]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: fjes: Use resource_size
[mirror_ubuntu-jammy-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
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>
40d29af0 45
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
ee122c79 52#include <net/dst_metadata.h>
d342894c 53
54#define VXLAN_VERSION "0.1"
55
553675fb 56#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
23c578bf 61/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 63 * for compatibility with early adopters.
23c578bf 64 */
9daaa397
SH		 65static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 67MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
60d9d4c6 73static int vxlan_net_id;
0dfbdf41 74static struct rtnl_link_ops vxlan_link_ops;
553675fb 75
7256eac1 76static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 77
205f356d 78static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 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
d342894c 87/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
3e61aa8f 93 struct list_head remotes;
7177a3b0 94 u8 eth_addr[ETH_ALEN];
d342894c 95 u16 state; /* see ndm_state */
ae884082 96 u8 flags; /* see ndm_flags */
d342894c 97};
98
d342894c 99/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
758c57d1 101static struct workqueue_struct *vxlan_wq;
d342894c 102
ee122c79
TG		 103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
e7030878
TG		 105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
ee122c79
TG		 107}
108
e4c7ed41
CW		 109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
f0ef3126
JB		 113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
f0ef3126
JB		 123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
f0ef3126
JB		 131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
f0ef3126 139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
e4c7ed41
CW		 150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 153 const union vxlan_addr *ip)
e4c7ed41 154{
f0ef3126 155 if (ip->sa.sa_family == AF_INET6)
930345ea 156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 157 else
930345ea 158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
f0ef3126 166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
f0ef3126 171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
f0ef3126 176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
f0ef3126
JB		 181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
e4c7ed41
CW		 190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 193 const union vxlan_addr *ip)
e4c7ed41 194{
930345ea 195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 196}
197#endif
198
553675fb 199/* Virtual Network hash table head */
54bfd872 200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 201{
54bfd872 202 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
553675fb 210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
3e61aa8f
SH		 213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
5ca5461c 216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 217{
5ca5461c 218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 224}
225
ac5132d1
TG		 226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
553675fb 231{
232 struct vxlan_sock *vs;
af33c1ad
TH		 233
234 flags &= VXLAN_F_RCV_FLAGS;
553675fb 235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 238 vxlan_get_sk_family(vs) == family &&
af33c1ad 239 vs->flags == flags)
553675fb 240 return vs;
241 }
242 return NULL;
d342894c 243}
244
54bfd872 245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 246{
247 struct vxlan_dev *vxlan;
d342894c 248
b9167b2e
JB		 249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs->flags & VXLAN_F_COLLECT_METADATA)
251 vni = 0;
252
54bfd872
JB		 253 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
254 if (vxlan->default_dst.remote_vni == vni)
d342894c 255 return vxlan;
256 }
257
258 return NULL;
259}
260
5cfccc5a 261/* Look up VNI in a per net namespace table */
54bfd872 262static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 263 sa_family_t family, __be16 port,
264 u32 flags)
5cfccc5a
PS		 265{
266 struct vxlan_sock *vs;
267
ac5132d1 268 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 269 if (!vs)
270 return NULL;
271
54bfd872 272 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 273}
274
d342894c 275/* Fill in neighbour message in skbuff. */
276static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 277 const struct vxlan_fdb *fdb,
278 u32 portid, u32 seq, int type, unsigned int flags,
279 const struct vxlan_rdst *rdst)
d342894c 280{
281 unsigned long now = jiffies;
282 struct nda_cacheinfo ci;
283 struct nlmsghdr *nlh;
284 struct ndmsg *ndm;
e4f67add 285 bool send_ip, send_eth;
d342894c 286
287 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
288 if (nlh == NULL)
289 return -EMSGSIZE;
290
291 ndm = nlmsg_data(nlh);
292 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 293
294 send_eth = send_ip = true;
295
296 if (type == RTM_GETNEIGH) {
297 ndm->ndm_family = AF_INET;
e4c7ed41 298 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 299 send_eth = !is_zero_ether_addr(fdb->eth_addr);
300 } else
301 ndm->ndm_family = AF_BRIDGE;
d342894c 302 ndm->ndm_state = fdb->state;
303 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 304 ndm->ndm_flags = fdb->flags;
545469f7 305 ndm->ndm_type = RTN_UNICAST;
d342894c 306
193523bf 307 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 308 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 309 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 310 goto nla_put_failure;
311
e4f67add 312 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 313 goto nla_put_failure;
314
e4c7ed41 315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 316 goto nla_put_failure;
317
0dfbdf41 318 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 319 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
320 goto nla_put_failure;
c7995c43 321 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 322 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d
DS		 323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
053c095a
JB		 336 nlmsg_end(skb, nlh);
337 return 0;
d342894c 338
339nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342}
343
344static inline size_t vxlan_nlmsg_size(void)
345{
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 353 + nla_total_size(sizeof(struct nda_cacheinfo));
354}
355
9e4b93f9
ND		 356static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
d342894c 358{
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
9e4b93f9 367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380}
381
e4c7ed41 382static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 383{
384 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
e4c7ed41 389 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 391 };
3e61aa8f 392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 394}
395
396static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397{
bb3fd687
SH		 398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
9e4b93f9 401 struct vxlan_rdst remote = { };
e4f67add 402
e4f67add
DS		 403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
9e4b93f9 405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 406}
407
d342894c 408/* Hash Ethernet address */
409static u32 eth_hash(const unsigned char *addr)
410{
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414#ifdef __BIG_ENDIAN
d342894c 415 value >>= 16;
321fb991 416#else
417 value <<= 16;
d342894c 418#endif
419 return hash_64(value, FDB_HASH_BITS);
420}
421
422/* Hash chain to use given mac address */
423static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
424 const u8 *mac)
425{
426 return &vxlan->fdb_head[eth_hash(mac)];
427}
428
429/* Look up Ethernet address in forwarding table */
014be2c8 430static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 431 const u8 *mac)
d342894c 432{
433 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
434 struct vxlan_fdb *f;
d342894c 435
b67bfe0d 436 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 437 if (ether_addr_equal(mac, f->eth_addr))
d342894c 438 return f;
439 }
440
441 return NULL;
442}
443
014be2c8
SS		 444static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
445 const u8 *mac)
446{
447 struct vxlan_fdb *f;
448
449 f = __vxlan_find_mac(vxlan, mac);
450 if (f)
451 f->used = jiffies;
452
453 return f;
454}
455
a5e7c10a
MR		 456/* caller should hold vxlan->hash_lock */
457static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 458 union vxlan_addr *ip, __be16 port,
54bfd872 459 __be32 vni, __u32 ifindex)
6681712d 460{
3e61aa8f 461 struct vxlan_rdst *rd;
6681712d 462
3e61aa8f 463 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 464 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 465 rd->remote_port == port &&
466 rd->remote_vni == vni &&
467 rd->remote_ifindex == ifindex)
a5e7c10a 468 return rd;
6681712d 469 }
3e61aa8f 470
a5e7c10a
MR		 471 return NULL;
472}
473
906dc186
TR		 474/* Replace destination of unicast mac */
475static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 476 union vxlan_addr *ip, __be16 port, __be32 vni,
477 __u32 ifindex)
906dc186
TR		 478{
479 struct vxlan_rdst *rd;
480
481 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
482 if (rd)
483 return 0;
484
485 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
486 if (!rd)
487 return 0;
0c1d70af
PA		 488
489 dst_cache_reset(&rd->dst_cache);
e4c7ed41 490 rd->remote_ip = *ip;
906dc186
TR		 491 rd->remote_port = port;
492 rd->remote_vni = vni;
493 rd->remote_ifindex = ifindex;
494 return 1;
495}
496
a5e7c10a
MR		 497/* Add/update destinations for multicast */
498static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 499 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 500 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 501{
502 struct vxlan_rdst *rd;
503
504 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
505 if (rd)
506 return 0;
507
6681712d
DS		 508 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
509 if (rd == NULL)
510 return -ENOBUFS;
0c1d70af
PA		 511
512 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
513 kfree(rd);
514 return -ENOBUFS;
515 }
516
e4c7ed41 517 rd->remote_ip = *ip;
6681712d
DS		 518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 521
522 list_add_tail_rcu(&rd->list, &f->remotes);
523
9e4b93f9 524 *rdp = rd;
6681712d
DS		 525 return 1;
526}
527
dfd8645e
TH		 528static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
529 unsigned int off,
530 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 531 __be32 vni_field,
532 struct gro_remcsum *grc,
0ace2ca8 533 bool nopartial)
dfd8645e 534{
b7fe10e5 535 size_t start, offset;
dfd8645e
TH		 536
537 if (skb->remcsum_offload)
b7fe10e5 538 return vh;
dfd8645e
TH		 539
540 if (!NAPI_GRO_CB(skb)->csum_valid)
541 return NULL;
542
54bfd872
JB		 543 start = vxlan_rco_start(vni_field);
544 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 545
b7fe10e5
TH		 546 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
547 start, offset, grc, nopartial);
dfd8645e
TH		 548
549 skb->remcsum_offload = 1;
550
551 return vh;
552}
553
5602c48c
TH		 554static struct sk_buff **vxlan_gro_receive(struct sock *sk,
555 struct sk_buff **head,
556 struct sk_buff *skb)
dc01e7d3
OG		 557{
558 struct sk_buff *p, **pp = NULL;
559 struct vxlanhdr *vh, *vh2;
9b174d88 560 unsigned int hlen, off_vx;
dc01e7d3 561 int flush = 1;
5602c48c 562 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 563 __be32 flags;
26c4f7da
TH		 564 struct gro_remcsum grc;
565
566 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 567
568 off_vx = skb_gro_offset(skb);
569 hlen = off_vx + sizeof(*vh);
570 vh = skb_gro_header_fast(skb, off_vx);
571 if (skb_gro_header_hard(skb, hlen)) {
572 vh = skb_gro_header_slow(skb, hlen, off_vx);
573 if (unlikely(!vh))
574 goto out;
575 }
dc01e7d3 576
dfd8645e
TH		 577 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
578
54bfd872 579 flags = vh->vx_flags;
dfd8645e
TH		 580
581 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
582 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 583 vh->vx_vni, &grc,
0ace2ca8
TH		 584 !!(vs->flags &
585 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 586
587 if (!vh)
588 goto out;
589 }
590
b7fe10e5
TH		 591 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
592
dc01e7d3
OG		 593 for (p = *head; p; p = p->next) {
594 if (!NAPI_GRO_CB(p)->same_flow)
595 continue;
596
597 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 598 if (vh->vx_flags != vh2->vx_flags ||
599 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 600 NAPI_GRO_CB(p)->same_flow = 0;
601 continue;
602 }
dc01e7d3
OG		 603 }
604
9b174d88 605 pp = eth_gro_receive(head, skb);
c194cf93 606 flush = 0;
dc01e7d3 607
dc01e7d3 608out:
26c4f7da 609 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 610 NAPI_GRO_CB(skb)->flush |= flush;
611
612 return pp;
613}
614
5602c48c 615static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 616{
cfdf1e1b
JG		 617 udp_tunnel_gro_complete(skb, nhoff);
618
9b174d88 619 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 620}
621
53cf5275 622/* Notify netdevs that UDP port started listening */
dc01e7d3 623static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 624{
625 struct net_device *dev;
dc01e7d3 626 struct sock *sk = vs->sock->sk;
53cf5275 627 struct net *net = sock_net(sk);
705cc62f 628 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 629 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 630
631 rcu_read_lock();
632 for_each_netdev_rcu(net, dev) {
633 if (dev->netdev_ops->ndo_add_vxlan_port)
634 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
635 port);
636 }
637 rcu_read_unlock();
638}
639
640/* Notify netdevs that UDP port is no more listening */
dc01e7d3 641static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 642{
643 struct net_device *dev;
dc01e7d3 644 struct sock *sk = vs->sock->sk;
53cf5275 645 struct net *net = sock_net(sk);
705cc62f 646 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 647 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 648
649 rcu_read_lock();
650 for_each_netdev_rcu(net, dev) {
651 if (dev->netdev_ops->ndo_del_vxlan_port)
652 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
653 port);
654 }
655 rcu_read_unlock();
656}
657
d342894c 658/* Add new entry to forwarding table -- assumes lock held */
659static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 660 const u8 *mac, union vxlan_addr *ip,
6681712d 661 __u16 state, __u16 flags,
54bfd872 662 __be16 port, __be32 vni, __u32 ifindex,
ae884082 663 __u8 ndm_flags)
d342894c 664{
9e4b93f9 665 struct vxlan_rdst *rd = NULL;
d342894c 666 struct vxlan_fdb *f;
667 int notify = 0;
668
014be2c8 669 f = __vxlan_find_mac(vxlan, mac);
d342894c 670 if (f) {
671 if (flags & NLM_F_EXCL) {
672 netdev_dbg(vxlan->dev,
673 "lost race to create %pM\n", mac);
674 return -EEXIST;
675 }
676 if (f->state != state) {
677 f->state = state;
678 f->updated = jiffies;
679 notify = 1;
680 }
ae884082
DS		 681 if (f->flags != ndm_flags) {
682 f->flags = ndm_flags;
683 f->updated = jiffies;
684 notify = 1;
685 }
906dc186
TR		 686 if ((flags & NLM_F_REPLACE)) {
687 /* Only change unicasts */
688 if (!(is_multicast_ether_addr(f->eth_addr) ||
689 is_zero_ether_addr(f->eth_addr))) {
60840429 690 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 691 ifindex);
906dc186
TR		 692 } else
693 return -EOPNOTSUPP;
694 }
6681712d 695 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 696 (is_multicast_ether_addr(f->eth_addr) ||
697 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 698 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
699 &rd);
6681712d
DS		 700
701 if (rc < 0)
702 return rc;
703 notify |= rc;
704 }
d342894c 705 } else {
706 if (!(flags & NLM_F_CREATE))
707 return -ENOENT;
708
0dfbdf41
TG		 709 if (vxlan->cfg.addrmax &&
710 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 711 return -ENOSPC;
712
906dc186
TR		 713 /* Disallow replace to add a multicast entry */
714 if ((flags & NLM_F_REPLACE) &&
715 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
716 return -EOPNOTSUPP;
717
e4c7ed41 718 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 719 f = kmalloc(sizeof(*f), GFP_ATOMIC);
720 if (!f)
721 return -ENOMEM;
722
723 notify = 1;
d342894c 724 f->state = state;
ae884082 725 f->flags = ndm_flags;
d342894c 726 f->updated = f->used = jiffies;
3e61aa8f 727 INIT_LIST_HEAD(&f->remotes);
d342894c 728 memcpy(f->eth_addr, mac, ETH_ALEN);
729
9e4b93f9 730 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 731
d342894c 732 ++vxlan->addrcnt;
733 hlist_add_head_rcu(&f->hlist,
734 vxlan_fdb_head(vxlan, mac));
735 }
736
9e4b93f9
ND		 737 if (notify) {
738 if (rd == NULL)
739 rd = first_remote_rtnl(f);
740 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
741 }
d342894c 742
743 return 0;
744}
745
6706c82e 746static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 747{
748 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 749 struct vxlan_rdst *rd, *nd;
6681712d 750
0c1d70af
PA		 751 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
752 dst_cache_destroy(&rd->dst_cache);
6681712d 753 kfree(rd);
0c1d70af 754 }
6681712d
DS		 755 kfree(f);
756}
757
d342894c 758static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
759{
760 netdev_dbg(vxlan->dev,
761 "delete %pM\n", f->eth_addr);
762
763 --vxlan->addrcnt;
9e4b93f9 764 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 765
766 hlist_del_rcu(&f->hlist);
6681712d 767 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 768}
769
f0b074be 770static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB		 771 union vxlan_addr *ip, __be16 *port, __be32 *vni,
772 u32 *ifindex)
d342894c 773{
6681712d 774 struct net *net = dev_net(vxlan->dev);
e4c7ed41 775 int err;
d342894c 776
f0b074be 777 if (tb[NDA_DST]) {
e4c7ed41
CW		 778 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
779 if (err)
780 return err;
f0b074be 781 } else {
e4c7ed41
CW		 782 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
783 if (remote->sa.sa_family == AF_INET) {
784 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
785 ip->sa.sa_family = AF_INET;
786#if IS_ENABLED(CONFIG_IPV6)
787 } else {
788 ip->sin6.sin6_addr = in6addr_any;
789 ip->sa.sa_family = AF_INET6;
790#endif
791 }
f0b074be 792 }
d342894c 793
6681712d 794 if (tb[NDA_PORT]) {
73cf3317 795 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 796 return -EINVAL;
f0b074be
MR		 797 *port = nla_get_be16(tb[NDA_PORT]);
798 } else {
0dfbdf41 799 *port = vxlan->cfg.dst_port;
f0b074be 800 }
6681712d
DS		 801
802 if (tb[NDA_VNI]) {
803 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
804 return -EINVAL;
54bfd872 805 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 806 } else {
807 *vni = vxlan->default_dst.remote_vni;
808 }
6681712d
DS		 809
810 if (tb[NDA_IFINDEX]) {
5abb0029 811 struct net_device *tdev;
6681712d
DS		 812
813 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
814 return -EINVAL;
f0b074be 815 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 816 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 817 if (!tdev)
6681712d 818 return -EADDRNOTAVAIL;
f0b074be
MR		 819 } else {
820 *ifindex = 0;
821 }
822
823 return 0;
824}
825
826/* Add static entry (via netlink) */
827static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
828 struct net_device *dev,
f6f6424b 829 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 830{
831 struct vxlan_dev *vxlan = netdev_priv(dev);
832 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 833 union vxlan_addr ip;
f0b074be 834 __be16 port;
54bfd872
JB		 835 __be32 vni;
836 u32 ifindex;
f0b074be
MR		 837 int err;
838
839 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
840 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
841 ndm->ndm_state);
842 return -EINVAL;
843 }
844
845 if (tb[NDA_DST] == NULL)
846 return -EINVAL;
847
848 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
849 if (err)
850 return err;
6681712d 851
5933a7bb
MR		 852 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
853 return -EAFNOSUPPORT;
854
d342894c 855 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 856 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 857 port, vni, ifindex, ndm->ndm_flags);
d342894c 858 spin_unlock_bh(&vxlan->hash_lock);
859
860 return err;
861}
862
863/* Delete entry (via netlink) */
1690be63
VY		 864static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
865 struct net_device *dev,
f6f6424b 866 const unsigned char *addr, u16 vid)
d342894c 867{
868 struct vxlan_dev *vxlan = netdev_priv(dev);
869 struct vxlan_fdb *f;
bc7892ba 870 struct vxlan_rdst *rd = NULL;
e4c7ed41 871 union vxlan_addr ip;
bc7892ba 872 __be16 port;
54bfd872
JB		 873 __be32 vni;
874 u32 ifindex;
bc7892ba
MR		 875 int err;
876
877 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
878 if (err)
879 return err;
880
881 err = -ENOENT;
d342894c 882
883 spin_lock_bh(&vxlan->hash_lock);
884 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 885 if (!f)
886 goto out;
887
e4c7ed41
CW		 888 if (!vxlan_addr_any(&ip)) {
889 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 890 if (!rd)
891 goto out;
892 }
893
894 err = 0;
895
896 /* remove a destination if it's not the only one on the list,
897 * otherwise destroy the fdb entry
898 */
899 if (rd && !list_is_singular(&f->remotes)) {
900 list_del_rcu(&rd->list);
9e4b93f9 901 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 902 kfree_rcu(rd, rcu);
bc7892ba 903 goto out;
d342894c 904 }
bc7892ba
MR		 905
906 vxlan_fdb_destroy(vxlan, f);
907
908out:
d342894c 909 spin_unlock_bh(&vxlan->hash_lock);
910
911 return err;
912}
913
914/* Dump forwarding table */
915static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 916 struct net_device *dev,
917 struct net_device *filter_dev, int idx)
d342894c 918{
919 struct vxlan_dev *vxlan = netdev_priv(dev);
920 unsigned int h;
921
922 for (h = 0; h < FDB_HASH_SIZE; ++h) {
923 struct vxlan_fdb *f;
d342894c 924 int err;
925
b67bfe0d 926 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 927 struct vxlan_rdst *rd;
6681712d 928
3e61aa8f 929 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 930 if (idx < cb->args[0])
931 goto skip;
932
6681712d
DS		 933 err = vxlan_fdb_info(skb, vxlan, f,
934 NETLINK_CB(cb->skb).portid,
935 cb->nlh->nlmsg_seq,
936 RTM_NEWNEIGH,
937 NLM_F_MULTI, rd);
472681d5
MM		 938 if (err < 0) {
939 cb->args[1] = err;
3e61aa8f 940 goto out;
472681d5 941 }
3e61aa8f 942skip:
07a51cd3
AW		 943 ++idx;
944 }
d342894c 945 }
946 }
3e61aa8f 947out:
d342894c 948 return idx;
949}
950
951/* Watch incoming packets to learn mapping between Ethernet address
952 * and Tunnel endpoint.
c4b49512 953 * Return true if packet is bogus and should be dropped.
d342894c 954 */
26a41ae6 955static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 956 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 957{
958 struct vxlan_dev *vxlan = netdev_priv(dev);
959 struct vxlan_fdb *f;
d342894c 960
961 f = vxlan_find_mac(vxlan, src_mac);
962 if (likely(f)) {
5ca5461c 963 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 964
e4c7ed41 965 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 966 return false;
967
968 /* Don't migrate static entries, drop packets */
eb064c3b 969 if (f->state & NUD_NOARP)
26a41ae6 970 return true;
d342894c 971
972 if (net_ratelimit())
973 netdev_info(dev,
e4c7ed41 974 "%pM migrated from %pIS to %pIS\n",
a4870f79 975 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 976
e4c7ed41 977 rdst->remote_ip = *src_ip;
d342894c 978 f->updated = jiffies;
9e4b93f9 979 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 980 } else {
981 /* learned new entry */
982 spin_lock(&vxlan->hash_lock);
3bf74b1a 983
984 /* close off race between vxlan_flush and incoming packets */
985 if (netif_running(dev))
986 vxlan_fdb_create(vxlan, src_mac, src_ip,
987 NUD_REACHABLE,
988 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 989 vxlan->cfg.dst_port,
3bf74b1a 990 vxlan->default_dst.remote_vni,
991 0, NTF_SELF);
d342894c 992 spin_unlock(&vxlan->hash_lock);
993 }
26a41ae6 994
995 return false;
d342894c 996}
997
d342894c 998/* See if multicast group is already in use by other ID */
95ab0991 999static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1000{
553675fb 1001 struct vxlan_dev *vxlan;
b1be00a6 1002 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1003
95ab0991
G		 1004 /* The vxlan_sock is only used by dev, leaving group has
1005 * no effect on other vxlan devices.
1006 */
b1be00a6
JB		 1007 if (family == AF_INET && dev->vn4_sock &&
1008 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1009 return false;
b1be00a6
JB		 1010#if IS_ENABLED(CONFIG_IPV6)
1011 if (family == AF_INET6 && dev->vn6_sock &&
1012 atomic_read(&dev->vn6_sock->refcnt) == 1)
1013 return false;
1014#endif
95ab0991 1015
553675fb 1016 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1017 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1018 continue;
d342894c 1019
b1be00a6 1020 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1021 continue;
b1be00a6
JB		 1022#if IS_ENABLED(CONFIG_IPV6)
1023 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1024 continue;
1025#endif
95ab0991
G		 1026
1027 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1028 &dev->default_dst.remote_ip))
1029 continue;
1030
1031 if (vxlan->default_dst.remote_ifindex !=
1032 dev->default_dst.remote_ifindex)
1033 continue;
1034
1035 return true;
553675fb 1036 }
d342894c 1037
1038 return false;
1039}
1040
b1be00a6 1041static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1042{
b1be00a6 1043 struct vxlan_net *vn;
012a5729 1044
b1be00a6
JB		 1045 if (!vs)
1046 return;
7c47cedf
SH		 1047 if (!atomic_dec_and_test(&vs->refcnt))
1048 return;
d342894c 1049
b1be00a6 1050 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1051 spin_lock(&vn->sock_lock);
7c47cedf 1052 hlist_del_rcu(&vs->hlist);
dc01e7d3 1053 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1054 spin_unlock(&vn->sock_lock);
1055
7c47cedf 1056 queue_work(vxlan_wq, &vs->del_work);
d342894c 1057}
1058
b1be00a6
JB		 1059static void vxlan_sock_release(struct vxlan_dev *vxlan)
1060{
1061 __vxlan_sock_release(vxlan->vn4_sock);
1062#if IS_ENABLED(CONFIG_IPV6)
1063 __vxlan_sock_release(vxlan->vn6_sock);
1064#endif
1065}
1066
56ef9c90 1067/* Update multicast group membership when first VNI on
c4b49512 1068 * multicast address is brought up
7c47cedf 1069 */
56ef9c90 1070static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1071{
b1be00a6 1072 struct sock *sk;
e4c7ed41
CW		 1073 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1074 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1075 int ret = -EINVAL;
d342894c 1076
e4c7ed41
CW		 1077 if (ip->sa.sa_family == AF_INET) {
1078 struct ip_mreqn mreq = {
1079 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1080 .imr_ifindex = ifindex,
1081 };
1082
b1be00a6
JB		 1083 sk = vxlan->vn4_sock->sock->sk;
1084 lock_sock(sk);
56ef9c90 1085 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1086 release_sock(sk);
e4c7ed41
CW		 1087#if IS_ENABLED(CONFIG_IPV6)
1088 } else {
b1be00a6
JB		 1089 sk = vxlan->vn6_sock->sock->sk;
1090 lock_sock(sk);
56ef9c90
MRL		 1091 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1092 &ip->sin6.sin6_addr);
b1be00a6 1093 release_sock(sk);
e4c7ed41
CW		 1094#endif
1095 }
3fc2de2f 1096
56ef9c90 1097 return ret;
3fc2de2f 1098}
1099
1100/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1101static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1102{
b1be00a6 1103 struct sock *sk;
e4c7ed41
CW		 1104 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1105 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1106 int ret = -EINVAL;
3fc2de2f 1107
e4c7ed41
CW		 1108 if (ip->sa.sa_family == AF_INET) {
1109 struct ip_mreqn mreq = {
1110 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1111 .imr_ifindex = ifindex,
1112 };
1113
b1be00a6
JB		 1114 sk = vxlan->vn4_sock->sock->sk;
1115 lock_sock(sk);
56ef9c90 1116 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1117 release_sock(sk);
e4c7ed41
CW		 1118#if IS_ENABLED(CONFIG_IPV6)
1119 } else {
b1be00a6
JB		 1120 sk = vxlan->vn6_sock->sock->sk;
1121 lock_sock(sk);
56ef9c90
MRL		 1122 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1123 &ip->sin6.sin6_addr);
b1be00a6 1124 release_sock(sk);
e4c7ed41
CW		 1125#endif
1126 }
d342894c 1127
56ef9c90 1128 return ret;
d342894c 1129}
1130
f14ecebb
JB		 1131static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1132 struct sk_buff *skb, u32 vxflags)
dfd8645e 1133{
7d34fa75 1134 size_t start, offset;
dfd8645e 1135
f14ecebb
JB		 1136 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1137 goto out;
b7fe10e5 1138
f14ecebb
JB		 1139 start = vxlan_rco_start(unparsed->vx_vni);
1140 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1141
7d34fa75 1142 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1143 return false;
dfd8645e 1144
be5cfeab
JB		 1145 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1146 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1147out:
1148 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1149 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1150 return true;
dfd8645e
TH		 1151}
1152
f14ecebb 1153static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1154 struct sk_buff *skb, u32 vxflags,
10a5af23 1155 struct vxlan_metadata *md)
3288af08 1156{
f14ecebb 1157 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1158 struct metadata_dst *tun_dst;
f14ecebb
JB		 1159
1160 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1161 goto out;
3288af08 1162
3288af08
JB		 1163 md->gbp = ntohs(gbp->policy_id);
1164
10a5af23 1165 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1166 if (tun_dst) {
3288af08 1167 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1168 tun_dst->u.tun_info.options_len = sizeof(*md);
1169 }
3288af08
JB		 1170 if (gbp->dont_learn)
1171 md->gbp |= VXLAN_GBP_DONT_LEARN;
1172
1173 if (gbp->policy_applied)
1174 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1175
64f87d36
JB		 1176 /* In flow-based mode, GBP is carried in dst_metadata */
1177 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1178 skb->mark = md->gbp;
f14ecebb
JB		 1179out:
1180 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1181}
1182
e1e5314d
JB		 1183static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1184 __be32 *protocol,
1185 struct sk_buff *skb, u32 vxflags)
1186{
1187 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1188
1189 /* Need to have Next Protocol set for interfaces in GPE mode. */
1190 if (!gpe->np_applied)
1191 return false;
1192 /* "The initial version is 0. If a receiver does not support the
1193 * version indicated it MUST drop the packet.
1194 */
1195 if (gpe->version != 0)
1196 return false;
1197 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1198 * processing MUST occur." However, we don't implement OAM
1199 * processing, thus drop the packet.
1200 */
1201 if (gpe->oam_flag)
1202 return false;
1203
1204 switch (gpe->next_protocol) {
1205 case VXLAN_GPE_NP_IPV4:
1206 *protocol = htons(ETH_P_IP);
1207 break;
1208 case VXLAN_GPE_NP_IPV6:
1209 *protocol = htons(ETH_P_IPV6);
1210 break;
1211 case VXLAN_GPE_NP_ETHERNET:
1212 *protocol = htons(ETH_P_TEB);
1213 break;
1214 default:
1215 return false;
1216 }
1217
1218 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1219 return true;
1220}
1221
1ab016e2
JB		 1222static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1223 struct vxlan_sock *vs,
1224 struct sk_buff *skb)
614732ea 1225{
614732ea 1226 union vxlan_addr saddr;
614732ea 1227
614732ea 1228 skb_reset_mac_header(skb);
614732ea
TG		 1229 skb->protocol = eth_type_trans(skb, vxlan->dev);
1230 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1231
1232 /* Ignore packet loops (and multicast echo) */
1233 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1234 return false;
614732ea 1235
760c6805 1236 /* Get address from the outer IP header */
ce212d0f 1237 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1238 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1239 saddr.sa.sa_family = AF_INET;
1240#if IS_ENABLED(CONFIG_IPV6)
1241 } else {
1ab016e2 1242 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1243 saddr.sa.sa_family = AF_INET6;
1244#endif
1245 }
1246
1ab016e2
JB		 1247 if ((vxlan->flags & VXLAN_F_LEARN) &&
1248 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1249 return false;
1250
1251 return true;
1252}
1253
760c6805
JB		 1254static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1255 struct sk_buff *skb)
1256{
1257 int err = 0;
1258
1259 if (vxlan_get_sk_family(vs) == AF_INET)
1260 err = IP_ECN_decapsulate(oiph, skb);
1261#if IS_ENABLED(CONFIG_IPV6)
1262 else
1263 err = IP6_ECN_decapsulate(oiph, skb);
1264#endif
1265
1266 if (unlikely(err) && log_ecn_error) {
1267 if (vxlan_get_sk_family(vs) == AF_INET)
1268 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1269 &((struct iphdr *)oiph)->saddr,
1270 ((struct iphdr *)oiph)->tos);
1271 else
1272 net_info_ratelimited("non-ECT from %pI6\n",
1273 &((struct ipv6hdr *)oiph)->saddr);
1274 }
1275 return err <= 1;
1276}
1277
d342894c 1278/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1279static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1280{
f2d1968e 1281 struct pcpu_sw_netstats *stats;
c9e78efb 1282 struct vxlan_dev *vxlan;
5cfccc5a 1283 struct vxlan_sock *vs;
f14ecebb 1284 struct vxlanhdr unparsed;
ee122c79
TG		 1285 struct vxlan_metadata _md;
1286 struct vxlan_metadata *md = &_md;
e1e5314d
JB		 1287 __be32 protocol = htons(ETH_P_TEB);
1288 bool raw_proto = false;
f2d1968e 1289 void *oiph;
d342894c 1290
e1e5314d 1291 /* Need UDP and VXLAN header to be present */
7ce04758 1292 if (!pskb_may_pull(skb, VXLAN_HLEN))
288b01c8 1293 return 1;
d342894c 1294
f14ecebb 1295 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1296 /* VNI flag always required to be set */
1297 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1298 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1299 ntohl(vxlan_hdr(skb)->vx_flags),
1300 ntohl(vxlan_hdr(skb)->vx_vni));
1301 /* Return non vxlan pkt */
1302 return 1;
d342894c 1303 }
288b01c8
JB		 1304 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1305 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1306
559835ea 1307 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1308 if (!vs)
d342894c 1309 goto drop;
d342894c 1310
c9e78efb
JB		 1311 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1312 if (!vxlan)
1313 goto drop;
1314
e1e5314d
JB		 1315 /* For backwards compatibility, only allow reserved fields to be
1316 * used by VXLAN extensions if explicitly requested.
1317 */
1318 if (vs->flags & VXLAN_F_GPE) {
1319 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1320 goto drop;
1321 raw_proto = true;
1322 }
1323
1324 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1325 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1326 goto drop;
c9e78efb 1327
ee122c79 1328 if (vxlan_collect_metadata(vs)) {
07dabf20 1329 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
10a5af23 1330 struct metadata_dst *tun_dst;
07dabf20 1331
c29a70d2 1332 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
07dabf20 1333 vxlan_vni_to_tun_id(vni), sizeof(*md));
c29a70d2 1334
ee122c79
TG		 1335 if (!tun_dst)
1336 goto drop;
1337
0f1b7354 1338 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1339
1340 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1341 } else {
1342 memset(md, 0, sizeof(*md));
1343 }
1344
f14ecebb
JB		 1345 if (vs->flags & VXLAN_F_REMCSUM_RX)
1346 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1347 goto drop;
1348 if (vs->flags & VXLAN_F_GBP)
10a5af23 1349 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1350 /* Note that GBP and GPE can never be active together. This is
1351 * ensured in vxlan_dev_configure.
1352 */
3511494c 1353
f14ecebb 1354 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1355 /* If there are any unprocessed flags remaining treat
1356 * this as a malformed packet. This behavior diverges from
1357 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1358 * in reserved fields are to be ignored. The approach here
c4b49512 1359 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1360 * is more robust and provides a little more security in
1361 * adding extensions to VXLAN.
1362 */
288b01c8 1363 goto drop;
3bf39475
TH		 1364 }
1365
e1e5314d
JB		 1366 if (!raw_proto) {
1367 if (!vxlan_set_mac(vxlan, vs, skb))
1368 goto drop;
1369 } else {
1370 skb->dev = vxlan->dev;
1371 skb->pkt_type = PACKET_HOST;
1372 }
f2d1968e 1373
f2d1968e
JB		 1374 oiph = skb_network_header(skb);
1375 skb_reset_network_header(skb);
1376
1377 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1378 ++vxlan->dev->stats.rx_frame_errors;
1379 ++vxlan->dev->stats.rx_errors;
1380 goto drop;
1381 }
1382
1383 stats = this_cpu_ptr(vxlan->dev->tstats);
1384 u64_stats_update_begin(&stats->syncp);
1385 stats->rx_packets++;
1386 stats->rx_bytes += skb->len;
1387 u64_stats_update_end(&stats->syncp);
1388
1389 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1390 return 0;
1391
1392drop:
288b01c8
JB		 1393 /* Consume bad packet */
1394 kfree_skb(skb);
1395 return 0;
5cfccc5a
PS		 1396}
1397
e4f67add
DS		 1398static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1399{
1400 struct vxlan_dev *vxlan = netdev_priv(dev);
1401 struct arphdr *parp;
1402 u8 *arpptr, *sha;
1403 __be32 sip, tip;
1404 struct neighbour *n;
1405
1406 if (dev->flags & IFF_NOARP)
1407 goto out;
1408
1409 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1410 dev->stats.tx_dropped++;
1411 goto out;
1412 }
1413 parp = arp_hdr(skb);
1414
1415 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1416 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1417 parp->ar_pro != htons(ETH_P_IP) ||
1418 parp->ar_op != htons(ARPOP_REQUEST) ||
1419 parp->ar_hln != dev->addr_len ||
1420 parp->ar_pln != 4)
1421 goto out;
1422 arpptr = (u8 *)parp + sizeof(struct arphdr);
1423 sha = arpptr;
1424 arpptr += dev->addr_len; /* sha */
1425 memcpy(&sip, arpptr, sizeof(sip));
1426 arpptr += sizeof(sip);
1427 arpptr += dev->addr_len; /* tha */
1428 memcpy(&tip, arpptr, sizeof(tip));
1429
1430 if (ipv4_is_loopback(tip) ||
1431 ipv4_is_multicast(tip))
1432 goto out;
1433
1434 n = neigh_lookup(&arp_tbl, &tip, dev);
1435
1436 if (n) {
e4f67add
DS		 1437 struct vxlan_fdb *f;
1438 struct sk_buff *reply;
1439
1440 if (!(n->nud_state & NUD_CONNECTED)) {
1441 neigh_release(n);
1442 goto out;
1443 }
1444
1445 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1446 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1447 /* bridge-local neighbor */
1448 neigh_release(n);
1449 goto out;
1450 }
1451
1452 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1453 n->ha, sha);
1454
1455 neigh_release(n);
1456
7346135d
DS		 1457 if (reply == NULL)
1458 goto out;
1459
e4f67add
DS		 1460 skb_reset_mac_header(reply);
1461 __skb_pull(reply, skb_network_offset(reply));
1462 reply->ip_summed = CHECKSUM_UNNECESSARY;
1463 reply->pkt_type = PACKET_HOST;
1464
1465 if (netif_rx_ni(reply) == NET_RX_DROP)
1466 dev->stats.rx_dropped++;
e4c7ed41
CW		 1467 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1468 union vxlan_addr ipa = {
1469 .sin.sin_addr.s_addr = tip,
a45e92a5 1470 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1471 };
1472
1473 vxlan_ip_miss(dev, &ipa);
1474 }
e4f67add
DS		 1475out:
1476 consume_skb(skb);
1477 return NETDEV_TX_OK;
1478}
1479
f564f45c 1480#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1481static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1482 struct neighbour *n, bool isrouter)
1483{
1484 struct net_device *dev = request->dev;
1485 struct sk_buff *reply;
1486 struct nd_msg *ns, *na;
1487 struct ipv6hdr *pip6;
1488 u8 *daddr;
1489 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1490 int ns_olen;
1491 int i, len;
1492
1493 if (dev == NULL)
1494 return NULL;
1495
1496 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1497 sizeof(*na) + na_olen + dev->needed_tailroom;
1498 reply = alloc_skb(len, GFP_ATOMIC);
1499 if (reply == NULL)
1500 return NULL;
1501
1502 reply->protocol = htons(ETH_P_IPV6);
1503 reply->dev = dev;
1504 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1505 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1506 skb_reset_mac_header(reply);
4b29dba9
DS		 1507
1508 ns = (struct nd_msg *)skb_transport_header(request);
1509
1510 daddr = eth_hdr(request)->h_source;
1511 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1512 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1513 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1514 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1515 break;
1516 }
1517 }
1518
1519 /* Ethernet header */
1520 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1521 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1522 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1523 reply->protocol = htons(ETH_P_IPV6);
1524
1525 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1526 skb_reset_network_header(reply);
4b29dba9
DS		 1527 skb_put(reply, sizeof(struct ipv6hdr));
1528
1529 /* IPv6 header */
1530
1531 pip6 = ipv6_hdr(reply);
1532 memset(pip6, 0, sizeof(struct ipv6hdr));
1533 pip6->version = 6;
1534 pip6->priority = ipv6_hdr(request)->priority;
1535 pip6->nexthdr = IPPROTO_ICMPV6;
1536 pip6->hop_limit = 255;
1537 pip6->daddr = ipv6_hdr(request)->saddr;
1538 pip6->saddr = *(struct in6_addr *)n->primary_key;
1539
1540 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1541 skb_reset_transport_header(reply);
4b29dba9
DS		 1542
1543 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1544
1545 /* Neighbor Advertisement */
1546 memset(na, 0, sizeof(*na)+na_olen);
1547 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1548 na->icmph.icmp6_router = isrouter;
1549 na->icmph.icmp6_override = 1;
1550 na->icmph.icmp6_solicited = 1;
1551 na->target = ns->target;
1552 ether_addr_copy(&na->opt[2], n->ha);
1553 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1554 na->opt[1] = na_olen >> 3;
1555
1556 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1557 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1558 csum_partial(na, sizeof(*na)+na_olen, 0));
1559
1560 pip6->payload_len = htons(sizeof(*na)+na_olen);
1561
1562 skb_push(reply, sizeof(struct ipv6hdr));
1563
1564 reply->ip_summed = CHECKSUM_UNNECESSARY;
1565
1566 return reply;
1567}
1568
f564f45c
CW		 1569static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1570{
1571 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1572 struct nd_msg *msg;
f564f45c
CW		 1573 const struct ipv6hdr *iphdr;
1574 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1575 struct neighbour *n;
1576 struct inet6_dev *in6_dev;
f564f45c
CW		 1577
1578 in6_dev = __in6_dev_get(dev);
1579 if (!in6_dev)
1580 goto out;
1581
f564f45c
CW		 1582 iphdr = ipv6_hdr(skb);
1583 saddr = &iphdr->saddr;
1584 daddr = &iphdr->daddr;
1585
f564f45c
CW		 1586 msg = (struct nd_msg *)skb_transport_header(skb);
1587 if (msg->icmph.icmp6_code != 0 ||
1588 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1589 goto out;
1590
4b29dba9
DS		 1591 if (ipv6_addr_loopback(daddr) ||
1592 ipv6_addr_is_multicast(&msg->target))
1593 goto out;
1594
1595 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1596
1597 if (n) {
1598 struct vxlan_fdb *f;
4b29dba9 1599 struct sk_buff *reply;
f564f45c
CW		 1600
1601 if (!(n->nud_state & NUD_CONNECTED)) {
1602 neigh_release(n);
1603 goto out;
1604 }
1605
1606 f = vxlan_find_mac(vxlan, n->ha);
1607 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1608 /* bridge-local neighbor */
1609 neigh_release(n);
1610 goto out;
1611 }
1612
4b29dba9
DS		 1613 reply = vxlan_na_create(skb, n,
1614 !!(f ? f->flags & NTF_ROUTER : 0));
1615
f564f45c 1616 neigh_release(n);
4b29dba9
DS		 1617
1618 if (reply == NULL)
1619 goto out;
1620
1621 if (netif_rx_ni(reply) == NET_RX_DROP)
1622 dev->stats.rx_dropped++;
1623
f564f45c 1624 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1625 union vxlan_addr ipa = {
1626 .sin6.sin6_addr = msg->target,
a45e92a5 1627 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1628 };
1629
f564f45c
CW		 1630 vxlan_ip_miss(dev, &ipa);
1631 }
1632
1633out:
1634 consume_skb(skb);
1635 return NETDEV_TX_OK;
1636}
1637#endif
1638
e4f67add
DS		 1639static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1640{
1641 struct vxlan_dev *vxlan = netdev_priv(dev);
1642 struct neighbour *n;
e4f67add
DS		 1643
1644 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1645 return false;
1646
1647 n = NULL;
1648 switch (ntohs(eth_hdr(skb)->h_proto)) {
1649 case ETH_P_IP:
e15a00aa
CW		 1650 {
1651 struct iphdr *pip;
1652
e4f67add
DS		 1653 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1654 return false;
1655 pip = ip_hdr(skb);
1656 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1657 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1658 union vxlan_addr ipa = {
1659 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1660 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1661 };
1662
1663 vxlan_ip_miss(dev, &ipa);
1664 return false;
1665 }
1666
e4f67add 1667 break;
e15a00aa
CW		 1668 }
1669#if IS_ENABLED(CONFIG_IPV6)
1670 case ETH_P_IPV6:
1671 {
1672 struct ipv6hdr *pip6;
1673
1674 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1675 return false;
1676 pip6 = ipv6_hdr(skb);
1677 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1678 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1679 union vxlan_addr ipa = {
1680 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1681 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1682 };
1683
1684 vxlan_ip_miss(dev, &ipa);
1685 return false;
1686 }
1687
1688 break;
1689 }
1690#endif
e4f67add
DS		 1691 default:
1692 return false;
1693 }
1694
1695 if (n) {
1696 bool diff;
1697
7367d0b5 1698 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1699 if (diff) {
1700 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1701 dev->addr_len);
1702 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1703 }
1704 neigh_release(n);
1705 return diff;
e4c7ed41
CW		 1706 }
1707
e4f67add
DS		 1708 return false;
1709}
1710
af33c1ad 1711static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1712 struct vxlan_metadata *md)
1713{
1714 struct vxlanhdr_gbp *gbp;
1715
db79a621
TG		 1716 if (!md->gbp)
1717 return;
1718
3511494c 1719 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1720 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1721
1722 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1723 gbp->dont_learn = 1;
1724
1725 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1726 gbp->policy_applied = 1;
1727
1728 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1729}
1730
e1e5314d
JB		 1731static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1732 __be16 protocol)
1733{
1734 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1735
1736 gpe->np_applied = 1;
1737
1738 switch (protocol) {
1739 case htons(ETH_P_IP):
1740 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1741 return 0;
1742 case htons(ETH_P_IPV6):
1743 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1744 return 0;
1745 case htons(ETH_P_TEB):
1746 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1747 return 0;
1748 }
1749 return -EPFNOSUPPORT;
1750}
1751
f491e56d
JB		 1752static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1753 int iphdr_len, __be32 vni,
1754 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1755 bool udp_sum)
e4c7ed41 1756{
e4c7ed41 1757 struct vxlanhdr *vxh;
e4c7ed41
CW		 1758 int min_headroom;
1759 int err;
dfd8645e 1760 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1761 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1762
af33c1ad 1763 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1764 skb->ip_summed == CHECKSUM_PARTIAL) {
1765 int csum_start = skb_checksum_start_offset(skb);
1766
1767 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1768 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1769 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1770 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1771 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1772 }
e4c7ed41 1773
e4c7ed41 1774 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1775 + VXLAN_HLEN + iphdr_len
df8a39de 1776 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1777
1778 /* Need space for new headers (invalidates iph ptr) */
1779 err = skb_cow_head(skb, min_headroom);
e1e5314d
JB		 1780 if (unlikely(err))
1781 goto out_free;
649c5b8b 1782
5968250c
JP		 1783 skb = vlan_hwaccel_push_inside(skb);
1784 if (WARN_ON(!skb))
1785 return -ENOMEM;
1eaa8178 1786
6fa79666 1787 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG		 1788 if (IS_ERR(skb))
1789 return PTR_ERR(skb);
1790
49560532 1791 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1792 vxh->vx_flags = VXLAN_HF_VNI;
1793 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1794
dfd8645e 1795 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1796 unsigned int start;
dfd8645e 1797
54bfd872
JB		 1798 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1799 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1800 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1801
1802 if (!skb_is_gso(skb)) {
1803 skb->ip_summed = CHECKSUM_NONE;
1804 skb->encapsulation = 0;
1805 }
1806 }
1807
af33c1ad
TH		 1808 if (vxflags & VXLAN_F_GBP)
1809 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1810 if (vxflags & VXLAN_F_GPE) {
1811 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1812 if (err < 0)
1813 goto out_free;
1814 inner_protocol = skb->protocol;
1815 }
3511494c 1816
e1e5314d 1817 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1818 return 0;
e1e5314d
JB		 1819
1820out_free:
1821 kfree_skb(skb);
1822 return err;
49560532 1823}
49560532 1824
1a8496ba
JB		 1825static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1826 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1827 __be32 daddr, __be32 *saddr,
1828 struct dst_cache *dst_cache,
db3c6139 1829 const struct ip_tunnel_info *info)
1a8496ba 1830{
db3c6139 1831 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1832 struct rtable *rt = NULL;
1833 struct flowi4 fl4;
1834
db3c6139
DB		 1835 if (tos && !info)
1836 use_cache = false;
1837 if (use_cache) {
0c1d70af
PA		 1838 rt = dst_cache_get_ip4(dst_cache, saddr);
1839 if (rt)
1840 return rt;
1841 }
1842
1a8496ba
JB		 1843 memset(&fl4, 0, sizeof(fl4));
1844 fl4.flowi4_oif = oif;
1845 fl4.flowi4_tos = RT_TOS(tos);
1846 fl4.flowi4_mark = skb->mark;
1847 fl4.flowi4_proto = IPPROTO_UDP;
1848 fl4.daddr = daddr;
1849 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1850
1851 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1852 if (!IS_ERR(rt)) {
1a8496ba 1853 *saddr = fl4.saddr;
0c1d70af
PA		 1854 if (use_cache)
1855 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1856 }
1a8496ba
JB		 1857 return rt;
1858}
1859
e5d4b29f
JB		 1860#if IS_ENABLED(CONFIG_IPV6)
1861static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1400615d 1862 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1863 __be32 label,
e5d4b29f 1864 const struct in6_addr *daddr,
0c1d70af 1865 struct in6_addr *saddr,
db3c6139
DB		 1866 struct dst_cache *dst_cache,
1867 const struct ip_tunnel_info *info)
e5d4b29f 1868{
db3c6139 1869 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1870 struct dst_entry *ndst;
1871 struct flowi6 fl6;
1872 int err;
1873
1400615d
DB		 1874 if (tos && !info)
1875 use_cache = false;
db3c6139 1876 if (use_cache) {
0c1d70af
PA		 1877 ndst = dst_cache_get_ip6(dst_cache, saddr);
1878 if (ndst)
1879 return ndst;
1880 }
1881
e5d4b29f
JB		 1882 memset(&fl6, 0, sizeof(fl6));
1883 fl6.flowi6_oif = oif;
1884 fl6.daddr = *daddr;
1885 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
eaa93bf4 1886 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1887 fl6.flowi6_mark = skb->mark;
1888 fl6.flowi6_proto = IPPROTO_UDP;
1889
1890 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1891 vxlan->vn6_sock->sock->sk,
1892 &ndst, &fl6);
1893 if (err < 0)
1894 return ERR_PTR(err);
1895
1896 *saddr = fl6.saddr;
db3c6139 1897 if (use_cache)
0c1d70af 1898 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1899 return ndst;
1900}
1901#endif
1902
9dcc71e1
SS		 1903/* Bypass encapsulation if the destination is local */
1904static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1905 struct vxlan_dev *dst_vxlan)
1906{
8f84985f 1907 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1908 union vxlan_addr loopback;
1909 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1910 struct net_device *dev = skb->dev;
1911 int len = skb->len;
9dcc71e1 1912
8f84985f
LR		 1913 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1914 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1915 skb->pkt_type = PACKET_HOST;
1916 skb->encapsulation = 0;
1917 skb->dev = dst_vxlan->dev;
1918 __skb_pull(skb, skb_network_offset(skb));
1919
e4c7ed41
CW		 1920 if (remote_ip->sa.sa_family == AF_INET) {
1921 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1922 loopback.sa.sa_family = AF_INET;
1923#if IS_ENABLED(CONFIG_IPV6)
1924 } else {
1925 loopback.sin6.sin6_addr = in6addr_loopback;
1926 loopback.sa.sa_family = AF_INET6;
1927#endif
1928 }
1929
9dcc71e1 1930 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1931 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1932
1933 u64_stats_update_begin(&tx_stats->syncp);
1934 tx_stats->tx_packets++;
ce6502a8 1935 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1936 u64_stats_update_end(&tx_stats->syncp);
1937
1938 if (netif_rx(skb) == NET_RX_SUCCESS) {
1939 u64_stats_update_begin(&rx_stats->syncp);
1940 rx_stats->rx_packets++;
ce6502a8 1941 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1942 u64_stats_update_end(&rx_stats->syncp);
1943 } else {
ce6502a8 1944 dev->stats.rx_dropped++;
9dcc71e1
SS		 1945 }
1946}
1947
4ad16930
SH		 1948static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1949 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1950{
d71785ff 1951 struct dst_cache *dst_cache;
3093fbe7 1952 struct ip_tunnel_info *info;
d342894c 1953 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1954 struct sock *sk;
e4c7ed41 1955 struct rtable *rt = NULL;
d342894c 1956 const struct iphdr *old_iph;
e4c7ed41 1957 union vxlan_addr *dst;
ee122c79
TG		 1958 union vxlan_addr remote_ip;
1959 struct vxlan_metadata _md;
1960 struct vxlan_metadata *md = &_md;
e4c7ed41 1961 __be16 src_port = 0, dst_port;
e7f70af1 1962 __be32 vni, label;
d342894c 1963 __be16 df = 0;
1964 __u8 tos, ttl;
0e6fbc5b 1965 int err;
ee122c79 1966 u32 flags = vxlan->flags;
b4ed5cad 1967 bool udp_sum = false;
f491e56d 1968 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1969
61adedf3 1970 info = skb_tunnel_info(skb);
3093fbe7 1971
ee122c79 1972 if (rdst) {
0dfbdf41 1973 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1974 vni = rdst->remote_vni;
1975 dst = &rdst->remote_ip;
d71785ff 1976 dst_cache = &rdst->dst_cache;
ee122c79
TG		 1977 } else {
1978 if (!info) {
1979 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1980 dev->name);
1981 goto drop;
1982 }
0dfbdf41 1983 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
54bfd872 1984 vni = vxlan_tun_id_to_vni(info->key.tun_id);
b1be00a6
JB		 1985 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1986 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1987 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1988 else
1989 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1990 dst = &remote_ip;
d71785ff 1991 dst_cache = &info->dst_cache;
ee122c79 1992 }
e4f67add 1993
e4c7ed41 1994 if (vxlan_addr_any(dst)) {
e4f67add 1995 if (did_rsc) {
e4f67add 1996 /* short-circuited back to local bridge */
9dcc71e1 1997 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1998 return;
e4f67add 1999 }
ef59febe 2000 goto drop;
e4f67add 2001 }
ef59febe 2002
d342894c 2003 old_iph = ip_hdr(skb);
2004
0dfbdf41 2005 ttl = vxlan->cfg.ttl;
e4c7ed41 2006 if (!ttl && vxlan_addr_multicast(dst))
d342894c 2007 ttl = 1;
2008
0dfbdf41 2009 tos = vxlan->cfg.tos;
d342894c 2010 if (tos == 1)
206aaafc 2011 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 2012
e7f70af1 2013 label = vxlan->cfg.label;
0dfbdf41
TG		 2014 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2015 vxlan->cfg.port_max, true);
d342894c 2016
a725e514 2017 if (info) {
a725e514
JB		 2018 ttl = info->key.ttl;
2019 tos = info->key.tos;
e7f70af1 2020 label = info->key.label;
b4ed5cad 2021 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB		 2022
2023 if (info->options_len)
4c222798 2024 md = ip_tunnel_info_opts(info);
a725e514
JB		 2025 } else {
2026 md->gbp = skb->mark;
2027 }
2028
e4c7ed41 2029 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB		 2030 __be32 saddr;
2031
b1be00a6
JB		 2032 if (!vxlan->vn4_sock)
2033 goto drop;
2034 sk = vxlan->vn4_sock->sock->sk;
2035
1a8496ba
JB		 2036 rt = vxlan_get_route(vxlan, skb,
2037 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 2038 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 2039 dst_cache, info);
e4c7ed41
CW		 2040 if (IS_ERR(rt)) {
2041 netdev_dbg(dev, "no route to %pI4\n",
2042 &dst->sin.sin_addr.s_addr);
2043 dev->stats.tx_carrier_errors++;
2044 goto tx_error;
2045 }
d342894c 2046
e4c7ed41
CW		 2047 if (rt->dst.dev == dev) {
2048 netdev_dbg(dev, "circular route to %pI4\n",
2049 &dst->sin.sin_addr.s_addr);
2050 dev->stats.collisions++;
fffc15a5 2051 goto rt_tx_error;
e4c7ed41
CW		 2052 }
2053
2054 /* Bypass encapsulation if the destination is local */
2055 if (rt->rt_flags & RTCF_LOCAL &&
2056 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2057 struct vxlan_dev *dst_vxlan;
2058
2059 ip_rt_put(rt);
19ca9fc1 2060 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2061 dst->sa.sa_family, dst_port,
2062 vxlan->flags);
e4c7ed41
CW		 2063 if (!dst_vxlan)
2064 goto tx_error;
2065 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2066 return;
2067 }
2068
6ceb31ca
AD		 2069 if (!info)
2070 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2071 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2072 df = htons(IP_DF);
2073
e4c7ed41
CW		 2074 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2075 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d 2076 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
54bfd872 2077 vni, md, flags, udp_sum);
f491e56d
JB		 2078 if (err < 0)
2079 goto xmit_tx_error;
2080
2081 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2082 dst->sin.sin_addr.s_addr, tos, ttl, df,
2083 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2084#if IS_ENABLED(CONFIG_IPV6)
2085 } else {
e4c7ed41 2086 struct dst_entry *ndst;
e5d4b29f 2087 struct in6_addr saddr;
6f264e47 2088 u32 rt6i_flags;
e4c7ed41 2089
b1be00a6
JB		 2090 if (!vxlan->vn6_sock)
2091 goto drop;
2092 sk = vxlan->vn6_sock->sock->sk;
2093
e5d4b29f 2094 ndst = vxlan6_get_route(vxlan, skb,
1400615d 2095 rdst ? rdst->remote_ifindex : 0, tos,
e7f70af1 2096 label, &dst->sin6.sin6_addr, &saddr,
db3c6139 2097 dst_cache, info);
e5d4b29f 2098 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2099 netdev_dbg(dev, "no route to %pI6\n",
2100 &dst->sin6.sin6_addr);
2101 dev->stats.tx_carrier_errors++;
9dcc71e1 2102 goto tx_error;
e4c7ed41 2103 }
d342894c 2104
e4c7ed41
CW		 2105 if (ndst->dev == dev) {
2106 netdev_dbg(dev, "circular route to %pI6\n",
2107 &dst->sin6.sin6_addr);
2108 dst_release(ndst);
2109 dev->stats.collisions++;
2110 goto tx_error;
2111 }
0e6fbc5b 2112
e4c7ed41 2113 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2114 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2115 if (rt6i_flags & RTF_LOCAL &&
2116 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2117 struct vxlan_dev *dst_vxlan;
2118
2119 dst_release(ndst);
19ca9fc1 2120 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2121 dst->sa.sa_family, dst_port,
2122 vxlan->flags);
e4c7ed41
CW		 2123 if (!dst_vxlan)
2124 goto tx_error;
2125 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2126 return;
2127 }
49560532 2128
b4ed5cad
JB		 2129 if (!info)
2130 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2131
1400615d 2132 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2133 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2134 skb_scrub_packet(skb, xnet);
2135 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2136 vni, md, flags, udp_sum);
f491e56d
JB		 2137 if (err < 0) {
2138 dst_release(ndst);
2139 return;
2140 }
2141 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
e7f70af1
DB		 2142 &saddr, &dst->sin6.sin6_addr, tos, ttl,
2143 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2144#endif
2145 }
0e6fbc5b 2146
4ad16930 2147 return;
d342894c 2148
2149drop:
2150 dev->stats.tx_dropped++;
2151 goto tx_free;
2152
f491e56d
JB		 2153xmit_tx_error:
2154 /* skb is already freed. */
2155 skb = NULL;
49560532
PS		 2156rt_tx_error:
2157 ip_rt_put(rt);
d342894c 2158tx_error:
2159 dev->stats.tx_errors++;
2160tx_free:
2161 dev_kfree_skb(skb);
d342894c 2162}
2163
6681712d
DS		 2164/* Transmit local packets over Vxlan
2165 *
2166 * Outer IP header inherits ECN and DF from inner header.
2167 * Outer UDP destination is the VXLAN assigned port.
2168 * source port is based on hash of flow
2169 */
2170static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2171{
2172 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2173 const struct ip_tunnel_info *info;
6681712d
DS		 2174 struct ethhdr *eth;
2175 bool did_rsc = false;
8f646c92 2176 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2177 struct vxlan_fdb *f;
6681712d 2178
61adedf3 2179 info = skb_tunnel_info(skb);
3093fbe7 2180
6681712d 2181 skb_reset_mac_header(skb);
6681712d 2182
47e5d1b0
JB		 2183 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2184 if (info && info->mode & IP_TUNNEL_INFO_TX)
2185 vxlan_xmit_one(skb, dev, NULL, false);
2186 else
2187 kfree_skb(skb);
2188 return NETDEV_TX_OK;
2189 }
2190
2191 if (vxlan->flags & VXLAN_F_PROXY) {
2192 eth = eth_hdr(skb);
f564f45c
CW		 2193 if (ntohs(eth->h_proto) == ETH_P_ARP)
2194 return arp_reduce(dev, skb);
2195#if IS_ENABLED(CONFIG_IPV6)
2196 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2197 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2198 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2199 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2200 struct nd_msg *msg;
2201
2202 msg = (struct nd_msg *)skb_transport_header(skb);
2203 if (msg->icmph.icmp6_code == 0 &&
2204 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2205 return neigh_reduce(dev, skb);
2206 }
2207#endif
2208 }
6681712d 2209
47e5d1b0 2210 eth = eth_hdr(skb);
6681712d 2211 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2212 did_rsc = false;
2213
2214 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2215 (ntohs(eth->h_proto) == ETH_P_IP ||
2216 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2217 did_rsc = route_shortcircuit(dev, skb);
2218 if (did_rsc)
2219 f = vxlan_find_mac(vxlan, eth->h_dest);
2220 }
2221
6681712d 2222 if (f == NULL) {
afbd8bae
MR		 2223 f = vxlan_find_mac(vxlan, all_zeros_mac);
2224 if (f == NULL) {
2225 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2226 !is_multicast_ether_addr(eth->h_dest))
2227 vxlan_fdb_miss(vxlan, eth->h_dest);
2228
2229 dev->stats.tx_dropped++;
8f646c92 2230 kfree_skb(skb);
afbd8bae
MR		 2231 return NETDEV_TX_OK;
2232 }
2233 }
6681712d 2234
afbd8bae
MR		 2235 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2236 struct sk_buff *skb1;
6681712d 2237
8f646c92
ED		 2238 if (!fdst) {
2239 fdst = rdst;
2240 continue;
2241 }
afbd8bae
MR		 2242 skb1 = skb_clone(skb, GFP_ATOMIC);
2243 if (skb1)
2244 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2245 }
2246
8f646c92
ED		 2247 if (fdst)
2248 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2249 else
2250 kfree_skb(skb);
4ad16930 2251 return NETDEV_TX_OK;
6681712d
DS		 2252}
2253
d342894c 2254/* Walk the forwarding table and purge stale entries */
2255static void vxlan_cleanup(unsigned long arg)
2256{
2257 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2258 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2259 unsigned int h;
2260
2261 if (!netif_running(vxlan->dev))
2262 return;
2263
d342894c 2264 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2265 struct hlist_node *p, *n;
14e1d0fa
SD		 2266
2267 spin_lock_bh(&vxlan->hash_lock);
d342894c 2268 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2269 struct vxlan_fdb *f
2270 = container_of(p, struct vxlan_fdb, hlist);
2271 unsigned long timeout;
2272
3c172868 2273 if (f->state & NUD_PERMANENT)
d342894c 2274 continue;
2275
0dfbdf41 2276 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2277 if (time_before_eq(timeout, jiffies)) {
2278 netdev_dbg(vxlan->dev,
2279 "garbage collect %pM\n",
2280 f->eth_addr);
2281 f->state = NUD_STALE;
2282 vxlan_fdb_destroy(vxlan, f);
2283 } else if (time_before(timeout, next_timer))
2284 next_timer = timeout;
2285 }
14e1d0fa 2286 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2287 }
d342894c 2288
2289 mod_timer(&vxlan->age_timer, next_timer);
2290}
2291
9c2e24e1
PS		 2292static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2293{
56ef9c90 2294 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2295 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2296
56ef9c90 2297 spin_lock(&vn->sock_lock);
9c2e24e1 2298 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2299 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2300}
2301
d342894c 2302/* Setup stats when device is created */
2303static int vxlan_init(struct net_device *dev)
2304{
1c213bd2 2305 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2306 if (!dev->tstats)
d342894c 2307 return -ENOMEM;
2308
2309 return 0;
2310}
2311
ba609e9b 2312static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2313{
2314 struct vxlan_fdb *f;
2315
2316 spin_lock_bh(&vxlan->hash_lock);
2317 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2318 if (f)
2319 vxlan_fdb_destroy(vxlan, f);
2320 spin_unlock_bh(&vxlan->hash_lock);
2321}
2322
ebf4063e
SH		 2323static void vxlan_uninit(struct net_device *dev)
2324{
2325 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2326
ba609e9b 2327 vxlan_fdb_delete_default(vxlan);
afbd8bae 2328
ebf4063e
SH		 2329 free_percpu(dev->tstats);
2330}
2331
d342894c 2332/* Start ageing timer and join group when device is brought up */
2333static int vxlan_open(struct net_device *dev)
2334{
2335 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2336 int ret;
56ef9c90 2337
205f356d
JB		 2338 ret = vxlan_sock_add(vxlan);
2339 if (ret < 0)
2340 return ret;
d342894c 2341
79d4a94f 2342 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2343 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2344 if (ret == -EADDRINUSE)
2345 ret = 0;
56ef9c90 2346 if (ret) {
205f356d 2347 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2348 return ret;
2349 }
d342894c 2350 }
2351
0dfbdf41 2352 if (vxlan->cfg.age_interval)
d342894c 2353 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2354
56ef9c90 2355 return ret;
d342894c 2356}
2357
2358/* Purge the forwarding table */
2359static void vxlan_flush(struct vxlan_dev *vxlan)
2360{
31fec5aa 2361 unsigned int h;
d342894c 2362
2363 spin_lock_bh(&vxlan->hash_lock);
2364 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2365 struct hlist_node *p, *n;
2366 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2367 struct vxlan_fdb *f
2368 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2369 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2370 if (!is_zero_ether_addr(f->eth_addr))
2371 vxlan_fdb_destroy(vxlan, f);
d342894c 2372 }
2373 }
2374 spin_unlock_bh(&vxlan->hash_lock);
2375}
2376
2377/* Cleanup timer and forwarding table on shutdown */
2378static int vxlan_stop(struct net_device *dev)
2379{
2380 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2381 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2382 int ret = 0;
d342894c 2383
24c0e683 2384 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2385 !vxlan_group_used(vn, vxlan))
56ef9c90 2386 ret = vxlan_igmp_leave(vxlan);
d342894c 2387
2388 del_timer_sync(&vxlan->age_timer);
2389
2390 vxlan_flush(vxlan);
205f356d 2391 vxlan_sock_release(vxlan);
d342894c 2392
56ef9c90 2393 return ret;
d342894c 2394}
2395
d342894c 2396/* Stub, nothing needs to be done. */
2397static void vxlan_set_multicast_list(struct net_device *dev)
2398{
2399}
2400
72564b59
DW		 2401static int __vxlan_change_mtu(struct net_device *dev,
2402 struct net_device *lowerdev,
2403 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2404{
72564b59 2405 int max_mtu = IP_MAX_MTU;
345010b5 2406
72564b59
DW		 2407 if (lowerdev)
2408 max_mtu = lowerdev->mtu;
345010b5
DB		 2409
2410 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2411 max_mtu -= VXLAN6_HEADROOM;
345010b5 2412 else
72564b59 2413 max_mtu -= VXLAN_HEADROOM;
345010b5 2414
72564b59 2415 if (new_mtu < 68)
345010b5
DB		 2416 return -EINVAL;
2417
72564b59
DW		 2418 if (new_mtu > max_mtu) {
2419 if (strict)
2420 return -EINVAL;
2421
2422 new_mtu = max_mtu;
2423 }
2424
345010b5
DB		 2425 dev->mtu = new_mtu;
2426 return 0;
2427}
2428
72564b59
DW		 2429static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2430{
2431 struct vxlan_dev *vxlan = netdev_priv(dev);
2432 struct vxlan_rdst *dst = &vxlan->default_dst;
2433 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2434 dst->remote_ifindex);
2435 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2436}
2437
fc4099f1
PS		 2438static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2439{
2440 struct vxlan_dev *vxlan = netdev_priv(dev);
2441 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2442 __be16 sport, dport;
2443
2444 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2445 vxlan->cfg.port_max, true);
2446 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2447
239e944f 2448 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB		 2449 struct rtable *rt;
2450
239e944f
JB		 2451 if (!vxlan->vn4_sock)
2452 return -EINVAL;
1a8496ba
JB		 2453 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2454 info->key.u.ipv4.dst,
0c1d70af 2455 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2456 if (IS_ERR(rt))
2457 return PTR_ERR(rt);
2458 ip_rt_put(rt);
239e944f
JB		 2459 } else {
2460#if IS_ENABLED(CONFIG_IPV6)
2461 struct dst_entry *ndst;
2462
2463 if (!vxlan->vn6_sock)
2464 return -EINVAL;
1400615d 2465 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
e7f70af1 2466 info->key.label, &info->key.u.ipv6.dst,
db3c6139 2467 &info->key.u.ipv6.src, NULL, info);
239e944f
JB		 2468 if (IS_ERR(ndst))
2469 return PTR_ERR(ndst);
2470 dst_release(ndst);
239e944f
JB		 2471#else /* !CONFIG_IPV6 */
2472 return -EPFNOSUPPORT;
2473#endif
2474 }
1a8496ba
JB		 2475 info->key.tp_src = sport;
2476 info->key.tp_dst = dport;
239e944f 2477 return 0;
fc4099f1
PS		 2478}
2479
0c867c9b 2480static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2481 .ndo_init = vxlan_init,
ebf4063e 2482 .ndo_uninit = vxlan_uninit,
d342894c 2483 .ndo_open = vxlan_open,
2484 .ndo_stop = vxlan_stop,
2485 .ndo_start_xmit = vxlan_xmit,
e8171045 2486 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2487 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2488 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2489 .ndo_validate_addr = eth_validate_addr,
2490 .ndo_set_mac_address = eth_mac_addr,
2491 .ndo_fdb_add = vxlan_fdb_add,
2492 .ndo_fdb_del = vxlan_fdb_delete,
2493 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2494 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2495};
2496
e1e5314d
JB		 2497static const struct net_device_ops vxlan_netdev_raw_ops = {
2498 .ndo_init = vxlan_init,
2499 .ndo_uninit = vxlan_uninit,
2500 .ndo_open = vxlan_open,
2501 .ndo_stop = vxlan_stop,
2502 .ndo_start_xmit = vxlan_xmit,
2503 .ndo_get_stats64 = ip_tunnel_get_stats64,
2504 .ndo_change_mtu = vxlan_change_mtu,
2505 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2506};
2507
d342894c 2508/* Info for udev, that this is a virtual tunnel endpoint */
2509static struct device_type vxlan_type = {
2510 .name = "vxlan",
2511};
2512
53cf5275 2513/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2514 * supply the listening VXLAN udp ports. Callers are expected
2515 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2516 */
2517void vxlan_get_rx_port(struct net_device *dev)
2518{
2519 struct vxlan_sock *vs;
2520 struct net *net = dev_net(dev);
2521 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2522 sa_family_t sa_family;
35e42379
JG		 2523 __be16 port;
2524 unsigned int i;
53cf5275
JG		 2525
2526 spin_lock(&vn->sock_lock);
2527 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2528 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2529 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2530 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2531 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2532 port);
2533 }
2534 }
2535 spin_unlock(&vn->sock_lock);
2536}
2537EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2538
d342894c 2539/* Initialize the device structure. */
2540static void vxlan_setup(struct net_device *dev)
2541{
2542 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2543 unsigned int h;
d342894c 2544
ebf4063e 2545 dev->destructor = free_netdev;
d342894c 2546 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2547
d342894c 2548 dev->features |= NETIF_F_LLTX;
d6727fe3 2549 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2550 dev->features |= NETIF_F_RXCSUM;
05c0db08 2551 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2552
1eaa8178
PS		 2553 dev->vlan_features = dev->features;
2554 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2555 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2556 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2557 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2558 netif_keep_dst(dev);
0c867c9b 2559 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2560
553675fb 2561 INIT_LIST_HEAD(&vxlan->next);
d342894c 2562 spin_lock_init(&vxlan->hash_lock);
2563
2564 init_timer_deferrable(&vxlan->age_timer);
2565 vxlan->age_timer.function = vxlan_cleanup;
2566 vxlan->age_timer.data = (unsigned long) vxlan;
2567
0dfbdf41 2568 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2569
d342894c 2570 vxlan->dev = dev;
2571
58ce31cc
TH		 2572 gro_cells_init(&vxlan->gro_cells, dev);
2573
d342894c 2574 for (h = 0; h < FDB_HASH_SIZE; ++h)
2575 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2576}
2577
0c867c9b
JB		 2578static void vxlan_ether_setup(struct net_device *dev)
2579{
2580 eth_hw_addr_random(dev);
2581 ether_setup(dev);
2582 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2583 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2584 dev->netdev_ops = &vxlan_netdev_ether_ops;
2585}
2586
e1e5314d
JB		 2587static void vxlan_raw_setup(struct net_device *dev)
2588{
2589 dev->type = ARPHRD_NONE;
2590 dev->hard_header_len = 0;
2591 dev->addr_len = 0;
2592 dev->mtu = ETH_DATA_LEN;
2593 dev->tx_queue_len = 1000;
2594 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2595 dev->netdev_ops = &vxlan_netdev_raw_ops;
2596}
2597
d342894c 2598static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2599 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2600 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2601 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2602 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2603 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2604 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2605 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2606 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2607 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2608 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2609 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2610 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2611 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2612 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2613 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2614 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2615 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2616 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2617 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2618 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2619 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2620 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2621 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2622 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2623 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2624 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2625 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2626};
2627
2628static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2629{
2630 if (tb[IFLA_ADDRESS]) {
2631 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2632 pr_debug("invalid link address (not ethernet)\n");
2633 return -EINVAL;
2634 }
2635
2636 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2637 pr_debug("invalid all zero ethernet address\n");
2638 return -EADDRNOTAVAIL;
2639 }
2640 }
2641
2642 if (!data)
2643 return -EINVAL;
2644
2645 if (data[IFLA_VXLAN_ID]) {
2646 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2647 if (id >= VXLAN_VID_MASK)
2648 return -ERANGE;
2649 }
2650
05f47d69 2651 if (data[IFLA_VXLAN_PORT_RANGE]) {
2652 const struct ifla_vxlan_port_range *p
2653 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2654
2655 if (ntohs(p->high) < ntohs(p->low)) {
2656 pr_debug("port range %u .. %u not valid\n",
2657 ntohs(p->low), ntohs(p->high));
2658 return -EINVAL;
2659 }
2660 }
2661
d342894c 2662 return 0;
2663}
2664
1b13c97f
YB		 2665static void vxlan_get_drvinfo(struct net_device *netdev,
2666 struct ethtool_drvinfo *drvinfo)
2667{
2668 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2669 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2670}
2671
2672static const struct ethtool_ops vxlan_ethtool_ops = {
2673 .get_drvinfo = vxlan_get_drvinfo,
2674 .get_link = ethtool_op_get_link,
2675};
2676
553675fb 2677static void vxlan_del_work(struct work_struct *work)
2678{
2679 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2680 udp_tunnel_sock_release(vs->sock);
553675fb 2681 kfree_rcu(vs, rcu);
2682}
2683
3ee64f39
TH		 2684static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2685 __be16 port, u32 flags)
553675fb 2686{
e4c7ed41 2687 struct socket *sock;
3ee64f39
TH		 2688 struct udp_port_cfg udp_conf;
2689 int err;
e4c7ed41 2690
3ee64f39 2691 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2692
3ee64f39
TH		 2693 if (ipv6) {
2694 udp_conf.family = AF_INET6;
3ee64f39 2695 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2696 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2697 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2698 } else {
2699 udp_conf.family = AF_INET;
e4c7ed41 2700 }
e4c7ed41 2701
3ee64f39 2702 udp_conf.local_udp_port = port;
553675fb 2703
3ee64f39
TH		 2704 /* Open UDP socket */
2705 err = udp_sock_create(net, &udp_conf, &sock);
2706 if (err < 0)
2707 return ERR_PTR(err);
e4c7ed41 2708
39deb2c7 2709 return sock;
e4c7ed41
CW		 2710}
2711
2712/* Create new listen socket if needed */
b1be00a6
JB		 2713static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2714 __be16 port, u32 flags)
e4c7ed41
CW		 2715{
2716 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2717 struct vxlan_sock *vs;
2718 struct socket *sock;
e4c7ed41 2719 unsigned int h;
acbf74a7 2720 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2721
dc01e7d3 2722 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2723 if (!vs)
2724 return ERR_PTR(-ENOMEM);
2725
2726 for (h = 0; h < VNI_HASH_SIZE; ++h)
2727 INIT_HLIST_HEAD(&vs->vni_list[h]);
2728
2729 INIT_WORK(&vs->del_work, vxlan_del_work);
2730
3ee64f39 2731 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2732 if (IS_ERR(sock)) {
56ef9c90
MRL		 2733 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2734 PTR_ERR(sock));
553675fb 2735 kfree(vs);
e50fddc8 2736 return ERR_CAST(sock);
553675fb 2737 }
e4c7ed41
CW		 2738
2739 vs->sock = sock;
9c2e24e1 2740 atomic_set(&vs->refcnt, 1);
af33c1ad 2741 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2742
9c2e24e1
PS		 2743 spin_lock(&vn->sock_lock);
2744 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2745 vxlan_notify_add_rx_port(vs);
9c2e24e1 2746 spin_unlock(&vn->sock_lock);
553675fb 2747
2748 /* Mark socket as an encapsulation socket. */
5602c48c 2749 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2750 tunnel_cfg.sk_user_data = vs;
2751 tunnel_cfg.encap_type = 1;
f2d1968e 2752 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2753 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2754 tunnel_cfg.gro_receive = vxlan_gro_receive;
2755 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2756
2757 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2758
9c2e24e1
PS		 2759 return vs;
2760}
2761
b1be00a6 2762static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2763{
205f356d
JB		 2764 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2765 struct vxlan_sock *vs = NULL;
9c2e24e1 2766
205f356d 2767 if (!vxlan->cfg.no_share) {
56ef9c90 2768 spin_lock(&vn->sock_lock);
205f356d
JB		 2769 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2770 vxlan->cfg.dst_port, vxlan->flags);
2771 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2772 spin_unlock(&vn->sock_lock);
205f356d 2773 return -EBUSY;
56ef9c90
MRL		 2774 }
2775 spin_unlock(&vn->sock_lock);
2776 }
205f356d 2777 if (!vs)
b1be00a6
JB		 2778 vs = vxlan_socket_create(vxlan->net, ipv6,
2779 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2780 if (IS_ERR(vs))
2781 return PTR_ERR(vs);
b1be00a6
JB		 2782#if IS_ENABLED(CONFIG_IPV6)
2783 if (ipv6)
2784 vxlan->vn6_sock = vs;
2785 else
2786#endif
2787 vxlan->vn4_sock = vs;
205f356d
JB		 2788 vxlan_vs_add_dev(vs, vxlan);
2789 return 0;
553675fb 2790}
2791
b1be00a6
JB		 2792static int vxlan_sock_add(struct vxlan_dev *vxlan)
2793{
2794 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2795 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2796 int ret = 0;
2797
2798 vxlan->vn4_sock = NULL;
2799#if IS_ENABLED(CONFIG_IPV6)
2800 vxlan->vn6_sock = NULL;
2801 if (ipv6 || metadata)
2802 ret = __vxlan_sock_add(vxlan, true);
2803#endif
2804 if (!ret && (!ipv6 || metadata))
2805 ret = __vxlan_sock_add(vxlan, false);
2806 if (ret < 0)
2807 vxlan_sock_release(vxlan);
2808 return ret;
2809}
2810
0dfbdf41
TG		 2811static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2812 struct vxlan_config *conf)
d342894c 2813{
33564bbb 2814 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2815 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2816 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2817 unsigned short needed_headroom = ETH_HLEN;
d342894c 2818 int err;
e4c7ed41 2819 bool use_ipv6 = false;
0dfbdf41 2820 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2821 struct net_device *lowerdev = NULL;
d342894c 2822
e1e5314d
JB		 2823 if (conf->flags & VXLAN_F_GPE) {
2824 if (conf->flags & ~VXLAN_F_ALLOWED_GPE)
2825 return -EINVAL;
2826 /* For now, allow GPE only together with COLLECT_METADATA.
2827 * This can be relaxed later; in such case, the other side
2828 * of the PtP link will have to be provided.
2829 */
2830 if (!(conf->flags & VXLAN_F_COLLECT_METADATA))
2831 return -EINVAL;
2832
2833 vxlan_raw_setup(dev);
2834 } else {
2835 vxlan_ether_setup(dev);
2836 }
0c867c9b 2837
33564bbb 2838 vxlan->net = src_net;
f01ec1c0 2839
0dfbdf41 2840 dst->remote_vni = conf->vni;
d342894c 2841
0dfbdf41 2842 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2843
0dfbdf41
TG		 2844 /* Unless IPv6 is explicitly requested, assume IPv4 */
2845 if (!dst->remote_ip.sa.sa_family)
2846 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2847
0dfbdf41 2848 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2849 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2850 if (!IS_ENABLED(CONFIG_IPV6))
2851 return -EPFNOSUPPORT;
e4c7ed41 2852 use_ipv6 = true;
b1be00a6 2853 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2854 }
d342894c 2855
e7f70af1
DB		 2856 if (conf->label && !use_ipv6) {
2857 pr_info("label only supported in use with IPv6\n");
2858 return -EINVAL;
2859 }
2860
0dfbdf41 2861 if (conf->remote_ifindex) {
7e059158 2862 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2863 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2864
2865 if (!lowerdev) {
c7995c43 2866 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2867 return -ENODEV;
2868 }
d342894c 2869
e4c7ed41
CW		 2870#if IS_ENABLED(CONFIG_IPV6)
2871 if (use_ipv6) {
2872 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2873 if (idev && idev->cnf.disable_ipv6) {
2874 pr_info("IPv6 is disabled via sysctl\n");
2875 return -EPERM;
2876 }
e4c7ed41
CW		 2877 }
2878#endif
2879
0dfbdf41 2880 if (!conf->mtu)
e4c7ed41 2881 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2882
b1be00a6 2883 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2884 }
d342894c 2885
7e059158
DW		 2886 if (conf->mtu) {
2887 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2888 if (err)
2889 return err;
2890 }
2891
b1be00a6
JB		 2892 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2893 needed_headroom += VXLAN6_HEADROOM;
2894 else
2895 needed_headroom += VXLAN_HEADROOM;
2896 dev->needed_headroom = needed_headroom;
2897
0dfbdf41 2898 memcpy(&vxlan->cfg, conf, sizeof(*conf));
e1e5314d
JB		 2899 if (!vxlan->cfg.dst_port) {
2900 if (conf->flags & VXLAN_F_GPE)
2901 vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */
2902 else
2903 vxlan->cfg.dst_port = default_port;
2904 }
0dfbdf41
TG		 2905 vxlan->flags |= conf->flags;
2906
2907 if (!vxlan->cfg.age_interval)
2908 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2909
07b9b37c
ND		 2910 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2911 if (tmp->cfg.vni == conf->vni &&
2912 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2913 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2914 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2915 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2916 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2917 return -EEXIST;
07b9b37c 2918 }
0dfbdf41
TG		 2919
2920 dev->ethtool_ops = &vxlan_ethtool_ops;
2921
2922 /* create an fdb entry for a valid default destination */
2923 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2924 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2925 &vxlan->default_dst.remote_ip,
2926 NUD_REACHABLE|NUD_PERMANENT,
2927 NLM_F_EXCL|NLM_F_CREATE,
2928 vxlan->cfg.dst_port,
2929 vxlan->default_dst.remote_vni,
2930 vxlan->default_dst.remote_ifindex,
2931 NTF_SELF);
2932 if (err)
2933 return err;
2934 }
2935
2936 err = register_netdevice(dev);
2937 if (err) {
2938 vxlan_fdb_delete_default(vxlan);
2939 return err;
2940 }
2941
2942 list_add(&vxlan->next, &vn->vxlan_list);
2943
2944 return 0;
2945}
2946
2947struct net_device *vxlan_dev_create(struct net *net, const char *name,
2948 u8 name_assign_type, struct vxlan_config *conf)
2949{
2950 struct nlattr *tb[IFLA_MAX+1];
2951 struct net_device *dev;
2952 int err;
2953
2954 memset(&tb, 0, sizeof(tb));
2955
2956 dev = rtnl_create_link(net, name, name_assign_type,
2957 &vxlan_link_ops, tb);
2958 if (IS_ERR(dev))
2959 return dev;
2960
2961 err = vxlan_dev_configure(net, dev, conf);
2962 if (err < 0) {
2963 free_netdev(dev);
2964 return ERR_PTR(err);
2965 }
2966
2967 return dev;
2968}
2969EXPORT_SYMBOL_GPL(vxlan_dev_create);
2970
2971static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2972 struct nlattr *tb[], struct nlattr *data[])
2973{
2974 struct vxlan_config conf;
2975 int err;
2976
0dfbdf41 2977 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2978
2979 if (data[IFLA_VXLAN_ID])
54bfd872 2980 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG		 2981
2982 if (data[IFLA_VXLAN_GROUP]) {
2983 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2984 } else if (data[IFLA_VXLAN_GROUP6]) {
2985 if (!IS_ENABLED(CONFIG_IPV6))
2986 return -EPFNOSUPPORT;
2987
2988 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2989 conf.remote_ip.sa.sa_family = AF_INET6;
2990 }
2991
2992 if (data[IFLA_VXLAN_LOCAL]) {
2993 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2994 conf.saddr.sa.sa_family = AF_INET;
2995 } else if (data[IFLA_VXLAN_LOCAL6]) {
2996 if (!IS_ENABLED(CONFIG_IPV6))
2997 return -EPFNOSUPPORT;
2998
2999 /* TODO: respect scope id */
3000 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3001 conf.saddr.sa.sa_family = AF_INET6;
3002 }
3003
3004 if (data[IFLA_VXLAN_LINK])
3005 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3006
d342894c 3007 if (data[IFLA_VXLAN_TOS])
0dfbdf41 3008 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3009
afb97186 3010 if (data[IFLA_VXLAN_TTL])
0dfbdf41 3011 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3012
e7f70af1
DB		 3013 if (data[IFLA_VXLAN_LABEL])
3014 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3015 IPV6_FLOWLABEL_MASK;
3016
d342894c 3017 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 3018 conf.flags |= VXLAN_F_LEARN;
d342894c 3019
3020 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 3021 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 3022
e4f67add 3023 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 3024 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 3025
3026 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 3027 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 3028
3029 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 3030 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 3031
3032 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 3033 conf.flags |= VXLAN_F_L3MISS;
e4f67add 3034
d342894c 3035 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 3036 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 3037
f8a9b1bc
AS		 3038 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
3039 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3040 conf.flags |= VXLAN_F_COLLECT_METADATA;
3041
05f47d69 3042 if (data[IFLA_VXLAN_PORT_RANGE]) {
3043 const struct ifla_vxlan_port_range *p
3044 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 3045 conf.port_min = ntohs(p->low);
3046 conf.port_max = ntohs(p->high);
05f47d69 3047 }
3048
823aa873 3049 if (data[IFLA_VXLAN_PORT])
0dfbdf41 3050 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 3051
6ceb31ca
AD		 3052 if (data[IFLA_VXLAN_UDP_CSUM] &&
3053 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3054 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH		 3055
3056 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
3057 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 3058 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 3059
3060 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
3061 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 3062 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 3063
dfd8645e
TH		 3064 if (data[IFLA_VXLAN_REMCSUM_TX] &&
3065 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 3066 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 3067
3068 if (data[IFLA_VXLAN_REMCSUM_RX] &&
3069 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 3070 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 3071
3511494c 3072 if (data[IFLA_VXLAN_GBP])
0dfbdf41 3073 conf.flags |= VXLAN_F_GBP;
3511494c 3074
e1e5314d
JB		 3075 if (data[IFLA_VXLAN_GPE])
3076 conf.flags |= VXLAN_F_GPE;
3077
0ace2ca8 3078 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 3079 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 3080
0dfbdf41
TG		 3081 err = vxlan_dev_configure(src_net, dev, &conf);
3082 switch (err) {
3083 case -ENODEV:
3084 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
3085 break;
1b13c97f 3086
0dfbdf41
TG		 3087 case -EPERM:
3088 pr_info("IPv6 is disabled via sysctl\n");
3089 break;
d342894c 3090
0dfbdf41 3091 case -EEXIST:
54bfd872 3092 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
0dfbdf41 3093 break;
e1e5314d
JB		 3094
3095 case -EINVAL:
3096 pr_info("unsupported combination of extensions\n");
3097 break;
afbd8bae
MR		 3098 }
3099
0dfbdf41 3100 return err;
d342894c 3101}
3102
3103static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3104{
3105 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3106 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 3107
fe5c3561 3108 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 3109 if (!hlist_unhashed(&vxlan->hlist))
3110 hlist_del_rcu(&vxlan->hlist);
fe5c3561 3111 spin_unlock(&vn->sock_lock);
3112
58ce31cc 3113 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3114 list_del(&vxlan->next);
d342894c 3115 unregister_netdevice_queue(dev, head);
3116}
3117
3118static size_t vxlan_get_size(const struct net_device *dev)
3119{
3120
3121 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3122 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3123 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3124 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3125 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3126 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3127 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3128 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3129 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3130 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3131 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3132 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3133 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3134 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3135 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3136 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3137 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3138 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3139 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3140 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3141 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3142 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3143 0;
3144}
3145
3146static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3147{
3148 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3149 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3150 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3151 .low = htons(vxlan->cfg.port_min),
3152 .high = htons(vxlan->cfg.port_max),
05f47d69 3153 };
d342894c 3154
54bfd872 3155 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3156 goto nla_put_failure;
3157
e4c7ed41
CW		 3158 if (!vxlan_addr_any(&dst->remote_ip)) {
3159 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3160 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3161 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3162 goto nla_put_failure;
3163#if IS_ENABLED(CONFIG_IPV6)
3164 } else {
930345ea
JB		 3165 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3166 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3167 goto nla_put_failure;
3168#endif
3169 }
3170 }
d342894c 3171
c7995c43 3172 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3173 goto nla_put_failure;
3174
0dfbdf41
TG		 3175 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3176 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3177 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3178 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3179 goto nla_put_failure;
3180#if IS_ENABLED(CONFIG_IPV6)
3181 } else {
930345ea 3182 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3183 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3184 goto nla_put_failure;
3185#endif
3186 }
3187 }
d342894c 3188
0dfbdf41
TG		 3189 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3190 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3191 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3192 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3193 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3194 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3195 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3196 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3197 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3198 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3199 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3200 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3201 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3202 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3203 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3204 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3205 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3206 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3207 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3208 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3209 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3210 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3211 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3212 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3213 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3214 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3215 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3216 goto nla_put_failure;
3217
05f47d69 3218 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3219 goto nla_put_failure;
3220
3511494c
TG		 3221 if (vxlan->flags & VXLAN_F_GBP &&
3222 nla_put_flag(skb, IFLA_VXLAN_GBP))
3223 goto nla_put_failure;
3224
e1e5314d
JB		 3225 if (vxlan->flags & VXLAN_F_GPE &&
3226 nla_put_flag(skb, IFLA_VXLAN_GPE))
3227 goto nla_put_failure;
3228
0ace2ca8
TH		 3229 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3230 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3231 goto nla_put_failure;
3232
d342894c 3233 return 0;
3234
3235nla_put_failure:
3236 return -EMSGSIZE;
3237}
3238
1728d4fa
ND		 3239static struct net *vxlan_get_link_net(const struct net_device *dev)
3240{
3241 struct vxlan_dev *vxlan = netdev_priv(dev);
3242
3243 return vxlan->net;
3244}
3245
d342894c 3246static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3247 .kind = "vxlan",
3248 .maxtype = IFLA_VXLAN_MAX,
3249 .policy = vxlan_policy,
3250 .priv_size = sizeof(struct vxlan_dev),
3251 .setup = vxlan_setup,
3252 .validate = vxlan_validate,
3253 .newlink = vxlan_newlink,
3254 .dellink = vxlan_dellink,
3255 .get_size = vxlan_get_size,
3256 .fill_info = vxlan_fill_info,
1728d4fa 3257 .get_link_net = vxlan_get_link_net,
d342894c 3258};
3259
acaf4e70
DB		 3260static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3261 struct net_device *dev)
3262{
3263 struct vxlan_dev *vxlan, *next;
3264 LIST_HEAD(list_kill);
3265
3266 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3267 struct vxlan_rdst *dst = &vxlan->default_dst;
3268
3269 /* In case we created vxlan device with carrier
3270 * and we loose the carrier due to module unload
3271 * we also need to remove vxlan device. In other
3272 * cases, it's not necessary and remote_ifindex
3273 * is 0 here, so no matches.
3274 */
3275 if (dst->remote_ifindex == dev->ifindex)
3276 vxlan_dellink(vxlan->dev, &list_kill);
3277 }
3278
3279 unregister_netdevice_many(&list_kill);
3280}
3281
3282static int vxlan_lowerdev_event(struct notifier_block *unused,
3283 unsigned long event, void *ptr)
3284{
3285 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3286 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3287
783c1463 3288 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3289 vxlan_handle_lowerdev_unregister(vn, dev);
3290
3291 return NOTIFY_DONE;
3292}
3293
3294static struct notifier_block vxlan_notifier_block __read_mostly = {
3295 .notifier_call = vxlan_lowerdev_event,
3296};
3297
d342894c 3298static __net_init int vxlan_init_net(struct net *net)
3299{
3300 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3301 unsigned int h;
d342894c 3302
553675fb 3303 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3304 spin_lock_init(&vn->sock_lock);
d342894c 3305
553675fb 3306 for (h = 0; h < PORT_HASH_SIZE; ++h)
3307 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3308
3309 return 0;
3310}
3311
f01ec1c0
ND		 3312static void __net_exit vxlan_exit_net(struct net *net)
3313{
3314 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3315 struct vxlan_dev *vxlan, *next;
3316 struct net_device *dev, *aux;
3317 LIST_HEAD(list);
3318
3319 rtnl_lock();
3320 for_each_netdev_safe(net, dev, aux)
3321 if (dev->rtnl_link_ops == &vxlan_link_ops)
3322 unregister_netdevice_queue(dev, &list);
3323
3324 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3325 /* If vxlan->dev is in the same netns, it has already been added
3326 * to the list by the previous loop.
3327 */
58ce31cc
TH		 3328 if (!net_eq(dev_net(vxlan->dev), net)) {
3329 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3330 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3331 }
f01ec1c0
ND		 3332 }
3333
3334 unregister_netdevice_many(&list);
3335 rtnl_unlock();
3336}
3337
d342894c 3338static struct pernet_operations vxlan_net_ops = {
3339 .init = vxlan_init_net,
f01ec1c0 3340 .exit = vxlan_exit_net,
d342894c 3341 .id = &vxlan_net_id,
3342 .size = sizeof(struct vxlan_net),
3343};
3344
3345static int __init vxlan_init_module(void)
3346{
3347 int rc;
3348
758c57d1
SH		 3349 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3350 if (!vxlan_wq)
3351 return -ENOMEM;
3352
d342894c 3353 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3354
783c1463 3355 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3356 if (rc)
3357 goto out1;
3358
acaf4e70 3359 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3360 if (rc)
3361 goto out2;
3362
acaf4e70
DB		 3363 rc = rtnl_link_register(&vxlan_link_ops);
3364 if (rc)
3365 goto out3;
d342894c 3366
acaf4e70
DB		 3367 return 0;
3368out3:
3369 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3370out2:
783c1463 3371 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3372out1:
758c57d1 3373 destroy_workqueue(vxlan_wq);
d342894c 3374 return rc;
3375}
7332a13b 3376late_initcall(vxlan_init_module);
d342894c 3377
3378static void __exit vxlan_cleanup_module(void)
3379{
b7153984 3380 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3381 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3382 destroy_workqueue(vxlan_wq);
783c1463
DB		 3383 unregister_pernet_subsys(&vxlan_net_ops);
3384 /* rcu_barrier() is called by netns */
d342894c 3385}
3386module_exit(vxlan_cleanup_module);
3387
3388MODULE_LICENSE("GPL");
3389MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3390MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3391MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3392MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror