg
->target
,
1882 .sin6
.sin6_family
= AF_INET6
,
1885 vxlan_ip_miss(dev
, &ipa
);
1890 return NETDEV_TX_OK
;
1894 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
1896 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1897 struct neighbour
*n
;
1899 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
1903 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
1908 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
1911 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
1912 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
1913 union vxlan_addr ipa
= {
1914 .sin
.sin_addr
.s_addr
= pip
->daddr
,
1915 .sin
.sin_family
= AF_INET
,
1918 vxlan_ip_miss(dev
, &ipa
);
1924 #if IS_ENABLED(CONFIG_IPV6)
1927 struct ipv6hdr
*pip6
;
1929 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
1931 pip6
= ipv6_hdr(skb
);
1932 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
1933 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
1934 union vxlan_addr ipa
= {
1935 .sin6
.sin6_addr
= pip6
->daddr
,
1936 .sin6
.sin6_family
= AF_INET6
,
1939 vxlan_ip_miss(dev
, &ipa
);
1953 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
1955 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
1957 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
1966 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1967 struct vxlan_metadata
*md
)
1969 struct vxlanhdr_gbp
*gbp
;
1974 gbp
= (struct vxlanhdr_gbp
*)vxh
;
1975 vxh
->vx_flags
|= VXLAN_HF_GBP
;
1977 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
1978 gbp
->dont_learn
= 1;
1980 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
1981 gbp
->policy_applied
= 1;
1983 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
1986 static int vxlan_build_gpe_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
1989 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)vxh
;
1991 gpe
->np_applied
= 1;
1992 gpe
->next_protocol
= tun_p_from_eth_p(protocol
);
1993 if (!gpe
->next_protocol
)
1994 return -EPFNOSUPPORT
;
1998 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
1999 int iphdr_len
, __be32 vni
,
2000 struct vxlan_metadata
*md
, u32 vxflags
,
2003 struct vxlanhdr
*vxh
;
2006 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
2007 __be16 inner_protocol
= htons(ETH_P_TEB
);
2009 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
2010 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
2011 int csum_start
= skb_checksum_start_offset(skb
);
2013 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
2014 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
2015 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
2016 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
2017 type
|= SKB_GSO_TUNNEL_REMCSUM
;
2020 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
2021 + VXLAN_HLEN
+ iphdr_len
;
2023 /* Need space for new headers (invalidates iph ptr) */
2024 err
= skb_cow_head(skb
, min_headroom
);
2028 err
= iptunnel_handle_offloads(skb
, type
);
2032 vxh
= __skb_push(skb
, sizeof(*vxh
));
2033 vxh
->vx_flags
= VXLAN_HF_VNI
;
2034 vxh
->vx_vni
= vxlan_vni_field(vni
);
2036 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
2039 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
2040 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
2041 vxh
->vx_flags
|= VXLAN_HF_RCO
;
2043 if (!skb_is_gso(skb
)) {
2044 skb
->ip_summed
= CHECKSUM_NONE
;
2045 skb
->encapsulation
= 0;
2049 if (vxflags
& VXLAN_F_GBP
)
2050 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
2051 if (vxflags
& VXLAN_F_GPE
) {
2052 err
= vxlan_build_gpe_hdr(vxh
, vxflags
, skb
->protocol
);
2055 inner_protocol
= skb
->protocol
;
2058 skb_set_inner_protocol(skb
, inner_protocol
);
2062 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
, struct net_device
*dev
,
2063 struct vxlan_sock
*sock4
,
2064 struct sk_buff
*skb
, int oif
, u8 tos
,
2065 __be32 daddr
, __be32
*saddr
, __be16 dport
, __be16 sport
,
2066 struct dst_cache
*dst_cache
,
2067 const struct ip_tunnel_info
*info
)
2069 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
2070 struct rtable
*rt
= NULL
;
2074 return ERR_PTR(-EIO
);
2079 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
2084 memset(&fl4
, 0, sizeof(fl4
));
2085 fl4
.flowi4_oif
= oif
;
2086 fl4
.flowi4_tos
= RT_TOS(tos
);
2087 fl4
.flowi4_mark
= skb
->mark
;
2088 fl4
.flowi4_proto
= IPPROTO_UDP
;
2091 fl4
.fl4_dport
= dport
;
2092 fl4
.fl4_sport
= sport
;
2094 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
2095 if (likely(!IS_ERR(rt
))) {
2096 if (rt
->dst
.dev
== dev
) {
2097 netdev_dbg(dev
, "circular route to %pI4\n", &daddr
);
2099 return ERR_PTR(-ELOOP
);
2104 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
2106 netdev_dbg(dev
, "no route to %pI4\n", &daddr
);
2107 return ERR_PTR(-ENETUNREACH
);
2112 #if IS_ENABLED(CONFIG_IPV6)
2113 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
2114 struct net_device
*dev
,
2115 struct vxlan_sock
*sock6
,
2116 struct sk_buff
*skb
, int oif
, u8 tos
,
2118 const struct in6_addr
*daddr
,
2119 struct in6_addr
*saddr
,
2120 __be16 dport
, __be16 sport
,
2121 struct dst_cache
*dst_cache
,
2122 const struct ip_tunnel_info
*info
)
2124 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
2125 struct dst_entry
*ndst
;
2130 return ERR_PTR(-EIO
);
2135 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
2140 memset(&fl6
, 0, sizeof(fl6
));
2141 fl6
.flowi6_oif
= oif
;
2144 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
2145 fl6
.flowi6_mark
= skb
->mark
;
2146 fl6
.flowi6_proto
= IPPROTO_UDP
;
2147 fl6
.fl6_dport
= dport
;
2148 fl6
.fl6_sport
= sport
;
2150 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
2153 if (unlikely(err
< 0)) {
2154 netdev_dbg(dev
, "no route to %pI6\n", daddr
);
2155 return ERR_PTR(-ENETUNREACH
);
2158 if (unlikely(ndst
->dev
== dev
)) {
2159 netdev_dbg(dev
, "circular route to %pI6\n", daddr
);
2161 return ERR_PTR(-ELOOP
);
2166 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
2171 /* Bypass encapsulation if the destination is local */
2172 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
2173 struct vxlan_dev
*dst_vxlan
, __be32 vni
)
2175 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
2176 union vxlan_addr loopback
;
2177 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
2178 struct net_device
*dev
;
2181 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
2182 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
2183 skb
->pkt_type
= PACKET_HOST
;
2184 skb
->encapsulation
= 0;
2185 skb
->dev
= dst_vxlan
->dev
;
2186 __skb_pull(skb
, skb_network_offset(skb
));
2188 if (remote_ip
->sa
.sa_family
== AF_INET
) {
2189 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
2190 loopback
.sa
.sa_family
= AF_INET
;
2191 #if IS_ENABLED(CONFIG_IPV6)
2193 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
2194 loopback
.sa
.sa_family
= AF_INET6
;
2200 if (unlikely(!(dev
->flags
& IFF_UP
))) {
2205 if (dst_vxlan
->cfg
.flags
& VXLAN_F_LEARN
)
2206 vxlan_snoop(dev
, &loopback
, eth_hdr(skb
)->h_source
, 0, vni
);
2208 u64_stats_update_begin(&tx_stats
->syncp
);
2209 tx_stats
->tx_packets
++;
2210 tx_stats
->tx_bytes
+= len
;
2211 u64_stats_update_end(&tx_stats
->syncp
);
2213 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
2214 u64_stats_update_begin(&rx_stats
->syncp
);
2215 rx_stats
->rx_packets
++;
2216 rx_stats
->rx_bytes
+= len
;
2217 u64_stats_update_end(&rx_stats
->syncp
);
2220 dev
->stats
.rx_dropped
++;
2225 static int encap_bypass_if_local(struct sk_buff
*skb
, struct net_device
*dev
,
2226 struct vxlan_dev
*vxlan
,
2227 union vxlan_addr
*daddr
,
2228 __be16 dst_port
, int dst_ifindex
, __be32 vni
,
2229 struct dst_entry
*dst
,
2232 #if IS_ENABLED(CONFIG_IPV6)
2233 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2234 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2235 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2237 BUILD_BUG_ON(RTCF_LOCAL
!= RTF_LOCAL
);
2239 /* Bypass encapsulation if the destination is local */
2240 if (rt_flags
& RTCF_LOCAL
&&
2241 !(rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2242 struct vxlan_dev
*dst_vxlan
;
2245 dst_vxlan
= vxlan_find_vni(vxlan
->net
, dst_ifindex
, vni
,
2246 daddr
->sa
.sa_family
, dst_port
,
2249 dev
->stats
.tx_errors
++;
2254 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
, vni
);
2261 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
2262 __be32 default_vni
, struct vxlan_rdst
*rdst
,
2265 struct dst_cache
*dst_cache
;
2266 struct ip_tunnel_info
*info
;
2267 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2268 const struct iphdr
*old_iph
= ip_hdr(skb
);
2269 union vxlan_addr
*dst
;
2270 union vxlan_addr remote_ip
, local_ip
;
2271 struct vxlan_metadata _md
;
2272 struct vxlan_metadata
*md
= &_md
;
2273 __be16 src_port
= 0, dst_port
;
2274 struct dst_entry
*ndst
= NULL
;
2279 u32 flags
= vxlan
->cfg
.flags
;
2280 bool udp_sum
= false;
2281 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
2283 info
= skb_tunnel_info(skb
);
2286 dst
= &rdst
->remote_ip
;
2287 if (vxlan_addr_any(dst
)) {
2289 /* short-circuited back to local bridge */
2290 vxlan_encap_bypass(skb
, vxlan
, vxlan
, default_vni
);
2296 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
2297 vni
= (rdst
->remote_vni
) ? : default_vni
;
2298 ifindex
= rdst
->remote_ifindex
;
2299 local_ip
= vxlan
->cfg
.saddr
;
2300 dst_cache
= &rdst
->dst_cache
;
2301 md
->gbp
= skb
->mark
;
2302 ttl
= vxlan
->cfg
.ttl
;
2303 if (!ttl
&& vxlan_addr_multicast(dst
))
2306 tos
= vxlan
->cfg
.tos
;
2308 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
2310 if (dst
->sa
.sa_family
== AF_INET
)
2311 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
2313 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2314 label
= vxlan
->cfg
.label
;
2317 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2321 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
2322 if (remote_ip
.sa
.sa_family
== AF_INET
) {
2323 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
2324 local_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.src
;
2326 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
2327 local_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.src
;
2330 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2331 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2333 dst_cache
= &info
->dst_cache
;
2334 if (info
->options_len
)
2335 md
= ip_tunnel_info_opts(info
);
2336 ttl
= info
->key
.ttl
;
2337 tos
= info
->key
.tos
;
2338 label
= info
->key
.label
;
2339 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
2341 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2342 vxlan
->cfg
.port_max
, true);
2345 if (dst
->sa
.sa_family
== AF_INET
) {
2346 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2350 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, ifindex
, tos
,
2351 dst
->sin
.sin_addr
.s_addr
,
2352 &local_ip
.sin
.sin_addr
.s_addr
,
2360 if (fan_has_map(&vxlan
->fan
) && rt
->rt_flags
& RTCF_LOCAL
) {
2361 netdev_dbg(dev
, "discard fan to localhost %pI4\n",
2362 &dst
->sin
.sin_addr
.s_addr
);
2367 /* Bypass encapsulation if the destination is local */
2369 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2370 dst_port
, ifindex
, vni
,
2371 &rt
->dst
, rt
->rt_flags
);
2374 } else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
) {
2380 int mtu
= dst_mtu(ndst
) - VXLAN_HEADROOM
;
2382 skb_dst_update_pmtu(skb
, mtu
);
2385 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2386 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2387 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct iphdr
),
2388 vni
, md
, flags
, udp_sum
);
2392 udp_tunnel_xmit_skb(rt
, sock4
->sock
->sk
, skb
, local_ip
.sin
.sin_addr
.s_addr
,
2393 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2394 src_port
, dst_port
, xnet
, !udp_sum
);
2395 #if IS_ENABLED(CONFIG_IPV6)
2397 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2399 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, ifindex
, tos
,
2400 label
, &dst
->sin6
.sin6_addr
,
2401 &local_ip
.sin6
.sin6_addr
,
2405 err
= PTR_ERR(ndst
);
2411 u32 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2413 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2414 dst_port
, ifindex
, vni
,
2421 int mtu
= dst_mtu(ndst
) - VXLAN6_HEADROOM
;
2423 skb_dst_update_pmtu(skb
, mtu
);
2426 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2427 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2428 skb_scrub_packet(skb
, xnet
);
2429 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2430 vni
, md
, flags
, udp_sum
);
2434 udp_tunnel6_xmit_skb(ndst
, sock6
->sock
->sk
, skb
, dev
,
2435 &local_ip
.sin6
.sin6_addr
,
2436 &dst
->sin6
.sin6_addr
, tos
, ttl
,
2437 label
, src_port
, dst_port
, !udp_sum
);
2445 dev
->stats
.tx_dropped
++;
2452 dev
->stats
.collisions
++;
2453 else if (err
== -ENETUNREACH
)
2454 dev
->stats
.tx_carrier_errors
++;
2456 dev
->stats
.tx_errors
++;
2461 /* Transmit local packets over Vxlan
2463 * Outer IP header inherits ECN and DF from inner header.
2464 * Outer UDP destination is the VXLAN assigned port.
2465 * source port is based on hash of flow
2467 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2469 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2470 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2471 const struct ip_tunnel_info
*info
;
2472 bool did_rsc
= false;
2473 struct vxlan_fdb
*f
;
2477 info
= skb_tunnel_info(skb
);
2479 skb_reset_mac_header(skb
);
2481 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
2482 if (info
&& info
->mode
& IP_TUNNEL_INFO_BRIDGE
&&
2483 info
->mode
& IP_TUNNEL_INFO_TX
) {
2484 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2486 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2487 vxlan_xmit_one(skb
, dev
, vni
, NULL
, false);
2490 return NETDEV_TX_OK
;
2494 if (vxlan
->cfg
.flags
& VXLAN_F_PROXY
) {
2496 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2497 return arp_reduce(dev
, skb
, vni
);
2498 #if IS_ENABLED(CONFIG_IPV6)
2499 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2500 pskb_may_pull(skb
, sizeof(struct ipv6hdr
) +
2501 sizeof(struct nd_msg
)) &&
2502 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2503 struct nd_msg
*m
= (struct nd_msg
*)(ipv6_hdr(skb
) + 1);
2505 if (m
->icmph
.icmp6_code
== 0 &&
2506 m
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2507 return neigh_reduce(dev
, skb
, vni
);
2512 if (fan_has_map(&vxlan
->fan
)) {
2513 struct vxlan_rdst fan_rdst
;
2515 netdev_dbg(vxlan
->dev
, "vxlan_xmit p %x d %pM\n",
2516 eth
->h_proto
, eth
->h_dest
);
2517 if (vxlan_fan_build_rdst(vxlan
, skb
, &fan_rdst
)) {
2518 dev
->stats
.tx_dropped
++;
2520 return NETDEV_TX_OK
;
2522 vxlan_xmit_one(skb
, dev
, vni
, &fan_rdst
, 0);
2523 return NETDEV_TX_OK
;
2527 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2530 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->cfg
.flags
& VXLAN_F_RSC
) &&
2531 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2532 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2533 did_rsc
= route_shortcircuit(dev
, skb
);
2535 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2539 f
= vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2541 if ((vxlan
->cfg
.flags
& VXLAN_F_L2MISS
) &&
2542 !is_multicast_ether_addr(eth
->h_dest
))
2543 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2545 dev
->stats
.tx_dropped
++;
2547 return NETDEV_TX_OK
;
2551 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2552 struct sk_buff
*skb1
;
2558 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2560 vxlan_xmit_one(skb1
, dev
, vni
, rdst
, did_rsc
);
2564 vxlan_xmit_one(skb
, dev
, vni
, fdst
, did_rsc
);
2567 return NETDEV_TX_OK
;
2570 /* Walk the forwarding table and purge stale entries */
2571 static void vxlan_cleanup(struct timer_list
*t
)
2573 struct vxlan_dev
*vxlan
= from_timer(vxlan
, t
, age_timer
);
2574 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2577 if (!netif_running(vxlan
->dev
))
2580 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2581 struct hlist_node
*p
, *n
;
2583 spin_lock_bh(&vxlan
->hash_lock
);
2584 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2586 = container_of(p
, struct vxlan_fdb
, hlist
);
2587 unsigned long timeout
;
2589 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
2592 if (f
->flags
& NTF_EXT_LEARNED
)
2595 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2596 if (time_before_eq(timeout
, jiffies
)) {
2597 netdev_dbg(vxlan
->dev
,
2598 "garbage collect %pM\n",
2600 f
->state
= NUD_STALE
;
2601 vxlan_fdb_destroy(vxlan
, f
, true);
2602 } else if (time_before(timeout
, next_timer
))
2603 next_timer
= timeout
;
2605 spin_unlock_bh(&vxlan
->hash_lock
);
2608 mod_timer(&vxlan
->age_timer
, next_timer
);
2611 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
)
2613 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2615 spin_lock(&vn
->sock_lock
);
2616 hlist_del_init_rcu(&vxlan
->hlist4
.hlist
);
2617 #if IS_ENABLED(CONFIG_IPV6)
2618 hlist_del_init_rcu(&vxlan
->hlist6
.hlist
);
2620 spin_unlock(&vn
->sock_lock
);
2623 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
,
2624 struct vxlan_dev_node
*node
)
2626 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2627 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2629 node
->vxlan
= vxlan
;
2630 spin_lock(&vn
->sock_lock
);
2631 hlist_add_head_rcu(&node
->hlist
, vni_head(vs
, vni
));
2632 spin_unlock(&vn
->sock_lock
);
2635 /* Setup stats when device is created */
2636 static int vxlan_init(struct net_device
*dev
)
2638 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2645 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
, __be32 vni
)
2647 struct vxlan_fdb
*f
;
2649 spin_lock_bh(&vxlan
->hash_lock
);
2650 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2652 vxlan_fdb_destroy(vxlan
, f
, true);
2653 spin_unlock_bh(&vxlan
->hash_lock
);
2656 static void vxlan_uninit(struct net_device
*dev
)
2658 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2660 gro_cells_destroy(&vxlan
->gro_cells
);
2662 vxlan_fdb_delete_default(vxlan
, vxlan
->cfg
.vni
);
2664 free_percpu(dev
->tstats
);
2667 /* Start ageing timer and join group when device is brought up */
2668 static int vxlan_open(struct net_device
*dev
)
2670 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2673 ret
= vxlan_sock_add(vxlan
);
2677 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2678 ret
= vxlan_igmp_join(vxlan
);
2679 if (ret
== -EADDRINUSE
)
2682 vxlan_sock_release(vxlan
);
2687 if (vxlan
->cfg
.age_interval
)
2688 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2693 /* Purge the forwarding table */
2694 static void vxlan_flush(struct vxlan_dev
*vxlan
, bool do_all
)
2698 spin_lock_bh(&vxlan
->hash_lock
);
2699 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2700 struct hlist_node
*p
, *n
;
2701 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2703 = container_of(p
, struct vxlan_fdb
, hlist
);
2704 if (!do_all
&& (f
->state
& (NUD_PERMANENT
| NUD_NOARP
)))
2706 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2707 if (!is_zero_ether_addr(f
->eth_addr
))
2708 vxlan_fdb_destroy(vxlan
, f
, true);
2711 spin_unlock_bh(&vxlan
->hash_lock
);
2714 /* Cleanup timer and forwarding table on shutdown */
2715 static int vxlan_stop(struct net_device
*dev
)
2717 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2718 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2721 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2722 !vxlan_group_used(vn
, vxlan
))
2723 ret
= vxlan_igmp_leave(vxlan
);
2725 del_timer_sync(&vxlan
->age_timer
);
2727 vxlan_flush(vxlan
, false);
2728 vxlan_sock_release(vxlan
);
2733 /* Stub, nothing needs to be done. */
2734 static void vxlan_set_multicast_list(struct net_device
*dev
)
2738 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2740 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2741 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2742 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2743 dst
->remote_ifindex
);
2744 bool use_ipv6
= !!(vxlan
->cfg
.flags
& VXLAN_F_IPV6
);
2746 /* This check is different than dev->max_mtu, because it looks at
2747 * the lowerdev->mtu, rather than the static dev->max_mtu
2750 int max_mtu
= lowerdev
->mtu
-
2751 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2752 if (new_mtu
> max_mtu
)
2760 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2762 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2763 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2764 __be16 sport
, dport
;
2766 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2767 vxlan
->cfg
.port_max
, true);
2768 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2770 if (ip_tunnel_info_af(info
) == AF_INET
) {
2771 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2774 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, 0, info
->key
.tos
,
2775 info
->key
.u
.ipv4
.dst
,
2776 &info
->key
.u
.ipv4
.src
, dport
, sport
,
2777 &info
->dst_cache
, info
);
2782 #if IS_ENABLED(CONFIG_IPV6)
2783 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2784 struct dst_entry
*ndst
;
2786 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, 0, info
->key
.tos
,
2787 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2788 &info
->key
.u
.ipv6
.src
, dport
, sport
,
2789 &info
->dst_cache
, info
);
2791 return PTR_ERR(ndst
);
2793 #else /* !CONFIG_IPV6 */
2794 return -EPFNOSUPPORT
;
2797 info
->key
.tp_src
= sport
;
2798 info
->key
.tp_dst
= dport
;
2802 static const struct net_device_ops vxlan_netdev_ether_ops
= {
2803 .ndo_init
= vxlan_init
,
2804 .ndo_uninit
= vxlan_uninit
,
2805 .ndo_open
= vxlan_open
,
2806 .ndo_stop
= vxlan_stop
,
2807 .ndo_start_xmit
= vxlan_xmit
,
2808 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2809 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2810 .ndo_change_mtu
= vxlan_change_mtu
,
2811 .ndo_validate_addr
= eth_validate_addr
,
2812 .ndo_set_mac_address
= eth_mac_addr
,
2813 .ndo_fdb_add
= vxlan_fdb_add
,
2814 .ndo_fdb_del
= vxlan_fdb_delete
,
2815 .ndo_fdb_dump
= vxlan_fdb_dump
,
2816 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2819 static const struct net_device_ops vxlan_netdev_raw_ops
= {
2820 .ndo_init
= vxlan_init
,
2821 .ndo_uninit
= vxlan_uninit
,
2822 .ndo_open
= vxlan_open
,
2823 .ndo_stop
= vxlan_stop
,
2824 .ndo_start_xmit
= vxlan_xmit
,
2825 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2826 .ndo_change_mtu
= vxlan_change_mtu
,
2827 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2830 /* Info for udev, that this is a virtual tunnel endpoint */
2831 static struct device_type vxlan_type
= {
2835 /* Calls the ndo_udp_tunnel_add of the caller in order to
2836 * supply the listening VXLAN udp ports. Callers are expected
2837 * to implement the ndo_udp_tunnel_add.
2839 static void vxlan_offload_rx_ports(struct net_device
*dev
, bool push
)
2841 struct vxlan_sock
*vs
;
2842 struct net
*net
= dev_net(dev
);
2843 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2846 spin_lock(&vn
->sock_lock
);
2847 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2848 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2849 unsigned short type
;
2851 if (vs
->flags
& VXLAN_F_GPE
)
2852 type
= UDP_TUNNEL_TYPE_VXLAN_GPE
;
2854 type
= UDP_TUNNEL_TYPE_VXLAN
;
2857 udp_tunnel_push_rx_port(dev
, vs
->sock
, type
);
2859 udp_tunnel_drop_rx_port(dev
, vs
->sock
, type
);
2862 spin_unlock(&vn
->sock_lock
);
2865 /* Initialize the device structure. */
2866 static void vxlan_setup(struct net_device
*dev
)
2868 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2871 eth_hw_addr_random(dev
);
2874 dev
->needs_free_netdev
= true;
2875 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
2877 dev
->features
|= NETIF_F_LLTX
;
2878 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
2879 dev
->features
|= NETIF_F_RXCSUM
;
2880 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
2882 dev
->vlan_features
= dev
->features
;
2883 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
2884 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
2885 netif_keep_dst(dev
);
2886 dev
->priv_flags
|= IFF_NO_QUEUE
;
2888 /* MTU range: 68 - 65535 */
2889 dev
->min_mtu
= ETH_MIN_MTU
;
2890 dev
->max_mtu
= ETH_MAX_MTU
;
2892 INIT_LIST_HEAD(&vxlan
->next
);
2893 spin_lock_init(&vxlan
->hash_lock
);
2895 timer_setup(&vxlan
->age_timer
, vxlan_cleanup
, TIMER_DEFERRABLE
);
2899 gro_cells_init(&vxlan
->gro_cells
, dev
);
2901 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
2902 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
2904 INIT_LIST_HEAD(&vxlan
->fan
.fan_maps
);
2907 static void vxlan_ether_setup(struct net_device
*dev
)
2909 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
2910 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
2911 dev
->netdev_ops
= &vxlan_netdev_ether_ops
;
2914 static void vxlan_raw_setup(struct net_device
*dev
)
2916 dev
->header_ops
= NULL
;
2917 dev
->type
= ARPHRD_NONE
;
2918 dev
->hard_header_len
= 0;
2920 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
2921 dev
->netdev_ops
= &vxlan_netdev_raw_ops
;
2924 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
2925 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
2926 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
2927 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2928 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
2929 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
2930 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
2931 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
2932 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
2933 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
2934 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
2935 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
2936 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
2937 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
2938 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
2939 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
2940 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
2941 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
2942 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
2943 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
2944 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
2945 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
2946 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
2947 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
2948 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
2949 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
2950 [IFLA_VXLAN_GPE
] = { .type
= NLA_FLAG
, },
2951 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
2954 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
2955 struct netlink_ext_ack
*extack
)
2957 if (tb
[IFLA_ADDRESS
]) {
2958 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
2959 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
2960 "Provided link layer address is not Ethernet");
2964 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
2965 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
2966 "Provided Ethernet address is not unicast");
2967 return -EADDRNOTAVAIL
;
2972 u32 mtu
= nla_get_u32(tb
[IFLA_MTU
]);
2974 if (mtu
< ETH_MIN_MTU
|| mtu
> ETH_MAX_MTU
) {
2975 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_MTU
],
2976 "MTU must be between 68 and 65535");
2982 NL_SET_ERR_MSG(extack
,
2983 "Required attributes not provided to perform the operation");
2987 if (data
[IFLA_VXLAN_ID
]) {
2988 u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
2990 if (id
>= VXLAN_N_VID
) {
2991 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_ID
],
2992 "VXLAN ID must be lower than 16777216");
2997 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
2998 const struct ifla_vxlan_port_range
*p
2999 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3001 if (ntohs(p
->high
) < ntohs(p
->low
)) {
3002 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_PORT_RANGE
],
3003 "Invalid source port range");
3011 static void vxlan_get_drvinfo(struct net_device
*netdev
,
3012 struct ethtool_drvinfo
*drvinfo
)
3014 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
3015 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
3018 static const struct ethtool_ops vxlan_ethtool_ops
= {
3019 .get_drvinfo
= vxlan_get_drvinfo
,
3020 .get_link
= ethtool_op_get_link
,
3023 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
3024 __be16 port
, u32 flags
)
3026 struct socket
*sock
;
3027 struct udp_port_cfg udp_conf
;
3030 memset(&udp_conf
, 0, sizeof(udp_conf
));
3033 udp_conf
.family
= AF_INET6
;
3034 udp_conf
.use_udp6_rx_checksums
=
3035 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
3036 udp_conf
.ipv6_v6only
= 1;
3038 udp_conf
.family
= AF_INET
;
3041 udp_conf
.local_udp_port
= port
;
3043 /* Open UDP socket */
3044 err
= udp_sock_create(net
, &udp_conf
, &sock
);
3046 return ERR_PTR(err
);
3051 /* Create new listen socket if needed */
3052 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
3053 __be16 port
, u32 flags
)
3055 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3056 struct vxlan_sock
*vs
;
3057 struct socket
*sock
;
3059 struct udp_tunnel_sock_cfg tunnel_cfg
;
3061 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
3063 return ERR_PTR(-ENOMEM
);
3065 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
3066 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
3068 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
);
3071 return ERR_CAST(sock
);
3075 refcount_set(&vs
->refcnt
, 1);
3076 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
3078 spin_lock(&vn
->sock_lock
);
3079 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
3080 udp_tunnel_notify_add_rx_port(sock
,
3081 (vs
->flags
& VXLAN_F_GPE
) ?
3082 UDP_TUNNEL_TYPE_VXLAN_GPE
:
3083 UDP_TUNNEL_TYPE_VXLAN
);
3084 spin_unlock(&vn
->sock_lock
);
3086 /* Mark socket as an encapsulation socket. */
3087 memset(&tunnel_cfg
, 0, sizeof(tunnel_cfg
));
3088 tunnel_cfg
.sk_user_data
= vs
;
3089 tunnel_cfg
.encap_type
= 1;
3090 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
3091 tunnel_cfg
.encap_destroy
= NULL
;
3092 tunnel_cfg
.gro_receive
= vxlan_gro_receive
;
3093 tunnel_cfg
.gro_complete
= vxlan_gro_complete
;
3095 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
3100 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
3102 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
3103 struct vxlan_sock
*vs
= NULL
;
3104 struct vxlan_dev_node
*node
;
3106 if (!vxlan
->cfg
.no_share
) {
3107 spin_lock(&vn
->sock_lock
);
3108 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
3109 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
);
3110 if (vs
&& !refcount_inc_not_zero(&vs
->refcnt
)) {
3111 spin_unlock(&vn
->sock_lock
);
3114 spin_unlock(&vn
->sock_lock
);
3117 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
3118 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
);
3121 #if IS_ENABLED(CONFIG_IPV6)
3123 rcu_assign_pointer(vxlan
->vn6_sock
, vs
);
3124 node
= &vxlan
->hlist6
;
3128 rcu_assign_pointer(vxlan
->vn4_sock
, vs
);
3129 node
= &vxlan
->hlist4
;
3131 vxlan_vs_add_dev(vs
, vxlan
, node
);
3135 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
3137 bool metadata
= vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
;
3138 bool ipv6
= vxlan
->cfg
.flags
& VXLAN_F_IPV6
|| metadata
;
3139 bool ipv4
= !ipv6
|| metadata
;
3142 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
3143 #if IS_ENABLED(CONFIG_IPV6)
3144 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
3146 ret
= __vxlan_sock_add(vxlan
, true);
3147 if (ret
< 0 && ret
!= -EAFNOSUPPORT
)
3152 ret
= __vxlan_sock_add(vxlan
, false);
3154 vxlan_sock_release(vxlan
);
3158 static int vxlan_config_validate(struct net
*src_net
, struct vxlan_config
*conf
,
3159 struct net_device
**lower
,
3160 struct vxlan_dev
*old
,
3161 struct netlink_ext_ack
*extack
)
3163 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
3164 struct vxlan_dev
*tmp
;
3165 bool use_ipv6
= false;
3167 if (conf
->flags
& VXLAN_F_GPE
) {
3168 /* For now, allow GPE only together with
3169 * COLLECT_METADATA. This can be relaxed later; in such
3170 * case, the other side of the PtP link will have to be
3173 if ((conf
->flags
& ~VXLAN_F_ALLOWED_GPE
) ||
3174 !(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
3175 NL_SET_ERR_MSG(extack
,
3176 "VXLAN GPE does not support this combination of attributes");
3181 if (!conf
->remote_ip
.sa
.sa_family
&& !conf
->saddr
.sa
.sa_family
) {
3182 /* Unless IPv6 is explicitly requested, assume IPv4 */
3183 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
3184 conf
->saddr
.sa
.sa_family
= AF_INET
;
3185 } else if (!conf
->remote_ip
.sa
.sa_family
) {
3186 conf
->remote_ip
.sa
.sa_family
= conf
->saddr
.sa
.sa_family
;
3187 } else if (!conf
->saddr
.sa
.sa_family
) {
3188 conf
->saddr
.sa
.sa_family
= conf
->remote_ip
.sa
.sa_family
;
3191 if (conf
->saddr
.sa
.sa_family
!= conf
->remote_ip
.sa
.sa_family
) {
3192 NL_SET_ERR_MSG(extack
,
3193 "Local and remote address must be from the same family");
3197 if (vxlan_addr_multicast(&conf
->saddr
)) {
3198 NL_SET_ERR_MSG(extack
, "Local address cannot be multicast");
3202 if (conf
->saddr
.sa
.sa_family
== AF_INET6
) {
3203 if (!IS_ENABLED(CONFIG_IPV6
)) {
3204 NL_SET_ERR_MSG(extack
,
3205 "IPv6 support not enabled in the kernel");
3206 return -EPFNOSUPPORT
;
3209 conf
->flags
|= VXLAN_F_IPV6
;
3211 if (!(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
3213 ipv6_addr_type(&conf
->saddr
.sin6
.sin6_addr
);
3215 ipv6_addr_type(&conf
->remote_ip
.sin6
.sin6_addr
);
3217 if (local_type
& IPV6_ADDR_LINKLOCAL
) {
3218 if (!(remote_type
& IPV6_ADDR_LINKLOCAL
) &&
3219 (remote_type
!= IPV6_ADDR_ANY
)) {
3220 NL_SET_ERR_MSG(extack
,
3221 "Invalid combination of local and remote address scopes");
3225 conf
->flags
|= VXLAN_F_IPV6_LINKLOCAL
;
3228 (IPV6_ADDR_UNICAST
| IPV6_ADDR_LINKLOCAL
)) {
3229 NL_SET_ERR_MSG(extack
,
3230 "Invalid combination of local and remote address scopes");
3234 conf
->flags
&= ~VXLAN_F_IPV6_LINKLOCAL
;
3239 if (conf
->label
&& !use_ipv6
) {
3240 NL_SET_ERR_MSG(extack
,
3241 "Label attribute only applies to IPv6 VXLAN devices");
3245 if (conf
->remote_ifindex
) {
3246 struct net_device
*lowerdev
;
3248 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
3250 NL_SET_ERR_MSG(extack
,
3251 "Invalid local interface, device not found");
3255 #if IS_ENABLED(CONFIG_IPV6)
3257 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
3258 if (idev
&& idev
->cnf
.disable_ipv6
) {
3259 NL_SET_ERR_MSG(extack
,
3260 "IPv6 support disabled by administrator");
3268 if (vxlan_addr_multicast(&conf
->remote_ip
)) {
3269 NL_SET_ERR_MSG(extack
,
3270 "Local interface required for multicast remote destination");
3275 #if IS_ENABLED(CONFIG_IPV6)
3276 if (conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) {
3277 NL_SET_ERR_MSG(extack
,
3278 "Local interface required for link-local local/remote addresses");
3286 if (!conf
->dst_port
) {
3287 if (conf
->flags
& VXLAN_F_GPE
)
3288 conf
->dst_port
= htons(4790); /* IANA VXLAN-GPE port */
3290 conf
->dst_port
= htons(vxlan_port
);
3293 if (!conf
->age_interval
)
3294 conf
->age_interval
= FDB_AGE_DEFAULT
;
3296 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
3300 if (tmp
->cfg
.vni
!= conf
->vni
)
3302 if (tmp
->cfg
.dst_port
!= conf
->dst_port
)
3304 if ((tmp
->cfg
.flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)) !=
3305 (conf
->flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)))
3308 if ((conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
3309 tmp
->cfg
.remote_ifindex
!= conf
->remote_ifindex
)
3312 NL_SET_ERR_MSG(extack
,
3313 "A VXLAN device with the specified VNI already exists");
3320 static void vxlan_config_apply(struct net_device
*dev
,
3321 struct vxlan_config
*conf
,
3322 struct net_device
*lowerdev
,
3323 struct net
*src_net
,
3326 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3327 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3328 unsigned short needed_headroom
= ETH_HLEN
;
3329 bool use_ipv6
= !!(conf
->flags
& VXLAN_F_IPV6
);
3330 int max_mtu
= ETH_MAX_MTU
;
3333 if (conf
->flags
& VXLAN_F_GPE
)
3334 vxlan_raw_setup(dev
);
3336 vxlan_ether_setup(dev
);
3339 dev
->mtu
= conf
->mtu
;
3341 vxlan
->net
= src_net
;
3344 dst
->remote_vni
= conf
->vni
;
3346 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
3349 dst
->remote_ifindex
= conf
->remote_ifindex
;
3351 dev
->gso_max_size
= lowerdev
->gso_max_size
;
3352 dev
->gso_max_segs
= lowerdev
->gso_max_segs
;
3354 needed_headroom
= lowerdev
->hard_header_len
;
3356 max_mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
:
3358 if (max_mtu
< ETH_MIN_MTU
)
3359 max_mtu
= ETH_MIN_MTU
;
3361 if (!changelink
&& !conf
->mtu
)
3365 if (dev
->mtu
> max_mtu
)
3368 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
3369 needed_headroom
+= VXLAN6_HEADROOM
;
3371 needed_headroom
+= VXLAN_HEADROOM
;
3372 dev
->needed_headroom
= needed_headroom
;
3374 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
3377 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
3378 struct vxlan_config
*conf
, bool changelink
,
3379 struct netlink_ext_ack
*extack
)
3381 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3382 struct net_device
*lowerdev
;
3385 ret
= vxlan_config_validate(src_net
, conf
, &lowerdev
, vxlan
, extack
);
3389 vxlan_config_apply(dev
, conf
, lowerdev
, src_net
, changelink
);
3394 static int __vxlan_dev_create(struct net
*net
, struct net_device
*dev
,
3395 struct vxlan_config
*conf
,
3396 struct netlink_ext_ack
*extack
)
3398 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3399 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3400 struct vxlan_fdb
*f
= NULL
;
3403 err
= vxlan_dev_configure(net
, dev
, conf
, false, extack
);
3407 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
3409 /* create an fdb entry for a valid default destination */
3410 if (!vxlan_addr_any(&vxlan
->default_dst
.remote_ip
)) {
3411 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3412 &vxlan
->default_dst
.remote_ip
,
3413 NUD_REACHABLE
| NUD_PERMANENT
,
3414 vxlan
->cfg
.dst_port
,
3415 vxlan
->default_dst
.remote_vni
,
3416 vxlan
->default_dst
.remote_vni
,
3417 vxlan
->default_dst
.remote_ifindex
,
3423 err
= register_netdevice(dev
);
3427 err
= rtnl_configure_link(dev
, NULL
);
3429 unregister_netdevice(dev
);
3433 /* notify default fdb entry */
3435 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_NEWNEIGH
);
3437 list_add(&vxlan
->next
, &vn
->vxlan_list
);
3441 vxlan_fdb_destroy(vxlan
, f
, false);
3445 static int vxlan_nl2conf(struct nlattr
*tb
[], struct nlattr
*data
[],
3446 struct net_device
*dev
, struct vxlan_config
*conf
,
3449 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3452 memset(conf
, 0, sizeof(*conf
));
3454 /* if changelink operation, start with old existing cfg */
3456 memcpy(conf
, &vxlan
->cfg
, sizeof(*conf
));
3458 if (data
[IFLA_VXLAN_ID
]) {
3459 __be32 vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3461 if (changelink
&& (vni
!= conf
->vni
))
3463 conf
->vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3466 if (data
[IFLA_VXLAN_GROUP
]) {
3467 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET
))
3470 conf
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
3471 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
3472 } else if (data
[IFLA_VXLAN_GROUP6
]) {
3473 if (!IS_ENABLED(CONFIG_IPV6
))
3474 return -EPFNOSUPPORT
;
3476 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET6
))
3479 conf
->remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
3480 conf
->remote_ip
.sa
.sa_family
= AF_INET6
;
3483 if (data
[IFLA_VXLAN_FAN_MAP
]) {
3484 err
= vxlan_parse_fan_map(data
, vxlan
);
3489 if (data
[IFLA_VXLAN_LOCAL
]) {
3490 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET
))
3493 conf
->saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
3494 conf
->saddr
.sa
.sa_family
= AF_INET
;
3495 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
3496 if (!IS_ENABLED(CONFIG_IPV6
))
3497 return -EPFNOSUPPORT
;
3499 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET6
))
3502 /* TODO: respect scope id */
3503 conf
->saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
3504 conf
->saddr
.sa
.sa_family
= AF_INET6
;
3507 if (data
[IFLA_VXLAN_LINK
])
3508 conf
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
3510 if (data
[IFLA_VXLAN_TOS
])
3511 conf
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
3513 if (data
[IFLA_VXLAN_TTL
])
3514 conf
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
3516 if (data
[IFLA_VXLAN_LABEL
])
3517 conf
->label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
3518 IPV6_FLOWLABEL_MASK
;
3520 if (data
[IFLA_VXLAN_LEARNING
]) {
3521 if (nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
3522 conf
->flags
|= VXLAN_F_LEARN
;
3524 conf
->flags
&= ~VXLAN_F_LEARN
;
3525 } else if (!changelink
) {
3526 /* default to learn on a new device */
3527 conf
->flags
|= VXLAN_F_LEARN
;
3530 if (data
[IFLA_VXLAN_AGEING
]) {
3533 conf
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
3536 if (data
[IFLA_VXLAN_PROXY
]) {
3539 if (nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
3540 conf
->flags
|= VXLAN_F_PROXY
;
3543 if (data
[IFLA_VXLAN_RSC
]) {
3546 if (nla_get_u8(data
[IFLA_VXLAN_RSC
]))
3547 conf
->flags
|= VXLAN_F_RSC
;
3550 if (data
[IFLA_VXLAN_L2MISS
]) {
3553 if (nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
3554 conf
->flags
|= VXLAN_F_L2MISS
;
3557 if (data
[IFLA_VXLAN_L3MISS
]) {
3560 if (nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
3561 conf
->flags
|= VXLAN_F_L3MISS
;
3564 if (data
[IFLA_VXLAN_LIMIT
]) {
3567 conf
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
3570 if (data
[IFLA_VXLAN_COLLECT_METADATA
]) {
3573 if (nla_get_u8(data
[IFLA_VXLAN_COLLECT_METADATA
]))
3574 conf
->flags
|= VXLAN_F_COLLECT_METADATA
;
3577 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3579 const struct ifla_vxlan_port_range
*p
3580 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3581 conf
->port_min
= ntohs(p
->low
);
3582 conf
->port_max
= ntohs(p
->high
);
3588 if (data
[IFLA_VXLAN_PORT
]) {
3591 conf
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
3594 if (data
[IFLA_VXLAN_UDP_CSUM
]) {
3597 if (!nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
3598 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
3601 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]) {
3604 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]))
3605 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_TX
;
3608 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]) {
3611 if (nla_get_u8(data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]))
3612 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM6_RX
;
3615 if (data
[IFLA_VXLAN_REMCSUM_TX
]) {
3618 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_TX
]))
3619 conf
->flags
|= VXLAN_F_REMCSUM_TX
;
3622 if (data
[IFLA_VXLAN_REMCSUM_RX
]) {
3625 if (nla_get_u8(data
[IFLA_VXLAN_REMCSUM_RX
]))
3626 conf
->flags
|= VXLAN_F_REMCSUM_RX
;
3629 if (data
[IFLA_VXLAN_GBP
]) {
3632 conf
->flags
|= VXLAN_F_GBP
;
3635 if (data
[IFLA_VXLAN_GPE
]) {
3638 conf
->flags
|= VXLAN_F_GPE
;
3641 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
]) {
3644 conf
->flags
|= VXLAN_F_REMCSUM_NOPARTIAL
;
3650 conf
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3656 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
3657 struct nlattr
*tb
[], struct nlattr
*data
[],
3658 struct netlink_ext_ack
*extack
)
3660 struct vxlan_config conf
;
3663 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, false);
3667 return __vxlan_dev_create(src_net
, dev
, &conf
, extack
);
3670 static int vxlan_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
3671 struct nlattr
*data
[],
3672 struct netlink_ext_ack
*extack
)
3674 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3675 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3676 struct vxlan_rdst old_dst
;
3677 struct vxlan_config conf
;
3678 struct vxlan_fdb
*f
= NULL
;
3681 err
= vxlan_nl2conf(tb
, data
,
3686 memcpy(&old_dst
, dst
, sizeof(struct vxlan_rdst
));
3688 err
= vxlan_dev_configure(vxlan
->net
, dev
, &conf
, true, extack
);
3692 /* handle default dst entry */
3693 if (!vxlan_addr_equal(&dst
->remote_ip
, &old_dst
.remote_ip
)) {
3694 spin_lock_bh(&vxlan
->hash_lock
);
3695 if (!vxlan_addr_any(&old_dst
.remote_ip
))
3696 __vxlan_fdb_delete(vxlan
, all_zeros_mac
,
3698 vxlan
->cfg
.dst_port
,
3701 old_dst
.remote_ifindex
, 0);
3703 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3704 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3706 NUD_REACHABLE
| NUD_PERMANENT
,
3707 vxlan
->cfg
.dst_port
,
3710 dst
->remote_ifindex
,
3713 spin_unlock_bh(&vxlan
->hash_lock
);
3716 vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_NEWNEIGH
);
3718 spin_unlock_bh(&vxlan
->hash_lock
);
3724 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
3726 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3728 vxlan_flush(vxlan
, true);
3730 list_del(&vxlan
->next
);
3731 unregister_netdevice_queue(dev
, head
);
3734 static size_t vxlan_get_size(const struct net_device
*dev
)
3737 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
3738 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
3739 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
3740 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
3741 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
3742 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
3743 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
3744 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
3745 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
3746 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
3747 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
3748 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
3749 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
3750 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
3751 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
3752 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
3753 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
3754 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
3755 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3756 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3757 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
3758 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
3759 nla_total_size(sizeof(struct ip_fan_map
) * 256) +
3763 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
3765 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3766 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3767 struct ifla_vxlan_port_range ports
= {
3768 .low
= htons(vxlan
->cfg
.port_min
),
3769 .high
= htons(vxlan
->cfg
.port_max
),
3772 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
3773 goto nla_put_failure
;
3775 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3776 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
3777 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
3778 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
3779 goto nla_put_failure
;
3780 #if IS_ENABLED(CONFIG_IPV6)
3782 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
3783 &dst
->remote_ip
.sin6
.sin6_addr
))
3784 goto nla_put_failure
;
3789 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
3790 goto nla_put_failure
;
3792 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
3793 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
3794 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
3795 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
3796 goto nla_put_failure
;
3797 #if IS_ENABLED(CONFIG_IPV6)
3799 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
3800 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
3801 goto nla_put_failure
;
3806 if (fan_has_map(&vxlan
->fan
)) {
3807 struct nlattr
*fan_nest
;
3808 struct ip_fan_map
*fan_map
;
3810 fan_nest
= nla_nest_start(skb
, IFLA_VXLAN_FAN_MAP
);
3812 goto nla_put_failure
;
3813 list_for_each_entry_rcu(fan_map
, &vxlan
->fan
.fan_maps
, list
) {
3814 struct ifla_fan_map map
;
3816 map
.underlay
= fan_map
->underlay
;
3817 map
.underlay_prefix
= fan_map
->underlay_prefix
;
3818 map
.overlay
= fan_map
->overlay
;
3819 map
.overlay_prefix
= fan_map
->overlay_prefix
;
3820 if (nla_put(skb
, IFLA_FAN_MAPPING
, sizeof(map
), &map
))
3821 goto nla_put_failure
;
3823 nla_nest_end(skb
, fan_nest
);
3826 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
3827 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
3828 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
3829 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
3830 !!(vxlan
->cfg
.flags
& VXLAN_F_LEARN
)) ||
3831 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
3832 !!(vxlan
->cfg
.flags
& VXLAN_F_PROXY
)) ||
3833 nla_put_u8(skb
, IFLA_VXLAN_RSC
,
3834 !!(vxlan
->cfg
.flags
& VXLAN_F_RSC
)) ||
3835 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
3836 !!(vxlan
->cfg
.flags
& VXLAN_F_L2MISS
)) ||
3837 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
3838 !!(vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) ||
3839 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
3840 !!(vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)) ||
3841 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
3842 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
3843 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
3844 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
3845 !(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
3846 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3847 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
3848 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3849 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
3850 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
3851 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_TX
)) ||
3852 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
3853 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_RX
)))
3854 goto nla_put_failure
;
3856 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
3857 goto nla_put_failure
;
3859 if (vxlan
->cfg
.flags
& VXLAN_F_GBP
&&
3860 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
3861 goto nla_put_failure
;
3863 if (vxlan
->cfg
.flags
& VXLAN_F_GPE
&&
3864 nla_put_flag(skb
, IFLA_VXLAN_GPE
))
3865 goto nla_put_failure
;
3867 if (vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
3868 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
3869 goto nla_put_failure
;
3877 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
3879 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3884 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
3886 .maxtype
= IFLA_VXLAN_MAX
,
3887 .policy
= vxlan_policy
,
3888 .priv_size
= sizeof(struct vxlan_dev
),
3889 .setup
= vxlan_setup
,
3890 .validate
= vxlan_validate
,
3891 .newlink
= vxlan_newlink
,
3892 .changelink
= vxlan_changelink
,
3893 .dellink
= vxlan_dellink
,
3894 .get_size
= vxlan_get_size
,
3895 .fill_info
= vxlan_fill_info
,
3896 .get_link_net
= vxlan_get_link_net
,
3899 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
3900 u8 name_assign_type
,
3901 struct vxlan_config
*conf
)
3903 struct nlattr
*tb
[IFLA_MAX
+ 1];
3904 struct net_device
*dev
;
3907 memset(&tb
, 0, sizeof(tb
));
3909 dev
= rtnl_create_link(net
, name
, name_assign_type
,
3910 &vxlan_link_ops
, tb
);
3914 err
= __vxlan_dev_create(net
, dev
, conf
, NULL
);
3917 return ERR_PTR(err
);
3920 err
= rtnl_configure_link(dev
, NULL
);
3922 LIST_HEAD(list_kill
);
3924 vxlan_dellink(dev
, &list_kill
);
3925 unregister_netdevice_many(&list_kill
);
3926 return ERR_PTR(err
);
3931 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
3933 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
3934 struct net_device
*dev
)
3936 struct vxlan_dev
*vxlan
, *next
;
3937 LIST_HEAD(list_kill
);
3939 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
3940 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3942 /* In case we created vxlan device with carrier
3943 * and we loose the carrier due to module unload
3944 * we also need to remove vxlan device. In other
3945 * cases, it's not necessary and remote_ifindex
3946 * is 0 here, so no matches.
3948 if (dst
->remote_ifindex
== dev
->ifindex
)
3949 vxlan_dellink(vxlan
->dev
, &list_kill
);
3952 unregister_netdevice_many(&list_kill
);
3955 static int vxlan_netdevice_event(struct notifier_block
*unused
,
3956 unsigned long event
, void *ptr
)
3958 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3959 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
3961 if (event
== NETDEV_UNREGISTER
) {
3962 vxlan_offload_rx_ports(dev
, false);
3963 vxlan_handle_lowerdev_unregister(vn
, dev
);
3964 } else if (event
== NETDEV_REGISTER
) {
3965 vxlan_offload_rx_ports(dev
, true);
3966 } else if (event
== NETDEV_UDP_TUNNEL_PUSH_INFO
||
3967 event
== NETDEV_UDP_TUNNEL_DROP_INFO
) {
3968 vxlan_offload_rx_ports(dev
, event
== NETDEV_UDP_TUNNEL_PUSH_INFO
);
3974 static struct notifier_block vxlan_notifier_block __read_mostly
= {
3975 .notifier_call
= vxlan_netdevice_event
,
3978 static __net_init
int vxlan_init_net(struct net
*net
)
3980 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3983 INIT_LIST_HEAD(&vn
->vxlan_list
);
3984 spin_lock_init(&vn
->sock_lock
);
3986 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
3987 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
3992 #ifdef CONFIG_SYSCTL
3993 static struct ctl_table_header
*vxlan_fan_header
;
3994 static unsigned int vxlan_fan_version
= 4;
3996 static struct ctl_table vxlan_fan_sysctls
[] = {
3998 .procname
= "vxlan",
3999 .data
= &vxlan_fan_version
,
4000 .maxlen
= sizeof(vxlan_fan_version
),
4002 .proc_handler
= proc_dointvec
,
4006 #endif /* CONFIG_SYSCTL */
4008 static void __net_exit
vxlan_exit_net(struct net
*net
)
4010 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
4011 struct vxlan_dev
*vxlan
, *next
;
4012 struct net_device
*dev
, *aux
;
4017 for_each_netdev_safe(net
, dev
, aux
)
4018 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
4019 unregister_netdevice_queue(dev
, &list
);
4021 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
4022 /* If vxlan->dev is in the same netns, it has already been added
4023 * to the list by the previous loop.
4025 if (!net_eq(dev_net(vxlan
->dev
), net
))
4026 unregister_netdevice_queue(vxlan
->dev
, &list
);
4029 unregister_netdevice_many(&list
);
4032 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
4033 WARN_ON_ONCE(!hlist_empty(&vn
->sock_list
[h
]));
4036 static struct pernet_operations vxlan_net_ops
= {
4037 .init
= vxlan_init_net
,
4038 .exit
= vxlan_exit_net
,
4039 .id
= &vxlan_net_id
,
4040 .size
= sizeof(struct vxlan_net
),
4043 static int __init
vxlan_init_module(void)
4047 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
4049 rc
= register_pernet_subsys(&vxlan_net_ops
);
4053 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
4057 rc
= rtnl_link_register(&vxlan_link_ops
);
4061 #ifdef CONFIG_SYSCTL
4062 vxlan_fan_header
= register_net_sysctl(&init_net
, "net/fan",
4064 if (!vxlan_fan_header
) {
4068 #endif /* CONFIG_SYSCTL */
4071 #ifdef CONFIG_SYSCTL
4073 rtnl_link_unregister(&vxlan_link_ops
);
4074 #endif /* CONFIG_SYSCTL */
4076 unregister_netdevice_notifier(&vxlan_notifier_block
);
4078 unregister_pernet_subsys(&vxlan_net_ops
);
4082 late_initcall(vxlan_init_module
);
4084 static void __exit
vxlan_cleanup_module(void)
4086 #ifdef CONFIG_SYSCTL
4087 unregister_net_sysctl_table(vxlan_fan_header
);
4088 #endif /* CONFIG_SYSCTL */
4089 rtnl_link_unregister(&vxlan_link_ops
);
4090 unregister_netdevice_notifier(&vxlan_notifier_block
);
4091 unregister_pernet_subsys(&vxlan_net_ops
);
4092 /* rcu_barrier() is called by netns */
4094 module_exit(vxlan_cleanup_module
);
4096 MODULE_LICENSE("GPL");
4097 MODULE_VERSION(VXLAN_VERSION
);
4098 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4099 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4100 MODULE_ALIAS_RTNL_LINK("vxlan");