2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012-2013 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/errno.h>
20 #include <linux/slab.h>
21 #include <linux/skbuff.h>
22 #include <linux/rculist.h>
23 #include <linux/netdevice.h>
26 #include <linux/udp.h>
27 #include <linux/igmp.h>
28 #include <linux/etherdevice.h>
29 #include <linux/if_ether.h>
30 #include <linux/hash.h>
31 #include <linux/ethtool.h>
33 #include <net/ndisc.h>
35 #include <net/ip_tunnels.h>
38 #include <net/rtnetlink.h>
39 #include <net/route.h>
40 #include <net/dsfield.h>
41 #include <net/inet_ecn.h>
42 #include <net/net_namespace.h>
43 #include <net/netns/generic.h>
45 #define VXLAN_VERSION "0.1"
47 #define PORT_HASH_BITS 8
48 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
49 #define VNI_HASH_BITS 10
50 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
51 #define FDB_HASH_BITS 8
52 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
53 #define FDB_AGE_DEFAULT 300 /* 5 min */
54 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
56 #define VXLAN_N_VID (1u << 24)
57 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
58 /* IP header + UDP + VXLAN + Ethernet header */
59 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
61 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
63 /* VXLAN protocol header */
69 /* UDP port for VXLAN traffic.
70 * The IANA assigned port is 4789, but the Linux default is 8472
71 * for compatability with early adopters.
73 static unsigned int vxlan_port __read_mostly
= 8472;
74 module_param_named(udp_port
, vxlan_port
, uint
, 0444);
75 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
77 static bool log_ecn_error
= true;
78 module_param(log_ecn_error
, bool, 0644);
79 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
81 static unsigned int vxlan_net_id
;
83 /* per UDP socket information */
85 struct hlist_node hlist
;
87 struct work_struct del_work
;
90 struct hlist_head vni_list
[VNI_HASH_SIZE
];
93 /* per-network namespace private data for this module */
95 struct list_head vxlan_list
;
96 struct hlist_head sock_list
[PORT_HASH_SIZE
];
104 struct vxlan_rdst
*remote_next
;
107 /* Forwarding table entry */
109 struct hlist_node hlist
; /* linked list of entries */
111 unsigned long updated
; /* jiffies */
113 struct vxlan_rdst remote
;
114 u16 state
; /* see ndm_state */
115 u8 flags
; /* see ndm_flags */
116 u8 eth_addr
[ETH_ALEN
];
119 /* Pseudo network device */
121 struct hlist_node hlist
; /* vni hash table */
122 struct list_head next
; /* vxlan's per namespace list */
123 struct vxlan_sock
*vn_sock
; /* listening socket */
124 struct net_device
*dev
;
125 struct vxlan_rdst default_dst
; /* default destination */
126 __be32 saddr
; /* source address */
128 __u16 port_min
; /* source port range */
130 __u8 tos
; /* TOS override */
132 u32 flags
; /* VXLAN_F_* below */
134 unsigned long age_interval
;
135 struct timer_list age_timer
;
136 spinlock_t hash_lock
;
137 unsigned int addrcnt
;
138 unsigned int addrmax
;
140 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
143 #define VXLAN_F_LEARN 0x01
144 #define VXLAN_F_PROXY 0x02
145 #define VXLAN_F_RSC 0x04
146 #define VXLAN_F_L2MISS 0x08
147 #define VXLAN_F_L3MISS 0x10
149 /* salt for hash table */
150 static u32 vxlan_salt __read_mostly
;
152 /* Virtual Network hash table head */
153 static inline struct hlist_head
*vni_head(struct vxlan_sock
*vs
, u32 id
)
155 return &vs
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
158 /* Socket hash table head */
159 static inline struct hlist_head
*vs_head(struct net
*net
, __be16 port
)
161 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
163 return &vn
->sock_list
[hash_32(ntohs(port
), PORT_HASH_BITS
)];
166 /* Find VXLAN socket based on network namespace and UDP port */
167 static struct vxlan_sock
*vxlan_find_port(struct net
*net
, __be16 port
)
169 struct vxlan_sock
*vs
;
171 hlist_for_each_entry_rcu(vs
, vs_head(net
, port
), hlist
) {
172 if (inet_sk(vs
->sock
->sk
)->inet_sport
== port
)
178 /* Look up VNI in a per net namespace table */
179 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
, __be16 port
)
181 struct vxlan_sock
*vs
;
182 struct vxlan_dev
*vxlan
;
184 vs
= vxlan_find_port(net
, port
);
188 hlist_for_each_entry_rcu(vxlan
, vni_head(vs
, id
), hlist
) {
189 if (vxlan
->default_dst
.remote_vni
== id
)
196 /* Fill in neighbour message in skbuff. */
197 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
198 const struct vxlan_fdb
*fdb
,
199 u32 portid
, u32 seq
, int type
, unsigned int flags
,
200 const struct vxlan_rdst
*rdst
)
202 unsigned long now
= jiffies
;
203 struct nda_cacheinfo ci
;
204 struct nlmsghdr
*nlh
;
206 bool send_ip
, send_eth
;
208 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
212 ndm
= nlmsg_data(nlh
);
213 memset(ndm
, 0, sizeof(*ndm
));
215 send_eth
= send_ip
= true;
217 if (type
== RTM_GETNEIGH
) {
218 ndm
->ndm_family
= AF_INET
;
219 send_ip
= rdst
->remote_ip
!= htonl(INADDR_ANY
);
220 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
222 ndm
->ndm_family
= AF_BRIDGE
;
223 ndm
->ndm_state
= fdb
->state
;
224 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
225 ndm
->ndm_flags
= fdb
->flags
;
226 ndm
->ndm_type
= NDA_DST
;
228 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
229 goto nla_put_failure
;
231 if (send_ip
&& nla_put_be32(skb
, NDA_DST
, rdst
->remote_ip
))
232 goto nla_put_failure
;
234 if (rdst
->remote_port
&& rdst
->remote_port
!= vxlan
->dst_port
&&
235 nla_put_be16(skb
, NDA_PORT
, rdst
->remote_port
))
236 goto nla_put_failure
;
237 if (rdst
->remote_vni
!= vxlan
->default_dst
.remote_vni
&&
238 nla_put_be32(skb
, NDA_VNI
, rdst
->remote_vni
))
239 goto nla_put_failure
;
240 if (rdst
->remote_ifindex
&&
241 nla_put_u32(skb
, NDA_IFINDEX
, rdst
->remote_ifindex
))
242 goto nla_put_failure
;
244 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
245 ci
.ndm_confirmed
= 0;
246 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
249 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
250 goto nla_put_failure
;
252 return nlmsg_end(skb
, nlh
);
255 nlmsg_cancel(skb
, nlh
);
259 static inline size_t vxlan_nlmsg_size(void)
261 return NLMSG_ALIGN(sizeof(struct ndmsg
))
262 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
263 + nla_total_size(sizeof(__be32
)) /* NDA_DST */
264 + nla_total_size(sizeof(__be16
)) /* NDA_PORT */
265 + nla_total_size(sizeof(__be32
)) /* NDA_VNI */
266 + nla_total_size(sizeof(__u32
)) /* NDA_IFINDEX */
267 + nla_total_size(sizeof(struct nda_cacheinfo
));
270 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
,
271 const struct vxlan_fdb
*fdb
, int type
)
273 struct net
*net
= dev_net(vxlan
->dev
);
277 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
281 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0, &fdb
->remote
);
283 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
284 WARN_ON(err
== -EMSGSIZE
);
289 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
293 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
296 static void vxlan_ip_miss(struct net_device
*dev
, __be32 ipa
)
298 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
301 memset(&f
, 0, sizeof f
);
303 f
.remote
.remote_ip
= ipa
; /* goes to NDA_DST */
304 f
.remote
.remote_vni
= VXLAN_N_VID
;
306 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
309 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
313 memset(&f
, 0, sizeof f
);
315 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
317 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
320 /* Hash Ethernet address */
321 static u32
eth_hash(const unsigned char *addr
)
323 u64 value
= get_unaligned((u64
*)addr
);
325 /* only want 6 bytes */
331 return hash_64(value
, FDB_HASH_BITS
);
334 /* Hash chain to use given mac address */
335 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
338 return &vxlan
->fdb_head
[eth_hash(mac
)];
341 /* Look up Ethernet address in forwarding table */
342 static struct vxlan_fdb
*__vxlan_find_mac(struct vxlan_dev
*vxlan
,
346 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
349 hlist_for_each_entry_rcu(f
, head
, hlist
) {
350 if (compare_ether_addr(mac
, f
->eth_addr
) == 0)
357 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
362 f
= __vxlan_find_mac(vxlan
, mac
);
369 /* Add/update destinations for multicast */
370 static int vxlan_fdb_append(struct vxlan_fdb
*f
,
371 __be32 ip
, __be16 port
, __u32 vni
, __u32 ifindex
)
373 struct vxlan_rdst
*rd_prev
, *rd
;
376 for (rd
= &f
->remote
; rd
; rd
= rd
->remote_next
) {
377 if (rd
->remote_ip
== ip
&&
378 rd
->remote_port
== port
&&
379 rd
->remote_vni
== vni
&&
380 rd
->remote_ifindex
== ifindex
)
384 rd
= kmalloc(sizeof(*rd
), GFP_ATOMIC
);
388 rd
->remote_port
= port
;
389 rd
->remote_vni
= vni
;
390 rd
->remote_ifindex
= ifindex
;
391 rd
->remote_next
= NULL
;
392 rd_prev
->remote_next
= rd
;
396 /* Add new entry to forwarding table -- assumes lock held */
397 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
398 const u8
*mac
, __be32 ip
,
399 __u16 state
, __u16 flags
,
400 __be16 port
, __u32 vni
, __u32 ifindex
,
406 f
= __vxlan_find_mac(vxlan
, mac
);
408 if (flags
& NLM_F_EXCL
) {
409 netdev_dbg(vxlan
->dev
,
410 "lost race to create %pM\n", mac
);
413 if (f
->state
!= state
) {
415 f
->updated
= jiffies
;
418 if (f
->flags
!= ndm_flags
) {
419 f
->flags
= ndm_flags
;
420 f
->updated
= jiffies
;
423 if ((flags
& NLM_F_APPEND
) &&
424 is_multicast_ether_addr(f
->eth_addr
)) {
425 int rc
= vxlan_fdb_append(f
, ip
, port
, vni
, ifindex
);
432 if (!(flags
& NLM_F_CREATE
))
435 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
438 netdev_dbg(vxlan
->dev
, "add %pM -> %pI4\n", mac
, &ip
);
439 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
444 f
->remote
.remote_ip
= ip
;
445 f
->remote
.remote_port
= port
;
446 f
->remote
.remote_vni
= vni
;
447 f
->remote
.remote_ifindex
= ifindex
;
448 f
->remote
.remote_next
= NULL
;
450 f
->flags
= ndm_flags
;
451 f
->updated
= f
->used
= jiffies
;
452 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
455 hlist_add_head_rcu(&f
->hlist
,
456 vxlan_fdb_head(vxlan
, mac
));
460 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
465 static void vxlan_fdb_free(struct rcu_head
*head
)
467 struct vxlan_fdb
*f
= container_of(head
, struct vxlan_fdb
, rcu
);
469 while (f
->remote
.remote_next
) {
470 struct vxlan_rdst
*rd
= f
->remote
.remote_next
;
472 f
->remote
.remote_next
= rd
->remote_next
;
478 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
480 netdev_dbg(vxlan
->dev
,
481 "delete %pM\n", f
->eth_addr
);
484 vxlan_fdb_notify(vxlan
, f
, RTM_DELNEIGH
);
486 hlist_del_rcu(&f
->hlist
);
487 call_rcu(&f
->rcu
, vxlan_fdb_free
);
490 /* Add static entry (via netlink) */
491 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
492 struct net_device
*dev
,
493 const unsigned char *addr
, u16 flags
)
495 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
496 struct net
*net
= dev_net(vxlan
->dev
);
502 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
503 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
508 if (tb
[NDA_DST
] == NULL
)
511 if (nla_len(tb
[NDA_DST
]) != sizeof(__be32
))
512 return -EAFNOSUPPORT
;
514 ip
= nla_get_be32(tb
[NDA_DST
]);
517 if (nla_len(tb
[NDA_PORT
]) != sizeof(__be16
))
519 port
= nla_get_be16(tb
[NDA_PORT
]);
521 port
= vxlan
->dst_port
;
524 if (nla_len(tb
[NDA_VNI
]) != sizeof(u32
))
526 vni
= nla_get_u32(tb
[NDA_VNI
]);
528 vni
= vxlan
->default_dst
.remote_vni
;
530 if (tb
[NDA_IFINDEX
]) {
531 struct net_device
*tdev
;
533 if (nla_len(tb
[NDA_IFINDEX
]) != sizeof(u32
))
535 ifindex
= nla_get_u32(tb
[NDA_IFINDEX
]);
536 tdev
= dev_get_by_index(net
, ifindex
);
538 return -EADDRNOTAVAIL
;
543 spin_lock_bh(&vxlan
->hash_lock
);
544 err
= vxlan_fdb_create(vxlan
, addr
, ip
, ndm
->ndm_state
, flags
,
545 port
, vni
, ifindex
, ndm
->ndm_flags
);
546 spin_unlock_bh(&vxlan
->hash_lock
);
551 /* Delete entry (via netlink) */
552 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
553 struct net_device
*dev
,
554 const unsigned char *addr
)
556 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
560 spin_lock_bh(&vxlan
->hash_lock
);
561 f
= vxlan_find_mac(vxlan
, addr
);
563 vxlan_fdb_destroy(vxlan
, f
);
566 spin_unlock_bh(&vxlan
->hash_lock
);
571 /* Dump forwarding table */
572 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
573 struct net_device
*dev
, int idx
)
575 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
578 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
582 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
583 struct vxlan_rdst
*rd
;
584 for (rd
= &f
->remote
; rd
; rd
= rd
->remote_next
) {
585 if (idx
< cb
->args
[0])
588 err
= vxlan_fdb_info(skb
, vxlan
, f
,
589 NETLINK_CB(cb
->skb
).portid
,
604 /* Watch incoming packets to learn mapping between Ethernet address
605 * and Tunnel endpoint.
607 static void vxlan_snoop(struct net_device
*dev
,
608 __be32 src_ip
, const u8
*src_mac
)
610 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
614 f
= vxlan_find_mac(vxlan
, src_mac
);
616 if (likely(f
->remote
.remote_ip
== src_ip
))
621 "%pM migrated from %pI4 to %pI4\n",
622 src_mac
, &f
->remote
.remote_ip
, &src_ip
);
624 f
->remote
.remote_ip
= src_ip
;
625 f
->updated
= jiffies
;
627 /* learned new entry */
628 spin_lock(&vxlan
->hash_lock
);
629 err
= vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
631 NLM_F_EXCL
|NLM_F_CREATE
,
633 vxlan
->default_dst
.remote_vni
,
635 spin_unlock(&vxlan
->hash_lock
);
640 /* See if multicast group is already in use by other ID */
641 static bool vxlan_group_used(struct vxlan_net
*vn
,
642 const struct vxlan_dev
*this)
644 struct vxlan_dev
*vxlan
;
646 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
) {
650 if (!netif_running(vxlan
->dev
))
653 if (vxlan
->default_dst
.remote_ip
== this->default_dst
.remote_ip
)
660 /* kernel equivalent to IP_ADD_MEMBERSHIP */
661 static int vxlan_join_group(struct net_device
*dev
)
663 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
664 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
665 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
666 struct ip_mreqn mreq
= {
667 .imr_multiaddr
.s_addr
= vxlan
->default_dst
.remote_ip
,
668 .imr_ifindex
= vxlan
->default_dst
.remote_ifindex
,
672 /* Already a member of group */
673 if (vxlan_group_used(vn
, vxlan
))
676 /* Need to drop RTNL to call multicast join */
679 err
= ip_mc_join_group(sk
, &mreq
);
687 /* kernel equivalent to IP_DROP_MEMBERSHIP */
688 static int vxlan_leave_group(struct net_device
*dev
)
690 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
691 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
693 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
694 struct ip_mreqn mreq
= {
695 .imr_multiaddr
.s_addr
= vxlan
->default_dst
.remote_ip
,
696 .imr_ifindex
= vxlan
->default_dst
.remote_ifindex
,
699 /* Only leave group when last vxlan is done. */
700 if (vxlan_group_used(vn
, vxlan
))
703 /* Need to drop RTNL to call multicast leave */
706 err
= ip_mc_leave_group(sk
, &mreq
);
713 /* Callback from net/ipv4/udp.c to receive packets */
714 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
717 struct vxlanhdr
*vxh
;
718 struct vxlan_dev
*vxlan
;
719 struct pcpu_tstats
*stats
;
724 /* pop off outer UDP header */
725 __skb_pull(skb
, sizeof(struct udphdr
));
727 /* Need Vxlan and inner Ethernet header to be present */
728 if (!pskb_may_pull(skb
, sizeof(struct vxlanhdr
)))
731 /* Drop packets with reserved bits set */
732 vxh
= (struct vxlanhdr
*) skb
->data
;
733 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
734 (vxh
->vx_vni
& htonl(0xff))) {
735 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
736 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
740 __skb_pull(skb
, sizeof(struct vxlanhdr
));
742 /* Is this VNI defined? */
743 vni
= ntohl(vxh
->vx_vni
) >> 8;
744 port
= inet_sk(sk
)->inet_sport
;
745 vxlan
= vxlan_find_vni(sock_net(sk
), vni
, port
);
747 netdev_dbg(skb
->dev
, "unknown vni %d port %u\n",
752 if (!pskb_may_pull(skb
, ETH_HLEN
)) {
753 vxlan
->dev
->stats
.rx_length_errors
++;
754 vxlan
->dev
->stats
.rx_errors
++;
758 skb_reset_mac_header(skb
);
760 /* Re-examine inner Ethernet packet */
762 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
764 /* Ignore packet loops (and multicast echo) */
765 if (compare_ether_addr(eth_hdr(skb
)->h_source
,
766 vxlan
->dev
->dev_addr
) == 0)
769 if (vxlan
->flags
& VXLAN_F_LEARN
)
770 vxlan_snoop(skb
->dev
, oip
->saddr
, eth_hdr(skb
)->h_source
);
772 __skb_tunnel_rx(skb
, vxlan
->dev
);
773 skb_reset_network_header(skb
);
775 /* If the NIC driver gave us an encapsulated packet with
776 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
777 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
778 * for us. Otherwise force the upper layers to verify it.
780 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
|| !skb
->encapsulation
||
781 !(vxlan
->dev
->features
& NETIF_F_RXCSUM
))
782 skb
->ip_summed
= CHECKSUM_NONE
;
784 skb
->encapsulation
= 0;
786 err
= IP_ECN_decapsulate(oip
, skb
);
789 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
790 &oip
->saddr
, oip
->tos
);
792 ++vxlan
->dev
->stats
.rx_frame_errors
;
793 ++vxlan
->dev
->stats
.rx_errors
;
798 stats
= this_cpu_ptr(vxlan
->dev
->tstats
);
799 u64_stats_update_begin(&stats
->syncp
);
801 stats
->rx_bytes
+= skb
->len
;
802 u64_stats_update_end(&stats
->syncp
);
808 /* Put UDP header back */
809 __skb_push(skb
, sizeof(struct udphdr
));
813 /* Consume bad packet */
818 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
820 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
826 if (dev
->flags
& IFF_NOARP
)
829 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
830 dev
->stats
.tx_dropped
++;
835 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
836 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
837 parp
->ar_pro
!= htons(ETH_P_IP
) ||
838 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
839 parp
->ar_hln
!= dev
->addr_len
||
842 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
844 arpptr
+= dev
->addr_len
; /* sha */
845 memcpy(&sip
, arpptr
, sizeof(sip
));
846 arpptr
+= sizeof(sip
);
847 arpptr
+= dev
->addr_len
; /* tha */
848 memcpy(&tip
, arpptr
, sizeof(tip
));
850 if (ipv4_is_loopback(tip
) ||
851 ipv4_is_multicast(tip
))
854 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
858 struct sk_buff
*reply
;
860 if (!(n
->nud_state
& NUD_CONNECTED
)) {
865 f
= vxlan_find_mac(vxlan
, n
->ha
);
866 if (f
&& f
->remote
.remote_ip
== htonl(INADDR_ANY
)) {
867 /* bridge-local neighbor */
872 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
877 skb_reset_mac_header(reply
);
878 __skb_pull(reply
, skb_network_offset(reply
));
879 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
880 reply
->pkt_type
= PACKET_HOST
;
882 if (netif_rx_ni(reply
) == NET_RX_DROP
)
883 dev
->stats
.rx_dropped
++;
884 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
885 vxlan_ip_miss(dev
, tip
);
891 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
893 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
897 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
901 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
903 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
906 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
915 diff
= compare_ether_addr(eth_hdr(skb
)->h_dest
, n
->ha
) != 0;
917 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
919 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
923 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
924 vxlan_ip_miss(dev
, pip
->daddr
);
928 static void vxlan_sock_put(struct sk_buff
*skb
)
933 /* On transmit, associate with the tunnel socket */
934 static void vxlan_set_owner(struct net_device
*dev
, struct sk_buff
*skb
)
936 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
937 struct sock
*sk
= vxlan
->vn_sock
->sock
->sk
;
942 skb
->destructor
= vxlan_sock_put
;
945 /* Compute source port for outgoing packet
946 * first choice to use L4 flow hash since it will spread
947 * better and maybe available from hardware
948 * secondary choice is to use jhash on the Ethernet header
950 static __be16
vxlan_src_port(const struct vxlan_dev
*vxlan
, struct sk_buff
*skb
)
952 unsigned int range
= (vxlan
->port_max
- vxlan
->port_min
) + 1;
955 hash
= skb_get_rxhash(skb
);
957 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
958 (__force u32
) skb
->protocol
);
960 return htons((((u64
) hash
* range
) >> 32) + vxlan
->port_min
);
963 static int handle_offloads(struct sk_buff
*skb
)
965 if (skb_is_gso(skb
)) {
966 int err
= skb_unclone(skb
, GFP_ATOMIC
);
970 skb_shinfo(skb
)->gso_type
|= SKB_GSO_UDP_TUNNEL
;
971 } else if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
972 skb
->ip_summed
= CHECKSUM_NONE
;
977 /* Bypass encapsulation if the destination is local */
978 static void vxlan_encap_bypass(struct sk_buff
*skb
, struct vxlan_dev
*src_vxlan
,
979 struct vxlan_dev
*dst_vxlan
)
981 struct pcpu_tstats
*tx_stats
= this_cpu_ptr(src_vxlan
->dev
->tstats
);
982 struct pcpu_tstats
*rx_stats
= this_cpu_ptr(dst_vxlan
->dev
->tstats
);
984 skb
->pkt_type
= PACKET_HOST
;
985 skb
->encapsulation
= 0;
986 skb
->dev
= dst_vxlan
->dev
;
987 __skb_pull(skb
, skb_network_offset(skb
));
989 if (dst_vxlan
->flags
& VXLAN_F_LEARN
)
990 vxlan_snoop(skb
->dev
, htonl(INADDR_LOOPBACK
),
991 eth_hdr(skb
)->h_source
);
993 u64_stats_update_begin(&tx_stats
->syncp
);
994 tx_stats
->tx_packets
++;
995 tx_stats
->tx_bytes
+= skb
->len
;
996 u64_stats_update_end(&tx_stats
->syncp
);
998 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
999 u64_stats_update_begin(&rx_stats
->syncp
);
1000 rx_stats
->rx_packets
++;
1001 rx_stats
->rx_bytes
+= skb
->len
;
1002 u64_stats_update_end(&rx_stats
->syncp
);
1004 skb
->dev
->stats
.rx_dropped
++;
1008 static netdev_tx_t
vxlan_xmit_one(struct sk_buff
*skb
, struct net_device
*dev
,
1009 struct vxlan_rdst
*rdst
, bool did_rsc
)
1011 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1013 const struct iphdr
*old_iph
;
1015 struct vxlanhdr
*vxh
;
1019 __be16 src_port
, dst_port
;
1024 dst_port
= rdst
->remote_port
? rdst
->remote_port
: vxlan
->dst_port
;
1025 vni
= rdst
->remote_vni
;
1026 dst
= rdst
->remote_ip
;
1030 /* short-circuited back to local bridge */
1031 vxlan_encap_bypass(skb
, vxlan
, vxlan
);
1032 return NETDEV_TX_OK
;
1037 if (!skb
->encapsulation
) {
1038 skb_reset_inner_headers(skb
);
1039 skb
->encapsulation
= 1;
1042 /* Need space for new headers (invalidates iph ptr) */
1043 if (skb_cow_head(skb
, VXLAN_HEADROOM
))
1046 old_iph
= ip_hdr(skb
);
1049 if (!ttl
&& IN_MULTICAST(ntohl(dst
)))
1054 tos
= ip_tunnel_get_dsfield(old_iph
, skb
);
1056 src_port
= vxlan_src_port(vxlan
, skb
);
1058 memset(&fl4
, 0, sizeof(fl4
));
1059 fl4
.flowi4_oif
= rdst
->remote_ifindex
;
1060 fl4
.flowi4_tos
= RT_TOS(tos
);
1062 fl4
.saddr
= vxlan
->saddr
;
1064 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
1066 netdev_dbg(dev
, "no route to %pI4\n", &dst
);
1067 dev
->stats
.tx_carrier_errors
++;
1071 if (rt
->dst
.dev
== dev
) {
1072 netdev_dbg(dev
, "circular route to %pI4\n", &dst
);
1074 dev
->stats
.collisions
++;
1078 /* Bypass encapsulation if the destination is local */
1079 if (rt
->rt_flags
& RTCF_LOCAL
&&
1080 !(rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))) {
1081 struct vxlan_dev
*dst_vxlan
;
1084 dst_vxlan
= vxlan_find_vni(dev_net(dev
), vni
, dst_port
);
1087 vxlan_encap_bypass(skb
, vxlan
, dst_vxlan
);
1088 return NETDEV_TX_OK
;
1091 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1092 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
1095 skb_dst_set(skb
, &rt
->dst
);
1097 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
1098 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
1099 vxh
->vx_vni
= htonl(vni
<< 8);
1101 __skb_push(skb
, sizeof(*uh
));
1102 skb_reset_transport_header(skb
);
1105 uh
->dest
= dst_port
;
1106 uh
->source
= src_port
;
1108 uh
->len
= htons(skb
->len
);
1111 __skb_push(skb
, sizeof(*iph
));
1112 skb_reset_network_header(skb
);
1115 iph
->ihl
= sizeof(struct iphdr
) >> 2;
1117 iph
->protocol
= IPPROTO_UDP
;
1118 iph
->tos
= ip_tunnel_ecn_encap(tos
, old_iph
, skb
);
1120 iph
->saddr
= fl4
.saddr
;
1121 iph
->ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
1122 tunnel_ip_select_ident(skb
, old_iph
, &rt
->dst
);
1126 vxlan_set_owner(dev
, skb
);
1128 if (handle_offloads(skb
))
1131 iptunnel_xmit(skb
, dev
);
1132 return NETDEV_TX_OK
;
1135 dev
->stats
.tx_dropped
++;
1139 dev
->stats
.tx_errors
++;
1142 return NETDEV_TX_OK
;
1145 /* Transmit local packets over Vxlan
1147 * Outer IP header inherits ECN and DF from inner header.
1148 * Outer UDP destination is the VXLAN assigned port.
1149 * source port is based on hash of flow
1151 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1153 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1155 bool did_rsc
= false;
1156 struct vxlan_rdst
*rdst0
, *rdst
;
1157 struct vxlan_fdb
*f
;
1160 skb_reset_mac_header(skb
);
1163 if ((vxlan
->flags
& VXLAN_F_PROXY
) && ntohs(eth
->h_proto
) == ETH_P_ARP
)
1164 return arp_reduce(dev
, skb
);
1166 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1169 if (f
&& (f
->flags
& NTF_ROUTER
) && (vxlan
->flags
& VXLAN_F_RSC
) &&
1170 ntohs(eth
->h_proto
) == ETH_P_IP
) {
1171 did_rsc
= route_shortcircuit(dev
, skb
);
1173 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
1177 rdst0
= &vxlan
->default_dst
;
1179 if (rdst0
->remote_ip
== htonl(INADDR_ANY
) &&
1180 (vxlan
->flags
& VXLAN_F_L2MISS
) &&
1181 !is_multicast_ether_addr(eth
->h_dest
))
1182 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
1188 /* if there are multiple destinations, send copies */
1189 for (rdst
= rdst0
->remote_next
; rdst
; rdst
= rdst
->remote_next
) {
1190 struct sk_buff
*skb1
;
1192 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1193 rc1
= vxlan_xmit_one(skb1
, dev
, rdst
, did_rsc
);
1194 if (rc
== NETDEV_TX_OK
)
1198 rc1
= vxlan_xmit_one(skb
, dev
, rdst0
, did_rsc
);
1199 if (rc
== NETDEV_TX_OK
)
1204 /* Walk the forwarding table and purge stale entries */
1205 static void vxlan_cleanup(unsigned long arg
)
1207 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1208 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1211 if (!netif_running(vxlan
->dev
))
1214 spin_lock_bh(&vxlan
->hash_lock
);
1215 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1216 struct hlist_node
*p
, *n
;
1217 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1219 = container_of(p
, struct vxlan_fdb
, hlist
);
1220 unsigned long timeout
;
1222 if (f
->state
& NUD_PERMANENT
)
1225 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
1226 if (time_before_eq(timeout
, jiffies
)) {
1227 netdev_dbg(vxlan
->dev
,
1228 "garbage collect %pM\n",
1230 f
->state
= NUD_STALE
;
1231 vxlan_fdb_destroy(vxlan
, f
);
1232 } else if (time_before(timeout
, next_timer
))
1233 next_timer
= timeout
;
1236 spin_unlock_bh(&vxlan
->hash_lock
);
1238 mod_timer(&vxlan
->age_timer
, next_timer
);
1241 /* Setup stats when device is created */
1242 static int vxlan_init(struct net_device
*dev
)
1244 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1251 /* Start ageing timer and join group when device is brought up */
1252 static int vxlan_open(struct net_device
*dev
)
1254 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1257 if (IN_MULTICAST(ntohl(vxlan
->default_dst
.remote_ip
))) {
1258 err
= vxlan_join_group(dev
);
1263 if (vxlan
->age_interval
)
1264 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
1269 /* Purge the forwarding table */
1270 static void vxlan_flush(struct vxlan_dev
*vxlan
)
1274 spin_lock_bh(&vxlan
->hash_lock
);
1275 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1276 struct hlist_node
*p
, *n
;
1277 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1279 = container_of(p
, struct vxlan_fdb
, hlist
);
1280 vxlan_fdb_destroy(vxlan
, f
);
1283 spin_unlock_bh(&vxlan
->hash_lock
);
1286 /* Cleanup timer and forwarding table on shutdown */
1287 static int vxlan_stop(struct net_device
*dev
)
1289 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1291 if (IN_MULTICAST(ntohl(vxlan
->default_dst
.remote_ip
)))
1292 vxlan_leave_group(dev
);
1294 del_timer_sync(&vxlan
->age_timer
);
1301 /* Stub, nothing needs to be done. */
1302 static void vxlan_set_multicast_list(struct net_device
*dev
)
1306 static const struct net_device_ops vxlan_netdev_ops
= {
1307 .ndo_init
= vxlan_init
,
1308 .ndo_open
= vxlan_open
,
1309 .ndo_stop
= vxlan_stop
,
1310 .ndo_start_xmit
= vxlan_xmit
,
1311 .ndo_get_stats64
= ip_tunnel_get_stats64
,
1312 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
1313 .ndo_change_mtu
= eth_change_mtu
,
1314 .ndo_validate_addr
= eth_validate_addr
,
1315 .ndo_set_mac_address
= eth_mac_addr
,
1316 .ndo_fdb_add
= vxlan_fdb_add
,
1317 .ndo_fdb_del
= vxlan_fdb_delete
,
1318 .ndo_fdb_dump
= vxlan_fdb_dump
,
1321 /* Info for udev, that this is a virtual tunnel endpoint */
1322 static struct device_type vxlan_type
= {
1326 static void vxlan_free(struct net_device
*dev
)
1328 free_percpu(dev
->tstats
);
1332 /* Initialize the device structure. */
1333 static void vxlan_setup(struct net_device
*dev
)
1335 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1339 eth_hw_addr_random(dev
);
1341 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
1343 dev
->netdev_ops
= &vxlan_netdev_ops
;
1344 dev
->destructor
= vxlan_free
;
1345 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
1347 dev
->tx_queue_len
= 0;
1348 dev
->features
|= NETIF_F_LLTX
;
1349 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1350 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1351 dev
->features
|= NETIF_F_RXCSUM
;
1352 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
1354 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
1355 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
1356 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1357 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1359 INIT_LIST_HEAD(&vxlan
->next
);
1360 spin_lock_init(&vxlan
->hash_lock
);
1362 init_timer_deferrable(&vxlan
->age_timer
);
1363 vxlan
->age_timer
.function
= vxlan_cleanup
;
1364 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
1366 inet_get_local_port_range(&low
, &high
);
1367 vxlan
->port_min
= low
;
1368 vxlan
->port_max
= high
;
1369 vxlan
->dst_port
= htons(vxlan_port
);
1373 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
1374 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
1377 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
1378 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
1379 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
1380 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
1381 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
1382 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
1383 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
1384 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
1385 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
1386 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
1387 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
1388 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
1389 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
1390 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
1391 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
1392 [IFLA_VXLAN_PORT
] = { .type
= NLA_U16
},
1395 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
1397 if (tb
[IFLA_ADDRESS
]) {
1398 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
1399 pr_debug("invalid link address (not ethernet)\n");
1403 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
1404 pr_debug("invalid all zero ethernet address\n");
1405 return -EADDRNOTAVAIL
;
1412 if (data
[IFLA_VXLAN_ID
]) {
1413 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1414 if (id
>= VXLAN_VID_MASK
)
1418 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1419 const struct ifla_vxlan_port_range
*p
1420 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1422 if (ntohs(p
->high
) < ntohs(p
->low
)) {
1423 pr_debug("port range %u .. %u not valid\n",
1424 ntohs(p
->low
), ntohs(p
->high
));
1432 static void vxlan_get_drvinfo(struct net_device
*netdev
,
1433 struct ethtool_drvinfo
*drvinfo
)
1435 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
1436 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
1439 static const struct ethtool_ops vxlan_ethtool_ops
= {
1440 .get_drvinfo
= vxlan_get_drvinfo
,
1441 .get_link
= ethtool_op_get_link
,
1444 static void vxlan_del_work(struct work_struct
*work
)
1446 struct vxlan_sock
*vs
= container_of(work
, struct vxlan_sock
, del_work
);
1448 sk_release_kernel(vs
->sock
->sk
);
1452 /* Create new listen socket if needed */
1453 static struct vxlan_sock
*vxlan_socket_create(struct net
*net
, __be16 port
)
1455 struct vxlan_sock
*vs
;
1457 struct sockaddr_in vxlan_addr
= {
1458 .sin_family
= AF_INET
,
1459 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
1464 vs
= kmalloc(sizeof(*vs
), GFP_KERNEL
);
1466 return ERR_PTR(-ENOMEM
);
1468 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1469 INIT_HLIST_HEAD(&vs
->vni_list
[h
]);
1471 INIT_WORK(&vs
->del_work
, vxlan_del_work
);
1473 /* Create UDP socket for encapsulation receive. */
1474 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &vs
->sock
);
1476 pr_debug("UDP socket create failed\n");
1481 /* Put in proper namespace */
1483 sk_change_net(sk
, net
);
1485 vxlan_addr
.sin_port
= port
;
1487 rc
= kernel_bind(vs
->sock
, (struct sockaddr
*) &vxlan_addr
,
1488 sizeof(vxlan_addr
));
1490 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1491 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
1492 sk_release_kernel(sk
);
1497 /* Disable multicast loopback */
1498 inet_sk(sk
)->mc_loop
= 0;
1500 /* Mark socket as an encapsulation socket. */
1501 udp_sk(sk
)->encap_type
= 1;
1502 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
1509 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
1510 struct nlattr
*tb
[], struct nlattr
*data
[])
1512 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1513 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1514 struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
1515 struct vxlan_sock
*vs
;
1519 if (!data
[IFLA_VXLAN_ID
])
1522 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1523 dst
->remote_vni
= vni
;
1525 if (data
[IFLA_VXLAN_GROUP
])
1526 dst
->remote_ip
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1528 if (data
[IFLA_VXLAN_LOCAL
])
1529 vxlan
->saddr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
1531 if (data
[IFLA_VXLAN_LINK
] &&
1532 (dst
->remote_ifindex
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
1533 struct net_device
*lowerdev
1534 = __dev_get_by_index(net
, dst
->remote_ifindex
);
1537 pr_info("ifindex %d does not exist\n", dst
->remote_ifindex
);
1542 dev
->mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
1544 /* update header length based on lower device */
1545 dev
->hard_header_len
= lowerdev
->hard_header_len
+
1549 if (data
[IFLA_VXLAN_TOS
])
1550 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
1552 if (data
[IFLA_VXLAN_TTL
])
1553 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
1555 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
1556 vxlan
->flags
|= VXLAN_F_LEARN
;
1558 if (data
[IFLA_VXLAN_AGEING
])
1559 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
1561 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
1563 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
1564 vxlan
->flags
|= VXLAN_F_PROXY
;
1566 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
1567 vxlan
->flags
|= VXLAN_F_RSC
;
1569 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
1570 vxlan
->flags
|= VXLAN_F_L2MISS
;
1572 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
1573 vxlan
->flags
|= VXLAN_F_L3MISS
;
1575 if (data
[IFLA_VXLAN_LIMIT
])
1576 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
1578 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1579 const struct ifla_vxlan_port_range
*p
1580 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1581 vxlan
->port_min
= ntohs(p
->low
);
1582 vxlan
->port_max
= ntohs(p
->high
);
1585 if (data
[IFLA_VXLAN_PORT
])
1586 vxlan
->dst_port
= nla_get_be16(data
[IFLA_VXLAN_PORT
]);
1588 if (vxlan_find_vni(net
, vni
, vxlan
->dst_port
)) {
1589 pr_info("duplicate VNI %u\n", vni
);
1593 vs
= vxlan_find_port(net
, vxlan
->dst_port
);
1597 /* Drop lock because socket create acquires RTNL lock */
1599 vs
= vxlan_socket_create(net
, vxlan
->dst_port
);
1604 hlist_add_head_rcu(&vs
->hlist
, vs_head(net
, vxlan
->dst_port
));
1606 vxlan
->vn_sock
= vs
;
1608 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
1610 err
= register_netdevice(dev
);
1612 if (--vs
->refcnt
== 0) {
1614 sk_release_kernel(vs
->sock
->sk
);
1621 list_add(&vxlan
->next
, &vn
->vxlan_list
);
1622 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(vs
, vni
));
1627 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
1629 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1630 struct vxlan_sock
*vs
= vxlan
->vn_sock
;
1632 hlist_del_rcu(&vxlan
->hlist
);
1633 list_del(&vxlan
->next
);
1634 unregister_netdevice_queue(dev
, head
);
1636 if (--vs
->refcnt
== 0) {
1637 hlist_del_rcu(&vs
->hlist
);
1638 schedule_work(&vs
->del_work
);
1642 static size_t vxlan_get_size(const struct net_device
*dev
)
1645 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
1646 nla_total_size(sizeof(__be32
)) +/* IFLA_VXLAN_GROUP */
1647 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
1648 nla_total_size(sizeof(__be32
))+ /* IFLA_VXLAN_LOCAL */
1649 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
1650 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
1651 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
1652 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
1653 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
1654 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
1655 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
1656 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
1657 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
1658 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
1659 nla_total_size(sizeof(__be16
))+ /* IFLA_VXLAN_PORT */
1663 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1665 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1666 const struct vxlan_rdst
*dst
= &vxlan
->default_dst
;
1667 struct ifla_vxlan_port_range ports
= {
1668 .low
= htons(vxlan
->port_min
),
1669 .high
= htons(vxlan
->port_max
),
1672 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, dst
->remote_vni
))
1673 goto nla_put_failure
;
1675 if (dst
->remote_ip
&& nla_put_be32(skb
, IFLA_VXLAN_GROUP
, dst
->remote_ip
))
1676 goto nla_put_failure
;
1678 if (dst
->remote_ifindex
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, dst
->remote_ifindex
))
1679 goto nla_put_failure
;
1681 if (vxlan
->saddr
&& nla_put_be32(skb
, IFLA_VXLAN_LOCAL
, vxlan
->saddr
))
1682 goto nla_put_failure
;
1684 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
1685 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
1686 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
1687 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
1688 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
1689 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
1690 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
1691 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
1692 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
1693 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
1694 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
1695 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
1696 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
) ||
1697 nla_put_be16(skb
, IFLA_VXLAN_PORT
, vxlan
->dst_port
))
1698 goto nla_put_failure
;
1700 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
1701 goto nla_put_failure
;
1709 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
1711 .maxtype
= IFLA_VXLAN_MAX
,
1712 .policy
= vxlan_policy
,
1713 .priv_size
= sizeof(struct vxlan_dev
),
1714 .setup
= vxlan_setup
,
1715 .validate
= vxlan_validate
,
1716 .newlink
= vxlan_newlink
,
1717 .dellink
= vxlan_dellink
,
1718 .get_size
= vxlan_get_size
,
1719 .fill_info
= vxlan_fill_info
,
1722 static __net_init
int vxlan_init_net(struct net
*net
)
1724 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1727 INIT_LIST_HEAD(&vn
->vxlan_list
);
1729 for (h
= 0; h
< PORT_HASH_SIZE
; ++h
)
1730 INIT_HLIST_HEAD(&vn
->sock_list
[h
]);
1735 static __net_exit
void vxlan_exit_net(struct net
*net
)
1737 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1738 struct vxlan_dev
*vxlan
;
1741 list_for_each_entry(vxlan
, &vn
->vxlan_list
, next
)
1742 dev_close(vxlan
->dev
);
1746 static struct pernet_operations vxlan_net_ops
= {
1747 .init
= vxlan_init_net
,
1748 .exit
= vxlan_exit_net
,
1749 .id
= &vxlan_net_id
,
1750 .size
= sizeof(struct vxlan_net
),
1753 static int __init
vxlan_init_module(void)
1757 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
1759 rc
= register_pernet_device(&vxlan_net_ops
);
1763 rc
= rtnl_link_register(&vxlan_link_ops
);
1770 unregister_pernet_device(&vxlan_net_ops
);
1774 late_initcall(vxlan_init_module
);
1776 static void __exit
vxlan_cleanup_module(void)
1778 rtnl_link_unregister(&vxlan_link_ops
);
1779 unregister_pernet_device(&vxlan_net_ops
);
1782 module_exit(vxlan_cleanup_module
);
1784 MODULE_LICENSE("GPL");
1785 MODULE_VERSION(VXLAN_VERSION
);
1786 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
1787 MODULE_ALIAS_RTNL_LINK("vxlan");