]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: Add switchdev notifications
[mirror_ubuntu-jammy-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
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;
ae884082 82 u8 flags; /* see ndm_flags */
d342894c 83};
84
d342894c 85/* salt for hash table */
86static u32 vxlan_salt __read_mostly;
87
ee122c79
TG		 88static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
89{
e7030878
TG		 90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
ee122c79
TG		 92}
93
e4c7ed41
CW		 94#if IS_ENABLED(CONFIG_IPV6)
95static inline
96bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
97{
f0ef3126
JB		 98 if (a->sa.sa_family != b->sa.sa_family)
99 return false;
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
102 else
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 104}
105
e4c7ed41
CW		 106static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
107{
f0ef3126 108 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 109 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 110 ip->sa.sa_family = AF_INET6;
111 return 0;
112 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 113 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 114 ip->sa.sa_family = AF_INET;
115 return 0;
116 } else {
117 return -EAFNOSUPPORT;
118 }
e4c7ed41
CW		 119}
120
121static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 122 const union vxlan_addr *ip)
e4c7ed41 123{
f0ef3126 124 if (ip->sa.sa_family == AF_INET6)
930345ea 125 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 126 else
930345ea 127 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 128}
129
130#else /* !CONFIG_IPV6 */
131
132static inline
133bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
134{
f0ef3126 135 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 136}
137
e4c7ed41
CW		 138static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
139{
f0ef3126
JB		 140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
141 return -EAFNOSUPPORT;
142 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 143 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 144 ip->sa.sa_family = AF_INET;
145 return 0;
146 } else {
147 return -EAFNOSUPPORT;
148 }
e4c7ed41
CW		 149}
150
151static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 152 const union vxlan_addr *ip)
e4c7ed41 153{
930345ea 154 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 155}
156#endif
157
553675fb 158/* Virtual Network hash table head */
54bfd872 159static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 160{
54bfd872 161 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 162}
163
164/* Socket hash table head */
165static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 166{
167 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
168
553675fb 169 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
170}
171
3e61aa8f
SH		 172/* First remote destination for a forwarding entry.
173 * Guaranteed to be non-NULL because remotes are never deleted.
174 */
5ca5461c 175static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 176{
5ca5461c 177 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
178}
179
180static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
181{
182 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 183}
184
ac5132d1
TG		 185/* Find VXLAN socket based on network namespace, address family and UDP port
186 * and enabled unshareable flags.
187 */
188static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
189 __be16 port, u32 flags)
553675fb 190{
191 struct vxlan_sock *vs;
af33c1ad
TH		 192
193 flags &= VXLAN_F_RCV_FLAGS;
553675fb 194
195 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 196 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 197 vxlan_get_sk_family(vs) == family &&
af33c1ad 198 vs->flags == flags)
553675fb 199 return vs;
200 }
201 return NULL;
d342894c 202}
203
49f810f0
MS		 204static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
205 __be32 vni)
d342894c 206{
69e76661 207 struct vxlan_dev_node *node;
d342894c 208
b9167b2e
JB		 209 /* For flow based devices, map all packets to VNI 0 */
210 if (vs->flags & VXLAN_F_COLLECT_METADATA)
211 vni = 0;
212
69e76661
JB		 213 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
214 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS		 215 continue;
216
217 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 218 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS		 219
220 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
221 cfg->remote_ifindex != ifindex)
222 continue;
223 }
224
69e76661 225 return node->vxlan;
d342894c 226 }
227
228 return NULL;
229}
230
5cfccc5a 231/* Look up VNI in a per net namespace table */
49f810f0
MS		 232static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
233 __be32 vni, sa_family_t family,
234 __be16 port, u32 flags)
5cfccc5a
PS		 235{
236 struct vxlan_sock *vs;
237
ac5132d1 238 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 239 if (!vs)
240 return NULL;
241
49f810f0 242 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS		 243}
244
d342894c 245/* Fill in neighbour message in skbuff. */
246static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 247 const struct vxlan_fdb *fdb,
248 u32 portid, u32 seq, int type, unsigned int flags,
249 const struct vxlan_rdst *rdst)
d342894c 250{
251 unsigned long now = jiffies;
252 struct nda_cacheinfo ci;
253 struct nlmsghdr *nlh;
254 struct ndmsg *ndm;
e4f67add 255 bool send_ip, send_eth;
d342894c 256
257 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
258 if (nlh == NULL)
259 return -EMSGSIZE;
260
261 ndm = nlmsg_data(nlh);
262 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 263
264 send_eth = send_ip = true;
265
266 if (type == RTM_GETNEIGH) {
e4c7ed41 267 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 268 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 269 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 270 } else
271 ndm->ndm_family = AF_BRIDGE;
d342894c 272 ndm->ndm_state = fdb->state;
273 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 274 ndm->ndm_flags = fdb->flags;
545469f7 275 ndm->ndm_type = RTN_UNICAST;
d342894c 276
193523bf 277 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 278 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 279 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 280 goto nla_put_failure;
281
e4f67add 282 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 283 goto nla_put_failure;
284
e4c7ed41 285 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 286 goto nla_put_failure;
287
0dfbdf41 288 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 289 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
290 goto nla_put_failure;
c7995c43 291 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 292 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 293 goto nla_put_failure;
dc5321d7 294 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 295 nla_put_u32(skb, NDA_SRC_VNI,
296 be32_to_cpu(fdb->vni)))
297 goto nla_put_failure;
6681712d
DS		 298 if (rdst->remote_ifindex &&
299 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 300 goto nla_put_failure;
301
302 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
303 ci.ndm_confirmed = 0;
304 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
305 ci.ndm_refcnt = 0;
306
307 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
308 goto nla_put_failure;
309
053c095a
JB		 310 nlmsg_end(skb, nlh);
311 return 0;
d342894c 312
313nla_put_failure:
314 nlmsg_cancel(skb, nlh);
315 return -EMSGSIZE;
316}
317
318static inline size_t vxlan_nlmsg_size(void)
319{
320 return NLMSG_ALIGN(sizeof(struct ndmsg))
321 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 322 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 323 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 324 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
325 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 326 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 327 + nla_total_size(sizeof(struct nda_cacheinfo));
328}
329
9a997353
PM		 330static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
331 struct vxlan_rdst *rd, int type)
d342894c 332{
333 struct net *net = dev_net(vxlan->dev);
334 struct sk_buff *skb;
335 int err = -ENOBUFS;
336
337 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
338 if (skb == NULL)
339 goto errout;
340
9e4b93f9 341 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 342 if (err < 0) {
343 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
344 WARN_ON(err == -EMSGSIZE);
345 kfree_skb(skb);
346 goto errout;
347 }
348
349 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
350 return;
351errout:
352 if (err < 0)
353 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
354}
355
9a997353
PM		 356static void vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
357 struct vxlan_fdb *fdb,
358 struct vxlan_rdst *rd,
359 bool adding)
360{
361 struct switchdev_notifier_vxlan_fdb_info info;
362 enum switchdev_notifier_type notifier_type;
363
364 if (WARN_ON(!rd))
365 return;
366
367 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
368 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
369
370 info = (struct switchdev_notifier_vxlan_fdb_info){
371 .remote_ip = rd->remote_ip,
372 .remote_port = rd->remote_port,
373 .remote_vni = rd->remote_vni,
374 .remote_ifindex = rd->remote_ifindex,
375 .vni = fdb->vni,
376 };
377 memcpy(info.eth_addr, fdb->eth_addr, ETH_ALEN);
378
379 call_switchdev_notifiers(notifier_type, vxlan->dev,
380 &info.info);
381}
382
383static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
384 struct vxlan_rdst *rd, int type)
385{
386 switch (type) {
387 case RTM_NEWNEIGH:
388 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, true);
389 break;
390 case RTM_DELNEIGH:
391 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd, false);
392 break;
393 }
394
395 __vxlan_fdb_notify(vxlan, fdb, rd, type);
396}
397
e4c7ed41 398static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 399{
400 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 401 struct vxlan_fdb f = {
402 .state = NUD_STALE,
403 };
404 struct vxlan_rdst remote = {
e4c7ed41 405 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 406 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 407 };
3e61aa8f 408
9e4b93f9 409 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 410}
411
412static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
413{
bb3fd687
SH		 414 struct vxlan_fdb f = {
415 .state = NUD_STALE,
416 };
9e4b93f9 417 struct vxlan_rdst remote = { };
e4f67add 418
e4f67add
DS		 419 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
420
9e4b93f9 421 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 422}
423
d342894c 424/* Hash Ethernet address */
425static u32 eth_hash(const unsigned char *addr)
426{
427 u64 value = get_unaligned((u64 *)addr);
428
429 /* only want 6 bytes */
430#ifdef __BIG_ENDIAN
d342894c 431 value >>= 16;
321fb991 432#else
433 value <<= 16;
d342894c 434#endif
435 return hash_64(value, FDB_HASH_BITS);
436}
437
3ad7a4b1
RP		 438static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
439{
440 /* use 1 byte of OUI and 3 bytes of NIC */
441 u32 key = get_unaligned((u32 *)(addr + 2));
442
443 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
444}
445
d342894c 446/* Hash chain to use given mac address */
447static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 448 const u8 *mac, __be32 vni)
d342894c 449{
dc5321d7 450 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
3ad7a4b1
RP		 451 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
452 else
453 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 454}
455
456/* Look up Ethernet address in forwarding table */
014be2c8 457static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 458 const u8 *mac, __be32 vni)
d342894c 459{
3ad7a4b1 460 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 461 struct vxlan_fdb *f;
d342894c 462
b67bfe0d 463 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 464 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 465 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 466 if (vni == f->vni)
467 return f;
468 } else {
469 return f;
470 }
471 }
d342894c 472 }
473
474 return NULL;
475}
476
014be2c8 477static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 478 const u8 *mac, __be32 vni)
014be2c8
SS		 479{
480 struct vxlan_fdb *f;
481
3ad7a4b1 482 f = __vxlan_find_mac(vxlan, mac, vni);
016f3d18 483 if (f && f->used != jiffies)
014be2c8
SS		 484 f->used = jiffies;
485
486 return f;
487}
488
a5e7c10a
MR		 489/* caller should hold vxlan->hash_lock */
490static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 491 union vxlan_addr *ip, __be16 port,
54bfd872 492 __be32 vni, __u32 ifindex)
6681712d 493{
3e61aa8f 494 struct vxlan_rdst *rd;
6681712d 495
3e61aa8f 496 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 497 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 498 rd->remote_port == port &&
499 rd->remote_vni == vni &&
500 rd->remote_ifindex == ifindex)
a5e7c10a 501 return rd;
6681712d 502 }
3e61aa8f 503
a5e7c10a
MR		 504 return NULL;
505}
506
906dc186
TR		 507/* Replace destination of unicast mac */
508static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 509 union vxlan_addr *ip, __be16 port, __be32 vni,
510 __u32 ifindex)
906dc186
TR		 511{
512 struct vxlan_rdst *rd;
513
514 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
515 if (rd)
516 return 0;
517
518 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
519 if (!rd)
520 return 0;
0c1d70af
PA		 521
522 dst_cache_reset(&rd->dst_cache);
e4c7ed41 523 rd->remote_ip = *ip;
906dc186
TR		 524 rd->remote_port = port;
525 rd->remote_vni = vni;
526 rd->remote_ifindex = ifindex;
527 return 1;
528}
529
a5e7c10a
MR		 530/* Add/update destinations for multicast */
531static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 532 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 533 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 534{
535 struct vxlan_rdst *rd;
536
537 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
538 if (rd)
539 return 0;
540
6681712d
DS		 541 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
542 if (rd == NULL)
543 return -ENOBUFS;
0c1d70af
PA		 544
545 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
546 kfree(rd);
547 return -ENOBUFS;
548 }
549
e4c7ed41 550 rd->remote_ip = *ip;
6681712d
DS		 551 rd->remote_port = port;
552 rd->remote_vni = vni;
553 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 554
555 list_add_tail_rcu(&rd->list, &f->remotes);
556
9e4b93f9 557 *rdp = rd;
6681712d
DS		 558 return 1;
559}
560
dfd8645e
TH		 561static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
562 unsigned int off,
563 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 564 __be32 vni_field,
565 struct gro_remcsum *grc,
0ace2ca8 566 bool nopartial)
dfd8645e 567{
b7fe10e5 568 size_t start, offset;
dfd8645e
TH		 569
570 if (skb->remcsum_offload)
b7fe10e5 571 return vh;
dfd8645e
TH		 572
573 if (!NAPI_GRO_CB(skb)->csum_valid)
574 return NULL;
575
54bfd872
JB		 576 start = vxlan_rco_start(vni_field);
577 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 578
b7fe10e5
TH		 579 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
580 start, offset, grc, nopartial);
dfd8645e
TH		 581
582 skb->remcsum_offload = 1;
583
584 return vh;
585}
586
d4546c25
DM		 587static struct sk_buff *vxlan_gro_receive(struct sock *sk,
588 struct list_head *head,
589 struct sk_buff *skb)
dc01e7d3 590{
d4546c25
DM		 591 struct sk_buff *pp = NULL;
592 struct sk_buff *p;
dc01e7d3 593 struct vxlanhdr *vh, *vh2;
9b174d88 594 unsigned int hlen, off_vx;
dc01e7d3 595 int flush = 1;
5602c48c 596 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 597 __be32 flags;
26c4f7da
TH		 598 struct gro_remcsum grc;
599
600 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 601
602 off_vx = skb_gro_offset(skb);
603 hlen = off_vx + sizeof(*vh);
604 vh = skb_gro_header_fast(skb, off_vx);
605 if (skb_gro_header_hard(skb, hlen)) {
606 vh = skb_gro_header_slow(skb, hlen, off_vx);
607 if (unlikely(!vh))
608 goto out;
609 }
dc01e7d3 610
dfd8645e
TH		 611 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
612
54bfd872 613 flags = vh->vx_flags;
dfd8645e
TH		 614
615 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
616 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 617 vh->vx_vni, &grc,
0ace2ca8
TH		 618 !!(vs->flags &
619 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 620
621 if (!vh)
622 goto out;
623 }
624
b7fe10e5
TH		 625 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
626
d4546c25 627 list_for_each_entry(p, head, list) {
dc01e7d3
OG		 628 if (!NAPI_GRO_CB(p)->same_flow)
629 continue;
630
631 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 632 if (vh->vx_flags != vh2->vx_flags ||
633 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 634 NAPI_GRO_CB(p)->same_flow = 0;
635 continue;
636 }
dc01e7d3
OG		 637 }
638
fcd91dd4 639 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 640 flush = 0;
dc01e7d3 641
dc01e7d3 642out:
603d4cf8 643 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 644
645 return pp;
646}
647
5602c48c 648static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 649{
229740c6
JR		 650 /* Sets 'skb->inner_mac_header' since we are always called with
651 * 'skb->encapsulation' set.
652 */
9b174d88 653 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 654}
655
25e20e73
RP		 656static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
657 const u8 *mac, __u16 state,
658 __be32 src_vni, __u8 ndm_flags)
659{
660 struct vxlan_fdb *f;
661
662 f = kmalloc(sizeof(*f), GFP_ATOMIC);
663 if (!f)
664 return NULL;
665 f->state = state;
666 f->flags = ndm_flags;
667 f->updated = f->used = jiffies;
668 f->vni = src_vni;
669 INIT_LIST_HEAD(&f->remotes);
670 memcpy(f->eth_addr, mac, ETH_ALEN);
671
672 return f;
673}
674
d342894c 675static int vxlan_fdb_create(struct vxlan_dev *vxlan,
25e20e73
RP		 676 const u8 *mac, union vxlan_addr *ip,
677 __u16 state, __be16 port, __be32 src_vni,
678 __be32 vni, __u32 ifindex, __u8 ndm_flags,
679 struct vxlan_fdb **fdb)
680{
681 struct vxlan_rdst *rd = NULL;
682 struct vxlan_fdb *f;
683 int rc;
684
685 if (vxlan->cfg.addrmax &&
686 vxlan->addrcnt >= vxlan->cfg.addrmax)
687 return -ENOSPC;
688
689 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
690 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
691 if (!f)
692 return -ENOMEM;
693
694 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
695 if (rc < 0) {
696 kfree(f);
697 return rc;
698 }
699
700 ++vxlan->addrcnt;
701 hlist_add_head_rcu(&f->hlist,
702 vxlan_fdb_head(vxlan, mac, src_vni));
703
704 *fdb = f;
705
706 return 0;
707}
708
709/* Add new entry to forwarding table -- assumes lock held */
710static int vxlan_fdb_update(struct vxlan_dev *vxlan,
e4c7ed41 711 const u8 *mac, union vxlan_addr *ip,
6681712d 712 __u16 state, __u16 flags,
3ad7a4b1
RP		 713 __be16 port, __be32 src_vni, __be32 vni,
714 __u32 ifindex, __u8 ndm_flags)
d342894c 715{
0813e957 716 __u8 fdb_flags = (ndm_flags & ~NTF_USE);
9e4b93f9 717 struct vxlan_rdst *rd = NULL;
d342894c 718 struct vxlan_fdb *f;
719 int notify = 0;
17b46365 720 int rc;
d342894c 721
3ad7a4b1 722 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 723 if (f) {
724 if (flags & NLM_F_EXCL) {
725 netdev_dbg(vxlan->dev,
726 "lost race to create %pM\n", mac);
727 return -EEXIST;
728 }
729 if (f->state != state) {
730 f->state = state;
731 f->updated = jiffies;
732 notify = 1;
733 }
0813e957
RP		 734 if (f->flags != fdb_flags) {
735 f->flags = fdb_flags;
ae884082
DS		 736 f->updated = jiffies;
737 notify = 1;
738 }
906dc186
TR		 739 if ((flags & NLM_F_REPLACE)) {
740 /* Only change unicasts */
741 if (!(is_multicast_ether_addr(f->eth_addr) ||
742 is_zero_ether_addr(f->eth_addr))) {
60840429 743 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 744 ifindex);
906dc186
TR		 745 } else
746 return -EOPNOTSUPP;
747 }
6681712d 748 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 749 (is_multicast_ether_addr(f->eth_addr) ||
750 is_zero_ether_addr(f->eth_addr))) {
17b46365 751 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 752
753 if (rc < 0)
754 return rc;
755 notify |= rc;
756 }
0813e957
RP		 757
758 if (ndm_flags & NTF_USE)
759 f->used = jiffies;
d342894c 760 } else {
761 if (!(flags & NLM_F_CREATE))
762 return -ENOENT;
763
906dc186
TR		 764 /* Disallow replace to add a multicast entry */
765 if ((flags & NLM_F_REPLACE) &&
766 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
767 return -EOPNOTSUPP;
768
e4c7ed41 769 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
25e20e73 770 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
0813e957 771 vni, ifindex, fdb_flags, &f);
25e20e73 772 if (rc < 0)
17b46365 773 return rc;
25e20e73 774 notify = 1;
d342894c 775 }
776
9e4b93f9
ND		 777 if (notify) {
778 if (rd == NULL)
779 rd = first_remote_rtnl(f);
780 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
781 }
d342894c 782
783 return 0;
784}
785
6706c82e 786static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 787{
788 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 789 struct vxlan_rdst *rd, *nd;
6681712d 790
0c1d70af
PA		 791 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
792 dst_cache_destroy(&rd->dst_cache);
6681712d 793 kfree(rd);
0c1d70af 794 }
6681712d
DS		 795 kfree(f);
796}
797
4c2438ba
RP		 798static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
799 bool do_notify)
d342894c 800{
801 netdev_dbg(vxlan->dev,
802 "delete %pM\n", f->eth_addr);
803
804 --vxlan->addrcnt;
4c2438ba
RP		 805 if (do_notify)
806 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 807
808 hlist_del_rcu(&f->hlist);
6681712d 809 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 810}
811
35cf2845
LR		 812static void vxlan_dst_free(struct rcu_head *head)
813{
814 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
815
816 dst_cache_destroy(&rd->dst_cache);
817 kfree(rd);
818}
819
820static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
821 struct vxlan_rdst *rd)
822{
823 list_del_rcu(&rd->list);
824 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
825 call_rcu(&rd->rcu, vxlan_dst_free);
826}
827
f0b074be 828static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 829 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
830 __be32 *vni, u32 *ifindex)
d342894c 831{
6681712d 832 struct net *net = dev_net(vxlan->dev);
e4c7ed41 833 int err;
d342894c 834
f0b074be 835 if (tb[NDA_DST]) {
e4c7ed41
CW		 836 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
837 if (err)
838 return err;
f0b074be 839 } else {
e4c7ed41
CW		 840 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
841 if (remote->sa.sa_family == AF_INET) {
842 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
843 ip->sa.sa_family = AF_INET;
844#if IS_ENABLED(CONFIG_IPV6)
845 } else {
846 ip->sin6.sin6_addr = in6addr_any;
847 ip->sa.sa_family = AF_INET6;
848#endif
849 }
f0b074be 850 }
d342894c 851
6681712d 852 if (tb[NDA_PORT]) {
73cf3317 853 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 854 return -EINVAL;
f0b074be
MR		 855 *port = nla_get_be16(tb[NDA_PORT]);
856 } else {
0dfbdf41 857 *port = vxlan->cfg.dst_port;
f0b074be 858 }
6681712d
DS		 859
860 if (tb[NDA_VNI]) {
861 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
862 return -EINVAL;
54bfd872 863 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 864 } else {
865 *vni = vxlan->default_dst.remote_vni;
866 }
6681712d 867
3ad7a4b1
RP		 868 if (tb[NDA_SRC_VNI]) {
869 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
870 return -EINVAL;
871 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
872 } else {
873 *src_vni = vxlan->default_dst.remote_vni;
874 }
875
6681712d 876 if (tb[NDA_IFINDEX]) {
5abb0029 877 struct net_device *tdev;
6681712d
DS		 878
879 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
880 return -EINVAL;
f0b074be 881 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 882 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 883 if (!tdev)
6681712d 884 return -EADDRNOTAVAIL;
f0b074be
MR		 885 } else {
886 *ifindex = 0;
887 }
888
889 return 0;
890}
891
892/* Add static entry (via netlink) */
893static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
894 struct net_device *dev,
f6f6424b 895 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 896{
897 struct vxlan_dev *vxlan = netdev_priv(dev);
898 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 899 union vxlan_addr ip;
f0b074be 900 __be16 port;
3ad7a4b1 901 __be32 src_vni, vni;
54bfd872 902 u32 ifindex;
f0b074be
MR		 903 int err;
904
905 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
906 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
907 ndm->ndm_state);
908 return -EINVAL;
909 }
910
911 if (tb[NDA_DST] == NULL)
912 return -EINVAL;
913
3ad7a4b1 914 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 915 if (err)
916 return err;
6681712d 917
5933a7bb
MR		 918 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
919 return -EAFNOSUPPORT;
920
d342894c 921 spin_lock_bh(&vxlan->hash_lock);
25e20e73 922 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 923 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 924 spin_unlock_bh(&vxlan->hash_lock);
925
926 return err;
927}
928
3ad7a4b1
RP		 929static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
930 const unsigned char *addr, union vxlan_addr ip,
fc39c38b
XL		 931 __be16 port, __be32 src_vni, __be32 vni,
932 u32 ifindex, u16 vid)
d342894c 933{
d342894c 934 struct vxlan_fdb *f;
bc7892ba 935 struct vxlan_rdst *rd = NULL;
3ad7a4b1 936 int err = -ENOENT;
bc7892ba 937
3ad7a4b1 938 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 939 if (!f)
3ad7a4b1 940 return err;
bc7892ba 941
e4c7ed41
CW		 942 if (!vxlan_addr_any(&ip)) {
943 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 944 if (!rd)
945 goto out;
946 }
947
bc7892ba
MR		 948 /* remove a destination if it's not the only one on the list,
949 * otherwise destroy the fdb entry
950 */
951 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 952 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 953 goto out;
d342894c 954 }
bc7892ba 955
4c2438ba 956 vxlan_fdb_destroy(vxlan, f, true);
bc7892ba
MR		 957
958out:
3ad7a4b1
RP		 959 return 0;
960}
961
962/* Delete entry (via netlink) */
963static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
964 struct net_device *dev,
965 const unsigned char *addr, u16 vid)
966{
967 struct vxlan_dev *vxlan = netdev_priv(dev);
968 union vxlan_addr ip;
969 __be32 src_vni, vni;
970 __be16 port;
971 u32 ifindex;
972 int err;
973
974 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
975 if (err)
976 return err;
977
978 spin_lock_bh(&vxlan->hash_lock);
979 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
980 vid);
d342894c 981 spin_unlock_bh(&vxlan->hash_lock);
982
983 return err;
984}
985
986/* Dump forwarding table */
987static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 988 struct net_device *dev,
d297653d 989 struct net_device *filter_dev, int *idx)
d342894c 990{
991 struct vxlan_dev *vxlan = netdev_priv(dev);
992 unsigned int h;
d297653d 993 int err = 0;
d342894c 994
995 for (h = 0; h < FDB_HASH_SIZE; ++h) {
996 struct vxlan_fdb *f;
d342894c 997
b67bfe0d 998 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 999 struct vxlan_rdst *rd;
6681712d 1000
3e61aa8f 1001 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 1002 if (*idx < cb->args[2])
07a51cd3
AW		 1003 goto skip;
1004
6681712d
DS		 1005 err = vxlan_fdb_info(skb, vxlan, f,
1006 NETLINK_CB(cb->skb).portid,
1007 cb->nlh->nlmsg_seq,
1008 RTM_NEWNEIGH,
1009 NLM_F_MULTI, rd);
d297653d 1010 if (err < 0)
3e61aa8f 1011 goto out;
3e61aa8f 1012skip:
d297653d 1013 *idx += 1;
07a51cd3 1014 }
d342894c 1015 }
1016 }
3e61aa8f 1017out:
d297653d 1018 return err;
d342894c 1019}
1020
1021/* Watch incoming packets to learn mapping between Ethernet address
1022 * and Tunnel endpoint.
c4b49512 1023 * Return true if packet is bogus and should be dropped.
d342894c 1024 */
26a41ae6 1025static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1026 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1027 u32 src_ifindex, __be32 vni)
d342894c 1028{
1029 struct vxlan_dev *vxlan = netdev_priv(dev);
1030 struct vxlan_fdb *f;
87613de9
MS		 1031 u32 ifindex = 0;
1032
1033#if IS_ENABLED(CONFIG_IPV6)
1034 if (src_ip->sa.sa_family == AF_INET6 &&
1035 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1036 ifindex = src_ifindex;
1037#endif
d342894c 1038
3ad7a4b1 1039 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1040 if (likely(f)) {
5ca5461c 1041 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1042
87613de9
MS		 1043 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1044 rdst->remote_ifindex == ifindex))
26a41ae6 1045 return false;
1046
1047 /* Don't migrate static entries, drop packets */
e0090a9e 1048 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1049 return true;
d342894c 1050
1051 if (net_ratelimit())
1052 netdev_info(dev,
e4c7ed41 1053 "%pM migrated from %pIS to %pIS\n",
a4870f79 1054 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1055
e4c7ed41 1056 rdst->remote_ip = *src_ip;
d342894c 1057 f->updated = jiffies;
9e4b93f9 1058 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1059 } else {
1060 /* learned new entry */
1061 spin_lock(&vxlan->hash_lock);
3bf74b1a 1062
1063 /* close off race between vxlan_flush and incoming packets */
1064 if (netif_running(dev))
25e20e73 1065 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1066 NUD_REACHABLE,
1067 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1068 vxlan->cfg.dst_port,
3ad7a4b1 1069 vni,
3bf74b1a 1070 vxlan->default_dst.remote_vni,
87613de9 1071 ifindex, NTF_SELF);
d342894c 1072 spin_unlock(&vxlan->hash_lock);
1073 }
26a41ae6 1074
1075 return false;
d342894c 1076}
1077
d342894c 1078/* See if multicast group is already in use by other ID */
95ab0991 1079static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1080{
553675fb 1081 struct vxlan_dev *vxlan;
c6fcc4fc 1082 struct vxlan_sock *sock4;
4053ab1b
AB		 1083#if IS_ENABLED(CONFIG_IPV6)
1084 struct vxlan_sock *sock6;
1085#endif
b1be00a6 1086 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1087
c6fcc4fc 1088 sock4 = rtnl_dereference(dev->vn4_sock);
1089
95ab0991
G		 1090 /* The vxlan_sock is only used by dev, leaving group has
1091 * no effect on other vxlan devices.
1092 */
66af846f 1093 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1094 return false;
b1be00a6 1095#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1096 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1097 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1098 return false;
1099#endif
95ab0991 1100
553675fb 1101 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1102 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1103 continue;
d342894c 1104
c6fcc4fc 1105 if (family == AF_INET &&
1106 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1107 continue;
b1be00a6 1108#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1109 if (family == AF_INET6 &&
1110 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1111 continue;
1112#endif
95ab0991
G		 1113
1114 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1115 &dev->default_dst.remote_ip))
1116 continue;
1117
1118 if (vxlan->default_dst.remote_ifindex !=
1119 dev->default_dst.remote_ifindex)
1120 continue;
1121
1122 return true;
553675fb 1123 }
d342894c 1124
1125 return false;
1126}
1127
544a773a 1128static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1129{
b1be00a6 1130 struct vxlan_net *vn;
012a5729 1131
b1be00a6 1132 if (!vs)
544a773a 1133 return false;
66af846f 1134 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1135 return false;
d342894c 1136
b1be00a6 1137 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1138 spin_lock(&vn->sock_lock);
7c47cedf 1139 hlist_del_rcu(&vs->hlist);
e7b3db5e 1140 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1141 (vs->flags & VXLAN_F_GPE) ?
1142 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1143 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1144 spin_unlock(&vn->sock_lock);
1145
544a773a 1146 return true;
d342894c 1147}
1148
b1be00a6
JB		 1149static void vxlan_sock_release(struct vxlan_dev *vxlan)
1150{
c6fcc4fc 1151 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1152#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1153 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1154
57d88182 1155 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1156#endif
1157
57d88182 1158 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1159 synchronize_net();
1160
a53cb29b
MB		 1161 vxlan_vs_del_dev(vxlan);
1162
c6fcc4fc 1163 if (__vxlan_sock_release_prep(sock4)) {
1164 udp_tunnel_sock_release(sock4->sock);
1165 kfree(sock4);
544a773a
HFS		 1166 }
1167
1168#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1169 if (__vxlan_sock_release_prep(sock6)) {
1170 udp_tunnel_sock_release(sock6->sock);
1171 kfree(sock6);
544a773a 1172 }
b1be00a6
JB		 1173#endif
1174}
1175
56ef9c90 1176/* Update multicast group membership when first VNI on
c4b49512 1177 * multicast address is brought up
7c47cedf 1178 */
56ef9c90 1179static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1180{
b1be00a6 1181 struct sock *sk;
e4c7ed41
CW		 1182 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1183 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1184 int ret = -EINVAL;
d342894c 1185
e4c7ed41 1186 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1187 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1188 struct ip_mreqn mreq = {
1189 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1190 .imr_ifindex = ifindex,
1191 };
1192
c6fcc4fc 1193 sk = sock4->sock->sk;
b1be00a6 1194 lock_sock(sk);
56ef9c90 1195 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1196 release_sock(sk);
e4c7ed41
CW		 1197#if IS_ENABLED(CONFIG_IPV6)
1198 } else {
c6fcc4fc 1199 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1200
1201 sk = sock6->sock->sk;
b1be00a6 1202 lock_sock(sk);
56ef9c90
MRL		 1203 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1204 &ip->sin6.sin6_addr);
b1be00a6 1205 release_sock(sk);
e4c7ed41
CW		 1206#endif
1207 }
3fc2de2f 1208
56ef9c90 1209 return ret;
3fc2de2f 1210}
1211
1212/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1213static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1214{
b1be00a6 1215 struct sock *sk;
e4c7ed41
CW		 1216 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1217 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1218 int ret = -EINVAL;
3fc2de2f 1219
e4c7ed41 1220 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1221 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1222 struct ip_mreqn mreq = {
1223 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1224 .imr_ifindex = ifindex,
1225 };
1226
c6fcc4fc 1227 sk = sock4->sock->sk;
b1be00a6 1228 lock_sock(sk);
56ef9c90 1229 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1230 release_sock(sk);
e4c7ed41
CW		 1231#if IS_ENABLED(CONFIG_IPV6)
1232 } else {
c6fcc4fc 1233 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1234
1235 sk = sock6->sock->sk;
b1be00a6 1236 lock_sock(sk);
56ef9c90
MRL		 1237 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1238 &ip->sin6.sin6_addr);
b1be00a6 1239 release_sock(sk);
e4c7ed41
CW		 1240#endif
1241 }
d342894c 1242
56ef9c90 1243 return ret;
d342894c 1244}
1245
f14ecebb
JB		 1246static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1247 struct sk_buff *skb, u32 vxflags)
dfd8645e 1248{
7d34fa75 1249 size_t start, offset;
dfd8645e 1250
f14ecebb
JB		 1251 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1252 goto out;
b7fe10e5 1253
f14ecebb
JB		 1254 start = vxlan_rco_start(unparsed->vx_vni);
1255 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1256
7d34fa75 1257 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1258 return false;
dfd8645e 1259
be5cfeab
JB		 1260 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1261 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1262out:
1263 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1264 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1265 return true;
dfd8645e
TH		 1266}
1267
f14ecebb 1268static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1269 struct sk_buff *skb, u32 vxflags,
10a5af23 1270 struct vxlan_metadata *md)
3288af08 1271{
f14ecebb 1272 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1273 struct metadata_dst *tun_dst;
f14ecebb
JB		 1274
1275 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1276 goto out;
3288af08 1277
3288af08
JB		 1278 md->gbp = ntohs(gbp->policy_id);
1279
10a5af23 1280 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1281 if (tun_dst) {
3288af08 1282 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1283 tun_dst->u.tun_info.options_len = sizeof(*md);
1284 }
3288af08
JB		 1285 if (gbp->dont_learn)
1286 md->gbp |= VXLAN_GBP_DONT_LEARN;
1287
1288 if (gbp->policy_applied)
1289 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1290
64f87d36
JB		 1291 /* In flow-based mode, GBP is carried in dst_metadata */
1292 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1293 skb->mark = md->gbp;
f14ecebb
JB		 1294out:
1295 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1296}
1297
e1e5314d 1298static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1299 __be16 *protocol,
e1e5314d
JB		 1300 struct sk_buff *skb, u32 vxflags)
1301{
1302 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1303
1304 /* Need to have Next Protocol set for interfaces in GPE mode. */
1305 if (!gpe->np_applied)
1306 return false;
1307 /* "The initial version is 0. If a receiver does not support the
1308 * version indicated it MUST drop the packet.
1309 */
1310 if (gpe->version != 0)
1311 return false;
1312 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1313 * processing MUST occur." However, we don't implement OAM
1314 * processing, thus drop the packet.
1315 */
1316 if (gpe->oam_flag)
1317 return false;
1318
fa20e0e3
JB		 1319 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1320 if (!*protocol)
e1e5314d 1321 return false;
e1e5314d
JB		 1322
1323 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1324 return true;
1325}
1326
1ab016e2
JB		 1327static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1328 struct vxlan_sock *vs,
3ad7a4b1 1329 struct sk_buff *skb, __be32 vni)
614732ea 1330{
614732ea 1331 union vxlan_addr saddr;
87613de9 1332 u32 ifindex = skb->dev->ifindex;
614732ea 1333
614732ea 1334 skb_reset_mac_header(skb);
614732ea
TG		 1335 skb->protocol = eth_type_trans(skb, vxlan->dev);
1336 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1337
1338 /* Ignore packet loops (and multicast echo) */
1339 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1340 return false;
614732ea 1341
760c6805 1342 /* Get address from the outer IP header */
ce212d0f 1343 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1344 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1345 saddr.sa.sa_family = AF_INET;
1346#if IS_ENABLED(CONFIG_IPV6)
1347 } else {
1ab016e2 1348 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1349 saddr.sa.sa_family = AF_INET6;
1350#endif
1351 }
1352
dc5321d7 1353 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1354 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1355 return false;
1356
1357 return true;
1358}
1359
760c6805
JB		 1360static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1361 struct sk_buff *skb)
1362{
1363 int err = 0;
1364
1365 if (vxlan_get_sk_family(vs) == AF_INET)
1366 err = IP_ECN_decapsulate(oiph, skb);
1367#if IS_ENABLED(CONFIG_IPV6)
1368 else
1369 err = IP6_ECN_decapsulate(oiph, skb);
1370#endif
1371
1372 if (unlikely(err) && log_ecn_error) {
1373 if (vxlan_get_sk_family(vs) == AF_INET)
1374 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1375 &((struct iphdr *)oiph)->saddr,
1376 ((struct iphdr *)oiph)->tos);
1377 else
1378 net_info_ratelimited("non-ECT from %pI6\n",
1379 &((struct ipv6hdr *)oiph)->saddr);
1380 }
1381 return err <= 1;
1382}
1383
d342894c 1384/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1385static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1386{
f2d1968e 1387 struct pcpu_sw_netstats *stats;
c9e78efb 1388 struct vxlan_dev *vxlan;
5cfccc5a 1389 struct vxlan_sock *vs;
f14ecebb 1390 struct vxlanhdr unparsed;
ee122c79
TG		 1391 struct vxlan_metadata _md;
1392 struct vxlan_metadata *md = &_md;
61618eea 1393 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1394 bool raw_proto = false;
f2d1968e 1395 void *oiph;
3ad7a4b1 1396 __be32 vni = 0;
d342894c 1397
e1e5314d 1398 /* Need UDP and VXLAN header to be present */
7ce04758 1399 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1400 goto drop;
d342894c 1401
f14ecebb 1402 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1403 /* VNI flag always required to be set */
1404 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1405 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1406 ntohl(vxlan_hdr(skb)->vx_flags),
1407 ntohl(vxlan_hdr(skb)->vx_vni));
1408 /* Return non vxlan pkt */
e5aed006 1409 goto drop;
d342894c 1410 }
288b01c8
JB		 1411 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1412 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1413
559835ea 1414 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1415 if (!vs)
d342894c 1416 goto drop;
d342894c 1417
3ad7a4b1
RP		 1418 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1419
49f810f0 1420 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1421 if (!vxlan)
1422 goto drop;
1423
e1e5314d
JB		 1424 /* For backwards compatibility, only allow reserved fields to be
1425 * used by VXLAN extensions if explicitly requested.
1426 */
1427 if (vs->flags & VXLAN_F_GPE) {
1428 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1429 goto drop;
1430 raw_proto = true;
1431 }
1432
1433 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1434 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1435 goto drop;
c9e78efb 1436
ee122c79 1437 if (vxlan_collect_metadata(vs)) {
10a5af23 1438 struct metadata_dst *tun_dst;
07dabf20 1439
c29a70d2 1440 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1441 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1442
ee122c79
TG		 1443 if (!tun_dst)
1444 goto drop;
1445
0f1b7354 1446 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1447
1448 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1449 } else {
1450 memset(md, 0, sizeof(*md));
1451 }
1452
f14ecebb
JB		 1453 if (vs->flags & VXLAN_F_REMCSUM_RX)
1454 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1455 goto drop;
1456 if (vs->flags & VXLAN_F_GBP)
10a5af23 1457 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1458 /* Note that GBP and GPE can never be active together. This is
1459 * ensured in vxlan_dev_configure.
1460 */
3511494c 1461
f14ecebb 1462 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1463 /* If there are any unprocessed flags remaining treat
1464 * this as a malformed packet. This behavior diverges from
1465 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1466 * in reserved fields are to be ignored. The approach here
c4b49512 1467 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1468 * is more robust and provides a little more security in
1469 * adding extensions to VXLAN.
1470 */
288b01c8 1471 goto drop;
3bf39475
TH		 1472 }
1473
e1e5314d 1474 if (!raw_proto) {
3ad7a4b1 1475 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1476 goto drop;
1477 } else {
8be0cfa4 1478 skb_reset_mac_header(skb);
e1e5314d
JB		 1479 skb->dev = vxlan->dev;
1480 skb->pkt_type = PACKET_HOST;
1481 }
f2d1968e 1482
f2d1968e
JB		 1483 oiph = skb_network_header(skb);
1484 skb_reset_network_header(skb);
1485
1486 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1487 ++vxlan->dev->stats.rx_frame_errors;
1488 ++vxlan->dev->stats.rx_errors;
1489 goto drop;
1490 }
1491
1492 stats = this_cpu_ptr(vxlan->dev->tstats);
1493 u64_stats_update_begin(&stats->syncp);
1494 stats->rx_packets++;
1495 stats->rx_bytes += skb->len;
1496 u64_stats_update_end(&stats->syncp);
1497
1498 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1499 return 0;
1500
1501drop:
288b01c8
JB		 1502 /* Consume bad packet */
1503 kfree_skb(skb);
1504 return 0;
5cfccc5a
PS		 1505}
1506
3ad7a4b1 1507static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1508{
1509 struct vxlan_dev *vxlan = netdev_priv(dev);
1510 struct arphdr *parp;
1511 u8 *arpptr, *sha;
1512 __be32 sip, tip;
1513 struct neighbour *n;
1514
1515 if (dev->flags & IFF_NOARP)
1516 goto out;
1517
1518 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1519 dev->stats.tx_dropped++;
1520 goto out;
1521 }
1522 parp = arp_hdr(skb);
1523
1524 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1525 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1526 parp->ar_pro != htons(ETH_P_IP) ||
1527 parp->ar_op != htons(ARPOP_REQUEST) ||
1528 parp->ar_hln != dev->addr_len ||
1529 parp->ar_pln != 4)
1530 goto out;
1531 arpptr = (u8 *)parp + sizeof(struct arphdr);
1532 sha = arpptr;
1533 arpptr += dev->addr_len; /* sha */
1534 memcpy(&sip, arpptr, sizeof(sip));
1535 arpptr += sizeof(sip);
1536 arpptr += dev->addr_len; /* tha */
1537 memcpy(&tip, arpptr, sizeof(tip));
1538
1539 if (ipv4_is_loopback(tip) ||
1540 ipv4_is_multicast(tip))
1541 goto out;
1542
1543 n = neigh_lookup(&arp_tbl, &tip, dev);
1544
1545 if (n) {
e4f67add
DS		 1546 struct vxlan_fdb *f;
1547 struct sk_buff *reply;
1548
1549 if (!(n->nud_state & NUD_CONNECTED)) {
1550 neigh_release(n);
1551 goto out;
1552 }
1553
3ad7a4b1 1554 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1555 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1556 /* bridge-local neighbor */
1557 neigh_release(n);
1558 goto out;
1559 }
1560
1561 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1562 n->ha, sha);
1563
1564 neigh_release(n);
1565
7346135d
DS		 1566 if (reply == NULL)
1567 goto out;
1568
e4f67add
DS		 1569 skb_reset_mac_header(reply);
1570 __skb_pull(reply, skb_network_offset(reply));
1571 reply->ip_summed = CHECKSUM_UNNECESSARY;
1572 reply->pkt_type = PACKET_HOST;
1573
1574 if (netif_rx_ni(reply) == NET_RX_DROP)
1575 dev->stats.rx_dropped++;
dc5321d7 1576 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 1577 union vxlan_addr ipa = {
1578 .sin.sin_addr.s_addr = tip,
a45e92a5 1579 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1580 };
1581
1582 vxlan_ip_miss(dev, &ipa);
1583 }
e4f67add
DS		 1584out:
1585 consume_skb(skb);
1586 return NETDEV_TX_OK;
1587}
1588
f564f45c 1589#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1590static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1591 struct neighbour *n, bool isrouter)
1592{
1593 struct net_device *dev = request->dev;
1594 struct sk_buff *reply;
1595 struct nd_msg *ns, *na;
1596 struct ipv6hdr *pip6;
1597 u8 *daddr;
1598 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1599 int ns_olen;
1600 int i, len;
1601
f1fb08f6 1602 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1603 return NULL;
1604
1605 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1606 sizeof(*na) + na_olen + dev->needed_tailroom;
1607 reply = alloc_skb(len, GFP_ATOMIC);
1608 if (reply == NULL)
1609 return NULL;
1610
1611 reply->protocol = htons(ETH_P_IPV6);
1612 reply->dev = dev;
1613 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1614 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1615 skb_reset_mac_header(reply);
4b29dba9 1616
f1fb08f6 1617 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1618
1619 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1620 ns_olen = request->len - skb_network_offset(request) -
1621 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1622 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1623 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1624 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1625 break;
1626 }
1627 }
1628
1629 /* Ethernet header */
1630 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1631 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1632 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1633 reply->protocol = htons(ETH_P_IPV6);
1634
1635 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1636 skb_reset_network_header(reply);
4b29dba9
DS		 1637 skb_put(reply, sizeof(struct ipv6hdr));
1638
1639 /* IPv6 header */
1640
1641 pip6 = ipv6_hdr(reply);
1642 memset(pip6, 0, sizeof(struct ipv6hdr));
1643 pip6->version = 6;
1644 pip6->priority = ipv6_hdr(request)->priority;
1645 pip6->nexthdr = IPPROTO_ICMPV6;
1646 pip6->hop_limit = 255;
1647 pip6->daddr = ipv6_hdr(request)->saddr;
1648 pip6->saddr = *(struct in6_addr *)n->primary_key;
1649
1650 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1651 skb_reset_transport_header(reply);
4b29dba9 1652
4b29dba9 1653 /* Neighbor Advertisement */
b080db58 1654 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1655 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1656 na->icmph.icmp6_router = isrouter;
1657 na->icmph.icmp6_override = 1;
1658 na->icmph.icmp6_solicited = 1;
1659 na->target = ns->target;
1660 ether_addr_copy(&na->opt[2], n->ha);
1661 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1662 na->opt[1] = na_olen >> 3;
1663
1664 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1665 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1666 csum_partial(na, sizeof(*na)+na_olen, 0));
1667
1668 pip6->payload_len = htons(sizeof(*na)+na_olen);
1669
1670 skb_push(reply, sizeof(struct ipv6hdr));
1671
1672 reply->ip_summed = CHECKSUM_UNNECESSARY;
1673
1674 return reply;
1675}
1676
3ad7a4b1 1677static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1678{
1679 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 1680 const struct in6_addr *daddr;
8bff3685 1681 const struct ipv6hdr *iphdr;
4b29dba9 1682 struct inet6_dev *in6_dev;
8bff3685
XL		 1683 struct neighbour *n;
1684 struct nd_msg *msg;
f564f45c
CW		 1685
1686 in6_dev = __in6_dev_get(dev);
1687 if (!in6_dev)
1688 goto out;
1689
f564f45c 1690 iphdr = ipv6_hdr(skb);
f564f45c 1691 daddr = &iphdr->daddr;
f1fb08f6 1692 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 1693
4b29dba9
DS		 1694 if (ipv6_addr_loopback(daddr) ||
1695 ipv6_addr_is_multicast(&msg->target))
1696 goto out;
1697
1698 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1699
1700 if (n) {
1701 struct vxlan_fdb *f;
4b29dba9 1702 struct sk_buff *reply;
f564f45c
CW		 1703
1704 if (!(n->nud_state & NUD_CONNECTED)) {
1705 neigh_release(n);
1706 goto out;
1707 }
1708
3ad7a4b1 1709 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1710 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1711 /* bridge-local neighbor */
1712 neigh_release(n);
1713 goto out;
1714 }
1715
4b29dba9
DS		 1716 reply = vxlan_na_create(skb, n,
1717 !!(f ? f->flags & NTF_ROUTER : 0));
1718
f564f45c 1719 neigh_release(n);
4b29dba9
DS		 1720
1721 if (reply == NULL)
1722 goto out;
1723
1724 if (netif_rx_ni(reply) == NET_RX_DROP)
1725 dev->stats.rx_dropped++;
1726
dc5321d7 1727 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1728 union vxlan_addr ipa = {
1729 .sin6.sin6_addr = msg->target,
a45e92a5 1730 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1731 };
1732
f564f45c
CW		 1733 vxlan_ip_miss(dev, &ipa);
1734 }
1735
1736out:
1737 consume_skb(skb);
1738 return NETDEV_TX_OK;
1739}
1740#endif
1741
e4f67add
DS		 1742static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1743{
1744 struct vxlan_dev *vxlan = netdev_priv(dev);
1745 struct neighbour *n;
e4f67add
DS		 1746
1747 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1748 return false;
1749
1750 n = NULL;
1751 switch (ntohs(eth_hdr(skb)->h_proto)) {
1752 case ETH_P_IP:
e15a00aa
CW		 1753 {
1754 struct iphdr *pip;
1755
e4f67add
DS		 1756 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1757 return false;
1758 pip = ip_hdr(skb);
1759 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 1760 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 1761 union vxlan_addr ipa = {
1762 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1763 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1764 };
1765
1766 vxlan_ip_miss(dev, &ipa);
1767 return false;
1768 }
1769
e4f67add 1770 break;
e15a00aa
CW		 1771 }
1772#if IS_ENABLED(CONFIG_IPV6)
1773 case ETH_P_IPV6:
1774 {
1775 struct ipv6hdr *pip6;
1776
1777 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1778 return false;
1779 pip6 = ipv6_hdr(skb);
1780 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 1781 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 1782 union vxlan_addr ipa = {
1783 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1784 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1785 };
1786
1787 vxlan_ip_miss(dev, &ipa);
1788 return false;
1789 }
1790
1791 break;
1792 }
1793#endif
e4f67add
DS		 1794 default:
1795 return false;
1796 }
1797
1798 if (n) {
1799 bool diff;
1800
7367d0b5 1801 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1802 if (diff) {
1803 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1804 dev->addr_len);
1805 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1806 }
1807 neigh_release(n);
1808 return diff;
e4c7ed41
CW		 1809 }
1810
e4f67add
DS		 1811 return false;
1812}
1813
af33c1ad 1814static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1815 struct vxlan_metadata *md)
1816{
1817 struct vxlanhdr_gbp *gbp;
1818
db79a621
TG		 1819 if (!md->gbp)
1820 return;
1821
3511494c 1822 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1823 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1824
1825 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1826 gbp->dont_learn = 1;
1827
1828 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1829 gbp->policy_applied = 1;
1830
1831 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1832}
1833
e1e5314d
JB		 1834static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1835 __be16 protocol)
1836{
1837 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1838
1839 gpe->np_applied = 1;
fa20e0e3
JB		 1840 gpe->next_protocol = tun_p_from_eth_p(protocol);
1841 if (!gpe->next_protocol)
1842 return -EPFNOSUPPORT;
1843 return 0;
e1e5314d
JB		 1844}
1845
f491e56d
JB		 1846static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1847 int iphdr_len, __be32 vni,
1848 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1849 bool udp_sum)
e4c7ed41 1850{
e4c7ed41 1851 struct vxlanhdr *vxh;
e4c7ed41
CW		 1852 int min_headroom;
1853 int err;
dfd8645e 1854 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1855 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1856
af33c1ad 1857 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1858 skb->ip_summed == CHECKSUM_PARTIAL) {
1859 int csum_start = skb_checksum_start_offset(skb);
1860
1861 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1862 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1863 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1864 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1865 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1866 }
e4c7ed41 1867
e4c7ed41 1868 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1869 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1870
1871 /* Need space for new headers (invalidates iph ptr) */
1872 err = skb_cow_head(skb, min_headroom);
e1e5314d 1873 if (unlikely(err))
c46b7897 1874 return err;
649c5b8b 1875
aed069df
AD		 1876 err = iptunnel_handle_offloads(skb, type);
1877 if (err)
c46b7897 1878 return err;
b736a623 1879
d58ff351 1880 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1881 vxh->vx_flags = VXLAN_HF_VNI;
1882 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1883
dfd8645e 1884 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1885 unsigned int start;
dfd8645e 1886
54bfd872
JB		 1887 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1888 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1889 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1890
1891 if (!skb_is_gso(skb)) {
1892 skb->ip_summed = CHECKSUM_NONE;
1893 skb->encapsulation = 0;
1894 }
1895 }
1896
af33c1ad
TH		 1897 if (vxflags & VXLAN_F_GBP)
1898 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1899 if (vxflags & VXLAN_F_GPE) {
1900 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1901 if (err < 0)
c46b7897 1902 return err;
e1e5314d
JB		 1903 inner_protocol = skb->protocol;
1904 }
3511494c 1905
e1e5314d 1906 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1907 return 0;
49560532 1908}
49560532 1909
655c3de1 1910static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1911 struct vxlan_sock *sock4,
1a8496ba 1912 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1913 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1914 struct dst_cache *dst_cache,
db3c6139 1915 const struct ip_tunnel_info *info)
1a8496ba 1916{
db3c6139 1917 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1918 struct rtable *rt = NULL;
1919 struct flowi4 fl4;
1920
655c3de1 1921 if (!sock4)
1922 return ERR_PTR(-EIO);
1923
db3c6139
DB		 1924 if (tos && !info)
1925 use_cache = false;
1926 if (use_cache) {
0c1d70af
PA		 1927 rt = dst_cache_get_ip4(dst_cache, saddr);
1928 if (rt)
1929 return rt;
1930 }
1931
1a8496ba
JB		 1932 memset(&fl4, 0, sizeof(fl4));
1933 fl4.flowi4_oif = oif;
1934 fl4.flowi4_tos = RT_TOS(tos);
1935 fl4.flowi4_mark = skb->mark;
1936 fl4.flowi4_proto = IPPROTO_UDP;
1937 fl4.daddr = daddr;
272d96a5 1938 fl4.saddr = *saddr;
4ecb1d83
MP		 1939 fl4.fl4_dport = dport;
1940 fl4.fl4_sport = sport;
1a8496ba
JB		 1941
1942 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1943 if (likely(!IS_ERR(rt))) {
1944 if (rt->dst.dev == dev) {
1945 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1946 ip_rt_put(rt);
1947 return ERR_PTR(-ELOOP);
1948 }
1949
1a8496ba 1950 *saddr = fl4.saddr;
0c1d70af
PA		 1951 if (use_cache)
1952 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1953 } else {
1954 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1955 return ERR_PTR(-ENETUNREACH);
0c1d70af 1956 }
1a8496ba
JB		 1957 return rt;
1958}
1959
e5d4b29f
JB		 1960#if IS_ENABLED(CONFIG_IPV6)
1961static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1962 struct net_device *dev,
03dc52a8 1963 struct vxlan_sock *sock6,
1400615d 1964 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1965 __be32 label,
e5d4b29f 1966 const struct in6_addr *daddr,
0c1d70af 1967 struct in6_addr *saddr,
4ecb1d83 1968 __be16 dport, __be16 sport,
db3c6139
DB		 1969 struct dst_cache *dst_cache,
1970 const struct ip_tunnel_info *info)
e5d4b29f 1971{
db3c6139 1972 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1973 struct dst_entry *ndst;
1974 struct flowi6 fl6;
1975 int err;
1976
c6fcc4fc 1977 if (!sock6)
1978 return ERR_PTR(-EIO);
1979
1400615d
DB		 1980 if (tos && !info)
1981 use_cache = false;
db3c6139 1982 if (use_cache) {
0c1d70af
PA		 1983 ndst = dst_cache_get_ip6(dst_cache, saddr);
1984 if (ndst)
1985 return ndst;
1986 }
1987
e5d4b29f
JB		 1988 memset(&fl6, 0, sizeof(fl6));
1989 fl6.flowi6_oif = oif;
1990 fl6.daddr = *daddr;
272d96a5 1991 fl6.saddr = *saddr;
eaa93bf4 1992 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1993 fl6.flowi6_mark = skb->mark;
1994 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1995 fl6.fl6_dport = dport;
1996 fl6.fl6_sport = sport;
e5d4b29f
JB		 1997
1998 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1999 sock6->sock->sk,
e5d4b29f 2000 &ndst, &fl6);
655c3de1 2001 if (unlikely(err < 0)) {
2002 netdev_dbg(dev, "no route to %pI6\n", daddr);
2003 return ERR_PTR(-ENETUNREACH);
2004 }
2005
2006 if (unlikely(ndst->dev == dev)) {
2007 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2008 dst_release(ndst);
2009 return ERR_PTR(-ELOOP);
2010 }
e5d4b29f
JB		 2011
2012 *saddr = fl6.saddr;
db3c6139 2013 if (use_cache)
0c1d70af 2014 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 2015 return ndst;
2016}
2017#endif
2018
9dcc71e1
SS		 2019/* Bypass encapsulation if the destination is local */
2020static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 2021 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 2022{
8f84985f 2023 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2024 union vxlan_addr loopback;
2025 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 2026 struct net_device *dev = skb->dev;
2027 int len = skb->len;
9dcc71e1 2028
8f84985f
LR		 2029 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2030 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2031 skb->pkt_type = PACKET_HOST;
2032 skb->encapsulation = 0;
2033 skb->dev = dst_vxlan->dev;
2034 __skb_pull(skb, skb_network_offset(skb));
2035
e4c7ed41
CW		 2036 if (remote_ip->sa.sa_family == AF_INET) {
2037 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2038 loopback.sa.sa_family = AF_INET;
2039#if IS_ENABLED(CONFIG_IPV6)
2040 } else {
2041 loopback.sin6.sin6_addr = in6addr_loopback;
2042 loopback.sa.sa_family = AF_INET6;
2043#endif
2044 }
2045
dc5321d7 2046 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
87613de9
MS		 2047 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
2048 vni);
9dcc71e1
SS		 2049
2050 u64_stats_update_begin(&tx_stats->syncp);
2051 tx_stats->tx_packets++;
ce6502a8 2052 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2053 u64_stats_update_end(&tx_stats->syncp);
2054
2055 if (netif_rx(skb) == NET_RX_SUCCESS) {
2056 u64_stats_update_begin(&rx_stats->syncp);
2057 rx_stats->rx_packets++;
ce6502a8 2058 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2059 u64_stats_update_end(&rx_stats->syncp);
2060 } else {
ce6502a8 2061 dev->stats.rx_dropped++;
9dcc71e1
SS		 2062 }
2063}
2064
fee1fad7 2065static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2066 struct vxlan_dev *vxlan,
2067 union vxlan_addr *daddr,
2068 __be16 dst_port, int dst_ifindex, __be32 vni,
2069 struct dst_entry *dst,
fee1fad7 2070 u32 rt_flags)
2071{
2072#if IS_ENABLED(CONFIG_IPV6)
2073 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2074 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2075 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2076 */
2077 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2078#endif
2079 /* Bypass encapsulation if the destination is local */
2080 if (rt_flags & RTCF_LOCAL &&
2081 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2082 struct vxlan_dev *dst_vxlan;
2083
2084 dst_release(dst);
49f810f0 2085 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2086 daddr->sa.sa_family, dst_port,
dc5321d7 2087 vxlan->cfg.flags);
fee1fad7 2088 if (!dst_vxlan) {
2089 dev->stats.tx_errors++;
2090 kfree_skb(skb);
2091
2092 return -ENOENT;
2093 }
3ad7a4b1 2094 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2095 return 1;
2096 }
2097
2098 return 0;
2099}
2100
4ad16930 2101static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2102 __be32 default_vni, struct vxlan_rdst *rdst,
2103 bool did_rsc)
d342894c 2104{
d71785ff 2105 struct dst_cache *dst_cache;
3093fbe7 2106 struct ip_tunnel_info *info;
d342894c 2107 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2108 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2109 union vxlan_addr *dst;
272d96a5 2110 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2111 struct vxlan_metadata _md;
2112 struct vxlan_metadata *md = &_md;
e4c7ed41 2113 __be16 src_port = 0, dst_port;
655c3de1 2114 struct dst_entry *ndst = NULL;
e7f70af1 2115 __be32 vni, label;
d342894c 2116 __u8 tos, ttl;
49f810f0 2117 int ifindex;
0e6fbc5b 2118 int err;
dc5321d7 2119 u32 flags = vxlan->cfg.flags;
b4ed5cad 2120 bool udp_sum = false;
f491e56d 2121 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2122
61adedf3 2123 info = skb_tunnel_info(skb);
3093fbe7 2124
ee122c79 2125 if (rdst) {
0770b53b 2126 dst = &rdst->remote_ip;
2127 if (vxlan_addr_any(dst)) {
2128 if (did_rsc) {
2129 /* short-circuited back to local bridge */
3ad7a4b1 2130 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2131 return;
2132 }
2133 goto drop;
2134 }
2135
0dfbdf41 2136 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2137 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2138 ifindex = rdst->remote_ifindex;
1158632b 2139 local_ip = vxlan->cfg.saddr;
d71785ff 2140 dst_cache = &rdst->dst_cache;
0770b53b 2141 md->gbp = skb->mark;
72f6d71e
HL		 2142 if (flags & VXLAN_F_TTL_INHERIT) {
2143 ttl = ip_tunnel_get_ttl(old_iph, skb);
2144 } else {
2145 ttl = vxlan->cfg.ttl;
2146 if (!ttl && vxlan_addr_multicast(dst))
2147 ttl = 1;
2148 }
0770b53b 2149
2150 tos = vxlan->cfg.tos;
2151 if (tos == 1)
2152 tos = ip_tunnel_get_dsfield(old_iph, skb);
2153
2154 if (dst->sa.sa_family == AF_INET)
2155 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2156 else
2157 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2158 label = vxlan->cfg.label;
ee122c79
TG		 2159 } else {
2160 if (!info) {
2161 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2162 dev->name);
2163 goto drop;
2164 }
b1be00a6 2165 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2166 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2167 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2168 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2169 } else {
a725e514 2170 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2171 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2172 }
ee122c79 2173 dst = &remote_ip;
0770b53b 2174 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2175 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2176 ifindex = 0;
d71785ff 2177 dst_cache = &info->dst_cache;
256c87c1
PJV		 2178 if (info->options_len &&
2179 info->key.tun_flags & TUNNEL_VXLAN_OPT)
0770b53b 2180 md = ip_tunnel_info_opts(info);
a725e514
JB		 2181 ttl = info->key.ttl;
2182 tos = info->key.tos;
e7f70af1 2183 label = info->key.label;
b4ed5cad 2184 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2185 }
0770b53b 2186 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2187 vxlan->cfg.port_max, true);
a725e514 2188
56de859e 2189 rcu_read_lock();
e4c7ed41 2190 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2191 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2192 struct rtable *rt;
0770b53b 2193 __be16 df = 0;
c6fcc4fc 2194
49f810f0 2195 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2196 dst->sin.sin_addr.s_addr,
1158632b 2197 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2198 dst_port, src_port,
d71785ff 2199 dst_cache, info);
8ebd115b
DM		 2200 if (IS_ERR(rt)) {
2201 err = PTR_ERR(rt);
c46b7897 2202 goto tx_error;
8ebd115b 2203 }
e4c7ed41
CW		 2204
2205 /* Bypass encapsulation if the destination is local */
fee1fad7 2206 if (!info) {
2207 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2208 dst_port, ifindex, vni,
2209 &rt->dst, rt->rt_flags);
fee1fad7 2210 if (err)
56de859e 2211 goto out_unlock;
fee1fad7 2212 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2213 df = htons(IP_DF);
fee1fad7 2214 }
6ceb31ca 2215
c46b7897 2216 ndst = &rt->dst;
a93bf0ff
XL		 2217 if (skb_dst(skb)) {
2218 int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2219
f15ca723 2220 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2221 }
2222
e4c7ed41
CW		 2223 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2224 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2225 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2226 vni, md, flags, udp_sum);
f491e56d 2227 if (err < 0)
c46b7897 2228 goto tx_error;
f491e56d 2229
1158632b 2230 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2231 dst->sin.sin_addr.s_addr, tos, ttl, df,
2232 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2233#if IS_ENABLED(CONFIG_IPV6)
2234 } else {
c6fcc4fc 2235 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2236
49f810f0 2237 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2238 label, &dst->sin6.sin6_addr,
1158632b 2239 &local_ip.sin6.sin6_addr,
4ecb1d83 2240 dst_port, src_port,
db3c6139 2241 dst_cache, info);
e5d4b29f 2242 if (IS_ERR(ndst)) {
8ebd115b 2243 err = PTR_ERR(ndst);
c46b7897 2244 ndst = NULL;
9dcc71e1 2245 goto tx_error;
e4c7ed41 2246 }
655c3de1 2247
fee1fad7 2248 if (!info) {
2249 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2250
fee1fad7 2251 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2252 dst_port, ifindex, vni,
2253 ndst, rt6i_flags);
fee1fad7 2254 if (err)
56de859e 2255 goto out_unlock;
fee1fad7 2256 }
35e2d115 2257
a93bf0ff
XL		 2258 if (skb_dst(skb)) {
2259 int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
2260
f15ca723 2261 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2262 }
2263
1400615d 2264 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2265 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2266 skb_scrub_packet(skb, xnet);
2267 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2268 vni, md, flags, udp_sum);
c46b7897 2269 if (err < 0)
2270 goto tx_error;
2271
0770b53b 2272 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2273 &local_ip.sin6.sin6_addr,
272d96a5 2274 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2275 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2276#endif
2277 }
56de859e
JK		 2278out_unlock:
2279 rcu_read_unlock();
4ad16930 2280 return;
d342894c 2281
2282drop:
2283 dev->stats.tx_dropped++;
c46b7897 2284 dev_kfree_skb(skb);
2285 return;
d342894c 2286
2287tx_error:
56de859e 2288 rcu_read_unlock();
655c3de1 2289 if (err == -ELOOP)
2290 dev->stats.collisions++;
2291 else if (err == -ENETUNREACH)
2292 dev->stats.tx_carrier_errors++;
c46b7897 2293 dst_release(ndst);
d342894c 2294 dev->stats.tx_errors++;
c46b7897 2295 kfree_skb(skb);
d342894c 2296}
2297
6681712d
DS		 2298/* Transmit local packets over Vxlan
2299 *
2300 * Outer IP header inherits ECN and DF from inner header.
2301 * Outer UDP destination is the VXLAN assigned port.
2302 * source port is based on hash of flow
2303 */
2304static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2305{
2306 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2307 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2308 const struct ip_tunnel_info *info;
6681712d 2309 bool did_rsc = false;
6681712d 2310 struct vxlan_fdb *f;
8bff3685 2311 struct ethhdr *eth;
3ad7a4b1 2312 __be32 vni = 0;
6681712d 2313
61adedf3 2314 info = skb_tunnel_info(skb);
3093fbe7 2315
6681712d 2316 skb_reset_mac_header(skb);
6681712d 2317
dc5321d7 2318 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2319 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2320 info->mode & IP_TUNNEL_INFO_TX) {
2321 vni = tunnel_id_to_key32(info->key.tun_id);
2322 } else {
2323 if (info && info->mode & IP_TUNNEL_INFO_TX)
2324 vxlan_xmit_one(skb, dev, vni, NULL, false);
2325 else
2326 kfree_skb(skb);
2327 return NETDEV_TX_OK;
2328 }
47e5d1b0
JB		 2329 }
2330
dc5321d7 2331 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2332 eth = eth_hdr(skb);
f564f45c 2333 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2334 return arp_reduce(dev, skb, vni);
f564f45c 2335#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 2336 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2337 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2338 sizeof(struct nd_msg)) &&
2339 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2340 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2341
2342 if (m->icmph.icmp6_code == 0 &&
2343 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2344 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2345 }
2346#endif
2347 }
6681712d 2348
47e5d1b0 2349 eth = eth_hdr(skb);
3ad7a4b1 2350 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2351 did_rsc = false;
2352
dc5321d7 2353 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2354 (ntohs(eth->h_proto) == ETH_P_IP ||
2355 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2356 did_rsc = route_shortcircuit(dev, skb);
2357 if (did_rsc)
3ad7a4b1 2358 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2359 }
2360
6681712d 2361 if (f == NULL) {
3ad7a4b1 2362 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2363 if (f == NULL) {
dc5321d7 2364 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2365 !is_multicast_ether_addr(eth->h_dest))
2366 vxlan_fdb_miss(vxlan, eth->h_dest);
2367
2368 dev->stats.tx_dropped++;
8f646c92 2369 kfree_skb(skb);
afbd8bae
MR		 2370 return NETDEV_TX_OK;
2371 }
2372 }
6681712d 2373
afbd8bae
MR		 2374 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2375 struct sk_buff *skb1;
6681712d 2376
8f646c92
ED		 2377 if (!fdst) {
2378 fdst = rdst;
2379 continue;
2380 }
afbd8bae
MR		 2381 skb1 = skb_clone(skb, GFP_ATOMIC);
2382 if (skb1)
3ad7a4b1 2383 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2384 }
2385
8f646c92 2386 if (fdst)
3ad7a4b1 2387 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2388 else
2389 kfree_skb(skb);
4ad16930 2390 return NETDEV_TX_OK;
6681712d
DS		 2391}
2392
d342894c 2393/* Walk the forwarding table and purge stale entries */
df7e828c 2394static void vxlan_cleanup(struct timer_list *t)
d342894c 2395{
df7e828c 2396 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2397 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2398 unsigned int h;
2399
2400 if (!netif_running(vxlan->dev))
2401 return;
2402
d342894c 2403 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2404 struct hlist_node *p, *n;
14e1d0fa
SD		 2405
2406 spin_lock_bh(&vxlan->hash_lock);
d342894c 2407 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2408 struct vxlan_fdb *f
2409 = container_of(p, struct vxlan_fdb, hlist);
2410 unsigned long timeout;
2411
efb5f68f 2412 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2413 continue;
2414
def499c9
RP		 2415 if (f->flags & NTF_EXT_LEARNED)
2416 continue;
2417
0dfbdf41 2418 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2419 if (time_before_eq(timeout, jiffies)) {
2420 netdev_dbg(vxlan->dev,
2421 "garbage collect %pM\n",
2422 f->eth_addr);
2423 f->state = NUD_STALE;
4c2438ba 2424 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2425 } else if (time_before(timeout, next_timer))
2426 next_timer = timeout;
2427 }
14e1d0fa 2428 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2429 }
d342894c 2430
2431 mod_timer(&vxlan->age_timer, next_timer);
2432}
2433
a53cb29b
MB		 2434static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2435{
2436 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2437
2438 spin_lock(&vn->sock_lock);
69e76661
JB		 2439 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2440#if IS_ENABLED(CONFIG_IPV6)
2441 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2442#endif
a53cb29b
MB		 2443 spin_unlock(&vn->sock_lock);
2444}
2445
69e76661
JB		 2446static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2447 struct vxlan_dev_node *node)
9c2e24e1 2448{
56ef9c90 2449 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2450 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2451
69e76661 2452 node->vxlan = vxlan;
56ef9c90 2453 spin_lock(&vn->sock_lock);
69e76661 2454 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 2455 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2456}
2457
d342894c 2458/* Setup stats when device is created */
2459static int vxlan_init(struct net_device *dev)
2460{
1c213bd2 2461 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2462 if (!dev->tstats)
d342894c 2463 return -ENOMEM;
2464
2465 return 0;
2466}
2467
3ad7a4b1 2468static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2469{
2470 struct vxlan_fdb *f;
2471
2472 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2473 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2474 if (f)
4c2438ba 2475 vxlan_fdb_destroy(vxlan, f, true);
afbd8bae
MR		 2476 spin_unlock_bh(&vxlan->hash_lock);
2477}
2478
ebf4063e
SH		 2479static void vxlan_uninit(struct net_device *dev)
2480{
2481 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2482
3ad7a4b1 2483 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2484
ebf4063e
SH		 2485 free_percpu(dev->tstats);
2486}
2487
d342894c 2488/* Start ageing timer and join group when device is brought up */
2489static int vxlan_open(struct net_device *dev)
2490{
2491 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2492 int ret;
56ef9c90 2493
205f356d
JB		 2494 ret = vxlan_sock_add(vxlan);
2495 if (ret < 0)
2496 return ret;
d342894c 2497
79d4a94f 2498 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2499 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2500 if (ret == -EADDRINUSE)
2501 ret = 0;
56ef9c90 2502 if (ret) {
205f356d 2503 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2504 return ret;
2505 }
d342894c 2506 }
2507
0dfbdf41 2508 if (vxlan->cfg.age_interval)
d342894c 2509 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2510
56ef9c90 2511 return ret;
d342894c 2512}
2513
2514/* Purge the forwarding table */
8b3f9337 2515static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2516{
31fec5aa 2517 unsigned int h;
d342894c 2518
2519 spin_lock_bh(&vxlan->hash_lock);
2520 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2521 struct hlist_node *p, *n;
2522 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2523 struct vxlan_fdb *f
2524 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2525 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2526 continue;
afbd8bae
MR		 2527 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2528 if (!is_zero_ether_addr(f->eth_addr))
4c2438ba 2529 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2530 }
2531 }
2532 spin_unlock_bh(&vxlan->hash_lock);
2533}
2534
2535/* Cleanup timer and forwarding table on shutdown */
2536static int vxlan_stop(struct net_device *dev)
2537{
2538 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2539 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2540 int ret = 0;
d342894c 2541
24c0e683 2542 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2543 !vxlan_group_used(vn, vxlan))
56ef9c90 2544 ret = vxlan_igmp_leave(vxlan);
d342894c 2545
2546 del_timer_sync(&vxlan->age_timer);
2547
8b3f9337 2548 vxlan_flush(vxlan, false);
205f356d 2549 vxlan_sock_release(vxlan);
d342894c 2550
56ef9c90 2551 return ret;
d342894c 2552}
2553
d342894c 2554/* Stub, nothing needs to be done. */
2555static void vxlan_set_multicast_list(struct net_device *dev)
2556{
2557}
2558
91572088 2559static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2560{
91572088
JW		 2561 struct vxlan_dev *vxlan = netdev_priv(dev);
2562 struct vxlan_rdst *dst = &vxlan->default_dst;
2563 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2564 dst->remote_ifindex);
ce44a4ae 2565 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2566
91572088
JW		 2567 /* This check is different than dev->max_mtu, because it looks at
2568 * the lowerdev->mtu, rather than the static dev->max_mtu
2569 */
2570 if (lowerdev) {
2571 int max_mtu = lowerdev->mtu -
2572 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2573 if (new_mtu > max_mtu)
72564b59 2574 return -EINVAL;
72564b59
DW		 2575 }
2576
345010b5
DB		 2577 dev->mtu = new_mtu;
2578 return 0;
2579}
2580
fc4099f1
PS		 2581static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2582{
2583 struct vxlan_dev *vxlan = netdev_priv(dev);
2584 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2585 __be16 sport, dport;
2586
2587 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2588 vxlan->cfg.port_max, true);
2589 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2590
239e944f 2591 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2592 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2593 struct rtable *rt;
2594
655c3de1 2595 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2596 info->key.u.ipv4.dst,
22f0708a
PA		 2597 &info->key.u.ipv4.src, dport, sport,
2598 &info->dst_cache, info);
1a8496ba
JB		 2599 if (IS_ERR(rt))
2600 return PTR_ERR(rt);
2601 ip_rt_put(rt);
239e944f
JB		 2602 } else {
2603#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2604 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2605 struct dst_entry *ndst;
2606
655c3de1 2607 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2608 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2609 &info->key.u.ipv6.src, dport, sport,
2610 &info->dst_cache, info);
239e944f
JB		 2611 if (IS_ERR(ndst))
2612 return PTR_ERR(ndst);
2613 dst_release(ndst);
239e944f
JB		 2614#else /* !CONFIG_IPV6 */
2615 return -EPFNOSUPPORT;
2616#endif
2617 }
1a8496ba
JB		 2618 info->key.tp_src = sport;
2619 info->key.tp_dst = dport;
239e944f 2620 return 0;
fc4099f1
PS		 2621}
2622
0c867c9b 2623static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2624 .ndo_init = vxlan_init,
ebf4063e 2625 .ndo_uninit = vxlan_uninit,
d342894c 2626 .ndo_open = vxlan_open,
2627 .ndo_stop = vxlan_stop,
2628 .ndo_start_xmit = vxlan_xmit,
e8171045 2629 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2630 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2631 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2632 .ndo_validate_addr = eth_validate_addr,
2633 .ndo_set_mac_address = eth_mac_addr,
2634 .ndo_fdb_add = vxlan_fdb_add,
2635 .ndo_fdb_del = vxlan_fdb_delete,
2636 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2637 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2638};
2639
e1e5314d
JB		 2640static const struct net_device_ops vxlan_netdev_raw_ops = {
2641 .ndo_init = vxlan_init,
2642 .ndo_uninit = vxlan_uninit,
2643 .ndo_open = vxlan_open,
2644 .ndo_stop = vxlan_stop,
2645 .ndo_start_xmit = vxlan_xmit,
2646 .ndo_get_stats64 = ip_tunnel_get_stats64,
2647 .ndo_change_mtu = vxlan_change_mtu,
2648 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2649};
2650
d342894c 2651/* Info for udev, that this is a virtual tunnel endpoint */
2652static struct device_type vxlan_type = {
2653 .name = "vxlan",
2654};
2655
e5de25dc 2656/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2657 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2658 * to implement the ndo_udp_tunnel_add.
53cf5275 2659 */
2d2b13fc 2660static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 2661{
2662 struct vxlan_sock *vs;
2663 struct net *net = dev_net(dev);
2664 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2665 unsigned int i;
53cf5275
JG		 2666
2667 spin_lock(&vn->sock_lock);
2668 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 2669 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2670 unsigned short type;
2671
2672 if (vs->flags & VXLAN_F_GPE)
2673 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2674 else
2675 type = UDP_TUNNEL_TYPE_VXLAN;
2676
2677 if (push)
2678 udp_tunnel_push_rx_port(dev, vs->sock, type);
2679 else
2680 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2681 }
53cf5275
JG		 2682 }
2683 spin_unlock(&vn->sock_lock);
2684}
53cf5275 2685
d342894c 2686/* Initialize the device structure. */
2687static void vxlan_setup(struct net_device *dev)
2688{
2689 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2690 unsigned int h;
d342894c 2691
65226ef8
JB		 2692 eth_hw_addr_random(dev);
2693 ether_setup(dev);
2694
cf124db5 2695 dev->needs_free_netdev = true;
d342894c 2696 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2697
d342894c 2698 dev->features |= NETIF_F_LLTX;
d6727fe3 2699 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2700 dev->features |= NETIF_F_RXCSUM;
05c0db08 2701 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2702
1eaa8178 2703 dev->vlan_features = dev->features;
0afb1666 2704 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2705 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2706 netif_keep_dst(dev);
0c867c9b 2707 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2708
a985343b
MS		 2709 /* MTU range: 68 - 65535 */
2710 dev->min_mtu = ETH_MIN_MTU;
2711 dev->max_mtu = ETH_MAX_MTU;
2712
553675fb 2713 INIT_LIST_HEAD(&vxlan->next);
d342894c 2714 spin_lock_init(&vxlan->hash_lock);
2715
df7e828c 2716 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 2717
2718 vxlan->dev = dev;
2719
58ce31cc
TH		 2720 gro_cells_init(&vxlan->gro_cells, dev);
2721
d342894c 2722 for (h = 0; h < FDB_HASH_SIZE; ++h)
2723 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2724}
2725
0c867c9b
JB		 2726static void vxlan_ether_setup(struct net_device *dev)
2727{
0c867c9b
JB		 2728 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2729 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2730 dev->netdev_ops = &vxlan_netdev_ether_ops;
2731}
2732
e1e5314d
JB		 2733static void vxlan_raw_setup(struct net_device *dev)
2734{
65226ef8 2735 dev->header_ops = NULL;
e1e5314d
JB		 2736 dev->type = ARPHRD_NONE;
2737 dev->hard_header_len = 0;
2738 dev->addr_len = 0;
e1e5314d
JB		 2739 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2740 dev->netdev_ops = &vxlan_netdev_raw_ops;
2741}
2742
d342894c 2743static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2744 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2745 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2746 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2747 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2748 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2749 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2750 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2751 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2752 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2753 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2754 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2755 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2756 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2757 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2758 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2759 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2760 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2761 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2762 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2763 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2764 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2765 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2766 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2767 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2768 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2769 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2770 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
72f6d71e 2771 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
d342894c 2772};
2773
a8b8a889
MS		 2774static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2775 struct netlink_ext_ack *extack)
d342894c 2776{
2777 if (tb[IFLA_ADDRESS]) {
2778 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 2779 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2780 "Provided link layer address is not Ethernet");
d342894c 2781 return -EINVAL;
2782 }
2783
2784 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 2785 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2786 "Provided Ethernet address is not unicast");
d342894c 2787 return -EADDRNOTAVAIL;
2788 }
2789 }
2790
a985343b 2791 if (tb[IFLA_MTU]) {
019b13ae 2792 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 2793
653ef6a3
GM		 2794 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2795 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2796 "MTU must be between 68 and 65535");
a985343b 2797 return -EINVAL;
653ef6a3 2798 }
a985343b
MS		 2799 }
2800
653ef6a3
GM		 2801 if (!data) {
2802 NL_SET_ERR_MSG(extack,
2803 "Required attributes not provided to perform the operation");
d342894c 2804 return -EINVAL;
653ef6a3 2805 }
d342894c 2806
2807 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 2808 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2809
653ef6a3
GM		 2810 if (id >= VXLAN_N_VID) {
2811 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2812 "VXLAN ID must be lower than 16777216");
d342894c 2813 return -ERANGE;
653ef6a3 2814 }
d342894c 2815 }
2816
05f47d69 2817 if (data[IFLA_VXLAN_PORT_RANGE]) {
2818 const struct ifla_vxlan_port_range *p
2819 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2820
2821 if (ntohs(p->high) < ntohs(p->low)) {
653ef6a3
GM		 2822 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2823 "Invalid source port range");
05f47d69 2824 return -EINVAL;
2825 }
2826 }
2827
d342894c 2828 return 0;
2829}
2830
1b13c97f
YB		 2831static void vxlan_get_drvinfo(struct net_device *netdev,
2832 struct ethtool_drvinfo *drvinfo)
2833{
2834 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2835 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2836}
2837
2838static const struct ethtool_ops vxlan_ethtool_ops = {
2839 .get_drvinfo = vxlan_get_drvinfo,
2840 .get_link = ethtool_op_get_link,
2841};
2842
3ee64f39
TH		 2843static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2844 __be16 port, u32 flags)
553675fb 2845{
e4c7ed41 2846 struct socket *sock;
3ee64f39
TH		 2847 struct udp_port_cfg udp_conf;
2848 int err;
e4c7ed41 2849
3ee64f39 2850 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2851
3ee64f39
TH		 2852 if (ipv6) {
2853 udp_conf.family = AF_INET6;
3ee64f39 2854 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2855 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2856 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2857 } else {
2858 udp_conf.family = AF_INET;
e4c7ed41 2859 }
e4c7ed41 2860
3ee64f39 2861 udp_conf.local_udp_port = port;
553675fb 2862
3ee64f39
TH		 2863 /* Open UDP socket */
2864 err = udp_sock_create(net, &udp_conf, &sock);
2865 if (err < 0)
2866 return ERR_PTR(err);
e4c7ed41 2867
39deb2c7 2868 return sock;
e4c7ed41
CW		 2869}
2870
2871/* Create new listen socket if needed */
b1be00a6
JB		 2872static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2873 __be16 port, u32 flags)
e4c7ed41
CW		 2874{
2875 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2876 struct vxlan_sock *vs;
2877 struct socket *sock;
e4c7ed41 2878 unsigned int h;
acbf74a7 2879 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2880
dc01e7d3 2881 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2882 if (!vs)
2883 return ERR_PTR(-ENOMEM);
2884
2885 for (h = 0; h < VNI_HASH_SIZE; ++h)
2886 INIT_HLIST_HEAD(&vs->vni_list[h]);
2887
3ee64f39 2888 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2889 if (IS_ERR(sock)) {
553675fb 2890 kfree(vs);
e50fddc8 2891 return ERR_CAST(sock);
553675fb 2892 }
e4c7ed41
CW		 2893
2894 vs->sock = sock;
66af846f 2895 refcount_set(&vs->refcnt, 1);
af33c1ad 2896 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2897
9c2e24e1
PS		 2898 spin_lock(&vn->sock_lock);
2899 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2900 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2901 (vs->flags & VXLAN_F_GPE) ?
2902 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2903 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2904 spin_unlock(&vn->sock_lock);
553675fb 2905
2906 /* Mark socket as an encapsulation socket. */
5602c48c 2907 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2908 tunnel_cfg.sk_user_data = vs;
2909 tunnel_cfg.encap_type = 1;
f2d1968e 2910 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2911 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2912 tunnel_cfg.gro_receive = vxlan_gro_receive;
2913 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2914
2915 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2916
9c2e24e1
PS		 2917 return vs;
2918}
2919
b1be00a6 2920static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2921{
205f356d
JB		 2922 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2923 struct vxlan_sock *vs = NULL;
69e76661 2924 struct vxlan_dev_node *node;
9c2e24e1 2925
205f356d 2926 if (!vxlan->cfg.no_share) {
56ef9c90 2927 spin_lock(&vn->sock_lock);
205f356d 2928 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
dc5321d7 2929 vxlan->cfg.dst_port, vxlan->cfg.flags);
66af846f 2930 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 2931 spin_unlock(&vn->sock_lock);
205f356d 2932 return -EBUSY;
56ef9c90
MRL		 2933 }
2934 spin_unlock(&vn->sock_lock);
2935 }
205f356d 2936 if (!vs)
b1be00a6 2937 vs = vxlan_socket_create(vxlan->net, ipv6,
dc5321d7 2938 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d
JB		 2939 if (IS_ERR(vs))
2940 return PTR_ERR(vs);
b1be00a6 2941#if IS_ENABLED(CONFIG_IPV6)
69e76661 2942 if (ipv6) {
c6fcc4fc 2943 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 2944 node = &vxlan->hlist6;
2945 } else
b1be00a6 2946#endif
69e76661 2947 {
c6fcc4fc 2948 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 2949 node = &vxlan->hlist4;
2950 }
2951 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 2952 return 0;
553675fb 2953}
2954
b1be00a6
JB		 2955static int vxlan_sock_add(struct vxlan_dev *vxlan)
2956{
dc5321d7
MS		 2957 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2958 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 2959 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2960 int ret = 0;
2961
c6fcc4fc 2962 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2963#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2964 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2965 if (ipv6) {
b1be00a6 2966 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2967 if (ret < 0 && ret != -EAFNOSUPPORT)
2968 ipv4 = false;
2969 }
b1be00a6 2970#endif
d074bf96 2971 if (ipv4)
b1be00a6
JB		 2972 ret = __vxlan_sock_add(vxlan, false);
2973 if (ret < 0)
2974 vxlan_sock_release(vxlan);
2975 return ret;
2976}
2977
a985343b
MS		 2978static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2979 struct net_device **lower,
653ef6a3
GM		 2980 struct vxlan_dev *old,
2981 struct netlink_ext_ack *extack)
d342894c 2982{
33564bbb 2983 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 2984 struct vxlan_dev *tmp;
e4c7ed41 2985 bool use_ipv6 = false;
d342894c 2986
a985343b
MS		 2987 if (conf->flags & VXLAN_F_GPE) {
2988 /* For now, allow GPE only together with
2989 * COLLECT_METADATA. This can be relaxed later; in such
2990 * case, the other side of the PtP link will have to be
2991 * provided.
2992 */
2993 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2994 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 2995 NL_SET_ERR_MSG(extack,
2996 "VXLAN GPE does not support this combination of attributes");
a985343b 2997 return -EINVAL;
3555621d 2998 }
e1e5314d 2999 }
0c867c9b 3000
ce44a4ae
MS		 3001 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3002 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 3003 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 3004 conf->saddr.sa.sa_family = AF_INET;
3005 } else if (!conf->remote_ip.sa.sa_family) {
3006 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3007 } else if (!conf->saddr.sa.sa_family) {
3008 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3009 }
3010
653ef6a3
GM		 3011 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3012 NL_SET_ERR_MSG(extack,
3013 "Local and remote address must be from the same family");
ce44a4ae 3014 return -EINVAL;
653ef6a3 3015 }
e4c7ed41 3016
653ef6a3
GM		 3017 if (vxlan_addr_multicast(&conf->saddr)) {
3018 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 3019 return -EINVAL;
653ef6a3 3020 }
0f22a3c6 3021
ce44a4ae 3022 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3023 if (!IS_ENABLED(CONFIG_IPV6)) {
3024 NL_SET_ERR_MSG(extack,
3025 "IPv6 support not enabled in the kernel");
057ba29b 3026 return -EPFNOSUPPORT;
653ef6a3 3027 }
e4c7ed41 3028 use_ipv6 = true;
a985343b 3029 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3030
3031 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3032 int local_type =
3033 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3034 int remote_type =
3035 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3036
3037 if (local_type & IPV6_ADDR_LINKLOCAL) {
3038 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3039 (remote_type != IPV6_ADDR_ANY)) {
3040 NL_SET_ERR_MSG(extack,
3041 "Invalid combination of local and remote address scopes");
0f22a3c6 3042 return -EINVAL;
653ef6a3 3043 }
0f22a3c6
MS		 3044
3045 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3046 } else {
3047 if (remote_type ==
653ef6a3
GM		 3048 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3049 NL_SET_ERR_MSG(extack,
3050 "Invalid combination of local and remote address scopes");
0f22a3c6 3051 return -EINVAL;
653ef6a3 3052 }
0f22a3c6
MS		 3053
3054 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3055 }
3056 }
057ba29b 3057 }
d342894c 3058
653ef6a3
GM		 3059 if (conf->label && !use_ipv6) {
3060 NL_SET_ERR_MSG(extack,
3061 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3062 return -EINVAL;
653ef6a3 3063 }
e7f70af1 3064
a985343b
MS		 3065 if (conf->remote_ifindex) {
3066 struct net_device *lowerdev;
34e02aa1 3067
a985343b 3068 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3069 if (!lowerdev) {
3070 NL_SET_ERR_MSG(extack,
3071 "Invalid local interface, device not found");
34e02aa1 3072 return -ENODEV;
653ef6a3 3073 }
d342894c 3074
e4c7ed41
CW		 3075#if IS_ENABLED(CONFIG_IPV6)
3076 if (use_ipv6) {
3077 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3078 if (idev && idev->cnf.disable_ipv6) {
3079 NL_SET_ERR_MSG(extack,
3080 "IPv6 support disabled by administrator");
e4c7ed41 3081 return -EPERM;
653ef6a3 3082 }
e4c7ed41
CW		 3083 }
3084#endif
3085
a985343b
MS		 3086 *lower = lowerdev;
3087 } else {
653ef6a3
GM		 3088 if (vxlan_addr_multicast(&conf->remote_ip)) {
3089 NL_SET_ERR_MSG(extack,
3090 "Local interface required for multicast remote destination");
3091
a985343b 3092 return -EINVAL;
653ef6a3 3093 }
1ba56fb4 3094
0f22a3c6 3095#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3096 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3097 NL_SET_ERR_MSG(extack,
3098 "Local interface required for link-local local/remote addresses");
0f22a3c6 3099 return -EINVAL;
653ef6a3 3100 }
0f22a3c6
MS		 3101#endif
3102
a985343b 3103 *lower = NULL;
9dc2ad10 3104 }
d342894c 3105
a985343b
MS		 3106 if (!conf->dst_port) {
3107 if (conf->flags & VXLAN_F_GPE)
3108 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3109 else
3110 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3111 }
3112
a985343b
MS		 3113 if (!conf->age_interval)
3114 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3115
a985343b
MS		 3116 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3117 if (tmp == old)
3118 continue;
91572088 3119
49f810f0
MS		 3120 if (tmp->cfg.vni != conf->vni)
3121 continue;
3122 if (tmp->cfg.dst_port != conf->dst_port)
3123 continue;
3124 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3125 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3126 continue;
3127
3128 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3129 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3130 continue;
3131
653ef6a3
GM		 3132 NL_SET_ERR_MSG(extack,
3133 "A VXLAN device with the specified VNI already exists");
49f810f0 3134 return -EEXIST;
a985343b 3135 }
91572088 3136
a985343b
MS		 3137 return 0;
3138}
3139
3140static void vxlan_config_apply(struct net_device *dev,
3141 struct vxlan_config *conf,
889ce937
SD		 3142 struct net_device *lowerdev,
3143 struct net *src_net,
3144 bool changelink)
a985343b
MS		 3145{
3146 struct vxlan_dev *vxlan = netdev_priv(dev);
3147 struct vxlan_rdst *dst = &vxlan->default_dst;
3148 unsigned short needed_headroom = ETH_HLEN;
3149 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3150 int max_mtu = ETH_MAX_MTU;
3151
3152 if (!changelink) {
3153 if (conf->flags & VXLAN_F_GPE)
3154 vxlan_raw_setup(dev);
3155 else
3156 vxlan_ether_setup(dev);
3157
3158 if (conf->mtu)
3159 dev->mtu = conf->mtu;
889ce937
SD		 3160
3161 vxlan->net = src_net;
a985343b
MS		 3162 }
3163
3164 dst->remote_vni = conf->vni;
3165
3166 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3167
3168 if (lowerdev) {
3169 dst->remote_ifindex = conf->remote_ifindex;
3170
3171 dev->gso_max_size = lowerdev->gso_max_size;
3172 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3173
a985343b 3174 needed_headroom = lowerdev->hard_header_len;
91572088 3175
a985343b
MS		 3176 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3177 VXLAN_HEADROOM);
f870c1ff
AK		 3178 if (max_mtu < ETH_MIN_MTU)
3179 max_mtu = ETH_MIN_MTU;
3180
3181 if (!changelink && !conf->mtu)
3182 dev->mtu = max_mtu;
7e059158
DW		 3183 }
3184
a985343b
MS		 3185 if (dev->mtu > max_mtu)
3186 dev->mtu = max_mtu;
3187
b1be00a6
JB		 3188 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3189 needed_headroom += VXLAN6_HEADROOM;
3190 else
3191 needed_headroom += VXLAN_HEADROOM;
3192 dev->needed_headroom = needed_headroom;
3193
0dfbdf41 3194 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3195}
0dfbdf41 3196
a985343b 3197static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 3198 struct vxlan_config *conf, bool changelink,
3199 struct netlink_ext_ack *extack)
a985343b
MS		 3200{
3201 struct vxlan_dev *vxlan = netdev_priv(dev);
3202 struct net_device *lowerdev;
3203 int ret;
0dfbdf41 3204
653ef6a3 3205 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 3206 if (ret)
3207 return ret;
8bcdc4f3 3208
889ce937 3209 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3210
0dfbdf41
TG		 3211 return 0;
3212}
3213
c80498e3 3214static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 3215 struct vxlan_config *conf,
3216 struct netlink_ext_ack *extack)
c80498e3
ND		 3217{
3218 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3219 struct vxlan_dev *vxlan = netdev_priv(dev);
0241b836 3220 struct vxlan_fdb *f = NULL;
c80498e3
ND		 3221 int err;
3222
653ef6a3 3223 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 3224 if (err)
3225 return err;
3226
3227 dev->ethtool_ops = &vxlan_ethtool_ops;
3228
3229 /* create an fdb entry for a valid default destination */
3230 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
0241b836 3231 err = vxlan_fdb_create(vxlan, all_zeros_mac,
c80498e3
ND		 3232 &vxlan->default_dst.remote_ip,
3233 NUD_REACHABLE | NUD_PERMANENT,
c80498e3
ND		 3234 vxlan->cfg.dst_port,
3235 vxlan->default_dst.remote_vni,
3236 vxlan->default_dst.remote_vni,
3237 vxlan->default_dst.remote_ifindex,
0241b836 3238 NTF_SELF, &f);
c80498e3
ND		 3239 if (err)
3240 return err;
3241 }
3242
3243 err = register_netdevice(dev);
0241b836
RP		 3244 if (err)
3245 goto errout;
3246
3247 err = rtnl_configure_link(dev, NULL);
c80498e3 3248 if (err) {
0241b836
RP		 3249 unregister_netdevice(dev);
3250 goto errout;
c80498e3
ND		 3251 }
3252
0241b836
RP		 3253 /* notify default fdb entry */
3254 if (f)
3255 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH);
3256
c80498e3
ND		 3257 list_add(&vxlan->next, &vn->vxlan_list);
3258 return 0;
0241b836
RP		 3259errout:
3260 if (f)
3261 vxlan_fdb_destroy(vxlan, f, false);
3262 return err;
c80498e3
ND		 3263}
3264
8bcdc4f3
RP		 3265static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3266 struct net_device *dev, struct vxlan_config *conf,
3267 bool changelink)
0dfbdf41 3268{
8bcdc4f3 3269 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3270
8bcdc4f3 3271 memset(conf, 0, sizeof(*conf));
e277de5f 3272
8bcdc4f3
RP		 3273 /* if changelink operation, start with old existing cfg */
3274 if (changelink)
3275 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3276
3277 if (data[IFLA_VXLAN_ID]) {
3278 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3279
3280 if (changelink && (vni != conf->vni))
3281 return -EOPNOTSUPP;
3282 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3283 }
0dfbdf41
TG		 3284
3285 if (data[IFLA_VXLAN_GROUP]) {
ce44a4ae
MS		 3286 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3287 return -EOPNOTSUPP;
3288
8bcdc4f3 3289 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3290 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41
TG		 3291 } else if (data[IFLA_VXLAN_GROUP6]) {
3292 if (!IS_ENABLED(CONFIG_IPV6))
3293 return -EPFNOSUPPORT;
3294
ce44a4ae
MS		 3295 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3296 return -EOPNOTSUPP;
3297
8bcdc4f3
RP		 3298 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3299 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3300 }
3301
3302 if (data[IFLA_VXLAN_LOCAL]) {
ce44a4ae
MS		 3303 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3304 return -EOPNOTSUPP;
3305
8bcdc4f3
RP		 3306 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3307 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3308 } else if (data[IFLA_VXLAN_LOCAL6]) {
3309 if (!IS_ENABLED(CONFIG_IPV6))
3310 return -EPFNOSUPPORT;
3311
ce44a4ae
MS		 3312 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3313 return -EOPNOTSUPP;
3314
0dfbdf41 3315 /* TODO: respect scope id */
8bcdc4f3
RP		 3316 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3317 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3318 }
3319
3320 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3321 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3322
d342894c 3323 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3324 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3325
afb97186 3326 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3327 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3328
72f6d71e
HL		 3329 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3330 if (changelink)
3331 return -EOPNOTSUPP;
3332 conf->flags |= VXLAN_F_TTL_INHERIT;
3333 }
3334
e7f70af1 3335 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3336 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3337 IPV6_FLOWLABEL_MASK;
3338
8bcdc4f3 3339 if (data[IFLA_VXLAN_LEARNING]) {
dc5321d7 3340 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
8bcdc4f3 3341 conf->flags |= VXLAN_F_LEARN;
dc5321d7 3342 else
8bcdc4f3 3343 conf->flags &= ~VXLAN_F_LEARN;
8bcdc4f3
RP		 3344 } else if (!changelink) {
3345 /* default to learn on a new device */
3346 conf->flags |= VXLAN_F_LEARN;
3347 }
d342894c 3348
8bcdc4f3
RP		 3349 if (data[IFLA_VXLAN_AGEING]) {
3350 if (changelink)
3351 return -EOPNOTSUPP;
3352 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3353 }
d342894c 3354
8bcdc4f3
RP		 3355 if (data[IFLA_VXLAN_PROXY]) {
3356 if (changelink)
3357 return -EOPNOTSUPP;
3358 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3359 conf->flags |= VXLAN_F_PROXY;
3360 }
e4f67add 3361
8bcdc4f3
RP		 3362 if (data[IFLA_VXLAN_RSC]) {
3363 if (changelink)
3364 return -EOPNOTSUPP;
3365 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3366 conf->flags |= VXLAN_F_RSC;
3367 }
e4f67add 3368
8bcdc4f3
RP		 3369 if (data[IFLA_VXLAN_L2MISS]) {
3370 if (changelink)
3371 return -EOPNOTSUPP;
3372 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3373 conf->flags |= VXLAN_F_L2MISS;
3374 }
e4f67add 3375
8bcdc4f3
RP		 3376 if (data[IFLA_VXLAN_L3MISS]) {
3377 if (changelink)
3378 return -EOPNOTSUPP;
3379 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3380 conf->flags |= VXLAN_F_L3MISS;
3381 }
e4f67add 3382
8bcdc4f3
RP		 3383 if (data[IFLA_VXLAN_LIMIT]) {
3384 if (changelink)
3385 return -EOPNOTSUPP;
3386 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3387 }
d342894c 3388
8bcdc4f3
RP		 3389 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3390 if (changelink)
3391 return -EOPNOTSUPP;
3392 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3393 conf->flags |= VXLAN_F_COLLECT_METADATA;
3394 }
f8a9b1bc 3395
05f47d69 3396 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3397 if (!changelink) {
3398 const struct ifla_vxlan_port_range *p
3399 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3400 conf->port_min = ntohs(p->low);
3401 conf->port_max = ntohs(p->high);
3402 } else {
3403 return -EOPNOTSUPP;
3404 }
05f47d69 3405 }
3406
8bcdc4f3
RP		 3407 if (data[IFLA_VXLAN_PORT]) {
3408 if (changelink)
3409 return -EOPNOTSUPP;
3410 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3411 }
823aa873 3412
8bcdc4f3
RP		 3413 if (data[IFLA_VXLAN_UDP_CSUM]) {
3414 if (changelink)
3415 return -EOPNOTSUPP;
3416 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3417 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3418 }
359a0ea9 3419
8bcdc4f3
RP		 3420 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3421 if (changelink)
3422 return -EOPNOTSUPP;
3423 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3424 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3425 }
359a0ea9 3426
8bcdc4f3
RP		 3427 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3428 if (changelink)
3429 return -EOPNOTSUPP;
3430 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3431 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3432 }
359a0ea9 3433
8bcdc4f3
RP		 3434 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3435 if (changelink)
3436 return -EOPNOTSUPP;
3437 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3438 conf->flags |= VXLAN_F_REMCSUM_TX;
3439 }
dfd8645e 3440
8bcdc4f3
RP		 3441 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3442 if (changelink)
3443 return -EOPNOTSUPP;
3444 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3445 conf->flags |= VXLAN_F_REMCSUM_RX;
3446 }
3447
3448 if (data[IFLA_VXLAN_GBP]) {
3449 if (changelink)
3450 return -EOPNOTSUPP;
3451 conf->flags |= VXLAN_F_GBP;
3452 }
3453
3454 if (data[IFLA_VXLAN_GPE]) {
3455 if (changelink)
3456 return -EOPNOTSUPP;
3457 conf->flags |= VXLAN_F_GPE;
3458 }
3459
3460 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3461 if (changelink)
3462 return -EOPNOTSUPP;
3463 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3464 }
3465
3466 if (tb[IFLA_MTU]) {
3467 if (changelink)
3468 return -EOPNOTSUPP;
3469 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3470 }
3471
3472 return 0;
3473}
3474
3475static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 3476 struct nlattr *tb[], struct nlattr *data[],
3477 struct netlink_ext_ack *extack)
8bcdc4f3 3478{
8bcdc4f3
RP		 3479 struct vxlan_config conf;
3480 int err;
3481
3482 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3483 if (err)
3484 return err;
3485
653ef6a3 3486 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 3487}
dfd8645e 3488
8bcdc4f3 3489static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 3490 struct nlattr *data[],
3491 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 3492{
3493 struct vxlan_dev *vxlan = netdev_priv(dev);
3494 struct vxlan_rdst *dst = &vxlan->default_dst;
3495 struct vxlan_rdst old_dst;
3496 struct vxlan_config conf;
0241b836 3497 struct vxlan_fdb *f = NULL;
8bcdc4f3
RP		 3498 int err;
3499
3500 err = vxlan_nl2conf(tb, data,
3501 dev, &conf, true);
3502 if (err)
3503 return err;
3511494c 3504
8bcdc4f3 3505 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3506
653ef6a3 3507 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
8bcdc4f3
RP		 3508 if (err)
3509 return err;
0ace2ca8 3510
8bcdc4f3
RP		 3511 /* handle default dst entry */
3512 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3513 spin_lock_bh(&vxlan->hash_lock);
3514 if (!vxlan_addr_any(&old_dst.remote_ip))
3515 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3516 old_dst.remote_ip,
3517 vxlan->cfg.dst_port,
3518 old_dst.remote_vni,
3519 old_dst.remote_vni,
3520 old_dst.remote_ifindex, 0);
3521
3522 if (!vxlan_addr_any(&dst->remote_ip)) {
0241b836 3523 err = vxlan_fdb_create(vxlan, all_zeros_mac,
8bcdc4f3
RP		 3524 &dst->remote_ip,
3525 NUD_REACHABLE | NUD_PERMANENT,
8bcdc4f3
RP		 3526 vxlan->cfg.dst_port,
3527 dst->remote_vni,
3528 dst->remote_vni,
3529 dst->remote_ifindex,
0241b836 3530 NTF_SELF, &f);
8bcdc4f3
RP		 3531 if (err) {
3532 spin_unlock_bh(&vxlan->hash_lock);
3533 return err;
3534 }
0241b836 3535 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH);
8bcdc4f3
RP		 3536 }
3537 spin_unlock_bh(&vxlan->hash_lock);
3538 }
ce577668 3539
8bcdc4f3 3540 return 0;
d342894c 3541}
3542
3543static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3544{
3545 struct vxlan_dev *vxlan = netdev_priv(dev);
3546
8b3f9337
RP		 3547 vxlan_flush(vxlan, true);
3548
58ce31cc 3549 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3550 list_del(&vxlan->next);
d342894c 3551 unregister_netdevice_queue(dev, head);
3552}
3553
3554static size_t vxlan_get_size(const struct net_device *dev)
3555{
3556
3557 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3558 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3559 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3560 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3561 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
8fd78069 3562 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
d342894c 3563 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3564 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3565 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3566 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3567 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3568 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3569 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3570 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3571 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3572 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3573 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3574 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3575 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3576 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3577 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3578 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3579 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3580 0;
3581}
3582
3583static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3584{
3585 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3586 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3587 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3588 .low = htons(vxlan->cfg.port_min),
3589 .high = htons(vxlan->cfg.port_max),
05f47d69 3590 };
d342894c 3591
54bfd872 3592 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3593 goto nla_put_failure;
3594
e4c7ed41
CW		 3595 if (!vxlan_addr_any(&dst->remote_ip)) {
3596 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3597 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3598 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3599 goto nla_put_failure;
3600#if IS_ENABLED(CONFIG_IPV6)
3601 } else {
930345ea
JB		 3602 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3603 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3604 goto nla_put_failure;
3605#endif
3606 }
3607 }
d342894c 3608
c7995c43 3609 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3610 goto nla_put_failure;
3611
0dfbdf41
TG		 3612 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3613 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3614 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3615 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3616 goto nla_put_failure;
3617#if IS_ENABLED(CONFIG_IPV6)
3618 } else {
930345ea 3619 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3620 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3621 goto nla_put_failure;
3622#endif
3623 }
3624 }
d342894c 3625
0dfbdf41 3626 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
8fd78069
HL		 3627 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
3628 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
0dfbdf41 3629 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3630 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 3631 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 3632 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 3633 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS		 3634 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3635 nla_put_u8(skb, IFLA_VXLAN_RSC,
3636 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 3637 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 3638 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 3639 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 3640 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 3641 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 3642 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3643 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3644 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3645 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3646 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 3647 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 3648 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 3649 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 3650 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 3651 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 3652 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 3653 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 3654 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 3655 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 3656 goto nla_put_failure;
3657
05f47d69 3658 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3659 goto nla_put_failure;
3660
dc5321d7 3661 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 3662 nla_put_flag(skb, IFLA_VXLAN_GBP))
3663 goto nla_put_failure;
3664
dc5321d7 3665 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 3666 nla_put_flag(skb, IFLA_VXLAN_GPE))
3667 goto nla_put_failure;
3668
dc5321d7 3669 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 3670 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3671 goto nla_put_failure;
3672
d342894c 3673 return 0;
3674
3675nla_put_failure:
3676 return -EMSGSIZE;
3677}
3678
1728d4fa
ND		 3679static struct net *vxlan_get_link_net(const struct net_device *dev)
3680{
3681 struct vxlan_dev *vxlan = netdev_priv(dev);
3682
3683 return vxlan->net;
3684}
3685
d342894c 3686static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3687 .kind = "vxlan",
3688 .maxtype = IFLA_VXLAN_MAX,
3689 .policy = vxlan_policy,
3690 .priv_size = sizeof(struct vxlan_dev),
3691 .setup = vxlan_setup,
3692 .validate = vxlan_validate,
3693 .newlink = vxlan_newlink,
8bcdc4f3 3694 .changelink = vxlan_changelink,
d342894c 3695 .dellink = vxlan_dellink,
3696 .get_size = vxlan_get_size,
3697 .fill_info = vxlan_fill_info,
1728d4fa 3698 .get_link_net = vxlan_get_link_net,
d342894c 3699};
3700
cf5da330
ND		 3701struct net_device *vxlan_dev_create(struct net *net, const char *name,
3702 u8 name_assign_type,
3703 struct vxlan_config *conf)
3704{
3705 struct nlattr *tb[IFLA_MAX + 1];
3706 struct net_device *dev;
3707 int err;
3708
3709 memset(&tb, 0, sizeof(tb));
3710
3711 dev = rtnl_create_link(net, name, name_assign_type,
3712 &vxlan_link_ops, tb);
3713 if (IS_ERR(dev))
3714 return dev;
3715
653ef6a3 3716 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 3717 if (err < 0) {
3718 free_netdev(dev);
3719 return ERR_PTR(err);
3720 }
3721
3722 err = rtnl_configure_link(dev, NULL);
3723 if (err < 0) {
3724 LIST_HEAD(list_kill);
3725
3726 vxlan_dellink(dev, &list_kill);
3727 unregister_netdevice_many(&list_kill);
3728 return ERR_PTR(err);
3729 }
3730
3731 return dev;
3732}
3733EXPORT_SYMBOL_GPL(vxlan_dev_create);
3734
acaf4e70
DB		 3735static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3736 struct net_device *dev)
3737{
3738 struct vxlan_dev *vxlan, *next;
3739 LIST_HEAD(list_kill);
3740
3741 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3742 struct vxlan_rdst *dst = &vxlan->default_dst;
3743
3744 /* In case we created vxlan device with carrier
3745 * and we loose the carrier due to module unload
3746 * we also need to remove vxlan device. In other
3747 * cases, it's not necessary and remote_ifindex
3748 * is 0 here, so no matches.
3749 */
3750 if (dst->remote_ifindex == dev->ifindex)
3751 vxlan_dellink(vxlan->dev, &list_kill);
3752 }
3753
3754 unregister_netdevice_many(&list_kill);
3755}
3756
b7aade15
HFS		 3757static int vxlan_netdevice_event(struct notifier_block *unused,
3758 unsigned long event, void *ptr)
acaf4e70
DB		 3759{
3760 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3761 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3762
04584957
SD		 3763 if (event == NETDEV_UNREGISTER) {
3764 vxlan_offload_rx_ports(dev, false);
acaf4e70 3765 vxlan_handle_lowerdev_unregister(vn, dev);
04584957
SD		 3766 } else if (event == NETDEV_REGISTER) {
3767 vxlan_offload_rx_ports(dev, true);
3768 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3769 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 3770 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 3771 }
acaf4e70
DB		 3772
3773 return NOTIFY_DONE;
3774}
3775
3776static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3777 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3778};
3779
d342894c 3780static __net_init int vxlan_init_net(struct net *net)
3781{
3782 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3783 unsigned int h;
d342894c 3784
553675fb 3785 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3786 spin_lock_init(&vn->sock_lock);
d342894c 3787
553675fb 3788 for (h = 0; h < PORT_HASH_SIZE; ++h)
3789 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3790
3791 return 0;
3792}
3793
57b61127 3794static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
f01ec1c0
ND		 3795{
3796 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3797 struct vxlan_dev *vxlan, *next;
3798 struct net_device *dev, *aux;
0e4ec5ac 3799 unsigned int h;
f01ec1c0 3800
f01ec1c0
ND		 3801 for_each_netdev_safe(net, dev, aux)
3802 if (dev->rtnl_link_ops == &vxlan_link_ops)
57b61127 3803 unregister_netdevice_queue(dev, head);
f01ec1c0
ND		 3804
3805 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3806 /* If vxlan->dev is in the same netns, it has already been added
3807 * to the list by the previous loop.
3808 */
58ce31cc
TH		 3809 if (!net_eq(dev_net(vxlan->dev), net)) {
3810 gro_cells_destroy(&vxlan->gro_cells);
57b61127 3811 unregister_netdevice_queue(vxlan->dev, head);
58ce31cc 3812 }
f01ec1c0
ND		 3813 }
3814
0e4ec5ac
VA		 3815 for (h = 0; h < PORT_HASH_SIZE; ++h)
3816 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND		 3817}
3818
57b61127
HY		 3819static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
3820{
3821 struct net *net;
3822 LIST_HEAD(list);
3823
3824 rtnl_lock();
3825 list_for_each_entry(net, net_list, exit_list)
3826 vxlan_destroy_tunnels(net, &list);
3827
3828 unregister_netdevice_many(&list);
3829 rtnl_unlock();
3830}
3831
d342894c 3832static struct pernet_operations vxlan_net_ops = {
3833 .init = vxlan_init_net,
57b61127 3834 .exit_batch = vxlan_exit_batch_net,
d342894c 3835 .id = &vxlan_net_id,
3836 .size = sizeof(struct vxlan_net),
3837};
3838
3839static int __init vxlan_init_module(void)
3840{
3841 int rc;
3842
3843 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3844
783c1463 3845 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3846 if (rc)
3847 goto out1;
3848
acaf4e70 3849 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3850 if (rc)
3851 goto out2;
3852
acaf4e70
DB		 3853 rc = rtnl_link_register(&vxlan_link_ops);
3854 if (rc)
3855 goto out3;
d342894c 3856
acaf4e70
DB		 3857 return 0;
3858out3:
3859 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3860out2:
783c1463 3861 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3862out1:
3863 return rc;
3864}
7332a13b 3865late_initcall(vxlan_init_module);
d342894c 3866
3867static void __exit vxlan_cleanup_module(void)
3868{
b7153984 3869 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3870 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3871 unregister_pernet_subsys(&vxlan_net_ops);
3872 /* rcu_barrier() is called by netns */
d342894c 3873}
3874module_exit(vxlan_cleanup_module);
3875
3876MODULE_LICENSE("GPL");
3877MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3878MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3879MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3880MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror