]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: net_secret should not depend on TCP
[mirror_ubuntu-jammy-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)
52b702ff
AD		 63/* IP header + UDP + VXLAN + Ethernet header */
64#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
e4c7ed41
CW		 65/* IPv6 header + UDP + VXLAN + Ethernet header */
66#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
7ce04758 67#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
d342894c 68
69#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
70
71/* VXLAN protocol header */
72struct vxlanhdr {
73 __be32 vx_flags;
74 __be32 vx_vni;
75};
76
23c578bf 77/* UDP port for VXLAN traffic.
78 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 79 * for compatibility with early adopters.
23c578bf 80 */
9daaa397
SH		 81static unsigned short vxlan_port __read_mostly = 8472;
82module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 83MODULE_PARM_DESC(udp_port, "Destination UDP port");
84
85static bool log_ecn_error = true;
86module_param(log_ecn_error, bool, 0644);
87MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
88
60d9d4c6 89static int vxlan_net_id;
553675fb 90
afbd8bae
MR		 91static const u8 all_zeros_mac[ETH_ALEN];
92
553675fb 93/* per-network namespace private data for this module */
94struct vxlan_net {
95 struct list_head vxlan_list;
96 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 97 spinlock_t sock_lock;
553675fb 98};
99
e4c7ed41
CW		 100union vxlan_addr {
101 struct sockaddr_in sin;
102 struct sockaddr_in6 sin6;
103 struct sockaddr sa;
104};
105
6681712d 106struct vxlan_rdst {
e4c7ed41 107 union vxlan_addr remote_ip;
6681712d
DS		 108 __be16 remote_port;
109 u32 remote_vni;
110 u32 remote_ifindex;
3e61aa8f 111 struct list_head list;
bc7892ba 112 struct rcu_head rcu;
6681712d
DS		 113};
114
d342894c 115/* Forwarding table entry */
116struct vxlan_fdb {
117 struct hlist_node hlist; /* linked list of entries */
118 struct rcu_head rcu;
119 unsigned long updated; /* jiffies */
120 unsigned long used;
3e61aa8f 121 struct list_head remotes;
d342894c 122 u16 state; /* see ndm_state */
ae884082 123 u8 flags; /* see ndm_flags */
d342894c 124 u8 eth_addr[ETH_ALEN];
125};
126
d342894c 127/* Pseudo network device */
128struct vxlan_dev {
553675fb 129 struct hlist_node hlist; /* vni hash table */
130 struct list_head next; /* vxlan's per namespace list */
131 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 132 struct net_device *dev;
c7995c43 133 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 134 union vxlan_addr saddr; /* source address */
823aa873 135 __be16 dst_port;
05f47d69 136 __u16 port_min; /* source port range */
137 __u16 port_max;
d342894c 138 __u8 tos; /* TOS override */
139 __u8 ttl;
e4f67add 140 u32 flags; /* VXLAN_F_* below */
d342894c 141
1c51a915 142 struct work_struct sock_work;
3fc2de2f 143 struct work_struct igmp_join;
144 struct work_struct igmp_leave;
1c51a915 145
d342894c 146 unsigned long age_interval;
147 struct timer_list age_timer;
148 spinlock_t hash_lock;
149 unsigned int addrcnt;
150 unsigned int addrmax;
d342894c 151
152 struct hlist_head fdb_head[FDB_HASH_SIZE];
153};
154
e4f67add
DS		 155#define VXLAN_F_LEARN 0x01
156#define VXLAN_F_PROXY 0x02
157#define VXLAN_F_RSC 0x04
158#define VXLAN_F_L2MISS 0x08
159#define VXLAN_F_L3MISS 0x10
e4c7ed41 160#define VXLAN_F_IPV6 0x20 /* internal flag */
e4f67add 161
d342894c 162/* salt for hash table */
163static u32 vxlan_salt __read_mostly;
758c57d1 164static struct workqueue_struct *vxlan_wq;
d342894c 165
1c51a915
SH		 166static void vxlan_sock_work(struct work_struct *work);
167
e4c7ed41
CW		 168#if IS_ENABLED(CONFIG_IPV6)
169static inline
170bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
171{
172 if (a->sa.sa_family != b->sa.sa_family)
173 return false;
174 if (a->sa.sa_family == AF_INET6)
175 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
176 else
177 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
178}
179
180static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
181{
182 if (ipa->sa.sa_family == AF_INET6)
183 return ipv6_addr_any(&ipa->sin6.sin6_addr);
184 else
185 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
186}
187
188static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
189{
190 if (ipa->sa.sa_family == AF_INET6)
191 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
192 else
193 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
194}
195
196static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
197{
198 if (nla_len(nla) >= sizeof(struct in6_addr)) {
199 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
200 ip->sa.sa_family = AF_INET6;
201 return 0;
202 } else if (nla_len(nla) >= sizeof(__be32)) {
203 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
204 ip->sa.sa_family = AF_INET;
205 return 0;
206 } else {
207 return -EAFNOSUPPORT;
208 }
209}
210
211static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
212 const union vxlan_addr *ip)
213{
214 if (ip->sa.sa_family == AF_INET6)
215 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
216 else
217 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
218}
219
220#else /* !CONFIG_IPV6 */
221
222static inline
223bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
224{
225 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
226}
227
228static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
229{
230 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
231}
232
233static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
234{
235 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
236}
237
238static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
239{
240 if (nla_len(nla) >= sizeof(struct in6_addr)) {
241 return -EAFNOSUPPORT;
242 } else if (nla_len(nla) >= sizeof(__be32)) {
243 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
244 ip->sa.sa_family = AF_INET;
245 return 0;
246 } else {
247 return -EAFNOSUPPORT;
248 }
249}
250
251static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
252 const union vxlan_addr *ip)
253{
254 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
255}
256#endif
257
553675fb 258/* Virtual Network hash table head */
259static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
260{
261 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
262}
263
264/* Socket hash table head */
265static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 266{
267 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
268
553675fb 269 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
270}
271
3e61aa8f
SH		 272/* First remote destination for a forwarding entry.
273 * Guaranteed to be non-NULL because remotes are never deleted.
274 */
5ca5461c 275static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 276{
5ca5461c 277 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
278}
279
280static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
281{
282 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 283}
284
553675fb 285/* Find VXLAN socket based on network namespace and UDP port */
9c2e24e1 286static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
553675fb 287{
288 struct vxlan_sock *vs;
289
290 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
291 if (inet_sk(vs->sock->sk)->inet_sport == port)
292 return vs;
293 }
294 return NULL;
d342894c 295}
296
5cfccc5a 297static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 298{
299 struct vxlan_dev *vxlan;
d342894c 300
553675fb 301 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 302 if (vxlan->default_dst.remote_vni == id)
d342894c 303 return vxlan;
304 }
305
306 return NULL;
307}
308
5cfccc5a
PS		 309/* Look up VNI in a per net namespace table */
310static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
311{
312 struct vxlan_sock *vs;
313
314 vs = vxlan_find_sock(net, port);
315 if (!vs)
316 return NULL;
317
318 return vxlan_vs_find_vni(vs, id);
319}
320
d342894c 321/* Fill in neighbour message in skbuff. */
322static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 323 const struct vxlan_fdb *fdb,
324 u32 portid, u32 seq, int type, unsigned int flags,
325 const struct vxlan_rdst *rdst)
d342894c 326{
327 unsigned long now = jiffies;
328 struct nda_cacheinfo ci;
329 struct nlmsghdr *nlh;
330 struct ndmsg *ndm;
e4f67add 331 bool send_ip, send_eth;
d342894c 332
333 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
334 if (nlh == NULL)
335 return -EMSGSIZE;
336
337 ndm = nlmsg_data(nlh);
338 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 339
340 send_eth = send_ip = true;
341
342 if (type == RTM_GETNEIGH) {
343 ndm->ndm_family = AF_INET;
e4c7ed41 344 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 345 send_eth = !is_zero_ether_addr(fdb->eth_addr);
346 } else
347 ndm->ndm_family = AF_BRIDGE;
d342894c 348 ndm->ndm_state = fdb->state;
349 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 350 ndm->ndm_flags = fdb->flags;
d342894c 351 ndm->ndm_type = NDA_DST;
352
e4f67add 353 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 354 goto nla_put_failure;
355
e4c7ed41 356 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 357 goto nla_put_failure;
358
823aa873 359 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 360 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
361 goto nla_put_failure;
c7995c43 362 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 363 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 364 goto nla_put_failure;
365 if (rdst->remote_ifindex &&
366 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 367 goto nla_put_failure;
368
369 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
370 ci.ndm_confirmed = 0;
371 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
372 ci.ndm_refcnt = 0;
373
374 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
375 goto nla_put_failure;
376
377 return nlmsg_end(skb, nlh);
378
379nla_put_failure:
380 nlmsg_cancel(skb, nlh);
381 return -EMSGSIZE;
382}
383
384static inline size_t vxlan_nlmsg_size(void)
385{
386 return NLMSG_ALIGN(sizeof(struct ndmsg))
387 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 388 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 389 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 390 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
391 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 392 + nla_total_size(sizeof(struct nda_cacheinfo));
393}
394
395static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
3e61aa8f 396 struct vxlan_fdb *fdb, int type)
d342894c 397{
398 struct net *net = dev_net(vxlan->dev);
399 struct sk_buff *skb;
400 int err = -ENOBUFS;
401
402 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
403 if (skb == NULL)
404 goto errout;
405
5ca5461c 406 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
407 first_remote_rtnl(fdb));
d342894c 408 if (err < 0) {
409 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
410 WARN_ON(err == -EMSGSIZE);
411 kfree_skb(skb);
412 goto errout;
413 }
414
415 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
416 return;
417errout:
418 if (err < 0)
419 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
420}
421
e4c7ed41 422static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 423{
424 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 425 struct vxlan_fdb f = {
426 .state = NUD_STALE,
427 };
428 struct vxlan_rdst remote = {
e4c7ed41 429 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 430 .remote_vni = VXLAN_N_VID,
431 };
3e61aa8f
SH		 432
433 INIT_LIST_HEAD(&f.remotes);
434 list_add_rcu(&remote.list, &f.remotes);
e4f67add
DS		 435
436 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
437}
438
439static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
440{
bb3fd687
SH		 441 struct vxlan_fdb f = {
442 .state = NUD_STALE,
443 };
e4f67add 444
3e61aa8f 445 INIT_LIST_HEAD(&f.remotes);
e4f67add
DS		 446 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
447
448 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
449}
450
d342894c 451/* Hash Ethernet address */
452static u32 eth_hash(const unsigned char *addr)
453{
454 u64 value = get_unaligned((u64 *)addr);
455
456 /* only want 6 bytes */
457#ifdef __BIG_ENDIAN
d342894c 458 value >>= 16;
321fb991 459#else
460 value <<= 16;
d342894c 461#endif
462 return hash_64(value, FDB_HASH_BITS);
463}
464
465/* Hash chain to use given mac address */
466static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
467 const u8 *mac)
468{
469 return &vxlan->fdb_head[eth_hash(mac)];
470}
471
472/* Look up Ethernet address in forwarding table */
014be2c8 473static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 474 const u8 *mac)
475
476{
477 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
478 struct vxlan_fdb *f;
d342894c 479
b67bfe0d 480 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 481 if (ether_addr_equal(mac, f->eth_addr))
d342894c 482 return f;
483 }
484
485 return NULL;
486}
487
014be2c8
SS		 488static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
489 const u8 *mac)
490{
491 struct vxlan_fdb *f;
492
493 f = __vxlan_find_mac(vxlan, mac);
494 if (f)
495 f->used = jiffies;
496
497 return f;
498}
499
a5e7c10a
MR		 500/* caller should hold vxlan->hash_lock */
501static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 502 union vxlan_addr *ip, __be16 port,
a5e7c10a 503 __u32 vni, __u32 ifindex)
6681712d 504{
3e61aa8f 505 struct vxlan_rdst *rd;
6681712d 506
3e61aa8f 507 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 508 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 509 rd->remote_port == port &&
510 rd->remote_vni == vni &&
511 rd->remote_ifindex == ifindex)
a5e7c10a 512 return rd;
6681712d 513 }
3e61aa8f 514
a5e7c10a
MR		 515 return NULL;
516}
517
906dc186
TR		 518/* Replace destination of unicast mac */
519static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 520 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 521{
522 struct vxlan_rdst *rd;
523
524 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
525 if (rd)
526 return 0;
527
528 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
529 if (!rd)
530 return 0;
e4c7ed41 531 rd->remote_ip = *ip;
906dc186
TR		 532 rd->remote_port = port;
533 rd->remote_vni = vni;
534 rd->remote_ifindex = ifindex;
535 return 1;
536}
537
a5e7c10a
MR		 538/* Add/update destinations for multicast */
539static int vxlan_fdb_append(struct vxlan_fdb *f,
e4c7ed41 540 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
a5e7c10a
MR		 541{
542 struct vxlan_rdst *rd;
543
544 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
545 if (rd)
546 return 0;
547
6681712d
DS		 548 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
549 if (rd == NULL)
550 return -ENOBUFS;
e4c7ed41 551 rd->remote_ip = *ip;
6681712d
DS		 552 rd->remote_port = port;
553 rd->remote_vni = vni;
554 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 555
556 list_add_tail_rcu(&rd->list, &f->remotes);
557
6681712d
DS		 558 return 1;
559}
560
53cf5275
JG		 561/* Notify netdevs that UDP port started listening */
562static void vxlan_notify_add_rx_port(struct sock *sk)
563{
564 struct net_device *dev;
565 struct net *net = sock_net(sk);
566 sa_family_t sa_family = sk->sk_family;
35e42379 567 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 568
569 rcu_read_lock();
570 for_each_netdev_rcu(net, dev) {
571 if (dev->netdev_ops->ndo_add_vxlan_port)
572 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
573 port);
574 }
575 rcu_read_unlock();
576}
577
578/* Notify netdevs that UDP port is no more listening */
579static void vxlan_notify_del_rx_port(struct sock *sk)
580{
581 struct net_device *dev;
582 struct net *net = sock_net(sk);
583 sa_family_t sa_family = sk->sk_family;
35e42379 584 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 585
586 rcu_read_lock();
587 for_each_netdev_rcu(net, dev) {
588 if (dev->netdev_ops->ndo_del_vxlan_port)
589 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
590 port);
591 }
592 rcu_read_unlock();
593}
594
d342894c 595/* Add new entry to forwarding table -- assumes lock held */
596static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 597 const u8 *mac, union vxlan_addr *ip,
6681712d 598 __u16 state, __u16 flags,
73cf3317 599 __be16 port, __u32 vni, __u32 ifindex,
ae884082 600 __u8 ndm_flags)
d342894c 601{
602 struct vxlan_fdb *f;
603 int notify = 0;
604
014be2c8 605 f = __vxlan_find_mac(vxlan, mac);
d342894c 606 if (f) {
607 if (flags & NLM_F_EXCL) {
608 netdev_dbg(vxlan->dev,
609 "lost race to create %pM\n", mac);
610 return -EEXIST;
611 }
612 if (f->state != state) {
613 f->state = state;
614 f->updated = jiffies;
615 notify = 1;
616 }
ae884082
DS		 617 if (f->flags != ndm_flags) {
618 f->flags = ndm_flags;
619 f->updated = jiffies;
620 notify = 1;
621 }
906dc186
TR		 622 if ((flags & NLM_F_REPLACE)) {
623 /* Only change unicasts */
624 if (!(is_multicast_ether_addr(f->eth_addr) ||
625 is_zero_ether_addr(f->eth_addr))) {
626 int rc = vxlan_fdb_replace(f, ip, port, vni,
627 ifindex);
628
629 if (rc < 0)
630 return rc;
631 notify |= rc;
632 } else
633 return -EOPNOTSUPP;
634 }
6681712d 635 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 636 (is_multicast_ether_addr(f->eth_addr) ||
637 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 638 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
639
640 if (rc < 0)
641 return rc;
642 notify |= rc;
643 }
d342894c 644 } else {
645 if (!(flags & NLM_F_CREATE))
646 return -ENOENT;
647
648 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
649 return -ENOSPC;
650
906dc186
TR		 651 /* Disallow replace to add a multicast entry */
652 if ((flags & NLM_F_REPLACE) &&
653 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
654 return -EOPNOTSUPP;
655
e4c7ed41 656 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 657 f = kmalloc(sizeof(*f), GFP_ATOMIC);
658 if (!f)
659 return -ENOMEM;
660
661 notify = 1;
d342894c 662 f->state = state;
ae884082 663 f->flags = ndm_flags;
d342894c 664 f->updated = f->used = jiffies;
3e61aa8f 665 INIT_LIST_HEAD(&f->remotes);
d342894c 666 memcpy(f->eth_addr, mac, ETH_ALEN);
667
3e61aa8f
SH		 668 vxlan_fdb_append(f, ip, port, vni, ifindex);
669
d342894c 670 ++vxlan->addrcnt;
671 hlist_add_head_rcu(&f->hlist,
672 vxlan_fdb_head(vxlan, mac));
673 }
674
675 if (notify)
676 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
677
678 return 0;
679}
680
6706c82e 681static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 682{
683 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 684 struct vxlan_rdst *rd, *nd;
6681712d 685
3e61aa8f 686 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 687 kfree(rd);
6681712d
DS		 688 kfree(f);
689}
690
d342894c 691static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
692{
693 netdev_dbg(vxlan->dev,
694 "delete %pM\n", f->eth_addr);
695
696 --vxlan->addrcnt;
697 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
698
699 hlist_del_rcu(&f->hlist);
6681712d 700 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 701}
702
f0b074be 703static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 704 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 705{
6681712d 706 struct net *net = dev_net(vxlan->dev);
e4c7ed41 707 int err;
d342894c 708
f0b074be 709 if (tb[NDA_DST]) {
e4c7ed41
CW		 710 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
711 if (err)
712 return err;
f0b074be 713 } else {
e4c7ed41
CW		 714 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
715 if (remote->sa.sa_family == AF_INET) {
716 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
717 ip->sa.sa_family = AF_INET;
718#if IS_ENABLED(CONFIG_IPV6)
719 } else {
720 ip->sin6.sin6_addr = in6addr_any;
721 ip->sa.sa_family = AF_INET6;
722#endif
723 }
f0b074be 724 }
d342894c 725
6681712d 726 if (tb[NDA_PORT]) {
73cf3317 727 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 728 return -EINVAL;
f0b074be
MR		 729 *port = nla_get_be16(tb[NDA_PORT]);
730 } else {
731 *port = vxlan->dst_port;
732 }
6681712d
DS		 733
734 if (tb[NDA_VNI]) {
735 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
736 return -EINVAL;
f0b074be
MR		 737 *vni = nla_get_u32(tb[NDA_VNI]);
738 } else {
739 *vni = vxlan->default_dst.remote_vni;
740 }
6681712d
DS		 741
742 if (tb[NDA_IFINDEX]) {
5abb0029 743 struct net_device *tdev;
6681712d
DS		 744
745 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
746 return -EINVAL;
f0b074be
MR		 747 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
748 tdev = dev_get_by_index(net, *ifindex);
5abb0029 749 if (!tdev)
6681712d 750 return -EADDRNOTAVAIL;
5abb0029 751 dev_put(tdev);
f0b074be
MR		 752 } else {
753 *ifindex = 0;
754 }
755
756 return 0;
757}
758
759/* Add static entry (via netlink) */
760static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
761 struct net_device *dev,
762 const unsigned char *addr, u16 flags)
763{
764 struct vxlan_dev *vxlan = netdev_priv(dev);
765 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 766 union vxlan_addr ip;
f0b074be
MR		 767 __be16 port;
768 u32 vni, ifindex;
769 int err;
770
771 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
772 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
773 ndm->ndm_state);
774 return -EINVAL;
775 }
776
777 if (tb[NDA_DST] == NULL)
778 return -EINVAL;
779
780 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
781 if (err)
782 return err;
6681712d 783
d342894c 784 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 785 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 786 port, vni, ifindex, ndm->ndm_flags);
d342894c 787 spin_unlock_bh(&vxlan->hash_lock);
788
789 return err;
790}
791
792/* Delete entry (via netlink) */
1690be63
VY		 793static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
794 struct net_device *dev,
d342894c 795 const unsigned char *addr)
796{
797 struct vxlan_dev *vxlan = netdev_priv(dev);
798 struct vxlan_fdb *f;
bc7892ba 799 struct vxlan_rdst *rd = NULL;
e4c7ed41 800 union vxlan_addr ip;
bc7892ba
MR		 801 __be16 port;
802 u32 vni, ifindex;
803 int err;
804
805 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
806 if (err)
807 return err;
808
809 err = -ENOENT;
d342894c 810
811 spin_lock_bh(&vxlan->hash_lock);
812 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 813 if (!f)
814 goto out;
815
e4c7ed41
CW		 816 if (!vxlan_addr_any(&ip)) {
817 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 818 if (!rd)
819 goto out;
820 }
821
822 err = 0;
823
824 /* remove a destination if it's not the only one on the list,
825 * otherwise destroy the fdb entry
826 */
827 if (rd && !list_is_singular(&f->remotes)) {
828 list_del_rcu(&rd->list);
dcdc7a59 829 kfree_rcu(rd, rcu);
bc7892ba 830 goto out;
d342894c 831 }
bc7892ba
MR		 832
833 vxlan_fdb_destroy(vxlan, f);
834
835out:
d342894c 836 spin_unlock_bh(&vxlan->hash_lock);
837
838 return err;
839}
840
841/* Dump forwarding table */
842static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
843 struct net_device *dev, int idx)
844{
845 struct vxlan_dev *vxlan = netdev_priv(dev);
846 unsigned int h;
847
848 for (h = 0; h < FDB_HASH_SIZE; ++h) {
849 struct vxlan_fdb *f;
d342894c 850 int err;
851
b67bfe0d 852 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 853 struct vxlan_rdst *rd;
6681712d 854
3e61aa8f
SH		 855 if (idx < cb->args[0])
856 goto skip;
857
858 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 859 err = vxlan_fdb_info(skb, vxlan, f,
860 NETLINK_CB(cb->skb).portid,
861 cb->nlh->nlmsg_seq,
862 RTM_NEWNEIGH,
863 NLM_F_MULTI, rd);
864 if (err < 0)
3e61aa8f 865 goto out;
6681712d 866 }
3e61aa8f
SH		 867skip:
868 ++idx;
d342894c 869 }
870 }
3e61aa8f 871out:
d342894c 872 return idx;
873}
874
875/* Watch incoming packets to learn mapping between Ethernet address
876 * and Tunnel endpoint.
26a41ae6 877 * Return true if packet is bogus and should be droppped.
d342894c 878 */
26a41ae6 879static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 880 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 881{
882 struct vxlan_dev *vxlan = netdev_priv(dev);
883 struct vxlan_fdb *f;
d342894c 884
885 f = vxlan_find_mac(vxlan, src_mac);
886 if (likely(f)) {
5ca5461c 887 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 888
e4c7ed41 889 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 890 return false;
891
892 /* Don't migrate static entries, drop packets */
eb064c3b 893 if (f->state & NUD_NOARP)
26a41ae6 894 return true;
d342894c 895
896 if (net_ratelimit())
897 netdev_info(dev,
e4c7ed41 898 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 899 src_mac, &rdst->remote_ip, &src_ip);
d342894c 900
e4c7ed41 901 rdst->remote_ip = *src_ip;
d342894c 902 f->updated = jiffies;
8385f50a 903 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 904 } else {
905 /* learned new entry */
906 spin_lock(&vxlan->hash_lock);
3bf74b1a 907
908 /* close off race between vxlan_flush and incoming packets */
909 if (netif_running(dev))
910 vxlan_fdb_create(vxlan, src_mac, src_ip,
911 NUD_REACHABLE,
912 NLM_F_EXCL|NLM_F_CREATE,
913 vxlan->dst_port,
914 vxlan->default_dst.remote_vni,
915 0, NTF_SELF);
d342894c 916 spin_unlock(&vxlan->hash_lock);
917 }
26a41ae6 918
919 return false;
d342894c 920}
921
d342894c 922/* See if multicast group is already in use by other ID */
e4c7ed41 923static bool vxlan_group_used(struct vxlan_net *vn, union vxlan_addr *remote_ip)
d342894c 924{
553675fb 925 struct vxlan_dev *vxlan;
d342894c 926
553675fb 927 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
553675fb 928 if (!netif_running(vxlan->dev))
929 continue;
d342894c 930
e4c7ed41
CW		 931 if (vxlan_addr_equal(&vxlan->default_dst.remote_ip,
932 remote_ip))
553675fb 933 return true;
934 }
d342894c 935
936 return false;
937}
938
7c47cedf 939static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 940{
7c47cedf
SH		 941 atomic_inc(&vs->refcnt);
942}
d342894c 943
012a5729 944void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 945{
53cf5275
JG		 946 struct sock *sk = vs->sock->sk;
947 struct net *net = sock_net(sk);
948 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 949
7c47cedf
SH		 950 if (!atomic_dec_and_test(&vs->refcnt))
951 return;
d342894c 952
1c51a915 953 spin_lock(&vn->sock_lock);
7c47cedf 954 hlist_del_rcu(&vs->hlist);
430eda6d
PS		 955 smp_wmb();
956 vs->sock->sk->sk_user_data = NULL;
53cf5275 957 vxlan_notify_del_rx_port(sk);
1c51a915
SH		 958 spin_unlock(&vn->sock_lock);
959
7c47cedf 960 queue_work(vxlan_wq, &vs->del_work);
d342894c 961}
012a5729 962EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 963
3fc2de2f 964/* Callback to update multicast group membership when first VNI on
965 * multicast asddress is brought up
966 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 967 */
3fc2de2f 968static void vxlan_igmp_join(struct work_struct *work)
d342894c 969{
3fc2de2f 970 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 971 struct vxlan_sock *vs = vxlan->vn_sock;
972 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 973 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
974 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 975
d342894c 976 lock_sock(sk);
e4c7ed41
CW		 977 if (ip->sa.sa_family == AF_INET) {
978 struct ip_mreqn mreq = {
979 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
980 .imr_ifindex = ifindex,
981 };
982
983 ip_mc_join_group(sk, &mreq);
984#if IS_ENABLED(CONFIG_IPV6)
985 } else {
986 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
987 &ip->sin6.sin6_addr);
988#endif
989 }
3fc2de2f 990 release_sock(sk);
991
012a5729 992 vxlan_sock_release(vs);
3fc2de2f 993 dev_put(vxlan->dev);
994}
995
996/* Inverse of vxlan_igmp_join when last VNI is brought down */
997static void vxlan_igmp_leave(struct work_struct *work)
998{
999 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1000 struct vxlan_sock *vs = vxlan->vn_sock;
1001 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1002 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1003 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1004
1005 lock_sock(sk);
e4c7ed41
CW		 1006 if (ip->sa.sa_family == AF_INET) {
1007 struct ip_mreqn mreq = {
1008 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1009 .imr_ifindex = ifindex,
1010 };
1011
1012 ip_mc_leave_group(sk, &mreq);
1013#if IS_ENABLED(CONFIG_IPV6)
1014 } else {
1015 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1016 &ip->sin6.sin6_addr);
1017#endif
1018 }
1019
d342894c 1020 release_sock(sk);
d342894c 1021
012a5729 1022 vxlan_sock_release(vs);
7c47cedf 1023 dev_put(vxlan->dev);
d342894c 1024}
1025
1026/* Callback from net/ipv4/udp.c to receive packets */
1027static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1028{
5cfccc5a 1029 struct vxlan_sock *vs;
d342894c 1030 struct vxlanhdr *vxh;
553675fb 1031 __be16 port;
d342894c 1032
d342894c 1033 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1034 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1035 goto error;
1036
7ce04758
PS		 1037 /* Return packets with reserved bits set */
1038 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 1039 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1040 (vxh->vx_vni & htonl(0xff))) {
1041 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1042 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1043 goto error;
1044 }
1045
5cfccc5a
PS		 1046 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1047 goto drop;
1048
553675fb 1049 port = inet_sk(sk)->inet_sport;
5cfccc5a 1050
430eda6d
PS		 1051 smp_read_barrier_depends();
1052 vs = (struct vxlan_sock *)sk->sk_user_data;
5cfccc5a 1053 if (!vs)
d342894c 1054 goto drop;
d342894c 1055
5cfccc5a
PS		 1056 vs->rcv(vs, skb, vxh->vx_vni);
1057 return 0;
1058
1059drop:
1060 /* Consume bad packet */
1061 kfree_skb(skb);
1062 return 0;
1063
1064error:
1065 /* Return non vxlan pkt */
1066 return 1;
1067}
1068
1069static void vxlan_rcv(struct vxlan_sock *vs,
1070 struct sk_buff *skb, __be32 vx_vni)
1071{
e4c7ed41
CW		 1072 struct iphdr *oip = NULL;
1073 struct ipv6hdr *oip6 = NULL;
5cfccc5a
PS		 1074 struct vxlan_dev *vxlan;
1075 struct pcpu_tstats *stats;
e4c7ed41 1076 union vxlan_addr saddr;
5cfccc5a 1077 __u32 vni;
e4c7ed41
CW		 1078 int err = 0;
1079 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1080
1081 vni = ntohl(vx_vni) >> 8;
1082 /* Is this VNI defined? */
1083 vxlan = vxlan_vs_find_vni(vs, vni);
1084 if (!vxlan)
d342894c 1085 goto drop;
d342894c 1086
e4c7ed41 1087 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1088 skb_reset_mac_header(skb);
d342894c 1089 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 1090
1091 /* Ignore packet loops (and multicast echo) */
7367d0b5 1092 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1093 goto drop;
1094
7ce04758 1095 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1096 if (remote_ip->sa.sa_family == AF_INET) {
1097 oip = ip_hdr(skb);
1098 saddr.sin.sin_addr.s_addr = oip->saddr;
1099 saddr.sa.sa_family = AF_INET;
1100#if IS_ENABLED(CONFIG_IPV6)
1101 } else {
1102 oip6 = ipv6_hdr(skb);
1103 saddr.sin6.sin6_addr = oip6->saddr;
1104 saddr.sa.sa_family = AF_INET6;
1105#endif
1106 }
1107
26a41ae6 1108 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1109 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1110 goto drop;
d342894c 1111
d342894c 1112 skb_reset_network_header(skb);
0afb1666
JG		 1113
1114 /* If the NIC driver gave us an encapsulated packet with
1115 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1116 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1117 * for us. Otherwise force the upper layers to verify it.
1118 */
1119 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1120 !(vxlan->dev->features & NETIF_F_RXCSUM))
1121 skb->ip_summed = CHECKSUM_NONE;
1122
1123 skb->encapsulation = 0;
d342894c 1124
e4c7ed41
CW		 1125 if (oip6)
1126 err = IP6_ECN_decapsulate(oip6, skb);
1127 if (oip)
1128 err = IP_ECN_decapsulate(oip, skb);
1129
d342894c 1130 if (unlikely(err)) {
e4c7ed41
CW		 1131 if (log_ecn_error) {
1132 if (oip6)
1133 net_info_ratelimited("non-ECT from %pI6\n",
1134 &oip6->saddr);
1135 if (oip)
1136 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1137 &oip->saddr, oip->tos);
1138 }
d342894c 1139 if (err > 1) {
1140 ++vxlan->dev->stats.rx_frame_errors;
1141 ++vxlan->dev->stats.rx_errors;
1142 goto drop;
1143 }
1144 }
1145
e8171045 1146 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1147 u64_stats_update_begin(&stats->syncp);
1148 stats->rx_packets++;
1149 stats->rx_bytes += skb->len;
1150 u64_stats_update_end(&stats->syncp);
1151
1152 netif_rx(skb);
1153
5cfccc5a 1154 return;
d342894c 1155drop:
1156 /* Consume bad packet */
1157 kfree_skb(skb);
d342894c 1158}
1159
e4f67add
DS		 1160static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1161{
1162 struct vxlan_dev *vxlan = netdev_priv(dev);
1163 struct arphdr *parp;
1164 u8 *arpptr, *sha;
1165 __be32 sip, tip;
1166 struct neighbour *n;
1167
1168 if (dev->flags & IFF_NOARP)
1169 goto out;
1170
1171 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1172 dev->stats.tx_dropped++;
1173 goto out;
1174 }
1175 parp = arp_hdr(skb);
1176
1177 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1178 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1179 parp->ar_pro != htons(ETH_P_IP) ||
1180 parp->ar_op != htons(ARPOP_REQUEST) ||
1181 parp->ar_hln != dev->addr_len ||
1182 parp->ar_pln != 4)
1183 goto out;
1184 arpptr = (u8 *)parp + sizeof(struct arphdr);
1185 sha = arpptr;
1186 arpptr += dev->addr_len; /* sha */
1187 memcpy(&sip, arpptr, sizeof(sip));
1188 arpptr += sizeof(sip);
1189 arpptr += dev->addr_len; /* tha */
1190 memcpy(&tip, arpptr, sizeof(tip));
1191
1192 if (ipv4_is_loopback(tip) ||
1193 ipv4_is_multicast(tip))
1194 goto out;
1195
1196 n = neigh_lookup(&arp_tbl, &tip, dev);
1197
1198 if (n) {
e4f67add
DS		 1199 struct vxlan_fdb *f;
1200 struct sk_buff *reply;
1201
1202 if (!(n->nud_state & NUD_CONNECTED)) {
1203 neigh_release(n);
1204 goto out;
1205 }
1206
1207 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1208 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1209 /* bridge-local neighbor */
1210 neigh_release(n);
1211 goto out;
1212 }
1213
1214 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1215 n->ha, sha);
1216
1217 neigh_release(n);
1218
1219 skb_reset_mac_header(reply);
1220 __skb_pull(reply, skb_network_offset(reply));
1221 reply->ip_summed = CHECKSUM_UNNECESSARY;
1222 reply->pkt_type = PACKET_HOST;
1223
1224 if (netif_rx_ni(reply) == NET_RX_DROP)
1225 dev->stats.rx_dropped++;
e4c7ed41
CW		 1226 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1227 union vxlan_addr ipa = {
1228 .sin.sin_addr.s_addr = tip,
1229 .sa.sa_family = AF_INET,
1230 };
1231
1232 vxlan_ip_miss(dev, &ipa);
1233 }
e4f67add
DS		 1234out:
1235 consume_skb(skb);
1236 return NETDEV_TX_OK;
1237}
1238
f564f45c
CW		 1239#if IS_ENABLED(CONFIG_IPV6)
1240static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1241{
1242 struct vxlan_dev *vxlan = netdev_priv(dev);
1243 struct neighbour *n;
1244 union vxlan_addr ipa;
1245 const struct ipv6hdr *iphdr;
1246 const struct in6_addr *saddr, *daddr;
1247 struct nd_msg *msg;
1248 struct inet6_dev *in6_dev = NULL;
1249
1250 in6_dev = __in6_dev_get(dev);
1251 if (!in6_dev)
1252 goto out;
1253
1254 if (!pskb_may_pull(skb, skb->len))
1255 goto out;
1256
1257 iphdr = ipv6_hdr(skb);
1258 saddr = &iphdr->saddr;
1259 daddr = &iphdr->daddr;
1260
1261 if (ipv6_addr_loopback(daddr) ||
1262 ipv6_addr_is_multicast(daddr))
1263 goto out;
1264
1265 msg = (struct nd_msg *)skb_transport_header(skb);
1266 if (msg->icmph.icmp6_code != 0 ||
1267 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1268 goto out;
1269
1270 n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1271
1272 if (n) {
1273 struct vxlan_fdb *f;
1274
1275 if (!(n->nud_state & NUD_CONNECTED)) {
1276 neigh_release(n);
1277 goto out;
1278 }
1279
1280 f = vxlan_find_mac(vxlan, n->ha);
1281 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1282 /* bridge-local neighbor */
1283 neigh_release(n);
1284 goto out;
1285 }
1286
1287 ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1288 !!in6_dev->cnf.forwarding,
1289 true, false, false);
1290 neigh_release(n);
1291 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1292 ipa.sin6.sin6_addr = *daddr;
1293 ipa.sa.sa_family = AF_INET6;
1294 vxlan_ip_miss(dev, &ipa);
1295 }
1296
1297out:
1298 consume_skb(skb);
1299 return NETDEV_TX_OK;
1300}
1301#endif
1302
e4f67add
DS		 1303static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1304{
1305 struct vxlan_dev *vxlan = netdev_priv(dev);
1306 struct neighbour *n;
e4f67add
DS		 1307
1308 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1309 return false;
1310
1311 n = NULL;
1312 switch (ntohs(eth_hdr(skb)->h_proto)) {
1313 case ETH_P_IP:
e15a00aa
CW		 1314 {
1315 struct iphdr *pip;
1316
e4f67add
DS		 1317 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1318 return false;
1319 pip = ip_hdr(skb);
1320 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1321 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1322 union vxlan_addr ipa = {
1323 .sin.sin_addr.s_addr = pip->daddr,
1324 .sa.sa_family = AF_INET,
1325 };
1326
1327 vxlan_ip_miss(dev, &ipa);
1328 return false;
1329 }
1330
e4f67add 1331 break;
e15a00aa
CW		 1332 }
1333#if IS_ENABLED(CONFIG_IPV6)
1334 case ETH_P_IPV6:
1335 {
1336 struct ipv6hdr *pip6;
1337
1338 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1339 return false;
1340 pip6 = ipv6_hdr(skb);
1341 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1342 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1343 union vxlan_addr ipa = {
1344 .sin6.sin6_addr = pip6->daddr,
1345 .sa.sa_family = AF_INET6,
1346 };
1347
1348 vxlan_ip_miss(dev, &ipa);
1349 return false;
1350 }
1351
1352 break;
1353 }
1354#endif
e4f67add
DS		 1355 default:
1356 return false;
1357 }
1358
1359 if (n) {
1360 bool diff;
1361
7367d0b5 1362 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1363 if (diff) {
1364 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1365 dev->addr_len);
1366 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1367 }
1368 neigh_release(n);
1369 return diff;
e4c7ed41
CW		 1370 }
1371
e4f67add
DS		 1372 return false;
1373}
1374
553675fb 1375static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1376{
1377 sock_put(skb->sk);
1378}
1379
1380/* On transmit, associate with the tunnel socket */
49560532 1381static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1cad8715 1382{
1cad8715 1383 skb_orphan(skb);
1384 sock_hold(sk);
1385 skb->sk = sk;
553675fb 1386 skb->destructor = vxlan_sock_put;
1cad8715 1387}
1388
05f47d69 1389/* Compute source port for outgoing packet
1390 * first choice to use L4 flow hash since it will spread
1391 * better and maybe available from hardware
1392 * secondary choice is to use jhash on the Ethernet header
1393 */
49560532 1394__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
05f47d69 1395{
49560532 1396 unsigned int range = (port_max - port_min) + 1;
05f47d69 1397 u32 hash;
1398
1399 hash = skb_get_rxhash(skb);
1400 if (!hash)
1401 hash = jhash(skb->data, 2 * ETH_ALEN,
1402 (__force u32) skb->protocol);
1403
49560532 1404 return htons((((u64) hash * range) >> 32) + port_min);
05f47d69 1405}
49560532 1406EXPORT_SYMBOL_GPL(vxlan_src_port);
05f47d69 1407
05c0db08
PS		 1408static int handle_offloads(struct sk_buff *skb)
1409{
1410 if (skb_is_gso(skb)) {
1411 int err = skb_unclone(skb, GFP_ATOMIC);
1412 if (unlikely(err))
1413 return err;
1414
f6ace502 1415 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1416 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1417 skb->ip_summed = CHECKSUM_NONE;
1418
1419 return 0;
1420}
1421
e4c7ed41 1422#if IS_ENABLED(CONFIG_IPV6)
11796187 1423static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1424 struct dst_entry *dst, struct sk_buff *skb,
1425 struct net_device *dev, struct in6_addr *saddr,
1426 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1427 __be16 src_port, __be16 dst_port, __be32 vni)
1428{
1429 struct ipv6hdr *ip6h;
1430 struct vxlanhdr *vxh;
1431 struct udphdr *uh;
1432 int min_headroom;
1433 int err;
1434
1435 if (!skb->encapsulation) {
1436 skb_reset_inner_headers(skb);
1437 skb->encapsulation = 1;
1438 }
1439
963a88b3
ND		 1440 skb_scrub_packet(skb, false);
1441
e4c7ed41
CW		 1442 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1443 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1444 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1445
1446 /* Need space for new headers (invalidates iph ptr) */
1447 err = skb_cow_head(skb, min_headroom);
1448 if (unlikely(err))
1449 return err;
1450
1451 if (vlan_tx_tag_present(skb)) {
1452 if (WARN_ON(!__vlan_put_tag(skb,
1453 skb->vlan_proto,
1454 vlan_tx_tag_get(skb))))
1455 return -ENOMEM;
1456
1457 skb->vlan_tci = 0;
1458 }
1459
1460 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1461 vxh->vx_flags = htonl(VXLAN_FLAGS);
1462 vxh->vx_vni = vni;
1463
1464 __skb_push(skb, sizeof(*uh));
1465 skb_reset_transport_header(skb);
1466 uh = udp_hdr(skb);
1467
1468 uh->dest = dst_port;
1469 uh->source = src_port;
1470
1471 uh->len = htons(skb->len);
1472 uh->check = 0;
1473
1474 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1475 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1476 IPSKB_REROUTED);
e4c7ed41
CW		 1477 skb_dst_set(skb, dst);
1478
1479 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1480 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1481 skb->ip_summed = CHECKSUM_UNNECESSARY;
1482 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1483 IPPROTO_UDP, csum);
1484 if (uh->check == 0)
1485 uh->check = CSUM_MANGLED_0;
1486 } else {
1487 skb->ip_summed = CHECKSUM_PARTIAL;
1488 skb->csum_start = skb_transport_header(skb) - skb->head;
1489 skb->csum_offset = offsetof(struct udphdr, check);
1490 uh->check = ~csum_ipv6_magic(saddr, daddr,
1491 skb->len, IPPROTO_UDP, 0);
1492 }
1493
1494 __skb_push(skb, sizeof(*ip6h));
1495 skb_reset_network_header(skb);
1496 ip6h = ipv6_hdr(skb);
1497 ip6h->version = 6;
1498 ip6h->priority = prio;
1499 ip6h->flow_lbl[0] = 0;
1500 ip6h->flow_lbl[1] = 0;
1501 ip6h->flow_lbl[2] = 0;
1502 ip6h->payload_len = htons(skb->len);
1503 ip6h->nexthdr = IPPROTO_UDP;
1504 ip6h->hop_limit = ttl;
1505 ip6h->daddr = *daddr;
1506 ip6h->saddr = *saddr;
1507
1508 vxlan_set_owner(vs->sock->sk, skb);
1509
1510 err = handle_offloads(skb);
1511 if (err)
1512 return err;
1513
1514 ip6tunnel_xmit(skb, dev);
1515 return 0;
1516}
1517#endif
1518
11796187 1519int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1520 struct rtable *rt, struct sk_buff *skb,
1521 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1522 __be16 src_port, __be16 dst_port, __be32 vni)
1523{
1524 struct vxlanhdr *vxh;
1525 struct udphdr *uh;
649c5b8b 1526 int min_headroom;
49560532
PS		 1527 int err;
1528
1529 if (!skb->encapsulation) {
1530 skb_reset_inner_headers(skb);
1531 skb->encapsulation = 1;
1532 }
1533
649c5b8b 1534 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1535 + VXLAN_HLEN + sizeof(struct iphdr)
1536 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1537
1538 /* Need space for new headers (invalidates iph ptr) */
1539 err = skb_cow_head(skb, min_headroom);
1540 if (unlikely(err))
1541 return err;
1542
1eaa8178
PS		 1543 if (vlan_tx_tag_present(skb)) {
1544 if (WARN_ON(!__vlan_put_tag(skb,
1545 skb->vlan_proto,
1546 vlan_tx_tag_get(skb))))
1547 return -ENOMEM;
1548
1549 skb->vlan_tci = 0;
1550 }
1551
49560532
PS		 1552 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1553 vxh->vx_flags = htonl(VXLAN_FLAGS);
1554 vxh->vx_vni = vni;
1555
1556 __skb_push(skb, sizeof(*uh));
1557 skb_reset_transport_header(skb);
1558 uh = udp_hdr(skb);
1559
1560 uh->dest = dst_port;
1561 uh->source = src_port;
1562
1563 uh->len = htons(skb->len);
1564 uh->check = 0;
1565
1566 vxlan_set_owner(vs->sock->sk, skb);
1567
1568 err = handle_offloads(skb);
1569 if (err)
1570 return err;
1571
963a88b3
ND		 1572 return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
1573 false);
49560532
PS		 1574}
1575EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1576
9dcc71e1
SS		 1577/* Bypass encapsulation if the destination is local */
1578static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1579 struct vxlan_dev *dst_vxlan)
1580{
1581 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1582 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
e4c7ed41
CW		 1583 union vxlan_addr loopback;
1584 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1
SS		 1585
1586 skb->pkt_type = PACKET_HOST;
1587 skb->encapsulation = 0;
1588 skb->dev = dst_vxlan->dev;
1589 __skb_pull(skb, skb_network_offset(skb));
1590
e4c7ed41
CW		 1591 if (remote_ip->sa.sa_family == AF_INET) {
1592 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1593 loopback.sa.sa_family = AF_INET;
1594#if IS_ENABLED(CONFIG_IPV6)
1595 } else {
1596 loopback.sin6.sin6_addr = in6addr_loopback;
1597 loopback.sa.sa_family = AF_INET6;
1598#endif
1599 }
1600
9dcc71e1 1601 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1602 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1603
1604 u64_stats_update_begin(&tx_stats->syncp);
1605 tx_stats->tx_packets++;
1606 tx_stats->tx_bytes += skb->len;
1607 u64_stats_update_end(&tx_stats->syncp);
1608
1609 if (netif_rx(skb) == NET_RX_SUCCESS) {
1610 u64_stats_update_begin(&rx_stats->syncp);
1611 rx_stats->rx_packets++;
1612 rx_stats->rx_bytes += skb->len;
1613 u64_stats_update_end(&rx_stats->syncp);
1614 } else {
1615 skb->dev->stats.rx_dropped++;
1616 }
1617}
1618
4ad16930
SH		 1619static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1620 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1621{
1622 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1623 struct rtable *rt = NULL;
d342894c 1624 const struct iphdr *old_iph;
d342894c 1625 struct flowi4 fl4;
e4c7ed41
CW		 1626 union vxlan_addr *dst;
1627 __be16 src_port = 0, dst_port;
234f5b73 1628 u32 vni;
d342894c 1629 __be16 df = 0;
1630 __u8 tos, ttl;
0e6fbc5b 1631 int err;
d342894c 1632
823aa873 1633 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1634 vni = rdst->remote_vni;
e4c7ed41 1635 dst = &rdst->remote_ip;
e4f67add 1636
e4c7ed41 1637 if (vxlan_addr_any(dst)) {
e4f67add 1638 if (did_rsc) {
e4f67add 1639 /* short-circuited back to local bridge */
9dcc71e1 1640 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1641 return;
e4f67add 1642 }
ef59febe 1643 goto drop;
e4f67add 1644 }
ef59febe 1645
d342894c 1646 old_iph = ip_hdr(skb);
1647
d342894c 1648 ttl = vxlan->ttl;
e4c7ed41 1649 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1650 ttl = 1;
1651
1652 tos = vxlan->tos;
1653 if (tos == 1)
206aaafc 1654 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1655
49560532 1656 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
d342894c 1657
e4c7ed41
CW		 1658 if (dst->sa.sa_family == AF_INET) {
1659 memset(&fl4, 0, sizeof(fl4));
1660 fl4.flowi4_oif = rdst->remote_ifindex;
1661 fl4.flowi4_tos = RT_TOS(tos);
1662 fl4.daddr = dst->sin.sin_addr.s_addr;
1663 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1664
1665 rt = ip_route_output_key(dev_net(dev), &fl4);
1666 if (IS_ERR(rt)) {
1667 netdev_dbg(dev, "no route to %pI4\n",
1668 &dst->sin.sin_addr.s_addr);
1669 dev->stats.tx_carrier_errors++;
1670 goto tx_error;
1671 }
d342894c 1672
e4c7ed41
CW		 1673 if (rt->dst.dev == dev) {
1674 netdev_dbg(dev, "circular route to %pI4\n",
1675 &dst->sin.sin_addr.s_addr);
1676 dev->stats.collisions++;
1677 goto tx_error;
1678 }
1679
1680 /* Bypass encapsulation if the destination is local */
1681 if (rt->rt_flags & RTCF_LOCAL &&
1682 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1683 struct vxlan_dev *dst_vxlan;
1684
1685 ip_rt_put(rt);
1686 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1687 if (!dst_vxlan)
1688 goto tx_error;
1689 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1690 return;
1691 }
1692
1693 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1694 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1695
11796187 1696 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
e4c7ed41
CW		 1697 fl4.saddr, dst->sin.sin_addr.s_addr,
1698 tos, ttl, df, src_port, dst_port,
1699 htonl(vni << 8));
9dcc71e1 1700
e4c7ed41
CW		 1701 if (err < 0)
1702 goto rt_tx_error;
1703 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1704#if IS_ENABLED(CONFIG_IPV6)
1705 } else {
1706 struct sock *sk = vxlan->vn_sock->sock->sk;
1707 struct dst_entry *ndst;
1708 struct flowi6 fl6;
1709 u32 flags;
1710
1711 memset(&fl6, 0, sizeof(fl6));
1712 fl6.flowi6_oif = rdst->remote_ifindex;
1713 fl6.daddr = dst->sin6.sin6_addr;
1714 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1715 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1716
1717 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1718 netdev_dbg(dev, "no route to %pI6\n",
1719 &dst->sin6.sin6_addr);
1720 dev->stats.tx_carrier_errors++;
9dcc71e1 1721 goto tx_error;
e4c7ed41 1722 }
d342894c 1723
e4c7ed41
CW		 1724 if (ndst->dev == dev) {
1725 netdev_dbg(dev, "circular route to %pI6\n",
1726 &dst->sin6.sin6_addr);
1727 dst_release(ndst);
1728 dev->stats.collisions++;
1729 goto tx_error;
1730 }
0e6fbc5b 1731
e4c7ed41
CW		 1732 /* Bypass encapsulation if the destination is local */
1733 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1734 if (flags & RTF_LOCAL &&
1735 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1736 struct vxlan_dev *dst_vxlan;
1737
1738 dst_release(ndst);
1739 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1740 if (!dst_vxlan)
1741 goto tx_error;
1742 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1743 return;
1744 }
49560532 1745
e4c7ed41
CW		 1746 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1747
11796187 1748 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41
CW		 1749 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1750 src_port, dst_port, htonl(vni << 8));
1751#endif
1752 }
0e6fbc5b 1753
4ad16930 1754 return;
d342894c 1755
1756drop:
1757 dev->stats.tx_dropped++;
1758 goto tx_free;
1759
49560532
PS		 1760rt_tx_error:
1761 ip_rt_put(rt);
d342894c 1762tx_error:
1763 dev->stats.tx_errors++;
1764tx_free:
1765 dev_kfree_skb(skb);
d342894c 1766}
1767
6681712d
DS		 1768/* Transmit local packets over Vxlan
1769 *
1770 * Outer IP header inherits ECN and DF from inner header.
1771 * Outer UDP destination is the VXLAN assigned port.
1772 * source port is based on hash of flow
1773 */
1774static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1775{
1776 struct vxlan_dev *vxlan = netdev_priv(dev);
1777 struct ethhdr *eth;
1778 bool did_rsc = false;
afbd8bae 1779 struct vxlan_rdst *rdst;
6681712d 1780 struct vxlan_fdb *f;
6681712d
DS		 1781
1782 skb_reset_mac_header(skb);
1783 eth = eth_hdr(skb);
1784
f564f45c
CW		 1785 if ((vxlan->flags & VXLAN_F_PROXY)) {
1786 if (ntohs(eth->h_proto) == ETH_P_ARP)
1787 return arp_reduce(dev, skb);
1788#if IS_ENABLED(CONFIG_IPV6)
1789 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1790 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1791 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1792 struct nd_msg *msg;
1793
1794 msg = (struct nd_msg *)skb_transport_header(skb);
1795 if (msg->icmph.icmp6_code == 0 &&
1796 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1797 return neigh_reduce(dev, skb);
1798 }
1799#endif
1800 }
6681712d
DS		 1801
1802 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1803 did_rsc = false;
1804
1805 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1806 (ntohs(eth->h_proto) == ETH_P_IP ||
1807 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1808 did_rsc = route_shortcircuit(dev, skb);
1809 if (did_rsc)
1810 f = vxlan_find_mac(vxlan, eth->h_dest);
1811 }
1812
6681712d 1813 if (f == NULL) {
afbd8bae
MR		 1814 f = vxlan_find_mac(vxlan, all_zeros_mac);
1815 if (f == NULL) {
1816 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1817 !is_multicast_ether_addr(eth->h_dest))
1818 vxlan_fdb_miss(vxlan, eth->h_dest);
1819
1820 dev->stats.tx_dropped++;
1821 dev_kfree_skb(skb);
1822 return NETDEV_TX_OK;
1823 }
1824 }
6681712d 1825
afbd8bae
MR		 1826 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1827 struct sk_buff *skb1;
6681712d 1828
afbd8bae
MR		 1829 skb1 = skb_clone(skb, GFP_ATOMIC);
1830 if (skb1)
1831 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1832 }
1833
afbd8bae 1834 dev_kfree_skb(skb);
4ad16930 1835 return NETDEV_TX_OK;
6681712d
DS		 1836}
1837
d342894c 1838/* Walk the forwarding table and purge stale entries */
1839static void vxlan_cleanup(unsigned long arg)
1840{
1841 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1842 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1843 unsigned int h;
1844
1845 if (!netif_running(vxlan->dev))
1846 return;
1847
1848 spin_lock_bh(&vxlan->hash_lock);
1849 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1850 struct hlist_node *p, *n;
1851 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1852 struct vxlan_fdb *f
1853 = container_of(p, struct vxlan_fdb, hlist);
1854 unsigned long timeout;
1855
3c172868 1856 if (f->state & NUD_PERMANENT)
d342894c 1857 continue;
1858
1859 timeout = f->used + vxlan->age_interval * HZ;
1860 if (time_before_eq(timeout, jiffies)) {
1861 netdev_dbg(vxlan->dev,
1862 "garbage collect %pM\n",
1863 f->eth_addr);
1864 f->state = NUD_STALE;
1865 vxlan_fdb_destroy(vxlan, f);
1866 } else if (time_before(timeout, next_timer))
1867 next_timer = timeout;
1868 }
1869 }
1870 spin_unlock_bh(&vxlan->hash_lock);
1871
1872 mod_timer(&vxlan->age_timer, next_timer);
1873}
1874
9c2e24e1
PS		 1875static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1876{
1877 __u32 vni = vxlan->default_dst.remote_vni;
1878
1879 vxlan->vn_sock = vs;
1880 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1881}
1882
d342894c 1883/* Setup stats when device is created */
1884static int vxlan_init(struct net_device *dev)
1885{
1c51a915
SH		 1886 struct vxlan_dev *vxlan = netdev_priv(dev);
1887 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1888 struct vxlan_sock *vs;
1c51a915 1889
e8171045
PS		 1890 dev->tstats = alloc_percpu(struct pcpu_tstats);
1891 if (!dev->tstats)
d342894c 1892 return -ENOMEM;
1893
1c51a915 1894 spin_lock(&vn->sock_lock);
9c2e24e1 1895 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1c51a915
SH		 1896 if (vs) {
1897 /* If we have a socket with same port already, reuse it */
1898 atomic_inc(&vs->refcnt);
9c2e24e1 1899 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1900 } else {
1901 /* otherwise make new socket outside of RTNL */
1902 dev_hold(dev);
1903 queue_work(vxlan_wq, &vxlan->sock_work);
1904 }
1905 spin_unlock(&vn->sock_lock);
1906
d342894c 1907 return 0;
1908}
1909
ba609e9b 1910static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1911{
1912 struct vxlan_fdb *f;
1913
1914 spin_lock_bh(&vxlan->hash_lock);
1915 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1916 if (f)
1917 vxlan_fdb_destroy(vxlan, f);
1918 spin_unlock_bh(&vxlan->hash_lock);
1919}
1920
ebf4063e
SH		 1921static void vxlan_uninit(struct net_device *dev)
1922{
1923 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 1924 struct vxlan_sock *vs = vxlan->vn_sock;
1925
ba609e9b 1926 vxlan_fdb_delete_default(vxlan);
afbd8bae 1927
ebf4063e 1928 if (vs)
012a5729 1929 vxlan_sock_release(vs);
ebf4063e
SH		 1930 free_percpu(dev->tstats);
1931}
1932
d342894c 1933/* Start ageing timer and join group when device is brought up */
1934static int vxlan_open(struct net_device *dev)
1935{
3fc2de2f 1936 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1937 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1938 struct vxlan_sock *vs = vxlan->vn_sock;
1939
1940 /* socket hasn't been created */
1941 if (!vs)
1942 return -ENOTCONN;
d342894c 1943
e4c7ed41
CW		 1944 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1945 vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1946 vxlan_sock_hold(vs);
7c47cedf 1947 dev_hold(dev);
3fc2de2f 1948 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 1949 }
1950
1951 if (vxlan->age_interval)
1952 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1953
1954 return 0;
1955}
1956
1957/* Purge the forwarding table */
1958static void vxlan_flush(struct vxlan_dev *vxlan)
1959{
31fec5aa 1960 unsigned int h;
d342894c 1961
1962 spin_lock_bh(&vxlan->hash_lock);
1963 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1964 struct hlist_node *p, *n;
1965 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1966 struct vxlan_fdb *f
1967 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 1968 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1969 if (!is_zero_ether_addr(f->eth_addr))
1970 vxlan_fdb_destroy(vxlan, f);
d342894c 1971 }
1972 }
1973 spin_unlock_bh(&vxlan->hash_lock);
1974}
1975
1976/* Cleanup timer and forwarding table on shutdown */
1977static int vxlan_stop(struct net_device *dev)
1978{
3fc2de2f 1979 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1980 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1981 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1982
e4c7ed41
CW		 1983 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1984 ! vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1985 vxlan_sock_hold(vs);
7c47cedf 1986 dev_hold(dev);
3fc2de2f 1987 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 1988 }
d342894c 1989
1990 del_timer_sync(&vxlan->age_timer);
1991
1992 vxlan_flush(vxlan);
1993
1994 return 0;
1995}
1996
d342894c 1997/* Stub, nothing needs to be done. */
1998static void vxlan_set_multicast_list(struct net_device *dev)
1999{
2000}
2001
2002static const struct net_device_ops vxlan_netdev_ops = {
2003 .ndo_init = vxlan_init,
ebf4063e 2004 .ndo_uninit = vxlan_uninit,
d342894c 2005 .ndo_open = vxlan_open,
2006 .ndo_stop = vxlan_stop,
2007 .ndo_start_xmit = vxlan_xmit,
e8171045 2008 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2009 .ndo_set_rx_mode = vxlan_set_multicast_list,
2010 .ndo_change_mtu = eth_change_mtu,
2011 .ndo_validate_addr = eth_validate_addr,
2012 .ndo_set_mac_address = eth_mac_addr,
2013 .ndo_fdb_add = vxlan_fdb_add,
2014 .ndo_fdb_del = vxlan_fdb_delete,
2015 .ndo_fdb_dump = vxlan_fdb_dump,
2016};
2017
2018/* Info for udev, that this is a virtual tunnel endpoint */
2019static struct device_type vxlan_type = {
2020 .name = "vxlan",
2021};
2022
53cf5275 2023/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2024 * supply the listening VXLAN udp ports. Callers are expected
2025 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2026 */
2027void vxlan_get_rx_port(struct net_device *dev)
2028{
2029 struct vxlan_sock *vs;
2030 struct net *net = dev_net(dev);
2031 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2032 sa_family_t sa_family;
35e42379
JG		 2033 __be16 port;
2034 unsigned int i;
53cf5275
JG		 2035
2036 spin_lock(&vn->sock_lock);
2037 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2038 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2039 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2040 sa_family = vs->sock->sk->sk_family;
2041 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2042 port);
2043 }
2044 }
2045 spin_unlock(&vn->sock_lock);
2046}
2047EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2048
d342894c 2049/* Initialize the device structure. */
2050static void vxlan_setup(struct net_device *dev)
2051{
2052 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2053 unsigned int h;
05f47d69 2054 int low, high;
d342894c 2055
2056 eth_hw_addr_random(dev);
2057 ether_setup(dev);
e4c7ed41
CW		 2058 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2059 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
2060 else
2061 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2062
2063 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2064 dev->destructor = free_netdev;
d342894c 2065 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2066
2067 dev->tx_queue_len = 0;
2068 dev->features |= NETIF_F_LLTX;
2069 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 2070 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2071 dev->features |= NETIF_F_RXCSUM;
05c0db08 2072 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2073
1eaa8178
PS		 2074 dev->vlan_features = dev->features;
2075 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2076 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2077 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2078 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2079 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2080 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2081
553675fb 2082 INIT_LIST_HEAD(&vxlan->next);
d342894c 2083 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2084 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2085 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2086 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2087
2088 init_timer_deferrable(&vxlan->age_timer);
2089 vxlan->age_timer.function = vxlan_cleanup;
2090 vxlan->age_timer.data = (unsigned long) vxlan;
2091
05f47d69 2092 inet_get_local_port_range(&low, &high);
2093 vxlan->port_min = low;
2094 vxlan->port_max = high;
823aa873 2095 vxlan->dst_port = htons(vxlan_port);
05f47d69 2096
d342894c 2097 vxlan->dev = dev;
2098
2099 for (h = 0; h < FDB_HASH_SIZE; ++h)
2100 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2101}
2102
2103static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2104 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2105 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2106 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2107 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2108 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2109 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2110 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2111 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2112 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2113 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2114 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2115 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2116 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2117 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2118 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2119 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2120 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2121};
2122
2123static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2124{
2125 if (tb[IFLA_ADDRESS]) {
2126 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2127 pr_debug("invalid link address (not ethernet)\n");
2128 return -EINVAL;
2129 }
2130
2131 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2132 pr_debug("invalid all zero ethernet address\n");
2133 return -EADDRNOTAVAIL;
2134 }
2135 }
2136
2137 if (!data)
2138 return -EINVAL;
2139
2140 if (data[IFLA_VXLAN_ID]) {
2141 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2142 if (id >= VXLAN_VID_MASK)
2143 return -ERANGE;
2144 }
2145
05f47d69 2146 if (data[IFLA_VXLAN_PORT_RANGE]) {
2147 const struct ifla_vxlan_port_range *p
2148 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2149
2150 if (ntohs(p->high) < ntohs(p->low)) {
2151 pr_debug("port range %u .. %u not valid\n",
2152 ntohs(p->low), ntohs(p->high));
2153 return -EINVAL;
2154 }
2155 }
2156
d342894c 2157 return 0;
2158}
2159
1b13c97f
YB		 2160static void vxlan_get_drvinfo(struct net_device *netdev,
2161 struct ethtool_drvinfo *drvinfo)
2162{
2163 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2164 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2165}
2166
2167static const struct ethtool_ops vxlan_ethtool_ops = {
2168 .get_drvinfo = vxlan_get_drvinfo,
2169 .get_link = ethtool_op_get_link,
2170};
2171
553675fb 2172static void vxlan_del_work(struct work_struct *work)
2173{
2174 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2175
2176 sk_release_kernel(vs->sock->sk);
2177 kfree_rcu(vs, rcu);
2178}
2179
e4c7ed41
CW		 2180#if IS_ENABLED(CONFIG_IPV6)
2181/* Create UDP socket for encapsulation receive. AF_INET6 socket
2182 * could be used for both IPv4 and IPv6 communications, but
2183 * users may set bindv6only=1.
2184 */
2185static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
553675fb 2186{
553675fb 2187 struct sock *sk;
e4c7ed41
CW		 2188 struct socket *sock;
2189 struct sockaddr_in6 vxlan_addr = {
2190 .sin6_family = AF_INET6,
2191 .sin6_port = port,
2192 };
2193 int rc, val = 1;
2194
2195 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2196 if (rc < 0) {
2197 pr_debug("UDPv6 socket create failed\n");
2198 return rc;
2199 }
2200
2201 /* Put in proper namespace */
2202 sk = sock->sk;
2203 sk_change_net(sk, net);
2204
2205 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2206 (char *)&val, sizeof(val));
2207 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2208 sizeof(struct sockaddr_in6));
2209 if (rc < 0) {
2210 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2211 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2212 sk_release_kernel(sk);
2213 return rc;
2214 }
2215 /* At this point, IPv6 module should have been loaded in
2216 * sock_create_kern().
2217 */
2218 BUG_ON(!ipv6_stub);
2219
2220 *psock = sock;
2221 /* Disable multicast loopback */
2222 inet_sk(sk)->mc_loop = 0;
2223 return 0;
2224}
2225
2226#else
2227
2228static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2229{
2230 return -EPFNOSUPPORT;
2231}
2232#endif
2233
2234static int create_v4_sock(struct net *net, __be16 port, struct socket **psock)
2235{
2236 struct sock *sk;
2237 struct socket *sock;
553675fb 2238 struct sockaddr_in vxlan_addr = {
2239 .sin_family = AF_INET,
2240 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2241 .sin_port = port,
553675fb 2242 };
2243 int rc;
553675fb 2244
2245 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2246 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2247 if (rc < 0) {
2248 pr_debug("UDP socket create failed\n");
e4c7ed41 2249 return rc;
553675fb 2250 }
2251
2252 /* Put in proper namespace */
e4c7ed41 2253 sk = sock->sk;
553675fb 2254 sk_change_net(sk, net);
2255
e4c7ed41 2256 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2257 sizeof(vxlan_addr));
2258 if (rc < 0) {
2259 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2260 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2261 sk_release_kernel(sk);
e4c7ed41
CW		 2262 return rc;
2263 }
2264
2265 *psock = sock;
2266 /* Disable multicast loopback */
2267 inet_sk(sk)->mc_loop = 0;
2268 return 0;
2269}
2270
2271/* Create new listen socket if needed */
2272static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2273 vxlan_rcv_t *rcv, void *data, bool ipv6)
2274{
2275 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2276 struct vxlan_sock *vs;
2277 struct socket *sock;
2278 struct sock *sk;
2279 int rc = 0;
2280 unsigned int h;
2281
2282 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2283 if (!vs)
2284 return ERR_PTR(-ENOMEM);
2285
2286 for (h = 0; h < VNI_HASH_SIZE; ++h)
2287 INIT_HLIST_HEAD(&vs->vni_list[h]);
2288
2289 INIT_WORK(&vs->del_work, vxlan_del_work);
2290
2291 if (ipv6)
2292 rc = create_v6_sock(net, port, &sock);
2293 else
2294 rc = create_v4_sock(net, port, &sock);
2295 if (rc < 0) {
553675fb 2296 kfree(vs);
2297 return ERR_PTR(rc);
2298 }
e4c7ed41
CW		 2299
2300 vs->sock = sock;
2301 sk = sock->sk;
9c2e24e1 2302 atomic_set(&vs->refcnt, 1);
5cfccc5a 2303 vs->rcv = rcv;
012a5729 2304 vs->data = data;
430eda6d
PS		 2305 smp_wmb();
2306 vs->sock->sk->sk_user_data = vs;
553675fb 2307
9c2e24e1
PS		 2308 spin_lock(&vn->sock_lock);
2309 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
53cf5275 2310 vxlan_notify_add_rx_port(sk);
9c2e24e1 2311 spin_unlock(&vn->sock_lock);
553675fb 2312
2313 /* Mark socket as an encapsulation socket. */
2314 udp_sk(sk)->encap_type = 1;
2315 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2316#if IS_ENABLED(CONFIG_IPV6)
2317 if (ipv6)
2318 ipv6_stub->udpv6_encap_enable();
2319 else
2320#endif
2321 udp_encap_enable();
2322
9c2e24e1
PS		 2323 return vs;
2324}
2325
012a5729
PS		 2326struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2327 vxlan_rcv_t *rcv, void *data,
e4c7ed41 2328 bool no_share, bool ipv6)
9c2e24e1
PS		 2329{
2330 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2331 struct vxlan_sock *vs;
2332
e4c7ed41 2333 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
9c2e24e1
PS		 2334 if (!IS_ERR(vs))
2335 return vs;
553675fb 2336
012a5729
PS		 2337 if (no_share) /* Return error if sharing is not allowed. */
2338 return vs;
2339
9c2e24e1
PS		 2340 spin_lock(&vn->sock_lock);
2341 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2342 if (vs) {
2343 if (vs->rcv == rcv)
2344 atomic_inc(&vs->refcnt);
2345 else
2346 vs = ERR_PTR(-EBUSY);
2347 }
2348 spin_unlock(&vn->sock_lock);
2349
2350 if (!vs)
9c2e24e1
PS		 2351 vs = ERR_PTR(-EINVAL);
2352
553675fb 2353 return vs;
2354}
012a5729 2355EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2356
1c51a915
SH		 2357/* Scheduled at device creation to bind to a socket */
2358static void vxlan_sock_work(struct work_struct *work)
2359{
9c2e24e1
PS		 2360 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2361 struct net *net = dev_net(vxlan->dev);
1c51a915 2362 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2363 __be16 port = vxlan->dst_port;
2364 struct vxlan_sock *nvs;
1c51a915 2365
e4c7ed41 2366 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
1c51a915 2367 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2368 if (!IS_ERR(nvs))
2369 vxlan_vs_add_dev(nvs, vxlan);
2370 spin_unlock(&vn->sock_lock);
2371
2372 dev_put(vxlan->dev);
1c51a915
SH		 2373}
2374
d342894c 2375static int vxlan_newlink(struct net *net, struct net_device *dev,
2376 struct nlattr *tb[], struct nlattr *data[])
2377{
553675fb 2378 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2379 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2380 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2381 __u32 vni;
2382 int err;
e4c7ed41 2383 bool use_ipv6 = false;
d342894c 2384
2385 if (!data[IFLA_VXLAN_ID])
2386 return -EINVAL;
2387
2388 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2389 dst->remote_vni = vni;
d342894c 2390
e4c7ed41
CW		 2391 if (data[IFLA_VXLAN_GROUP]) {
2392 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2393 dst->remote_ip.sa.sa_family = AF_INET;
2394 } else if (data[IFLA_VXLAN_GROUP6]) {
2395 if (!IS_ENABLED(CONFIG_IPV6))
2396 return -EPFNOSUPPORT;
2397
2398 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2399 sizeof(struct in6_addr));
2400 dst->remote_ip.sa.sa_family = AF_INET6;
2401 use_ipv6 = true;
2402 }
d342894c 2403
e4c7ed41
CW		 2404 if (data[IFLA_VXLAN_LOCAL]) {
2405 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2406 vxlan->saddr.sa.sa_family = AF_INET;
2407 } else if (data[IFLA_VXLAN_LOCAL6]) {
2408 if (!IS_ENABLED(CONFIG_IPV6))
2409 return -EPFNOSUPPORT;
2410
2411 /* TODO: respect scope id */
2412 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2413 sizeof(struct in6_addr));
2414 vxlan->saddr.sa.sa_family = AF_INET6;
2415 use_ipv6 = true;
2416 }
d342894c 2417
34e02aa1 2418 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2419 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2420 struct net_device *lowerdev
c7995c43 2421 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2422
2423 if (!lowerdev) {
c7995c43 2424 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2425 return -ENODEV;
2426 }
d342894c 2427
e4c7ed41
CW		 2428#if IS_ENABLED(CONFIG_IPV6)
2429 if (use_ipv6) {
2430 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2431 if (idev && idev->cnf.disable_ipv6) {
2432 pr_info("IPv6 is disabled via sysctl\n");
2433 return -EPERM;
2434 }
2435 vxlan->flags |= VXLAN_F_IPV6;
2436 }
2437#endif
2438
34e02aa1 2439 if (!tb[IFLA_MTU])
e4c7ed41 2440 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4
AD		 2441
2442 /* update header length based on lower device */
2443 dev->hard_header_len = lowerdev->hard_header_len +
e4c7ed41 2444 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
d342894c 2445 }
2446
2447 if (data[IFLA_VXLAN_TOS])
2448 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2449
afb97186
VB		 2450 if (data[IFLA_VXLAN_TTL])
2451 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2452
d342894c 2453 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2454 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2455
2456 if (data[IFLA_VXLAN_AGEING])
2457 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2458 else
2459 vxlan->age_interval = FDB_AGE_DEFAULT;
2460
e4f67add
DS		 2461 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2462 vxlan->flags |= VXLAN_F_PROXY;
2463
2464 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2465 vxlan->flags |= VXLAN_F_RSC;
2466
2467 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2468 vxlan->flags |= VXLAN_F_L2MISS;
2469
2470 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2471 vxlan->flags |= VXLAN_F_L3MISS;
2472
d342894c 2473 if (data[IFLA_VXLAN_LIMIT])
2474 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2475
05f47d69 2476 if (data[IFLA_VXLAN_PORT_RANGE]) {
2477 const struct ifla_vxlan_port_range *p
2478 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2479 vxlan->port_min = ntohs(p->low);
2480 vxlan->port_max = ntohs(p->high);
2481 }
2482
823aa873 2483 if (data[IFLA_VXLAN_PORT])
2484 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2485
553675fb 2486 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2487 pr_info("duplicate VNI %u\n", vni);
2488 return -EEXIST;
2489 }
2490
1b13c97f
YB		 2491 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2492
2936b6ab
SS		 2493 /* create an fdb entry for a valid default destination */
2494 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2495 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2496 &vxlan->default_dst.remote_ip,
2497 NUD_REACHABLE|NUD_PERMANENT,
2498 NLM_F_EXCL|NLM_F_CREATE,
2499 vxlan->dst_port,
2500 vxlan->default_dst.remote_vni,
2501 vxlan->default_dst.remote_ifindex,
2502 NTF_SELF);
2503 if (err)
2504 return err;
2505 }
d342894c 2506
afbd8bae
MR		 2507 err = register_netdevice(dev);
2508 if (err) {
ba609e9b 2509 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2510 return err;
2511 }
2512
553675fb 2513 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2514
2515 return 0;
d342894c 2516}
2517
2518static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2519{
fe5c3561 2520 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 2521 struct vxlan_dev *vxlan = netdev_priv(dev);
2522
fe5c3561 2523 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2524 if (!hlist_unhashed(&vxlan->hlist))
2525 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2526 spin_unlock(&vn->sock_lock);
2527
553675fb 2528 list_del(&vxlan->next);
d342894c 2529 unregister_netdevice_queue(dev, head);
2530}
2531
2532static size_t vxlan_get_size(const struct net_device *dev)
2533{
2534
2535 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2536 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2537 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2538 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2540 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2541 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2542 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2543 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2544 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2545 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2546 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2547 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2548 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 2549 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 2550 0;
2551}
2552
2553static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2554{
2555 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2556 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2557 struct ifla_vxlan_port_range ports = {
2558 .low = htons(vxlan->port_min),
2559 .high = htons(vxlan->port_max),
2560 };
d342894c 2561
c7995c43 2562 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2563 goto nla_put_failure;
2564
e4c7ed41
CW		 2565 if (!vxlan_addr_any(&dst->remote_ip)) {
2566 if (dst->remote_ip.sa.sa_family == AF_INET) {
2567 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2568 dst->remote_ip.sin.sin_addr.s_addr))
2569 goto nla_put_failure;
2570#if IS_ENABLED(CONFIG_IPV6)
2571 } else {
2572 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2573 &dst->remote_ip.sin6.sin6_addr))
2574 goto nla_put_failure;
2575#endif
2576 }
2577 }
d342894c 2578
c7995c43 2579 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2580 goto nla_put_failure;
2581
e4c7ed41
CW		 2582 if (!vxlan_addr_any(&vxlan->saddr)) {
2583 if (vxlan->saddr.sa.sa_family == AF_INET) {
2584 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2585 vxlan->saddr.sin.sin_addr.s_addr))
2586 goto nla_put_failure;
2587#if IS_ENABLED(CONFIG_IPV6)
2588 } else {
2589 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2590 &vxlan->saddr.sin6.sin6_addr))
2591 goto nla_put_failure;
2592#endif
2593 }
2594 }
d342894c 2595
2596 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2597 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2598 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2599 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2600 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2601 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2602 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2603 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2604 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2605 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2606 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2607 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2608 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2609 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 2610 goto nla_put_failure;
2611
05f47d69 2612 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2613 goto nla_put_failure;
2614
d342894c 2615 return 0;
2616
2617nla_put_failure:
2618 return -EMSGSIZE;
2619}
2620
2621static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2622 .kind = "vxlan",
2623 .maxtype = IFLA_VXLAN_MAX,
2624 .policy = vxlan_policy,
2625 .priv_size = sizeof(struct vxlan_dev),
2626 .setup = vxlan_setup,
2627 .validate = vxlan_validate,
2628 .newlink = vxlan_newlink,
2629 .dellink = vxlan_dellink,
2630 .get_size = vxlan_get_size,
2631 .fill_info = vxlan_fill_info,
2632};
2633
2634static __net_init int vxlan_init_net(struct net *net)
2635{
2636 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2637 unsigned int h;
d342894c 2638
553675fb 2639 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2640 spin_lock_init(&vn->sock_lock);
d342894c 2641
553675fb 2642 for (h = 0; h < PORT_HASH_SIZE; ++h)
2643 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2644
2645 return 0;
2646}
2647
2648static __net_exit void vxlan_exit_net(struct net *net)
2649{
2650 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 2651 struct vxlan_dev *vxlan;
372675a4 2652 LIST_HEAD(list);
9cb6cb7e
ZM		 2653
2654 rtnl_lock();
553675fb 2655 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 2656 unregister_netdevice_queue(vxlan->dev, &list);
2657 unregister_netdevice_many(&list);
9cb6cb7e 2658 rtnl_unlock();
d342894c 2659}
2660
2661static struct pernet_operations vxlan_net_ops = {
2662 .init = vxlan_init_net,
2663 .exit = vxlan_exit_net,
2664 .id = &vxlan_net_id,
2665 .size = sizeof(struct vxlan_net),
2666};
2667
2668static int __init vxlan_init_module(void)
2669{
2670 int rc;
2671
758c57d1
SH		 2672 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2673 if (!vxlan_wq)
2674 return -ENOMEM;
2675
d342894c 2676 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2677
2678 rc = register_pernet_device(&vxlan_net_ops);
2679 if (rc)
2680 goto out1;
2681
2682 rc = rtnl_link_register(&vxlan_link_ops);
2683 if (rc)
2684 goto out2;
2685
2686 return 0;
2687
2688out2:
2689 unregister_pernet_device(&vxlan_net_ops);
2690out1:
758c57d1 2691 destroy_workqueue(vxlan_wq);
d342894c 2692 return rc;
2693}
7332a13b 2694late_initcall(vxlan_init_module);
d342894c 2695
2696static void __exit vxlan_cleanup_module(void)
2697{
b7153984 2698 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 2699 destroy_workqueue(vxlan_wq);
f89e57c4 2700 unregister_pernet_device(&vxlan_net_ops);
6681712d 2701 rcu_barrier();
d342894c 2702}
2703module_exit(vxlan_cleanup_module);
2704
2705MODULE_LICENSE("GPL");
2706MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2707MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 2708MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror