]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tcp: out_of_order_queue do not use its lock
[mirror_ubuntu-hirsute-kernel.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>
660d98ca 47#include <net/ip6_checksum.h>
e4c7ed41 48#endif
d342894c 49
50#define VXLAN_VERSION "0.1"
51
553675fb 52#define PORT_HASH_BITS 8
53#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 54#define VNI_HASH_BITS 10
55#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
56#define FDB_HASH_BITS 8
57#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
61#define VXLAN_N_VID (1u << 24)
62#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
7ce04758 63#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
d342894c 64
65#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
66
67/* VXLAN protocol header */
68struct vxlanhdr {
69 __be32 vx_flags;
70 __be32 vx_vni;
71};
72
23c578bf 73/* UDP port for VXLAN traffic.
74 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 75 * for compatibility with early adopters.
23c578bf 76 */
9daaa397
SH		 77static unsigned short vxlan_port __read_mostly = 8472;
78module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 79MODULE_PARM_DESC(udp_port, "Destination UDP port");
80
81static bool log_ecn_error = true;
82module_param(log_ecn_error, bool, 0644);
83MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
84
60d9d4c6 85static int vxlan_net_id;
553675fb 86
afbd8bae
MR		 87static const u8 all_zeros_mac[ETH_ALEN];
88
553675fb 89/* per-network namespace private data for this module */
90struct vxlan_net {
91 struct list_head vxlan_list;
92 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 93 spinlock_t sock_lock;
553675fb 94};
95
e4c7ed41
CW		 96union vxlan_addr {
97 struct sockaddr_in sin;
98 struct sockaddr_in6 sin6;
99 struct sockaddr sa;
100};
101
6681712d 102struct vxlan_rdst {
e4c7ed41 103 union vxlan_addr remote_ip;
6681712d
DS		 104 __be16 remote_port;
105 u32 remote_vni;
106 u32 remote_ifindex;
3e61aa8f 107 struct list_head list;
bc7892ba 108 struct rcu_head rcu;
6681712d
DS		 109};
110
d342894c 111/* Forwarding table entry */
112struct vxlan_fdb {
113 struct hlist_node hlist; /* linked list of entries */
114 struct rcu_head rcu;
115 unsigned long updated; /* jiffies */
116 unsigned long used;
3e61aa8f 117 struct list_head remotes;
d342894c 118 u16 state; /* see ndm_state */
ae884082 119 u8 flags; /* see ndm_flags */
d342894c 120 u8 eth_addr[ETH_ALEN];
121};
122
d342894c 123/* Pseudo network device */
124struct vxlan_dev {
553675fb 125 struct hlist_node hlist; /* vni hash table */
126 struct list_head next; /* vxlan's per namespace list */
127 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 128 struct net_device *dev;
c7995c43 129 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 130 union vxlan_addr saddr; /* source address */
823aa873 131 __be16 dst_port;
05f47d69 132 __u16 port_min; /* source port range */
133 __u16 port_max;
d342894c 134 __u8 tos; /* TOS override */
135 __u8 ttl;
e4f67add 136 u32 flags; /* VXLAN_F_* below */
d342894c 137
1c51a915 138 struct work_struct sock_work;
3fc2de2f 139 struct work_struct igmp_join;
140 struct work_struct igmp_leave;
1c51a915 141
d342894c 142 unsigned long age_interval;
143 struct timer_list age_timer;
144 spinlock_t hash_lock;
145 unsigned int addrcnt;
146 unsigned int addrmax;
d342894c 147
148 struct hlist_head fdb_head[FDB_HASH_SIZE];
149};
150
e4f67add
DS		 151#define VXLAN_F_LEARN 0x01
152#define VXLAN_F_PROXY 0x02
153#define VXLAN_F_RSC 0x04
154#define VXLAN_F_L2MISS 0x08
155#define VXLAN_F_L3MISS 0x10
e4c7ed41 156#define VXLAN_F_IPV6 0x20 /* internal flag */
e4f67add 157
d342894c 158/* salt for hash table */
159static u32 vxlan_salt __read_mostly;
758c57d1 160static struct workqueue_struct *vxlan_wq;
d342894c 161
1c51a915
SH		 162static void vxlan_sock_work(struct work_struct *work);
163
e4c7ed41
CW		 164#if IS_ENABLED(CONFIG_IPV6)
165static inline
166bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
167{
168 if (a->sa.sa_family != b->sa.sa_family)
169 return false;
170 if (a->sa.sa_family == AF_INET6)
171 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
172 else
173 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
174}
175
176static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
177{
178 if (ipa->sa.sa_family == AF_INET6)
179 return ipv6_addr_any(&ipa->sin6.sin6_addr);
180 else
181 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
182}
183
184static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
185{
186 if (ipa->sa.sa_family == AF_INET6)
187 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
188 else
189 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
190}
191
192static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
193{
194 if (nla_len(nla) >= sizeof(struct in6_addr)) {
195 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
196 ip->sa.sa_family = AF_INET6;
197 return 0;
198 } else if (nla_len(nla) >= sizeof(__be32)) {
199 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
200 ip->sa.sa_family = AF_INET;
201 return 0;
202 } else {
203 return -EAFNOSUPPORT;
204 }
205}
206
207static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
208 const union vxlan_addr *ip)
209{
210 if (ip->sa.sa_family == AF_INET6)
211 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
212 else
213 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
214}
215
216#else /* !CONFIG_IPV6 */
217
218static inline
219bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
220{
221 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
222}
223
224static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
225{
226 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
227}
228
229static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
230{
231 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
232}
233
234static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
235{
236 if (nla_len(nla) >= sizeof(struct in6_addr)) {
237 return -EAFNOSUPPORT;
238 } else if (nla_len(nla) >= sizeof(__be32)) {
239 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
240 ip->sa.sa_family = AF_INET;
241 return 0;
242 } else {
243 return -EAFNOSUPPORT;
244 }
245}
246
247static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
248 const union vxlan_addr *ip)
249{
250 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
251}
252#endif
253
553675fb 254/* Virtual Network hash table head */
255static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
256{
257 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
258}
259
260/* Socket hash table head */
261static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 262{
263 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
264
553675fb 265 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
266}
267
3e61aa8f
SH		 268/* First remote destination for a forwarding entry.
269 * Guaranteed to be non-NULL because remotes are never deleted.
270 */
5ca5461c 271static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 272{
5ca5461c 273 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
274}
275
276static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
277{
278 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 279}
280
553675fb 281/* Find VXLAN socket based on network namespace and UDP port */
9c2e24e1 282static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
553675fb 283{
284 struct vxlan_sock *vs;
285
286 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
287 if (inet_sk(vs->sock->sk)->inet_sport == port)
288 return vs;
289 }
290 return NULL;
d342894c 291}
292
5cfccc5a 293static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 294{
295 struct vxlan_dev *vxlan;
d342894c 296
553675fb 297 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 298 if (vxlan->default_dst.remote_vni == id)
d342894c 299 return vxlan;
300 }
301
302 return NULL;
303}
304
5cfccc5a
PS		 305/* Look up VNI in a per net namespace table */
306static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
307{
308 struct vxlan_sock *vs;
309
310 vs = vxlan_find_sock(net, port);
311 if (!vs)
312 return NULL;
313
314 return vxlan_vs_find_vni(vs, id);
315}
316
d342894c 317/* Fill in neighbour message in skbuff. */
318static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 319 const struct vxlan_fdb *fdb,
320 u32 portid, u32 seq, int type, unsigned int flags,
321 const struct vxlan_rdst *rdst)
d342894c 322{
323 unsigned long now = jiffies;
324 struct nda_cacheinfo ci;
325 struct nlmsghdr *nlh;
326 struct ndmsg *ndm;
e4f67add 327 bool send_ip, send_eth;
d342894c 328
329 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
330 if (nlh == NULL)
331 return -EMSGSIZE;
332
333 ndm = nlmsg_data(nlh);
334 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 335
336 send_eth = send_ip = true;
337
338 if (type == RTM_GETNEIGH) {
339 ndm->ndm_family = AF_INET;
e4c7ed41 340 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 341 send_eth = !is_zero_ether_addr(fdb->eth_addr);
342 } else
343 ndm->ndm_family = AF_BRIDGE;
d342894c 344 ndm->ndm_state = fdb->state;
345 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 346 ndm->ndm_flags = fdb->flags;
d342894c 347 ndm->ndm_type = NDA_DST;
348
e4f67add 349 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 350 goto nla_put_failure;
351
e4c7ed41 352 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 353 goto nla_put_failure;
354
823aa873 355 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 356 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
357 goto nla_put_failure;
c7995c43 358 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 359 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 360 goto nla_put_failure;
361 if (rdst->remote_ifindex &&
362 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 363 goto nla_put_failure;
364
365 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
366 ci.ndm_confirmed = 0;
367 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
368 ci.ndm_refcnt = 0;
369
370 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
371 goto nla_put_failure;
372
373 return nlmsg_end(skb, nlh);
374
375nla_put_failure:
376 nlmsg_cancel(skb, nlh);
377 return -EMSGSIZE;
378}
379
380static inline size_t vxlan_nlmsg_size(void)
381{
382 return NLMSG_ALIGN(sizeof(struct ndmsg))
383 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 384 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 385 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 386 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
387 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 388 + nla_total_size(sizeof(struct nda_cacheinfo));
389}
390
391static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
3e61aa8f 392 struct vxlan_fdb *fdb, int type)
d342894c 393{
394 struct net *net = dev_net(vxlan->dev);
395 struct sk_buff *skb;
396 int err = -ENOBUFS;
397
398 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
399 if (skb == NULL)
400 goto errout;
401
5ca5461c 402 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
403 first_remote_rtnl(fdb));
d342894c 404 if (err < 0) {
405 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
406 WARN_ON(err == -EMSGSIZE);
407 kfree_skb(skb);
408 goto errout;
409 }
410
411 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
412 return;
413errout:
414 if (err < 0)
415 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
416}
417
e4c7ed41 418static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 419{
420 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 421 struct vxlan_fdb f = {
422 .state = NUD_STALE,
423 };
424 struct vxlan_rdst remote = {
e4c7ed41 425 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 426 .remote_vni = VXLAN_N_VID,
427 };
3e61aa8f
SH		 428
429 INIT_LIST_HEAD(&f.remotes);
430 list_add_rcu(&remote.list, &f.remotes);
e4f67add
DS		 431
432 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
433}
434
435static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
436{
bb3fd687
SH		 437 struct vxlan_fdb f = {
438 .state = NUD_STALE,
439 };
e4f67add 440
3e61aa8f 441 INIT_LIST_HEAD(&f.remotes);
e4f67add
DS		 442 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
443
444 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
445}
446
d342894c 447/* Hash Ethernet address */
448static u32 eth_hash(const unsigned char *addr)
449{
450 u64 value = get_unaligned((u64 *)addr);
451
452 /* only want 6 bytes */
453#ifdef __BIG_ENDIAN
d342894c 454 value >>= 16;
321fb991 455#else
456 value <<= 16;
d342894c 457#endif
458 return hash_64(value, FDB_HASH_BITS);
459}
460
461/* Hash chain to use given mac address */
462static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
463 const u8 *mac)
464{
465 return &vxlan->fdb_head[eth_hash(mac)];
466}
467
468/* Look up Ethernet address in forwarding table */
014be2c8 469static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 470 const u8 *mac)
471
472{
473 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
474 struct vxlan_fdb *f;
d342894c 475
b67bfe0d 476 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 477 if (ether_addr_equal(mac, f->eth_addr))
d342894c 478 return f;
479 }
480
481 return NULL;
482}
483
014be2c8
SS		 484static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
485 const u8 *mac)
486{
487 struct vxlan_fdb *f;
488
489 f = __vxlan_find_mac(vxlan, mac);
490 if (f)
491 f->used = jiffies;
492
493 return f;
494}
495
a5e7c10a
MR		 496/* caller should hold vxlan->hash_lock */
497static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 498 union vxlan_addr *ip, __be16 port,
a5e7c10a 499 __u32 vni, __u32 ifindex)
6681712d 500{
3e61aa8f 501 struct vxlan_rdst *rd;
6681712d 502
3e61aa8f 503 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 504 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 505 rd->remote_port == port &&
506 rd->remote_vni == vni &&
507 rd->remote_ifindex == ifindex)
a5e7c10a 508 return rd;
6681712d 509 }
3e61aa8f 510
a5e7c10a
MR		 511 return NULL;
512}
513
906dc186
TR		 514/* Replace destination of unicast mac */
515static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 516 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 517{
518 struct vxlan_rdst *rd;
519
520 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
521 if (rd)
522 return 0;
523
524 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
525 if (!rd)
526 return 0;
e4c7ed41 527 rd->remote_ip = *ip;
906dc186
TR		 528 rd->remote_port = port;
529 rd->remote_vni = vni;
530 rd->remote_ifindex = ifindex;
531 return 1;
532}
533
a5e7c10a
MR		 534/* Add/update destinations for multicast */
535static int vxlan_fdb_append(struct vxlan_fdb *f,
e4c7ed41 536 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
a5e7c10a
MR		 537{
538 struct vxlan_rdst *rd;
539
540 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
541 if (rd)
542 return 0;
543
6681712d
DS		 544 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
545 if (rd == NULL)
546 return -ENOBUFS;
e4c7ed41 547 rd->remote_ip = *ip;
6681712d
DS		 548 rd->remote_port = port;
549 rd->remote_vni = vni;
550 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 551
552 list_add_tail_rcu(&rd->list, &f->remotes);
553
6681712d
DS		 554 return 1;
555}
556
53cf5275
JG		 557/* Notify netdevs that UDP port started listening */
558static void vxlan_notify_add_rx_port(struct sock *sk)
559{
560 struct net_device *dev;
561 struct net *net = sock_net(sk);
562 sa_family_t sa_family = sk->sk_family;
35e42379 563 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 564
565 rcu_read_lock();
566 for_each_netdev_rcu(net, dev) {
567 if (dev->netdev_ops->ndo_add_vxlan_port)
568 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
569 port);
570 }
571 rcu_read_unlock();
572}
573
574/* Notify netdevs that UDP port is no more listening */
575static void vxlan_notify_del_rx_port(struct sock *sk)
576{
577 struct net_device *dev;
578 struct net *net = sock_net(sk);
579 sa_family_t sa_family = sk->sk_family;
35e42379 580 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 581
582 rcu_read_lock();
583 for_each_netdev_rcu(net, dev) {
584 if (dev->netdev_ops->ndo_del_vxlan_port)
585 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
586 port);
587 }
588 rcu_read_unlock();
589}
590
d342894c 591/* Add new entry to forwarding table -- assumes lock held */
592static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 593 const u8 *mac, union vxlan_addr *ip,
6681712d 594 __u16 state, __u16 flags,
73cf3317 595 __be16 port, __u32 vni, __u32 ifindex,
ae884082 596 __u8 ndm_flags)
d342894c 597{
598 struct vxlan_fdb *f;
599 int notify = 0;
600
014be2c8 601 f = __vxlan_find_mac(vxlan, mac);
d342894c 602 if (f) {
603 if (flags & NLM_F_EXCL) {
604 netdev_dbg(vxlan->dev,
605 "lost race to create %pM\n", mac);
606 return -EEXIST;
607 }
608 if (f->state != state) {
609 f->state = state;
610 f->updated = jiffies;
611 notify = 1;
612 }
ae884082
DS		 613 if (f->flags != ndm_flags) {
614 f->flags = ndm_flags;
615 f->updated = jiffies;
616 notify = 1;
617 }
906dc186
TR		 618 if ((flags & NLM_F_REPLACE)) {
619 /* Only change unicasts */
620 if (!(is_multicast_ether_addr(f->eth_addr) ||
621 is_zero_ether_addr(f->eth_addr))) {
622 int rc = vxlan_fdb_replace(f, ip, port, vni,
623 ifindex);
624
625 if (rc < 0)
626 return rc;
627 notify |= rc;
628 } else
629 return -EOPNOTSUPP;
630 }
6681712d 631 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 632 (is_multicast_ether_addr(f->eth_addr) ||
633 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 634 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
635
636 if (rc < 0)
637 return rc;
638 notify |= rc;
639 }
d342894c 640 } else {
641 if (!(flags & NLM_F_CREATE))
642 return -ENOENT;
643
644 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
645 return -ENOSPC;
646
906dc186
TR		 647 /* Disallow replace to add a multicast entry */
648 if ((flags & NLM_F_REPLACE) &&
649 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
650 return -EOPNOTSUPP;
651
e4c7ed41 652 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 653 f = kmalloc(sizeof(*f), GFP_ATOMIC);
654 if (!f)
655 return -ENOMEM;
656
657 notify = 1;
d342894c 658 f->state = state;
ae884082 659 f->flags = ndm_flags;
d342894c 660 f->updated = f->used = jiffies;
3e61aa8f 661 INIT_LIST_HEAD(&f->remotes);
d342894c 662 memcpy(f->eth_addr, mac, ETH_ALEN);
663
3e61aa8f
SH		 664 vxlan_fdb_append(f, ip, port, vni, ifindex);
665
d342894c 666 ++vxlan->addrcnt;
667 hlist_add_head_rcu(&f->hlist,
668 vxlan_fdb_head(vxlan, mac));
669 }
670
671 if (notify)
672 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
673
674 return 0;
675}
676
6706c82e 677static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 678{
679 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 680 struct vxlan_rdst *rd, *nd;
6681712d 681
3e61aa8f 682 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 683 kfree(rd);
6681712d
DS		 684 kfree(f);
685}
686
d342894c 687static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
688{
689 netdev_dbg(vxlan->dev,
690 "delete %pM\n", f->eth_addr);
691
692 --vxlan->addrcnt;
693 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
694
695 hlist_del_rcu(&f->hlist);
6681712d 696 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 697}
698
f0b074be 699static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 700 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 701{
6681712d 702 struct net *net = dev_net(vxlan->dev);
e4c7ed41 703 int err;
d342894c 704
f0b074be 705 if (tb[NDA_DST]) {
e4c7ed41
CW		 706 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
707 if (err)
708 return err;
f0b074be 709 } else {
e4c7ed41
CW		 710 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
711 if (remote->sa.sa_family == AF_INET) {
712 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
713 ip->sa.sa_family = AF_INET;
714#if IS_ENABLED(CONFIG_IPV6)
715 } else {
716 ip->sin6.sin6_addr = in6addr_any;
717 ip->sa.sa_family = AF_INET6;
718#endif
719 }
f0b074be 720 }
d342894c 721
6681712d 722 if (tb[NDA_PORT]) {
73cf3317 723 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 724 return -EINVAL;
f0b074be
MR		 725 *port = nla_get_be16(tb[NDA_PORT]);
726 } else {
727 *port = vxlan->dst_port;
728 }
6681712d
DS		 729
730 if (tb[NDA_VNI]) {
731 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
732 return -EINVAL;
f0b074be
MR		 733 *vni = nla_get_u32(tb[NDA_VNI]);
734 } else {
735 *vni = vxlan->default_dst.remote_vni;
736 }
6681712d
DS		 737
738 if (tb[NDA_IFINDEX]) {
5abb0029 739 struct net_device *tdev;
6681712d
DS		 740
741 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
742 return -EINVAL;
f0b074be
MR		 743 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
744 tdev = dev_get_by_index(net, *ifindex);
5abb0029 745 if (!tdev)
6681712d 746 return -EADDRNOTAVAIL;
5abb0029 747 dev_put(tdev);
f0b074be
MR		 748 } else {
749 *ifindex = 0;
750 }
751
752 return 0;
753}
754
755/* Add static entry (via netlink) */
756static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
757 struct net_device *dev,
758 const unsigned char *addr, u16 flags)
759{
760 struct vxlan_dev *vxlan = netdev_priv(dev);
761 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 762 union vxlan_addr ip;
f0b074be
MR		 763 __be16 port;
764 u32 vni, ifindex;
765 int err;
766
767 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
768 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
769 ndm->ndm_state);
770 return -EINVAL;
771 }
772
773 if (tb[NDA_DST] == NULL)
774 return -EINVAL;
775
776 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
777 if (err)
778 return err;
6681712d 779
d342894c 780 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 781 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 782 port, vni, ifindex, ndm->ndm_flags);
d342894c 783 spin_unlock_bh(&vxlan->hash_lock);
784
785 return err;
786}
787
788/* Delete entry (via netlink) */
1690be63
VY		 789static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
790 struct net_device *dev,
d342894c 791 const unsigned char *addr)
792{
793 struct vxlan_dev *vxlan = netdev_priv(dev);
794 struct vxlan_fdb *f;
bc7892ba 795 struct vxlan_rdst *rd = NULL;
e4c7ed41 796 union vxlan_addr ip;
bc7892ba
MR		 797 __be16 port;
798 u32 vni, ifindex;
799 int err;
800
801 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
802 if (err)
803 return err;
804
805 err = -ENOENT;
d342894c 806
807 spin_lock_bh(&vxlan->hash_lock);
808 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 809 if (!f)
810 goto out;
811
e4c7ed41
CW		 812 if (!vxlan_addr_any(&ip)) {
813 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 814 if (!rd)
815 goto out;
816 }
817
818 err = 0;
819
820 /* remove a destination if it's not the only one on the list,
821 * otherwise destroy the fdb entry
822 */
823 if (rd && !list_is_singular(&f->remotes)) {
824 list_del_rcu(&rd->list);
dcdc7a59 825 kfree_rcu(rd, rcu);
bc7892ba 826 goto out;
d342894c 827 }
bc7892ba
MR		 828
829 vxlan_fdb_destroy(vxlan, f);
830
831out:
d342894c 832 spin_unlock_bh(&vxlan->hash_lock);
833
834 return err;
835}
836
837/* Dump forwarding table */
838static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
839 struct net_device *dev, int idx)
840{
841 struct vxlan_dev *vxlan = netdev_priv(dev);
842 unsigned int h;
843
844 for (h = 0; h < FDB_HASH_SIZE; ++h) {
845 struct vxlan_fdb *f;
d342894c 846 int err;
847
b67bfe0d 848 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 849 struct vxlan_rdst *rd;
6681712d 850
3e61aa8f
SH		 851 if (idx < cb->args[0])
852 goto skip;
853
854 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 855 err = vxlan_fdb_info(skb, vxlan, f,
856 NETLINK_CB(cb->skb).portid,
857 cb->nlh->nlmsg_seq,
858 RTM_NEWNEIGH,
859 NLM_F_MULTI, rd);
860 if (err < 0)
3e61aa8f 861 goto out;
6681712d 862 }
3e61aa8f
SH		 863skip:
864 ++idx;
d342894c 865 }
866 }
3e61aa8f 867out:
d342894c 868 return idx;
869}
870
871/* Watch incoming packets to learn mapping between Ethernet address
872 * and Tunnel endpoint.
26a41ae6 873 * Return true if packet is bogus and should be droppped.
d342894c 874 */
26a41ae6 875static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 876 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 877{
878 struct vxlan_dev *vxlan = netdev_priv(dev);
879 struct vxlan_fdb *f;
d342894c 880
881 f = vxlan_find_mac(vxlan, src_mac);
882 if (likely(f)) {
5ca5461c 883 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 884
e4c7ed41 885 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 886 return false;
887
888 /* Don't migrate static entries, drop packets */
eb064c3b 889 if (f->state & NUD_NOARP)
26a41ae6 890 return true;
d342894c 891
892 if (net_ratelimit())
893 netdev_info(dev,
e4c7ed41 894 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 895 src_mac, &rdst->remote_ip, &src_ip);
d342894c 896
e4c7ed41 897 rdst->remote_ip = *src_ip;
d342894c 898 f->updated = jiffies;
8385f50a 899 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 900 } else {
901 /* learned new entry */
902 spin_lock(&vxlan->hash_lock);
3bf74b1a 903
904 /* close off race between vxlan_flush and incoming packets */
905 if (netif_running(dev))
906 vxlan_fdb_create(vxlan, src_mac, src_ip,
907 NUD_REACHABLE,
908 NLM_F_EXCL|NLM_F_CREATE,
909 vxlan->dst_port,
910 vxlan->default_dst.remote_vni,
911 0, NTF_SELF);
d342894c 912 spin_unlock(&vxlan->hash_lock);
913 }
26a41ae6 914
915 return false;
d342894c 916}
917
d342894c 918/* See if multicast group is already in use by other ID */
95ab0991 919static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 920{
553675fb 921 struct vxlan_dev *vxlan;
d342894c 922
95ab0991
G		 923 /* The vxlan_sock is only used by dev, leaving group has
924 * no effect on other vxlan devices.
925 */
926 if (atomic_read(&dev->vn_sock->refcnt) == 1)
927 return false;
928
553675fb 929 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 930 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 931 continue;
d342894c 932
95ab0991
G		 933 if (vxlan->vn_sock != dev->vn_sock)
934 continue;
935
936 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
937 &dev->default_dst.remote_ip))
938 continue;
939
940 if (vxlan->default_dst.remote_ifindex !=
941 dev->default_dst.remote_ifindex)
942 continue;
943
944 return true;
553675fb 945 }
d342894c 946
947 return false;
948}
949
7c47cedf 950static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 951{
7c47cedf
SH		 952 atomic_inc(&vs->refcnt);
953}
d342894c 954
012a5729 955void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 956{
53cf5275
JG		 957 struct sock *sk = vs->sock->sk;
958 struct net *net = sock_net(sk);
959 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 960
7c47cedf
SH		 961 if (!atomic_dec_and_test(&vs->refcnt))
962 return;
d342894c 963
1c51a915 964 spin_lock(&vn->sock_lock);
7c47cedf 965 hlist_del_rcu(&vs->hlist);
559835ea 966 rcu_assign_sk_user_data(vs->sock->sk, NULL);
53cf5275 967 vxlan_notify_del_rx_port(sk);
1c51a915
SH		 968 spin_unlock(&vn->sock_lock);
969
7c47cedf 970 queue_work(vxlan_wq, &vs->del_work);
d342894c 971}
012a5729 972EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 973
3fc2de2f 974/* Callback to update multicast group membership when first VNI on
975 * multicast asddress is brought up
976 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 977 */
3fc2de2f 978static void vxlan_igmp_join(struct work_struct *work)
d342894c 979{
3fc2de2f 980 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 981 struct vxlan_sock *vs = vxlan->vn_sock;
982 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 983 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
984 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 985
d342894c 986 lock_sock(sk);
e4c7ed41
CW		 987 if (ip->sa.sa_family == AF_INET) {
988 struct ip_mreqn mreq = {
989 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
990 .imr_ifindex = ifindex,
991 };
992
993 ip_mc_join_group(sk, &mreq);
994#if IS_ENABLED(CONFIG_IPV6)
995 } else {
996 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
997 &ip->sin6.sin6_addr);
998#endif
999 }
3fc2de2f 1000 release_sock(sk);
1001
012a5729 1002 vxlan_sock_release(vs);
3fc2de2f 1003 dev_put(vxlan->dev);
1004}
1005
1006/* Inverse of vxlan_igmp_join when last VNI is brought down */
1007static void vxlan_igmp_leave(struct work_struct *work)
1008{
1009 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1010 struct vxlan_sock *vs = vxlan->vn_sock;
1011 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1012 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1013 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1014
1015 lock_sock(sk);
e4c7ed41
CW		 1016 if (ip->sa.sa_family == AF_INET) {
1017 struct ip_mreqn mreq = {
1018 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1019 .imr_ifindex = ifindex,
1020 };
1021
1022 ip_mc_leave_group(sk, &mreq);
1023#if IS_ENABLED(CONFIG_IPV6)
1024 } else {
1025 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1026 &ip->sin6.sin6_addr);
1027#endif
1028 }
1029
d342894c 1030 release_sock(sk);
d342894c 1031
012a5729 1032 vxlan_sock_release(vs);
7c47cedf 1033 dev_put(vxlan->dev);
d342894c 1034}
1035
1036/* Callback from net/ipv4/udp.c to receive packets */
1037static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1038{
5cfccc5a 1039 struct vxlan_sock *vs;
d342894c 1040 struct vxlanhdr *vxh;
553675fb 1041 __be16 port;
d342894c 1042
d342894c 1043 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1044 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1045 goto error;
1046
7ce04758
PS		 1047 /* Return packets with reserved bits set */
1048 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 1049 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1050 (vxh->vx_vni & htonl(0xff))) {
1051 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1052 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1053 goto error;
1054 }
1055
5cfccc5a
PS		 1056 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1057 goto drop;
1058
553675fb 1059 port = inet_sk(sk)->inet_sport;
5cfccc5a 1060
559835ea 1061 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1062 if (!vs)
d342894c 1063 goto drop;
d342894c 1064
5cfccc5a
PS		 1065 vs->rcv(vs, skb, vxh->vx_vni);
1066 return 0;
1067
1068drop:
1069 /* Consume bad packet */
1070 kfree_skb(skb);
1071 return 0;
1072
1073error:
1074 /* Return non vxlan pkt */
1075 return 1;
1076}
1077
1078static void vxlan_rcv(struct vxlan_sock *vs,
1079 struct sk_buff *skb, __be32 vx_vni)
1080{
e4c7ed41
CW		 1081 struct iphdr *oip = NULL;
1082 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1083 struct vxlan_dev *vxlan;
8f84985f 1084 struct pcpu_sw_netstats *stats;
e4c7ed41 1085 union vxlan_addr saddr;
5cfccc5a 1086 __u32 vni;
e4c7ed41
CW		 1087 int err = 0;
1088 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1089
1090 vni = ntohl(vx_vni) >> 8;
1091 /* Is this VNI defined? */
1092 vxlan = vxlan_vs_find_vni(vs, vni);
1093 if (!vxlan)
d342894c 1094 goto drop;
d342894c 1095
e4c7ed41 1096 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1097 skb_reset_mac_header(skb);
d342894c 1098 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 1099
1100 /* Ignore packet loops (and multicast echo) */
7367d0b5 1101 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1102 goto drop;
1103
7ce04758 1104 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1105 if (remote_ip->sa.sa_family == AF_INET) {
1106 oip = ip_hdr(skb);
1107 saddr.sin.sin_addr.s_addr = oip->saddr;
1108 saddr.sa.sa_family = AF_INET;
1109#if IS_ENABLED(CONFIG_IPV6)
1110 } else {
1111 oip6 = ipv6_hdr(skb);
1112 saddr.sin6.sin6_addr = oip6->saddr;
1113 saddr.sa.sa_family = AF_INET6;
1114#endif
1115 }
1116
26a41ae6 1117 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1118 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1119 goto drop;
d342894c 1120
d342894c 1121 skb_reset_network_header(skb);
0afb1666
JG		 1122
1123 /* If the NIC driver gave us an encapsulated packet with
1124 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1125 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1126 * for us. Otherwise force the upper layers to verify it.
1127 */
1128 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1129 !(vxlan->dev->features & NETIF_F_RXCSUM))
1130 skb->ip_summed = CHECKSUM_NONE;
1131
1132 skb->encapsulation = 0;
d342894c 1133
e4c7ed41
CW		 1134 if (oip6)
1135 err = IP6_ECN_decapsulate(oip6, skb);
1136 if (oip)
1137 err = IP_ECN_decapsulate(oip, skb);
1138
d342894c 1139 if (unlikely(err)) {
e4c7ed41
CW		 1140 if (log_ecn_error) {
1141 if (oip6)
1142 net_info_ratelimited("non-ECT from %pI6\n",
1143 &oip6->saddr);
1144 if (oip)
1145 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1146 &oip->saddr, oip->tos);
1147 }
d342894c 1148 if (err > 1) {
1149 ++vxlan->dev->stats.rx_frame_errors;
1150 ++vxlan->dev->stats.rx_errors;
1151 goto drop;
1152 }
1153 }
1154
e8171045 1155 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1156 u64_stats_update_begin(&stats->syncp);
1157 stats->rx_packets++;
1158 stats->rx_bytes += skb->len;
1159 u64_stats_update_end(&stats->syncp);
1160
1161 netif_rx(skb);
1162
5cfccc5a 1163 return;
d342894c 1164drop:
1165 /* Consume bad packet */
1166 kfree_skb(skb);
d342894c 1167}
1168
e4f67add
DS		 1169static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1170{
1171 struct vxlan_dev *vxlan = netdev_priv(dev);
1172 struct arphdr *parp;
1173 u8 *arpptr, *sha;
1174 __be32 sip, tip;
1175 struct neighbour *n;
1176
1177 if (dev->flags & IFF_NOARP)
1178 goto out;
1179
1180 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1181 dev->stats.tx_dropped++;
1182 goto out;
1183 }
1184 parp = arp_hdr(skb);
1185
1186 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1187 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1188 parp->ar_pro != htons(ETH_P_IP) ||
1189 parp->ar_op != htons(ARPOP_REQUEST) ||
1190 parp->ar_hln != dev->addr_len ||
1191 parp->ar_pln != 4)
1192 goto out;
1193 arpptr = (u8 *)parp + sizeof(struct arphdr);
1194 sha = arpptr;
1195 arpptr += dev->addr_len; /* sha */
1196 memcpy(&sip, arpptr, sizeof(sip));
1197 arpptr += sizeof(sip);
1198 arpptr += dev->addr_len; /* tha */
1199 memcpy(&tip, arpptr, sizeof(tip));
1200
1201 if (ipv4_is_loopback(tip) ||
1202 ipv4_is_multicast(tip))
1203 goto out;
1204
1205 n = neigh_lookup(&arp_tbl, &tip, dev);
1206
1207 if (n) {
e4f67add
DS		 1208 struct vxlan_fdb *f;
1209 struct sk_buff *reply;
1210
1211 if (!(n->nud_state & NUD_CONNECTED)) {
1212 neigh_release(n);
1213 goto out;
1214 }
1215
1216 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1217 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1218 /* bridge-local neighbor */
1219 neigh_release(n);
1220 goto out;
1221 }
1222
1223 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1224 n->ha, sha);
1225
1226 neigh_release(n);
1227
1228 skb_reset_mac_header(reply);
1229 __skb_pull(reply, skb_network_offset(reply));
1230 reply->ip_summed = CHECKSUM_UNNECESSARY;
1231 reply->pkt_type = PACKET_HOST;
1232
1233 if (netif_rx_ni(reply) == NET_RX_DROP)
1234 dev->stats.rx_dropped++;
e4c7ed41
CW		 1235 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1236 union vxlan_addr ipa = {
1237 .sin.sin_addr.s_addr = tip,
1238 .sa.sa_family = AF_INET,
1239 };
1240
1241 vxlan_ip_miss(dev, &ipa);
1242 }
e4f67add
DS		 1243out:
1244 consume_skb(skb);
1245 return NETDEV_TX_OK;
1246}
1247
f564f45c
CW		 1248#if IS_ENABLED(CONFIG_IPV6)
1249static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1250{
1251 struct vxlan_dev *vxlan = netdev_priv(dev);
1252 struct neighbour *n;
1253 union vxlan_addr ipa;
1254 const struct ipv6hdr *iphdr;
1255 const struct in6_addr *saddr, *daddr;
1256 struct nd_msg *msg;
1257 struct inet6_dev *in6_dev = NULL;
1258
1259 in6_dev = __in6_dev_get(dev);
1260 if (!in6_dev)
1261 goto out;
1262
1263 if (!pskb_may_pull(skb, skb->len))
1264 goto out;
1265
1266 iphdr = ipv6_hdr(skb);
1267 saddr = &iphdr->saddr;
1268 daddr = &iphdr->daddr;
1269
1270 if (ipv6_addr_loopback(daddr) ||
1271 ipv6_addr_is_multicast(daddr))
1272 goto out;
1273
1274 msg = (struct nd_msg *)skb_transport_header(skb);
1275 if (msg->icmph.icmp6_code != 0 ||
1276 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1277 goto out;
1278
1279 n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1280
1281 if (n) {
1282 struct vxlan_fdb *f;
1283
1284 if (!(n->nud_state & NUD_CONNECTED)) {
1285 neigh_release(n);
1286 goto out;
1287 }
1288
1289 f = vxlan_find_mac(vxlan, n->ha);
1290 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1291 /* bridge-local neighbor */
1292 neigh_release(n);
1293 goto out;
1294 }
1295
1296 ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1297 !!in6_dev->cnf.forwarding,
1298 true, false, false);
1299 neigh_release(n);
1300 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1301 ipa.sin6.sin6_addr = *daddr;
1302 ipa.sa.sa_family = AF_INET6;
1303 vxlan_ip_miss(dev, &ipa);
1304 }
1305
1306out:
1307 consume_skb(skb);
1308 return NETDEV_TX_OK;
1309}
1310#endif
1311
e4f67add
DS		 1312static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1313{
1314 struct vxlan_dev *vxlan = netdev_priv(dev);
1315 struct neighbour *n;
e4f67add
DS		 1316
1317 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1318 return false;
1319
1320 n = NULL;
1321 switch (ntohs(eth_hdr(skb)->h_proto)) {
1322 case ETH_P_IP:
e15a00aa
CW		 1323 {
1324 struct iphdr *pip;
1325
e4f67add
DS		 1326 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1327 return false;
1328 pip = ip_hdr(skb);
1329 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1330 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1331 union vxlan_addr ipa = {
1332 .sin.sin_addr.s_addr = pip->daddr,
1333 .sa.sa_family = AF_INET,
1334 };
1335
1336 vxlan_ip_miss(dev, &ipa);
1337 return false;
1338 }
1339
e4f67add 1340 break;
e15a00aa
CW		 1341 }
1342#if IS_ENABLED(CONFIG_IPV6)
1343 case ETH_P_IPV6:
1344 {
1345 struct ipv6hdr *pip6;
1346
1347 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1348 return false;
1349 pip6 = ipv6_hdr(skb);
1350 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1351 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1352 union vxlan_addr ipa = {
1353 .sin6.sin6_addr = pip6->daddr,
1354 .sa.sa_family = AF_INET6,
1355 };
1356
1357 vxlan_ip_miss(dev, &ipa);
1358 return false;
1359 }
1360
1361 break;
1362 }
1363#endif
e4f67add
DS		 1364 default:
1365 return false;
1366 }
1367
1368 if (n) {
1369 bool diff;
1370
7367d0b5 1371 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1372 if (diff) {
1373 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1374 dev->addr_len);
1375 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1376 }
1377 neigh_release(n);
1378 return diff;
e4c7ed41
CW		 1379 }
1380
e4f67add
DS		 1381 return false;
1382}
1383
553675fb 1384static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1385{
1386 sock_put(skb->sk);
1387}
1388
1389/* On transmit, associate with the tunnel socket */
49560532 1390static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1cad8715 1391{
1cad8715 1392 skb_orphan(skb);
1393 sock_hold(sk);
1394 skb->sk = sk;
553675fb 1395 skb->destructor = vxlan_sock_put;
1cad8715 1396}
1397
05f47d69 1398/* Compute source port for outgoing packet
1399 * first choice to use L4 flow hash since it will spread
1400 * better and maybe available from hardware
1401 * secondary choice is to use jhash on the Ethernet header
1402 */
49560532 1403__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
05f47d69 1404{
49560532 1405 unsigned int range = (port_max - port_min) + 1;
05f47d69 1406 u32 hash;
1407
3958afa1 1408 hash = skb_get_hash(skb);
05f47d69 1409 if (!hash)
1410 hash = jhash(skb->data, 2 * ETH_ALEN,
1411 (__force u32) skb->protocol);
1412
49560532 1413 return htons((((u64) hash * range) >> 32) + port_min);
05f47d69 1414}
49560532 1415EXPORT_SYMBOL_GPL(vxlan_src_port);
05f47d69 1416
05c0db08
PS		 1417static int handle_offloads(struct sk_buff *skb)
1418{
1419 if (skb_is_gso(skb)) {
1420 int err = skb_unclone(skb, GFP_ATOMIC);
1421 if (unlikely(err))
1422 return err;
1423
f6ace502 1424 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1425 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1426 skb->ip_summed = CHECKSUM_NONE;
1427
1428 return 0;
1429}
1430
e4c7ed41 1431#if IS_ENABLED(CONFIG_IPV6)
11796187 1432static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1433 struct dst_entry *dst, struct sk_buff *skb,
1434 struct net_device *dev, struct in6_addr *saddr,
1435 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1436 __be16 src_port, __be16 dst_port, __be32 vni)
1437{
1438 struct ipv6hdr *ip6h;
1439 struct vxlanhdr *vxh;
1440 struct udphdr *uh;
1441 int min_headroom;
1442 int err;
1443
1444 if (!skb->encapsulation) {
1445 skb_reset_inner_headers(skb);
1446 skb->encapsulation = 1;
1447 }
1448
963a88b3
ND		 1449 skb_scrub_packet(skb, false);
1450
e4c7ed41
CW		 1451 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1452 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1453 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1454
1455 /* Need space for new headers (invalidates iph ptr) */
1456 err = skb_cow_head(skb, min_headroom);
1457 if (unlikely(err))
1458 return err;
1459
1460 if (vlan_tx_tag_present(skb)) {
1461 if (WARN_ON(!__vlan_put_tag(skb,
1462 skb->vlan_proto,
1463 vlan_tx_tag_get(skb))))
1464 return -ENOMEM;
1465
1466 skb->vlan_tci = 0;
1467 }
1468
1469 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1470 vxh->vx_flags = htonl(VXLAN_FLAGS);
1471 vxh->vx_vni = vni;
1472
1473 __skb_push(skb, sizeof(*uh));
1474 skb_reset_transport_header(skb);
1475 uh = udp_hdr(skb);
1476
1477 uh->dest = dst_port;
1478 uh->source = src_port;
1479
1480 uh->len = htons(skb->len);
1481 uh->check = 0;
1482
1483 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1484 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1485 IPSKB_REROUTED);
e4c7ed41
CW		 1486 skb_dst_set(skb, dst);
1487
1488 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1489 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1490 skb->ip_summed = CHECKSUM_UNNECESSARY;
1491 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1492 IPPROTO_UDP, csum);
1493 if (uh->check == 0)
1494 uh->check = CSUM_MANGLED_0;
1495 } else {
1496 skb->ip_summed = CHECKSUM_PARTIAL;
1497 skb->csum_start = skb_transport_header(skb) - skb->head;
1498 skb->csum_offset = offsetof(struct udphdr, check);
1499 uh->check = ~csum_ipv6_magic(saddr, daddr,
1500 skb->len, IPPROTO_UDP, 0);
1501 }
1502
1503 __skb_push(skb, sizeof(*ip6h));
1504 skb_reset_network_header(skb);
1505 ip6h = ipv6_hdr(skb);
1506 ip6h->version = 6;
1507 ip6h->priority = prio;
1508 ip6h->flow_lbl[0] = 0;
1509 ip6h->flow_lbl[1] = 0;
1510 ip6h->flow_lbl[2] = 0;
1511 ip6h->payload_len = htons(skb->len);
1512 ip6h->nexthdr = IPPROTO_UDP;
1513 ip6h->hop_limit = ttl;
1514 ip6h->daddr = *daddr;
1515 ip6h->saddr = *saddr;
1516
1517 vxlan_set_owner(vs->sock->sk, skb);
1518
1519 err = handle_offloads(skb);
1520 if (err)
1521 return err;
1522
1523 ip6tunnel_xmit(skb, dev);
1524 return 0;
1525}
1526#endif
1527
11796187 1528int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1529 struct rtable *rt, struct sk_buff *skb,
1530 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1531 __be16 src_port, __be16 dst_port, __be32 vni)
1532{
1533 struct vxlanhdr *vxh;
1534 struct udphdr *uh;
649c5b8b 1535 int min_headroom;
49560532
PS		 1536 int err;
1537
1538 if (!skb->encapsulation) {
1539 skb_reset_inner_headers(skb);
1540 skb->encapsulation = 1;
1541 }
1542
649c5b8b 1543 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1544 + VXLAN_HLEN + sizeof(struct iphdr)
1545 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1546
1547 /* Need space for new headers (invalidates iph ptr) */
1548 err = skb_cow_head(skb, min_headroom);
1549 if (unlikely(err))
1550 return err;
1551
1eaa8178
PS		 1552 if (vlan_tx_tag_present(skb)) {
1553 if (WARN_ON(!__vlan_put_tag(skb,
1554 skb->vlan_proto,
1555 vlan_tx_tag_get(skb))))
1556 return -ENOMEM;
1557
1558 skb->vlan_tci = 0;
1559 }
1560
49560532
PS		 1561 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1562 vxh->vx_flags = htonl(VXLAN_FLAGS);
1563 vxh->vx_vni = vni;
1564
1565 __skb_push(skb, sizeof(*uh));
1566 skb_reset_transport_header(skb);
1567 uh = udp_hdr(skb);
1568
1569 uh->dest = dst_port;
1570 uh->source = src_port;
1571
1572 uh->len = htons(skb->len);
1573 uh->check = 0;
1574
1575 vxlan_set_owner(vs->sock->sk, skb);
1576
1577 err = handle_offloads(skb);
1578 if (err)
1579 return err;
1580
963a88b3
ND		 1581 return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
1582 false);
49560532
PS		 1583}
1584EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1585
9dcc71e1
SS		 1586/* Bypass encapsulation if the destination is local */
1587static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1588 struct vxlan_dev *dst_vxlan)
1589{
8f84985f 1590 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1591 union vxlan_addr loopback;
1592 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1 1593
8f84985f
LR		 1594 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1595 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1596 skb->pkt_type = PACKET_HOST;
1597 skb->encapsulation = 0;
1598 skb->dev = dst_vxlan->dev;
1599 __skb_pull(skb, skb_network_offset(skb));
1600
e4c7ed41
CW		 1601 if (remote_ip->sa.sa_family == AF_INET) {
1602 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1603 loopback.sa.sa_family = AF_INET;
1604#if IS_ENABLED(CONFIG_IPV6)
1605 } else {
1606 loopback.sin6.sin6_addr = in6addr_loopback;
1607 loopback.sa.sa_family = AF_INET6;
1608#endif
1609 }
1610
9dcc71e1 1611 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1612 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1613
1614 u64_stats_update_begin(&tx_stats->syncp);
1615 tx_stats->tx_packets++;
1616 tx_stats->tx_bytes += skb->len;
1617 u64_stats_update_end(&tx_stats->syncp);
1618
1619 if (netif_rx(skb) == NET_RX_SUCCESS) {
1620 u64_stats_update_begin(&rx_stats->syncp);
1621 rx_stats->rx_packets++;
1622 rx_stats->rx_bytes += skb->len;
1623 u64_stats_update_end(&rx_stats->syncp);
1624 } else {
1625 skb->dev->stats.rx_dropped++;
1626 }
1627}
1628
4ad16930
SH		 1629static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1630 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1631{
1632 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1633 struct rtable *rt = NULL;
d342894c 1634 const struct iphdr *old_iph;
d342894c 1635 struct flowi4 fl4;
e4c7ed41
CW		 1636 union vxlan_addr *dst;
1637 __be16 src_port = 0, dst_port;
234f5b73 1638 u32 vni;
d342894c 1639 __be16 df = 0;
1640 __u8 tos, ttl;
0e6fbc5b 1641 int err;
d342894c 1642
823aa873 1643 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1644 vni = rdst->remote_vni;
e4c7ed41 1645 dst = &rdst->remote_ip;
e4f67add 1646
e4c7ed41 1647 if (vxlan_addr_any(dst)) {
e4f67add 1648 if (did_rsc) {
e4f67add 1649 /* short-circuited back to local bridge */
9dcc71e1 1650 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1651 return;
e4f67add 1652 }
ef59febe 1653 goto drop;
e4f67add 1654 }
ef59febe 1655
d342894c 1656 old_iph = ip_hdr(skb);
1657
d342894c 1658 ttl = vxlan->ttl;
e4c7ed41 1659 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1660 ttl = 1;
1661
1662 tos = vxlan->tos;
1663 if (tos == 1)
206aaafc 1664 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1665
49560532 1666 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
d342894c 1667
e4c7ed41
CW		 1668 if (dst->sa.sa_family == AF_INET) {
1669 memset(&fl4, 0, sizeof(fl4));
1670 fl4.flowi4_oif = rdst->remote_ifindex;
1671 fl4.flowi4_tos = RT_TOS(tos);
1672 fl4.daddr = dst->sin.sin_addr.s_addr;
1673 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1674
1675 rt = ip_route_output_key(dev_net(dev), &fl4);
1676 if (IS_ERR(rt)) {
1677 netdev_dbg(dev, "no route to %pI4\n",
1678 &dst->sin.sin_addr.s_addr);
1679 dev->stats.tx_carrier_errors++;
1680 goto tx_error;
1681 }
d342894c 1682
e4c7ed41
CW		 1683 if (rt->dst.dev == dev) {
1684 netdev_dbg(dev, "circular route to %pI4\n",
1685 &dst->sin.sin_addr.s_addr);
1686 dev->stats.collisions++;
fffc15a5 1687 goto rt_tx_error;
e4c7ed41
CW		 1688 }
1689
1690 /* Bypass encapsulation if the destination is local */
1691 if (rt->rt_flags & RTCF_LOCAL &&
1692 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1693 struct vxlan_dev *dst_vxlan;
1694
1695 ip_rt_put(rt);
1696 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1697 if (!dst_vxlan)
1698 goto tx_error;
1699 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1700 return;
1701 }
1702
1703 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1704 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1705
11796187 1706 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
e4c7ed41
CW		 1707 fl4.saddr, dst->sin.sin_addr.s_addr,
1708 tos, ttl, df, src_port, dst_port,
1709 htonl(vni << 8));
9dcc71e1 1710
e4c7ed41
CW		 1711 if (err < 0)
1712 goto rt_tx_error;
1713 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1714#if IS_ENABLED(CONFIG_IPV6)
1715 } else {
1716 struct sock *sk = vxlan->vn_sock->sock->sk;
1717 struct dst_entry *ndst;
1718 struct flowi6 fl6;
1719 u32 flags;
1720
1721 memset(&fl6, 0, sizeof(fl6));
1722 fl6.flowi6_oif = rdst->remote_ifindex;
1723 fl6.daddr = dst->sin6.sin6_addr;
1724 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1725 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1726
1727 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1728 netdev_dbg(dev, "no route to %pI6\n",
1729 &dst->sin6.sin6_addr);
1730 dev->stats.tx_carrier_errors++;
9dcc71e1 1731 goto tx_error;
e4c7ed41 1732 }
d342894c 1733
e4c7ed41
CW		 1734 if (ndst->dev == dev) {
1735 netdev_dbg(dev, "circular route to %pI6\n",
1736 &dst->sin6.sin6_addr);
1737 dst_release(ndst);
1738 dev->stats.collisions++;
1739 goto tx_error;
1740 }
0e6fbc5b 1741
e4c7ed41
CW		 1742 /* Bypass encapsulation if the destination is local */
1743 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1744 if (flags & RTF_LOCAL &&
1745 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1746 struct vxlan_dev *dst_vxlan;
1747
1748 dst_release(ndst);
1749 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1750 if (!dst_vxlan)
1751 goto tx_error;
1752 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1753 return;
1754 }
49560532 1755
e4c7ed41
CW		 1756 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1757
11796187 1758 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41
CW		 1759 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1760 src_port, dst_port, htonl(vni << 8));
1761#endif
1762 }
0e6fbc5b 1763
4ad16930 1764 return;
d342894c 1765
1766drop:
1767 dev->stats.tx_dropped++;
1768 goto tx_free;
1769
49560532
PS		 1770rt_tx_error:
1771 ip_rt_put(rt);
d342894c 1772tx_error:
1773 dev->stats.tx_errors++;
1774tx_free:
1775 dev_kfree_skb(skb);
d342894c 1776}
1777
6681712d
DS		 1778/* Transmit local packets over Vxlan
1779 *
1780 * Outer IP header inherits ECN and DF from inner header.
1781 * Outer UDP destination is the VXLAN assigned port.
1782 * source port is based on hash of flow
1783 */
1784static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1785{
1786 struct vxlan_dev *vxlan = netdev_priv(dev);
1787 struct ethhdr *eth;
1788 bool did_rsc = false;
afbd8bae 1789 struct vxlan_rdst *rdst;
6681712d 1790 struct vxlan_fdb *f;
6681712d
DS		 1791
1792 skb_reset_mac_header(skb);
1793 eth = eth_hdr(skb);
1794
f564f45c
CW		 1795 if ((vxlan->flags & VXLAN_F_PROXY)) {
1796 if (ntohs(eth->h_proto) == ETH_P_ARP)
1797 return arp_reduce(dev, skb);
1798#if IS_ENABLED(CONFIG_IPV6)
1799 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1800 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1801 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1802 struct nd_msg *msg;
1803
1804 msg = (struct nd_msg *)skb_transport_header(skb);
1805 if (msg->icmph.icmp6_code == 0 &&
1806 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1807 return neigh_reduce(dev, skb);
1808 }
1809#endif
1810 }
6681712d
DS		 1811
1812 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1813 did_rsc = false;
1814
1815 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1816 (ntohs(eth->h_proto) == ETH_P_IP ||
1817 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1818 did_rsc = route_shortcircuit(dev, skb);
1819 if (did_rsc)
1820 f = vxlan_find_mac(vxlan, eth->h_dest);
1821 }
1822
6681712d 1823 if (f == NULL) {
afbd8bae
MR		 1824 f = vxlan_find_mac(vxlan, all_zeros_mac);
1825 if (f == NULL) {
1826 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1827 !is_multicast_ether_addr(eth->h_dest))
1828 vxlan_fdb_miss(vxlan, eth->h_dest);
1829
1830 dev->stats.tx_dropped++;
1831 dev_kfree_skb(skb);
1832 return NETDEV_TX_OK;
1833 }
1834 }
6681712d 1835
afbd8bae
MR		 1836 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1837 struct sk_buff *skb1;
6681712d 1838
afbd8bae
MR		 1839 skb1 = skb_clone(skb, GFP_ATOMIC);
1840 if (skb1)
1841 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1842 }
1843
afbd8bae 1844 dev_kfree_skb(skb);
4ad16930 1845 return NETDEV_TX_OK;
6681712d
DS		 1846}
1847
d342894c 1848/* Walk the forwarding table and purge stale entries */
1849static void vxlan_cleanup(unsigned long arg)
1850{
1851 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1852 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1853 unsigned int h;
1854
1855 if (!netif_running(vxlan->dev))
1856 return;
1857
1858 spin_lock_bh(&vxlan->hash_lock);
1859 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1860 struct hlist_node *p, *n;
1861 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1862 struct vxlan_fdb *f
1863 = container_of(p, struct vxlan_fdb, hlist);
1864 unsigned long timeout;
1865
3c172868 1866 if (f->state & NUD_PERMANENT)
d342894c 1867 continue;
1868
1869 timeout = f->used + vxlan->age_interval * HZ;
1870 if (time_before_eq(timeout, jiffies)) {
1871 netdev_dbg(vxlan->dev,
1872 "garbage collect %pM\n",
1873 f->eth_addr);
1874 f->state = NUD_STALE;
1875 vxlan_fdb_destroy(vxlan, f);
1876 } else if (time_before(timeout, next_timer))
1877 next_timer = timeout;
1878 }
1879 }
1880 spin_unlock_bh(&vxlan->hash_lock);
1881
1882 mod_timer(&vxlan->age_timer, next_timer);
1883}
1884
9c2e24e1
PS		 1885static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1886{
1887 __u32 vni = vxlan->default_dst.remote_vni;
1888
1889 vxlan->vn_sock = vs;
1890 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1891}
1892
d342894c 1893/* Setup stats when device is created */
1894static int vxlan_init(struct net_device *dev)
1895{
1c51a915
SH		 1896 struct vxlan_dev *vxlan = netdev_priv(dev);
1897 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1898 struct vxlan_sock *vs;
827da44c 1899 int i;
1c51a915 1900
8f84985f 1901 dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
e8171045 1902 if (!dev->tstats)
d342894c 1903 return -ENOMEM;
1904
827da44c 1905 for_each_possible_cpu(i) {
8f84985f 1906 struct pcpu_sw_netstats *vxlan_stats;
827da44c
JS		 1907 vxlan_stats = per_cpu_ptr(dev->tstats, i);
1908 u64_stats_init(&vxlan_stats->syncp);
1909 }
1910
1911
1c51a915 1912 spin_lock(&vn->sock_lock);
9c2e24e1 1913 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1c51a915
SH		 1914 if (vs) {
1915 /* If we have a socket with same port already, reuse it */
1916 atomic_inc(&vs->refcnt);
9c2e24e1 1917 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1918 } else {
1919 /* otherwise make new socket outside of RTNL */
1920 dev_hold(dev);
1921 queue_work(vxlan_wq, &vxlan->sock_work);
1922 }
1923 spin_unlock(&vn->sock_lock);
1924
d342894c 1925 return 0;
1926}
1927
ba609e9b 1928static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1929{
1930 struct vxlan_fdb *f;
1931
1932 spin_lock_bh(&vxlan->hash_lock);
1933 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1934 if (f)
1935 vxlan_fdb_destroy(vxlan, f);
1936 spin_unlock_bh(&vxlan->hash_lock);
1937}
1938
ebf4063e
SH		 1939static void vxlan_uninit(struct net_device *dev)
1940{
1941 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 1942 struct vxlan_sock *vs = vxlan->vn_sock;
1943
ba609e9b 1944 vxlan_fdb_delete_default(vxlan);
afbd8bae 1945
ebf4063e 1946 if (vs)
012a5729 1947 vxlan_sock_release(vs);
ebf4063e
SH		 1948 free_percpu(dev->tstats);
1949}
1950
d342894c 1951/* Start ageing timer and join group when device is brought up */
1952static int vxlan_open(struct net_device *dev)
1953{
1954 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1955 struct vxlan_sock *vs = vxlan->vn_sock;
1956
1957 /* socket hasn't been created */
1958 if (!vs)
1959 return -ENOTCONN;
d342894c 1960
79d4a94f 1961 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 1962 vxlan_sock_hold(vs);
7c47cedf 1963 dev_hold(dev);
3fc2de2f 1964 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 1965 }
1966
1967 if (vxlan->age_interval)
1968 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1969
1970 return 0;
1971}
1972
1973/* Purge the forwarding table */
1974static void vxlan_flush(struct vxlan_dev *vxlan)
1975{
31fec5aa 1976 unsigned int h;
d342894c 1977
1978 spin_lock_bh(&vxlan->hash_lock);
1979 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1980 struct hlist_node *p, *n;
1981 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1982 struct vxlan_fdb *f
1983 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 1984 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1985 if (!is_zero_ether_addr(f->eth_addr))
1986 vxlan_fdb_destroy(vxlan, f);
d342894c 1987 }
1988 }
1989 spin_unlock_bh(&vxlan->hash_lock);
1990}
1991
1992/* Cleanup timer and forwarding table on shutdown */
1993static int vxlan_stop(struct net_device *dev)
1994{
3fc2de2f 1995 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1996 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1997 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1998
e4c7ed41 1999 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2000 !vxlan_group_used(vn, vxlan)) {
1c51a915 2001 vxlan_sock_hold(vs);
7c47cedf 2002 dev_hold(dev);
3fc2de2f 2003 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2004 }
d342894c 2005
2006 del_timer_sync(&vxlan->age_timer);
2007
2008 vxlan_flush(vxlan);
2009
2010 return 0;
2011}
2012
d342894c 2013/* Stub, nothing needs to be done. */
2014static void vxlan_set_multicast_list(struct net_device *dev)
2015{
2016}
2017
345010b5
DB		 2018static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2019{
2020 struct vxlan_dev *vxlan = netdev_priv(dev);
2021 struct vxlan_rdst *dst = &vxlan->default_dst;
2022 struct net_device *lowerdev;
2023 int max_mtu;
2024
2025 lowerdev = __dev_get_by_index(dev_net(dev), dst->remote_ifindex);
2026 if (lowerdev == NULL)
2027 return eth_change_mtu(dev, new_mtu);
2028
2029 if (dst->remote_ip.sa.sa_family == AF_INET6)
2030 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2031 else
2032 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2033
2034 if (new_mtu < 68 || new_mtu > max_mtu)
2035 return -EINVAL;
2036
2037 dev->mtu = new_mtu;
2038 return 0;
2039}
2040
d342894c 2041static const struct net_device_ops vxlan_netdev_ops = {
2042 .ndo_init = vxlan_init,
ebf4063e 2043 .ndo_uninit = vxlan_uninit,
d342894c 2044 .ndo_open = vxlan_open,
2045 .ndo_stop = vxlan_stop,
2046 .ndo_start_xmit = vxlan_xmit,
e8171045 2047 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2048 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2049 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2050 .ndo_validate_addr = eth_validate_addr,
2051 .ndo_set_mac_address = eth_mac_addr,
2052 .ndo_fdb_add = vxlan_fdb_add,
2053 .ndo_fdb_del = vxlan_fdb_delete,
2054 .ndo_fdb_dump = vxlan_fdb_dump,
2055};
2056
2057/* Info for udev, that this is a virtual tunnel endpoint */
2058static struct device_type vxlan_type = {
2059 .name = "vxlan",
2060};
2061
53cf5275 2062/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2063 * supply the listening VXLAN udp ports. Callers are expected
2064 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2065 */
2066void vxlan_get_rx_port(struct net_device *dev)
2067{
2068 struct vxlan_sock *vs;
2069 struct net *net = dev_net(dev);
2070 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2071 sa_family_t sa_family;
35e42379
JG		 2072 __be16 port;
2073 unsigned int i;
53cf5275
JG		 2074
2075 spin_lock(&vn->sock_lock);
2076 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2077 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2078 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2079 sa_family = vs->sock->sk->sk_family;
2080 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2081 port);
2082 }
2083 }
2084 spin_unlock(&vn->sock_lock);
2085}
2086EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2087
d342894c 2088/* Initialize the device structure. */
2089static void vxlan_setup(struct net_device *dev)
2090{
2091 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2092 unsigned int h;
05f47d69 2093 int low, high;
d342894c 2094
2095 eth_hw_addr_random(dev);
2096 ether_setup(dev);
e4c7ed41
CW		 2097 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2098 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
2099 else
2100 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2101
2102 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2103 dev->destructor = free_netdev;
d342894c 2104 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2105
2106 dev->tx_queue_len = 0;
2107 dev->features |= NETIF_F_LLTX;
2108 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 2109 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2110 dev->features |= NETIF_F_RXCSUM;
05c0db08 2111 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2112
1eaa8178
PS		 2113 dev->vlan_features = dev->features;
2114 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2115 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2116 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2117 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2118 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2119 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2120
553675fb 2121 INIT_LIST_HEAD(&vxlan->next);
d342894c 2122 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2123 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2124 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2125 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2126
2127 init_timer_deferrable(&vxlan->age_timer);
2128 vxlan->age_timer.function = vxlan_cleanup;
2129 vxlan->age_timer.data = (unsigned long) vxlan;
2130
0bbf87d8 2131 inet_get_local_port_range(dev_net(dev), &low, &high);
05f47d69 2132 vxlan->port_min = low;
2133 vxlan->port_max = high;
823aa873 2134 vxlan->dst_port = htons(vxlan_port);
05f47d69 2135
d342894c 2136 vxlan->dev = dev;
2137
2138 for (h = 0; h < FDB_HASH_SIZE; ++h)
2139 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2140}
2141
2142static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2143 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2144 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2145 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2146 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2147 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2148 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2149 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2150 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2151 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2152 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2153 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2154 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2155 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2156 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2157 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2158 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2159 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2160};
2161
2162static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2163{
2164 if (tb[IFLA_ADDRESS]) {
2165 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2166 pr_debug("invalid link address (not ethernet)\n");
2167 return -EINVAL;
2168 }
2169
2170 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2171 pr_debug("invalid all zero ethernet address\n");
2172 return -EADDRNOTAVAIL;
2173 }
2174 }
2175
2176 if (!data)
2177 return -EINVAL;
2178
2179 if (data[IFLA_VXLAN_ID]) {
2180 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2181 if (id >= VXLAN_VID_MASK)
2182 return -ERANGE;
2183 }
2184
05f47d69 2185 if (data[IFLA_VXLAN_PORT_RANGE]) {
2186 const struct ifla_vxlan_port_range *p
2187 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2188
2189 if (ntohs(p->high) < ntohs(p->low)) {
2190 pr_debug("port range %u .. %u not valid\n",
2191 ntohs(p->low), ntohs(p->high));
2192 return -EINVAL;
2193 }
2194 }
2195
d342894c 2196 return 0;
2197}
2198
1b13c97f
YB		 2199static void vxlan_get_drvinfo(struct net_device *netdev,
2200 struct ethtool_drvinfo *drvinfo)
2201{
2202 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2203 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2204}
2205
2206static const struct ethtool_ops vxlan_ethtool_ops = {
2207 .get_drvinfo = vxlan_get_drvinfo,
2208 .get_link = ethtool_op_get_link,
2209};
2210
553675fb 2211static void vxlan_del_work(struct work_struct *work)
2212{
2213 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2214
2215 sk_release_kernel(vs->sock->sk);
2216 kfree_rcu(vs, rcu);
2217}
2218
e4c7ed41
CW		 2219#if IS_ENABLED(CONFIG_IPV6)
2220/* Create UDP socket for encapsulation receive. AF_INET6 socket
2221 * could be used for both IPv4 and IPv6 communications, but
2222 * users may set bindv6only=1.
2223 */
39deb2c7 2224static struct socket *create_v6_sock(struct net *net, __be16 port)
553675fb 2225{
553675fb 2226 struct sock *sk;
e4c7ed41
CW		 2227 struct socket *sock;
2228 struct sockaddr_in6 vxlan_addr = {
2229 .sin6_family = AF_INET6,
2230 .sin6_port = port,
2231 };
2232 int rc, val = 1;
2233
2234 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2235 if (rc < 0) {
2236 pr_debug("UDPv6 socket create failed\n");
39deb2c7 2237 return ERR_PTR(rc);
e4c7ed41
CW		 2238 }
2239
2240 /* Put in proper namespace */
2241 sk = sock->sk;
2242 sk_change_net(sk, net);
2243
2244 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2245 (char *)&val, sizeof(val));
2246 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2247 sizeof(struct sockaddr_in6));
2248 if (rc < 0) {
2249 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2250 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2251 sk_release_kernel(sk);
39deb2c7 2252 return ERR_PTR(rc);
e4c7ed41
CW		 2253 }
2254 /* At this point, IPv6 module should have been loaded in
2255 * sock_create_kern().
2256 */
2257 BUG_ON(!ipv6_stub);
2258
e4c7ed41
CW		 2259 /* Disable multicast loopback */
2260 inet_sk(sk)->mc_loop = 0;
39deb2c7 2261 return sock;
e4c7ed41
CW		 2262}
2263
2264#else
2265
39deb2c7 2266static struct socket *create_v6_sock(struct net *net, __be16 port)
e4c7ed41 2267{
39deb2c7 2268 return ERR_PTR(-EPFNOSUPPORT);
e4c7ed41
CW		 2269}
2270#endif
2271
39deb2c7 2272static struct socket *create_v4_sock(struct net *net, __be16 port)
e4c7ed41
CW		 2273{
2274 struct sock *sk;
2275 struct socket *sock;
553675fb 2276 struct sockaddr_in vxlan_addr = {
2277 .sin_family = AF_INET,
2278 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2279 .sin_port = port,
553675fb 2280 };
2281 int rc;
553675fb 2282
2283 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2284 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2285 if (rc < 0) {
2286 pr_debug("UDP socket create failed\n");
39deb2c7 2287 return ERR_PTR(rc);
553675fb 2288 }
2289
2290 /* Put in proper namespace */
e4c7ed41 2291 sk = sock->sk;
553675fb 2292 sk_change_net(sk, net);
2293
e4c7ed41 2294 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2295 sizeof(vxlan_addr));
2296 if (rc < 0) {
2297 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2298 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2299 sk_release_kernel(sk);
39deb2c7 2300 return ERR_PTR(rc);
e4c7ed41
CW		 2301 }
2302
e4c7ed41
CW		 2303 /* Disable multicast loopback */
2304 inet_sk(sk)->mc_loop = 0;
39deb2c7 2305 return sock;
e4c7ed41
CW		 2306}
2307
2308/* Create new listen socket if needed */
2309static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2310 vxlan_rcv_t *rcv, void *data, bool ipv6)
2311{
2312 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2313 struct vxlan_sock *vs;
2314 struct socket *sock;
2315 struct sock *sk;
e4c7ed41
CW		 2316 unsigned int h;
2317
2318 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2319 if (!vs)
2320 return ERR_PTR(-ENOMEM);
2321
2322 for (h = 0; h < VNI_HASH_SIZE; ++h)
2323 INIT_HLIST_HEAD(&vs->vni_list[h]);
2324
2325 INIT_WORK(&vs->del_work, vxlan_del_work);
2326
2327 if (ipv6)
39deb2c7 2328 sock = create_v6_sock(net, port);
e4c7ed41 2329 else
39deb2c7
ZYW		 2330 sock = create_v4_sock(net, port);
2331 if (IS_ERR(sock)) {
553675fb 2332 kfree(vs);
e50fddc8 2333 return ERR_CAST(sock);
553675fb 2334 }
e4c7ed41
CW		 2335
2336 vs->sock = sock;
2337 sk = sock->sk;
9c2e24e1 2338 atomic_set(&vs->refcnt, 1);
5cfccc5a 2339 vs->rcv = rcv;
012a5729 2340 vs->data = data;
559835ea 2341 rcu_assign_sk_user_data(vs->sock->sk, vs);
553675fb 2342
9c2e24e1
PS		 2343 spin_lock(&vn->sock_lock);
2344 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
53cf5275 2345 vxlan_notify_add_rx_port(sk);
9c2e24e1 2346 spin_unlock(&vn->sock_lock);
553675fb 2347
2348 /* Mark socket as an encapsulation socket. */
2349 udp_sk(sk)->encap_type = 1;
2350 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2351#if IS_ENABLED(CONFIG_IPV6)
2352 if (ipv6)
2353 ipv6_stub->udpv6_encap_enable();
2354 else
2355#endif
2356 udp_encap_enable();
2357
9c2e24e1
PS		 2358 return vs;
2359}
2360
012a5729
PS		 2361struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2362 vxlan_rcv_t *rcv, void *data,
e4c7ed41 2363 bool no_share, bool ipv6)
9c2e24e1
PS		 2364{
2365 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2366 struct vxlan_sock *vs;
2367
e4c7ed41 2368 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
9c2e24e1
PS		 2369 if (!IS_ERR(vs))
2370 return vs;
553675fb 2371
012a5729
PS		 2372 if (no_share) /* Return error if sharing is not allowed. */
2373 return vs;
2374
9c2e24e1
PS		 2375 spin_lock(&vn->sock_lock);
2376 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2377 if (vs) {
2378 if (vs->rcv == rcv)
2379 atomic_inc(&vs->refcnt);
2380 else
2381 vs = ERR_PTR(-EBUSY);
2382 }
2383 spin_unlock(&vn->sock_lock);
2384
2385 if (!vs)
9c2e24e1
PS		 2386 vs = ERR_PTR(-EINVAL);
2387
553675fb 2388 return vs;
2389}
012a5729 2390EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2391
1c51a915
SH		 2392/* Scheduled at device creation to bind to a socket */
2393static void vxlan_sock_work(struct work_struct *work)
2394{
9c2e24e1
PS		 2395 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2396 struct net *net = dev_net(vxlan->dev);
1c51a915 2397 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2398 __be16 port = vxlan->dst_port;
2399 struct vxlan_sock *nvs;
1c51a915 2400
e4c7ed41 2401 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
1c51a915 2402 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2403 if (!IS_ERR(nvs))
2404 vxlan_vs_add_dev(nvs, vxlan);
2405 spin_unlock(&vn->sock_lock);
2406
2407 dev_put(vxlan->dev);
1c51a915
SH		 2408}
2409
d342894c 2410static int vxlan_newlink(struct net *net, struct net_device *dev,
2411 struct nlattr *tb[], struct nlattr *data[])
2412{
553675fb 2413 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2414 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2415 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2416 __u32 vni;
2417 int err;
e4c7ed41 2418 bool use_ipv6 = false;
d342894c 2419
2420 if (!data[IFLA_VXLAN_ID])
2421 return -EINVAL;
2422
2423 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2424 dst->remote_vni = vni;
d342894c 2425
e4c7ed41
CW		 2426 if (data[IFLA_VXLAN_GROUP]) {
2427 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2428 dst->remote_ip.sa.sa_family = AF_INET;
2429 } else if (data[IFLA_VXLAN_GROUP6]) {
2430 if (!IS_ENABLED(CONFIG_IPV6))
2431 return -EPFNOSUPPORT;
2432
2433 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2434 sizeof(struct in6_addr));
2435 dst->remote_ip.sa.sa_family = AF_INET6;
2436 use_ipv6 = true;
2437 }
d342894c 2438
e4c7ed41
CW		 2439 if (data[IFLA_VXLAN_LOCAL]) {
2440 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2441 vxlan->saddr.sa.sa_family = AF_INET;
2442 } else if (data[IFLA_VXLAN_LOCAL6]) {
2443 if (!IS_ENABLED(CONFIG_IPV6))
2444 return -EPFNOSUPPORT;
2445
2446 /* TODO: respect scope id */
2447 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2448 sizeof(struct in6_addr));
2449 vxlan->saddr.sa.sa_family = AF_INET6;
2450 use_ipv6 = true;
2451 }
d342894c 2452
34e02aa1 2453 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2454 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2455 struct net_device *lowerdev
c7995c43 2456 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2457
2458 if (!lowerdev) {
c7995c43 2459 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2460 return -ENODEV;
2461 }
d342894c 2462
e4c7ed41
CW		 2463#if IS_ENABLED(CONFIG_IPV6)
2464 if (use_ipv6) {
2465 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2466 if (idev && idev->cnf.disable_ipv6) {
2467 pr_info("IPv6 is disabled via sysctl\n");
2468 return -EPERM;
2469 }
2470 vxlan->flags |= VXLAN_F_IPV6;
2471 }
2472#endif
2473
34e02aa1 2474 if (!tb[IFLA_MTU])
e4c7ed41 2475 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4
AD		 2476
2477 /* update header length based on lower device */
2478 dev->hard_header_len = lowerdev->hard_header_len +
e4c7ed41 2479 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
d342894c 2480 }
2481
2482 if (data[IFLA_VXLAN_TOS])
2483 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2484
afb97186
VB		 2485 if (data[IFLA_VXLAN_TTL])
2486 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2487
d342894c 2488 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2489 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2490
2491 if (data[IFLA_VXLAN_AGEING])
2492 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2493 else
2494 vxlan->age_interval = FDB_AGE_DEFAULT;
2495
e4f67add
DS		 2496 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2497 vxlan->flags |= VXLAN_F_PROXY;
2498
2499 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2500 vxlan->flags |= VXLAN_F_RSC;
2501
2502 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2503 vxlan->flags |= VXLAN_F_L2MISS;
2504
2505 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2506 vxlan->flags |= VXLAN_F_L3MISS;
2507
d342894c 2508 if (data[IFLA_VXLAN_LIMIT])
2509 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2510
05f47d69 2511 if (data[IFLA_VXLAN_PORT_RANGE]) {
2512 const struct ifla_vxlan_port_range *p
2513 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2514 vxlan->port_min = ntohs(p->low);
2515 vxlan->port_max = ntohs(p->high);
2516 }
2517
823aa873 2518 if (data[IFLA_VXLAN_PORT])
2519 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2520
553675fb 2521 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2522 pr_info("duplicate VNI %u\n", vni);
2523 return -EEXIST;
2524 }
2525
1b13c97f
YB		 2526 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2527
2936b6ab
SS		 2528 /* create an fdb entry for a valid default destination */
2529 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2530 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2531 &vxlan->default_dst.remote_ip,
2532 NUD_REACHABLE|NUD_PERMANENT,
2533 NLM_F_EXCL|NLM_F_CREATE,
2534 vxlan->dst_port,
2535 vxlan->default_dst.remote_vni,
2536 vxlan->default_dst.remote_ifindex,
2537 NTF_SELF);
2538 if (err)
2539 return err;
2540 }
d342894c 2541
afbd8bae
MR		 2542 err = register_netdevice(dev);
2543 if (err) {
ba609e9b 2544 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2545 return err;
2546 }
2547
553675fb 2548 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2549
2550 return 0;
d342894c 2551}
2552
2553static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2554{
fe5c3561 2555 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 2556 struct vxlan_dev *vxlan = netdev_priv(dev);
2557
fe5c3561 2558 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2559 if (!hlist_unhashed(&vxlan->hlist))
2560 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2561 spin_unlock(&vn->sock_lock);
2562
553675fb 2563 list_del(&vxlan->next);
d342894c 2564 unregister_netdevice_queue(dev, head);
2565}
2566
2567static size_t vxlan_get_size(const struct net_device *dev)
2568{
2569
2570 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2571 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2572 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2573 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2574 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2575 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2576 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2577 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2578 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2579 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2580 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2581 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2582 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2583 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 2584 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 2585 0;
2586}
2587
2588static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2589{
2590 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2591 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2592 struct ifla_vxlan_port_range ports = {
2593 .low = htons(vxlan->port_min),
2594 .high = htons(vxlan->port_max),
2595 };
d342894c 2596
c7995c43 2597 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2598 goto nla_put_failure;
2599
e4c7ed41
CW		 2600 if (!vxlan_addr_any(&dst->remote_ip)) {
2601 if (dst->remote_ip.sa.sa_family == AF_INET) {
2602 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2603 dst->remote_ip.sin.sin_addr.s_addr))
2604 goto nla_put_failure;
2605#if IS_ENABLED(CONFIG_IPV6)
2606 } else {
2607 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2608 &dst->remote_ip.sin6.sin6_addr))
2609 goto nla_put_failure;
2610#endif
2611 }
2612 }
d342894c 2613
c7995c43 2614 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2615 goto nla_put_failure;
2616
e4c7ed41
CW		 2617 if (!vxlan_addr_any(&vxlan->saddr)) {
2618 if (vxlan->saddr.sa.sa_family == AF_INET) {
2619 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2620 vxlan->saddr.sin.sin_addr.s_addr))
2621 goto nla_put_failure;
2622#if IS_ENABLED(CONFIG_IPV6)
2623 } else {
2624 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2625 &vxlan->saddr.sin6.sin6_addr))
2626 goto nla_put_failure;
2627#endif
2628 }
2629 }
d342894c 2630
2631 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2632 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2633 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2634 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2635 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2636 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2637 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2638 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2639 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2640 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2641 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2642 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2643 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2644 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 2645 goto nla_put_failure;
2646
05f47d69 2647 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2648 goto nla_put_failure;
2649
d342894c 2650 return 0;
2651
2652nla_put_failure:
2653 return -EMSGSIZE;
2654}
2655
2656static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2657 .kind = "vxlan",
2658 .maxtype = IFLA_VXLAN_MAX,
2659 .policy = vxlan_policy,
2660 .priv_size = sizeof(struct vxlan_dev),
2661 .setup = vxlan_setup,
2662 .validate = vxlan_validate,
2663 .newlink = vxlan_newlink,
2664 .dellink = vxlan_dellink,
2665 .get_size = vxlan_get_size,
2666 .fill_info = vxlan_fill_info,
2667};
2668
2669static __net_init int vxlan_init_net(struct net *net)
2670{
2671 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2672 unsigned int h;
d342894c 2673
553675fb 2674 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2675 spin_lock_init(&vn->sock_lock);
d342894c 2676
553675fb 2677 for (h = 0; h < PORT_HASH_SIZE; ++h)
2678 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2679
2680 return 0;
2681}
2682
2683static __net_exit void vxlan_exit_net(struct net *net)
2684{
2685 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 2686 struct vxlan_dev *vxlan;
372675a4 2687 LIST_HEAD(list);
9cb6cb7e
ZM		 2688
2689 rtnl_lock();
553675fb 2690 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 2691 unregister_netdevice_queue(vxlan->dev, &list);
2692 unregister_netdevice_many(&list);
9cb6cb7e 2693 rtnl_unlock();
d342894c 2694}
2695
2696static struct pernet_operations vxlan_net_ops = {
2697 .init = vxlan_init_net,
2698 .exit = vxlan_exit_net,
2699 .id = &vxlan_net_id,
2700 .size = sizeof(struct vxlan_net),
2701};
2702
2703static int __init vxlan_init_module(void)
2704{
2705 int rc;
2706
758c57d1
SH		 2707 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2708 if (!vxlan_wq)
2709 return -ENOMEM;
2710
d342894c 2711 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2712
2713 rc = register_pernet_device(&vxlan_net_ops);
2714 if (rc)
2715 goto out1;
2716
2717 rc = rtnl_link_register(&vxlan_link_ops);
2718 if (rc)
2719 goto out2;
2720
2721 return 0;
2722
2723out2:
2724 unregister_pernet_device(&vxlan_net_ops);
2725out1:
758c57d1 2726 destroy_workqueue(vxlan_wq);
d342894c 2727 return rc;
2728}
7332a13b 2729late_initcall(vxlan_init_module);
d342894c 2730
2731static void __exit vxlan_cleanup_module(void)
2732{
b7153984 2733 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 2734 destroy_workqueue(vxlan_wq);
f89e57c4 2735 unregister_pernet_device(&vxlan_net_ops);
6681712d 2736 rcu_barrier();
d342894c 2737}
2738module_exit(vxlan_cleanup_module);
2739
2740MODULE_LICENSE("GPL");
2741MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2742MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 2743MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Hirsute Linux kernel mirror