]> 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: refactor verification and application of configuration
[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>
012a5729 29#include <net/vxlan.h>
40d29af0 30
e4c7ed41 31#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 32#include <net/ip6_tunnel.h>
660d98ca 33#include <net/ip6_checksum.h>
e4c7ed41 34#endif
d342894c 35
36#define VXLAN_VERSION "0.1"
37
553675fb 38#define PORT_HASH_BITS 8
39#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 40#define FDB_AGE_DEFAULT 300 /* 5 min */
41#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42
23c578bf 43/* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 45 * for compatibility with early adopters.
23c578bf 46 */
9daaa397
SH		 47static unsigned short vxlan_port __read_mostly = 8472;
48module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 49MODULE_PARM_DESC(udp_port, "Destination UDP port");
50
51static bool log_ecn_error = true;
52module_param(log_ecn_error, bool, 0644);
53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54
c7d03a00 55static unsigned int vxlan_net_id;
0dfbdf41 56static struct rtnl_link_ops vxlan_link_ops;
553675fb 57
7256eac1 58static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 59
205f356d 60static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 61
a53cb29b
MB		 62static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
63
553675fb 64/* per-network namespace private data for this module */
65struct vxlan_net {
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 68 spinlock_t sock_lock;
553675fb 69};
70
d342894c 71/* Forwarding table entry */
72struct vxlan_fdb {
73 struct hlist_node hlist; /* linked list of entries */
74 struct rcu_head rcu;
75 unsigned long updated; /* jiffies */
76 unsigned long used;
3e61aa8f 77 struct list_head remotes;
7177a3b0 78 u8 eth_addr[ETH_ALEN];
d342894c 79 u16 state; /* see ndm_state */
3ad7a4b1 80 __be32 vni;
ae884082 81 u8 flags; /* see ndm_flags */
d342894c 82};
83
d342894c 84/* salt for hash table */
85static u32 vxlan_salt __read_mostly;
86
ee122c79
TG		 87static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
88{
e7030878
TG		 89 return vs->flags & VXLAN_F_COLLECT_METADATA ||
90 ip_tunnel_collect_metadata();
ee122c79
TG		 91}
92
e4c7ed41
CW		 93#if IS_ENABLED(CONFIG_IPV6)
94static inline
95bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
96{
f0ef3126
JB		 97 if (a->sa.sa_family != b->sa.sa_family)
98 return false;
99 if (a->sa.sa_family == AF_INET6)
100 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
101 else
102 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 103}
104
105static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
106{
f0ef3126
JB		 107 if (ipa->sa.sa_family == AF_INET6)
108 return ipv6_addr_any(&ipa->sin6.sin6_addr);
109 else
110 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 111}
112
113static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
114{
f0ef3126
JB		 115 if (ipa->sa.sa_family == AF_INET6)
116 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
117 else
118 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 119}
120
121static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
122{
f0ef3126 123 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 124 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 125 ip->sa.sa_family = AF_INET6;
126 return 0;
127 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 128 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 129 ip->sa.sa_family = AF_INET;
130 return 0;
131 } else {
132 return -EAFNOSUPPORT;
133 }
e4c7ed41
CW		 134}
135
136static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 137 const union vxlan_addr *ip)
e4c7ed41 138{
f0ef3126 139 if (ip->sa.sa_family == AF_INET6)
930345ea 140 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 141 else
930345ea 142 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 143}
144
145#else /* !CONFIG_IPV6 */
146
147static inline
148bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
149{
f0ef3126 150 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 151}
152
153static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
154{
f0ef3126 155 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 156}
157
158static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
159{
f0ef3126 160 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 161}
162
163static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
164{
f0ef3126
JB		 165 if (nla_len(nla) >= sizeof(struct in6_addr)) {
166 return -EAFNOSUPPORT;
167 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 168 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 169 ip->sa.sa_family = AF_INET;
170 return 0;
171 } else {
172 return -EAFNOSUPPORT;
173 }
e4c7ed41
CW		 174}
175
176static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 177 const union vxlan_addr *ip)
e4c7ed41 178{
930345ea 179 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 180}
181#endif
182
553675fb 183/* Virtual Network hash table head */
54bfd872 184static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 185{
54bfd872 186 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 187}
188
189/* Socket hash table head */
190static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 191{
192 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
193
553675fb 194 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
195}
196
3e61aa8f
SH		 197/* First remote destination for a forwarding entry.
198 * Guaranteed to be non-NULL because remotes are never deleted.
199 */
5ca5461c 200static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 201{
5ca5461c 202 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
203}
204
205static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
206{
207 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 208}
209
ac5132d1
TG		 210/* Find VXLAN socket based on network namespace, address family and UDP port
211 * and enabled unshareable flags.
212 */
213static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
214 __be16 port, u32 flags)
553675fb 215{
216 struct vxlan_sock *vs;
af33c1ad
TH		 217
218 flags &= VXLAN_F_RCV_FLAGS;
553675fb 219
220 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 221 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 222 vxlan_get_sk_family(vs) == family &&
af33c1ad 223 vs->flags == flags)
553675fb 224 return vs;
225 }
226 return NULL;
d342894c 227}
228
54bfd872 229static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 230{
231 struct vxlan_dev *vxlan;
d342894c 232
b9167b2e
JB		 233 /* For flow based devices, map all packets to VNI 0 */
234 if (vs->flags & VXLAN_F_COLLECT_METADATA)
235 vni = 0;
236
54bfd872
JB		 237 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
238 if (vxlan->default_dst.remote_vni == vni)
d342894c 239 return vxlan;
240 }
241
242 return NULL;
243}
244
5cfccc5a 245/* Look up VNI in a per net namespace table */
54bfd872 246static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 247 sa_family_t family, __be16 port,
248 u32 flags)
5cfccc5a
PS		 249{
250 struct vxlan_sock *vs;
251
ac5132d1 252 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 253 if (!vs)
254 return NULL;
255
54bfd872 256 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 257}
258
d342894c 259/* Fill in neighbour message in skbuff. */
260static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 261 const struct vxlan_fdb *fdb,
262 u32 portid, u32 seq, int type, unsigned int flags,
263 const struct vxlan_rdst *rdst)
d342894c 264{
265 unsigned long now = jiffies;
266 struct nda_cacheinfo ci;
267 struct nlmsghdr *nlh;
268 struct ndmsg *ndm;
e4f67add 269 bool send_ip, send_eth;
d342894c 270
271 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
272 if (nlh == NULL)
273 return -EMSGSIZE;
274
275 ndm = nlmsg_data(nlh);
276 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 277
278 send_eth = send_ip = true;
279
280 if (type == RTM_GETNEIGH) {
e4c7ed41 281 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 282 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 283 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 284 } else
285 ndm->ndm_family = AF_BRIDGE;
d342894c 286 ndm->ndm_state = fdb->state;
287 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 288 ndm->ndm_flags = fdb->flags;
545469f7 289 ndm->ndm_type = RTN_UNICAST;
d342894c 290
193523bf 291 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 292 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 293 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 294 goto nla_put_failure;
295
e4f67add 296 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 297 goto nla_put_failure;
298
e4c7ed41 299 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 300 goto nla_put_failure;
301
0dfbdf41 302 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 303 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
304 goto nla_put_failure;
c7995c43 305 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 306 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 307 goto nla_put_failure;
3ad7a4b1
RP		 308 if ((vxlan->flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
309 nla_put_u32(skb, NDA_SRC_VNI,
310 be32_to_cpu(fdb->vni)))
311 goto nla_put_failure;
6681712d
DS		 312 if (rdst->remote_ifindex &&
313 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 314 goto nla_put_failure;
315
316 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
317 ci.ndm_confirmed = 0;
318 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
319 ci.ndm_refcnt = 0;
320
321 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
322 goto nla_put_failure;
323
053c095a
JB		 324 nlmsg_end(skb, nlh);
325 return 0;
d342894c 326
327nla_put_failure:
328 nlmsg_cancel(skb, nlh);
329 return -EMSGSIZE;
330}
331
332static inline size_t vxlan_nlmsg_size(void)
333{
334 return NLMSG_ALIGN(sizeof(struct ndmsg))
335 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 336 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 337 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 338 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
339 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 340 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 341 + nla_total_size(sizeof(struct nda_cacheinfo));
342}
343
9e4b93f9
ND		 344static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
345 struct vxlan_rdst *rd, int type)
d342894c 346{
347 struct net *net = dev_net(vxlan->dev);
348 struct sk_buff *skb;
349 int err = -ENOBUFS;
350
351 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
352 if (skb == NULL)
353 goto errout;
354
9e4b93f9 355 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 356 if (err < 0) {
357 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
358 WARN_ON(err == -EMSGSIZE);
359 kfree_skb(skb);
360 goto errout;
361 }
362
363 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
364 return;
365errout:
366 if (err < 0)
367 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
368}
369
e4c7ed41 370static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 371{
372 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 373 struct vxlan_fdb f = {
374 .state = NUD_STALE,
375 };
376 struct vxlan_rdst remote = {
e4c7ed41 377 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 378 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 379 };
3e61aa8f 380
9e4b93f9 381 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 382}
383
384static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
385{
bb3fd687
SH		 386 struct vxlan_fdb f = {
387 .state = NUD_STALE,
388 };
9e4b93f9 389 struct vxlan_rdst remote = { };
e4f67add 390
e4f67add
DS		 391 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 394}
395
d342894c 396/* Hash Ethernet address */
397static u32 eth_hash(const unsigned char *addr)
398{
399 u64 value = get_unaligned((u64 *)addr);
400
401 /* only want 6 bytes */
402#ifdef __BIG_ENDIAN
d342894c 403 value >>= 16;
321fb991 404#else
405 value <<= 16;
d342894c 406#endif
407 return hash_64(value, FDB_HASH_BITS);
408}
409
3ad7a4b1
RP		 410static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
411{
412 /* use 1 byte of OUI and 3 bytes of NIC */
413 u32 key = get_unaligned((u32 *)(addr + 2));
414
415 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
416}
417
d342894c 418/* Hash chain to use given mac address */
419static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 420 const u8 *mac, __be32 vni)
d342894c 421{
3ad7a4b1
RP		 422 if (vxlan->flags & VXLAN_F_COLLECT_METADATA)
423 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
424 else
425 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 426}
427
428/* Look up Ethernet address in forwarding table */
014be2c8 429static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 430 const u8 *mac, __be32 vni)
d342894c 431{
3ad7a4b1 432 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 433 struct vxlan_fdb *f;
d342894c 434
b67bfe0d 435 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1
RP		 436 if (ether_addr_equal(mac, f->eth_addr)) {
437 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
438 if (vni == f->vni)
439 return f;
440 } else {
441 return f;
442 }
443 }
d342894c 444 }
445
446 return NULL;
447}
448
014be2c8 449static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 450 const u8 *mac, __be32 vni)
014be2c8
SS		 451{
452 struct vxlan_fdb *f;
453
3ad7a4b1 454 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS		 455 if (f)
456 f->used = jiffies;
457
458 return f;
459}
460
a5e7c10a
MR		 461/* caller should hold vxlan->hash_lock */
462static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 463 union vxlan_addr *ip, __be16 port,
54bfd872 464 __be32 vni, __u32 ifindex)
6681712d 465{
3e61aa8f 466 struct vxlan_rdst *rd;
6681712d 467
3e61aa8f 468 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 469 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 470 rd->remote_port == port &&
471 rd->remote_vni == vni &&
472 rd->remote_ifindex == ifindex)
a5e7c10a 473 return rd;
6681712d 474 }
3e61aa8f 475
a5e7c10a
MR		 476 return NULL;
477}
478
906dc186
TR		 479/* Replace destination of unicast mac */
480static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 481 union vxlan_addr *ip, __be16 port, __be32 vni,
482 __u32 ifindex)
906dc186
TR		 483{
484 struct vxlan_rdst *rd;
485
486 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
487 if (rd)
488 return 0;
489
490 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
491 if (!rd)
492 return 0;
0c1d70af
PA		 493
494 dst_cache_reset(&rd->dst_cache);
e4c7ed41 495 rd->remote_ip = *ip;
906dc186
TR		 496 rd->remote_port = port;
497 rd->remote_vni = vni;
498 rd->remote_ifindex = ifindex;
499 return 1;
500}
501
a5e7c10a
MR		 502/* Add/update destinations for multicast */
503static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 504 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 505 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 506{
507 struct vxlan_rdst *rd;
508
509 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
510 if (rd)
511 return 0;
512
6681712d
DS		 513 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
514 if (rd == NULL)
515 return -ENOBUFS;
0c1d70af
PA		 516
517 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
518 kfree(rd);
519 return -ENOBUFS;
520 }
521
e4c7ed41 522 rd->remote_ip = *ip;
6681712d
DS		 523 rd->remote_port = port;
524 rd->remote_vni = vni;
525 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 526
527 list_add_tail_rcu(&rd->list, &f->remotes);
528
9e4b93f9 529 *rdp = rd;
6681712d
DS		 530 return 1;
531}
532
dfd8645e
TH		 533static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
534 unsigned int off,
535 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 536 __be32 vni_field,
537 struct gro_remcsum *grc,
0ace2ca8 538 bool nopartial)
dfd8645e 539{
b7fe10e5 540 size_t start, offset;
dfd8645e
TH		 541
542 if (skb->remcsum_offload)
b7fe10e5 543 return vh;
dfd8645e
TH		 544
545 if (!NAPI_GRO_CB(skb)->csum_valid)
546 return NULL;
547
54bfd872
JB		 548 start = vxlan_rco_start(vni_field);
549 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 550
b7fe10e5
TH		 551 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
552 start, offset, grc, nopartial);
dfd8645e
TH		 553
554 skb->remcsum_offload = 1;
555
556 return vh;
557}
558
5602c48c
TH		 559static struct sk_buff **vxlan_gro_receive(struct sock *sk,
560 struct sk_buff **head,
561 struct sk_buff *skb)
dc01e7d3
OG		 562{
563 struct sk_buff *p, **pp = NULL;
564 struct vxlanhdr *vh, *vh2;
9b174d88 565 unsigned int hlen, off_vx;
dc01e7d3 566 int flush = 1;
5602c48c 567 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 568 __be32 flags;
26c4f7da
TH		 569 struct gro_remcsum grc;
570
571 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 572
573 off_vx = skb_gro_offset(skb);
574 hlen = off_vx + sizeof(*vh);
575 vh = skb_gro_header_fast(skb, off_vx);
576 if (skb_gro_header_hard(skb, hlen)) {
577 vh = skb_gro_header_slow(skb, hlen, off_vx);
578 if (unlikely(!vh))
579 goto out;
580 }
dc01e7d3 581
dfd8645e
TH		 582 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
583
54bfd872 584 flags = vh->vx_flags;
dfd8645e
TH		 585
586 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
587 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 588 vh->vx_vni, &grc,
0ace2ca8
TH		 589 !!(vs->flags &
590 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 591
592 if (!vh)
593 goto out;
594 }
595
b7fe10e5
TH		 596 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
597
dc01e7d3
OG		 598 for (p = *head; p; p = p->next) {
599 if (!NAPI_GRO_CB(p)->same_flow)
600 continue;
601
602 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 603 if (vh->vx_flags != vh2->vx_flags ||
604 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 605 NAPI_GRO_CB(p)->same_flow = 0;
606 continue;
607 }
dc01e7d3
OG		 608 }
609
fcd91dd4 610 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 611 flush = 0;
dc01e7d3 612
dc01e7d3 613out:
26c4f7da 614 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 615 NAPI_GRO_CB(skb)->flush |= flush;
616
617 return pp;
618}
619
5602c48c 620static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 621{
229740c6
JR		 622 /* Sets 'skb->inner_mac_header' since we are always called with
623 * 'skb->encapsulation' set.
624 */
9b174d88 625 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 626}
627
d342894c 628/* Add new entry to forwarding table -- assumes lock held */
629static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 630 const u8 *mac, union vxlan_addr *ip,
6681712d 631 __u16 state, __u16 flags,
3ad7a4b1
RP		 632 __be16 port, __be32 src_vni, __be32 vni,
633 __u32 ifindex, __u8 ndm_flags)
d342894c 634{
9e4b93f9 635 struct vxlan_rdst *rd = NULL;
d342894c 636 struct vxlan_fdb *f;
637 int notify = 0;
17b46365 638 int rc;
d342894c 639
3ad7a4b1 640 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 641 if (f) {
642 if (flags & NLM_F_EXCL) {
643 netdev_dbg(vxlan->dev,
644 "lost race to create %pM\n", mac);
645 return -EEXIST;
646 }
647 if (f->state != state) {
648 f->state = state;
649 f->updated = jiffies;
650 notify = 1;
651 }
ae884082
DS		 652 if (f->flags != ndm_flags) {
653 f->flags = ndm_flags;
654 f->updated = jiffies;
655 notify = 1;
656 }
906dc186
TR		 657 if ((flags & NLM_F_REPLACE)) {
658 /* Only change unicasts */
659 if (!(is_multicast_ether_addr(f->eth_addr) ||
660 is_zero_ether_addr(f->eth_addr))) {
60840429 661 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 662 ifindex);
906dc186
TR		 663 } else
664 return -EOPNOTSUPP;
665 }
6681712d 666 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 667 (is_multicast_ether_addr(f->eth_addr) ||
668 is_zero_ether_addr(f->eth_addr))) {
17b46365 669 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 670
671 if (rc < 0)
672 return rc;
673 notify |= rc;
674 }
d342894c 675 } else {
676 if (!(flags & NLM_F_CREATE))
677 return -ENOENT;
678
0dfbdf41
TG		 679 if (vxlan->cfg.addrmax &&
680 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 681 return -ENOSPC;
682
906dc186
TR		 683 /* Disallow replace to add a multicast entry */
684 if ((flags & NLM_F_REPLACE) &&
685 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
686 return -EOPNOTSUPP;
687
e4c7ed41 688 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 689 f = kmalloc(sizeof(*f), GFP_ATOMIC);
690 if (!f)
691 return -ENOMEM;
692
693 notify = 1;
d342894c 694 f->state = state;
ae884082 695 f->flags = ndm_flags;
d342894c 696 f->updated = f->used = jiffies;
3ad7a4b1 697 f->vni = src_vni;
3e61aa8f 698 INIT_LIST_HEAD(&f->remotes);
d342894c 699 memcpy(f->eth_addr, mac, ETH_ALEN);
700
17b46365
HY		 701 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
702 if (rc < 0) {
703 kfree(f);
704 return rc;
705 }
3e61aa8f 706
d342894c 707 ++vxlan->addrcnt;
708 hlist_add_head_rcu(&f->hlist,
3ad7a4b1 709 vxlan_fdb_head(vxlan, mac, src_vni));
d342894c 710 }
711
9e4b93f9
ND		 712 if (notify) {
713 if (rd == NULL)
714 rd = first_remote_rtnl(f);
715 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
716 }
d342894c 717
718 return 0;
719}
720
6706c82e 721static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 722{
723 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 724 struct vxlan_rdst *rd, *nd;
6681712d 725
0c1d70af
PA		 726 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
727 dst_cache_destroy(&rd->dst_cache);
6681712d 728 kfree(rd);
0c1d70af 729 }
6681712d
DS		 730 kfree(f);
731}
732
d342894c 733static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
734{
735 netdev_dbg(vxlan->dev,
736 "delete %pM\n", f->eth_addr);
737
738 --vxlan->addrcnt;
9e4b93f9 739 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 740
741 hlist_del_rcu(&f->hlist);
6681712d 742 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 743}
744
35cf2845
LR		 745static void vxlan_dst_free(struct rcu_head *head)
746{
747 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
748
749 dst_cache_destroy(&rd->dst_cache);
750 kfree(rd);
751}
752
753static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
754 struct vxlan_rdst *rd)
755{
756 list_del_rcu(&rd->list);
757 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
758 call_rcu(&rd->rcu, vxlan_dst_free);
759}
760
f0b074be 761static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 762 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
763 __be32 *vni, u32 *ifindex)
d342894c 764{
6681712d 765 struct net *net = dev_net(vxlan->dev);
e4c7ed41 766 int err;
d342894c 767
f0b074be 768 if (tb[NDA_DST]) {
e4c7ed41
CW		 769 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
770 if (err)
771 return err;
f0b074be 772 } else {
e4c7ed41
CW		 773 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
774 if (remote->sa.sa_family == AF_INET) {
775 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
776 ip->sa.sa_family = AF_INET;
777#if IS_ENABLED(CONFIG_IPV6)
778 } else {
779 ip->sin6.sin6_addr = in6addr_any;
780 ip->sa.sa_family = AF_INET6;
781#endif
782 }
f0b074be 783 }
d342894c 784
6681712d 785 if (tb[NDA_PORT]) {
73cf3317 786 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 787 return -EINVAL;
f0b074be
MR		 788 *port = nla_get_be16(tb[NDA_PORT]);
789 } else {
0dfbdf41 790 *port = vxlan->cfg.dst_port;
f0b074be 791 }
6681712d
DS		 792
793 if (tb[NDA_VNI]) {
794 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
795 return -EINVAL;
54bfd872 796 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 797 } else {
798 *vni = vxlan->default_dst.remote_vni;
799 }
6681712d 800
3ad7a4b1
RP		 801 if (tb[NDA_SRC_VNI]) {
802 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
803 return -EINVAL;
804 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
805 } else {
806 *src_vni = vxlan->default_dst.remote_vni;
807 }
808
6681712d 809 if (tb[NDA_IFINDEX]) {
5abb0029 810 struct net_device *tdev;
6681712d
DS		 811
812 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
813 return -EINVAL;
f0b074be 814 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 815 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 816 if (!tdev)
6681712d 817 return -EADDRNOTAVAIL;
f0b074be
MR		 818 } else {
819 *ifindex = 0;
820 }
821
822 return 0;
823}
824
825/* Add static entry (via netlink) */
826static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
827 struct net_device *dev,
f6f6424b 828 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 829{
830 struct vxlan_dev *vxlan = netdev_priv(dev);
831 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 832 union vxlan_addr ip;
f0b074be 833 __be16 port;
3ad7a4b1 834 __be32 src_vni, vni;
54bfd872 835 u32 ifindex;
f0b074be
MR		 836 int err;
837
838 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
839 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
840 ndm->ndm_state);
841 return -EINVAL;
842 }
843
844 if (tb[NDA_DST] == NULL)
845 return -EINVAL;
846
3ad7a4b1 847 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 848 if (err)
849 return err;
6681712d 850
5933a7bb
MR		 851 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
852 return -EAFNOSUPPORT;
853
d342894c 854 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 855 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 856 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 857 spin_unlock_bh(&vxlan->hash_lock);
858
859 return err;
860}
861
3ad7a4b1
RP		 862static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
863 const unsigned char *addr, union vxlan_addr ip,
864 __be16 port, __be32 src_vni, u32 vni, u32 ifindex,
865 u16 vid)
d342894c 866{
d342894c 867 struct vxlan_fdb *f;
bc7892ba 868 struct vxlan_rdst *rd = NULL;
3ad7a4b1 869 int err = -ENOENT;
bc7892ba 870
3ad7a4b1 871 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 872 if (!f)
3ad7a4b1 873 return err;
bc7892ba 874
e4c7ed41
CW		 875 if (!vxlan_addr_any(&ip)) {
876 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 877 if (!rd)
878 goto out;
879 }
880
bc7892ba
MR		 881 /* remove a destination if it's not the only one on the list,
882 * otherwise destroy the fdb entry
883 */
884 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 885 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 886 goto out;
d342894c 887 }
bc7892ba
MR		 888
889 vxlan_fdb_destroy(vxlan, f);
890
891out:
3ad7a4b1
RP		 892 return 0;
893}
894
895/* Delete entry (via netlink) */
896static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
897 struct net_device *dev,
898 const unsigned char *addr, u16 vid)
899{
900 struct vxlan_dev *vxlan = netdev_priv(dev);
901 union vxlan_addr ip;
902 __be32 src_vni, vni;
903 __be16 port;
904 u32 ifindex;
905 int err;
906
907 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
908 if (err)
909 return err;
910
911 spin_lock_bh(&vxlan->hash_lock);
912 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
913 vid);
d342894c 914 spin_unlock_bh(&vxlan->hash_lock);
915
916 return err;
917}
918
919/* Dump forwarding table */
920static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 921 struct net_device *dev,
d297653d 922 struct net_device *filter_dev, int *idx)
d342894c 923{
924 struct vxlan_dev *vxlan = netdev_priv(dev);
925 unsigned int h;
d297653d 926 int err = 0;
d342894c 927
928 for (h = 0; h < FDB_HASH_SIZE; ++h) {
929 struct vxlan_fdb *f;
d342894c 930
b67bfe0d 931 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 932 struct vxlan_rdst *rd;
6681712d 933
3e61aa8f 934 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 935 if (*idx < cb->args[2])
07a51cd3
AW		 936 goto skip;
937
6681712d
DS		 938 err = vxlan_fdb_info(skb, vxlan, f,
939 NETLINK_CB(cb->skb).portid,
940 cb->nlh->nlmsg_seq,
941 RTM_NEWNEIGH,
942 NLM_F_MULTI, rd);
d297653d 943 if (err < 0)
3e61aa8f 944 goto out;
3e61aa8f 945skip:
d297653d 946 *idx += 1;
07a51cd3 947 }
d342894c 948 }
949 }
3e61aa8f 950out:
d297653d 951 return err;
d342894c 952}
953
954/* Watch incoming packets to learn mapping between Ethernet address
955 * and Tunnel endpoint.
c4b49512 956 * Return true if packet is bogus and should be dropped.
d342894c 957 */
26a41ae6 958static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1
RP		 959 union vxlan_addr *src_ip, const u8 *src_mac,
960 __be32 vni)
d342894c 961{
962 struct vxlan_dev *vxlan = netdev_priv(dev);
963 struct vxlan_fdb *f;
d342894c 964
3ad7a4b1 965 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 966 if (likely(f)) {
5ca5461c 967 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 968
e4c7ed41 969 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 970 return false;
971
972 /* Don't migrate static entries, drop packets */
e0090a9e 973 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 974 return true;
d342894c 975
976 if (net_ratelimit())
977 netdev_info(dev,
e4c7ed41 978 "%pM migrated from %pIS to %pIS\n",
a4870f79 979 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 980
e4c7ed41 981 rdst->remote_ip = *src_ip;
d342894c 982 f->updated = jiffies;
9e4b93f9 983 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 984 } else {
985 /* learned new entry */
986 spin_lock(&vxlan->hash_lock);
3bf74b1a 987
988 /* close off race between vxlan_flush and incoming packets */
989 if (netif_running(dev))
990 vxlan_fdb_create(vxlan, src_mac, src_ip,
991 NUD_REACHABLE,
992 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 993 vxlan->cfg.dst_port,
3ad7a4b1 994 vni,
3bf74b1a 995 vxlan->default_dst.remote_vni,
996 0, NTF_SELF);
d342894c 997 spin_unlock(&vxlan->hash_lock);
998 }
26a41ae6 999
1000 return false;
d342894c 1001}
1002
d342894c 1003/* See if multicast group is already in use by other ID */
95ab0991 1004static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1005{
553675fb 1006 struct vxlan_dev *vxlan;
c6fcc4fc 1007 struct vxlan_sock *sock4;
4053ab1b
AB		 1008#if IS_ENABLED(CONFIG_IPV6)
1009 struct vxlan_sock *sock6;
1010#endif
b1be00a6 1011 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1012
c6fcc4fc 1013 sock4 = rtnl_dereference(dev->vn4_sock);
1014
95ab0991
G		 1015 /* The vxlan_sock is only used by dev, leaving group has
1016 * no effect on other vxlan devices.
1017 */
c6fcc4fc 1018 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
95ab0991 1019 return false;
b1be00a6 1020#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1021 sock6 = rtnl_dereference(dev->vn6_sock);
1022 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1023 return false;
1024#endif
95ab0991 1025
553675fb 1026 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1027 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1028 continue;
d342894c 1029
c6fcc4fc 1030 if (family == AF_INET &&
1031 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1032 continue;
b1be00a6 1033#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1034 if (family == AF_INET6 &&
1035 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1036 continue;
1037#endif
95ab0991
G		 1038
1039 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1040 &dev->default_dst.remote_ip))
1041 continue;
1042
1043 if (vxlan->default_dst.remote_ifindex !=
1044 dev->default_dst.remote_ifindex)
1045 continue;
1046
1047 return true;
553675fb 1048 }
d342894c 1049
1050 return false;
1051}
1052
544a773a 1053static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1054{
b1be00a6 1055 struct vxlan_net *vn;
012a5729 1056
b1be00a6 1057 if (!vs)
544a773a 1058 return false;
7c47cedf 1059 if (!atomic_dec_and_test(&vs->refcnt))
544a773a 1060 return false;
d342894c 1061
b1be00a6 1062 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1063 spin_lock(&vn->sock_lock);
7c47cedf 1064 hlist_del_rcu(&vs->hlist);
e7b3db5e 1065 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1066 (vs->flags & VXLAN_F_GPE) ?
1067 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1068 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1069 spin_unlock(&vn->sock_lock);
1070
544a773a 1071 return true;
d342894c 1072}
1073
b1be00a6
JB		 1074static void vxlan_sock_release(struct vxlan_dev *vxlan)
1075{
c6fcc4fc 1076 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1077#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1078 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1079
57d88182 1080 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1081#endif
1082
57d88182 1083 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1084 synchronize_net();
1085
a53cb29b
MB		 1086 vxlan_vs_del_dev(vxlan);
1087
c6fcc4fc 1088 if (__vxlan_sock_release_prep(sock4)) {
1089 udp_tunnel_sock_release(sock4->sock);
1090 kfree(sock4);
544a773a
HFS		 1091 }
1092
1093#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1094 if (__vxlan_sock_release_prep(sock6)) {
1095 udp_tunnel_sock_release(sock6->sock);
1096 kfree(sock6);
544a773a 1097 }
b1be00a6
JB		 1098#endif
1099}
1100
56ef9c90 1101/* Update multicast group membership when first VNI on
c4b49512 1102 * multicast address is brought up
7c47cedf 1103 */
56ef9c90 1104static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1105{
b1be00a6 1106 struct sock *sk;
e4c7ed41
CW		 1107 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1108 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1109 int ret = -EINVAL;
d342894c 1110
e4c7ed41 1111 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1112 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1113 struct ip_mreqn mreq = {
1114 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1115 .imr_ifindex = ifindex,
1116 };
1117
c6fcc4fc 1118 sk = sock4->sock->sk;
b1be00a6 1119 lock_sock(sk);
56ef9c90 1120 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1121 release_sock(sk);
e4c7ed41
CW		 1122#if IS_ENABLED(CONFIG_IPV6)
1123 } else {
c6fcc4fc 1124 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1125
1126 sk = sock6->sock->sk;
b1be00a6 1127 lock_sock(sk);
56ef9c90
MRL		 1128 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1129 &ip->sin6.sin6_addr);
b1be00a6 1130 release_sock(sk);
e4c7ed41
CW		 1131#endif
1132 }
3fc2de2f 1133
56ef9c90 1134 return ret;
3fc2de2f 1135}
1136
1137/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1138static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1139{
b1be00a6 1140 struct sock *sk;
e4c7ed41
CW		 1141 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1142 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1143 int ret = -EINVAL;
3fc2de2f 1144
e4c7ed41 1145 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1146 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1147 struct ip_mreqn mreq = {
1148 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1149 .imr_ifindex = ifindex,
1150 };
1151
c6fcc4fc 1152 sk = sock4->sock->sk;
b1be00a6 1153 lock_sock(sk);
56ef9c90 1154 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1155 release_sock(sk);
e4c7ed41
CW		 1156#if IS_ENABLED(CONFIG_IPV6)
1157 } else {
c6fcc4fc 1158 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1159
1160 sk = sock6->sock->sk;
b1be00a6 1161 lock_sock(sk);
56ef9c90
MRL		 1162 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1163 &ip->sin6.sin6_addr);
b1be00a6 1164 release_sock(sk);
e4c7ed41
CW		 1165#endif
1166 }
d342894c 1167
56ef9c90 1168 return ret;
d342894c 1169}
1170
f14ecebb
JB		 1171static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1172 struct sk_buff *skb, u32 vxflags)
dfd8645e 1173{
7d34fa75 1174 size_t start, offset;
dfd8645e 1175
f14ecebb
JB		 1176 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1177 goto out;
b7fe10e5 1178
f14ecebb
JB		 1179 start = vxlan_rco_start(unparsed->vx_vni);
1180 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1181
7d34fa75 1182 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1183 return false;
dfd8645e 1184
be5cfeab
JB		 1185 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1186 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1187out:
1188 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1189 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1190 return true;
dfd8645e
TH		 1191}
1192
f14ecebb 1193static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1194 struct sk_buff *skb, u32 vxflags,
10a5af23 1195 struct vxlan_metadata *md)
3288af08 1196{
f14ecebb 1197 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1198 struct metadata_dst *tun_dst;
f14ecebb
JB		 1199
1200 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1201 goto out;
3288af08 1202
3288af08
JB		 1203 md->gbp = ntohs(gbp->policy_id);
1204
10a5af23 1205 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1206 if (tun_dst) {
3288af08 1207 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1208 tun_dst->u.tun_info.options_len = sizeof(*md);
1209 }
3288af08
JB		 1210 if (gbp->dont_learn)
1211 md->gbp |= VXLAN_GBP_DONT_LEARN;
1212
1213 if (gbp->policy_applied)
1214 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1215
64f87d36
JB		 1216 /* In flow-based mode, GBP is carried in dst_metadata */
1217 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1218 skb->mark = md->gbp;
f14ecebb
JB		 1219out:
1220 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1221}
1222
e1e5314d 1223static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1224 __be16 *protocol,
e1e5314d
JB		 1225 struct sk_buff *skb, u32 vxflags)
1226{
1227 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1228
1229 /* Need to have Next Protocol set for interfaces in GPE mode. */
1230 if (!gpe->np_applied)
1231 return false;
1232 /* "The initial version is 0. If a receiver does not support the
1233 * version indicated it MUST drop the packet.
1234 */
1235 if (gpe->version != 0)
1236 return false;
1237 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1238 * processing MUST occur." However, we don't implement OAM
1239 * processing, thus drop the packet.
1240 */
1241 if (gpe->oam_flag)
1242 return false;
1243
1244 switch (gpe->next_protocol) {
1245 case VXLAN_GPE_NP_IPV4:
1246 *protocol = htons(ETH_P_IP);
1247 break;
1248 case VXLAN_GPE_NP_IPV6:
1249 *protocol = htons(ETH_P_IPV6);
1250 break;
1251 case VXLAN_GPE_NP_ETHERNET:
1252 *protocol = htons(ETH_P_TEB);
1253 break;
1254 default:
1255 return false;
1256 }
1257
1258 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1259 return true;
1260}
1261
1ab016e2
JB		 1262static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1263 struct vxlan_sock *vs,
3ad7a4b1 1264 struct sk_buff *skb, __be32 vni)
614732ea 1265{
614732ea 1266 union vxlan_addr saddr;
614732ea 1267
614732ea 1268 skb_reset_mac_header(skb);
614732ea
TG		 1269 skb->protocol = eth_type_trans(skb, vxlan->dev);
1270 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1271
1272 /* Ignore packet loops (and multicast echo) */
1273 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1274 return false;
614732ea 1275
760c6805 1276 /* Get address from the outer IP header */
ce212d0f 1277 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1278 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1279 saddr.sa.sa_family = AF_INET;
1280#if IS_ENABLED(CONFIG_IPV6)
1281 } else {
1ab016e2 1282 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1283 saddr.sa.sa_family = AF_INET6;
1284#endif
1285 }
1286
1ab016e2 1287 if ((vxlan->flags & VXLAN_F_LEARN) &&
3ad7a4b1 1288 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, vni))
1ab016e2
JB		 1289 return false;
1290
1291 return true;
1292}
1293
760c6805
JB		 1294static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1295 struct sk_buff *skb)
1296{
1297 int err = 0;
1298
1299 if (vxlan_get_sk_family(vs) == AF_INET)
1300 err = IP_ECN_decapsulate(oiph, skb);
1301#if IS_ENABLED(CONFIG_IPV6)
1302 else
1303 err = IP6_ECN_decapsulate(oiph, skb);
1304#endif
1305
1306 if (unlikely(err) && log_ecn_error) {
1307 if (vxlan_get_sk_family(vs) == AF_INET)
1308 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1309 &((struct iphdr *)oiph)->saddr,
1310 ((struct iphdr *)oiph)->tos);
1311 else
1312 net_info_ratelimited("non-ECT from %pI6\n",
1313 &((struct ipv6hdr *)oiph)->saddr);
1314 }
1315 return err <= 1;
1316}
1317
d342894c 1318/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1319static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1320{
f2d1968e 1321 struct pcpu_sw_netstats *stats;
c9e78efb 1322 struct vxlan_dev *vxlan;
5cfccc5a 1323 struct vxlan_sock *vs;
f14ecebb 1324 struct vxlanhdr unparsed;
ee122c79
TG		 1325 struct vxlan_metadata _md;
1326 struct vxlan_metadata *md = &_md;
61618eea 1327 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1328 bool raw_proto = false;
f2d1968e 1329 void *oiph;
3ad7a4b1 1330 __be32 vni = 0;
d342894c 1331
e1e5314d 1332 /* Need UDP and VXLAN header to be present */
7ce04758 1333 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1334 goto drop;
d342894c 1335
f14ecebb 1336 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1337 /* VNI flag always required to be set */
1338 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1339 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1340 ntohl(vxlan_hdr(skb)->vx_flags),
1341 ntohl(vxlan_hdr(skb)->vx_vni));
1342 /* Return non vxlan pkt */
e5aed006 1343 goto drop;
d342894c 1344 }
288b01c8
JB		 1345 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1346 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1347
559835ea 1348 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1349 if (!vs)
d342894c 1350 goto drop;
d342894c 1351
3ad7a4b1
RP		 1352 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1353
3ad7a4b1 1354 vxlan = vxlan_vs_find_vni(vs, vni);
c9e78efb
JB		 1355 if (!vxlan)
1356 goto drop;
1357
e1e5314d
JB		 1358 /* For backwards compatibility, only allow reserved fields to be
1359 * used by VXLAN extensions if explicitly requested.
1360 */
1361 if (vs->flags & VXLAN_F_GPE) {
1362 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1363 goto drop;
1364 raw_proto = true;
1365 }
1366
1367 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1368 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1369 goto drop;
c9e78efb 1370
ee122c79 1371 if (vxlan_collect_metadata(vs)) {
10a5af23 1372 struct metadata_dst *tun_dst;
07dabf20 1373
c29a70d2 1374 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1375 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1376
ee122c79
TG		 1377 if (!tun_dst)
1378 goto drop;
1379
0f1b7354 1380 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1381
1382 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1383 } else {
1384 memset(md, 0, sizeof(*md));
1385 }
1386
f14ecebb
JB		 1387 if (vs->flags & VXLAN_F_REMCSUM_RX)
1388 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1389 goto drop;
1390 if (vs->flags & VXLAN_F_GBP)
10a5af23 1391 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1392 /* Note that GBP and GPE can never be active together. This is
1393 * ensured in vxlan_dev_configure.
1394 */
3511494c 1395
f14ecebb 1396 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1397 /* If there are any unprocessed flags remaining treat
1398 * this as a malformed packet. This behavior diverges from
1399 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1400 * in reserved fields are to be ignored. The approach here
c4b49512 1401 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1402 * is more robust and provides a little more security in
1403 * adding extensions to VXLAN.
1404 */
288b01c8 1405 goto drop;
3bf39475
TH		 1406 }
1407
e1e5314d 1408 if (!raw_proto) {
3ad7a4b1 1409 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1410 goto drop;
1411 } else {
8be0cfa4 1412 skb_reset_mac_header(skb);
e1e5314d
JB		 1413 skb->dev = vxlan->dev;
1414 skb->pkt_type = PACKET_HOST;
1415 }
f2d1968e 1416
f2d1968e
JB		 1417 oiph = skb_network_header(skb);
1418 skb_reset_network_header(skb);
1419
1420 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1421 ++vxlan->dev->stats.rx_frame_errors;
1422 ++vxlan->dev->stats.rx_errors;
1423 goto drop;
1424 }
1425
1426 stats = this_cpu_ptr(vxlan->dev->tstats);
1427 u64_stats_update_begin(&stats->syncp);
1428 stats->rx_packets++;
1429 stats->rx_bytes += skb->len;
1430 u64_stats_update_end(&stats->syncp);
1431
1432 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1433 return 0;
1434
1435drop:
288b01c8
JB		 1436 /* Consume bad packet */
1437 kfree_skb(skb);
1438 return 0;
5cfccc5a
PS		 1439}
1440
3ad7a4b1 1441static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1442{
1443 struct vxlan_dev *vxlan = netdev_priv(dev);
1444 struct arphdr *parp;
1445 u8 *arpptr, *sha;
1446 __be32 sip, tip;
1447 struct neighbour *n;
1448
1449 if (dev->flags & IFF_NOARP)
1450 goto out;
1451
1452 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1453 dev->stats.tx_dropped++;
1454 goto out;
1455 }
1456 parp = arp_hdr(skb);
1457
1458 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1459 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1460 parp->ar_pro != htons(ETH_P_IP) ||
1461 parp->ar_op != htons(ARPOP_REQUEST) ||
1462 parp->ar_hln != dev->addr_len ||
1463 parp->ar_pln != 4)
1464 goto out;
1465 arpptr = (u8 *)parp + sizeof(struct arphdr);
1466 sha = arpptr;
1467 arpptr += dev->addr_len; /* sha */
1468 memcpy(&sip, arpptr, sizeof(sip));
1469 arpptr += sizeof(sip);
1470 arpptr += dev->addr_len; /* tha */
1471 memcpy(&tip, arpptr, sizeof(tip));
1472
1473 if (ipv4_is_loopback(tip) ||
1474 ipv4_is_multicast(tip))
1475 goto out;
1476
1477 n = neigh_lookup(&arp_tbl, &tip, dev);
1478
1479 if (n) {
e4f67add
DS		 1480 struct vxlan_fdb *f;
1481 struct sk_buff *reply;
1482
1483 if (!(n->nud_state & NUD_CONNECTED)) {
1484 neigh_release(n);
1485 goto out;
1486 }
1487
3ad7a4b1 1488 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1489 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1490 /* bridge-local neighbor */
1491 neigh_release(n);
1492 goto out;
1493 }
1494
1495 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1496 n->ha, sha);
1497
1498 neigh_release(n);
1499
7346135d
DS		 1500 if (reply == NULL)
1501 goto out;
1502
e4f67add
DS		 1503 skb_reset_mac_header(reply);
1504 __skb_pull(reply, skb_network_offset(reply));
1505 reply->ip_summed = CHECKSUM_UNNECESSARY;
1506 reply->pkt_type = PACKET_HOST;
1507
1508 if (netif_rx_ni(reply) == NET_RX_DROP)
1509 dev->stats.rx_dropped++;
e4c7ed41
CW		 1510 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1511 union vxlan_addr ipa = {
1512 .sin.sin_addr.s_addr = tip,
a45e92a5 1513 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1514 };
1515
1516 vxlan_ip_miss(dev, &ipa);
1517 }
e4f67add
DS		 1518out:
1519 consume_skb(skb);
1520 return NETDEV_TX_OK;
1521}
1522
f564f45c 1523#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1524static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1525 struct neighbour *n, bool isrouter)
1526{
1527 struct net_device *dev = request->dev;
1528 struct sk_buff *reply;
1529 struct nd_msg *ns, *na;
1530 struct ipv6hdr *pip6;
1531 u8 *daddr;
1532 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1533 int ns_olen;
1534 int i, len;
1535
f1fb08f6 1536 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1537 return NULL;
1538
1539 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1540 sizeof(*na) + na_olen + dev->needed_tailroom;
1541 reply = alloc_skb(len, GFP_ATOMIC);
1542 if (reply == NULL)
1543 return NULL;
1544
1545 reply->protocol = htons(ETH_P_IPV6);
1546 reply->dev = dev;
1547 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1548 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1549 skb_reset_mac_header(reply);
4b29dba9 1550
f1fb08f6 1551 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1552
1553 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1554 ns_olen = request->len - skb_network_offset(request) -
1555 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1556 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1557 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1558 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1559 break;
1560 }
1561 }
1562
1563 /* Ethernet header */
1564 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1565 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1566 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1567 reply->protocol = htons(ETH_P_IPV6);
1568
1569 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1570 skb_reset_network_header(reply);
4b29dba9
DS		 1571 skb_put(reply, sizeof(struct ipv6hdr));
1572
1573 /* IPv6 header */
1574
1575 pip6 = ipv6_hdr(reply);
1576 memset(pip6, 0, sizeof(struct ipv6hdr));
1577 pip6->version = 6;
1578 pip6->priority = ipv6_hdr(request)->priority;
1579 pip6->nexthdr = IPPROTO_ICMPV6;
1580 pip6->hop_limit = 255;
1581 pip6->daddr = ipv6_hdr(request)->saddr;
1582 pip6->saddr = *(struct in6_addr *)n->primary_key;
1583
1584 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1585 skb_reset_transport_header(reply);
4b29dba9 1586
4b29dba9 1587 /* Neighbor Advertisement */
b080db58 1588 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1589 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1590 na->icmph.icmp6_router = isrouter;
1591 na->icmph.icmp6_override = 1;
1592 na->icmph.icmp6_solicited = 1;
1593 na->target = ns->target;
1594 ether_addr_copy(&na->opt[2], n->ha);
1595 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1596 na->opt[1] = na_olen >> 3;
1597
1598 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1599 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1600 csum_partial(na, sizeof(*na)+na_olen, 0));
1601
1602 pip6->payload_len = htons(sizeof(*na)+na_olen);
1603
1604 skb_push(reply, sizeof(struct ipv6hdr));
1605
1606 reply->ip_summed = CHECKSUM_UNNECESSARY;
1607
1608 return reply;
1609}
1610
3ad7a4b1 1611static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1612{
1613 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1614 struct nd_msg *msg;
f564f45c 1615 const struct ipv6hdr *iphdr;
8dcd81a9 1616 const struct in6_addr *daddr;
4b29dba9
DS		 1617 struct neighbour *n;
1618 struct inet6_dev *in6_dev;
f564f45c
CW		 1619
1620 in6_dev = __in6_dev_get(dev);
1621 if (!in6_dev)
1622 goto out;
1623
f1fb08f6
VB		 1624 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + sizeof(struct nd_msg)))
1625 goto out;
1626
f564f45c 1627 iphdr = ipv6_hdr(skb);
f564f45c
CW		 1628 daddr = &iphdr->daddr;
1629
f1fb08f6 1630 msg = (struct nd_msg *)(iphdr + 1);
f564f45c
CW		 1631 if (msg->icmph.icmp6_code != 0 ||
1632 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1633 goto out;
1634
4b29dba9
DS		 1635 if (ipv6_addr_loopback(daddr) ||
1636 ipv6_addr_is_multicast(&msg->target))
1637 goto out;
1638
1639 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1640
1641 if (n) {
1642 struct vxlan_fdb *f;
4b29dba9 1643 struct sk_buff *reply;
f564f45c
CW		 1644
1645 if (!(n->nud_state & NUD_CONNECTED)) {
1646 neigh_release(n);
1647 goto out;
1648 }
1649
3ad7a4b1 1650 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1651 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1652 /* bridge-local neighbor */
1653 neigh_release(n);
1654 goto out;
1655 }
1656
4b29dba9
DS		 1657 reply = vxlan_na_create(skb, n,
1658 !!(f ? f->flags & NTF_ROUTER : 0));
1659
f564f45c 1660 neigh_release(n);
4b29dba9
DS		 1661
1662 if (reply == NULL)
1663 goto out;
1664
1665 if (netif_rx_ni(reply) == NET_RX_DROP)
1666 dev->stats.rx_dropped++;
1667
f564f45c 1668 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1669 union vxlan_addr ipa = {
1670 .sin6.sin6_addr = msg->target,
a45e92a5 1671 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1672 };
1673
f564f45c
CW		 1674 vxlan_ip_miss(dev, &ipa);
1675 }
1676
1677out:
1678 consume_skb(skb);
1679 return NETDEV_TX_OK;
1680}
1681#endif
1682
e4f67add
DS		 1683static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1684{
1685 struct vxlan_dev *vxlan = netdev_priv(dev);
1686 struct neighbour *n;
e4f67add
DS		 1687
1688 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1689 return false;
1690
1691 n = NULL;
1692 switch (ntohs(eth_hdr(skb)->h_proto)) {
1693 case ETH_P_IP:
e15a00aa
CW		 1694 {
1695 struct iphdr *pip;
1696
e4f67add
DS		 1697 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1698 return false;
1699 pip = ip_hdr(skb);
1700 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1701 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1702 union vxlan_addr ipa = {
1703 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1704 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1705 };
1706
1707 vxlan_ip_miss(dev, &ipa);
1708 return false;
1709 }
1710
e4f67add 1711 break;
e15a00aa
CW		 1712 }
1713#if IS_ENABLED(CONFIG_IPV6)
1714 case ETH_P_IPV6:
1715 {
1716 struct ipv6hdr *pip6;
1717
1718 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1719 return false;
1720 pip6 = ipv6_hdr(skb);
1721 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1722 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1723 union vxlan_addr ipa = {
1724 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1725 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1726 };
1727
1728 vxlan_ip_miss(dev, &ipa);
1729 return false;
1730 }
1731
1732 break;
1733 }
1734#endif
e4f67add
DS		 1735 default:
1736 return false;
1737 }
1738
1739 if (n) {
1740 bool diff;
1741
7367d0b5 1742 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1743 if (diff) {
1744 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1745 dev->addr_len);
1746 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1747 }
1748 neigh_release(n);
1749 return diff;
e4c7ed41
CW		 1750 }
1751
e4f67add
DS		 1752 return false;
1753}
1754
af33c1ad 1755static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1756 struct vxlan_metadata *md)
1757{
1758 struct vxlanhdr_gbp *gbp;
1759
db79a621
TG		 1760 if (!md->gbp)
1761 return;
1762
3511494c 1763 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1764 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1765
1766 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1767 gbp->dont_learn = 1;
1768
1769 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1770 gbp->policy_applied = 1;
1771
1772 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1773}
1774
e1e5314d
JB		 1775static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1776 __be16 protocol)
1777{
1778 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1779
1780 gpe->np_applied = 1;
1781
1782 switch (protocol) {
1783 case htons(ETH_P_IP):
1784 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1785 return 0;
1786 case htons(ETH_P_IPV6):
1787 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1788 return 0;
1789 case htons(ETH_P_TEB):
1790 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1791 return 0;
1792 }
1793 return -EPFNOSUPPORT;
1794}
1795
f491e56d
JB		 1796static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1797 int iphdr_len, __be32 vni,
1798 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1799 bool udp_sum)
e4c7ed41 1800{
e4c7ed41 1801 struct vxlanhdr *vxh;
e4c7ed41
CW		 1802 int min_headroom;
1803 int err;
dfd8645e 1804 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1805 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1806
af33c1ad 1807 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1808 skb->ip_summed == CHECKSUM_PARTIAL) {
1809 int csum_start = skb_checksum_start_offset(skb);
1810
1811 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1812 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1813 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1814 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1815 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1816 }
e4c7ed41 1817
e4c7ed41 1818 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1819 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1820
1821 /* Need space for new headers (invalidates iph ptr) */
1822 err = skb_cow_head(skb, min_headroom);
e1e5314d 1823 if (unlikely(err))
c46b7897 1824 return err;
649c5b8b 1825
aed069df
AD		 1826 err = iptunnel_handle_offloads(skb, type);
1827 if (err)
c46b7897 1828 return err;
b736a623 1829
d58ff351 1830 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1831 vxh->vx_flags = VXLAN_HF_VNI;
1832 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1833
dfd8645e 1834 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1835 unsigned int start;
dfd8645e 1836
54bfd872
JB		 1837 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1838 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1839 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1840
1841 if (!skb_is_gso(skb)) {
1842 skb->ip_summed = CHECKSUM_NONE;
1843 skb->encapsulation = 0;
1844 }
1845 }
1846
af33c1ad
TH		 1847 if (vxflags & VXLAN_F_GBP)
1848 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1849 if (vxflags & VXLAN_F_GPE) {
1850 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1851 if (err < 0)
c46b7897 1852 return err;
e1e5314d
JB		 1853 inner_protocol = skb->protocol;
1854 }
3511494c 1855
e1e5314d 1856 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1857 return 0;
49560532 1858}
49560532 1859
655c3de1 1860static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1861 struct vxlan_sock *sock4,
1a8496ba 1862 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1863 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1864 struct dst_cache *dst_cache,
db3c6139 1865 const struct ip_tunnel_info *info)
1a8496ba 1866{
db3c6139 1867 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1868 struct rtable *rt = NULL;
1869 struct flowi4 fl4;
1870
655c3de1 1871 if (!sock4)
1872 return ERR_PTR(-EIO);
1873
db3c6139
DB		 1874 if (tos && !info)
1875 use_cache = false;
1876 if (use_cache) {
0c1d70af
PA		 1877 rt = dst_cache_get_ip4(dst_cache, saddr);
1878 if (rt)
1879 return rt;
1880 }
1881
1a8496ba
JB		 1882 memset(&fl4, 0, sizeof(fl4));
1883 fl4.flowi4_oif = oif;
1884 fl4.flowi4_tos = RT_TOS(tos);
1885 fl4.flowi4_mark = skb->mark;
1886 fl4.flowi4_proto = IPPROTO_UDP;
1887 fl4.daddr = daddr;
272d96a5 1888 fl4.saddr = *saddr;
4ecb1d83
MP		 1889 fl4.fl4_dport = dport;
1890 fl4.fl4_sport = sport;
1a8496ba
JB		 1891
1892 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1893 if (likely(!IS_ERR(rt))) {
1894 if (rt->dst.dev == dev) {
1895 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1896 ip_rt_put(rt);
1897 return ERR_PTR(-ELOOP);
1898 }
1899
1a8496ba 1900 *saddr = fl4.saddr;
0c1d70af
PA		 1901 if (use_cache)
1902 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1903 } else {
1904 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1905 return ERR_PTR(-ENETUNREACH);
0c1d70af 1906 }
1a8496ba
JB		 1907 return rt;
1908}
1909
e5d4b29f
JB		 1910#if IS_ENABLED(CONFIG_IPV6)
1911static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1912 struct net_device *dev,
03dc52a8 1913 struct vxlan_sock *sock6,
1400615d 1914 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1915 __be32 label,
e5d4b29f 1916 const struct in6_addr *daddr,
0c1d70af 1917 struct in6_addr *saddr,
4ecb1d83 1918 __be16 dport, __be16 sport,
db3c6139
DB		 1919 struct dst_cache *dst_cache,
1920 const struct ip_tunnel_info *info)
e5d4b29f 1921{
db3c6139 1922 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1923 struct dst_entry *ndst;
1924 struct flowi6 fl6;
1925 int err;
1926
c6fcc4fc 1927 if (!sock6)
1928 return ERR_PTR(-EIO);
1929
1400615d
DB		 1930 if (tos && !info)
1931 use_cache = false;
db3c6139 1932 if (use_cache) {
0c1d70af
PA		 1933 ndst = dst_cache_get_ip6(dst_cache, saddr);
1934 if (ndst)
1935 return ndst;
1936 }
1937
e5d4b29f
JB		 1938 memset(&fl6, 0, sizeof(fl6));
1939 fl6.flowi6_oif = oif;
1940 fl6.daddr = *daddr;
272d96a5 1941 fl6.saddr = *saddr;
eaa93bf4 1942 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1943 fl6.flowi6_mark = skb->mark;
1944 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1945 fl6.fl6_dport = dport;
1946 fl6.fl6_sport = sport;
e5d4b29f
JB		 1947
1948 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1949 sock6->sock->sk,
e5d4b29f 1950 &ndst, &fl6);
655c3de1 1951 if (unlikely(err < 0)) {
1952 netdev_dbg(dev, "no route to %pI6\n", daddr);
1953 return ERR_PTR(-ENETUNREACH);
1954 }
1955
1956 if (unlikely(ndst->dev == dev)) {
1957 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1958 dst_release(ndst);
1959 return ERR_PTR(-ELOOP);
1960 }
e5d4b29f
JB		 1961
1962 *saddr = fl6.saddr;
db3c6139 1963 if (use_cache)
0c1d70af 1964 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1965 return ndst;
1966}
1967#endif
1968
9dcc71e1
SS		 1969/* Bypass encapsulation if the destination is local */
1970static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 1971 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 1972{
8f84985f 1973 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1974 union vxlan_addr loopback;
1975 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1976 struct net_device *dev = skb->dev;
1977 int len = skb->len;
9dcc71e1 1978
8f84985f
LR		 1979 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1980 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1981 skb->pkt_type = PACKET_HOST;
1982 skb->encapsulation = 0;
1983 skb->dev = dst_vxlan->dev;
1984 __skb_pull(skb, skb_network_offset(skb));
1985
e4c7ed41
CW		 1986 if (remote_ip->sa.sa_family == AF_INET) {
1987 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1988 loopback.sa.sa_family = AF_INET;
1989#if IS_ENABLED(CONFIG_IPV6)
1990 } else {
1991 loopback.sin6.sin6_addr = in6addr_loopback;
1992 loopback.sa.sa_family = AF_INET6;
1993#endif
1994 }
1995
9dcc71e1 1996 if (dst_vxlan->flags & VXLAN_F_LEARN)
3ad7a4b1 1997 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, vni);
9dcc71e1
SS		 1998
1999 u64_stats_update_begin(&tx_stats->syncp);
2000 tx_stats->tx_packets++;
ce6502a8 2001 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2002 u64_stats_update_end(&tx_stats->syncp);
2003
2004 if (netif_rx(skb) == NET_RX_SUCCESS) {
2005 u64_stats_update_begin(&rx_stats->syncp);
2006 rx_stats->rx_packets++;
ce6502a8 2007 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2008 u64_stats_update_end(&rx_stats->syncp);
2009 } else {
ce6502a8 2010 dev->stats.rx_dropped++;
9dcc71e1
SS		 2011 }
2012}
2013
fee1fad7 2014static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2015 struct vxlan_dev *vxlan, union vxlan_addr *daddr,
1f6cc07e 2016 __be16 dst_port, __be32 vni, struct dst_entry *dst,
fee1fad7 2017 u32 rt_flags)
2018{
2019#if IS_ENABLED(CONFIG_IPV6)
2020 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2021 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2022 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2023 */
2024 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2025#endif
2026 /* Bypass encapsulation if the destination is local */
2027 if (rt_flags & RTCF_LOCAL &&
2028 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2029 struct vxlan_dev *dst_vxlan;
2030
2031 dst_release(dst);
2032 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2033 daddr->sa.sa_family, dst_port,
2034 vxlan->flags);
2035 if (!dst_vxlan) {
2036 dev->stats.tx_errors++;
2037 kfree_skb(skb);
2038
2039 return -ENOENT;
2040 }
3ad7a4b1 2041 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2042 return 1;
2043 }
2044
2045 return 0;
2046}
2047
4ad16930 2048static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2049 __be32 default_vni, struct vxlan_rdst *rdst,
2050 bool did_rsc)
d342894c 2051{
d71785ff 2052 struct dst_cache *dst_cache;
3093fbe7 2053 struct ip_tunnel_info *info;
d342894c 2054 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2055 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2056 union vxlan_addr *dst;
272d96a5 2057 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2058 struct vxlan_metadata _md;
2059 struct vxlan_metadata *md = &_md;
e4c7ed41 2060 __be16 src_port = 0, dst_port;
655c3de1 2061 struct dst_entry *ndst = NULL;
e7f70af1 2062 __be32 vni, label;
d342894c 2063 __u8 tos, ttl;
0e6fbc5b 2064 int err;
ee122c79 2065 u32 flags = vxlan->flags;
b4ed5cad 2066 bool udp_sum = false;
f491e56d 2067 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2068
61adedf3 2069 info = skb_tunnel_info(skb);
3093fbe7 2070
ee122c79 2071 if (rdst) {
0770b53b 2072 dst = &rdst->remote_ip;
2073 if (vxlan_addr_any(dst)) {
2074 if (did_rsc) {
2075 /* short-circuited back to local bridge */
3ad7a4b1 2076 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2077 return;
2078 }
2079 goto drop;
2080 }
2081
0dfbdf41 2082 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2083 vni = (rdst->remote_vni) ? : default_vni;
1158632b 2084 local_ip = vxlan->cfg.saddr;
d71785ff 2085 dst_cache = &rdst->dst_cache;
0770b53b 2086 md->gbp = skb->mark;
2087 ttl = vxlan->cfg.ttl;
2088 if (!ttl && vxlan_addr_multicast(dst))
2089 ttl = 1;
2090
2091 tos = vxlan->cfg.tos;
2092 if (tos == 1)
2093 tos = ip_tunnel_get_dsfield(old_iph, skb);
2094
2095 if (dst->sa.sa_family == AF_INET)
2096 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2097 else
2098 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2099 label = vxlan->cfg.label;
ee122c79
TG		 2100 } else {
2101 if (!info) {
2102 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2103 dev->name);
2104 goto drop;
2105 }
b1be00a6 2106 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2107 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2108 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2109 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2110 } else {
a725e514 2111 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2112 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2113 }
ee122c79 2114 dst = &remote_ip;
0770b53b 2115 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2116 vni = tunnel_id_to_key32(info->key.tun_id);
d71785ff 2117 dst_cache = &info->dst_cache;
0770b53b 2118 if (info->options_len)
2119 md = ip_tunnel_info_opts(info);
a725e514
JB		 2120 ttl = info->key.ttl;
2121 tos = info->key.tos;
e7f70af1 2122 label = info->key.label;
b4ed5cad 2123 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2124 }
0770b53b 2125 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2126 vxlan->cfg.port_max, true);
a725e514 2127
56de859e 2128 rcu_read_lock();
e4c7ed41 2129 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2130 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2131 struct rtable *rt;
0770b53b 2132 __be16 df = 0;
c6fcc4fc 2133
655c3de1 2134 rt = vxlan_get_route(vxlan, dev, sock4, skb,
1a8496ba 2135 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2136 dst->sin.sin_addr.s_addr,
1158632b 2137 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2138 dst_port, src_port,
d71785ff 2139 dst_cache, info);
8ebd115b
DM		 2140 if (IS_ERR(rt)) {
2141 err = PTR_ERR(rt);
c46b7897 2142 goto tx_error;
8ebd115b 2143 }
e4c7ed41
CW		 2144
2145 /* Bypass encapsulation if the destination is local */
fee1fad7 2146 if (!info) {
2147 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2148 dst_port, vni, &rt->dst,
2149 rt->rt_flags);
2150 if (err)
56de859e 2151 goto out_unlock;
fee1fad7 2152 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2153 df = htons(IP_DF);
fee1fad7 2154 }
6ceb31ca 2155
c46b7897 2156 ndst = &rt->dst;
e4c7ed41
CW		 2157 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2158 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2159 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2160 vni, md, flags, udp_sum);
f491e56d 2161 if (err < 0)
c46b7897 2162 goto tx_error;
f491e56d 2163
1158632b 2164 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2165 dst->sin.sin_addr.s_addr, tos, ttl, df,
2166 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2167#if IS_ENABLED(CONFIG_IPV6)
2168 } else {
c6fcc4fc 2169 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2170
655c3de1 2171 ndst = vxlan6_get_route(vxlan, dev, sock6, skb,
1400615d 2172 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2173 label, &dst->sin6.sin6_addr,
1158632b 2174 &local_ip.sin6.sin6_addr,
4ecb1d83 2175 dst_port, src_port,
db3c6139 2176 dst_cache, info);
e5d4b29f 2177 if (IS_ERR(ndst)) {
8ebd115b 2178 err = PTR_ERR(ndst);
c46b7897 2179 ndst = NULL;
9dcc71e1 2180 goto tx_error;
e4c7ed41 2181 }
655c3de1 2182
fee1fad7 2183 if (!info) {
2184 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2185
fee1fad7 2186 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2187 dst_port, vni, ndst,
2188 rt6i_flags);
2189 if (err)
56de859e 2190 goto out_unlock;
fee1fad7 2191 }
35e2d115 2192
1400615d 2193 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2194 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2195 skb_scrub_packet(skb, xnet);
2196 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2197 vni, md, flags, udp_sum);
c46b7897 2198 if (err < 0)
2199 goto tx_error;
2200
0770b53b 2201 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2202 &local_ip.sin6.sin6_addr,
272d96a5 2203 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2204 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2205#endif
2206 }
56de859e
JK		 2207out_unlock:
2208 rcu_read_unlock();
4ad16930 2209 return;
d342894c 2210
2211drop:
2212 dev->stats.tx_dropped++;
c46b7897 2213 dev_kfree_skb(skb);
2214 return;
d342894c 2215
2216tx_error:
56de859e 2217 rcu_read_unlock();
655c3de1 2218 if (err == -ELOOP)
2219 dev->stats.collisions++;
2220 else if (err == -ENETUNREACH)
2221 dev->stats.tx_carrier_errors++;
c46b7897 2222 dst_release(ndst);
d342894c 2223 dev->stats.tx_errors++;
c46b7897 2224 kfree_skb(skb);
d342894c 2225}
2226
6681712d
DS		 2227/* Transmit local packets over Vxlan
2228 *
2229 * Outer IP header inherits ECN and DF from inner header.
2230 * Outer UDP destination is the VXLAN assigned port.
2231 * source port is based on hash of flow
2232 */
2233static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2234{
2235 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2236 const struct ip_tunnel_info *info;
6681712d
DS		 2237 struct ethhdr *eth;
2238 bool did_rsc = false;
8f646c92 2239 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2240 struct vxlan_fdb *f;
3ad7a4b1 2241 __be32 vni = 0;
6681712d 2242
61adedf3 2243 info = skb_tunnel_info(skb);
3093fbe7 2244
6681712d 2245 skb_reset_mac_header(skb);
6681712d 2246
47e5d1b0 2247 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2248 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2249 info->mode & IP_TUNNEL_INFO_TX) {
2250 vni = tunnel_id_to_key32(info->key.tun_id);
2251 } else {
2252 if (info && info->mode & IP_TUNNEL_INFO_TX)
2253 vxlan_xmit_one(skb, dev, vni, NULL, false);
2254 else
2255 kfree_skb(skb);
2256 return NETDEV_TX_OK;
2257 }
47e5d1b0
JB		 2258 }
2259
2260 if (vxlan->flags & VXLAN_F_PROXY) {
2261 eth = eth_hdr(skb);
f564f45c 2262 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2263 return arp_reduce(dev, skb, vni);
f564f45c 2264#if IS_ENABLED(CONFIG_IPV6)
f1fb08f6
VB		 2265 else if (ntohs(eth->h_proto) == ETH_P_IPV6) {
2266 struct ipv6hdr *hdr, _hdr;
2267 if ((hdr = skb_header_pointer(skb,
2268 skb_network_offset(skb),
2269 sizeof(_hdr), &_hdr)) &&
2270 hdr->nexthdr == IPPROTO_ICMPV6)
2271 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2272 }
2273#endif
2274 }
6681712d 2275
47e5d1b0 2276 eth = eth_hdr(skb);
3ad7a4b1 2277 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2278 did_rsc = false;
2279
2280 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2281 (ntohs(eth->h_proto) == ETH_P_IP ||
2282 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2283 did_rsc = route_shortcircuit(dev, skb);
2284 if (did_rsc)
3ad7a4b1 2285 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2286 }
2287
6681712d 2288 if (f == NULL) {
3ad7a4b1 2289 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2290 if (f == NULL) {
2291 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2292 !is_multicast_ether_addr(eth->h_dest))
2293 vxlan_fdb_miss(vxlan, eth->h_dest);
2294
2295 dev->stats.tx_dropped++;
8f646c92 2296 kfree_skb(skb);
afbd8bae
MR		 2297 return NETDEV_TX_OK;
2298 }
2299 }
6681712d 2300
afbd8bae
MR		 2301 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2302 struct sk_buff *skb1;
6681712d 2303
8f646c92
ED		 2304 if (!fdst) {
2305 fdst = rdst;
2306 continue;
2307 }
afbd8bae
MR		 2308 skb1 = skb_clone(skb, GFP_ATOMIC);
2309 if (skb1)
3ad7a4b1 2310 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2311 }
2312
8f646c92 2313 if (fdst)
3ad7a4b1 2314 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2315 else
2316 kfree_skb(skb);
4ad16930 2317 return NETDEV_TX_OK;
6681712d
DS		 2318}
2319
d342894c 2320/* Walk the forwarding table and purge stale entries */
2321static void vxlan_cleanup(unsigned long arg)
2322{
2323 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2324 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2325 unsigned int h;
2326
2327 if (!netif_running(vxlan->dev))
2328 return;
2329
d342894c 2330 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2331 struct hlist_node *p, *n;
14e1d0fa
SD		 2332
2333 spin_lock_bh(&vxlan->hash_lock);
d342894c 2334 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2335 struct vxlan_fdb *f
2336 = container_of(p, struct vxlan_fdb, hlist);
2337 unsigned long timeout;
2338
efb5f68f 2339 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2340 continue;
2341
def499c9
RP		 2342 if (f->flags & NTF_EXT_LEARNED)
2343 continue;
2344
0dfbdf41 2345 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2346 if (time_before_eq(timeout, jiffies)) {
2347 netdev_dbg(vxlan->dev,
2348 "garbage collect %pM\n",
2349 f->eth_addr);
2350 f->state = NUD_STALE;
2351 vxlan_fdb_destroy(vxlan, f);
2352 } else if (time_before(timeout, next_timer))
2353 next_timer = timeout;
2354 }
14e1d0fa 2355 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2356 }
d342894c 2357
2358 mod_timer(&vxlan->age_timer, next_timer);
2359}
2360
a53cb29b
MB		 2361static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2362{
2363 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2364
2365 spin_lock(&vn->sock_lock);
2366 hlist_del_init_rcu(&vxlan->hlist);
2367 spin_unlock(&vn->sock_lock);
2368}
2369
9c2e24e1
PS		 2370static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2371{
56ef9c90 2372 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2373 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2374
56ef9c90 2375 spin_lock(&vn->sock_lock);
9c2e24e1 2376 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2377 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2378}
2379
d342894c 2380/* Setup stats when device is created */
2381static int vxlan_init(struct net_device *dev)
2382{
1c213bd2 2383 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2384 if (!dev->tstats)
d342894c 2385 return -ENOMEM;
2386
2387 return 0;
2388}
2389
3ad7a4b1 2390static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2391{
2392 struct vxlan_fdb *f;
2393
2394 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2395 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2396 if (f)
2397 vxlan_fdb_destroy(vxlan, f);
2398 spin_unlock_bh(&vxlan->hash_lock);
2399}
2400
ebf4063e
SH		 2401static void vxlan_uninit(struct net_device *dev)
2402{
2403 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2404
3ad7a4b1 2405 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2406
ebf4063e
SH		 2407 free_percpu(dev->tstats);
2408}
2409
d342894c 2410/* Start ageing timer and join group when device is brought up */
2411static int vxlan_open(struct net_device *dev)
2412{
2413 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2414 int ret;
56ef9c90 2415
205f356d
JB		 2416 ret = vxlan_sock_add(vxlan);
2417 if (ret < 0)
2418 return ret;
d342894c 2419
79d4a94f 2420 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2421 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2422 if (ret == -EADDRINUSE)
2423 ret = 0;
56ef9c90 2424 if (ret) {
205f356d 2425 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2426 return ret;
2427 }
d342894c 2428 }
2429
0dfbdf41 2430 if (vxlan->cfg.age_interval)
d342894c 2431 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2432
56ef9c90 2433 return ret;
d342894c 2434}
2435
2436/* Purge the forwarding table */
8b3f9337 2437static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2438{
31fec5aa 2439 unsigned int h;
d342894c 2440
2441 spin_lock_bh(&vxlan->hash_lock);
2442 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2443 struct hlist_node *p, *n;
2444 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2445 struct vxlan_fdb *f
2446 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2447 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2448 continue;
afbd8bae
MR		 2449 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2450 if (!is_zero_ether_addr(f->eth_addr))
2451 vxlan_fdb_destroy(vxlan, f);
d342894c 2452 }
2453 }
2454 spin_unlock_bh(&vxlan->hash_lock);
2455}
2456
2457/* Cleanup timer and forwarding table on shutdown */
2458static int vxlan_stop(struct net_device *dev)
2459{
2460 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2461 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2462 int ret = 0;
d342894c 2463
24c0e683 2464 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2465 !vxlan_group_used(vn, vxlan))
56ef9c90 2466 ret = vxlan_igmp_leave(vxlan);
d342894c 2467
2468 del_timer_sync(&vxlan->age_timer);
2469
8b3f9337 2470 vxlan_flush(vxlan, false);
205f356d 2471 vxlan_sock_release(vxlan);
d342894c 2472
56ef9c90 2473 return ret;
d342894c 2474}
2475
d342894c 2476/* Stub, nothing needs to be done. */
2477static void vxlan_set_multicast_list(struct net_device *dev)
2478{
2479}
2480
91572088 2481static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2482{
91572088
JW		 2483 struct vxlan_dev *vxlan = netdev_priv(dev);
2484 struct vxlan_rdst *dst = &vxlan->default_dst;
2485 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2486 dst->remote_ifindex);
2487 bool use_ipv6 = false;
345010b5
DB		 2488
2489 if (dst->remote_ip.sa.sa_family == AF_INET6)
91572088 2490 use_ipv6 = true;
345010b5 2491
91572088
JW		 2492 /* This check is different than dev->max_mtu, because it looks at
2493 * the lowerdev->mtu, rather than the static dev->max_mtu
2494 */
2495 if (lowerdev) {
2496 int max_mtu = lowerdev->mtu -
2497 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2498 if (new_mtu > max_mtu)
72564b59 2499 return -EINVAL;
72564b59
DW		 2500 }
2501
345010b5
DB		 2502 dev->mtu = new_mtu;
2503 return 0;
2504}
2505
fc4099f1
PS		 2506static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2507{
2508 struct vxlan_dev *vxlan = netdev_priv(dev);
2509 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2510 __be16 sport, dport;
2511
2512 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2513 vxlan->cfg.port_max, true);
2514 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2515
239e944f 2516 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2517 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2518 struct rtable *rt;
2519
655c3de1 2520 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2521 info->key.u.ipv4.dst,
22f0708a
PA		 2522 &info->key.u.ipv4.src, dport, sport,
2523 &info->dst_cache, info);
1a8496ba
JB		 2524 if (IS_ERR(rt))
2525 return PTR_ERR(rt);
2526 ip_rt_put(rt);
239e944f
JB		 2527 } else {
2528#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2529 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2530 struct dst_entry *ndst;
2531
655c3de1 2532 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2533 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2534 &info->key.u.ipv6.src, dport, sport,
2535 &info->dst_cache, info);
239e944f
JB		 2536 if (IS_ERR(ndst))
2537 return PTR_ERR(ndst);
2538 dst_release(ndst);
239e944f
JB		 2539#else /* !CONFIG_IPV6 */
2540 return -EPFNOSUPPORT;
2541#endif
2542 }
1a8496ba
JB		 2543 info->key.tp_src = sport;
2544 info->key.tp_dst = dport;
239e944f 2545 return 0;
fc4099f1
PS		 2546}
2547
0c867c9b 2548static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2549 .ndo_init = vxlan_init,
ebf4063e 2550 .ndo_uninit = vxlan_uninit,
d342894c 2551 .ndo_open = vxlan_open,
2552 .ndo_stop = vxlan_stop,
2553 .ndo_start_xmit = vxlan_xmit,
e8171045 2554 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2555 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2556 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2557 .ndo_validate_addr = eth_validate_addr,
2558 .ndo_set_mac_address = eth_mac_addr,
2559 .ndo_fdb_add = vxlan_fdb_add,
2560 .ndo_fdb_del = vxlan_fdb_delete,
2561 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2562 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2563};
2564
e1e5314d
JB		 2565static const struct net_device_ops vxlan_netdev_raw_ops = {
2566 .ndo_init = vxlan_init,
2567 .ndo_uninit = vxlan_uninit,
2568 .ndo_open = vxlan_open,
2569 .ndo_stop = vxlan_stop,
2570 .ndo_start_xmit = vxlan_xmit,
2571 .ndo_get_stats64 = ip_tunnel_get_stats64,
2572 .ndo_change_mtu = vxlan_change_mtu,
2573 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2574};
2575
d342894c 2576/* Info for udev, that this is a virtual tunnel endpoint */
2577static struct device_type vxlan_type = {
2578 .name = "vxlan",
2579};
2580
e5de25dc 2581/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2582 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2583 * to implement the ndo_udp_tunnel_add.
53cf5275 2584 */
b7aade15 2585static void vxlan_push_rx_ports(struct net_device *dev)
53cf5275
JG		 2586{
2587 struct vxlan_sock *vs;
2588 struct net *net = dev_net(dev);
2589 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2590 unsigned int i;
53cf5275
JG		 2591
2592 spin_lock(&vn->sock_lock);
2593 for (i = 0; i < PORT_HASH_SIZE; ++i) {
e7b3db5e
AD		 2594 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2595 udp_tunnel_push_rx_port(dev, vs->sock,
b9adcd69
AD		 2596 (vs->flags & VXLAN_F_GPE) ?
2597 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2598 UDP_TUNNEL_TYPE_VXLAN);
53cf5275
JG		 2599 }
2600 spin_unlock(&vn->sock_lock);
2601}
53cf5275 2602
d342894c 2603/* Initialize the device structure. */
2604static void vxlan_setup(struct net_device *dev)
2605{
2606 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2607 unsigned int h;
d342894c 2608
65226ef8
JB		 2609 eth_hw_addr_random(dev);
2610 ether_setup(dev);
2611
cf124db5 2612 dev->needs_free_netdev = true;
d342894c 2613 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2614
d342894c 2615 dev->features |= NETIF_F_LLTX;
d6727fe3 2616 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2617 dev->features |= NETIF_F_RXCSUM;
05c0db08 2618 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2619
1eaa8178 2620 dev->vlan_features = dev->features;
0afb1666 2621 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2622 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2623 netif_keep_dst(dev);
0c867c9b 2624 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2625
a985343b
MS		 2626 /* MTU range: 68 - 65535 */
2627 dev->min_mtu = ETH_MIN_MTU;
2628 dev->max_mtu = ETH_MAX_MTU;
2629
553675fb 2630 INIT_LIST_HEAD(&vxlan->next);
d342894c 2631 spin_lock_init(&vxlan->hash_lock);
2632
2633 init_timer_deferrable(&vxlan->age_timer);
2634 vxlan->age_timer.function = vxlan_cleanup;
2635 vxlan->age_timer.data = (unsigned long) vxlan;
2636
2637 vxlan->dev = dev;
a985343b 2638 vxlan->net = dev_net(dev);
d342894c 2639
58ce31cc
TH		 2640 gro_cells_init(&vxlan->gro_cells, dev);
2641
d342894c 2642 for (h = 0; h < FDB_HASH_SIZE; ++h)
2643 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2644}
2645
0c867c9b
JB		 2646static void vxlan_ether_setup(struct net_device *dev)
2647{
0c867c9b
JB		 2648 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2649 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2650 dev->netdev_ops = &vxlan_netdev_ether_ops;
2651}
2652
e1e5314d
JB		 2653static void vxlan_raw_setup(struct net_device *dev)
2654{
65226ef8 2655 dev->header_ops = NULL;
e1e5314d
JB		 2656 dev->type = ARPHRD_NONE;
2657 dev->hard_header_len = 0;
2658 dev->addr_len = 0;
e1e5314d
JB		 2659 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2660 dev->netdev_ops = &vxlan_netdev_raw_ops;
2661}
2662
d342894c 2663static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2664 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2665 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2666 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2667 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2668 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2669 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2670 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2671 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2672 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2673 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2674 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2675 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2676 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2677 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2678 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2679 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2680 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2681 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2682 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2683 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2684 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2685 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2686 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2687 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2688 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2689 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2690 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2691};
2692
2693static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2694{
2695 if (tb[IFLA_ADDRESS]) {
2696 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2697 pr_debug("invalid link address (not ethernet)\n");
2698 return -EINVAL;
2699 }
2700
2701 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2702 pr_debug("invalid all zero ethernet address\n");
2703 return -EADDRNOTAVAIL;
2704 }
2705 }
2706
a985343b
MS		 2707 if (tb[IFLA_MTU]) {
2708 u32 mtu = nla_get_u32(data[IFLA_MTU]);
2709
2710 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU)
2711 return -EINVAL;
2712 }
2713
d342894c 2714 if (!data)
2715 return -EINVAL;
2716
2717 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 2718 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2719
4e37d691 2720 if (id >= VXLAN_N_VID)
d342894c 2721 return -ERANGE;
2722 }
2723
05f47d69 2724 if (data[IFLA_VXLAN_PORT_RANGE]) {
2725 const struct ifla_vxlan_port_range *p
2726 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2727
2728 if (ntohs(p->high) < ntohs(p->low)) {
2729 pr_debug("port range %u .. %u not valid\n",
2730 ntohs(p->low), ntohs(p->high));
2731 return -EINVAL;
2732 }
2733 }
2734
d342894c 2735 return 0;
2736}
2737
1b13c97f
YB		 2738static void vxlan_get_drvinfo(struct net_device *netdev,
2739 struct ethtool_drvinfo *drvinfo)
2740{
2741 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2742 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2743}
2744
2745static const struct ethtool_ops vxlan_ethtool_ops = {
2746 .get_drvinfo = vxlan_get_drvinfo,
2747 .get_link = ethtool_op_get_link,
2748};
2749
3ee64f39
TH		 2750static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2751 __be16 port, u32 flags)
553675fb 2752{
e4c7ed41 2753 struct socket *sock;
3ee64f39
TH		 2754 struct udp_port_cfg udp_conf;
2755 int err;
e4c7ed41 2756
3ee64f39 2757 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2758
3ee64f39
TH		 2759 if (ipv6) {
2760 udp_conf.family = AF_INET6;
3ee64f39 2761 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2762 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2763 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2764 } else {
2765 udp_conf.family = AF_INET;
e4c7ed41 2766 }
e4c7ed41 2767
3ee64f39 2768 udp_conf.local_udp_port = port;
553675fb 2769
3ee64f39
TH		 2770 /* Open UDP socket */
2771 err = udp_sock_create(net, &udp_conf, &sock);
2772 if (err < 0)
2773 return ERR_PTR(err);
e4c7ed41 2774
39deb2c7 2775 return sock;
e4c7ed41
CW		 2776}
2777
2778/* Create new listen socket if needed */
b1be00a6
JB		 2779static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2780 __be16 port, u32 flags)
e4c7ed41
CW		 2781{
2782 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2783 struct vxlan_sock *vs;
2784 struct socket *sock;
e4c7ed41 2785 unsigned int h;
acbf74a7 2786 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2787
dc01e7d3 2788 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2789 if (!vs)
2790 return ERR_PTR(-ENOMEM);
2791
2792 for (h = 0; h < VNI_HASH_SIZE; ++h)
2793 INIT_HLIST_HEAD(&vs->vni_list[h]);
2794
3ee64f39 2795 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2796 if (IS_ERR(sock)) {
553675fb 2797 kfree(vs);
e50fddc8 2798 return ERR_CAST(sock);
553675fb 2799 }
e4c7ed41
CW		 2800
2801 vs->sock = sock;
9c2e24e1 2802 atomic_set(&vs->refcnt, 1);
af33c1ad 2803 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2804
9c2e24e1
PS		 2805 spin_lock(&vn->sock_lock);
2806 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2807 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2808 (vs->flags & VXLAN_F_GPE) ?
2809 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2810 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2811 spin_unlock(&vn->sock_lock);
553675fb 2812
2813 /* Mark socket as an encapsulation socket. */
5602c48c 2814 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2815 tunnel_cfg.sk_user_data = vs;
2816 tunnel_cfg.encap_type = 1;
f2d1968e 2817 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2818 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2819 tunnel_cfg.gro_receive = vxlan_gro_receive;
2820 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2821
2822 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2823
9c2e24e1
PS		 2824 return vs;
2825}
2826
b1be00a6 2827static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2828{
205f356d
JB		 2829 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2830 struct vxlan_sock *vs = NULL;
9c2e24e1 2831
205f356d 2832 if (!vxlan->cfg.no_share) {
56ef9c90 2833 spin_lock(&vn->sock_lock);
205f356d
JB		 2834 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2835 vxlan->cfg.dst_port, vxlan->flags);
2836 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2837 spin_unlock(&vn->sock_lock);
205f356d 2838 return -EBUSY;
56ef9c90
MRL		 2839 }
2840 spin_unlock(&vn->sock_lock);
2841 }
205f356d 2842 if (!vs)
b1be00a6
JB		 2843 vs = vxlan_socket_create(vxlan->net, ipv6,
2844 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2845 if (IS_ERR(vs))
2846 return PTR_ERR(vs);
b1be00a6
JB		 2847#if IS_ENABLED(CONFIG_IPV6)
2848 if (ipv6)
c6fcc4fc 2849 rcu_assign_pointer(vxlan->vn6_sock, vs);
b1be00a6
JB		 2850 else
2851#endif
c6fcc4fc 2852 rcu_assign_pointer(vxlan->vn4_sock, vs);
205f356d
JB		 2853 vxlan_vs_add_dev(vs, vxlan);
2854 return 0;
553675fb 2855}
2856
b1be00a6
JB		 2857static int vxlan_sock_add(struct vxlan_dev *vxlan)
2858{
b1be00a6 2859 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
d074bf96
JB		 2860 bool ipv6 = vxlan->flags & VXLAN_F_IPV6 || metadata;
2861 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2862 int ret = 0;
2863
c6fcc4fc 2864 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2865#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2866 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2867 if (ipv6) {
b1be00a6 2868 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2869 if (ret < 0 && ret != -EAFNOSUPPORT)
2870 ipv4 = false;
2871 }
b1be00a6 2872#endif
d074bf96 2873 if (ipv4)
b1be00a6
JB		 2874 ret = __vxlan_sock_add(vxlan, false);
2875 if (ret < 0)
2876 vxlan_sock_release(vxlan);
2877 return ret;
2878}
2879
a985343b
MS		 2880static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2881 struct net_device **lower,
2882 struct vxlan_dev *old)
d342894c 2883{
33564bbb 2884 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 2885 struct vxlan_dev *tmp;
e4c7ed41 2886 bool use_ipv6 = false;
d342894c 2887
a985343b
MS		 2888 if (conf->flags & VXLAN_F_GPE) {
2889 /* For now, allow GPE only together with
2890 * COLLECT_METADATA. This can be relaxed later; in such
2891 * case, the other side of the PtP link will have to be
2892 * provided.
2893 */
2894 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2895 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2896 return -EINVAL;
3555621d 2897 }
e1e5314d 2898 }
0c867c9b 2899
a985343b
MS		 2900 if (!conf->remote_ip.sa.sa_family)
2901 conf->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2902
a985343b
MS		 2903 if (conf->remote_ip.sa.sa_family == AF_INET6 ||
2904 conf->saddr.sa.sa_family == AF_INET6) {
057ba29b
JB		 2905 if (!IS_ENABLED(CONFIG_IPV6))
2906 return -EPFNOSUPPORT;
e4c7ed41 2907 use_ipv6 = true;
a985343b 2908 conf->flags |= VXLAN_F_IPV6;
057ba29b 2909 }
d342894c 2910
a985343b 2911 if (conf->label && !use_ipv6)
e7f70af1 2912 return -EINVAL;
e7f70af1 2913
a985343b
MS		 2914 if (conf->remote_ifindex) {
2915 struct net_device *lowerdev;
34e02aa1 2916
a985343b
MS		 2917 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
2918 if (!lowerdev)
34e02aa1 2919 return -ENODEV;
d342894c 2920
e4c7ed41
CW		 2921#if IS_ENABLED(CONFIG_IPV6)
2922 if (use_ipv6) {
2923 struct inet6_dev *idev = __in6_dev_get(lowerdev);
a985343b 2924 if (idev && idev->cnf.disable_ipv6)
e4c7ed41 2925 return -EPERM;
e4c7ed41
CW		 2926 }
2927#endif
2928
a985343b
MS		 2929 *lower = lowerdev;
2930 } else {
2931 if (vxlan_addr_multicast(&conf->remote_ip))
2932 return -EINVAL;
1ba56fb4 2933
a985343b 2934 *lower = NULL;
9dc2ad10 2935 }
d342894c 2936
a985343b
MS		 2937 if (!conf->dst_port) {
2938 if (conf->flags & VXLAN_F_GPE)
2939 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
2940 else
2941 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 2942 }
2943
a985343b
MS		 2944 if (!conf->age_interval)
2945 conf->age_interval = FDB_AGE_DEFAULT;
91572088 2946
a985343b
MS		 2947 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2948 if (tmp == old)
2949 continue;
91572088 2950
a985343b
MS		 2951 if (tmp->cfg.vni == conf->vni &&
2952 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2953 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2954 tmp->cfg.dst_port == conf->dst_port &&
2955 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2956 (conf->flags & VXLAN_F_RCV_FLAGS))
2957 return -EEXIST;
2958 }
91572088 2959
a985343b
MS		 2960 return 0;
2961}
2962
2963static void vxlan_config_apply(struct net_device *dev,
2964 struct vxlan_config *conf,
2965 struct net_device *lowerdev, bool changelink)
2966{
2967 struct vxlan_dev *vxlan = netdev_priv(dev);
2968 struct vxlan_rdst *dst = &vxlan->default_dst;
2969 unsigned short needed_headroom = ETH_HLEN;
2970 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
2971 int max_mtu = ETH_MAX_MTU;
2972
2973 if (!changelink) {
2974 if (conf->flags & VXLAN_F_GPE)
2975 vxlan_raw_setup(dev);
2976 else
2977 vxlan_ether_setup(dev);
2978
2979 if (conf->mtu)
2980 dev->mtu = conf->mtu;
2981 }
2982
2983 dst->remote_vni = conf->vni;
2984
2985 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
2986
2987 if (lowerdev) {
2988 dst->remote_ifindex = conf->remote_ifindex;
2989
2990 dev->gso_max_size = lowerdev->gso_max_size;
2991 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 2992
a985343b 2993 needed_headroom = lowerdev->hard_header_len;
91572088 2994
a985343b
MS		 2995 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
2996 VXLAN_HEADROOM);
7e059158
DW		 2997 }
2998
a985343b
MS		 2999 if (dev->mtu > max_mtu)
3000 dev->mtu = max_mtu;
3001
b1be00a6
JB		 3002 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3003 needed_headroom += VXLAN6_HEADROOM;
3004 else
3005 needed_headroom += VXLAN_HEADROOM;
3006 dev->needed_headroom = needed_headroom;
3007
0dfbdf41 3008 memcpy(&vxlan->cfg, conf, sizeof(*conf));
0dfbdf41 3009 vxlan->flags |= conf->flags;
a985343b 3010}
0dfbdf41 3011
a985343b
MS		 3012static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3013 struct vxlan_config *conf,
3014 bool changelink)
3015{
3016 struct vxlan_dev *vxlan = netdev_priv(dev);
3017 struct net_device *lowerdev;
3018 int ret;
0dfbdf41 3019
a985343b
MS		 3020 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan);
3021 if (ret)
3022 return ret;
8bcdc4f3 3023
a985343b 3024 vxlan_config_apply(dev, conf, lowerdev, changelink);
0dfbdf41 3025
0dfbdf41
TG		 3026 return 0;
3027}
3028
c80498e3
ND		 3029static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3030 struct vxlan_config *conf)
3031{
3032 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3033 struct vxlan_dev *vxlan = netdev_priv(dev);
3034 int err;
3035
3036 err = vxlan_dev_configure(net, dev, conf, false);
3037 if (err)
3038 return err;
3039
3040 dev->ethtool_ops = &vxlan_ethtool_ops;
3041
3042 /* create an fdb entry for a valid default destination */
3043 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3044 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3045 &vxlan->default_dst.remote_ip,
3046 NUD_REACHABLE | NUD_PERMANENT,
3047 NLM_F_EXCL | NLM_F_CREATE,
3048 vxlan->cfg.dst_port,
3049 vxlan->default_dst.remote_vni,
3050 vxlan->default_dst.remote_vni,
3051 vxlan->default_dst.remote_ifindex,
3052 NTF_SELF);
3053 if (err)
3054 return err;
3055 }
3056
3057 err = register_netdevice(dev);
3058 if (err) {
3059 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3060 return err;
3061 }
3062
3063 list_add(&vxlan->next, &vn->vxlan_list);
3064 return 0;
3065}
3066
8bcdc4f3
RP		 3067static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3068 struct net_device *dev, struct vxlan_config *conf,
3069 bool changelink)
0dfbdf41 3070{
8bcdc4f3 3071 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3072
8bcdc4f3 3073 memset(conf, 0, sizeof(*conf));
e277de5f 3074
8bcdc4f3
RP		 3075 /* if changelink operation, start with old existing cfg */
3076 if (changelink)
3077 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3078
3079 if (data[IFLA_VXLAN_ID]) {
3080 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3081
3082 if (changelink && (vni != conf->vni))
3083 return -EOPNOTSUPP;
3084 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3085 }
0dfbdf41
TG		 3086
3087 if (data[IFLA_VXLAN_GROUP]) {
8bcdc4f3 3088 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
0dfbdf41
TG		 3089 } else if (data[IFLA_VXLAN_GROUP6]) {
3090 if (!IS_ENABLED(CONFIG_IPV6))
3091 return -EPFNOSUPPORT;
3092
8bcdc4f3
RP		 3093 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3094 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3095 }
3096
3097 if (data[IFLA_VXLAN_LOCAL]) {
8bcdc4f3
RP		 3098 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3099 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3100 } else if (data[IFLA_VXLAN_LOCAL6]) {
3101 if (!IS_ENABLED(CONFIG_IPV6))
3102 return -EPFNOSUPPORT;
3103
3104 /* TODO: respect scope id */
8bcdc4f3
RP		 3105 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3106 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3107 }
3108
3109 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3110 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3111
d342894c 3112 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3113 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3114
afb97186 3115 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3116 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3117
e7f70af1 3118 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3119 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3120 IPV6_FLOWLABEL_MASK;
3121
8bcdc4f3
RP		 3122 if (data[IFLA_VXLAN_LEARNING]) {
3123 if (nla_get_u8(data[IFLA_VXLAN_LEARNING])) {
3124 conf->flags |= VXLAN_F_LEARN;
3125 } else {
3126 conf->flags &= ~VXLAN_F_LEARN;
3127 vxlan->flags &= ~VXLAN_F_LEARN;
3128 }
3129 } else if (!changelink) {
3130 /* default to learn on a new device */
3131 conf->flags |= VXLAN_F_LEARN;
3132 }
d342894c 3133
8bcdc4f3
RP		 3134 if (data[IFLA_VXLAN_AGEING]) {
3135 if (changelink)
3136 return -EOPNOTSUPP;
3137 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3138 }
d342894c 3139
8bcdc4f3
RP		 3140 if (data[IFLA_VXLAN_PROXY]) {
3141 if (changelink)
3142 return -EOPNOTSUPP;
3143 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3144 conf->flags |= VXLAN_F_PROXY;
3145 }
e4f67add 3146
8bcdc4f3
RP		 3147 if (data[IFLA_VXLAN_RSC]) {
3148 if (changelink)
3149 return -EOPNOTSUPP;
3150 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3151 conf->flags |= VXLAN_F_RSC;
3152 }
e4f67add 3153
8bcdc4f3
RP		 3154 if (data[IFLA_VXLAN_L2MISS]) {
3155 if (changelink)
3156 return -EOPNOTSUPP;
3157 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3158 conf->flags |= VXLAN_F_L2MISS;
3159 }
e4f67add 3160
8bcdc4f3
RP		 3161 if (data[IFLA_VXLAN_L3MISS]) {
3162 if (changelink)
3163 return -EOPNOTSUPP;
3164 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3165 conf->flags |= VXLAN_F_L3MISS;
3166 }
e4f67add 3167
8bcdc4f3
RP		 3168 if (data[IFLA_VXLAN_LIMIT]) {
3169 if (changelink)
3170 return -EOPNOTSUPP;
3171 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3172 }
d342894c 3173
8bcdc4f3
RP		 3174 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3175 if (changelink)
3176 return -EOPNOTSUPP;
3177 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3178 conf->flags |= VXLAN_F_COLLECT_METADATA;
3179 }
f8a9b1bc 3180
05f47d69 3181 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3182 if (!changelink) {
3183 const struct ifla_vxlan_port_range *p
3184 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3185 conf->port_min = ntohs(p->low);
3186 conf->port_max = ntohs(p->high);
3187 } else {
3188 return -EOPNOTSUPP;
3189 }
05f47d69 3190 }
3191
8bcdc4f3
RP		 3192 if (data[IFLA_VXLAN_PORT]) {
3193 if (changelink)
3194 return -EOPNOTSUPP;
3195 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3196 }
823aa873 3197
8bcdc4f3
RP		 3198 if (data[IFLA_VXLAN_UDP_CSUM]) {
3199 if (changelink)
3200 return -EOPNOTSUPP;
3201 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3202 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3203 }
359a0ea9 3204
8bcdc4f3
RP		 3205 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3206 if (changelink)
3207 return -EOPNOTSUPP;
3208 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3209 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3210 }
359a0ea9 3211
8bcdc4f3
RP		 3212 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3213 if (changelink)
3214 return -EOPNOTSUPP;
3215 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3216 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3217 }
359a0ea9 3218
8bcdc4f3
RP		 3219 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3220 if (changelink)
3221 return -EOPNOTSUPP;
3222 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3223 conf->flags |= VXLAN_F_REMCSUM_TX;
3224 }
dfd8645e 3225
8bcdc4f3
RP		 3226 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3227 if (changelink)
3228 return -EOPNOTSUPP;
3229 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3230 conf->flags |= VXLAN_F_REMCSUM_RX;
3231 }
3232
3233 if (data[IFLA_VXLAN_GBP]) {
3234 if (changelink)
3235 return -EOPNOTSUPP;
3236 conf->flags |= VXLAN_F_GBP;
3237 }
3238
3239 if (data[IFLA_VXLAN_GPE]) {
3240 if (changelink)
3241 return -EOPNOTSUPP;
3242 conf->flags |= VXLAN_F_GPE;
3243 }
3244
3245 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3246 if (changelink)
3247 return -EOPNOTSUPP;
3248 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3249 }
3250
3251 if (tb[IFLA_MTU]) {
3252 if (changelink)
3253 return -EOPNOTSUPP;
3254 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3255 }
3256
3257 return 0;
3258}
3259
3260static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3261 struct nlattr *tb[], struct nlattr *data[])
3262{
8bcdc4f3
RP		 3263 struct vxlan_config conf;
3264 int err;
3265
3266 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3267 if (err)
3268 return err;
3269
c80498e3 3270 return __vxlan_dev_create(src_net, dev, &conf);
8bcdc4f3 3271}
dfd8645e 3272
8bcdc4f3
RP		 3273static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3274 struct nlattr *data[])
3275{
3276 struct vxlan_dev *vxlan = netdev_priv(dev);
3277 struct vxlan_rdst *dst = &vxlan->default_dst;
3278 struct vxlan_rdst old_dst;
3279 struct vxlan_config conf;
3280 int err;
3281
3282 err = vxlan_nl2conf(tb, data,
3283 dev, &conf, true);
3284 if (err)
3285 return err;
3511494c 3286
8bcdc4f3 3287 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3288
8bcdc4f3
RP		 3289 err = vxlan_dev_configure(vxlan->net, dev, &conf, true);
3290 if (err)
3291 return err;
0ace2ca8 3292
8bcdc4f3
RP		 3293 /* handle default dst entry */
3294 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3295 spin_lock_bh(&vxlan->hash_lock);
3296 if (!vxlan_addr_any(&old_dst.remote_ip))
3297 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3298 old_dst.remote_ip,
3299 vxlan->cfg.dst_port,
3300 old_dst.remote_vni,
3301 old_dst.remote_vni,
3302 old_dst.remote_ifindex, 0);
3303
3304 if (!vxlan_addr_any(&dst->remote_ip)) {
3305 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3306 &dst->remote_ip,
3307 NUD_REACHABLE | NUD_PERMANENT,
3308 NLM_F_CREATE | NLM_F_APPEND,
3309 vxlan->cfg.dst_port,
3310 dst->remote_vni,
3311 dst->remote_vni,
3312 dst->remote_ifindex,
3313 NTF_SELF);
3314 if (err) {
3315 spin_unlock_bh(&vxlan->hash_lock);
3316 return err;
3317 }
3318 }
3319 spin_unlock_bh(&vxlan->hash_lock);
3320 }
ce577668 3321
8bcdc4f3 3322 return 0;
d342894c 3323}
3324
3325static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3326{
3327 struct vxlan_dev *vxlan = netdev_priv(dev);
3328
8b3f9337
RP		 3329 vxlan_flush(vxlan, true);
3330
58ce31cc 3331 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3332 list_del(&vxlan->next);
d342894c 3333 unregister_netdevice_queue(dev, head);
3334}
3335
3336static size_t vxlan_get_size(const struct net_device *dev)
3337{
3338
3339 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3340 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3341 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3342 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3343 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3344 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3345 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3346 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3347 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3348 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3349 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3350 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3351 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3352 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3353 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3354 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3355 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3356 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3357 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3358 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3359 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3360 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3361 0;
3362}
3363
3364static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3365{
3366 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3367 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3368 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3369 .low = htons(vxlan->cfg.port_min),
3370 .high = htons(vxlan->cfg.port_max),
05f47d69 3371 };
d342894c 3372
54bfd872 3373 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3374 goto nla_put_failure;
3375
e4c7ed41
CW		 3376 if (!vxlan_addr_any(&dst->remote_ip)) {
3377 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3378 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3379 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3380 goto nla_put_failure;
3381#if IS_ENABLED(CONFIG_IPV6)
3382 } else {
930345ea
JB		 3383 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3384 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3385 goto nla_put_failure;
3386#endif
3387 }
3388 }
d342894c 3389
c7995c43 3390 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3391 goto nla_put_failure;
3392
0dfbdf41
TG		 3393 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3394 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3395 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3396 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3397 goto nla_put_failure;
3398#if IS_ENABLED(CONFIG_IPV6)
3399 } else {
930345ea 3400 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3401 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3402 goto nla_put_failure;
3403#endif
3404 }
3405 }
d342894c 3406
0dfbdf41
TG		 3407 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3408 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3409 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3410 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3411 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3412 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3413 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3414 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3415 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3416 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3417 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3418 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3419 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3420 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3421 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3422 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3423 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3424 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3425 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3426 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3427 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3428 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3429 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3430 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3431 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3432 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3433 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3434 goto nla_put_failure;
3435
05f47d69 3436 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3437 goto nla_put_failure;
3438
3511494c
TG		 3439 if (vxlan->flags & VXLAN_F_GBP &&
3440 nla_put_flag(skb, IFLA_VXLAN_GBP))
3441 goto nla_put_failure;
3442
e1e5314d
JB		 3443 if (vxlan->flags & VXLAN_F_GPE &&
3444 nla_put_flag(skb, IFLA_VXLAN_GPE))
3445 goto nla_put_failure;
3446
0ace2ca8
TH		 3447 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3448 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3449 goto nla_put_failure;
3450
d342894c 3451 return 0;
3452
3453nla_put_failure:
3454 return -EMSGSIZE;
3455}
3456
1728d4fa
ND		 3457static struct net *vxlan_get_link_net(const struct net_device *dev)
3458{
3459 struct vxlan_dev *vxlan = netdev_priv(dev);
3460
3461 return vxlan->net;
3462}
3463
d342894c 3464static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3465 .kind = "vxlan",
3466 .maxtype = IFLA_VXLAN_MAX,
3467 .policy = vxlan_policy,
3468 .priv_size = sizeof(struct vxlan_dev),
3469 .setup = vxlan_setup,
3470 .validate = vxlan_validate,
3471 .newlink = vxlan_newlink,
8bcdc4f3 3472 .changelink = vxlan_changelink,
d342894c 3473 .dellink = vxlan_dellink,
3474 .get_size = vxlan_get_size,
3475 .fill_info = vxlan_fill_info,
1728d4fa 3476 .get_link_net = vxlan_get_link_net,
d342894c 3477};
3478
cf5da330
ND		 3479struct net_device *vxlan_dev_create(struct net *net, const char *name,
3480 u8 name_assign_type,
3481 struct vxlan_config *conf)
3482{
3483 struct nlattr *tb[IFLA_MAX + 1];
3484 struct net_device *dev;
3485 int err;
3486
3487 memset(&tb, 0, sizeof(tb));
3488
3489 dev = rtnl_create_link(net, name, name_assign_type,
3490 &vxlan_link_ops, tb);
3491 if (IS_ERR(dev))
3492 return dev;
3493
c80498e3 3494 err = __vxlan_dev_create(net, dev, conf);
cf5da330
ND		 3495 if (err < 0) {
3496 free_netdev(dev);
3497 return ERR_PTR(err);
3498 }
3499
3500 err = rtnl_configure_link(dev, NULL);
3501 if (err < 0) {
3502 LIST_HEAD(list_kill);
3503
3504 vxlan_dellink(dev, &list_kill);
3505 unregister_netdevice_many(&list_kill);
3506 return ERR_PTR(err);
3507 }
3508
3509 return dev;
3510}
3511EXPORT_SYMBOL_GPL(vxlan_dev_create);
3512
acaf4e70
DB		 3513static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3514 struct net_device *dev)
3515{
3516 struct vxlan_dev *vxlan, *next;
3517 LIST_HEAD(list_kill);
3518
3519 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3520 struct vxlan_rdst *dst = &vxlan->default_dst;
3521
3522 /* In case we created vxlan device with carrier
3523 * and we loose the carrier due to module unload
3524 * we also need to remove vxlan device. In other
3525 * cases, it's not necessary and remote_ifindex
3526 * is 0 here, so no matches.
3527 */
3528 if (dst->remote_ifindex == dev->ifindex)
3529 vxlan_dellink(vxlan->dev, &list_kill);
3530 }
3531
3532 unregister_netdevice_many(&list_kill);
3533}
3534
b7aade15
HFS		 3535static int vxlan_netdevice_event(struct notifier_block *unused,
3536 unsigned long event, void *ptr)
acaf4e70
DB		 3537{
3538 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3539 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3540
783c1463 3541 if (event == NETDEV_UNREGISTER)
acaf4e70 3542 vxlan_handle_lowerdev_unregister(vn, dev);
7c46a640 3543 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
b7aade15 3544 vxlan_push_rx_ports(dev);
acaf4e70
DB		 3545
3546 return NOTIFY_DONE;
3547}
3548
3549static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3550 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3551};
3552
d342894c 3553static __net_init int vxlan_init_net(struct net *net)
3554{
3555 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3556 unsigned int h;
d342894c 3557
553675fb 3558 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3559 spin_lock_init(&vn->sock_lock);
d342894c 3560
553675fb 3561 for (h = 0; h < PORT_HASH_SIZE; ++h)
3562 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3563
3564 return 0;
3565}
3566
f01ec1c0
ND		 3567static void __net_exit vxlan_exit_net(struct net *net)
3568{
3569 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3570 struct vxlan_dev *vxlan, *next;
3571 struct net_device *dev, *aux;
3572 LIST_HEAD(list);
3573
3574 rtnl_lock();
3575 for_each_netdev_safe(net, dev, aux)
3576 if (dev->rtnl_link_ops == &vxlan_link_ops)
3577 unregister_netdevice_queue(dev, &list);
3578
3579 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3580 /* If vxlan->dev is in the same netns, it has already been added
3581 * to the list by the previous loop.
3582 */
58ce31cc
TH		 3583 if (!net_eq(dev_net(vxlan->dev), net)) {
3584 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3585 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3586 }
f01ec1c0
ND		 3587 }
3588
3589 unregister_netdevice_many(&list);
3590 rtnl_unlock();
3591}
3592
d342894c 3593static struct pernet_operations vxlan_net_ops = {
3594 .init = vxlan_init_net,
f01ec1c0 3595 .exit = vxlan_exit_net,
d342894c 3596 .id = &vxlan_net_id,
3597 .size = sizeof(struct vxlan_net),
3598};
3599
3600static int __init vxlan_init_module(void)
3601{
3602 int rc;
3603
3604 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3605
783c1463 3606 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3607 if (rc)
3608 goto out1;
3609
acaf4e70 3610 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3611 if (rc)
3612 goto out2;
3613
acaf4e70
DB		 3614 rc = rtnl_link_register(&vxlan_link_ops);
3615 if (rc)
3616 goto out3;
d342894c 3617
acaf4e70
DB		 3618 return 0;
3619out3:
3620 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3621out2:
783c1463 3622 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3623out1:
3624 return rc;
3625}
7332a13b 3626late_initcall(vxlan_init_module);
d342894c 3627
3628static void __exit vxlan_cleanup_module(void)
3629{
b7153984 3630 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3631 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3632 unregister_pernet_subsys(&vxlan_net_ops);
3633 /* rcu_barrier() is called by netns */
d342894c 3634}
3635module_exit(vxlan_cleanup_module);
3636
3637MODULE_LICENSE("GPL");
3638MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3639MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3640MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3641MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror