2 * VXLAN: Virtual eXtensible Local Area Network
4 * Copyright (c) 2012 Vyatta Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 * - use IANA UDP port number (when defined)
15 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/module.h>
20 #include <linux/errno.h>
21 #include <linux/slab.h>
22 #include <linux/skbuff.h>
23 #include <linux/rculist.h>
24 #include <linux/netdevice.h>
27 #include <linux/udp.h>
28 #include <linux/igmp.h>
29 #include <linux/etherdevice.h>
30 #include <linux/if_ether.h>
31 #include <linux/hash.h>
32 #include <linux/ethtool.h>
34 #include <net/ndisc.h>
39 #include <net/rtnetlink.h>
40 #include <net/route.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/net_namespace.h>
44 #include <net/netns/generic.h>
46 #define VXLAN_VERSION "0.1"
48 #define VNI_HASH_BITS 10
49 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
50 #define FDB_HASH_BITS 8
51 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
52 #define FDB_AGE_DEFAULT 300 /* 5 min */
53 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
55 #define VXLAN_N_VID (1u << 24)
56 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
57 /* IP header + UDP + VXLAN + Ethernet header */
58 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
60 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
62 /* VXLAN protocol header */
68 /* UDP port for VXLAN traffic. */
69 static unsigned int vxlan_port __read_mostly
= 8472;
70 module_param_named(udp_port
, vxlan_port
, uint
, 0444);
71 MODULE_PARM_DESC(udp_port
, "Destination UDP port");
73 static bool log_ecn_error
= true;
74 module_param(log_ecn_error
, bool, 0644);
75 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
77 /* per-net private data for this module */
78 static unsigned int vxlan_net_id
;
80 struct socket
*sock
; /* UDP encap socket */
81 struct hlist_head vni_list
[VNI_HASH_SIZE
];
84 /* Forwarding table entry */
86 struct hlist_node hlist
; /* linked list of entries */
88 unsigned long updated
; /* jiffies */
91 u16 state
; /* see ndm_state */
92 u8 eth_addr
[ETH_ALEN
];
95 /* Per-cpu network traffic stats */
101 struct u64_stats_sync syncp
;
104 /* Pseudo network device */
106 struct hlist_node hlist
;
107 struct net_device
*dev
;
108 struct vxlan_stats __percpu
*stats
;
109 __u32 vni
; /* virtual network id */
110 __be32 gaddr
; /* multicast group */
111 __be32 saddr
; /* source address */
112 unsigned int link
; /* link to multicast over */
113 __u16 port_min
; /* source port range */
115 __u8 tos
; /* TOS override */
117 u32 flags
; /* VXLAN_F_* below */
119 unsigned long age_interval
;
120 struct timer_list age_timer
;
121 spinlock_t hash_lock
;
122 unsigned int addrcnt
;
123 unsigned int addrmax
;
125 struct hlist_head fdb_head
[FDB_HASH_SIZE
];
128 #define VXLAN_F_LEARN 0x01
129 #define VXLAN_F_PROXY 0x02
130 #define VXLAN_F_RSC 0x04
131 #define VXLAN_F_L2MISS 0x08
132 #define VXLAN_F_L3MISS 0x10
134 /* salt for hash table */
135 static u32 vxlan_salt __read_mostly
;
137 static inline struct hlist_head
*vni_head(struct net
*net
, u32 id
)
139 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
141 return &vn
->vni_list
[hash_32(id
, VNI_HASH_BITS
)];
144 /* Look up VNI in a per net namespace table */
145 static struct vxlan_dev
*vxlan_find_vni(struct net
*net
, u32 id
)
147 struct vxlan_dev
*vxlan
;
149 hlist_for_each_entry_rcu(vxlan
, vni_head(net
, id
), hlist
) {
150 if (vxlan
->vni
== id
)
157 /* Fill in neighbour message in skbuff. */
158 static int vxlan_fdb_info(struct sk_buff
*skb
, struct vxlan_dev
*vxlan
,
159 const struct vxlan_fdb
*fdb
,
160 u32 portid
, u32 seq
, int type
, unsigned int flags
)
162 unsigned long now
= jiffies
;
163 struct nda_cacheinfo ci
;
164 struct nlmsghdr
*nlh
;
166 bool send_ip
, send_eth
;
168 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*ndm
), flags
);
172 ndm
= nlmsg_data(nlh
);
173 memset(ndm
, 0, sizeof(*ndm
));
175 send_eth
= send_ip
= true;
177 if (type
== RTM_GETNEIGH
) {
178 ndm
->ndm_family
= AF_INET
;
179 send_ip
= fdb
->remote_ip
!= 0;
180 send_eth
= !is_zero_ether_addr(fdb
->eth_addr
);
182 ndm
->ndm_family
= AF_BRIDGE
;
183 ndm
->ndm_state
= fdb
->state
;
184 ndm
->ndm_ifindex
= vxlan
->dev
->ifindex
;
185 ndm
->ndm_flags
= NTF_SELF
;
186 ndm
->ndm_type
= NDA_DST
;
188 if (send_eth
&& nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, &fdb
->eth_addr
))
189 goto nla_put_failure
;
191 if (send_ip
&& nla_put_be32(skb
, NDA_DST
, fdb
->remote_ip
))
192 goto nla_put_failure
;
194 ci
.ndm_used
= jiffies_to_clock_t(now
- fdb
->used
);
195 ci
.ndm_confirmed
= 0;
196 ci
.ndm_updated
= jiffies_to_clock_t(now
- fdb
->updated
);
199 if (nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
200 goto nla_put_failure
;
202 return nlmsg_end(skb
, nlh
);
205 nlmsg_cancel(skb
, nlh
);
209 static inline size_t vxlan_nlmsg_size(void)
211 return NLMSG_ALIGN(sizeof(struct ndmsg
))
212 + nla_total_size(ETH_ALEN
) /* NDA_LLADDR */
213 + nla_total_size(sizeof(__be32
)) /* NDA_DST */
214 + nla_total_size(sizeof(struct nda_cacheinfo
));
217 static void vxlan_fdb_notify(struct vxlan_dev
*vxlan
,
218 const struct vxlan_fdb
*fdb
, int type
)
220 struct net
*net
= dev_net(vxlan
->dev
);
224 skb
= nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC
);
228 err
= vxlan_fdb_info(skb
, vxlan
, fdb
, 0, 0, type
, 0);
230 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
231 WARN_ON(err
== -EMSGSIZE
);
236 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
240 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
243 static void vxlan_ip_miss(struct net_device
*dev
, __be32 ipa
)
245 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
248 memset(&f
, 0, sizeof f
);
250 f
.remote_ip
= ipa
; /* goes to NDA_DST */
252 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
255 static void vxlan_fdb_miss(struct vxlan_dev
*vxlan
, const u8 eth_addr
[ETH_ALEN
])
259 memset(&f
, 0, sizeof f
);
261 memcpy(f
.eth_addr
, eth_addr
, ETH_ALEN
);
263 vxlan_fdb_notify(vxlan
, &f
, RTM_GETNEIGH
);
266 /* Hash Ethernet address */
267 static u32
eth_hash(const unsigned char *addr
)
269 u64 value
= get_unaligned((u64
*)addr
);
271 /* only want 6 bytes */
277 return hash_64(value
, FDB_HASH_BITS
);
280 /* Hash chain to use given mac address */
281 static inline struct hlist_head
*vxlan_fdb_head(struct vxlan_dev
*vxlan
,
284 return &vxlan
->fdb_head
[eth_hash(mac
)];
287 /* Look up Ethernet address in forwarding table */
288 static struct vxlan_fdb
*vxlan_find_mac(struct vxlan_dev
*vxlan
,
292 struct hlist_head
*head
= vxlan_fdb_head(vxlan
, mac
);
295 hlist_for_each_entry_rcu(f
, head
, hlist
) {
296 if (compare_ether_addr(mac
, f
->eth_addr
) == 0)
303 /* Add new entry to forwarding table -- assumes lock held */
304 static int vxlan_fdb_create(struct vxlan_dev
*vxlan
,
305 const u8
*mac
, __be32 ip
,
306 __u16 state
, __u16 flags
)
311 f
= vxlan_find_mac(vxlan
, mac
);
313 if (flags
& NLM_F_EXCL
) {
314 netdev_dbg(vxlan
->dev
,
315 "lost race to create %pM\n", mac
);
318 if (f
->state
!= state
) {
320 f
->updated
= jiffies
;
324 if (!(flags
& NLM_F_CREATE
))
327 if (vxlan
->addrmax
&& vxlan
->addrcnt
>= vxlan
->addrmax
)
330 netdev_dbg(vxlan
->dev
, "add %pM -> %pI4\n", mac
, &ip
);
331 f
= kmalloc(sizeof(*f
), GFP_ATOMIC
);
338 f
->updated
= f
->used
= jiffies
;
339 memcpy(f
->eth_addr
, mac
, ETH_ALEN
);
342 hlist_add_head_rcu(&f
->hlist
,
343 vxlan_fdb_head(vxlan
, mac
));
347 vxlan_fdb_notify(vxlan
, f
, RTM_NEWNEIGH
);
352 static void vxlan_fdb_destroy(struct vxlan_dev
*vxlan
, struct vxlan_fdb
*f
)
354 netdev_dbg(vxlan
->dev
,
355 "delete %pM\n", f
->eth_addr
);
358 vxlan_fdb_notify(vxlan
, f
, RTM_DELNEIGH
);
360 hlist_del_rcu(&f
->hlist
);
364 /* Add static entry (via netlink) */
365 static int vxlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
366 struct net_device
*dev
,
367 const unsigned char *addr
, u16 flags
)
369 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
373 if (!(ndm
->ndm_state
& (NUD_PERMANENT
|NUD_REACHABLE
))) {
374 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
379 if (tb
[NDA_DST
] == NULL
)
382 if (nla_len(tb
[NDA_DST
]) != sizeof(__be32
))
383 return -EAFNOSUPPORT
;
385 ip
= nla_get_be32(tb
[NDA_DST
]);
387 spin_lock_bh(&vxlan
->hash_lock
);
388 err
= vxlan_fdb_create(vxlan
, addr
, ip
, ndm
->ndm_state
, flags
);
389 spin_unlock_bh(&vxlan
->hash_lock
);
394 /* Delete entry (via netlink) */
395 static int vxlan_fdb_delete(struct ndmsg
*ndm
, struct nlattr
*tb
[],
396 struct net_device
*dev
,
397 const unsigned char *addr
)
399 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
403 spin_lock_bh(&vxlan
->hash_lock
);
404 f
= vxlan_find_mac(vxlan
, addr
);
406 vxlan_fdb_destroy(vxlan
, f
);
409 spin_unlock_bh(&vxlan
->hash_lock
);
414 /* Dump forwarding table */
415 static int vxlan_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
416 struct net_device
*dev
, int idx
)
418 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
421 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
425 hlist_for_each_entry_rcu(f
, &vxlan
->fdb_head
[h
], hlist
) {
426 if (idx
< cb
->args
[0])
429 err
= vxlan_fdb_info(skb
, vxlan
, f
,
430 NETLINK_CB(cb
->skb
).portid
,
444 /* Watch incoming packets to learn mapping between Ethernet address
445 * and Tunnel endpoint.
447 static void vxlan_snoop(struct net_device
*dev
,
448 __be32 src_ip
, const u8
*src_mac
)
450 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
454 f
= vxlan_find_mac(vxlan
, src_mac
);
457 if (likely(f
->remote_ip
== src_ip
))
462 "%pM migrated from %pI4 to %pI4\n",
463 src_mac
, &f
->remote_ip
, &src_ip
);
465 f
->remote_ip
= src_ip
;
466 f
->updated
= jiffies
;
468 /* learned new entry */
469 spin_lock(&vxlan
->hash_lock
);
470 err
= vxlan_fdb_create(vxlan
, src_mac
, src_ip
,
472 NLM_F_EXCL
|NLM_F_CREATE
);
473 spin_unlock(&vxlan
->hash_lock
);
478 /* See if multicast group is already in use by other ID */
479 static bool vxlan_group_used(struct vxlan_net
*vn
,
480 const struct vxlan_dev
*this)
482 const struct vxlan_dev
*vxlan
;
485 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
486 hlist_for_each_entry(vxlan
, &vn
->vni_list
[h
], hlist
) {
490 if (!netif_running(vxlan
->dev
))
493 if (vxlan
->gaddr
== this->gaddr
)
500 /* kernel equivalent to IP_ADD_MEMBERSHIP */
501 static int vxlan_join_group(struct net_device
*dev
)
503 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
504 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
505 struct sock
*sk
= vn
->sock
->sk
;
506 struct ip_mreqn mreq
= {
507 .imr_multiaddr
.s_addr
= vxlan
->gaddr
,
508 .imr_ifindex
= vxlan
->link
,
512 /* Already a member of group */
513 if (vxlan_group_used(vn
, vxlan
))
516 /* Need to drop RTNL to call multicast join */
519 err
= ip_mc_join_group(sk
, &mreq
);
527 /* kernel equivalent to IP_DROP_MEMBERSHIP */
528 static int vxlan_leave_group(struct net_device
*dev
)
530 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
531 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
533 struct sock
*sk
= vn
->sock
->sk
;
534 struct ip_mreqn mreq
= {
535 .imr_multiaddr
.s_addr
= vxlan
->gaddr
,
536 .imr_ifindex
= vxlan
->link
,
539 /* Only leave group when last vxlan is done. */
540 if (vxlan_group_used(vn
, vxlan
))
543 /* Need to drop RTNL to call multicast leave */
546 err
= ip_mc_leave_group(sk
, &mreq
);
553 /* Callback from net/ipv4/udp.c to receive packets */
554 static int vxlan_udp_encap_recv(struct sock
*sk
, struct sk_buff
*skb
)
557 struct vxlanhdr
*vxh
;
558 struct vxlan_dev
*vxlan
;
559 struct vxlan_stats
*stats
;
563 /* pop off outer UDP header */
564 __skb_pull(skb
, sizeof(struct udphdr
));
566 /* Need Vxlan and inner Ethernet header to be present */
567 if (!pskb_may_pull(skb
, sizeof(struct vxlanhdr
)))
570 /* Drop packets with reserved bits set */
571 vxh
= (struct vxlanhdr
*) skb
->data
;
572 if (vxh
->vx_flags
!= htonl(VXLAN_FLAGS
) ||
573 (vxh
->vx_vni
& htonl(0xff))) {
574 netdev_dbg(skb
->dev
, "invalid vxlan flags=%#x vni=%#x\n",
575 ntohl(vxh
->vx_flags
), ntohl(vxh
->vx_vni
));
579 __skb_pull(skb
, sizeof(struct vxlanhdr
));
581 /* Is this VNI defined? */
582 vni
= ntohl(vxh
->vx_vni
) >> 8;
583 vxlan
= vxlan_find_vni(sock_net(sk
), vni
);
585 netdev_dbg(skb
->dev
, "unknown vni %d\n", vni
);
589 if (!pskb_may_pull(skb
, ETH_HLEN
)) {
590 vxlan
->dev
->stats
.rx_length_errors
++;
591 vxlan
->dev
->stats
.rx_errors
++;
595 skb_reset_mac_header(skb
);
597 /* Re-examine inner Ethernet packet */
599 skb
->protocol
= eth_type_trans(skb
, vxlan
->dev
);
601 /* Ignore packet loops (and multicast echo) */
602 if (compare_ether_addr(eth_hdr(skb
)->h_source
,
603 vxlan
->dev
->dev_addr
) == 0)
606 if (vxlan
->flags
& VXLAN_F_LEARN
)
607 vxlan_snoop(skb
->dev
, oip
->saddr
, eth_hdr(skb
)->h_source
);
609 __skb_tunnel_rx(skb
, vxlan
->dev
);
610 skb_reset_network_header(skb
);
612 /* If the NIC driver gave us an encapsulated packet with
613 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
614 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
615 * for us. Otherwise force the upper layers to verify it.
617 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
|| !skb
->encapsulation
||
618 !(vxlan
->dev
->features
& NETIF_F_RXCSUM
))
619 skb
->ip_summed
= CHECKSUM_NONE
;
621 skb
->encapsulation
= 0;
623 err
= IP_ECN_decapsulate(oip
, skb
);
626 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
627 &oip
->saddr
, oip
->tos
);
629 ++vxlan
->dev
->stats
.rx_frame_errors
;
630 ++vxlan
->dev
->stats
.rx_errors
;
635 stats
= this_cpu_ptr(vxlan
->stats
);
636 u64_stats_update_begin(&stats
->syncp
);
638 stats
->rx_bytes
+= skb
->len
;
639 u64_stats_update_end(&stats
->syncp
);
645 /* Put UDP header back */
646 __skb_push(skb
, sizeof(struct udphdr
));
650 /* Consume bad packet */
655 static int arp_reduce(struct net_device
*dev
, struct sk_buff
*skb
)
657 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
663 if (dev
->flags
& IFF_NOARP
)
666 if (!pskb_may_pull(skb
, arp_hdr_len(dev
))) {
667 dev
->stats
.tx_dropped
++;
672 if ((parp
->ar_hrd
!= htons(ARPHRD_ETHER
) &&
673 parp
->ar_hrd
!= htons(ARPHRD_IEEE802
)) ||
674 parp
->ar_pro
!= htons(ETH_P_IP
) ||
675 parp
->ar_op
!= htons(ARPOP_REQUEST
) ||
676 parp
->ar_hln
!= dev
->addr_len
||
679 arpptr
= (u8
*)parp
+ sizeof(struct arphdr
);
681 arpptr
+= dev
->addr_len
; /* sha */
682 memcpy(&sip
, arpptr
, sizeof(sip
));
683 arpptr
+= sizeof(sip
);
684 arpptr
+= dev
->addr_len
; /* tha */
685 memcpy(&tip
, arpptr
, sizeof(tip
));
687 if (ipv4_is_loopback(tip
) ||
688 ipv4_is_multicast(tip
))
691 n
= neigh_lookup(&arp_tbl
, &tip
, dev
);
694 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
696 struct sk_buff
*reply
;
698 if (!(n
->nud_state
& NUD_CONNECTED
)) {
703 f
= vxlan_find_mac(vxlan
, n
->ha
);
704 if (f
&& f
->remote_ip
== 0) {
705 /* bridge-local neighbor */
710 reply
= arp_create(ARPOP_REPLY
, ETH_P_ARP
, sip
, dev
, tip
, sha
,
715 skb_reset_mac_header(reply
);
716 __skb_pull(reply
, skb_network_offset(reply
));
717 reply
->ip_summed
= CHECKSUM_UNNECESSARY
;
718 reply
->pkt_type
= PACKET_HOST
;
720 if (netif_rx_ni(reply
) == NET_RX_DROP
)
721 dev
->stats
.rx_dropped
++;
722 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
723 vxlan_ip_miss(dev
, tip
);
729 static bool route_shortcircuit(struct net_device
*dev
, struct sk_buff
*skb
)
731 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
735 if (is_multicast_ether_addr(eth_hdr(skb
)->h_dest
))
739 switch (ntohs(eth_hdr(skb
)->h_proto
)) {
741 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
744 n
= neigh_lookup(&arp_tbl
, &pip
->daddr
, dev
);
753 diff
= compare_ether_addr(eth_hdr(skb
)->h_dest
, n
->ha
) != 0;
755 memcpy(eth_hdr(skb
)->h_source
, eth_hdr(skb
)->h_dest
,
757 memcpy(eth_hdr(skb
)->h_dest
, n
->ha
, dev
->addr_len
);
761 } else if (vxlan
->flags
& VXLAN_F_L3MISS
)
762 vxlan_ip_miss(dev
, pip
->daddr
);
766 /* Extract dsfield from inner protocol */
767 static inline u8
vxlan_get_dsfield(const struct iphdr
*iph
,
768 const struct sk_buff
*skb
)
770 if (skb
->protocol
== htons(ETH_P_IP
))
772 else if (skb
->protocol
== htons(ETH_P_IPV6
))
773 return ipv6_get_dsfield((const struct ipv6hdr
*)iph
);
778 /* Propogate ECN bits out */
779 static inline u8
vxlan_ecn_encap(u8 tos
,
780 const struct iphdr
*iph
,
781 const struct sk_buff
*skb
)
783 u8 inner
= vxlan_get_dsfield(iph
, skb
);
785 return INET_ECN_encapsulate(tos
, inner
);
788 static void vxlan_sock_free(struct sk_buff
*skb
)
793 /* On transmit, associate with the tunnel socket */
794 static void vxlan_set_owner(struct net_device
*dev
, struct sk_buff
*skb
)
796 struct vxlan_net
*vn
= net_generic(dev_net(dev
), vxlan_net_id
);
797 struct sock
*sk
= vn
->sock
->sk
;
802 skb
->destructor
= vxlan_sock_free
;
805 /* Compute source port for outgoing packet
806 * first choice to use L4 flow hash since it will spread
807 * better and maybe available from hardware
808 * secondary choice is to use jhash on the Ethernet header
810 static u16
vxlan_src_port(const struct vxlan_dev
*vxlan
, struct sk_buff
*skb
)
812 unsigned int range
= (vxlan
->port_max
- vxlan
->port_min
) + 1;
815 hash
= skb_get_rxhash(skb
);
817 hash
= jhash(skb
->data
, 2 * ETH_ALEN
,
818 (__force u32
) skb
->protocol
);
820 return (((u64
) hash
* range
) >> 32) + vxlan
->port_min
;
823 /* Transmit local packets over Vxlan
825 * Outer IP header inherits ECN and DF from inner header.
826 * Outer UDP destination is the VXLAN assigned port.
827 * source port is based on hash of flow
829 static netdev_tx_t
vxlan_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
831 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
833 const struct iphdr
*old_iph
;
836 struct vxlanhdr
*vxh
;
839 unsigned int pkt_len
= skb
->len
;
845 bool did_rsc
= false;
846 const struct vxlan_fdb
*f
;
848 skb_reset_mac_header(skb
);
851 if ((vxlan
->flags
& VXLAN_F_PROXY
) && ntohs(eth
->h_proto
) == ETH_P_ARP
)
852 return arp_reduce(dev
, skb
);
853 else if ((vxlan
->flags
&VXLAN_F_RSC
) && ntohs(eth
->h_proto
) == ETH_P_IP
)
854 did_rsc
= route_shortcircuit(dev
, skb
);
856 f
= vxlan_find_mac(vxlan
, eth
->h_dest
);
860 if (!dst
&& (vxlan
->flags
& VXLAN_F_L2MISS
) &&
861 !is_multicast_ether_addr(eth
->h_dest
))
862 vxlan_fdb_miss(vxlan
, eth
->h_dest
);
868 __skb_pull(skb
, skb_network_offset(skb
));
869 skb
->ip_summed
= CHECKSUM_NONE
;
870 skb
->pkt_type
= PACKET_HOST
;
872 /* short-circuited back to local bridge */
873 if (netif_rx(skb
) == NET_RX_SUCCESS
) {
874 struct vxlan_stats
*stats
=
875 this_cpu_ptr(vxlan
->stats
);
877 u64_stats_update_begin(&stats
->syncp
);
879 stats
->tx_bytes
+= pkt_len
;
880 u64_stats_update_end(&stats
->syncp
);
882 dev
->stats
.tx_errors
++;
883 dev
->stats
.tx_aborted_errors
++;
890 if (!skb
->encapsulation
) {
891 skb_reset_inner_headers(skb
);
892 skb
->encapsulation
= 1;
895 /* Need space for new headers (invalidates iph ptr) */
896 if (skb_cow_head(skb
, VXLAN_HEADROOM
))
899 old_iph
= ip_hdr(skb
);
902 if (!ttl
&& IN_MULTICAST(ntohl(dst
)))
907 tos
= vxlan_get_dsfield(old_iph
, skb
);
909 src_port
= vxlan_src_port(vxlan
, skb
);
911 memset(&fl4
, 0, sizeof(fl4
));
912 fl4
.flowi4_oif
= vxlan
->link
;
913 fl4
.flowi4_tos
= RT_TOS(tos
);
915 fl4
.saddr
= vxlan
->saddr
;
917 rt
= ip_route_output_key(dev_net(dev
), &fl4
);
919 netdev_dbg(dev
, "no route to %pI4\n", &dst
);
920 dev
->stats
.tx_carrier_errors
++;
924 if (rt
->dst
.dev
== dev
) {
925 netdev_dbg(dev
, "circular route to %pI4\n", &dst
);
927 dev
->stats
.collisions
++;
931 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
932 IPCB(skb
)->flags
&= ~(IPSKB_XFRM_TUNNEL_SIZE
| IPSKB_XFRM_TRANSFORMED
|
935 skb_dst_set(skb
, &rt
->dst
);
937 vxh
= (struct vxlanhdr
*) __skb_push(skb
, sizeof(*vxh
));
938 vxh
->vx_flags
= htonl(VXLAN_FLAGS
);
939 vxh
->vx_vni
= htonl(vxlan
->vni
<< 8);
941 __skb_push(skb
, sizeof(*uh
));
942 skb_reset_transport_header(skb
);
945 uh
->dest
= htons(vxlan_port
);
946 uh
->source
= htons(src_port
);
948 uh
->len
= htons(skb
->len
);
951 __skb_push(skb
, sizeof(*iph
));
952 skb_reset_network_header(skb
);
955 iph
->ihl
= sizeof(struct iphdr
) >> 2;
957 iph
->protocol
= IPPROTO_UDP
;
958 iph
->tos
= vxlan_ecn_encap(tos
, old_iph
, skb
);
960 iph
->saddr
= fl4
.saddr
;
961 iph
->ttl
= ttl
? : ip4_dst_hoplimit(&rt
->dst
);
962 tunnel_ip_select_ident(skb
, old_iph
, &rt
->dst
);
966 vxlan_set_owner(dev
, skb
);
968 /* See iptunnel_xmit() */
969 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
970 skb
->ip_summed
= CHECKSUM_NONE
;
972 err
= ip_local_out(skb
);
973 if (likely(net_xmit_eval(err
) == 0)) {
974 struct vxlan_stats
*stats
= this_cpu_ptr(vxlan
->stats
);
976 u64_stats_update_begin(&stats
->syncp
);
978 stats
->tx_bytes
+= pkt_len
;
979 u64_stats_update_end(&stats
->syncp
);
981 dev
->stats
.tx_errors
++;
982 dev
->stats
.tx_aborted_errors
++;
987 dev
->stats
.tx_dropped
++;
991 dev
->stats
.tx_errors
++;
997 /* Walk the forwarding table and purge stale entries */
998 static void vxlan_cleanup(unsigned long arg
)
1000 struct vxlan_dev
*vxlan
= (struct vxlan_dev
*) arg
;
1001 unsigned long next_timer
= jiffies
+ FDB_AGE_INTERVAL
;
1004 if (!netif_running(vxlan
->dev
))
1007 spin_lock_bh(&vxlan
->hash_lock
);
1008 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1009 struct hlist_node
*p
, *n
;
1010 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1012 = container_of(p
, struct vxlan_fdb
, hlist
);
1013 unsigned long timeout
;
1015 if (f
->state
& NUD_PERMANENT
)
1018 timeout
= f
->used
+ vxlan
->age_interval
* HZ
;
1019 if (time_before_eq(timeout
, jiffies
)) {
1020 netdev_dbg(vxlan
->dev
,
1021 "garbage collect %pM\n",
1023 f
->state
= NUD_STALE
;
1024 vxlan_fdb_destroy(vxlan
, f
);
1025 } else if (time_before(timeout
, next_timer
))
1026 next_timer
= timeout
;
1029 spin_unlock_bh(&vxlan
->hash_lock
);
1031 mod_timer(&vxlan
->age_timer
, next_timer
);
1034 /* Setup stats when device is created */
1035 static int vxlan_init(struct net_device
*dev
)
1037 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1039 vxlan
->stats
= alloc_percpu(struct vxlan_stats
);
1046 /* Start ageing timer and join group when device is brought up */
1047 static int vxlan_open(struct net_device
*dev
)
1049 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1053 err
= vxlan_join_group(dev
);
1058 if (vxlan
->age_interval
)
1059 mod_timer(&vxlan
->age_timer
, jiffies
+ FDB_AGE_INTERVAL
);
1064 /* Purge the forwarding table */
1065 static void vxlan_flush(struct vxlan_dev
*vxlan
)
1069 spin_lock_bh(&vxlan
->hash_lock
);
1070 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
) {
1071 struct hlist_node
*p
, *n
;
1072 hlist_for_each_safe(p
, n
, &vxlan
->fdb_head
[h
]) {
1074 = container_of(p
, struct vxlan_fdb
, hlist
);
1075 vxlan_fdb_destroy(vxlan
, f
);
1078 spin_unlock_bh(&vxlan
->hash_lock
);
1081 /* Cleanup timer and forwarding table on shutdown */
1082 static int vxlan_stop(struct net_device
*dev
)
1084 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1087 vxlan_leave_group(dev
);
1089 del_timer_sync(&vxlan
->age_timer
);
1096 /* Merge per-cpu statistics */
1097 static struct rtnl_link_stats64
*vxlan_stats64(struct net_device
*dev
,
1098 struct rtnl_link_stats64
*stats
)
1100 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1101 struct vxlan_stats tmp
, sum
= { 0 };
1104 for_each_possible_cpu(cpu
) {
1106 const struct vxlan_stats
*stats
1107 = per_cpu_ptr(vxlan
->stats
, cpu
);
1110 start
= u64_stats_fetch_begin_bh(&stats
->syncp
);
1111 memcpy(&tmp
, stats
, sizeof(tmp
));
1112 } while (u64_stats_fetch_retry_bh(&stats
->syncp
, start
));
1114 sum
.tx_bytes
+= tmp
.tx_bytes
;
1115 sum
.tx_packets
+= tmp
.tx_packets
;
1116 sum
.rx_bytes
+= tmp
.rx_bytes
;
1117 sum
.rx_packets
+= tmp
.rx_packets
;
1120 stats
->tx_bytes
= sum
.tx_bytes
;
1121 stats
->tx_packets
= sum
.tx_packets
;
1122 stats
->rx_bytes
= sum
.rx_bytes
;
1123 stats
->rx_packets
= sum
.rx_packets
;
1125 stats
->multicast
= dev
->stats
.multicast
;
1126 stats
->rx_length_errors
= dev
->stats
.rx_length_errors
;
1127 stats
->rx_frame_errors
= dev
->stats
.rx_frame_errors
;
1128 stats
->rx_errors
= dev
->stats
.rx_errors
;
1130 stats
->tx_dropped
= dev
->stats
.tx_dropped
;
1131 stats
->tx_carrier_errors
= dev
->stats
.tx_carrier_errors
;
1132 stats
->tx_aborted_errors
= dev
->stats
.tx_aborted_errors
;
1133 stats
->collisions
= dev
->stats
.collisions
;
1134 stats
->tx_errors
= dev
->stats
.tx_errors
;
1139 /* Stub, nothing needs to be done. */
1140 static void vxlan_set_multicast_list(struct net_device
*dev
)
1144 static const struct net_device_ops vxlan_netdev_ops
= {
1145 .ndo_init
= vxlan_init
,
1146 .ndo_open
= vxlan_open
,
1147 .ndo_stop
= vxlan_stop
,
1148 .ndo_start_xmit
= vxlan_xmit
,
1149 .ndo_get_stats64
= vxlan_stats64
,
1150 .ndo_set_rx_mode
= vxlan_set_multicast_list
,
1151 .ndo_change_mtu
= eth_change_mtu
,
1152 .ndo_validate_addr
= eth_validate_addr
,
1153 .ndo_set_mac_address
= eth_mac_addr
,
1154 .ndo_fdb_add
= vxlan_fdb_add
,
1155 .ndo_fdb_del
= vxlan_fdb_delete
,
1156 .ndo_fdb_dump
= vxlan_fdb_dump
,
1159 /* Info for udev, that this is a virtual tunnel endpoint */
1160 static struct device_type vxlan_type
= {
1164 static void vxlan_free(struct net_device
*dev
)
1166 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1168 free_percpu(vxlan
->stats
);
1172 /* Initialize the device structure. */
1173 static void vxlan_setup(struct net_device
*dev
)
1175 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1179 eth_hw_addr_random(dev
);
1181 dev
->hard_header_len
= ETH_HLEN
+ VXLAN_HEADROOM
;
1183 dev
->netdev_ops
= &vxlan_netdev_ops
;
1184 dev
->destructor
= vxlan_free
;
1185 SET_NETDEV_DEVTYPE(dev
, &vxlan_type
);
1187 dev
->tx_queue_len
= 0;
1188 dev
->features
|= NETIF_F_LLTX
;
1189 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1190 dev
->features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1191 dev
->features
|= NETIF_F_RXCSUM
;
1193 dev
->hw_features
|= NETIF_F_SG
| NETIF_F_HW_CSUM
| NETIF_F_RXCSUM
;
1194 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1195 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1197 spin_lock_init(&vxlan
->hash_lock
);
1199 init_timer_deferrable(&vxlan
->age_timer
);
1200 vxlan
->age_timer
.function
= vxlan_cleanup
;
1201 vxlan
->age_timer
.data
= (unsigned long) vxlan
;
1203 inet_get_local_port_range(&low
, &high
);
1204 vxlan
->port_min
= low
;
1205 vxlan
->port_max
= high
;
1209 for (h
= 0; h
< FDB_HASH_SIZE
; ++h
)
1210 INIT_HLIST_HEAD(&vxlan
->fdb_head
[h
]);
1213 static const struct nla_policy vxlan_policy
[IFLA_VXLAN_MAX
+ 1] = {
1214 [IFLA_VXLAN_ID
] = { .type
= NLA_U32
},
1215 [IFLA_VXLAN_GROUP
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
1216 [IFLA_VXLAN_LINK
] = { .type
= NLA_U32
},
1217 [IFLA_VXLAN_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
1218 [IFLA_VXLAN_TOS
] = { .type
= NLA_U8
},
1219 [IFLA_VXLAN_TTL
] = { .type
= NLA_U8
},
1220 [IFLA_VXLAN_LEARNING
] = { .type
= NLA_U8
},
1221 [IFLA_VXLAN_AGEING
] = { .type
= NLA_U32
},
1222 [IFLA_VXLAN_LIMIT
] = { .type
= NLA_U32
},
1223 [IFLA_VXLAN_PORT_RANGE
] = { .len
= sizeof(struct ifla_vxlan_port_range
) },
1224 [IFLA_VXLAN_PROXY
] = { .type
= NLA_U8
},
1225 [IFLA_VXLAN_RSC
] = { .type
= NLA_U8
},
1226 [IFLA_VXLAN_L2MISS
] = { .type
= NLA_U8
},
1227 [IFLA_VXLAN_L3MISS
] = { .type
= NLA_U8
},
1230 static int vxlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
1232 if (tb
[IFLA_ADDRESS
]) {
1233 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
) {
1234 pr_debug("invalid link address (not ethernet)\n");
1238 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
]))) {
1239 pr_debug("invalid all zero ethernet address\n");
1240 return -EADDRNOTAVAIL
;
1247 if (data
[IFLA_VXLAN_ID
]) {
1248 __u32 id
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1249 if (id
>= VXLAN_VID_MASK
)
1253 if (data
[IFLA_VXLAN_GROUP
]) {
1254 __be32 gaddr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1255 if (!IN_MULTICAST(ntohl(gaddr
))) {
1256 pr_debug("group address is not IPv4 multicast\n");
1257 return -EADDRNOTAVAIL
;
1261 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1262 const struct ifla_vxlan_port_range
*p
1263 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1265 if (ntohs(p
->high
) < ntohs(p
->low
)) {
1266 pr_debug("port range %u .. %u not valid\n",
1267 ntohs(p
->low
), ntohs(p
->high
));
1275 static void vxlan_get_drvinfo(struct net_device
*netdev
,
1276 struct ethtool_drvinfo
*drvinfo
)
1278 strlcpy(drvinfo
->version
, VXLAN_VERSION
, sizeof(drvinfo
->version
));
1279 strlcpy(drvinfo
->driver
, "vxlan", sizeof(drvinfo
->driver
));
1282 static const struct ethtool_ops vxlan_ethtool_ops
= {
1283 .get_drvinfo
= vxlan_get_drvinfo
,
1284 .get_link
= ethtool_op_get_link
,
1287 static int vxlan_newlink(struct net
*net
, struct net_device
*dev
,
1288 struct nlattr
*tb
[], struct nlattr
*data
[])
1290 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1294 if (!data
[IFLA_VXLAN_ID
])
1297 vni
= nla_get_u32(data
[IFLA_VXLAN_ID
]);
1298 if (vxlan_find_vni(net
, vni
)) {
1299 pr_info("duplicate VNI %u\n", vni
);
1304 if (data
[IFLA_VXLAN_GROUP
])
1305 vxlan
->gaddr
= nla_get_be32(data
[IFLA_VXLAN_GROUP
]);
1307 if (data
[IFLA_VXLAN_LOCAL
])
1308 vxlan
->saddr
= nla_get_be32(data
[IFLA_VXLAN_LOCAL
]);
1310 if (data
[IFLA_VXLAN_LINK
] &&
1311 (vxlan
->link
= nla_get_u32(data
[IFLA_VXLAN_LINK
]))) {
1312 struct net_device
*lowerdev
1313 = __dev_get_by_index(net
, vxlan
->link
);
1316 pr_info("ifindex %d does not exist\n", vxlan
->link
);
1321 dev
->mtu
= lowerdev
->mtu
- VXLAN_HEADROOM
;
1323 /* update header length based on lower device */
1324 dev
->hard_header_len
= lowerdev
->hard_header_len
+
1328 if (data
[IFLA_VXLAN_TOS
])
1329 vxlan
->tos
= nla_get_u8(data
[IFLA_VXLAN_TOS
]);
1331 if (data
[IFLA_VXLAN_TTL
])
1332 vxlan
->ttl
= nla_get_u8(data
[IFLA_VXLAN_TTL
]);
1334 if (!data
[IFLA_VXLAN_LEARNING
] || nla_get_u8(data
[IFLA_VXLAN_LEARNING
]))
1335 vxlan
->flags
|= VXLAN_F_LEARN
;
1337 if (data
[IFLA_VXLAN_AGEING
])
1338 vxlan
->age_interval
= nla_get_u32(data
[IFLA_VXLAN_AGEING
]);
1340 vxlan
->age_interval
= FDB_AGE_DEFAULT
;
1342 if (data
[IFLA_VXLAN_PROXY
] && nla_get_u8(data
[IFLA_VXLAN_PROXY
]))
1343 vxlan
->flags
|= VXLAN_F_PROXY
;
1345 if (data
[IFLA_VXLAN_RSC
] && nla_get_u8(data
[IFLA_VXLAN_RSC
]))
1346 vxlan
->flags
|= VXLAN_F_RSC
;
1348 if (data
[IFLA_VXLAN_L2MISS
] && nla_get_u8(data
[IFLA_VXLAN_L2MISS
]))
1349 vxlan
->flags
|= VXLAN_F_L2MISS
;
1351 if (data
[IFLA_VXLAN_L3MISS
] && nla_get_u8(data
[IFLA_VXLAN_L3MISS
]))
1352 vxlan
->flags
|= VXLAN_F_L3MISS
;
1354 if (data
[IFLA_VXLAN_LIMIT
])
1355 vxlan
->addrmax
= nla_get_u32(data
[IFLA_VXLAN_LIMIT
]);
1357 if (data
[IFLA_VXLAN_PORT_RANGE
]) {
1358 const struct ifla_vxlan_port_range
*p
1359 = nla_data(data
[IFLA_VXLAN_PORT_RANGE
]);
1360 vxlan
->port_min
= ntohs(p
->low
);
1361 vxlan
->port_max
= ntohs(p
->high
);
1364 SET_ETHTOOL_OPS(dev
, &vxlan_ethtool_ops
);
1366 err
= register_netdevice(dev
);
1368 hlist_add_head_rcu(&vxlan
->hlist
, vni_head(net
, vxlan
->vni
));
1373 static void vxlan_dellink(struct net_device
*dev
, struct list_head
*head
)
1375 struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1377 hlist_del_rcu(&vxlan
->hlist
);
1379 unregister_netdevice_queue(dev
, head
);
1382 static size_t vxlan_get_size(const struct net_device
*dev
)
1385 return nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_ID */
1386 nla_total_size(sizeof(__be32
)) +/* IFLA_VXLAN_GROUP */
1387 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LINK */
1388 nla_total_size(sizeof(__be32
))+ /* IFLA_VXLAN_LOCAL */
1389 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TTL */
1390 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_TOS */
1391 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_LEARNING */
1392 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_PROXY */
1393 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_RSC */
1394 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L2MISS */
1395 nla_total_size(sizeof(__u8
)) + /* IFLA_VXLAN_L3MISS */
1396 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_AGEING */
1397 nla_total_size(sizeof(__u32
)) + /* IFLA_VXLAN_LIMIT */
1398 nla_total_size(sizeof(struct ifla_vxlan_port_range
)) +
1402 static int vxlan_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1404 const struct vxlan_dev
*vxlan
= netdev_priv(dev
);
1405 struct ifla_vxlan_port_range ports
= {
1406 .low
= htons(vxlan
->port_min
),
1407 .high
= htons(vxlan
->port_max
),
1410 if (nla_put_u32(skb
, IFLA_VXLAN_ID
, vxlan
->vni
))
1411 goto nla_put_failure
;
1413 if (vxlan
->gaddr
&& nla_put_be32(skb
, IFLA_VXLAN_GROUP
, vxlan
->gaddr
))
1414 goto nla_put_failure
;
1416 if (vxlan
->link
&& nla_put_u32(skb
, IFLA_VXLAN_LINK
, vxlan
->link
))
1417 goto nla_put_failure
;
1419 if (vxlan
->saddr
&& nla_put_be32(skb
, IFLA_VXLAN_LOCAL
, vxlan
->saddr
))
1420 goto nla_put_failure
;
1422 if (nla_put_u8(skb
, IFLA_VXLAN_TTL
, vxlan
->ttl
) ||
1423 nla_put_u8(skb
, IFLA_VXLAN_TOS
, vxlan
->tos
) ||
1424 nla_put_u8(skb
, IFLA_VXLAN_LEARNING
,
1425 !!(vxlan
->flags
& VXLAN_F_LEARN
)) ||
1426 nla_put_u8(skb
, IFLA_VXLAN_PROXY
,
1427 !!(vxlan
->flags
& VXLAN_F_PROXY
)) ||
1428 nla_put_u8(skb
, IFLA_VXLAN_RSC
, !!(vxlan
->flags
& VXLAN_F_RSC
)) ||
1429 nla_put_u8(skb
, IFLA_VXLAN_L2MISS
,
1430 !!(vxlan
->flags
& VXLAN_F_L2MISS
)) ||
1431 nla_put_u8(skb
, IFLA_VXLAN_L3MISS
,
1432 !!(vxlan
->flags
& VXLAN_F_L3MISS
)) ||
1433 nla_put_u32(skb
, IFLA_VXLAN_AGEING
, vxlan
->age_interval
) ||
1434 nla_put_u32(skb
, IFLA_VXLAN_LIMIT
, vxlan
->addrmax
))
1435 goto nla_put_failure
;
1437 if (nla_put(skb
, IFLA_VXLAN_PORT_RANGE
, sizeof(ports
), &ports
))
1438 goto nla_put_failure
;
1446 static struct rtnl_link_ops vxlan_link_ops __read_mostly
= {
1448 .maxtype
= IFLA_VXLAN_MAX
,
1449 .policy
= vxlan_policy
,
1450 .priv_size
= sizeof(struct vxlan_dev
),
1451 .setup
= vxlan_setup
,
1452 .validate
= vxlan_validate
,
1453 .newlink
= vxlan_newlink
,
1454 .dellink
= vxlan_dellink
,
1455 .get_size
= vxlan_get_size
,
1456 .fill_info
= vxlan_fill_info
,
1459 static __net_init
int vxlan_init_net(struct net
*net
)
1461 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1463 struct sockaddr_in vxlan_addr
= {
1464 .sin_family
= AF_INET
,
1465 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
1470 /* Create UDP socket for encapsulation receive. */
1471 rc
= sock_create_kern(AF_INET
, SOCK_DGRAM
, IPPROTO_UDP
, &vn
->sock
);
1473 pr_debug("UDP socket create failed\n");
1476 /* Put in proper namespace */
1478 sk_change_net(sk
, net
);
1480 vxlan_addr
.sin_port
= htons(vxlan_port
);
1482 rc
= kernel_bind(vn
->sock
, (struct sockaddr
*) &vxlan_addr
,
1483 sizeof(vxlan_addr
));
1485 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1486 &vxlan_addr
.sin_addr
, ntohs(vxlan_addr
.sin_port
), rc
);
1487 sk_release_kernel(sk
);
1492 /* Disable multicast loopback */
1493 inet_sk(sk
)->mc_loop
= 0;
1495 /* Mark socket as an encapsulation socket. */
1496 udp_sk(sk
)->encap_type
= 1;
1497 udp_sk(sk
)->encap_rcv
= vxlan_udp_encap_recv
;
1500 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1501 INIT_HLIST_HEAD(&vn
->vni_list
[h
]);
1506 static __net_exit
void vxlan_exit_net(struct net
*net
)
1508 struct vxlan_net
*vn
= net_generic(net
, vxlan_net_id
);
1509 struct vxlan_dev
*vxlan
;
1513 for (h
= 0; h
< VNI_HASH_SIZE
; ++h
)
1514 hlist_for_each_entry(vxlan
, &vn
->vni_list
[h
], hlist
)
1515 dev_close(vxlan
->dev
);
1519 sk_release_kernel(vn
->sock
->sk
);
1524 static struct pernet_operations vxlan_net_ops
= {
1525 .init
= vxlan_init_net
,
1526 .exit
= vxlan_exit_net
,
1527 .id
= &vxlan_net_id
,
1528 .size
= sizeof(struct vxlan_net
),
1531 static int __init
vxlan_init_module(void)
1535 get_random_bytes(&vxlan_salt
, sizeof(vxlan_salt
));
1537 rc
= register_pernet_device(&vxlan_net_ops
);
1541 rc
= rtnl_link_register(&vxlan_link_ops
);
1548 unregister_pernet_device(&vxlan_net_ops
);
1552 module_init(vxlan_init_module
);
1554 static void __exit
vxlan_cleanup_module(void)
1556 rtnl_link_unregister(&vxlan_link_ops
);
1557 unregister_pernet_device(&vxlan_net_ops
);
1559 module_exit(vxlan_cleanup_module
);
1561 MODULE_LICENSE("GPL");
1562 MODULE_VERSION(VXLAN_VERSION
);
1563 MODULE_AUTHOR("Stephen Hemminger <shemminger@vyatta.com>");
1564 MODULE_ALIAS_RTNL_LINK("vxlan");