projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
r8152: replace tasklet with NAPI
[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>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
acbf74a7 45#include <net/udp_tunnel.h>
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
d342894c 52
53#define VXLAN_VERSION "0.1"
54
553675fb 55#define PORT_HASH_BITS 8
56#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 57#define VNI_HASH_BITS 10
58#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
59#define FDB_HASH_BITS 8
60#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
61#define FDB_AGE_DEFAULT 300 /* 5 min */
62#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
63
23c578bf 64/* UDP port for VXLAN traffic.
65 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 66 * for compatibility with early adopters.
23c578bf 67 */
9daaa397
SH		 68static unsigned short vxlan_port __read_mostly = 8472;
69module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 70MODULE_PARM_DESC(udp_port, "Destination UDP port");
71
72static bool log_ecn_error = true;
73module_param(log_ecn_error, bool, 0644);
74MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
75
60d9d4c6 76static int vxlan_net_id;
553675fb 77
afbd8bae
MR		 78static const u8 all_zeros_mac[ETH_ALEN];
79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
e4c7ed41
CW		 87union vxlan_addr {
88 struct sockaddr_in sin;
89 struct sockaddr_in6 sin6;
90 struct sockaddr sa;
91};
92
6681712d 93struct vxlan_rdst {
e4c7ed41 94 union vxlan_addr remote_ip;
6681712d
DS		 95 __be16 remote_port;
96 u32 remote_vni;
97 u32 remote_ifindex;
3e61aa8f 98 struct list_head list;
bc7892ba 99 struct rcu_head rcu;
6681712d
DS		 100};
101
d342894c 102/* Forwarding table entry */
103struct vxlan_fdb {
104 struct hlist_node hlist; /* linked list of entries */
105 struct rcu_head rcu;
106 unsigned long updated; /* jiffies */
107 unsigned long used;
3e61aa8f 108 struct list_head remotes;
d342894c 109 u16 state; /* see ndm_state */
ae884082 110 u8 flags; /* see ndm_flags */
d342894c 111 u8 eth_addr[ETH_ALEN];
112};
113
d342894c 114/* Pseudo network device */
115struct vxlan_dev {
553675fb 116 struct hlist_node hlist; /* vni hash table */
117 struct list_head next; /* vxlan's per namespace list */
118 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 119 struct net_device *dev;
f01ec1c0 120 struct net *net; /* netns for packet i/o */
c7995c43 121 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 122 union vxlan_addr saddr; /* source address */
823aa873 123 __be16 dst_port;
05f47d69 124 __u16 port_min; /* source port range */
125 __u16 port_max;
d342894c 126 __u8 tos; /* TOS override */
127 __u8 ttl;
359a0ea9 128 u32 flags; /* VXLAN_F_* in vxlan.h */
d342894c 129
1c51a915 130 struct work_struct sock_work;
3fc2de2f 131 struct work_struct igmp_join;
132 struct work_struct igmp_leave;
1c51a915 133
d342894c 134 unsigned long age_interval;
135 struct timer_list age_timer;
136 spinlock_t hash_lock;
137 unsigned int addrcnt;
138 unsigned int addrmax;
d342894c 139
140 struct hlist_head fdb_head[FDB_HASH_SIZE];
141};
142
143/* salt for hash table */
144static u32 vxlan_salt __read_mostly;
758c57d1 145static struct workqueue_struct *vxlan_wq;
d342894c 146
1c51a915
SH		 147static void vxlan_sock_work(struct work_struct *work);
148
e4c7ed41
CW		 149#if IS_ENABLED(CONFIG_IPV6)
150static inline
151bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
152{
153 if (a->sa.sa_family != b->sa.sa_family)
154 return false;
155 if (a->sa.sa_family == AF_INET6)
156 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
157 else
158 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
159}
160
161static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
162{
163 if (ipa->sa.sa_family == AF_INET6)
164 return ipv6_addr_any(&ipa->sin6.sin6_addr);
165 else
166 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
167}
168
169static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
170{
171 if (ipa->sa.sa_family == AF_INET6)
172 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
173 else
174 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
175}
176
177static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
178{
179 if (nla_len(nla) >= sizeof(struct in6_addr)) {
180 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
181 ip->sa.sa_family = AF_INET6;
182 return 0;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
194{
195 if (ip->sa.sa_family == AF_INET6)
196 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
197 else
198 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
199}
200
201#else /* !CONFIG_IPV6 */
202
203static inline
204bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
205{
206 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
207}
208
209static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
210{
211 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
212}
213
214static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
215{
216 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
217}
218
219static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
220{
221 if (nla_len(nla) >= sizeof(struct in6_addr)) {
222 return -EAFNOSUPPORT;
223 } else if (nla_len(nla) >= sizeof(__be32)) {
224 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
225 ip->sa.sa_family = AF_INET;
226 return 0;
227 } else {
228 return -EAFNOSUPPORT;
229 }
230}
231
232static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
233 const union vxlan_addr *ip)
234{
235 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
236}
237#endif
238
553675fb 239/* Virtual Network hash table head */
240static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
241{
242 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
243}
244
245/* Socket hash table head */
246static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 247{
248 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
249
553675fb 250 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
251}
252
3e61aa8f
SH		 253/* First remote destination for a forwarding entry.
254 * Guaranteed to be non-NULL because remotes are never deleted.
255 */
5ca5461c 256static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 257{
5ca5461c 258 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
259}
260
261static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
262{
263 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 264}
265
19ca9fc1
ML		 266/* Find VXLAN socket based on network namespace, address family and UDP port */
267static struct vxlan_sock *vxlan_find_sock(struct net *net,
268 sa_family_t family, __be16 port)
553675fb 269{
270 struct vxlan_sock *vs;
271
272 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1
ML		 273 if (inet_sk(vs->sock->sk)->inet_sport == port &&
274 inet_sk(vs->sock->sk)->sk.sk_family == family)
553675fb 275 return vs;
276 }
277 return NULL;
d342894c 278}
279
5cfccc5a 280static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 281{
282 struct vxlan_dev *vxlan;
d342894c 283
553675fb 284 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 285 if (vxlan->default_dst.remote_vni == id)
d342894c 286 return vxlan;
287 }
288
289 return NULL;
290}
291
5cfccc5a 292/* Look up VNI in a per net namespace table */
19ca9fc1
ML		 293static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
294 sa_family_t family, __be16 port)
5cfccc5a
PS		 295{
296 struct vxlan_sock *vs;
297
19ca9fc1 298 vs = vxlan_find_sock(net, family, port);
5cfccc5a
PS		 299 if (!vs)
300 return NULL;
301
302 return vxlan_vs_find_vni(vs, id);
303}
304
d342894c 305/* Fill in neighbour message in skbuff. */
306static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 307 const struct vxlan_fdb *fdb,
308 u32 portid, u32 seq, int type, unsigned int flags,
309 const struct vxlan_rdst *rdst)
d342894c 310{
311 unsigned long now = jiffies;
312 struct nda_cacheinfo ci;
313 struct nlmsghdr *nlh;
314 struct ndmsg *ndm;
e4f67add 315 bool send_ip, send_eth;
d342894c 316
317 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
318 if (nlh == NULL)
319 return -EMSGSIZE;
320
321 ndm = nlmsg_data(nlh);
322 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 323
324 send_eth = send_ip = true;
325
326 if (type == RTM_GETNEIGH) {
327 ndm->ndm_family = AF_INET;
e4c7ed41 328 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 329 send_eth = !is_zero_ether_addr(fdb->eth_addr);
330 } else
331 ndm->ndm_family = AF_BRIDGE;
d342894c 332 ndm->ndm_state = fdb->state;
333 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 334 ndm->ndm_flags = fdb->flags;
545469f7 335 ndm->ndm_type = RTN_UNICAST;
d342894c 336
e4f67add 337 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 338 goto nla_put_failure;
339
e4c7ed41 340 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 341 goto nla_put_failure;
342
823aa873 343 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 344 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
345 goto nla_put_failure;
c7995c43 346 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 347 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 348 goto nla_put_failure;
349 if (rdst->remote_ifindex &&
350 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 351 goto nla_put_failure;
352
353 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
354 ci.ndm_confirmed = 0;
355 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
356 ci.ndm_refcnt = 0;
357
358 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
359 goto nla_put_failure;
360
361 return nlmsg_end(skb, nlh);
362
363nla_put_failure:
364 nlmsg_cancel(skb, nlh);
365 return -EMSGSIZE;
366}
367
368static inline size_t vxlan_nlmsg_size(void)
369{
370 return NLMSG_ALIGN(sizeof(struct ndmsg))
371 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 372 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 373 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 374 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
375 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 376 + nla_total_size(sizeof(struct nda_cacheinfo));
377}
378
9e4b93f9
ND		 379static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
380 struct vxlan_rdst *rd, int type)
d342894c 381{
382 struct net *net = dev_net(vxlan->dev);
383 struct sk_buff *skb;
384 int err = -ENOBUFS;
385
386 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
387 if (skb == NULL)
388 goto errout;
389
9e4b93f9 390 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 391 if (err < 0) {
392 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
393 WARN_ON(err == -EMSGSIZE);
394 kfree_skb(skb);
395 goto errout;
396 }
397
398 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
399 return;
400errout:
401 if (err < 0)
402 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
403}
404
e4c7ed41 405static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 406{
407 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 408 struct vxlan_fdb f = {
409 .state = NUD_STALE,
410 };
411 struct vxlan_rdst remote = {
e4c7ed41 412 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 413 .remote_vni = VXLAN_N_VID,
414 };
3e61aa8f 415
9e4b93f9 416 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 417}
418
419static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
420{
bb3fd687
SH		 421 struct vxlan_fdb f = {
422 .state = NUD_STALE,
423 };
9e4b93f9 424 struct vxlan_rdst remote = { };
e4f67add 425
e4f67add
DS		 426 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
427
9e4b93f9 428 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 429}
430
d342894c 431/* Hash Ethernet address */
432static u32 eth_hash(const unsigned char *addr)
433{
434 u64 value = get_unaligned((u64 *)addr);
435
436 /* only want 6 bytes */
437#ifdef __BIG_ENDIAN
d342894c 438 value >>= 16;
321fb991 439#else
440 value <<= 16;
d342894c 441#endif
442 return hash_64(value, FDB_HASH_BITS);
443}
444
445/* Hash chain to use given mac address */
446static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
447 const u8 *mac)
448{
449 return &vxlan->fdb_head[eth_hash(mac)];
450}
451
452/* Look up Ethernet address in forwarding table */
014be2c8 453static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 454 const u8 *mac)
d342894c 455{
456 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
457 struct vxlan_fdb *f;
d342894c 458
b67bfe0d 459 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 460 if (ether_addr_equal(mac, f->eth_addr))
d342894c 461 return f;
462 }
463
464 return NULL;
465}
466
014be2c8
SS		 467static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
468 const u8 *mac)
469{
470 struct vxlan_fdb *f;
471
472 f = __vxlan_find_mac(vxlan, mac);
473 if (f)
474 f->used = jiffies;
475
476 return f;
477}
478
a5e7c10a
MR		 479/* caller should hold vxlan->hash_lock */
480static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 481 union vxlan_addr *ip, __be16 port,
a5e7c10a 482 __u32 vni, __u32 ifindex)
6681712d 483{
3e61aa8f 484 struct vxlan_rdst *rd;
6681712d 485
3e61aa8f 486 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 487 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 488 rd->remote_port == port &&
489 rd->remote_vni == vni &&
490 rd->remote_ifindex == ifindex)
a5e7c10a 491 return rd;
6681712d 492 }
3e61aa8f 493
a5e7c10a
MR		 494 return NULL;
495}
496
906dc186
TR		 497/* Replace destination of unicast mac */
498static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 499 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 500{
501 struct vxlan_rdst *rd;
502
503 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
504 if (rd)
505 return 0;
506
507 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
508 if (!rd)
509 return 0;
e4c7ed41 510 rd->remote_ip = *ip;
906dc186
TR		 511 rd->remote_port = port;
512 rd->remote_vni = vni;
513 rd->remote_ifindex = ifindex;
514 return 1;
515}
516
a5e7c10a
MR		 517/* Add/update destinations for multicast */
518static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 519 union vxlan_addr *ip, __be16 port, __u32 vni,
520 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 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
6681712d
DS		 528 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
529 if (rd == NULL)
530 return -ENOBUFS;
e4c7ed41 531 rd->remote_ip = *ip;
6681712d
DS		 532 rd->remote_port = port;
533 rd->remote_vni = vni;
534 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 535
536 list_add_tail_rcu(&rd->list, &f->remotes);
537
9e4b93f9 538 *rdp = rd;
6681712d
DS		 539 return 1;
540}
541
dc01e7d3
OG		 542static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb)
543{
544 struct sk_buff *p, **pp = NULL;
545 struct vxlanhdr *vh, *vh2;
9b174d88 546 unsigned int hlen, off_vx;
dc01e7d3
OG		 547 int flush = 1;
548
549 off_vx = skb_gro_offset(skb);
550 hlen = off_vx + sizeof(*vh);
551 vh = skb_gro_header_fast(skb, off_vx);
552 if (skb_gro_header_hard(skb, hlen)) {
553 vh = skb_gro_header_slow(skb, hlen, off_vx);
554 if (unlikely(!vh))
555 goto out;
556 }
dc01e7d3
OG		 557
558 flush = 0;
559
560 for (p = *head; p; p = p->next) {
561 if (!NAPI_GRO_CB(p)->same_flow)
562 continue;
563
564 vh2 = (struct vxlanhdr *)(p->data + off_vx);
9b174d88 565 if (vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 566 NAPI_GRO_CB(p)->same_flow = 0;
567 continue;
568 }
dc01e7d3
OG		 569 }
570
9b174d88
JG		 571 skb_gro_pull(skb, sizeof(struct vxlanhdr));
572 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
573 pp = eth_gro_receive(head, skb);
dc01e7d3 574
dc01e7d3
OG		 575out:
576 NAPI_GRO_CB(skb)->flush |= flush;
577
578 return pp;
579}
580
581static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
582{
cfdf1e1b
JG		 583 udp_tunnel_gro_complete(skb, nhoff);
584
9b174d88 585 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 586}
587
53cf5275 588/* Notify netdevs that UDP port started listening */
dc01e7d3 589static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 590{
591 struct net_device *dev;
dc01e7d3 592 struct sock *sk = vs->sock->sk;
53cf5275
JG		 593 struct net *net = sock_net(sk);
594 sa_family_t sa_family = sk->sk_family;
35e42379 595 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 596 int err;
597
598 if (sa_family == AF_INET) {
599 err = udp_add_offload(&vs->udp_offloads);
600 if (err)
601 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
602 }
53cf5275
JG		 603
604 rcu_read_lock();
605 for_each_netdev_rcu(net, dev) {
606 if (dev->netdev_ops->ndo_add_vxlan_port)
607 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
608 port);
609 }
610 rcu_read_unlock();
611}
612
613/* Notify netdevs that UDP port is no more listening */
dc01e7d3 614static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 615{
616 struct net_device *dev;
dc01e7d3 617 struct sock *sk = vs->sock->sk;
53cf5275
JG		 618 struct net *net = sock_net(sk);
619 sa_family_t sa_family = sk->sk_family;
35e42379 620 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 621
622 rcu_read_lock();
623 for_each_netdev_rcu(net, dev) {
624 if (dev->netdev_ops->ndo_del_vxlan_port)
625 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
626 port);
627 }
628 rcu_read_unlock();
dc01e7d3
OG		 629
630 if (sa_family == AF_INET)
631 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 632}
633
d342894c 634/* Add new entry to forwarding table -- assumes lock held */
635static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 636 const u8 *mac, union vxlan_addr *ip,
6681712d 637 __u16 state, __u16 flags,
73cf3317 638 __be16 port, __u32 vni, __u32 ifindex,
ae884082 639 __u8 ndm_flags)
d342894c 640{
9e4b93f9 641 struct vxlan_rdst *rd = NULL;
d342894c 642 struct vxlan_fdb *f;
643 int notify = 0;
644
014be2c8 645 f = __vxlan_find_mac(vxlan, mac);
d342894c 646 if (f) {
647 if (flags & NLM_F_EXCL) {
648 netdev_dbg(vxlan->dev,
649 "lost race to create %pM\n", mac);
650 return -EEXIST;
651 }
652 if (f->state != state) {
653 f->state = state;
654 f->updated = jiffies;
655 notify = 1;
656 }
ae884082
DS		 657 if (f->flags != ndm_flags) {
658 f->flags = ndm_flags;
659 f->updated = jiffies;
660 notify = 1;
661 }
906dc186
TR		 662 if ((flags & NLM_F_REPLACE)) {
663 /* Only change unicasts */
664 if (!(is_multicast_ether_addr(f->eth_addr) ||
665 is_zero_ether_addr(f->eth_addr))) {
666 int rc = vxlan_fdb_replace(f, ip, port, vni,
667 ifindex);
668
669 if (rc < 0)
670 return rc;
671 notify |= rc;
672 } else
673 return -EOPNOTSUPP;
674 }
6681712d 675 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 676 (is_multicast_ether_addr(f->eth_addr) ||
677 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 678 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
679 &rd);
6681712d
DS		 680
681 if (rc < 0)
682 return rc;
683 notify |= rc;
684 }
d342894c 685 } else {
686 if (!(flags & NLM_F_CREATE))
687 return -ENOENT;
688
689 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
690 return -ENOSPC;
691
906dc186
TR		 692 /* Disallow replace to add a multicast entry */
693 if ((flags & NLM_F_REPLACE) &&
694 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
695 return -EOPNOTSUPP;
696
e4c7ed41 697 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 698 f = kmalloc(sizeof(*f), GFP_ATOMIC);
699 if (!f)
700 return -ENOMEM;
701
702 notify = 1;
d342894c 703 f->state = state;
ae884082 704 f->flags = ndm_flags;
d342894c 705 f->updated = f->used = jiffies;
3e61aa8f 706 INIT_LIST_HEAD(&f->remotes);
d342894c 707 memcpy(f->eth_addr, mac, ETH_ALEN);
708
9e4b93f9 709 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 710
d342894c 711 ++vxlan->addrcnt;
712 hlist_add_head_rcu(&f->hlist,
713 vxlan_fdb_head(vxlan, mac));
714 }
715
9e4b93f9
ND		 716 if (notify) {
717 if (rd == NULL)
718 rd = first_remote_rtnl(f);
719 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
720 }
d342894c 721
722 return 0;
723}
724
6706c82e 725static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 726{
727 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 728 struct vxlan_rdst *rd, *nd;
6681712d 729
3e61aa8f 730 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 731 kfree(rd);
6681712d
DS		 732 kfree(f);
733}
734
d342894c 735static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
736{
737 netdev_dbg(vxlan->dev,
738 "delete %pM\n", f->eth_addr);
739
740 --vxlan->addrcnt;
9e4b93f9 741 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 742
743 hlist_del_rcu(&f->hlist);
6681712d 744 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 745}
746
f0b074be 747static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 748 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 749{
6681712d 750 struct net *net = dev_net(vxlan->dev);
e4c7ed41 751 int err;
d342894c 752
f0b074be 753 if (tb[NDA_DST]) {
e4c7ed41
CW		 754 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
755 if (err)
756 return err;
f0b074be 757 } else {
e4c7ed41
CW		 758 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
759 if (remote->sa.sa_family == AF_INET) {
760 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
761 ip->sa.sa_family = AF_INET;
762#if IS_ENABLED(CONFIG_IPV6)
763 } else {
764 ip->sin6.sin6_addr = in6addr_any;
765 ip->sa.sa_family = AF_INET6;
766#endif
767 }
f0b074be 768 }
d342894c 769
6681712d 770 if (tb[NDA_PORT]) {
73cf3317 771 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 772 return -EINVAL;
f0b074be
MR		 773 *port = nla_get_be16(tb[NDA_PORT]);
774 } else {
775 *port = vxlan->dst_port;
776 }
6681712d
DS		 777
778 if (tb[NDA_VNI]) {
779 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
780 return -EINVAL;
f0b074be
MR		 781 *vni = nla_get_u32(tb[NDA_VNI]);
782 } else {
783 *vni = vxlan->default_dst.remote_vni;
784 }
6681712d
DS		 785
786 if (tb[NDA_IFINDEX]) {
5abb0029 787 struct net_device *tdev;
6681712d
DS		 788
789 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
790 return -EINVAL;
f0b074be 791 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 792 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 793 if (!tdev)
6681712d 794 return -EADDRNOTAVAIL;
f0b074be
MR		 795 } else {
796 *ifindex = 0;
797 }
798
799 return 0;
800}
801
802/* Add static entry (via netlink) */
803static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
804 struct net_device *dev,
f6f6424b 805 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 806{
807 struct vxlan_dev *vxlan = netdev_priv(dev);
808 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 809 union vxlan_addr ip;
f0b074be
MR		 810 __be16 port;
811 u32 vni, ifindex;
812 int err;
813
814 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
815 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
816 ndm->ndm_state);
817 return -EINVAL;
818 }
819
820 if (tb[NDA_DST] == NULL)
821 return -EINVAL;
822
823 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
824 if (err)
825 return err;
6681712d 826
5933a7bb
MR		 827 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
828 return -EAFNOSUPPORT;
829
d342894c 830 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 831 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 832 port, vni, ifindex, ndm->ndm_flags);
d342894c 833 spin_unlock_bh(&vxlan->hash_lock);
834
835 return err;
836}
837
838/* Delete entry (via netlink) */
1690be63
VY		 839static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
840 struct net_device *dev,
f6f6424b 841 const unsigned char *addr, u16 vid)
d342894c 842{
843 struct vxlan_dev *vxlan = netdev_priv(dev);
844 struct vxlan_fdb *f;
bc7892ba 845 struct vxlan_rdst *rd = NULL;
e4c7ed41 846 union vxlan_addr ip;
bc7892ba
MR		 847 __be16 port;
848 u32 vni, ifindex;
849 int err;
850
851 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
852 if (err)
853 return err;
854
855 err = -ENOENT;
d342894c 856
857 spin_lock_bh(&vxlan->hash_lock);
858 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 859 if (!f)
860 goto out;
861
e4c7ed41
CW		 862 if (!vxlan_addr_any(&ip)) {
863 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 864 if (!rd)
865 goto out;
866 }
867
868 err = 0;
869
870 /* remove a destination if it's not the only one on the list,
871 * otherwise destroy the fdb entry
872 */
873 if (rd && !list_is_singular(&f->remotes)) {
874 list_del_rcu(&rd->list);
9e4b93f9 875 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 876 kfree_rcu(rd, rcu);
bc7892ba 877 goto out;
d342894c 878 }
bc7892ba
MR		 879
880 vxlan_fdb_destroy(vxlan, f);
881
882out:
d342894c 883 spin_unlock_bh(&vxlan->hash_lock);
884
885 return err;
886}
887
888/* Dump forwarding table */
889static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 890 struct net_device *dev,
891 struct net_device *filter_dev, int idx)
d342894c 892{
893 struct vxlan_dev *vxlan = netdev_priv(dev);
894 unsigned int h;
895
896 for (h = 0; h < FDB_HASH_SIZE; ++h) {
897 struct vxlan_fdb *f;
d342894c 898 int err;
899
b67bfe0d 900 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 901 struct vxlan_rdst *rd;
6681712d 902
3e61aa8f
SH		 903 if (idx < cb->args[0])
904 goto skip;
905
906 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 907 err = vxlan_fdb_info(skb, vxlan, f,
908 NETLINK_CB(cb->skb).portid,
909 cb->nlh->nlmsg_seq,
910 RTM_NEWNEIGH,
911 NLM_F_MULTI, rd);
912 if (err < 0)
3e61aa8f 913 goto out;
6681712d 914 }
3e61aa8f
SH		 915skip:
916 ++idx;
d342894c 917 }
918 }
3e61aa8f 919out:
d342894c 920 return idx;
921}
922
923/* Watch incoming packets to learn mapping between Ethernet address
924 * and Tunnel endpoint.
26a41ae6 925 * Return true if packet is bogus and should be droppped.
d342894c 926 */
26a41ae6 927static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 928 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 929{
930 struct vxlan_dev *vxlan = netdev_priv(dev);
931 struct vxlan_fdb *f;
d342894c 932
933 f = vxlan_find_mac(vxlan, src_mac);
934 if (likely(f)) {
5ca5461c 935 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 936
e4c7ed41 937 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 938 return false;
939
940 /* Don't migrate static entries, drop packets */
eb064c3b 941 if (f->state & NUD_NOARP)
26a41ae6 942 return true;
d342894c 943
944 if (net_ratelimit())
945 netdev_info(dev,
e4c7ed41 946 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 947 src_mac, &rdst->remote_ip, &src_ip);
d342894c 948
e4c7ed41 949 rdst->remote_ip = *src_ip;
d342894c 950 f->updated = jiffies;
9e4b93f9 951 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 952 } else {
953 /* learned new entry */
954 spin_lock(&vxlan->hash_lock);
3bf74b1a 955
956 /* close off race between vxlan_flush and incoming packets */
957 if (netif_running(dev))
958 vxlan_fdb_create(vxlan, src_mac, src_ip,
959 NUD_REACHABLE,
960 NLM_F_EXCL|NLM_F_CREATE,
961 vxlan->dst_port,
962 vxlan->default_dst.remote_vni,
963 0, NTF_SELF);
d342894c 964 spin_unlock(&vxlan->hash_lock);
965 }
26a41ae6 966
967 return false;
d342894c 968}
969
d342894c 970/* See if multicast group is already in use by other ID */
95ab0991 971static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 972{
553675fb 973 struct vxlan_dev *vxlan;
d342894c 974
95ab0991
G		 975 /* The vxlan_sock is only used by dev, leaving group has
976 * no effect on other vxlan devices.
977 */
978 if (atomic_read(&dev->vn_sock->refcnt) == 1)
979 return false;
980
553675fb 981 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 982 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 983 continue;
d342894c 984
95ab0991
G		 985 if (vxlan->vn_sock != dev->vn_sock)
986 continue;
987
988 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
989 &dev->default_dst.remote_ip))
990 continue;
991
992 if (vxlan->default_dst.remote_ifindex !=
993 dev->default_dst.remote_ifindex)
994 continue;
995
996 return true;
553675fb 997 }
d342894c 998
999 return false;
1000}
1001
7c47cedf 1002static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1003{
7c47cedf
SH		 1004 atomic_inc(&vs->refcnt);
1005}
d342894c 1006
012a5729 1007void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1008{
53cf5275
JG		 1009 struct sock *sk = vs->sock->sk;
1010 struct net *net = sock_net(sk);
1011 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1012
7c47cedf
SH		 1013 if (!atomic_dec_and_test(&vs->refcnt))
1014 return;
d342894c 1015
1c51a915 1016 spin_lock(&vn->sock_lock);
7c47cedf 1017 hlist_del_rcu(&vs->hlist);
dc01e7d3 1018 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1019 spin_unlock(&vn->sock_lock);
1020
7c47cedf 1021 queue_work(vxlan_wq, &vs->del_work);
d342894c 1022}
012a5729 1023EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1024
3fc2de2f 1025/* Callback to update multicast group membership when first VNI on
1026 * multicast asddress is brought up
1027 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1028 */
3fc2de2f 1029static void vxlan_igmp_join(struct work_struct *work)
d342894c 1030{
3fc2de2f 1031 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1032 struct vxlan_sock *vs = vxlan->vn_sock;
1033 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1034 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1035 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1036
d342894c 1037 lock_sock(sk);
e4c7ed41
CW		 1038 if (ip->sa.sa_family == AF_INET) {
1039 struct ip_mreqn mreq = {
1040 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1041 .imr_ifindex = ifindex,
1042 };
1043
1044 ip_mc_join_group(sk, &mreq);
1045#if IS_ENABLED(CONFIG_IPV6)
1046 } else {
1047 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1048 &ip->sin6.sin6_addr);
1049#endif
1050 }
3fc2de2f 1051 release_sock(sk);
1052
012a5729 1053 vxlan_sock_release(vs);
3fc2de2f 1054 dev_put(vxlan->dev);
1055}
1056
1057/* Inverse of vxlan_igmp_join when last VNI is brought down */
1058static void vxlan_igmp_leave(struct work_struct *work)
1059{
1060 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1061 struct vxlan_sock *vs = vxlan->vn_sock;
1062 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1063 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1064 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1065
1066 lock_sock(sk);
e4c7ed41
CW		 1067 if (ip->sa.sa_family == AF_INET) {
1068 struct ip_mreqn mreq = {
1069 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1070 .imr_ifindex = ifindex,
1071 };
1072
1073 ip_mc_leave_group(sk, &mreq);
1074#if IS_ENABLED(CONFIG_IPV6)
1075 } else {
1076 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1077 &ip->sin6.sin6_addr);
1078#endif
1079 }
1080
d342894c 1081 release_sock(sk);
d342894c 1082
012a5729 1083 vxlan_sock_release(vs);
7c47cedf 1084 dev_put(vxlan->dev);
d342894c 1085}
1086
1087/* Callback from net/ipv4/udp.c to receive packets */
1088static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1089{
5cfccc5a 1090 struct vxlan_sock *vs;
d342894c 1091 struct vxlanhdr *vxh;
3bf39475 1092 u32 flags, vni;
d342894c 1093
d342894c 1094 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1095 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1096 goto error;
1097
7ce04758 1098 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1099 flags = ntohl(vxh->vx_flags);
1100 vni = ntohl(vxh->vx_vni);
1101
1102 if (flags & VXLAN_HF_VNI) {
1103 flags &= ~VXLAN_HF_VNI;
1104 } else {
1105 /* VNI flag always required to be set */
1106 goto bad_flags;
d342894c 1107 }
1108
5cfccc5a
PS		 1109 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1110 goto drop;
1111
559835ea 1112 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1113 if (!vs)
d342894c 1114 goto drop;
d342894c 1115
3bf39475
TH		 1116 if (flags || (vni & 0xff)) {
1117 /* If there are any unprocessed flags remaining treat
1118 * this as a malformed packet. This behavior diverges from
1119 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1120 * in reserved fields are to be ignored. The approach here
1121 * maintains compatbility with previous stack code, and also
1122 * is more robust and provides a little more security in
1123 * adding extensions to VXLAN.
1124 */
1125
1126 goto bad_flags;
1127 }
1128
5cfccc5a
PS		 1129 vs->rcv(vs, skb, vxh->vx_vni);
1130 return 0;
1131
1132drop:
1133 /* Consume bad packet */
1134 kfree_skb(skb);
1135 return 0;
1136
3bf39475
TH		 1137bad_flags:
1138 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1139 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1140
5cfccc5a
PS		 1141error:
1142 /* Return non vxlan pkt */
1143 return 1;
1144}
1145
1146static void vxlan_rcv(struct vxlan_sock *vs,
1147 struct sk_buff *skb, __be32 vx_vni)
1148{
e4c7ed41
CW		 1149 struct iphdr *oip = NULL;
1150 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1151 struct vxlan_dev *vxlan;
8f84985f 1152 struct pcpu_sw_netstats *stats;
e4c7ed41 1153 union vxlan_addr saddr;
5cfccc5a 1154 __u32 vni;
e4c7ed41
CW		 1155 int err = 0;
1156 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1157
1158 vni = ntohl(vx_vni) >> 8;
1159 /* Is this VNI defined? */
1160 vxlan = vxlan_vs_find_vni(vs, vni);
1161 if (!vxlan)
d342894c 1162 goto drop;
d342894c 1163
e4c7ed41 1164 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1165 skb_reset_mac_header(skb);
f01ec1c0 1166 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1167 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1168 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1169
1170 /* Ignore packet loops (and multicast echo) */
7367d0b5 1171 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1172 goto drop;
1173
7ce04758 1174 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1175 if (remote_ip->sa.sa_family == AF_INET) {
1176 oip = ip_hdr(skb);
1177 saddr.sin.sin_addr.s_addr = oip->saddr;
1178 saddr.sa.sa_family = AF_INET;
1179#if IS_ENABLED(CONFIG_IPV6)
1180 } else {
1181 oip6 = ipv6_hdr(skb);
1182 saddr.sin6.sin6_addr = oip6->saddr;
1183 saddr.sa.sa_family = AF_INET6;
1184#endif
1185 }
1186
26a41ae6 1187 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1188 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1189 goto drop;
d342894c 1190
d342894c 1191 skb_reset_network_header(skb);
0afb1666 1192
e4c7ed41
CW		 1193 if (oip6)
1194 err = IP6_ECN_decapsulate(oip6, skb);
1195 if (oip)
1196 err = IP_ECN_decapsulate(oip, skb);
1197
d342894c 1198 if (unlikely(err)) {
e4c7ed41
CW		 1199 if (log_ecn_error) {
1200 if (oip6)
1201 net_info_ratelimited("non-ECT from %pI6\n",
1202 &oip6->saddr);
1203 if (oip)
1204 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1205 &oip->saddr, oip->tos);
1206 }
d342894c 1207 if (err > 1) {
1208 ++vxlan->dev->stats.rx_frame_errors;
1209 ++vxlan->dev->stats.rx_errors;
1210 goto drop;
1211 }
1212 }
1213
e8171045 1214 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1215 u64_stats_update_begin(&stats->syncp);
1216 stats->rx_packets++;
1217 stats->rx_bytes += skb->len;
1218 u64_stats_update_end(&stats->syncp);
1219
1220 netif_rx(skb);
1221
5cfccc5a 1222 return;
d342894c 1223drop:
1224 /* Consume bad packet */
1225 kfree_skb(skb);
d342894c 1226}
1227
e4f67add
DS		 1228static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1229{
1230 struct vxlan_dev *vxlan = netdev_priv(dev);
1231 struct arphdr *parp;
1232 u8 *arpptr, *sha;
1233 __be32 sip, tip;
1234 struct neighbour *n;
1235
1236 if (dev->flags & IFF_NOARP)
1237 goto out;
1238
1239 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1240 dev->stats.tx_dropped++;
1241 goto out;
1242 }
1243 parp = arp_hdr(skb);
1244
1245 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1246 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1247 parp->ar_pro != htons(ETH_P_IP) ||
1248 parp->ar_op != htons(ARPOP_REQUEST) ||
1249 parp->ar_hln != dev->addr_len ||
1250 parp->ar_pln != 4)
1251 goto out;
1252 arpptr = (u8 *)parp + sizeof(struct arphdr);
1253 sha = arpptr;
1254 arpptr += dev->addr_len; /* sha */
1255 memcpy(&sip, arpptr, sizeof(sip));
1256 arpptr += sizeof(sip);
1257 arpptr += dev->addr_len; /* tha */
1258 memcpy(&tip, arpptr, sizeof(tip));
1259
1260 if (ipv4_is_loopback(tip) ||
1261 ipv4_is_multicast(tip))
1262 goto out;
1263
1264 n = neigh_lookup(&arp_tbl, &tip, dev);
1265
1266 if (n) {
e4f67add
DS		 1267 struct vxlan_fdb *f;
1268 struct sk_buff *reply;
1269
1270 if (!(n->nud_state & NUD_CONNECTED)) {
1271 neigh_release(n);
1272 goto out;
1273 }
1274
1275 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1276 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1277 /* bridge-local neighbor */
1278 neigh_release(n);
1279 goto out;
1280 }
1281
1282 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1283 n->ha, sha);
1284
1285 neigh_release(n);
1286
7346135d
DS		 1287 if (reply == NULL)
1288 goto out;
1289
e4f67add
DS		 1290 skb_reset_mac_header(reply);
1291 __skb_pull(reply, skb_network_offset(reply));
1292 reply->ip_summed = CHECKSUM_UNNECESSARY;
1293 reply->pkt_type = PACKET_HOST;
1294
1295 if (netif_rx_ni(reply) == NET_RX_DROP)
1296 dev->stats.rx_dropped++;
e4c7ed41
CW		 1297 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1298 union vxlan_addr ipa = {
1299 .sin.sin_addr.s_addr = tip,
a45e92a5 1300 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1301 };
1302
1303 vxlan_ip_miss(dev, &ipa);
1304 }
e4f67add
DS		 1305out:
1306 consume_skb(skb);
1307 return NETDEV_TX_OK;
1308}
1309
f564f45c 1310#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1311static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1312 struct neighbour *n, bool isrouter)
1313{
1314 struct net_device *dev = request->dev;
1315 struct sk_buff *reply;
1316 struct nd_msg *ns, *na;
1317 struct ipv6hdr *pip6;
1318 u8 *daddr;
1319 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1320 int ns_olen;
1321 int i, len;
1322
1323 if (dev == NULL)
1324 return NULL;
1325
1326 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1327 sizeof(*na) + na_olen + dev->needed_tailroom;
1328 reply = alloc_skb(len, GFP_ATOMIC);
1329 if (reply == NULL)
1330 return NULL;
1331
1332 reply->protocol = htons(ETH_P_IPV6);
1333 reply->dev = dev;
1334 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1335 skb_push(reply, sizeof(struct ethhdr));
1336 skb_set_mac_header(reply, 0);
1337
1338 ns = (struct nd_msg *)skb_transport_header(request);
1339
1340 daddr = eth_hdr(request)->h_source;
1341 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1342 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1343 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1344 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1345 break;
1346 }
1347 }
1348
1349 /* Ethernet header */
1350 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1351 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1352 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1353 reply->protocol = htons(ETH_P_IPV6);
1354
1355 skb_pull(reply, sizeof(struct ethhdr));
1356 skb_set_network_header(reply, 0);
1357 skb_put(reply, sizeof(struct ipv6hdr));
1358
1359 /* IPv6 header */
1360
1361 pip6 = ipv6_hdr(reply);
1362 memset(pip6, 0, sizeof(struct ipv6hdr));
1363 pip6->version = 6;
1364 pip6->priority = ipv6_hdr(request)->priority;
1365 pip6->nexthdr = IPPROTO_ICMPV6;
1366 pip6->hop_limit = 255;
1367 pip6->daddr = ipv6_hdr(request)->saddr;
1368 pip6->saddr = *(struct in6_addr *)n->primary_key;
1369
1370 skb_pull(reply, sizeof(struct ipv6hdr));
1371 skb_set_transport_header(reply, 0);
1372
1373 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1374
1375 /* Neighbor Advertisement */
1376 memset(na, 0, sizeof(*na)+na_olen);
1377 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1378 na->icmph.icmp6_router = isrouter;
1379 na->icmph.icmp6_override = 1;
1380 na->icmph.icmp6_solicited = 1;
1381 na->target = ns->target;
1382 ether_addr_copy(&na->opt[2], n->ha);
1383 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1384 na->opt[1] = na_olen >> 3;
1385
1386 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1387 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1388 csum_partial(na, sizeof(*na)+na_olen, 0));
1389
1390 pip6->payload_len = htons(sizeof(*na)+na_olen);
1391
1392 skb_push(reply, sizeof(struct ipv6hdr));
1393
1394 reply->ip_summed = CHECKSUM_UNNECESSARY;
1395
1396 return reply;
1397}
1398
f564f45c
CW		 1399static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1400{
1401 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1402 struct nd_msg *msg;
f564f45c
CW		 1403 const struct ipv6hdr *iphdr;
1404 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1405 struct neighbour *n;
1406 struct inet6_dev *in6_dev;
f564f45c
CW		 1407
1408 in6_dev = __in6_dev_get(dev);
1409 if (!in6_dev)
1410 goto out;
1411
f564f45c
CW		 1412 iphdr = ipv6_hdr(skb);
1413 saddr = &iphdr->saddr;
1414 daddr = &iphdr->daddr;
1415
f564f45c
CW		 1416 msg = (struct nd_msg *)skb_transport_header(skb);
1417 if (msg->icmph.icmp6_code != 0 ||
1418 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1419 goto out;
1420
4b29dba9
DS		 1421 if (ipv6_addr_loopback(daddr) ||
1422 ipv6_addr_is_multicast(&msg->target))
1423 goto out;
1424
1425 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1426
1427 if (n) {
1428 struct vxlan_fdb *f;
4b29dba9 1429 struct sk_buff *reply;
f564f45c
CW		 1430
1431 if (!(n->nud_state & NUD_CONNECTED)) {
1432 neigh_release(n);
1433 goto out;
1434 }
1435
1436 f = vxlan_find_mac(vxlan, n->ha);
1437 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1438 /* bridge-local neighbor */
1439 neigh_release(n);
1440 goto out;
1441 }
1442
4b29dba9
DS		 1443 reply = vxlan_na_create(skb, n,
1444 !!(f ? f->flags & NTF_ROUTER : 0));
1445
f564f45c 1446 neigh_release(n);
4b29dba9
DS		 1447
1448 if (reply == NULL)
1449 goto out;
1450
1451 if (netif_rx_ni(reply) == NET_RX_DROP)
1452 dev->stats.rx_dropped++;
1453
f564f45c 1454 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1455 union vxlan_addr ipa = {
1456 .sin6.sin6_addr = msg->target,
a45e92a5 1457 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1458 };
1459
f564f45c
CW		 1460 vxlan_ip_miss(dev, &ipa);
1461 }
1462
1463out:
1464 consume_skb(skb);
1465 return NETDEV_TX_OK;
1466}
1467#endif
1468
e4f67add
DS		 1469static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1470{
1471 struct vxlan_dev *vxlan = netdev_priv(dev);
1472 struct neighbour *n;
e4f67add
DS		 1473
1474 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1475 return false;
1476
1477 n = NULL;
1478 switch (ntohs(eth_hdr(skb)->h_proto)) {
1479 case ETH_P_IP:
e15a00aa
CW		 1480 {
1481 struct iphdr *pip;
1482
e4f67add
DS		 1483 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1484 return false;
1485 pip = ip_hdr(skb);
1486 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1487 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1488 union vxlan_addr ipa = {
1489 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1490 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1491 };
1492
1493 vxlan_ip_miss(dev, &ipa);
1494 return false;
1495 }
1496
e4f67add 1497 break;
e15a00aa
CW		 1498 }
1499#if IS_ENABLED(CONFIG_IPV6)
1500 case ETH_P_IPV6:
1501 {
1502 struct ipv6hdr *pip6;
1503
1504 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1505 return false;
1506 pip6 = ipv6_hdr(skb);
1507 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1508 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1509 union vxlan_addr ipa = {
1510 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1511 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1512 };
1513
1514 vxlan_ip_miss(dev, &ipa);
1515 return false;
1516 }
1517
1518 break;
1519 }
1520#endif
e4f67add
DS		 1521 default:
1522 return false;
1523 }
1524
1525 if (n) {
1526 bool diff;
1527
7367d0b5 1528 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1529 if (diff) {
1530 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1531 dev->addr_len);
1532 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1533 }
1534 neigh_release(n);
1535 return diff;
e4c7ed41
CW		 1536 }
1537
e4f67add
DS		 1538 return false;
1539}
1540
e4c7ed41 1541#if IS_ENABLED(CONFIG_IPV6)
11796187 1542static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1543 struct dst_entry *dst, struct sk_buff *skb,
1544 struct net_device *dev, struct in6_addr *saddr,
1545 struct in6_addr *daddr, __u8 prio, __u8 ttl,
f01ec1c0
ND		 1546 __be16 src_port, __be16 dst_port, __be32 vni,
1547 bool xnet)
e4c7ed41 1548{
e4c7ed41 1549 struct vxlanhdr *vxh;
e4c7ed41
CW		 1550 int min_headroom;
1551 int err;
acbf74a7 1552 bool udp_sum = !udp_get_no_check6_tx(vs->sock->sk);
e4c7ed41 1553
acbf74a7 1554 skb = udp_tunnel_handle_offloads(skb, udp_sum);
74f47278
PS		 1555 if (IS_ERR(skb)) {
1556 err = -EINVAL;
1557 goto err;
1558 }
e4c7ed41 1559
f01ec1c0 1560 skb_scrub_packet(skb, xnet);
963a88b3 1561
e4c7ed41
CW		 1562 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1563 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1564 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1565
1566 /* Need space for new headers (invalidates iph ptr) */
1567 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1568 if (unlikely(err)) {
1569 kfree_skb(skb);
1570 goto err;
1571 }
e4c7ed41 1572
5968250c 1573 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1574 if (WARN_ON(!skb)) {
1575 err = -ENOMEM;
1576 goto err;
1577 }
e4c7ed41
CW		 1578
1579 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1580 vxh->vx_flags = htonl(VXLAN_HF_VNI);
e4c7ed41
CW		 1581 vxh->vx_vni = vni;
1582
996c9fd1
TH		 1583 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1584
acbf74a7
AZ		 1585 udp_tunnel6_xmit_skb(vs->sock, dst, skb, dev, saddr, daddr, prio,
1586 ttl, src_port, dst_port);
e4c7ed41 1587 return 0;
74f47278
PS		 1588err:
1589 dst_release(dst);
1590 return err;
e4c7ed41
CW		 1591}
1592#endif
1593
11796187 1594int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1595 struct rtable *rt, struct sk_buff *skb,
1596 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
f01ec1c0 1597 __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
49560532
PS		 1598{
1599 struct vxlanhdr *vxh;
649c5b8b 1600 int min_headroom;
49560532 1601 int err;
acbf74a7 1602 bool udp_sum = !vs->sock->sk->sk_no_check_tx;
49560532 1603
acbf74a7 1604 skb = udp_tunnel_handle_offloads(skb, udp_sum);
359a0ea9 1605 if (IS_ERR(skb))
74f47278 1606 return PTR_ERR(skb);
49560532 1607
649c5b8b 1608 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1609 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1610 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1611
1612 /* Need space for new headers (invalidates iph ptr) */
1613 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1614 if (unlikely(err)) {
1615 kfree_skb(skb);
649c5b8b 1616 return err;
74f47278 1617 }
649c5b8b 1618
5968250c
JP		 1619 skb = vlan_hwaccel_push_inside(skb);
1620 if (WARN_ON(!skb))
1621 return -ENOMEM;
1eaa8178 1622
49560532 1623 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1624 vxh->vx_flags = htonl(VXLAN_HF_VNI);
49560532
PS		 1625 vxh->vx_vni = vni;
1626
996c9fd1
TH		 1627 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1628
acbf74a7
AZ		 1629 return udp_tunnel_xmit_skb(vs->sock, rt, skb, src, dst, tos,
1630 ttl, df, src_port, dst_port, xnet);
49560532
PS		 1631}
1632EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1633
9dcc71e1
SS		 1634/* Bypass encapsulation if the destination is local */
1635static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1636 struct vxlan_dev *dst_vxlan)
1637{
8f84985f 1638 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1639 union vxlan_addr loopback;
1640 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1641 struct net_device *dev = skb->dev;
1642 int len = skb->len;
9dcc71e1 1643
8f84985f
LR		 1644 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1645 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1646 skb->pkt_type = PACKET_HOST;
1647 skb->encapsulation = 0;
1648 skb->dev = dst_vxlan->dev;
1649 __skb_pull(skb, skb_network_offset(skb));
1650
e4c7ed41
CW		 1651 if (remote_ip->sa.sa_family == AF_INET) {
1652 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1653 loopback.sa.sa_family = AF_INET;
1654#if IS_ENABLED(CONFIG_IPV6)
1655 } else {
1656 loopback.sin6.sin6_addr = in6addr_loopback;
1657 loopback.sa.sa_family = AF_INET6;
1658#endif
1659 }
1660
9dcc71e1 1661 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1662 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1663
1664 u64_stats_update_begin(&tx_stats->syncp);
1665 tx_stats->tx_packets++;
ce6502a8 1666 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1667 u64_stats_update_end(&tx_stats->syncp);
1668
1669 if (netif_rx(skb) == NET_RX_SUCCESS) {
1670 u64_stats_update_begin(&rx_stats->syncp);
1671 rx_stats->rx_packets++;
ce6502a8 1672 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1673 u64_stats_update_end(&rx_stats->syncp);
1674 } else {
ce6502a8 1675 dev->stats.rx_dropped++;
9dcc71e1
SS		 1676 }
1677}
1678
4ad16930
SH		 1679static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1680 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1681{
1682 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1683 struct rtable *rt = NULL;
d342894c 1684 const struct iphdr *old_iph;
d342894c 1685 struct flowi4 fl4;
e4c7ed41
CW		 1686 union vxlan_addr *dst;
1687 __be16 src_port = 0, dst_port;
234f5b73 1688 u32 vni;
d342894c 1689 __be16 df = 0;
1690 __u8 tos, ttl;
0e6fbc5b 1691 int err;
d342894c 1692
823aa873 1693 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1694 vni = rdst->remote_vni;
e4c7ed41 1695 dst = &rdst->remote_ip;
e4f67add 1696
e4c7ed41 1697 if (vxlan_addr_any(dst)) {
e4f67add 1698 if (did_rsc) {
e4f67add 1699 /* short-circuited back to local bridge */
9dcc71e1 1700 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1701 return;
e4f67add 1702 }
ef59febe 1703 goto drop;
e4f67add 1704 }
ef59febe 1705
d342894c 1706 old_iph = ip_hdr(skb);
1707
d342894c 1708 ttl = vxlan->ttl;
e4c7ed41 1709 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1710 ttl = 1;
1711
1712 tos = vxlan->tos;
1713 if (tos == 1)
206aaafc 1714 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1715
535fb8d0
TH		 1716 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1717 vxlan->port_max, true);
d342894c 1718
e4c7ed41
CW		 1719 if (dst->sa.sa_family == AF_INET) {
1720 memset(&fl4, 0, sizeof(fl4));
1721 fl4.flowi4_oif = rdst->remote_ifindex;
1722 fl4.flowi4_tos = RT_TOS(tos);
1723 fl4.daddr = dst->sin.sin_addr.s_addr;
1724 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1725
f01ec1c0 1726 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1727 if (IS_ERR(rt)) {
1728 netdev_dbg(dev, "no route to %pI4\n",
1729 &dst->sin.sin_addr.s_addr);
1730 dev->stats.tx_carrier_errors++;
1731 goto tx_error;
1732 }
d342894c 1733
e4c7ed41
CW		 1734 if (rt->dst.dev == dev) {
1735 netdev_dbg(dev, "circular route to %pI4\n",
1736 &dst->sin.sin_addr.s_addr);
1737 dev->stats.collisions++;
fffc15a5 1738 goto rt_tx_error;
e4c7ed41
CW		 1739 }
1740
1741 /* Bypass encapsulation if the destination is local */
1742 if (rt->rt_flags & RTCF_LOCAL &&
1743 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1744 struct vxlan_dev *dst_vxlan;
1745
1746 ip_rt_put(rt);
19ca9fc1
ML		 1747 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1748 dst->sa.sa_family, dst_port);
e4c7ed41
CW		 1749 if (!dst_vxlan)
1750 goto tx_error;
1751 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1752 return;
1753 }
1754
1755 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1756 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1757
3c4d1dae
AZ		 1758 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1759 fl4.saddr, dst->sin.sin_addr.s_addr,
1760 tos, ttl, df, src_port, dst_port,
1761 htonl(vni << 8),
1762 !net_eq(vxlan->net, dev_net(vxlan->dev)));
74f47278
PS		 1763 if (err < 0) {
1764 /* skb is already freed. */
1765 skb = NULL;
e4c7ed41 1766 goto rt_tx_error;
74f47278
PS		 1767 }
1768
e4c7ed41
CW		 1769 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1770#if IS_ENABLED(CONFIG_IPV6)
1771 } else {
1772 struct sock *sk = vxlan->vn_sock->sock->sk;
1773 struct dst_entry *ndst;
1774 struct flowi6 fl6;
1775 u32 flags;
1776
1777 memset(&fl6, 0, sizeof(fl6));
1778 fl6.flowi6_oif = rdst->remote_ifindex;
1779 fl6.daddr = dst->sin6.sin6_addr;
1780 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1781 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1782
1783 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1784 netdev_dbg(dev, "no route to %pI6\n",
1785 &dst->sin6.sin6_addr);
1786 dev->stats.tx_carrier_errors++;
9dcc71e1 1787 goto tx_error;
e4c7ed41 1788 }
d342894c 1789
e4c7ed41
CW		 1790 if (ndst->dev == dev) {
1791 netdev_dbg(dev, "circular route to %pI6\n",
1792 &dst->sin6.sin6_addr);
1793 dst_release(ndst);
1794 dev->stats.collisions++;
1795 goto tx_error;
1796 }
0e6fbc5b 1797
e4c7ed41
CW		 1798 /* Bypass encapsulation if the destination is local */
1799 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1800 if (flags & RTF_LOCAL &&
1801 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1802 struct vxlan_dev *dst_vxlan;
1803
1804 dst_release(ndst);
19ca9fc1
ML		 1805 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1806 dst->sa.sa_family, dst_port);
e4c7ed41
CW		 1807 if (!dst_vxlan)
1808 goto tx_error;
1809 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1810 return;
1811 }
49560532 1812
e4c7ed41
CW		 1813 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1814
11796187 1815 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41 1816 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
f01ec1c0
ND		 1817 src_port, dst_port, htonl(vni << 8),
1818 !net_eq(vxlan->net, dev_net(vxlan->dev)));
e4c7ed41
CW		 1819#endif
1820 }
0e6fbc5b 1821
4ad16930 1822 return;
d342894c 1823
1824drop:
1825 dev->stats.tx_dropped++;
1826 goto tx_free;
1827
49560532
PS		 1828rt_tx_error:
1829 ip_rt_put(rt);
d342894c 1830tx_error:
1831 dev->stats.tx_errors++;
1832tx_free:
1833 dev_kfree_skb(skb);
d342894c 1834}
1835
6681712d
DS		 1836/* Transmit local packets over Vxlan
1837 *
1838 * Outer IP header inherits ECN and DF from inner header.
1839 * Outer UDP destination is the VXLAN assigned port.
1840 * source port is based on hash of flow
1841 */
1842static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1843{
1844 struct vxlan_dev *vxlan = netdev_priv(dev);
1845 struct ethhdr *eth;
1846 bool did_rsc = false;
8f646c92 1847 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 1848 struct vxlan_fdb *f;
6681712d
DS		 1849
1850 skb_reset_mac_header(skb);
1851 eth = eth_hdr(skb);
1852
f564f45c
CW		 1853 if ((vxlan->flags & VXLAN_F_PROXY)) {
1854 if (ntohs(eth->h_proto) == ETH_P_ARP)
1855 return arp_reduce(dev, skb);
1856#if IS_ENABLED(CONFIG_IPV6)
1857 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 1858 pskb_may_pull(skb, sizeof(struct ipv6hdr)
1859 + sizeof(struct nd_msg)) &&
f564f45c
CW		 1860 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1861 struct nd_msg *msg;
1862
1863 msg = (struct nd_msg *)skb_transport_header(skb);
1864 if (msg->icmph.icmp6_code == 0 &&
1865 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1866 return neigh_reduce(dev, skb);
1867 }
7a9f526f 1868 eth = eth_hdr(skb);
f564f45c
CW		 1869#endif
1870 }
6681712d
DS		 1871
1872 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1873 did_rsc = false;
1874
1875 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1876 (ntohs(eth->h_proto) == ETH_P_IP ||
1877 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1878 did_rsc = route_shortcircuit(dev, skb);
1879 if (did_rsc)
1880 f = vxlan_find_mac(vxlan, eth->h_dest);
1881 }
1882
6681712d 1883 if (f == NULL) {
afbd8bae
MR		 1884 f = vxlan_find_mac(vxlan, all_zeros_mac);
1885 if (f == NULL) {
1886 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1887 !is_multicast_ether_addr(eth->h_dest))
1888 vxlan_fdb_miss(vxlan, eth->h_dest);
1889
1890 dev->stats.tx_dropped++;
8f646c92 1891 kfree_skb(skb);
afbd8bae
MR		 1892 return NETDEV_TX_OK;
1893 }
1894 }
6681712d 1895
afbd8bae
MR		 1896 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1897 struct sk_buff *skb1;
6681712d 1898
8f646c92
ED		 1899 if (!fdst) {
1900 fdst = rdst;
1901 continue;
1902 }
afbd8bae
MR		 1903 skb1 = skb_clone(skb, GFP_ATOMIC);
1904 if (skb1)
1905 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1906 }
1907
8f646c92
ED		 1908 if (fdst)
1909 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1910 else
1911 kfree_skb(skb);
4ad16930 1912 return NETDEV_TX_OK;
6681712d
DS		 1913}
1914
d342894c 1915/* Walk the forwarding table and purge stale entries */
1916static void vxlan_cleanup(unsigned long arg)
1917{
1918 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1919 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1920 unsigned int h;
1921
1922 if (!netif_running(vxlan->dev))
1923 return;
1924
1925 spin_lock_bh(&vxlan->hash_lock);
1926 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1927 struct hlist_node *p, *n;
1928 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1929 struct vxlan_fdb *f
1930 = container_of(p, struct vxlan_fdb, hlist);
1931 unsigned long timeout;
1932
3c172868 1933 if (f->state & NUD_PERMANENT)
d342894c 1934 continue;
1935
1936 timeout = f->used + vxlan->age_interval * HZ;
1937 if (time_before_eq(timeout, jiffies)) {
1938 netdev_dbg(vxlan->dev,
1939 "garbage collect %pM\n",
1940 f->eth_addr);
1941 f->state = NUD_STALE;
1942 vxlan_fdb_destroy(vxlan, f);
1943 } else if (time_before(timeout, next_timer))
1944 next_timer = timeout;
1945 }
1946 }
1947 spin_unlock_bh(&vxlan->hash_lock);
1948
1949 mod_timer(&vxlan->age_timer, next_timer);
1950}
1951
9c2e24e1
PS		 1952static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1953{
1954 __u32 vni = vxlan->default_dst.remote_vni;
1955
1956 vxlan->vn_sock = vs;
1957 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1958}
1959
d342894c 1960/* Setup stats when device is created */
1961static int vxlan_init(struct net_device *dev)
1962{
1c51a915 1963 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 1964 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 1965 struct vxlan_sock *vs;
19ca9fc1 1966 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
1c51a915 1967
1c213bd2 1968 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 1969 if (!dev->tstats)
d342894c 1970 return -ENOMEM;
1971
1c51a915 1972 spin_lock(&vn->sock_lock);
19ca9fc1
ML		 1973 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
1974 vxlan->dst_port);
00c83b01 1975 if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
1c51a915 1976 /* If we have a socket with same port already, reuse it */
9c2e24e1 1977 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1978 } else {
1979 /* otherwise make new socket outside of RTNL */
1980 dev_hold(dev);
1981 queue_work(vxlan_wq, &vxlan->sock_work);
1982 }
1983 spin_unlock(&vn->sock_lock);
1984
d342894c 1985 return 0;
1986}
1987
ba609e9b 1988static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1989{
1990 struct vxlan_fdb *f;
1991
1992 spin_lock_bh(&vxlan->hash_lock);
1993 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1994 if (f)
1995 vxlan_fdb_destroy(vxlan, f);
1996 spin_unlock_bh(&vxlan->hash_lock);
1997}
1998
ebf4063e
SH		 1999static void vxlan_uninit(struct net_device *dev)
2000{
2001 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2002 struct vxlan_sock *vs = vxlan->vn_sock;
2003
ba609e9b 2004 vxlan_fdb_delete_default(vxlan);
afbd8bae 2005
ebf4063e 2006 if (vs)
012a5729 2007 vxlan_sock_release(vs);
ebf4063e
SH		 2008 free_percpu(dev->tstats);
2009}
2010
d342894c 2011/* Start ageing timer and join group when device is brought up */
2012static int vxlan_open(struct net_device *dev)
2013{
2014 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2015 struct vxlan_sock *vs = vxlan->vn_sock;
2016
2017 /* socket hasn't been created */
2018 if (!vs)
2019 return -ENOTCONN;
d342894c 2020
79d4a94f 2021 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2022 vxlan_sock_hold(vs);
7c47cedf 2023 dev_hold(dev);
3fc2de2f 2024 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2025 }
2026
2027 if (vxlan->age_interval)
2028 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2029
2030 return 0;
2031}
2032
2033/* Purge the forwarding table */
2034static void vxlan_flush(struct vxlan_dev *vxlan)
2035{
31fec5aa 2036 unsigned int h;
d342894c 2037
2038 spin_lock_bh(&vxlan->hash_lock);
2039 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2040 struct hlist_node *p, *n;
2041 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2042 struct vxlan_fdb *f
2043 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2044 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2045 if (!is_zero_ether_addr(f->eth_addr))
2046 vxlan_fdb_destroy(vxlan, f);
d342894c 2047 }
2048 }
2049 spin_unlock_bh(&vxlan->hash_lock);
2050}
2051
2052/* Cleanup timer and forwarding table on shutdown */
2053static int vxlan_stop(struct net_device *dev)
2054{
2055 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2056 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2057 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2058
e4c7ed41 2059 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2060 !vxlan_group_used(vn, vxlan)) {
1c51a915 2061 vxlan_sock_hold(vs);
7c47cedf 2062 dev_hold(dev);
3fc2de2f 2063 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2064 }
d342894c 2065
2066 del_timer_sync(&vxlan->age_timer);
2067
2068 vxlan_flush(vxlan);
2069
2070 return 0;
2071}
2072
d342894c 2073/* Stub, nothing needs to be done. */
2074static void vxlan_set_multicast_list(struct net_device *dev)
2075{
2076}
2077
345010b5
DB		 2078static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2079{
2080 struct vxlan_dev *vxlan = netdev_priv(dev);
2081 struct vxlan_rdst *dst = &vxlan->default_dst;
2082 struct net_device *lowerdev;
2083 int max_mtu;
2084
f01ec1c0 2085 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2086 if (lowerdev == NULL)
2087 return eth_change_mtu(dev, new_mtu);
2088
2089 if (dst->remote_ip.sa.sa_family == AF_INET6)
2090 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2091 else
2092 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2093
2094 if (new_mtu < 68 || new_mtu > max_mtu)
2095 return -EINVAL;
2096
2097 dev->mtu = new_mtu;
2098 return 0;
2099}
2100
d342894c 2101static const struct net_device_ops vxlan_netdev_ops = {
2102 .ndo_init = vxlan_init,
ebf4063e 2103 .ndo_uninit = vxlan_uninit,
d342894c 2104 .ndo_open = vxlan_open,
2105 .ndo_stop = vxlan_stop,
2106 .ndo_start_xmit = vxlan_xmit,
e8171045 2107 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2108 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2109 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2110 .ndo_validate_addr = eth_validate_addr,
2111 .ndo_set_mac_address = eth_mac_addr,
2112 .ndo_fdb_add = vxlan_fdb_add,
2113 .ndo_fdb_del = vxlan_fdb_delete,
2114 .ndo_fdb_dump = vxlan_fdb_dump,
2115};
2116
2117/* Info for udev, that this is a virtual tunnel endpoint */
2118static struct device_type vxlan_type = {
2119 .name = "vxlan",
2120};
2121
53cf5275 2122/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2123 * supply the listening VXLAN udp ports. Callers are expected
2124 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2125 */
2126void vxlan_get_rx_port(struct net_device *dev)
2127{
2128 struct vxlan_sock *vs;
2129 struct net *net = dev_net(dev);
2130 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2131 sa_family_t sa_family;
35e42379
JG		 2132 __be16 port;
2133 unsigned int i;
53cf5275
JG		 2134
2135 spin_lock(&vn->sock_lock);
2136 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2137 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2138 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2139 sa_family = vs->sock->sk->sk_family;
2140 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2141 port);
2142 }
2143 }
2144 spin_unlock(&vn->sock_lock);
2145}
2146EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2147
d342894c 2148/* Initialize the device structure. */
2149static void vxlan_setup(struct net_device *dev)
2150{
2151 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2152 unsigned int h;
d342894c 2153
2154 eth_hw_addr_random(dev);
2155 ether_setup(dev);
e4c7ed41 2156 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2157 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2158 else
2853af6a 2159 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2160
2161 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2162 dev->destructor = free_netdev;
d342894c 2163 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2164
2165 dev->tx_queue_len = 0;
2166 dev->features |= NETIF_F_LLTX;
d6727fe3 2167 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2168 dev->features |= NETIF_F_RXCSUM;
05c0db08 2169 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2170
1eaa8178
PS		 2171 dev->vlan_features = dev->features;
2172 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2173 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2174 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2175 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2176 netif_keep_dst(dev);
6602d007 2177 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2178
553675fb 2179 INIT_LIST_HEAD(&vxlan->next);
d342894c 2180 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2181 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2182 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2183 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2184
2185 init_timer_deferrable(&vxlan->age_timer);
2186 vxlan->age_timer.function = vxlan_cleanup;
2187 vxlan->age_timer.data = (unsigned long) vxlan;
2188
823aa873 2189 vxlan->dst_port = htons(vxlan_port);
05f47d69 2190
d342894c 2191 vxlan->dev = dev;
2192
2193 for (h = 0; h < FDB_HASH_SIZE; ++h)
2194 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2195}
2196
2197static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2198 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2199 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2200 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2201 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2202 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2203 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2204 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2205 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2206 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2207 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2208 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2209 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2210 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2211 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2212 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2213 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2214 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2215 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2216 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2217 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
d342894c 2218};
2219
2220static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2221{
2222 if (tb[IFLA_ADDRESS]) {
2223 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2224 pr_debug("invalid link address (not ethernet)\n");
2225 return -EINVAL;
2226 }
2227
2228 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2229 pr_debug("invalid all zero ethernet address\n");
2230 return -EADDRNOTAVAIL;
2231 }
2232 }
2233
2234 if (!data)
2235 return -EINVAL;
2236
2237 if (data[IFLA_VXLAN_ID]) {
2238 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2239 if (id >= VXLAN_VID_MASK)
2240 return -ERANGE;
2241 }
2242
05f47d69 2243 if (data[IFLA_VXLAN_PORT_RANGE]) {
2244 const struct ifla_vxlan_port_range *p
2245 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2246
2247 if (ntohs(p->high) < ntohs(p->low)) {
2248 pr_debug("port range %u .. %u not valid\n",
2249 ntohs(p->low), ntohs(p->high));
2250 return -EINVAL;
2251 }
2252 }
2253
d342894c 2254 return 0;
2255}
2256
1b13c97f
YB		 2257static void vxlan_get_drvinfo(struct net_device *netdev,
2258 struct ethtool_drvinfo *drvinfo)
2259{
2260 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2261 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2262}
2263
2264static const struct ethtool_ops vxlan_ethtool_ops = {
2265 .get_drvinfo = vxlan_get_drvinfo,
2266 .get_link = ethtool_op_get_link,
2267};
2268
553675fb 2269static void vxlan_del_work(struct work_struct *work)
2270{
2271 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2272 udp_tunnel_sock_release(vs->sock);
553675fb 2273 kfree_rcu(vs, rcu);
2274}
2275
3ee64f39
TH		 2276static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2277 __be16 port, u32 flags)
553675fb 2278{
e4c7ed41 2279 struct socket *sock;
3ee64f39
TH		 2280 struct udp_port_cfg udp_conf;
2281 int err;
e4c7ed41 2282
3ee64f39 2283 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2284
3ee64f39
TH		 2285 if (ipv6) {
2286 udp_conf.family = AF_INET6;
2287 udp_conf.use_udp6_tx_checksums =
3dc2b6a8 2288 !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
3ee64f39 2289 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2290 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2291 } else {
2292 udp_conf.family = AF_INET;
2293 udp_conf.local_ip.s_addr = INADDR_ANY;
2294 udp_conf.use_udp_checksums =
2295 !!(flags & VXLAN_F_UDP_CSUM);
e4c7ed41 2296 }
e4c7ed41 2297
3ee64f39 2298 udp_conf.local_udp_port = port;
553675fb 2299
3ee64f39
TH		 2300 /* Open UDP socket */
2301 err = udp_sock_create(net, &udp_conf, &sock);
2302 if (err < 0)
2303 return ERR_PTR(err);
e4c7ed41 2304
39deb2c7 2305 return sock;
e4c7ed41
CW		 2306}
2307
2308/* Create new listen socket if needed */
2309static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2310 vxlan_rcv_t *rcv, void *data,
2311 u32 flags)
e4c7ed41
CW		 2312{
2313 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2314 struct vxlan_sock *vs;
2315 struct socket *sock;
e4c7ed41 2316 unsigned int h;
359a0ea9 2317 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2318 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2319
dc01e7d3 2320 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2321 if (!vs)
2322 return ERR_PTR(-ENOMEM);
2323
2324 for (h = 0; h < VNI_HASH_SIZE; ++h)
2325 INIT_HLIST_HEAD(&vs->vni_list[h]);
2326
2327 INIT_WORK(&vs->del_work, vxlan_del_work);
2328
3ee64f39 2329 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2330 if (IS_ERR(sock)) {
553675fb 2331 kfree(vs);
e50fddc8 2332 return ERR_CAST(sock);
553675fb 2333 }
e4c7ed41
CW		 2334
2335 vs->sock = sock;
9c2e24e1 2336 atomic_set(&vs->refcnt, 1);
5cfccc5a 2337 vs->rcv = rcv;
012a5729 2338 vs->data = data;
553675fb 2339
dc01e7d3
OG		 2340 /* Initialize the vxlan udp offloads structure */
2341 vs->udp_offloads.port = port;
2342 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2343 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2344
9c2e24e1
PS		 2345 spin_lock(&vn->sock_lock);
2346 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2347 vxlan_notify_add_rx_port(vs);
9c2e24e1 2348 spin_unlock(&vn->sock_lock);
553675fb 2349
2350 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2351 tunnel_cfg.sk_user_data = vs;
2352 tunnel_cfg.encap_type = 1;
2353 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2354 tunnel_cfg.encap_destroy = NULL;
2355
2356 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2357
9c2e24e1
PS		 2358 return vs;
2359}
2360
012a5729
PS		 2361struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2362 vxlan_rcv_t *rcv, void *data,
359a0ea9 2363 bool no_share, u32 flags)
9c2e24e1
PS		 2364{
2365 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2366 struct vxlan_sock *vs;
19ca9fc1 2367 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2368
359a0ea9 2369 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2370 if (!IS_ERR(vs))
2371 return vs;
553675fb 2372
012a5729
PS		 2373 if (no_share) /* Return error if sharing is not allowed. */
2374 return vs;
2375
9c2e24e1 2376 spin_lock(&vn->sock_lock);
19ca9fc1 2377 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port);
00c83b01
ML		 2378 if (vs && ((vs->rcv != rcv) ||
2379 !atomic_add_unless(&vs->refcnt, 1, 0)))
5cfccc5a 2380 vs = ERR_PTR(-EBUSY);
5cfccc5a
PS		 2381 spin_unlock(&vn->sock_lock);
2382
2383 if (!vs)
9c2e24e1
PS		 2384 vs = ERR_PTR(-EINVAL);
2385
553675fb 2386 return vs;
2387}
012a5729 2388EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2389
1c51a915
SH		 2390/* Scheduled at device creation to bind to a socket */
2391static void vxlan_sock_work(struct work_struct *work)
2392{
9c2e24e1 2393 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2394 struct net *net = vxlan->net;
1c51a915 2395 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2396 __be16 port = vxlan->dst_port;
2397 struct vxlan_sock *nvs;
1c51a915 2398
359a0ea9 2399 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2400 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2401 if (!IS_ERR(nvs))
2402 vxlan_vs_add_dev(nvs, vxlan);
2403 spin_unlock(&vn->sock_lock);
2404
2405 dev_put(vxlan->dev);
1c51a915
SH		 2406}
2407
d342894c 2408static int vxlan_newlink(struct net *net, struct net_device *dev,
2409 struct nlattr *tb[], struct nlattr *data[])
2410{
553675fb 2411 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2412 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2413 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2414 __u32 vni;
2415 int err;
e4c7ed41 2416 bool use_ipv6 = false;
d342894c 2417
2418 if (!data[IFLA_VXLAN_ID])
2419 return -EINVAL;
2420
f01ec1c0
ND		 2421 vxlan->net = dev_net(dev);
2422
d342894c 2423 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2424 dst->remote_vni = vni;
d342894c 2425
5933a7bb
MR		 2426 /* Unless IPv6 is explicitly requested, assume IPv4 */
2427 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2428 if (data[IFLA_VXLAN_GROUP]) {
2429 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2430 } else if (data[IFLA_VXLAN_GROUP6]) {
2431 if (!IS_ENABLED(CONFIG_IPV6))
2432 return -EPFNOSUPPORT;
2433
2434 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2435 sizeof(struct in6_addr));
2436 dst->remote_ip.sa.sa_family = AF_INET6;
2437 use_ipv6 = true;
2438 }
d342894c 2439
e4c7ed41
CW		 2440 if (data[IFLA_VXLAN_LOCAL]) {
2441 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2442 vxlan->saddr.sa.sa_family = AF_INET;
2443 } else if (data[IFLA_VXLAN_LOCAL6]) {
2444 if (!IS_ENABLED(CONFIG_IPV6))
2445 return -EPFNOSUPPORT;
2446
2447 /* TODO: respect scope id */
2448 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2449 sizeof(struct in6_addr));
2450 vxlan->saddr.sa.sa_family = AF_INET6;
2451 use_ipv6 = true;
2452 }
d342894c 2453
34e02aa1 2454 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2455 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2456 struct net_device *lowerdev
c7995c43 2457 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2458
2459 if (!lowerdev) {
c7995c43 2460 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2461 return -ENODEV;
2462 }
d342894c 2463
e4c7ed41
CW		 2464#if IS_ENABLED(CONFIG_IPV6)
2465 if (use_ipv6) {
2466 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2467 if (idev && idev->cnf.disable_ipv6) {
2468 pr_info("IPv6 is disabled via sysctl\n");
2469 return -EPERM;
2470 }
2471 vxlan->flags |= VXLAN_F_IPV6;
2472 }
2473#endif
2474
34e02aa1 2475 if (!tb[IFLA_MTU])
e4c7ed41 2476 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2477
2853af6a 2478 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2479 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2480 } else if (use_ipv6)
2481 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2482
2483 if (data[IFLA_VXLAN_TOS])
2484 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2485
afb97186
VB		 2486 if (data[IFLA_VXLAN_TTL])
2487 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2488
d342894c 2489 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2490 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2491
2492 if (data[IFLA_VXLAN_AGEING])
2493 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2494 else
2495 vxlan->age_interval = FDB_AGE_DEFAULT;
2496
e4f67add
DS		 2497 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2498 vxlan->flags |= VXLAN_F_PROXY;
2499
2500 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2501 vxlan->flags |= VXLAN_F_RSC;
2502
2503 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2504 vxlan->flags |= VXLAN_F_L2MISS;
2505
2506 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2507 vxlan->flags |= VXLAN_F_L3MISS;
2508
d342894c 2509 if (data[IFLA_VXLAN_LIMIT])
2510 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2511
05f47d69 2512 if (data[IFLA_VXLAN_PORT_RANGE]) {
2513 const struct ifla_vxlan_port_range *p
2514 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2515 vxlan->port_min = ntohs(p->low);
2516 vxlan->port_max = ntohs(p->high);
2517 }
2518
823aa873 2519 if (data[IFLA_VXLAN_PORT])
2520 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2521
359a0ea9
TH		 2522 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2523 vxlan->flags |= VXLAN_F_UDP_CSUM;
2524
2525 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2526 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2527 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2528
2529 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2530 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2531 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2532
19ca9fc1
ML		 2533 if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
2534 vxlan->dst_port)) {
553675fb 2535 pr_info("duplicate VNI %u\n", vni);
2536 return -EEXIST;
2537 }
2538
7ad24ea4 2539 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2540
2936b6ab
SS		 2541 /* create an fdb entry for a valid default destination */
2542 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2543 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2544 &vxlan->default_dst.remote_ip,
2545 NUD_REACHABLE|NUD_PERMANENT,
2546 NLM_F_EXCL|NLM_F_CREATE,
2547 vxlan->dst_port,
2548 vxlan->default_dst.remote_vni,
2549 vxlan->default_dst.remote_ifindex,
2550 NTF_SELF);
2551 if (err)
2552 return err;
2553 }
d342894c 2554
afbd8bae
MR		 2555 err = register_netdevice(dev);
2556 if (err) {
ba609e9b 2557 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2558 return err;
2559 }
2560
553675fb 2561 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2562
2563 return 0;
d342894c 2564}
2565
2566static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2567{
2568 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2569 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2570
fe5c3561 2571 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2572 if (!hlist_unhashed(&vxlan->hlist))
2573 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2574 spin_unlock(&vn->sock_lock);
2575
553675fb 2576 list_del(&vxlan->next);
d342894c 2577 unregister_netdevice_queue(dev, head);
2578}
2579
2580static size_t vxlan_get_size(const struct net_device *dev)
2581{
2582
2583 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2584 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2585 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2586 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2587 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2588 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2589 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2590 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2591 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2592 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2593 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2594 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2595 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2596 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2597 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2598 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2599 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2600 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
d342894c 2601 0;
2602}
2603
2604static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2605{
2606 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2607 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2608 struct ifla_vxlan_port_range ports = {
2609 .low = htons(vxlan->port_min),
2610 .high = htons(vxlan->port_max),
2611 };
d342894c 2612
c7995c43 2613 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2614 goto nla_put_failure;
2615
e4c7ed41
CW		 2616 if (!vxlan_addr_any(&dst->remote_ip)) {
2617 if (dst->remote_ip.sa.sa_family == AF_INET) {
2618 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2619 dst->remote_ip.sin.sin_addr.s_addr))
2620 goto nla_put_failure;
2621#if IS_ENABLED(CONFIG_IPV6)
2622 } else {
2623 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2624 &dst->remote_ip.sin6.sin6_addr))
2625 goto nla_put_failure;
2626#endif
2627 }
2628 }
d342894c 2629
c7995c43 2630 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2631 goto nla_put_failure;
2632
e4c7ed41
CW		 2633 if (!vxlan_addr_any(&vxlan->saddr)) {
2634 if (vxlan->saddr.sa.sa_family == AF_INET) {
2635 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2636 vxlan->saddr.sin.sin_addr.s_addr))
2637 goto nla_put_failure;
2638#if IS_ENABLED(CONFIG_IPV6)
2639 } else {
2640 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2641 &vxlan->saddr.sin6.sin6_addr))
2642 goto nla_put_failure;
2643#endif
2644 }
2645 }
d342894c 2646
2647 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2648 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2649 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2650 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2651 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2652 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2653 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2654 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2655 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2656 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2657 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2658 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2659 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2660 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2661 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2662 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2663 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2664 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2665 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2666 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
d342894c 2667 goto nla_put_failure;
2668
05f47d69 2669 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2670 goto nla_put_failure;
2671
d342894c 2672 return 0;
2673
2674nla_put_failure:
2675 return -EMSGSIZE;
2676}
2677
2678static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2679 .kind = "vxlan",
2680 .maxtype = IFLA_VXLAN_MAX,
2681 .policy = vxlan_policy,
2682 .priv_size = sizeof(struct vxlan_dev),
2683 .setup = vxlan_setup,
2684 .validate = vxlan_validate,
2685 .newlink = vxlan_newlink,
2686 .dellink = vxlan_dellink,
2687 .get_size = vxlan_get_size,
2688 .fill_info = vxlan_fill_info,
2689};
2690
acaf4e70
DB		 2691static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2692 struct net_device *dev)
2693{
2694 struct vxlan_dev *vxlan, *next;
2695 LIST_HEAD(list_kill);
2696
2697 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2698 struct vxlan_rdst *dst = &vxlan->default_dst;
2699
2700 /* In case we created vxlan device with carrier
2701 * and we loose the carrier due to module unload
2702 * we also need to remove vxlan device. In other
2703 * cases, it's not necessary and remote_ifindex
2704 * is 0 here, so no matches.
2705 */
2706 if (dst->remote_ifindex == dev->ifindex)
2707 vxlan_dellink(vxlan->dev, &list_kill);
2708 }
2709
2710 unregister_netdevice_many(&list_kill);
2711}
2712
2713static int vxlan_lowerdev_event(struct notifier_block *unused,
2714 unsigned long event, void *ptr)
2715{
2716 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2717 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2718
783c1463 2719 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2720 vxlan_handle_lowerdev_unregister(vn, dev);
2721
2722 return NOTIFY_DONE;
2723}
2724
2725static struct notifier_block vxlan_notifier_block __read_mostly = {
2726 .notifier_call = vxlan_lowerdev_event,
2727};
2728
d342894c 2729static __net_init int vxlan_init_net(struct net *net)
2730{
2731 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2732 unsigned int h;
d342894c 2733
553675fb 2734 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2735 spin_lock_init(&vn->sock_lock);
d342894c 2736
553675fb 2737 for (h = 0; h < PORT_HASH_SIZE; ++h)
2738 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2739
2740 return 0;
2741}
2742
f01ec1c0
ND		 2743static void __net_exit vxlan_exit_net(struct net *net)
2744{
2745 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2746 struct vxlan_dev *vxlan, *next;
2747 struct net_device *dev, *aux;
2748 LIST_HEAD(list);
2749
2750 rtnl_lock();
2751 for_each_netdev_safe(net, dev, aux)
2752 if (dev->rtnl_link_ops == &vxlan_link_ops)
2753 unregister_netdevice_queue(dev, &list);
2754
2755 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2756 /* If vxlan->dev is in the same netns, it has already been added
2757 * to the list by the previous loop.
2758 */
2759 if (!net_eq(dev_net(vxlan->dev), net))
2760 unregister_netdevice_queue(dev, &list);
2761 }
2762
2763 unregister_netdevice_many(&list);
2764 rtnl_unlock();
2765}
2766
d342894c 2767static struct pernet_operations vxlan_net_ops = {
2768 .init = vxlan_init_net,
f01ec1c0 2769 .exit = vxlan_exit_net,
d342894c 2770 .id = &vxlan_net_id,
2771 .size = sizeof(struct vxlan_net),
2772};
2773
2774static int __init vxlan_init_module(void)
2775{
2776 int rc;
2777
758c57d1
SH		 2778 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2779 if (!vxlan_wq)
2780 return -ENOMEM;
2781
d342894c 2782 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2783
783c1463 2784 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 2785 if (rc)
2786 goto out1;
2787
acaf4e70 2788 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 2789 if (rc)
2790 goto out2;
2791
acaf4e70
DB		 2792 rc = rtnl_link_register(&vxlan_link_ops);
2793 if (rc)
2794 goto out3;
d342894c 2795
acaf4e70
DB		 2796 return 0;
2797out3:
2798 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 2799out2:
783c1463 2800 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 2801out1:
758c57d1 2802 destroy_workqueue(vxlan_wq);
d342894c 2803 return rc;
2804}
7332a13b 2805late_initcall(vxlan_init_module);
d342894c 2806
2807static void __exit vxlan_cleanup_module(void)
2808{
b7153984 2809 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 2810 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 2811 destroy_workqueue(vxlan_wq);
783c1463
DB		 2812 unregister_pernet_subsys(&vxlan_net_ops);
2813 /* rcu_barrier() is called by netns */
d342894c 2814}
2815module_exit(vxlan_cleanup_module);
2816
2817MODULE_LICENSE("GPL");
2818MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2819MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 2820MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 2821MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.390618 seconds and 5 git commands to generate.