]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - drivers/net/vxlan.c
projects / thirdparty / kernel / stable.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'smp-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[thirdparty/kernel/stable.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>
d342894c 14#include <linux/module.h>
15#include <linux/errno.h>
16#include <linux/slab.h>
d342894c 17#include <linux/udp.h>
18#include <linux/igmp.h>
d342894c 19#include <linux/if_ether.h>
1b13c97f 20#include <linux/ethtool.h>
e4f67add
DS		 21#include <net/arp.h>
22#include <net/ndisc.h>
d342894c 23#include <net/ip.h>
24#include <net/icmp.h>
d342894c 25#include <net/rtnetlink.h>
d342894c 26#include <net/inet_ecn.h>
27#include <net/net_namespace.h>
28#include <net/netns/generic.h>
fa20e0e3 29#include <net/tun_proto.h>
012a5729 30#include <net/vxlan.h>
40d29af0 31
e4c7ed41 32#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 33#include <net/ip6_tunnel.h>
660d98ca 34#include <net/ip6_checksum.h>
e4c7ed41 35#endif
d342894c 36
37#define VXLAN_VERSION "0.1"
38
553675fb 39#define PORT_HASH_BITS 8
40#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 41#define FDB_AGE_DEFAULT 300 /* 5 min */
42#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43
23c578bf 44/* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 46 * for compatibility with early adopters.
23c578bf 47 */
9daaa397
SH		 48static unsigned short vxlan_port __read_mostly = 8472;
49module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 50MODULE_PARM_DESC(udp_port, "Destination UDP port");
51
52static bool log_ecn_error = true;
53module_param(log_ecn_error, bool, 0644);
54MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55
c7d03a00 56static unsigned int vxlan_net_id;
0dfbdf41 57static struct rtnl_link_ops vxlan_link_ops;
553675fb 58
7256eac1 59static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 60
205f356d 61static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 62
a53cb29b
MB		 63static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64
553675fb 65/* per-network namespace private data for this module */
66struct vxlan_net {
67 struct list_head vxlan_list;
68 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 69 spinlock_t sock_lock;
553675fb 70};
71
d342894c 72/* Forwarding table entry */
73struct vxlan_fdb {
74 struct hlist_node hlist; /* linked list of entries */
75 struct rcu_head rcu;
76 unsigned long updated; /* jiffies */
77 unsigned long used;
3e61aa8f 78 struct list_head remotes;
7177a3b0 79 u8 eth_addr[ETH_ALEN];
d342894c 80 u16 state; /* see ndm_state */
3ad7a4b1 81 __be32 vni;
45598c1c 82 u16 flags; /* see ndm_flags and below */
d342894c 83};
84
45598c1c
PM		 85#define NTF_VXLAN_ADDED_BY_USER 0x100
86
d342894c 87/* salt for hash table */
88static u32 vxlan_salt __read_mostly;
89
ee122c79
TG		 90static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
91{
e7030878
TG		 92 return vs->flags & VXLAN_F_COLLECT_METADATA ||
93 ip_tunnel_collect_metadata();
ee122c79
TG		 94}
95
e4c7ed41
CW		 96#if IS_ENABLED(CONFIG_IPV6)
97static inline
98bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
99{
f0ef3126
JB		 100 if (a->sa.sa_family != b->sa.sa_family)
101 return false;
102 if (a->sa.sa_family == AF_INET6)
103 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
104 else
105 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 106}
107
e4c7ed41
CW		 108static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
109{
f0ef3126 110 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 111 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 112 ip->sa.sa_family = AF_INET6;
113 return 0;
114 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 115 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 116 ip->sa.sa_family = AF_INET;
117 return 0;
118 } else {
119 return -EAFNOSUPPORT;
120 }
e4c7ed41
CW		 121}
122
123static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 124 const union vxlan_addr *ip)
e4c7ed41 125{
f0ef3126 126 if (ip->sa.sa_family == AF_INET6)
930345ea 127 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 128 else
930345ea 129 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 130}
131
132#else /* !CONFIG_IPV6 */
133
134static inline
135bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
136{
f0ef3126 137 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 138}
139
e4c7ed41
CW		 140static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
141{
f0ef3126
JB		 142 if (nla_len(nla) >= sizeof(struct in6_addr)) {
143 return -EAFNOSUPPORT;
144 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 145 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 146 ip->sa.sa_family = AF_INET;
147 return 0;
148 } else {
149 return -EAFNOSUPPORT;
150 }
e4c7ed41
CW		 151}
152
153static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 154 const union vxlan_addr *ip)
e4c7ed41 155{
930345ea 156 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 157}
158#endif
159
553675fb 160/* Virtual Network hash table head */
54bfd872 161static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 162{
54bfd872 163 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 164}
165
166/* Socket hash table head */
167static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 168{
169 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
170
553675fb 171 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
172}
173
3e61aa8f
SH		 174/* First remote destination for a forwarding entry.
175 * Guaranteed to be non-NULL because remotes are never deleted.
176 */
5ca5461c 177static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 178{
5ca5461c 179 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
180}
181
182static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
183{
184 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 185}
186
ac5132d1
TG		 187/* Find VXLAN socket based on network namespace, address family and UDP port
188 * and enabled unshareable flags.
189 */
190static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
aab8cc36 191 __be16 port, u32 flags, int ifindex)
553675fb 192{
193 struct vxlan_sock *vs;
af33c1ad
TH		 194
195 flags &= VXLAN_F_RCV_FLAGS;
553675fb 196
197 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 198 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 199 vxlan_get_sk_family(vs) == family &&
aab8cc36
AB		 200 vs->flags == flags &&
201 vs->sock->sk->sk_bound_dev_if == ifindex)
553675fb 202 return vs;
203 }
204 return NULL;
d342894c 205}
206
49f810f0
MS		 207static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
208 __be32 vni)
d342894c 209{
69e76661 210 struct vxlan_dev_node *node;
d342894c 211
b9167b2e
JB		 212 /* For flow based devices, map all packets to VNI 0 */
213 if (vs->flags & VXLAN_F_COLLECT_METADATA)
214 vni = 0;
215
69e76661
JB		 216 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
217 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS		 218 continue;
219
220 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 221 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS		 222
223 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
224 cfg->remote_ifindex != ifindex)
225 continue;
226 }
227
69e76661 228 return node->vxlan;
d342894c 229 }
230
231 return NULL;
232}
233
5cfccc5a 234/* Look up VNI in a per net namespace table */
49f810f0
MS		 235static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
236 __be32 vni, sa_family_t family,
237 __be16 port, u32 flags)
5cfccc5a
PS		 238{
239 struct vxlan_sock *vs;
240
aab8cc36 241 vs = vxlan_find_sock(net, family, port, flags, ifindex);
5cfccc5a
PS		 242 if (!vs)
243 return NULL;
244
49f810f0 245 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS		 246}
247
d342894c 248/* Fill in neighbour message in skbuff. */
249static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 250 const struct vxlan_fdb *fdb,
251 u32 portid, u32 seq, int type, unsigned int flags,
252 const struct vxlan_rdst *rdst)
d342894c 253{
254 unsigned long now = jiffies;
255 struct nda_cacheinfo ci;
256 struct nlmsghdr *nlh;
257 struct ndmsg *ndm;
e4f67add 258 bool send_ip, send_eth;
d342894c 259
260 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
261 if (nlh == NULL)
262 return -EMSGSIZE;
263
264 ndm = nlmsg_data(nlh);
265 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 266
267 send_eth = send_ip = true;
268
269 if (type == RTM_GETNEIGH) {
e4c7ed41 270 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 271 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 272 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 273 } else
274 ndm->ndm_family = AF_BRIDGE;
d342894c 275 ndm->ndm_state = fdb->state;
276 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 277 ndm->ndm_flags = fdb->flags;
0efe1173
PM		 278 if (rdst->offloaded)
279 ndm->ndm_flags |= NTF_OFFLOADED;
545469f7 280 ndm->ndm_type = RTN_UNICAST;
d342894c 281
193523bf 282 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 283 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 284 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 285 goto nla_put_failure;
286
e4f67add 287 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 288 goto nla_put_failure;
289
e4c7ed41 290 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 291 goto nla_put_failure;
292
0dfbdf41 293 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 294 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
295 goto nla_put_failure;
c7995c43 296 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 297 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 298 goto nla_put_failure;
dc5321d7 299 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 300 nla_put_u32(skb, NDA_SRC_VNI,
301 be32_to_cpu(fdb->vni)))
302 goto nla_put_failure;
6681712d
DS		 303 if (rdst->remote_ifindex &&
304 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 305 goto nla_put_failure;
306
307 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
308 ci.ndm_confirmed = 0;
309 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
310 ci.ndm_refcnt = 0;
311
312 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
313 goto nla_put_failure;
314
053c095a
JB		 315 nlmsg_end(skb, nlh);
316 return 0;
d342894c 317
318nla_put_failure:
319 nlmsg_cancel(skb, nlh);
320 return -EMSGSIZE;
321}
322
323static inline size_t vxlan_nlmsg_size(void)
324{
325 return NLMSG_ALIGN(sizeof(struct ndmsg))
326 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 327 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 328 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 329 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
330 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 331 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 332 + nla_total_size(sizeof(struct nda_cacheinfo));
333}
334
9a997353
PM		 335static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
336 struct vxlan_rdst *rd, int type)
d342894c 337{
338 struct net *net = dev_net(vxlan->dev);
339 struct sk_buff *skb;
340 int err = -ENOBUFS;
341
342 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
343 if (skb == NULL)
344 goto errout;
345
9e4b93f9 346 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 347 if (err < 0) {
348 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
349 WARN_ON(err == -EMSGSIZE);
350 kfree_skb(skb);
351 goto errout;
352 }
353
354 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
355 return;
356errout:
357 if (err < 0)
358 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
359}
360
ff23b91c
PM		 361static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
362 const struct vxlan_fdb *fdb,
363 const struct vxlan_rdst *rd,
4c59b7d1 364 struct netlink_ext_ack *extack,
ff23b91c
PM		 365 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
366{
367 fdb_info->info.dev = vxlan->dev;
4c59b7d1 368 fdb_info->info.extack = extack;
ff23b91c
PM		 369 fdb_info->remote_ip = rd->remote_ip;
370 fdb_info->remote_port = rd->remote_port;
371 fdb_info->remote_vni = rd->remote_vni;
372 fdb_info->remote_ifindex = rd->remote_ifindex;
373 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
374 fdb_info->vni = fdb->vni;
375 fdb_info->offloaded = rd->offloaded;
376 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
377}
378
61f46fe8
PM		 379static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
380 struct vxlan_fdb *fdb,
381 struct vxlan_rdst *rd,
4c59b7d1
PM		 382 bool adding,
383 struct netlink_ext_ack *extack)
9a997353
PM		 384{
385 struct switchdev_notifier_vxlan_fdb_info info;
386 enum switchdev_notifier_type notifier_type;
61f46fe8 387 int ret;
9a997353
PM		 388
389 if (WARN_ON(!rd))
61f46fe8 390 return 0;
9a997353
PM		 391
392 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
393 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
4c59b7d1 394 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
61f46fe8 395 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
6685987c 396 &info.info, extack);
61f46fe8 397 return notifier_to_errno(ret);
9a997353
PM		 398}
399
61f46fe8 400static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
4c59b7d1
PM		 401 struct vxlan_rdst *rd, int type, bool swdev_notify,
402 struct netlink_ext_ack *extack)
9a997353 403{
61f46fe8
PM		 404 int err;
405
0e6160f3
PM		 406 if (swdev_notify) {
407 switch (type) {
408 case RTM_NEWNEIGH:
61f46fe8 409 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 410 true, extack);
61f46fe8
PM		 411 if (err)
412 return err;
0e6160f3
PM		 413 break;
414 case RTM_DELNEIGH:
415 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 416 false, extack);
0e6160f3
PM		 417 break;
418 }
9a997353
PM		 419 }
420
421 __vxlan_fdb_notify(vxlan, fdb, rd, type);
61f46fe8 422 return 0;
9a997353
PM		 423}
424
e4c7ed41 425static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 426{
427 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 428 struct vxlan_fdb f = {
429 .state = NUD_STALE,
430 };
431 struct vxlan_rdst remote = {
e4c7ed41 432 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 433 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 434 };
3e61aa8f 435
4c59b7d1 436 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 437}
438
439static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
440{
bb3fd687
SH		 441 struct vxlan_fdb f = {
442 .state = NUD_STALE,
443 };
9e4b93f9 444 struct vxlan_rdst remote = { };
e4f67add 445
e4f67add
DS		 446 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
447
4c59b7d1 448 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 449}
450
d342894c 451/* Hash Ethernet address */
452static u32 eth_hash(const unsigned char *addr)
453{
454 u64 value = get_unaligned((u64 *)addr);
455
456 /* only want 6 bytes */
457#ifdef __BIG_ENDIAN
d342894c 458 value >>= 16;
321fb991 459#else
460 value <<= 16;
d342894c 461#endif
462 return hash_64(value, FDB_HASH_BITS);
463}
464
3ad7a4b1
RP		 465static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
466{
467 /* use 1 byte of OUI and 3 bytes of NIC */
468 u32 key = get_unaligned((u32 *)(addr + 2));
469
470 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
471}
472
d342894c 473/* Hash chain to use given mac address */
474static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 475 const u8 *mac, __be32 vni)
d342894c 476{
dc5321d7 477 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
3ad7a4b1
RP		 478 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
479 else
480 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 481}
482
483/* Look up Ethernet address in forwarding table */
014be2c8 484static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 485 const u8 *mac, __be32 vni)
d342894c 486{
3ad7a4b1 487 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 488 struct vxlan_fdb *f;
d342894c 489
b67bfe0d 490 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 491 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 492 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 493 if (vni == f->vni)
494 return f;
495 } else {
496 return f;
497 }
498 }
d342894c 499 }
500
501 return NULL;
502}
503
014be2c8 504static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 505 const u8 *mac, __be32 vni)
014be2c8
SS		 506{
507 struct vxlan_fdb *f;
508
3ad7a4b1 509 f = __vxlan_find_mac(vxlan, mac, vni);
016f3d18 510 if (f && f->used != jiffies)
014be2c8
SS		 511 f->used = jiffies;
512
513 return f;
514}
515
a5e7c10a
MR		 516/* caller should hold vxlan->hash_lock */
517static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 518 union vxlan_addr *ip, __be16 port,
54bfd872 519 __be32 vni, __u32 ifindex)
6681712d 520{
3e61aa8f 521 struct vxlan_rdst *rd;
6681712d 522
3e61aa8f 523 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 524 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 525 rd->remote_port == port &&
526 rd->remote_vni == vni &&
527 rd->remote_ifindex == ifindex)
a5e7c10a 528 return rd;
6681712d 529 }
3e61aa8f 530
a5e7c10a
MR		 531 return NULL;
532}
533
1941f1d6
PM		 534int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
535 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
536{
537 struct vxlan_dev *vxlan = netdev_priv(dev);
538 u8 eth_addr[ETH_ALEN + 2] = { 0 };
539 struct vxlan_rdst *rdst;
540 struct vxlan_fdb *f;
541 int rc = 0;
542
543 if (is_multicast_ether_addr(mac) ||
544 is_zero_ether_addr(mac))
545 return -EINVAL;
546
547 ether_addr_copy(eth_addr, mac);
548
549 rcu_read_lock();
550
551 f = __vxlan_find_mac(vxlan, eth_addr, vni);
552 if (!f) {
553 rc = -ENOENT;
554 goto out;
555 }
556
557 rdst = first_remote_rcu(f);
4c59b7d1 558 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
1941f1d6
PM		 559
560out:
561 rcu_read_unlock();
562 return rc;
563}
564EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
565
4f89f5b5
PM		 566static int vxlan_fdb_notify_one(struct notifier_block *nb,
567 const struct vxlan_dev *vxlan,
568 const struct vxlan_fdb *f,
4c59b7d1
PM		 569 const struct vxlan_rdst *rdst,
570 struct netlink_ext_ack *extack)
4f89f5b5
PM		 571{
572 struct switchdev_notifier_vxlan_fdb_info fdb_info;
573 int rc;
574
4c59b7d1 575 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
4f89f5b5
PM		 576 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
577 &fdb_info);
578 return notifier_to_errno(rc);
579}
580
581int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4c59b7d1
PM		 582 struct notifier_block *nb,
583 struct netlink_ext_ack *extack)
4f89f5b5
PM		 584{
585 struct vxlan_dev *vxlan;
586 struct vxlan_rdst *rdst;
587 struct vxlan_fdb *f;
588 unsigned int h;
589 int rc = 0;
590
591 if (!netif_is_vxlan(dev))
592 return -EINVAL;
593 vxlan = netdev_priv(dev);
594
595 spin_lock_bh(&vxlan->hash_lock);
596 for (h = 0; h < FDB_HASH_SIZE; ++h) {
597 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
598 if (f->vni == vni) {
599 list_for_each_entry(rdst, &f->remotes, list) {
600 rc = vxlan_fdb_notify_one(nb, vxlan,
4c59b7d1
PM		 601 f, rdst,
602 extack);
4f89f5b5
PM		 603 if (rc)
604 goto out;
605 }
606 }
607 }
608 }
609
610out:
611 spin_unlock_bh(&vxlan->hash_lock);
612 return rc;
613}
614EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
615
e5ff4b19
PM		 616void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
617{
618 struct vxlan_dev *vxlan;
619 struct vxlan_rdst *rdst;
620 struct vxlan_fdb *f;
621 unsigned int h;
622
623 if (!netif_is_vxlan(dev))
624 return;
625 vxlan = netdev_priv(dev);
626
627 spin_lock_bh(&vxlan->hash_lock);
628 for (h = 0; h < FDB_HASH_SIZE; ++h) {
629 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
630 if (f->vni == vni)
631 list_for_each_entry(rdst, &f->remotes, list)
632 rdst->offloaded = false;
633 }
634 spin_unlock_bh(&vxlan->hash_lock);
635}
636EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
637
906dc186
TR		 638/* Replace destination of unicast mac */
639static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872 640 union vxlan_addr *ip, __be16 port, __be32 vni,
ccdfd4f7 641 __u32 ifindex, struct vxlan_rdst *oldrd)
906dc186
TR		 642{
643 struct vxlan_rdst *rd;
644
645 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
646 if (rd)
647 return 0;
648
649 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
650 if (!rd)
651 return 0;
0c1d70af 652
ccdfd4f7 653 *oldrd = *rd;
0c1d70af 654 dst_cache_reset(&rd->dst_cache);
e4c7ed41 655 rd->remote_ip = *ip;
906dc186
TR		 656 rd->remote_port = port;
657 rd->remote_vni = vni;
658 rd->remote_ifindex = ifindex;
6ad0b5a4 659 rd->offloaded = false;
906dc186
TR		 660 return 1;
661}
662
a5e7c10a
MR		 663/* Add/update destinations for multicast */
664static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 665 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 666 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 667{
668 struct vxlan_rdst *rd;
669
670 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
671 if (rd)
672 return 0;
673
6681712d
DS		 674 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
675 if (rd == NULL)
676 return -ENOBUFS;
0c1d70af
PA		 677
678 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
679 kfree(rd);
680 return -ENOBUFS;
681 }
682
e4c7ed41 683 rd->remote_ip = *ip;
6681712d 684 rd->remote_port = port;
0efe1173 685 rd->offloaded = false;
6681712d
DS		 686 rd->remote_vni = vni;
687 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 688
689 list_add_tail_rcu(&rd->list, &f->remotes);
690
9e4b93f9 691 *rdp = rd;
6681712d
DS		 692 return 1;
693}
694
dfd8645e
TH		 695static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
696 unsigned int off,
697 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 698 __be32 vni_field,
699 struct gro_remcsum *grc,
0ace2ca8 700 bool nopartial)
dfd8645e 701{
b7fe10e5 702 size_t start, offset;
dfd8645e
TH		 703
704 if (skb->remcsum_offload)
b7fe10e5 705 return vh;
dfd8645e
TH		 706
707 if (!NAPI_GRO_CB(skb)->csum_valid)
708 return NULL;
709
54bfd872
JB		 710 start = vxlan_rco_start(vni_field);
711 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 712
b7fe10e5
TH		 713 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
714 start, offset, grc, nopartial);
dfd8645e
TH		 715
716 skb->remcsum_offload = 1;
717
718 return vh;
719}
720
d4546c25
DM		 721static struct sk_buff *vxlan_gro_receive(struct sock *sk,
722 struct list_head *head,
723 struct sk_buff *skb)
dc01e7d3 724{
d4546c25
DM		 725 struct sk_buff *pp = NULL;
726 struct sk_buff *p;
dc01e7d3 727 struct vxlanhdr *vh, *vh2;
9b174d88 728 unsigned int hlen, off_vx;
dc01e7d3 729 int flush = 1;
5602c48c 730 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 731 __be32 flags;
26c4f7da
TH		 732 struct gro_remcsum grc;
733
734 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 735
736 off_vx = skb_gro_offset(skb);
737 hlen = off_vx + sizeof(*vh);
738 vh = skb_gro_header_fast(skb, off_vx);
739 if (skb_gro_header_hard(skb, hlen)) {
740 vh = skb_gro_header_slow(skb, hlen, off_vx);
741 if (unlikely(!vh))
742 goto out;
743 }
dc01e7d3 744
dfd8645e
TH		 745 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
746
54bfd872 747 flags = vh->vx_flags;
dfd8645e
TH		 748
749 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
750 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 751 vh->vx_vni, &grc,
0ace2ca8
TH		 752 !!(vs->flags &
753 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 754
755 if (!vh)
756 goto out;
757 }
758
b7fe10e5
TH		 759 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
760
d4546c25 761 list_for_each_entry(p, head, list) {
dc01e7d3
OG		 762 if (!NAPI_GRO_CB(p)->same_flow)
763 continue;
764
765 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 766 if (vh->vx_flags != vh2->vx_flags ||
767 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 768 NAPI_GRO_CB(p)->same_flow = 0;
769 continue;
770 }
dc01e7d3
OG		 771 }
772
fcd91dd4 773 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 774 flush = 0;
dc01e7d3 775
dc01e7d3 776out:
603d4cf8 777 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 778
779 return pp;
780}
781
5602c48c 782static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 783{
229740c6
JR		 784 /* Sets 'skb->inner_mac_header' since we are always called with
785 * 'skb->encapsulation' set.
786 */
9b174d88 787 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 788}
789
25e20e73
RP		 790static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
791 const u8 *mac, __u16 state,
45598c1c 792 __be32 src_vni, __u16 ndm_flags)
25e20e73
RP		 793{
794 struct vxlan_fdb *f;
795
796 f = kmalloc(sizeof(*f), GFP_ATOMIC);
797 if (!f)
798 return NULL;
799 f->state = state;
800 f->flags = ndm_flags;
801 f->updated = f->used = jiffies;
802 f->vni = src_vni;
803 INIT_LIST_HEAD(&f->remotes);
804 memcpy(f->eth_addr, mac, ETH_ALEN);
805
806 return f;
807}
808
d342894c 809static int vxlan_fdb_create(struct vxlan_dev *vxlan,
25e20e73
RP		 810 const u8 *mac, union vxlan_addr *ip,
811 __u16 state, __be16 port, __be32 src_vni,
45598c1c 812 __be32 vni, __u32 ifindex, __u16 ndm_flags,
25e20e73
RP		 813 struct vxlan_fdb **fdb)
814{
815 struct vxlan_rdst *rd = NULL;
816 struct vxlan_fdb *f;
817 int rc;
818
819 if (vxlan->cfg.addrmax &&
820 vxlan->addrcnt >= vxlan->cfg.addrmax)
821 return -ENOSPC;
822
823 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
824 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
825 if (!f)
826 return -ENOMEM;
827
828 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
829 if (rc < 0) {
830 kfree(f);
831 return rc;
832 }
833
834 ++vxlan->addrcnt;
835 hlist_add_head_rcu(&f->hlist,
836 vxlan_fdb_head(vxlan, mac, src_vni));
837
838 *fdb = f;
839
840 return 0;
841}
842
c2b200e0
PM		 843static void vxlan_fdb_free(struct rcu_head *head)
844{
845 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
846 struct vxlan_rdst *rd, *nd;
847
848 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
849 dst_cache_destroy(&rd->dst_cache);
850 kfree(rd);
851 }
852 kfree(f);
853}
854
855static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
856 bool do_notify, bool swdev_notify)
857{
858 struct vxlan_rdst *rd;
859
860 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
861
862 --vxlan->addrcnt;
863 if (do_notify)
864 list_for_each_entry(rd, &f->remotes, list)
865 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
4c59b7d1 866 swdev_notify, NULL);
c2b200e0
PM		 867
868 hlist_del_rcu(&f->hlist);
869 call_rcu(&f->rcu, vxlan_fdb_free);
870}
871
fc4aa1ca
PM		 872static void vxlan_dst_free(struct rcu_head *head)
873{
874 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
875
876 dst_cache_destroy(&rd->dst_cache);
877 kfree(rd);
878}
879
a76d1ca2
PM		 880static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
881 union vxlan_addr *ip,
882 __u16 state, __u16 flags,
883 __be16 port, __be32 vni,
884 __u32 ifindex, __u16 ndm_flags,
885 struct vxlan_fdb *f,
4c59b7d1
PM		 886 bool swdev_notify,
887 struct netlink_ext_ack *extack)
d342894c 888{
45598c1c 889 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
9e4b93f9 890 struct vxlan_rdst *rd = NULL;
ccdfd4f7 891 struct vxlan_rdst oldrd;
d342894c 892 int notify = 0;
61f46fe8
PM		 893 int rc = 0;
894 int err;
d342894c 895
a76d1ca2
PM		 896 /* Do not allow an externally learned entry to take over an entry added
897 * by the user.
898 */
899 if (!(fdb_flags & NTF_EXT_LEARNED) ||
900 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
901 if (f->state != state) {
902 f->state = state;
903 f->updated = jiffies;
904 notify = 1;
ae884082 905 }
a76d1ca2
PM		 906 if (f->flags != fdb_flags) {
907 f->flags = fdb_flags;
908 f->updated = jiffies;
909 notify = 1;
906dc186 910 }
a76d1ca2 911 }
6681712d 912
a76d1ca2
PM		 913 if ((flags & NLM_F_REPLACE)) {
914 /* Only change unicasts */
915 if (!(is_multicast_ether_addr(f->eth_addr) ||
916 is_zero_ether_addr(f->eth_addr))) {
917 rc = vxlan_fdb_replace(f, ip, port, vni,
ccdfd4f7 918 ifindex, &oldrd);
6681712d 919 notify |= rc;
a76d1ca2 920 } else {
906dc186 921 return -EOPNOTSUPP;
a76d1ca2
PM		 922 }
923 }
924 if ((flags & NLM_F_APPEND) &&
925 (is_multicast_ether_addr(f->eth_addr) ||
926 is_zero_ether_addr(f->eth_addr))) {
927 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
906dc186 928
25e20e73 929 if (rc < 0)
17b46365 930 return rc;
a76d1ca2 931 notify |= rc;
d342894c 932 }
933
a76d1ca2
PM		 934 if (ndm_flags & NTF_USE)
935 f->used = jiffies;
936
9e4b93f9
ND		 937 if (notify) {
938 if (rd == NULL)
939 rd = first_remote_rtnl(f);
a76d1ca2 940
61f46fe8 941 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
4c59b7d1 942 swdev_notify, extack);
61f46fe8
PM		 943 if (err)
944 goto err_notify;
9e4b93f9 945 }
d342894c 946
947 return 0;
61f46fe8
PM		 948
949err_notify:
950 if ((flags & NLM_F_REPLACE) && rc)
951 *rd = oldrd;
fc4aa1ca 952 else if ((flags & NLM_F_APPEND) && rc) {
61f46fe8 953 list_del_rcu(&rd->list);
fc4aa1ca
PM		 954 call_rcu(&rd->rcu, vxlan_dst_free);
955 }
61f46fe8 956 return err;
d342894c 957}
958
a76d1ca2
PM		 959static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
960 const u8 *mac, union vxlan_addr *ip,
961 __u16 state, __u16 flags,
962 __be16 port, __be32 src_vni, __be32 vni,
963 __u32 ifindex, __u16 ndm_flags,
4c59b7d1
PM		 964 bool swdev_notify,
965 struct netlink_ext_ack *extack)
a76d1ca2
PM		 966{
967 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
968 struct vxlan_fdb *f;
969 int rc;
970
971 /* Disallow replace to add a multicast entry */
972 if ((flags & NLM_F_REPLACE) &&
973 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
974 return -EOPNOTSUPP;
975
976 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
977 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
978 vni, ifindex, fdb_flags, &f);
979 if (rc < 0)
980 return rc;
981
61f46fe8 982 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
4c59b7d1 983 swdev_notify, extack);
61f46fe8
PM		 984 if (rc)
985 goto err_notify;
986
a76d1ca2 987 return 0;
61f46fe8
PM		 988
989err_notify:
990 vxlan_fdb_destroy(vxlan, f, false, false);
991 return rc;
a76d1ca2
PM		 992}
993
994/* Add new entry to forwarding table -- assumes lock held */
995static int vxlan_fdb_update(struct vxlan_dev *vxlan,
996 const u8 *mac, union vxlan_addr *ip,
997 __u16 state, __u16 flags,
998 __be16 port, __be32 src_vni, __be32 vni,
999 __u32 ifindex, __u16 ndm_flags,
4c59b7d1
PM		 1000 bool swdev_notify,
1001 struct netlink_ext_ack *extack)
a76d1ca2
PM		 1002{
1003 struct vxlan_fdb *f;
1004
1005 f = __vxlan_find_mac(vxlan, mac, src_vni);
1006 if (f) {
1007 if (flags & NLM_F_EXCL) {
1008 netdev_dbg(vxlan->dev,
1009 "lost race to create %pM\n", mac);
1010 return -EEXIST;
1011 }
1012
1013 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1014 vni, ifindex, ndm_flags, f,
4c59b7d1 1015 swdev_notify, extack);
a76d1ca2
PM		 1016 } else {
1017 if (!(flags & NLM_F_CREATE))
1018 return -ENOENT;
1019
1020 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1021 port, src_vni, vni, ifindex,
4c59b7d1 1022 ndm_flags, swdev_notify, extack);
a76d1ca2
PM		 1023 }
1024}
1025
35cf2845 1026static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
0e6160f3 1027 struct vxlan_rdst *rd, bool swdev_notify)
35cf2845
LR		 1028{
1029 list_del_rcu(&rd->list);
4c59b7d1 1030 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
35cf2845
LR		 1031 call_rcu(&rd->rcu, vxlan_dst_free);
1032}
1033
f0b074be 1034static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 1035 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1036 __be32 *vni, u32 *ifindex)
d342894c 1037{
6681712d 1038 struct net *net = dev_net(vxlan->dev);
e4c7ed41 1039 int err;
d342894c 1040
f0b074be 1041 if (tb[NDA_DST]) {
e4c7ed41
CW		 1042 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1043 if (err)
1044 return err;
f0b074be 1045 } else {
e4c7ed41
CW		 1046 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1047 if (remote->sa.sa_family == AF_INET) {
1048 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1049 ip->sa.sa_family = AF_INET;
1050#if IS_ENABLED(CONFIG_IPV6)
1051 } else {
1052 ip->sin6.sin6_addr = in6addr_any;
1053 ip->sa.sa_family = AF_INET6;
1054#endif
1055 }
f0b074be 1056 }
d342894c 1057
6681712d 1058 if (tb[NDA_PORT]) {
73cf3317 1059 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 1060 return -EINVAL;
f0b074be
MR		 1061 *port = nla_get_be16(tb[NDA_PORT]);
1062 } else {
0dfbdf41 1063 *port = vxlan->cfg.dst_port;
f0b074be 1064 }
6681712d
DS		 1065
1066 if (tb[NDA_VNI]) {
1067 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1068 return -EINVAL;
54bfd872 1069 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 1070 } else {
1071 *vni = vxlan->default_dst.remote_vni;
1072 }
6681712d 1073
3ad7a4b1
RP		 1074 if (tb[NDA_SRC_VNI]) {
1075 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1076 return -EINVAL;
1077 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1078 } else {
1079 *src_vni = vxlan->default_dst.remote_vni;
1080 }
1081
6681712d 1082 if (tb[NDA_IFINDEX]) {
5abb0029 1083 struct net_device *tdev;
6681712d
DS		 1084
1085 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1086 return -EINVAL;
f0b074be 1087 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 1088 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 1089 if (!tdev)
6681712d 1090 return -EADDRNOTAVAIL;
f0b074be
MR		 1091 } else {
1092 *ifindex = 0;
1093 }
1094
1095 return 0;
1096}
1097
1098/* Add static entry (via netlink) */
1099static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1100 struct net_device *dev,
87b0984e
PM		 1101 const unsigned char *addr, u16 vid, u16 flags,
1102 struct netlink_ext_ack *extack)
f0b074be
MR		 1103{
1104 struct vxlan_dev *vxlan = netdev_priv(dev);
1105 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 1106 union vxlan_addr ip;
f0b074be 1107 __be16 port;
3ad7a4b1 1108 __be32 src_vni, vni;
54bfd872 1109 u32 ifindex;
f0b074be
MR		 1110 int err;
1111
1112 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1113 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1114 ndm->ndm_state);
1115 return -EINVAL;
1116 }
1117
1118 if (tb[NDA_DST] == NULL)
1119 return -EINVAL;
1120
3ad7a4b1 1121 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 1122 if (err)
1123 return err;
6681712d 1124
5933a7bb
MR		 1125 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1126 return -EAFNOSUPPORT;
1127
d342894c 1128 spin_lock_bh(&vxlan->hash_lock);
25e20e73 1129 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
45598c1c
PM		 1130 port, src_vni, vni, ifindex,
1131 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
4c59b7d1 1132 true, extack);
d342894c 1133 spin_unlock_bh(&vxlan->hash_lock);
1134
1135 return err;
1136}
1137
3ad7a4b1
RP		 1138static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1139 const unsigned char *addr, union vxlan_addr ip,
fc39c38b 1140 __be16 port, __be32 src_vni, __be32 vni,
0e6160f3 1141 u32 ifindex, bool swdev_notify)
d342894c 1142{
d342894c 1143 struct vxlan_fdb *f;
bc7892ba 1144 struct vxlan_rdst *rd = NULL;
3ad7a4b1 1145 int err = -ENOENT;
bc7892ba 1146
3ad7a4b1 1147 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 1148 if (!f)
3ad7a4b1 1149 return err;
bc7892ba 1150
e4c7ed41
CW		 1151 if (!vxlan_addr_any(&ip)) {
1152 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 1153 if (!rd)
1154 goto out;
1155 }
1156
bc7892ba
MR		 1157 /* remove a destination if it's not the only one on the list,
1158 * otherwise destroy the fdb entry
1159 */
1160 if (rd && !list_is_singular(&f->remotes)) {
0e6160f3 1161 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
bc7892ba 1162 goto out;
d342894c 1163 }
bc7892ba 1164
0e6160f3 1165 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
bc7892ba
MR		 1166
1167out:
3ad7a4b1
RP		 1168 return 0;
1169}
1170
1171/* Delete entry (via netlink) */
1172static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1173 struct net_device *dev,
1174 const unsigned char *addr, u16 vid)
1175{
1176 struct vxlan_dev *vxlan = netdev_priv(dev);
1177 union vxlan_addr ip;
1178 __be32 src_vni, vni;
1179 __be16 port;
1180 u32 ifindex;
1181 int err;
1182
1183 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1184 if (err)
1185 return err;
1186
1187 spin_lock_bh(&vxlan->hash_lock);
0e6160f3
PM		 1188 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1189 true);
d342894c 1190 spin_unlock_bh(&vxlan->hash_lock);
1191
1192 return err;
1193}
1194
1195/* Dump forwarding table */
1196static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1197 struct net_device *dev,
d297653d 1198 struct net_device *filter_dev, int *idx)
d342894c 1199{
1200 struct vxlan_dev *vxlan = netdev_priv(dev);
1201 unsigned int h;
d297653d 1202 int err = 0;
d342894c 1203
1204 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1205 struct vxlan_fdb *f;
d342894c 1206
b67bfe0d 1207 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 1208 struct vxlan_rdst *rd;
6681712d 1209
3e61aa8f 1210 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 1211 if (*idx < cb->args[2])
07a51cd3
AW		 1212 goto skip;
1213
6681712d
DS		 1214 err = vxlan_fdb_info(skb, vxlan, f,
1215 NETLINK_CB(cb->skb).portid,
1216 cb->nlh->nlmsg_seq,
1217 RTM_NEWNEIGH,
1218 NLM_F_MULTI, rd);
d297653d 1219 if (err < 0)
3e61aa8f 1220 goto out;
3e61aa8f 1221skip:
d297653d 1222 *idx += 1;
07a51cd3 1223 }
d342894c 1224 }
1225 }
3e61aa8f 1226out:
d297653d 1227 return err;
d342894c 1228}
1229
474c3c89
RP		 1230static int vxlan_fdb_get(struct sk_buff *skb,
1231 struct nlattr *tb[],
1232 struct net_device *dev,
1233 const unsigned char *addr,
1234 u16 vid, u32 portid, u32 seq,
1235 struct netlink_ext_ack *extack)
1236{
1237 struct vxlan_dev *vxlan = netdev_priv(dev);
1238 struct vxlan_fdb *f;
1239 __be32 vni;
1240 int err;
1241
1242 if (tb[NDA_VNI])
1243 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1244 else
1245 vni = vxlan->default_dst.remote_vni;
1246
1247 rcu_read_lock();
1248
1249 f = __vxlan_find_mac(vxlan, addr, vni);
1250 if (!f) {
1251 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1252 err = -ENOENT;
1253 goto errout;
1254 }
1255
1256 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1257 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1258errout:
1259 rcu_read_unlock();
1260 return err;
1261}
1262
d342894c 1263/* Watch incoming packets to learn mapping between Ethernet address
1264 * and Tunnel endpoint.
c4b49512 1265 * Return true if packet is bogus and should be dropped.
d342894c 1266 */
26a41ae6 1267static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1268 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1269 u32 src_ifindex, __be32 vni)
d342894c 1270{
1271 struct vxlan_dev *vxlan = netdev_priv(dev);
1272 struct vxlan_fdb *f;
87613de9
MS		 1273 u32 ifindex = 0;
1274
1275#if IS_ENABLED(CONFIG_IPV6)
1276 if (src_ip->sa.sa_family == AF_INET6 &&
1277 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1278 ifindex = src_ifindex;
1279#endif
d342894c 1280
3ad7a4b1 1281 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1282 if (likely(f)) {
5ca5461c 1283 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1284
87613de9
MS		 1285 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1286 rdst->remote_ifindex == ifindex))
26a41ae6 1287 return false;
1288
1289 /* Don't migrate static entries, drop packets */
e0090a9e 1290 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1291 return true;
d342894c 1292
1293 if (net_ratelimit())
1294 netdev_info(dev,
e4c7ed41 1295 "%pM migrated from %pIS to %pIS\n",
a4870f79 1296 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1297
e4c7ed41 1298 rdst->remote_ip = *src_ip;
d342894c 1299 f->updated = jiffies;
4c59b7d1 1300 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
d342894c 1301 } else {
1302 /* learned new entry */
1303 spin_lock(&vxlan->hash_lock);
3bf74b1a 1304
1305 /* close off race between vxlan_flush and incoming packets */
1306 if (netif_running(dev))
25e20e73 1307 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1308 NUD_REACHABLE,
1309 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1310 vxlan->cfg.dst_port,
3ad7a4b1 1311 vni,
3bf74b1a 1312 vxlan->default_dst.remote_vni,
4c59b7d1 1313 ifindex, NTF_SELF, true, NULL);
d342894c 1314 spin_unlock(&vxlan->hash_lock);
1315 }
26a41ae6 1316
1317 return false;
d342894c 1318}
1319
d342894c 1320/* See if multicast group is already in use by other ID */
95ab0991 1321static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1322{
553675fb 1323 struct vxlan_dev *vxlan;
c6fcc4fc 1324 struct vxlan_sock *sock4;
4053ab1b
AB		 1325#if IS_ENABLED(CONFIG_IPV6)
1326 struct vxlan_sock *sock6;
1327#endif
b1be00a6 1328 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1329
c6fcc4fc 1330 sock4 = rtnl_dereference(dev->vn4_sock);
1331
95ab0991
G		 1332 /* The vxlan_sock is only used by dev, leaving group has
1333 * no effect on other vxlan devices.
1334 */
66af846f 1335 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1336 return false;
b1be00a6 1337#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1338 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1339 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1340 return false;
1341#endif
95ab0991 1342
553675fb 1343 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1344 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1345 continue;
d342894c 1346
c6fcc4fc 1347 if (family == AF_INET &&
1348 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1349 continue;
b1be00a6 1350#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1351 if (family == AF_INET6 &&
1352 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1353 continue;
1354#endif
95ab0991
G		 1355
1356 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1357 &dev->default_dst.remote_ip))
1358 continue;
1359
1360 if (vxlan->default_dst.remote_ifindex !=
1361 dev->default_dst.remote_ifindex)
1362 continue;
1363
1364 return true;
553675fb 1365 }
d342894c 1366
1367 return false;
1368}
1369
544a773a 1370static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1371{
b1be00a6 1372 struct vxlan_net *vn;
012a5729 1373
b1be00a6 1374 if (!vs)
544a773a 1375 return false;
66af846f 1376 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1377 return false;
d342894c 1378
b1be00a6 1379 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1380 spin_lock(&vn->sock_lock);
7c47cedf 1381 hlist_del_rcu(&vs->hlist);
e7b3db5e 1382 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1383 (vs->flags & VXLAN_F_GPE) ?
1384 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1385 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1386 spin_unlock(&vn->sock_lock);
1387
544a773a 1388 return true;
d342894c 1389}
1390
b1be00a6
JB		 1391static void vxlan_sock_release(struct vxlan_dev *vxlan)
1392{
c6fcc4fc 1393 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1394#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1395 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1396
57d88182 1397 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1398#endif
1399
57d88182 1400 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1401 synchronize_net();
1402
a53cb29b
MB		 1403 vxlan_vs_del_dev(vxlan);
1404
c6fcc4fc 1405 if (__vxlan_sock_release_prep(sock4)) {
1406 udp_tunnel_sock_release(sock4->sock);
1407 kfree(sock4);
544a773a
HFS		 1408 }
1409
1410#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1411 if (__vxlan_sock_release_prep(sock6)) {
1412 udp_tunnel_sock_release(sock6->sock);
1413 kfree(sock6);
544a773a 1414 }
b1be00a6
JB		 1415#endif
1416}
1417
56ef9c90 1418/* Update multicast group membership when first VNI on
c4b49512 1419 * multicast address is brought up
7c47cedf 1420 */
56ef9c90 1421static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1422{
b1be00a6 1423 struct sock *sk;
e4c7ed41
CW		 1424 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1425 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1426 int ret = -EINVAL;
d342894c 1427
e4c7ed41 1428 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1429 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1430 struct ip_mreqn mreq = {
1431 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1432 .imr_ifindex = ifindex,
1433 };
1434
c6fcc4fc 1435 sk = sock4->sock->sk;
b1be00a6 1436 lock_sock(sk);
56ef9c90 1437 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1438 release_sock(sk);
e4c7ed41
CW		 1439#if IS_ENABLED(CONFIG_IPV6)
1440 } else {
c6fcc4fc 1441 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1442
1443 sk = sock6->sock->sk;
b1be00a6 1444 lock_sock(sk);
56ef9c90
MRL		 1445 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1446 &ip->sin6.sin6_addr);
b1be00a6 1447 release_sock(sk);
e4c7ed41
CW		 1448#endif
1449 }
3fc2de2f 1450
56ef9c90 1451 return ret;
3fc2de2f 1452}
1453
1454/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1455static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1456{
b1be00a6 1457 struct sock *sk;
e4c7ed41
CW		 1458 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1459 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1460 int ret = -EINVAL;
3fc2de2f 1461
e4c7ed41 1462 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1463 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1464 struct ip_mreqn mreq = {
1465 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1466 .imr_ifindex = ifindex,
1467 };
1468
c6fcc4fc 1469 sk = sock4->sock->sk;
b1be00a6 1470 lock_sock(sk);
56ef9c90 1471 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1472 release_sock(sk);
e4c7ed41
CW		 1473#if IS_ENABLED(CONFIG_IPV6)
1474 } else {
c6fcc4fc 1475 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1476
1477 sk = sock6->sock->sk;
b1be00a6 1478 lock_sock(sk);
56ef9c90
MRL		 1479 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1480 &ip->sin6.sin6_addr);
b1be00a6 1481 release_sock(sk);
e4c7ed41
CW		 1482#endif
1483 }
d342894c 1484
56ef9c90 1485 return ret;
d342894c 1486}
1487
f14ecebb
JB		 1488static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1489 struct sk_buff *skb, u32 vxflags)
dfd8645e 1490{
7d34fa75 1491 size_t start, offset;
dfd8645e 1492
f14ecebb
JB		 1493 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1494 goto out;
b7fe10e5 1495
f14ecebb
JB		 1496 start = vxlan_rco_start(unparsed->vx_vni);
1497 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1498
7d34fa75 1499 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1500 return false;
dfd8645e 1501
be5cfeab
JB		 1502 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1503 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1504out:
1505 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1506 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1507 return true;
dfd8645e
TH		 1508}
1509
f14ecebb 1510static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1511 struct sk_buff *skb, u32 vxflags,
10a5af23 1512 struct vxlan_metadata *md)
3288af08 1513{
f14ecebb 1514 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1515 struct metadata_dst *tun_dst;
f14ecebb
JB		 1516
1517 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1518 goto out;
3288af08 1519
3288af08
JB		 1520 md->gbp = ntohs(gbp->policy_id);
1521
10a5af23 1522 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1523 if (tun_dst) {
3288af08 1524 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1525 tun_dst->u.tun_info.options_len = sizeof(*md);
1526 }
3288af08
JB		 1527 if (gbp->dont_learn)
1528 md->gbp |= VXLAN_GBP_DONT_LEARN;
1529
1530 if (gbp->policy_applied)
1531 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1532
64f87d36
JB		 1533 /* In flow-based mode, GBP is carried in dst_metadata */
1534 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1535 skb->mark = md->gbp;
f14ecebb
JB		 1536out:
1537 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1538}
1539
e1e5314d 1540static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1541 __be16 *protocol,
e1e5314d
JB		 1542 struct sk_buff *skb, u32 vxflags)
1543{
1544 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1545
1546 /* Need to have Next Protocol set for interfaces in GPE mode. */
1547 if (!gpe->np_applied)
1548 return false;
1549 /* "The initial version is 0. If a receiver does not support the
1550 * version indicated it MUST drop the packet.
1551 */
1552 if (gpe->version != 0)
1553 return false;
1554 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1555 * processing MUST occur." However, we don't implement OAM
1556 * processing, thus drop the packet.
1557 */
1558 if (gpe->oam_flag)
1559 return false;
1560
fa20e0e3
JB		 1561 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1562 if (!*protocol)
e1e5314d 1563 return false;
e1e5314d
JB		 1564
1565 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1566 return true;
1567}
1568
1ab016e2
JB		 1569static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1570 struct vxlan_sock *vs,
3ad7a4b1 1571 struct sk_buff *skb, __be32 vni)
614732ea 1572{
614732ea 1573 union vxlan_addr saddr;
87613de9 1574 u32 ifindex = skb->dev->ifindex;
614732ea 1575
614732ea 1576 skb_reset_mac_header(skb);
614732ea
TG		 1577 skb->protocol = eth_type_trans(skb, vxlan->dev);
1578 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1579
1580 /* Ignore packet loops (and multicast echo) */
1581 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1582 return false;
614732ea 1583
760c6805 1584 /* Get address from the outer IP header */
ce212d0f 1585 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1586 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1587 saddr.sa.sa_family = AF_INET;
1588#if IS_ENABLED(CONFIG_IPV6)
1589 } else {
1ab016e2 1590 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1591 saddr.sa.sa_family = AF_INET6;
1592#endif
1593 }
1594
dc5321d7 1595 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1596 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1597 return false;
1598
1599 return true;
1600}
1601
760c6805
JB		 1602static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1603 struct sk_buff *skb)
1604{
1605 int err = 0;
1606
1607 if (vxlan_get_sk_family(vs) == AF_INET)
1608 err = IP_ECN_decapsulate(oiph, skb);
1609#if IS_ENABLED(CONFIG_IPV6)
1610 else
1611 err = IP6_ECN_decapsulate(oiph, skb);
1612#endif
1613
1614 if (unlikely(err) && log_ecn_error) {
1615 if (vxlan_get_sk_family(vs) == AF_INET)
1616 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1617 &((struct iphdr *)oiph)->saddr,
1618 ((struct iphdr *)oiph)->tos);
1619 else
1620 net_info_ratelimited("non-ECT from %pI6\n",
1621 &((struct ipv6hdr *)oiph)->saddr);
1622 }
1623 return err <= 1;
1624}
1625
d342894c 1626/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1627static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1628{
f2d1968e 1629 struct pcpu_sw_netstats *stats;
c9e78efb 1630 struct vxlan_dev *vxlan;
5cfccc5a 1631 struct vxlan_sock *vs;
f14ecebb 1632 struct vxlanhdr unparsed;
ee122c79
TG		 1633 struct vxlan_metadata _md;
1634 struct vxlan_metadata *md = &_md;
61618eea 1635 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1636 bool raw_proto = false;
f2d1968e 1637 void *oiph;
3ad7a4b1 1638 __be32 vni = 0;
d342894c 1639
e1e5314d 1640 /* Need UDP and VXLAN header to be present */
7ce04758 1641 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1642 goto drop;
d342894c 1643
f14ecebb 1644 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1645 /* VNI flag always required to be set */
1646 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1647 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1648 ntohl(vxlan_hdr(skb)->vx_flags),
1649 ntohl(vxlan_hdr(skb)->vx_vni));
1650 /* Return non vxlan pkt */
e5aed006 1651 goto drop;
d342894c 1652 }
288b01c8
JB		 1653 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1654 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1655
559835ea 1656 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1657 if (!vs)
d342894c 1658 goto drop;
d342894c 1659
3ad7a4b1
RP		 1660 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1661
49f810f0 1662 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1663 if (!vxlan)
1664 goto drop;
1665
e1e5314d
JB		 1666 /* For backwards compatibility, only allow reserved fields to be
1667 * used by VXLAN extensions if explicitly requested.
1668 */
1669 if (vs->flags & VXLAN_F_GPE) {
1670 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1671 goto drop;
1672 raw_proto = true;
1673 }
1674
1675 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1676 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1677 goto drop;
c9e78efb 1678
ee122c79 1679 if (vxlan_collect_metadata(vs)) {
10a5af23 1680 struct metadata_dst *tun_dst;
07dabf20 1681
c29a70d2 1682 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1683 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1684
ee122c79
TG		 1685 if (!tun_dst)
1686 goto drop;
1687
0f1b7354 1688 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1689
1690 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1691 } else {
1692 memset(md, 0, sizeof(*md));
1693 }
1694
f14ecebb
JB		 1695 if (vs->flags & VXLAN_F_REMCSUM_RX)
1696 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1697 goto drop;
1698 if (vs->flags & VXLAN_F_GBP)
10a5af23 1699 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1700 /* Note that GBP and GPE can never be active together. This is
1701 * ensured in vxlan_dev_configure.
1702 */
3511494c 1703
f14ecebb 1704 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1705 /* If there are any unprocessed flags remaining treat
1706 * this as a malformed packet. This behavior diverges from
1707 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1708 * in reserved fields are to be ignored. The approach here
c4b49512 1709 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1710 * is more robust and provides a little more security in
1711 * adding extensions to VXLAN.
1712 */
288b01c8 1713 goto drop;
3bf39475
TH		 1714 }
1715
e1e5314d 1716 if (!raw_proto) {
3ad7a4b1 1717 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1718 goto drop;
1719 } else {
8be0cfa4 1720 skb_reset_mac_header(skb);
e1e5314d
JB		 1721 skb->dev = vxlan->dev;
1722 skb->pkt_type = PACKET_HOST;
1723 }
f2d1968e 1724
f2d1968e
JB		 1725 oiph = skb_network_header(skb);
1726 skb_reset_network_header(skb);
1727
1728 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1729 ++vxlan->dev->stats.rx_frame_errors;
1730 ++vxlan->dev->stats.rx_errors;
1731 goto drop;
1732 }
1733
59cbf56f
ED		 1734 rcu_read_lock();
1735
1736 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1737 rcu_read_unlock();
1738 atomic_long_inc(&vxlan->dev->rx_dropped);
1739 goto drop;
1740 }
1741
f2d1968e
JB		 1742 stats = this_cpu_ptr(vxlan->dev->tstats);
1743 u64_stats_update_begin(&stats->syncp);
1744 stats->rx_packets++;
1745 stats->rx_bytes += skb->len;
1746 u64_stats_update_end(&stats->syncp);
1747
1748 gro_cells_receive(&vxlan->gro_cells, skb);
59cbf56f
ED		 1749
1750 rcu_read_unlock();
1751
5cfccc5a
PS		 1752 return 0;
1753
1754drop:
288b01c8
JB		 1755 /* Consume bad packet */
1756 kfree_skb(skb);
1757 return 0;
5cfccc5a
PS		 1758}
1759
c3a43b9f
SB		 1760/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1761static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1762{
1763 struct vxlan_dev *vxlan;
1764 struct vxlan_sock *vs;
1765 struct vxlanhdr *hdr;
1766 __be32 vni;
1767
1768 if (skb->len < VXLAN_HLEN)
1769 return -EINVAL;
1770
1771 hdr = vxlan_hdr(skb);
1772
1773 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1774 return -EINVAL;
1775
1776 vs = rcu_dereference_sk_user_data(sk);
1777 if (!vs)
1778 return -ENOENT;
1779
1780 vni = vxlan_vni(hdr->vx_vni);
1781 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1782 if (!vxlan)
1783 return -ENOENT;
1784
1785 return 0;
1786}
1787
3ad7a4b1 1788static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1789{
1790 struct vxlan_dev *vxlan = netdev_priv(dev);
1791 struct arphdr *parp;
1792 u8 *arpptr, *sha;
1793 __be32 sip, tip;
1794 struct neighbour *n;
1795
1796 if (dev->flags & IFF_NOARP)
1797 goto out;
1798
1799 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1800 dev->stats.tx_dropped++;
1801 goto out;
1802 }
1803 parp = arp_hdr(skb);
1804
1805 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1806 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1807 parp->ar_pro != htons(ETH_P_IP) ||
1808 parp->ar_op != htons(ARPOP_REQUEST) ||
1809 parp->ar_hln != dev->addr_len ||
1810 parp->ar_pln != 4)
1811 goto out;
1812 arpptr = (u8 *)parp + sizeof(struct arphdr);
1813 sha = arpptr;
1814 arpptr += dev->addr_len; /* sha */
1815 memcpy(&sip, arpptr, sizeof(sip));
1816 arpptr += sizeof(sip);
1817 arpptr += dev->addr_len; /* tha */
1818 memcpy(&tip, arpptr, sizeof(tip));
1819
1820 if (ipv4_is_loopback(tip) ||
1821 ipv4_is_multicast(tip))
1822 goto out;
1823
1824 n = neigh_lookup(&arp_tbl, &tip, dev);
1825
1826 if (n) {
e4f67add
DS		 1827 struct vxlan_fdb *f;
1828 struct sk_buff *reply;
1829
1830 if (!(n->nud_state & NUD_CONNECTED)) {
1831 neigh_release(n);
1832 goto out;
1833 }
1834
3ad7a4b1 1835 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1836 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1837 /* bridge-local neighbor */
1838 neigh_release(n);
1839 goto out;
1840 }
1841
1842 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1843 n->ha, sha);
1844
1845 neigh_release(n);
1846
7346135d
DS		 1847 if (reply == NULL)
1848 goto out;
1849
e4f67add
DS		 1850 skb_reset_mac_header(reply);
1851 __skb_pull(reply, skb_network_offset(reply));
1852 reply->ip_summed = CHECKSUM_UNNECESSARY;
1853 reply->pkt_type = PACKET_HOST;
1854
1855 if (netif_rx_ni(reply) == NET_RX_DROP)
1856 dev->stats.rx_dropped++;
dc5321d7 1857 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 1858 union vxlan_addr ipa = {
1859 .sin.sin_addr.s_addr = tip,
a45e92a5 1860 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1861 };
1862
1863 vxlan_ip_miss(dev, &ipa);
1864 }
e4f67add
DS		 1865out:
1866 consume_skb(skb);
1867 return NETDEV_TX_OK;
1868}
1869
f564f45c 1870#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1871static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1872 struct neighbour *n, bool isrouter)
1873{
1874 struct net_device *dev = request->dev;
1875 struct sk_buff *reply;
1876 struct nd_msg *ns, *na;
1877 struct ipv6hdr *pip6;
1878 u8 *daddr;
1879 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1880 int ns_olen;
1881 int i, len;
1882
f1fb08f6 1883 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1884 return NULL;
1885
1886 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1887 sizeof(*na) + na_olen + dev->needed_tailroom;
1888 reply = alloc_skb(len, GFP_ATOMIC);
1889 if (reply == NULL)
1890 return NULL;
1891
1892 reply->protocol = htons(ETH_P_IPV6);
1893 reply->dev = dev;
1894 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1895 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1896 skb_reset_mac_header(reply);
4b29dba9 1897
f1fb08f6 1898 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1899
1900 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1901 ns_olen = request->len - skb_network_offset(request) -
1902 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1903 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1904 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1905 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1906 break;
1907 }
1908 }
1909
1910 /* Ethernet header */
1911 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1912 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1913 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1914 reply->protocol = htons(ETH_P_IPV6);
1915
1916 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1917 skb_reset_network_header(reply);
4b29dba9
DS		 1918 skb_put(reply, sizeof(struct ipv6hdr));
1919
1920 /* IPv6 header */
1921
1922 pip6 = ipv6_hdr(reply);
1923 memset(pip6, 0, sizeof(struct ipv6hdr));
1924 pip6->version = 6;
1925 pip6->priority = ipv6_hdr(request)->priority;
1926 pip6->nexthdr = IPPROTO_ICMPV6;
1927 pip6->hop_limit = 255;
1928 pip6->daddr = ipv6_hdr(request)->saddr;
1929 pip6->saddr = *(struct in6_addr *)n->primary_key;
1930
1931 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1932 skb_reset_transport_header(reply);
4b29dba9 1933
4b29dba9 1934 /* Neighbor Advertisement */
b080db58 1935 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1936 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1937 na->icmph.icmp6_router = isrouter;
1938 na->icmph.icmp6_override = 1;
1939 na->icmph.icmp6_solicited = 1;
1940 na->target = ns->target;
1941 ether_addr_copy(&na->opt[2], n->ha);
1942 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1943 na->opt[1] = na_olen >> 3;
1944
1945 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1946 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1947 csum_partial(na, sizeof(*na)+na_olen, 0));
1948
1949 pip6->payload_len = htons(sizeof(*na)+na_olen);
1950
1951 skb_push(reply, sizeof(struct ipv6hdr));
1952
1953 reply->ip_summed = CHECKSUM_UNNECESSARY;
1954
1955 return reply;
1956}
1957
3ad7a4b1 1958static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1959{
1960 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 1961 const struct in6_addr *daddr;
8bff3685 1962 const struct ipv6hdr *iphdr;
4b29dba9 1963 struct inet6_dev *in6_dev;
8bff3685
XL		 1964 struct neighbour *n;
1965 struct nd_msg *msg;
f564f45c
CW		 1966
1967 in6_dev = __in6_dev_get(dev);
1968 if (!in6_dev)
1969 goto out;
1970
f564f45c 1971 iphdr = ipv6_hdr(skb);
f564f45c 1972 daddr = &iphdr->daddr;
f1fb08f6 1973 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 1974
4b29dba9
DS		 1975 if (ipv6_addr_loopback(daddr) ||
1976 ipv6_addr_is_multicast(&msg->target))
1977 goto out;
1978
1979 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1980
1981 if (n) {
1982 struct vxlan_fdb *f;
4b29dba9 1983 struct sk_buff *reply;
f564f45c
CW		 1984
1985 if (!(n->nud_state & NUD_CONNECTED)) {
1986 neigh_release(n);
1987 goto out;
1988 }
1989
3ad7a4b1 1990 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1991 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1992 /* bridge-local neighbor */
1993 neigh_release(n);
1994 goto out;
1995 }
1996
4b29dba9
DS		 1997 reply = vxlan_na_create(skb, n,
1998 !!(f ? f->flags & NTF_ROUTER : 0));
1999
f564f45c 2000 neigh_release(n);
4b29dba9
DS		 2001
2002 if (reply == NULL)
2003 goto out;
2004
2005 if (netif_rx_ni(reply) == NET_RX_DROP)
2006 dev->stats.rx_dropped++;
2007
dc5321d7 2008 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 2009 union vxlan_addr ipa = {
2010 .sin6.sin6_addr = msg->target,
a45e92a5 2011 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 2012 };
2013
f564f45c
CW		 2014 vxlan_ip_miss(dev, &ipa);
2015 }
2016
2017out:
2018 consume_skb(skb);
2019 return NETDEV_TX_OK;
2020}
2021#endif
2022
e4f67add
DS		 2023static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2024{
2025 struct vxlan_dev *vxlan = netdev_priv(dev);
2026 struct neighbour *n;
e4f67add
DS		 2027
2028 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2029 return false;
2030
2031 n = NULL;
2032 switch (ntohs(eth_hdr(skb)->h_proto)) {
2033 case ETH_P_IP:
e15a00aa
CW		 2034 {
2035 struct iphdr *pip;
2036
e4f67add
DS		 2037 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2038 return false;
2039 pip = ip_hdr(skb);
2040 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 2041 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 2042 union vxlan_addr ipa = {
2043 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 2044 .sin.sin_family = AF_INET,
e4c7ed41
CW		 2045 };
2046
2047 vxlan_ip_miss(dev, &ipa);
2048 return false;
2049 }
2050
e4f67add 2051 break;
e15a00aa
CW		 2052 }
2053#if IS_ENABLED(CONFIG_IPV6)
2054 case ETH_P_IPV6:
2055 {
2056 struct ipv6hdr *pip6;
2057
2058 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2059 return false;
2060 pip6 = ipv6_hdr(skb);
2061 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 2062 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 2063 union vxlan_addr ipa = {
2064 .sin6.sin6_addr = pip6->daddr,
a45e92a5 2065 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 2066 };
2067
2068 vxlan_ip_miss(dev, &ipa);
2069 return false;
2070 }
2071
2072 break;
2073 }
2074#endif
e4f67add
DS		 2075 default:
2076 return false;
2077 }
2078
2079 if (n) {
2080 bool diff;
2081
7367d0b5 2082 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 2083 if (diff) {
2084 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2085 dev->addr_len);
2086 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2087 }
2088 neigh_release(n);
2089 return diff;
e4c7ed41
CW		 2090 }
2091
e4f67add
DS		 2092 return false;
2093}
2094
af33c1ad 2095static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 2096 struct vxlan_metadata *md)
2097{
2098 struct vxlanhdr_gbp *gbp;
2099
db79a621
TG		 2100 if (!md->gbp)
2101 return;
2102
3511494c 2103 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 2104 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 2105
2106 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2107 gbp->dont_learn = 1;
2108
2109 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2110 gbp->policy_applied = 1;
2111
2112 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2113}
2114
e1e5314d
JB		 2115static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2116 __be16 protocol)
2117{
2118 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2119
2120 gpe->np_applied = 1;
fa20e0e3
JB		 2121 gpe->next_protocol = tun_p_from_eth_p(protocol);
2122 if (!gpe->next_protocol)
2123 return -EPFNOSUPPORT;
2124 return 0;
e1e5314d
JB		 2125}
2126
f491e56d
JB		 2127static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2128 int iphdr_len, __be32 vni,
2129 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 2130 bool udp_sum)
e4c7ed41 2131{
e4c7ed41 2132 struct vxlanhdr *vxh;
e4c7ed41
CW		 2133 int min_headroom;
2134 int err;
dfd8645e 2135 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 2136 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 2137
af33c1ad 2138 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 2139 skb->ip_summed == CHECKSUM_PARTIAL) {
2140 int csum_start = skb_checksum_start_offset(skb);
2141
2142 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2143 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2144 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 2145 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 2146 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 2147 }
e4c7ed41 2148
e4c7ed41 2149 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 2150 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 2151
2152 /* Need space for new headers (invalidates iph ptr) */
2153 err = skb_cow_head(skb, min_headroom);
e1e5314d 2154 if (unlikely(err))
c46b7897 2155 return err;
649c5b8b 2156
aed069df
AD		 2157 err = iptunnel_handle_offloads(skb, type);
2158 if (err)
c46b7897 2159 return err;
b736a623 2160
d58ff351 2161 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 2162 vxh->vx_flags = VXLAN_HF_VNI;
2163 vxh->vx_vni = vxlan_vni_field(vni);
49560532 2164
dfd8645e 2165 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 2166 unsigned int start;
dfd8645e 2167
54bfd872
JB		 2168 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2169 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2170 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 2171
2172 if (!skb_is_gso(skb)) {
2173 skb->ip_summed = CHECKSUM_NONE;
2174 skb->encapsulation = 0;
2175 }
2176 }
2177
af33c1ad
TH		 2178 if (vxflags & VXLAN_F_GBP)
2179 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 2180 if (vxflags & VXLAN_F_GPE) {
2181 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2182 if (err < 0)
c46b7897 2183 return err;
e1e5314d
JB		 2184 inner_protocol = skb->protocol;
2185 }
3511494c 2186
e1e5314d 2187 skb_set_inner_protocol(skb, inner_protocol);
039f5062 2188 return 0;
49560532 2189}
49560532 2190
655c3de1 2191static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2192 struct vxlan_sock *sock4,
1a8496ba 2193 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 2194 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 2195 struct dst_cache *dst_cache,
db3c6139 2196 const struct ip_tunnel_info *info)
1a8496ba 2197{
db3c6139 2198 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 2199 struct rtable *rt = NULL;
2200 struct flowi4 fl4;
2201
655c3de1 2202 if (!sock4)
2203 return ERR_PTR(-EIO);
2204
db3c6139
DB		 2205 if (tos && !info)
2206 use_cache = false;
2207 if (use_cache) {
0c1d70af
PA		 2208 rt = dst_cache_get_ip4(dst_cache, saddr);
2209 if (rt)
2210 return rt;
2211 }
2212
1a8496ba
JB		 2213 memset(&fl4, 0, sizeof(fl4));
2214 fl4.flowi4_oif = oif;
2215 fl4.flowi4_tos = RT_TOS(tos);
2216 fl4.flowi4_mark = skb->mark;
2217 fl4.flowi4_proto = IPPROTO_UDP;
2218 fl4.daddr = daddr;
272d96a5 2219 fl4.saddr = *saddr;
4ecb1d83
MP		 2220 fl4.fl4_dport = dport;
2221 fl4.fl4_sport = sport;
1a8496ba
JB		 2222
2223 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 2224 if (likely(!IS_ERR(rt))) {
2225 if (rt->dst.dev == dev) {
2226 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2227 ip_rt_put(rt);
2228 return ERR_PTR(-ELOOP);
2229 }
2230
1a8496ba 2231 *saddr = fl4.saddr;
0c1d70af
PA		 2232 if (use_cache)
2233 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 2234 } else {
2235 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2236 return ERR_PTR(-ENETUNREACH);
0c1d70af 2237 }
1a8496ba
JB		 2238 return rt;
2239}
2240
e5d4b29f
JB		 2241#if IS_ENABLED(CONFIG_IPV6)
2242static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 2243 struct net_device *dev,
03dc52a8 2244 struct vxlan_sock *sock6,
1400615d 2245 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 2246 __be32 label,
e5d4b29f 2247 const struct in6_addr *daddr,
0c1d70af 2248 struct in6_addr *saddr,
4ecb1d83 2249 __be16 dport, __be16 sport,
db3c6139
DB		 2250 struct dst_cache *dst_cache,
2251 const struct ip_tunnel_info *info)
e5d4b29f 2252{
db3c6139 2253 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 2254 struct dst_entry *ndst;
2255 struct flowi6 fl6;
2256 int err;
2257
c6fcc4fc 2258 if (!sock6)
2259 return ERR_PTR(-EIO);
2260
1400615d
DB		 2261 if (tos && !info)
2262 use_cache = false;
db3c6139 2263 if (use_cache) {
0c1d70af
PA		 2264 ndst = dst_cache_get_ip6(dst_cache, saddr);
2265 if (ndst)
2266 return ndst;
2267 }
2268
e5d4b29f
JB		 2269 memset(&fl6, 0, sizeof(fl6));
2270 fl6.flowi6_oif = oif;
2271 fl6.daddr = *daddr;
272d96a5 2272 fl6.saddr = *saddr;
eaa93bf4 2273 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 2274 fl6.flowi6_mark = skb->mark;
2275 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 2276 fl6.fl6_dport = dport;
2277 fl6.fl6_sport = sport;
e5d4b29f
JB		 2278
2279 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 2280 sock6->sock->sk,
e5d4b29f 2281 &ndst, &fl6);
655c3de1 2282 if (unlikely(err < 0)) {
2283 netdev_dbg(dev, "no route to %pI6\n", daddr);
2284 return ERR_PTR(-ENETUNREACH);
2285 }
2286
2287 if (unlikely(ndst->dev == dev)) {
2288 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2289 dst_release(ndst);
2290 return ERR_PTR(-ELOOP);
2291 }
e5d4b29f
JB		 2292
2293 *saddr = fl6.saddr;
db3c6139 2294 if (use_cache)
0c1d70af 2295 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 2296 return ndst;
2297}
2298#endif
2299
9dcc71e1
SS		 2300/* Bypass encapsulation if the destination is local */
2301static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 2302 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 2303{
8f84985f 2304 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2305 union vxlan_addr loopback;
2306 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
4179cb5a 2307 struct net_device *dev;
ce6502a8 2308 int len = skb->len;
9dcc71e1 2309
8f84985f
LR		 2310 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2311 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2312 skb->pkt_type = PACKET_HOST;
2313 skb->encapsulation = 0;
2314 skb->dev = dst_vxlan->dev;
2315 __skb_pull(skb, skb_network_offset(skb));
2316
e4c7ed41
CW		 2317 if (remote_ip->sa.sa_family == AF_INET) {
2318 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2319 loopback.sa.sa_family = AF_INET;
2320#if IS_ENABLED(CONFIG_IPV6)
2321 } else {
2322 loopback.sin6.sin6_addr = in6addr_loopback;
2323 loopback.sa.sa_family = AF_INET6;
2324#endif
2325 }
2326
4179cb5a
ED		 2327 rcu_read_lock();
2328 dev = skb->dev;
2329 if (unlikely(!(dev->flags & IFF_UP))) {
2330 kfree_skb(skb);
2331 goto drop;
2332 }
2333
dc5321d7 2334 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
4179cb5a 2335 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
9dcc71e1
SS		 2336
2337 u64_stats_update_begin(&tx_stats->syncp);
2338 tx_stats->tx_packets++;
ce6502a8 2339 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2340 u64_stats_update_end(&tx_stats->syncp);
2341
2342 if (netif_rx(skb) == NET_RX_SUCCESS) {
2343 u64_stats_update_begin(&rx_stats->syncp);
2344 rx_stats->rx_packets++;
ce6502a8 2345 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2346 u64_stats_update_end(&rx_stats->syncp);
2347 } else {
4179cb5a 2348drop:
ce6502a8 2349 dev->stats.rx_dropped++;
9dcc71e1 2350 }
4179cb5a 2351 rcu_read_unlock();
9dcc71e1
SS		 2352}
2353
fee1fad7 2354static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2355 struct vxlan_dev *vxlan,
2356 union vxlan_addr *daddr,
2357 __be16 dst_port, int dst_ifindex, __be32 vni,
2358 struct dst_entry *dst,
fee1fad7 2359 u32 rt_flags)
2360{
2361#if IS_ENABLED(CONFIG_IPV6)
2362 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2363 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2364 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2365 */
2366 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2367#endif
2368 /* Bypass encapsulation if the destination is local */
2369 if (rt_flags & RTCF_LOCAL &&
2370 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2371 struct vxlan_dev *dst_vxlan;
2372
2373 dst_release(dst);
49f810f0 2374 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2375 daddr->sa.sa_family, dst_port,
dc5321d7 2376 vxlan->cfg.flags);
fee1fad7 2377 if (!dst_vxlan) {
2378 dev->stats.tx_errors++;
2379 kfree_skb(skb);
2380
2381 return -ENOENT;
2382 }
3ad7a4b1 2383 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2384 return 1;
2385 }
2386
2387 return 0;
2388}
2389
4ad16930 2390static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2391 __be32 default_vni, struct vxlan_rdst *rdst,
2392 bool did_rsc)
d342894c 2393{
d71785ff 2394 struct dst_cache *dst_cache;
3093fbe7 2395 struct ip_tunnel_info *info;
d342894c 2396 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2397 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2398 union vxlan_addr *dst;
272d96a5 2399 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2400 struct vxlan_metadata _md;
2401 struct vxlan_metadata *md = &_md;
e4c7ed41 2402 __be16 src_port = 0, dst_port;
655c3de1 2403 struct dst_entry *ndst = NULL;
e7f70af1 2404 __be32 vni, label;
d342894c 2405 __u8 tos, ttl;
49f810f0 2406 int ifindex;
0e6fbc5b 2407 int err;
dc5321d7 2408 u32 flags = vxlan->cfg.flags;
b4ed5cad 2409 bool udp_sum = false;
f491e56d 2410 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2411
61adedf3 2412 info = skb_tunnel_info(skb);
3093fbe7 2413
ee122c79 2414 if (rdst) {
0770b53b 2415 dst = &rdst->remote_ip;
2416 if (vxlan_addr_any(dst)) {
2417 if (did_rsc) {
2418 /* short-circuited back to local bridge */
3ad7a4b1 2419 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2420 return;
2421 }
2422 goto drop;
2423 }
2424
0dfbdf41 2425 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2426 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2427 ifindex = rdst->remote_ifindex;
1158632b 2428 local_ip = vxlan->cfg.saddr;
d71785ff 2429 dst_cache = &rdst->dst_cache;
0770b53b 2430 md->gbp = skb->mark;
72f6d71e
HL		 2431 if (flags & VXLAN_F_TTL_INHERIT) {
2432 ttl = ip_tunnel_get_ttl(old_iph, skb);
2433 } else {
2434 ttl = vxlan->cfg.ttl;
2435 if (!ttl && vxlan_addr_multicast(dst))
2436 ttl = 1;
2437 }
0770b53b 2438
2439 tos = vxlan->cfg.tos;
2440 if (tos == 1)
2441 tos = ip_tunnel_get_dsfield(old_iph, skb);
2442
2443 if (dst->sa.sa_family == AF_INET)
2444 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2445 else
2446 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2447 label = vxlan->cfg.label;
ee122c79
TG		 2448 } else {
2449 if (!info) {
2450 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2451 dev->name);
2452 goto drop;
2453 }
b1be00a6 2454 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2455 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2456 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2457 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2458 } else {
a725e514 2459 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2460 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2461 }
ee122c79 2462 dst = &remote_ip;
0770b53b 2463 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2464 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2465 ifindex = 0;
d71785ff 2466 dst_cache = &info->dst_cache;
256c87c1
PJV		 2467 if (info->options_len &&
2468 info->key.tun_flags & TUNNEL_VXLAN_OPT)
0770b53b 2469 md = ip_tunnel_info_opts(info);
a725e514
JB		 2470 ttl = info->key.ttl;
2471 tos = info->key.tos;
e7f70af1 2472 label = info->key.label;
b4ed5cad 2473 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2474 }
0770b53b 2475 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2476 vxlan->cfg.port_max, true);
a725e514 2477
56de859e 2478 rcu_read_lock();
e4c7ed41 2479 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2480 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2481 struct rtable *rt;
0770b53b 2482 __be16 df = 0;
c6fcc4fc 2483
aab8cc36
AB		 2484 if (!ifindex)
2485 ifindex = sock4->sock->sk->sk_bound_dev_if;
2486
49f810f0 2487 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2488 dst->sin.sin_addr.s_addr,
1158632b 2489 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2490 dst_port, src_port,
d71785ff 2491 dst_cache, info);
8ebd115b
DM		 2492 if (IS_ERR(rt)) {
2493 err = PTR_ERR(rt);
c46b7897 2494 goto tx_error;
8ebd115b 2495 }
e4c7ed41 2496
fee1fad7 2497 if (!info) {
b4d30697 2498 /* Bypass encapsulation if the destination is local */
fee1fad7 2499 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2500 dst_port, ifindex, vni,
2501 &rt->dst, rt->rt_flags);
fee1fad7 2502 if (err)
56de859e 2503 goto out_unlock;
b4d30697
SB		 2504
2505 if (vxlan->cfg.df == VXLAN_DF_SET) {
2506 df = htons(IP_DF);
2507 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2508 struct ethhdr *eth = eth_hdr(skb);
2509
2510 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2511 (ntohs(eth->h_proto) == ETH_P_IP &&
2512 old_iph->frag_off & htons(IP_DF)))
2513 df = htons(IP_DF);
2514 }
fee1fad7 2515 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2516 df = htons(IP_DF);
fee1fad7 2517 }
6ceb31ca 2518
c46b7897 2519 ndst = &rt->dst;
6b4f92af 2520 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
a93bf0ff 2521
e4c7ed41
CW		 2522 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2523 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2524 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2525 vni, md, flags, udp_sum);
f491e56d 2526 if (err < 0)
c46b7897 2527 goto tx_error;
f491e56d 2528
1158632b 2529 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2530 dst->sin.sin_addr.s_addr, tos, ttl, df,
2531 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2532#if IS_ENABLED(CONFIG_IPV6)
2533 } else {
c6fcc4fc 2534 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2535
aab8cc36
AB		 2536 if (!ifindex)
2537 ifindex = sock6->sock->sk->sk_bound_dev_if;
2538
49f810f0 2539 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2540 label, &dst->sin6.sin6_addr,
1158632b 2541 &local_ip.sin6.sin6_addr,
4ecb1d83 2542 dst_port, src_port,
db3c6139 2543 dst_cache, info);
e5d4b29f 2544 if (IS_ERR(ndst)) {
8ebd115b 2545 err = PTR_ERR(ndst);
c46b7897 2546 ndst = NULL;
9dcc71e1 2547 goto tx_error;
e4c7ed41 2548 }
655c3de1 2549
fee1fad7 2550 if (!info) {
2551 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2552
fee1fad7 2553 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2554 dst_port, ifindex, vni,
2555 ndst, rt6i_flags);
fee1fad7 2556 if (err)
56de859e 2557 goto out_unlock;
fee1fad7 2558 }
35e2d115 2559
6b4f92af 2560 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
a93bf0ff 2561
1400615d 2562 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2563 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2564 skb_scrub_packet(skb, xnet);
2565 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2566 vni, md, flags, udp_sum);
c46b7897 2567 if (err < 0)
2568 goto tx_error;
2569
0770b53b 2570 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2571 &local_ip.sin6.sin6_addr,
272d96a5 2572 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2573 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2574#endif
2575 }
56de859e
JK		 2576out_unlock:
2577 rcu_read_unlock();
4ad16930 2578 return;
d342894c 2579
2580drop:
2581 dev->stats.tx_dropped++;
c46b7897 2582 dev_kfree_skb(skb);
2583 return;
d342894c 2584
2585tx_error:
56de859e 2586 rcu_read_unlock();
655c3de1 2587 if (err == -ELOOP)
2588 dev->stats.collisions++;
2589 else if (err == -ENETUNREACH)
2590 dev->stats.tx_carrier_errors++;
c46b7897 2591 dst_release(ndst);
d342894c 2592 dev->stats.tx_errors++;
c46b7897 2593 kfree_skb(skb);
d342894c 2594}
2595
6681712d
DS		 2596/* Transmit local packets over Vxlan
2597 *
2598 * Outer IP header inherits ECN and DF from inner header.
2599 * Outer UDP destination is the VXLAN assigned port.
2600 * source port is based on hash of flow
2601 */
2602static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2603{
2604 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2605 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2606 const struct ip_tunnel_info *info;
6681712d 2607 bool did_rsc = false;
6681712d 2608 struct vxlan_fdb *f;
8bff3685 2609 struct ethhdr *eth;
3ad7a4b1 2610 __be32 vni = 0;
6681712d 2611
61adedf3 2612 info = skb_tunnel_info(skb);
3093fbe7 2613
6681712d 2614 skb_reset_mac_header(skb);
6681712d 2615
dc5321d7 2616 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2617 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2618 info->mode & IP_TUNNEL_INFO_TX) {
2619 vni = tunnel_id_to_key32(info->key.tun_id);
2620 } else {
2621 if (info && info->mode & IP_TUNNEL_INFO_TX)
2622 vxlan_xmit_one(skb, dev, vni, NULL, false);
2623 else
2624 kfree_skb(skb);
2625 return NETDEV_TX_OK;
2626 }
47e5d1b0
JB		 2627 }
2628
dc5321d7 2629 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2630 eth = eth_hdr(skb);
f564f45c 2631 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2632 return arp_reduce(dev, skb, vni);
f564f45c 2633#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 2634 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2635 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2636 sizeof(struct nd_msg)) &&
2637 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2638 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2639
2640 if (m->icmph.icmp6_code == 0 &&
2641 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2642 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2643 }
2644#endif
2645 }
6681712d 2646
47e5d1b0 2647 eth = eth_hdr(skb);
3ad7a4b1 2648 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2649 did_rsc = false;
2650
dc5321d7 2651 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2652 (ntohs(eth->h_proto) == ETH_P_IP ||
2653 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2654 did_rsc = route_shortcircuit(dev, skb);
2655 if (did_rsc)
3ad7a4b1 2656 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2657 }
2658
6681712d 2659 if (f == NULL) {
3ad7a4b1 2660 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2661 if (f == NULL) {
dc5321d7 2662 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2663 !is_multicast_ether_addr(eth->h_dest))
2664 vxlan_fdb_miss(vxlan, eth->h_dest);
2665
2666 dev->stats.tx_dropped++;
8f646c92 2667 kfree_skb(skb);
afbd8bae
MR		 2668 return NETDEV_TX_OK;
2669 }
2670 }
6681712d 2671
afbd8bae
MR		 2672 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2673 struct sk_buff *skb1;
6681712d 2674
8f646c92
ED		 2675 if (!fdst) {
2676 fdst = rdst;
2677 continue;
2678 }
afbd8bae
MR		 2679 skb1 = skb_clone(skb, GFP_ATOMIC);
2680 if (skb1)
3ad7a4b1 2681 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2682 }
2683
8f646c92 2684 if (fdst)
3ad7a4b1 2685 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2686 else
2687 kfree_skb(skb);
4ad16930 2688 return NETDEV_TX_OK;
6681712d
DS		 2689}
2690
d342894c 2691/* Walk the forwarding table and purge stale entries */
df7e828c 2692static void vxlan_cleanup(struct timer_list *t)
d342894c 2693{
df7e828c 2694 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2695 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2696 unsigned int h;
2697
2698 if (!netif_running(vxlan->dev))
2699 return;
2700
d342894c 2701 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2702 struct hlist_node *p, *n;
14e1d0fa 2703
f98ec788 2704 spin_lock(&vxlan->hash_lock);
d342894c 2705 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2706 struct vxlan_fdb *f
2707 = container_of(p, struct vxlan_fdb, hlist);
2708 unsigned long timeout;
2709
efb5f68f 2710 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2711 continue;
2712
def499c9
RP		 2713 if (f->flags & NTF_EXT_LEARNED)
2714 continue;
2715
0dfbdf41 2716 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2717 if (time_before_eq(timeout, jiffies)) {
2718 netdev_dbg(vxlan->dev,
2719 "garbage collect %pM\n",
2720 f->eth_addr);
2721 f->state = NUD_STALE;
0e6160f3 2722 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 2723 } else if (time_before(timeout, next_timer))
2724 next_timer = timeout;
2725 }
f98ec788 2726 spin_unlock(&vxlan->hash_lock);
d342894c 2727 }
d342894c 2728
2729 mod_timer(&vxlan->age_timer, next_timer);
2730}
2731
a53cb29b
MB		 2732static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2733{
2734 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2735
2736 spin_lock(&vn->sock_lock);
69e76661
JB		 2737 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2738#if IS_ENABLED(CONFIG_IPV6)
2739 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2740#endif
a53cb29b
MB		 2741 spin_unlock(&vn->sock_lock);
2742}
2743
69e76661
JB		 2744static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2745 struct vxlan_dev_node *node)
9c2e24e1 2746{
56ef9c90 2747 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2748 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2749
69e76661 2750 node->vxlan = vxlan;
56ef9c90 2751 spin_lock(&vn->sock_lock);
69e76661 2752 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 2753 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2754}
2755
d342894c 2756/* Setup stats when device is created */
2757static int vxlan_init(struct net_device *dev)
2758{
1c213bd2 2759 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2760 if (!dev->tstats)
d342894c 2761 return -ENOMEM;
2762
2763 return 0;
2764}
2765
3ad7a4b1 2766static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2767{
2768 struct vxlan_fdb *f;
2769
2770 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2771 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2772 if (f)
0e6160f3 2773 vxlan_fdb_destroy(vxlan, f, true, true);
afbd8bae
MR		 2774 spin_unlock_bh(&vxlan->hash_lock);
2775}
2776
ebf4063e
SH		 2777static void vxlan_uninit(struct net_device *dev)
2778{
2779 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2780
ad6c9986
SB		 2781 gro_cells_destroy(&vxlan->gro_cells);
2782
3ad7a4b1 2783 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2784
ebf4063e
SH		 2785 free_percpu(dev->tstats);
2786}
2787
d342894c 2788/* Start ageing timer and join group when device is brought up */
2789static int vxlan_open(struct net_device *dev)
2790{
2791 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2792 int ret;
56ef9c90 2793
205f356d
JB		 2794 ret = vxlan_sock_add(vxlan);
2795 if (ret < 0)
2796 return ret;
d342894c 2797
79d4a94f 2798 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2799 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2800 if (ret == -EADDRINUSE)
2801 ret = 0;
56ef9c90 2802 if (ret) {
205f356d 2803 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2804 return ret;
2805 }
d342894c 2806 }
2807
0dfbdf41 2808 if (vxlan->cfg.age_interval)
d342894c 2809 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2810
56ef9c90 2811 return ret;
d342894c 2812}
2813
2814/* Purge the forwarding table */
8b3f9337 2815static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2816{
31fec5aa 2817 unsigned int h;
d342894c 2818
2819 spin_lock_bh(&vxlan->hash_lock);
2820 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2821 struct hlist_node *p, *n;
2822 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2823 struct vxlan_fdb *f
2824 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2825 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2826 continue;
afbd8bae
MR		 2827 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2828 if (!is_zero_ether_addr(f->eth_addr))
0e6160f3 2829 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 2830 }
2831 }
2832 spin_unlock_bh(&vxlan->hash_lock);
2833}
2834
2835/* Cleanup timer and forwarding table on shutdown */
2836static int vxlan_stop(struct net_device *dev)
2837{
2838 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2839 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2840 int ret = 0;
d342894c 2841
24c0e683 2842 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2843 !vxlan_group_used(vn, vxlan))
56ef9c90 2844 ret = vxlan_igmp_leave(vxlan);
d342894c 2845
2846 del_timer_sync(&vxlan->age_timer);
2847
8b3f9337 2848 vxlan_flush(vxlan, false);
205f356d 2849 vxlan_sock_release(vxlan);
d342894c 2850
56ef9c90 2851 return ret;
d342894c 2852}
2853
d342894c 2854/* Stub, nothing needs to be done. */
2855static void vxlan_set_multicast_list(struct net_device *dev)
2856{
2857}
2858
91572088 2859static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2860{
91572088
JW		 2861 struct vxlan_dev *vxlan = netdev_priv(dev);
2862 struct vxlan_rdst *dst = &vxlan->default_dst;
2863 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2864 dst->remote_ifindex);
ce44a4ae 2865 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2866
91572088
JW		 2867 /* This check is different than dev->max_mtu, because it looks at
2868 * the lowerdev->mtu, rather than the static dev->max_mtu
2869 */
2870 if (lowerdev) {
2871 int max_mtu = lowerdev->mtu -
2872 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2873 if (new_mtu > max_mtu)
72564b59 2874 return -EINVAL;
72564b59
DW		 2875 }
2876
345010b5
DB		 2877 dev->mtu = new_mtu;
2878 return 0;
2879}
2880
fc4099f1
PS		 2881static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2882{
2883 struct vxlan_dev *vxlan = netdev_priv(dev);
2884 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2885 __be16 sport, dport;
2886
2887 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2888 vxlan->cfg.port_max, true);
2889 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2890
239e944f 2891 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2892 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2893 struct rtable *rt;
2894
655c3de1 2895 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2896 info->key.u.ipv4.dst,
22f0708a
PA		 2897 &info->key.u.ipv4.src, dport, sport,
2898 &info->dst_cache, info);
1a8496ba
JB		 2899 if (IS_ERR(rt))
2900 return PTR_ERR(rt);
2901 ip_rt_put(rt);
239e944f
JB		 2902 } else {
2903#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2904 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2905 struct dst_entry *ndst;
2906
655c3de1 2907 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2908 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2909 &info->key.u.ipv6.src, dport, sport,
2910 &info->dst_cache, info);
239e944f
JB		 2911 if (IS_ERR(ndst))
2912 return PTR_ERR(ndst);
2913 dst_release(ndst);
239e944f
JB		 2914#else /* !CONFIG_IPV6 */
2915 return -EPFNOSUPPORT;
2916#endif
2917 }
1a8496ba
JB		 2918 info->key.tp_src = sport;
2919 info->key.tp_dst = dport;
239e944f 2920 return 0;
fc4099f1
PS		 2921}
2922
0c867c9b 2923static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2924 .ndo_init = vxlan_init,
ebf4063e 2925 .ndo_uninit = vxlan_uninit,
d342894c 2926 .ndo_open = vxlan_open,
2927 .ndo_stop = vxlan_stop,
2928 .ndo_start_xmit = vxlan_xmit,
e8171045 2929 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2930 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2931 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2932 .ndo_validate_addr = eth_validate_addr,
2933 .ndo_set_mac_address = eth_mac_addr,
2934 .ndo_fdb_add = vxlan_fdb_add,
2935 .ndo_fdb_del = vxlan_fdb_delete,
2936 .ndo_fdb_dump = vxlan_fdb_dump,
474c3c89 2937 .ndo_fdb_get = vxlan_fdb_get,
fc4099f1 2938 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
8f1af75d 2939 .ndo_change_proto_down = dev_change_proto_down_generic,
d342894c 2940};
2941
e1e5314d
JB		 2942static const struct net_device_ops vxlan_netdev_raw_ops = {
2943 .ndo_init = vxlan_init,
2944 .ndo_uninit = vxlan_uninit,
2945 .ndo_open = vxlan_open,
2946 .ndo_stop = vxlan_stop,
2947 .ndo_start_xmit = vxlan_xmit,
2948 .ndo_get_stats64 = ip_tunnel_get_stats64,
2949 .ndo_change_mtu = vxlan_change_mtu,
2950 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2951};
2952
d342894c 2953/* Info for udev, that this is a virtual tunnel endpoint */
2954static struct device_type vxlan_type = {
2955 .name = "vxlan",
2956};
2957
e5de25dc 2958/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2959 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2960 * to implement the ndo_udp_tunnel_add.
53cf5275 2961 */
2d2b13fc 2962static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 2963{
2964 struct vxlan_sock *vs;
2965 struct net *net = dev_net(dev);
2966 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2967 unsigned int i;
53cf5275
JG		 2968
2969 spin_lock(&vn->sock_lock);
2970 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 2971 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2972 unsigned short type;
2973
2974 if (vs->flags & VXLAN_F_GPE)
2975 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2976 else
2977 type = UDP_TUNNEL_TYPE_VXLAN;
2978
2979 if (push)
2980 udp_tunnel_push_rx_port(dev, vs->sock, type);
2981 else
2982 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2983 }
53cf5275
JG		 2984 }
2985 spin_unlock(&vn->sock_lock);
2986}
53cf5275 2987
d342894c 2988/* Initialize the device structure. */
2989static void vxlan_setup(struct net_device *dev)
2990{
2991 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2992 unsigned int h;
d342894c 2993
65226ef8
JB		 2994 eth_hw_addr_random(dev);
2995 ether_setup(dev);
2996
cf124db5 2997 dev->needs_free_netdev = true;
d342894c 2998 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2999
d342894c 3000 dev->features |= NETIF_F_LLTX;
d6727fe3 3001 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 3002 dev->features |= NETIF_F_RXCSUM;
05c0db08 3003 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 3004
1eaa8178 3005 dev->vlan_features = dev->features;
0afb1666 3006 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 3007 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 3008 netif_keep_dst(dev);
0c867c9b 3009 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 3010
a985343b
MS		 3011 /* MTU range: 68 - 65535 */
3012 dev->min_mtu = ETH_MIN_MTU;
3013 dev->max_mtu = ETH_MAX_MTU;
3014
553675fb 3015 INIT_LIST_HEAD(&vxlan->next);
d342894c 3016 spin_lock_init(&vxlan->hash_lock);
3017
df7e828c 3018 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 3019
3020 vxlan->dev = dev;
3021
58ce31cc
TH		 3022 gro_cells_init(&vxlan->gro_cells, dev);
3023
d342894c 3024 for (h = 0; h < FDB_HASH_SIZE; ++h)
3025 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3026}
3027
0c867c9b
JB		 3028static void vxlan_ether_setup(struct net_device *dev)
3029{
0c867c9b
JB		 3030 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3031 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3032 dev->netdev_ops = &vxlan_netdev_ether_ops;
3033}
3034
e1e5314d
JB		 3035static void vxlan_raw_setup(struct net_device *dev)
3036{
65226ef8 3037 dev->header_ops = NULL;
e1e5314d
JB		 3038 dev->type = ARPHRD_NONE;
3039 dev->hard_header_len = 0;
3040 dev->addr_len = 0;
e1e5314d
JB		 3041 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3042 dev->netdev_ops = &vxlan_netdev_raw_ops;
3043}
3044
d342894c 3045static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3046 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 3047 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 3048 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 3049 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3050 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 3051 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 3052 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3053 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 3054 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 3055 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3056 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3057 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 3058 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 3059 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3060 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3061 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3062 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 3063 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 3064 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 3065 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3066 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3067 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 3068 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3069 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 3070 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 3071 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 3072 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
72f6d71e 3073 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
b4d30697 3074 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
d342894c 3075};
3076
a8b8a889
MS		 3077static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3078 struct netlink_ext_ack *extack)
d342894c 3079{
3080 if (tb[IFLA_ADDRESS]) {
3081 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 3082 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3083 "Provided link layer address is not Ethernet");
d342894c 3084 return -EINVAL;
3085 }
3086
3087 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 3088 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3089 "Provided Ethernet address is not unicast");
d342894c 3090 return -EADDRNOTAVAIL;
3091 }
3092 }
3093
a985343b 3094 if (tb[IFLA_MTU]) {
019b13ae 3095 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 3096
653ef6a3
GM		 3097 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3098 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3099 "MTU must be between 68 and 65535");
a985343b 3100 return -EINVAL;
653ef6a3 3101 }
a985343b
MS		 3102 }
3103
653ef6a3
GM		 3104 if (!data) {
3105 NL_SET_ERR_MSG(extack,
3106 "Required attributes not provided to perform the operation");
d342894c 3107 return -EINVAL;
653ef6a3 3108 }
d342894c 3109
3110 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 3111 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3112
653ef6a3
GM		 3113 if (id >= VXLAN_N_VID) {
3114 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
3115 "VXLAN ID must be lower than 16777216");
d342894c 3116 return -ERANGE;
653ef6a3 3117 }
d342894c 3118 }
3119
05f47d69 3120 if (data[IFLA_VXLAN_PORT_RANGE]) {
3121 const struct ifla_vxlan_port_range *p
3122 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3123
3124 if (ntohs(p->high) < ntohs(p->low)) {
653ef6a3
GM		 3125 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3126 "Invalid source port range");
05f47d69 3127 return -EINVAL;
3128 }
3129 }
3130
b4d30697
SB		 3131 if (data[IFLA_VXLAN_DF]) {
3132 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3133
3134 if (df < 0 || df > VXLAN_DF_MAX) {
3135 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_DF],
3136 "Invalid DF attribute");
3137 return -EINVAL;
3138 }
3139 }
3140
d342894c 3141 return 0;
3142}
3143
1b13c97f
YB		 3144static void vxlan_get_drvinfo(struct net_device *netdev,
3145 struct ethtool_drvinfo *drvinfo)
3146{
3147 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3148 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3149}
3150
3151static const struct ethtool_ops vxlan_ethtool_ops = {
3152 .get_drvinfo = vxlan_get_drvinfo,
3153 .get_link = ethtool_op_get_link,
3154};
3155
3ee64f39 3156static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
aab8cc36 3157 __be16 port, u32 flags, int ifindex)
553675fb 3158{
e4c7ed41 3159 struct socket *sock;
3ee64f39
TH		 3160 struct udp_port_cfg udp_conf;
3161 int err;
e4c7ed41 3162
3ee64f39 3163 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 3164
3ee64f39
TH		 3165 if (ipv6) {
3166 udp_conf.family = AF_INET6;
3ee64f39 3167 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 3168 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 3169 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 3170 } else {
3171 udp_conf.family = AF_INET;
e4c7ed41 3172 }
e4c7ed41 3173
3ee64f39 3174 udp_conf.local_udp_port = port;
aab8cc36 3175 udp_conf.bind_ifindex = ifindex;
553675fb 3176
3ee64f39
TH		 3177 /* Open UDP socket */
3178 err = udp_sock_create(net, &udp_conf, &sock);
3179 if (err < 0)
3180 return ERR_PTR(err);
e4c7ed41 3181
39deb2c7 3182 return sock;
e4c7ed41
CW		 3183}
3184
3185/* Create new listen socket if needed */
b1be00a6 3186static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
aab8cc36
AB		 3187 __be16 port, u32 flags,
3188 int ifindex)
e4c7ed41
CW		 3189{
3190 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3191 struct vxlan_sock *vs;
3192 struct socket *sock;
e4c7ed41 3193 unsigned int h;
acbf74a7 3194 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 3195
dc01e7d3 3196 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 3197 if (!vs)
3198 return ERR_PTR(-ENOMEM);
3199
3200 for (h = 0; h < VNI_HASH_SIZE; ++h)
3201 INIT_HLIST_HEAD(&vs->vni_list[h]);
3202
aab8cc36 3203 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
39deb2c7 3204 if (IS_ERR(sock)) {
553675fb 3205 kfree(vs);
e50fddc8 3206 return ERR_CAST(sock);
553675fb 3207 }
e4c7ed41
CW		 3208
3209 vs->sock = sock;
66af846f 3210 refcount_set(&vs->refcnt, 1);
af33c1ad 3211 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 3212
9c2e24e1
PS		 3213 spin_lock(&vn->sock_lock);
3214 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 3215 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 3216 (vs->flags & VXLAN_F_GPE) ?
3217 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 3218 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 3219 spin_unlock(&vn->sock_lock);
553675fb 3220
3221 /* Mark socket as an encapsulation socket. */
5602c48c 3222 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 3223 tunnel_cfg.sk_user_data = vs;
3224 tunnel_cfg.encap_type = 1;
f2d1968e 3225 tunnel_cfg.encap_rcv = vxlan_rcv;
c3a43b9f 3226 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
acbf74a7 3227 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 3228 tunnel_cfg.gro_receive = vxlan_gro_receive;
3229 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 3230
3231 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 3232
9c2e24e1
PS		 3233 return vs;
3234}
3235
b1be00a6 3236static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 3237{
205f356d
JB		 3238 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3239 struct vxlan_sock *vs = NULL;
69e76661 3240 struct vxlan_dev_node *node;
aab8cc36
AB		 3241 int l3mdev_index = 0;
3242
3243 if (vxlan->cfg.remote_ifindex)
3244 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3245 vxlan->net, vxlan->cfg.remote_ifindex);
9c2e24e1 3246
205f356d 3247 if (!vxlan->cfg.no_share) {
56ef9c90 3248 spin_lock(&vn->sock_lock);
205f356d 3249 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
aab8cc36
AB		 3250 vxlan->cfg.dst_port, vxlan->cfg.flags,
3251 l3mdev_index);
66af846f 3252 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 3253 spin_unlock(&vn->sock_lock);
205f356d 3254 return -EBUSY;
56ef9c90
MRL		 3255 }
3256 spin_unlock(&vn->sock_lock);
3257 }
205f356d 3258 if (!vs)
b1be00a6 3259 vs = vxlan_socket_create(vxlan->net, ipv6,
aab8cc36
AB		 3260 vxlan->cfg.dst_port, vxlan->cfg.flags,
3261 l3mdev_index);
205f356d
JB		 3262 if (IS_ERR(vs))
3263 return PTR_ERR(vs);
b1be00a6 3264#if IS_ENABLED(CONFIG_IPV6)
69e76661 3265 if (ipv6) {
c6fcc4fc 3266 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 3267 node = &vxlan->hlist6;
3268 } else
b1be00a6 3269#endif
69e76661 3270 {
c6fcc4fc 3271 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 3272 node = &vxlan->hlist4;
3273 }
3274 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 3275 return 0;
553675fb 3276}
3277
b1be00a6
JB		 3278static int vxlan_sock_add(struct vxlan_dev *vxlan)
3279{
dc5321d7
MS		 3280 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3281 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 3282 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 3283 int ret = 0;
3284
c6fcc4fc 3285 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 3286#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 3287 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 3288 if (ipv6) {
b1be00a6 3289 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 3290 if (ret < 0 && ret != -EAFNOSUPPORT)
3291 ipv4 = false;
3292 }
b1be00a6 3293#endif
d074bf96 3294 if (ipv4)
b1be00a6
JB		 3295 ret = __vxlan_sock_add(vxlan, false);
3296 if (ret < 0)
3297 vxlan_sock_release(vxlan);
3298 return ret;
3299}
3300
a985343b
MS		 3301static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3302 struct net_device **lower,
653ef6a3
GM		 3303 struct vxlan_dev *old,
3304 struct netlink_ext_ack *extack)
d342894c 3305{
33564bbb 3306 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 3307 struct vxlan_dev *tmp;
e4c7ed41 3308 bool use_ipv6 = false;
d342894c 3309
a985343b
MS		 3310 if (conf->flags & VXLAN_F_GPE) {
3311 /* For now, allow GPE only together with
3312 * COLLECT_METADATA. This can be relaxed later; in such
3313 * case, the other side of the PtP link will have to be
3314 * provided.
3315 */
3316 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3317 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 3318 NL_SET_ERR_MSG(extack,
3319 "VXLAN GPE does not support this combination of attributes");
a985343b 3320 return -EINVAL;
3555621d 3321 }
e1e5314d 3322 }
0c867c9b 3323
ce44a4ae
MS		 3324 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3325 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 3326 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 3327 conf->saddr.sa.sa_family = AF_INET;
3328 } else if (!conf->remote_ip.sa.sa_family) {
3329 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3330 } else if (!conf->saddr.sa.sa_family) {
3331 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3332 }
3333
653ef6a3
GM		 3334 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3335 NL_SET_ERR_MSG(extack,
3336 "Local and remote address must be from the same family");
ce44a4ae 3337 return -EINVAL;
653ef6a3 3338 }
e4c7ed41 3339
653ef6a3
GM		 3340 if (vxlan_addr_multicast(&conf->saddr)) {
3341 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 3342 return -EINVAL;
653ef6a3 3343 }
0f22a3c6 3344
ce44a4ae 3345 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3346 if (!IS_ENABLED(CONFIG_IPV6)) {
3347 NL_SET_ERR_MSG(extack,
3348 "IPv6 support not enabled in the kernel");
057ba29b 3349 return -EPFNOSUPPORT;
653ef6a3 3350 }
e4c7ed41 3351 use_ipv6 = true;
a985343b 3352 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3353
3354 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3355 int local_type =
3356 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3357 int remote_type =
3358 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3359
3360 if (local_type & IPV6_ADDR_LINKLOCAL) {
3361 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3362 (remote_type != IPV6_ADDR_ANY)) {
3363 NL_SET_ERR_MSG(extack,
3364 "Invalid combination of local and remote address scopes");
0f22a3c6 3365 return -EINVAL;
653ef6a3 3366 }
0f22a3c6
MS		 3367
3368 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3369 } else {
3370 if (remote_type ==
653ef6a3
GM		 3371 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3372 NL_SET_ERR_MSG(extack,
3373 "Invalid combination of local and remote address scopes");
0f22a3c6 3374 return -EINVAL;
653ef6a3 3375 }
0f22a3c6
MS		 3376
3377 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3378 }
3379 }
057ba29b 3380 }
d342894c 3381
653ef6a3
GM		 3382 if (conf->label && !use_ipv6) {
3383 NL_SET_ERR_MSG(extack,
3384 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3385 return -EINVAL;
653ef6a3 3386 }
e7f70af1 3387
a985343b
MS		 3388 if (conf->remote_ifindex) {
3389 struct net_device *lowerdev;
34e02aa1 3390
a985343b 3391 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3392 if (!lowerdev) {
3393 NL_SET_ERR_MSG(extack,
3394 "Invalid local interface, device not found");
34e02aa1 3395 return -ENODEV;
653ef6a3 3396 }
d342894c 3397
e4c7ed41
CW		 3398#if IS_ENABLED(CONFIG_IPV6)
3399 if (use_ipv6) {
3400 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3401 if (idev && idev->cnf.disable_ipv6) {
3402 NL_SET_ERR_MSG(extack,
3403 "IPv6 support disabled by administrator");
e4c7ed41 3404 return -EPERM;
653ef6a3 3405 }
e4c7ed41
CW		 3406 }
3407#endif
3408
a985343b
MS		 3409 *lower = lowerdev;
3410 } else {
653ef6a3
GM		 3411 if (vxlan_addr_multicast(&conf->remote_ip)) {
3412 NL_SET_ERR_MSG(extack,
3413 "Local interface required for multicast remote destination");
3414
a985343b 3415 return -EINVAL;
653ef6a3 3416 }
1ba56fb4 3417
0f22a3c6 3418#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3419 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3420 NL_SET_ERR_MSG(extack,
3421 "Local interface required for link-local local/remote addresses");
0f22a3c6 3422 return -EINVAL;
653ef6a3 3423 }
0f22a3c6
MS		 3424#endif
3425
a985343b 3426 *lower = NULL;
9dc2ad10 3427 }
d342894c 3428
a985343b
MS		 3429 if (!conf->dst_port) {
3430 if (conf->flags & VXLAN_F_GPE)
3431 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3432 else
3433 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3434 }
3435
a985343b
MS		 3436 if (!conf->age_interval)
3437 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3438
a985343b
MS		 3439 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3440 if (tmp == old)
3441 continue;
91572088 3442
49f810f0
MS		 3443 if (tmp->cfg.vni != conf->vni)
3444 continue;
3445 if (tmp->cfg.dst_port != conf->dst_port)
3446 continue;
3447 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3448 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3449 continue;
3450
3451 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3452 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3453 continue;
3454
653ef6a3
GM		 3455 NL_SET_ERR_MSG(extack,
3456 "A VXLAN device with the specified VNI already exists");
49f810f0 3457 return -EEXIST;
a985343b 3458 }
91572088 3459
a985343b
MS		 3460 return 0;
3461}
3462
3463static void vxlan_config_apply(struct net_device *dev,
3464 struct vxlan_config *conf,
889ce937
SD		 3465 struct net_device *lowerdev,
3466 struct net *src_net,
3467 bool changelink)
a985343b
MS		 3468{
3469 struct vxlan_dev *vxlan = netdev_priv(dev);
3470 struct vxlan_rdst *dst = &vxlan->default_dst;
3471 unsigned short needed_headroom = ETH_HLEN;
3472 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3473 int max_mtu = ETH_MAX_MTU;
3474
3475 if (!changelink) {
3476 if (conf->flags & VXLAN_F_GPE)
3477 vxlan_raw_setup(dev);
3478 else
3479 vxlan_ether_setup(dev);
3480
3481 if (conf->mtu)
3482 dev->mtu = conf->mtu;
889ce937
SD		 3483
3484 vxlan->net = src_net;
a985343b
MS		 3485 }
3486
3487 dst->remote_vni = conf->vni;
3488
3489 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3490
3491 if (lowerdev) {
3492 dst->remote_ifindex = conf->remote_ifindex;
3493
3494 dev->gso_max_size = lowerdev->gso_max_size;
3495 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3496
a985343b 3497 needed_headroom = lowerdev->hard_header_len;
91572088 3498
a985343b
MS		 3499 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3500 VXLAN_HEADROOM);
f870c1ff
AK		 3501 if (max_mtu < ETH_MIN_MTU)
3502 max_mtu = ETH_MIN_MTU;
3503
3504 if (!changelink && !conf->mtu)
3505 dev->mtu = max_mtu;
7e059158
DW		 3506 }
3507
a985343b
MS		 3508 if (dev->mtu > max_mtu)
3509 dev->mtu = max_mtu;
3510
b1be00a6
JB		 3511 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3512 needed_headroom += VXLAN6_HEADROOM;
3513 else
3514 needed_headroom += VXLAN_HEADROOM;
3515 dev->needed_headroom = needed_headroom;
3516
0dfbdf41 3517 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3518}
0dfbdf41 3519
a985343b 3520static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 3521 struct vxlan_config *conf, bool changelink,
3522 struct netlink_ext_ack *extack)
a985343b
MS		 3523{
3524 struct vxlan_dev *vxlan = netdev_priv(dev);
3525 struct net_device *lowerdev;
3526 int ret;
0dfbdf41 3527
653ef6a3 3528 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 3529 if (ret)
3530 return ret;
8bcdc4f3 3531
889ce937 3532 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3533
0dfbdf41
TG		 3534 return 0;
3535}
3536
c80498e3 3537static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 3538 struct vxlan_config *conf,
3539 struct netlink_ext_ack *extack)
c80498e3
ND		 3540{
3541 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3542 struct vxlan_dev *vxlan = netdev_priv(dev);
0241b836 3543 struct vxlan_fdb *f = NULL;
6db92468 3544 bool unregister = false;
c80498e3
ND		 3545 int err;
3546
653ef6a3 3547 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 3548 if (err)
3549 return err;
3550
3551 dev->ethtool_ops = &vxlan_ethtool_ops;
3552
3553 /* create an fdb entry for a valid default destination */
3554 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
0241b836 3555 err = vxlan_fdb_create(vxlan, all_zeros_mac,
c80498e3
ND		 3556 &vxlan->default_dst.remote_ip,
3557 NUD_REACHABLE | NUD_PERMANENT,
c80498e3
ND		 3558 vxlan->cfg.dst_port,
3559 vxlan->default_dst.remote_vni,
3560 vxlan->default_dst.remote_vni,
3561 vxlan->default_dst.remote_ifindex,
0241b836 3562 NTF_SELF, &f);
c80498e3
ND		 3563 if (err)
3564 return err;
3565 }
3566
3567 err = register_netdevice(dev);
0241b836
RP		 3568 if (err)
3569 goto errout;
6db92468 3570 unregister = true;
0241b836
RP		 3571
3572 err = rtnl_configure_link(dev, NULL);
6db92468 3573 if (err)
0241b836 3574 goto errout;
c80498e3 3575
0241b836 3576 /* notify default fdb entry */
61f46fe8
PM		 3577 if (f) {
3578 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4c59b7d1 3579 RTM_NEWNEIGH, true, extack);
61f46fe8
PM		 3580 if (err)
3581 goto errout;
3582 }
0241b836 3583
c80498e3
ND		 3584 list_add(&vxlan->next, &vn->vxlan_list);
3585 return 0;
6db92468 3586
0241b836 3587errout:
6db92468
PM		 3588 /* unregister_netdevice() destroys the default FDB entry with deletion
3589 * notification. But the addition notification was not sent yet, so
3590 * destroy the entry by hand here.
3591 */
0241b836 3592 if (f)
0e6160f3 3593 vxlan_fdb_destroy(vxlan, f, false, false);
6db92468
PM		 3594 if (unregister)
3595 unregister_netdevice(dev);
0241b836 3596 return err;
c80498e3
ND		 3597}
3598
70fb0828
RP		 3599/* Set/clear flags based on attribute */
3600static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3601 int attrtype, unsigned long mask, bool changelink,
3602 bool changelink_supported,
3603 struct netlink_ext_ack *extack)
3604{
3605 unsigned long flags;
3606
3607 if (!tb[attrtype])
3608 return 0;
3609
3610 if (changelink && !changelink_supported) {
3611 vxlan_flag_attr_error(attrtype, extack);
3612 return -EOPNOTSUPP;
3613 }
3614
3615 if (vxlan_policy[attrtype].type == NLA_FLAG)
3616 flags = conf->flags | mask;
3617 else if (nla_get_u8(tb[attrtype]))
3618 flags = conf->flags | mask;
3619 else
3620 flags = conf->flags & ~mask;
3621
3622 conf->flags = flags;
3623
3624 return 0;
3625}
3626
8bcdc4f3
RP		 3627static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3628 struct net_device *dev, struct vxlan_config *conf,
70fb0828 3629 bool changelink, struct netlink_ext_ack *extack)
0dfbdf41 3630{
8bcdc4f3 3631 struct vxlan_dev *vxlan = netdev_priv(dev);
70fb0828 3632 int err = 0;
0dfbdf41 3633
8bcdc4f3 3634 memset(conf, 0, sizeof(*conf));
e277de5f 3635
8bcdc4f3
RP		 3636 /* if changelink operation, start with old existing cfg */
3637 if (changelink)
3638 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3639
3640 if (data[IFLA_VXLAN_ID]) {
3641 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3642
70fb0828
RP		 3643 if (changelink && (vni != conf->vni)) {
3644 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
8bcdc4f3 3645 return -EOPNOTSUPP;
70fb0828 3646 }
8bcdc4f3
RP		 3647 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3648 }
0dfbdf41
TG		 3649
3650 if (data[IFLA_VXLAN_GROUP]) {
70fb0828
RP		 3651 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3652 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
ce44a4ae 3653 return -EOPNOTSUPP;
70fb0828 3654 }
ce44a4ae 3655
8bcdc4f3 3656 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3657 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41 3658 } else if (data[IFLA_VXLAN_GROUP6]) {
70fb0828
RP		 3659 if (!IS_ENABLED(CONFIG_IPV6)) {
3660 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
0dfbdf41 3661 return -EPFNOSUPPORT;
70fb0828 3662 }
0dfbdf41 3663
70fb0828
RP		 3664 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3665 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
ce44a4ae 3666 return -EOPNOTSUPP;
70fb0828 3667 }
ce44a4ae 3668
8bcdc4f3
RP		 3669 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3670 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3671 }
3672
3673 if (data[IFLA_VXLAN_LOCAL]) {
70fb0828
RP		 3674 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3675 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
ce44a4ae 3676 return -EOPNOTSUPP;
70fb0828 3677 }
ce44a4ae 3678
8bcdc4f3
RP		 3679 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3680 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41 3681 } else if (data[IFLA_VXLAN_LOCAL6]) {
70fb0828
RP		 3682 if (!IS_ENABLED(CONFIG_IPV6)) {
3683 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
0dfbdf41 3684 return -EPFNOSUPPORT;
70fb0828 3685 }
0dfbdf41 3686
70fb0828
RP		 3687 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3688 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
ce44a4ae 3689 return -EOPNOTSUPP;
70fb0828 3690 }
ce44a4ae 3691
0dfbdf41 3692 /* TODO: respect scope id */
8bcdc4f3
RP		 3693 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3694 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3695 }
3696
3697 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3698 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3699
d342894c 3700 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3701 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3702
afb97186 3703 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3704 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3705
72f6d71e 3706 if (data[IFLA_VXLAN_TTL_INHERIT]) {
70fb0828
RP		 3707 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
3708 VXLAN_F_TTL_INHERIT, changelink, false,
3709 extack);
3710 if (err)
3711 return err;
3712
72f6d71e
HL		 3713 }
3714
e7f70af1 3715 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3716 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3717 IPV6_FLOWLABEL_MASK;
3718
8bcdc4f3 3719 if (data[IFLA_VXLAN_LEARNING]) {
70fb0828
RP		 3720 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
3721 VXLAN_F_LEARN, changelink, true,
3722 extack);
3723 if (err)
3724 return err;
8bcdc4f3
RP		 3725 } else if (!changelink) {
3726 /* default to learn on a new device */
3727 conf->flags |= VXLAN_F_LEARN;
3728 }
d342894c 3729
40051c4d 3730 if (data[IFLA_VXLAN_AGEING])
8bcdc4f3 3731 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 3732
8bcdc4f3 3733 if (data[IFLA_VXLAN_PROXY]) {
70fb0828
RP		 3734 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
3735 VXLAN_F_PROXY, changelink, false,
3736 extack);
3737 if (err)
3738 return err;
8bcdc4f3 3739 }
e4f67add 3740
8bcdc4f3 3741 if (data[IFLA_VXLAN_RSC]) {
70fb0828
RP		 3742 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
3743 VXLAN_F_RSC, changelink, false,
3744 extack);
3745 if (err)
3746 return err;
8bcdc4f3 3747 }
e4f67add 3748
8bcdc4f3 3749 if (data[IFLA_VXLAN_L2MISS]) {
70fb0828
RP		 3750 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
3751 VXLAN_F_L2MISS, changelink, false,
3752 extack);
3753 if (err)
3754 return err;
8bcdc4f3 3755 }
e4f67add 3756
8bcdc4f3 3757 if (data[IFLA_VXLAN_L3MISS]) {
70fb0828
RP		 3758 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
3759 VXLAN_F_L3MISS, changelink, false,
3760 extack);
3761 if (err)
3762 return err;
8bcdc4f3 3763 }
e4f67add 3764
8bcdc4f3 3765 if (data[IFLA_VXLAN_LIMIT]) {
70fb0828
RP		 3766 if (changelink) {
3767 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
3768 "Cannot change limit");
8bcdc4f3 3769 return -EOPNOTSUPP;
70fb0828 3770 }
8bcdc4f3
RP		 3771 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3772 }
d342894c 3773
8bcdc4f3 3774 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
70fb0828
RP		 3775 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
3776 VXLAN_F_COLLECT_METADATA, changelink, false,
3777 extack);
3778 if (err)
3779 return err;
8bcdc4f3 3780 }
f8a9b1bc 3781
05f47d69 3782 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3783 if (!changelink) {
3784 const struct ifla_vxlan_port_range *p
3785 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3786 conf->port_min = ntohs(p->low);
3787 conf->port_max = ntohs(p->high);
3788 } else {
70fb0828
RP		 3789 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3790 "Cannot change port range");
8bcdc4f3
RP		 3791 return -EOPNOTSUPP;
3792 }
05f47d69 3793 }
3794
8bcdc4f3 3795 if (data[IFLA_VXLAN_PORT]) {
70fb0828
RP		 3796 if (changelink) {
3797 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
3798 "Cannot change port");
8bcdc4f3 3799 return -EOPNOTSUPP;
70fb0828 3800 }
8bcdc4f3
RP		 3801 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3802 }
823aa873 3803
8bcdc4f3 3804 if (data[IFLA_VXLAN_UDP_CSUM]) {
70fb0828
RP		 3805 if (changelink) {
3806 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
3807 "Cannot change UDP_CSUM flag");
8bcdc4f3 3808 return -EOPNOTSUPP;
70fb0828 3809 }
8bcdc4f3
RP		 3810 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3811 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3812 }
359a0ea9 3813
8bcdc4f3 3814 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
70fb0828
RP		 3815 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3816 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
3817 false, extack);
3818 if (err)
3819 return err;
8bcdc4f3 3820 }
359a0ea9 3821
8bcdc4f3 3822 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
70fb0828
RP		 3823 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3824 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
3825 false, extack);
3826 if (err)
3827 return err;
8bcdc4f3 3828 }
359a0ea9 3829
8bcdc4f3 3830 if (data[IFLA_VXLAN_REMCSUM_TX]) {
70fb0828
RP		 3831 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
3832 VXLAN_F_REMCSUM_TX, changelink, false,
3833 extack);
3834 if (err)
3835 return err;
8bcdc4f3 3836 }
dfd8645e 3837
8bcdc4f3 3838 if (data[IFLA_VXLAN_REMCSUM_RX]) {
70fb0828
RP		 3839 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
3840 VXLAN_F_REMCSUM_RX, changelink, false,
3841 extack);
3842 if (err)
3843 return err;
8bcdc4f3
RP		 3844 }
3845
3846 if (data[IFLA_VXLAN_GBP]) {
70fb0828
RP		 3847 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
3848 VXLAN_F_GBP, changelink, false, extack);
3849 if (err)
3850 return err;
8bcdc4f3
RP		 3851 }
3852
3853 if (data[IFLA_VXLAN_GPE]) {
70fb0828
RP		 3854 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
3855 VXLAN_F_GPE, changelink, false,
3856 extack);
3857 if (err)
3858 return err;
8bcdc4f3
RP		 3859 }
3860
3861 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
70fb0828
RP		 3862 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
3863 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
3864 false, extack);
3865 if (err)
3866 return err;
8bcdc4f3
RP		 3867 }
3868
3869 if (tb[IFLA_MTU]) {
70fb0828
RP		 3870 if (changelink) {
3871 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3872 "Cannot change mtu");
8bcdc4f3 3873 return -EOPNOTSUPP;
70fb0828 3874 }
8bcdc4f3
RP		 3875 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3876 }
3877
b4d30697
SB		 3878 if (data[IFLA_VXLAN_DF])
3879 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
3880
8bcdc4f3
RP		 3881 return 0;
3882}
3883
3884static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 3885 struct nlattr *tb[], struct nlattr *data[],
3886 struct netlink_ext_ack *extack)
8bcdc4f3 3887{
8bcdc4f3
RP		 3888 struct vxlan_config conf;
3889 int err;
3890
70fb0828 3891 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
8bcdc4f3
RP		 3892 if (err)
3893 return err;
3894
653ef6a3 3895 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 3896}
dfd8645e 3897
8bcdc4f3 3898static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 3899 struct nlattr *data[],
3900 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 3901{
3902 struct vxlan_dev *vxlan = netdev_priv(dev);
3903 struct vxlan_rdst *dst = &vxlan->default_dst;
8db9427d 3904 struct net_device *lowerdev;
8bcdc4f3
RP		 3905 struct vxlan_config conf;
3906 int err;
3907
70fb0828 3908 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
8bcdc4f3
RP		 3909 if (err)
3910 return err;
3511494c 3911
8db9427d
PM		 3912 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
3913 vxlan, extack);
8bcdc4f3
RP		 3914 if (err)
3915 return err;
0ace2ca8 3916
8bcdc4f3 3917 /* handle default dst entry */
038a5a99 3918 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
8bcdc4f3 3919 spin_lock_bh(&vxlan->hash_lock);
038a5a99 3920 if (!vxlan_addr_any(&conf.remote_ip)) {
ce5e098f 3921 err = vxlan_fdb_update(vxlan, all_zeros_mac,
038a5a99 3922 &conf.remote_ip,
8bcdc4f3 3923 NUD_REACHABLE | NUD_PERMANENT,
ce5e098f 3924 NLM_F_APPEND | NLM_F_CREATE,
8bcdc4f3 3925 vxlan->cfg.dst_port,
038a5a99
PM		 3926 conf.vni, conf.vni,
3927 conf.remote_ifindex,
4c59b7d1 3928 NTF_SELF, true, extack);
8bcdc4f3
RP		 3929 if (err) {
3930 spin_unlock_bh(&vxlan->hash_lock);
3931 return err;
3932 }
3933 }
1cdc98c2
PM		 3934 if (!vxlan_addr_any(&dst->remote_ip))
3935 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3936 dst->remote_ip,
3937 vxlan->cfg.dst_port,
3938 dst->remote_vni,
3939 dst->remote_vni,
3940 dst->remote_ifindex,
3941 true);
8bcdc4f3
RP		 3942 spin_unlock_bh(&vxlan->hash_lock);
3943 }
ce577668 3944
038a5a99
PM		 3945 if (conf.age_interval != vxlan->cfg.age_interval)
3946 mod_timer(&vxlan->age_timer, jiffies);
3947
3948 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
8bcdc4f3 3949 return 0;
d342894c 3950}
3951
3952static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3953{
3954 struct vxlan_dev *vxlan = netdev_priv(dev);
3955
8b3f9337
RP		 3956 vxlan_flush(vxlan, true);
3957
553675fb 3958 list_del(&vxlan->next);
d342894c 3959 unregister_netdevice_queue(dev, head);
3960}
3961
3962static size_t vxlan_get_size(const struct net_device *dev)
3963{
3964
3965 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3966 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3967 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3968 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3969 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
8fd78069 3970 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
d342894c 3971 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
b4d30697 3972 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
e7f70af1 3973 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3974 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3975 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3976 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3977 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3978 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3979 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3980 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3981 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3982 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3983 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3984 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3985 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3986 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3987 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3988 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3989 0;
3990}
3991
3992static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3993{
3994 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3995 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3996 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3997 .low = htons(vxlan->cfg.port_min),
3998 .high = htons(vxlan->cfg.port_max),
05f47d69 3999 };
d342894c 4000
54bfd872 4001 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 4002 goto nla_put_failure;
4003
e4c7ed41
CW		 4004 if (!vxlan_addr_any(&dst->remote_ip)) {
4005 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 4006 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4007 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 4008 goto nla_put_failure;
4009#if IS_ENABLED(CONFIG_IPV6)
4010 } else {
930345ea
JB		 4011 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4012 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 4013 goto nla_put_failure;
4014#endif
4015 }
4016 }
d342894c 4017
c7995c43 4018 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 4019 goto nla_put_failure;
4020
0dfbdf41
TG		 4021 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4022 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 4023 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 4024 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 4025 goto nla_put_failure;
4026#if IS_ENABLED(CONFIG_IPV6)
4027 } else {
930345ea 4028 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 4029 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 4030 goto nla_put_failure;
4031#endif
4032 }
4033 }
d342894c 4034
0dfbdf41 4035 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
8fd78069
HL		 4036 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4037 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
0dfbdf41 4038 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
b4d30697 4039 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
e7f70af1 4040 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 4041 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 4042 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 4043 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS		 4044 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4045 nla_put_u8(skb, IFLA_VXLAN_RSC,
4046 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 4047 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 4048 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 4049 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 4050 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 4051 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 4052 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 4053 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4054 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4055 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 4056 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 4057 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 4058 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 4059 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 4060 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 4061 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 4062 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 4063 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 4064 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 4065 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 4066 goto nla_put_failure;
4067
05f47d69 4068 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4069 goto nla_put_failure;
4070
dc5321d7 4071 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 4072 nla_put_flag(skb, IFLA_VXLAN_GBP))
4073 goto nla_put_failure;
4074
dc5321d7 4075 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 4076 nla_put_flag(skb, IFLA_VXLAN_GPE))
4077 goto nla_put_failure;
4078
dc5321d7 4079 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 4080 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4081 goto nla_put_failure;
4082
d342894c 4083 return 0;
4084
4085nla_put_failure:
4086 return -EMSGSIZE;
4087}
4088
1728d4fa
ND		 4089static struct net *vxlan_get_link_net(const struct net_device *dev)
4090{
4091 struct vxlan_dev *vxlan = netdev_priv(dev);
4092
4093 return vxlan->net;
4094}
4095
d342894c 4096static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4097 .kind = "vxlan",
4098 .maxtype = IFLA_VXLAN_MAX,
4099 .policy = vxlan_policy,
4100 .priv_size = sizeof(struct vxlan_dev),
4101 .setup = vxlan_setup,
4102 .validate = vxlan_validate,
4103 .newlink = vxlan_newlink,
8bcdc4f3 4104 .changelink = vxlan_changelink,
d342894c 4105 .dellink = vxlan_dellink,
4106 .get_size = vxlan_get_size,
4107 .fill_info = vxlan_fill_info,
1728d4fa 4108 .get_link_net = vxlan_get_link_net,
d342894c 4109};
4110
cf5da330
ND		 4111struct net_device *vxlan_dev_create(struct net *net, const char *name,
4112 u8 name_assign_type,
4113 struct vxlan_config *conf)
4114{
4115 struct nlattr *tb[IFLA_MAX + 1];
4116 struct net_device *dev;
4117 int err;
4118
4119 memset(&tb, 0, sizeof(tb));
4120
4121 dev = rtnl_create_link(net, name, name_assign_type,
d0522f1c 4122 &vxlan_link_ops, tb, NULL);
cf5da330
ND		 4123 if (IS_ERR(dev))
4124 return dev;
4125
653ef6a3 4126 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 4127 if (err < 0) {
4128 free_netdev(dev);
4129 return ERR_PTR(err);
4130 }
4131
4132 err = rtnl_configure_link(dev, NULL);
4133 if (err < 0) {
4134 LIST_HEAD(list_kill);
4135
4136 vxlan_dellink(dev, &list_kill);
4137 unregister_netdevice_many(&list_kill);
4138 return ERR_PTR(err);
4139 }
4140
4141 return dev;
4142}
4143EXPORT_SYMBOL_GPL(vxlan_dev_create);
4144
acaf4e70
DB		 4145static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4146 struct net_device *dev)
4147{
4148 struct vxlan_dev *vxlan, *next;
4149 LIST_HEAD(list_kill);
4150
4151 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4152 struct vxlan_rdst *dst = &vxlan->default_dst;
4153
4154 /* In case we created vxlan device with carrier
4155 * and we loose the carrier due to module unload
4156 * we also need to remove vxlan device. In other
4157 * cases, it's not necessary and remote_ifindex
4158 * is 0 here, so no matches.
4159 */
4160 if (dst->remote_ifindex == dev->ifindex)
4161 vxlan_dellink(vxlan->dev, &list_kill);
4162 }
4163
4164 unregister_netdevice_many(&list_kill);
4165}
4166
b7aade15
HFS		 4167static int vxlan_netdevice_event(struct notifier_block *unused,
4168 unsigned long event, void *ptr)
acaf4e70
DB		 4169{
4170 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 4171 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 4172
04584957
SD		 4173 if (event == NETDEV_UNREGISTER) {
4174 vxlan_offload_rx_ports(dev, false);
acaf4e70 4175 vxlan_handle_lowerdev_unregister(vn, dev);
04584957
SD		 4176 } else if (event == NETDEV_REGISTER) {
4177 vxlan_offload_rx_ports(dev, true);
4178 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4179 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 4180 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 4181 }
acaf4e70
DB		 4182
4183 return NOTIFY_DONE;
4184}
4185
4186static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 4187 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 4188};
4189
0efe1173
PM		 4190static void
4191vxlan_fdb_offloaded_set(struct net_device *dev,
4192 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4193{
4194 struct vxlan_dev *vxlan = netdev_priv(dev);
4195 struct vxlan_rdst *rdst;
4196 struct vxlan_fdb *f;
4197
4198 spin_lock_bh(&vxlan->hash_lock);
4199
4200 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4201 if (!f)
4202 goto out;
4203
4204 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4205 fdb_info->remote_port,
4206 fdb_info->remote_vni,
4207 fdb_info->remote_ifindex);
4208 if (!rdst)
4209 goto out;
4210
4211 rdst->offloaded = fdb_info->offloaded;
4212
4213out:
4214 spin_unlock_bh(&vxlan->hash_lock);
4215}
4216
5728ae0d
PM		 4217static int
4218vxlan_fdb_external_learn_add(struct net_device *dev,
4219 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4220{
4221 struct vxlan_dev *vxlan = netdev_priv(dev);
4c59b7d1 4222 struct netlink_ext_ack *extack;
5728ae0d
PM		 4223 int err;
4224
4c59b7d1
PM		 4225 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4226
5728ae0d
PM		 4227 spin_lock_bh(&vxlan->hash_lock);
4228 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4229 NUD_REACHABLE,
4230 NLM_F_CREATE | NLM_F_REPLACE,
4231 fdb_info->remote_port,
4232 fdb_info->vni,
4233 fdb_info->remote_vni,
4234 fdb_info->remote_ifindex,
4235 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4c59b7d1 4236 false, extack);
5728ae0d
PM		 4237 spin_unlock_bh(&vxlan->hash_lock);
4238
4239 return err;
4240}
4241
4242static int
4243vxlan_fdb_external_learn_del(struct net_device *dev,
4244 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4245{
4246 struct vxlan_dev *vxlan = netdev_priv(dev);
4247 struct vxlan_fdb *f;
4248 int err = 0;
4249
4250 spin_lock_bh(&vxlan->hash_lock);
4251
4252 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4253 if (!f)
4254 err = -ENOENT;
4255 else if (f->flags & NTF_EXT_LEARNED)
4256 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4257 fdb_info->remote_ip,
4258 fdb_info->remote_port,
4259 fdb_info->vni,
4260 fdb_info->remote_vni,
4261 fdb_info->remote_ifindex,
4262 false);
4263
4264 spin_unlock_bh(&vxlan->hash_lock);
4265
4266 return err;
4267}
4268
0efe1173
PM		 4269static int vxlan_switchdev_event(struct notifier_block *unused,
4270 unsigned long event, void *ptr)
4271{
4272 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
5728ae0d
PM		 4273 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4274 int err = 0;
0efe1173
PM		 4275
4276 switch (event) {
4277 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4278 vxlan_fdb_offloaded_set(dev, ptr);
4279 break;
5728ae0d
PM		 4280 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4281 fdb_info = ptr;
4282 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4283 if (err) {
4284 err = notifier_from_errno(err);
4285 break;
4286 }
4287 fdb_info->offloaded = true;
4288 vxlan_fdb_offloaded_set(dev, fdb_info);
4289 break;
4290 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4291 fdb_info = ptr;
4292 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4293 if (err) {
4294 err = notifier_from_errno(err);
4295 break;
4296 }
4297 fdb_info->offloaded = false;
4298 vxlan_fdb_offloaded_set(dev, fdb_info);
4299 break;
0efe1173
PM		 4300 }
4301
5728ae0d 4302 return err;
0efe1173
PM		 4303}
4304
4305static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4306 .notifier_call = vxlan_switchdev_event,
4307};
4308
d342894c 4309static __net_init int vxlan_init_net(struct net *net)
4310{
4311 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 4312 unsigned int h;
d342894c 4313
553675fb 4314 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 4315 spin_lock_init(&vn->sock_lock);
d342894c 4316
553675fb 4317 for (h = 0; h < PORT_HASH_SIZE; ++h)
4318 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 4319
4320 return 0;
4321}
4322
57b61127 4323static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
f01ec1c0
ND		 4324{
4325 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4326 struct vxlan_dev *vxlan, *next;
4327 struct net_device *dev, *aux;
0e4ec5ac 4328 unsigned int h;
f01ec1c0 4329
f01ec1c0
ND		 4330 for_each_netdev_safe(net, dev, aux)
4331 if (dev->rtnl_link_ops == &vxlan_link_ops)
57b61127 4332 unregister_netdevice_queue(dev, head);
f01ec1c0
ND		 4333
4334 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4335 /* If vxlan->dev is in the same netns, it has already been added
4336 * to the list by the previous loop.
4337 */
cc4807bb 4338 if (!net_eq(dev_net(vxlan->dev), net))
57b61127 4339 unregister_netdevice_queue(vxlan->dev, head);
f01ec1c0
ND		 4340 }
4341
0e4ec5ac
VA		 4342 for (h = 0; h < PORT_HASH_SIZE; ++h)
4343 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND		 4344}
4345
57b61127
HY		 4346static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4347{
4348 struct net *net;
4349 LIST_HEAD(list);
4350
4351 rtnl_lock();
4352 list_for_each_entry(net, net_list, exit_list)
4353 vxlan_destroy_tunnels(net, &list);
4354
4355 unregister_netdevice_many(&list);
4356 rtnl_unlock();
4357}
4358
d342894c 4359static struct pernet_operations vxlan_net_ops = {
4360 .init = vxlan_init_net,
57b61127 4361 .exit_batch = vxlan_exit_batch_net,
d342894c 4362 .id = &vxlan_net_id,
4363 .size = sizeof(struct vxlan_net),
4364};
4365
4366static int __init vxlan_init_module(void)
4367{
4368 int rc;
4369
4370 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4371
783c1463 4372 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 4373 if (rc)
4374 goto out1;
4375
acaf4e70 4376 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 4377 if (rc)
4378 goto out2;
4379
0efe1173 4380 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 4381 if (rc)
4382 goto out3;
d342894c 4383
0efe1173
PM		 4384 rc = rtnl_link_register(&vxlan_link_ops);
4385 if (rc)
4386 goto out4;
4387
acaf4e70 4388 return 0;
0efe1173
PM		 4389out4:
4390 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 4391out3:
4392 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 4393out2:
783c1463 4394 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 4395out1:
4396 return rc;
4397}
7332a13b 4398late_initcall(vxlan_init_module);
d342894c 4399
4400static void __exit vxlan_cleanup_module(void)
4401{
b7153984 4402 rtnl_link_unregister(&vxlan_link_ops);
0efe1173 4403 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70 4404 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 4405 unregister_pernet_subsys(&vxlan_net_ops);
4406 /* rcu_barrier() is called by netns */
d342894c 4407}
4408module_exit(vxlan_cleanup_module);
4409
4410MODULE_LICENSE("GPL");
4411MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 4412MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 4413MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 4414MODULE_ALIAS_RTNL_LINK("vxlan");
Mirror https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git