]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ip_tunnel: add support for setting flow label via collect metadata
[mirror_ubuntu-hirsute-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
40d29af0 45
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
ee122c79 52#include <net/dst_metadata.h>
d342894c 53
54#define VXLAN_VERSION "0.1"
55
553675fb 56#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
23c578bf 61/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 63 * for compatibility with early adopters.
23c578bf 64 */
9daaa397
SH		 65static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 67MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
60d9d4c6 73static int vxlan_net_id;
0dfbdf41 74static struct rtnl_link_ops vxlan_link_ops;
553675fb 75
7256eac1 76static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 77
205f356d 78static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
d342894c 87/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
3e61aa8f 93 struct list_head remotes;
7177a3b0 94 u8 eth_addr[ETH_ALEN];
d342894c 95 u16 state; /* see ndm_state */
ae884082 96 u8 flags; /* see ndm_flags */
d342894c 97};
98
d342894c 99/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
758c57d1 101static struct workqueue_struct *vxlan_wq;
d342894c 102
ee122c79
TG		 103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
e7030878
TG		 105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
ee122c79
TG		 107}
108
e4c7ed41
CW		 109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
f0ef3126
JB		 113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
f0ef3126
JB		 123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
f0ef3126
JB		 131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
f0ef3126 139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
e4c7ed41
CW		 150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 153 const union vxlan_addr *ip)
e4c7ed41 154{
f0ef3126 155 if (ip->sa.sa_family == AF_INET6)
930345ea 156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 157 else
930345ea 158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
f0ef3126 166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
f0ef3126 171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
f0ef3126 176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
f0ef3126
JB		 181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
e4c7ed41
CW		 190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 193 const union vxlan_addr *ip)
e4c7ed41 194{
930345ea 195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 196}
197#endif
198
553675fb 199/* Virtual Network hash table head */
54bfd872 200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 201{
54bfd872 202 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
553675fb 210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
3e61aa8f
SH		 213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
5ca5461c 216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 217{
5ca5461c 218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 224}
225
ac5132d1
TG		 226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
553675fb 231{
232 struct vxlan_sock *vs;
af33c1ad
TH		 233
234 flags &= VXLAN_F_RCV_FLAGS;
553675fb 235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 238 vxlan_get_sk_family(vs) == family &&
af33c1ad 239 vs->flags == flags)
553675fb 240 return vs;
241 }
242 return NULL;
d342894c 243}
244
54bfd872 245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 246{
247 struct vxlan_dev *vxlan;
d342894c 248
b9167b2e
JB		 249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs->flags & VXLAN_F_COLLECT_METADATA)
251 vni = 0;
252
54bfd872
JB		 253 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
254 if (vxlan->default_dst.remote_vni == vni)
d342894c 255 return vxlan;
256 }
257
258 return NULL;
259}
260
5cfccc5a 261/* Look up VNI in a per net namespace table */
54bfd872 262static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 263 sa_family_t family, __be16 port,
264 u32 flags)
5cfccc5a
PS		 265{
266 struct vxlan_sock *vs;
267
ac5132d1 268 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 269 if (!vs)
270 return NULL;
271
54bfd872 272 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 273}
274
d342894c 275/* Fill in neighbour message in skbuff. */
276static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 277 const struct vxlan_fdb *fdb,
278 u32 portid, u32 seq, int type, unsigned int flags,
279 const struct vxlan_rdst *rdst)
d342894c 280{
281 unsigned long now = jiffies;
282 struct nda_cacheinfo ci;
283 struct nlmsghdr *nlh;
284 struct ndmsg *ndm;
e4f67add 285 bool send_ip, send_eth;
d342894c 286
287 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
288 if (nlh == NULL)
289 return -EMSGSIZE;
290
291 ndm = nlmsg_data(nlh);
292 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 293
294 send_eth = send_ip = true;
295
296 if (type == RTM_GETNEIGH) {
297 ndm->ndm_family = AF_INET;
e4c7ed41 298 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 299 send_eth = !is_zero_ether_addr(fdb->eth_addr);
300 } else
301 ndm->ndm_family = AF_BRIDGE;
d342894c 302 ndm->ndm_state = fdb->state;
303 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 304 ndm->ndm_flags = fdb->flags;
545469f7 305 ndm->ndm_type = RTN_UNICAST;
d342894c 306
193523bf 307 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 308 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 309 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 310 goto nla_put_failure;
311
e4f67add 312 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 313 goto nla_put_failure;
314
e4c7ed41 315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 316 goto nla_put_failure;
317
0dfbdf41 318 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 319 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
320 goto nla_put_failure;
c7995c43 321 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 322 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d
DS		 323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
053c095a
JB		 336 nlmsg_end(skb, nlh);
337 return 0;
d342894c 338
339nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342}
343
344static inline size_t vxlan_nlmsg_size(void)
345{
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 353 + nla_total_size(sizeof(struct nda_cacheinfo));
354}
355
9e4b93f9
ND		 356static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
d342894c 358{
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
9e4b93f9 367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380}
381
e4c7ed41 382static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 383{
384 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
e4c7ed41 389 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 391 };
3e61aa8f 392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 394}
395
396static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397{
bb3fd687
SH		 398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
9e4b93f9 401 struct vxlan_rdst remote = { };
e4f67add 402
e4f67add
DS		 403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
9e4b93f9 405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 406}
407
d342894c 408/* Hash Ethernet address */
409static u32 eth_hash(const unsigned char *addr)
410{
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414#ifdef __BIG_ENDIAN
d342894c 415 value >>= 16;
321fb991 416#else
417 value <<= 16;
d342894c 418#endif
419 return hash_64(value, FDB_HASH_BITS);
420}
421
422/* Hash chain to use given mac address */
423static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
424 const u8 *mac)
425{
426 return &vxlan->fdb_head[eth_hash(mac)];
427}
428
429/* Look up Ethernet address in forwarding table */
014be2c8 430static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 431 const u8 *mac)
d342894c 432{
433 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
434 struct vxlan_fdb *f;
d342894c 435
b67bfe0d 436 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 437 if (ether_addr_equal(mac, f->eth_addr))
d342894c 438 return f;
439 }
440
441 return NULL;
442}
443
014be2c8
SS		 444static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
445 const u8 *mac)
446{
447 struct vxlan_fdb *f;
448
449 f = __vxlan_find_mac(vxlan, mac);
450 if (f)
451 f->used = jiffies;
452
453 return f;
454}
455
a5e7c10a
MR		 456/* caller should hold vxlan->hash_lock */
457static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 458 union vxlan_addr *ip, __be16 port,
54bfd872 459 __be32 vni, __u32 ifindex)
6681712d 460{
3e61aa8f 461 struct vxlan_rdst *rd;
6681712d 462
3e61aa8f 463 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 464 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 465 rd->remote_port == port &&
466 rd->remote_vni == vni &&
467 rd->remote_ifindex == ifindex)
a5e7c10a 468 return rd;
6681712d 469 }
3e61aa8f 470
a5e7c10a
MR		 471 return NULL;
472}
473
906dc186
TR		 474/* Replace destination of unicast mac */
475static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 476 union vxlan_addr *ip, __be16 port, __be32 vni,
477 __u32 ifindex)
906dc186
TR		 478{
479 struct vxlan_rdst *rd;
480
481 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
482 if (rd)
483 return 0;
484
485 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
486 if (!rd)
487 return 0;
0c1d70af
PA		 488
489 dst_cache_reset(&rd->dst_cache);
e4c7ed41 490 rd->remote_ip = *ip;
906dc186
TR		 491 rd->remote_port = port;
492 rd->remote_vni = vni;
493 rd->remote_ifindex = ifindex;
494 return 1;
495}
496
a5e7c10a
MR		 497/* Add/update destinations for multicast */
498static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 499 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 500 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 501{
502 struct vxlan_rdst *rd;
503
504 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
505 if (rd)
506 return 0;
507
6681712d
DS		 508 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
509 if (rd == NULL)
510 return -ENOBUFS;
0c1d70af
PA		 511
512 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
513 kfree(rd);
514 return -ENOBUFS;
515 }
516
e4c7ed41 517 rd->remote_ip = *ip;
6681712d
DS		 518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 521
522 list_add_tail_rcu(&rd->list, &f->remotes);
523
9e4b93f9 524 *rdp = rd;
6681712d
DS		 525 return 1;
526}
527
dfd8645e
TH		 528static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
529 unsigned int off,
530 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 531 __be32 vni_field,
532 struct gro_remcsum *grc,
0ace2ca8 533 bool nopartial)
dfd8645e 534{
b7fe10e5 535 size_t start, offset;
dfd8645e
TH		 536
537 if (skb->remcsum_offload)
b7fe10e5 538 return vh;
dfd8645e
TH		 539
540 if (!NAPI_GRO_CB(skb)->csum_valid)
541 return NULL;
542
54bfd872
JB		 543 start = vxlan_rco_start(vni_field);
544 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 545
b7fe10e5
TH		 546 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
547 start, offset, grc, nopartial);
dfd8645e
TH		 548
549 skb->remcsum_offload = 1;
550
551 return vh;
552}
553
a2b12f3c
TH		 554static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
555 struct sk_buff *skb,
556 struct udp_offload *uoff)
dc01e7d3
OG		 557{
558 struct sk_buff *p, **pp = NULL;
559 struct vxlanhdr *vh, *vh2;
9b174d88 560 unsigned int hlen, off_vx;
dc01e7d3 561 int flush = 1;
dfd8645e
TH		 562 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
563 udp_offloads);
54bfd872 564 __be32 flags;
26c4f7da
TH		 565 struct gro_remcsum grc;
566
567 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 568
569 off_vx = skb_gro_offset(skb);
570 hlen = off_vx + sizeof(*vh);
571 vh = skb_gro_header_fast(skb, off_vx);
572 if (skb_gro_header_hard(skb, hlen)) {
573 vh = skb_gro_header_slow(skb, hlen, off_vx);
574 if (unlikely(!vh))
575 goto out;
576 }
dc01e7d3 577
dfd8645e
TH		 578 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
579
54bfd872 580 flags = vh->vx_flags;
dfd8645e
TH		 581
582 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
583 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 584 vh->vx_vni, &grc,
0ace2ca8
TH		 585 !!(vs->flags &
586 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 587
588 if (!vh)
589 goto out;
590 }
591
b7fe10e5
TH		 592 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
593
dc01e7d3
OG		 594 flush = 0;
595
596 for (p = *head; p; p = p->next) {
597 if (!NAPI_GRO_CB(p)->same_flow)
598 continue;
599
600 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 601 if (vh->vx_flags != vh2->vx_flags ||
602 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 603 NAPI_GRO_CB(p)->same_flow = 0;
604 continue;
605 }
dc01e7d3
OG		 606 }
607
9b174d88 608 pp = eth_gro_receive(head, skb);
dc01e7d3 609
dc01e7d3 610out:
26c4f7da 611 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 612 NAPI_GRO_CB(skb)->flush |= flush;
613
614 return pp;
615}
616
a2b12f3c
TH		 617static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
618 struct udp_offload *uoff)
dc01e7d3 619{
cfdf1e1b
JG		 620 udp_tunnel_gro_complete(skb, nhoff);
621
9b174d88 622 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 623}
624
53cf5275 625/* Notify netdevs that UDP port started listening */
dc01e7d3 626static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 627{
628 struct net_device *dev;
dc01e7d3 629 struct sock *sk = vs->sock->sk;
53cf5275 630 struct net *net = sock_net(sk);
705cc62f 631 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 632 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 633 int err;
634
635 if (sa_family == AF_INET) {
787d7ac3 636 err = udp_add_offload(net, &vs->udp_offloads);
dc01e7d3
OG		 637 if (err)
638 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
639 }
53cf5275
JG		 640
641 rcu_read_lock();
642 for_each_netdev_rcu(net, dev) {
643 if (dev->netdev_ops->ndo_add_vxlan_port)
644 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
645 port);
646 }
647 rcu_read_unlock();
648}
649
650/* Notify netdevs that UDP port is no more listening */
dc01e7d3 651static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 652{
653 struct net_device *dev;
dc01e7d3 654 struct sock *sk = vs->sock->sk;
53cf5275 655 struct net *net = sock_net(sk);
705cc62f 656 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 657 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 658
659 rcu_read_lock();
660 for_each_netdev_rcu(net, dev) {
661 if (dev->netdev_ops->ndo_del_vxlan_port)
662 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
663 port);
664 }
665 rcu_read_unlock();
dc01e7d3
OG		 666
667 if (sa_family == AF_INET)
668 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 669}
670
d342894c 671/* Add new entry to forwarding table -- assumes lock held */
672static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 673 const u8 *mac, union vxlan_addr *ip,
6681712d 674 __u16 state, __u16 flags,
54bfd872 675 __be16 port, __be32 vni, __u32 ifindex,
ae884082 676 __u8 ndm_flags)
d342894c 677{
9e4b93f9 678 struct vxlan_rdst *rd = NULL;
d342894c 679 struct vxlan_fdb *f;
680 int notify = 0;
681
014be2c8 682 f = __vxlan_find_mac(vxlan, mac);
d342894c 683 if (f) {
684 if (flags & NLM_F_EXCL) {
685 netdev_dbg(vxlan->dev,
686 "lost race to create %pM\n", mac);
687 return -EEXIST;
688 }
689 if (f->state != state) {
690 f->state = state;
691 f->updated = jiffies;
692 notify = 1;
693 }
ae884082
DS		 694 if (f->flags != ndm_flags) {
695 f->flags = ndm_flags;
696 f->updated = jiffies;
697 notify = 1;
698 }
906dc186
TR		 699 if ((flags & NLM_F_REPLACE)) {
700 /* Only change unicasts */
701 if (!(is_multicast_ether_addr(f->eth_addr) ||
702 is_zero_ether_addr(f->eth_addr))) {
60840429 703 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 704 ifindex);
906dc186
TR		 705 } else
706 return -EOPNOTSUPP;
707 }
6681712d 708 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 709 (is_multicast_ether_addr(f->eth_addr) ||
710 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 711 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
712 &rd);
6681712d
DS		 713
714 if (rc < 0)
715 return rc;
716 notify |= rc;
717 }
d342894c 718 } else {
719 if (!(flags & NLM_F_CREATE))
720 return -ENOENT;
721
0dfbdf41
TG		 722 if (vxlan->cfg.addrmax &&
723 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 724 return -ENOSPC;
725
906dc186
TR		 726 /* Disallow replace to add a multicast entry */
727 if ((flags & NLM_F_REPLACE) &&
728 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
729 return -EOPNOTSUPP;
730
e4c7ed41 731 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 732 f = kmalloc(sizeof(*f), GFP_ATOMIC);
733 if (!f)
734 return -ENOMEM;
735
736 notify = 1;
d342894c 737 f->state = state;
ae884082 738 f->flags = ndm_flags;
d342894c 739 f->updated = f->used = jiffies;
3e61aa8f 740 INIT_LIST_HEAD(&f->remotes);
d342894c 741 memcpy(f->eth_addr, mac, ETH_ALEN);
742
9e4b93f9 743 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 744
d342894c 745 ++vxlan->addrcnt;
746 hlist_add_head_rcu(&f->hlist,
747 vxlan_fdb_head(vxlan, mac));
748 }
749
9e4b93f9
ND		 750 if (notify) {
751 if (rd == NULL)
752 rd = first_remote_rtnl(f);
753 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
754 }
d342894c 755
756 return 0;
757}
758
6706c82e 759static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 760{
761 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 762 struct vxlan_rdst *rd, *nd;
6681712d 763
0c1d70af
PA		 764 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
765 dst_cache_destroy(&rd->dst_cache);
6681712d 766 kfree(rd);
0c1d70af 767 }
6681712d
DS		 768 kfree(f);
769}
770
d342894c 771static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
772{
773 netdev_dbg(vxlan->dev,
774 "delete %pM\n", f->eth_addr);
775
776 --vxlan->addrcnt;
9e4b93f9 777 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 778
779 hlist_del_rcu(&f->hlist);
6681712d 780 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 781}
782
f0b074be 783static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB		 784 union vxlan_addr *ip, __be16 *port, __be32 *vni,
785 u32 *ifindex)
d342894c 786{
6681712d 787 struct net *net = dev_net(vxlan->dev);
e4c7ed41 788 int err;
d342894c 789
f0b074be 790 if (tb[NDA_DST]) {
e4c7ed41
CW		 791 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
792 if (err)
793 return err;
f0b074be 794 } else {
e4c7ed41
CW		 795 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
796 if (remote->sa.sa_family == AF_INET) {
797 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
798 ip->sa.sa_family = AF_INET;
799#if IS_ENABLED(CONFIG_IPV6)
800 } else {
801 ip->sin6.sin6_addr = in6addr_any;
802 ip->sa.sa_family = AF_INET6;
803#endif
804 }
f0b074be 805 }
d342894c 806
6681712d 807 if (tb[NDA_PORT]) {
73cf3317 808 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 809 return -EINVAL;
f0b074be
MR		 810 *port = nla_get_be16(tb[NDA_PORT]);
811 } else {
0dfbdf41 812 *port = vxlan->cfg.dst_port;
f0b074be 813 }
6681712d
DS		 814
815 if (tb[NDA_VNI]) {
816 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
817 return -EINVAL;
54bfd872 818 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 819 } else {
820 *vni = vxlan->default_dst.remote_vni;
821 }
6681712d
DS		 822
823 if (tb[NDA_IFINDEX]) {
5abb0029 824 struct net_device *tdev;
6681712d
DS		 825
826 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
827 return -EINVAL;
f0b074be 828 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 829 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 830 if (!tdev)
6681712d 831 return -EADDRNOTAVAIL;
f0b074be
MR		 832 } else {
833 *ifindex = 0;
834 }
835
836 return 0;
837}
838
839/* Add static entry (via netlink) */
840static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
841 struct net_device *dev,
f6f6424b 842 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 843{
844 struct vxlan_dev *vxlan = netdev_priv(dev);
845 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 846 union vxlan_addr ip;
f0b074be 847 __be16 port;
54bfd872
JB		 848 __be32 vni;
849 u32 ifindex;
f0b074be
MR		 850 int err;
851
852 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
853 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
854 ndm->ndm_state);
855 return -EINVAL;
856 }
857
858 if (tb[NDA_DST] == NULL)
859 return -EINVAL;
860
861 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
862 if (err)
863 return err;
6681712d 864
5933a7bb
MR		 865 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
866 return -EAFNOSUPPORT;
867
d342894c 868 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 869 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 870 port, vni, ifindex, ndm->ndm_flags);
d342894c 871 spin_unlock_bh(&vxlan->hash_lock);
872
873 return err;
874}
875
876/* Delete entry (via netlink) */
1690be63
VY		 877static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
878 struct net_device *dev,
f6f6424b 879 const unsigned char *addr, u16 vid)
d342894c 880{
881 struct vxlan_dev *vxlan = netdev_priv(dev);
882 struct vxlan_fdb *f;
bc7892ba 883 struct vxlan_rdst *rd = NULL;
e4c7ed41 884 union vxlan_addr ip;
bc7892ba 885 __be16 port;
54bfd872
JB		 886 __be32 vni;
887 u32 ifindex;
bc7892ba
MR		 888 int err;
889
890 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
891 if (err)
892 return err;
893
894 err = -ENOENT;
d342894c 895
896 spin_lock_bh(&vxlan->hash_lock);
897 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 898 if (!f)
899 goto out;
900
e4c7ed41
CW		 901 if (!vxlan_addr_any(&ip)) {
902 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 903 if (!rd)
904 goto out;
905 }
906
907 err = 0;
908
909 /* remove a destination if it's not the only one on the list,
910 * otherwise destroy the fdb entry
911 */
912 if (rd && !list_is_singular(&f->remotes)) {
913 list_del_rcu(&rd->list);
9e4b93f9 914 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 915 kfree_rcu(rd, rcu);
bc7892ba 916 goto out;
d342894c 917 }
bc7892ba
MR		 918
919 vxlan_fdb_destroy(vxlan, f);
920
921out:
d342894c 922 spin_unlock_bh(&vxlan->hash_lock);
923
924 return err;
925}
926
927/* Dump forwarding table */
928static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 929 struct net_device *dev,
930 struct net_device *filter_dev, int idx)
d342894c 931{
932 struct vxlan_dev *vxlan = netdev_priv(dev);
933 unsigned int h;
934
935 for (h = 0; h < FDB_HASH_SIZE; ++h) {
936 struct vxlan_fdb *f;
d342894c 937 int err;
938
b67bfe0d 939 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 940 struct vxlan_rdst *rd;
6681712d 941
3e61aa8f 942 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 943 if (idx < cb->args[0])
944 goto skip;
945
6681712d
DS		 946 err = vxlan_fdb_info(skb, vxlan, f,
947 NETLINK_CB(cb->skb).portid,
948 cb->nlh->nlmsg_seq,
949 RTM_NEWNEIGH,
950 NLM_F_MULTI, rd);
472681d5
MM		 951 if (err < 0) {
952 cb->args[1] = err;
3e61aa8f 953 goto out;
472681d5 954 }
3e61aa8f 955skip:
07a51cd3
AW		 956 ++idx;
957 }
d342894c 958 }
959 }
3e61aa8f 960out:
d342894c 961 return idx;
962}
963
964/* Watch incoming packets to learn mapping between Ethernet address
965 * and Tunnel endpoint.
c4b49512 966 * Return true if packet is bogus and should be dropped.
d342894c 967 */
26a41ae6 968static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 969 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 970{
971 struct vxlan_dev *vxlan = netdev_priv(dev);
972 struct vxlan_fdb *f;
d342894c 973
974 f = vxlan_find_mac(vxlan, src_mac);
975 if (likely(f)) {
5ca5461c 976 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 977
e4c7ed41 978 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 979 return false;
980
981 /* Don't migrate static entries, drop packets */
eb064c3b 982 if (f->state & NUD_NOARP)
26a41ae6 983 return true;
d342894c 984
985 if (net_ratelimit())
986 netdev_info(dev,
e4c7ed41 987 "%pM migrated from %pIS to %pIS\n",
a4870f79 988 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 989
e4c7ed41 990 rdst->remote_ip = *src_ip;
d342894c 991 f->updated = jiffies;
9e4b93f9 992 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 993 } else {
994 /* learned new entry */
995 spin_lock(&vxlan->hash_lock);
3bf74b1a 996
997 /* close off race between vxlan_flush and incoming packets */
998 if (netif_running(dev))
999 vxlan_fdb_create(vxlan, src_mac, src_ip,
1000 NUD_REACHABLE,
1001 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1002 vxlan->cfg.dst_port,
3bf74b1a 1003 vxlan->default_dst.remote_vni,
1004 0, NTF_SELF);
d342894c 1005 spin_unlock(&vxlan->hash_lock);
1006 }
26a41ae6 1007
1008 return false;
d342894c 1009}
1010
d342894c 1011/* See if multicast group is already in use by other ID */
95ab0991 1012static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1013{
553675fb 1014 struct vxlan_dev *vxlan;
b1be00a6 1015 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1016
95ab0991
G		 1017 /* The vxlan_sock is only used by dev, leaving group has
1018 * no effect on other vxlan devices.
1019 */
b1be00a6
JB		 1020 if (family == AF_INET && dev->vn4_sock &&
1021 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1022 return false;
b1be00a6
JB		 1023#if IS_ENABLED(CONFIG_IPV6)
1024 if (family == AF_INET6 && dev->vn6_sock &&
1025 atomic_read(&dev->vn6_sock->refcnt) == 1)
1026 return false;
1027#endif
95ab0991 1028
553675fb 1029 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1030 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1031 continue;
d342894c 1032
b1be00a6 1033 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1034 continue;
b1be00a6
JB		 1035#if IS_ENABLED(CONFIG_IPV6)
1036 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1037 continue;
1038#endif
95ab0991
G		 1039
1040 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1041 &dev->default_dst.remote_ip))
1042 continue;
1043
1044 if (vxlan->default_dst.remote_ifindex !=
1045 dev->default_dst.remote_ifindex)
1046 continue;
1047
1048 return true;
553675fb 1049 }
d342894c 1050
1051 return false;
1052}
1053
b1be00a6 1054static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1055{
b1be00a6 1056 struct vxlan_net *vn;
012a5729 1057
b1be00a6
JB		 1058 if (!vs)
1059 return;
7c47cedf
SH		 1060 if (!atomic_dec_and_test(&vs->refcnt))
1061 return;
d342894c 1062
b1be00a6 1063 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1064 spin_lock(&vn->sock_lock);
7c47cedf 1065 hlist_del_rcu(&vs->hlist);
dc01e7d3 1066 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1067 spin_unlock(&vn->sock_lock);
1068
7c47cedf 1069 queue_work(vxlan_wq, &vs->del_work);
d342894c 1070}
1071
b1be00a6
JB		 1072static void vxlan_sock_release(struct vxlan_dev *vxlan)
1073{
1074 __vxlan_sock_release(vxlan->vn4_sock);
1075#if IS_ENABLED(CONFIG_IPV6)
1076 __vxlan_sock_release(vxlan->vn6_sock);
1077#endif
1078}
1079
56ef9c90 1080/* Update multicast group membership when first VNI on
c4b49512 1081 * multicast address is brought up
7c47cedf 1082 */
56ef9c90 1083static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1084{
b1be00a6 1085 struct sock *sk;
e4c7ed41
CW		 1086 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1087 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1088 int ret = -EINVAL;
d342894c 1089
e4c7ed41
CW		 1090 if (ip->sa.sa_family == AF_INET) {
1091 struct ip_mreqn mreq = {
1092 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1093 .imr_ifindex = ifindex,
1094 };
1095
b1be00a6
JB		 1096 sk = vxlan->vn4_sock->sock->sk;
1097 lock_sock(sk);
56ef9c90 1098 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1099 release_sock(sk);
e4c7ed41
CW		 1100#if IS_ENABLED(CONFIG_IPV6)
1101 } else {
b1be00a6
JB		 1102 sk = vxlan->vn6_sock->sock->sk;
1103 lock_sock(sk);
56ef9c90
MRL		 1104 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1105 &ip->sin6.sin6_addr);
b1be00a6 1106 release_sock(sk);
e4c7ed41
CW		 1107#endif
1108 }
3fc2de2f 1109
56ef9c90 1110 return ret;
3fc2de2f 1111}
1112
1113/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1114static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1115{
b1be00a6 1116 struct sock *sk;
e4c7ed41
CW		 1117 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1118 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1119 int ret = -EINVAL;
3fc2de2f 1120
e4c7ed41
CW		 1121 if (ip->sa.sa_family == AF_INET) {
1122 struct ip_mreqn mreq = {
1123 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1124 .imr_ifindex = ifindex,
1125 };
1126
b1be00a6
JB		 1127 sk = vxlan->vn4_sock->sock->sk;
1128 lock_sock(sk);
56ef9c90 1129 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1130 release_sock(sk);
e4c7ed41
CW		 1131#if IS_ENABLED(CONFIG_IPV6)
1132 } else {
b1be00a6
JB		 1133 sk = vxlan->vn6_sock->sock->sk;
1134 lock_sock(sk);
56ef9c90
MRL		 1135 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1136 &ip->sin6.sin6_addr);
b1be00a6 1137 release_sock(sk);
e4c7ed41
CW		 1138#endif
1139 }
d342894c 1140
56ef9c90 1141 return ret;
d342894c 1142}
1143
f14ecebb
JB		 1144static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1145 struct sk_buff *skb, u32 vxflags)
dfd8645e
TH		 1146{
1147 size_t start, offset, plen;
dfd8645e 1148
f14ecebb
JB		 1149 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1150 goto out;
b7fe10e5 1151
f14ecebb
JB		 1152 start = vxlan_rco_start(unparsed->vx_vni);
1153 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1154
be5cfeab 1155 plen = sizeof(struct vxlanhdr) + offset + sizeof(u16);
dfd8645e
TH		 1156
1157 if (!pskb_may_pull(skb, plen))
be5cfeab 1158 return false;
dfd8645e 1159
be5cfeab
JB		 1160 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1161 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1162out:
1163 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1164 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1165 return true;
dfd8645e
TH		 1166}
1167
f14ecebb 1168static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1169 struct sk_buff *skb, u32 vxflags,
10a5af23 1170 struct vxlan_metadata *md)
3288af08 1171{
f14ecebb 1172 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1173 struct metadata_dst *tun_dst;
f14ecebb
JB		 1174
1175 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1176 goto out;
3288af08 1177
3288af08
JB		 1178 md->gbp = ntohs(gbp->policy_id);
1179
10a5af23 1180 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1181 if (tun_dst) {
3288af08 1182 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1183 tun_dst->u.tun_info.options_len = sizeof(*md);
1184 }
3288af08
JB		 1185 if (gbp->dont_learn)
1186 md->gbp |= VXLAN_GBP_DONT_LEARN;
1187
1188 if (gbp->policy_applied)
1189 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1190
64f87d36
JB		 1191 /* In flow-based mode, GBP is carried in dst_metadata */
1192 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1193 skb->mark = md->gbp;
f14ecebb
JB		 1194out:
1195 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1196}
1197
1ab016e2
JB		 1198static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1199 struct vxlan_sock *vs,
1200 struct sk_buff *skb)
614732ea 1201{
614732ea 1202 union vxlan_addr saddr;
614732ea 1203
614732ea 1204 skb_reset_mac_header(skb);
614732ea
TG		 1205 skb->protocol = eth_type_trans(skb, vxlan->dev);
1206 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1207
1208 /* Ignore packet loops (and multicast echo) */
1209 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1210 return false;
614732ea 1211
760c6805 1212 /* Get address from the outer IP header */
ce212d0f 1213 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1214 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1215 saddr.sa.sa_family = AF_INET;
1216#if IS_ENABLED(CONFIG_IPV6)
1217 } else {
1ab016e2 1218 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1219 saddr.sa.sa_family = AF_INET6;
1220#endif
1221 }
1222
1ab016e2
JB		 1223 if ((vxlan->flags & VXLAN_F_LEARN) &&
1224 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1225 return false;
1226
1227 return true;
1228}
1229
760c6805
JB		 1230static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1231 struct sk_buff *skb)
1232{
1233 int err = 0;
1234
1235 if (vxlan_get_sk_family(vs) == AF_INET)
1236 err = IP_ECN_decapsulate(oiph, skb);
1237#if IS_ENABLED(CONFIG_IPV6)
1238 else
1239 err = IP6_ECN_decapsulate(oiph, skb);
1240#endif
1241
1242 if (unlikely(err) && log_ecn_error) {
1243 if (vxlan_get_sk_family(vs) == AF_INET)
1244 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1245 &((struct iphdr *)oiph)->saddr,
1246 ((struct iphdr *)oiph)->tos);
1247 else
1248 net_info_ratelimited("non-ECT from %pI6\n",
1249 &((struct ipv6hdr *)oiph)->saddr);
1250 }
1251 return err <= 1;
1252}
1253
d342894c 1254/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1255static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1256{
f2d1968e 1257 struct pcpu_sw_netstats *stats;
c9e78efb 1258 struct vxlan_dev *vxlan;
5cfccc5a 1259 struct vxlan_sock *vs;
f14ecebb 1260 struct vxlanhdr unparsed;
ee122c79
TG		 1261 struct vxlan_metadata _md;
1262 struct vxlan_metadata *md = &_md;
f2d1968e 1263 void *oiph;
d342894c 1264
d342894c 1265 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1266 if (!pskb_may_pull(skb, VXLAN_HLEN))
288b01c8 1267 return 1;
d342894c 1268
f14ecebb 1269 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1270 /* VNI flag always required to be set */
1271 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1272 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1273 ntohl(vxlan_hdr(skb)->vx_flags),
1274 ntohl(vxlan_hdr(skb)->vx_vni));
1275 /* Return non vxlan pkt */
1276 return 1;
d342894c 1277 }
288b01c8
JB		 1278 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1279 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1280
559835ea 1281 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1282 if (!vs)
d342894c 1283 goto drop;
d342894c 1284
c9e78efb
JB		 1285 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1286 if (!vxlan)
1287 goto drop;
1288
7f290c94
JB		 1289 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB),
1290 !net_eq(vxlan->net, dev_net(vxlan->dev))))
c9e78efb
JB		 1291 goto drop;
1292
ee122c79 1293 if (vxlan_collect_metadata(vs)) {
07dabf20 1294 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
10a5af23 1295 struct metadata_dst *tun_dst;
07dabf20 1296
c29a70d2 1297 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
07dabf20 1298 vxlan_vni_to_tun_id(vni), sizeof(*md));
c29a70d2 1299
ee122c79
TG		 1300 if (!tun_dst)
1301 goto drop;
1302
0f1b7354 1303 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1304
1305 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1306 } else {
1307 memset(md, 0, sizeof(*md));
1308 }
1309
3511494c
TG		 1310 /* For backwards compatibility, only allow reserved fields to be
1311 * used by VXLAN extensions if explicitly requested.
1312 */
f14ecebb
JB		 1313 if (vs->flags & VXLAN_F_REMCSUM_RX)
1314 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1315 goto drop;
1316 if (vs->flags & VXLAN_F_GBP)
10a5af23 1317 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
3511494c 1318
f14ecebb 1319 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1320 /* If there are any unprocessed flags remaining treat
1321 * this as a malformed packet. This behavior diverges from
1322 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1323 * in reserved fields are to be ignored. The approach here
c4b49512 1324 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1325 * is more robust and provides a little more security in
1326 * adding extensions to VXLAN.
1327 */
288b01c8 1328 goto drop;
3bf39475
TH		 1329 }
1330
f2d1968e
JB		 1331 if (!vxlan_set_mac(vxlan, vs, skb))
1332 goto drop;
1333
f2d1968e
JB		 1334 oiph = skb_network_header(skb);
1335 skb_reset_network_header(skb);
1336
1337 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1338 ++vxlan->dev->stats.rx_frame_errors;
1339 ++vxlan->dev->stats.rx_errors;
1340 goto drop;
1341 }
1342
1343 stats = this_cpu_ptr(vxlan->dev->tstats);
1344 u64_stats_update_begin(&stats->syncp);
1345 stats->rx_packets++;
1346 stats->rx_bytes += skb->len;
1347 u64_stats_update_end(&stats->syncp);
1348
1349 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1350 return 0;
1351
1352drop:
288b01c8
JB		 1353 /* Consume bad packet */
1354 kfree_skb(skb);
1355 return 0;
5cfccc5a
PS		 1356}
1357
e4f67add
DS		 1358static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1359{
1360 struct vxlan_dev *vxlan = netdev_priv(dev);
1361 struct arphdr *parp;
1362 u8 *arpptr, *sha;
1363 __be32 sip, tip;
1364 struct neighbour *n;
1365
1366 if (dev->flags & IFF_NOARP)
1367 goto out;
1368
1369 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1370 dev->stats.tx_dropped++;
1371 goto out;
1372 }
1373 parp = arp_hdr(skb);
1374
1375 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1376 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1377 parp->ar_pro != htons(ETH_P_IP) ||
1378 parp->ar_op != htons(ARPOP_REQUEST) ||
1379 parp->ar_hln != dev->addr_len ||
1380 parp->ar_pln != 4)
1381 goto out;
1382 arpptr = (u8 *)parp + sizeof(struct arphdr);
1383 sha = arpptr;
1384 arpptr += dev->addr_len; /* sha */
1385 memcpy(&sip, arpptr, sizeof(sip));
1386 arpptr += sizeof(sip);
1387 arpptr += dev->addr_len; /* tha */
1388 memcpy(&tip, arpptr, sizeof(tip));
1389
1390 if (ipv4_is_loopback(tip) ||
1391 ipv4_is_multicast(tip))
1392 goto out;
1393
1394 n = neigh_lookup(&arp_tbl, &tip, dev);
1395
1396 if (n) {
e4f67add
DS		 1397 struct vxlan_fdb *f;
1398 struct sk_buff *reply;
1399
1400 if (!(n->nud_state & NUD_CONNECTED)) {
1401 neigh_release(n);
1402 goto out;
1403 }
1404
1405 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1406 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1407 /* bridge-local neighbor */
1408 neigh_release(n);
1409 goto out;
1410 }
1411
1412 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1413 n->ha, sha);
1414
1415 neigh_release(n);
1416
7346135d
DS		 1417 if (reply == NULL)
1418 goto out;
1419
e4f67add
DS		 1420 skb_reset_mac_header(reply);
1421 __skb_pull(reply, skb_network_offset(reply));
1422 reply->ip_summed = CHECKSUM_UNNECESSARY;
1423 reply->pkt_type = PACKET_HOST;
1424
1425 if (netif_rx_ni(reply) == NET_RX_DROP)
1426 dev->stats.rx_dropped++;
e4c7ed41
CW		 1427 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1428 union vxlan_addr ipa = {
1429 .sin.sin_addr.s_addr = tip,
a45e92a5 1430 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1431 };
1432
1433 vxlan_ip_miss(dev, &ipa);
1434 }
e4f67add
DS		 1435out:
1436 consume_skb(skb);
1437 return NETDEV_TX_OK;
1438}
1439
f564f45c 1440#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1441static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1442 struct neighbour *n, bool isrouter)
1443{
1444 struct net_device *dev = request->dev;
1445 struct sk_buff *reply;
1446 struct nd_msg *ns, *na;
1447 struct ipv6hdr *pip6;
1448 u8 *daddr;
1449 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1450 int ns_olen;
1451 int i, len;
1452
1453 if (dev == NULL)
1454 return NULL;
1455
1456 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1457 sizeof(*na) + na_olen + dev->needed_tailroom;
1458 reply = alloc_skb(len, GFP_ATOMIC);
1459 if (reply == NULL)
1460 return NULL;
1461
1462 reply->protocol = htons(ETH_P_IPV6);
1463 reply->dev = dev;
1464 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1465 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1466 skb_reset_mac_header(reply);
4b29dba9
DS		 1467
1468 ns = (struct nd_msg *)skb_transport_header(request);
1469
1470 daddr = eth_hdr(request)->h_source;
1471 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1472 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1473 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1474 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1475 break;
1476 }
1477 }
1478
1479 /* Ethernet header */
1480 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1481 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1482 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1483 reply->protocol = htons(ETH_P_IPV6);
1484
1485 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1486 skb_reset_network_header(reply);
4b29dba9
DS		 1487 skb_put(reply, sizeof(struct ipv6hdr));
1488
1489 /* IPv6 header */
1490
1491 pip6 = ipv6_hdr(reply);
1492 memset(pip6, 0, sizeof(struct ipv6hdr));
1493 pip6->version = 6;
1494 pip6->priority = ipv6_hdr(request)->priority;
1495 pip6->nexthdr = IPPROTO_ICMPV6;
1496 pip6->hop_limit = 255;
1497 pip6->daddr = ipv6_hdr(request)->saddr;
1498 pip6->saddr = *(struct in6_addr *)n->primary_key;
1499
1500 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1501 skb_reset_transport_header(reply);
4b29dba9
DS		 1502
1503 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1504
1505 /* Neighbor Advertisement */
1506 memset(na, 0, sizeof(*na)+na_olen);
1507 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1508 na->icmph.icmp6_router = isrouter;
1509 na->icmph.icmp6_override = 1;
1510 na->icmph.icmp6_solicited = 1;
1511 na->target = ns->target;
1512 ether_addr_copy(&na->opt[2], n->ha);
1513 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1514 na->opt[1] = na_olen >> 3;
1515
1516 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1517 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1518 csum_partial(na, sizeof(*na)+na_olen, 0));
1519
1520 pip6->payload_len = htons(sizeof(*na)+na_olen);
1521
1522 skb_push(reply, sizeof(struct ipv6hdr));
1523
1524 reply->ip_summed = CHECKSUM_UNNECESSARY;
1525
1526 return reply;
1527}
1528
f564f45c
CW		 1529static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1530{
1531 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1532 struct nd_msg *msg;
f564f45c
CW		 1533 const struct ipv6hdr *iphdr;
1534 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1535 struct neighbour *n;
1536 struct inet6_dev *in6_dev;
f564f45c
CW		 1537
1538 in6_dev = __in6_dev_get(dev);
1539 if (!in6_dev)
1540 goto out;
1541
f564f45c
CW		 1542 iphdr = ipv6_hdr(skb);
1543 saddr = &iphdr->saddr;
1544 daddr = &iphdr->daddr;
1545
f564f45c
CW		 1546 msg = (struct nd_msg *)skb_transport_header(skb);
1547 if (msg->icmph.icmp6_code != 0 ||
1548 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1549 goto out;
1550
4b29dba9
DS		 1551 if (ipv6_addr_loopback(daddr) ||
1552 ipv6_addr_is_multicast(&msg->target))
1553 goto out;
1554
1555 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1556
1557 if (n) {
1558 struct vxlan_fdb *f;
4b29dba9 1559 struct sk_buff *reply;
f564f45c
CW		 1560
1561 if (!(n->nud_state & NUD_CONNECTED)) {
1562 neigh_release(n);
1563 goto out;
1564 }
1565
1566 f = vxlan_find_mac(vxlan, n->ha);
1567 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1568 /* bridge-local neighbor */
1569 neigh_release(n);
1570 goto out;
1571 }
1572
4b29dba9
DS		 1573 reply = vxlan_na_create(skb, n,
1574 !!(f ? f->flags & NTF_ROUTER : 0));
1575
f564f45c 1576 neigh_release(n);
4b29dba9
DS		 1577
1578 if (reply == NULL)
1579 goto out;
1580
1581 if (netif_rx_ni(reply) == NET_RX_DROP)
1582 dev->stats.rx_dropped++;
1583
f564f45c 1584 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1585 union vxlan_addr ipa = {
1586 .sin6.sin6_addr = msg->target,
a45e92a5 1587 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1588 };
1589
f564f45c
CW		 1590 vxlan_ip_miss(dev, &ipa);
1591 }
1592
1593out:
1594 consume_skb(skb);
1595 return NETDEV_TX_OK;
1596}
1597#endif
1598
e4f67add
DS		 1599static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1600{
1601 struct vxlan_dev *vxlan = netdev_priv(dev);
1602 struct neighbour *n;
e4f67add
DS		 1603
1604 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1605 return false;
1606
1607 n = NULL;
1608 switch (ntohs(eth_hdr(skb)->h_proto)) {
1609 case ETH_P_IP:
e15a00aa
CW		 1610 {
1611 struct iphdr *pip;
1612
e4f67add
DS		 1613 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1614 return false;
1615 pip = ip_hdr(skb);
1616 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1617 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1618 union vxlan_addr ipa = {
1619 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1620 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1621 };
1622
1623 vxlan_ip_miss(dev, &ipa);
1624 return false;
1625 }
1626
e4f67add 1627 break;
e15a00aa
CW		 1628 }
1629#if IS_ENABLED(CONFIG_IPV6)
1630 case ETH_P_IPV6:
1631 {
1632 struct ipv6hdr *pip6;
1633
1634 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1635 return false;
1636 pip6 = ipv6_hdr(skb);
1637 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1638 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1639 union vxlan_addr ipa = {
1640 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1641 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1642 };
1643
1644 vxlan_ip_miss(dev, &ipa);
1645 return false;
1646 }
1647
1648 break;
1649 }
1650#endif
e4f67add
DS		 1651 default:
1652 return false;
1653 }
1654
1655 if (n) {
1656 bool diff;
1657
7367d0b5 1658 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1659 if (diff) {
1660 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1661 dev->addr_len);
1662 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1663 }
1664 neigh_release(n);
1665 return diff;
e4c7ed41
CW		 1666 }
1667
e4f67add
DS		 1668 return false;
1669}
1670
af33c1ad 1671static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1672 struct vxlan_metadata *md)
1673{
1674 struct vxlanhdr_gbp *gbp;
1675
db79a621
TG		 1676 if (!md->gbp)
1677 return;
1678
3511494c 1679 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1680 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1681
1682 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1683 gbp->dont_learn = 1;
1684
1685 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1686 gbp->policy_applied = 1;
1687
1688 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1689}
1690
f491e56d
JB		 1691static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1692 int iphdr_len, __be32 vni,
1693 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1694 bool udp_sum)
e4c7ed41 1695{
e4c7ed41 1696 struct vxlanhdr *vxh;
e4c7ed41
CW		 1697 int min_headroom;
1698 int err;
dfd8645e 1699 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
dfd8645e 1700
af33c1ad 1701 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1702 skb->ip_summed == CHECKSUM_PARTIAL) {
1703 int csum_start = skb_checksum_start_offset(skb);
1704
1705 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1706 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1707 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1708 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1709 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1710 }
e4c7ed41 1711
e4c7ed41 1712 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1713 + VXLAN_HLEN + iphdr_len
df8a39de 1714 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1715
1716 /* Need space for new headers (invalidates iph ptr) */
1717 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1718 if (unlikely(err)) {
1719 kfree_skb(skb);
649c5b8b 1720 return err;
74f47278 1721 }
649c5b8b 1722
5968250c
JP		 1723 skb = vlan_hwaccel_push_inside(skb);
1724 if (WARN_ON(!skb))
1725 return -ENOMEM;
1eaa8178 1726
6fa79666 1727 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG		 1728 if (IS_ERR(skb))
1729 return PTR_ERR(skb);
1730
49560532 1731 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1732 vxh->vx_flags = VXLAN_HF_VNI;
1733 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1734
dfd8645e 1735 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1736 unsigned int start;
dfd8645e 1737
54bfd872
JB		 1738 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1739 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1740 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1741
1742 if (!skb_is_gso(skb)) {
1743 skb->ip_summed = CHECKSUM_NONE;
1744 skb->encapsulation = 0;
1745 }
1746 }
1747
af33c1ad
TH		 1748 if (vxflags & VXLAN_F_GBP)
1749 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1750
996c9fd1 1751 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
039f5062 1752 return 0;
49560532 1753}
49560532 1754
1a8496ba
JB		 1755static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1756 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1757 __be32 daddr, __be32 *saddr,
1758 struct dst_cache *dst_cache,
db3c6139 1759 const struct ip_tunnel_info *info)
1a8496ba 1760{
db3c6139 1761 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1762 struct rtable *rt = NULL;
1763 struct flowi4 fl4;
1764
db3c6139
DB		 1765 if (tos && !info)
1766 use_cache = false;
1767 if (use_cache) {
0c1d70af
PA		 1768 rt = dst_cache_get_ip4(dst_cache, saddr);
1769 if (rt)
1770 return rt;
1771 }
1772
1a8496ba
JB		 1773 memset(&fl4, 0, sizeof(fl4));
1774 fl4.flowi4_oif = oif;
1775 fl4.flowi4_tos = RT_TOS(tos);
1776 fl4.flowi4_mark = skb->mark;
1777 fl4.flowi4_proto = IPPROTO_UDP;
1778 fl4.daddr = daddr;
1779 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1780
1781 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1782 if (!IS_ERR(rt)) {
1a8496ba 1783 *saddr = fl4.saddr;
0c1d70af
PA		 1784 if (use_cache)
1785 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1786 }
1a8496ba
JB		 1787 return rt;
1788}
1789
e5d4b29f
JB		 1790#if IS_ENABLED(CONFIG_IPV6)
1791static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1400615d 1792 struct sk_buff *skb, int oif, u8 tos,
e5d4b29f 1793 const struct in6_addr *daddr,
0c1d70af 1794 struct in6_addr *saddr,
db3c6139
DB		 1795 struct dst_cache *dst_cache,
1796 const struct ip_tunnel_info *info)
e5d4b29f 1797{
db3c6139 1798 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1799 struct dst_entry *ndst;
1800 struct flowi6 fl6;
1801 int err;
1802
1400615d
DB		 1803 if (tos && !info)
1804 use_cache = false;
db3c6139 1805 if (use_cache) {
0c1d70af
PA		 1806 ndst = dst_cache_get_ip6(dst_cache, saddr);
1807 if (ndst)
1808 return ndst;
1809 }
1810
e5d4b29f
JB		 1811 memset(&fl6, 0, sizeof(fl6));
1812 fl6.flowi6_oif = oif;
1400615d 1813 fl6.flowi6_tos = RT_TOS(tos);
e5d4b29f
JB		 1814 fl6.daddr = *daddr;
1815 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1816 fl6.flowi6_mark = skb->mark;
1817 fl6.flowi6_proto = IPPROTO_UDP;
1818
1819 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1820 vxlan->vn6_sock->sock->sk,
1821 &ndst, &fl6);
1822 if (err < 0)
1823 return ERR_PTR(err);
1824
1825 *saddr = fl6.saddr;
db3c6139 1826 if (use_cache)
0c1d70af 1827 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1828 return ndst;
1829}
1830#endif
1831
9dcc71e1
SS		 1832/* Bypass encapsulation if the destination is local */
1833static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1834 struct vxlan_dev *dst_vxlan)
1835{
8f84985f 1836 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1837 union vxlan_addr loopback;
1838 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1839 struct net_device *dev = skb->dev;
1840 int len = skb->len;
9dcc71e1 1841
8f84985f
LR		 1842 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1843 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1844 skb->pkt_type = PACKET_HOST;
1845 skb->encapsulation = 0;
1846 skb->dev = dst_vxlan->dev;
1847 __skb_pull(skb, skb_network_offset(skb));
1848
e4c7ed41
CW		 1849 if (remote_ip->sa.sa_family == AF_INET) {
1850 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1851 loopback.sa.sa_family = AF_INET;
1852#if IS_ENABLED(CONFIG_IPV6)
1853 } else {
1854 loopback.sin6.sin6_addr = in6addr_loopback;
1855 loopback.sa.sa_family = AF_INET6;
1856#endif
1857 }
1858
9dcc71e1 1859 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1860 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1861
1862 u64_stats_update_begin(&tx_stats->syncp);
1863 tx_stats->tx_packets++;
ce6502a8 1864 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1865 u64_stats_update_end(&tx_stats->syncp);
1866
1867 if (netif_rx(skb) == NET_RX_SUCCESS) {
1868 u64_stats_update_begin(&rx_stats->syncp);
1869 rx_stats->rx_packets++;
ce6502a8 1870 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1871 u64_stats_update_end(&rx_stats->syncp);
1872 } else {
ce6502a8 1873 dev->stats.rx_dropped++;
9dcc71e1
SS		 1874 }
1875}
1876
4ad16930
SH		 1877static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1878 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1879{
d71785ff 1880 struct dst_cache *dst_cache;
3093fbe7 1881 struct ip_tunnel_info *info;
d342894c 1882 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1883 struct sock *sk;
e4c7ed41 1884 struct rtable *rt = NULL;
d342894c 1885 const struct iphdr *old_iph;
e4c7ed41 1886 union vxlan_addr *dst;
ee122c79
TG		 1887 union vxlan_addr remote_ip;
1888 struct vxlan_metadata _md;
1889 struct vxlan_metadata *md = &_md;
e4c7ed41 1890 __be16 src_port = 0, dst_port;
54bfd872 1891 __be32 vni;
d342894c 1892 __be16 df = 0;
1893 __u8 tos, ttl;
0e6fbc5b 1894 int err;
ee122c79 1895 u32 flags = vxlan->flags;
b4ed5cad 1896 bool udp_sum = false;
f491e56d 1897 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1898
61adedf3 1899 info = skb_tunnel_info(skb);
3093fbe7 1900
ee122c79 1901 if (rdst) {
0dfbdf41 1902 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1903 vni = rdst->remote_vni;
1904 dst = &rdst->remote_ip;
d71785ff 1905 dst_cache = &rdst->dst_cache;
ee122c79
TG		 1906 } else {
1907 if (!info) {
1908 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1909 dev->name);
1910 goto drop;
1911 }
0dfbdf41 1912 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
54bfd872 1913 vni = vxlan_tun_id_to_vni(info->key.tun_id);
b1be00a6
JB		 1914 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1915 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1916 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1917 else
1918 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1919 dst = &remote_ip;
d71785ff 1920 dst_cache = &info->dst_cache;
ee122c79 1921 }
e4f67add 1922
e4c7ed41 1923 if (vxlan_addr_any(dst)) {
e4f67add 1924 if (did_rsc) {
e4f67add 1925 /* short-circuited back to local bridge */
9dcc71e1 1926 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1927 return;
e4f67add 1928 }
ef59febe 1929 goto drop;
e4f67add 1930 }
ef59febe 1931
d342894c 1932 old_iph = ip_hdr(skb);
1933
0dfbdf41 1934 ttl = vxlan->cfg.ttl;
e4c7ed41 1935 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1936 ttl = 1;
1937
0dfbdf41 1938 tos = vxlan->cfg.tos;
d342894c 1939 if (tos == 1)
206aaafc 1940 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1941
0dfbdf41
TG		 1942 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1943 vxlan->cfg.port_max, true);
d342894c 1944
a725e514 1945 if (info) {
a725e514
JB		 1946 ttl = info->key.ttl;
1947 tos = info->key.tos;
b4ed5cad 1948 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB		 1949
1950 if (info->options_len)
4c222798 1951 md = ip_tunnel_info_opts(info);
a725e514
JB		 1952 } else {
1953 md->gbp = skb->mark;
1954 }
1955
e4c7ed41 1956 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB		 1957 __be32 saddr;
1958
b1be00a6
JB		 1959 if (!vxlan->vn4_sock)
1960 goto drop;
1961 sk = vxlan->vn4_sock->sock->sk;
1962
1a8496ba
JB		 1963 rt = vxlan_get_route(vxlan, skb,
1964 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 1965 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 1966 dst_cache, info);
e4c7ed41
CW		 1967 if (IS_ERR(rt)) {
1968 netdev_dbg(dev, "no route to %pI4\n",
1969 &dst->sin.sin_addr.s_addr);
1970 dev->stats.tx_carrier_errors++;
1971 goto tx_error;
1972 }
d342894c 1973
e4c7ed41
CW		 1974 if (rt->dst.dev == dev) {
1975 netdev_dbg(dev, "circular route to %pI4\n",
1976 &dst->sin.sin_addr.s_addr);
1977 dev->stats.collisions++;
fffc15a5 1978 goto rt_tx_error;
e4c7ed41
CW		 1979 }
1980
1981 /* Bypass encapsulation if the destination is local */
1982 if (rt->rt_flags & RTCF_LOCAL &&
1983 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1984 struct vxlan_dev *dst_vxlan;
1985
1986 ip_rt_put(rt);
19ca9fc1 1987 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1988 dst->sa.sa_family, dst_port,
1989 vxlan->flags);
e4c7ed41
CW		 1990 if (!dst_vxlan)
1991 goto tx_error;
1992 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1993 return;
1994 }
1995
6ceb31ca
AD		 1996 if (!info)
1997 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
1998 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1999 df = htons(IP_DF);
2000
e4c7ed41
CW		 2001 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2002 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d 2003 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
54bfd872 2004 vni, md, flags, udp_sum);
f491e56d
JB		 2005 if (err < 0)
2006 goto xmit_tx_error;
2007
2008 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2009 dst->sin.sin_addr.s_addr, tos, ttl, df,
2010 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2011#if IS_ENABLED(CONFIG_IPV6)
2012 } else {
e4c7ed41 2013 struct dst_entry *ndst;
e5d4b29f 2014 struct in6_addr saddr;
6f264e47 2015 u32 rt6i_flags;
e4c7ed41 2016
b1be00a6
JB		 2017 if (!vxlan->vn6_sock)
2018 goto drop;
2019 sk = vxlan->vn6_sock->sock->sk;
2020
e5d4b29f 2021 ndst = vxlan6_get_route(vxlan, skb,
1400615d 2022 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 2023 &dst->sin6.sin6_addr, &saddr,
db3c6139 2024 dst_cache, info);
e5d4b29f 2025 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2026 netdev_dbg(dev, "no route to %pI6\n",
2027 &dst->sin6.sin6_addr);
2028 dev->stats.tx_carrier_errors++;
9dcc71e1 2029 goto tx_error;
e4c7ed41 2030 }
d342894c 2031
e4c7ed41
CW		 2032 if (ndst->dev == dev) {
2033 netdev_dbg(dev, "circular route to %pI6\n",
2034 &dst->sin6.sin6_addr);
2035 dst_release(ndst);
2036 dev->stats.collisions++;
2037 goto tx_error;
2038 }
0e6fbc5b 2039
e4c7ed41 2040 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2041 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2042 if (rt6i_flags & RTF_LOCAL &&
2043 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2044 struct vxlan_dev *dst_vxlan;
2045
2046 dst_release(ndst);
19ca9fc1 2047 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2048 dst->sa.sa_family, dst_port,
2049 vxlan->flags);
e4c7ed41
CW		 2050 if (!dst_vxlan)
2051 goto tx_error;
2052 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2053 return;
2054 }
49560532 2055
b4ed5cad
JB		 2056 if (!info)
2057 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2058
1400615d 2059 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2060 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2061 skb_scrub_packet(skb, xnet);
2062 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2063 vni, md, flags, udp_sum);
f491e56d
JB		 2064 if (err < 0) {
2065 dst_release(ndst);
2066 return;
2067 }
2068 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
13461144 2069 &saddr, &dst->sin6.sin6_addr, tos, ttl, 0,
1400615d 2070 src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2071#endif
2072 }
0e6fbc5b 2073
4ad16930 2074 return;
d342894c 2075
2076drop:
2077 dev->stats.tx_dropped++;
2078 goto tx_free;
2079
f491e56d
JB		 2080xmit_tx_error:
2081 /* skb is already freed. */
2082 skb = NULL;
49560532
PS		 2083rt_tx_error:
2084 ip_rt_put(rt);
d342894c 2085tx_error:
2086 dev->stats.tx_errors++;
2087tx_free:
2088 dev_kfree_skb(skb);
d342894c 2089}
2090
6681712d
DS		 2091/* Transmit local packets over Vxlan
2092 *
2093 * Outer IP header inherits ECN and DF from inner header.
2094 * Outer UDP destination is the VXLAN assigned port.
2095 * source port is based on hash of flow
2096 */
2097static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2098{
2099 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2100 const struct ip_tunnel_info *info;
6681712d
DS		 2101 struct ethhdr *eth;
2102 bool did_rsc = false;
8f646c92 2103 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2104 struct vxlan_fdb *f;
6681712d 2105
61adedf3 2106 info = skb_tunnel_info(skb);
3093fbe7 2107
6681712d
DS		 2108 skb_reset_mac_header(skb);
2109 eth = eth_hdr(skb);
2110
f564f45c
CW		 2111 if ((vxlan->flags & VXLAN_F_PROXY)) {
2112 if (ntohs(eth->h_proto) == ETH_P_ARP)
2113 return arp_reduce(dev, skb);
2114#if IS_ENABLED(CONFIG_IPV6)
2115 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2116 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2117 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2118 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2119 struct nd_msg *msg;
2120
2121 msg = (struct nd_msg *)skb_transport_header(skb);
2122 if (msg->icmph.icmp6_code == 0 &&
2123 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2124 return neigh_reduce(dev, skb);
2125 }
7a9f526f 2126 eth = eth_hdr(skb);
f564f45c
CW		 2127#endif
2128 }
6681712d 2129
f468a729
JB		 2130 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2131 if (info && info->mode & IP_TUNNEL_INFO_TX)
2132 vxlan_xmit_one(skb, dev, NULL, false);
2133 else
2134 kfree_skb(skb);
ee122c79
TG		 2135 return NETDEV_TX_OK;
2136 }
2137
6681712d 2138 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2139 did_rsc = false;
2140
2141 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2142 (ntohs(eth->h_proto) == ETH_P_IP ||
2143 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2144 did_rsc = route_shortcircuit(dev, skb);
2145 if (did_rsc)
2146 f = vxlan_find_mac(vxlan, eth->h_dest);
2147 }
2148
6681712d 2149 if (f == NULL) {
afbd8bae
MR		 2150 f = vxlan_find_mac(vxlan, all_zeros_mac);
2151 if (f == NULL) {
2152 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2153 !is_multicast_ether_addr(eth->h_dest))
2154 vxlan_fdb_miss(vxlan, eth->h_dest);
2155
2156 dev->stats.tx_dropped++;
8f646c92 2157 kfree_skb(skb);
afbd8bae
MR		 2158 return NETDEV_TX_OK;
2159 }
2160 }
6681712d 2161
afbd8bae
MR		 2162 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2163 struct sk_buff *skb1;
6681712d 2164
8f646c92
ED		 2165 if (!fdst) {
2166 fdst = rdst;
2167 continue;
2168 }
afbd8bae
MR		 2169 skb1 = skb_clone(skb, GFP_ATOMIC);
2170 if (skb1)
2171 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2172 }
2173
8f646c92
ED		 2174 if (fdst)
2175 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2176 else
2177 kfree_skb(skb);
4ad16930 2178 return NETDEV_TX_OK;
6681712d
DS		 2179}
2180
d342894c 2181/* Walk the forwarding table and purge stale entries */
2182static void vxlan_cleanup(unsigned long arg)
2183{
2184 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2185 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2186 unsigned int h;
2187
2188 if (!netif_running(vxlan->dev))
2189 return;
2190
d342894c 2191 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2192 struct hlist_node *p, *n;
14e1d0fa
SD		 2193
2194 spin_lock_bh(&vxlan->hash_lock);
d342894c 2195 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2196 struct vxlan_fdb *f
2197 = container_of(p, struct vxlan_fdb, hlist);
2198 unsigned long timeout;
2199
3c172868 2200 if (f->state & NUD_PERMANENT)
d342894c 2201 continue;
2202
0dfbdf41 2203 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2204 if (time_before_eq(timeout, jiffies)) {
2205 netdev_dbg(vxlan->dev,
2206 "garbage collect %pM\n",
2207 f->eth_addr);
2208 f->state = NUD_STALE;
2209 vxlan_fdb_destroy(vxlan, f);
2210 } else if (time_before(timeout, next_timer))
2211 next_timer = timeout;
2212 }
14e1d0fa 2213 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2214 }
d342894c 2215
2216 mod_timer(&vxlan->age_timer, next_timer);
2217}
2218
9c2e24e1
PS		 2219static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2220{
56ef9c90 2221 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2222 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2223
56ef9c90 2224 spin_lock(&vn->sock_lock);
9c2e24e1 2225 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2226 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2227}
2228
d342894c 2229/* Setup stats when device is created */
2230static int vxlan_init(struct net_device *dev)
2231{
1c213bd2 2232 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2233 if (!dev->tstats)
d342894c 2234 return -ENOMEM;
2235
2236 return 0;
2237}
2238
ba609e9b 2239static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2240{
2241 struct vxlan_fdb *f;
2242
2243 spin_lock_bh(&vxlan->hash_lock);
2244 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2245 if (f)
2246 vxlan_fdb_destroy(vxlan, f);
2247 spin_unlock_bh(&vxlan->hash_lock);
2248}
2249
ebf4063e
SH		 2250static void vxlan_uninit(struct net_device *dev)
2251{
2252 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2253
ba609e9b 2254 vxlan_fdb_delete_default(vxlan);
afbd8bae 2255
ebf4063e
SH		 2256 free_percpu(dev->tstats);
2257}
2258
d342894c 2259/* Start ageing timer and join group when device is brought up */
2260static int vxlan_open(struct net_device *dev)
2261{
2262 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2263 int ret;
56ef9c90 2264
205f356d
JB		 2265 ret = vxlan_sock_add(vxlan);
2266 if (ret < 0)
2267 return ret;
d342894c 2268
79d4a94f 2269 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2270 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2271 if (ret == -EADDRINUSE)
2272 ret = 0;
56ef9c90 2273 if (ret) {
205f356d 2274 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2275 return ret;
2276 }
d342894c 2277 }
2278
0dfbdf41 2279 if (vxlan->cfg.age_interval)
d342894c 2280 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2281
56ef9c90 2282 return ret;
d342894c 2283}
2284
2285/* Purge the forwarding table */
2286static void vxlan_flush(struct vxlan_dev *vxlan)
2287{
31fec5aa 2288 unsigned int h;
d342894c 2289
2290 spin_lock_bh(&vxlan->hash_lock);
2291 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2292 struct hlist_node *p, *n;
2293 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2294 struct vxlan_fdb *f
2295 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2296 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2297 if (!is_zero_ether_addr(f->eth_addr))
2298 vxlan_fdb_destroy(vxlan, f);
d342894c 2299 }
2300 }
2301 spin_unlock_bh(&vxlan->hash_lock);
2302}
2303
2304/* Cleanup timer and forwarding table on shutdown */
2305static int vxlan_stop(struct net_device *dev)
2306{
2307 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2308 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2309 int ret = 0;
d342894c 2310
24c0e683 2311 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2312 !vxlan_group_used(vn, vxlan))
56ef9c90 2313 ret = vxlan_igmp_leave(vxlan);
d342894c 2314
2315 del_timer_sync(&vxlan->age_timer);
2316
2317 vxlan_flush(vxlan);
205f356d 2318 vxlan_sock_release(vxlan);
d342894c 2319
56ef9c90 2320 return ret;
d342894c 2321}
2322
d342894c 2323/* Stub, nothing needs to be done. */
2324static void vxlan_set_multicast_list(struct net_device *dev)
2325{
2326}
2327
72564b59
DW		 2328static int __vxlan_change_mtu(struct net_device *dev,
2329 struct net_device *lowerdev,
2330 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2331{
72564b59 2332 int max_mtu = IP_MAX_MTU;
345010b5 2333
72564b59
DW		 2334 if (lowerdev)
2335 max_mtu = lowerdev->mtu;
345010b5
DB		 2336
2337 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2338 max_mtu -= VXLAN6_HEADROOM;
345010b5 2339 else
72564b59 2340 max_mtu -= VXLAN_HEADROOM;
345010b5 2341
72564b59 2342 if (new_mtu < 68)
345010b5
DB		 2343 return -EINVAL;
2344
72564b59
DW		 2345 if (new_mtu > max_mtu) {
2346 if (strict)
2347 return -EINVAL;
2348
2349 new_mtu = max_mtu;
2350 }
2351
345010b5
DB		 2352 dev->mtu = new_mtu;
2353 return 0;
2354}
2355
72564b59
DW		 2356static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2357{
2358 struct vxlan_dev *vxlan = netdev_priv(dev);
2359 struct vxlan_rdst *dst = &vxlan->default_dst;
2360 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2361 dst->remote_ifindex);
2362 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2363}
2364
fc4099f1
PS		 2365static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2366{
2367 struct vxlan_dev *vxlan = netdev_priv(dev);
2368 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2369 __be16 sport, dport;
2370
2371 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2372 vxlan->cfg.port_max, true);
2373 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2374
239e944f 2375 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB		 2376 struct rtable *rt;
2377
239e944f
JB		 2378 if (!vxlan->vn4_sock)
2379 return -EINVAL;
1a8496ba
JB		 2380 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2381 info->key.u.ipv4.dst,
0c1d70af 2382 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2383 if (IS_ERR(rt))
2384 return PTR_ERR(rt);
2385 ip_rt_put(rt);
239e944f
JB		 2386 } else {
2387#if IS_ENABLED(CONFIG_IPV6)
2388 struct dst_entry *ndst;
2389
2390 if (!vxlan->vn6_sock)
2391 return -EINVAL;
1400615d 2392 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
239e944f 2393 &info->key.u.ipv6.dst,
db3c6139 2394 &info->key.u.ipv6.src, NULL, info);
239e944f
JB		 2395 if (IS_ERR(ndst))
2396 return PTR_ERR(ndst);
2397 dst_release(ndst);
239e944f
JB		 2398#else /* !CONFIG_IPV6 */
2399 return -EPFNOSUPPORT;
2400#endif
2401 }
1a8496ba
JB		 2402 info->key.tp_src = sport;
2403 info->key.tp_dst = dport;
239e944f 2404 return 0;
fc4099f1
PS		 2405}
2406
d342894c 2407static const struct net_device_ops vxlan_netdev_ops = {
2408 .ndo_init = vxlan_init,
ebf4063e 2409 .ndo_uninit = vxlan_uninit,
d342894c 2410 .ndo_open = vxlan_open,
2411 .ndo_stop = vxlan_stop,
2412 .ndo_start_xmit = vxlan_xmit,
e8171045 2413 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2414 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2415 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2416 .ndo_validate_addr = eth_validate_addr,
2417 .ndo_set_mac_address = eth_mac_addr,
2418 .ndo_fdb_add = vxlan_fdb_add,
2419 .ndo_fdb_del = vxlan_fdb_delete,
2420 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2421 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2422};
2423
2424/* Info for udev, that this is a virtual tunnel endpoint */
2425static struct device_type vxlan_type = {
2426 .name = "vxlan",
2427};
2428
53cf5275 2429/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2430 * supply the listening VXLAN udp ports. Callers are expected
2431 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2432 */
2433void vxlan_get_rx_port(struct net_device *dev)
2434{
2435 struct vxlan_sock *vs;
2436 struct net *net = dev_net(dev);
2437 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2438 sa_family_t sa_family;
35e42379
JG		 2439 __be16 port;
2440 unsigned int i;
53cf5275
JG		 2441
2442 spin_lock(&vn->sock_lock);
2443 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2444 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2445 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2446 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2447 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2448 port);
2449 }
2450 }
2451 spin_unlock(&vn->sock_lock);
2452}
2453EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2454
d342894c 2455/* Initialize the device structure. */
2456static void vxlan_setup(struct net_device *dev)
2457{
2458 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2459 unsigned int h;
d342894c 2460
2461 eth_hw_addr_random(dev);
2462 ether_setup(dev);
2463
2464 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2465 dev->destructor = free_netdev;
d342894c 2466 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2467
d342894c 2468 dev->features |= NETIF_F_LLTX;
d6727fe3 2469 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2470 dev->features |= NETIF_F_RXCSUM;
05c0db08 2471 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2472
1eaa8178
PS		 2473 dev->vlan_features = dev->features;
2474 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2475 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2476 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2477 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2478 netif_keep_dst(dev);
82a0f6b4 2479 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
22dba393 2480 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2481
553675fb 2482 INIT_LIST_HEAD(&vxlan->next);
d342894c 2483 spin_lock_init(&vxlan->hash_lock);
2484
2485 init_timer_deferrable(&vxlan->age_timer);
2486 vxlan->age_timer.function = vxlan_cleanup;
2487 vxlan->age_timer.data = (unsigned long) vxlan;
2488
0dfbdf41 2489 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2490
d342894c 2491 vxlan->dev = dev;
2492
58ce31cc
TH		 2493 gro_cells_init(&vxlan->gro_cells, dev);
2494
d342894c 2495 for (h = 0; h < FDB_HASH_SIZE; ++h)
2496 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2497}
2498
2499static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2500 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2501 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2502 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2503 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2504 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2505 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2506 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2507 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2508 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2509 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2510 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2511 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2512 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2513 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2514 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2515 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2516 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2517 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2518 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2519 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2520 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2521 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2522 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2523 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2524 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2525};
2526
2527static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2528{
2529 if (tb[IFLA_ADDRESS]) {
2530 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2531 pr_debug("invalid link address (not ethernet)\n");
2532 return -EINVAL;
2533 }
2534
2535 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2536 pr_debug("invalid all zero ethernet address\n");
2537 return -EADDRNOTAVAIL;
2538 }
2539 }
2540
2541 if (!data)
2542 return -EINVAL;
2543
2544 if (data[IFLA_VXLAN_ID]) {
2545 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2546 if (id >= VXLAN_VID_MASK)
2547 return -ERANGE;
2548 }
2549
05f47d69 2550 if (data[IFLA_VXLAN_PORT_RANGE]) {
2551 const struct ifla_vxlan_port_range *p
2552 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2553
2554 if (ntohs(p->high) < ntohs(p->low)) {
2555 pr_debug("port range %u .. %u not valid\n",
2556 ntohs(p->low), ntohs(p->high));
2557 return -EINVAL;
2558 }
2559 }
2560
d342894c 2561 return 0;
2562}
2563
1b13c97f
YB		 2564static void vxlan_get_drvinfo(struct net_device *netdev,
2565 struct ethtool_drvinfo *drvinfo)
2566{
2567 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2568 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2569}
2570
2571static const struct ethtool_ops vxlan_ethtool_ops = {
2572 .get_drvinfo = vxlan_get_drvinfo,
2573 .get_link = ethtool_op_get_link,
2574};
2575
553675fb 2576static void vxlan_del_work(struct work_struct *work)
2577{
2578 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2579 udp_tunnel_sock_release(vs->sock);
553675fb 2580 kfree_rcu(vs, rcu);
2581}
2582
3ee64f39
TH		 2583static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2584 __be16 port, u32 flags)
553675fb 2585{
e4c7ed41 2586 struct socket *sock;
3ee64f39
TH		 2587 struct udp_port_cfg udp_conf;
2588 int err;
e4c7ed41 2589
3ee64f39 2590 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2591
3ee64f39
TH		 2592 if (ipv6) {
2593 udp_conf.family = AF_INET6;
3ee64f39 2594 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2595 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2596 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2597 } else {
2598 udp_conf.family = AF_INET;
e4c7ed41 2599 }
e4c7ed41 2600
3ee64f39 2601 udp_conf.local_udp_port = port;
553675fb 2602
3ee64f39
TH		 2603 /* Open UDP socket */
2604 err = udp_sock_create(net, &udp_conf, &sock);
2605 if (err < 0)
2606 return ERR_PTR(err);
e4c7ed41 2607
39deb2c7 2608 return sock;
e4c7ed41
CW		 2609}
2610
2611/* Create new listen socket if needed */
b1be00a6
JB		 2612static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2613 __be16 port, u32 flags)
e4c7ed41
CW		 2614{
2615 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2616 struct vxlan_sock *vs;
2617 struct socket *sock;
e4c7ed41 2618 unsigned int h;
acbf74a7 2619 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2620
dc01e7d3 2621 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2622 if (!vs)
2623 return ERR_PTR(-ENOMEM);
2624
2625 for (h = 0; h < VNI_HASH_SIZE; ++h)
2626 INIT_HLIST_HEAD(&vs->vni_list[h]);
2627
2628 INIT_WORK(&vs->del_work, vxlan_del_work);
2629
3ee64f39 2630 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2631 if (IS_ERR(sock)) {
56ef9c90
MRL		 2632 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2633 PTR_ERR(sock));
553675fb 2634 kfree(vs);
e50fddc8 2635 return ERR_CAST(sock);
553675fb 2636 }
e4c7ed41
CW		 2637
2638 vs->sock = sock;
9c2e24e1 2639 atomic_set(&vs->refcnt, 1);
af33c1ad 2640 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2641
dc01e7d3
OG		 2642 /* Initialize the vxlan udp offloads structure */
2643 vs->udp_offloads.port = port;
2644 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2645 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2646
9c2e24e1
PS		 2647 spin_lock(&vn->sock_lock);
2648 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2649 vxlan_notify_add_rx_port(vs);
9c2e24e1 2650 spin_unlock(&vn->sock_lock);
553675fb 2651
2652 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2653 tunnel_cfg.sk_user_data = vs;
2654 tunnel_cfg.encap_type = 1;
f2d1968e 2655 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7
AZ		 2656 tunnel_cfg.encap_destroy = NULL;
2657
2658 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2659
9c2e24e1
PS		 2660 return vs;
2661}
2662
b1be00a6 2663static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2664{
205f356d
JB		 2665 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2666 struct vxlan_sock *vs = NULL;
9c2e24e1 2667
205f356d 2668 if (!vxlan->cfg.no_share) {
56ef9c90 2669 spin_lock(&vn->sock_lock);
205f356d
JB		 2670 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2671 vxlan->cfg.dst_port, vxlan->flags);
2672 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2673 spin_unlock(&vn->sock_lock);
205f356d 2674 return -EBUSY;
56ef9c90
MRL		 2675 }
2676 spin_unlock(&vn->sock_lock);
2677 }
205f356d 2678 if (!vs)
b1be00a6
JB		 2679 vs = vxlan_socket_create(vxlan->net, ipv6,
2680 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2681 if (IS_ERR(vs))
2682 return PTR_ERR(vs);
b1be00a6
JB		 2683#if IS_ENABLED(CONFIG_IPV6)
2684 if (ipv6)
2685 vxlan->vn6_sock = vs;
2686 else
2687#endif
2688 vxlan->vn4_sock = vs;
205f356d
JB		 2689 vxlan_vs_add_dev(vs, vxlan);
2690 return 0;
553675fb 2691}
2692
b1be00a6
JB		 2693static int vxlan_sock_add(struct vxlan_dev *vxlan)
2694{
2695 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2696 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2697 int ret = 0;
2698
2699 vxlan->vn4_sock = NULL;
2700#if IS_ENABLED(CONFIG_IPV6)
2701 vxlan->vn6_sock = NULL;
2702 if (ipv6 || metadata)
2703 ret = __vxlan_sock_add(vxlan, true);
2704#endif
2705 if (!ret && (!ipv6 || metadata))
2706 ret = __vxlan_sock_add(vxlan, false);
2707 if (ret < 0)
2708 vxlan_sock_release(vxlan);
2709 return ret;
2710}
2711
0dfbdf41
TG		 2712static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2713 struct vxlan_config *conf)
d342894c 2714{
33564bbb 2715 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2716 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2717 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2718 unsigned short needed_headroom = ETH_HLEN;
d342894c 2719 int err;
e4c7ed41 2720 bool use_ipv6 = false;
0dfbdf41 2721 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2722 struct net_device *lowerdev = NULL;
d342894c 2723
33564bbb 2724 vxlan->net = src_net;
f01ec1c0 2725
0dfbdf41 2726 dst->remote_vni = conf->vni;
d342894c 2727
0dfbdf41 2728 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2729
0dfbdf41
TG		 2730 /* Unless IPv6 is explicitly requested, assume IPv4 */
2731 if (!dst->remote_ip.sa.sa_family)
2732 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2733
0dfbdf41 2734 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2735 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2736 if (!IS_ENABLED(CONFIG_IPV6))
2737 return -EPFNOSUPPORT;
e4c7ed41 2738 use_ipv6 = true;
b1be00a6 2739 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2740 }
d342894c 2741
0dfbdf41 2742 if (conf->remote_ifindex) {
7e059158 2743 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2744 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2745
2746 if (!lowerdev) {
c7995c43 2747 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2748 return -ENODEV;
2749 }
d342894c 2750
e4c7ed41
CW		 2751#if IS_ENABLED(CONFIG_IPV6)
2752 if (use_ipv6) {
2753 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2754 if (idev && idev->cnf.disable_ipv6) {
2755 pr_info("IPv6 is disabled via sysctl\n");
2756 return -EPERM;
2757 }
e4c7ed41
CW		 2758 }
2759#endif
2760
0dfbdf41 2761 if (!conf->mtu)
e4c7ed41 2762 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2763
b1be00a6 2764 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2765 }
d342894c 2766
7e059158
DW		 2767 if (conf->mtu) {
2768 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2769 if (err)
2770 return err;
2771 }
2772
b1be00a6
JB		 2773 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2774 needed_headroom += VXLAN6_HEADROOM;
2775 else
2776 needed_headroom += VXLAN_HEADROOM;
2777 dev->needed_headroom = needed_headroom;
2778
0dfbdf41
TG		 2779 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2780 if (!vxlan->cfg.dst_port)
2781 vxlan->cfg.dst_port = default_port;
2782 vxlan->flags |= conf->flags;
2783
2784 if (!vxlan->cfg.age_interval)
2785 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2786
07b9b37c
ND		 2787 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2788 if (tmp->cfg.vni == conf->vni &&
2789 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2790 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2791 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2792 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2793 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2794 return -EEXIST;
07b9b37c 2795 }
0dfbdf41
TG		 2796
2797 dev->ethtool_ops = &vxlan_ethtool_ops;
2798
2799 /* create an fdb entry for a valid default destination */
2800 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2801 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2802 &vxlan->default_dst.remote_ip,
2803 NUD_REACHABLE|NUD_PERMANENT,
2804 NLM_F_EXCL|NLM_F_CREATE,
2805 vxlan->cfg.dst_port,
2806 vxlan->default_dst.remote_vni,
2807 vxlan->default_dst.remote_ifindex,
2808 NTF_SELF);
2809 if (err)
2810 return err;
2811 }
2812
2813 err = register_netdevice(dev);
2814 if (err) {
2815 vxlan_fdb_delete_default(vxlan);
2816 return err;
2817 }
2818
2819 list_add(&vxlan->next, &vn->vxlan_list);
2820
2821 return 0;
2822}
2823
2824struct net_device *vxlan_dev_create(struct net *net, const char *name,
2825 u8 name_assign_type, struct vxlan_config *conf)
2826{
2827 struct nlattr *tb[IFLA_MAX+1];
2828 struct net_device *dev;
2829 int err;
2830
2831 memset(&tb, 0, sizeof(tb));
2832
2833 dev = rtnl_create_link(net, name, name_assign_type,
2834 &vxlan_link_ops, tb);
2835 if (IS_ERR(dev))
2836 return dev;
2837
2838 err = vxlan_dev_configure(net, dev, conf);
2839 if (err < 0) {
2840 free_netdev(dev);
2841 return ERR_PTR(err);
2842 }
2843
2844 return dev;
2845}
2846EXPORT_SYMBOL_GPL(vxlan_dev_create);
2847
2848static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2849 struct nlattr *tb[], struct nlattr *data[])
2850{
2851 struct vxlan_config conf;
2852 int err;
2853
0dfbdf41 2854 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2855
2856 if (data[IFLA_VXLAN_ID])
54bfd872 2857 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG		 2858
2859 if (data[IFLA_VXLAN_GROUP]) {
2860 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2861 } else if (data[IFLA_VXLAN_GROUP6]) {
2862 if (!IS_ENABLED(CONFIG_IPV6))
2863 return -EPFNOSUPPORT;
2864
2865 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2866 conf.remote_ip.sa.sa_family = AF_INET6;
2867 }
2868
2869 if (data[IFLA_VXLAN_LOCAL]) {
2870 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2871 conf.saddr.sa.sa_family = AF_INET;
2872 } else if (data[IFLA_VXLAN_LOCAL6]) {
2873 if (!IS_ENABLED(CONFIG_IPV6))
2874 return -EPFNOSUPPORT;
2875
2876 /* TODO: respect scope id */
2877 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2878 conf.saddr.sa.sa_family = AF_INET6;
2879 }
2880
2881 if (data[IFLA_VXLAN_LINK])
2882 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2883
d342894c 2884 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2885 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2886
afb97186 2887 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2888 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2889
d342894c 2890 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2891 conf.flags |= VXLAN_F_LEARN;
d342894c 2892
2893 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2894 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2895
e4f67add 2896 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2897 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2898
2899 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2900 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2901
2902 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2903 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2904
2905 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2906 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2907
d342894c 2908 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2909 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2910
f8a9b1bc
AS		 2911 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2912 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2913 conf.flags |= VXLAN_F_COLLECT_METADATA;
2914
05f47d69 2915 if (data[IFLA_VXLAN_PORT_RANGE]) {
2916 const struct ifla_vxlan_port_range *p
2917 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2918 conf.port_min = ntohs(p->low);
2919 conf.port_max = ntohs(p->high);
05f47d69 2920 }
2921
823aa873 2922 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2923 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2924
6ceb31ca
AD		 2925 if (data[IFLA_VXLAN_UDP_CSUM] &&
2926 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2927 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH		 2928
2929 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2930 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2931 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2932
2933 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2934 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2935 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2936
dfd8645e
TH		 2937 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2938 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2939 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 2940
2941 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2942 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2943 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2944
3511494c 2945 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2946 conf.flags |= VXLAN_F_GBP;
3511494c 2947
0ace2ca8 2948 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2949 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2950
0dfbdf41
TG		 2951 err = vxlan_dev_configure(src_net, dev, &conf);
2952 switch (err) {
2953 case -ENODEV:
2954 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2955 break;
1b13c97f 2956
0dfbdf41
TG		 2957 case -EPERM:
2958 pr_info("IPv6 is disabled via sysctl\n");
2959 break;
d342894c 2960
0dfbdf41 2961 case -EEXIST:
54bfd872 2962 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
0dfbdf41 2963 break;
afbd8bae
MR		 2964 }
2965
0dfbdf41 2966 return err;
d342894c 2967}
2968
2969static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2970{
2971 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2972 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2973
fe5c3561 2974 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2975 if (!hlist_unhashed(&vxlan->hlist))
2976 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2977 spin_unlock(&vn->sock_lock);
2978
58ce31cc 2979 gro_cells_destroy(&vxlan->gro_cells);
553675fb 2980 list_del(&vxlan->next);
d342894c 2981 unregister_netdevice_queue(dev, head);
2982}
2983
2984static size_t vxlan_get_size(const struct net_device *dev)
2985{
2986
2987 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2988 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2989 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2990 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2991 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2992 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2993 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2994 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2995 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2996 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2997 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 2998 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 2999 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3000 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3001 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3002 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3003 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3005 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3006 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3007 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3008 0;
3009}
3010
3011static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3012{
3013 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3014 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3015 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3016 .low = htons(vxlan->cfg.port_min),
3017 .high = htons(vxlan->cfg.port_max),
05f47d69 3018 };
d342894c 3019
54bfd872 3020 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3021 goto nla_put_failure;
3022
e4c7ed41
CW		 3023 if (!vxlan_addr_any(&dst->remote_ip)) {
3024 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3025 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3026 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3027 goto nla_put_failure;
3028#if IS_ENABLED(CONFIG_IPV6)
3029 } else {
930345ea
JB		 3030 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3031 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3032 goto nla_put_failure;
3033#endif
3034 }
3035 }
d342894c 3036
c7995c43 3037 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3038 goto nla_put_failure;
3039
0dfbdf41
TG		 3040 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3041 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3042 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3043 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3044 goto nla_put_failure;
3045#if IS_ENABLED(CONFIG_IPV6)
3046 } else {
930345ea 3047 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3048 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3049 goto nla_put_failure;
3050#endif
3051 }
3052 }
d342894c 3053
0dfbdf41
TG		 3054 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3055 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 3056 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3057 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3058 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3059 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3060 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3061 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3062 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3063 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3064 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3065 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3066 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3067 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3068 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3069 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3070 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3071 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3072 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3073 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3074 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3075 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3076 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3077 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3078 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3079 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3080 goto nla_put_failure;
3081
05f47d69 3082 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3083 goto nla_put_failure;
3084
3511494c
TG		 3085 if (vxlan->flags & VXLAN_F_GBP &&
3086 nla_put_flag(skb, IFLA_VXLAN_GBP))
3087 goto nla_put_failure;
3088
0ace2ca8
TH		 3089 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3090 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3091 goto nla_put_failure;
3092
d342894c 3093 return 0;
3094
3095nla_put_failure:
3096 return -EMSGSIZE;
3097}
3098
1728d4fa
ND		 3099static struct net *vxlan_get_link_net(const struct net_device *dev)
3100{
3101 struct vxlan_dev *vxlan = netdev_priv(dev);
3102
3103 return vxlan->net;
3104}
3105
d342894c 3106static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3107 .kind = "vxlan",
3108 .maxtype = IFLA_VXLAN_MAX,
3109 .policy = vxlan_policy,
3110 .priv_size = sizeof(struct vxlan_dev),
3111 .setup = vxlan_setup,
3112 .validate = vxlan_validate,
3113 .newlink = vxlan_newlink,
3114 .dellink = vxlan_dellink,
3115 .get_size = vxlan_get_size,
3116 .fill_info = vxlan_fill_info,
1728d4fa 3117 .get_link_net = vxlan_get_link_net,
d342894c 3118};
3119
acaf4e70
DB		 3120static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3121 struct net_device *dev)
3122{
3123 struct vxlan_dev *vxlan, *next;
3124 LIST_HEAD(list_kill);
3125
3126 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3127 struct vxlan_rdst *dst = &vxlan->default_dst;
3128
3129 /* In case we created vxlan device with carrier
3130 * and we loose the carrier due to module unload
3131 * we also need to remove vxlan device. In other
3132 * cases, it's not necessary and remote_ifindex
3133 * is 0 here, so no matches.
3134 */
3135 if (dst->remote_ifindex == dev->ifindex)
3136 vxlan_dellink(vxlan->dev, &list_kill);
3137 }
3138
3139 unregister_netdevice_many(&list_kill);
3140}
3141
3142static int vxlan_lowerdev_event(struct notifier_block *unused,
3143 unsigned long event, void *ptr)
3144{
3145 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3146 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3147
783c1463 3148 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3149 vxlan_handle_lowerdev_unregister(vn, dev);
3150
3151 return NOTIFY_DONE;
3152}
3153
3154static struct notifier_block vxlan_notifier_block __read_mostly = {
3155 .notifier_call = vxlan_lowerdev_event,
3156};
3157
d342894c 3158static __net_init int vxlan_init_net(struct net *net)
3159{
3160 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3161 unsigned int h;
d342894c 3162
553675fb 3163 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3164 spin_lock_init(&vn->sock_lock);
d342894c 3165
553675fb 3166 for (h = 0; h < PORT_HASH_SIZE; ++h)
3167 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3168
3169 return 0;
3170}
3171
f01ec1c0
ND		 3172static void __net_exit vxlan_exit_net(struct net *net)
3173{
3174 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3175 struct vxlan_dev *vxlan, *next;
3176 struct net_device *dev, *aux;
3177 LIST_HEAD(list);
3178
3179 rtnl_lock();
3180 for_each_netdev_safe(net, dev, aux)
3181 if (dev->rtnl_link_ops == &vxlan_link_ops)
3182 unregister_netdevice_queue(dev, &list);
3183
3184 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3185 /* If vxlan->dev is in the same netns, it has already been added
3186 * to the list by the previous loop.
3187 */
58ce31cc
TH		 3188 if (!net_eq(dev_net(vxlan->dev), net)) {
3189 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3190 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3191 }
f01ec1c0
ND		 3192 }
3193
3194 unregister_netdevice_many(&list);
3195 rtnl_unlock();
3196}
3197
d342894c 3198static struct pernet_operations vxlan_net_ops = {
3199 .init = vxlan_init_net,
f01ec1c0 3200 .exit = vxlan_exit_net,
d342894c 3201 .id = &vxlan_net_id,
3202 .size = sizeof(struct vxlan_net),
3203};
3204
3205static int __init vxlan_init_module(void)
3206{
3207 int rc;
3208
758c57d1
SH		 3209 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3210 if (!vxlan_wq)
3211 return -ENOMEM;
3212
d342894c 3213 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3214
783c1463 3215 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3216 if (rc)
3217 goto out1;
3218
acaf4e70 3219 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3220 if (rc)
3221 goto out2;
3222
acaf4e70
DB		 3223 rc = rtnl_link_register(&vxlan_link_ops);
3224 if (rc)
3225 goto out3;
d342894c 3226
acaf4e70
DB		 3227 return 0;
3228out3:
3229 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3230out2:
783c1463 3231 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3232out1:
758c57d1 3233 destroy_workqueue(vxlan_wq);
d342894c 3234 return rc;
3235}
7332a13b 3236late_initcall(vxlan_init_module);
d342894c 3237
3238static void __exit vxlan_cleanup_module(void)
3239{
b7153984 3240 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3241 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3242 destroy_workqueue(vxlan_wq);
783c1463
DB		 3243 unregister_pernet_subsys(&vxlan_net_ops);
3244 /* rcu_barrier() is called by netns */
d342894c 3245}
3246module_exit(vxlan_cleanup_module);
3247
3248MODULE_LICENSE("GPL");
3249MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3250MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3251MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3252MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Hirsute Linux kernel mirror