projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'qlcnic'
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
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.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
36#include <net/rtnetlink.h>
37#include <net/route.h>
38#include <net/dsfield.h>
39#include <net/inet_ecn.h>
40#include <net/net_namespace.h>
41#include <net/netns/generic.h>
012a5729 42#include <net/vxlan.h>
e4c7ed41
CW		 43#if IS_ENABLED(CONFIG_IPV6)
44#include <net/ipv6.h>
45#include <net/addrconf.h>
46#include <net/ip6_tunnel.h>
47#endif
d342894c 48
49#define VXLAN_VERSION "0.1"
50
553675fb 51#define PORT_HASH_BITS 8
52#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 53#define VNI_HASH_BITS 10
54#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
55#define FDB_HASH_BITS 8
56#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
57#define FDB_AGE_DEFAULT 300 /* 5 min */
58#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
59
60#define VXLAN_N_VID (1u << 24)
61#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
52b702ff
AD		 62/* IP header + UDP + VXLAN + Ethernet header */
63#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
e4c7ed41
CW		 64/* IPv6 header + UDP + VXLAN + Ethernet header */
65#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
7ce04758 66#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
d342894c 67
68#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
69
70/* VXLAN protocol header */
71struct vxlanhdr {
72 __be32 vx_flags;
73 __be32 vx_vni;
74};
75
23c578bf 76/* UDP port for VXLAN traffic.
77 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 78 * for compatibility with early adopters.
23c578bf 79 */
9daaa397
SH		 80static unsigned short vxlan_port __read_mostly = 8472;
81module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 82MODULE_PARM_DESC(udp_port, "Destination UDP port");
83
84static bool log_ecn_error = true;
85module_param(log_ecn_error, bool, 0644);
86MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
87
60d9d4c6 88static int vxlan_net_id;
553675fb 89
afbd8bae
MR		 90static const u8 all_zeros_mac[ETH_ALEN];
91
553675fb 92/* per-network namespace private data for this module */
93struct vxlan_net {
94 struct list_head vxlan_list;
95 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 96 spinlock_t sock_lock;
553675fb 97};
98
e4c7ed41
CW		 99union vxlan_addr {
100 struct sockaddr_in sin;
101 struct sockaddr_in6 sin6;
102 struct sockaddr sa;
103};
104
6681712d 105struct vxlan_rdst {
e4c7ed41 106 union vxlan_addr remote_ip;
6681712d
DS		 107 __be16 remote_port;
108 u32 remote_vni;
109 u32 remote_ifindex;
3e61aa8f 110 struct list_head list;
bc7892ba 111 struct rcu_head rcu;
6681712d
DS		 112};
113
d342894c 114/* Forwarding table entry */
115struct vxlan_fdb {
116 struct hlist_node hlist; /* linked list of entries */
117 struct rcu_head rcu;
118 unsigned long updated; /* jiffies */
119 unsigned long used;
3e61aa8f 120 struct list_head remotes;
d342894c 121 u16 state; /* see ndm_state */
ae884082 122 u8 flags; /* see ndm_flags */
d342894c 123 u8 eth_addr[ETH_ALEN];
124};
125
d342894c 126/* Pseudo network device */
127struct vxlan_dev {
553675fb 128 struct hlist_node hlist; /* vni hash table */
129 struct list_head next; /* vxlan's per namespace list */
130 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 131 struct net_device *dev;
c7995c43 132 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 133 union vxlan_addr saddr; /* source address */
823aa873 134 __be16 dst_port;
05f47d69 135 __u16 port_min; /* source port range */
136 __u16 port_max;
d342894c 137 __u8 tos; /* TOS override */
138 __u8 ttl;
e4f67add 139 u32 flags; /* VXLAN_F_* below */
d342894c 140
1c51a915 141 struct work_struct sock_work;
3fc2de2f 142 struct work_struct igmp_join;
143 struct work_struct igmp_leave;
1c51a915 144
d342894c 145 unsigned long age_interval;
146 struct timer_list age_timer;
147 spinlock_t hash_lock;
148 unsigned int addrcnt;
149 unsigned int addrmax;
d342894c 150
151 struct hlist_head fdb_head[FDB_HASH_SIZE];
152};
153
e4f67add
DS		 154#define VXLAN_F_LEARN 0x01
155#define VXLAN_F_PROXY 0x02
156#define VXLAN_F_RSC 0x04
157#define VXLAN_F_L2MISS 0x08
158#define VXLAN_F_L3MISS 0x10
e4c7ed41 159#define VXLAN_F_IPV6 0x20 /* internal flag */
e4f67add 160
d342894c 161/* salt for hash table */
162static u32 vxlan_salt __read_mostly;
758c57d1 163static struct workqueue_struct *vxlan_wq;
d342894c 164
1c51a915
SH		 165static void vxlan_sock_work(struct work_struct *work);
166
e4c7ed41
CW		 167#if IS_ENABLED(CONFIG_IPV6)
168static inline
169bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
170{
171 if (a->sa.sa_family != b->sa.sa_family)
172 return false;
173 if (a->sa.sa_family == AF_INET6)
174 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
175 else
176 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
177}
178
179static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
180{
181 if (ipa->sa.sa_family == AF_INET6)
182 return ipv6_addr_any(&ipa->sin6.sin6_addr);
183 else
184 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
185}
186
187static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
188{
189 if (ipa->sa.sa_family == AF_INET6)
190 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
191 else
192 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
193}
194
195static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
196{
197 if (nla_len(nla) >= sizeof(struct in6_addr)) {
198 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
199 ip->sa.sa_family = AF_INET6;
200 return 0;
201 } else if (nla_len(nla) >= sizeof(__be32)) {
202 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
203 ip->sa.sa_family = AF_INET;
204 return 0;
205 } else {
206 return -EAFNOSUPPORT;
207 }
208}
209
210static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
211 const union vxlan_addr *ip)
212{
213 if (ip->sa.sa_family == AF_INET6)
214 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
215 else
216 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
217}
218
219#else /* !CONFIG_IPV6 */
220
221static inline
222bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
223{
224 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
225}
226
227static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
228{
229 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
230}
231
232static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
233{
234 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
235}
236
237static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
238{
239 if (nla_len(nla) >= sizeof(struct in6_addr)) {
240 return -EAFNOSUPPORT;
241 } else if (nla_len(nla) >= sizeof(__be32)) {
242 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
243 ip->sa.sa_family = AF_INET;
244 return 0;
245 } else {
246 return -EAFNOSUPPORT;
247 }
248}
249
250static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
251 const union vxlan_addr *ip)
252{
253 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
254}
255#endif
256
553675fb 257/* Virtual Network hash table head */
258static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
259{
260 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
261}
262
263/* Socket hash table head */
264static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 265{
266 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
267
553675fb 268 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
269}
270
3e61aa8f
SH		 271/* First remote destination for a forwarding entry.
272 * Guaranteed to be non-NULL because remotes are never deleted.
273 */
5ca5461c 274static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 275{
5ca5461c 276 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
277}
278
279static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
280{
281 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 282}
283
553675fb 284/* Find VXLAN socket based on network namespace and UDP port */
9c2e24e1 285static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
553675fb 286{
287 struct vxlan_sock *vs;
288
289 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
290 if (inet_sk(vs->sock->sk)->inet_sport == port)
291 return vs;
292 }
293 return NULL;
d342894c 294}
295
5cfccc5a 296static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 297{
298 struct vxlan_dev *vxlan;
d342894c 299
553675fb 300 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 301 if (vxlan->default_dst.remote_vni == id)
d342894c 302 return vxlan;
303 }
304
305 return NULL;
306}
307
5cfccc5a
PS		 308/* Look up VNI in a per net namespace table */
309static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
310{
311 struct vxlan_sock *vs;
312
313 vs = vxlan_find_sock(net, port);
314 if (!vs)
315 return NULL;
316
317 return vxlan_vs_find_vni(vs, id);
318}
319
d342894c 320/* Fill in neighbour message in skbuff. */
321static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 322 const struct vxlan_fdb *fdb,
323 u32 portid, u32 seq, int type, unsigned int flags,
324 const struct vxlan_rdst *rdst)
d342894c 325{
326 unsigned long now = jiffies;
327 struct nda_cacheinfo ci;
328 struct nlmsghdr *nlh;
329 struct ndmsg *ndm;
e4f67add 330 bool send_ip, send_eth;
d342894c 331
332 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
333 if (nlh == NULL)
334 return -EMSGSIZE;
335
336 ndm = nlmsg_data(nlh);
337 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 338
339 send_eth = send_ip = true;
340
341 if (type == RTM_GETNEIGH) {
342 ndm->ndm_family = AF_INET;
e4c7ed41 343 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 344 send_eth = !is_zero_ether_addr(fdb->eth_addr);
345 } else
346 ndm->ndm_family = AF_BRIDGE;
d342894c 347 ndm->ndm_state = fdb->state;
348 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 349 ndm->ndm_flags = fdb->flags;
d342894c 350 ndm->ndm_type = NDA_DST;
351
e4f67add 352 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 353 goto nla_put_failure;
354
e4c7ed41 355 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 356 goto nla_put_failure;
357
823aa873 358 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 359 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
360 goto nla_put_failure;
c7995c43 361 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 362 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 363 goto nla_put_failure;
364 if (rdst->remote_ifindex &&
365 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 366 goto nla_put_failure;
367
368 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
369 ci.ndm_confirmed = 0;
370 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
371 ci.ndm_refcnt = 0;
372
373 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
374 goto nla_put_failure;
375
376 return nlmsg_end(skb, nlh);
377
378nla_put_failure:
379 nlmsg_cancel(skb, nlh);
380 return -EMSGSIZE;
381}
382
383static inline size_t vxlan_nlmsg_size(void)
384{
385 return NLMSG_ALIGN(sizeof(struct ndmsg))
386 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 387 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 388 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 389 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
390 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 391 + nla_total_size(sizeof(struct nda_cacheinfo));
392}
393
394static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
3e61aa8f 395 struct vxlan_fdb *fdb, int type)
d342894c 396{
397 struct net *net = dev_net(vxlan->dev);
398 struct sk_buff *skb;
399 int err = -ENOBUFS;
400
401 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
402 if (skb == NULL)
403 goto errout;
404
5ca5461c 405 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
406 first_remote_rtnl(fdb));
d342894c 407 if (err < 0) {
408 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
409 WARN_ON(err == -EMSGSIZE);
410 kfree_skb(skb);
411 goto errout;
412 }
413
414 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
415 return;
416errout:
417 if (err < 0)
418 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
419}
420
e4c7ed41 421static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 422{
423 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 424 struct vxlan_fdb f = {
425 .state = NUD_STALE,
426 };
427 struct vxlan_rdst remote = {
e4c7ed41 428 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 429 .remote_vni = VXLAN_N_VID,
430 };
3e61aa8f
SH		 431
432 INIT_LIST_HEAD(&f.remotes);
433 list_add_rcu(&remote.list, &f.remotes);
e4f67add
DS		 434
435 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
436}
437
438static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
439{
bb3fd687
SH		 440 struct vxlan_fdb f = {
441 .state = NUD_STALE,
442 };
e4f67add 443
3e61aa8f 444 INIT_LIST_HEAD(&f.remotes);
e4f67add
DS		 445 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
446
447 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
448}
449
d342894c 450/* Hash Ethernet address */
451static u32 eth_hash(const unsigned char *addr)
452{
453 u64 value = get_unaligned((u64 *)addr);
454
455 /* only want 6 bytes */
456#ifdef __BIG_ENDIAN
d342894c 457 value >>= 16;
321fb991 458#else
459 value <<= 16;
d342894c 460#endif
461 return hash_64(value, FDB_HASH_BITS);
462}
463
464/* Hash chain to use given mac address */
465static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
466 const u8 *mac)
467{
468 return &vxlan->fdb_head[eth_hash(mac)];
469}
470
471/* Look up Ethernet address in forwarding table */
014be2c8 472static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 473 const u8 *mac)
474
475{
476 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
477 struct vxlan_fdb *f;
d342894c 478
b67bfe0d 479 hlist_for_each_entry_rcu(f, head, hlist) {
d342894c 480 if (compare_ether_addr(mac, f->eth_addr) == 0)
481 return f;
482 }
483
484 return NULL;
485}
486
014be2c8
SS		 487static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
488 const u8 *mac)
489{
490 struct vxlan_fdb *f;
491
492 f = __vxlan_find_mac(vxlan, mac);
493 if (f)
494 f->used = jiffies;
495
496 return f;
497}
498
a5e7c10a
MR		 499/* caller should hold vxlan->hash_lock */
500static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 501 union vxlan_addr *ip, __be16 port,
a5e7c10a 502 __u32 vni, __u32 ifindex)
6681712d 503{
3e61aa8f 504 struct vxlan_rdst *rd;
6681712d 505
3e61aa8f 506 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 507 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 508 rd->remote_port == port &&
509 rd->remote_vni == vni &&
510 rd->remote_ifindex == ifindex)
a5e7c10a 511 return rd;
6681712d 512 }
3e61aa8f 513
a5e7c10a
MR		 514 return NULL;
515}
516
906dc186
TR		 517/* Replace destination of unicast mac */
518static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 519 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 520{
521 struct vxlan_rdst *rd;
522
523 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
524 if (rd)
525 return 0;
526
527 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
528 if (!rd)
529 return 0;
e4c7ed41 530 rd->remote_ip = *ip;
906dc186
TR		 531 rd->remote_port = port;
532 rd->remote_vni = vni;
533 rd->remote_ifindex = ifindex;
534 return 1;
535}
536
a5e7c10a
MR		 537/* Add/update destinations for multicast */
538static int vxlan_fdb_append(struct vxlan_fdb *f,
e4c7ed41 539 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
a5e7c10a
MR		 540{
541 struct vxlan_rdst *rd;
542
543 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
544 if (rd)
545 return 0;
546
6681712d
DS		 547 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
548 if (rd == NULL)
549 return -ENOBUFS;
e4c7ed41 550 rd->remote_ip = *ip;
6681712d
DS		 551 rd->remote_port = port;
552 rd->remote_vni = vni;
553 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 554
555 list_add_tail_rcu(&rd->list, &f->remotes);
556
6681712d
DS		 557 return 1;
558}
559
d342894c 560/* Add new entry to forwarding table -- assumes lock held */
561static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 562 const u8 *mac, union vxlan_addr *ip,
6681712d 563 __u16 state, __u16 flags,
73cf3317 564 __be16 port, __u32 vni, __u32 ifindex,
ae884082 565 __u8 ndm_flags)
d342894c 566{
567 struct vxlan_fdb *f;
568 int notify = 0;
569
014be2c8 570 f = __vxlan_find_mac(vxlan, mac);
d342894c 571 if (f) {
572 if (flags & NLM_F_EXCL) {
573 netdev_dbg(vxlan->dev,
574 "lost race to create %pM\n", mac);
575 return -EEXIST;
576 }
577 if (f->state != state) {
578 f->state = state;
579 f->updated = jiffies;
580 notify = 1;
581 }
ae884082
DS		 582 if (f->flags != ndm_flags) {
583 f->flags = ndm_flags;
584 f->updated = jiffies;
585 notify = 1;
586 }
906dc186
TR		 587 if ((flags & NLM_F_REPLACE)) {
588 /* Only change unicasts */
589 if (!(is_multicast_ether_addr(f->eth_addr) ||
590 is_zero_ether_addr(f->eth_addr))) {
591 int rc = vxlan_fdb_replace(f, ip, port, vni,
592 ifindex);
593
594 if (rc < 0)
595 return rc;
596 notify |= rc;
597 } else
598 return -EOPNOTSUPP;
599 }
6681712d 600 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 601 (is_multicast_ether_addr(f->eth_addr) ||
602 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 603 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
604
605 if (rc < 0)
606 return rc;
607 notify |= rc;
608 }
d342894c 609 } else {
610 if (!(flags & NLM_F_CREATE))
611 return -ENOENT;
612
613 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
614 return -ENOSPC;
615
906dc186
TR		 616 /* Disallow replace to add a multicast entry */
617 if ((flags & NLM_F_REPLACE) &&
618 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
619 return -EOPNOTSUPP;
620
e4c7ed41 621 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 622 f = kmalloc(sizeof(*f), GFP_ATOMIC);
623 if (!f)
624 return -ENOMEM;
625
626 notify = 1;
d342894c 627 f->state = state;
ae884082 628 f->flags = ndm_flags;
d342894c 629 f->updated = f->used = jiffies;
3e61aa8f 630 INIT_LIST_HEAD(&f->remotes);
d342894c 631 memcpy(f->eth_addr, mac, ETH_ALEN);
632
3e61aa8f
SH		 633 vxlan_fdb_append(f, ip, port, vni, ifindex);
634
d342894c 635 ++vxlan->addrcnt;
636 hlist_add_head_rcu(&f->hlist,
637 vxlan_fdb_head(vxlan, mac));
638 }
639
640 if (notify)
641 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
642
643 return 0;
644}
645
6706c82e 646static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 647{
648 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 649 struct vxlan_rdst *rd, *nd;
6681712d 650
3e61aa8f 651 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 652 kfree(rd);
6681712d
DS		 653 kfree(f);
654}
655
d342894c 656static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
657{
658 netdev_dbg(vxlan->dev,
659 "delete %pM\n", f->eth_addr);
660
661 --vxlan->addrcnt;
662 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
663
664 hlist_del_rcu(&f->hlist);
6681712d 665 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 666}
667
f0b074be 668static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 669 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 670{
6681712d 671 struct net *net = dev_net(vxlan->dev);
e4c7ed41 672 int err;
d342894c 673
f0b074be 674 if (tb[NDA_DST]) {
e4c7ed41
CW		 675 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
676 if (err)
677 return err;
f0b074be 678 } else {
e4c7ed41
CW		 679 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
680 if (remote->sa.sa_family == AF_INET) {
681 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
682 ip->sa.sa_family = AF_INET;
683#if IS_ENABLED(CONFIG_IPV6)
684 } else {
685 ip->sin6.sin6_addr = in6addr_any;
686 ip->sa.sa_family = AF_INET6;
687#endif
688 }
f0b074be 689 }
d342894c 690
6681712d 691 if (tb[NDA_PORT]) {
73cf3317 692 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 693 return -EINVAL;
f0b074be
MR		 694 *port = nla_get_be16(tb[NDA_PORT]);
695 } else {
696 *port = vxlan->dst_port;
697 }
6681712d
DS		 698
699 if (tb[NDA_VNI]) {
700 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
701 return -EINVAL;
f0b074be
MR		 702 *vni = nla_get_u32(tb[NDA_VNI]);
703 } else {
704 *vni = vxlan->default_dst.remote_vni;
705 }
6681712d
DS		 706
707 if (tb[NDA_IFINDEX]) {
5abb0029 708 struct net_device *tdev;
6681712d
DS		 709
710 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
711 return -EINVAL;
f0b074be
MR		 712 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
713 tdev = dev_get_by_index(net, *ifindex);
5abb0029 714 if (!tdev)
6681712d 715 return -EADDRNOTAVAIL;
5abb0029 716 dev_put(tdev);
f0b074be
MR		 717 } else {
718 *ifindex = 0;
719 }
720
721 return 0;
722}
723
724/* Add static entry (via netlink) */
725static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
726 struct net_device *dev,
727 const unsigned char *addr, u16 flags)
728{
729 struct vxlan_dev *vxlan = netdev_priv(dev);
730 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 731 union vxlan_addr ip;
f0b074be
MR		 732 __be16 port;
733 u32 vni, ifindex;
734 int err;
735
736 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
737 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
738 ndm->ndm_state);
739 return -EINVAL;
740 }
741
742 if (tb[NDA_DST] == NULL)
743 return -EINVAL;
744
745 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
746 if (err)
747 return err;
6681712d 748
d342894c 749 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 750 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 751 port, vni, ifindex, ndm->ndm_flags);
d342894c 752 spin_unlock_bh(&vxlan->hash_lock);
753
754 return err;
755}
756
757/* Delete entry (via netlink) */
1690be63
VY		 758static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
759 struct net_device *dev,
d342894c 760 const unsigned char *addr)
761{
762 struct vxlan_dev *vxlan = netdev_priv(dev);
763 struct vxlan_fdb *f;
bc7892ba 764 struct vxlan_rdst *rd = NULL;
e4c7ed41 765 union vxlan_addr ip;
bc7892ba
MR		 766 __be16 port;
767 u32 vni, ifindex;
768 int err;
769
770 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
771 if (err)
772 return err;
773
774 err = -ENOENT;
d342894c 775
776 spin_lock_bh(&vxlan->hash_lock);
777 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 778 if (!f)
779 goto out;
780
e4c7ed41
CW		 781 if (!vxlan_addr_any(&ip)) {
782 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 783 if (!rd)
784 goto out;
785 }
786
787 err = 0;
788
789 /* remove a destination if it's not the only one on the list,
790 * otherwise destroy the fdb entry
791 */
792 if (rd && !list_is_singular(&f->remotes)) {
793 list_del_rcu(&rd->list);
dcdc7a59 794 kfree_rcu(rd, rcu);
bc7892ba 795 goto out;
d342894c 796 }
bc7892ba
MR		 797
798 vxlan_fdb_destroy(vxlan, f);
799
800out:
d342894c 801 spin_unlock_bh(&vxlan->hash_lock);
802
803 return err;
804}
805
806/* Dump forwarding table */
807static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
808 struct net_device *dev, int idx)
809{
810 struct vxlan_dev *vxlan = netdev_priv(dev);
811 unsigned int h;
812
813 for (h = 0; h < FDB_HASH_SIZE; ++h) {
814 struct vxlan_fdb *f;
d342894c 815 int err;
816
b67bfe0d 817 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 818 struct vxlan_rdst *rd;
6681712d 819
3e61aa8f
SH		 820 if (idx < cb->args[0])
821 goto skip;
822
823 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 824 err = vxlan_fdb_info(skb, vxlan, f,
825 NETLINK_CB(cb->skb).portid,
826 cb->nlh->nlmsg_seq,
827 RTM_NEWNEIGH,
828 NLM_F_MULTI, rd);
829 if (err < 0)
3e61aa8f 830 goto out;
6681712d 831 }
3e61aa8f
SH		 832skip:
833 ++idx;
d342894c 834 }
835 }
3e61aa8f 836out:
d342894c 837 return idx;
838}
839
840/* Watch incoming packets to learn mapping between Ethernet address
841 * and Tunnel endpoint.
26a41ae6 842 * Return true if packet is bogus and should be droppped.
d342894c 843 */
26a41ae6 844static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 845 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 846{
847 struct vxlan_dev *vxlan = netdev_priv(dev);
848 struct vxlan_fdb *f;
d342894c 849
850 f = vxlan_find_mac(vxlan, src_mac);
851 if (likely(f)) {
5ca5461c 852 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 853
e4c7ed41 854 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 855 return false;
856
857 /* Don't migrate static entries, drop packets */
eb064c3b 858 if (f->state & NUD_NOARP)
26a41ae6 859 return true;
d342894c 860
861 if (net_ratelimit())
862 netdev_info(dev,
e4c7ed41 863 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 864 src_mac, &rdst->remote_ip, &src_ip);
d342894c 865
e4c7ed41 866 rdst->remote_ip = *src_ip;
d342894c 867 f->updated = jiffies;
8385f50a 868 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 869 } else {
870 /* learned new entry */
871 spin_lock(&vxlan->hash_lock);
3bf74b1a 872
873 /* close off race between vxlan_flush and incoming packets */
874 if (netif_running(dev))
875 vxlan_fdb_create(vxlan, src_mac, src_ip,
876 NUD_REACHABLE,
877 NLM_F_EXCL|NLM_F_CREATE,
878 vxlan->dst_port,
879 vxlan->default_dst.remote_vni,
880 0, NTF_SELF);
d342894c 881 spin_unlock(&vxlan->hash_lock);
882 }
26a41ae6 883
884 return false;
d342894c 885}
886
d342894c 887/* See if multicast group is already in use by other ID */
e4c7ed41 888static bool vxlan_group_used(struct vxlan_net *vn, union vxlan_addr *remote_ip)
d342894c 889{
553675fb 890 struct vxlan_dev *vxlan;
d342894c 891
553675fb 892 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
553675fb 893 if (!netif_running(vxlan->dev))
894 continue;
d342894c 895
e4c7ed41
CW		 896 if (vxlan_addr_equal(&vxlan->default_dst.remote_ip,
897 remote_ip))
553675fb 898 return true;
899 }
d342894c 900
901 return false;
902}
903
7c47cedf 904static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 905{
7c47cedf
SH		 906 atomic_inc(&vs->refcnt);
907}
d342894c 908
012a5729 909void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 910{
012a5729
PS		 911 struct vxlan_net *vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
912
7c47cedf
SH		 913 if (!atomic_dec_and_test(&vs->refcnt))
914 return;
d342894c 915
1c51a915 916 spin_lock(&vn->sock_lock);
7c47cedf 917 hlist_del_rcu(&vs->hlist);
1c51a915
SH		 918 spin_unlock(&vn->sock_lock);
919
7c47cedf 920 queue_work(vxlan_wq, &vs->del_work);
d342894c 921}
012a5729 922EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 923
3fc2de2f 924/* Callback to update multicast group membership when first VNI on
925 * multicast asddress is brought up
926 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 927 */
3fc2de2f 928static void vxlan_igmp_join(struct work_struct *work)
d342894c 929{
3fc2de2f 930 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 931 struct vxlan_sock *vs = vxlan->vn_sock;
932 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 933 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
934 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 935
d342894c 936 lock_sock(sk);
e4c7ed41
CW		 937 if (ip->sa.sa_family == AF_INET) {
938 struct ip_mreqn mreq = {
939 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
940 .imr_ifindex = ifindex,
941 };
942
943 ip_mc_join_group(sk, &mreq);
944#if IS_ENABLED(CONFIG_IPV6)
945 } else {
946 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
947 &ip->sin6.sin6_addr);
948#endif
949 }
3fc2de2f 950 release_sock(sk);
951
012a5729 952 vxlan_sock_release(vs);
3fc2de2f 953 dev_put(vxlan->dev);
954}
955
956/* Inverse of vxlan_igmp_join when last VNI is brought down */
957static void vxlan_igmp_leave(struct work_struct *work)
958{
959 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 960 struct vxlan_sock *vs = vxlan->vn_sock;
961 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 962 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
963 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 964
965 lock_sock(sk);
e4c7ed41
CW		 966 if (ip->sa.sa_family == AF_INET) {
967 struct ip_mreqn mreq = {
968 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
969 .imr_ifindex = ifindex,
970 };
971
972 ip_mc_leave_group(sk, &mreq);
973#if IS_ENABLED(CONFIG_IPV6)
974 } else {
975 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
976 &ip->sin6.sin6_addr);
977#endif
978 }
979
d342894c 980 release_sock(sk);
d342894c 981
012a5729 982 vxlan_sock_release(vs);
7c47cedf 983 dev_put(vxlan->dev);
d342894c 984}
985
986/* Callback from net/ipv4/udp.c to receive packets */
987static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
988{
5cfccc5a 989 struct vxlan_sock *vs;
d342894c 990 struct vxlanhdr *vxh;
553675fb 991 __be16 port;
d342894c 992
d342894c 993 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 994 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 995 goto error;
996
7ce04758
PS		 997 /* Return packets with reserved bits set */
998 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 999 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1000 (vxh->vx_vni & htonl(0xff))) {
1001 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1002 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1003 goto error;
1004 }
1005
5cfccc5a
PS		 1006 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1007 goto drop;
1008
553675fb 1009 port = inet_sk(sk)->inet_sport;
5cfccc5a
PS		 1010
1011 vs = vxlan_find_sock(sock_net(sk), port);
1012 if (!vs)
d342894c 1013 goto drop;
d342894c 1014
5cfccc5a
PS		 1015 vs->rcv(vs, skb, vxh->vx_vni);
1016 return 0;
1017
1018drop:
1019 /* Consume bad packet */
1020 kfree_skb(skb);
1021 return 0;
1022
1023error:
1024 /* Return non vxlan pkt */
1025 return 1;
1026}
1027
1028static void vxlan_rcv(struct vxlan_sock *vs,
1029 struct sk_buff *skb, __be32 vx_vni)
1030{
e4c7ed41
CW		 1031 struct iphdr *oip = NULL;
1032 struct ipv6hdr *oip6 = NULL;
5cfccc5a
PS		 1033 struct vxlan_dev *vxlan;
1034 struct pcpu_tstats *stats;
e4c7ed41 1035 union vxlan_addr saddr;
5cfccc5a 1036 __u32 vni;
e4c7ed41
CW		 1037 int err = 0;
1038 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1039
1040 vni = ntohl(vx_vni) >> 8;
1041 /* Is this VNI defined? */
1042 vxlan = vxlan_vs_find_vni(vs, vni);
1043 if (!vxlan)
d342894c 1044 goto drop;
d342894c 1045
e4c7ed41 1046 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1047 skb_reset_mac_header(skb);
d342894c 1048 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 1049
1050 /* Ignore packet loops (and multicast echo) */
1051 if (compare_ether_addr(eth_hdr(skb)->h_source,
1052 vxlan->dev->dev_addr) == 0)
1053 goto drop;
1054
7ce04758 1055 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1056 if (remote_ip->sa.sa_family == AF_INET) {
1057 oip = ip_hdr(skb);
1058 saddr.sin.sin_addr.s_addr = oip->saddr;
1059 saddr.sa.sa_family = AF_INET;
1060#if IS_ENABLED(CONFIG_IPV6)
1061 } else {
1062 oip6 = ipv6_hdr(skb);
1063 saddr.sin6.sin6_addr = oip6->saddr;
1064 saddr.sa.sa_family = AF_INET6;
1065#endif
1066 }
1067
26a41ae6 1068 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1069 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1070 goto drop;
d342894c 1071
d342894c 1072 skb_reset_network_header(skb);
0afb1666
JG		 1073
1074 /* If the NIC driver gave us an encapsulated packet with
1075 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1076 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1077 * for us. Otherwise force the upper layers to verify it.
1078 */
1079 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1080 !(vxlan->dev->features & NETIF_F_RXCSUM))
1081 skb->ip_summed = CHECKSUM_NONE;
1082
1083 skb->encapsulation = 0;
d342894c 1084
e4c7ed41
CW		 1085 if (oip6)
1086 err = IP6_ECN_decapsulate(oip6, skb);
1087 if (oip)
1088 err = IP_ECN_decapsulate(oip, skb);
1089
d342894c 1090 if (unlikely(err)) {
e4c7ed41
CW		 1091 if (log_ecn_error) {
1092 if (oip6)
1093 net_info_ratelimited("non-ECT from %pI6\n",
1094 &oip6->saddr);
1095 if (oip)
1096 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1097 &oip->saddr, oip->tos);
1098 }
d342894c 1099 if (err > 1) {
1100 ++vxlan->dev->stats.rx_frame_errors;
1101 ++vxlan->dev->stats.rx_errors;
1102 goto drop;
1103 }
1104 }
1105
e8171045 1106 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1107 u64_stats_update_begin(&stats->syncp);
1108 stats->rx_packets++;
1109 stats->rx_bytes += skb->len;
1110 u64_stats_update_end(&stats->syncp);
1111
1112 netif_rx(skb);
1113
5cfccc5a 1114 return;
d342894c 1115drop:
1116 /* Consume bad packet */
1117 kfree_skb(skb);
d342894c 1118}
1119
e4f67add
DS		 1120static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1121{
1122 struct vxlan_dev *vxlan = netdev_priv(dev);
1123 struct arphdr *parp;
1124 u8 *arpptr, *sha;
1125 __be32 sip, tip;
1126 struct neighbour *n;
1127
1128 if (dev->flags & IFF_NOARP)
1129 goto out;
1130
1131 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1132 dev->stats.tx_dropped++;
1133 goto out;
1134 }
1135 parp = arp_hdr(skb);
1136
1137 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1138 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1139 parp->ar_pro != htons(ETH_P_IP) ||
1140 parp->ar_op != htons(ARPOP_REQUEST) ||
1141 parp->ar_hln != dev->addr_len ||
1142 parp->ar_pln != 4)
1143 goto out;
1144 arpptr = (u8 *)parp + sizeof(struct arphdr);
1145 sha = arpptr;
1146 arpptr += dev->addr_len; /* sha */
1147 memcpy(&sip, arpptr, sizeof(sip));
1148 arpptr += sizeof(sip);
1149 arpptr += dev->addr_len; /* tha */
1150 memcpy(&tip, arpptr, sizeof(tip));
1151
1152 if (ipv4_is_loopback(tip) ||
1153 ipv4_is_multicast(tip))
1154 goto out;
1155
1156 n = neigh_lookup(&arp_tbl, &tip, dev);
1157
1158 if (n) {
e4f67add
DS		 1159 struct vxlan_fdb *f;
1160 struct sk_buff *reply;
1161
1162 if (!(n->nud_state & NUD_CONNECTED)) {
1163 neigh_release(n);
1164 goto out;
1165 }
1166
1167 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1168 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1169 /* bridge-local neighbor */
1170 neigh_release(n);
1171 goto out;
1172 }
1173
1174 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1175 n->ha, sha);
1176
1177 neigh_release(n);
1178
1179 skb_reset_mac_header(reply);
1180 __skb_pull(reply, skb_network_offset(reply));
1181 reply->ip_summed = CHECKSUM_UNNECESSARY;
1182 reply->pkt_type = PACKET_HOST;
1183
1184 if (netif_rx_ni(reply) == NET_RX_DROP)
1185 dev->stats.rx_dropped++;
e4c7ed41
CW		 1186 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1187 union vxlan_addr ipa = {
1188 .sin.sin_addr.s_addr = tip,
1189 .sa.sa_family = AF_INET,
1190 };
1191
1192 vxlan_ip_miss(dev, &ipa);
1193 }
e4f67add
DS		 1194out:
1195 consume_skb(skb);
1196 return NETDEV_TX_OK;
1197}
1198
f564f45c
CW		 1199#if IS_ENABLED(CONFIG_IPV6)
1200static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1201{
1202 struct vxlan_dev *vxlan = netdev_priv(dev);
1203 struct neighbour *n;
1204 union vxlan_addr ipa;
1205 const struct ipv6hdr *iphdr;
1206 const struct in6_addr *saddr, *daddr;
1207 struct nd_msg *msg;
1208 struct inet6_dev *in6_dev = NULL;
1209
1210 in6_dev = __in6_dev_get(dev);
1211 if (!in6_dev)
1212 goto out;
1213
1214 if (!pskb_may_pull(skb, skb->len))
1215 goto out;
1216
1217 iphdr = ipv6_hdr(skb);
1218 saddr = &iphdr->saddr;
1219 daddr = &iphdr->daddr;
1220
1221 if (ipv6_addr_loopback(daddr) ||
1222 ipv6_addr_is_multicast(daddr))
1223 goto out;
1224
1225 msg = (struct nd_msg *)skb_transport_header(skb);
1226 if (msg->icmph.icmp6_code != 0 ||
1227 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1228 goto out;
1229
1230 n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1231
1232 if (n) {
1233 struct vxlan_fdb *f;
1234
1235 if (!(n->nud_state & NUD_CONNECTED)) {
1236 neigh_release(n);
1237 goto out;
1238 }
1239
1240 f = vxlan_find_mac(vxlan, n->ha);
1241 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1242 /* bridge-local neighbor */
1243 neigh_release(n);
1244 goto out;
1245 }
1246
1247 ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1248 !!in6_dev->cnf.forwarding,
1249 true, false, false);
1250 neigh_release(n);
1251 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1252 ipa.sin6.sin6_addr = *daddr;
1253 ipa.sa.sa_family = AF_INET6;
1254 vxlan_ip_miss(dev, &ipa);
1255 }
1256
1257out:
1258 consume_skb(skb);
1259 return NETDEV_TX_OK;
1260}
1261#endif
1262
e4f67add
DS		 1263static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1264{
1265 struct vxlan_dev *vxlan = netdev_priv(dev);
1266 struct neighbour *n;
e4f67add
DS		 1267
1268 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1269 return false;
1270
1271 n = NULL;
1272 switch (ntohs(eth_hdr(skb)->h_proto)) {
1273 case ETH_P_IP:
e15a00aa
CW		 1274 {
1275 struct iphdr *pip;
1276
e4f67add
DS		 1277 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1278 return false;
1279 pip = ip_hdr(skb);
1280 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1281 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1282 union vxlan_addr ipa = {
1283 .sin.sin_addr.s_addr = pip->daddr,
1284 .sa.sa_family = AF_INET,
1285 };
1286
1287 vxlan_ip_miss(dev, &ipa);
1288 return false;
1289 }
1290
e4f67add 1291 break;
e15a00aa
CW		 1292 }
1293#if IS_ENABLED(CONFIG_IPV6)
1294 case ETH_P_IPV6:
1295 {
1296 struct ipv6hdr *pip6;
1297
1298 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1299 return false;
1300 pip6 = ipv6_hdr(skb);
1301 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1302 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1303 union vxlan_addr ipa = {
1304 .sin6.sin6_addr = pip6->daddr,
1305 .sa.sa_family = AF_INET6,
1306 };
1307
1308 vxlan_ip_miss(dev, &ipa);
1309 return false;
1310 }
1311
1312 break;
1313 }
1314#endif
e4f67add
DS		 1315 default:
1316 return false;
1317 }
1318
1319 if (n) {
1320 bool diff;
1321
1322 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
1323 if (diff) {
1324 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1325 dev->addr_len);
1326 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1327 }
1328 neigh_release(n);
1329 return diff;
e4c7ed41
CW		 1330 }
1331
e4f67add
DS		 1332 return false;
1333}
1334
553675fb 1335static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1336{
1337 sock_put(skb->sk);
1338}
1339
1340/* On transmit, associate with the tunnel socket */
49560532 1341static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1cad8715 1342{
1cad8715 1343 skb_orphan(skb);
1344 sock_hold(sk);
1345 skb->sk = sk;
553675fb 1346 skb->destructor = vxlan_sock_put;
1cad8715 1347}
1348
05f47d69 1349/* Compute source port for outgoing packet
1350 * first choice to use L4 flow hash since it will spread
1351 * better and maybe available from hardware
1352 * secondary choice is to use jhash on the Ethernet header
1353 */
49560532 1354__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
05f47d69 1355{
49560532 1356 unsigned int range = (port_max - port_min) + 1;
05f47d69 1357 u32 hash;
1358
1359 hash = skb_get_rxhash(skb);
1360 if (!hash)
1361 hash = jhash(skb->data, 2 * ETH_ALEN,
1362 (__force u32) skb->protocol);
1363
49560532 1364 return htons((((u64) hash * range) >> 32) + port_min);
05f47d69 1365}
49560532 1366EXPORT_SYMBOL_GPL(vxlan_src_port);
05f47d69 1367
05c0db08
PS		 1368static int handle_offloads(struct sk_buff *skb)
1369{
1370 if (skb_is_gso(skb)) {
1371 int err = skb_unclone(skb, GFP_ATOMIC);
1372 if (unlikely(err))
1373 return err;
1374
f6ace502 1375 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1376 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1377 skb->ip_summed = CHECKSUM_NONE;
1378
1379 return 0;
1380}
1381
e4c7ed41
CW		 1382#if IS_ENABLED(CONFIG_IPV6)
1383static int vxlan6_xmit_skb(struct net *net, struct vxlan_sock *vs,
1384 struct dst_entry *dst, struct sk_buff *skb,
1385 struct net_device *dev, struct in6_addr *saddr,
1386 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1387 __be16 src_port, __be16 dst_port, __be32 vni)
1388{
1389 struct ipv6hdr *ip6h;
1390 struct vxlanhdr *vxh;
1391 struct udphdr *uh;
1392 int min_headroom;
1393 int err;
1394
1395 if (!skb->encapsulation) {
1396 skb_reset_inner_headers(skb);
1397 skb->encapsulation = 1;
1398 }
1399
1400 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1401 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1402 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1403
1404 /* Need space for new headers (invalidates iph ptr) */
1405 err = skb_cow_head(skb, min_headroom);
1406 if (unlikely(err))
1407 return err;
1408
1409 if (vlan_tx_tag_present(skb)) {
1410 if (WARN_ON(!__vlan_put_tag(skb,
1411 skb->vlan_proto,
1412 vlan_tx_tag_get(skb))))
1413 return -ENOMEM;
1414
1415 skb->vlan_tci = 0;
1416 }
1417
1418 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1419 vxh->vx_flags = htonl(VXLAN_FLAGS);
1420 vxh->vx_vni = vni;
1421
1422 __skb_push(skb, sizeof(*uh));
1423 skb_reset_transport_header(skb);
1424 uh = udp_hdr(skb);
1425
1426 uh->dest = dst_port;
1427 uh->source = src_port;
1428
1429 uh->len = htons(skb->len);
1430 uh->check = 0;
1431
1432 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1433 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1434 IPSKB_REROUTED);
1435 skb_dst_drop(skb);
1436 skb_dst_set(skb, dst);
1437
1438 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1439 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1440 skb->ip_summed = CHECKSUM_UNNECESSARY;
1441 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1442 IPPROTO_UDP, csum);
1443 if (uh->check == 0)
1444 uh->check = CSUM_MANGLED_0;
1445 } else {
1446 skb->ip_summed = CHECKSUM_PARTIAL;
1447 skb->csum_start = skb_transport_header(skb) - skb->head;
1448 skb->csum_offset = offsetof(struct udphdr, check);
1449 uh->check = ~csum_ipv6_magic(saddr, daddr,
1450 skb->len, IPPROTO_UDP, 0);
1451 }
1452
1453 __skb_push(skb, sizeof(*ip6h));
1454 skb_reset_network_header(skb);
1455 ip6h = ipv6_hdr(skb);
1456 ip6h->version = 6;
1457 ip6h->priority = prio;
1458 ip6h->flow_lbl[0] = 0;
1459 ip6h->flow_lbl[1] = 0;
1460 ip6h->flow_lbl[2] = 0;
1461 ip6h->payload_len = htons(skb->len);
1462 ip6h->nexthdr = IPPROTO_UDP;
1463 ip6h->hop_limit = ttl;
1464 ip6h->daddr = *daddr;
1465 ip6h->saddr = *saddr;
1466
1467 vxlan_set_owner(vs->sock->sk, skb);
1468
1469 err = handle_offloads(skb);
1470 if (err)
1471 return err;
1472
1473 ip6tunnel_xmit(skb, dev);
1474 return 0;
1475}
1476#endif
1477
49560532
PS		 1478int vxlan_xmit_skb(struct net *net, struct vxlan_sock *vs,
1479 struct rtable *rt, struct sk_buff *skb,
1480 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1481 __be16 src_port, __be16 dst_port, __be32 vni)
1482{
1483 struct vxlanhdr *vxh;
1484 struct udphdr *uh;
649c5b8b 1485 int min_headroom;
49560532
PS		 1486 int err;
1487
1488 if (!skb->encapsulation) {
1489 skb_reset_inner_headers(skb);
1490 skb->encapsulation = 1;
1491 }
1492
649c5b8b 1493 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1494 + VXLAN_HLEN + sizeof(struct iphdr)
1495 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1496
1497 /* Need space for new headers (invalidates iph ptr) */
1498 err = skb_cow_head(skb, min_headroom);
1499 if (unlikely(err))
1500 return err;
1501
1eaa8178
PS		 1502 if (vlan_tx_tag_present(skb)) {
1503 if (WARN_ON(!__vlan_put_tag(skb,
1504 skb->vlan_proto,
1505 vlan_tx_tag_get(skb))))
1506 return -ENOMEM;
1507
1508 skb->vlan_tci = 0;
1509 }
1510
49560532
PS		 1511 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1512 vxh->vx_flags = htonl(VXLAN_FLAGS);
1513 vxh->vx_vni = vni;
1514
1515 __skb_push(skb, sizeof(*uh));
1516 skb_reset_transport_header(skb);
1517 uh = udp_hdr(skb);
1518
1519 uh->dest = dst_port;
1520 uh->source = src_port;
1521
1522 uh->len = htons(skb->len);
1523 uh->check = 0;
1524
1525 vxlan_set_owner(vs->sock->sk, skb);
1526
1527 err = handle_offloads(skb);
1528 if (err)
1529 return err;
1530
1531 return iptunnel_xmit(net, rt, skb, src, dst,
1532 IPPROTO_UDP, tos, ttl, df);
1533}
1534EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1535
9dcc71e1
SS		 1536/* Bypass encapsulation if the destination is local */
1537static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1538 struct vxlan_dev *dst_vxlan)
1539{
1540 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1541 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
e4c7ed41
CW		 1542 union vxlan_addr loopback;
1543 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1
SS		 1544
1545 skb->pkt_type = PACKET_HOST;
1546 skb->encapsulation = 0;
1547 skb->dev = dst_vxlan->dev;
1548 __skb_pull(skb, skb_network_offset(skb));
1549
e4c7ed41
CW		 1550 if (remote_ip->sa.sa_family == AF_INET) {
1551 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1552 loopback.sa.sa_family = AF_INET;
1553#if IS_ENABLED(CONFIG_IPV6)
1554 } else {
1555 loopback.sin6.sin6_addr = in6addr_loopback;
1556 loopback.sa.sa_family = AF_INET6;
1557#endif
1558 }
1559
9dcc71e1 1560 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1561 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1562
1563 u64_stats_update_begin(&tx_stats->syncp);
1564 tx_stats->tx_packets++;
1565 tx_stats->tx_bytes += skb->len;
1566 u64_stats_update_end(&tx_stats->syncp);
1567
1568 if (netif_rx(skb) == NET_RX_SUCCESS) {
1569 u64_stats_update_begin(&rx_stats->syncp);
1570 rx_stats->rx_packets++;
1571 rx_stats->rx_bytes += skb->len;
1572 u64_stats_update_end(&rx_stats->syncp);
1573 } else {
1574 skb->dev->stats.rx_dropped++;
1575 }
1576}
1577
4ad16930
SH		 1578static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1579 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1580{
1581 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1582 struct rtable *rt = NULL;
d342894c 1583 const struct iphdr *old_iph;
d342894c 1584 struct flowi4 fl4;
e4c7ed41
CW		 1585 union vxlan_addr *dst;
1586 __be16 src_port = 0, dst_port;
234f5b73 1587 u32 vni;
d342894c 1588 __be16 df = 0;
1589 __u8 tos, ttl;
0e6fbc5b 1590 int err;
d342894c 1591
823aa873 1592 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1593 vni = rdst->remote_vni;
e4c7ed41 1594 dst = &rdst->remote_ip;
e4f67add 1595
e4c7ed41 1596 if (vxlan_addr_any(dst)) {
e4f67add 1597 if (did_rsc) {
e4f67add 1598 /* short-circuited back to local bridge */
9dcc71e1 1599 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1600 return;
e4f67add 1601 }
ef59febe 1602 goto drop;
e4f67add 1603 }
ef59febe 1604
d342894c 1605 old_iph = ip_hdr(skb);
1606
d342894c 1607 ttl = vxlan->ttl;
e4c7ed41 1608 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1609 ttl = 1;
1610
1611 tos = vxlan->tos;
1612 if (tos == 1)
206aaafc 1613 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1614
49560532 1615 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
d342894c 1616
e4c7ed41
CW		 1617 if (dst->sa.sa_family == AF_INET) {
1618 memset(&fl4, 0, sizeof(fl4));
1619 fl4.flowi4_oif = rdst->remote_ifindex;
1620 fl4.flowi4_tos = RT_TOS(tos);
1621 fl4.daddr = dst->sin.sin_addr.s_addr;
1622 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1623
1624 rt = ip_route_output_key(dev_net(dev), &fl4);
1625 if (IS_ERR(rt)) {
1626 netdev_dbg(dev, "no route to %pI4\n",
1627 &dst->sin.sin_addr.s_addr);
1628 dev->stats.tx_carrier_errors++;
1629 goto tx_error;
1630 }
d342894c 1631
e4c7ed41
CW		 1632 if (rt->dst.dev == dev) {
1633 netdev_dbg(dev, "circular route to %pI4\n",
1634 &dst->sin.sin_addr.s_addr);
1635 dev->stats.collisions++;
1636 goto tx_error;
1637 }
1638
1639 /* Bypass encapsulation if the destination is local */
1640 if (rt->rt_flags & RTCF_LOCAL &&
1641 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1642 struct vxlan_dev *dst_vxlan;
1643
1644 ip_rt_put(rt);
1645 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1646 if (!dst_vxlan)
1647 goto tx_error;
1648 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1649 return;
1650 }
1651
1652 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1653 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1654
e4c7ed41
CW		 1655 err = vxlan_xmit_skb(dev_net(dev), vxlan->vn_sock, rt, skb,
1656 fl4.saddr, dst->sin.sin_addr.s_addr,
1657 tos, ttl, df, src_port, dst_port,
1658 htonl(vni << 8));
9dcc71e1 1659
e4c7ed41
CW		 1660 if (err < 0)
1661 goto rt_tx_error;
1662 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1663#if IS_ENABLED(CONFIG_IPV6)
1664 } else {
1665 struct sock *sk = vxlan->vn_sock->sock->sk;
1666 struct dst_entry *ndst;
1667 struct flowi6 fl6;
1668 u32 flags;
1669
1670 memset(&fl6, 0, sizeof(fl6));
1671 fl6.flowi6_oif = rdst->remote_ifindex;
1672 fl6.daddr = dst->sin6.sin6_addr;
1673 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1674 fl6.flowi6_proto = skb->protocol;
1675
1676 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1677 netdev_dbg(dev, "no route to %pI6\n",
1678 &dst->sin6.sin6_addr);
1679 dev->stats.tx_carrier_errors++;
9dcc71e1 1680 goto tx_error;
e4c7ed41 1681 }
d342894c 1682
e4c7ed41
CW		 1683 if (ndst->dev == dev) {
1684 netdev_dbg(dev, "circular route to %pI6\n",
1685 &dst->sin6.sin6_addr);
1686 dst_release(ndst);
1687 dev->stats.collisions++;
1688 goto tx_error;
1689 }
0e6fbc5b 1690
e4c7ed41
CW		 1691 /* Bypass encapsulation if the destination is local */
1692 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1693 if (flags & RTF_LOCAL &&
1694 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1695 struct vxlan_dev *dst_vxlan;
1696
1697 dst_release(ndst);
1698 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1699 if (!dst_vxlan)
1700 goto tx_error;
1701 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1702 return;
1703 }
49560532 1704
e4c7ed41
CW		 1705 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1706
1707 err = vxlan6_xmit_skb(dev_net(dev), vxlan->vn_sock, ndst, skb,
1708 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1709 src_port, dst_port, htonl(vni << 8));
1710#endif
1711 }
0e6fbc5b 1712
4ad16930 1713 return;
d342894c 1714
1715drop:
1716 dev->stats.tx_dropped++;
1717 goto tx_free;
1718
49560532
PS		 1719rt_tx_error:
1720 ip_rt_put(rt);
d342894c 1721tx_error:
1722 dev->stats.tx_errors++;
1723tx_free:
1724 dev_kfree_skb(skb);
d342894c 1725}
1726
6681712d
DS		 1727/* Transmit local packets over Vxlan
1728 *
1729 * Outer IP header inherits ECN and DF from inner header.
1730 * Outer UDP destination is the VXLAN assigned port.
1731 * source port is based on hash of flow
1732 */
1733static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1734{
1735 struct vxlan_dev *vxlan = netdev_priv(dev);
1736 struct ethhdr *eth;
1737 bool did_rsc = false;
afbd8bae 1738 struct vxlan_rdst *rdst;
6681712d 1739 struct vxlan_fdb *f;
6681712d
DS		 1740
1741 skb_reset_mac_header(skb);
1742 eth = eth_hdr(skb);
1743
f564f45c
CW		 1744 if ((vxlan->flags & VXLAN_F_PROXY)) {
1745 if (ntohs(eth->h_proto) == ETH_P_ARP)
1746 return arp_reduce(dev, skb);
1747#if IS_ENABLED(CONFIG_IPV6)
1748 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1749 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1750 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1751 struct nd_msg *msg;
1752
1753 msg = (struct nd_msg *)skb_transport_header(skb);
1754 if (msg->icmph.icmp6_code == 0 &&
1755 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1756 return neigh_reduce(dev, skb);
1757 }
1758#endif
1759 }
6681712d
DS		 1760
1761 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1762 did_rsc = false;
1763
1764 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1765 (ntohs(eth->h_proto) == ETH_P_IP ||
1766 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1767 did_rsc = route_shortcircuit(dev, skb);
1768 if (did_rsc)
1769 f = vxlan_find_mac(vxlan, eth->h_dest);
1770 }
1771
6681712d 1772 if (f == NULL) {
afbd8bae
MR		 1773 f = vxlan_find_mac(vxlan, all_zeros_mac);
1774 if (f == NULL) {
1775 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1776 !is_multicast_ether_addr(eth->h_dest))
1777 vxlan_fdb_miss(vxlan, eth->h_dest);
1778
1779 dev->stats.tx_dropped++;
1780 dev_kfree_skb(skb);
1781 return NETDEV_TX_OK;
1782 }
1783 }
6681712d 1784
afbd8bae
MR		 1785 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1786 struct sk_buff *skb1;
6681712d 1787
afbd8bae
MR		 1788 skb1 = skb_clone(skb, GFP_ATOMIC);
1789 if (skb1)
1790 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1791 }
1792
afbd8bae 1793 dev_kfree_skb(skb);
4ad16930 1794 return NETDEV_TX_OK;
6681712d
DS		 1795}
1796
d342894c 1797/* Walk the forwarding table and purge stale entries */
1798static void vxlan_cleanup(unsigned long arg)
1799{
1800 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1801 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1802 unsigned int h;
1803
1804 if (!netif_running(vxlan->dev))
1805 return;
1806
1807 spin_lock_bh(&vxlan->hash_lock);
1808 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1809 struct hlist_node *p, *n;
1810 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1811 struct vxlan_fdb *f
1812 = container_of(p, struct vxlan_fdb, hlist);
1813 unsigned long timeout;
1814
3c172868 1815 if (f->state & NUD_PERMANENT)
d342894c 1816 continue;
1817
1818 timeout = f->used + vxlan->age_interval * HZ;
1819 if (time_before_eq(timeout, jiffies)) {
1820 netdev_dbg(vxlan->dev,
1821 "garbage collect %pM\n",
1822 f->eth_addr);
1823 f->state = NUD_STALE;
1824 vxlan_fdb_destroy(vxlan, f);
1825 } else if (time_before(timeout, next_timer))
1826 next_timer = timeout;
1827 }
1828 }
1829 spin_unlock_bh(&vxlan->hash_lock);
1830
1831 mod_timer(&vxlan->age_timer, next_timer);
1832}
1833
9c2e24e1
PS		 1834static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1835{
1836 __u32 vni = vxlan->default_dst.remote_vni;
1837
1838 vxlan->vn_sock = vs;
1839 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1840}
1841
d342894c 1842/* Setup stats when device is created */
1843static int vxlan_init(struct net_device *dev)
1844{
1c51a915
SH		 1845 struct vxlan_dev *vxlan = netdev_priv(dev);
1846 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1847 struct vxlan_sock *vs;
1c51a915 1848
e8171045
PS		 1849 dev->tstats = alloc_percpu(struct pcpu_tstats);
1850 if (!dev->tstats)
d342894c 1851 return -ENOMEM;
1852
1c51a915 1853 spin_lock(&vn->sock_lock);
9c2e24e1 1854 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1c51a915
SH		 1855 if (vs) {
1856 /* If we have a socket with same port already, reuse it */
1857 atomic_inc(&vs->refcnt);
9c2e24e1 1858 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1859 } else {
1860 /* otherwise make new socket outside of RTNL */
1861 dev_hold(dev);
1862 queue_work(vxlan_wq, &vxlan->sock_work);
1863 }
1864 spin_unlock(&vn->sock_lock);
1865
d342894c 1866 return 0;
1867}
1868
ba609e9b 1869static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1870{
1871 struct vxlan_fdb *f;
1872
1873 spin_lock_bh(&vxlan->hash_lock);
1874 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1875 if (f)
1876 vxlan_fdb_destroy(vxlan, f);
1877 spin_unlock_bh(&vxlan->hash_lock);
1878}
1879
ebf4063e
SH		 1880static void vxlan_uninit(struct net_device *dev)
1881{
1882 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 1883 struct vxlan_sock *vs = vxlan->vn_sock;
1884
ba609e9b 1885 vxlan_fdb_delete_default(vxlan);
afbd8bae 1886
ebf4063e 1887 if (vs)
012a5729 1888 vxlan_sock_release(vs);
ebf4063e
SH		 1889 free_percpu(dev->tstats);
1890}
1891
d342894c 1892/* Start ageing timer and join group when device is brought up */
1893static int vxlan_open(struct net_device *dev)
1894{
3fc2de2f 1895 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1896 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1897 struct vxlan_sock *vs = vxlan->vn_sock;
1898
1899 /* socket hasn't been created */
1900 if (!vs)
1901 return -ENOTCONN;
d342894c 1902
e4c7ed41
CW		 1903 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1904 vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1905 vxlan_sock_hold(vs);
7c47cedf 1906 dev_hold(dev);
3fc2de2f 1907 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 1908 }
1909
1910 if (vxlan->age_interval)
1911 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1912
1913 return 0;
1914}
1915
1916/* Purge the forwarding table */
1917static void vxlan_flush(struct vxlan_dev *vxlan)
1918{
31fec5aa 1919 unsigned int h;
d342894c 1920
1921 spin_lock_bh(&vxlan->hash_lock);
1922 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1923 struct hlist_node *p, *n;
1924 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1925 struct vxlan_fdb *f
1926 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 1927 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1928 if (!is_zero_ether_addr(f->eth_addr))
1929 vxlan_fdb_destroy(vxlan, f);
d342894c 1930 }
1931 }
1932 spin_unlock_bh(&vxlan->hash_lock);
1933}
1934
1935/* Cleanup timer and forwarding table on shutdown */
1936static int vxlan_stop(struct net_device *dev)
1937{
3fc2de2f 1938 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1939 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1940 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1941
e4c7ed41
CW		 1942 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1943 ! vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1944 vxlan_sock_hold(vs);
7c47cedf 1945 dev_hold(dev);
3fc2de2f 1946 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 1947 }
d342894c 1948
1949 del_timer_sync(&vxlan->age_timer);
1950
1951 vxlan_flush(vxlan);
1952
1953 return 0;
1954}
1955
d342894c 1956/* Stub, nothing needs to be done. */
1957static void vxlan_set_multicast_list(struct net_device *dev)
1958{
1959}
1960
1961static const struct net_device_ops vxlan_netdev_ops = {
1962 .ndo_init = vxlan_init,
ebf4063e 1963 .ndo_uninit = vxlan_uninit,
d342894c 1964 .ndo_open = vxlan_open,
1965 .ndo_stop = vxlan_stop,
1966 .ndo_start_xmit = vxlan_xmit,
e8171045 1967 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 1968 .ndo_set_rx_mode = vxlan_set_multicast_list,
1969 .ndo_change_mtu = eth_change_mtu,
1970 .ndo_validate_addr = eth_validate_addr,
1971 .ndo_set_mac_address = eth_mac_addr,
1972 .ndo_fdb_add = vxlan_fdb_add,
1973 .ndo_fdb_del = vxlan_fdb_delete,
1974 .ndo_fdb_dump = vxlan_fdb_dump,
1975};
1976
1977/* Info for udev, that this is a virtual tunnel endpoint */
1978static struct device_type vxlan_type = {
1979 .name = "vxlan",
1980};
1981
d342894c 1982/* Initialize the device structure. */
1983static void vxlan_setup(struct net_device *dev)
1984{
1985 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 1986 unsigned int h;
05f47d69 1987 int low, high;
d342894c 1988
1989 eth_hw_addr_random(dev);
1990 ether_setup(dev);
e4c7ed41
CW		 1991 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
1992 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
1993 else
1994 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1995
1996 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 1997 dev->destructor = free_netdev;
d342894c 1998 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1999
2000 dev->tx_queue_len = 0;
2001 dev->features |= NETIF_F_LLTX;
2002 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 2003 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2004 dev->features |= NETIF_F_RXCSUM;
05c0db08 2005 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2006
1eaa8178
PS		 2007 dev->vlan_features = dev->features;
2008 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2009 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2010 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2011 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2012 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2013 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2014
553675fb 2015 INIT_LIST_HEAD(&vxlan->next);
d342894c 2016 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2017 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2018 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2019 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2020
2021 init_timer_deferrable(&vxlan->age_timer);
2022 vxlan->age_timer.function = vxlan_cleanup;
2023 vxlan->age_timer.data = (unsigned long) vxlan;
2024
05f47d69 2025 inet_get_local_port_range(&low, &high);
2026 vxlan->port_min = low;
2027 vxlan->port_max = high;
823aa873 2028 vxlan->dst_port = htons(vxlan_port);
05f47d69 2029
d342894c 2030 vxlan->dev = dev;
2031
2032 for (h = 0; h < FDB_HASH_SIZE; ++h)
2033 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2034}
2035
2036static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2037 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2038 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2039 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2040 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2041 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2042 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2043 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2044 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2045 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2046 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2047 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2048 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2049 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2050 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2051 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2052 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2053 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2054};
2055
2056static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2057{
2058 if (tb[IFLA_ADDRESS]) {
2059 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2060 pr_debug("invalid link address (not ethernet)\n");
2061 return -EINVAL;
2062 }
2063
2064 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2065 pr_debug("invalid all zero ethernet address\n");
2066 return -EADDRNOTAVAIL;
2067 }
2068 }
2069
2070 if (!data)
2071 return -EINVAL;
2072
2073 if (data[IFLA_VXLAN_ID]) {
2074 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2075 if (id >= VXLAN_VID_MASK)
2076 return -ERANGE;
2077 }
2078
05f47d69 2079 if (data[IFLA_VXLAN_PORT_RANGE]) {
2080 const struct ifla_vxlan_port_range *p
2081 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2082
2083 if (ntohs(p->high) < ntohs(p->low)) {
2084 pr_debug("port range %u .. %u not valid\n",
2085 ntohs(p->low), ntohs(p->high));
2086 return -EINVAL;
2087 }
2088 }
2089
d342894c 2090 return 0;
2091}
2092
1b13c97f
YB		 2093static void vxlan_get_drvinfo(struct net_device *netdev,
2094 struct ethtool_drvinfo *drvinfo)
2095{
2096 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2097 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2098}
2099
2100static const struct ethtool_ops vxlan_ethtool_ops = {
2101 .get_drvinfo = vxlan_get_drvinfo,
2102 .get_link = ethtool_op_get_link,
2103};
2104
553675fb 2105static void vxlan_del_work(struct work_struct *work)
2106{
2107 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2108
2109 sk_release_kernel(vs->sock->sk);
2110 kfree_rcu(vs, rcu);
2111}
2112
e4c7ed41
CW		 2113#if IS_ENABLED(CONFIG_IPV6)
2114/* Create UDP socket for encapsulation receive. AF_INET6 socket
2115 * could be used for both IPv4 and IPv6 communications, but
2116 * users may set bindv6only=1.
2117 */
2118static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
553675fb 2119{
553675fb 2120 struct sock *sk;
e4c7ed41
CW		 2121 struct socket *sock;
2122 struct sockaddr_in6 vxlan_addr = {
2123 .sin6_family = AF_INET6,
2124 .sin6_port = port,
2125 };
2126 int rc, val = 1;
2127
2128 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2129 if (rc < 0) {
2130 pr_debug("UDPv6 socket create failed\n");
2131 return rc;
2132 }
2133
2134 /* Put in proper namespace */
2135 sk = sock->sk;
2136 sk_change_net(sk, net);
2137
2138 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2139 (char *)&val, sizeof(val));
2140 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2141 sizeof(struct sockaddr_in6));
2142 if (rc < 0) {
2143 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2144 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2145 sk_release_kernel(sk);
2146 return rc;
2147 }
2148 /* At this point, IPv6 module should have been loaded in
2149 * sock_create_kern().
2150 */
2151 BUG_ON(!ipv6_stub);
2152
2153 *psock = sock;
2154 /* Disable multicast loopback */
2155 inet_sk(sk)->mc_loop = 0;
2156 return 0;
2157}
2158
2159#else
2160
2161static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2162{
2163 return -EPFNOSUPPORT;
2164}
2165#endif
2166
2167static int create_v4_sock(struct net *net, __be16 port, struct socket **psock)
2168{
2169 struct sock *sk;
2170 struct socket *sock;
553675fb 2171 struct sockaddr_in vxlan_addr = {
2172 .sin_family = AF_INET,
2173 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2174 .sin_port = port,
553675fb 2175 };
2176 int rc;
553675fb 2177
2178 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2179 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2180 if (rc < 0) {
2181 pr_debug("UDP socket create failed\n");
e4c7ed41 2182 return rc;
553675fb 2183 }
2184
2185 /* Put in proper namespace */
e4c7ed41 2186 sk = sock->sk;
553675fb 2187 sk_change_net(sk, net);
2188
e4c7ed41 2189 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2190 sizeof(vxlan_addr));
2191 if (rc < 0) {
2192 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2193 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2194 sk_release_kernel(sk);
e4c7ed41
CW		 2195 return rc;
2196 }
2197
2198 *psock = sock;
2199 /* Disable multicast loopback */
2200 inet_sk(sk)->mc_loop = 0;
2201 return 0;
2202}
2203
2204/* Create new listen socket if needed */
2205static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2206 vxlan_rcv_t *rcv, void *data, bool ipv6)
2207{
2208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2209 struct vxlan_sock *vs;
2210 struct socket *sock;
2211 struct sock *sk;
2212 int rc = 0;
2213 unsigned int h;
2214
2215 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2216 if (!vs)
2217 return ERR_PTR(-ENOMEM);
2218
2219 for (h = 0; h < VNI_HASH_SIZE; ++h)
2220 INIT_HLIST_HEAD(&vs->vni_list[h]);
2221
2222 INIT_WORK(&vs->del_work, vxlan_del_work);
2223
2224 if (ipv6)
2225 rc = create_v6_sock(net, port, &sock);
2226 else
2227 rc = create_v4_sock(net, port, &sock);
2228 if (rc < 0) {
553675fb 2229 kfree(vs);
2230 return ERR_PTR(rc);
2231 }
e4c7ed41
CW		 2232
2233 vs->sock = sock;
2234 sk = sock->sk;
9c2e24e1 2235 atomic_set(&vs->refcnt, 1);
5cfccc5a 2236 vs->rcv = rcv;
012a5729 2237 vs->data = data;
553675fb 2238
9c2e24e1
PS		 2239 spin_lock(&vn->sock_lock);
2240 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2241 spin_unlock(&vn->sock_lock);
553675fb 2242
2243 /* Mark socket as an encapsulation socket. */
2244 udp_sk(sk)->encap_type = 1;
2245 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2246#if IS_ENABLED(CONFIG_IPV6)
2247 if (ipv6)
2248 ipv6_stub->udpv6_encap_enable();
2249 else
2250#endif
2251 udp_encap_enable();
2252
9c2e24e1
PS		 2253 return vs;
2254}
2255
012a5729
PS		 2256struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2257 vxlan_rcv_t *rcv, void *data,
e4c7ed41 2258 bool no_share, bool ipv6)
9c2e24e1
PS		 2259{
2260 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2261 struct vxlan_sock *vs;
2262
e4c7ed41 2263 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
9c2e24e1
PS		 2264 if (!IS_ERR(vs))
2265 return vs;
553675fb 2266
012a5729
PS		 2267 if (no_share) /* Return error if sharing is not allowed. */
2268 return vs;
2269
9c2e24e1
PS		 2270 spin_lock(&vn->sock_lock);
2271 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2272 if (vs) {
2273 if (vs->rcv == rcv)
2274 atomic_inc(&vs->refcnt);
2275 else
2276 vs = ERR_PTR(-EBUSY);
2277 }
2278 spin_unlock(&vn->sock_lock);
2279
2280 if (!vs)
9c2e24e1
PS		 2281 vs = ERR_PTR(-EINVAL);
2282
553675fb 2283 return vs;
2284}
012a5729 2285EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2286
1c51a915
SH		 2287/* Scheduled at device creation to bind to a socket */
2288static void vxlan_sock_work(struct work_struct *work)
2289{
9c2e24e1
PS		 2290 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2291 struct net *net = dev_net(vxlan->dev);
1c51a915 2292 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2293 __be16 port = vxlan->dst_port;
2294 struct vxlan_sock *nvs;
1c51a915 2295
e4c7ed41 2296 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
1c51a915 2297 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2298 if (!IS_ERR(nvs))
2299 vxlan_vs_add_dev(nvs, vxlan);
2300 spin_unlock(&vn->sock_lock);
2301
2302 dev_put(vxlan->dev);
1c51a915
SH		 2303}
2304
d342894c 2305static int vxlan_newlink(struct net *net, struct net_device *dev,
2306 struct nlattr *tb[], struct nlattr *data[])
2307{
553675fb 2308 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2309 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2310 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2311 __u32 vni;
2312 int err;
e4c7ed41 2313 bool use_ipv6 = false;
d342894c 2314
2315 if (!data[IFLA_VXLAN_ID])
2316 return -EINVAL;
2317
2318 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2319 dst->remote_vni = vni;
d342894c 2320
e4c7ed41
CW		 2321 if (data[IFLA_VXLAN_GROUP]) {
2322 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2323 dst->remote_ip.sa.sa_family = AF_INET;
2324 } else if (data[IFLA_VXLAN_GROUP6]) {
2325 if (!IS_ENABLED(CONFIG_IPV6))
2326 return -EPFNOSUPPORT;
2327
2328 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2329 sizeof(struct in6_addr));
2330 dst->remote_ip.sa.sa_family = AF_INET6;
2331 use_ipv6 = true;
2332 }
d342894c 2333
e4c7ed41
CW		 2334 if (data[IFLA_VXLAN_LOCAL]) {
2335 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2336 vxlan->saddr.sa.sa_family = AF_INET;
2337 } else if (data[IFLA_VXLAN_LOCAL6]) {
2338 if (!IS_ENABLED(CONFIG_IPV6))
2339 return -EPFNOSUPPORT;
2340
2341 /* TODO: respect scope id */
2342 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2343 sizeof(struct in6_addr));
2344 vxlan->saddr.sa.sa_family = AF_INET6;
2345 use_ipv6 = true;
2346 }
d342894c 2347
34e02aa1 2348 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2349 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2350 struct net_device *lowerdev
c7995c43 2351 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2352
2353 if (!lowerdev) {
c7995c43 2354 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2355 return -ENODEV;
2356 }
d342894c 2357
e4c7ed41
CW		 2358#if IS_ENABLED(CONFIG_IPV6)
2359 if (use_ipv6) {
2360 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2361 if (idev && idev->cnf.disable_ipv6) {
2362 pr_info("IPv6 is disabled via sysctl\n");
2363 return -EPERM;
2364 }
2365 vxlan->flags |= VXLAN_F_IPV6;
2366 }
2367#endif
2368
34e02aa1 2369 if (!tb[IFLA_MTU])
e4c7ed41 2370 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4
AD		 2371
2372 /* update header length based on lower device */
2373 dev->hard_header_len = lowerdev->hard_header_len +
e4c7ed41 2374 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
d342894c 2375 }
2376
2377 if (data[IFLA_VXLAN_TOS])
2378 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2379
afb97186
VB		 2380 if (data[IFLA_VXLAN_TTL])
2381 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2382
d342894c 2383 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2384 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2385
2386 if (data[IFLA_VXLAN_AGEING])
2387 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2388 else
2389 vxlan->age_interval = FDB_AGE_DEFAULT;
2390
e4f67add
DS		 2391 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2392 vxlan->flags |= VXLAN_F_PROXY;
2393
2394 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2395 vxlan->flags |= VXLAN_F_RSC;
2396
2397 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2398 vxlan->flags |= VXLAN_F_L2MISS;
2399
2400 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2401 vxlan->flags |= VXLAN_F_L3MISS;
2402
d342894c 2403 if (data[IFLA_VXLAN_LIMIT])
2404 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2405
05f47d69 2406 if (data[IFLA_VXLAN_PORT_RANGE]) {
2407 const struct ifla_vxlan_port_range *p
2408 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2409 vxlan->port_min = ntohs(p->low);
2410 vxlan->port_max = ntohs(p->high);
2411 }
2412
823aa873 2413 if (data[IFLA_VXLAN_PORT])
2414 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2415
553675fb 2416 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2417 pr_info("duplicate VNI %u\n", vni);
2418 return -EEXIST;
2419 }
2420
1b13c97f
YB		 2421 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2422
afbd8bae
MR		 2423 /* create an fdb entry for default destination */
2424 err = vxlan_fdb_create(vxlan, all_zeros_mac,
e4c7ed41 2425 &vxlan->default_dst.remote_ip,
afbd8bae
MR		 2426 NUD_REACHABLE|NUD_PERMANENT,
2427 NLM_F_EXCL|NLM_F_CREATE,
2428 vxlan->dst_port, vxlan->default_dst.remote_vni,
2429 vxlan->default_dst.remote_ifindex, NTF_SELF);
1c51a915 2430 if (err)
553675fb 2431 return err;
d342894c 2432
afbd8bae
MR		 2433 err = register_netdevice(dev);
2434 if (err) {
ba609e9b 2435 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2436 return err;
2437 }
2438
553675fb 2439 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2440
2441 return 0;
d342894c 2442}
2443
2444static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2445{
fe5c3561 2446 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 2447 struct vxlan_dev *vxlan = netdev_priv(dev);
2448
fe5c3561 2449 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2450 if (!hlist_unhashed(&vxlan->hlist))
2451 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2452 spin_unlock(&vn->sock_lock);
2453
553675fb 2454 list_del(&vxlan->next);
d342894c 2455 unregister_netdevice_queue(dev, head);
2456}
2457
2458static size_t vxlan_get_size(const struct net_device *dev)
2459{
2460
2461 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2462 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2463 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2464 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2465 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2466 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2467 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2468 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2469 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2470 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2471 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2472 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2473 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2474 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 2475 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 2476 0;
2477}
2478
2479static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2480{
2481 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2482 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2483 struct ifla_vxlan_port_range ports = {
2484 .low = htons(vxlan->port_min),
2485 .high = htons(vxlan->port_max),
2486 };
d342894c 2487
c7995c43 2488 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2489 goto nla_put_failure;
2490
e4c7ed41
CW		 2491 if (!vxlan_addr_any(&dst->remote_ip)) {
2492 if (dst->remote_ip.sa.sa_family == AF_INET) {
2493 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2494 dst->remote_ip.sin.sin_addr.s_addr))
2495 goto nla_put_failure;
2496#if IS_ENABLED(CONFIG_IPV6)
2497 } else {
2498 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2499 &dst->remote_ip.sin6.sin6_addr))
2500 goto nla_put_failure;
2501#endif
2502 }
2503 }
d342894c 2504
c7995c43 2505 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2506 goto nla_put_failure;
2507
e4c7ed41
CW		 2508 if (!vxlan_addr_any(&vxlan->saddr)) {
2509 if (vxlan->saddr.sa.sa_family == AF_INET) {
2510 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2511 vxlan->saddr.sin.sin_addr.s_addr))
2512 goto nla_put_failure;
2513#if IS_ENABLED(CONFIG_IPV6)
2514 } else {
2515 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2516 &vxlan->saddr.sin6.sin6_addr))
2517 goto nla_put_failure;
2518#endif
2519 }
2520 }
d342894c 2521
2522 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2523 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2524 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2525 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2526 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2527 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2528 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2529 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2530 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2531 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2532 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2533 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2534 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2535 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 2536 goto nla_put_failure;
2537
05f47d69 2538 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2539 goto nla_put_failure;
2540
d342894c 2541 return 0;
2542
2543nla_put_failure:
2544 return -EMSGSIZE;
2545}
2546
2547static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2548 .kind = "vxlan",
2549 .maxtype = IFLA_VXLAN_MAX,
2550 .policy = vxlan_policy,
2551 .priv_size = sizeof(struct vxlan_dev),
2552 .setup = vxlan_setup,
2553 .validate = vxlan_validate,
2554 .newlink = vxlan_newlink,
2555 .dellink = vxlan_dellink,
2556 .get_size = vxlan_get_size,
2557 .fill_info = vxlan_fill_info,
2558};
2559
2560static __net_init int vxlan_init_net(struct net *net)
2561{
2562 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2563 unsigned int h;
d342894c 2564
553675fb 2565 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2566 spin_lock_init(&vn->sock_lock);
d342894c 2567
553675fb 2568 for (h = 0; h < PORT_HASH_SIZE; ++h)
2569 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2570
2571 return 0;
2572}
2573
2574static __net_exit void vxlan_exit_net(struct net *net)
2575{
2576 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 2577 struct vxlan_dev *vxlan;
372675a4 2578 LIST_HEAD(list);
9cb6cb7e
ZM		 2579
2580 rtnl_lock();
553675fb 2581 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 2582 unregister_netdevice_queue(vxlan->dev, &list);
2583 unregister_netdevice_many(&list);
9cb6cb7e 2584 rtnl_unlock();
d342894c 2585}
2586
2587static struct pernet_operations vxlan_net_ops = {
2588 .init = vxlan_init_net,
2589 .exit = vxlan_exit_net,
2590 .id = &vxlan_net_id,
2591 .size = sizeof(struct vxlan_net),
2592};
2593
2594static int __init vxlan_init_module(void)
2595{
2596 int rc;
2597
758c57d1
SH		 2598 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2599 if (!vxlan_wq)
2600 return -ENOMEM;
2601
d342894c 2602 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2603
2604 rc = register_pernet_device(&vxlan_net_ops);
2605 if (rc)
2606 goto out1;
2607
2608 rc = rtnl_link_register(&vxlan_link_ops);
2609 if (rc)
2610 goto out2;
2611
2612 return 0;
2613
2614out2:
2615 unregister_pernet_device(&vxlan_net_ops);
2616out1:
758c57d1 2617 destroy_workqueue(vxlan_wq);
d342894c 2618 return rc;
2619}
7332a13b 2620late_initcall(vxlan_init_module);
d342894c 2621
2622static void __exit vxlan_cleanup_module(void)
2623{
b7153984 2624 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 2625 destroy_workqueue(vxlan_wq);
f89e57c4 2626 unregister_pernet_device(&vxlan_net_ops);
6681712d 2627 rcu_barrier();
d342894c 2628}
2629module_exit(vxlan_cleanup_module);
2630
2631MODULE_LICENSE("GPL");
2632MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2633MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 2634MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)