projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tunnels: advertise link netns via netlink
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
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
ac5132d1
TG		 266/* Find VXLAN socket based on network namespace, address family and UDP port
267 * and enabled unshareable flags.
268 */
269static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
270 __be16 port, u32 flags)
553675fb 271{
272 struct vxlan_sock *vs;
ac5132d1 273 u32 match_flags = flags & VXLAN_F_UNSHAREABLE;
553675fb 274
275 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 276 if (inet_sk(vs->sock->sk)->inet_sport == port &&
ac5132d1
TG		 277 inet_sk(vs->sock->sk)->sk.sk_family == family &&
278 (vs->flags & VXLAN_F_UNSHAREABLE) == match_flags)
553675fb 279 return vs;
280 }
281 return NULL;
d342894c 282}
283
5cfccc5a 284static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 285{
286 struct vxlan_dev *vxlan;
d342894c 287
553675fb 288 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 289 if (vxlan->default_dst.remote_vni == id)
d342894c 290 return vxlan;
291 }
292
293 return NULL;
294}
295
5cfccc5a 296/* Look up VNI in a per net namespace table */
19ca9fc1 297static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 298 sa_family_t family, __be16 port,
299 u32 flags)
5cfccc5a
PS		 300{
301 struct vxlan_sock *vs;
302
ac5132d1 303 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 304 if (!vs)
305 return NULL;
306
307 return vxlan_vs_find_vni(vs, id);
308}
309
d342894c 310/* Fill in neighbour message in skbuff. */
311static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 312 const struct vxlan_fdb *fdb,
313 u32 portid, u32 seq, int type, unsigned int flags,
314 const struct vxlan_rdst *rdst)
d342894c 315{
316 unsigned long now = jiffies;
317 struct nda_cacheinfo ci;
318 struct nlmsghdr *nlh;
319 struct ndmsg *ndm;
e4f67add 320 bool send_ip, send_eth;
d342894c 321
322 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
323 if (nlh == NULL)
324 return -EMSGSIZE;
325
326 ndm = nlmsg_data(nlh);
327 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 328
329 send_eth = send_ip = true;
330
331 if (type == RTM_GETNEIGH) {
332 ndm->ndm_family = AF_INET;
e4c7ed41 333 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 334 send_eth = !is_zero_ether_addr(fdb->eth_addr);
335 } else
336 ndm->ndm_family = AF_BRIDGE;
d342894c 337 ndm->ndm_state = fdb->state;
338 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 339 ndm->ndm_flags = fdb->flags;
545469f7 340 ndm->ndm_type = RTN_UNICAST;
d342894c 341
e4f67add 342 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 343 goto nla_put_failure;
344
e4c7ed41 345 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 346 goto nla_put_failure;
347
823aa873 348 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 349 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
350 goto nla_put_failure;
c7995c43 351 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 352 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 353 goto nla_put_failure;
354 if (rdst->remote_ifindex &&
355 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 356 goto nla_put_failure;
357
358 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
359 ci.ndm_confirmed = 0;
360 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
361 ci.ndm_refcnt = 0;
362
363 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
364 goto nla_put_failure;
365
053c095a
JB		 366 nlmsg_end(skb, nlh);
367 return 0;
d342894c 368
369nla_put_failure:
370 nlmsg_cancel(skb, nlh);
371 return -EMSGSIZE;
372}
373
374static inline size_t vxlan_nlmsg_size(void)
375{
376 return NLMSG_ALIGN(sizeof(struct ndmsg))
377 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 378 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 379 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 380 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
381 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 382 + nla_total_size(sizeof(struct nda_cacheinfo));
383}
384
9e4b93f9
ND		 385static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
386 struct vxlan_rdst *rd, int type)
d342894c 387{
388 struct net *net = dev_net(vxlan->dev);
389 struct sk_buff *skb;
390 int err = -ENOBUFS;
391
392 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
393 if (skb == NULL)
394 goto errout;
395
9e4b93f9 396 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 397 if (err < 0) {
398 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
399 WARN_ON(err == -EMSGSIZE);
400 kfree_skb(skb);
401 goto errout;
402 }
403
404 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
405 return;
406errout:
407 if (err < 0)
408 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
409}
410
e4c7ed41 411static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 412{
413 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 414 struct vxlan_fdb f = {
415 .state = NUD_STALE,
416 };
417 struct vxlan_rdst remote = {
e4c7ed41 418 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 419 .remote_vni = VXLAN_N_VID,
420 };
3e61aa8f 421
9e4b93f9 422 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 423}
424
425static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
426{
bb3fd687
SH		 427 struct vxlan_fdb f = {
428 .state = NUD_STALE,
429 };
9e4b93f9 430 struct vxlan_rdst remote = { };
e4f67add 431
e4f67add
DS		 432 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
433
9e4b93f9 434 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 435}
436
d342894c 437/* Hash Ethernet address */
438static u32 eth_hash(const unsigned char *addr)
439{
440 u64 value = get_unaligned((u64 *)addr);
441
442 /* only want 6 bytes */
443#ifdef __BIG_ENDIAN
d342894c 444 value >>= 16;
321fb991 445#else
446 value <<= 16;
d342894c 447#endif
448 return hash_64(value, FDB_HASH_BITS);
449}
450
451/* Hash chain to use given mac address */
452static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
453 const u8 *mac)
454{
455 return &vxlan->fdb_head[eth_hash(mac)];
456}
457
458/* Look up Ethernet address in forwarding table */
014be2c8 459static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 460 const u8 *mac)
d342894c 461{
462 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
463 struct vxlan_fdb *f;
d342894c 464
b67bfe0d 465 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 466 if (ether_addr_equal(mac, f->eth_addr))
d342894c 467 return f;
468 }
469
470 return NULL;
471}
472
014be2c8
SS		 473static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
474 const u8 *mac)
475{
476 struct vxlan_fdb *f;
477
478 f = __vxlan_find_mac(vxlan, mac);
479 if (f)
480 f->used = jiffies;
481
482 return f;
483}
484
a5e7c10a
MR		 485/* caller should hold vxlan->hash_lock */
486static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 487 union vxlan_addr *ip, __be16 port,
a5e7c10a 488 __u32 vni, __u32 ifindex)
6681712d 489{
3e61aa8f 490 struct vxlan_rdst *rd;
6681712d 491
3e61aa8f 492 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 493 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 494 rd->remote_port == port &&
495 rd->remote_vni == vni &&
496 rd->remote_ifindex == ifindex)
a5e7c10a 497 return rd;
6681712d 498 }
3e61aa8f 499
a5e7c10a
MR		 500 return NULL;
501}
502
906dc186
TR		 503/* Replace destination of unicast mac */
504static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 505 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 506{
507 struct vxlan_rdst *rd;
508
509 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
510 if (rd)
511 return 0;
512
513 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
514 if (!rd)
515 return 0;
e4c7ed41 516 rd->remote_ip = *ip;
906dc186
TR		 517 rd->remote_port = port;
518 rd->remote_vni = vni;
519 rd->remote_ifindex = ifindex;
520 return 1;
521}
522
a5e7c10a
MR		 523/* Add/update destinations for multicast */
524static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 525 union vxlan_addr *ip, __be16 port, __u32 vni,
526 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 527{
528 struct vxlan_rdst *rd;
529
530 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
531 if (rd)
532 return 0;
533
6681712d
DS		 534 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
535 if (rd == NULL)
536 return -ENOBUFS;
e4c7ed41 537 rd->remote_ip = *ip;
6681712d
DS		 538 rd->remote_port = port;
539 rd->remote_vni = vni;
540 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 541
542 list_add_tail_rcu(&rd->list, &f->remotes);
543
9e4b93f9 544 *rdp = rd;
6681712d
DS		 545 return 1;
546}
547
dfd8645e
TH		 548static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
549 unsigned int off,
550 struct vxlanhdr *vh, size_t hdrlen,
551 u32 data)
552{
553 size_t start, offset, plen;
554 __wsum delta;
555
556 if (skb->remcsum_offload)
557 return vh;
558
559 if (!NAPI_GRO_CB(skb)->csum_valid)
560 return NULL;
561
562 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
563 offset = start + ((data & VXLAN_RCO_UDP) ?
564 offsetof(struct udphdr, check) :
565 offsetof(struct tcphdr, check));
566
567 plen = hdrlen + offset + sizeof(u16);
568
569 /* Pull checksum that will be written */
570 if (skb_gro_header_hard(skb, off + plen)) {
571 vh = skb_gro_header_slow(skb, off + plen, off);
572 if (!vh)
573 return NULL;
574 }
575
576 delta = remcsum_adjust((void *)vh + hdrlen,
577 NAPI_GRO_CB(skb)->csum, start, offset);
578
579 /* Adjust skb->csum since we changed the packet */
580 skb->csum = csum_add(skb->csum, delta);
581 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
582
583 skb->remcsum_offload = 1;
584
585 return vh;
586}
587
a2b12f3c
TH		 588static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
589 struct sk_buff *skb,
590 struct udp_offload *uoff)
dc01e7d3
OG		 591{
592 struct sk_buff *p, **pp = NULL;
593 struct vxlanhdr *vh, *vh2;
9b174d88 594 unsigned int hlen, off_vx;
dc01e7d3 595 int flush = 1;
dfd8645e
TH		 596 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
597 udp_offloads);
598 u32 flags;
dc01e7d3
OG		 599
600 off_vx = skb_gro_offset(skb);
601 hlen = off_vx + sizeof(*vh);
602 vh = skb_gro_header_fast(skb, off_vx);
603 if (skb_gro_header_hard(skb, hlen)) {
604 vh = skb_gro_header_slow(skb, hlen, off_vx);
605 if (unlikely(!vh))
606 goto out;
607 }
dc01e7d3 608
dfd8645e
TH		 609 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
610 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
611
612 flags = ntohl(vh->vx_flags);
613
614 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
615 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
616 ntohl(vh->vx_vni));
617
618 if (!vh)
619 goto out;
620 }
621
dc01e7d3
OG		 622 flush = 0;
623
624 for (p = *head; p; p = p->next) {
625 if (!NAPI_GRO_CB(p)->same_flow)
626 continue;
627
628 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 629 if (vh->vx_flags != vh2->vx_flags ||
630 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 631 NAPI_GRO_CB(p)->same_flow = 0;
632 continue;
633 }
dc01e7d3
OG		 634 }
635
9b174d88 636 pp = eth_gro_receive(head, skb);
dc01e7d3 637
dc01e7d3
OG		 638out:
639 NAPI_GRO_CB(skb)->flush |= flush;
640
641 return pp;
642}
643
a2b12f3c
TH		 644static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
645 struct udp_offload *uoff)
dc01e7d3 646{
cfdf1e1b
JG		 647 udp_tunnel_gro_complete(skb, nhoff);
648
9b174d88 649 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 650}
651
53cf5275 652/* Notify netdevs that UDP port started listening */
dc01e7d3 653static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 654{
655 struct net_device *dev;
dc01e7d3 656 struct sock *sk = vs->sock->sk;
53cf5275
JG		 657 struct net *net = sock_net(sk);
658 sa_family_t sa_family = sk->sk_family;
35e42379 659 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 660 int err;
661
662 if (sa_family == AF_INET) {
663 err = udp_add_offload(&vs->udp_offloads);
664 if (err)
665 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
666 }
53cf5275
JG		 667
668 rcu_read_lock();
669 for_each_netdev_rcu(net, dev) {
670 if (dev->netdev_ops->ndo_add_vxlan_port)
671 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
672 port);
673 }
674 rcu_read_unlock();
675}
676
677/* Notify netdevs that UDP port is no more listening */
dc01e7d3 678static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 679{
680 struct net_device *dev;
dc01e7d3 681 struct sock *sk = vs->sock->sk;
53cf5275
JG		 682 struct net *net = sock_net(sk);
683 sa_family_t sa_family = sk->sk_family;
35e42379 684 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 685
686 rcu_read_lock();
687 for_each_netdev_rcu(net, dev) {
688 if (dev->netdev_ops->ndo_del_vxlan_port)
689 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
690 port);
691 }
692 rcu_read_unlock();
dc01e7d3
OG		 693
694 if (sa_family == AF_INET)
695 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 696}
697
d342894c 698/* Add new entry to forwarding table -- assumes lock held */
699static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 700 const u8 *mac, union vxlan_addr *ip,
6681712d 701 __u16 state, __u16 flags,
73cf3317 702 __be16 port, __u32 vni, __u32 ifindex,
ae884082 703 __u8 ndm_flags)
d342894c 704{
9e4b93f9 705 struct vxlan_rdst *rd = NULL;
d342894c 706 struct vxlan_fdb *f;
707 int notify = 0;
708
014be2c8 709 f = __vxlan_find_mac(vxlan, mac);
d342894c 710 if (f) {
711 if (flags & NLM_F_EXCL) {
712 netdev_dbg(vxlan->dev,
713 "lost race to create %pM\n", mac);
714 return -EEXIST;
715 }
716 if (f->state != state) {
717 f->state = state;
718 f->updated = jiffies;
719 notify = 1;
720 }
ae884082
DS		 721 if (f->flags != ndm_flags) {
722 f->flags = ndm_flags;
723 f->updated = jiffies;
724 notify = 1;
725 }
906dc186
TR		 726 if ((flags & NLM_F_REPLACE)) {
727 /* Only change unicasts */
728 if (!(is_multicast_ether_addr(f->eth_addr) ||
729 is_zero_ether_addr(f->eth_addr))) {
730 int rc = vxlan_fdb_replace(f, ip, port, vni,
731 ifindex);
732
733 if (rc < 0)
734 return rc;
735 notify |= rc;
736 } else
737 return -EOPNOTSUPP;
738 }
6681712d 739 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 740 (is_multicast_ether_addr(f->eth_addr) ||
741 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 742 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
743 &rd);
6681712d
DS		 744
745 if (rc < 0)
746 return rc;
747 notify |= rc;
748 }
d342894c 749 } else {
750 if (!(flags & NLM_F_CREATE))
751 return -ENOENT;
752
753 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
754 return -ENOSPC;
755
906dc186
TR		 756 /* Disallow replace to add a multicast entry */
757 if ((flags & NLM_F_REPLACE) &&
758 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
759 return -EOPNOTSUPP;
760
e4c7ed41 761 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 762 f = kmalloc(sizeof(*f), GFP_ATOMIC);
763 if (!f)
764 return -ENOMEM;
765
766 notify = 1;
d342894c 767 f->state = state;
ae884082 768 f->flags = ndm_flags;
d342894c 769 f->updated = f->used = jiffies;
3e61aa8f 770 INIT_LIST_HEAD(&f->remotes);
d342894c 771 memcpy(f->eth_addr, mac, ETH_ALEN);
772
9e4b93f9 773 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 774
d342894c 775 ++vxlan->addrcnt;
776 hlist_add_head_rcu(&f->hlist,
777 vxlan_fdb_head(vxlan, mac));
778 }
779
9e4b93f9
ND		 780 if (notify) {
781 if (rd == NULL)
782 rd = first_remote_rtnl(f);
783 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
784 }
d342894c 785
786 return 0;
787}
788
6706c82e 789static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 790{
791 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 792 struct vxlan_rdst *rd, *nd;
6681712d 793
3e61aa8f 794 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 795 kfree(rd);
6681712d
DS		 796 kfree(f);
797}
798
d342894c 799static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
800{
801 netdev_dbg(vxlan->dev,
802 "delete %pM\n", f->eth_addr);
803
804 --vxlan->addrcnt;
9e4b93f9 805 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 806
807 hlist_del_rcu(&f->hlist);
6681712d 808 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 809}
810
f0b074be 811static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 812 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 813{
6681712d 814 struct net *net = dev_net(vxlan->dev);
e4c7ed41 815 int err;
d342894c 816
f0b074be 817 if (tb[NDA_DST]) {
e4c7ed41
CW		 818 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
819 if (err)
820 return err;
f0b074be 821 } else {
e4c7ed41
CW		 822 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
823 if (remote->sa.sa_family == AF_INET) {
824 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
825 ip->sa.sa_family = AF_INET;
826#if IS_ENABLED(CONFIG_IPV6)
827 } else {
828 ip->sin6.sin6_addr = in6addr_any;
829 ip->sa.sa_family = AF_INET6;
830#endif
831 }
f0b074be 832 }
d342894c 833
6681712d 834 if (tb[NDA_PORT]) {
73cf3317 835 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 836 return -EINVAL;
f0b074be
MR		 837 *port = nla_get_be16(tb[NDA_PORT]);
838 } else {
839 *port = vxlan->dst_port;
840 }
6681712d
DS		 841
842 if (tb[NDA_VNI]) {
843 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
844 return -EINVAL;
f0b074be
MR		 845 *vni = nla_get_u32(tb[NDA_VNI]);
846 } else {
847 *vni = vxlan->default_dst.remote_vni;
848 }
6681712d
DS		 849
850 if (tb[NDA_IFINDEX]) {
5abb0029 851 struct net_device *tdev;
6681712d
DS		 852
853 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
854 return -EINVAL;
f0b074be 855 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 856 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 857 if (!tdev)
6681712d 858 return -EADDRNOTAVAIL;
f0b074be
MR		 859 } else {
860 *ifindex = 0;
861 }
862
863 return 0;
864}
865
866/* Add static entry (via netlink) */
867static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
868 struct net_device *dev,
f6f6424b 869 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 870{
871 struct vxlan_dev *vxlan = netdev_priv(dev);
872 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 873 union vxlan_addr ip;
f0b074be
MR		 874 __be16 port;
875 u32 vni, ifindex;
876 int err;
877
878 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
879 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
880 ndm->ndm_state);
881 return -EINVAL;
882 }
883
884 if (tb[NDA_DST] == NULL)
885 return -EINVAL;
886
887 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
888 if (err)
889 return err;
6681712d 890
5933a7bb
MR		 891 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
892 return -EAFNOSUPPORT;
893
d342894c 894 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 895 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 896 port, vni, ifindex, ndm->ndm_flags);
d342894c 897 spin_unlock_bh(&vxlan->hash_lock);
898
899 return err;
900}
901
902/* Delete entry (via netlink) */
1690be63
VY		 903static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
904 struct net_device *dev,
f6f6424b 905 const unsigned char *addr, u16 vid)
d342894c 906{
907 struct vxlan_dev *vxlan = netdev_priv(dev);
908 struct vxlan_fdb *f;
bc7892ba 909 struct vxlan_rdst *rd = NULL;
e4c7ed41 910 union vxlan_addr ip;
bc7892ba
MR		 911 __be16 port;
912 u32 vni, ifindex;
913 int err;
914
915 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
916 if (err)
917 return err;
918
919 err = -ENOENT;
d342894c 920
921 spin_lock_bh(&vxlan->hash_lock);
922 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 923 if (!f)
924 goto out;
925
e4c7ed41
CW		 926 if (!vxlan_addr_any(&ip)) {
927 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 928 if (!rd)
929 goto out;
930 }
931
932 err = 0;
933
934 /* remove a destination if it's not the only one on the list,
935 * otherwise destroy the fdb entry
936 */
937 if (rd && !list_is_singular(&f->remotes)) {
938 list_del_rcu(&rd->list);
9e4b93f9 939 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 940 kfree_rcu(rd, rcu);
bc7892ba 941 goto out;
d342894c 942 }
bc7892ba
MR		 943
944 vxlan_fdb_destroy(vxlan, f);
945
946out:
d342894c 947 spin_unlock_bh(&vxlan->hash_lock);
948
949 return err;
950}
951
952/* Dump forwarding table */
953static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 954 struct net_device *dev,
955 struct net_device *filter_dev, int idx)
d342894c 956{
957 struct vxlan_dev *vxlan = netdev_priv(dev);
958 unsigned int h;
959
960 for (h = 0; h < FDB_HASH_SIZE; ++h) {
961 struct vxlan_fdb *f;
d342894c 962 int err;
963
b67bfe0d 964 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 965 struct vxlan_rdst *rd;
6681712d 966
3e61aa8f
SH		 967 if (idx < cb->args[0])
968 goto skip;
969
970 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 971 err = vxlan_fdb_info(skb, vxlan, f,
972 NETLINK_CB(cb->skb).portid,
973 cb->nlh->nlmsg_seq,
974 RTM_NEWNEIGH,
975 NLM_F_MULTI, rd);
976 if (err < 0)
3e61aa8f 977 goto out;
6681712d 978 }
3e61aa8f
SH		 979skip:
980 ++idx;
d342894c 981 }
982 }
3e61aa8f 983out:
d342894c 984 return idx;
985}
986
987/* Watch incoming packets to learn mapping between Ethernet address
988 * and Tunnel endpoint.
26a41ae6 989 * Return true if packet is bogus and should be droppped.
d342894c 990 */
26a41ae6 991static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 992 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 993{
994 struct vxlan_dev *vxlan = netdev_priv(dev);
995 struct vxlan_fdb *f;
d342894c 996
997 f = vxlan_find_mac(vxlan, src_mac);
998 if (likely(f)) {
5ca5461c 999 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1000
e4c7ed41 1001 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 1002 return false;
1003
1004 /* Don't migrate static entries, drop packets */
eb064c3b 1005 if (f->state & NUD_NOARP)
26a41ae6 1006 return true;
d342894c 1007
1008 if (net_ratelimit())
1009 netdev_info(dev,
e4c7ed41 1010 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 1011 src_mac, &rdst->remote_ip, &src_ip);
d342894c 1012
e4c7ed41 1013 rdst->remote_ip = *src_ip;
d342894c 1014 f->updated = jiffies;
9e4b93f9 1015 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1016 } else {
1017 /* learned new entry */
1018 spin_lock(&vxlan->hash_lock);
3bf74b1a 1019
1020 /* close off race between vxlan_flush and incoming packets */
1021 if (netif_running(dev))
1022 vxlan_fdb_create(vxlan, src_mac, src_ip,
1023 NUD_REACHABLE,
1024 NLM_F_EXCL|NLM_F_CREATE,
1025 vxlan->dst_port,
1026 vxlan->default_dst.remote_vni,
1027 0, NTF_SELF);
d342894c 1028 spin_unlock(&vxlan->hash_lock);
1029 }
26a41ae6 1030
1031 return false;
d342894c 1032}
1033
d342894c 1034/* See if multicast group is already in use by other ID */
95ab0991 1035static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1036{
553675fb 1037 struct vxlan_dev *vxlan;
d342894c 1038
95ab0991
G		 1039 /* The vxlan_sock is only used by dev, leaving group has
1040 * no effect on other vxlan devices.
1041 */
1042 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1043 return false;
1044
553675fb 1045 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1046 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1047 continue;
d342894c 1048
95ab0991
G		 1049 if (vxlan->vn_sock != dev->vn_sock)
1050 continue;
1051
1052 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1053 &dev->default_dst.remote_ip))
1054 continue;
1055
1056 if (vxlan->default_dst.remote_ifindex !=
1057 dev->default_dst.remote_ifindex)
1058 continue;
1059
1060 return true;
553675fb 1061 }
d342894c 1062
1063 return false;
1064}
1065
7c47cedf 1066static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1067{
7c47cedf
SH		 1068 atomic_inc(&vs->refcnt);
1069}
d342894c 1070
012a5729 1071void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1072{
53cf5275
JG		 1073 struct sock *sk = vs->sock->sk;
1074 struct net *net = sock_net(sk);
1075 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1076
7c47cedf
SH		 1077 if (!atomic_dec_and_test(&vs->refcnt))
1078 return;
d342894c 1079
1c51a915 1080 spin_lock(&vn->sock_lock);
7c47cedf 1081 hlist_del_rcu(&vs->hlist);
dc01e7d3 1082 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1083 spin_unlock(&vn->sock_lock);
1084
7c47cedf 1085 queue_work(vxlan_wq, &vs->del_work);
d342894c 1086}
012a5729 1087EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1088
3fc2de2f 1089/* Callback to update multicast group membership when first VNI on
1090 * multicast asddress is brought up
1091 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1092 */
3fc2de2f 1093static void vxlan_igmp_join(struct work_struct *work)
d342894c 1094{
3fc2de2f 1095 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1096 struct vxlan_sock *vs = vxlan->vn_sock;
1097 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1098 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1099 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1100
d342894c 1101 lock_sock(sk);
e4c7ed41
CW		 1102 if (ip->sa.sa_family == AF_INET) {
1103 struct ip_mreqn mreq = {
1104 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1105 .imr_ifindex = ifindex,
1106 };
1107
1108 ip_mc_join_group(sk, &mreq);
1109#if IS_ENABLED(CONFIG_IPV6)
1110 } else {
1111 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1112 &ip->sin6.sin6_addr);
1113#endif
1114 }
3fc2de2f 1115 release_sock(sk);
1116
012a5729 1117 vxlan_sock_release(vs);
3fc2de2f 1118 dev_put(vxlan->dev);
1119}
1120
1121/* Inverse of vxlan_igmp_join when last VNI is brought down */
1122static void vxlan_igmp_leave(struct work_struct *work)
1123{
1124 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1125 struct vxlan_sock *vs = vxlan->vn_sock;
1126 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1127 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1128 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1129
1130 lock_sock(sk);
e4c7ed41
CW		 1131 if (ip->sa.sa_family == AF_INET) {
1132 struct ip_mreqn mreq = {
1133 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1134 .imr_ifindex = ifindex,
1135 };
1136
1137 ip_mc_leave_group(sk, &mreq);
1138#if IS_ENABLED(CONFIG_IPV6)
1139 } else {
1140 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1141 &ip->sin6.sin6_addr);
1142#endif
1143 }
1144
d342894c 1145 release_sock(sk);
d342894c 1146
012a5729 1147 vxlan_sock_release(vs);
7c47cedf 1148 dev_put(vxlan->dev);
d342894c 1149}
1150
dfd8645e
TH		 1151static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
1152 size_t hdrlen, u32 data)
1153{
1154 size_t start, offset, plen;
1155 __wsum delta;
1156
1157 if (skb->remcsum_offload) {
1158 /* Already processed in GRO path */
1159 skb->remcsum_offload = 0;
1160 return vh;
1161 }
1162
1163 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1164 offset = start + ((data & VXLAN_RCO_UDP) ?
1165 offsetof(struct udphdr, check) :
1166 offsetof(struct tcphdr, check));
1167
1168 plen = hdrlen + offset + sizeof(u16);
1169
1170 if (!pskb_may_pull(skb, plen))
1171 return NULL;
1172
1173 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1174
1175 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE))
1176 __skb_checksum_complete(skb);
1177
1178 delta = remcsum_adjust((void *)vh + hdrlen,
1179 skb->csum, start, offset);
1180
1181 /* Adjust skb->csum since we changed the packet */
1182 skb->csum = csum_add(skb->csum, delta);
1183
1184 return vh;
1185}
1186
d342894c 1187/* Callback from net/ipv4/udp.c to receive packets */
1188static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1189{
5cfccc5a 1190 struct vxlan_sock *vs;
d342894c 1191 struct vxlanhdr *vxh;
3bf39475 1192 u32 flags, vni;
3511494c 1193 struct vxlan_metadata md = {0};
d342894c 1194
d342894c 1195 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1196 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1197 goto error;
1198
7ce04758 1199 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1200 flags = ntohl(vxh->vx_flags);
1201 vni = ntohl(vxh->vx_vni);
1202
1203 if (flags & VXLAN_HF_VNI) {
1204 flags &= ~VXLAN_HF_VNI;
1205 } else {
1206 /* VNI flag always required to be set */
1207 goto bad_flags;
d342894c 1208 }
1209
5cfccc5a
PS		 1210 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1211 goto drop;
dfd8645e 1212 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1213
559835ea 1214 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1215 if (!vs)
d342894c 1216 goto drop;
d342894c 1217
dfd8645e
TH		 1218 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
1219 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni);
1220 if (!vxh)
1221 goto drop;
1222
1223 flags &= ~VXLAN_HF_RCO;
1224 vni &= VXLAN_VID_MASK;
1225 }
1226
3511494c
TG		 1227 /* For backwards compatibility, only allow reserved fields to be
1228 * used by VXLAN extensions if explicitly requested.
1229 */
1230 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1231 struct vxlanhdr_gbp *gbp;
1232
1233 gbp = (struct vxlanhdr_gbp *)vxh;
1234 md.gbp = ntohs(gbp->policy_id);
1235
1236 if (gbp->dont_learn)
1237 md.gbp |= VXLAN_GBP_DONT_LEARN;
1238
1239 if (gbp->policy_applied)
1240 md.gbp |= VXLAN_GBP_POLICY_APPLIED;
1241
1242 flags &= ~VXLAN_GBP_USED_BITS;
1243 }
1244
dfd8645e 1245 if (flags || (vni & ~VXLAN_VID_MASK)) {
3bf39475
TH		 1246 /* If there are any unprocessed flags remaining treat
1247 * this as a malformed packet. This behavior diverges from
1248 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1249 * in reserved fields are to be ignored. The approach here
1250 * maintains compatbility with previous stack code, and also
1251 * is more robust and provides a little more security in
1252 * adding extensions to VXLAN.
1253 */
1254
1255 goto bad_flags;
1256 }
1257
3511494c
TG		 1258 md.vni = vxh->vx_vni;
1259 vs->rcv(vs, skb, &md);
5cfccc5a
PS		 1260 return 0;
1261
1262drop:
1263 /* Consume bad packet */
1264 kfree_skb(skb);
1265 return 0;
1266
3bf39475
TH		 1267bad_flags:
1268 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1269 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1270
5cfccc5a
PS		 1271error:
1272 /* Return non vxlan pkt */
1273 return 1;
1274}
1275
3511494c
TG		 1276static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1277 struct vxlan_metadata *md)
5cfccc5a 1278{
e4c7ed41
CW		 1279 struct iphdr *oip = NULL;
1280 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1281 struct vxlan_dev *vxlan;
8f84985f 1282 struct pcpu_sw_netstats *stats;
e4c7ed41 1283 union vxlan_addr saddr;
5cfccc5a 1284 __u32 vni;
e4c7ed41
CW		 1285 int err = 0;
1286 union vxlan_addr *remote_ip;
5cfccc5a 1287
3511494c 1288 vni = ntohl(md->vni) >> 8;
5cfccc5a
PS		 1289 /* Is this VNI defined? */
1290 vxlan = vxlan_vs_find_vni(vs, vni);
1291 if (!vxlan)
d342894c 1292 goto drop;
d342894c 1293
e4c7ed41 1294 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1295 skb_reset_mac_header(skb);
f01ec1c0 1296 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1297 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1298 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1299
1300 /* Ignore packet loops (and multicast echo) */
7367d0b5 1301 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1302 goto drop;
1303
7ce04758 1304 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1305 if (remote_ip->sa.sa_family == AF_INET) {
1306 oip = ip_hdr(skb);
1307 saddr.sin.sin_addr.s_addr = oip->saddr;
1308 saddr.sa.sa_family = AF_INET;
1309#if IS_ENABLED(CONFIG_IPV6)
1310 } else {
1311 oip6 = ipv6_hdr(skb);
1312 saddr.sin6.sin6_addr = oip6->saddr;
1313 saddr.sa.sa_family = AF_INET6;
1314#endif
1315 }
1316
26a41ae6 1317 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1318 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1319 goto drop;
d342894c 1320
d342894c 1321 skb_reset_network_header(skb);
3511494c 1322 skb->mark = md->gbp;
0afb1666 1323
e4c7ed41
CW		 1324 if (oip6)
1325 err = IP6_ECN_decapsulate(oip6, skb);
1326 if (oip)
1327 err = IP_ECN_decapsulate(oip, skb);
1328
d342894c 1329 if (unlikely(err)) {
e4c7ed41
CW		 1330 if (log_ecn_error) {
1331 if (oip6)
1332 net_info_ratelimited("non-ECT from %pI6\n",
1333 &oip6->saddr);
1334 if (oip)
1335 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1336 &oip->saddr, oip->tos);
1337 }
d342894c 1338 if (err > 1) {
1339 ++vxlan->dev->stats.rx_frame_errors;
1340 ++vxlan->dev->stats.rx_errors;
1341 goto drop;
1342 }
1343 }
1344
e8171045 1345 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1346 u64_stats_update_begin(&stats->syncp);
1347 stats->rx_packets++;
1348 stats->rx_bytes += skb->len;
1349 u64_stats_update_end(&stats->syncp);
1350
1351 netif_rx(skb);
1352
5cfccc5a 1353 return;
d342894c 1354drop:
1355 /* Consume bad packet */
1356 kfree_skb(skb);
d342894c 1357}
1358
e4f67add
DS		 1359static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1360{
1361 struct vxlan_dev *vxlan = netdev_priv(dev);
1362 struct arphdr *parp;
1363 u8 *arpptr, *sha;
1364 __be32 sip, tip;
1365 struct neighbour *n;
1366
1367 if (dev->flags & IFF_NOARP)
1368 goto out;
1369
1370 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1371 dev->stats.tx_dropped++;
1372 goto out;
1373 }
1374 parp = arp_hdr(skb);
1375
1376 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1377 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1378 parp->ar_pro != htons(ETH_P_IP) ||
1379 parp->ar_op != htons(ARPOP_REQUEST) ||
1380 parp->ar_hln != dev->addr_len ||
1381 parp->ar_pln != 4)
1382 goto out;
1383 arpptr = (u8 *)parp + sizeof(struct arphdr);
1384 sha = arpptr;
1385 arpptr += dev->addr_len; /* sha */
1386 memcpy(&sip, arpptr, sizeof(sip));
1387 arpptr += sizeof(sip);
1388 arpptr += dev->addr_len; /* tha */
1389 memcpy(&tip, arpptr, sizeof(tip));
1390
1391 if (ipv4_is_loopback(tip) ||
1392 ipv4_is_multicast(tip))
1393 goto out;
1394
1395 n = neigh_lookup(&arp_tbl, &tip, dev);
1396
1397 if (n) {
e4f67add
DS		 1398 struct vxlan_fdb *f;
1399 struct sk_buff *reply;
1400
1401 if (!(n->nud_state & NUD_CONNECTED)) {
1402 neigh_release(n);
1403 goto out;
1404 }
1405
1406 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1407 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1408 /* bridge-local neighbor */
1409 neigh_release(n);
1410 goto out;
1411 }
1412
1413 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1414 n->ha, sha);
1415
1416 neigh_release(n);
1417
7346135d
DS		 1418 if (reply == NULL)
1419 goto out;
1420
e4f67add
DS		 1421 skb_reset_mac_header(reply);
1422 __skb_pull(reply, skb_network_offset(reply));
1423 reply->ip_summed = CHECKSUM_UNNECESSARY;
1424 reply->pkt_type = PACKET_HOST;
1425
1426 if (netif_rx_ni(reply) == NET_RX_DROP)
1427 dev->stats.rx_dropped++;
e4c7ed41
CW		 1428 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1429 union vxlan_addr ipa = {
1430 .sin.sin_addr.s_addr = tip,
a45e92a5 1431 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1432 };
1433
1434 vxlan_ip_miss(dev, &ipa);
1435 }
e4f67add
DS		 1436out:
1437 consume_skb(skb);
1438 return NETDEV_TX_OK;
1439}
1440
f564f45c 1441#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1442static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1443 struct neighbour *n, bool isrouter)
1444{
1445 struct net_device *dev = request->dev;
1446 struct sk_buff *reply;
1447 struct nd_msg *ns, *na;
1448 struct ipv6hdr *pip6;
1449 u8 *daddr;
1450 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1451 int ns_olen;
1452 int i, len;
1453
1454 if (dev == NULL)
1455 return NULL;
1456
1457 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1458 sizeof(*na) + na_olen + dev->needed_tailroom;
1459 reply = alloc_skb(len, GFP_ATOMIC);
1460 if (reply == NULL)
1461 return NULL;
1462
1463 reply->protocol = htons(ETH_P_IPV6);
1464 reply->dev = dev;
1465 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1466 skb_push(reply, sizeof(struct ethhdr));
1467 skb_set_mac_header(reply, 0);
1468
1469 ns = (struct nd_msg *)skb_transport_header(request);
1470
1471 daddr = eth_hdr(request)->h_source;
1472 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1473 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1474 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1475 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1476 break;
1477 }
1478 }
1479
1480 /* Ethernet header */
1481 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1482 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1483 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1484 reply->protocol = htons(ETH_P_IPV6);
1485
1486 skb_pull(reply, sizeof(struct ethhdr));
1487 skb_set_network_header(reply, 0);
1488 skb_put(reply, sizeof(struct ipv6hdr));
1489
1490 /* IPv6 header */
1491
1492 pip6 = ipv6_hdr(reply);
1493 memset(pip6, 0, sizeof(struct ipv6hdr));
1494 pip6->version = 6;
1495 pip6->priority = ipv6_hdr(request)->priority;
1496 pip6->nexthdr = IPPROTO_ICMPV6;
1497 pip6->hop_limit = 255;
1498 pip6->daddr = ipv6_hdr(request)->saddr;
1499 pip6->saddr = *(struct in6_addr *)n->primary_key;
1500
1501 skb_pull(reply, sizeof(struct ipv6hdr));
1502 skb_set_transport_header(reply, 0);
1503
1504 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1505
1506 /* Neighbor Advertisement */
1507 memset(na, 0, sizeof(*na)+na_olen);
1508 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1509 na->icmph.icmp6_router = isrouter;
1510 na->icmph.icmp6_override = 1;
1511 na->icmph.icmp6_solicited = 1;
1512 na->target = ns->target;
1513 ether_addr_copy(&na->opt[2], n->ha);
1514 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1515 na->opt[1] = na_olen >> 3;
1516
1517 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1518 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1519 csum_partial(na, sizeof(*na)+na_olen, 0));
1520
1521 pip6->payload_len = htons(sizeof(*na)+na_olen);
1522
1523 skb_push(reply, sizeof(struct ipv6hdr));
1524
1525 reply->ip_summed = CHECKSUM_UNNECESSARY;
1526
1527 return reply;
1528}
1529
f564f45c
CW		 1530static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1531{
1532 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1533 struct nd_msg *msg;
f564f45c
CW		 1534 const struct ipv6hdr *iphdr;
1535 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1536 struct neighbour *n;
1537 struct inet6_dev *in6_dev;
f564f45c
CW		 1538
1539 in6_dev = __in6_dev_get(dev);
1540 if (!in6_dev)
1541 goto out;
1542
f564f45c
CW		 1543 iphdr = ipv6_hdr(skb);
1544 saddr = &iphdr->saddr;
1545 daddr = &iphdr->daddr;
1546
f564f45c
CW		 1547 msg = (struct nd_msg *)skb_transport_header(skb);
1548 if (msg->icmph.icmp6_code != 0 ||
1549 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1550 goto out;
1551
4b29dba9
DS		 1552 if (ipv6_addr_loopback(daddr) ||
1553 ipv6_addr_is_multicast(&msg->target))
1554 goto out;
1555
1556 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1557
1558 if (n) {
1559 struct vxlan_fdb *f;
4b29dba9 1560 struct sk_buff *reply;
f564f45c
CW		 1561
1562 if (!(n->nud_state & NUD_CONNECTED)) {
1563 neigh_release(n);
1564 goto out;
1565 }
1566
1567 f = vxlan_find_mac(vxlan, n->ha);
1568 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1569 /* bridge-local neighbor */
1570 neigh_release(n);
1571 goto out;
1572 }
1573
4b29dba9
DS		 1574 reply = vxlan_na_create(skb, n,
1575 !!(f ? f->flags & NTF_ROUTER : 0));
1576
f564f45c 1577 neigh_release(n);
4b29dba9
DS		 1578
1579 if (reply == NULL)
1580 goto out;
1581
1582 if (netif_rx_ni(reply) == NET_RX_DROP)
1583 dev->stats.rx_dropped++;
1584
f564f45c 1585 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1586 union vxlan_addr ipa = {
1587 .sin6.sin6_addr = msg->target,
a45e92a5 1588 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1589 };
1590
f564f45c
CW		 1591 vxlan_ip_miss(dev, &ipa);
1592 }
1593
1594out:
1595 consume_skb(skb);
1596 return NETDEV_TX_OK;
1597}
1598#endif
1599
e4f67add
DS		 1600static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1601{
1602 struct vxlan_dev *vxlan = netdev_priv(dev);
1603 struct neighbour *n;
e4f67add
DS		 1604
1605 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1606 return false;
1607
1608 n = NULL;
1609 switch (ntohs(eth_hdr(skb)->h_proto)) {
1610 case ETH_P_IP:
e15a00aa
CW		 1611 {
1612 struct iphdr *pip;
1613
e4f67add
DS		 1614 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1615 return false;
1616 pip = ip_hdr(skb);
1617 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1618 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1619 union vxlan_addr ipa = {
1620 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1621 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1622 };
1623
1624 vxlan_ip_miss(dev, &ipa);
1625 return false;
1626 }
1627
e4f67add 1628 break;
e15a00aa
CW		 1629 }
1630#if IS_ENABLED(CONFIG_IPV6)
1631 case ETH_P_IPV6:
1632 {
1633 struct ipv6hdr *pip6;
1634
1635 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1636 return false;
1637 pip6 = ipv6_hdr(skb);
1638 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1639 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1640 union vxlan_addr ipa = {
1641 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1642 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1643 };
1644
1645 vxlan_ip_miss(dev, &ipa);
1646 return false;
1647 }
1648
1649 break;
1650 }
1651#endif
e4f67add
DS		 1652 default:
1653 return false;
1654 }
1655
1656 if (n) {
1657 bool diff;
1658
7367d0b5 1659 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1660 if (diff) {
1661 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1662 dev->addr_len);
1663 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1664 }
1665 neigh_release(n);
1666 return diff;
e4c7ed41
CW		 1667 }
1668
e4f67add
DS		 1669 return false;
1670}
1671
3511494c
TG		 1672static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, struct vxlan_sock *vs,
1673 struct vxlan_metadata *md)
1674{
1675 struct vxlanhdr_gbp *gbp;
1676
1677 gbp = (struct vxlanhdr_gbp *)vxh;
1678 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1679
1680 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1681 gbp->dont_learn = 1;
1682
1683 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1684 gbp->policy_applied = 1;
1685
1686 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1687}
1688
e4c7ed41 1689#if IS_ENABLED(CONFIG_IPV6)
11796187 1690static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1691 struct dst_entry *dst, struct sk_buff *skb,
1692 struct net_device *dev, struct in6_addr *saddr,
1693 struct in6_addr *daddr, __u8 prio, __u8 ttl,
3511494c
TG		 1694 __be16 src_port, __be16 dst_port,
1695 struct vxlan_metadata *md, bool xnet)
e4c7ed41 1696{
e4c7ed41 1697 struct vxlanhdr *vxh;
e4c7ed41
CW		 1698 int min_headroom;
1699 int err;
acbf74a7 1700 bool udp_sum = !udp_get_no_check6_tx(vs->sock->sk);
dfd8645e
TH		 1701 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1702 u16 hdrlen = sizeof(struct vxlanhdr);
1703
1704 if ((vs->flags & VXLAN_F_REMCSUM_TX) &&
1705 skb->ip_summed == CHECKSUM_PARTIAL) {
1706 int csum_start = skb_checksum_start_offset(skb);
1707
1708 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1709 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1710 (skb->csum_offset == offsetof(struct udphdr, check) ||
1711 skb->csum_offset == offsetof(struct tcphdr, check))) {
1712 udp_sum = false;
1713 type |= SKB_GSO_TUNNEL_REMCSUM;
1714 }
1715 }
e4c7ed41 1716
dfd8645e 1717 skb = iptunnel_handle_offloads(skb, udp_sum, type);
74f47278
PS		 1718 if (IS_ERR(skb)) {
1719 err = -EINVAL;
1720 goto err;
1721 }
e4c7ed41 1722
f01ec1c0 1723 skb_scrub_packet(skb, xnet);
963a88b3 1724
e4c7ed41
CW		 1725 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1726 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1727 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1728
1729 /* Need space for new headers (invalidates iph ptr) */
1730 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1731 if (unlikely(err)) {
1732 kfree_skb(skb);
1733 goto err;
1734 }
e4c7ed41 1735
5968250c 1736 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1737 if (WARN_ON(!skb)) {
1738 err = -ENOMEM;
1739 goto err;
1740 }
e4c7ed41
CW		 1741
1742 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1743 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1744 vxh->vx_vni = md->vni;
e4c7ed41 1745
dfd8645e
TH		 1746 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1747 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1748 VXLAN_RCO_SHIFT;
1749
1750 if (skb->csum_offset == offsetof(struct udphdr, check))
1751 data |= VXLAN_RCO_UDP;
1752
1753 vxh->vx_vni |= htonl(data);
1754 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1755
1756 if (!skb_is_gso(skb)) {
1757 skb->ip_summed = CHECKSUM_NONE;
1758 skb->encapsulation = 0;
1759 }
1760 }
1761
3511494c
TG		 1762 if (vs->flags & VXLAN_F_GBP)
1763 vxlan_build_gbp_hdr(vxh, vs, md);
1764
996c9fd1
TH		 1765 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1766
acbf74a7
AZ		 1767 udp_tunnel6_xmit_skb(vs->sock, dst, skb, dev, saddr, daddr, prio,
1768 ttl, src_port, dst_port);
e4c7ed41 1769 return 0;
74f47278
PS		 1770err:
1771 dst_release(dst);
1772 return err;
e4c7ed41
CW		 1773}
1774#endif
1775
11796187 1776int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1777 struct rtable *rt, struct sk_buff *skb,
1778 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
3511494c
TG		 1779 __be16 src_port, __be16 dst_port,
1780 struct vxlan_metadata *md, bool xnet)
49560532
PS		 1781{
1782 struct vxlanhdr *vxh;
649c5b8b 1783 int min_headroom;
49560532 1784 int err;
acbf74a7 1785 bool udp_sum = !vs->sock->sk->sk_no_check_tx;
dfd8645e
TH		 1786 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1787 u16 hdrlen = sizeof(struct vxlanhdr);
1788
1789 if ((vs->flags & VXLAN_F_REMCSUM_TX) &&
1790 skb->ip_summed == CHECKSUM_PARTIAL) {
1791 int csum_start = skb_checksum_start_offset(skb);
1792
1793 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1794 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1795 (skb->csum_offset == offsetof(struct udphdr, check) ||
1796 skb->csum_offset == offsetof(struct tcphdr, check))) {
1797 udp_sum = false;
1798 type |= SKB_GSO_TUNNEL_REMCSUM;
1799 }
1800 }
49560532 1801
dfd8645e 1802 skb = iptunnel_handle_offloads(skb, udp_sum, type);
359a0ea9 1803 if (IS_ERR(skb))
74f47278 1804 return PTR_ERR(skb);
49560532 1805
649c5b8b 1806 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1807 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1808 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1809
1810 /* Need space for new headers (invalidates iph ptr) */
1811 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1812 if (unlikely(err)) {
1813 kfree_skb(skb);
649c5b8b 1814 return err;
74f47278 1815 }
649c5b8b 1816
5968250c
JP		 1817 skb = vlan_hwaccel_push_inside(skb);
1818 if (WARN_ON(!skb))
1819 return -ENOMEM;
1eaa8178 1820
49560532 1821 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1822 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1823 vxh->vx_vni = md->vni;
49560532 1824
dfd8645e
TH		 1825 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1826 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1827 VXLAN_RCO_SHIFT;
1828
1829 if (skb->csum_offset == offsetof(struct udphdr, check))
1830 data |= VXLAN_RCO_UDP;
1831
1832 vxh->vx_vni |= htonl(data);
1833 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1834
1835 if (!skb_is_gso(skb)) {
1836 skb->ip_summed = CHECKSUM_NONE;
1837 skb->encapsulation = 0;
1838 }
1839 }
1840
3511494c
TG		 1841 if (vs->flags & VXLAN_F_GBP)
1842 vxlan_build_gbp_hdr(vxh, vs, md);
1843
996c9fd1
TH		 1844 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1845
acbf74a7
AZ		 1846 return udp_tunnel_xmit_skb(vs->sock, rt, skb, src, dst, tos,
1847 ttl, df, src_port, dst_port, xnet);
49560532
PS		 1848}
1849EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1850
9dcc71e1
SS		 1851/* Bypass encapsulation if the destination is local */
1852static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1853 struct vxlan_dev *dst_vxlan)
1854{
8f84985f 1855 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1856 union vxlan_addr loopback;
1857 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1858 struct net_device *dev = skb->dev;
1859 int len = skb->len;
9dcc71e1 1860
8f84985f
LR		 1861 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1862 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1863 skb->pkt_type = PACKET_HOST;
1864 skb->encapsulation = 0;
1865 skb->dev = dst_vxlan->dev;
1866 __skb_pull(skb, skb_network_offset(skb));
1867
e4c7ed41
CW		 1868 if (remote_ip->sa.sa_family == AF_INET) {
1869 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1870 loopback.sa.sa_family = AF_INET;
1871#if IS_ENABLED(CONFIG_IPV6)
1872 } else {
1873 loopback.sin6.sin6_addr = in6addr_loopback;
1874 loopback.sa.sa_family = AF_INET6;
1875#endif
1876 }
1877
9dcc71e1 1878 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1879 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1880
1881 u64_stats_update_begin(&tx_stats->syncp);
1882 tx_stats->tx_packets++;
ce6502a8 1883 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1884 u64_stats_update_end(&tx_stats->syncp);
1885
1886 if (netif_rx(skb) == NET_RX_SUCCESS) {
1887 u64_stats_update_begin(&rx_stats->syncp);
1888 rx_stats->rx_packets++;
ce6502a8 1889 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1890 u64_stats_update_end(&rx_stats->syncp);
1891 } else {
ce6502a8 1892 dev->stats.rx_dropped++;
9dcc71e1
SS		 1893 }
1894}
1895
4ad16930
SH		 1896static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1897 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1898{
1899 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1900 struct rtable *rt = NULL;
d342894c 1901 const struct iphdr *old_iph;
d342894c 1902 struct flowi4 fl4;
e4c7ed41 1903 union vxlan_addr *dst;
3511494c 1904 struct vxlan_metadata md;
e4c7ed41 1905 __be16 src_port = 0, dst_port;
234f5b73 1906 u32 vni;
d342894c 1907 __be16 df = 0;
1908 __u8 tos, ttl;
0e6fbc5b 1909 int err;
d342894c 1910
823aa873 1911 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1912 vni = rdst->remote_vni;
e4c7ed41 1913 dst = &rdst->remote_ip;
e4f67add 1914
e4c7ed41 1915 if (vxlan_addr_any(dst)) {
e4f67add 1916 if (did_rsc) {
e4f67add 1917 /* short-circuited back to local bridge */
9dcc71e1 1918 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1919 return;
e4f67add 1920 }
ef59febe 1921 goto drop;
e4f67add 1922 }
ef59febe 1923
d342894c 1924 old_iph = ip_hdr(skb);
1925
d342894c 1926 ttl = vxlan->ttl;
e4c7ed41 1927 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1928 ttl = 1;
1929
1930 tos = vxlan->tos;
1931 if (tos == 1)
206aaafc 1932 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1933
535fb8d0
TH		 1934 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1935 vxlan->port_max, true);
d342894c 1936
e4c7ed41
CW		 1937 if (dst->sa.sa_family == AF_INET) {
1938 memset(&fl4, 0, sizeof(fl4));
1939 fl4.flowi4_oif = rdst->remote_ifindex;
1940 fl4.flowi4_tos = RT_TOS(tos);
1941 fl4.daddr = dst->sin.sin_addr.s_addr;
1942 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1943
f01ec1c0 1944 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1945 if (IS_ERR(rt)) {
1946 netdev_dbg(dev, "no route to %pI4\n",
1947 &dst->sin.sin_addr.s_addr);
1948 dev->stats.tx_carrier_errors++;
1949 goto tx_error;
1950 }
d342894c 1951
e4c7ed41
CW		 1952 if (rt->dst.dev == dev) {
1953 netdev_dbg(dev, "circular route to %pI4\n",
1954 &dst->sin.sin_addr.s_addr);
1955 dev->stats.collisions++;
fffc15a5 1956 goto rt_tx_error;
e4c7ed41
CW		 1957 }
1958
1959 /* Bypass encapsulation if the destination is local */
1960 if (rt->rt_flags & RTCF_LOCAL &&
1961 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1962 struct vxlan_dev *dst_vxlan;
1963
1964 ip_rt_put(rt);
19ca9fc1 1965 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1966 dst->sa.sa_family, dst_port,
1967 vxlan->flags);
e4c7ed41
CW		 1968 if (!dst_vxlan)
1969 goto tx_error;
1970 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1971 return;
1972 }
1973
1974 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1975 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
3511494c
TG		 1976 md.vni = htonl(vni << 8);
1977 md.gbp = skb->mark;
d342894c 1978
3c4d1dae
AZ		 1979 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1980 fl4.saddr, dst->sin.sin_addr.s_addr,
3511494c 1981 tos, ttl, df, src_port, dst_port, &md,
3c4d1dae 1982 !net_eq(vxlan->net, dev_net(vxlan->dev)));
74f47278
PS		 1983 if (err < 0) {
1984 /* skb is already freed. */
1985 skb = NULL;
e4c7ed41 1986 goto rt_tx_error;
74f47278
PS		 1987 }
1988
e4c7ed41
CW		 1989 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1990#if IS_ENABLED(CONFIG_IPV6)
1991 } else {
1992 struct sock *sk = vxlan->vn_sock->sock->sk;
1993 struct dst_entry *ndst;
1994 struct flowi6 fl6;
1995 u32 flags;
1996
1997 memset(&fl6, 0, sizeof(fl6));
1998 fl6.flowi6_oif = rdst->remote_ifindex;
1999 fl6.daddr = dst->sin6.sin6_addr;
2000 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 2001 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 2002
2003 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
2004 netdev_dbg(dev, "no route to %pI6\n",
2005 &dst->sin6.sin6_addr);
2006 dev->stats.tx_carrier_errors++;
9dcc71e1 2007 goto tx_error;
e4c7ed41 2008 }
d342894c 2009
e4c7ed41
CW		 2010 if (ndst->dev == dev) {
2011 netdev_dbg(dev, "circular route to %pI6\n",
2012 &dst->sin6.sin6_addr);
2013 dst_release(ndst);
2014 dev->stats.collisions++;
2015 goto tx_error;
2016 }
0e6fbc5b 2017
e4c7ed41
CW		 2018 /* Bypass encapsulation if the destination is local */
2019 flags = ((struct rt6_info *)ndst)->rt6i_flags;
2020 if (flags & RTF_LOCAL &&
2021 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2022 struct vxlan_dev *dst_vxlan;
2023
2024 dst_release(ndst);
19ca9fc1 2025 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2026 dst->sa.sa_family, dst_port,
2027 vxlan->flags);
e4c7ed41
CW		 2028 if (!dst_vxlan)
2029 goto tx_error;
2030 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2031 return;
2032 }
49560532 2033
e4c7ed41 2034 ttl = ttl ? : ip6_dst_hoplimit(ndst);
3511494c
TG		 2035 md.vni = htonl(vni << 8);
2036 md.gbp = skb->mark;
e4c7ed41 2037
11796187 2038 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41 2039 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
3511494c 2040 src_port, dst_port, &md,
f01ec1c0 2041 !net_eq(vxlan->net, dev_net(vxlan->dev)));
e4c7ed41
CW		 2042#endif
2043 }
0e6fbc5b 2044
4ad16930 2045 return;
d342894c 2046
2047drop:
2048 dev->stats.tx_dropped++;
2049 goto tx_free;
2050
49560532
PS		 2051rt_tx_error:
2052 ip_rt_put(rt);
d342894c 2053tx_error:
2054 dev->stats.tx_errors++;
2055tx_free:
2056 dev_kfree_skb(skb);
d342894c 2057}
2058
6681712d
DS		 2059/* Transmit local packets over Vxlan
2060 *
2061 * Outer IP header inherits ECN and DF from inner header.
2062 * Outer UDP destination is the VXLAN assigned port.
2063 * source port is based on hash of flow
2064 */
2065static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2066{
2067 struct vxlan_dev *vxlan = netdev_priv(dev);
2068 struct ethhdr *eth;
2069 bool did_rsc = false;
8f646c92 2070 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2071 struct vxlan_fdb *f;
6681712d
DS		 2072
2073 skb_reset_mac_header(skb);
2074 eth = eth_hdr(skb);
2075
f564f45c
CW		 2076 if ((vxlan->flags & VXLAN_F_PROXY)) {
2077 if (ntohs(eth->h_proto) == ETH_P_ARP)
2078 return arp_reduce(dev, skb);
2079#if IS_ENABLED(CONFIG_IPV6)
2080 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2081 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2082 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2083 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2084 struct nd_msg *msg;
2085
2086 msg = (struct nd_msg *)skb_transport_header(skb);
2087 if (msg->icmph.icmp6_code == 0 &&
2088 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2089 return neigh_reduce(dev, skb);
2090 }
7a9f526f 2091 eth = eth_hdr(skb);
f564f45c
CW		 2092#endif
2093 }
6681712d
DS		 2094
2095 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2096 did_rsc = false;
2097
2098 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2099 (ntohs(eth->h_proto) == ETH_P_IP ||
2100 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2101 did_rsc = route_shortcircuit(dev, skb);
2102 if (did_rsc)
2103 f = vxlan_find_mac(vxlan, eth->h_dest);
2104 }
2105
6681712d 2106 if (f == NULL) {
afbd8bae
MR		 2107 f = vxlan_find_mac(vxlan, all_zeros_mac);
2108 if (f == NULL) {
2109 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2110 !is_multicast_ether_addr(eth->h_dest))
2111 vxlan_fdb_miss(vxlan, eth->h_dest);
2112
2113 dev->stats.tx_dropped++;
8f646c92 2114 kfree_skb(skb);
afbd8bae
MR		 2115 return NETDEV_TX_OK;
2116 }
2117 }
6681712d 2118
afbd8bae
MR		 2119 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2120 struct sk_buff *skb1;
6681712d 2121
8f646c92
ED		 2122 if (!fdst) {
2123 fdst = rdst;
2124 continue;
2125 }
afbd8bae
MR		 2126 skb1 = skb_clone(skb, GFP_ATOMIC);
2127 if (skb1)
2128 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2129 }
2130
8f646c92
ED		 2131 if (fdst)
2132 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2133 else
2134 kfree_skb(skb);
4ad16930 2135 return NETDEV_TX_OK;
6681712d
DS		 2136}
2137
d342894c 2138/* Walk the forwarding table and purge stale entries */
2139static void vxlan_cleanup(unsigned long arg)
2140{
2141 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2142 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2143 unsigned int h;
2144
2145 if (!netif_running(vxlan->dev))
2146 return;
2147
2148 spin_lock_bh(&vxlan->hash_lock);
2149 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2150 struct hlist_node *p, *n;
2151 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2152 struct vxlan_fdb *f
2153 = container_of(p, struct vxlan_fdb, hlist);
2154 unsigned long timeout;
2155
3c172868 2156 if (f->state & NUD_PERMANENT)
d342894c 2157 continue;
2158
2159 timeout = f->used + vxlan->age_interval * HZ;
2160 if (time_before_eq(timeout, jiffies)) {
2161 netdev_dbg(vxlan->dev,
2162 "garbage collect %pM\n",
2163 f->eth_addr);
2164 f->state = NUD_STALE;
2165 vxlan_fdb_destroy(vxlan, f);
2166 } else if (time_before(timeout, next_timer))
2167 next_timer = timeout;
2168 }
2169 }
2170 spin_unlock_bh(&vxlan->hash_lock);
2171
2172 mod_timer(&vxlan->age_timer, next_timer);
2173}
2174
9c2e24e1
PS		 2175static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2176{
2177 __u32 vni = vxlan->default_dst.remote_vni;
2178
2179 vxlan->vn_sock = vs;
2180 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2181}
2182
d342894c 2183/* Setup stats when device is created */
2184static int vxlan_init(struct net_device *dev)
2185{
1c51a915 2186 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2187 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2188 struct vxlan_sock *vs;
19ca9fc1 2189 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
1c51a915 2190
1c213bd2 2191 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2192 if (!dev->tstats)
d342894c 2193 return -ENOMEM;
2194
1c51a915 2195 spin_lock(&vn->sock_lock);
19ca9fc1 2196 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
ac5132d1 2197 vxlan->dst_port, vxlan->flags);
00c83b01 2198 if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
1c51a915 2199 /* If we have a socket with same port already, reuse it */
9c2e24e1 2200 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 2201 } else {
2202 /* otherwise make new socket outside of RTNL */
2203 dev_hold(dev);
2204 queue_work(vxlan_wq, &vxlan->sock_work);
2205 }
2206 spin_unlock(&vn->sock_lock);
2207
d342894c 2208 return 0;
2209}
2210
ba609e9b 2211static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2212{
2213 struct vxlan_fdb *f;
2214
2215 spin_lock_bh(&vxlan->hash_lock);
2216 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2217 if (f)
2218 vxlan_fdb_destroy(vxlan, f);
2219 spin_unlock_bh(&vxlan->hash_lock);
2220}
2221
ebf4063e
SH		 2222static void vxlan_uninit(struct net_device *dev)
2223{
2224 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2225 struct vxlan_sock *vs = vxlan->vn_sock;
2226
ba609e9b 2227 vxlan_fdb_delete_default(vxlan);
afbd8bae 2228
ebf4063e 2229 if (vs)
012a5729 2230 vxlan_sock_release(vs);
ebf4063e
SH		 2231 free_percpu(dev->tstats);
2232}
2233
d342894c 2234/* Start ageing timer and join group when device is brought up */
2235static int vxlan_open(struct net_device *dev)
2236{
2237 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2238 struct vxlan_sock *vs = vxlan->vn_sock;
2239
2240 /* socket hasn't been created */
2241 if (!vs)
2242 return -ENOTCONN;
d342894c 2243
79d4a94f 2244 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2245 vxlan_sock_hold(vs);
7c47cedf 2246 dev_hold(dev);
3fc2de2f 2247 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2248 }
2249
2250 if (vxlan->age_interval)
2251 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2252
2253 return 0;
2254}
2255
2256/* Purge the forwarding table */
2257static void vxlan_flush(struct vxlan_dev *vxlan)
2258{
31fec5aa 2259 unsigned int h;
d342894c 2260
2261 spin_lock_bh(&vxlan->hash_lock);
2262 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2263 struct hlist_node *p, *n;
2264 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2265 struct vxlan_fdb *f
2266 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2267 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2268 if (!is_zero_ether_addr(f->eth_addr))
2269 vxlan_fdb_destroy(vxlan, f);
d342894c 2270 }
2271 }
2272 spin_unlock_bh(&vxlan->hash_lock);
2273}
2274
2275/* Cleanup timer and forwarding table on shutdown */
2276static int vxlan_stop(struct net_device *dev)
2277{
2278 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2279 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2280 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2281
e4c7ed41 2282 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2283 !vxlan_group_used(vn, vxlan)) {
1c51a915 2284 vxlan_sock_hold(vs);
7c47cedf 2285 dev_hold(dev);
3fc2de2f 2286 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2287 }
d342894c 2288
2289 del_timer_sync(&vxlan->age_timer);
2290
2291 vxlan_flush(vxlan);
2292
2293 return 0;
2294}
2295
d342894c 2296/* Stub, nothing needs to be done. */
2297static void vxlan_set_multicast_list(struct net_device *dev)
2298{
2299}
2300
345010b5
DB		 2301static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2302{
2303 struct vxlan_dev *vxlan = netdev_priv(dev);
2304 struct vxlan_rdst *dst = &vxlan->default_dst;
2305 struct net_device *lowerdev;
2306 int max_mtu;
2307
f01ec1c0 2308 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2309 if (lowerdev == NULL)
2310 return eth_change_mtu(dev, new_mtu);
2311
2312 if (dst->remote_ip.sa.sa_family == AF_INET6)
2313 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2314 else
2315 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2316
2317 if (new_mtu < 68 || new_mtu > max_mtu)
2318 return -EINVAL;
2319
2320 dev->mtu = new_mtu;
2321 return 0;
2322}
2323
d342894c 2324static const struct net_device_ops vxlan_netdev_ops = {
2325 .ndo_init = vxlan_init,
ebf4063e 2326 .ndo_uninit = vxlan_uninit,
d342894c 2327 .ndo_open = vxlan_open,
2328 .ndo_stop = vxlan_stop,
2329 .ndo_start_xmit = vxlan_xmit,
e8171045 2330 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2331 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2332 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2333 .ndo_validate_addr = eth_validate_addr,
2334 .ndo_set_mac_address = eth_mac_addr,
2335 .ndo_fdb_add = vxlan_fdb_add,
2336 .ndo_fdb_del = vxlan_fdb_delete,
2337 .ndo_fdb_dump = vxlan_fdb_dump,
2338};
2339
2340/* Info for udev, that this is a virtual tunnel endpoint */
2341static struct device_type vxlan_type = {
2342 .name = "vxlan",
2343};
2344
53cf5275 2345/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2346 * supply the listening VXLAN udp ports. Callers are expected
2347 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2348 */
2349void vxlan_get_rx_port(struct net_device *dev)
2350{
2351 struct vxlan_sock *vs;
2352 struct net *net = dev_net(dev);
2353 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2354 sa_family_t sa_family;
35e42379
JG		 2355 __be16 port;
2356 unsigned int i;
53cf5275
JG		 2357
2358 spin_lock(&vn->sock_lock);
2359 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2360 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2361 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2362 sa_family = vs->sock->sk->sk_family;
2363 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2364 port);
2365 }
2366 }
2367 spin_unlock(&vn->sock_lock);
2368}
2369EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2370
d342894c 2371/* Initialize the device structure. */
2372static void vxlan_setup(struct net_device *dev)
2373{
2374 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2375 unsigned int h;
d342894c 2376
2377 eth_hw_addr_random(dev);
2378 ether_setup(dev);
e4c7ed41 2379 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2380 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2381 else
2853af6a 2382 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2383
2384 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2385 dev->destructor = free_netdev;
d342894c 2386 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2387
2388 dev->tx_queue_len = 0;
2389 dev->features |= NETIF_F_LLTX;
d6727fe3 2390 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2391 dev->features |= NETIF_F_RXCSUM;
05c0db08 2392 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2393
1eaa8178
PS		 2394 dev->vlan_features = dev->features;
2395 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2396 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2397 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2398 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2399 netif_keep_dst(dev);
6602d007 2400 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2401
553675fb 2402 INIT_LIST_HEAD(&vxlan->next);
d342894c 2403 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2404 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2405 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2406 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2407
2408 init_timer_deferrable(&vxlan->age_timer);
2409 vxlan->age_timer.function = vxlan_cleanup;
2410 vxlan->age_timer.data = (unsigned long) vxlan;
2411
823aa873 2412 vxlan->dst_port = htons(vxlan_port);
05f47d69 2413
d342894c 2414 vxlan->dev = dev;
2415
2416 for (h = 0; h < FDB_HASH_SIZE; ++h)
2417 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2418}
2419
2420static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2421 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2422 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2423 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2424 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2425 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2426 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2427 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2428 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2429 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2430 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2431 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2432 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2433 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2434 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2435 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2436 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2437 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2438 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2439 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2440 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2441 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2442 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2443 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
d342894c 2444};
2445
2446static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2447{
2448 if (tb[IFLA_ADDRESS]) {
2449 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2450 pr_debug("invalid link address (not ethernet)\n");
2451 return -EINVAL;
2452 }
2453
2454 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2455 pr_debug("invalid all zero ethernet address\n");
2456 return -EADDRNOTAVAIL;
2457 }
2458 }
2459
2460 if (!data)
2461 return -EINVAL;
2462
2463 if (data[IFLA_VXLAN_ID]) {
2464 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2465 if (id >= VXLAN_VID_MASK)
2466 return -ERANGE;
2467 }
2468
05f47d69 2469 if (data[IFLA_VXLAN_PORT_RANGE]) {
2470 const struct ifla_vxlan_port_range *p
2471 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2472
2473 if (ntohs(p->high) < ntohs(p->low)) {
2474 pr_debug("port range %u .. %u not valid\n",
2475 ntohs(p->low), ntohs(p->high));
2476 return -EINVAL;
2477 }
2478 }
2479
d342894c 2480 return 0;
2481}
2482
1b13c97f
YB		 2483static void vxlan_get_drvinfo(struct net_device *netdev,
2484 struct ethtool_drvinfo *drvinfo)
2485{
2486 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2487 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2488}
2489
2490static const struct ethtool_ops vxlan_ethtool_ops = {
2491 .get_drvinfo = vxlan_get_drvinfo,
2492 .get_link = ethtool_op_get_link,
2493};
2494
553675fb 2495static void vxlan_del_work(struct work_struct *work)
2496{
2497 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2498 udp_tunnel_sock_release(vs->sock);
553675fb 2499 kfree_rcu(vs, rcu);
2500}
2501
3ee64f39
TH		 2502static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2503 __be16 port, u32 flags)
553675fb 2504{
e4c7ed41 2505 struct socket *sock;
3ee64f39
TH		 2506 struct udp_port_cfg udp_conf;
2507 int err;
e4c7ed41 2508
3ee64f39 2509 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2510
3ee64f39
TH		 2511 if (ipv6) {
2512 udp_conf.family = AF_INET6;
2513 udp_conf.use_udp6_tx_checksums =
3dc2b6a8 2514 !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
3ee64f39 2515 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2516 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2517 } else {
2518 udp_conf.family = AF_INET;
2519 udp_conf.local_ip.s_addr = INADDR_ANY;
2520 udp_conf.use_udp_checksums =
2521 !!(flags & VXLAN_F_UDP_CSUM);
e4c7ed41 2522 }
e4c7ed41 2523
3ee64f39 2524 udp_conf.local_udp_port = port;
553675fb 2525
3ee64f39
TH		 2526 /* Open UDP socket */
2527 err = udp_sock_create(net, &udp_conf, &sock);
2528 if (err < 0)
2529 return ERR_PTR(err);
e4c7ed41 2530
39deb2c7 2531 return sock;
e4c7ed41
CW		 2532}
2533
2534/* Create new listen socket if needed */
2535static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2536 vxlan_rcv_t *rcv, void *data,
2537 u32 flags)
e4c7ed41
CW		 2538{
2539 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2540 struct vxlan_sock *vs;
2541 struct socket *sock;
e4c7ed41 2542 unsigned int h;
359a0ea9 2543 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2544 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2545
dc01e7d3 2546 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2547 if (!vs)
2548 return ERR_PTR(-ENOMEM);
2549
2550 for (h = 0; h < VNI_HASH_SIZE; ++h)
2551 INIT_HLIST_HEAD(&vs->vni_list[h]);
2552
2553 INIT_WORK(&vs->del_work, vxlan_del_work);
2554
3ee64f39 2555 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2556 if (IS_ERR(sock)) {
553675fb 2557 kfree(vs);
e50fddc8 2558 return ERR_CAST(sock);
553675fb 2559 }
e4c7ed41
CW		 2560
2561 vs->sock = sock;
9c2e24e1 2562 atomic_set(&vs->refcnt, 1);
5cfccc5a 2563 vs->rcv = rcv;
012a5729 2564 vs->data = data;
dfd8645e 2565 vs->flags = flags;
553675fb 2566
dc01e7d3
OG		 2567 /* Initialize the vxlan udp offloads structure */
2568 vs->udp_offloads.port = port;
2569 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2570 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2571
9c2e24e1
PS		 2572 spin_lock(&vn->sock_lock);
2573 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2574 vxlan_notify_add_rx_port(vs);
9c2e24e1 2575 spin_unlock(&vn->sock_lock);
553675fb 2576
2577 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2578 tunnel_cfg.sk_user_data = vs;
2579 tunnel_cfg.encap_type = 1;
2580 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2581 tunnel_cfg.encap_destroy = NULL;
2582
2583 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2584
9c2e24e1
PS		 2585 return vs;
2586}
2587
012a5729
PS		 2588struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2589 vxlan_rcv_t *rcv, void *data,
359a0ea9 2590 bool no_share, u32 flags)
9c2e24e1
PS		 2591{
2592 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2593 struct vxlan_sock *vs;
19ca9fc1 2594 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2595
359a0ea9 2596 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2597 if (!IS_ERR(vs))
2598 return vs;
553675fb 2599
012a5729
PS		 2600 if (no_share) /* Return error if sharing is not allowed. */
2601 return vs;
2602
9c2e24e1 2603 spin_lock(&vn->sock_lock);
ac5132d1 2604 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port, flags);
00c83b01
ML		 2605 if (vs && ((vs->rcv != rcv) ||
2606 !atomic_add_unless(&vs->refcnt, 1, 0)))
5cfccc5a 2607 vs = ERR_PTR(-EBUSY);
5cfccc5a
PS		 2608 spin_unlock(&vn->sock_lock);
2609
2610 if (!vs)
9c2e24e1
PS		 2611 vs = ERR_PTR(-EINVAL);
2612
553675fb 2613 return vs;
2614}
012a5729 2615EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2616
1c51a915
SH		 2617/* Scheduled at device creation to bind to a socket */
2618static void vxlan_sock_work(struct work_struct *work)
2619{
9c2e24e1 2620 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2621 struct net *net = vxlan->net;
1c51a915 2622 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2623 __be16 port = vxlan->dst_port;
2624 struct vxlan_sock *nvs;
1c51a915 2625
359a0ea9 2626 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2627 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2628 if (!IS_ERR(nvs))
2629 vxlan_vs_add_dev(nvs, vxlan);
2630 spin_unlock(&vn->sock_lock);
2631
2632 dev_put(vxlan->dev);
1c51a915
SH		 2633}
2634
d342894c 2635static int vxlan_newlink(struct net *net, struct net_device *dev,
2636 struct nlattr *tb[], struct nlattr *data[])
2637{
553675fb 2638 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2639 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2640 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2641 __u32 vni;
2642 int err;
e4c7ed41 2643 bool use_ipv6 = false;
d342894c 2644
2645 if (!data[IFLA_VXLAN_ID])
2646 return -EINVAL;
2647
f01ec1c0
ND		 2648 vxlan->net = dev_net(dev);
2649
d342894c 2650 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2651 dst->remote_vni = vni;
d342894c 2652
5933a7bb
MR		 2653 /* Unless IPv6 is explicitly requested, assume IPv4 */
2654 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2655 if (data[IFLA_VXLAN_GROUP]) {
2656 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2657 } else if (data[IFLA_VXLAN_GROUP6]) {
2658 if (!IS_ENABLED(CONFIG_IPV6))
2659 return -EPFNOSUPPORT;
2660
2661 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2662 sizeof(struct in6_addr));
2663 dst->remote_ip.sa.sa_family = AF_INET6;
2664 use_ipv6 = true;
2665 }
d342894c 2666
e4c7ed41
CW		 2667 if (data[IFLA_VXLAN_LOCAL]) {
2668 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2669 vxlan->saddr.sa.sa_family = AF_INET;
2670 } else if (data[IFLA_VXLAN_LOCAL6]) {
2671 if (!IS_ENABLED(CONFIG_IPV6))
2672 return -EPFNOSUPPORT;
2673
2674 /* TODO: respect scope id */
2675 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2676 sizeof(struct in6_addr));
2677 vxlan->saddr.sa.sa_family = AF_INET6;
2678 use_ipv6 = true;
2679 }
d342894c 2680
34e02aa1 2681 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2682 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2683 struct net_device *lowerdev
c7995c43 2684 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2685
2686 if (!lowerdev) {
c7995c43 2687 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2688 return -ENODEV;
2689 }
d342894c 2690
e4c7ed41
CW		 2691#if IS_ENABLED(CONFIG_IPV6)
2692 if (use_ipv6) {
2693 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2694 if (idev && idev->cnf.disable_ipv6) {
2695 pr_info("IPv6 is disabled via sysctl\n");
2696 return -EPERM;
2697 }
2698 vxlan->flags |= VXLAN_F_IPV6;
2699 }
2700#endif
2701
34e02aa1 2702 if (!tb[IFLA_MTU])
e4c7ed41 2703 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2704
2853af6a 2705 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2706 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2707 } else if (use_ipv6)
2708 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2709
2710 if (data[IFLA_VXLAN_TOS])
2711 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2712
afb97186
VB		 2713 if (data[IFLA_VXLAN_TTL])
2714 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2715
d342894c 2716 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2717 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2718
2719 if (data[IFLA_VXLAN_AGEING])
2720 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2721 else
2722 vxlan->age_interval = FDB_AGE_DEFAULT;
2723
e4f67add
DS		 2724 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2725 vxlan->flags |= VXLAN_F_PROXY;
2726
2727 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2728 vxlan->flags |= VXLAN_F_RSC;
2729
2730 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2731 vxlan->flags |= VXLAN_F_L2MISS;
2732
2733 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2734 vxlan->flags |= VXLAN_F_L3MISS;
2735
d342894c 2736 if (data[IFLA_VXLAN_LIMIT])
2737 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2738
05f47d69 2739 if (data[IFLA_VXLAN_PORT_RANGE]) {
2740 const struct ifla_vxlan_port_range *p
2741 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2742 vxlan->port_min = ntohs(p->low);
2743 vxlan->port_max = ntohs(p->high);
2744 }
2745
823aa873 2746 if (data[IFLA_VXLAN_PORT])
2747 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2748
359a0ea9
TH		 2749 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2750 vxlan->flags |= VXLAN_F_UDP_CSUM;
2751
2752 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2753 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2754 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2755
2756 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2757 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2758 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2759
dfd8645e
TH		 2760 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2761 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2762 vxlan->flags |= VXLAN_F_REMCSUM_TX;
2763
2764 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2765 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2766 vxlan->flags |= VXLAN_F_REMCSUM_RX;
2767
3511494c
TG		 2768 if (data[IFLA_VXLAN_GBP])
2769 vxlan->flags |= VXLAN_F_GBP;
2770
19ca9fc1 2771 if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
ac5132d1 2772 vxlan->dst_port, vxlan->flags)) {
553675fb 2773 pr_info("duplicate VNI %u\n", vni);
2774 return -EEXIST;
2775 }
2776
7ad24ea4 2777 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2778
2936b6ab
SS		 2779 /* create an fdb entry for a valid default destination */
2780 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2781 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2782 &vxlan->default_dst.remote_ip,
2783 NUD_REACHABLE|NUD_PERMANENT,
2784 NLM_F_EXCL|NLM_F_CREATE,
2785 vxlan->dst_port,
2786 vxlan->default_dst.remote_vni,
2787 vxlan->default_dst.remote_ifindex,
2788 NTF_SELF);
2789 if (err)
2790 return err;
2791 }
d342894c 2792
afbd8bae
MR		 2793 err = register_netdevice(dev);
2794 if (err) {
ba609e9b 2795 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2796 return err;
2797 }
2798
553675fb 2799 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2800
2801 return 0;
d342894c 2802}
2803
2804static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2805{
2806 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2807 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2808
fe5c3561 2809 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2810 if (!hlist_unhashed(&vxlan->hlist))
2811 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2812 spin_unlock(&vn->sock_lock);
2813
553675fb 2814 list_del(&vxlan->next);
d342894c 2815 unregister_netdevice_queue(dev, head);
2816}
2817
2818static size_t vxlan_get_size(const struct net_device *dev)
2819{
2820
2821 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2822 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2823 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2824 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2825 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2826 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2827 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2828 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2829 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2830 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2831 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2832 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2833 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2834 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2835 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2836 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2837 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2838 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2839 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2840 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2841 0;
2842}
2843
2844static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2845{
2846 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2847 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2848 struct ifla_vxlan_port_range ports = {
2849 .low = htons(vxlan->port_min),
2850 .high = htons(vxlan->port_max),
2851 };
d342894c 2852
c7995c43 2853 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2854 goto nla_put_failure;
2855
e4c7ed41
CW		 2856 if (!vxlan_addr_any(&dst->remote_ip)) {
2857 if (dst->remote_ip.sa.sa_family == AF_INET) {
2858 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2859 dst->remote_ip.sin.sin_addr.s_addr))
2860 goto nla_put_failure;
2861#if IS_ENABLED(CONFIG_IPV6)
2862 } else {
2863 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2864 &dst->remote_ip.sin6.sin6_addr))
2865 goto nla_put_failure;
2866#endif
2867 }
2868 }
d342894c 2869
c7995c43 2870 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2871 goto nla_put_failure;
2872
e4c7ed41
CW		 2873 if (!vxlan_addr_any(&vxlan->saddr)) {
2874 if (vxlan->saddr.sa.sa_family == AF_INET) {
2875 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2876 vxlan->saddr.sin.sin_addr.s_addr))
2877 goto nla_put_failure;
2878#if IS_ENABLED(CONFIG_IPV6)
2879 } else {
2880 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2881 &vxlan->saddr.sin6.sin6_addr))
2882 goto nla_put_failure;
2883#endif
2884 }
2885 }
d342894c 2886
2887 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2888 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2889 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2890 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2891 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2892 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2893 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2894 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2895 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2896 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2897 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2898 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2899 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2900 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2901 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2902 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2903 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2904 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2905 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 2906 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2907 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2908 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2909 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2910 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 2911 goto nla_put_failure;
2912
05f47d69 2913 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2914 goto nla_put_failure;
2915
3511494c
TG		 2916 if (vxlan->flags & VXLAN_F_GBP &&
2917 nla_put_flag(skb, IFLA_VXLAN_GBP))
2918 goto nla_put_failure;
2919
d342894c 2920 return 0;
2921
2922nla_put_failure:
2923 return -EMSGSIZE;
2924}
2925
1728d4fa
ND		 2926static struct net *vxlan_get_link_net(const struct net_device *dev)
2927{
2928 struct vxlan_dev *vxlan = netdev_priv(dev);
2929
2930 return vxlan->net;
2931}
2932
d342894c 2933static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2934 .kind = "vxlan",
2935 .maxtype = IFLA_VXLAN_MAX,
2936 .policy = vxlan_policy,
2937 .priv_size = sizeof(struct vxlan_dev),
2938 .setup = vxlan_setup,
2939 .validate = vxlan_validate,
2940 .newlink = vxlan_newlink,
2941 .dellink = vxlan_dellink,
2942 .get_size = vxlan_get_size,
2943 .fill_info = vxlan_fill_info,
1728d4fa 2944 .get_link_net = vxlan_get_link_net,
d342894c 2945};
2946
acaf4e70
DB		 2947static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2948 struct net_device *dev)
2949{
2950 struct vxlan_dev *vxlan, *next;
2951 LIST_HEAD(list_kill);
2952
2953 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2954 struct vxlan_rdst *dst = &vxlan->default_dst;
2955
2956 /* In case we created vxlan device with carrier
2957 * and we loose the carrier due to module unload
2958 * we also need to remove vxlan device. In other
2959 * cases, it's not necessary and remote_ifindex
2960 * is 0 here, so no matches.
2961 */
2962 if (dst->remote_ifindex == dev->ifindex)
2963 vxlan_dellink(vxlan->dev, &list_kill);
2964 }
2965
2966 unregister_netdevice_many(&list_kill);
2967}
2968
2969static int vxlan_lowerdev_event(struct notifier_block *unused,
2970 unsigned long event, void *ptr)
2971{
2972 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2973 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2974
783c1463 2975 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2976 vxlan_handle_lowerdev_unregister(vn, dev);
2977
2978 return NOTIFY_DONE;
2979}
2980
2981static struct notifier_block vxlan_notifier_block __read_mostly = {
2982 .notifier_call = vxlan_lowerdev_event,
2983};
2984
d342894c 2985static __net_init int vxlan_init_net(struct net *net)
2986{
2987 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2988 unsigned int h;
d342894c 2989
553675fb 2990 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2991 spin_lock_init(&vn->sock_lock);
d342894c 2992
553675fb 2993 for (h = 0; h < PORT_HASH_SIZE; ++h)
2994 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2995
2996 return 0;
2997}
2998
f01ec1c0
ND		 2999static void __net_exit vxlan_exit_net(struct net *net)
3000{
3001 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3002 struct vxlan_dev *vxlan, *next;
3003 struct net_device *dev, *aux;
3004 LIST_HEAD(list);
3005
3006 rtnl_lock();
3007 for_each_netdev_safe(net, dev, aux)
3008 if (dev->rtnl_link_ops == &vxlan_link_ops)
3009 unregister_netdevice_queue(dev, &list);
3010
3011 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3012 /* If vxlan->dev is in the same netns, it has already been added
3013 * to the list by the previous loop.
3014 */
3015 if (!net_eq(dev_net(vxlan->dev), net))
3016 unregister_netdevice_queue(dev, &list);
3017 }
3018
3019 unregister_netdevice_many(&list);
3020 rtnl_unlock();
3021}
3022
d342894c 3023static struct pernet_operations vxlan_net_ops = {
3024 .init = vxlan_init_net,
f01ec1c0 3025 .exit = vxlan_exit_net,
d342894c 3026 .id = &vxlan_net_id,
3027 .size = sizeof(struct vxlan_net),
3028};
3029
3030static int __init vxlan_init_module(void)
3031{
3032 int rc;
3033
758c57d1
SH		 3034 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3035 if (!vxlan_wq)
3036 return -ENOMEM;
3037
d342894c 3038 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3039
783c1463 3040 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3041 if (rc)
3042 goto out1;
3043
acaf4e70 3044 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3045 if (rc)
3046 goto out2;
3047
acaf4e70
DB		 3048 rc = rtnl_link_register(&vxlan_link_ops);
3049 if (rc)
3050 goto out3;
d342894c 3051
acaf4e70
DB		 3052 return 0;
3053out3:
3054 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3055out2:
783c1463 3056 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3057out1:
758c57d1 3058 destroy_workqueue(vxlan_wq);
d342894c 3059 return rc;
3060}
7332a13b 3061late_initcall(vxlan_init_module);
d342894c 3062
3063static void __exit vxlan_cleanup_module(void)
3064{
b7153984 3065 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3066 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3067 destroy_workqueue(vxlan_wq);
783c1463
DB		 3068 unregister_pernet_subsys(&vxlan_net_ops);
3069 /* rcu_barrier() is called by netns */
d342894c 3070}
3071module_exit(vxlan_cleanup_module);
3072
3073MODULE_LICENSE("GPL");
3074MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3075MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3076MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3077MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)