projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
USB2NET : SR9700 : One chip USB 1.1 USB2NET SR9700Device Driver Support
[deliverable/linux.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) {
d342894c 481 if (compare_ether_addr(mac, f->eth_addr) == 0)
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
d342894c 561/* Add new entry to forwarding table -- assumes lock held */
562static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 563 const u8 *mac, union vxlan_addr *ip,
6681712d 564 __u16 state, __u16 flags,
73cf3317 565 __be16 port, __u32 vni, __u32 ifindex,
ae884082 566 __u8 ndm_flags)
d342894c 567{
568 struct vxlan_fdb *f;
569 int notify = 0;
570
014be2c8 571 f = __vxlan_find_mac(vxlan, mac);
d342894c 572 if (f) {
573 if (flags & NLM_F_EXCL) {
574 netdev_dbg(vxlan->dev,
575 "lost race to create %pM\n", mac);
576 return -EEXIST;
577 }
578 if (f->state != state) {
579 f->state = state;
580 f->updated = jiffies;
581 notify = 1;
582 }
ae884082
DS		 583 if (f->flags != ndm_flags) {
584 f->flags = ndm_flags;
585 f->updated = jiffies;
586 notify = 1;
587 }
906dc186
TR		 588 if ((flags & NLM_F_REPLACE)) {
589 /* Only change unicasts */
590 if (!(is_multicast_ether_addr(f->eth_addr) ||
591 is_zero_ether_addr(f->eth_addr))) {
592 int rc = vxlan_fdb_replace(f, ip, port, vni,
593 ifindex);
594
595 if (rc < 0)
596 return rc;
597 notify |= rc;
598 } else
599 return -EOPNOTSUPP;
600 }
6681712d 601 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 602 (is_multicast_ether_addr(f->eth_addr) ||
603 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 604 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
605
606 if (rc < 0)
607 return rc;
608 notify |= rc;
609 }
d342894c 610 } else {
611 if (!(flags & NLM_F_CREATE))
612 return -ENOENT;
613
614 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
615 return -ENOSPC;
616
906dc186
TR		 617 /* Disallow replace to add a multicast entry */
618 if ((flags & NLM_F_REPLACE) &&
619 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
620 return -EOPNOTSUPP;
621
e4c7ed41 622 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 623 f = kmalloc(sizeof(*f), GFP_ATOMIC);
624 if (!f)
625 return -ENOMEM;
626
627 notify = 1;
d342894c 628 f->state = state;
ae884082 629 f->flags = ndm_flags;
d342894c 630 f->updated = f->used = jiffies;
3e61aa8f 631 INIT_LIST_HEAD(&f->remotes);
d342894c 632 memcpy(f->eth_addr, mac, ETH_ALEN);
633
3e61aa8f
SH		 634 vxlan_fdb_append(f, ip, port, vni, ifindex);
635
d342894c 636 ++vxlan->addrcnt;
637 hlist_add_head_rcu(&f->hlist,
638 vxlan_fdb_head(vxlan, mac));
639 }
640
641 if (notify)
642 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
643
644 return 0;
645}
646
6706c82e 647static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 648{
649 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 650 struct vxlan_rdst *rd, *nd;
6681712d 651
3e61aa8f 652 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 653 kfree(rd);
6681712d
DS		 654 kfree(f);
655}
656
d342894c 657static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
658{
659 netdev_dbg(vxlan->dev,
660 "delete %pM\n", f->eth_addr);
661
662 --vxlan->addrcnt;
663 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
664
665 hlist_del_rcu(&f->hlist);
6681712d 666 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 667}
668
f0b074be 669static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 670 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 671{
6681712d 672 struct net *net = dev_net(vxlan->dev);
e4c7ed41 673 int err;
d342894c 674
f0b074be 675 if (tb[NDA_DST]) {
e4c7ed41
CW		 676 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
677 if (err)
678 return err;
f0b074be 679 } else {
e4c7ed41
CW		 680 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
681 if (remote->sa.sa_family == AF_INET) {
682 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
683 ip->sa.sa_family = AF_INET;
684#if IS_ENABLED(CONFIG_IPV6)
685 } else {
686 ip->sin6.sin6_addr = in6addr_any;
687 ip->sa.sa_family = AF_INET6;
688#endif
689 }
f0b074be 690 }
d342894c 691
6681712d 692 if (tb[NDA_PORT]) {
73cf3317 693 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 694 return -EINVAL;
f0b074be
MR		 695 *port = nla_get_be16(tb[NDA_PORT]);
696 } else {
697 *port = vxlan->dst_port;
698 }
6681712d
DS		 699
700 if (tb[NDA_VNI]) {
701 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
702 return -EINVAL;
f0b074be
MR		 703 *vni = nla_get_u32(tb[NDA_VNI]);
704 } else {
705 *vni = vxlan->default_dst.remote_vni;
706 }
6681712d
DS		 707
708 if (tb[NDA_IFINDEX]) {
5abb0029 709 struct net_device *tdev;
6681712d
DS		 710
711 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
712 return -EINVAL;
f0b074be
MR		 713 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
714 tdev = dev_get_by_index(net, *ifindex);
5abb0029 715 if (!tdev)
6681712d 716 return -EADDRNOTAVAIL;
5abb0029 717 dev_put(tdev);
f0b074be
MR		 718 } else {
719 *ifindex = 0;
720 }
721
722 return 0;
723}
724
725/* Add static entry (via netlink) */
726static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
727 struct net_device *dev,
728 const unsigned char *addr, u16 flags)
729{
730 struct vxlan_dev *vxlan = netdev_priv(dev);
731 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 732 union vxlan_addr ip;
f0b074be
MR		 733 __be16 port;
734 u32 vni, ifindex;
735 int err;
736
737 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
738 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
739 ndm->ndm_state);
740 return -EINVAL;
741 }
742
743 if (tb[NDA_DST] == NULL)
744 return -EINVAL;
745
746 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
747 if (err)
748 return err;
6681712d 749
d342894c 750 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 751 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 752 port, vni, ifindex, ndm->ndm_flags);
d342894c 753 spin_unlock_bh(&vxlan->hash_lock);
754
755 return err;
756}
757
758/* Delete entry (via netlink) */
1690be63
VY		 759static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
760 struct net_device *dev,
d342894c 761 const unsigned char *addr)
762{
763 struct vxlan_dev *vxlan = netdev_priv(dev);
764 struct vxlan_fdb *f;
bc7892ba 765 struct vxlan_rdst *rd = NULL;
e4c7ed41 766 union vxlan_addr ip;
bc7892ba
MR		 767 __be16 port;
768 u32 vni, ifindex;
769 int err;
770
771 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
772 if (err)
773 return err;
774
775 err = -ENOENT;
d342894c 776
777 spin_lock_bh(&vxlan->hash_lock);
778 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 779 if (!f)
780 goto out;
781
e4c7ed41
CW		 782 if (!vxlan_addr_any(&ip)) {
783 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 784 if (!rd)
785 goto out;
786 }
787
788 err = 0;
789
790 /* remove a destination if it's not the only one on the list,
791 * otherwise destroy the fdb entry
792 */
793 if (rd && !list_is_singular(&f->remotes)) {
794 list_del_rcu(&rd->list);
dcdc7a59 795 kfree_rcu(rd, rcu);
bc7892ba 796 goto out;
d342894c 797 }
bc7892ba
MR		 798
799 vxlan_fdb_destroy(vxlan, f);
800
801out:
d342894c 802 spin_unlock_bh(&vxlan->hash_lock);
803
804 return err;
805}
806
807/* Dump forwarding table */
808static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
809 struct net_device *dev, int idx)
810{
811 struct vxlan_dev *vxlan = netdev_priv(dev);
812 unsigned int h;
813
814 for (h = 0; h < FDB_HASH_SIZE; ++h) {
815 struct vxlan_fdb *f;
d342894c 816 int err;
817
b67bfe0d 818 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 819 struct vxlan_rdst *rd;
6681712d 820
3e61aa8f
SH		 821 if (idx < cb->args[0])
822 goto skip;
823
824 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 825 err = vxlan_fdb_info(skb, vxlan, f,
826 NETLINK_CB(cb->skb).portid,
827 cb->nlh->nlmsg_seq,
828 RTM_NEWNEIGH,
829 NLM_F_MULTI, rd);
830 if (err < 0)
3e61aa8f 831 goto out;
6681712d 832 }
3e61aa8f
SH		 833skip:
834 ++idx;
d342894c 835 }
836 }
3e61aa8f 837out:
d342894c 838 return idx;
839}
840
841/* Watch incoming packets to learn mapping between Ethernet address
842 * and Tunnel endpoint.
26a41ae6 843 * Return true if packet is bogus and should be droppped.
d342894c 844 */
26a41ae6 845static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 846 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 847{
848 struct vxlan_dev *vxlan = netdev_priv(dev);
849 struct vxlan_fdb *f;
d342894c 850
851 f = vxlan_find_mac(vxlan, src_mac);
852 if (likely(f)) {
5ca5461c 853 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 854
e4c7ed41 855 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 856 return false;
857
858 /* Don't migrate static entries, drop packets */
eb064c3b 859 if (f->state & NUD_NOARP)
26a41ae6 860 return true;
d342894c 861
862 if (net_ratelimit())
863 netdev_info(dev,
e4c7ed41 864 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 865 src_mac, &rdst->remote_ip, &src_ip);
d342894c 866
e4c7ed41 867 rdst->remote_ip = *src_ip;
d342894c 868 f->updated = jiffies;
8385f50a 869 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 870 } else {
871 /* learned new entry */
872 spin_lock(&vxlan->hash_lock);
3bf74b1a 873
874 /* close off race between vxlan_flush and incoming packets */
875 if (netif_running(dev))
876 vxlan_fdb_create(vxlan, src_mac, src_ip,
877 NUD_REACHABLE,
878 NLM_F_EXCL|NLM_F_CREATE,
879 vxlan->dst_port,
880 vxlan->default_dst.remote_vni,
881 0, NTF_SELF);
d342894c 882 spin_unlock(&vxlan->hash_lock);
883 }
26a41ae6 884
885 return false;
d342894c 886}
887
d342894c 888/* See if multicast group is already in use by other ID */
e4c7ed41 889static bool vxlan_group_used(struct vxlan_net *vn, union vxlan_addr *remote_ip)
d342894c 890{
553675fb 891 struct vxlan_dev *vxlan;
d342894c 892
553675fb 893 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
553675fb 894 if (!netif_running(vxlan->dev))
895 continue;
d342894c 896
e4c7ed41
CW		 897 if (vxlan_addr_equal(&vxlan->default_dst.remote_ip,
898 remote_ip))
553675fb 899 return true;
900 }
d342894c 901
902 return false;
903}
904
7c47cedf 905static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 906{
7c47cedf
SH		 907 atomic_inc(&vs->refcnt);
908}
d342894c 909
012a5729 910void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 911{
012a5729
PS		 912 struct vxlan_net *vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
913
7c47cedf
SH		 914 if (!atomic_dec_and_test(&vs->refcnt))
915 return;
d342894c 916
1c51a915 917 spin_lock(&vn->sock_lock);
7c47cedf 918 hlist_del_rcu(&vs->hlist);
1c51a915
SH		 919 spin_unlock(&vn->sock_lock);
920
7c47cedf 921 queue_work(vxlan_wq, &vs->del_work);
d342894c 922}
012a5729 923EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 924
3fc2de2f 925/* Callback to update multicast group membership when first VNI on
926 * multicast asddress is brought up
927 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 928 */
3fc2de2f 929static void vxlan_igmp_join(struct work_struct *work)
d342894c 930{
3fc2de2f 931 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 932 struct vxlan_sock *vs = vxlan->vn_sock;
933 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 934 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
935 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 936
d342894c 937 lock_sock(sk);
e4c7ed41
CW		 938 if (ip->sa.sa_family == AF_INET) {
939 struct ip_mreqn mreq = {
940 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
941 .imr_ifindex = ifindex,
942 };
943
944 ip_mc_join_group(sk, &mreq);
945#if IS_ENABLED(CONFIG_IPV6)
946 } else {
947 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
948 &ip->sin6.sin6_addr);
949#endif
950 }
3fc2de2f 951 release_sock(sk);
952
012a5729 953 vxlan_sock_release(vs);
3fc2de2f 954 dev_put(vxlan->dev);
955}
956
957/* Inverse of vxlan_igmp_join when last VNI is brought down */
958static void vxlan_igmp_leave(struct work_struct *work)
959{
960 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 961 struct vxlan_sock *vs = vxlan->vn_sock;
962 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 963 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
964 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 965
966 lock_sock(sk);
e4c7ed41
CW		 967 if (ip->sa.sa_family == AF_INET) {
968 struct ip_mreqn mreq = {
969 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
970 .imr_ifindex = ifindex,
971 };
972
973 ip_mc_leave_group(sk, &mreq);
974#if IS_ENABLED(CONFIG_IPV6)
975 } else {
976 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
977 &ip->sin6.sin6_addr);
978#endif
979 }
980
d342894c 981 release_sock(sk);
d342894c 982
012a5729 983 vxlan_sock_release(vs);
7c47cedf 984 dev_put(vxlan->dev);
d342894c 985}
986
987/* Callback from net/ipv4/udp.c to receive packets */
988static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
989{
5cfccc5a 990 struct vxlan_sock *vs;
d342894c 991 struct vxlanhdr *vxh;
553675fb 992 __be16 port;
d342894c 993
d342894c 994 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 995 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 996 goto error;
997
7ce04758
PS		 998 /* Return packets with reserved bits set */
999 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 1000 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1001 (vxh->vx_vni & htonl(0xff))) {
1002 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1003 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1004 goto error;
1005 }
1006
5cfccc5a
PS		 1007 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1008 goto drop;
1009
553675fb 1010 port = inet_sk(sk)->inet_sport;
5cfccc5a
PS		 1011
1012 vs = vxlan_find_sock(sock_net(sk), port);
1013 if (!vs)
d342894c 1014 goto drop;
d342894c 1015
5cfccc5a
PS		 1016 vs->rcv(vs, skb, vxh->vx_vni);
1017 return 0;
1018
1019drop:
1020 /* Consume bad packet */
1021 kfree_skb(skb);
1022 return 0;
1023
1024error:
1025 /* Return non vxlan pkt */
1026 return 1;
1027}
1028
1029static void vxlan_rcv(struct vxlan_sock *vs,
1030 struct sk_buff *skb, __be32 vx_vni)
1031{
e4c7ed41
CW		 1032 struct iphdr *oip = NULL;
1033 struct ipv6hdr *oip6 = NULL;
5cfccc5a
PS		 1034 struct vxlan_dev *vxlan;
1035 struct pcpu_tstats *stats;
e4c7ed41 1036 union vxlan_addr saddr;
5cfccc5a 1037 __u32 vni;
e4c7ed41
CW		 1038 int err = 0;
1039 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1040
1041 vni = ntohl(vx_vni) >> 8;
1042 /* Is this VNI defined? */
1043 vxlan = vxlan_vs_find_vni(vs, vni);
1044 if (!vxlan)
d342894c 1045 goto drop;
d342894c 1046
e4c7ed41 1047 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1048 skb_reset_mac_header(skb);
d342894c 1049 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 1050
1051 /* Ignore packet loops (and multicast echo) */
1052 if (compare_ether_addr(eth_hdr(skb)->h_source,
1053 vxlan->dev->dev_addr) == 0)
1054 goto drop;
1055
7ce04758 1056 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1057 if (remote_ip->sa.sa_family == AF_INET) {
1058 oip = ip_hdr(skb);
1059 saddr.sin.sin_addr.s_addr = oip->saddr;
1060 saddr.sa.sa_family = AF_INET;
1061#if IS_ENABLED(CONFIG_IPV6)
1062 } else {
1063 oip6 = ipv6_hdr(skb);
1064 saddr.sin6.sin6_addr = oip6->saddr;
1065 saddr.sa.sa_family = AF_INET6;
1066#endif
1067 }
1068
26a41ae6 1069 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1070 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1071 goto drop;
d342894c 1072
d342894c 1073 skb_reset_network_header(skb);
0afb1666
JG		 1074
1075 /* If the NIC driver gave us an encapsulated packet with
1076 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1077 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1078 * for us. Otherwise force the upper layers to verify it.
1079 */
1080 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1081 !(vxlan->dev->features & NETIF_F_RXCSUM))
1082 skb->ip_summed = CHECKSUM_NONE;
1083
1084 skb->encapsulation = 0;
d342894c 1085
e4c7ed41
CW		 1086 if (oip6)
1087 err = IP6_ECN_decapsulate(oip6, skb);
1088 if (oip)
1089 err = IP_ECN_decapsulate(oip, skb);
1090
d342894c 1091 if (unlikely(err)) {
e4c7ed41
CW		 1092 if (log_ecn_error) {
1093 if (oip6)
1094 net_info_ratelimited("non-ECT from %pI6\n",
1095 &oip6->saddr);
1096 if (oip)
1097 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1098 &oip->saddr, oip->tos);
1099 }
d342894c 1100 if (err > 1) {
1101 ++vxlan->dev->stats.rx_frame_errors;
1102 ++vxlan->dev->stats.rx_errors;
1103 goto drop;
1104 }
1105 }
1106
e8171045 1107 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1108 u64_stats_update_begin(&stats->syncp);
1109 stats->rx_packets++;
1110 stats->rx_bytes += skb->len;
1111 u64_stats_update_end(&stats->syncp);
1112
1113 netif_rx(skb);
1114
5cfccc5a 1115 return;
d342894c 1116drop:
1117 /* Consume bad packet */
1118 kfree_skb(skb);
d342894c 1119}
1120
e4f67add
DS		 1121static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1122{
1123 struct vxlan_dev *vxlan = netdev_priv(dev);
1124 struct arphdr *parp;
1125 u8 *arpptr, *sha;
1126 __be32 sip, tip;
1127 struct neighbour *n;
1128
1129 if (dev->flags & IFF_NOARP)
1130 goto out;
1131
1132 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1133 dev->stats.tx_dropped++;
1134 goto out;
1135 }
1136 parp = arp_hdr(skb);
1137
1138 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1139 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1140 parp->ar_pro != htons(ETH_P_IP) ||
1141 parp->ar_op != htons(ARPOP_REQUEST) ||
1142 parp->ar_hln != dev->addr_len ||
1143 parp->ar_pln != 4)
1144 goto out;
1145 arpptr = (u8 *)parp + sizeof(struct arphdr);
1146 sha = arpptr;
1147 arpptr += dev->addr_len; /* sha */
1148 memcpy(&sip, arpptr, sizeof(sip));
1149 arpptr += sizeof(sip);
1150 arpptr += dev->addr_len; /* tha */
1151 memcpy(&tip, arpptr, sizeof(tip));
1152
1153 if (ipv4_is_loopback(tip) ||
1154 ipv4_is_multicast(tip))
1155 goto out;
1156
1157 n = neigh_lookup(&arp_tbl, &tip, dev);
1158
1159 if (n) {
e4f67add
DS		 1160 struct vxlan_fdb *f;
1161 struct sk_buff *reply;
1162
1163 if (!(n->nud_state & NUD_CONNECTED)) {
1164 neigh_release(n);
1165 goto out;
1166 }
1167
1168 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1169 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1170 /* bridge-local neighbor */
1171 neigh_release(n);
1172 goto out;
1173 }
1174
1175 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1176 n->ha, sha);
1177
1178 neigh_release(n);
1179
1180 skb_reset_mac_header(reply);
1181 __skb_pull(reply, skb_network_offset(reply));
1182 reply->ip_summed = CHECKSUM_UNNECESSARY;
1183 reply->pkt_type = PACKET_HOST;
1184
1185 if (netif_rx_ni(reply) == NET_RX_DROP)
1186 dev->stats.rx_dropped++;
e4c7ed41
CW		 1187 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1188 union vxlan_addr ipa = {
1189 .sin.sin_addr.s_addr = tip,
1190 .sa.sa_family = AF_INET,
1191 };
1192
1193 vxlan_ip_miss(dev, &ipa);
1194 }
e4f67add
DS		 1195out:
1196 consume_skb(skb);
1197 return NETDEV_TX_OK;
1198}
1199
f564f45c
CW		 1200#if IS_ENABLED(CONFIG_IPV6)
1201static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1202{
1203 struct vxlan_dev *vxlan = netdev_priv(dev);
1204 struct neighbour *n;
1205 union vxlan_addr ipa;
1206 const struct ipv6hdr *iphdr;
1207 const struct in6_addr *saddr, *daddr;
1208 struct nd_msg *msg;
1209 struct inet6_dev *in6_dev = NULL;
1210
1211 in6_dev = __in6_dev_get(dev);
1212 if (!in6_dev)
1213 goto out;
1214
1215 if (!pskb_may_pull(skb, skb->len))
1216 goto out;
1217
1218 iphdr = ipv6_hdr(skb);
1219 saddr = &iphdr->saddr;
1220 daddr = &iphdr->daddr;
1221
1222 if (ipv6_addr_loopback(daddr) ||
1223 ipv6_addr_is_multicast(daddr))
1224 goto out;
1225
1226 msg = (struct nd_msg *)skb_transport_header(skb);
1227 if (msg->icmph.icmp6_code != 0 ||
1228 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1229 goto out;
1230
1231 n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1232
1233 if (n) {
1234 struct vxlan_fdb *f;
1235
1236 if (!(n->nud_state & NUD_CONNECTED)) {
1237 neigh_release(n);
1238 goto out;
1239 }
1240
1241 f = vxlan_find_mac(vxlan, n->ha);
1242 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1243 /* bridge-local neighbor */
1244 neigh_release(n);
1245 goto out;
1246 }
1247
1248 ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1249 !!in6_dev->cnf.forwarding,
1250 true, false, false);
1251 neigh_release(n);
1252 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1253 ipa.sin6.sin6_addr = *daddr;
1254 ipa.sa.sa_family = AF_INET6;
1255 vxlan_ip_miss(dev, &ipa);
1256 }
1257
1258out:
1259 consume_skb(skb);
1260 return NETDEV_TX_OK;
1261}
1262#endif
1263
e4f67add
DS		 1264static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1265{
1266 struct vxlan_dev *vxlan = netdev_priv(dev);
1267 struct neighbour *n;
e4f67add
DS		 1268
1269 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1270 return false;
1271
1272 n = NULL;
1273 switch (ntohs(eth_hdr(skb)->h_proto)) {
1274 case ETH_P_IP:
e15a00aa
CW		 1275 {
1276 struct iphdr *pip;
1277
e4f67add
DS		 1278 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1279 return false;
1280 pip = ip_hdr(skb);
1281 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1282 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1283 union vxlan_addr ipa = {
1284 .sin.sin_addr.s_addr = pip->daddr,
1285 .sa.sa_family = AF_INET,
1286 };
1287
1288 vxlan_ip_miss(dev, &ipa);
1289 return false;
1290 }
1291
e4f67add 1292 break;
e15a00aa
CW		 1293 }
1294#if IS_ENABLED(CONFIG_IPV6)
1295 case ETH_P_IPV6:
1296 {
1297 struct ipv6hdr *pip6;
1298
1299 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1300 return false;
1301 pip6 = ipv6_hdr(skb);
1302 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1303 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1304 union vxlan_addr ipa = {
1305 .sin6.sin6_addr = pip6->daddr,
1306 .sa.sa_family = AF_INET6,
1307 };
1308
1309 vxlan_ip_miss(dev, &ipa);
1310 return false;
1311 }
1312
1313 break;
1314 }
1315#endif
e4f67add
DS		 1316 default:
1317 return false;
1318 }
1319
1320 if (n) {
1321 bool diff;
1322
1323 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
1324 if (diff) {
1325 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1326 dev->addr_len);
1327 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1328 }
1329 neigh_release(n);
1330 return diff;
e4c7ed41
CW		 1331 }
1332
e4f67add
DS		 1333 return false;
1334}
1335
553675fb 1336static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1337{
1338 sock_put(skb->sk);
1339}
1340
1341/* On transmit, associate with the tunnel socket */
49560532 1342static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1cad8715 1343{
1cad8715 1344 skb_orphan(skb);
1345 sock_hold(sk);
1346 skb->sk = sk;
553675fb 1347 skb->destructor = vxlan_sock_put;
1cad8715 1348}
1349
05f47d69 1350/* Compute source port for outgoing packet
1351 * first choice to use L4 flow hash since it will spread
1352 * better and maybe available from hardware
1353 * secondary choice is to use jhash on the Ethernet header
1354 */
49560532 1355__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
05f47d69 1356{
49560532 1357 unsigned int range = (port_max - port_min) + 1;
05f47d69 1358 u32 hash;
1359
1360 hash = skb_get_rxhash(skb);
1361 if (!hash)
1362 hash = jhash(skb->data, 2 * ETH_ALEN,
1363 (__force u32) skb->protocol);
1364
49560532 1365 return htons((((u64) hash * range) >> 32) + port_min);
05f47d69 1366}
49560532 1367EXPORT_SYMBOL_GPL(vxlan_src_port);
05f47d69 1368
05c0db08
PS		 1369static int handle_offloads(struct sk_buff *skb)
1370{
1371 if (skb_is_gso(skb)) {
1372 int err = skb_unclone(skb, GFP_ATOMIC);
1373 if (unlikely(err))
1374 return err;
1375
f6ace502 1376 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1377 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1378 skb->ip_summed = CHECKSUM_NONE;
1379
1380 return 0;
1381}
1382
e4c7ed41
CW		 1383#if IS_ENABLED(CONFIG_IPV6)
1384static int vxlan6_xmit_skb(struct net *net, struct vxlan_sock *vs,
1385 struct dst_entry *dst, struct sk_buff *skb,
1386 struct net_device *dev, struct in6_addr *saddr,
1387 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1388 __be16 src_port, __be16 dst_port, __be32 vni)
1389{
1390 struct ipv6hdr *ip6h;
1391 struct vxlanhdr *vxh;
1392 struct udphdr *uh;
1393 int min_headroom;
1394 int err;
1395
1396 if (!skb->encapsulation) {
1397 skb_reset_inner_headers(skb);
1398 skb->encapsulation = 1;
1399 }
1400
1401 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1402 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1403 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1404
1405 /* Need space for new headers (invalidates iph ptr) */
1406 err = skb_cow_head(skb, min_headroom);
1407 if (unlikely(err))
1408 return err;
1409
1410 if (vlan_tx_tag_present(skb)) {
1411 if (WARN_ON(!__vlan_put_tag(skb,
1412 skb->vlan_proto,
1413 vlan_tx_tag_get(skb))))
1414 return -ENOMEM;
1415
1416 skb->vlan_tci = 0;
1417 }
1418
1419 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1420 vxh->vx_flags = htonl(VXLAN_FLAGS);
1421 vxh->vx_vni = vni;
1422
1423 __skb_push(skb, sizeof(*uh));
1424 skb_reset_transport_header(skb);
1425 uh = udp_hdr(skb);
1426
1427 uh->dest = dst_port;
1428 uh->source = src_port;
1429
1430 uh->len = htons(skb->len);
1431 uh->check = 0;
1432
1433 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1434 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1435 IPSKB_REROUTED);
1436 skb_dst_drop(skb);
1437 skb_dst_set(skb, dst);
1438
1439 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1440 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1441 skb->ip_summed = CHECKSUM_UNNECESSARY;
1442 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1443 IPPROTO_UDP, csum);
1444 if (uh->check == 0)
1445 uh->check = CSUM_MANGLED_0;
1446 } else {
1447 skb->ip_summed = CHECKSUM_PARTIAL;
1448 skb->csum_start = skb_transport_header(skb) - skb->head;
1449 skb->csum_offset = offsetof(struct udphdr, check);
1450 uh->check = ~csum_ipv6_magic(saddr, daddr,
1451 skb->len, IPPROTO_UDP, 0);
1452 }
1453
1454 __skb_push(skb, sizeof(*ip6h));
1455 skb_reset_network_header(skb);
1456 ip6h = ipv6_hdr(skb);
1457 ip6h->version = 6;
1458 ip6h->priority = prio;
1459 ip6h->flow_lbl[0] = 0;
1460 ip6h->flow_lbl[1] = 0;
1461 ip6h->flow_lbl[2] = 0;
1462 ip6h->payload_len = htons(skb->len);
1463 ip6h->nexthdr = IPPROTO_UDP;
1464 ip6h->hop_limit = ttl;
1465 ip6h->daddr = *daddr;
1466 ip6h->saddr = *saddr;
1467
1468 vxlan_set_owner(vs->sock->sk, skb);
1469
1470 err = handle_offloads(skb);
1471 if (err)
1472 return err;
1473
1474 ip6tunnel_xmit(skb, dev);
1475 return 0;
1476}
1477#endif
1478
49560532
PS		 1479int vxlan_xmit_skb(struct net *net, struct vxlan_sock *vs,
1480 struct rtable *rt, struct sk_buff *skb,
1481 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1482 __be16 src_port, __be16 dst_port, __be32 vni)
1483{
1484 struct vxlanhdr *vxh;
1485 struct udphdr *uh;
649c5b8b 1486 int min_headroom;
49560532
PS		 1487 int err;
1488
1489 if (!skb->encapsulation) {
1490 skb_reset_inner_headers(skb);
1491 skb->encapsulation = 1;
1492 }
1493
649c5b8b 1494 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1495 + VXLAN_HLEN + sizeof(struct iphdr)
1496 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1497
1498 /* Need space for new headers (invalidates iph ptr) */
1499 err = skb_cow_head(skb, min_headroom);
1500 if (unlikely(err))
1501 return err;
1502
1eaa8178
PS		 1503 if (vlan_tx_tag_present(skb)) {
1504 if (WARN_ON(!__vlan_put_tag(skb,
1505 skb->vlan_proto,
1506 vlan_tx_tag_get(skb))))
1507 return -ENOMEM;
1508
1509 skb->vlan_tci = 0;
1510 }
1511
49560532
PS		 1512 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1513 vxh->vx_flags = htonl(VXLAN_FLAGS);
1514 vxh->vx_vni = vni;
1515
1516 __skb_push(skb, sizeof(*uh));
1517 skb_reset_transport_header(skb);
1518 uh = udp_hdr(skb);
1519
1520 uh->dest = dst_port;
1521 uh->source = src_port;
1522
1523 uh->len = htons(skb->len);
1524 uh->check = 0;
1525
1526 vxlan_set_owner(vs->sock->sk, skb);
1527
1528 err = handle_offloads(skb);
1529 if (err)
1530 return err;
1531
1532 return iptunnel_xmit(net, rt, skb, src, dst,
1533 IPPROTO_UDP, tos, ttl, df);
1534}
1535EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1536
9dcc71e1
SS		 1537/* Bypass encapsulation if the destination is local */
1538static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1539 struct vxlan_dev *dst_vxlan)
1540{
1541 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1542 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
e4c7ed41
CW		 1543 union vxlan_addr loopback;
1544 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1
SS		 1545
1546 skb->pkt_type = PACKET_HOST;
1547 skb->encapsulation = 0;
1548 skb->dev = dst_vxlan->dev;
1549 __skb_pull(skb, skb_network_offset(skb));
1550
e4c7ed41
CW		 1551 if (remote_ip->sa.sa_family == AF_INET) {
1552 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1553 loopback.sa.sa_family = AF_INET;
1554#if IS_ENABLED(CONFIG_IPV6)
1555 } else {
1556 loopback.sin6.sin6_addr = in6addr_loopback;
1557 loopback.sa.sa_family = AF_INET6;
1558#endif
1559 }
1560
9dcc71e1 1561 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1562 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1563
1564 u64_stats_update_begin(&tx_stats->syncp);
1565 tx_stats->tx_packets++;
1566 tx_stats->tx_bytes += skb->len;
1567 u64_stats_update_end(&tx_stats->syncp);
1568
1569 if (netif_rx(skb) == NET_RX_SUCCESS) {
1570 u64_stats_update_begin(&rx_stats->syncp);
1571 rx_stats->rx_packets++;
1572 rx_stats->rx_bytes += skb->len;
1573 u64_stats_update_end(&rx_stats->syncp);
1574 } else {
1575 skb->dev->stats.rx_dropped++;
1576 }
1577}
1578
4ad16930
SH		 1579static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1580 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1581{
1582 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1583 struct rtable *rt = NULL;
d342894c 1584 const struct iphdr *old_iph;
d342894c 1585 struct flowi4 fl4;
e4c7ed41
CW		 1586 union vxlan_addr *dst;
1587 __be16 src_port = 0, dst_port;
234f5b73 1588 u32 vni;
d342894c 1589 __be16 df = 0;
1590 __u8 tos, ttl;
0e6fbc5b 1591 int err;
d342894c 1592
823aa873 1593 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1594 vni = rdst->remote_vni;
e4c7ed41 1595 dst = &rdst->remote_ip;
e4f67add 1596
e4c7ed41 1597 if (vxlan_addr_any(dst)) {
e4f67add 1598 if (did_rsc) {
e4f67add 1599 /* short-circuited back to local bridge */
9dcc71e1 1600 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1601 return;
e4f67add 1602 }
ef59febe 1603 goto drop;
e4f67add 1604 }
ef59febe 1605
d342894c 1606 old_iph = ip_hdr(skb);
1607
d342894c 1608 ttl = vxlan->ttl;
e4c7ed41 1609 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1610 ttl = 1;
1611
1612 tos = vxlan->tos;
1613 if (tos == 1)
206aaafc 1614 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1615
49560532 1616 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
d342894c 1617
e4c7ed41
CW		 1618 if (dst->sa.sa_family == AF_INET) {
1619 memset(&fl4, 0, sizeof(fl4));
1620 fl4.flowi4_oif = rdst->remote_ifindex;
1621 fl4.flowi4_tos = RT_TOS(tos);
1622 fl4.daddr = dst->sin.sin_addr.s_addr;
1623 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1624
1625 rt = ip_route_output_key(dev_net(dev), &fl4);
1626 if (IS_ERR(rt)) {
1627 netdev_dbg(dev, "no route to %pI4\n",
1628 &dst->sin.sin_addr.s_addr);
1629 dev->stats.tx_carrier_errors++;
1630 goto tx_error;
1631 }
d342894c 1632
e4c7ed41
CW		 1633 if (rt->dst.dev == dev) {
1634 netdev_dbg(dev, "circular route to %pI4\n",
1635 &dst->sin.sin_addr.s_addr);
1636 dev->stats.collisions++;
1637 goto tx_error;
1638 }
1639
1640 /* Bypass encapsulation if the destination is local */
1641 if (rt->rt_flags & RTCF_LOCAL &&
1642 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1643 struct vxlan_dev *dst_vxlan;
1644
1645 ip_rt_put(rt);
1646 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1647 if (!dst_vxlan)
1648 goto tx_error;
1649 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1650 return;
1651 }
1652
1653 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1654 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1655
e4c7ed41
CW		 1656 err = vxlan_xmit_skb(dev_net(dev), vxlan->vn_sock, rt, skb,
1657 fl4.saddr, dst->sin.sin_addr.s_addr,
1658 tos, ttl, df, src_port, dst_port,
1659 htonl(vni << 8));
9dcc71e1 1660
e4c7ed41
CW		 1661 if (err < 0)
1662 goto rt_tx_error;
1663 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1664#if IS_ENABLED(CONFIG_IPV6)
1665 } else {
1666 struct sock *sk = vxlan->vn_sock->sock->sk;
1667 struct dst_entry *ndst;
1668 struct flowi6 fl6;
1669 u32 flags;
1670
1671 memset(&fl6, 0, sizeof(fl6));
1672 fl6.flowi6_oif = rdst->remote_ifindex;
1673 fl6.daddr = dst->sin6.sin6_addr;
1674 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1675 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1676
1677 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1678 netdev_dbg(dev, "no route to %pI6\n",
1679 &dst->sin6.sin6_addr);
1680 dev->stats.tx_carrier_errors++;
9dcc71e1 1681 goto tx_error;
e4c7ed41 1682 }
d342894c 1683
e4c7ed41
CW		 1684 if (ndst->dev == dev) {
1685 netdev_dbg(dev, "circular route to %pI6\n",
1686 &dst->sin6.sin6_addr);
1687 dst_release(ndst);
1688 dev->stats.collisions++;
1689 goto tx_error;
1690 }
0e6fbc5b 1691
e4c7ed41
CW		 1692 /* Bypass encapsulation if the destination is local */
1693 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1694 if (flags & RTF_LOCAL &&
1695 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1696 struct vxlan_dev *dst_vxlan;
1697
1698 dst_release(ndst);
1699 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1700 if (!dst_vxlan)
1701 goto tx_error;
1702 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1703 return;
1704 }
49560532 1705
e4c7ed41
CW		 1706 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1707
1708 err = vxlan6_xmit_skb(dev_net(dev), vxlan->vn_sock, ndst, skb,
1709 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1710 src_port, dst_port, htonl(vni << 8));
1711#endif
1712 }
0e6fbc5b 1713
4ad16930 1714 return;
d342894c 1715
1716drop:
1717 dev->stats.tx_dropped++;
1718 goto tx_free;
1719
49560532
PS		 1720rt_tx_error:
1721 ip_rt_put(rt);
d342894c 1722tx_error:
1723 dev->stats.tx_errors++;
1724tx_free:
1725 dev_kfree_skb(skb);
d342894c 1726}
1727
6681712d
DS		 1728/* Transmit local packets over Vxlan
1729 *
1730 * Outer IP header inherits ECN and DF from inner header.
1731 * Outer UDP destination is the VXLAN assigned port.
1732 * source port is based on hash of flow
1733 */
1734static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1735{
1736 struct vxlan_dev *vxlan = netdev_priv(dev);
1737 struct ethhdr *eth;
1738 bool did_rsc = false;
afbd8bae 1739 struct vxlan_rdst *rdst;
6681712d 1740 struct vxlan_fdb *f;
6681712d
DS		 1741
1742 skb_reset_mac_header(skb);
1743 eth = eth_hdr(skb);
1744
f564f45c
CW		 1745 if ((vxlan->flags & VXLAN_F_PROXY)) {
1746 if (ntohs(eth->h_proto) == ETH_P_ARP)
1747 return arp_reduce(dev, skb);
1748#if IS_ENABLED(CONFIG_IPV6)
1749 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1750 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1751 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1752 struct nd_msg *msg;
1753
1754 msg = (struct nd_msg *)skb_transport_header(skb);
1755 if (msg->icmph.icmp6_code == 0 &&
1756 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1757 return neigh_reduce(dev, skb);
1758 }
1759#endif
1760 }
6681712d
DS		 1761
1762 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1763 did_rsc = false;
1764
1765 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1766 (ntohs(eth->h_proto) == ETH_P_IP ||
1767 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1768 did_rsc = route_shortcircuit(dev, skb);
1769 if (did_rsc)
1770 f = vxlan_find_mac(vxlan, eth->h_dest);
1771 }
1772
6681712d 1773 if (f == NULL) {
afbd8bae
MR		 1774 f = vxlan_find_mac(vxlan, all_zeros_mac);
1775 if (f == NULL) {
1776 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1777 !is_multicast_ether_addr(eth->h_dest))
1778 vxlan_fdb_miss(vxlan, eth->h_dest);
1779
1780 dev->stats.tx_dropped++;
1781 dev_kfree_skb(skb);
1782 return NETDEV_TX_OK;
1783 }
1784 }
6681712d 1785
afbd8bae
MR		 1786 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1787 struct sk_buff *skb1;
6681712d 1788
afbd8bae
MR		 1789 skb1 = skb_clone(skb, GFP_ATOMIC);
1790 if (skb1)
1791 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1792 }
1793
afbd8bae 1794 dev_kfree_skb(skb);
4ad16930 1795 return NETDEV_TX_OK;
6681712d
DS		 1796}
1797
d342894c 1798/* Walk the forwarding table and purge stale entries */
1799static void vxlan_cleanup(unsigned long arg)
1800{
1801 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1802 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1803 unsigned int h;
1804
1805 if (!netif_running(vxlan->dev))
1806 return;
1807
1808 spin_lock_bh(&vxlan->hash_lock);
1809 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1810 struct hlist_node *p, *n;
1811 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1812 struct vxlan_fdb *f
1813 = container_of(p, struct vxlan_fdb, hlist);
1814 unsigned long timeout;
1815
3c172868 1816 if (f->state & NUD_PERMANENT)
d342894c 1817 continue;
1818
1819 timeout = f->used + vxlan->age_interval * HZ;
1820 if (time_before_eq(timeout, jiffies)) {
1821 netdev_dbg(vxlan->dev,
1822 "garbage collect %pM\n",
1823 f->eth_addr);
1824 f->state = NUD_STALE;
1825 vxlan_fdb_destroy(vxlan, f);
1826 } else if (time_before(timeout, next_timer))
1827 next_timer = timeout;
1828 }
1829 }
1830 spin_unlock_bh(&vxlan->hash_lock);
1831
1832 mod_timer(&vxlan->age_timer, next_timer);
1833}
1834
9c2e24e1
PS		 1835static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1836{
1837 __u32 vni = vxlan->default_dst.remote_vni;
1838
1839 vxlan->vn_sock = vs;
1840 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1841}
1842
d342894c 1843/* Setup stats when device is created */
1844static int vxlan_init(struct net_device *dev)
1845{
1c51a915
SH		 1846 struct vxlan_dev *vxlan = netdev_priv(dev);
1847 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1848 struct vxlan_sock *vs;
1c51a915 1849
e8171045
PS		 1850 dev->tstats = alloc_percpu(struct pcpu_tstats);
1851 if (!dev->tstats)
d342894c 1852 return -ENOMEM;
1853
1c51a915 1854 spin_lock(&vn->sock_lock);
9c2e24e1 1855 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1c51a915
SH		 1856 if (vs) {
1857 /* If we have a socket with same port already, reuse it */
1858 atomic_inc(&vs->refcnt);
9c2e24e1 1859 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1860 } else {
1861 /* otherwise make new socket outside of RTNL */
1862 dev_hold(dev);
1863 queue_work(vxlan_wq, &vxlan->sock_work);
1864 }
1865 spin_unlock(&vn->sock_lock);
1866
d342894c 1867 return 0;
1868}
1869
ba609e9b 1870static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1871{
1872 struct vxlan_fdb *f;
1873
1874 spin_lock_bh(&vxlan->hash_lock);
1875 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1876 if (f)
1877 vxlan_fdb_destroy(vxlan, f);
1878 spin_unlock_bh(&vxlan->hash_lock);
1879}
1880
ebf4063e
SH		 1881static void vxlan_uninit(struct net_device *dev)
1882{
1883 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 1884 struct vxlan_sock *vs = vxlan->vn_sock;
1885
ba609e9b 1886 vxlan_fdb_delete_default(vxlan);
afbd8bae 1887
ebf4063e 1888 if (vs)
012a5729 1889 vxlan_sock_release(vs);
ebf4063e
SH		 1890 free_percpu(dev->tstats);
1891}
1892
d342894c 1893/* Start ageing timer and join group when device is brought up */
1894static int vxlan_open(struct net_device *dev)
1895{
3fc2de2f 1896 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1897 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1898 struct vxlan_sock *vs = vxlan->vn_sock;
1899
1900 /* socket hasn't been created */
1901 if (!vs)
1902 return -ENOTCONN;
d342894c 1903
e4c7ed41
CW		 1904 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1905 vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1906 vxlan_sock_hold(vs);
7c47cedf 1907 dev_hold(dev);
3fc2de2f 1908 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 1909 }
1910
1911 if (vxlan->age_interval)
1912 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1913
1914 return 0;
1915}
1916
1917/* Purge the forwarding table */
1918static void vxlan_flush(struct vxlan_dev *vxlan)
1919{
31fec5aa 1920 unsigned int h;
d342894c 1921
1922 spin_lock_bh(&vxlan->hash_lock);
1923 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1924 struct hlist_node *p, *n;
1925 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1926 struct vxlan_fdb *f
1927 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 1928 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1929 if (!is_zero_ether_addr(f->eth_addr))
1930 vxlan_fdb_destroy(vxlan, f);
d342894c 1931 }
1932 }
1933 spin_unlock_bh(&vxlan->hash_lock);
1934}
1935
1936/* Cleanup timer and forwarding table on shutdown */
1937static int vxlan_stop(struct net_device *dev)
1938{
3fc2de2f 1939 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1940 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1941 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1942
e4c7ed41
CW		 1943 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1944 ! vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1c51a915 1945 vxlan_sock_hold(vs);
7c47cedf 1946 dev_hold(dev);
3fc2de2f 1947 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 1948 }
d342894c 1949
1950 del_timer_sync(&vxlan->age_timer);
1951
1952 vxlan_flush(vxlan);
1953
1954 return 0;
1955}
1956
d342894c 1957/* Stub, nothing needs to be done. */
1958static void vxlan_set_multicast_list(struct net_device *dev)
1959{
1960}
1961
1962static const struct net_device_ops vxlan_netdev_ops = {
1963 .ndo_init = vxlan_init,
ebf4063e 1964 .ndo_uninit = vxlan_uninit,
d342894c 1965 .ndo_open = vxlan_open,
1966 .ndo_stop = vxlan_stop,
1967 .ndo_start_xmit = vxlan_xmit,
e8171045 1968 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 1969 .ndo_set_rx_mode = vxlan_set_multicast_list,
1970 .ndo_change_mtu = eth_change_mtu,
1971 .ndo_validate_addr = eth_validate_addr,
1972 .ndo_set_mac_address = eth_mac_addr,
1973 .ndo_fdb_add = vxlan_fdb_add,
1974 .ndo_fdb_del = vxlan_fdb_delete,
1975 .ndo_fdb_dump = vxlan_fdb_dump,
1976};
1977
1978/* Info for udev, that this is a virtual tunnel endpoint */
1979static struct device_type vxlan_type = {
1980 .name = "vxlan",
1981};
1982
d342894c 1983/* Initialize the device structure. */
1984static void vxlan_setup(struct net_device *dev)
1985{
1986 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 1987 unsigned int h;
05f47d69 1988 int low, high;
d342894c 1989
1990 eth_hw_addr_random(dev);
1991 ether_setup(dev);
e4c7ed41
CW		 1992 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
1993 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
1994 else
1995 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1996
1997 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 1998 dev->destructor = free_netdev;
d342894c 1999 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2000
2001 dev->tx_queue_len = 0;
2002 dev->features |= NETIF_F_LLTX;
2003 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 2004 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2005 dev->features |= NETIF_F_RXCSUM;
05c0db08 2006 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2007
1eaa8178
PS		 2008 dev->vlan_features = dev->features;
2009 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2010 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2011 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2012 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2013 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2014 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2015
553675fb 2016 INIT_LIST_HEAD(&vxlan->next);
d342894c 2017 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2018 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2019 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2020 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2021
2022 init_timer_deferrable(&vxlan->age_timer);
2023 vxlan->age_timer.function = vxlan_cleanup;
2024 vxlan->age_timer.data = (unsigned long) vxlan;
2025
05f47d69 2026 inet_get_local_port_range(&low, &high);
2027 vxlan->port_min = low;
2028 vxlan->port_max = high;
823aa873 2029 vxlan->dst_port = htons(vxlan_port);
05f47d69 2030
d342894c 2031 vxlan->dev = dev;
2032
2033 for (h = 0; h < FDB_HASH_SIZE; ++h)
2034 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2035}
2036
2037static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2038 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2039 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2040 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2041 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2042 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2043 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2044 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2045 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2046 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2047 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2048 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2049 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2050 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2051 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2052 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2053 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2054 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2055};
2056
2057static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2058{
2059 if (tb[IFLA_ADDRESS]) {
2060 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2061 pr_debug("invalid link address (not ethernet)\n");
2062 return -EINVAL;
2063 }
2064
2065 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2066 pr_debug("invalid all zero ethernet address\n");
2067 return -EADDRNOTAVAIL;
2068 }
2069 }
2070
2071 if (!data)
2072 return -EINVAL;
2073
2074 if (data[IFLA_VXLAN_ID]) {
2075 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2076 if (id >= VXLAN_VID_MASK)
2077 return -ERANGE;
2078 }
2079
05f47d69 2080 if (data[IFLA_VXLAN_PORT_RANGE]) {
2081 const struct ifla_vxlan_port_range *p
2082 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2083
2084 if (ntohs(p->high) < ntohs(p->low)) {
2085 pr_debug("port range %u .. %u not valid\n",
2086 ntohs(p->low), ntohs(p->high));
2087 return -EINVAL;
2088 }
2089 }
2090
d342894c 2091 return 0;
2092}
2093
1b13c97f
YB		 2094static void vxlan_get_drvinfo(struct net_device *netdev,
2095 struct ethtool_drvinfo *drvinfo)
2096{
2097 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2098 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2099}
2100
2101static const struct ethtool_ops vxlan_ethtool_ops = {
2102 .get_drvinfo = vxlan_get_drvinfo,
2103 .get_link = ethtool_op_get_link,
2104};
2105
553675fb 2106static void vxlan_del_work(struct work_struct *work)
2107{
2108 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2109
2110 sk_release_kernel(vs->sock->sk);
2111 kfree_rcu(vs, rcu);
2112}
2113
e4c7ed41
CW		 2114#if IS_ENABLED(CONFIG_IPV6)
2115/* Create UDP socket for encapsulation receive. AF_INET6 socket
2116 * could be used for both IPv4 and IPv6 communications, but
2117 * users may set bindv6only=1.
2118 */
2119static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
553675fb 2120{
553675fb 2121 struct sock *sk;
e4c7ed41
CW		 2122 struct socket *sock;
2123 struct sockaddr_in6 vxlan_addr = {
2124 .sin6_family = AF_INET6,
2125 .sin6_port = port,
2126 };
2127 int rc, val = 1;
2128
2129 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2130 if (rc < 0) {
2131 pr_debug("UDPv6 socket create failed\n");
2132 return rc;
2133 }
2134
2135 /* Put in proper namespace */
2136 sk = sock->sk;
2137 sk_change_net(sk, net);
2138
2139 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2140 (char *)&val, sizeof(val));
2141 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2142 sizeof(struct sockaddr_in6));
2143 if (rc < 0) {
2144 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2145 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2146 sk_release_kernel(sk);
2147 return rc;
2148 }
2149 /* At this point, IPv6 module should have been loaded in
2150 * sock_create_kern().
2151 */
2152 BUG_ON(!ipv6_stub);
2153
2154 *psock = sock;
2155 /* Disable multicast loopback */
2156 inet_sk(sk)->mc_loop = 0;
2157 return 0;
2158}
2159
2160#else
2161
2162static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2163{
2164 return -EPFNOSUPPORT;
2165}
2166#endif
2167
2168static int create_v4_sock(struct net *net, __be16 port, struct socket **psock)
2169{
2170 struct sock *sk;
2171 struct socket *sock;
553675fb 2172 struct sockaddr_in vxlan_addr = {
2173 .sin_family = AF_INET,
2174 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2175 .sin_port = port,
553675fb 2176 };
2177 int rc;
553675fb 2178
2179 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2180 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2181 if (rc < 0) {
2182 pr_debug("UDP socket create failed\n");
e4c7ed41 2183 return rc;
553675fb 2184 }
2185
2186 /* Put in proper namespace */
e4c7ed41 2187 sk = sock->sk;
553675fb 2188 sk_change_net(sk, net);
2189
e4c7ed41 2190 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2191 sizeof(vxlan_addr));
2192 if (rc < 0) {
2193 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2194 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2195 sk_release_kernel(sk);
e4c7ed41
CW		 2196 return rc;
2197 }
2198
2199 *psock = sock;
2200 /* Disable multicast loopback */
2201 inet_sk(sk)->mc_loop = 0;
2202 return 0;
2203}
2204
2205/* Create new listen socket if needed */
2206static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2207 vxlan_rcv_t *rcv, void *data, bool ipv6)
2208{
2209 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2210 struct vxlan_sock *vs;
2211 struct socket *sock;
2212 struct sock *sk;
2213 int rc = 0;
2214 unsigned int h;
2215
2216 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2217 if (!vs)
2218 return ERR_PTR(-ENOMEM);
2219
2220 for (h = 0; h < VNI_HASH_SIZE; ++h)
2221 INIT_HLIST_HEAD(&vs->vni_list[h]);
2222
2223 INIT_WORK(&vs->del_work, vxlan_del_work);
2224
2225 if (ipv6)
2226 rc = create_v6_sock(net, port, &sock);
2227 else
2228 rc = create_v4_sock(net, port, &sock);
2229 if (rc < 0) {
553675fb 2230 kfree(vs);
2231 return ERR_PTR(rc);
2232 }
e4c7ed41
CW		 2233
2234 vs->sock = sock;
2235 sk = sock->sk;
9c2e24e1 2236 atomic_set(&vs->refcnt, 1);
5cfccc5a 2237 vs->rcv = rcv;
012a5729 2238 vs->data = data;
553675fb 2239
9c2e24e1
PS		 2240 spin_lock(&vn->sock_lock);
2241 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2242 spin_unlock(&vn->sock_lock);
553675fb 2243
2244 /* Mark socket as an encapsulation socket. */
2245 udp_sk(sk)->encap_type = 1;
2246 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2247#if IS_ENABLED(CONFIG_IPV6)
2248 if (ipv6)
2249 ipv6_stub->udpv6_encap_enable();
2250 else
2251#endif
2252 udp_encap_enable();
2253
9c2e24e1
PS		 2254 return vs;
2255}
2256
012a5729
PS		 2257struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2258 vxlan_rcv_t *rcv, void *data,
e4c7ed41 2259 bool no_share, bool ipv6)
9c2e24e1
PS		 2260{
2261 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2262 struct vxlan_sock *vs;
2263
e4c7ed41 2264 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
9c2e24e1
PS		 2265 if (!IS_ERR(vs))
2266 return vs;
553675fb 2267
012a5729
PS		 2268 if (no_share) /* Return error if sharing is not allowed. */
2269 return vs;
2270
9c2e24e1
PS		 2271 spin_lock(&vn->sock_lock);
2272 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2273 if (vs) {
2274 if (vs->rcv == rcv)
2275 atomic_inc(&vs->refcnt);
2276 else
2277 vs = ERR_PTR(-EBUSY);
2278 }
2279 spin_unlock(&vn->sock_lock);
2280
2281 if (!vs)
9c2e24e1
PS		 2282 vs = ERR_PTR(-EINVAL);
2283
553675fb 2284 return vs;
2285}
012a5729 2286EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2287
1c51a915
SH		 2288/* Scheduled at device creation to bind to a socket */
2289static void vxlan_sock_work(struct work_struct *work)
2290{
9c2e24e1
PS		 2291 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2292 struct net *net = dev_net(vxlan->dev);
1c51a915 2293 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2294 __be16 port = vxlan->dst_port;
2295 struct vxlan_sock *nvs;
1c51a915 2296
e4c7ed41 2297 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
1c51a915 2298 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2299 if (!IS_ERR(nvs))
2300 vxlan_vs_add_dev(nvs, vxlan);
2301 spin_unlock(&vn->sock_lock);
2302
2303 dev_put(vxlan->dev);
1c51a915
SH		 2304}
2305
d342894c 2306static int vxlan_newlink(struct net *net, struct net_device *dev,
2307 struct nlattr *tb[], struct nlattr *data[])
2308{
553675fb 2309 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2310 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2311 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2312 __u32 vni;
2313 int err;
e4c7ed41 2314 bool use_ipv6 = false;
d342894c 2315
2316 if (!data[IFLA_VXLAN_ID])
2317 return -EINVAL;
2318
2319 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2320 dst->remote_vni = vni;
d342894c 2321
e4c7ed41
CW		 2322 if (data[IFLA_VXLAN_GROUP]) {
2323 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2324 dst->remote_ip.sa.sa_family = AF_INET;
2325 } else if (data[IFLA_VXLAN_GROUP6]) {
2326 if (!IS_ENABLED(CONFIG_IPV6))
2327 return -EPFNOSUPPORT;
2328
2329 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2330 sizeof(struct in6_addr));
2331 dst->remote_ip.sa.sa_family = AF_INET6;
2332 use_ipv6 = true;
2333 }
d342894c 2334
e4c7ed41
CW		 2335 if (data[IFLA_VXLAN_LOCAL]) {
2336 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2337 vxlan->saddr.sa.sa_family = AF_INET;
2338 } else if (data[IFLA_VXLAN_LOCAL6]) {
2339 if (!IS_ENABLED(CONFIG_IPV6))
2340 return -EPFNOSUPPORT;
2341
2342 /* TODO: respect scope id */
2343 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2344 sizeof(struct in6_addr));
2345 vxlan->saddr.sa.sa_family = AF_INET6;
2346 use_ipv6 = true;
2347 }
d342894c 2348
34e02aa1 2349 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2350 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2351 struct net_device *lowerdev
c7995c43 2352 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2353
2354 if (!lowerdev) {
c7995c43 2355 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2356 return -ENODEV;
2357 }
d342894c 2358
e4c7ed41
CW		 2359#if IS_ENABLED(CONFIG_IPV6)
2360 if (use_ipv6) {
2361 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2362 if (idev && idev->cnf.disable_ipv6) {
2363 pr_info("IPv6 is disabled via sysctl\n");
2364 return -EPERM;
2365 }
2366 vxlan->flags |= VXLAN_F_IPV6;
2367 }
2368#endif
2369
34e02aa1 2370 if (!tb[IFLA_MTU])
e4c7ed41 2371 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4
AD		 2372
2373 /* update header length based on lower device */
2374 dev->hard_header_len = lowerdev->hard_header_len +
e4c7ed41 2375 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
d342894c 2376 }
2377
2378 if (data[IFLA_VXLAN_TOS])
2379 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2380
afb97186
VB		 2381 if (data[IFLA_VXLAN_TTL])
2382 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2383
d342894c 2384 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2385 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2386
2387 if (data[IFLA_VXLAN_AGEING])
2388 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2389 else
2390 vxlan->age_interval = FDB_AGE_DEFAULT;
2391
e4f67add
DS		 2392 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2393 vxlan->flags |= VXLAN_F_PROXY;
2394
2395 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2396 vxlan->flags |= VXLAN_F_RSC;
2397
2398 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2399 vxlan->flags |= VXLAN_F_L2MISS;
2400
2401 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2402 vxlan->flags |= VXLAN_F_L3MISS;
2403
d342894c 2404 if (data[IFLA_VXLAN_LIMIT])
2405 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2406
05f47d69 2407 if (data[IFLA_VXLAN_PORT_RANGE]) {
2408 const struct ifla_vxlan_port_range *p
2409 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2410 vxlan->port_min = ntohs(p->low);
2411 vxlan->port_max = ntohs(p->high);
2412 }
2413
823aa873 2414 if (data[IFLA_VXLAN_PORT])
2415 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2416
553675fb 2417 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2418 pr_info("duplicate VNI %u\n", vni);
2419 return -EEXIST;
2420 }
2421
1b13c97f
YB		 2422 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2423
afbd8bae
MR		 2424 /* create an fdb entry for default destination */
2425 err = vxlan_fdb_create(vxlan, all_zeros_mac,
e4c7ed41 2426 &vxlan->default_dst.remote_ip,
afbd8bae
MR		 2427 NUD_REACHABLE|NUD_PERMANENT,
2428 NLM_F_EXCL|NLM_F_CREATE,
2429 vxlan->dst_port, vxlan->default_dst.remote_vni,
2430 vxlan->default_dst.remote_ifindex, NTF_SELF);
1c51a915 2431 if (err)
553675fb 2432 return err;
d342894c 2433
afbd8bae
MR		 2434 err = register_netdevice(dev);
2435 if (err) {
ba609e9b 2436 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2437 return err;
2438 }
2439
553675fb 2440 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2441
2442 return 0;
d342894c 2443}
2444
2445static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2446{
fe5c3561 2447 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 2448 struct vxlan_dev *vxlan = netdev_priv(dev);
2449
fe5c3561 2450 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2451 if (!hlist_unhashed(&vxlan->hlist))
2452 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2453 spin_unlock(&vn->sock_lock);
2454
553675fb 2455 list_del(&vxlan->next);
d342894c 2456 unregister_netdevice_queue(dev, head);
2457}
2458
2459static size_t vxlan_get_size(const struct net_device *dev)
2460{
2461
2462 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2463 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2464 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2465 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2466 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2467 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2468 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2469 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2470 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2471 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2472 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2473 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2474 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2475 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 2476 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 2477 0;
2478}
2479
2480static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2481{
2482 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2483 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2484 struct ifla_vxlan_port_range ports = {
2485 .low = htons(vxlan->port_min),
2486 .high = htons(vxlan->port_max),
2487 };
d342894c 2488
c7995c43 2489 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2490 goto nla_put_failure;
2491
e4c7ed41
CW		 2492 if (!vxlan_addr_any(&dst->remote_ip)) {
2493 if (dst->remote_ip.sa.sa_family == AF_INET) {
2494 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2495 dst->remote_ip.sin.sin_addr.s_addr))
2496 goto nla_put_failure;
2497#if IS_ENABLED(CONFIG_IPV6)
2498 } else {
2499 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2500 &dst->remote_ip.sin6.sin6_addr))
2501 goto nla_put_failure;
2502#endif
2503 }
2504 }
d342894c 2505
c7995c43 2506 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2507 goto nla_put_failure;
2508
e4c7ed41
CW		 2509 if (!vxlan_addr_any(&vxlan->saddr)) {
2510 if (vxlan->saddr.sa.sa_family == AF_INET) {
2511 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2512 vxlan->saddr.sin.sin_addr.s_addr))
2513 goto nla_put_failure;
2514#if IS_ENABLED(CONFIG_IPV6)
2515 } else {
2516 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2517 &vxlan->saddr.sin6.sin6_addr))
2518 goto nla_put_failure;
2519#endif
2520 }
2521 }
d342894c 2522
2523 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2524 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2525 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2526 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2527 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2528 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2529 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2530 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2531 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2532 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2533 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2534 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2535 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2536 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 2537 goto nla_put_failure;
2538
05f47d69 2539 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2540 goto nla_put_failure;
2541
d342894c 2542 return 0;
2543
2544nla_put_failure:
2545 return -EMSGSIZE;
2546}
2547
2548static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2549 .kind = "vxlan",
2550 .maxtype = IFLA_VXLAN_MAX,
2551 .policy = vxlan_policy,
2552 .priv_size = sizeof(struct vxlan_dev),
2553 .setup = vxlan_setup,
2554 .validate = vxlan_validate,
2555 .newlink = vxlan_newlink,
2556 .dellink = vxlan_dellink,
2557 .get_size = vxlan_get_size,
2558 .fill_info = vxlan_fill_info,
2559};
2560
2561static __net_init int vxlan_init_net(struct net *net)
2562{
2563 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2564 unsigned int h;
d342894c 2565
553675fb 2566 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2567 spin_lock_init(&vn->sock_lock);
d342894c 2568
553675fb 2569 for (h = 0; h < PORT_HASH_SIZE; ++h)
2570 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2571
2572 return 0;
2573}
2574
2575static __net_exit void vxlan_exit_net(struct net *net)
2576{
2577 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 2578 struct vxlan_dev *vxlan;
372675a4 2579 LIST_HEAD(list);
9cb6cb7e
ZM		 2580
2581 rtnl_lock();
553675fb 2582 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 2583 unregister_netdevice_queue(vxlan->dev, &list);
2584 unregister_netdevice_many(&list);
9cb6cb7e 2585 rtnl_unlock();
d342894c 2586}
2587
2588static struct pernet_operations vxlan_net_ops = {
2589 .init = vxlan_init_net,
2590 .exit = vxlan_exit_net,
2591 .id = &vxlan_net_id,
2592 .size = sizeof(struct vxlan_net),
2593};
2594
2595static int __init vxlan_init_module(void)
2596{
2597 int rc;
2598
758c57d1
SH		 2599 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2600 if (!vxlan_wq)
2601 return -ENOMEM;
2602
d342894c 2603 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2604
2605 rc = register_pernet_device(&vxlan_net_ops);
2606 if (rc)
2607 goto out1;
2608
2609 rc = rtnl_link_register(&vxlan_link_ops);
2610 if (rc)
2611 goto out2;
2612
2613 return 0;
2614
2615out2:
2616 unregister_pernet_device(&vxlan_net_ops);
2617out1:
758c57d1 2618 destroy_workqueue(vxlan_wq);
d342894c 2619 return rc;
2620}
7332a13b 2621late_initcall(vxlan_init_module);
d342894c 2622
2623static void __exit vxlan_cleanup_module(void)
2624{
b7153984 2625 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 2626 destroy_workqueue(vxlan_wq);
f89e57c4 2627 unregister_pernet_device(&vxlan_net_ops);
6681712d 2628 rcu_barrier();
d342894c 2629}
2630module_exit(vxlan_cleanup_module);
2631
2632MODULE_LICENSE("GPL");
2633MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2634MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 2635MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.287739 seconds and 5 git commands to generate.