projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: synchronously and race-free destruction of vxlan sockets
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
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;
101
ee122c79
TG		 102static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
103{
e7030878
TG		 104 return vs->flags & VXLAN_F_COLLECT_METADATA ||
105 ip_tunnel_collect_metadata();
ee122c79
TG		 106}
107
e4c7ed41
CW		 108#if IS_ENABLED(CONFIG_IPV6)
109static inline
110bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
111{
f0ef3126
JB		 112 if (a->sa.sa_family != b->sa.sa_family)
113 return false;
114 if (a->sa.sa_family == AF_INET6)
115 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
116 else
117 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 118}
119
120static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
121{
f0ef3126
JB		 122 if (ipa->sa.sa_family == AF_INET6)
123 return ipv6_addr_any(&ipa->sin6.sin6_addr);
124 else
125 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 126}
127
128static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
129{
f0ef3126
JB		 130 if (ipa->sa.sa_family == AF_INET6)
131 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
132 else
133 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 134}
135
136static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
137{
f0ef3126 138 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 139 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 140 ip->sa.sa_family = AF_INET6;
141 return 0;
142 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 143 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 144 ip->sa.sa_family = AF_INET;
145 return 0;
146 } else {
147 return -EAFNOSUPPORT;
148 }
e4c7ed41
CW		 149}
150
151static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 152 const union vxlan_addr *ip)
e4c7ed41 153{
f0ef3126 154 if (ip->sa.sa_family == AF_INET6)
930345ea 155 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 156 else
930345ea 157 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 158}
159
160#else /* !CONFIG_IPV6 */
161
162static inline
163bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
164{
f0ef3126 165 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 166}
167
168static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
169{
f0ef3126 170 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 171}
172
173static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
174{
f0ef3126 175 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 176}
177
178static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
179{
f0ef3126
JB		 180 if (nla_len(nla) >= sizeof(struct in6_addr)) {
181 return -EAFNOSUPPORT;
182 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 183 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 184 ip->sa.sa_family = AF_INET;
185 return 0;
186 } else {
187 return -EAFNOSUPPORT;
188 }
e4c7ed41
CW		 189}
190
191static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 192 const union vxlan_addr *ip)
e4c7ed41 193{
930345ea 194 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 195}
196#endif
197
553675fb 198/* Virtual Network hash table head */
54bfd872 199static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 200{
54bfd872 201 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 202}
203
204/* Socket hash table head */
205static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 206{
207 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
208
553675fb 209 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
210}
211
3e61aa8f
SH		 212/* First remote destination for a forwarding entry.
213 * Guaranteed to be non-NULL because remotes are never deleted.
214 */
5ca5461c 215static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 216{
5ca5461c 217 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
218}
219
220static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
221{
222 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 223}
224
ac5132d1
TG		 225/* Find VXLAN socket based on network namespace, address family and UDP port
226 * and enabled unshareable flags.
227 */
228static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
229 __be16 port, u32 flags)
553675fb 230{
231 struct vxlan_sock *vs;
af33c1ad
TH		 232
233 flags &= VXLAN_F_RCV_FLAGS;
553675fb 234
235 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 236 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 237 vxlan_get_sk_family(vs) == family &&
af33c1ad 238 vs->flags == flags)
553675fb 239 return vs;
240 }
241 return NULL;
d342894c 242}
243
54bfd872 244static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 245{
246 struct vxlan_dev *vxlan;
d342894c 247
b9167b2e
JB		 248 /* For flow based devices, map all packets to VNI 0 */
249 if (vs->flags & VXLAN_F_COLLECT_METADATA)
250 vni = 0;
251
54bfd872
JB		 252 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
253 if (vxlan->default_dst.remote_vni == vni)
d342894c 254 return vxlan;
255 }
256
257 return NULL;
258}
259
5cfccc5a 260/* Look up VNI in a per net namespace table */
54bfd872 261static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 262 sa_family_t family, __be16 port,
263 u32 flags)
5cfccc5a
PS		 264{
265 struct vxlan_sock *vs;
266
ac5132d1 267 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 268 if (!vs)
269 return NULL;
270
54bfd872 271 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 272}
273
d342894c 274/* Fill in neighbour message in skbuff. */
275static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 276 const struct vxlan_fdb *fdb,
277 u32 portid, u32 seq, int type, unsigned int flags,
278 const struct vxlan_rdst *rdst)
d342894c 279{
280 unsigned long now = jiffies;
281 struct nda_cacheinfo ci;
282 struct nlmsghdr *nlh;
283 struct ndmsg *ndm;
e4f67add 284 bool send_ip, send_eth;
d342894c 285
286 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
287 if (nlh == NULL)
288 return -EMSGSIZE;
289
290 ndm = nlmsg_data(nlh);
291 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 292
293 send_eth = send_ip = true;
294
295 if (type == RTM_GETNEIGH) {
296 ndm->ndm_family = AF_INET;
e4c7ed41 297 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 298 send_eth = !is_zero_ether_addr(fdb->eth_addr);
299 } else
300 ndm->ndm_family = AF_BRIDGE;
d342894c 301 ndm->ndm_state = fdb->state;
302 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 303 ndm->ndm_flags = fdb->flags;
545469f7 304 ndm->ndm_type = RTN_UNICAST;
d342894c 305
193523bf 306 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 307 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 308 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 309 goto nla_put_failure;
310
e4f67add 311 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 312 goto nla_put_failure;
313
e4c7ed41 314 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 315 goto nla_put_failure;
316
0dfbdf41 317 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 318 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
319 goto nla_put_failure;
c7995c43 320 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 321 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d
DS		 322 goto nla_put_failure;
323 if (rdst->remote_ifindex &&
324 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 325 goto nla_put_failure;
326
327 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
328 ci.ndm_confirmed = 0;
329 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
330 ci.ndm_refcnt = 0;
331
332 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
333 goto nla_put_failure;
334
053c095a
JB		 335 nlmsg_end(skb, nlh);
336 return 0;
d342894c 337
338nla_put_failure:
339 nlmsg_cancel(skb, nlh);
340 return -EMSGSIZE;
341}
342
343static inline size_t vxlan_nlmsg_size(void)
344{
345 return NLMSG_ALIGN(sizeof(struct ndmsg))
346 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 347 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 348 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 349 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
350 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 351 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 352 + nla_total_size(sizeof(struct nda_cacheinfo));
353}
354
9e4b93f9
ND		 355static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
356 struct vxlan_rdst *rd, int type)
d342894c 357{
358 struct net *net = dev_net(vxlan->dev);
359 struct sk_buff *skb;
360 int err = -ENOBUFS;
361
362 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
363 if (skb == NULL)
364 goto errout;
365
9e4b93f9 366 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 367 if (err < 0) {
368 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
369 WARN_ON(err == -EMSGSIZE);
370 kfree_skb(skb);
371 goto errout;
372 }
373
374 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
375 return;
376errout:
377 if (err < 0)
378 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
379}
380
e4c7ed41 381static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 382{
383 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 384 struct vxlan_fdb f = {
385 .state = NUD_STALE,
386 };
387 struct vxlan_rdst remote = {
e4c7ed41 388 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 389 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 390 };
3e61aa8f 391
9e4b93f9 392 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 393}
394
395static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
396{
bb3fd687
SH		 397 struct vxlan_fdb f = {
398 .state = NUD_STALE,
399 };
9e4b93f9 400 struct vxlan_rdst remote = { };
e4f67add 401
e4f67add
DS		 402 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
403
9e4b93f9 404 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 405}
406
d342894c 407/* Hash Ethernet address */
408static u32 eth_hash(const unsigned char *addr)
409{
410 u64 value = get_unaligned((u64 *)addr);
411
412 /* only want 6 bytes */
413#ifdef __BIG_ENDIAN
d342894c 414 value >>= 16;
321fb991 415#else
416 value <<= 16;
d342894c 417#endif
418 return hash_64(value, FDB_HASH_BITS);
419}
420
421/* Hash chain to use given mac address */
422static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
423 const u8 *mac)
424{
425 return &vxlan->fdb_head[eth_hash(mac)];
426}
427
428/* Look up Ethernet address in forwarding table */
014be2c8 429static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 430 const u8 *mac)
d342894c 431{
432 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
433 struct vxlan_fdb *f;
d342894c 434
b67bfe0d 435 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 436 if (ether_addr_equal(mac, f->eth_addr))
d342894c 437 return f;
438 }
439
440 return NULL;
441}
442
014be2c8
SS		 443static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
444 const u8 *mac)
445{
446 struct vxlan_fdb *f;
447
448 f = __vxlan_find_mac(vxlan, mac);
449 if (f)
450 f->used = jiffies;
451
452 return f;
453}
454
a5e7c10a
MR		 455/* caller should hold vxlan->hash_lock */
456static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 457 union vxlan_addr *ip, __be16 port,
54bfd872 458 __be32 vni, __u32 ifindex)
6681712d 459{
3e61aa8f 460 struct vxlan_rdst *rd;
6681712d 461
3e61aa8f 462 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 463 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 464 rd->remote_port == port &&
465 rd->remote_vni == vni &&
466 rd->remote_ifindex == ifindex)
a5e7c10a 467 return rd;
6681712d 468 }
3e61aa8f 469
a5e7c10a
MR		 470 return NULL;
471}
472
906dc186
TR		 473/* Replace destination of unicast mac */
474static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 475 union vxlan_addr *ip, __be16 port, __be32 vni,
476 __u32 ifindex)
906dc186
TR		 477{
478 struct vxlan_rdst *rd;
479
480 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
481 if (rd)
482 return 0;
483
484 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
485 if (!rd)
486 return 0;
0c1d70af
PA		 487
488 dst_cache_reset(&rd->dst_cache);
e4c7ed41 489 rd->remote_ip = *ip;
906dc186
TR		 490 rd->remote_port = port;
491 rd->remote_vni = vni;
492 rd->remote_ifindex = ifindex;
493 return 1;
494}
495
a5e7c10a
MR		 496/* Add/update destinations for multicast */
497static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 498 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 499 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 500{
501 struct vxlan_rdst *rd;
502
503 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
504 if (rd)
505 return 0;
506
6681712d
DS		 507 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
508 if (rd == NULL)
509 return -ENOBUFS;
0c1d70af
PA		 510
511 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
512 kfree(rd);
513 return -ENOBUFS;
514 }
515
e4c7ed41 516 rd->remote_ip = *ip;
6681712d
DS		 517 rd->remote_port = port;
518 rd->remote_vni = vni;
519 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 520
521 list_add_tail_rcu(&rd->list, &f->remotes);
522
9e4b93f9 523 *rdp = rd;
6681712d
DS		 524 return 1;
525}
526
dfd8645e
TH		 527static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
528 unsigned int off,
529 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 530 __be32 vni_field,
531 struct gro_remcsum *grc,
0ace2ca8 532 bool nopartial)
dfd8645e 533{
b7fe10e5 534 size_t start, offset;
dfd8645e
TH		 535
536 if (skb->remcsum_offload)
b7fe10e5 537 return vh;
dfd8645e
TH		 538
539 if (!NAPI_GRO_CB(skb)->csum_valid)
540 return NULL;
541
54bfd872
JB		 542 start = vxlan_rco_start(vni_field);
543 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 544
b7fe10e5
TH		 545 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
546 start, offset, grc, nopartial);
dfd8645e
TH		 547
548 skb->remcsum_offload = 1;
549
550 return vh;
551}
552
5602c48c
TH		 553static struct sk_buff **vxlan_gro_receive(struct sock *sk,
554 struct sk_buff **head,
555 struct sk_buff *skb)
dc01e7d3
OG		 556{
557 struct sk_buff *p, **pp = NULL;
558 struct vxlanhdr *vh, *vh2;
9b174d88 559 unsigned int hlen, off_vx;
dc01e7d3 560 int flush = 1;
5602c48c 561 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 562 __be32 flags;
26c4f7da
TH		 563 struct gro_remcsum grc;
564
565 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 566
567 off_vx = skb_gro_offset(skb);
568 hlen = off_vx + sizeof(*vh);
569 vh = skb_gro_header_fast(skb, off_vx);
570 if (skb_gro_header_hard(skb, hlen)) {
571 vh = skb_gro_header_slow(skb, hlen, off_vx);
572 if (unlikely(!vh))
573 goto out;
574 }
dc01e7d3 575
dfd8645e
TH		 576 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
577
54bfd872 578 flags = vh->vx_flags;
dfd8645e
TH		 579
580 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
581 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 582 vh->vx_vni, &grc,
0ace2ca8
TH		 583 !!(vs->flags &
584 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 585
586 if (!vh)
587 goto out;
588 }
589
b7fe10e5
TH		 590 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
591
dc01e7d3
OG		 592 for (p = *head; p; p = p->next) {
593 if (!NAPI_GRO_CB(p)->same_flow)
594 continue;
595
596 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 597 if (vh->vx_flags != vh2->vx_flags ||
598 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 599 NAPI_GRO_CB(p)->same_flow = 0;
600 continue;
601 }
dc01e7d3
OG		 602 }
603
9b174d88 604 pp = eth_gro_receive(head, skb);
c194cf93 605 flush = 0;
dc01e7d3 606
dc01e7d3 607out:
26c4f7da 608 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 609 NAPI_GRO_CB(skb)->flush |= flush;
610
611 return pp;
612}
613
5602c48c 614static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 615{
cfdf1e1b
JG		 616 udp_tunnel_gro_complete(skb, nhoff);
617
9b174d88 618 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 619}
620
53cf5275 621/* Notify netdevs that UDP port started listening */
dc01e7d3 622static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 623{
624 struct net_device *dev;
dc01e7d3 625 struct sock *sk = vs->sock->sk;
53cf5275 626 struct net *net = sock_net(sk);
705cc62f 627 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 628 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 629
630 rcu_read_lock();
631 for_each_netdev_rcu(net, dev) {
632 if (dev->netdev_ops->ndo_add_vxlan_port)
633 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
634 port);
635 }
636 rcu_read_unlock();
637}
638
639/* Notify netdevs that UDP port is no more listening */
dc01e7d3 640static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 641{
642 struct net_device *dev;
dc01e7d3 643 struct sock *sk = vs->sock->sk;
53cf5275 644 struct net *net = sock_net(sk);
705cc62f 645 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 646 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 647
648 rcu_read_lock();
649 for_each_netdev_rcu(net, dev) {
650 if (dev->netdev_ops->ndo_del_vxlan_port)
651 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
652 port);
653 }
654 rcu_read_unlock();
655}
656
d342894c 657/* Add new entry to forwarding table -- assumes lock held */
658static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 659 const u8 *mac, union vxlan_addr *ip,
6681712d 660 __u16 state, __u16 flags,
54bfd872 661 __be16 port, __be32 vni, __u32 ifindex,
ae884082 662 __u8 ndm_flags)
d342894c 663{
9e4b93f9 664 struct vxlan_rdst *rd = NULL;
d342894c 665 struct vxlan_fdb *f;
666 int notify = 0;
667
014be2c8 668 f = __vxlan_find_mac(vxlan, mac);
d342894c 669 if (f) {
670 if (flags & NLM_F_EXCL) {
671 netdev_dbg(vxlan->dev,
672 "lost race to create %pM\n", mac);
673 return -EEXIST;
674 }
675 if (f->state != state) {
676 f->state = state;
677 f->updated = jiffies;
678 notify = 1;
679 }
ae884082
DS		 680 if (f->flags != ndm_flags) {
681 f->flags = ndm_flags;
682 f->updated = jiffies;
683 notify = 1;
684 }
906dc186
TR		 685 if ((flags & NLM_F_REPLACE)) {
686 /* Only change unicasts */
687 if (!(is_multicast_ether_addr(f->eth_addr) ||
688 is_zero_ether_addr(f->eth_addr))) {
60840429 689 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 690 ifindex);
906dc186
TR		 691 } else
692 return -EOPNOTSUPP;
693 }
6681712d 694 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 695 (is_multicast_ether_addr(f->eth_addr) ||
696 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 697 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
698 &rd);
6681712d
DS		 699
700 if (rc < 0)
701 return rc;
702 notify |= rc;
703 }
d342894c 704 } else {
705 if (!(flags & NLM_F_CREATE))
706 return -ENOENT;
707
0dfbdf41
TG		 708 if (vxlan->cfg.addrmax &&
709 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 710 return -ENOSPC;
711
906dc186
TR		 712 /* Disallow replace to add a multicast entry */
713 if ((flags & NLM_F_REPLACE) &&
714 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
715 return -EOPNOTSUPP;
716
e4c7ed41 717 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 718 f = kmalloc(sizeof(*f), GFP_ATOMIC);
719 if (!f)
720 return -ENOMEM;
721
722 notify = 1;
d342894c 723 f->state = state;
ae884082 724 f->flags = ndm_flags;
d342894c 725 f->updated = f->used = jiffies;
3e61aa8f 726 INIT_LIST_HEAD(&f->remotes);
d342894c 727 memcpy(f->eth_addr, mac, ETH_ALEN);
728
9e4b93f9 729 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 730
d342894c 731 ++vxlan->addrcnt;
732 hlist_add_head_rcu(&f->hlist,
733 vxlan_fdb_head(vxlan, mac));
734 }
735
9e4b93f9
ND		 736 if (notify) {
737 if (rd == NULL)
738 rd = first_remote_rtnl(f);
739 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
740 }
d342894c 741
742 return 0;
743}
744
6706c82e 745static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 746{
747 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 748 struct vxlan_rdst *rd, *nd;
6681712d 749
0c1d70af
PA		 750 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
751 dst_cache_destroy(&rd->dst_cache);
6681712d 752 kfree(rd);
0c1d70af 753 }
6681712d
DS		 754 kfree(f);
755}
756
d342894c 757static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
758{
759 netdev_dbg(vxlan->dev,
760 "delete %pM\n", f->eth_addr);
761
762 --vxlan->addrcnt;
9e4b93f9 763 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 764
765 hlist_del_rcu(&f->hlist);
6681712d 766 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 767}
768
f0b074be 769static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB		 770 union vxlan_addr *ip, __be16 *port, __be32 *vni,
771 u32 *ifindex)
d342894c 772{
6681712d 773 struct net *net = dev_net(vxlan->dev);
e4c7ed41 774 int err;
d342894c 775
f0b074be 776 if (tb[NDA_DST]) {
e4c7ed41
CW		 777 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
778 if (err)
779 return err;
f0b074be 780 } else {
e4c7ed41
CW		 781 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
782 if (remote->sa.sa_family == AF_INET) {
783 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
784 ip->sa.sa_family = AF_INET;
785#if IS_ENABLED(CONFIG_IPV6)
786 } else {
787 ip->sin6.sin6_addr = in6addr_any;
788 ip->sa.sa_family = AF_INET6;
789#endif
790 }
f0b074be 791 }
d342894c 792
6681712d 793 if (tb[NDA_PORT]) {
73cf3317 794 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 795 return -EINVAL;
f0b074be
MR		 796 *port = nla_get_be16(tb[NDA_PORT]);
797 } else {
0dfbdf41 798 *port = vxlan->cfg.dst_port;
f0b074be 799 }
6681712d
DS		 800
801 if (tb[NDA_VNI]) {
802 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
803 return -EINVAL;
54bfd872 804 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 805 } else {
806 *vni = vxlan->default_dst.remote_vni;
807 }
6681712d
DS		 808
809 if (tb[NDA_IFINDEX]) {
5abb0029 810 struct net_device *tdev;
6681712d
DS		 811
812 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
813 return -EINVAL;
f0b074be 814 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 815 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 816 if (!tdev)
6681712d 817 return -EADDRNOTAVAIL;
f0b074be
MR		 818 } else {
819 *ifindex = 0;
820 }
821
822 return 0;
823}
824
825/* Add static entry (via netlink) */
826static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
827 struct net_device *dev,
f6f6424b 828 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 829{
830 struct vxlan_dev *vxlan = netdev_priv(dev);
831 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 832 union vxlan_addr ip;
f0b074be 833 __be16 port;
54bfd872
JB		 834 __be32 vni;
835 u32 ifindex;
f0b074be
MR		 836 int err;
837
838 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
839 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
840 ndm->ndm_state);
841 return -EINVAL;
842 }
843
844 if (tb[NDA_DST] == NULL)
845 return -EINVAL;
846
847 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
848 if (err)
849 return err;
6681712d 850
5933a7bb
MR		 851 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
852 return -EAFNOSUPPORT;
853
d342894c 854 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 855 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 856 port, vni, ifindex, ndm->ndm_flags);
d342894c 857 spin_unlock_bh(&vxlan->hash_lock);
858
859 return err;
860}
861
862/* Delete entry (via netlink) */
1690be63
VY		 863static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
864 struct net_device *dev,
f6f6424b 865 const unsigned char *addr, u16 vid)
d342894c 866{
867 struct vxlan_dev *vxlan = netdev_priv(dev);
868 struct vxlan_fdb *f;
bc7892ba 869 struct vxlan_rdst *rd = NULL;
e4c7ed41 870 union vxlan_addr ip;
bc7892ba 871 __be16 port;
54bfd872
JB		 872 __be32 vni;
873 u32 ifindex;
bc7892ba
MR		 874 int err;
875
876 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
877 if (err)
878 return err;
879
880 err = -ENOENT;
d342894c 881
882 spin_lock_bh(&vxlan->hash_lock);
883 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 884 if (!f)
885 goto out;
886
e4c7ed41
CW		 887 if (!vxlan_addr_any(&ip)) {
888 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 889 if (!rd)
890 goto out;
891 }
892
893 err = 0;
894
895 /* remove a destination if it's not the only one on the list,
896 * otherwise destroy the fdb entry
897 */
898 if (rd && !list_is_singular(&f->remotes)) {
899 list_del_rcu(&rd->list);
9e4b93f9 900 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 901 kfree_rcu(rd, rcu);
bc7892ba 902 goto out;
d342894c 903 }
bc7892ba
MR		 904
905 vxlan_fdb_destroy(vxlan, f);
906
907out:
d342894c 908 spin_unlock_bh(&vxlan->hash_lock);
909
910 return err;
911}
912
913/* Dump forwarding table */
914static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 915 struct net_device *dev,
916 struct net_device *filter_dev, int idx)
d342894c 917{
918 struct vxlan_dev *vxlan = netdev_priv(dev);
919 unsigned int h;
920
921 for (h = 0; h < FDB_HASH_SIZE; ++h) {
922 struct vxlan_fdb *f;
d342894c 923 int err;
924
b67bfe0d 925 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 926 struct vxlan_rdst *rd;
6681712d 927
3e61aa8f 928 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 929 if (idx < cb->args[0])
930 goto skip;
931
6681712d
DS		 932 err = vxlan_fdb_info(skb, vxlan, f,
933 NETLINK_CB(cb->skb).portid,
934 cb->nlh->nlmsg_seq,
935 RTM_NEWNEIGH,
936 NLM_F_MULTI, rd);
472681d5
MM		 937 if (err < 0) {
938 cb->args[1] = err;
3e61aa8f 939 goto out;
472681d5 940 }
3e61aa8f 941skip:
07a51cd3
AW		 942 ++idx;
943 }
d342894c 944 }
945 }
3e61aa8f 946out:
d342894c 947 return idx;
948}
949
950/* Watch incoming packets to learn mapping between Ethernet address
951 * and Tunnel endpoint.
c4b49512 952 * Return true if packet is bogus and should be dropped.
d342894c 953 */
26a41ae6 954static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 955 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 956{
957 struct vxlan_dev *vxlan = netdev_priv(dev);
958 struct vxlan_fdb *f;
d342894c 959
960 f = vxlan_find_mac(vxlan, src_mac);
961 if (likely(f)) {
5ca5461c 962 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 963
e4c7ed41 964 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 965 return false;
966
967 /* Don't migrate static entries, drop packets */
eb064c3b 968 if (f->state & NUD_NOARP)
26a41ae6 969 return true;
d342894c 970
971 if (net_ratelimit())
972 netdev_info(dev,
e4c7ed41 973 "%pM migrated from %pIS to %pIS\n",
a4870f79 974 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 975
e4c7ed41 976 rdst->remote_ip = *src_ip;
d342894c 977 f->updated = jiffies;
9e4b93f9 978 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 979 } else {
980 /* learned new entry */
981 spin_lock(&vxlan->hash_lock);
3bf74b1a 982
983 /* close off race between vxlan_flush and incoming packets */
984 if (netif_running(dev))
985 vxlan_fdb_create(vxlan, src_mac, src_ip,
986 NUD_REACHABLE,
987 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 988 vxlan->cfg.dst_port,
3bf74b1a 989 vxlan->default_dst.remote_vni,
990 0, NTF_SELF);
d342894c 991 spin_unlock(&vxlan->hash_lock);
992 }
26a41ae6 993
994 return false;
d342894c 995}
996
d342894c 997/* See if multicast group is already in use by other ID */
95ab0991 998static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 999{
553675fb 1000 struct vxlan_dev *vxlan;
b1be00a6 1001 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1002
95ab0991
G		 1003 /* The vxlan_sock is only used by dev, leaving group has
1004 * no effect on other vxlan devices.
1005 */
b1be00a6
JB		 1006 if (family == AF_INET && dev->vn4_sock &&
1007 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1008 return false;
b1be00a6
JB		 1009#if IS_ENABLED(CONFIG_IPV6)
1010 if (family == AF_INET6 && dev->vn6_sock &&
1011 atomic_read(&dev->vn6_sock->refcnt) == 1)
1012 return false;
1013#endif
95ab0991 1014
553675fb 1015 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1016 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1017 continue;
d342894c 1018
b1be00a6 1019 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1020 continue;
b1be00a6
JB		 1021#if IS_ENABLED(CONFIG_IPV6)
1022 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1023 continue;
1024#endif
95ab0991
G		 1025
1026 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1027 &dev->default_dst.remote_ip))
1028 continue;
1029
1030 if (vxlan->default_dst.remote_ifindex !=
1031 dev->default_dst.remote_ifindex)
1032 continue;
1033
1034 return true;
553675fb 1035 }
d342894c 1036
1037 return false;
1038}
1039
b1be00a6 1040static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1041{
b1be00a6 1042 struct vxlan_net *vn;
012a5729 1043
b1be00a6
JB		 1044 if (!vs)
1045 return;
7c47cedf
SH		 1046 if (!atomic_dec_and_test(&vs->refcnt))
1047 return;
d342894c 1048
b1be00a6 1049 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1050 spin_lock(&vn->sock_lock);
7c47cedf 1051 hlist_del_rcu(&vs->hlist);
dc01e7d3 1052 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1053 spin_unlock(&vn->sock_lock);
1054
0412bd93
HFS		 1055 synchronize_net();
1056 udp_tunnel_sock_release(vs->sock);
1057 kfree(vs);
d342894c 1058}
1059
b1be00a6
JB		 1060static void vxlan_sock_release(struct vxlan_dev *vxlan)
1061{
1062 __vxlan_sock_release(vxlan->vn4_sock);
1063#if IS_ENABLED(CONFIG_IPV6)
1064 __vxlan_sock_release(vxlan->vn6_sock);
1065#endif
1066}
1067
56ef9c90 1068/* Update multicast group membership when first VNI on
c4b49512 1069 * multicast address is brought up
7c47cedf 1070 */
56ef9c90 1071static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1072{
b1be00a6 1073 struct sock *sk;
e4c7ed41
CW		 1074 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1075 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1076 int ret = -EINVAL;
d342894c 1077
e4c7ed41
CW		 1078 if (ip->sa.sa_family == AF_INET) {
1079 struct ip_mreqn mreq = {
1080 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1081 .imr_ifindex = ifindex,
1082 };
1083
b1be00a6
JB		 1084 sk = vxlan->vn4_sock->sock->sk;
1085 lock_sock(sk);
56ef9c90 1086 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1087 release_sock(sk);
e4c7ed41
CW		 1088#if IS_ENABLED(CONFIG_IPV6)
1089 } else {
b1be00a6
JB		 1090 sk = vxlan->vn6_sock->sock->sk;
1091 lock_sock(sk);
56ef9c90
MRL		 1092 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1093 &ip->sin6.sin6_addr);
b1be00a6 1094 release_sock(sk);
e4c7ed41
CW		 1095#endif
1096 }
3fc2de2f 1097
56ef9c90 1098 return ret;
3fc2de2f 1099}
1100
1101/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1102static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1103{
b1be00a6 1104 struct sock *sk;
e4c7ed41
CW		 1105 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1106 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1107 int ret = -EINVAL;
3fc2de2f 1108
e4c7ed41
CW		 1109 if (ip->sa.sa_family == AF_INET) {
1110 struct ip_mreqn mreq = {
1111 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1112 .imr_ifindex = ifindex,
1113 };
1114
b1be00a6
JB		 1115 sk = vxlan->vn4_sock->sock->sk;
1116 lock_sock(sk);
56ef9c90 1117 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1118 release_sock(sk);
e4c7ed41
CW		 1119#if IS_ENABLED(CONFIG_IPV6)
1120 } else {
b1be00a6
JB		 1121 sk = vxlan->vn6_sock->sock->sk;
1122 lock_sock(sk);
56ef9c90
MRL		 1123 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1124 &ip->sin6.sin6_addr);
b1be00a6 1125 release_sock(sk);
e4c7ed41
CW		 1126#endif
1127 }
d342894c 1128
56ef9c90 1129 return ret;
d342894c 1130}
1131
f14ecebb
JB		 1132static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1133 struct sk_buff *skb, u32 vxflags)
dfd8645e 1134{
7d34fa75 1135 size_t start, offset;
dfd8645e 1136
f14ecebb
JB		 1137 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1138 goto out;
b7fe10e5 1139
f14ecebb
JB		 1140 start = vxlan_rco_start(unparsed->vx_vni);
1141 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1142
7d34fa75 1143 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1144 return false;
dfd8645e 1145
be5cfeab
JB		 1146 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1147 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1148out:
1149 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1150 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1151 return true;
dfd8645e
TH		 1152}
1153
f14ecebb 1154static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1155 struct sk_buff *skb, u32 vxflags,
10a5af23 1156 struct vxlan_metadata *md)
3288af08 1157{
f14ecebb 1158 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1159 struct metadata_dst *tun_dst;
f14ecebb
JB		 1160
1161 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1162 goto out;
3288af08 1163
3288af08
JB		 1164 md->gbp = ntohs(gbp->policy_id);
1165
10a5af23 1166 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1167 if (tun_dst) {
3288af08 1168 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1169 tun_dst->u.tun_info.options_len = sizeof(*md);
1170 }
3288af08
JB		 1171 if (gbp->dont_learn)
1172 md->gbp |= VXLAN_GBP_DONT_LEARN;
1173
1174 if (gbp->policy_applied)
1175 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1176
64f87d36
JB		 1177 /* In flow-based mode, GBP is carried in dst_metadata */
1178 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1179 skb->mark = md->gbp;
f14ecebb
JB		 1180out:
1181 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1182}
1183
e1e5314d 1184static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1185 __be16 *protocol,
e1e5314d
JB		 1186 struct sk_buff *skb, u32 vxflags)
1187{
1188 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1189
1190 /* Need to have Next Protocol set for interfaces in GPE mode. */
1191 if (!gpe->np_applied)
1192 return false;
1193 /* "The initial version is 0. If a receiver does not support the
1194 * version indicated it MUST drop the packet.
1195 */
1196 if (gpe->version != 0)
1197 return false;
1198 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1199 * processing MUST occur." However, we don't implement OAM
1200 * processing, thus drop the packet.
1201 */
1202 if (gpe->oam_flag)
1203 return false;
1204
1205 switch (gpe->next_protocol) {
1206 case VXLAN_GPE_NP_IPV4:
1207 *protocol = htons(ETH_P_IP);
1208 break;
1209 case VXLAN_GPE_NP_IPV6:
1210 *protocol = htons(ETH_P_IPV6);
1211 break;
1212 case VXLAN_GPE_NP_ETHERNET:
1213 *protocol = htons(ETH_P_TEB);
1214 break;
1215 default:
1216 return false;
1217 }
1218
1219 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1220 return true;
1221}
1222
1ab016e2
JB		 1223static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1224 struct vxlan_sock *vs,
1225 struct sk_buff *skb)
614732ea 1226{
614732ea 1227 union vxlan_addr saddr;
614732ea 1228
614732ea 1229 skb_reset_mac_header(skb);
614732ea
TG		 1230 skb->protocol = eth_type_trans(skb, vxlan->dev);
1231 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1232
1233 /* Ignore packet loops (and multicast echo) */
1234 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1235 return false;
614732ea 1236
760c6805 1237 /* Get address from the outer IP header */
ce212d0f 1238 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1239 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1240 saddr.sa.sa_family = AF_INET;
1241#if IS_ENABLED(CONFIG_IPV6)
1242 } else {
1ab016e2 1243 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1244 saddr.sa.sa_family = AF_INET6;
1245#endif
1246 }
1247
1ab016e2
JB		 1248 if ((vxlan->flags & VXLAN_F_LEARN) &&
1249 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1250 return false;
1251
1252 return true;
1253}
1254
760c6805
JB		 1255static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1256 struct sk_buff *skb)
1257{
1258 int err = 0;
1259
1260 if (vxlan_get_sk_family(vs) == AF_INET)
1261 err = IP_ECN_decapsulate(oiph, skb);
1262#if IS_ENABLED(CONFIG_IPV6)
1263 else
1264 err = IP6_ECN_decapsulate(oiph, skb);
1265#endif
1266
1267 if (unlikely(err) && log_ecn_error) {
1268 if (vxlan_get_sk_family(vs) == AF_INET)
1269 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1270 &((struct iphdr *)oiph)->saddr,
1271 ((struct iphdr *)oiph)->tos);
1272 else
1273 net_info_ratelimited("non-ECT from %pI6\n",
1274 &((struct ipv6hdr *)oiph)->saddr);
1275 }
1276 return err <= 1;
1277}
1278
d342894c 1279/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1280static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1281{
f2d1968e 1282 struct pcpu_sw_netstats *stats;
c9e78efb 1283 struct vxlan_dev *vxlan;
5cfccc5a 1284 struct vxlan_sock *vs;
f14ecebb 1285 struct vxlanhdr unparsed;
ee122c79
TG		 1286 struct vxlan_metadata _md;
1287 struct vxlan_metadata *md = &_md;
61618eea 1288 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1289 bool raw_proto = false;
f2d1968e 1290 void *oiph;
d342894c 1291
e1e5314d 1292 /* Need UDP and VXLAN header to be present */
7ce04758 1293 if (!pskb_may_pull(skb, VXLAN_HLEN))
288b01c8 1294 return 1;
d342894c 1295
f14ecebb 1296 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1297 /* VNI flag always required to be set */
1298 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1299 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1300 ntohl(vxlan_hdr(skb)->vx_flags),
1301 ntohl(vxlan_hdr(skb)->vx_vni));
1302 /* Return non vxlan pkt */
1303 return 1;
d342894c 1304 }
288b01c8
JB		 1305 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1306 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1307
559835ea 1308 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1309 if (!vs)
d342894c 1310 goto drop;
d342894c 1311
c9e78efb
JB		 1312 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1313 if (!vxlan)
1314 goto drop;
1315
e1e5314d
JB		 1316 /* For backwards compatibility, only allow reserved fields to be
1317 * used by VXLAN extensions if explicitly requested.
1318 */
1319 if (vs->flags & VXLAN_F_GPE) {
1320 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1321 goto drop;
1322 raw_proto = true;
1323 }
1324
1325 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1326 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1327 goto drop;
c9e78efb 1328
ee122c79 1329 if (vxlan_collect_metadata(vs)) {
07dabf20 1330 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
10a5af23 1331 struct metadata_dst *tun_dst;
07dabf20 1332
c29a70d2 1333 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
07dabf20 1334 vxlan_vni_to_tun_id(vni), sizeof(*md));
c29a70d2 1335
ee122c79
TG		 1336 if (!tun_dst)
1337 goto drop;
1338
0f1b7354 1339 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1340
1341 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1342 } else {
1343 memset(md, 0, sizeof(*md));
1344 }
1345
f14ecebb
JB		 1346 if (vs->flags & VXLAN_F_REMCSUM_RX)
1347 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1348 goto drop;
1349 if (vs->flags & VXLAN_F_GBP)
10a5af23 1350 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1351 /* Note that GBP and GPE can never be active together. This is
1352 * ensured in vxlan_dev_configure.
1353 */
3511494c 1354
f14ecebb 1355 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1356 /* If there are any unprocessed flags remaining treat
1357 * this as a malformed packet. This behavior diverges from
1358 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1359 * in reserved fields are to be ignored. The approach here
c4b49512 1360 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1361 * is more robust and provides a little more security in
1362 * adding extensions to VXLAN.
1363 */
288b01c8 1364 goto drop;
3bf39475
TH		 1365 }
1366
e1e5314d
JB		 1367 if (!raw_proto) {
1368 if (!vxlan_set_mac(vxlan, vs, skb))
1369 goto drop;
1370 } else {
1371 skb->dev = vxlan->dev;
1372 skb->pkt_type = PACKET_HOST;
1373 }
f2d1968e 1374
f2d1968e
JB		 1375 oiph = skb_network_header(skb);
1376 skb_reset_network_header(skb);
1377
1378 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1379 ++vxlan->dev->stats.rx_frame_errors;
1380 ++vxlan->dev->stats.rx_errors;
1381 goto drop;
1382 }
1383
1384 stats = this_cpu_ptr(vxlan->dev->tstats);
1385 u64_stats_update_begin(&stats->syncp);
1386 stats->rx_packets++;
1387 stats->rx_bytes += skb->len;
1388 u64_stats_update_end(&stats->syncp);
1389
1390 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1391 return 0;
1392
1393drop:
288b01c8
JB		 1394 /* Consume bad packet */
1395 kfree_skb(skb);
1396 return 0;
5cfccc5a
PS		 1397}
1398
e4f67add
DS		 1399static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1400{
1401 struct vxlan_dev *vxlan = netdev_priv(dev);
1402 struct arphdr *parp;
1403 u8 *arpptr, *sha;
1404 __be32 sip, tip;
1405 struct neighbour *n;
1406
1407 if (dev->flags & IFF_NOARP)
1408 goto out;
1409
1410 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1411 dev->stats.tx_dropped++;
1412 goto out;
1413 }
1414 parp = arp_hdr(skb);
1415
1416 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1417 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1418 parp->ar_pro != htons(ETH_P_IP) ||
1419 parp->ar_op != htons(ARPOP_REQUEST) ||
1420 parp->ar_hln != dev->addr_len ||
1421 parp->ar_pln != 4)
1422 goto out;
1423 arpptr = (u8 *)parp + sizeof(struct arphdr);
1424 sha = arpptr;
1425 arpptr += dev->addr_len; /* sha */
1426 memcpy(&sip, arpptr, sizeof(sip));
1427 arpptr += sizeof(sip);
1428 arpptr += dev->addr_len; /* tha */
1429 memcpy(&tip, arpptr, sizeof(tip));
1430
1431 if (ipv4_is_loopback(tip) ||
1432 ipv4_is_multicast(tip))
1433 goto out;
1434
1435 n = neigh_lookup(&arp_tbl, &tip, dev);
1436
1437 if (n) {
e4f67add
DS		 1438 struct vxlan_fdb *f;
1439 struct sk_buff *reply;
1440
1441 if (!(n->nud_state & NUD_CONNECTED)) {
1442 neigh_release(n);
1443 goto out;
1444 }
1445
1446 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1447 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1448 /* bridge-local neighbor */
1449 neigh_release(n);
1450 goto out;
1451 }
1452
1453 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1454 n->ha, sha);
1455
1456 neigh_release(n);
1457
7346135d
DS		 1458 if (reply == NULL)
1459 goto out;
1460
e4f67add
DS		 1461 skb_reset_mac_header(reply);
1462 __skb_pull(reply, skb_network_offset(reply));
1463 reply->ip_summed = CHECKSUM_UNNECESSARY;
1464 reply->pkt_type = PACKET_HOST;
1465
1466 if (netif_rx_ni(reply) == NET_RX_DROP)
1467 dev->stats.rx_dropped++;
e4c7ed41
CW		 1468 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1469 union vxlan_addr ipa = {
1470 .sin.sin_addr.s_addr = tip,
a45e92a5 1471 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1472 };
1473
1474 vxlan_ip_miss(dev, &ipa);
1475 }
e4f67add
DS		 1476out:
1477 consume_skb(skb);
1478 return NETDEV_TX_OK;
1479}
1480
f564f45c 1481#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1482static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1483 struct neighbour *n, bool isrouter)
1484{
1485 struct net_device *dev = request->dev;
1486 struct sk_buff *reply;
1487 struct nd_msg *ns, *na;
1488 struct ipv6hdr *pip6;
1489 u8 *daddr;
1490 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1491 int ns_olen;
1492 int i, len;
1493
1494 if (dev == NULL)
1495 return NULL;
1496
1497 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1498 sizeof(*na) + na_olen + dev->needed_tailroom;
1499 reply = alloc_skb(len, GFP_ATOMIC);
1500 if (reply == NULL)
1501 return NULL;
1502
1503 reply->protocol = htons(ETH_P_IPV6);
1504 reply->dev = dev;
1505 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1506 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1507 skb_reset_mac_header(reply);
4b29dba9
DS		 1508
1509 ns = (struct nd_msg *)skb_transport_header(request);
1510
1511 daddr = eth_hdr(request)->h_source;
1512 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1513 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1514 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1515 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1516 break;
1517 }
1518 }
1519
1520 /* Ethernet header */
1521 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1522 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1523 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1524 reply->protocol = htons(ETH_P_IPV6);
1525
1526 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1527 skb_reset_network_header(reply);
4b29dba9
DS		 1528 skb_put(reply, sizeof(struct ipv6hdr));
1529
1530 /* IPv6 header */
1531
1532 pip6 = ipv6_hdr(reply);
1533 memset(pip6, 0, sizeof(struct ipv6hdr));
1534 pip6->version = 6;
1535 pip6->priority = ipv6_hdr(request)->priority;
1536 pip6->nexthdr = IPPROTO_ICMPV6;
1537 pip6->hop_limit = 255;
1538 pip6->daddr = ipv6_hdr(request)->saddr;
1539 pip6->saddr = *(struct in6_addr *)n->primary_key;
1540
1541 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1542 skb_reset_transport_header(reply);
4b29dba9
DS		 1543
1544 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1545
1546 /* Neighbor Advertisement */
1547 memset(na, 0, sizeof(*na)+na_olen);
1548 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1549 na->icmph.icmp6_router = isrouter;
1550 na->icmph.icmp6_override = 1;
1551 na->icmph.icmp6_solicited = 1;
1552 na->target = ns->target;
1553 ether_addr_copy(&na->opt[2], n->ha);
1554 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1555 na->opt[1] = na_olen >> 3;
1556
1557 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1558 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1559 csum_partial(na, sizeof(*na)+na_olen, 0));
1560
1561 pip6->payload_len = htons(sizeof(*na)+na_olen);
1562
1563 skb_push(reply, sizeof(struct ipv6hdr));
1564
1565 reply->ip_summed = CHECKSUM_UNNECESSARY;
1566
1567 return reply;
1568}
1569
f564f45c
CW		 1570static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1571{
1572 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1573 struct nd_msg *msg;
f564f45c
CW		 1574 const struct ipv6hdr *iphdr;
1575 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1576 struct neighbour *n;
1577 struct inet6_dev *in6_dev;
f564f45c
CW		 1578
1579 in6_dev = __in6_dev_get(dev);
1580 if (!in6_dev)
1581 goto out;
1582
f564f45c
CW		 1583 iphdr = ipv6_hdr(skb);
1584 saddr = &iphdr->saddr;
1585 daddr = &iphdr->daddr;
1586
f564f45c
CW		 1587 msg = (struct nd_msg *)skb_transport_header(skb);
1588 if (msg->icmph.icmp6_code != 0 ||
1589 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1590 goto out;
1591
4b29dba9
DS		 1592 if (ipv6_addr_loopback(daddr) ||
1593 ipv6_addr_is_multicast(&msg->target))
1594 goto out;
1595
1596 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1597
1598 if (n) {
1599 struct vxlan_fdb *f;
4b29dba9 1600 struct sk_buff *reply;
f564f45c
CW		 1601
1602 if (!(n->nud_state & NUD_CONNECTED)) {
1603 neigh_release(n);
1604 goto out;
1605 }
1606
1607 f = vxlan_find_mac(vxlan, n->ha);
1608 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1609 /* bridge-local neighbor */
1610 neigh_release(n);
1611 goto out;
1612 }
1613
4b29dba9
DS		 1614 reply = vxlan_na_create(skb, n,
1615 !!(f ? f->flags & NTF_ROUTER : 0));
1616
f564f45c 1617 neigh_release(n);
4b29dba9
DS		 1618
1619 if (reply == NULL)
1620 goto out;
1621
1622 if (netif_rx_ni(reply) == NET_RX_DROP)
1623 dev->stats.rx_dropped++;
1624
f564f45c 1625 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1626 union vxlan_addr ipa = {
1627 .sin6.sin6_addr = msg->target,
a45e92a5 1628 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1629 };
1630
f564f45c
CW		 1631 vxlan_ip_miss(dev, &ipa);
1632 }
1633
1634out:
1635 consume_skb(skb);
1636 return NETDEV_TX_OK;
1637}
1638#endif
1639
e4f67add
DS		 1640static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1641{
1642 struct vxlan_dev *vxlan = netdev_priv(dev);
1643 struct neighbour *n;
e4f67add
DS		 1644
1645 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1646 return false;
1647
1648 n = NULL;
1649 switch (ntohs(eth_hdr(skb)->h_proto)) {
1650 case ETH_P_IP:
e15a00aa
CW		 1651 {
1652 struct iphdr *pip;
1653
e4f67add
DS		 1654 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1655 return false;
1656 pip = ip_hdr(skb);
1657 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1658 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1659 union vxlan_addr ipa = {
1660 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1661 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1662 };
1663
1664 vxlan_ip_miss(dev, &ipa);
1665 return false;
1666 }
1667
e4f67add 1668 break;
e15a00aa
CW		 1669 }
1670#if IS_ENABLED(CONFIG_IPV6)
1671 case ETH_P_IPV6:
1672 {
1673 struct ipv6hdr *pip6;
1674
1675 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1676 return false;
1677 pip6 = ipv6_hdr(skb);
1678 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1679 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1680 union vxlan_addr ipa = {
1681 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1682 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1683 };
1684
1685 vxlan_ip_miss(dev, &ipa);
1686 return false;
1687 }
1688
1689 break;
1690 }
1691#endif
e4f67add
DS		 1692 default:
1693 return false;
1694 }
1695
1696 if (n) {
1697 bool diff;
1698
7367d0b5 1699 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1700 if (diff) {
1701 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1702 dev->addr_len);
1703 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1704 }
1705 neigh_release(n);
1706 return diff;
e4c7ed41
CW		 1707 }
1708
e4f67add
DS		 1709 return false;
1710}
1711
af33c1ad 1712static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1713 struct vxlan_metadata *md)
1714{
1715 struct vxlanhdr_gbp *gbp;
1716
db79a621
TG		 1717 if (!md->gbp)
1718 return;
1719
3511494c 1720 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1721 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1722
1723 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1724 gbp->dont_learn = 1;
1725
1726 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1727 gbp->policy_applied = 1;
1728
1729 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1730}
1731
e1e5314d
JB		 1732static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1733 __be16 protocol)
1734{
1735 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1736
1737 gpe->np_applied = 1;
1738
1739 switch (protocol) {
1740 case htons(ETH_P_IP):
1741 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1742 return 0;
1743 case htons(ETH_P_IPV6):
1744 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1745 return 0;
1746 case htons(ETH_P_TEB):
1747 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1748 return 0;
1749 }
1750 return -EPFNOSUPPORT;
1751}
1752
f491e56d
JB		 1753static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1754 int iphdr_len, __be32 vni,
1755 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1756 bool udp_sum)
e4c7ed41 1757{
e4c7ed41 1758 struct vxlanhdr *vxh;
e4c7ed41
CW		 1759 int min_headroom;
1760 int err;
dfd8645e 1761 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1762 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1763
af33c1ad 1764 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1765 skb->ip_summed == CHECKSUM_PARTIAL) {
1766 int csum_start = skb_checksum_start_offset(skb);
1767
1768 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1769 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1770 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1771 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1772 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1773 }
e4c7ed41 1774
e4c7ed41 1775 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1776 + VXLAN_HLEN + iphdr_len
df8a39de 1777 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1778
1779 /* Need space for new headers (invalidates iph ptr) */
1780 err = skb_cow_head(skb, min_headroom);
e1e5314d
JB		 1781 if (unlikely(err))
1782 goto out_free;
649c5b8b 1783
5968250c
JP		 1784 skb = vlan_hwaccel_push_inside(skb);
1785 if (WARN_ON(!skb))
1786 return -ENOMEM;
1eaa8178 1787
6fa79666 1788 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG		 1789 if (IS_ERR(skb))
1790 return PTR_ERR(skb);
1791
49560532 1792 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1793 vxh->vx_flags = VXLAN_HF_VNI;
1794 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1795
dfd8645e 1796 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1797 unsigned int start;
dfd8645e 1798
54bfd872
JB		 1799 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1800 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1801 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1802
1803 if (!skb_is_gso(skb)) {
1804 skb->ip_summed = CHECKSUM_NONE;
1805 skb->encapsulation = 0;
1806 }
1807 }
1808
af33c1ad
TH		 1809 if (vxflags & VXLAN_F_GBP)
1810 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1811 if (vxflags & VXLAN_F_GPE) {
1812 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1813 if (err < 0)
1814 goto out_free;
1815 inner_protocol = skb->protocol;
1816 }
3511494c 1817
e1e5314d 1818 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1819 return 0;
e1e5314d
JB		 1820
1821out_free:
1822 kfree_skb(skb);
1823 return err;
49560532 1824}
49560532 1825
1a8496ba
JB		 1826static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1827 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1828 __be32 daddr, __be32 *saddr,
1829 struct dst_cache *dst_cache,
db3c6139 1830 const struct ip_tunnel_info *info)
1a8496ba 1831{
db3c6139 1832 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1833 struct rtable *rt = NULL;
1834 struct flowi4 fl4;
1835
db3c6139
DB		 1836 if (tos && !info)
1837 use_cache = false;
1838 if (use_cache) {
0c1d70af
PA		 1839 rt = dst_cache_get_ip4(dst_cache, saddr);
1840 if (rt)
1841 return rt;
1842 }
1843
1a8496ba
JB		 1844 memset(&fl4, 0, sizeof(fl4));
1845 fl4.flowi4_oif = oif;
1846 fl4.flowi4_tos = RT_TOS(tos);
1847 fl4.flowi4_mark = skb->mark;
1848 fl4.flowi4_proto = IPPROTO_UDP;
1849 fl4.daddr = daddr;
1850 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1851
1852 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1853 if (!IS_ERR(rt)) {
1a8496ba 1854 *saddr = fl4.saddr;
0c1d70af
PA		 1855 if (use_cache)
1856 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1857 }
1a8496ba
JB		 1858 return rt;
1859}
1860
e5d4b29f
JB		 1861#if IS_ENABLED(CONFIG_IPV6)
1862static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1400615d 1863 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1864 __be32 label,
e5d4b29f 1865 const struct in6_addr *daddr,
0c1d70af 1866 struct in6_addr *saddr,
db3c6139
DB		 1867 struct dst_cache *dst_cache,
1868 const struct ip_tunnel_info *info)
e5d4b29f 1869{
db3c6139 1870 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1871 struct dst_entry *ndst;
1872 struct flowi6 fl6;
1873 int err;
1874
1400615d
DB		 1875 if (tos && !info)
1876 use_cache = false;
db3c6139 1877 if (use_cache) {
0c1d70af
PA		 1878 ndst = dst_cache_get_ip6(dst_cache, saddr);
1879 if (ndst)
1880 return ndst;
1881 }
1882
e5d4b29f
JB		 1883 memset(&fl6, 0, sizeof(fl6));
1884 fl6.flowi6_oif = oif;
1885 fl6.daddr = *daddr;
1886 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
eaa93bf4 1887 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1888 fl6.flowi6_mark = skb->mark;
1889 fl6.flowi6_proto = IPPROTO_UDP;
1890
1891 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1892 vxlan->vn6_sock->sock->sk,
1893 &ndst, &fl6);
1894 if (err < 0)
1895 return ERR_PTR(err);
1896
1897 *saddr = fl6.saddr;
db3c6139 1898 if (use_cache)
0c1d70af 1899 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1900 return ndst;
1901}
1902#endif
1903
9dcc71e1
SS		 1904/* Bypass encapsulation if the destination is local */
1905static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1906 struct vxlan_dev *dst_vxlan)
1907{
8f84985f 1908 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1909 union vxlan_addr loopback;
1910 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1911 struct net_device *dev = skb->dev;
1912 int len = skb->len;
9dcc71e1 1913
8f84985f
LR		 1914 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1915 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1916 skb->pkt_type = PACKET_HOST;
1917 skb->encapsulation = 0;
1918 skb->dev = dst_vxlan->dev;
1919 __skb_pull(skb, skb_network_offset(skb));
1920
e4c7ed41
CW		 1921 if (remote_ip->sa.sa_family == AF_INET) {
1922 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1923 loopback.sa.sa_family = AF_INET;
1924#if IS_ENABLED(CONFIG_IPV6)
1925 } else {
1926 loopback.sin6.sin6_addr = in6addr_loopback;
1927 loopback.sa.sa_family = AF_INET6;
1928#endif
1929 }
1930
9dcc71e1 1931 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1932 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1933
1934 u64_stats_update_begin(&tx_stats->syncp);
1935 tx_stats->tx_packets++;
ce6502a8 1936 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1937 u64_stats_update_end(&tx_stats->syncp);
1938
1939 if (netif_rx(skb) == NET_RX_SUCCESS) {
1940 u64_stats_update_begin(&rx_stats->syncp);
1941 rx_stats->rx_packets++;
ce6502a8 1942 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1943 u64_stats_update_end(&rx_stats->syncp);
1944 } else {
ce6502a8 1945 dev->stats.rx_dropped++;
9dcc71e1
SS		 1946 }
1947}
1948
4ad16930
SH		 1949static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1950 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1951{
d71785ff 1952 struct dst_cache *dst_cache;
3093fbe7 1953 struct ip_tunnel_info *info;
d342894c 1954 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1955 struct sock *sk;
e4c7ed41 1956 struct rtable *rt = NULL;
d342894c 1957 const struct iphdr *old_iph;
e4c7ed41 1958 union vxlan_addr *dst;
ee122c79
TG		 1959 union vxlan_addr remote_ip;
1960 struct vxlan_metadata _md;
1961 struct vxlan_metadata *md = &_md;
e4c7ed41 1962 __be16 src_port = 0, dst_port;
e7f70af1 1963 __be32 vni, label;
d342894c 1964 __be16 df = 0;
1965 __u8 tos, ttl;
0e6fbc5b 1966 int err;
ee122c79 1967 u32 flags = vxlan->flags;
b4ed5cad 1968 bool udp_sum = false;
f491e56d 1969 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1970
61adedf3 1971 info = skb_tunnel_info(skb);
3093fbe7 1972
ee122c79 1973 if (rdst) {
0dfbdf41 1974 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1975 vni = rdst->remote_vni;
1976 dst = &rdst->remote_ip;
d71785ff 1977 dst_cache = &rdst->dst_cache;
ee122c79
TG		 1978 } else {
1979 if (!info) {
1980 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1981 dev->name);
1982 goto drop;
1983 }
0dfbdf41 1984 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
54bfd872 1985 vni = vxlan_tun_id_to_vni(info->key.tun_id);
b1be00a6
JB		 1986 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1987 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1988 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1989 else
1990 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1991 dst = &remote_ip;
d71785ff 1992 dst_cache = &info->dst_cache;
ee122c79 1993 }
e4f67add 1994
e4c7ed41 1995 if (vxlan_addr_any(dst)) {
e4f67add 1996 if (did_rsc) {
e4f67add 1997 /* short-circuited back to local bridge */
9dcc71e1 1998 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1999 return;
e4f67add 2000 }
ef59febe 2001 goto drop;
e4f67add 2002 }
ef59febe 2003
d342894c 2004 old_iph = ip_hdr(skb);
2005
0dfbdf41 2006 ttl = vxlan->cfg.ttl;
e4c7ed41 2007 if (!ttl && vxlan_addr_multicast(dst))
d342894c 2008 ttl = 1;
2009
0dfbdf41 2010 tos = vxlan->cfg.tos;
d342894c 2011 if (tos == 1)
206aaafc 2012 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 2013
e7f70af1 2014 label = vxlan->cfg.label;
0dfbdf41
TG		 2015 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2016 vxlan->cfg.port_max, true);
d342894c 2017
a725e514 2018 if (info) {
a725e514
JB		 2019 ttl = info->key.ttl;
2020 tos = info->key.tos;
e7f70af1 2021 label = info->key.label;
b4ed5cad 2022 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB		 2023
2024 if (info->options_len)
4c222798 2025 md = ip_tunnel_info_opts(info);
a725e514
JB		 2026 } else {
2027 md->gbp = skb->mark;
2028 }
2029
e4c7ed41 2030 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB		 2031 __be32 saddr;
2032
b1be00a6
JB		 2033 if (!vxlan->vn4_sock)
2034 goto drop;
2035 sk = vxlan->vn4_sock->sock->sk;
2036
1a8496ba
JB		 2037 rt = vxlan_get_route(vxlan, skb,
2038 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 2039 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 2040 dst_cache, info);
e4c7ed41
CW		 2041 if (IS_ERR(rt)) {
2042 netdev_dbg(dev, "no route to %pI4\n",
2043 &dst->sin.sin_addr.s_addr);
2044 dev->stats.tx_carrier_errors++;
2045 goto tx_error;
2046 }
d342894c 2047
e4c7ed41
CW		 2048 if (rt->dst.dev == dev) {
2049 netdev_dbg(dev, "circular route to %pI4\n",
2050 &dst->sin.sin_addr.s_addr);
2051 dev->stats.collisions++;
fffc15a5 2052 goto rt_tx_error;
e4c7ed41
CW		 2053 }
2054
2055 /* Bypass encapsulation if the destination is local */
2056 if (rt->rt_flags & RTCF_LOCAL &&
2057 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2058 struct vxlan_dev *dst_vxlan;
2059
2060 ip_rt_put(rt);
19ca9fc1 2061 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2062 dst->sa.sa_family, dst_port,
2063 vxlan->flags);
e4c7ed41
CW		 2064 if (!dst_vxlan)
2065 goto tx_error;
2066 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2067 return;
2068 }
2069
6ceb31ca
AD		 2070 if (!info)
2071 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2072 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2073 df = htons(IP_DF);
2074
e4c7ed41
CW		 2075 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2076 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d 2077 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
54bfd872 2078 vni, md, flags, udp_sum);
f491e56d
JB		 2079 if (err < 0)
2080 goto xmit_tx_error;
2081
2082 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2083 dst->sin.sin_addr.s_addr, tos, ttl, df,
2084 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2085#if IS_ENABLED(CONFIG_IPV6)
2086 } else {
e4c7ed41 2087 struct dst_entry *ndst;
e5d4b29f 2088 struct in6_addr saddr;
6f264e47 2089 u32 rt6i_flags;
e4c7ed41 2090
b1be00a6
JB		 2091 if (!vxlan->vn6_sock)
2092 goto drop;
2093 sk = vxlan->vn6_sock->sock->sk;
2094
e5d4b29f 2095 ndst = vxlan6_get_route(vxlan, skb,
1400615d 2096 rdst ? rdst->remote_ifindex : 0, tos,
e7f70af1 2097 label, &dst->sin6.sin6_addr, &saddr,
db3c6139 2098 dst_cache, info);
e5d4b29f 2099 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2100 netdev_dbg(dev, "no route to %pI6\n",
2101 &dst->sin6.sin6_addr);
2102 dev->stats.tx_carrier_errors++;
9dcc71e1 2103 goto tx_error;
e4c7ed41 2104 }
d342894c 2105
e4c7ed41
CW		 2106 if (ndst->dev == dev) {
2107 netdev_dbg(dev, "circular route to %pI6\n",
2108 &dst->sin6.sin6_addr);
2109 dst_release(ndst);
2110 dev->stats.collisions++;
2111 goto tx_error;
2112 }
0e6fbc5b 2113
e4c7ed41 2114 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2115 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2116 if (rt6i_flags & RTF_LOCAL &&
2117 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2118 struct vxlan_dev *dst_vxlan;
2119
2120 dst_release(ndst);
19ca9fc1 2121 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2122 dst->sa.sa_family, dst_port,
2123 vxlan->flags);
e4c7ed41
CW		 2124 if (!dst_vxlan)
2125 goto tx_error;
2126 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2127 return;
2128 }
49560532 2129
b4ed5cad
JB		 2130 if (!info)
2131 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2132
1400615d 2133 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2134 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2135 skb_scrub_packet(skb, xnet);
2136 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2137 vni, md, flags, udp_sum);
f491e56d
JB		 2138 if (err < 0) {
2139 dst_release(ndst);
2140 return;
2141 }
2142 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
e7f70af1
DB		 2143 &saddr, &dst->sin6.sin6_addr, tos, ttl,
2144 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2145#endif
2146 }
0e6fbc5b 2147
4ad16930 2148 return;
d342894c 2149
2150drop:
2151 dev->stats.tx_dropped++;
2152 goto tx_free;
2153
f491e56d
JB		 2154xmit_tx_error:
2155 /* skb is already freed. */
2156 skb = NULL;
49560532
PS		 2157rt_tx_error:
2158 ip_rt_put(rt);
d342894c 2159tx_error:
2160 dev->stats.tx_errors++;
2161tx_free:
2162 dev_kfree_skb(skb);
d342894c 2163}
2164
6681712d
DS		 2165/* Transmit local packets over Vxlan
2166 *
2167 * Outer IP header inherits ECN and DF from inner header.
2168 * Outer UDP destination is the VXLAN assigned port.
2169 * source port is based on hash of flow
2170 */
2171static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2172{
2173 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2174 const struct ip_tunnel_info *info;
6681712d
DS		 2175 struct ethhdr *eth;
2176 bool did_rsc = false;
8f646c92 2177 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2178 struct vxlan_fdb *f;
6681712d 2179
61adedf3 2180 info = skb_tunnel_info(skb);
3093fbe7 2181
6681712d 2182 skb_reset_mac_header(skb);
6681712d 2183
47e5d1b0
JB		 2184 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2185 if (info && info->mode & IP_TUNNEL_INFO_TX)
2186 vxlan_xmit_one(skb, dev, NULL, false);
2187 else
2188 kfree_skb(skb);
2189 return NETDEV_TX_OK;
2190 }
2191
2192 if (vxlan->flags & VXLAN_F_PROXY) {
2193 eth = eth_hdr(skb);
f564f45c
CW		 2194 if (ntohs(eth->h_proto) == ETH_P_ARP)
2195 return arp_reduce(dev, skb);
2196#if IS_ENABLED(CONFIG_IPV6)
2197 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2198 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2199 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2200 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2201 struct nd_msg *msg;
2202
2203 msg = (struct nd_msg *)skb_transport_header(skb);
2204 if (msg->icmph.icmp6_code == 0 &&
2205 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2206 return neigh_reduce(dev, skb);
2207 }
2208#endif
2209 }
6681712d 2210
47e5d1b0 2211 eth = eth_hdr(skb);
6681712d 2212 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2213 did_rsc = false;
2214
2215 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2216 (ntohs(eth->h_proto) == ETH_P_IP ||
2217 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2218 did_rsc = route_shortcircuit(dev, skb);
2219 if (did_rsc)
2220 f = vxlan_find_mac(vxlan, eth->h_dest);
2221 }
2222
6681712d 2223 if (f == NULL) {
afbd8bae
MR		 2224 f = vxlan_find_mac(vxlan, all_zeros_mac);
2225 if (f == NULL) {
2226 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2227 !is_multicast_ether_addr(eth->h_dest))
2228 vxlan_fdb_miss(vxlan, eth->h_dest);
2229
2230 dev->stats.tx_dropped++;
8f646c92 2231 kfree_skb(skb);
afbd8bae
MR		 2232 return NETDEV_TX_OK;
2233 }
2234 }
6681712d 2235
afbd8bae
MR		 2236 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2237 struct sk_buff *skb1;
6681712d 2238
8f646c92
ED		 2239 if (!fdst) {
2240 fdst = rdst;
2241 continue;
2242 }
afbd8bae
MR		 2243 skb1 = skb_clone(skb, GFP_ATOMIC);
2244 if (skb1)
2245 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2246 }
2247
8f646c92
ED		 2248 if (fdst)
2249 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2250 else
2251 kfree_skb(skb);
4ad16930 2252 return NETDEV_TX_OK;
6681712d
DS		 2253}
2254
d342894c 2255/* Walk the forwarding table and purge stale entries */
2256static void vxlan_cleanup(unsigned long arg)
2257{
2258 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2259 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2260 unsigned int h;
2261
2262 if (!netif_running(vxlan->dev))
2263 return;
2264
d342894c 2265 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2266 struct hlist_node *p, *n;
14e1d0fa
SD		 2267
2268 spin_lock_bh(&vxlan->hash_lock);
d342894c 2269 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2270 struct vxlan_fdb *f
2271 = container_of(p, struct vxlan_fdb, hlist);
2272 unsigned long timeout;
2273
3c172868 2274 if (f->state & NUD_PERMANENT)
d342894c 2275 continue;
2276
0dfbdf41 2277 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2278 if (time_before_eq(timeout, jiffies)) {
2279 netdev_dbg(vxlan->dev,
2280 "garbage collect %pM\n",
2281 f->eth_addr);
2282 f->state = NUD_STALE;
2283 vxlan_fdb_destroy(vxlan, f);
2284 } else if (time_before(timeout, next_timer))
2285 next_timer = timeout;
2286 }
14e1d0fa 2287 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2288 }
d342894c 2289
2290 mod_timer(&vxlan->age_timer, next_timer);
2291}
2292
9c2e24e1
PS		 2293static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2294{
56ef9c90 2295 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2296 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2297
56ef9c90 2298 spin_lock(&vn->sock_lock);
9c2e24e1 2299 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2300 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2301}
2302
d342894c 2303/* Setup stats when device is created */
2304static int vxlan_init(struct net_device *dev)
2305{
1c213bd2 2306 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2307 if (!dev->tstats)
d342894c 2308 return -ENOMEM;
2309
2310 return 0;
2311}
2312
ba609e9b 2313static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2314{
2315 struct vxlan_fdb *f;
2316
2317 spin_lock_bh(&vxlan->hash_lock);
2318 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2319 if (f)
2320 vxlan_fdb_destroy(vxlan, f);
2321 spin_unlock_bh(&vxlan->hash_lock);
2322}
2323
ebf4063e
SH		 2324static void vxlan_uninit(struct net_device *dev)
2325{
2326 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2327
ba609e9b 2328 vxlan_fdb_delete_default(vxlan);
afbd8bae 2329
ebf4063e
SH		 2330 free_percpu(dev->tstats);
2331}
2332
d342894c 2333/* Start ageing timer and join group when device is brought up */
2334static int vxlan_open(struct net_device *dev)
2335{
2336 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2337 int ret;
56ef9c90 2338
205f356d
JB		 2339 ret = vxlan_sock_add(vxlan);
2340 if (ret < 0)
2341 return ret;
d342894c 2342
79d4a94f 2343 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2344 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2345 if (ret == -EADDRINUSE)
2346 ret = 0;
56ef9c90 2347 if (ret) {
205f356d 2348 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2349 return ret;
2350 }
d342894c 2351 }
2352
0dfbdf41 2353 if (vxlan->cfg.age_interval)
d342894c 2354 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2355
56ef9c90 2356 return ret;
d342894c 2357}
2358
2359/* Purge the forwarding table */
2360static void vxlan_flush(struct vxlan_dev *vxlan)
2361{
31fec5aa 2362 unsigned int h;
d342894c 2363
2364 spin_lock_bh(&vxlan->hash_lock);
2365 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2366 struct hlist_node *p, *n;
2367 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2368 struct vxlan_fdb *f
2369 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2370 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2371 if (!is_zero_ether_addr(f->eth_addr))
2372 vxlan_fdb_destroy(vxlan, f);
d342894c 2373 }
2374 }
2375 spin_unlock_bh(&vxlan->hash_lock);
2376}
2377
2378/* Cleanup timer and forwarding table on shutdown */
2379static int vxlan_stop(struct net_device *dev)
2380{
2381 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2382 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2383 int ret = 0;
d342894c 2384
24c0e683 2385 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2386 !vxlan_group_used(vn, vxlan))
56ef9c90 2387 ret = vxlan_igmp_leave(vxlan);
d342894c 2388
2389 del_timer_sync(&vxlan->age_timer);
2390
2391 vxlan_flush(vxlan);
205f356d 2392 vxlan_sock_release(vxlan);
d342894c 2393
56ef9c90 2394 return ret;
d342894c 2395}
2396
d342894c 2397/* Stub, nothing needs to be done. */
2398static void vxlan_set_multicast_list(struct net_device *dev)
2399{
2400}
2401
72564b59
DW		 2402static int __vxlan_change_mtu(struct net_device *dev,
2403 struct net_device *lowerdev,
2404 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2405{
72564b59 2406 int max_mtu = IP_MAX_MTU;
345010b5 2407
72564b59
DW		 2408 if (lowerdev)
2409 max_mtu = lowerdev->mtu;
345010b5
DB		 2410
2411 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2412 max_mtu -= VXLAN6_HEADROOM;
345010b5 2413 else
72564b59 2414 max_mtu -= VXLAN_HEADROOM;
345010b5 2415
72564b59 2416 if (new_mtu < 68)
345010b5
DB		 2417 return -EINVAL;
2418
72564b59
DW		 2419 if (new_mtu > max_mtu) {
2420 if (strict)
2421 return -EINVAL;
2422
2423 new_mtu = max_mtu;
2424 }
2425
345010b5
DB		 2426 dev->mtu = new_mtu;
2427 return 0;
2428}
2429
72564b59
DW		 2430static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2431{
2432 struct vxlan_dev *vxlan = netdev_priv(dev);
2433 struct vxlan_rdst *dst = &vxlan->default_dst;
2434 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2435 dst->remote_ifindex);
2436 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2437}
2438
fc4099f1
PS		 2439static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2440{
2441 struct vxlan_dev *vxlan = netdev_priv(dev);
2442 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2443 __be16 sport, dport;
2444
2445 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2446 vxlan->cfg.port_max, true);
2447 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2448
239e944f 2449 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB		 2450 struct rtable *rt;
2451
239e944f
JB		 2452 if (!vxlan->vn4_sock)
2453 return -EINVAL;
1a8496ba
JB		 2454 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2455 info->key.u.ipv4.dst,
0c1d70af 2456 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2457 if (IS_ERR(rt))
2458 return PTR_ERR(rt);
2459 ip_rt_put(rt);
239e944f
JB		 2460 } else {
2461#if IS_ENABLED(CONFIG_IPV6)
2462 struct dst_entry *ndst;
2463
2464 if (!vxlan->vn6_sock)
2465 return -EINVAL;
1400615d 2466 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
e7f70af1 2467 info->key.label, &info->key.u.ipv6.dst,
db3c6139 2468 &info->key.u.ipv6.src, NULL, info);
239e944f
JB		 2469 if (IS_ERR(ndst))
2470 return PTR_ERR(ndst);
2471 dst_release(ndst);
239e944f
JB		 2472#else /* !CONFIG_IPV6 */
2473 return -EPFNOSUPPORT;
2474#endif
2475 }
1a8496ba
JB		 2476 info->key.tp_src = sport;
2477 info->key.tp_dst = dport;
239e944f 2478 return 0;
fc4099f1
PS		 2479}
2480
0c867c9b 2481static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2482 .ndo_init = vxlan_init,
ebf4063e 2483 .ndo_uninit = vxlan_uninit,
d342894c 2484 .ndo_open = vxlan_open,
2485 .ndo_stop = vxlan_stop,
2486 .ndo_start_xmit = vxlan_xmit,
e8171045 2487 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2488 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2489 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2490 .ndo_validate_addr = eth_validate_addr,
2491 .ndo_set_mac_address = eth_mac_addr,
2492 .ndo_fdb_add = vxlan_fdb_add,
2493 .ndo_fdb_del = vxlan_fdb_delete,
2494 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2495 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2496};
2497
e1e5314d
JB		 2498static const struct net_device_ops vxlan_netdev_raw_ops = {
2499 .ndo_init = vxlan_init,
2500 .ndo_uninit = vxlan_uninit,
2501 .ndo_open = vxlan_open,
2502 .ndo_stop = vxlan_stop,
2503 .ndo_start_xmit = vxlan_xmit,
2504 .ndo_get_stats64 = ip_tunnel_get_stats64,
2505 .ndo_change_mtu = vxlan_change_mtu,
2506 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2507};
2508
d342894c 2509/* Info for udev, that this is a virtual tunnel endpoint */
2510static struct device_type vxlan_type = {
2511 .name = "vxlan",
2512};
2513
53cf5275 2514/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2515 * supply the listening VXLAN udp ports. Callers are expected
2516 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2517 */
2518void vxlan_get_rx_port(struct net_device *dev)
2519{
2520 struct vxlan_sock *vs;
2521 struct net *net = dev_net(dev);
2522 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2523 sa_family_t sa_family;
35e42379
JG		 2524 __be16 port;
2525 unsigned int i;
53cf5275
JG		 2526
2527 spin_lock(&vn->sock_lock);
2528 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2529 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2530 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2531 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2532 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2533 port);
2534 }
2535 }
2536 spin_unlock(&vn->sock_lock);
2537}
2538EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2539
d342894c 2540/* Initialize the device structure. */
2541static void vxlan_setup(struct net_device *dev)
2542{
2543 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2544 unsigned int h;
d342894c 2545
ebf4063e 2546 dev->destructor = free_netdev;
d342894c 2547 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2548
d342894c 2549 dev->features |= NETIF_F_LLTX;
d6727fe3 2550 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2551 dev->features |= NETIF_F_RXCSUM;
05c0db08 2552 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2553
1eaa8178
PS		 2554 dev->vlan_features = dev->features;
2555 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2556 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2557 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2558 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2559 netif_keep_dst(dev);
0c867c9b 2560 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2561
553675fb 2562 INIT_LIST_HEAD(&vxlan->next);
d342894c 2563 spin_lock_init(&vxlan->hash_lock);
2564
2565 init_timer_deferrable(&vxlan->age_timer);
2566 vxlan->age_timer.function = vxlan_cleanup;
2567 vxlan->age_timer.data = (unsigned long) vxlan;
2568
0dfbdf41 2569 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2570
d342894c 2571 vxlan->dev = dev;
2572
58ce31cc
TH		 2573 gro_cells_init(&vxlan->gro_cells, dev);
2574
d342894c 2575 for (h = 0; h < FDB_HASH_SIZE; ++h)
2576 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2577}
2578
0c867c9b
JB		 2579static void vxlan_ether_setup(struct net_device *dev)
2580{
2581 eth_hw_addr_random(dev);
2582 ether_setup(dev);
2583 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2584 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2585 dev->netdev_ops = &vxlan_netdev_ether_ops;
2586}
2587
e1e5314d
JB		 2588static void vxlan_raw_setup(struct net_device *dev)
2589{
2590 dev->type = ARPHRD_NONE;
2591 dev->hard_header_len = 0;
2592 dev->addr_len = 0;
2593 dev->mtu = ETH_DATA_LEN;
2594 dev->tx_queue_len = 1000;
2595 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2596 dev->netdev_ops = &vxlan_netdev_raw_ops;
2597}
2598
d342894c 2599static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2600 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2601 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2602 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2603 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2604 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2605 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2606 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2607 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2608 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2609 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2610 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2611 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2612 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2613 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2614 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2615 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2616 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2617 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2618 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2619 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2620 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2621 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2622 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2623 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2624 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2625 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2626 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2627};
2628
2629static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2630{
2631 if (tb[IFLA_ADDRESS]) {
2632 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2633 pr_debug("invalid link address (not ethernet)\n");
2634 return -EINVAL;
2635 }
2636
2637 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2638 pr_debug("invalid all zero ethernet address\n");
2639 return -EADDRNOTAVAIL;
2640 }
2641 }
2642
2643 if (!data)
2644 return -EINVAL;
2645
2646 if (data[IFLA_VXLAN_ID]) {
2647 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2648 if (id >= VXLAN_VID_MASK)
2649 return -ERANGE;
2650 }
2651
05f47d69 2652 if (data[IFLA_VXLAN_PORT_RANGE]) {
2653 const struct ifla_vxlan_port_range *p
2654 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2655
2656 if (ntohs(p->high) < ntohs(p->low)) {
2657 pr_debug("port range %u .. %u not valid\n",
2658 ntohs(p->low), ntohs(p->high));
2659 return -EINVAL;
2660 }
2661 }
2662
d342894c 2663 return 0;
2664}
2665
1b13c97f
YB		 2666static void vxlan_get_drvinfo(struct net_device *netdev,
2667 struct ethtool_drvinfo *drvinfo)
2668{
2669 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2670 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2671}
2672
2673static const struct ethtool_ops vxlan_ethtool_ops = {
2674 .get_drvinfo = vxlan_get_drvinfo,
2675 .get_link = ethtool_op_get_link,
2676};
2677
3ee64f39
TH		 2678static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2679 __be16 port, u32 flags)
553675fb 2680{
e4c7ed41 2681 struct socket *sock;
3ee64f39
TH		 2682 struct udp_port_cfg udp_conf;
2683 int err;
e4c7ed41 2684
3ee64f39 2685 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2686
3ee64f39
TH		 2687 if (ipv6) {
2688 udp_conf.family = AF_INET6;
3ee64f39 2689 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2690 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2691 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2692 } else {
2693 udp_conf.family = AF_INET;
e4c7ed41 2694 }
e4c7ed41 2695
3ee64f39 2696 udp_conf.local_udp_port = port;
553675fb 2697
3ee64f39
TH		 2698 /* Open UDP socket */
2699 err = udp_sock_create(net, &udp_conf, &sock);
2700 if (err < 0)
2701 return ERR_PTR(err);
e4c7ed41 2702
39deb2c7 2703 return sock;
e4c7ed41
CW		 2704}
2705
2706/* Create new listen socket if needed */
b1be00a6
JB		 2707static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2708 __be16 port, u32 flags)
e4c7ed41
CW		 2709{
2710 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2711 struct vxlan_sock *vs;
2712 struct socket *sock;
e4c7ed41 2713 unsigned int h;
acbf74a7 2714 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2715
dc01e7d3 2716 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2717 if (!vs)
2718 return ERR_PTR(-ENOMEM);
2719
2720 for (h = 0; h < VNI_HASH_SIZE; ++h)
2721 INIT_HLIST_HEAD(&vs->vni_list[h]);
2722
3ee64f39 2723 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2724 if (IS_ERR(sock)) {
56ef9c90
MRL		 2725 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2726 PTR_ERR(sock));
553675fb 2727 kfree(vs);
e50fddc8 2728 return ERR_CAST(sock);
553675fb 2729 }
e4c7ed41
CW		 2730
2731 vs->sock = sock;
9c2e24e1 2732 atomic_set(&vs->refcnt, 1);
af33c1ad 2733 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2734
9c2e24e1
PS		 2735 spin_lock(&vn->sock_lock);
2736 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2737 vxlan_notify_add_rx_port(vs);
9c2e24e1 2738 spin_unlock(&vn->sock_lock);
553675fb 2739
2740 /* Mark socket as an encapsulation socket. */
5602c48c 2741 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2742 tunnel_cfg.sk_user_data = vs;
2743 tunnel_cfg.encap_type = 1;
f2d1968e 2744 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2745 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2746 tunnel_cfg.gro_receive = vxlan_gro_receive;
2747 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2748
2749 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2750
9c2e24e1
PS		 2751 return vs;
2752}
2753
b1be00a6 2754static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2755{
205f356d
JB		 2756 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2757 struct vxlan_sock *vs = NULL;
9c2e24e1 2758
205f356d 2759 if (!vxlan->cfg.no_share) {
56ef9c90 2760 spin_lock(&vn->sock_lock);
205f356d
JB		 2761 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2762 vxlan->cfg.dst_port, vxlan->flags);
2763 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2764 spin_unlock(&vn->sock_lock);
205f356d 2765 return -EBUSY;
56ef9c90
MRL		 2766 }
2767 spin_unlock(&vn->sock_lock);
2768 }
205f356d 2769 if (!vs)
b1be00a6
JB		 2770 vs = vxlan_socket_create(vxlan->net, ipv6,
2771 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2772 if (IS_ERR(vs))
2773 return PTR_ERR(vs);
b1be00a6
JB		 2774#if IS_ENABLED(CONFIG_IPV6)
2775 if (ipv6)
2776 vxlan->vn6_sock = vs;
2777 else
2778#endif
2779 vxlan->vn4_sock = vs;
205f356d
JB		 2780 vxlan_vs_add_dev(vs, vxlan);
2781 return 0;
553675fb 2782}
2783
b1be00a6
JB		 2784static int vxlan_sock_add(struct vxlan_dev *vxlan)
2785{
2786 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2787 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2788 int ret = 0;
2789
2790 vxlan->vn4_sock = NULL;
2791#if IS_ENABLED(CONFIG_IPV6)
2792 vxlan->vn6_sock = NULL;
2793 if (ipv6 || metadata)
2794 ret = __vxlan_sock_add(vxlan, true);
2795#endif
2796 if (!ret && (!ipv6 || metadata))
2797 ret = __vxlan_sock_add(vxlan, false);
2798 if (ret < 0)
2799 vxlan_sock_release(vxlan);
2800 return ret;
2801}
2802
0dfbdf41
TG		 2803static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2804 struct vxlan_config *conf)
d342894c 2805{
33564bbb 2806 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2807 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2808 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2809 unsigned short needed_headroom = ETH_HLEN;
d342894c 2810 int err;
e4c7ed41 2811 bool use_ipv6 = false;
0dfbdf41 2812 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2813 struct net_device *lowerdev = NULL;
d342894c 2814
e1e5314d
JB		 2815 if (conf->flags & VXLAN_F_GPE) {
2816 if (conf->flags & ~VXLAN_F_ALLOWED_GPE)
2817 return -EINVAL;
2818 /* For now, allow GPE only together with COLLECT_METADATA.
2819 * This can be relaxed later; in such case, the other side
2820 * of the PtP link will have to be provided.
2821 */
2822 if (!(conf->flags & VXLAN_F_COLLECT_METADATA))
2823 return -EINVAL;
2824
2825 vxlan_raw_setup(dev);
2826 } else {
2827 vxlan_ether_setup(dev);
2828 }
0c867c9b 2829
33564bbb 2830 vxlan->net = src_net;
f01ec1c0 2831
0dfbdf41 2832 dst->remote_vni = conf->vni;
d342894c 2833
0dfbdf41 2834 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2835
0dfbdf41
TG		 2836 /* Unless IPv6 is explicitly requested, assume IPv4 */
2837 if (!dst->remote_ip.sa.sa_family)
2838 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2839
0dfbdf41 2840 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2841 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2842 if (!IS_ENABLED(CONFIG_IPV6))
2843 return -EPFNOSUPPORT;
e4c7ed41 2844 use_ipv6 = true;
b1be00a6 2845 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2846 }
d342894c 2847
e7f70af1
DB		 2848 if (conf->label && !use_ipv6) {
2849 pr_info("label only supported in use with IPv6\n");
2850 return -EINVAL;
2851 }
2852
0dfbdf41 2853 if (conf->remote_ifindex) {
7e059158 2854 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2855 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2856
2857 if (!lowerdev) {
c7995c43 2858 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2859 return -ENODEV;
2860 }
d342894c 2861
e4c7ed41
CW		 2862#if IS_ENABLED(CONFIG_IPV6)
2863 if (use_ipv6) {
2864 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2865 if (idev && idev->cnf.disable_ipv6) {
2866 pr_info("IPv6 is disabled via sysctl\n");
2867 return -EPERM;
2868 }
e4c7ed41
CW		 2869 }
2870#endif
2871
0dfbdf41 2872 if (!conf->mtu)
e4c7ed41 2873 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2874
b1be00a6 2875 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2876 }
d342894c 2877
7e059158
DW		 2878 if (conf->mtu) {
2879 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2880 if (err)
2881 return err;
2882 }
2883
b1be00a6
JB		 2884 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2885 needed_headroom += VXLAN6_HEADROOM;
2886 else
2887 needed_headroom += VXLAN_HEADROOM;
2888 dev->needed_headroom = needed_headroom;
2889
0dfbdf41 2890 memcpy(&vxlan->cfg, conf, sizeof(*conf));
e1e5314d
JB		 2891 if (!vxlan->cfg.dst_port) {
2892 if (conf->flags & VXLAN_F_GPE)
2893 vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */
2894 else
2895 vxlan->cfg.dst_port = default_port;
2896 }
0dfbdf41
TG		 2897 vxlan->flags |= conf->flags;
2898
2899 if (!vxlan->cfg.age_interval)
2900 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2901
07b9b37c
ND		 2902 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2903 if (tmp->cfg.vni == conf->vni &&
2904 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2905 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2906 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2907 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2908 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2909 return -EEXIST;
07b9b37c 2910 }
0dfbdf41
TG		 2911
2912 dev->ethtool_ops = &vxlan_ethtool_ops;
2913
2914 /* create an fdb entry for a valid default destination */
2915 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2916 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2917 &vxlan->default_dst.remote_ip,
2918 NUD_REACHABLE|NUD_PERMANENT,
2919 NLM_F_EXCL|NLM_F_CREATE,
2920 vxlan->cfg.dst_port,
2921 vxlan->default_dst.remote_vni,
2922 vxlan->default_dst.remote_ifindex,
2923 NTF_SELF);
2924 if (err)
2925 return err;
2926 }
2927
2928 err = register_netdevice(dev);
2929 if (err) {
2930 vxlan_fdb_delete_default(vxlan);
2931 return err;
2932 }
2933
2934 list_add(&vxlan->next, &vn->vxlan_list);
2935
2936 return 0;
2937}
2938
2939struct net_device *vxlan_dev_create(struct net *net, const char *name,
2940 u8 name_assign_type, struct vxlan_config *conf)
2941{
2942 struct nlattr *tb[IFLA_MAX+1];
2943 struct net_device *dev;
2944 int err;
2945
2946 memset(&tb, 0, sizeof(tb));
2947
2948 dev = rtnl_create_link(net, name, name_assign_type,
2949 &vxlan_link_ops, tb);
2950 if (IS_ERR(dev))
2951 return dev;
2952
2953 err = vxlan_dev_configure(net, dev, conf);
2954 if (err < 0) {
2955 free_netdev(dev);
2956 return ERR_PTR(err);
2957 }
2958
2959 return dev;
2960}
2961EXPORT_SYMBOL_GPL(vxlan_dev_create);
2962
2963static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2964 struct nlattr *tb[], struct nlattr *data[])
2965{
2966 struct vxlan_config conf;
2967 int err;
2968
0dfbdf41 2969 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2970
2971 if (data[IFLA_VXLAN_ID])
54bfd872 2972 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG		 2973
2974 if (data[IFLA_VXLAN_GROUP]) {
2975 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2976 } else if (data[IFLA_VXLAN_GROUP6]) {
2977 if (!IS_ENABLED(CONFIG_IPV6))
2978 return -EPFNOSUPPORT;
2979
2980 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2981 conf.remote_ip.sa.sa_family = AF_INET6;
2982 }
2983
2984 if (data[IFLA_VXLAN_LOCAL]) {
2985 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2986 conf.saddr.sa.sa_family = AF_INET;
2987 } else if (data[IFLA_VXLAN_LOCAL6]) {
2988 if (!IS_ENABLED(CONFIG_IPV6))
2989 return -EPFNOSUPPORT;
2990
2991 /* TODO: respect scope id */
2992 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2993 conf.saddr.sa.sa_family = AF_INET6;
2994 }
2995
2996 if (data[IFLA_VXLAN_LINK])
2997 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2998
d342894c 2999 if (data[IFLA_VXLAN_TOS])
0dfbdf41 3000 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3001
afb97186 3002 if (data[IFLA_VXLAN_TTL])
0dfbdf41 3003 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3004
e7f70af1
DB		 3005 if (data[IFLA_VXLAN_LABEL])
3006 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3007 IPV6_FLOWLABEL_MASK;
3008
d342894c 3009 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 3010 conf.flags |= VXLAN_F_LEARN;
d342894c 3011
3012 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 3013 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 3014
e4f67add 3015 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 3016 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 3017
3018 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 3019 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 3020
3021 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 3022 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 3023
3024 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 3025 conf.flags |= VXLAN_F_L3MISS;
e4f67add 3026
d342894c 3027 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 3028 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 3029
f8a9b1bc
AS		 3030 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
3031 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3032 conf.flags |= VXLAN_F_COLLECT_METADATA;
3033
05f47d69 3034 if (data[IFLA_VXLAN_PORT_RANGE]) {
3035 const struct ifla_vxlan_port_range *p
3036 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 3037 conf.port_min = ntohs(p->low);
3038 conf.port_max = ntohs(p->high);
05f47d69 3039 }
3040
823aa873 3041 if (data[IFLA_VXLAN_PORT])
0dfbdf41 3042 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 3043
6ceb31ca
AD		 3044 if (data[IFLA_VXLAN_UDP_CSUM] &&
3045 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3046 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH		 3047
3048 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
3049 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 3050 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 3051
3052 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
3053 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 3054 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 3055
dfd8645e
TH		 3056 if (data[IFLA_VXLAN_REMCSUM_TX] &&
3057 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 3058 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 3059
3060 if (data[IFLA_VXLAN_REMCSUM_RX] &&
3061 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 3062 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 3063
3511494c 3064 if (data[IFLA_VXLAN_GBP])
0dfbdf41 3065 conf.flags |= VXLAN_F_GBP;
3511494c 3066
e1e5314d
JB		 3067 if (data[IFLA_VXLAN_GPE])
3068 conf.flags |= VXLAN_F_GPE;
3069
0ace2ca8 3070 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 3071 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 3072
0dfbdf41
TG		 3073 err = vxlan_dev_configure(src_net, dev, &conf);
3074 switch (err) {
3075 case -ENODEV:
3076 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
3077 break;
1b13c97f 3078
0dfbdf41
TG		 3079 case -EPERM:
3080 pr_info("IPv6 is disabled via sysctl\n");
3081 break;
d342894c 3082
0dfbdf41 3083 case -EEXIST:
54bfd872 3084 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
0dfbdf41 3085 break;
e1e5314d
JB		 3086
3087 case -EINVAL:
3088 pr_info("unsupported combination of extensions\n");
3089 break;
afbd8bae
MR		 3090 }
3091
0dfbdf41 3092 return err;
d342894c 3093}
3094
3095static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3096{
3097 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3098 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 3099
fe5c3561 3100 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 3101 if (!hlist_unhashed(&vxlan->hlist))
3102 hlist_del_rcu(&vxlan->hlist);
fe5c3561 3103 spin_unlock(&vn->sock_lock);
3104
58ce31cc 3105 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3106 list_del(&vxlan->next);
d342894c 3107 unregister_netdevice_queue(dev, head);
3108}
3109
3110static size_t vxlan_get_size(const struct net_device *dev)
3111{
3112
3113 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3114 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3115 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3116 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3117 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3118 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3119 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3120 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3121 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3122 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3123 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3124 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3125 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3126 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3127 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3128 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3129 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3130 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3131 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3132 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3133 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3134 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3135 0;
3136}
3137
3138static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3139{
3140 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3141 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3142 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3143 .low = htons(vxlan->cfg.port_min),
3144 .high = htons(vxlan->cfg.port_max),
05f47d69 3145 };
d342894c 3146
54bfd872 3147 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3148 goto nla_put_failure;
3149
e4c7ed41
CW		 3150 if (!vxlan_addr_any(&dst->remote_ip)) {
3151 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3152 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3153 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3154 goto nla_put_failure;
3155#if IS_ENABLED(CONFIG_IPV6)
3156 } else {
930345ea
JB		 3157 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3158 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3159 goto nla_put_failure;
3160#endif
3161 }
3162 }
d342894c 3163
c7995c43 3164 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3165 goto nla_put_failure;
3166
0dfbdf41
TG		 3167 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3168 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3169 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3170 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3171 goto nla_put_failure;
3172#if IS_ENABLED(CONFIG_IPV6)
3173 } else {
930345ea 3174 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3175 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3176 goto nla_put_failure;
3177#endif
3178 }
3179 }
d342894c 3180
0dfbdf41
TG		 3181 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3182 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3183 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3184 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3185 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3186 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3187 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3188 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3189 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3190 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3191 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3192 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3193 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3194 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3195 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3196 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3197 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3198 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3199 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3200 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3201 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3202 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3203 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3204 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3205 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3206 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3207 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3208 goto nla_put_failure;
3209
05f47d69 3210 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3211 goto nla_put_failure;
3212
3511494c
TG		 3213 if (vxlan->flags & VXLAN_F_GBP &&
3214 nla_put_flag(skb, IFLA_VXLAN_GBP))
3215 goto nla_put_failure;
3216
e1e5314d
JB		 3217 if (vxlan->flags & VXLAN_F_GPE &&
3218 nla_put_flag(skb, IFLA_VXLAN_GPE))
3219 goto nla_put_failure;
3220
0ace2ca8
TH		 3221 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3222 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3223 goto nla_put_failure;
3224
d342894c 3225 return 0;
3226
3227nla_put_failure:
3228 return -EMSGSIZE;
3229}
3230
1728d4fa
ND		 3231static struct net *vxlan_get_link_net(const struct net_device *dev)
3232{
3233 struct vxlan_dev *vxlan = netdev_priv(dev);
3234
3235 return vxlan->net;
3236}
3237
d342894c 3238static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3239 .kind = "vxlan",
3240 .maxtype = IFLA_VXLAN_MAX,
3241 .policy = vxlan_policy,
3242 .priv_size = sizeof(struct vxlan_dev),
3243 .setup = vxlan_setup,
3244 .validate = vxlan_validate,
3245 .newlink = vxlan_newlink,
3246 .dellink = vxlan_dellink,
3247 .get_size = vxlan_get_size,
3248 .fill_info = vxlan_fill_info,
1728d4fa 3249 .get_link_net = vxlan_get_link_net,
d342894c 3250};
3251
acaf4e70
DB		 3252static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3253 struct net_device *dev)
3254{
3255 struct vxlan_dev *vxlan, *next;
3256 LIST_HEAD(list_kill);
3257
3258 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3259 struct vxlan_rdst *dst = &vxlan->default_dst;
3260
3261 /* In case we created vxlan device with carrier
3262 * and we loose the carrier due to module unload
3263 * we also need to remove vxlan device. In other
3264 * cases, it's not necessary and remote_ifindex
3265 * is 0 here, so no matches.
3266 */
3267 if (dst->remote_ifindex == dev->ifindex)
3268 vxlan_dellink(vxlan->dev, &list_kill);
3269 }
3270
3271 unregister_netdevice_many(&list_kill);
3272}
3273
3274static int vxlan_lowerdev_event(struct notifier_block *unused,
3275 unsigned long event, void *ptr)
3276{
3277 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3278 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3279
783c1463 3280 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3281 vxlan_handle_lowerdev_unregister(vn, dev);
3282
3283 return NOTIFY_DONE;
3284}
3285
3286static struct notifier_block vxlan_notifier_block __read_mostly = {
3287 .notifier_call = vxlan_lowerdev_event,
3288};
3289
d342894c 3290static __net_init int vxlan_init_net(struct net *net)
3291{
3292 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3293 unsigned int h;
d342894c 3294
553675fb 3295 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3296 spin_lock_init(&vn->sock_lock);
d342894c 3297
553675fb 3298 for (h = 0; h < PORT_HASH_SIZE; ++h)
3299 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3300
3301 return 0;
3302}
3303
f01ec1c0
ND		 3304static void __net_exit vxlan_exit_net(struct net *net)
3305{
3306 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3307 struct vxlan_dev *vxlan, *next;
3308 struct net_device *dev, *aux;
3309 LIST_HEAD(list);
3310
3311 rtnl_lock();
3312 for_each_netdev_safe(net, dev, aux)
3313 if (dev->rtnl_link_ops == &vxlan_link_ops)
3314 unregister_netdevice_queue(dev, &list);
3315
3316 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3317 /* If vxlan->dev is in the same netns, it has already been added
3318 * to the list by the previous loop.
3319 */
58ce31cc
TH		 3320 if (!net_eq(dev_net(vxlan->dev), net)) {
3321 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3322 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3323 }
f01ec1c0
ND		 3324 }
3325
3326 unregister_netdevice_many(&list);
3327 rtnl_unlock();
3328}
3329
d342894c 3330static struct pernet_operations vxlan_net_ops = {
3331 .init = vxlan_init_net,
f01ec1c0 3332 .exit = vxlan_exit_net,
d342894c 3333 .id = &vxlan_net_id,
3334 .size = sizeof(struct vxlan_net),
3335};
3336
3337static int __init vxlan_init_module(void)
3338{
3339 int rc;
3340
3341 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3342
783c1463 3343 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3344 if (rc)
3345 goto out1;
3346
acaf4e70 3347 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3348 if (rc)
3349 goto out2;
3350
acaf4e70
DB		 3351 rc = rtnl_link_register(&vxlan_link_ops);
3352 if (rc)
3353 goto out3;
d342894c 3354
acaf4e70
DB		 3355 return 0;
3356out3:
3357 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3358out2:
783c1463 3359 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3360out1:
3361 return rc;
3362}
7332a13b 3363late_initcall(vxlan_init_module);
d342894c 3364
3365static void __exit vxlan_cleanup_module(void)
3366{
b7153984 3367 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3368 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3369 unregister_pernet_subsys(&vxlan_net_ops);
3370 /* rcu_barrier() is called by netns */
d342894c 3371}
3372module_exit(vxlan_cleanup_module);
3373
3374MODULE_LICENSE("GPL");
3375MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3376MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3377MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3378MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)