1 // SPDX-License-Identifier: GPL-2.0-only
3 * VXLAN: Virtual eXtensible Local Area Network
5 * Copyright (c) 2012-2013 Vyatta Inc.
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
19 #include <net/ndisc.h>
20 #include <net/ipv6_stubs.h>
23 #include <net/rtnetlink.h>
24 #include <net/inet_ecn.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27 #include <net/tun_proto.h>
28 #include <net/vxlan.h>
30 #if IS_ENABLED(CONFIG_IPV6)
31 #include <net/ip6_tunnel.h>
32 #include <net/ip6_checksum.h>
35 #define VXLAN_VERSION "0.1"
37 #define PORT_HASH_BITS 8
38 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
39 #define FDB_AGE_DEFAULT 300 /* 5 min */
40 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42 /* UDP port for VXLAN traffic.
43 * The IANA assigned port is 4789, but the Linux default is 8472
44 * for compatibility with early adopters.
46 static unsigned short vxlan_port __read_mostly
= 8472;
47 module_param_named(udp_port
, vxlan_port
, ushort
, 0444);
48 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
50 static bool log_ecn_error
= true;
51 module_param(log_ecn_error
, bool, 0644);
52 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
54 static unsigned int vxlan_net_id
;
55 static struct rtnl_link_ops vxlan_link_ops
;
57 static const u8 all_zeros_mac
[ETH_ALEN
+ 2];
59 static int vxlan_sock_add(struct vxlan_dev
*vxlan
);
61 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
);
63 /* per-network namespace private data for this module */
65 struct list_head vxlan_list
;
66 struct hlist_head sock_list
[PORT_HASH_SIZE
];
70 /* Forwarding table entry */
72 struct hlist_node hlist
; /* linked list of entries */
74 unsigned long updated
; /* jiffies */
76 struct list_head remotes
;
77 u8 eth_addr
[ETH_ALEN
];
78 u16 state
; /* see ndm_state */
80 u16 flags
; /* see ndm_flags and below */
83 #define NTF_VXLAN_ADDED_BY_USER 0x100
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly
;
88 static inline bool vxlan_collect_metadata(struct vxlan_sock
*vs
)
90 return vs
->flags
& VXLAN_F_COLLECT_METADATA
||
91 ip_tunnel_collect_metadata();
94 #if IS_ENABLED(CONFIG_IPV6)
96 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
98 if (a
->sa
.sa_family
!= b
->sa
.sa_family
)
100 if (a
->sa
.sa_family
== AF_INET6
)
101 return ipv6_addr_equal(&a
->sin6
.sin6_addr
, &b
->sin6
.sin6_addr
);
103 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
106 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
108 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
109 ip
->sin6
.sin6_addr
= nla_get_in6_addr(nla
);
110 ip
->sa
.sa_family
= AF_INET6
;
112 } else if (nla_len(nla
) >= sizeof(__be32
)) {
113 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
114 ip
->sa
.sa_family
= AF_INET
;
117 return -EAFNOSUPPORT
;
121 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
122 const union vxlan_addr
*ip
)
124 if (ip
->sa
.sa_family
== AF_INET6
)
125 return nla_put_in6_addr(skb
, attr
, &ip
->sin6
.sin6_addr
);
127 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
130 #else /* !CONFIG_IPV6 */
133 bool vxlan_addr_equal(const union vxlan_addr
*a
, const union vxlan_addr
*b
)
135 return a
->sin
.sin_addr
.s_addr
== b
->sin
.sin_addr
.s_addr
;
138 static int vxlan_nla_get_addr(union vxlan_addr
*ip
, struct nlattr
*nla
)
140 if (nla_len(nla
) >= sizeof(struct in6_addr
)) {
141 return -EAFNOSUPPORT
;
142 } else if (nla_len(nla
) >= sizeof(__be32
)) {
143 ip
->sin
.sin_addr
.s_addr
= nla_get_in_addr(nla
);
144 ip
->sa
.sa_family
= AF_INET
;
147 return -EAFNOSUPPORT
;
151 static int vxlan_nla_put_addr(struct sk_buff
*skb
, int attr
,
152 const union vxlan_addr
*ip
)
154 return nla_put_in_addr(skb
, attr
, ip
->sin
.sin_addr
.s_addr
);
158 /* Virtual Network hash table head */
159 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, __be32 vni
)
161 return &vs
->vni_list
[hash_32((__force u32
)vni
, VNI_HASH_BITS
)];
164 /* Socket hash table head */
165 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
167 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
169 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
172 /* First remote destination for a forwarding entry.
173 * Guaranteed to be non-NULL because remotes are never deleted.
175 static inline struct vxlan_rdst
*first_remote_rcu(struct vxlan_fdb
*fdb
)
177 return list_entry_rcu(fdb
->remotes
.next
, struct vxlan_rdst
, list
);
180 static inline struct vxlan_rdst
*first_remote_rtnl(struct vxlan_fdb
*fdb
)
182 return list_first_entry(&fdb
->remotes
, struct vxlan_rdst
, list
);
185 /* Find VXLAN socket based on network namespace, address family and UDP port
186 * and enabled unshareable flags.
188 static struct vxlan_sock
*vxlan_find_sock(struct net
*net
, sa_family_t family
,
189 __be16 port
, u32 flags
, int ifindex
)
191 struct vxlan_sock
*vs
;
193 flags
&= VXLAN_F_RCV_FLAGS
;
195 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
196 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
&&
197 vxlan_get_sk_family(vs
) == family
&&
198 vs
->flags
== flags
&&
199 vs
->sock
->sk
->sk_bound_dev_if
== ifindex
)
205 static struct vxlan_dev
*vxlan_vs_find_vni(struct vxlan_sock
*vs
, int ifindex
,
208 struct vxlan_dev_node
*node
;
210 /* For flow based devices, map all packets to VNI 0 */
211 if (vs
->flags
& VXLAN_F_COLLECT_METADATA
)
214 hlist_for_each_entry_rcu(node
, vni_head(vs
, vni
), hlist
) {
215 if (node
->vxlan
->default_dst
.remote_vni
!= vni
)
218 if (IS_ENABLED(CONFIG_IPV6
)) {
219 const struct vxlan_config
*cfg
= &node
->vxlan
->cfg
;
221 if ((cfg
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
222 cfg
->remote_ifindex
!= ifindex
)
232 /* Look up VNI in a per net namespace table */
233 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, int ifindex
,
234 __be32 vni
, sa_family_t family
,
235 __be16 port
, u32 flags
)
237 struct vxlan_sock
*vs
;
239 vs
= vxlan_find_sock(net
, family
, port
, flags
, ifindex
);
243 return vxlan_vs_find_vni(vs
, ifindex
, vni
);
246 /* Fill in neighbour message in skbuff. */
247 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
248 const struct vxlan_fdb
*fdb
,
249 u32 portid
, u32 seq
, int type
, unsigned int flags
,
250 const struct vxlan_rdst
*rdst
)
252 unsigned long now
= jiffies
;
253 struct nda_cacheinfo ci
;
254 struct nlmsghdr
*nlh
;
256 bool send_ip
, send_eth
;
258 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
262 ndm
= nlmsg_data(nlh
);
263 memset(ndm
, 0, sizeof(*ndm
));
265 send_eth
= send_ip
= true;
267 if (type
== RTM_GETNEIGH
) {
268 send_ip
= !vxlan_addr_any(&rdst
->remote_ip
);
269 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
270 ndm
->ndm_family
= send_ip
? rdst
->remote_ip
.sa
.sa_family
: AF_INET
;
272 ndm
->ndm_family
= AF_BRIDGE
;
273 ndm
->ndm_state
= fdb
->state
;
274 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
275 ndm
->ndm_flags
= fdb
->flags
;
277 ndm
->ndm_flags
|= NTF_OFFLOADED
;
278 ndm
->ndm_type
= RTN_UNICAST
;
280 if (!net_eq(dev_net(vxlan
->dev
), vxlan
->net
) &&
281 nla_put_s32(skb
, NDA_LINK_NETNSID
,
282 peernet2id(dev_net(vxlan
->dev
), vxlan
->net
)))
283 goto nla_put_failure
;
285 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
286 goto nla_put_failure
;
288 if (send_ip
&& vxlan_nla_put_addr(skb
, NDA_DST
, &rdst
->remote_ip
))
289 goto nla_put_failure
;
291 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->cfg
.dst_port
&&
292 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
293 goto nla_put_failure
;
294 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
295 nla_put_u32(skb
, NDA_VNI
, be32_to_cpu(rdst
->remote_vni
)))
296 goto nla_put_failure
;
297 if ((vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) && fdb
->vni
&&
298 nla_put_u32(skb
, NDA_SRC_VNI
,
299 be32_to_cpu(fdb
->vni
)))
300 goto nla_put_failure
;
301 if (rdst
->remote_ifindex
&&
302 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
303 goto nla_put_failure
;
305 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
306 ci
.ndm_confirmed
= 0;
307 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
310 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
311 goto nla_put_failure
;
317 nlmsg_cancel(skb
, nlh
);
321 static inline size_t vxlan_nlmsg_size(void)
323 return NLMSG_ALIGN(sizeof(struct ndmsg
))
324 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
325 + nla_total_size(sizeof(struct in6_addr
)) /* NDA_DST */
326 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
327 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
328 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
329 + nla_total_size(sizeof(__s32
)) /* NDA_LINK_NETNSID */
330 + nla_total_size(sizeof(struct nda_cacheinfo
));
333 static void __vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
334 struct vxlan_rdst
*rd
, int type
)
336 struct net
*net
= dev_net(vxlan
->dev
);
340 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
344 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, rd
);
346 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
347 WARN_ON(err
== -EMSGSIZE
);
352 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
356 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
359 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev
*vxlan
,
360 const struct vxlan_fdb
*fdb
,
361 const struct vxlan_rdst
*rd
,
362 struct netlink_ext_ack
*extack
,
363 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
)
365 fdb_info
->info
.dev
= vxlan
->dev
;
366 fdb_info
->info
.extack
= extack
;
367 fdb_info
->remote_ip
= rd
->remote_ip
;
368 fdb_info
->remote_port
= rd
->remote_port
;
369 fdb_info
->remote_vni
= rd
->remote_vni
;
370 fdb_info
->remote_ifindex
= rd
->remote_ifindex
;
371 memcpy(fdb_info
->eth_addr
, fdb
->eth_addr
, ETH_ALEN
);
372 fdb_info
->vni
= fdb
->vni
;
373 fdb_info
->offloaded
= rd
->offloaded
;
374 fdb_info
->added_by_user
= fdb
->flags
& NTF_VXLAN_ADDED_BY_USER
;
377 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev
*vxlan
,
378 struct vxlan_fdb
*fdb
,
379 struct vxlan_rdst
*rd
,
381 struct netlink_ext_ack
*extack
)
383 struct switchdev_notifier_vxlan_fdb_info info
;
384 enum switchdev_notifier_type notifier_type
;
390 notifier_type
= adding
? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
391 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE
;
392 vxlan_fdb_switchdev_notifier_info(vxlan
, fdb
, rd
, NULL
, &info
);
393 ret
= call_switchdev_notifiers(notifier_type
, vxlan
->dev
,
395 return notifier_to_errno(ret
);
398 static int vxlan_fdb_notify(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*fdb
,
399 struct vxlan_rdst
*rd
, int type
, bool swdev_notify
,
400 struct netlink_ext_ack
*extack
)
407 err
= vxlan_fdb_switchdev_call_notifiers(vxlan
, fdb
, rd
,
413 vxlan_fdb_switchdev_call_notifiers(vxlan
, fdb
, rd
,
419 __vxlan_fdb_notify(vxlan
, fdb
, rd
, type
);
423 static void vxlan_ip_miss(struct net_device
*dev
, union vxlan_addr
*ipa
)
425 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
426 struct vxlan_fdb f
= {
429 struct vxlan_rdst remote
= {
430 .remote_ip
= *ipa
, /* goes to NDA_DST */
431 .remote_vni
= cpu_to_be32(VXLAN_N_VID
),
434 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
, true, NULL
);
437 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
439 struct vxlan_fdb f
= {
442 struct vxlan_rdst remote
= { };
444 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
446 vxlan_fdb_notify(vxlan
, &f
, &remote
, RTM_GETNEIGH
, true, NULL
);
449 /* Hash Ethernet address */
450 static u32
eth_hash(const unsigned char *addr
)
452 u64 value
= get_unaligned((u64
*)addr
);
454 /* only want 6 bytes */
460 return hash_64(value
, FDB_HASH_BITS
);
463 static u32
eth_vni_hash(const unsigned char *addr
, __be32 vni
)
465 /* use 1 byte of OUI and 3 bytes of NIC */
466 u32 key
= get_unaligned((u32
*)(addr
+ 2));
468 return jhash_2words(key
, vni
, vxlan_salt
) & (FDB_HASH_SIZE
- 1);
471 static u32
fdb_head_index(struct vxlan_dev
*vxlan
, const u8
*mac
, __be32 vni
)
473 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)
474 return eth_vni_hash(mac
, vni
);
476 return eth_hash(mac
);
479 /* Hash chain to use given mac address */
480 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
481 const u8
*mac
, __be32 vni
)
483 return &vxlan
->fdb_head
[fdb_head_index(vxlan
, mac
, vni
)];
486 /* Look up Ethernet address in forwarding table */
487 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
488 const u8
*mac
, __be32 vni
)
490 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
, vni
);
493 hlist_for_each_entry_rcu(f
, head
, hlist
) {
494 if (ether_addr_equal(mac
, f
->eth_addr
)) {
495 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
507 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
508 const u8
*mac
, __be32 vni
)
512 f
= __vxlan_find_mac(vxlan
, mac
, vni
);
513 if (f
&& f
->used
!= jiffies
)
519 /* caller should hold vxlan->hash_lock */
520 static struct vxlan_rdst
*vxlan_fdb_find_rdst(struct vxlan_fdb
*f
,
521 union vxlan_addr
*ip
, __be16 port
,
522 __be32 vni
, __u32 ifindex
)
524 struct vxlan_rdst
*rd
;
526 list_for_each_entry(rd
, &f
->remotes
, list
) {
527 if (vxlan_addr_equal(&rd
->remote_ip
, ip
) &&
528 rd
->remote_port
== port
&&
529 rd
->remote_vni
== vni
&&
530 rd
->remote_ifindex
== ifindex
)
537 int vxlan_fdb_find_uc(struct net_device
*dev
, const u8
*mac
, __be32 vni
,
538 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
)
540 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
541 u8 eth_addr
[ETH_ALEN
+ 2] = { 0 };
542 struct vxlan_rdst
*rdst
;
546 if (is_multicast_ether_addr(mac
) ||
547 is_zero_ether_addr(mac
))
550 ether_addr_copy(eth_addr
, mac
);
554 f
= __vxlan_find_mac(vxlan
, eth_addr
, vni
);
560 rdst
= first_remote_rcu(f
);
561 vxlan_fdb_switchdev_notifier_info(vxlan
, f
, rdst
, NULL
, fdb_info
);
567 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc
);
569 static int vxlan_fdb_notify_one(struct notifier_block
*nb
,
570 const struct vxlan_dev
*vxlan
,
571 const struct vxlan_fdb
*f
,
572 const struct vxlan_rdst
*rdst
,
573 struct netlink_ext_ack
*extack
)
575 struct switchdev_notifier_vxlan_fdb_info fdb_info
;
578 vxlan_fdb_switchdev_notifier_info(vxlan
, f
, rdst
, extack
, &fdb_info
);
579 rc
= nb
->notifier_call(nb
, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
,
581 return notifier_to_errno(rc
);
584 int vxlan_fdb_replay(const struct net_device
*dev
, __be32 vni
,
585 struct notifier_block
*nb
,
586 struct netlink_ext_ack
*extack
)
588 struct vxlan_dev
*vxlan
;
589 struct vxlan_rdst
*rdst
;
594 if (!netif_is_vxlan(dev
))
596 vxlan
= netdev_priv(dev
);
598 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
599 spin_lock_bh(&vxlan
->hash_lock
[h
]);
600 hlist_for_each_entry(f
, &vxlan
->fdb_head
[h
], hlist
) {
602 list_for_each_entry(rdst
, &f
->remotes
, list
) {
603 rc
= vxlan_fdb_notify_one(nb
, vxlan
,
611 spin_unlock_bh(&vxlan
->hash_lock
[h
]);
616 spin_unlock_bh(&vxlan
->hash_lock
[h
]);
619 EXPORT_SYMBOL_GPL(vxlan_fdb_replay
);
621 void vxlan_fdb_clear_offload(const struct net_device
*dev
, __be32 vni
)
623 struct vxlan_dev
*vxlan
;
624 struct vxlan_rdst
*rdst
;
628 if (!netif_is_vxlan(dev
))
630 vxlan
= netdev_priv(dev
);
632 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
633 spin_lock_bh(&vxlan
->hash_lock
[h
]);
634 hlist_for_each_entry(f
, &vxlan
->fdb_head
[h
], hlist
)
636 list_for_each_entry(rdst
, &f
->remotes
, list
)
637 rdst
->offloaded
= false;
638 spin_unlock_bh(&vxlan
->hash_lock
[h
]);
642 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload
);
644 /* Replace destination of unicast mac */
645 static int vxlan_fdb_replace(struct vxlan_fdb
*f
,
646 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
647 __u32 ifindex
, struct vxlan_rdst
*oldrd
)
649 struct vxlan_rdst
*rd
;
651 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
655 rd
= list_first_entry_or_null(&f
->remotes
, struct vxlan_rdst
, list
);
660 dst_cache_reset(&rd
->dst_cache
);
662 rd
->remote_port
= port
;
663 rd
->remote_vni
= vni
;
664 rd
->remote_ifindex
= ifindex
;
665 rd
->offloaded
= false;
669 /* Add/update destinations for multicast */
670 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
671 union vxlan_addr
*ip
, __be16 port
, __be32 vni
,
672 __u32 ifindex
, struct vxlan_rdst
**rdp
)
674 struct vxlan_rdst
*rd
;
676 rd
= vxlan_fdb_find_rdst(f
, ip
, port
, vni
, ifindex
);
680 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
684 if (dst_cache_init(&rd
->dst_cache
, GFP_ATOMIC
)) {
690 rd
->remote_port
= port
;
691 rd
->offloaded
= false;
692 rd
->remote_vni
= vni
;
693 rd
->remote_ifindex
= ifindex
;
695 list_add_tail_rcu(&rd
->list
, &f
->remotes
);
701 static struct vxlanhdr
*vxlan_gro_remcsum(struct sk_buff
*skb
,
703 struct vxlanhdr
*vh
, size_t hdrlen
,
705 struct gro_remcsum
*grc
,
708 size_t start
, offset
;
710 if (skb
->remcsum_offload
)
713 if (!NAPI_GRO_CB(skb
)->csum_valid
)
716 start
= vxlan_rco_start(vni_field
);
717 offset
= start
+ vxlan_rco_offset(vni_field
);
719 vh
= skb_gro_remcsum_process(skb
, (void *)vh
, off
, hdrlen
,
720 start
, offset
, grc
, nopartial
);
722 skb
->remcsum_offload
= 1;
727 static struct sk_buff
*vxlan_gro_receive(struct sock
*sk
,
728 struct list_head
*head
,
731 struct sk_buff
*pp
= NULL
;
733 struct vxlanhdr
*vh
, *vh2
;
734 unsigned int hlen
, off_vx
;
736 struct vxlan_sock
*vs
= rcu_dereference_sk_user_data(sk
);
738 struct gro_remcsum grc
;
740 skb_gro_remcsum_init(&grc
);
742 off_vx
= skb_gro_offset(skb
);
743 hlen
= off_vx
+ sizeof(*vh
);
744 vh
= skb_gro_header_fast(skb
, off_vx
);
745 if (skb_gro_header_hard(skb
, hlen
)) {
746 vh
= skb_gro_header_slow(skb
, hlen
, off_vx
);
751 skb_gro_postpull_rcsum(skb
, vh
, sizeof(struct vxlanhdr
));
753 flags
= vh
->vx_flags
;
755 if ((flags
& VXLAN_HF_RCO
) && (vs
->flags
& VXLAN_F_REMCSUM_RX
)) {
756 vh
= vxlan_gro_remcsum(skb
, off_vx
, vh
, sizeof(struct vxlanhdr
),
759 VXLAN_F_REMCSUM_NOPARTIAL
));
765 skb_gro_pull(skb
, sizeof(struct vxlanhdr
)); /* pull vxlan header */
767 list_for_each_entry(p
, head
, list
) {
768 if (!NAPI_GRO_CB(p
)->same_flow
)
771 vh2
= (struct vxlanhdr
*)(p
->data
+ off_vx
);
772 if (vh
->vx_flags
!= vh2
->vx_flags
||
773 vh
->vx_vni
!= vh2
->vx_vni
) {
774 NAPI_GRO_CB(p
)->same_flow
= 0;
779 pp
= call_gro_receive(eth_gro_receive
, head
, skb
);
783 skb_gro_flush_final_remcsum(skb
, pp
, flush
, &grc
);
788 static int vxlan_gro_complete(struct sock
*sk
, struct sk_buff
*skb
, int nhoff
)
790 /* Sets 'skb->inner_mac_header' since we are always called with
791 * 'skb->encapsulation' set.
793 return eth_gro_complete(skb
, nhoff
+ sizeof(struct vxlanhdr
));
796 static struct vxlan_fdb
*vxlan_fdb_alloc(const u8
*mac
, __u16 state
,
797 __be32 src_vni
, __u16 ndm_flags
)
801 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
805 f
->flags
= ndm_flags
;
806 f
->updated
= f
->used
= jiffies
;
808 INIT_LIST_HEAD(&f
->remotes
);
809 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
814 static void vxlan_fdb_insert(struct vxlan_dev
*vxlan
, const u8
*mac
,
815 __be32 src_vni
, struct vxlan_fdb
*f
)
818 hlist_add_head_rcu(&f
->hlist
,
819 vxlan_fdb_head(vxlan
, mac
, src_vni
));
822 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
823 const u8
*mac
, union vxlan_addr
*ip
,
824 __u16 state
, __be16 port
, __be32 src_vni
,
825 __be32 vni
, __u32 ifindex
, __u16 ndm_flags
,
826 struct vxlan_fdb
**fdb
)
828 struct vxlan_rdst
*rd
= NULL
;
832 if (vxlan
->cfg
.addrmax
&&
833 vxlan
->addrcnt
>= vxlan
->cfg
.addrmax
)
836 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
837 f
= vxlan_fdb_alloc(mac
, state
, src_vni
, ndm_flags
);
841 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
852 static void __vxlan_fdb_free(struct vxlan_fdb
*f
)
854 struct vxlan_rdst
*rd
, *nd
;
856 list_for_each_entry_safe(rd
, nd
, &f
->remotes
, list
) {
857 dst_cache_destroy(&rd
->dst_cache
);
863 static void vxlan_fdb_free(struct rcu_head
*head
)
865 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
870 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
871 bool do_notify
, bool swdev_notify
)
873 struct vxlan_rdst
*rd
;
875 netdev_dbg(vxlan
->dev
, "delete %pM\n", f
->eth_addr
);
879 list_for_each_entry(rd
, &f
->remotes
, list
)
880 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
,
883 hlist_del_rcu(&f
->hlist
);
884 call_rcu(&f
->rcu
, vxlan_fdb_free
);
887 static void vxlan_dst_free(struct rcu_head
*head
)
889 struct vxlan_rdst
*rd
= container_of(head
, struct vxlan_rdst
, rcu
);
891 dst_cache_destroy(&rd
->dst_cache
);
895 static int vxlan_fdb_update_existing(struct vxlan_dev
*vxlan
,
896 union vxlan_addr
*ip
,
897 __u16 state
, __u16 flags
,
898 __be16 port
, __be32 vni
,
899 __u32 ifindex
, __u16 ndm_flags
,
902 struct netlink_ext_ack
*extack
)
904 __u16 fdb_flags
= (ndm_flags
& ~NTF_USE
);
905 struct vxlan_rdst
*rd
= NULL
;
906 struct vxlan_rdst oldrd
;
911 /* Do not allow an externally learned entry to take over an entry added
914 if (!(fdb_flags
& NTF_EXT_LEARNED
) ||
915 !(f
->flags
& NTF_VXLAN_ADDED_BY_USER
)) {
916 if (f
->state
!= state
) {
918 f
->updated
= jiffies
;
921 if (f
->flags
!= fdb_flags
) {
922 f
->flags
= fdb_flags
;
923 f
->updated
= jiffies
;
928 if ((flags
& NLM_F_REPLACE
)) {
929 /* Only change unicasts */
930 if (!(is_multicast_ether_addr(f
->eth_addr
) ||
931 is_zero_ether_addr(f
->eth_addr
))) {
932 rc
= vxlan_fdb_replace(f
, ip
, port
, vni
,
939 if ((flags
& NLM_F_APPEND
) &&
940 (is_multicast_ether_addr(f
->eth_addr
) ||
941 is_zero_ether_addr(f
->eth_addr
))) {
942 rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
, &rd
);
949 if (ndm_flags
& NTF_USE
)
954 rd
= first_remote_rtnl(f
);
956 err
= vxlan_fdb_notify(vxlan
, f
, rd
, RTM_NEWNEIGH
,
957 swdev_notify
, extack
);
965 if ((flags
& NLM_F_REPLACE
) && rc
)
967 else if ((flags
& NLM_F_APPEND
) && rc
) {
968 list_del_rcu(&rd
->list
);
969 call_rcu(&rd
->rcu
, vxlan_dst_free
);
974 static int vxlan_fdb_update_create(struct vxlan_dev
*vxlan
,
975 const u8
*mac
, union vxlan_addr
*ip
,
976 __u16 state
, __u16 flags
,
977 __be16 port
, __be32 src_vni
, __be32 vni
,
978 __u32 ifindex
, __u16 ndm_flags
,
980 struct netlink_ext_ack
*extack
)
982 __u16 fdb_flags
= (ndm_flags
& ~NTF_USE
);
986 /* Disallow replace to add a multicast entry */
987 if ((flags
& NLM_F_REPLACE
) &&
988 (is_multicast_ether_addr(mac
) || is_zero_ether_addr(mac
)))
991 netdev_dbg(vxlan
->dev
, "add %pM -> %pIS\n", mac
, ip
);
992 rc
= vxlan_fdb_create(vxlan
, mac
, ip
, state
, port
, src_vni
,
993 vni
, ifindex
, fdb_flags
, &f
);
997 vxlan_fdb_insert(vxlan
, mac
, src_vni
, f
);
998 rc
= vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
), RTM_NEWNEIGH
,
999 swdev_notify
, extack
);
1006 vxlan_fdb_destroy(vxlan
, f
, false, false);
1010 /* Add new entry to forwarding table -- assumes lock held */
1011 static int vxlan_fdb_update(struct vxlan_dev
*vxlan
,
1012 const u8
*mac
, union vxlan_addr
*ip
,
1013 __u16 state
, __u16 flags
,
1014 __be16 port
, __be32 src_vni
, __be32 vni
,
1015 __u32 ifindex
, __u16 ndm_flags
,
1017 struct netlink_ext_ack
*extack
)
1019 struct vxlan_fdb
*f
;
1021 f
= __vxlan_find_mac(vxlan
, mac
, src_vni
);
1023 if (flags
& NLM_F_EXCL
) {
1024 netdev_dbg(vxlan
->dev
,
1025 "lost race to create %pM\n", mac
);
1029 return vxlan_fdb_update_existing(vxlan
, ip
, state
, flags
, port
,
1030 vni
, ifindex
, ndm_flags
, f
,
1031 swdev_notify
, extack
);
1033 if (!(flags
& NLM_F_CREATE
))
1036 return vxlan_fdb_update_create(vxlan
, mac
, ip
, state
, flags
,
1037 port
, src_vni
, vni
, ifindex
,
1038 ndm_flags
, swdev_notify
, extack
);
1042 static void vxlan_fdb_dst_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
,
1043 struct vxlan_rdst
*rd
, bool swdev_notify
)
1045 list_del_rcu(&rd
->list
);
1046 vxlan_fdb_notify(vxlan
, f
, rd
, RTM_DELNEIGH
, swdev_notify
, NULL
);
1047 call_rcu(&rd
->rcu
, vxlan_dst_free
);
1050 static int vxlan_fdb_parse(struct nlattr
*tb
[], struct vxlan_dev
*vxlan
,
1051 union vxlan_addr
*ip
, __be16
*port
, __be32
*src_vni
,
1052 __be32
*vni
, u32
*ifindex
)
1054 struct net
*net
= dev_net(vxlan
->dev
);
1058 err
= vxlan_nla_get_addr(ip
, tb
[NDA_DST
]);
1062 union vxlan_addr
*remote
= &vxlan
->default_dst
.remote_ip
;
1063 if (remote
->sa
.sa_family
== AF_INET
) {
1064 ip
->sin
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
1065 ip
->sa
.sa_family
= AF_INET
;
1066 #if IS_ENABLED(CONFIG_IPV6)
1068 ip
->sin6
.sin6_addr
= in6addr_any
;
1069 ip
->sa
.sa_family
= AF_INET6
;
1075 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
1077 *port
= nla_get_be16(tb
[NDA_PORT
]);
1079 *port
= vxlan
->cfg
.dst_port
;
1083 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
1085 *vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
1087 *vni
= vxlan
->default_dst
.remote_vni
;
1090 if (tb
[NDA_SRC_VNI
]) {
1091 if (nla_len(tb
[NDA_SRC_VNI
]) != sizeof(u32
))
1093 *src_vni
= cpu_to_be32(nla_get_u32(tb
[NDA_SRC_VNI
]));
1095 *src_vni
= vxlan
->default_dst
.remote_vni
;
1098 if (tb
[NDA_IFINDEX
]) {
1099 struct net_device
*tdev
;
1101 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
1103 *ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
1104 tdev
= __dev_get_by_index(net
, *ifindex
);
1106 return -EADDRNOTAVAIL
;
1114 /* Add static entry (via netlink) */
1115 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
1116 struct net_device
*dev
,
1117 const unsigned char *addr
, u16 vid
, u16 flags
,
1118 struct netlink_ext_ack
*extack
)
1120 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1121 /* struct net *net = dev_net(vxlan->dev); */
1122 union vxlan_addr ip
;
1124 __be32 src_vni
, vni
;
1129 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
1130 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1135 if (tb
[NDA_DST
] == NULL
)
1138 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
1142 if (vxlan
->default_dst
.remote_ip
.sa
.sa_family
!= ip
.sa
.sa_family
)
1143 return -EAFNOSUPPORT
;
1145 hash_index
= fdb_head_index(vxlan
, addr
, src_vni
);
1146 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
1147 err
= vxlan_fdb_update(vxlan
, addr
, &ip
, ndm
->ndm_state
, flags
,
1148 port
, src_vni
, vni
, ifindex
,
1149 ndm
->ndm_flags
| NTF_VXLAN_ADDED_BY_USER
,
1151 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
1156 static int __vxlan_fdb_delete(struct vxlan_dev
*vxlan
,
1157 const unsigned char *addr
, union vxlan_addr ip
,
1158 __be16 port
, __be32 src_vni
, __be32 vni
,
1159 u32 ifindex
, bool swdev_notify
)
1161 struct vxlan_fdb
*f
;
1162 struct vxlan_rdst
*rd
= NULL
;
1165 f
= vxlan_find_mac(vxlan
, addr
, src_vni
);
1169 if (!vxlan_addr_any(&ip
)) {
1170 rd
= vxlan_fdb_find_rdst(f
, &ip
, port
, vni
, ifindex
);
1175 /* remove a destination if it's not the only one on the list,
1176 * otherwise destroy the fdb entry
1178 if (rd
&& !list_is_singular(&f
->remotes
)) {
1179 vxlan_fdb_dst_destroy(vxlan
, f
, rd
, swdev_notify
);
1183 vxlan_fdb_destroy(vxlan
, f
, true, swdev_notify
);
1189 /* Delete entry (via netlink) */
1190 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
1191 struct net_device
*dev
,
1192 const unsigned char *addr
, u16 vid
)
1194 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1195 union vxlan_addr ip
;
1196 __be32 src_vni
, vni
;
1202 err
= vxlan_fdb_parse(tb
, vxlan
, &ip
, &port
, &src_vni
, &vni
, &ifindex
);
1206 hash_index
= fdb_head_index(vxlan
, addr
, src_vni
);
1207 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
1208 err
= __vxlan_fdb_delete(vxlan
, addr
, ip
, port
, src_vni
, vni
, ifindex
,
1210 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
1215 /* Dump forwarding table */
1216 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
1217 struct net_device
*dev
,
1218 struct net_device
*filter_dev
, int *idx
)
1220 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1224 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1225 struct vxlan_fdb
*f
;
1227 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
1228 struct vxlan_rdst
*rd
;
1230 list_for_each_entry_rcu(rd
, &f
->remotes
, list
) {
1231 if (*idx
< cb
->args
[2])
1234 err
= vxlan_fdb_info(skb
, vxlan
, f
,
1235 NETLINK_CB(cb
->skb
).portid
,
1250 static int vxlan_fdb_get(struct sk_buff
*skb
,
1251 struct nlattr
*tb
[],
1252 struct net_device
*dev
,
1253 const unsigned char *addr
,
1254 u16 vid
, u32 portid
, u32 seq
,
1255 struct netlink_ext_ack
*extack
)
1257 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1258 struct vxlan_fdb
*f
;
1263 vni
= cpu_to_be32(nla_get_u32(tb
[NDA_VNI
]));
1265 vni
= vxlan
->default_dst
.remote_vni
;
1269 f
= __vxlan_find_mac(vxlan
, addr
, vni
);
1271 NL_SET_ERR_MSG(extack
, "Fdb entry not found");
1276 err
= vxlan_fdb_info(skb
, vxlan
, f
, portid
, seq
,
1277 RTM_NEWNEIGH
, 0, first_remote_rcu(f
));
1283 /* Watch incoming packets to learn mapping between Ethernet address
1284 * and Tunnel endpoint.
1285 * Return true if packet is bogus and should be dropped.
1287 static bool vxlan_snoop(struct net_device
*dev
,
1288 union vxlan_addr
*src_ip
, const u8
*src_mac
,
1289 u32 src_ifindex
, __be32 vni
)
1291 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1292 struct vxlan_fdb
*f
;
1295 #if IS_ENABLED(CONFIG_IPV6)
1296 if (src_ip
->sa
.sa_family
== AF_INET6
&&
1297 (ipv6_addr_type(&src_ip
->sin6
.sin6_addr
) & IPV6_ADDR_LINKLOCAL
))
1298 ifindex
= src_ifindex
;
1301 f
= vxlan_find_mac(vxlan
, src_mac
, vni
);
1303 struct vxlan_rdst
*rdst
= first_remote_rcu(f
);
1305 if (likely(vxlan_addr_equal(&rdst
->remote_ip
, src_ip
) &&
1306 rdst
->remote_ifindex
== ifindex
))
1309 /* Don't migrate static entries, drop packets */
1310 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
1313 if (net_ratelimit())
1315 "%pM migrated from %pIS to %pIS\n",
1316 src_mac
, &rdst
->remote_ip
.sa
, &src_ip
->sa
);
1318 rdst
->remote_ip
= *src_ip
;
1319 f
->updated
= jiffies
;
1320 vxlan_fdb_notify(vxlan
, f
, rdst
, RTM_NEWNEIGH
, true, NULL
);
1322 u32 hash_index
= fdb_head_index(vxlan
, src_mac
, vni
);
1324 /* learned new entry */
1325 spin_lock(&vxlan
->hash_lock
[hash_index
]);
1327 /* close off race between vxlan_flush and incoming packets */
1328 if (netif_running(dev
))
1329 vxlan_fdb_update(vxlan
, src_mac
, src_ip
,
1331 NLM_F_EXCL
|NLM_F_CREATE
,
1332 vxlan
->cfg
.dst_port
,
1334 vxlan
->default_dst
.remote_vni
,
1335 ifindex
, NTF_SELF
, true, NULL
);
1336 spin_unlock(&vxlan
->hash_lock
[hash_index
]);
1342 /* See if multicast group is already in use by other ID */
1343 static bool vxlan_group_used(struct vxlan_net
*vn
, struct vxlan_dev
*dev
)
1345 struct vxlan_dev
*vxlan
;
1346 struct vxlan_sock
*sock4
;
1347 #if IS_ENABLED(CONFIG_IPV6)
1348 struct vxlan_sock
*sock6
;
1350 unsigned short family
= dev
->default_dst
.remote_ip
.sa
.sa_family
;
1352 sock4
= rtnl_dereference(dev
->vn4_sock
);
1354 /* The vxlan_sock is only used by dev, leaving group has
1355 * no effect on other vxlan devices.
1357 if (family
== AF_INET
&& sock4
&& refcount_read(&sock4
->refcnt
) == 1)
1359 #if IS_ENABLED(CONFIG_IPV6)
1360 sock6
= rtnl_dereference(dev
->vn6_sock
);
1361 if (family
== AF_INET6
&& sock6
&& refcount_read(&sock6
->refcnt
) == 1)
1365 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
1366 if (!netif_running(vxlan
->dev
) || vxlan
== dev
)
1369 if (family
== AF_INET
&&
1370 rtnl_dereference(vxlan
->vn4_sock
) != sock4
)
1372 #if IS_ENABLED(CONFIG_IPV6)
1373 if (family
== AF_INET6
&&
1374 rtnl_dereference(vxlan
->vn6_sock
) != sock6
)
1378 if (!vxlan_addr_equal(&vxlan
->default_dst
.remote_ip
,
1379 &dev
->default_dst
.remote_ip
))
1382 if (vxlan
->default_dst
.remote_ifindex
!=
1383 dev
->default_dst
.remote_ifindex
)
1392 static bool __vxlan_sock_release_prep(struct vxlan_sock
*vs
)
1394 struct vxlan_net
*vn
;
1398 if (!refcount_dec_and_test(&vs
->refcnt
))
1401 vn
= net_generic(sock_net(vs
->sock
->sk
), vxlan_net_id
);
1402 spin_lock(&vn
->sock_lock
);
1403 hlist_del_rcu(&vs
->hlist
);
1404 udp_tunnel_notify_del_rx_port(vs
->sock
,
1405 (vs
->flags
& VXLAN_F_GPE
) ?
1406 UDP_TUNNEL_TYPE_VXLAN_GPE
:
1407 UDP_TUNNEL_TYPE_VXLAN
);
1408 spin_unlock(&vn
->sock_lock
);
1413 static void vxlan_sock_release(struct vxlan_dev
*vxlan
)
1415 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1416 #if IS_ENABLED(CONFIG_IPV6)
1417 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1419 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
1422 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
1425 vxlan_vs_del_dev(vxlan
);
1427 if (__vxlan_sock_release_prep(sock4
)) {
1428 udp_tunnel_sock_release(sock4
->sock
);
1432 #if IS_ENABLED(CONFIG_IPV6)
1433 if (__vxlan_sock_release_prep(sock6
)) {
1434 udp_tunnel_sock_release(sock6
->sock
);
1440 /* Update multicast group membership when first VNI on
1441 * multicast address is brought up
1443 static int vxlan_igmp_join(struct vxlan_dev
*vxlan
)
1446 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1447 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1450 if (ip
->sa
.sa_family
== AF_INET
) {
1451 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1452 struct ip_mreqn mreq
= {
1453 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1454 .imr_ifindex
= ifindex
,
1457 sk
= sock4
->sock
->sk
;
1459 ret
= ip_mc_join_group(sk
, &mreq
);
1461 #if IS_ENABLED(CONFIG_IPV6)
1463 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1465 sk
= sock6
->sock
->sk
;
1467 ret
= ipv6_stub
->ipv6_sock_mc_join(sk
, ifindex
,
1468 &ip
->sin6
.sin6_addr
);
1476 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1477 static int vxlan_igmp_leave(struct vxlan_dev
*vxlan
)
1480 union vxlan_addr
*ip
= &vxlan
->default_dst
.remote_ip
;
1481 int ifindex
= vxlan
->default_dst
.remote_ifindex
;
1484 if (ip
->sa
.sa_family
== AF_INET
) {
1485 struct vxlan_sock
*sock4
= rtnl_dereference(vxlan
->vn4_sock
);
1486 struct ip_mreqn mreq
= {
1487 .imr_multiaddr
.s_addr
= ip
->sin
.sin_addr
.s_addr
,
1488 .imr_ifindex
= ifindex
,
1491 sk
= sock4
->sock
->sk
;
1493 ret
= ip_mc_leave_group(sk
, &mreq
);
1495 #if IS_ENABLED(CONFIG_IPV6)
1497 struct vxlan_sock
*sock6
= rtnl_dereference(vxlan
->vn6_sock
);
1499 sk
= sock6
->sock
->sk
;
1501 ret
= ipv6_stub
->ipv6_sock_mc_drop(sk
, ifindex
,
1502 &ip
->sin6
.sin6_addr
);
1510 static bool vxlan_remcsum(struct vxlanhdr
*unparsed
,
1511 struct sk_buff
*skb
, u32 vxflags
)
1513 size_t start
, offset
;
1515 if (!(unparsed
->vx_flags
& VXLAN_HF_RCO
) || skb
->remcsum_offload
)
1518 start
= vxlan_rco_start(unparsed
->vx_vni
);
1519 offset
= start
+ vxlan_rco_offset(unparsed
->vx_vni
);
1521 if (!pskb_may_pull(skb
, offset
+ sizeof(u16
)))
1524 skb_remcsum_process(skb
, (void *)(vxlan_hdr(skb
) + 1), start
, offset
,
1525 !!(vxflags
& VXLAN_F_REMCSUM_NOPARTIAL
));
1527 unparsed
->vx_flags
&= ~VXLAN_HF_RCO
;
1528 unparsed
->vx_vni
&= VXLAN_VNI_MASK
;
1532 static void vxlan_parse_gbp_hdr(struct vxlanhdr
*unparsed
,
1533 struct sk_buff
*skb
, u32 vxflags
,
1534 struct vxlan_metadata
*md
)
1536 struct vxlanhdr_gbp
*gbp
= (struct vxlanhdr_gbp
*)unparsed
;
1537 struct metadata_dst
*tun_dst
;
1539 if (!(unparsed
->vx_flags
& VXLAN_HF_GBP
))
1542 md
->gbp
= ntohs(gbp
->policy_id
);
1544 tun_dst
= (struct metadata_dst
*)skb_dst(skb
);
1546 tun_dst
->u
.tun_info
.key
.tun_flags
|= TUNNEL_VXLAN_OPT
;
1547 tun_dst
->u
.tun_info
.options_len
= sizeof(*md
);
1549 if (gbp
->dont_learn
)
1550 md
->gbp
|= VXLAN_GBP_DONT_LEARN
;
1552 if (gbp
->policy_applied
)
1553 md
->gbp
|= VXLAN_GBP_POLICY_APPLIED
;
1555 /* In flow-based mode, GBP is carried in dst_metadata */
1556 if (!(vxflags
& VXLAN_F_COLLECT_METADATA
))
1557 skb
->mark
= md
->gbp
;
1559 unparsed
->vx_flags
&= ~VXLAN_GBP_USED_BITS
;
1562 static bool vxlan_parse_gpe_hdr(struct vxlanhdr
*unparsed
,
1564 struct sk_buff
*skb
, u32 vxflags
)
1566 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)unparsed
;
1568 /* Need to have Next Protocol set for interfaces in GPE mode. */
1569 if (!gpe
->np_applied
)
1571 /* "The initial version is 0. If a receiver does not support the
1572 * version indicated it MUST drop the packet.
1574 if (gpe
->version
!= 0)
1576 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1577 * processing MUST occur." However, we don't implement OAM
1578 * processing, thus drop the packet.
1583 *protocol
= tun_p_to_eth_p(gpe
->next_protocol
);
1587 unparsed
->vx_flags
&= ~VXLAN_GPE_USED_BITS
;
1591 static bool vxlan_set_mac(struct vxlan_dev
*vxlan
,
1592 struct vxlan_sock
*vs
,
1593 struct sk_buff
*skb
, __be32 vni
)
1595 union vxlan_addr saddr
;
1596 u32 ifindex
= skb
->dev
->ifindex
;
1598 skb_reset_mac_header(skb
);
1599 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
1600 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
1602 /* Ignore packet loops (and multicast echo) */
1603 if (ether_addr_equal(eth_hdr(skb
)->h_source
, vxlan
->dev
->dev_addr
))
1606 /* Get address from the outer IP header */
1607 if (vxlan_get_sk_family(vs
) == AF_INET
) {
1608 saddr
.sin
.sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
1609 saddr
.sa
.sa_family
= AF_INET
;
1610 #if IS_ENABLED(CONFIG_IPV6)
1612 saddr
.sin6
.sin6_addr
= ipv6_hdr(skb
)->saddr
;
1613 saddr
.sa
.sa_family
= AF_INET6
;
1617 if ((vxlan
->cfg
.flags
& VXLAN_F_LEARN
) &&
1618 vxlan_snoop(skb
->dev
, &saddr
, eth_hdr(skb
)->h_source
, ifindex
, vni
))
1624 static bool vxlan_ecn_decapsulate(struct vxlan_sock
*vs
, void *oiph
,
1625 struct sk_buff
*skb
)
1629 if (vxlan_get_sk_family(vs
) == AF_INET
)
1630 err
= IP_ECN_decapsulate(oiph
, skb
);
1631 #if IS_ENABLED(CONFIG_IPV6)
1633 err
= IP6_ECN_decapsulate(oiph
, skb
);
1636 if (unlikely(err
) && log_ecn_error
) {
1637 if (vxlan_get_sk_family(vs
) == AF_INET
)
1638 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1639 &((struct iphdr
*)oiph
)->saddr
,
1640 ((struct iphdr
*)oiph
)->tos
);
1642 net_info_ratelimited("non-ECT from %pI6\n",
1643 &((struct ipv6hdr
*)oiph
)->saddr
);
1648 /* Callback from net/ipv4/udp.c to receive packets */
1649 static int vxlan_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1651 struct pcpu_sw_netstats
*stats
;
1652 struct vxlan_dev
*vxlan
;
1653 struct vxlan_sock
*vs
;
1654 struct vxlanhdr unparsed
;
1655 struct vxlan_metadata _md
;
1656 struct vxlan_metadata
*md
= &_md
;
1657 __be16 protocol
= htons(ETH_P_TEB
);
1658 bool raw_proto
= false;
1662 /* Need UDP and VXLAN header to be present */
1663 if (!pskb_may_pull(skb
, VXLAN_HLEN
))
1666 unparsed
= *vxlan_hdr(skb
);
1667 /* VNI flag always required to be set */
1668 if (!(unparsed
.vx_flags
& VXLAN_HF_VNI
)) {
1669 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
1670 ntohl(vxlan_hdr(skb
)->vx_flags
),
1671 ntohl(vxlan_hdr(skb
)->vx_vni
));
1672 /* Return non vxlan pkt */
1675 unparsed
.vx_flags
&= ~VXLAN_HF_VNI
;
1676 unparsed
.vx_vni
&= ~VXLAN_VNI_MASK
;
1678 vs
= rcu_dereference_sk_user_data(sk
);
1682 vni
= vxlan_vni(vxlan_hdr(skb
)->vx_vni
);
1684 vxlan
= vxlan_vs_find_vni(vs
, skb
->dev
->ifindex
, vni
);
1688 /* For backwards compatibility, only allow reserved fields to be
1689 * used by VXLAN extensions if explicitly requested.
1691 if (vs
->flags
& VXLAN_F_GPE
) {
1692 if (!vxlan_parse_gpe_hdr(&unparsed
, &protocol
, skb
, vs
->flags
))
1697 if (__iptunnel_pull_header(skb
, VXLAN_HLEN
, protocol
, raw_proto
,
1698 !net_eq(vxlan
->net
, dev_net(vxlan
->dev
))))
1701 if (vxlan_collect_metadata(vs
)) {
1702 struct metadata_dst
*tun_dst
;
1704 tun_dst
= udp_tun_rx_dst(skb
, vxlan_get_sk_family(vs
), TUNNEL_KEY
,
1705 key32_to_tunnel_id(vni
), sizeof(*md
));
1710 md
= ip_tunnel_info_opts(&tun_dst
->u
.tun_info
);
1712 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
1714 memset(md
, 0, sizeof(*md
));
1717 if (vs
->flags
& VXLAN_F_REMCSUM_RX
)
1718 if (!vxlan_remcsum(&unparsed
, skb
, vs
->flags
))
1720 if (vs
->flags
& VXLAN_F_GBP
)
1721 vxlan_parse_gbp_hdr(&unparsed
, skb
, vs
->flags
, md
);
1722 /* Note that GBP and GPE can never be active together. This is
1723 * ensured in vxlan_dev_configure.
1726 if (unparsed
.vx_flags
|| unparsed
.vx_vni
) {
1727 /* If there are any unprocessed flags remaining treat
1728 * this as a malformed packet. This behavior diverges from
1729 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1730 * in reserved fields are to be ignored. The approach here
1731 * maintains compatibility with previous stack code, and also
1732 * is more robust and provides a little more security in
1733 * adding extensions to VXLAN.
1739 if (!vxlan_set_mac(vxlan
, vs
, skb
, vni
))
1742 skb_reset_mac_header(skb
);
1743 skb
->dev
= vxlan
->dev
;
1744 skb
->pkt_type
= PACKET_HOST
;
1747 oiph
= skb_network_header(skb
);
1748 skb_reset_network_header(skb
);
1750 if (!vxlan_ecn_decapsulate(vs
, oiph
, skb
)) {
1751 ++vxlan
->dev
->stats
.rx_frame_errors
;
1752 ++vxlan
->dev
->stats
.rx_errors
;
1758 if (unlikely(!(vxlan
->dev
->flags
& IFF_UP
))) {
1760 atomic_long_inc(&vxlan
->dev
->rx_dropped
);
1764 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
1765 u64_stats_update_begin(&stats
->syncp
);
1766 stats
->rx_packets
++;
1767 stats
->rx_bytes
+= skb
->len
;
1768 u64_stats_update_end(&stats
->syncp
);
1770 gro_cells_receive(&vxlan
->gro_cells
, skb
);
1777 /* Consume bad packet */
1782 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1783 static int vxlan_err_lookup(struct sock
*sk
, struct sk_buff
*skb
)
1785 struct vxlan_dev
*vxlan
;
1786 struct vxlan_sock
*vs
;
1787 struct vxlanhdr
*hdr
;
1790 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) + VXLAN_HLEN
))
1793 hdr
= vxlan_hdr(skb
);
1795 if (!(hdr
->vx_flags
& VXLAN_HF_VNI
))
1798 vs
= rcu_dereference_sk_user_data(sk
);
1802 vni
= vxlan_vni(hdr
->vx_vni
);
1803 vxlan
= vxlan_vs_find_vni(vs
, skb
->dev
->ifindex
, vni
);
1810 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1812 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1813 struct arphdr
*parp
;
1816 struct neighbour
*n
;
1818 if (dev
->flags
& IFF_NOARP
)
1821 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
1822 dev
->stats
.tx_dropped
++;
1825 parp
= arp_hdr(skb
);
1827 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
1828 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
1829 parp
->ar_pro
!= htons(ETH_P_IP
) ||
1830 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
1831 parp
->ar_hln
!= dev
->addr_len
||
1834 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
1836 arpptr
+= dev
->addr_len
; /* sha */
1837 memcpy(&sip
, arpptr
, sizeof(sip
));
1838 arpptr
+= sizeof(sip
);
1839 arpptr
+= dev
->addr_len
; /* tha */
1840 memcpy(&tip
, arpptr
, sizeof(tip
));
1842 if (ipv4_is_loopback(tip
) ||
1843 ipv4_is_multicast(tip
))
1846 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
1849 struct vxlan_fdb
*f
;
1850 struct sk_buff
*reply
;
1852 if (!(n
->nud_state
& NUD_CONNECTED
)) {
1857 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
1858 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
1859 /* bridge-local neighbor */
1864 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
1872 skb_reset_mac_header(reply
);
1873 __skb_pull(reply
, skb_network_offset(reply
));
1874 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1875 reply
->pkt_type
= PACKET_HOST
;
1877 if (netif_rx_ni(reply
) == NET_RX_DROP
)
1878 dev
->stats
.rx_dropped
++;
1879 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
1880 union vxlan_addr ipa
= {
1881 .sin
.sin_addr
.s_addr
= tip
,
1882 .sin
.sin_family
= AF_INET
,
1885 vxlan_ip_miss(dev
, &ipa
);
1889 return NETDEV_TX_OK
;
1892 #if IS_ENABLED(CONFIG_IPV6)
1893 static struct sk_buff
*vxlan_na_create(struct sk_buff
*request
,
1894 struct neighbour
*n
, bool isrouter
)
1896 struct net_device
*dev
= request
->dev
;
1897 struct sk_buff
*reply
;
1898 struct nd_msg
*ns
, *na
;
1899 struct ipv6hdr
*pip6
;
1901 int na_olen
= 8; /* opt hdr + ETH_ALEN for target */
1905 if (dev
== NULL
|| !pskb_may_pull(request
, request
->len
))
1908 len
= LL_RESERVED_SPACE(dev
) + sizeof(struct ipv6hdr
) +
1909 sizeof(*na
) + na_olen
+ dev
->needed_tailroom
;
1910 reply
= alloc_skb(len
, GFP_ATOMIC
);
1914 reply
->protocol
= htons(ETH_P_IPV6
);
1916 skb_reserve(reply
, LL_RESERVED_SPACE(request
->dev
));
1917 skb_push(reply
, sizeof(struct ethhdr
));
1918 skb_reset_mac_header(reply
);
1920 ns
= (struct nd_msg
*)(ipv6_hdr(request
) + 1);
1922 daddr
= eth_hdr(request
)->h_source
;
1923 ns_olen
= request
->len
- skb_network_offset(request
) -
1924 sizeof(struct ipv6hdr
) - sizeof(*ns
);
1925 for (i
= 0; i
< ns_olen
-1; i
+= (ns
->opt
[i
+1]<<3)) {
1926 if (ns
->opt
[i
] == ND_OPT_SOURCE_LL_ADDR
) {
1927 daddr
= ns
->opt
+ i
+ sizeof(struct nd_opt_hdr
);
1932 /* Ethernet header */
1933 ether_addr_copy(eth_hdr(reply
)->h_dest
, daddr
);
1934 ether_addr_copy(eth_hdr(reply
)->h_source
, n
->ha
);
1935 eth_hdr(reply
)->h_proto
= htons(ETH_P_IPV6
);
1936 reply
->protocol
= htons(ETH_P_IPV6
);
1938 skb_pull(reply
, sizeof(struct ethhdr
));
1939 skb_reset_network_header(reply
);
1940 skb_put(reply
, sizeof(struct ipv6hdr
));
1944 pip6
= ipv6_hdr(reply
);
1945 memset(pip6
, 0, sizeof(struct ipv6hdr
));
1947 pip6
->priority
= ipv6_hdr(request
)->priority
;
1948 pip6
->nexthdr
= IPPROTO_ICMPV6
;
1949 pip6
->hop_limit
= 255;
1950 pip6
->daddr
= ipv6_hdr(request
)->saddr
;
1951 pip6
->saddr
= *(struct in6_addr
*)n
->primary_key
;
1953 skb_pull(reply
, sizeof(struct ipv6hdr
));
1954 skb_reset_transport_header(reply
);
1956 /* Neighbor Advertisement */
1957 na
= skb_put_zero(reply
, sizeof(*na
) + na_olen
);
1958 na
->icmph
.icmp6_type
= NDISC_NEIGHBOUR_ADVERTISEMENT
;
1959 na
->icmph
.icmp6_router
= isrouter
;
1960 na
->icmph
.icmp6_override
= 1;
1961 na
->icmph
.icmp6_solicited
= 1;
1962 na
->target
= ns
->target
;
1963 ether_addr_copy(&na
->opt
[2], n
->ha
);
1964 na
->opt
[0] = ND_OPT_TARGET_LL_ADDR
;
1965 na
->opt
[1] = na_olen
>> 3;
1967 na
->icmph
.icmp6_cksum
= csum_ipv6_magic(&pip6
->saddr
,
1968 &pip6
->daddr
, sizeof(*na
)+na_olen
, IPPROTO_ICMPV6
,
1969 csum_partial(na
, sizeof(*na
)+na_olen
, 0));
1971 pip6
->payload_len
= htons(sizeof(*na
)+na_olen
);
1973 skb_push(reply
, sizeof(struct ipv6hdr
));
1975 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
1980 static int neigh_reduce(struct net_device
*dev
, struct sk_buff
*skb
, __be32 vni
)
1982 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1983 const struct in6_addr
*daddr
;
1984 const struct ipv6hdr
*iphdr
;
1985 struct inet6_dev
*in6_dev
;
1986 struct neighbour
*n
;
1989 in6_dev
= __in6_dev_get(dev
);
1993 iphdr
= ipv6_hdr(skb
);
1994 daddr
= &iphdr
->daddr
;
1995 msg
= (struct nd_msg
*)(iphdr
+ 1);
1997 if (ipv6_addr_loopback(daddr
) ||
1998 ipv6_addr_is_multicast(&msg
->target
))
2001 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &msg
->target
, dev
);
2004 struct vxlan_fdb
*f
;
2005 struct sk_buff
*reply
;
2007 if (!(n
->nud_state
& NUD_CONNECTED
)) {
2012 f
= vxlan_find_mac(vxlan
, n
->ha
, vni
);
2013 if (f
&& vxlan_addr_any(&(first_remote_rcu(f
)->remote_ip
))) {
2014 /* bridge-local neighbor */
2019 reply
= vxlan_na_create(skb
, n
,
2020 !!(f
? f
->flags
& NTF_ROUTER
: 0));
2027 if (netif_rx_ni(reply
) == NET_RX_DROP
)
2028 dev
->stats
.rx_dropped
++;
2030 } else if (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
) {
2031 union vxlan_addr ipa
= {
2032 .sin6
.sin6_addr
= msg
->target
,
2033 .sin6
.sin6_family
= AF_INET6
,
2036 vxlan_ip_miss(dev
, &ipa
);
2041 return NETDEV_TX_OK
;
2045 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
2047 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2048 struct neighbour
*n
;
2050 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
2054 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
2059 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
2062 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
2063 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
2064 union vxlan_addr ipa
= {
2065 .sin
.sin_addr
.s_addr
= pip
->daddr
,
2066 .sin
.sin_family
= AF_INET
,
2069 vxlan_ip_miss(dev
, &ipa
);
2075 #if IS_ENABLED(CONFIG_IPV6)
2078 struct ipv6hdr
*pip6
;
2080 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
2082 pip6
= ipv6_hdr(skb
);
2083 n
= neigh_lookup(ipv6_stub
->nd_tbl
, &pip6
->daddr
, dev
);
2084 if (!n
&& (vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) {
2085 union vxlan_addr ipa
= {
2086 .sin6
.sin6_addr
= pip6
->daddr
,
2087 .sin6
.sin6_family
= AF_INET6
,
2090 vxlan_ip_miss(dev
, &ipa
);
2104 diff
= !ether_addr_equal(eth_hdr(skb
)->h_dest
, n
->ha
);
2106 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
2108 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
2117 static void vxlan_build_gbp_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
2118 struct vxlan_metadata
*md
)
2120 struct vxlanhdr_gbp
*gbp
;
2125 gbp
= (struct vxlanhdr_gbp
*)vxh
;
2126 vxh
->vx_flags
|= VXLAN_HF_GBP
;
2128 if (md
->gbp
& VXLAN_GBP_DONT_LEARN
)
2129 gbp
->dont_learn
= 1;
2131 if (md
->gbp
& VXLAN_GBP_POLICY_APPLIED
)
2132 gbp
->policy_applied
= 1;
2134 gbp
->policy_id
= htons(md
->gbp
& VXLAN_GBP_ID_MASK
);
2137 static int vxlan_build_gpe_hdr(struct vxlanhdr
*vxh
, u32 vxflags
,
2140 struct vxlanhdr_gpe
*gpe
= (struct vxlanhdr_gpe
*)vxh
;
2142 gpe
->np_applied
= 1;
2143 gpe
->next_protocol
= tun_p_from_eth_p(protocol
);
2144 if (!gpe
->next_protocol
)
2145 return -EPFNOSUPPORT
;
2149 static int vxlan_build_skb(struct sk_buff
*skb
, struct dst_entry
*dst
,
2150 int iphdr_len
, __be32 vni
,
2151 struct vxlan_metadata
*md
, u32 vxflags
,
2154 struct vxlanhdr
*vxh
;
2157 int type
= udp_sum
? SKB_GSO_UDP_TUNNEL_CSUM
: SKB_GSO_UDP_TUNNEL
;
2158 __be16 inner_protocol
= htons(ETH_P_TEB
);
2160 if ((vxflags
& VXLAN_F_REMCSUM_TX
) &&
2161 skb
->ip_summed
== CHECKSUM_PARTIAL
) {
2162 int csum_start
= skb_checksum_start_offset(skb
);
2164 if (csum_start
<= VXLAN_MAX_REMCSUM_START
&&
2165 !(csum_start
& VXLAN_RCO_SHIFT_MASK
) &&
2166 (skb
->csum_offset
== offsetof(struct udphdr
, check
) ||
2167 skb
->csum_offset
== offsetof(struct tcphdr
, check
)))
2168 type
|= SKB_GSO_TUNNEL_REMCSUM
;
2171 min_headroom
= LL_RESERVED_SPACE(dst
->dev
) + dst
->header_len
2172 + VXLAN_HLEN
+ iphdr_len
;
2174 /* Need space for new headers (invalidates iph ptr) */
2175 err
= skb_cow_head(skb
, min_headroom
);
2179 err
= iptunnel_handle_offloads(skb
, type
);
2183 vxh
= __skb_push(skb
, sizeof(*vxh
));
2184 vxh
->vx_flags
= VXLAN_HF_VNI
;
2185 vxh
->vx_vni
= vxlan_vni_field(vni
);
2187 if (type
& SKB_GSO_TUNNEL_REMCSUM
) {
2190 start
= skb_checksum_start_offset(skb
) - sizeof(struct vxlanhdr
);
2191 vxh
->vx_vni
|= vxlan_compute_rco(start
, skb
->csum_offset
);
2192 vxh
->vx_flags
|= VXLAN_HF_RCO
;
2194 if (!skb_is_gso(skb
)) {
2195 skb
->ip_summed
= CHECKSUM_NONE
;
2196 skb
->encapsulation
= 0;
2200 if (vxflags
& VXLAN_F_GBP
)
2201 vxlan_build_gbp_hdr(vxh
, vxflags
, md
);
2202 if (vxflags
& VXLAN_F_GPE
) {
2203 err
= vxlan_build_gpe_hdr(vxh
, vxflags
, skb
->protocol
);
2206 inner_protocol
= skb
->protocol
;
2209 skb_set_inner_protocol(skb
, inner_protocol
);
2213 static struct rtable
*vxlan_get_route(struct vxlan_dev
*vxlan
, struct net_device
*dev
,
2214 struct vxlan_sock
*sock4
,
2215 struct sk_buff
*skb
, int oif
, u8 tos
,
2216 __be32 daddr
, __be32
*saddr
, __be16 dport
, __be16 sport
,
2217 struct dst_cache
*dst_cache
,
2218 const struct ip_tunnel_info
*info
)
2220 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
2221 struct rtable
*rt
= NULL
;
2225 return ERR_PTR(-EIO
);
2230 rt
= dst_cache_get_ip4(dst_cache
, saddr
);
2235 memset(&fl4
, 0, sizeof(fl4
));
2236 fl4
.flowi4_oif
= oif
;
2237 fl4
.flowi4_tos
= RT_TOS(tos
);
2238 fl4
.flowi4_mark
= skb
->mark
;
2239 fl4
.flowi4_proto
= IPPROTO_UDP
;
2242 fl4
.fl4_dport
= dport
;
2243 fl4
.fl4_sport
= sport
;
2245 rt
= ip_route_output_key(vxlan
->net
, &fl4
);
2247 if (rt
->dst
.dev
== dev
) {
2248 netdev_dbg(dev
, "circular route to %pI4\n", &daddr
);
2250 return ERR_PTR(-ELOOP
);
2255 dst_cache_set_ip4(dst_cache
, &rt
->dst
, fl4
.saddr
);
2257 netdev_dbg(dev
, "no route to %pI4\n", &daddr
);
2258 return ERR_PTR(-ENETUNREACH
);
2263 #if IS_ENABLED(CONFIG_IPV6)
2264 static struct dst_entry
*vxlan6_get_route(struct vxlan_dev
*vxlan
,
2265 struct net_device
*dev
,
2266 struct vxlan_sock
*sock6
,
2267 struct sk_buff
*skb
, int oif
, u8 tos
,
2269 const struct in6_addr
*daddr
,
2270 struct in6_addr
*saddr
,
2271 __be16 dport
, __be16 sport
,
2272 struct dst_cache
*dst_cache
,
2273 const struct ip_tunnel_info
*info
)
2275 bool use_cache
= ip_tunnel_dst_cache_usable(skb
, info
);
2276 struct dst_entry
*ndst
;
2281 return ERR_PTR(-EIO
);
2286 ndst
= dst_cache_get_ip6(dst_cache
, saddr
);
2291 memset(&fl6
, 0, sizeof(fl6
));
2292 fl6
.flowi6_oif
= oif
;
2295 fl6
.flowlabel
= ip6_make_flowinfo(RT_TOS(tos
), label
);
2296 fl6
.flowi6_mark
= skb
->mark
;
2297 fl6
.flowi6_proto
= IPPROTO_UDP
;
2298 fl6
.fl6_dport
= dport
;
2299 fl6
.fl6_sport
= sport
;
2301 err
= ipv6_stub
->ipv6_dst_lookup(vxlan
->net
,
2304 if (unlikely(err
< 0)) {
2305 netdev_dbg(dev
, "no route to %pI6\n", daddr
);
2306 return ERR_PTR(-ENETUNREACH
);
2309 if (unlikely(ndst
->dev
== dev
)) {
2310 netdev_dbg(dev
, "circular route to %pI6\n", daddr
);
2312 return ERR_PTR(-ELOOP
);
2317 dst_cache_set_ip6(dst_cache
, ndst
, saddr
);
2322 /* Bypass encapsulation if the destination is local */
2323 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
2324 struct vxlan_dev
*dst_vxlan
, __be32 vni
)
2326 struct pcpu_sw_netstats
*tx_stats
, *rx_stats
;
2327 union vxlan_addr loopback
;
2328 union vxlan_addr
*remote_ip
= &dst_vxlan
->default_dst
.remote_ip
;
2329 struct net_device
*dev
;
2332 tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
2333 rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
2334 skb
->pkt_type
= PACKET_HOST
;
2335 skb
->encapsulation
= 0;
2336 skb
->dev
= dst_vxlan
->dev
;
2337 __skb_pull(skb
, skb_network_offset(skb
));
2339 if (remote_ip
->sa
.sa_family
== AF_INET
) {
2340 loopback
.sin
.sin_addr
.s_addr
= htonl(INADDR_LOOPBACK
);
2341 loopback
.sa
.sa_family
= AF_INET
;
2342 #if IS_ENABLED(CONFIG_IPV6)
2344 loopback
.sin6
.sin6_addr
= in6addr_loopback
;
2345 loopback
.sa
.sa_family
= AF_INET6
;
2351 if (unlikely(!(dev
->flags
& IFF_UP
))) {
2356 if (dst_vxlan
->cfg
.flags
& VXLAN_F_LEARN
)
2357 vxlan_snoop(dev
, &loopback
, eth_hdr(skb
)->h_source
, 0, vni
);
2359 u64_stats_update_begin(&tx_stats
->syncp
);
2360 tx_stats
->tx_packets
++;
2361 tx_stats
->tx_bytes
+= len
;
2362 u64_stats_update_end(&tx_stats
->syncp
);
2364 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
2365 u64_stats_update_begin(&rx_stats
->syncp
);
2366 rx_stats
->rx_packets
++;
2367 rx_stats
->rx_bytes
+= len
;
2368 u64_stats_update_end(&rx_stats
->syncp
);
2371 dev
->stats
.rx_dropped
++;
2376 static int encap_bypass_if_local(struct sk_buff
*skb
, struct net_device
*dev
,
2377 struct vxlan_dev
*vxlan
,
2378 union vxlan_addr
*daddr
,
2379 __be16 dst_port
, int dst_ifindex
, __be32 vni
,
2380 struct dst_entry
*dst
,
2383 #if IS_ENABLED(CONFIG_IPV6)
2384 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2385 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2386 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2388 BUILD_BUG_ON(RTCF_LOCAL
!= RTF_LOCAL
);
2390 /* Bypass encapsulation if the destination is local */
2391 if (rt_flags
& RTCF_LOCAL
&&
2392 !(rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
2393 struct vxlan_dev
*dst_vxlan
;
2396 dst_vxlan
= vxlan_find_vni(vxlan
->net
, dst_ifindex
, vni
,
2397 daddr
->sa
.sa_family
, dst_port
,
2400 dev
->stats
.tx_errors
++;
2405 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
, vni
);
2412 static void vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
2413 __be32 default_vni
, struct vxlan_rdst
*rdst
,
2416 struct dst_cache
*dst_cache
;
2417 struct ip_tunnel_info
*info
;
2418 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2419 const struct iphdr
*old_iph
= ip_hdr(skb
);
2420 union vxlan_addr
*dst
;
2421 union vxlan_addr remote_ip
, local_ip
;
2422 struct vxlan_metadata _md
;
2423 struct vxlan_metadata
*md
= &_md
;
2424 __be16 src_port
= 0, dst_port
;
2425 struct dst_entry
*ndst
= NULL
;
2430 u32 flags
= vxlan
->cfg
.flags
;
2431 bool udp_sum
= false;
2432 bool xnet
= !net_eq(vxlan
->net
, dev_net(vxlan
->dev
));
2434 info
= skb_tunnel_info(skb
);
2437 dst
= &rdst
->remote_ip
;
2438 if (vxlan_addr_any(dst
)) {
2440 /* short-circuited back to local bridge */
2441 vxlan_encap_bypass(skb
, vxlan
, vxlan
, default_vni
);
2447 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->cfg
.dst_port
;
2448 vni
= (rdst
->remote_vni
) ? : default_vni
;
2449 ifindex
= rdst
->remote_ifindex
;
2450 local_ip
= vxlan
->cfg
.saddr
;
2451 dst_cache
= &rdst
->dst_cache
;
2452 md
->gbp
= skb
->mark
;
2453 if (flags
& VXLAN_F_TTL_INHERIT
) {
2454 ttl
= ip_tunnel_get_ttl(old_iph
, skb
);
2456 ttl
= vxlan
->cfg
.ttl
;
2457 if (!ttl
&& vxlan_addr_multicast(dst
))
2461 tos
= vxlan
->cfg
.tos
;
2463 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
2465 if (dst
->sa
.sa_family
== AF_INET
)
2466 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM_TX
);
2468 udp_sum
= !(flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
);
2469 label
= vxlan
->cfg
.label
;
2472 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2476 remote_ip
.sa
.sa_family
= ip_tunnel_info_af(info
);
2477 if (remote_ip
.sa
.sa_family
== AF_INET
) {
2478 remote_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.dst
;
2479 local_ip
.sin
.sin_addr
.s_addr
= info
->key
.u
.ipv4
.src
;
2481 remote_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.dst
;
2482 local_ip
.sin6
.sin6_addr
= info
->key
.u
.ipv6
.src
;
2485 dst_port
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2486 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2488 dst_cache
= &info
->dst_cache
;
2489 if (info
->key
.tun_flags
& TUNNEL_VXLAN_OPT
) {
2490 if (info
->options_len
< sizeof(*md
))
2492 md
= ip_tunnel_info_opts(info
);
2494 ttl
= info
->key
.ttl
;
2495 tos
= info
->key
.tos
;
2496 label
= info
->key
.label
;
2497 udp_sum
= !!(info
->key
.tun_flags
& TUNNEL_CSUM
);
2499 src_port
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2500 vxlan
->cfg
.port_max
, true);
2503 if (dst
->sa
.sa_family
== AF_INET
) {
2504 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2509 ifindex
= sock4
->sock
->sk
->sk_bound_dev_if
;
2511 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, ifindex
, tos
,
2512 dst
->sin
.sin_addr
.s_addr
,
2513 &local_ip
.sin
.sin_addr
.s_addr
,
2522 /* Bypass encapsulation if the destination is local */
2523 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2524 dst_port
, ifindex
, vni
,
2525 &rt
->dst
, rt
->rt_flags
);
2529 if (vxlan
->cfg
.df
== VXLAN_DF_SET
) {
2531 } else if (vxlan
->cfg
.df
== VXLAN_DF_INHERIT
) {
2532 struct ethhdr
*eth
= eth_hdr(skb
);
2534 if (ntohs(eth
->h_proto
) == ETH_P_IPV6
||
2535 (ntohs(eth
->h_proto
) == ETH_P_IP
&&
2536 old_iph
->frag_off
& htons(IP_DF
)))
2539 } else if (info
->key
.tun_flags
& TUNNEL_DONT_FRAGMENT
) {
2544 skb_tunnel_check_pmtu(skb
, ndst
, VXLAN_HEADROOM
);
2546 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2547 ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
2548 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct iphdr
),
2549 vni
, md
, flags
, udp_sum
);
2553 udp_tunnel_xmit_skb(rt
, sock4
->sock
->sk
, skb
, local_ip
.sin
.sin_addr
.s_addr
,
2554 dst
->sin
.sin_addr
.s_addr
, tos
, ttl
, df
,
2555 src_port
, dst_port
, xnet
, !udp_sum
);
2556 #if IS_ENABLED(CONFIG_IPV6)
2558 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2561 ifindex
= sock6
->sock
->sk
->sk_bound_dev_if
;
2563 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, ifindex
, tos
,
2564 label
, &dst
->sin6
.sin6_addr
,
2565 &local_ip
.sin6
.sin6_addr
,
2569 err
= PTR_ERR(ndst
);
2575 u32 rt6i_flags
= ((struct rt6_info
*)ndst
)->rt6i_flags
;
2577 err
= encap_bypass_if_local(skb
, dev
, vxlan
, dst
,
2578 dst_port
, ifindex
, vni
,
2584 skb_tunnel_check_pmtu(skb
, ndst
, VXLAN6_HEADROOM
);
2586 tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
2587 ttl
= ttl
? : ip6_dst_hoplimit(ndst
);
2588 skb_scrub_packet(skb
, xnet
);
2589 err
= vxlan_build_skb(skb
, ndst
, sizeof(struct ipv6hdr
),
2590 vni
, md
, flags
, udp_sum
);
2594 udp_tunnel6_xmit_skb(ndst
, sock6
->sock
->sk
, skb
, dev
,
2595 &local_ip
.sin6
.sin6_addr
,
2596 &dst
->sin6
.sin6_addr
, tos
, ttl
,
2597 label
, src_port
, dst_port
, !udp_sum
);
2605 dev
->stats
.tx_dropped
++;
2612 dev
->stats
.collisions
++;
2613 else if (err
== -ENETUNREACH
)
2614 dev
->stats
.tx_carrier_errors
++;
2616 dev
->stats
.tx_errors
++;
2620 /* Transmit local packets over Vxlan
2622 * Outer IP header inherits ECN and DF from inner header.
2623 * Outer UDP destination is the VXLAN assigned port.
2624 * source port is based on hash of flow
2626 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
2628 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2629 struct vxlan_rdst
*rdst
, *fdst
= NULL
;
2630 const struct ip_tunnel_info
*info
;
2631 bool did_rsc
= false;
2632 struct vxlan_fdb
*f
;
2636 info
= skb_tunnel_info(skb
);
2638 skb_reset_mac_header(skb
);
2640 if (vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
) {
2641 if (info
&& info
->mode
& IP_TUNNEL_INFO_BRIDGE
&&
2642 info
->mode
& IP_TUNNEL_INFO_TX
) {
2643 vni
= tunnel_id_to_key32(info
->key
.tun_id
);
2645 if (info
&& info
->mode
& IP_TUNNEL_INFO_TX
)
2646 vxlan_xmit_one(skb
, dev
, vni
, NULL
, false);
2649 return NETDEV_TX_OK
;
2653 if (vxlan
->cfg
.flags
& VXLAN_F_PROXY
) {
2655 if (ntohs(eth
->h_proto
) == ETH_P_ARP
)
2656 return arp_reduce(dev
, skb
, vni
);
2657 #if IS_ENABLED(CONFIG_IPV6)
2658 else if (ntohs(eth
->h_proto
) == ETH_P_IPV6
&&
2659 pskb_may_pull(skb
, sizeof(struct ipv6hdr
) +
2660 sizeof(struct nd_msg
)) &&
2661 ipv6_hdr(skb
)->nexthdr
== IPPROTO_ICMPV6
) {
2662 struct nd_msg
*m
= (struct nd_msg
*)(ipv6_hdr(skb
) + 1);
2664 if (m
->icmph
.icmp6_code
== 0 &&
2665 m
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
)
2666 return neigh_reduce(dev
, skb
, vni
);
2672 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2675 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->cfg
.flags
& VXLAN_F_RSC
) &&
2676 (ntohs(eth
->h_proto
) == ETH_P_IP
||
2677 ntohs(eth
->h_proto
) == ETH_P_IPV6
)) {
2678 did_rsc
= route_shortcircuit(dev
, skb
);
2680 f
= vxlan_find_mac(vxlan
, eth
->h_dest
, vni
);
2684 f
= vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2686 if ((vxlan
->cfg
.flags
& VXLAN_F_L2MISS
) &&
2687 !is_multicast_ether_addr(eth
->h_dest
))
2688 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
2690 dev
->stats
.tx_dropped
++;
2692 return NETDEV_TX_OK
;
2696 list_for_each_entry_rcu(rdst
, &f
->remotes
, list
) {
2697 struct sk_buff
*skb1
;
2703 skb1
= skb_clone(skb
, GFP_ATOMIC
);
2705 vxlan_xmit_one(skb1
, dev
, vni
, rdst
, did_rsc
);
2709 vxlan_xmit_one(skb
, dev
, vni
, fdst
, did_rsc
);
2712 return NETDEV_TX_OK
;
2715 /* Walk the forwarding table and purge stale entries */
2716 static void vxlan_cleanup(struct timer_list
*t
)
2718 struct vxlan_dev
*vxlan
= from_timer(vxlan
, t
, age_timer
);
2719 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
2722 if (!netif_running(vxlan
->dev
))
2725 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2726 struct hlist_node
*p
, *n
;
2728 spin_lock(&vxlan
->hash_lock
[h
]);
2729 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2731 = container_of(p
, struct vxlan_fdb
, hlist
);
2732 unsigned long timeout
;
2734 if (f
->state
& (NUD_PERMANENT
| NUD_NOARP
))
2737 if (f
->flags
& NTF_EXT_LEARNED
)
2740 timeout
= f
->used
+ vxlan
->cfg
.age_interval
* HZ
;
2741 if (time_before_eq(timeout
, jiffies
)) {
2742 netdev_dbg(vxlan
->dev
,
2743 "garbage collect %pM\n",
2745 f
->state
= NUD_STALE
;
2746 vxlan_fdb_destroy(vxlan
, f
, true, true);
2747 } else if (time_before(timeout
, next_timer
))
2748 next_timer
= timeout
;
2750 spin_unlock(&vxlan
->hash_lock
[h
]);
2753 mod_timer(&vxlan
->age_timer
, next_timer
);
2756 static void vxlan_vs_del_dev(struct vxlan_dev
*vxlan
)
2758 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2760 spin_lock(&vn
->sock_lock
);
2761 hlist_del_init_rcu(&vxlan
->hlist4
.hlist
);
2762 #if IS_ENABLED(CONFIG_IPV6)
2763 hlist_del_init_rcu(&vxlan
->hlist6
.hlist
);
2765 spin_unlock(&vn
->sock_lock
);
2768 static void vxlan_vs_add_dev(struct vxlan_sock
*vs
, struct vxlan_dev
*vxlan
,
2769 struct vxlan_dev_node
*node
)
2771 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2772 __be32 vni
= vxlan
->default_dst
.remote_vni
;
2774 node
->vxlan
= vxlan
;
2775 spin_lock(&vn
->sock_lock
);
2776 hlist_add_head_rcu(&node
->hlist
, vni_head(vs
, vni
));
2777 spin_unlock(&vn
->sock_lock
);
2780 /* Setup stats when device is created */
2781 static int vxlan_init(struct net_device
*dev
)
2783 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
2790 static void vxlan_fdb_delete_default(struct vxlan_dev
*vxlan
, __be32 vni
)
2792 struct vxlan_fdb
*f
;
2793 u32 hash_index
= fdb_head_index(vxlan
, all_zeros_mac
, vni
);
2795 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
2796 f
= __vxlan_find_mac(vxlan
, all_zeros_mac
, vni
);
2798 vxlan_fdb_destroy(vxlan
, f
, true, true);
2799 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
2802 static void vxlan_uninit(struct net_device
*dev
)
2804 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2806 gro_cells_destroy(&vxlan
->gro_cells
);
2808 vxlan_fdb_delete_default(vxlan
, vxlan
->cfg
.vni
);
2810 free_percpu(dev
->tstats
);
2813 /* Start ageing timer and join group when device is brought up */
2814 static int vxlan_open(struct net_device
*dev
)
2816 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2819 ret
= vxlan_sock_add(vxlan
);
2823 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
)) {
2824 ret
= vxlan_igmp_join(vxlan
);
2825 if (ret
== -EADDRINUSE
)
2828 vxlan_sock_release(vxlan
);
2833 if (vxlan
->cfg
.age_interval
)
2834 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
2839 /* Purge the forwarding table */
2840 static void vxlan_flush(struct vxlan_dev
*vxlan
, bool do_all
)
2844 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
2845 struct hlist_node
*p
, *n
;
2847 spin_lock_bh(&vxlan
->hash_lock
[h
]);
2848 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
2850 = container_of(p
, struct vxlan_fdb
, hlist
);
2851 if (!do_all
&& (f
->state
& (NUD_PERMANENT
| NUD_NOARP
)))
2853 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2854 if (!is_zero_ether_addr(f
->eth_addr
))
2855 vxlan_fdb_destroy(vxlan
, f
, true, true);
2857 spin_unlock_bh(&vxlan
->hash_lock
[h
]);
2861 /* Cleanup timer and forwarding table on shutdown */
2862 static int vxlan_stop(struct net_device
*dev
)
2864 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2865 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
2868 if (vxlan_addr_multicast(&vxlan
->default_dst
.remote_ip
) &&
2869 !vxlan_group_used(vn
, vxlan
))
2870 ret
= vxlan_igmp_leave(vxlan
);
2872 del_timer_sync(&vxlan
->age_timer
);
2874 vxlan_flush(vxlan
, false);
2875 vxlan_sock_release(vxlan
);
2880 /* Stub, nothing needs to be done. */
2881 static void vxlan_set_multicast_list(struct net_device
*dev
)
2885 static int vxlan_change_mtu(struct net_device
*dev
, int new_mtu
)
2887 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2888 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
2889 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
2890 dst
->remote_ifindex
);
2891 bool use_ipv6
= !!(vxlan
->cfg
.flags
& VXLAN_F_IPV6
);
2893 /* This check is different than dev->max_mtu, because it looks at
2894 * the lowerdev->mtu, rather than the static dev->max_mtu
2897 int max_mtu
= lowerdev
->mtu
-
2898 (use_ipv6
? VXLAN6_HEADROOM
: VXLAN_HEADROOM
);
2899 if (new_mtu
> max_mtu
)
2907 static int vxlan_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
2909 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
2910 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
2911 __be16 sport
, dport
;
2913 sport
= udp_flow_src_port(dev_net(dev
), skb
, vxlan
->cfg
.port_min
,
2914 vxlan
->cfg
.port_max
, true);
2915 dport
= info
->key
.tp_dst
? : vxlan
->cfg
.dst_port
;
2917 if (ip_tunnel_info_af(info
) == AF_INET
) {
2918 struct vxlan_sock
*sock4
= rcu_dereference(vxlan
->vn4_sock
);
2921 rt
= vxlan_get_route(vxlan
, dev
, sock4
, skb
, 0, info
->key
.tos
,
2922 info
->key
.u
.ipv4
.dst
,
2923 &info
->key
.u
.ipv4
.src
, dport
, sport
,
2924 &info
->dst_cache
, info
);
2929 #if IS_ENABLED(CONFIG_IPV6)
2930 struct vxlan_sock
*sock6
= rcu_dereference(vxlan
->vn6_sock
);
2931 struct dst_entry
*ndst
;
2933 ndst
= vxlan6_get_route(vxlan
, dev
, sock6
, skb
, 0, info
->key
.tos
,
2934 info
->key
.label
, &info
->key
.u
.ipv6
.dst
,
2935 &info
->key
.u
.ipv6
.src
, dport
, sport
,
2936 &info
->dst_cache
, info
);
2938 return PTR_ERR(ndst
);
2940 #else /* !CONFIG_IPV6 */
2941 return -EPFNOSUPPORT
;
2944 info
->key
.tp_src
= sport
;
2945 info
->key
.tp_dst
= dport
;
2949 static const struct net_device_ops vxlan_netdev_ether_ops
= {
2950 .ndo_init
= vxlan_init
,
2951 .ndo_uninit
= vxlan_uninit
,
2952 .ndo_open
= vxlan_open
,
2953 .ndo_stop
= vxlan_stop
,
2954 .ndo_start_xmit
= vxlan_xmit
,
2955 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2956 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
2957 .ndo_change_mtu
= vxlan_change_mtu
,
2958 .ndo_validate_addr
= eth_validate_addr
,
2959 .ndo_set_mac_address
= eth_mac_addr
,
2960 .ndo_fdb_add
= vxlan_fdb_add
,
2961 .ndo_fdb_del
= vxlan_fdb_delete
,
2962 .ndo_fdb_dump
= vxlan_fdb_dump
,
2963 .ndo_fdb_get
= vxlan_fdb_get
,
2964 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2965 .ndo_change_proto_down
= dev_change_proto_down_generic
,
2968 static const struct net_device_ops vxlan_netdev_raw_ops
= {
2969 .ndo_init
= vxlan_init
,
2970 .ndo_uninit
= vxlan_uninit
,
2971 .ndo_open
= vxlan_open
,
2972 .ndo_stop
= vxlan_stop
,
2973 .ndo_start_xmit
= vxlan_xmit
,
2974 .ndo_get_stats64
= ip_tunnel_get_stats64
,
2975 .ndo_change_mtu
= vxlan_change_mtu
,
2976 .ndo_fill_metadata_dst
= vxlan_fill_metadata_dst
,
2979 /* Info for udev, that this is a virtual tunnel endpoint */
2980 static struct device_type vxlan_type
= {
2984 /* Calls the ndo_udp_tunnel_add of the caller in order to
2985 * supply the listening VXLAN udp ports. Callers are expected
2986 * to implement the ndo_udp_tunnel_add.
2988 static void vxlan_offload_rx_ports(struct net_device
*dev
, bool push
)
2990 struct vxlan_sock
*vs
;
2991 struct net
*net
= dev_net(dev
);
2992 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
2995 spin_lock(&vn
->sock_lock
);
2996 for (i
= 0; i
< PORT_HASH_SIZE
; ++i
) {
2997 hlist_for_each_entry_rcu(vs
, &vn
->sock_list
[i
], hlist
) {
2998 unsigned short type
;
3000 if (vs
->flags
& VXLAN_F_GPE
)
3001 type
= UDP_TUNNEL_TYPE_VXLAN_GPE
;
3003 type
= UDP_TUNNEL_TYPE_VXLAN
;
3006 udp_tunnel_push_rx_port(dev
, vs
->sock
, type
);
3008 udp_tunnel_drop_rx_port(dev
, vs
->sock
, type
);
3011 spin_unlock(&vn
->sock_lock
);
3014 /* Initialize the device structure. */
3015 static void vxlan_setup(struct net_device
*dev
)
3017 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3020 eth_hw_addr_random(dev
);
3023 dev
->needs_free_netdev
= true;
3024 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
3026 dev
->features
|= NETIF_F_LLTX
;
3027 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
3028 dev
->features
|= NETIF_F_RXCSUM
;
3029 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
3031 dev
->vlan_features
= dev
->features
;
3032 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
3033 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
3034 netif_keep_dst(dev
);
3035 dev
->priv_flags
|= IFF_NO_QUEUE
;
3037 /* MTU range: 68 - 65535 */
3038 dev
->min_mtu
= ETH_MIN_MTU
;
3039 dev
->max_mtu
= ETH_MAX_MTU
;
3041 INIT_LIST_HEAD(&vxlan
->next
);
3043 timer_setup(&vxlan
->age_timer
, vxlan_cleanup
, TIMER_DEFERRABLE
);
3047 gro_cells_init(&vxlan
->gro_cells
, dev
);
3049 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
3050 spin_lock_init(&vxlan
->hash_lock
[h
]);
3051 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
3055 static void vxlan_ether_setup(struct net_device
*dev
)
3057 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
3058 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
3059 dev
->netdev_ops
= &vxlan_netdev_ether_ops
;
3062 static void vxlan_raw_setup(struct net_device
*dev
)
3064 dev
->header_ops
= NULL
;
3065 dev
->type
= ARPHRD_NONE
;
3066 dev
->hard_header_len
= 0;
3068 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
3069 dev
->netdev_ops
= &vxlan_netdev_raw_ops
;
3072 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
3073 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
3074 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
3075 [IFLA_VXLAN_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
3076 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
3077 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
3078 [IFLA_VXLAN_LOCAL6
] = { .len
= sizeof(struct in6_addr
) },
3079 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
3080 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
3081 [IFLA_VXLAN_LABEL
] = { .type
= NLA_U32
},
3082 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
3083 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
3084 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
3085 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
3086 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
3087 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
3088 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
3089 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
3090 [IFLA_VXLAN_COLLECT_METADATA
] = { .type
= NLA_U8
},
3091 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
3092 [IFLA_VXLAN_UDP_CSUM
] = { .type
= NLA_U8
},
3093 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX
] = { .type
= NLA_U8
},
3094 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX
] = { .type
= NLA_U8
},
3095 [IFLA_VXLAN_REMCSUM_TX
] = { .type
= NLA_U8
},
3096 [IFLA_VXLAN_REMCSUM_RX
] = { .type
= NLA_U8
},
3097 [IFLA_VXLAN_GBP
] = { .type
= NLA_FLAG
, },
3098 [IFLA_VXLAN_GPE
] = { .type
= NLA_FLAG
, },
3099 [IFLA_VXLAN_REMCSUM_NOPARTIAL
] = { .type
= NLA_FLAG
},
3100 [IFLA_VXLAN_TTL_INHERIT
] = { .type
= NLA_FLAG
},
3101 [IFLA_VXLAN_DF
] = { .type
= NLA_U8
},
3104 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
3105 struct netlink_ext_ack
*extack
)
3107 if (tb
[IFLA_ADDRESS
]) {
3108 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
3109 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
3110 "Provided link layer address is not Ethernet");
3114 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
3115 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_ADDRESS
],
3116 "Provided Ethernet address is not unicast");
3117 return -EADDRNOTAVAIL
;
3122 u32 mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3124 if (mtu
< ETH_MIN_MTU
|| mtu
> ETH_MAX_MTU
) {
3125 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_MTU
],
3126 "MTU must be between 68 and 65535");
3132 NL_SET_ERR_MSG(extack
,
3133 "Required attributes not provided to perform the operation");
3137 if (data
[IFLA_VXLAN_ID
]) {
3138 u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
3140 if (id
>= VXLAN_N_VID
) {
3141 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_ID
],
3142 "VXLAN ID must be lower than 16777216");
3147 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3148 const struct ifla_vxlan_port_range
*p
3149 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3151 if (ntohs(p
->high
) < ntohs(p
->low
)) {
3152 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_PORT_RANGE
],
3153 "Invalid source port range");
3158 if (data
[IFLA_VXLAN_DF
]) {
3159 enum ifla_vxlan_df df
= nla_get_u8(data
[IFLA_VXLAN_DF
]);
3161 if (df
< 0 || df
> VXLAN_DF_MAX
) {
3162 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_DF
],
3163 "Invalid DF attribute");
3171 static void vxlan_get_drvinfo(struct net_device
*netdev
,
3172 struct ethtool_drvinfo
*drvinfo
)
3174 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
3175 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
3178 static int vxlan_get_link_ksettings(struct net_device
*dev
,
3179 struct ethtool_link_ksettings
*cmd
)
3181 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3182 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3183 struct net_device
*lowerdev
= __dev_get_by_index(vxlan
->net
,
3184 dst
->remote_ifindex
);
3187 cmd
->base
.duplex
= DUPLEX_UNKNOWN
;
3188 cmd
->base
.port
= PORT_OTHER
;
3189 cmd
->base
.speed
= SPEED_UNKNOWN
;
3194 return __ethtool_get_link_ksettings(lowerdev
, cmd
);
3197 static const struct ethtool_ops vxlan_ethtool_ops
= {
3198 .get_drvinfo
= vxlan_get_drvinfo
,
3199 .get_link
= ethtool_op_get_link
,
3200 .get_link_ksettings
= vxlan_get_link_ksettings
,
3203 static struct socket
*vxlan_create_sock(struct net
*net
, bool ipv6
,
3204 __be16 port
, u32 flags
, int ifindex
)
3206 struct socket
*sock
;
3207 struct udp_port_cfg udp_conf
;
3210 memset(&udp_conf
, 0, sizeof(udp_conf
));
3213 udp_conf
.family
= AF_INET6
;
3214 udp_conf
.use_udp6_rx_checksums
=
3215 !(flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
);
3216 udp_conf
.ipv6_v6only
= 1;
3218 udp_conf
.family
= AF_INET
;
3221 udp_conf
.local_udp_port
= port
;
3222 udp_conf
.bind_ifindex
= ifindex
;
3224 /* Open UDP socket */
3225 err
= udp_sock_create(net
, &udp_conf
, &sock
);
3227 return ERR_PTR(err
);
3232 /* Create new listen socket if needed */
3233 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, bool ipv6
,
3234 __be16 port
, u32 flags
,
3237 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3238 struct vxlan_sock
*vs
;
3239 struct socket
*sock
;
3241 struct udp_tunnel_sock_cfg tunnel_cfg
;
3243 vs
= kzalloc(sizeof(*vs
), GFP_KERNEL
);
3245 return ERR_PTR(-ENOMEM
);
3247 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
3248 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
3250 sock
= vxlan_create_sock(net
, ipv6
, port
, flags
, ifindex
);
3253 return ERR_CAST(sock
);
3257 refcount_set(&vs
->refcnt
, 1);
3258 vs
->flags
= (flags
& VXLAN_F_RCV_FLAGS
);
3260 spin_lock(&vn
->sock_lock
);
3261 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, port
));
3262 udp_tunnel_notify_add_rx_port(sock
,
3263 (vs
->flags
& VXLAN_F_GPE
) ?
3264 UDP_TUNNEL_TYPE_VXLAN_GPE
:
3265 UDP_TUNNEL_TYPE_VXLAN
);
3266 spin_unlock(&vn
->sock_lock
);
3268 /* Mark socket as an encapsulation socket. */
3269 memset(&tunnel_cfg
, 0, sizeof(tunnel_cfg
));
3270 tunnel_cfg
.sk_user_data
= vs
;
3271 tunnel_cfg
.encap_type
= 1;
3272 tunnel_cfg
.encap_rcv
= vxlan_rcv
;
3273 tunnel_cfg
.encap_err_lookup
= vxlan_err_lookup
;
3274 tunnel_cfg
.encap_destroy
= NULL
;
3275 tunnel_cfg
.gro_receive
= vxlan_gro_receive
;
3276 tunnel_cfg
.gro_complete
= vxlan_gro_complete
;
3278 setup_udp_tunnel_sock(net
, sock
, &tunnel_cfg
);
3283 static int __vxlan_sock_add(struct vxlan_dev
*vxlan
, bool ipv6
)
3285 struct vxlan_net
*vn
= net_generic(vxlan
->net
, vxlan_net_id
);
3286 struct vxlan_sock
*vs
= NULL
;
3287 struct vxlan_dev_node
*node
;
3288 int l3mdev_index
= 0;
3290 if (vxlan
->cfg
.remote_ifindex
)
3291 l3mdev_index
= l3mdev_master_upper_ifindex_by_index(
3292 vxlan
->net
, vxlan
->cfg
.remote_ifindex
);
3294 if (!vxlan
->cfg
.no_share
) {
3295 spin_lock(&vn
->sock_lock
);
3296 vs
= vxlan_find_sock(vxlan
->net
, ipv6
? AF_INET6
: AF_INET
,
3297 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
,
3299 if (vs
&& !refcount_inc_not_zero(&vs
->refcnt
)) {
3300 spin_unlock(&vn
->sock_lock
);
3303 spin_unlock(&vn
->sock_lock
);
3306 vs
= vxlan_socket_create(vxlan
->net
, ipv6
,
3307 vxlan
->cfg
.dst_port
, vxlan
->cfg
.flags
,
3311 #if IS_ENABLED(CONFIG_IPV6)
3313 rcu_assign_pointer(vxlan
->vn6_sock
, vs
);
3314 node
= &vxlan
->hlist6
;
3318 rcu_assign_pointer(vxlan
->vn4_sock
, vs
);
3319 node
= &vxlan
->hlist4
;
3321 vxlan_vs_add_dev(vs
, vxlan
, node
);
3325 static int vxlan_sock_add(struct vxlan_dev
*vxlan
)
3327 bool metadata
= vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
;
3328 bool ipv6
= vxlan
->cfg
.flags
& VXLAN_F_IPV6
|| metadata
;
3329 bool ipv4
= !ipv6
|| metadata
;
3332 RCU_INIT_POINTER(vxlan
->vn4_sock
, NULL
);
3333 #if IS_ENABLED(CONFIG_IPV6)
3334 RCU_INIT_POINTER(vxlan
->vn6_sock
, NULL
);
3336 ret
= __vxlan_sock_add(vxlan
, true);
3337 if (ret
< 0 && ret
!= -EAFNOSUPPORT
)
3342 ret
= __vxlan_sock_add(vxlan
, false);
3344 vxlan_sock_release(vxlan
);
3348 static int vxlan_config_validate(struct net
*src_net
, struct vxlan_config
*conf
,
3349 struct net_device
**lower
,
3350 struct vxlan_dev
*old
,
3351 struct netlink_ext_ack
*extack
)
3353 struct vxlan_net
*vn
= net_generic(src_net
, vxlan_net_id
);
3354 struct vxlan_dev
*tmp
;
3355 bool use_ipv6
= false;
3357 if (conf
->flags
& VXLAN_F_GPE
) {
3358 /* For now, allow GPE only together with
3359 * COLLECT_METADATA. This can be relaxed later; in such
3360 * case, the other side of the PtP link will have to be
3363 if ((conf
->flags
& ~VXLAN_F_ALLOWED_GPE
) ||
3364 !(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
3365 NL_SET_ERR_MSG(extack
,
3366 "VXLAN GPE does not support this combination of attributes");
3371 if (!conf
->remote_ip
.sa
.sa_family
&& !conf
->saddr
.sa
.sa_family
) {
3372 /* Unless IPv6 is explicitly requested, assume IPv4 */
3373 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
3374 conf
->saddr
.sa
.sa_family
= AF_INET
;
3375 } else if (!conf
->remote_ip
.sa
.sa_family
) {
3376 conf
->remote_ip
.sa
.sa_family
= conf
->saddr
.sa
.sa_family
;
3377 } else if (!conf
->saddr
.sa
.sa_family
) {
3378 conf
->saddr
.sa
.sa_family
= conf
->remote_ip
.sa
.sa_family
;
3381 if (conf
->saddr
.sa
.sa_family
!= conf
->remote_ip
.sa
.sa_family
) {
3382 NL_SET_ERR_MSG(extack
,
3383 "Local and remote address must be from the same family");
3387 if (vxlan_addr_multicast(&conf
->saddr
)) {
3388 NL_SET_ERR_MSG(extack
, "Local address cannot be multicast");
3392 if (conf
->saddr
.sa
.sa_family
== AF_INET6
) {
3393 if (!IS_ENABLED(CONFIG_IPV6
)) {
3394 NL_SET_ERR_MSG(extack
,
3395 "IPv6 support not enabled in the kernel");
3396 return -EPFNOSUPPORT
;
3399 conf
->flags
|= VXLAN_F_IPV6
;
3401 if (!(conf
->flags
& VXLAN_F_COLLECT_METADATA
)) {
3403 ipv6_addr_type(&conf
->saddr
.sin6
.sin6_addr
);
3405 ipv6_addr_type(&conf
->remote_ip
.sin6
.sin6_addr
);
3407 if (local_type
& IPV6_ADDR_LINKLOCAL
) {
3408 if (!(remote_type
& IPV6_ADDR_LINKLOCAL
) &&
3409 (remote_type
!= IPV6_ADDR_ANY
)) {
3410 NL_SET_ERR_MSG(extack
,
3411 "Invalid combination of local and remote address scopes");
3415 conf
->flags
|= VXLAN_F_IPV6_LINKLOCAL
;
3418 (IPV6_ADDR_UNICAST
| IPV6_ADDR_LINKLOCAL
)) {
3419 NL_SET_ERR_MSG(extack
,
3420 "Invalid combination of local and remote address scopes");
3424 conf
->flags
&= ~VXLAN_F_IPV6_LINKLOCAL
;
3429 if (conf
->label
&& !use_ipv6
) {
3430 NL_SET_ERR_MSG(extack
,
3431 "Label attribute only applies to IPv6 VXLAN devices");
3435 if (conf
->remote_ifindex
) {
3436 struct net_device
*lowerdev
;
3438 lowerdev
= __dev_get_by_index(src_net
, conf
->remote_ifindex
);
3440 NL_SET_ERR_MSG(extack
,
3441 "Invalid local interface, device not found");
3445 #if IS_ENABLED(CONFIG_IPV6)
3447 struct inet6_dev
*idev
= __in6_dev_get(lowerdev
);
3448 if (idev
&& idev
->cnf
.disable_ipv6
) {
3449 NL_SET_ERR_MSG(extack
,
3450 "IPv6 support disabled by administrator");
3458 if (vxlan_addr_multicast(&conf
->remote_ip
)) {
3459 NL_SET_ERR_MSG(extack
,
3460 "Local interface required for multicast remote destination");
3465 #if IS_ENABLED(CONFIG_IPV6)
3466 if (conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) {
3467 NL_SET_ERR_MSG(extack
,
3468 "Local interface required for link-local local/remote addresses");
3476 if (!conf
->dst_port
) {
3477 if (conf
->flags
& VXLAN_F_GPE
)
3478 conf
->dst_port
= htons(4790); /* IANA VXLAN-GPE port */
3480 conf
->dst_port
= htons(vxlan_port
);
3483 if (!conf
->age_interval
)
3484 conf
->age_interval
= FDB_AGE_DEFAULT
;
3486 list_for_each_entry(tmp
, &vn
->vxlan_list
, next
) {
3490 if (tmp
->cfg
.vni
!= conf
->vni
)
3492 if (tmp
->cfg
.dst_port
!= conf
->dst_port
)
3494 if ((tmp
->cfg
.flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)) !=
3495 (conf
->flags
& (VXLAN_F_RCV_FLAGS
| VXLAN_F_IPV6
)))
3498 if ((conf
->flags
& VXLAN_F_IPV6_LINKLOCAL
) &&
3499 tmp
->cfg
.remote_ifindex
!= conf
->remote_ifindex
)
3502 NL_SET_ERR_MSG(extack
,
3503 "A VXLAN device with the specified VNI already exists");
3510 static void vxlan_config_apply(struct net_device
*dev
,
3511 struct vxlan_config
*conf
,
3512 struct net_device
*lowerdev
,
3513 struct net
*src_net
,
3516 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3517 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
3518 unsigned short needed_headroom
= ETH_HLEN
;
3519 bool use_ipv6
= !!(conf
->flags
& VXLAN_F_IPV6
);
3520 int max_mtu
= ETH_MAX_MTU
;
3523 if (conf
->flags
& VXLAN_F_GPE
)
3524 vxlan_raw_setup(dev
);
3526 vxlan_ether_setup(dev
);
3529 dev
->mtu
= conf
->mtu
;
3531 vxlan
->net
= src_net
;
3534 dst
->remote_vni
= conf
->vni
;
3536 memcpy(&dst
->remote_ip
, &conf
->remote_ip
, sizeof(conf
->remote_ip
));
3539 dst
->remote_ifindex
= conf
->remote_ifindex
;
3541 dev
->gso_max_size
= lowerdev
->gso_max_size
;
3542 dev
->gso_max_segs
= lowerdev
->gso_max_segs
;
3544 needed_headroom
= lowerdev
->hard_header_len
;
3546 max_mtu
= lowerdev
->mtu
- (use_ipv6
? VXLAN6_HEADROOM
:
3548 if (max_mtu
< ETH_MIN_MTU
)
3549 max_mtu
= ETH_MIN_MTU
;
3551 if (!changelink
&& !conf
->mtu
)
3555 if (dev
->mtu
> max_mtu
)
3558 if (use_ipv6
|| conf
->flags
& VXLAN_F_COLLECT_METADATA
)
3559 needed_headroom
+= VXLAN6_HEADROOM
;
3561 needed_headroom
+= VXLAN_HEADROOM
;
3562 dev
->needed_headroom
= needed_headroom
;
3564 memcpy(&vxlan
->cfg
, conf
, sizeof(*conf
));
3567 static int vxlan_dev_configure(struct net
*src_net
, struct net_device
*dev
,
3568 struct vxlan_config
*conf
, bool changelink
,
3569 struct netlink_ext_ack
*extack
)
3571 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3572 struct net_device
*lowerdev
;
3575 ret
= vxlan_config_validate(src_net
, conf
, &lowerdev
, vxlan
, extack
);
3579 vxlan_config_apply(dev
, conf
, lowerdev
, src_net
, changelink
);
3584 static int __vxlan_dev_create(struct net
*net
, struct net_device
*dev
,
3585 struct vxlan_config
*conf
,
3586 struct netlink_ext_ack
*extack
)
3588 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
3589 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3590 struct net_device
*remote_dev
= NULL
;
3591 struct vxlan_fdb
*f
= NULL
;
3592 bool unregister
= false;
3593 struct vxlan_rdst
*dst
;
3596 dst
= &vxlan
->default_dst
;
3597 err
= vxlan_dev_configure(net
, dev
, conf
, false, extack
);
3601 dev
->ethtool_ops
= &vxlan_ethtool_ops
;
3603 /* create an fdb entry for a valid default destination */
3604 if (!vxlan_addr_any(&dst
->remote_ip
)) {
3605 err
= vxlan_fdb_create(vxlan
, all_zeros_mac
,
3607 NUD_REACHABLE
| NUD_PERMANENT
,
3608 vxlan
->cfg
.dst_port
,
3611 dst
->remote_ifindex
,
3617 err
= register_netdevice(dev
);
3622 if (dst
->remote_ifindex
) {
3623 remote_dev
= __dev_get_by_index(net
, dst
->remote_ifindex
);
3627 err
= netdev_upper_dev_link(remote_dev
, dev
, extack
);
3632 err
= rtnl_configure_link(dev
, NULL
);
3637 vxlan_fdb_insert(vxlan
, all_zeros_mac
, dst
->remote_vni
, f
);
3639 /* notify default fdb entry */
3640 err
= vxlan_fdb_notify(vxlan
, f
, first_remote_rtnl(f
),
3641 RTM_NEWNEIGH
, true, extack
);
3643 vxlan_fdb_destroy(vxlan
, f
, false, false);
3645 netdev_upper_dev_unlink(remote_dev
, dev
);
3650 list_add(&vxlan
->next
, &vn
->vxlan_list
);
3652 dst
->remote_dev
= remote_dev
;
3656 netdev_upper_dev_unlink(remote_dev
, dev
);
3658 /* unregister_netdevice() destroys the default FDB entry with deletion
3659 * notification. But the addition notification was not sent yet, so
3660 * destroy the entry by hand here.
3663 __vxlan_fdb_free(f
);
3666 unregister_netdevice(dev
);
3670 /* Set/clear flags based on attribute */
3671 static int vxlan_nl2flag(struct vxlan_config
*conf
, struct nlattr
*tb
[],
3672 int attrtype
, unsigned long mask
, bool changelink
,
3673 bool changelink_supported
,
3674 struct netlink_ext_ack
*extack
)
3676 unsigned long flags
;
3681 if (changelink
&& !changelink_supported
) {
3682 vxlan_flag_attr_error(attrtype
, extack
);
3686 if (vxlan_policy
[attrtype
].type
== NLA_FLAG
)
3687 flags
= conf
->flags
| mask
;
3688 else if (nla_get_u8(tb
[attrtype
]))
3689 flags
= conf
->flags
| mask
;
3691 flags
= conf
->flags
& ~mask
;
3693 conf
->flags
= flags
;
3698 static int vxlan_nl2conf(struct nlattr
*tb
[], struct nlattr
*data
[],
3699 struct net_device
*dev
, struct vxlan_config
*conf
,
3700 bool changelink
, struct netlink_ext_ack
*extack
)
3702 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3705 memset(conf
, 0, sizeof(*conf
));
3707 /* if changelink operation, start with old existing cfg */
3709 memcpy(conf
, &vxlan
->cfg
, sizeof(*conf
));
3711 if (data
[IFLA_VXLAN_ID
]) {
3712 __be32 vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3714 if (changelink
&& (vni
!= conf
->vni
)) {
3715 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_ID
], "Cannot change VNI");
3718 conf
->vni
= cpu_to_be32(nla_get_u32(data
[IFLA_VXLAN_ID
]));
3721 if (data
[IFLA_VXLAN_GROUP
]) {
3722 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET
)) {
3723 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_GROUP
], "New group address family does not match old group");
3727 conf
->remote_ip
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_GROUP
]);
3728 conf
->remote_ip
.sa
.sa_family
= AF_INET
;
3729 } else if (data
[IFLA_VXLAN_GROUP6
]) {
3730 if (!IS_ENABLED(CONFIG_IPV6
)) {
3731 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_GROUP6
], "IPv6 support not enabled in the kernel");
3732 return -EPFNOSUPPORT
;
3735 if (changelink
&& (conf
->remote_ip
.sa
.sa_family
!= AF_INET6
)) {
3736 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_GROUP6
], "New group address family does not match old group");
3740 conf
->remote_ip
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_GROUP6
]);
3741 conf
->remote_ip
.sa
.sa_family
= AF_INET6
;
3744 if (data
[IFLA_VXLAN_LOCAL
]) {
3745 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET
)) {
3746 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_LOCAL
], "New local address family does not match old");
3750 conf
->saddr
.sin
.sin_addr
.s_addr
= nla_get_in_addr(data
[IFLA_VXLAN_LOCAL
]);
3751 conf
->saddr
.sa
.sa_family
= AF_INET
;
3752 } else if (data
[IFLA_VXLAN_LOCAL6
]) {
3753 if (!IS_ENABLED(CONFIG_IPV6
)) {
3754 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_LOCAL6
], "IPv6 support not enabled in the kernel");
3755 return -EPFNOSUPPORT
;
3758 if (changelink
&& (conf
->saddr
.sa
.sa_family
!= AF_INET6
)) {
3759 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_LOCAL6
], "New local address family does not match old");
3763 /* TODO: respect scope id */
3764 conf
->saddr
.sin6
.sin6_addr
= nla_get_in6_addr(data
[IFLA_VXLAN_LOCAL6
]);
3765 conf
->saddr
.sa
.sa_family
= AF_INET6
;
3768 if (data
[IFLA_VXLAN_LINK
])
3769 conf
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]);
3771 if (data
[IFLA_VXLAN_TOS
])
3772 conf
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
3774 if (data
[IFLA_VXLAN_TTL
])
3775 conf
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
3777 if (data
[IFLA_VXLAN_TTL_INHERIT
]) {
3778 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_TTL_INHERIT
,
3779 VXLAN_F_TTL_INHERIT
, changelink
, false,
3786 if (data
[IFLA_VXLAN_LABEL
])
3787 conf
->label
= nla_get_be32(data
[IFLA_VXLAN_LABEL
]) &
3788 IPV6_FLOWLABEL_MASK
;
3790 if (data
[IFLA_VXLAN_LEARNING
]) {
3791 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_LEARNING
,
3792 VXLAN_F_LEARN
, changelink
, true,
3796 } else if (!changelink
) {
3797 /* default to learn on a new device */
3798 conf
->flags
|= VXLAN_F_LEARN
;
3801 if (data
[IFLA_VXLAN_AGEING
])
3802 conf
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
3804 if (data
[IFLA_VXLAN_PROXY
]) {
3805 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_PROXY
,
3806 VXLAN_F_PROXY
, changelink
, false,
3812 if (data
[IFLA_VXLAN_RSC
]) {
3813 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_RSC
,
3814 VXLAN_F_RSC
, changelink
, false,
3820 if (data
[IFLA_VXLAN_L2MISS
]) {
3821 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_L2MISS
,
3822 VXLAN_F_L2MISS
, changelink
, false,
3828 if (data
[IFLA_VXLAN_L3MISS
]) {
3829 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_L3MISS
,
3830 VXLAN_F_L3MISS
, changelink
, false,
3836 if (data
[IFLA_VXLAN_LIMIT
]) {
3838 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_LIMIT
],
3839 "Cannot change limit");
3842 conf
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
3845 if (data
[IFLA_VXLAN_COLLECT_METADATA
]) {
3846 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_COLLECT_METADATA
,
3847 VXLAN_F_COLLECT_METADATA
, changelink
, false,
3853 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
3855 const struct ifla_vxlan_port_range
*p
3856 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
3857 conf
->port_min
= ntohs(p
->low
);
3858 conf
->port_max
= ntohs(p
->high
);
3860 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_PORT_RANGE
],
3861 "Cannot change port range");
3866 if (data
[IFLA_VXLAN_PORT
]) {
3868 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_PORT
],
3869 "Cannot change port");
3872 conf
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
3875 if (data
[IFLA_VXLAN_UDP_CSUM
]) {
3877 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_VXLAN_UDP_CSUM
],
3878 "Cannot change UDP_CSUM flag");
3881 if (!nla_get_u8(data
[IFLA_VXLAN_UDP_CSUM
]))
3882 conf
->flags
|= VXLAN_F_UDP_ZERO_CSUM_TX
;
3885 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_TX
]) {
3886 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
3887 VXLAN_F_UDP_ZERO_CSUM6_TX
, changelink
,
3893 if (data
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX
]) {
3894 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
3895 VXLAN_F_UDP_ZERO_CSUM6_RX
, changelink
,
3901 if (data
[IFLA_VXLAN_REMCSUM_TX
]) {
3902 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_REMCSUM_TX
,
3903 VXLAN_F_REMCSUM_TX
, changelink
, false,
3909 if (data
[IFLA_VXLAN_REMCSUM_RX
]) {
3910 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_REMCSUM_RX
,
3911 VXLAN_F_REMCSUM_RX
, changelink
, false,
3917 if (data
[IFLA_VXLAN_GBP
]) {
3918 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_GBP
,
3919 VXLAN_F_GBP
, changelink
, false, extack
);
3924 if (data
[IFLA_VXLAN_GPE
]) {
3925 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_GPE
,
3926 VXLAN_F_GPE
, changelink
, false,
3932 if (data
[IFLA_VXLAN_REMCSUM_NOPARTIAL
]) {
3933 err
= vxlan_nl2flag(conf
, data
, IFLA_VXLAN_REMCSUM_NOPARTIAL
,
3934 VXLAN_F_REMCSUM_NOPARTIAL
, changelink
,
3942 NL_SET_ERR_MSG_ATTR(extack
, tb
[IFLA_MTU
],
3943 "Cannot change mtu");
3946 conf
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
3949 if (data
[IFLA_VXLAN_DF
])
3950 conf
->df
= nla_get_u8(data
[IFLA_VXLAN_DF
]);
3955 static int vxlan_newlink(struct net
*src_net
, struct net_device
*dev
,
3956 struct nlattr
*tb
[], struct nlattr
*data
[],
3957 struct netlink_ext_ack
*extack
)
3959 struct vxlan_config conf
;
3962 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, false, extack
);
3966 return __vxlan_dev_create(src_net
, dev
, &conf
, extack
);
3969 static int vxlan_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
3970 struct nlattr
*data
[],
3971 struct netlink_ext_ack
*extack
)
3973 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
3974 struct net_device
*lowerdev
;
3975 struct vxlan_config conf
;
3976 struct vxlan_rdst
*dst
;
3979 dst
= &vxlan
->default_dst
;
3980 err
= vxlan_nl2conf(tb
, data
, dev
, &conf
, true, extack
);
3984 err
= vxlan_config_validate(vxlan
->net
, &conf
, &lowerdev
,
3989 if (dst
->remote_dev
== lowerdev
)
3992 err
= netdev_adjacent_change_prepare(dst
->remote_dev
, lowerdev
, dev
,
3997 /* handle default dst entry */
3998 if (!vxlan_addr_equal(&conf
.remote_ip
, &dst
->remote_ip
)) {
3999 u32 hash_index
= fdb_head_index(vxlan
, all_zeros_mac
, conf
.vni
);
4001 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
4002 if (!vxlan_addr_any(&conf
.remote_ip
)) {
4003 err
= vxlan_fdb_update(vxlan
, all_zeros_mac
,
4005 NUD_REACHABLE
| NUD_PERMANENT
,
4006 NLM_F_APPEND
| NLM_F_CREATE
,
4007 vxlan
->cfg
.dst_port
,
4009 conf
.remote_ifindex
,
4010 NTF_SELF
, true, extack
);
4012 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
4013 netdev_adjacent_change_abort(dst
->remote_dev
,
4018 if (!vxlan_addr_any(&dst
->remote_ip
))
4019 __vxlan_fdb_delete(vxlan
, all_zeros_mac
,
4021 vxlan
->cfg
.dst_port
,
4024 dst
->remote_ifindex
,
4026 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
4029 if (conf
.age_interval
!= vxlan
->cfg
.age_interval
)
4030 mod_timer(&vxlan
->age_timer
, jiffies
);
4032 netdev_adjacent_change_commit(dst
->remote_dev
, lowerdev
, dev
);
4033 if (lowerdev
&& lowerdev
!= dst
->remote_dev
) {
4034 dst
->remote_dev
= lowerdev
;
4035 netdev_update_lockdep_key(lowerdev
);
4037 vxlan_config_apply(dev
, &conf
, lowerdev
, vxlan
->net
, true);
4041 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
4043 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4045 vxlan_flush(vxlan
, true);
4047 list_del(&vxlan
->next
);
4048 unregister_netdevice_queue(dev
, head
);
4049 if (vxlan
->default_dst
.remote_dev
)
4050 netdev_upper_dev_unlink(vxlan
->default_dst
.remote_dev
, dev
);
4053 static size_t vxlan_get_size(const struct net_device
*dev
)
4056 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
4057 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_GROUP{6} */
4058 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
4059 nla_total_size(sizeof(struct in6_addr
)) + /* IFLA_VXLAN_LOCAL{6} */
4060 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
4061 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL_INHERIT */
4062 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
4063 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_DF */
4064 nla_total_size(sizeof(__be32
)) + /* IFLA_VXLAN_LABEL */
4065 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
4066 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
4067 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
4068 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
4069 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
4070 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_COLLECT_METADATA */
4071 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
4072 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
4073 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
4074 nla_total_size(sizeof(__be16
)) + /* IFLA_VXLAN_PORT */
4075 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_CSUM */
4076 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4077 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4078 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_TX */
4079 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_REMCSUM_RX */
4083 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
4085 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4086 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
4087 struct ifla_vxlan_port_range ports
= {
4088 .low
= htons(vxlan
->cfg
.port_min
),
4089 .high
= htons(vxlan
->cfg
.port_max
),
4092 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, be32_to_cpu(dst
->remote_vni
)))
4093 goto nla_put_failure
;
4095 if (!vxlan_addr_any(&dst
->remote_ip
)) {
4096 if (dst
->remote_ip
.sa
.sa_family
== AF_INET
) {
4097 if (nla_put_in_addr(skb
, IFLA_VXLAN_GROUP
,
4098 dst
->remote_ip
.sin
.sin_addr
.s_addr
))
4099 goto nla_put_failure
;
4100 #if IS_ENABLED(CONFIG_IPV6)
4102 if (nla_put_in6_addr(skb
, IFLA_VXLAN_GROUP6
,
4103 &dst
->remote_ip
.sin6
.sin6_addr
))
4104 goto nla_put_failure
;
4109 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
4110 goto nla_put_failure
;
4112 if (!vxlan_addr_any(&vxlan
->cfg
.saddr
)) {
4113 if (vxlan
->cfg
.saddr
.sa
.sa_family
== AF_INET
) {
4114 if (nla_put_in_addr(skb
, IFLA_VXLAN_LOCAL
,
4115 vxlan
->cfg
.saddr
.sin
.sin_addr
.s_addr
))
4116 goto nla_put_failure
;
4117 #if IS_ENABLED(CONFIG_IPV6)
4119 if (nla_put_in6_addr(skb
, IFLA_VXLAN_LOCAL6
,
4120 &vxlan
->cfg
.saddr
.sin6
.sin6_addr
))
4121 goto nla_put_failure
;
4126 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->cfg
.ttl
) ||
4127 nla_put_u8(skb
, IFLA_VXLAN_TTL_INHERIT
,
4128 !!(vxlan
->cfg
.flags
& VXLAN_F_TTL_INHERIT
)) ||
4129 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->cfg
.tos
) ||
4130 nla_put_u8(skb
, IFLA_VXLAN_DF
, vxlan
->cfg
.df
) ||
4131 nla_put_be32(skb
, IFLA_VXLAN_LABEL
, vxlan
->cfg
.label
) ||
4132 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
4133 !!(vxlan
->cfg
.flags
& VXLAN_F_LEARN
)) ||
4134 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
4135 !!(vxlan
->cfg
.flags
& VXLAN_F_PROXY
)) ||
4136 nla_put_u8(skb
, IFLA_VXLAN_RSC
,
4137 !!(vxlan
->cfg
.flags
& VXLAN_F_RSC
)) ||
4138 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
4139 !!(vxlan
->cfg
.flags
& VXLAN_F_L2MISS
)) ||
4140 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
4141 !!(vxlan
->cfg
.flags
& VXLAN_F_L3MISS
)) ||
4142 nla_put_u8(skb
, IFLA_VXLAN_COLLECT_METADATA
,
4143 !!(vxlan
->cfg
.flags
& VXLAN_F_COLLECT_METADATA
)) ||
4144 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->cfg
.age_interval
) ||
4145 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->cfg
.addrmax
) ||
4146 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->cfg
.dst_port
) ||
4147 nla_put_u8(skb
, IFLA_VXLAN_UDP_CSUM
,
4148 !(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM_TX
)) ||
4149 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_TX
,
4150 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_TX
)) ||
4151 nla_put_u8(skb
, IFLA_VXLAN_UDP_ZERO_CSUM6_RX
,
4152 !!(vxlan
->cfg
.flags
& VXLAN_F_UDP_ZERO_CSUM6_RX
)) ||
4153 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_TX
,
4154 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_TX
)) ||
4155 nla_put_u8(skb
, IFLA_VXLAN_REMCSUM_RX
,
4156 !!(vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_RX
)))
4157 goto nla_put_failure
;
4159 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
4160 goto nla_put_failure
;
4162 if (vxlan
->cfg
.flags
& VXLAN_F_GBP
&&
4163 nla_put_flag(skb
, IFLA_VXLAN_GBP
))
4164 goto nla_put_failure
;
4166 if (vxlan
->cfg
.flags
& VXLAN_F_GPE
&&
4167 nla_put_flag(skb
, IFLA_VXLAN_GPE
))
4168 goto nla_put_failure
;
4170 if (vxlan
->cfg
.flags
& VXLAN_F_REMCSUM_NOPARTIAL
&&
4171 nla_put_flag(skb
, IFLA_VXLAN_REMCSUM_NOPARTIAL
))
4172 goto nla_put_failure
;
4180 static struct net
*vxlan_get_link_net(const struct net_device
*dev
)
4182 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4187 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
4189 .maxtype
= IFLA_VXLAN_MAX
,
4190 .policy
= vxlan_policy
,
4191 .priv_size
= sizeof(struct vxlan_dev
),
4192 .setup
= vxlan_setup
,
4193 .validate
= vxlan_validate
,
4194 .newlink
= vxlan_newlink
,
4195 .changelink
= vxlan_changelink
,
4196 .dellink
= vxlan_dellink
,
4197 .get_size
= vxlan_get_size
,
4198 .fill_info
= vxlan_fill_info
,
4199 .get_link_net
= vxlan_get_link_net
,
4202 struct net_device
*vxlan_dev_create(struct net
*net
, const char *name
,
4203 u8 name_assign_type
,
4204 struct vxlan_config
*conf
)
4206 struct nlattr
*tb
[IFLA_MAX
+ 1];
4207 struct net_device
*dev
;
4210 memset(&tb
, 0, sizeof(tb
));
4212 dev
= rtnl_create_link(net
, name
, name_assign_type
,
4213 &vxlan_link_ops
, tb
, NULL
);
4217 err
= __vxlan_dev_create(net
, dev
, conf
, NULL
);
4220 return ERR_PTR(err
);
4223 err
= rtnl_configure_link(dev
, NULL
);
4225 LIST_HEAD(list_kill
);
4227 vxlan_dellink(dev
, &list_kill
);
4228 unregister_netdevice_many(&list_kill
);
4229 return ERR_PTR(err
);
4234 EXPORT_SYMBOL_GPL(vxlan_dev_create
);
4236 static void vxlan_handle_lowerdev_unregister(struct vxlan_net
*vn
,
4237 struct net_device
*dev
)
4239 struct vxlan_dev
*vxlan
, *next
;
4240 LIST_HEAD(list_kill
);
4242 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
4243 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
4245 /* In case we created vxlan device with carrier
4246 * and we loose the carrier due to module unload
4247 * we also need to remove vxlan device. In other
4248 * cases, it's not necessary and remote_ifindex
4249 * is 0 here, so no matches.
4251 if (dst
->remote_ifindex
== dev
->ifindex
)
4252 vxlan_dellink(vxlan
->dev
, &list_kill
);
4255 unregister_netdevice_many(&list_kill
);
4258 static int vxlan_netdevice_event(struct notifier_block
*unused
,
4259 unsigned long event
, void *ptr
)
4261 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
4262 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
4264 if (event
== NETDEV_UNREGISTER
) {
4265 vxlan_offload_rx_ports(dev
, false);
4266 vxlan_handle_lowerdev_unregister(vn
, dev
);
4267 } else if (event
== NETDEV_REGISTER
) {
4268 vxlan_offload_rx_ports(dev
, true);
4269 } else if (event
== NETDEV_UDP_TUNNEL_PUSH_INFO
||
4270 event
== NETDEV_UDP_TUNNEL_DROP_INFO
) {
4271 vxlan_offload_rx_ports(dev
, event
== NETDEV_UDP_TUNNEL_PUSH_INFO
);
4277 static struct notifier_block vxlan_notifier_block __read_mostly
= {
4278 .notifier_call
= vxlan_netdevice_event
,
4282 vxlan_fdb_offloaded_set(struct net_device
*dev
,
4283 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
)
4285 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4286 struct vxlan_rdst
*rdst
;
4287 struct vxlan_fdb
*f
;
4290 hash_index
= fdb_head_index(vxlan
, fdb_info
->eth_addr
, fdb_info
->vni
);
4292 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
4294 f
= vxlan_find_mac(vxlan
, fdb_info
->eth_addr
, fdb_info
->vni
);
4298 rdst
= vxlan_fdb_find_rdst(f
, &fdb_info
->remote_ip
,
4299 fdb_info
->remote_port
,
4300 fdb_info
->remote_vni
,
4301 fdb_info
->remote_ifindex
);
4305 rdst
->offloaded
= fdb_info
->offloaded
;
4308 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
4312 vxlan_fdb_external_learn_add(struct net_device
*dev
,
4313 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
)
4315 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4316 struct netlink_ext_ack
*extack
;
4320 hash_index
= fdb_head_index(vxlan
, fdb_info
->eth_addr
, fdb_info
->vni
);
4321 extack
= switchdev_notifier_info_to_extack(&fdb_info
->info
);
4323 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
4324 err
= vxlan_fdb_update(vxlan
, fdb_info
->eth_addr
, &fdb_info
->remote_ip
,
4326 NLM_F_CREATE
| NLM_F_REPLACE
,
4327 fdb_info
->remote_port
,
4329 fdb_info
->remote_vni
,
4330 fdb_info
->remote_ifindex
,
4331 NTF_USE
| NTF_SELF
| NTF_EXT_LEARNED
,
4333 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
4339 vxlan_fdb_external_learn_del(struct net_device
*dev
,
4340 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
)
4342 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
4343 struct vxlan_fdb
*f
;
4347 hash_index
= fdb_head_index(vxlan
, fdb_info
->eth_addr
, fdb_info
->vni
);
4348 spin_lock_bh(&vxlan
->hash_lock
[hash_index
]);
4350 f
= vxlan_find_mac(vxlan
, fdb_info
->eth_addr
, fdb_info
->vni
);
4353 else if (f
->flags
& NTF_EXT_LEARNED
)
4354 err
= __vxlan_fdb_delete(vxlan
, fdb_info
->eth_addr
,
4355 fdb_info
->remote_ip
,
4356 fdb_info
->remote_port
,
4358 fdb_info
->remote_vni
,
4359 fdb_info
->remote_ifindex
,
4362 spin_unlock_bh(&vxlan
->hash_lock
[hash_index
]);
4367 static int vxlan_switchdev_event(struct notifier_block
*unused
,
4368 unsigned long event
, void *ptr
)
4370 struct net_device
*dev
= switchdev_notifier_info_to_dev(ptr
);
4371 struct switchdev_notifier_vxlan_fdb_info
*fdb_info
;
4375 case SWITCHDEV_VXLAN_FDB_OFFLOADED
:
4376 vxlan_fdb_offloaded_set(dev
, ptr
);
4378 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE
:
4380 err
= vxlan_fdb_external_learn_add(dev
, fdb_info
);
4382 err
= notifier_from_errno(err
);
4385 fdb_info
->offloaded
= true;
4386 vxlan_fdb_offloaded_set(dev
, fdb_info
);
4388 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE
:
4390 err
= vxlan_fdb_external_learn_del(dev
, fdb_info
);
4392 err
= notifier_from_errno(err
);
4395 fdb_info
->offloaded
= false;
4396 vxlan_fdb_offloaded_set(dev
, fdb_info
);
4403 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly
= {
4404 .notifier_call
= vxlan_switchdev_event
,
4407 static __net_init
int vxlan_init_net(struct net
*net
)
4409 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
4412 INIT_LIST_HEAD(&vn
->vxlan_list
);
4413 spin_lock_init(&vn
->sock_lock
);
4415 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
4416 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
4421 static void vxlan_destroy_tunnels(struct net
*net
, struct list_head
*head
)
4423 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
4424 struct vxlan_dev
*vxlan
, *next
;
4425 struct net_device
*dev
, *aux
;
4428 for_each_netdev_safe(net
, dev
, aux
)
4429 if (dev
->rtnl_link_ops
== &vxlan_link_ops
)
4430 unregister_netdevice_queue(dev
, head
);
4432 list_for_each_entry_safe(vxlan
, next
, &vn
->vxlan_list
, next
) {
4433 /* If vxlan->dev is in the same netns, it has already been added
4434 * to the list by the previous loop.
4436 if (!net_eq(dev_net(vxlan
->dev
), net
))
4437 unregister_netdevice_queue(vxlan
->dev
, head
);
4440 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
4441 WARN_ON_ONCE(!hlist_empty(&vn
->sock_list
[h
]));
4444 static void __net_exit
vxlan_exit_batch_net(struct list_head
*net_list
)
4450 list_for_each_entry(net
, net_list
, exit_list
)
4451 vxlan_destroy_tunnels(net
, &list
);
4453 unregister_netdevice_many(&list
);
4457 static struct pernet_operations vxlan_net_ops
= {
4458 .init
= vxlan_init_net
,
4459 .exit_batch
= vxlan_exit_batch_net
,
4460 .id
= &vxlan_net_id
,
4461 .size
= sizeof(struct vxlan_net
),
4464 static int __init
vxlan_init_module(void)
4468 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
4470 rc
= register_pernet_subsys(&vxlan_net_ops
);
4474 rc
= register_netdevice_notifier(&vxlan_notifier_block
);
4478 rc
= register_switchdev_notifier(&vxlan_switchdev_notifier_block
);
4482 rc
= rtnl_link_register(&vxlan_link_ops
);
4488 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block
);
4490 unregister_netdevice_notifier(&vxlan_notifier_block
);
4492 unregister_pernet_subsys(&vxlan_net_ops
);
4496 late_initcall(vxlan_init_module
);
4498 static void __exit
vxlan_cleanup_module(void)
4500 rtnl_link_unregister(&vxlan_link_ops
);
4501 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block
);
4502 unregister_netdevice_notifier(&vxlan_notifier_block
);
4503 unregister_pernet_subsys(&vxlan_net_ops
);
4504 /* rcu_barrier() is called by netns */
4506 module_exit(vxlan_cleanup_module
);
4508 MODULE_LICENSE("GPL");
4509 MODULE_VERSION(VXLAN_VERSION
);
4510 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4511 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4512 MODULE_ALIAS_RTNL_LINK("vxlan");