projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: thunderx: Fix for Qset error due to CQ full
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
acbf74a7 45#include <net/udp_tunnel.h>
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
ee122c79 52#include <net/dst_metadata.h>
d342894c 53
54#define VXLAN_VERSION "0.1"
55
553675fb 56#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
23c578bf 61/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 63 * for compatibility with early adopters.
23c578bf 64 */
9daaa397
SH		 65static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 67MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
60d9d4c6 73static int vxlan_net_id;
0dfbdf41 74static struct rtnl_link_ops vxlan_link_ops;
553675fb 75
7256eac1 76static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 77
205f356d 78static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
d342894c 87/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
3e61aa8f 93 struct list_head remotes;
7177a3b0 94 u8 eth_addr[ETH_ALEN];
d342894c 95 u16 state; /* see ndm_state */
ae884082 96 u8 flags; /* see ndm_flags */
d342894c 97};
98
d342894c 99/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
758c57d1 101static struct workqueue_struct *vxlan_wq;
d342894c 102
ee122c79
TG		 103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
e7030878
TG		 105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
ee122c79
TG		 107}
108
e4c7ed41
CW		 109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
f0ef3126
JB		 113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
f0ef3126
JB		 123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
f0ef3126
JB		 131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
f0ef3126 139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
e4c7ed41
CW		 150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 153 const union vxlan_addr *ip)
e4c7ed41 154{
f0ef3126 155 if (ip->sa.sa_family == AF_INET6)
930345ea 156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 157 else
930345ea 158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
f0ef3126 166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
f0ef3126 171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
f0ef3126 176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
f0ef3126
JB		 181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
e4c7ed41
CW		 190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 193 const union vxlan_addr *ip)
e4c7ed41 194{
930345ea 195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 196}
197#endif
198
553675fb 199/* Virtual Network hash table head */
200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
201{
202 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
553675fb 210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
3e61aa8f
SH		 213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
5ca5461c 216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 217{
5ca5461c 218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 224}
225
ac5132d1
TG		 226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
553675fb 231{
232 struct vxlan_sock *vs;
af33c1ad
TH		 233
234 flags &= VXLAN_F_RCV_FLAGS;
553675fb 235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 238 vxlan_get_sk_family(vs) == family &&
af33c1ad 239 vs->flags == flags)
553675fb 240 return vs;
241 }
242 return NULL;
d342894c 243}
244
5cfccc5a 245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 246{
247 struct vxlan_dev *vxlan;
d342894c 248
553675fb 249 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 250 if (vxlan->default_dst.remote_vni == id)
d342894c 251 return vxlan;
252 }
253
254 return NULL;
255}
256
5cfccc5a 257/* Look up VNI in a per net namespace table */
19ca9fc1 258static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 259 sa_family_t family, __be16 port,
260 u32 flags)
5cfccc5a
PS		 261{
262 struct vxlan_sock *vs;
263
ac5132d1 264 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 265 if (!vs)
266 return NULL;
267
268 return vxlan_vs_find_vni(vs, id);
269}
270
d342894c 271/* Fill in neighbour message in skbuff. */
272static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 273 const struct vxlan_fdb *fdb,
274 u32 portid, u32 seq, int type, unsigned int flags,
275 const struct vxlan_rdst *rdst)
d342894c 276{
277 unsigned long now = jiffies;
278 struct nda_cacheinfo ci;
279 struct nlmsghdr *nlh;
280 struct ndmsg *ndm;
e4f67add 281 bool send_ip, send_eth;
d342894c 282
283 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
284 if (nlh == NULL)
285 return -EMSGSIZE;
286
287 ndm = nlmsg_data(nlh);
288 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 289
290 send_eth = send_ip = true;
291
292 if (type == RTM_GETNEIGH) {
293 ndm->ndm_family = AF_INET;
e4c7ed41 294 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
d342894c 298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 300 ndm->ndm_flags = fdb->flags;
545469f7 301 ndm->ndm_type = RTN_UNICAST;
d342894c 302
193523bf 303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 304 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 305 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 306 goto nla_put_failure;
307
e4f67add 308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 309 goto nla_put_failure;
310
e4c7ed41 311 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 312 goto nla_put_failure;
313
0dfbdf41 314 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 315 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
316 goto nla_put_failure;
c7995c43 317 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 318 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 319 goto nla_put_failure;
320 if (rdst->remote_ifindex &&
321 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 322 goto nla_put_failure;
323
324 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
325 ci.ndm_confirmed = 0;
326 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
327 ci.ndm_refcnt = 0;
328
329 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
330 goto nla_put_failure;
331
053c095a
JB		 332 nlmsg_end(skb, nlh);
333 return 0;
d342894c 334
335nla_put_failure:
336 nlmsg_cancel(skb, nlh);
337 return -EMSGSIZE;
338}
339
340static inline size_t vxlan_nlmsg_size(void)
341{
342 return NLMSG_ALIGN(sizeof(struct ndmsg))
343 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 344 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 345 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 346 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
347 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 348 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 349 + nla_total_size(sizeof(struct nda_cacheinfo));
350}
351
9e4b93f9
ND		 352static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
353 struct vxlan_rdst *rd, int type)
d342894c 354{
355 struct net *net = dev_net(vxlan->dev);
356 struct sk_buff *skb;
357 int err = -ENOBUFS;
358
359 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
360 if (skb == NULL)
361 goto errout;
362
9e4b93f9 363 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 364 if (err < 0) {
365 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
366 WARN_ON(err == -EMSGSIZE);
367 kfree_skb(skb);
368 goto errout;
369 }
370
371 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
372 return;
373errout:
374 if (err < 0)
375 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
376}
377
e4c7ed41 378static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 379{
380 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 381 struct vxlan_fdb f = {
382 .state = NUD_STALE,
383 };
384 struct vxlan_rdst remote = {
e4c7ed41 385 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 386 .remote_vni = VXLAN_N_VID,
387 };
3e61aa8f 388
9e4b93f9 389 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 390}
391
392static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
393{
bb3fd687
SH		 394 struct vxlan_fdb f = {
395 .state = NUD_STALE,
396 };
9e4b93f9 397 struct vxlan_rdst remote = { };
e4f67add 398
e4f67add
DS		 399 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
400
9e4b93f9 401 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 402}
403
d342894c 404/* Hash Ethernet address */
405static u32 eth_hash(const unsigned char *addr)
406{
407 u64 value = get_unaligned((u64 *)addr);
408
409 /* only want 6 bytes */
410#ifdef __BIG_ENDIAN
d342894c 411 value >>= 16;
321fb991 412#else
413 value <<= 16;
d342894c 414#endif
415 return hash_64(value, FDB_HASH_BITS);
416}
417
418/* Hash chain to use given mac address */
419static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
420 const u8 *mac)
421{
422 return &vxlan->fdb_head[eth_hash(mac)];
423}
424
425/* Look up Ethernet address in forwarding table */
014be2c8 426static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 427 const u8 *mac)
d342894c 428{
429 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
430 struct vxlan_fdb *f;
d342894c 431
b67bfe0d 432 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 433 if (ether_addr_equal(mac, f->eth_addr))
d342894c 434 return f;
435 }
436
437 return NULL;
438}
439
014be2c8
SS		 440static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
441 const u8 *mac)
442{
443 struct vxlan_fdb *f;
444
445 f = __vxlan_find_mac(vxlan, mac);
446 if (f)
447 f->used = jiffies;
448
449 return f;
450}
451
a5e7c10a
MR		 452/* caller should hold vxlan->hash_lock */
453static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 454 union vxlan_addr *ip, __be16 port,
a5e7c10a 455 __u32 vni, __u32 ifindex)
6681712d 456{
3e61aa8f 457 struct vxlan_rdst *rd;
6681712d 458
3e61aa8f 459 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 460 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 461 rd->remote_port == port &&
462 rd->remote_vni == vni &&
463 rd->remote_ifindex == ifindex)
a5e7c10a 464 return rd;
6681712d 465 }
3e61aa8f 466
a5e7c10a
MR		 467 return NULL;
468}
469
906dc186
TR		 470/* Replace destination of unicast mac */
471static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 472 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 473{
474 struct vxlan_rdst *rd;
475
476 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
477 if (rd)
478 return 0;
479
480 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
481 if (!rd)
482 return 0;
e4c7ed41 483 rd->remote_ip = *ip;
906dc186
TR		 484 rd->remote_port = port;
485 rd->remote_vni = vni;
486 rd->remote_ifindex = ifindex;
487 return 1;
488}
489
a5e7c10a
MR		 490/* Add/update destinations for multicast */
491static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 492 union vxlan_addr *ip, __be16 port, __u32 vni,
493 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 494{
495 struct vxlan_rdst *rd;
496
497 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
498 if (rd)
499 return 0;
500
6681712d
DS		 501 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
502 if (rd == NULL)
503 return -ENOBUFS;
e4c7ed41 504 rd->remote_ip = *ip;
6681712d
DS		 505 rd->remote_port = port;
506 rd->remote_vni = vni;
507 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 508
509 list_add_tail_rcu(&rd->list, &f->remotes);
510
9e4b93f9 511 *rdp = rd;
6681712d
DS		 512 return 1;
513}
514
dfd8645e
TH		 515static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
516 unsigned int off,
517 struct vxlanhdr *vh, size_t hdrlen,
0ace2ca8
TH		 518 u32 data, struct gro_remcsum *grc,
519 bool nopartial)
dfd8645e 520{
b7fe10e5 521 size_t start, offset;
dfd8645e
TH		 522
523 if (skb->remcsum_offload)
b7fe10e5 524 return vh;
dfd8645e
TH		 525
526 if (!NAPI_GRO_CB(skb)->csum_valid)
527 return NULL;
528
529 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
530 offset = start + ((data & VXLAN_RCO_UDP) ?
531 offsetof(struct udphdr, check) :
532 offsetof(struct tcphdr, check));
533
b7fe10e5
TH		 534 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
535 start, offset, grc, nopartial);
dfd8645e
TH		 536
537 skb->remcsum_offload = 1;
538
539 return vh;
540}
541
a2b12f3c
TH		 542static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
543 struct sk_buff *skb,
544 struct udp_offload *uoff)
dc01e7d3
OG		 545{
546 struct sk_buff *p, **pp = NULL;
547 struct vxlanhdr *vh, *vh2;
9b174d88 548 unsigned int hlen, off_vx;
dc01e7d3 549 int flush = 1;
dfd8645e
TH		 550 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
551 udp_offloads);
552 u32 flags;
26c4f7da
TH		 553 struct gro_remcsum grc;
554
555 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 556
557 off_vx = skb_gro_offset(skb);
558 hlen = off_vx + sizeof(*vh);
559 vh = skb_gro_header_fast(skb, off_vx);
560 if (skb_gro_header_hard(skb, hlen)) {
561 vh = skb_gro_header_slow(skb, hlen, off_vx);
562 if (unlikely(!vh))
563 goto out;
564 }
dc01e7d3 565
dfd8645e
TH		 566 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
567
568 flags = ntohl(vh->vx_flags);
569
570 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
571 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
0ace2ca8
TH		 572 ntohl(vh->vx_vni), &grc,
573 !!(vs->flags &
574 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 575
576 if (!vh)
577 goto out;
578 }
579
b7fe10e5
TH		 580 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
581
dc01e7d3
OG		 582 flush = 0;
583
584 for (p = *head; p; p = p->next) {
585 if (!NAPI_GRO_CB(p)->same_flow)
586 continue;
587
588 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 589 if (vh->vx_flags != vh2->vx_flags ||
590 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 591 NAPI_GRO_CB(p)->same_flow = 0;
592 continue;
593 }
dc01e7d3
OG		 594 }
595
9b174d88 596 pp = eth_gro_receive(head, skb);
dc01e7d3 597
dc01e7d3 598out:
26c4f7da 599 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 600 NAPI_GRO_CB(skb)->flush |= flush;
601
602 return pp;
603}
604
a2b12f3c
TH		 605static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
606 struct udp_offload *uoff)
dc01e7d3 607{
cfdf1e1b
JG		 608 udp_tunnel_gro_complete(skb, nhoff);
609
9b174d88 610 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 611}
612
53cf5275 613/* Notify netdevs that UDP port started listening */
dc01e7d3 614static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 615{
616 struct net_device *dev;
dc01e7d3 617 struct sock *sk = vs->sock->sk;
53cf5275 618 struct net *net = sock_net(sk);
705cc62f 619 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 620 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 621 int err;
622
623 if (sa_family == AF_INET) {
787d7ac3 624 err = udp_add_offload(net, &vs->udp_offloads);
dc01e7d3
OG		 625 if (err)
626 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
627 }
53cf5275
JG		 628
629 rcu_read_lock();
630 for_each_netdev_rcu(net, dev) {
631 if (dev->netdev_ops->ndo_add_vxlan_port)
632 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
633 port);
634 }
635 rcu_read_unlock();
636}
637
638/* Notify netdevs that UDP port is no more listening */
dc01e7d3 639static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 640{
641 struct net_device *dev;
dc01e7d3 642 struct sock *sk = vs->sock->sk;
53cf5275 643 struct net *net = sock_net(sk);
705cc62f 644 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 645 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 646
647 rcu_read_lock();
648 for_each_netdev_rcu(net, dev) {
649 if (dev->netdev_ops->ndo_del_vxlan_port)
650 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
651 port);
652 }
653 rcu_read_unlock();
dc01e7d3
OG		 654
655 if (sa_family == AF_INET)
656 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 657}
658
d342894c 659/* Add new entry to forwarding table -- assumes lock held */
660static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 661 const u8 *mac, union vxlan_addr *ip,
6681712d 662 __u16 state, __u16 flags,
73cf3317 663 __be16 port, __u32 vni, __u32 ifindex,
ae884082 664 __u8 ndm_flags)
d342894c 665{
9e4b93f9 666 struct vxlan_rdst *rd = NULL;
d342894c 667 struct vxlan_fdb *f;
668 int notify = 0;
669
014be2c8 670 f = __vxlan_find_mac(vxlan, mac);
d342894c 671 if (f) {
672 if (flags & NLM_F_EXCL) {
673 netdev_dbg(vxlan->dev,
674 "lost race to create %pM\n", mac);
675 return -EEXIST;
676 }
677 if (f->state != state) {
678 f->state = state;
679 f->updated = jiffies;
680 notify = 1;
681 }
ae884082
DS		 682 if (f->flags != ndm_flags) {
683 f->flags = ndm_flags;
684 f->updated = jiffies;
685 notify = 1;
686 }
906dc186
TR		 687 if ((flags & NLM_F_REPLACE)) {
688 /* Only change unicasts */
689 if (!(is_multicast_ether_addr(f->eth_addr) ||
690 is_zero_ether_addr(f->eth_addr))) {
60840429 691 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 692 ifindex);
906dc186
TR		 693 } else
694 return -EOPNOTSUPP;
695 }
6681712d 696 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 697 (is_multicast_ether_addr(f->eth_addr) ||
698 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 699 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
700 &rd);
6681712d
DS		 701
702 if (rc < 0)
703 return rc;
704 notify |= rc;
705 }
d342894c 706 } else {
707 if (!(flags & NLM_F_CREATE))
708 return -ENOENT;
709
0dfbdf41
TG		 710 if (vxlan->cfg.addrmax &&
711 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 712 return -ENOSPC;
713
906dc186
TR		 714 /* Disallow replace to add a multicast entry */
715 if ((flags & NLM_F_REPLACE) &&
716 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
717 return -EOPNOTSUPP;
718
e4c7ed41 719 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 720 f = kmalloc(sizeof(*f), GFP_ATOMIC);
721 if (!f)
722 return -ENOMEM;
723
724 notify = 1;
d342894c 725 f->state = state;
ae884082 726 f->flags = ndm_flags;
d342894c 727 f->updated = f->used = jiffies;
3e61aa8f 728 INIT_LIST_HEAD(&f->remotes);
d342894c 729 memcpy(f->eth_addr, mac, ETH_ALEN);
730
9e4b93f9 731 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 732
d342894c 733 ++vxlan->addrcnt;
734 hlist_add_head_rcu(&f->hlist,
735 vxlan_fdb_head(vxlan, mac));
736 }
737
9e4b93f9
ND		 738 if (notify) {
739 if (rd == NULL)
740 rd = first_remote_rtnl(f);
741 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
742 }
d342894c 743
744 return 0;
745}
746
6706c82e 747static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 748{
749 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 750 struct vxlan_rdst *rd, *nd;
6681712d 751
3e61aa8f 752 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 753 kfree(rd);
6681712d
DS		 754 kfree(f);
755}
756
d342894c 757static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
758{
759 netdev_dbg(vxlan->dev,
760 "delete %pM\n", f->eth_addr);
761
762 --vxlan->addrcnt;
9e4b93f9 763 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 764
765 hlist_del_rcu(&f->hlist);
6681712d 766 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 767}
768
f0b074be 769static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 770 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 771{
6681712d 772 struct net *net = dev_net(vxlan->dev);
e4c7ed41 773 int err;
d342894c 774
f0b074be 775 if (tb[NDA_DST]) {
e4c7ed41
CW		 776 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
777 if (err)
778 return err;
f0b074be 779 } else {
e4c7ed41
CW		 780 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
781 if (remote->sa.sa_family == AF_INET) {
782 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
783 ip->sa.sa_family = AF_INET;
784#if IS_ENABLED(CONFIG_IPV6)
785 } else {
786 ip->sin6.sin6_addr = in6addr_any;
787 ip->sa.sa_family = AF_INET6;
788#endif
789 }
f0b074be 790 }
d342894c 791
6681712d 792 if (tb[NDA_PORT]) {
73cf3317 793 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 794 return -EINVAL;
f0b074be
MR		 795 *port = nla_get_be16(tb[NDA_PORT]);
796 } else {
0dfbdf41 797 *port = vxlan->cfg.dst_port;
f0b074be 798 }
6681712d
DS		 799
800 if (tb[NDA_VNI]) {
801 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
802 return -EINVAL;
f0b074be
MR		 803 *vni = nla_get_u32(tb[NDA_VNI]);
804 } else {
805 *vni = vxlan->default_dst.remote_vni;
806 }
6681712d
DS		 807
808 if (tb[NDA_IFINDEX]) {
5abb0029 809 struct net_device *tdev;
6681712d
DS		 810
811 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
812 return -EINVAL;
f0b074be 813 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 814 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 815 if (!tdev)
6681712d 816 return -EADDRNOTAVAIL;
f0b074be
MR		 817 } else {
818 *ifindex = 0;
819 }
820
821 return 0;
822}
823
824/* Add static entry (via netlink) */
825static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
826 struct net_device *dev,
f6f6424b 827 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 828{
829 struct vxlan_dev *vxlan = netdev_priv(dev);
830 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 831 union vxlan_addr ip;
f0b074be
MR		 832 __be16 port;
833 u32 vni, ifindex;
834 int err;
835
836 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
837 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
838 ndm->ndm_state);
839 return -EINVAL;
840 }
841
842 if (tb[NDA_DST] == NULL)
843 return -EINVAL;
844
845 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
846 if (err)
847 return err;
6681712d 848
5933a7bb
MR		 849 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
850 return -EAFNOSUPPORT;
851
d342894c 852 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 853 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 854 port, vni, ifindex, ndm->ndm_flags);
d342894c 855 spin_unlock_bh(&vxlan->hash_lock);
856
857 return err;
858}
859
860/* Delete entry (via netlink) */
1690be63
VY		 861static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
862 struct net_device *dev,
f6f6424b 863 const unsigned char *addr, u16 vid)
d342894c 864{
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 struct vxlan_fdb *f;
bc7892ba 867 struct vxlan_rdst *rd = NULL;
e4c7ed41 868 union vxlan_addr ip;
bc7892ba
MR		 869 __be16 port;
870 u32 vni, ifindex;
871 int err;
872
873 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
874 if (err)
875 return err;
876
877 err = -ENOENT;
d342894c 878
879 spin_lock_bh(&vxlan->hash_lock);
880 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 881 if (!f)
882 goto out;
883
e4c7ed41
CW		 884 if (!vxlan_addr_any(&ip)) {
885 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 886 if (!rd)
887 goto out;
888 }
889
890 err = 0;
891
892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
894 */
895 if (rd && !list_is_singular(&f->remotes)) {
896 list_del_rcu(&rd->list);
9e4b93f9 897 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 898 kfree_rcu(rd, rcu);
bc7892ba 899 goto out;
d342894c 900 }
bc7892ba
MR		 901
902 vxlan_fdb_destroy(vxlan, f);
903
904out:
d342894c 905 spin_unlock_bh(&vxlan->hash_lock);
906
907 return err;
908}
909
910/* Dump forwarding table */
911static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 912 struct net_device *dev,
913 struct net_device *filter_dev, int idx)
d342894c 914{
915 struct vxlan_dev *vxlan = netdev_priv(dev);
916 unsigned int h;
917
918 for (h = 0; h < FDB_HASH_SIZE; ++h) {
919 struct vxlan_fdb *f;
d342894c 920 int err;
921
b67bfe0d 922 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 923 struct vxlan_rdst *rd;
6681712d 924
3e61aa8f 925 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 926 if (idx < cb->args[0])
927 goto skip;
928
6681712d
DS		 929 err = vxlan_fdb_info(skb, vxlan, f,
930 NETLINK_CB(cb->skb).portid,
931 cb->nlh->nlmsg_seq,
932 RTM_NEWNEIGH,
933 NLM_F_MULTI, rd);
934 if (err < 0)
3e61aa8f 935 goto out;
3e61aa8f 936skip:
07a51cd3
AW		 937 ++idx;
938 }
d342894c 939 }
940 }
3e61aa8f 941out:
d342894c 942 return idx;
943}
944
945/* Watch incoming packets to learn mapping between Ethernet address
946 * and Tunnel endpoint.
c4b49512 947 * Return true if packet is bogus and should be dropped.
d342894c 948 */
26a41ae6 949static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 950 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 951{
952 struct vxlan_dev *vxlan = netdev_priv(dev);
953 struct vxlan_fdb *f;
d342894c 954
955 f = vxlan_find_mac(vxlan, src_mac);
956 if (likely(f)) {
5ca5461c 957 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 958
e4c7ed41 959 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 960 return false;
961
962 /* Don't migrate static entries, drop packets */
eb064c3b 963 if (f->state & NUD_NOARP)
26a41ae6 964 return true;
d342894c 965
966 if (net_ratelimit())
967 netdev_info(dev,
e4c7ed41 968 "%pM migrated from %pIS to %pIS\n",
a4870f79 969 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 970
e4c7ed41 971 rdst->remote_ip = *src_ip;
d342894c 972 f->updated = jiffies;
9e4b93f9 973 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 974 } else {
975 /* learned new entry */
976 spin_lock(&vxlan->hash_lock);
3bf74b1a 977
978 /* close off race between vxlan_flush and incoming packets */
979 if (netif_running(dev))
980 vxlan_fdb_create(vxlan, src_mac, src_ip,
981 NUD_REACHABLE,
982 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 983 vxlan->cfg.dst_port,
3bf74b1a 984 vxlan->default_dst.remote_vni,
985 0, NTF_SELF);
d342894c 986 spin_unlock(&vxlan->hash_lock);
987 }
26a41ae6 988
989 return false;
d342894c 990}
991
d342894c 992/* See if multicast group is already in use by other ID */
95ab0991 993static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 994{
553675fb 995 struct vxlan_dev *vxlan;
b1be00a6 996 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 997
95ab0991
G		 998 /* The vxlan_sock is only used by dev, leaving group has
999 * no effect on other vxlan devices.
1000 */
b1be00a6
JB		 1001 if (family == AF_INET && dev->vn4_sock &&
1002 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1003 return false;
b1be00a6
JB		 1004#if IS_ENABLED(CONFIG_IPV6)
1005 if (family == AF_INET6 && dev->vn6_sock &&
1006 atomic_read(&dev->vn6_sock->refcnt) == 1)
1007 return false;
1008#endif
95ab0991 1009
553675fb 1010 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1011 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1012 continue;
d342894c 1013
b1be00a6 1014 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1015 continue;
b1be00a6
JB		 1016#if IS_ENABLED(CONFIG_IPV6)
1017 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1018 continue;
1019#endif
95ab0991
G		 1020
1021 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1022 &dev->default_dst.remote_ip))
1023 continue;
1024
1025 if (vxlan->default_dst.remote_ifindex !=
1026 dev->default_dst.remote_ifindex)
1027 continue;
1028
1029 return true;
553675fb 1030 }
d342894c 1031
1032 return false;
1033}
1034
b1be00a6 1035static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1036{
b1be00a6 1037 struct vxlan_net *vn;
012a5729 1038
b1be00a6
JB		 1039 if (!vs)
1040 return;
7c47cedf
SH		 1041 if (!atomic_dec_and_test(&vs->refcnt))
1042 return;
d342894c 1043
b1be00a6 1044 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1045 spin_lock(&vn->sock_lock);
7c47cedf 1046 hlist_del_rcu(&vs->hlist);
dc01e7d3 1047 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1048 spin_unlock(&vn->sock_lock);
1049
7c47cedf 1050 queue_work(vxlan_wq, &vs->del_work);
d342894c 1051}
1052
b1be00a6
JB		 1053static void vxlan_sock_release(struct vxlan_dev *vxlan)
1054{
1055 __vxlan_sock_release(vxlan->vn4_sock);
1056#if IS_ENABLED(CONFIG_IPV6)
1057 __vxlan_sock_release(vxlan->vn6_sock);
1058#endif
1059}
1060
56ef9c90 1061/* Update multicast group membership when first VNI on
c4b49512 1062 * multicast address is brought up
7c47cedf 1063 */
56ef9c90 1064static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1065{
b1be00a6 1066 struct sock *sk;
e4c7ed41
CW		 1067 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1068 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1069 int ret = -EINVAL;
d342894c 1070
e4c7ed41
CW		 1071 if (ip->sa.sa_family == AF_INET) {
1072 struct ip_mreqn mreq = {
1073 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1074 .imr_ifindex = ifindex,
1075 };
1076
b1be00a6
JB		 1077 sk = vxlan->vn4_sock->sock->sk;
1078 lock_sock(sk);
56ef9c90 1079 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1080 release_sock(sk);
e4c7ed41
CW		 1081#if IS_ENABLED(CONFIG_IPV6)
1082 } else {
b1be00a6
JB		 1083 sk = vxlan->vn6_sock->sock->sk;
1084 lock_sock(sk);
56ef9c90
MRL		 1085 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1086 &ip->sin6.sin6_addr);
b1be00a6 1087 release_sock(sk);
e4c7ed41
CW		 1088#endif
1089 }
3fc2de2f 1090
56ef9c90 1091 return ret;
3fc2de2f 1092}
1093
1094/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1095static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1096{
b1be00a6 1097 struct sock *sk;
e4c7ed41
CW		 1098 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1099 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1100 int ret = -EINVAL;
3fc2de2f 1101
e4c7ed41
CW		 1102 if (ip->sa.sa_family == AF_INET) {
1103 struct ip_mreqn mreq = {
1104 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1105 .imr_ifindex = ifindex,
1106 };
1107
b1be00a6
JB		 1108 sk = vxlan->vn4_sock->sock->sk;
1109 lock_sock(sk);
56ef9c90 1110 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1111 release_sock(sk);
e4c7ed41
CW		 1112#if IS_ENABLED(CONFIG_IPV6)
1113 } else {
b1be00a6
JB		 1114 sk = vxlan->vn6_sock->sock->sk;
1115 lock_sock(sk);
56ef9c90
MRL		 1116 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1117 &ip->sin6.sin6_addr);
b1be00a6 1118 release_sock(sk);
e4c7ed41
CW		 1119#endif
1120 }
d342894c 1121
56ef9c90 1122 return ret;
d342894c 1123}
1124
dfd8645e 1125static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
0ace2ca8 1126 size_t hdrlen, u32 data, bool nopartial)
dfd8645e
TH		 1127{
1128 size_t start, offset, plen;
dfd8645e 1129
b7fe10e5
TH		 1130 if (skb->remcsum_offload)
1131 return vh;
1132
dfd8645e
TH		 1133 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1134 offset = start + ((data & VXLAN_RCO_UDP) ?
1135 offsetof(struct udphdr, check) :
1136 offsetof(struct tcphdr, check));
1137
1138 plen = hdrlen + offset + sizeof(u16);
1139
1140 if (!pskb_may_pull(skb, plen))
1141 return NULL;
1142
1143 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1144
0ace2ca8
TH		 1145 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1146 nopartial);
dfd8645e
TH		 1147
1148 return vh;
1149}
1150
614732ea
TG		 1151static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1152 struct vxlan_metadata *md, u32 vni,
1153 struct metadata_dst *tun_dst)
1154{
1155 struct iphdr *oip = NULL;
1156 struct ipv6hdr *oip6 = NULL;
1157 struct vxlan_dev *vxlan;
1158 struct pcpu_sw_netstats *stats;
1159 union vxlan_addr saddr;
1160 int err = 0;
614732ea
TG		 1161
1162 /* For flow based devices, map all packets to VNI 0 */
da8b43c0 1163 if (vs->flags & VXLAN_F_COLLECT_METADATA)
614732ea
TG		 1164 vni = 0;
1165
1166 /* Is this VNI defined? */
1167 vxlan = vxlan_vs_find_vni(vs, vni);
1168 if (!vxlan)
1169 goto drop;
1170
614732ea
TG		 1171 skb_reset_mac_header(skb);
1172 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1173 skb->protocol = eth_type_trans(skb, vxlan->dev);
1174 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1175
1176 /* Ignore packet loops (and multicast echo) */
1177 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1178 goto drop;
1179
ce212d0f
JB		 1180 /* Get data from the outer IP header */
1181 if (vxlan_get_sk_family(vs) == AF_INET) {
614732ea
TG		 1182 oip = ip_hdr(skb);
1183 saddr.sin.sin_addr.s_addr = oip->saddr;
1184 saddr.sa.sa_family = AF_INET;
1185#if IS_ENABLED(CONFIG_IPV6)
1186 } else {
1187 oip6 = ipv6_hdr(skb);
1188 saddr.sin6.sin6_addr = oip6->saddr;
1189 saddr.sa.sa_family = AF_INET6;
1190#endif
1191 }
1192
1193 if (tun_dst) {
1194 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1195 tun_dst = NULL;
1196 }
1197
1198 if ((vxlan->flags & VXLAN_F_LEARN) &&
1199 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1200 goto drop;
1201
1202 skb_reset_network_header(skb);
1203 /* In flow-based mode, GBP is carried in dst_metadata */
da8b43c0 1204 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
614732ea
TG		 1205 skb->mark = md->gbp;
1206
1207 if (oip6)
1208 err = IP6_ECN_decapsulate(oip6, skb);
1209 if (oip)
1210 err = IP_ECN_decapsulate(oip, skb);
1211
1212 if (unlikely(err)) {
1213 if (log_ecn_error) {
1214 if (oip6)
1215 net_info_ratelimited("non-ECT from %pI6\n",
1216 &oip6->saddr);
1217 if (oip)
1218 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1219 &oip->saddr, oip->tos);
1220 }
1221 if (err > 1) {
1222 ++vxlan->dev->stats.rx_frame_errors;
1223 ++vxlan->dev->stats.rx_errors;
1224 goto drop;
1225 }
1226 }
1227
1228 stats = this_cpu_ptr(vxlan->dev->tstats);
1229 u64_stats_update_begin(&stats->syncp);
1230 stats->rx_packets++;
1231 stats->rx_bytes += skb->len;
1232 u64_stats_update_end(&stats->syncp);
1233
58ce31cc 1234 gro_cells_receive(&vxlan->gro_cells, skb);
614732ea
TG		 1235
1236 return;
1237drop:
1238 if (tun_dst)
1239 dst_release((struct dst_entry *)tun_dst);
1240
1241 /* Consume bad packet */
1242 kfree_skb(skb);
1243}
1244
d342894c 1245/* Callback from net/ipv4/udp.c to receive packets */
1246static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1247{
ee122c79 1248 struct metadata_dst *tun_dst = NULL;
5cfccc5a 1249 struct vxlan_sock *vs;
d342894c 1250 struct vxlanhdr *vxh;
3bf39475 1251 u32 flags, vni;
ee122c79
TG		 1252 struct vxlan_metadata _md;
1253 struct vxlan_metadata *md = &_md;
d342894c 1254
d342894c 1255 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1256 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1257 goto error;
1258
7ce04758 1259 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1260 flags = ntohl(vxh->vx_flags);
1261 vni = ntohl(vxh->vx_vni);
1262
1263 if (flags & VXLAN_HF_VNI) {
1264 flags &= ~VXLAN_HF_VNI;
1265 } else {
1266 /* VNI flag always required to be set */
1267 goto bad_flags;
d342894c 1268 }
1269
5cfccc5a
PS		 1270 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1271 goto drop;
dfd8645e 1272 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1273
559835ea 1274 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1275 if (!vs)
d342894c 1276 goto drop;
d342894c 1277
dfd8645e 1278 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
0ace2ca8
TH		 1279 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1280 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 1281 if (!vxh)
1282 goto drop;
1283
1284 flags &= ~VXLAN_HF_RCO;
40fb70f3 1285 vni &= VXLAN_VNI_MASK;
dfd8645e
TH		 1286 }
1287
ee122c79 1288 if (vxlan_collect_metadata(vs)) {
c29a70d2
PS		 1289 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1290 cpu_to_be64(vni >> 8), sizeof(*md));
1291
ee122c79
TG		 1292 if (!tun_dst)
1293 goto drop;
1294
0f1b7354 1295 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
ee122c79
TG		 1296 } else {
1297 memset(md, 0, sizeof(*md));
1298 }
1299
3511494c
TG		 1300 /* For backwards compatibility, only allow reserved fields to be
1301 * used by VXLAN extensions if explicitly requested.
1302 */
1303 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1304 struct vxlanhdr_gbp *gbp;
1305
1306 gbp = (struct vxlanhdr_gbp *)vxh;
ee122c79
TG		 1307 md->gbp = ntohs(gbp->policy_id);
1308
1309 if (tun_dst)
0f1b7354 1310 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
3511494c
TG		 1311
1312 if (gbp->dont_learn)
ee122c79 1313 md->gbp |= VXLAN_GBP_DONT_LEARN;
3511494c
TG		 1314
1315 if (gbp->policy_applied)
ee122c79 1316 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
3511494c
TG		 1317
1318 flags &= ~VXLAN_GBP_USED_BITS;
1319 }
1320
40fb70f3 1321 if (flags || vni & ~VXLAN_VNI_MASK) {
3bf39475
TH		 1322 /* If there are any unprocessed flags remaining treat
1323 * this as a malformed packet. This behavior diverges from
1324 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1325 * in reserved fields are to be ignored. The approach here
c4b49512 1326 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1327 * is more robust and provides a little more security in
1328 * adding extensions to VXLAN.
1329 */
1330
1331 goto bad_flags;
1332 }
1333
614732ea 1334 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
5cfccc5a
PS		 1335 return 0;
1336
1337drop:
1338 /* Consume bad packet */
1339 kfree_skb(skb);
1340 return 0;
1341
3bf39475
TH		 1342bad_flags:
1343 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1344 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1345
5cfccc5a 1346error:
ee122c79
TG		 1347 if (tun_dst)
1348 dst_release((struct dst_entry *)tun_dst);
1349
5cfccc5a
PS		 1350 /* Return non vxlan pkt */
1351 return 1;
1352}
1353
e4f67add
DS		 1354static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1355{
1356 struct vxlan_dev *vxlan = netdev_priv(dev);
1357 struct arphdr *parp;
1358 u8 *arpptr, *sha;
1359 __be32 sip, tip;
1360 struct neighbour *n;
1361
1362 if (dev->flags & IFF_NOARP)
1363 goto out;
1364
1365 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1366 dev->stats.tx_dropped++;
1367 goto out;
1368 }
1369 parp = arp_hdr(skb);
1370
1371 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1372 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1373 parp->ar_pro != htons(ETH_P_IP) ||
1374 parp->ar_op != htons(ARPOP_REQUEST) ||
1375 parp->ar_hln != dev->addr_len ||
1376 parp->ar_pln != 4)
1377 goto out;
1378 arpptr = (u8 *)parp + sizeof(struct arphdr);
1379 sha = arpptr;
1380 arpptr += dev->addr_len; /* sha */
1381 memcpy(&sip, arpptr, sizeof(sip));
1382 arpptr += sizeof(sip);
1383 arpptr += dev->addr_len; /* tha */
1384 memcpy(&tip, arpptr, sizeof(tip));
1385
1386 if (ipv4_is_loopback(tip) ||
1387 ipv4_is_multicast(tip))
1388 goto out;
1389
1390 n = neigh_lookup(&arp_tbl, &tip, dev);
1391
1392 if (n) {
e4f67add
DS		 1393 struct vxlan_fdb *f;
1394 struct sk_buff *reply;
1395
1396 if (!(n->nud_state & NUD_CONNECTED)) {
1397 neigh_release(n);
1398 goto out;
1399 }
1400
1401 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1402 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1403 /* bridge-local neighbor */
1404 neigh_release(n);
1405 goto out;
1406 }
1407
1408 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1409 n->ha, sha);
1410
1411 neigh_release(n);
1412
7346135d
DS		 1413 if (reply == NULL)
1414 goto out;
1415
e4f67add
DS		 1416 skb_reset_mac_header(reply);
1417 __skb_pull(reply, skb_network_offset(reply));
1418 reply->ip_summed = CHECKSUM_UNNECESSARY;
1419 reply->pkt_type = PACKET_HOST;
1420
1421 if (netif_rx_ni(reply) == NET_RX_DROP)
1422 dev->stats.rx_dropped++;
e4c7ed41
CW		 1423 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1424 union vxlan_addr ipa = {
1425 .sin.sin_addr.s_addr = tip,
a45e92a5 1426 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1427 };
1428
1429 vxlan_ip_miss(dev, &ipa);
1430 }
e4f67add
DS		 1431out:
1432 consume_skb(skb);
1433 return NETDEV_TX_OK;
1434}
1435
f564f45c 1436#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1437static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1438 struct neighbour *n, bool isrouter)
1439{
1440 struct net_device *dev = request->dev;
1441 struct sk_buff *reply;
1442 struct nd_msg *ns, *na;
1443 struct ipv6hdr *pip6;
1444 u8 *daddr;
1445 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1446 int ns_olen;
1447 int i, len;
1448
1449 if (dev == NULL)
1450 return NULL;
1451
1452 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1453 sizeof(*na) + na_olen + dev->needed_tailroom;
1454 reply = alloc_skb(len, GFP_ATOMIC);
1455 if (reply == NULL)
1456 return NULL;
1457
1458 reply->protocol = htons(ETH_P_IPV6);
1459 reply->dev = dev;
1460 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1461 skb_push(reply, sizeof(struct ethhdr));
1462 skb_set_mac_header(reply, 0);
1463
1464 ns = (struct nd_msg *)skb_transport_header(request);
1465
1466 daddr = eth_hdr(request)->h_source;
1467 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1468 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1469 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1470 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1471 break;
1472 }
1473 }
1474
1475 /* Ethernet header */
1476 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1477 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1478 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1479 reply->protocol = htons(ETH_P_IPV6);
1480
1481 skb_pull(reply, sizeof(struct ethhdr));
1482 skb_set_network_header(reply, 0);
1483 skb_put(reply, sizeof(struct ipv6hdr));
1484
1485 /* IPv6 header */
1486
1487 pip6 = ipv6_hdr(reply);
1488 memset(pip6, 0, sizeof(struct ipv6hdr));
1489 pip6->version = 6;
1490 pip6->priority = ipv6_hdr(request)->priority;
1491 pip6->nexthdr = IPPROTO_ICMPV6;
1492 pip6->hop_limit = 255;
1493 pip6->daddr = ipv6_hdr(request)->saddr;
1494 pip6->saddr = *(struct in6_addr *)n->primary_key;
1495
1496 skb_pull(reply, sizeof(struct ipv6hdr));
1497 skb_set_transport_header(reply, 0);
1498
1499 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1500
1501 /* Neighbor Advertisement */
1502 memset(na, 0, sizeof(*na)+na_olen);
1503 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1504 na->icmph.icmp6_router = isrouter;
1505 na->icmph.icmp6_override = 1;
1506 na->icmph.icmp6_solicited = 1;
1507 na->target = ns->target;
1508 ether_addr_copy(&na->opt[2], n->ha);
1509 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1510 na->opt[1] = na_olen >> 3;
1511
1512 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1513 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1514 csum_partial(na, sizeof(*na)+na_olen, 0));
1515
1516 pip6->payload_len = htons(sizeof(*na)+na_olen);
1517
1518 skb_push(reply, sizeof(struct ipv6hdr));
1519
1520 reply->ip_summed = CHECKSUM_UNNECESSARY;
1521
1522 return reply;
1523}
1524
f564f45c
CW		 1525static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1526{
1527 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1528 struct nd_msg *msg;
f564f45c
CW		 1529 const struct ipv6hdr *iphdr;
1530 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1531 struct neighbour *n;
1532 struct inet6_dev *in6_dev;
f564f45c
CW		 1533
1534 in6_dev = __in6_dev_get(dev);
1535 if (!in6_dev)
1536 goto out;
1537
f564f45c
CW		 1538 iphdr = ipv6_hdr(skb);
1539 saddr = &iphdr->saddr;
1540 daddr = &iphdr->daddr;
1541
f564f45c
CW		 1542 msg = (struct nd_msg *)skb_transport_header(skb);
1543 if (msg->icmph.icmp6_code != 0 ||
1544 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1545 goto out;
1546
4b29dba9
DS		 1547 if (ipv6_addr_loopback(daddr) ||
1548 ipv6_addr_is_multicast(&msg->target))
1549 goto out;
1550
1551 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1552
1553 if (n) {
1554 struct vxlan_fdb *f;
4b29dba9 1555 struct sk_buff *reply;
f564f45c
CW		 1556
1557 if (!(n->nud_state & NUD_CONNECTED)) {
1558 neigh_release(n);
1559 goto out;
1560 }
1561
1562 f = vxlan_find_mac(vxlan, n->ha);
1563 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1564 /* bridge-local neighbor */
1565 neigh_release(n);
1566 goto out;
1567 }
1568
4b29dba9
DS		 1569 reply = vxlan_na_create(skb, n,
1570 !!(f ? f->flags & NTF_ROUTER : 0));
1571
f564f45c 1572 neigh_release(n);
4b29dba9
DS		 1573
1574 if (reply == NULL)
1575 goto out;
1576
1577 if (netif_rx_ni(reply) == NET_RX_DROP)
1578 dev->stats.rx_dropped++;
1579
f564f45c 1580 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1581 union vxlan_addr ipa = {
1582 .sin6.sin6_addr = msg->target,
a45e92a5 1583 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1584 };
1585
f564f45c
CW		 1586 vxlan_ip_miss(dev, &ipa);
1587 }
1588
1589out:
1590 consume_skb(skb);
1591 return NETDEV_TX_OK;
1592}
1593#endif
1594
e4f67add
DS		 1595static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1596{
1597 struct vxlan_dev *vxlan = netdev_priv(dev);
1598 struct neighbour *n;
e4f67add
DS		 1599
1600 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1601 return false;
1602
1603 n = NULL;
1604 switch (ntohs(eth_hdr(skb)->h_proto)) {
1605 case ETH_P_IP:
e15a00aa
CW		 1606 {
1607 struct iphdr *pip;
1608
e4f67add
DS		 1609 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1610 return false;
1611 pip = ip_hdr(skb);
1612 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1613 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1614 union vxlan_addr ipa = {
1615 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1616 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1617 };
1618
1619 vxlan_ip_miss(dev, &ipa);
1620 return false;
1621 }
1622
e4f67add 1623 break;
e15a00aa
CW		 1624 }
1625#if IS_ENABLED(CONFIG_IPV6)
1626 case ETH_P_IPV6:
1627 {
1628 struct ipv6hdr *pip6;
1629
1630 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1631 return false;
1632 pip6 = ipv6_hdr(skb);
1633 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1634 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1635 union vxlan_addr ipa = {
1636 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1637 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1638 };
1639
1640 vxlan_ip_miss(dev, &ipa);
1641 return false;
1642 }
1643
1644 break;
1645 }
1646#endif
e4f67add
DS		 1647 default:
1648 return false;
1649 }
1650
1651 if (n) {
1652 bool diff;
1653
7367d0b5 1654 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1655 if (diff) {
1656 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1657 dev->addr_len);
1658 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1659 }
1660 neigh_release(n);
1661 return diff;
e4c7ed41
CW		 1662 }
1663
e4f67add
DS		 1664 return false;
1665}
1666
af33c1ad 1667static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1668 struct vxlan_metadata *md)
1669{
1670 struct vxlanhdr_gbp *gbp;
1671
db79a621
TG		 1672 if (!md->gbp)
1673 return;
1674
3511494c
TG		 1675 gbp = (struct vxlanhdr_gbp *)vxh;
1676 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1677
1678 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1679 gbp->dont_learn = 1;
1680
1681 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1682 gbp->policy_applied = 1;
1683
1684 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1685}
1686
e4c7ed41 1687#if IS_ENABLED(CONFIG_IPV6)
79b16aad
DM		 1688static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1689 struct sk_buff *skb,
e4c7ed41
CW		 1690 struct net_device *dev, struct in6_addr *saddr,
1691 struct in6_addr *daddr, __u8 prio, __u8 ttl,
6b6948dd 1692 __be16 src_port, __be16 dst_port, __be32 vni,
af33c1ad 1693 struct vxlan_metadata *md, bool xnet, u32 vxflags)
e4c7ed41 1694{
e4c7ed41 1695 struct vxlanhdr *vxh;
e4c7ed41
CW		 1696 int min_headroom;
1697 int err;
af33c1ad 1698 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
dfd8645e
TH		 1699 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1700 u16 hdrlen = sizeof(struct vxlanhdr);
1701
af33c1ad 1702 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1703 skb->ip_summed == CHECKSUM_PARTIAL) {
1704 int csum_start = skb_checksum_start_offset(skb);
1705
1706 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1707 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1708 (skb->csum_offset == offsetof(struct udphdr, check) ||
1709 skb->csum_offset == offsetof(struct tcphdr, check))) {
1710 udp_sum = false;
1711 type |= SKB_GSO_TUNNEL_REMCSUM;
1712 }
1713 }
e4c7ed41 1714
f01ec1c0 1715 skb_scrub_packet(skb, xnet);
963a88b3 1716
e4c7ed41
CW		 1717 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1718 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1719 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1720
1721 /* Need space for new headers (invalidates iph ptr) */
1722 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1723 if (unlikely(err)) {
1724 kfree_skb(skb);
1725 goto err;
1726 }
e4c7ed41 1727
5968250c 1728 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1729 if (WARN_ON(!skb)) {
1730 err = -ENOMEM;
1731 goto err;
1732 }
e4c7ed41 1733
b736a623
JG		 1734 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1735 if (IS_ERR(skb)) {
1736 err = -EINVAL;
1737 goto err;
1738 }
1739
e4c7ed41 1740 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1741 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1742 vxh->vx_vni = vni;
e4c7ed41 1743
dfd8645e
TH		 1744 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1745 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1746 VXLAN_RCO_SHIFT;
1747
1748 if (skb->csum_offset == offsetof(struct udphdr, check))
1749 data |= VXLAN_RCO_UDP;
1750
1751 vxh->vx_vni |= htonl(data);
1752 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1753
1754 if (!skb_is_gso(skb)) {
1755 skb->ip_summed = CHECKSUM_NONE;
1756 skb->encapsulation = 0;
1757 }
1758 }
1759
af33c1ad
TH		 1760 if (vxflags & VXLAN_F_GBP)
1761 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1762
996c9fd1
TH		 1763 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1764
79b16aad 1765 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
d998f8ef 1766 ttl, src_port, dst_port,
af33c1ad 1767 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
e4c7ed41 1768 return 0;
74f47278
PS		 1769err:
1770 dst_release(dst);
1771 return err;
e4c7ed41
CW		 1772}
1773#endif
1774
614732ea
TG		 1775static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1776 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
6b6948dd 1777 __be16 src_port, __be16 dst_port, __be32 vni,
614732ea 1778 struct vxlan_metadata *md, bool xnet, u32 vxflags)
49560532
PS		 1779{
1780 struct vxlanhdr *vxh;
649c5b8b 1781 int min_headroom;
49560532 1782 int err;
af33c1ad 1783 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
dfd8645e
TH		 1784 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1785 u16 hdrlen = sizeof(struct vxlanhdr);
1786
af33c1ad 1787 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1788 skb->ip_summed == CHECKSUM_PARTIAL) {
1789 int csum_start = skb_checksum_start_offset(skb);
1790
1791 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1792 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1793 (skb->csum_offset == offsetof(struct udphdr, check) ||
1794 skb->csum_offset == offsetof(struct tcphdr, check))) {
1795 udp_sum = false;
1796 type |= SKB_GSO_TUNNEL_REMCSUM;
1797 }
1798 }
49560532 1799
649c5b8b 1800 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1801 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1802 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1803
1804 /* Need space for new headers (invalidates iph ptr) */
1805 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1806 if (unlikely(err)) {
1807 kfree_skb(skb);
649c5b8b 1808 return err;
74f47278 1809 }
649c5b8b 1810
5968250c
JP		 1811 skb = vlan_hwaccel_push_inside(skb);
1812 if (WARN_ON(!skb))
1813 return -ENOMEM;
1eaa8178 1814
b736a623
JG		 1815 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1816 if (IS_ERR(skb))
1817 return PTR_ERR(skb);
1818
49560532 1819 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1820 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1821 vxh->vx_vni = vni;
49560532 1822
dfd8645e
TH		 1823 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1824 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1825 VXLAN_RCO_SHIFT;
1826
1827 if (skb->csum_offset == offsetof(struct udphdr, check))
1828 data |= VXLAN_RCO_UDP;
1829
1830 vxh->vx_vni |= htonl(data);
1831 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1832
1833 if (!skb_is_gso(skb)) {
1834 skb->ip_summed = CHECKSUM_NONE;
1835 skb->encapsulation = 0;
1836 }
1837 }
1838
af33c1ad
TH		 1839 if (vxflags & VXLAN_F_GBP)
1840 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1841
996c9fd1
TH		 1842 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1843
039f5062
PS		 1844 udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos, ttl, df,
1845 src_port, dst_port, xnet,
1846 !(vxflags & VXLAN_F_UDP_CSUM));
1847 return 0;
49560532 1848}
49560532 1849
e5d4b29f
JB		 1850#if IS_ENABLED(CONFIG_IPV6)
1851static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1852 struct sk_buff *skb, int oif,
1853 const struct in6_addr *daddr,
1854 struct in6_addr *saddr)
1855{
1856 struct dst_entry *ndst;
1857 struct flowi6 fl6;
1858 int err;
1859
1860 memset(&fl6, 0, sizeof(fl6));
1861 fl6.flowi6_oif = oif;
1862 fl6.daddr = *daddr;
1863 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1864 fl6.flowi6_mark = skb->mark;
1865 fl6.flowi6_proto = IPPROTO_UDP;
1866
1867 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1868 vxlan->vn6_sock->sock->sk,
1869 &ndst, &fl6);
1870 if (err < 0)
1871 return ERR_PTR(err);
1872
1873 *saddr = fl6.saddr;
1874 return ndst;
1875}
1876#endif
1877
9dcc71e1
SS		 1878/* Bypass encapsulation if the destination is local */
1879static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1880 struct vxlan_dev *dst_vxlan)
1881{
8f84985f 1882 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1883 union vxlan_addr loopback;
1884 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1885 struct net_device *dev = skb->dev;
1886 int len = skb->len;
9dcc71e1 1887
8f84985f
LR		 1888 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1889 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1890 skb->pkt_type = PACKET_HOST;
1891 skb->encapsulation = 0;
1892 skb->dev = dst_vxlan->dev;
1893 __skb_pull(skb, skb_network_offset(skb));
1894
e4c7ed41
CW		 1895 if (remote_ip->sa.sa_family == AF_INET) {
1896 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1897 loopback.sa.sa_family = AF_INET;
1898#if IS_ENABLED(CONFIG_IPV6)
1899 } else {
1900 loopback.sin6.sin6_addr = in6addr_loopback;
1901 loopback.sa.sa_family = AF_INET6;
1902#endif
1903 }
1904
9dcc71e1 1905 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1906 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1907
1908 u64_stats_update_begin(&tx_stats->syncp);
1909 tx_stats->tx_packets++;
ce6502a8 1910 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1911 u64_stats_update_end(&tx_stats->syncp);
1912
1913 if (netif_rx(skb) == NET_RX_SUCCESS) {
1914 u64_stats_update_begin(&rx_stats->syncp);
1915 rx_stats->rx_packets++;
ce6502a8 1916 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1917 u64_stats_update_end(&rx_stats->syncp);
1918 } else {
ce6502a8 1919 dev->stats.rx_dropped++;
9dcc71e1
SS		 1920 }
1921}
1922
4ad16930
SH		 1923static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1924 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1925{
3093fbe7 1926 struct ip_tunnel_info *info;
d342894c 1927 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1928 struct sock *sk;
e4c7ed41 1929 struct rtable *rt = NULL;
d342894c 1930 const struct iphdr *old_iph;
d342894c 1931 struct flowi4 fl4;
e4c7ed41 1932 union vxlan_addr *dst;
ee122c79
TG		 1933 union vxlan_addr remote_ip;
1934 struct vxlan_metadata _md;
1935 struct vxlan_metadata *md = &_md;
e4c7ed41 1936 __be16 src_port = 0, dst_port;
234f5b73 1937 u32 vni;
d342894c 1938 __be16 df = 0;
1939 __u8 tos, ttl;
0e6fbc5b 1940 int err;
ee122c79 1941 u32 flags = vxlan->flags;
d342894c 1942
61adedf3 1943 info = skb_tunnel_info(skb);
3093fbe7 1944
ee122c79 1945 if (rdst) {
0dfbdf41 1946 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1947 vni = rdst->remote_vni;
1948 dst = &rdst->remote_ip;
1949 } else {
1950 if (!info) {
1951 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1952 dev->name);
1953 goto drop;
1954 }
0dfbdf41 1955 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
ee122c79 1956 vni = be64_to_cpu(info->key.tun_id);
b1be00a6
JB		 1957 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1958 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1959 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1960 else
1961 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79
TG		 1962 dst = &remote_ip;
1963 }
e4f67add 1964
e4c7ed41 1965 if (vxlan_addr_any(dst)) {
e4f67add 1966 if (did_rsc) {
e4f67add 1967 /* short-circuited back to local bridge */
9dcc71e1 1968 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1969 return;
e4f67add 1970 }
ef59febe 1971 goto drop;
e4f67add 1972 }
ef59febe 1973
d342894c 1974 old_iph = ip_hdr(skb);
1975
0dfbdf41 1976 ttl = vxlan->cfg.ttl;
e4c7ed41 1977 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1978 ttl = 1;
1979
0dfbdf41 1980 tos = vxlan->cfg.tos;
d342894c 1981 if (tos == 1)
206aaafc 1982 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1983
0dfbdf41
TG		 1984 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1985 vxlan->cfg.port_max, true);
d342894c 1986
a725e514 1987 if (info) {
a725e514
JB		 1988 ttl = info->key.ttl;
1989 tos = info->key.tos;
1990
1991 if (info->options_len)
4c222798 1992 md = ip_tunnel_info_opts(info);
a725e514
JB		 1993 } else {
1994 md->gbp = skb->mark;
1995 }
1996
e4c7ed41 1997 if (dst->sa.sa_family == AF_INET) {
b1be00a6
JB		 1998 if (!vxlan->vn4_sock)
1999 goto drop;
2000 sk = vxlan->vn4_sock->sock->sk;
2001
35e2d115
JG		 2002 if (info) {
2003 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2004 df = htons(IP_DF);
2005
2006 if (info->key.tun_flags & TUNNEL_CSUM)
2007 flags |= VXLAN_F_UDP_CSUM;
2008 else
2009 flags &= ~VXLAN_F_UDP_CSUM;
2010 }
ee122c79 2011
e4c7ed41 2012 memset(&fl4, 0, sizeof(fl4));
ee122c79 2013 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
e4c7ed41 2014 fl4.flowi4_tos = RT_TOS(tos);
239fb791
TG		 2015 fl4.flowi4_mark = skb->mark;
2016 fl4.flowi4_proto = IPPROTO_UDP;
e4c7ed41 2017 fl4.daddr = dst->sin.sin_addr.s_addr;
0dfbdf41 2018 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
e4c7ed41 2019
f01ec1c0 2020 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 2021 if (IS_ERR(rt)) {
2022 netdev_dbg(dev, "no route to %pI4\n",
2023 &dst->sin.sin_addr.s_addr);
2024 dev->stats.tx_carrier_errors++;
2025 goto tx_error;
2026 }
d342894c 2027
e4c7ed41
CW		 2028 if (rt->dst.dev == dev) {
2029 netdev_dbg(dev, "circular route to %pI4\n",
2030 &dst->sin.sin_addr.s_addr);
2031 dev->stats.collisions++;
fffc15a5 2032 goto rt_tx_error;
e4c7ed41
CW		 2033 }
2034
2035 /* Bypass encapsulation if the destination is local */
2036 if (rt->rt_flags & RTCF_LOCAL &&
2037 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2038 struct vxlan_dev *dst_vxlan;
2039
2040 ip_rt_put(rt);
19ca9fc1 2041 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2042 dst->sa.sa_family, dst_port,
2043 vxlan->flags);
e4c7ed41
CW		 2044 if (!dst_vxlan)
2045 goto tx_error;
2046 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2047 return;
2048 }
2049
2050 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2051 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
79b16aad 2052 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
af33c1ad 2053 dst->sin.sin_addr.s_addr, tos, ttl, df,
614732ea 2054 src_port, dst_port, htonl(vni << 8), md,
af33c1ad 2055 !net_eq(vxlan->net, dev_net(vxlan->dev)),
ee122c79 2056 flags);
74f47278
PS		 2057 if (err < 0) {
2058 /* skb is already freed. */
2059 skb = NULL;
e4c7ed41 2060 goto rt_tx_error;
74f47278 2061 }
e4c7ed41
CW		 2062#if IS_ENABLED(CONFIG_IPV6)
2063 } else {
e4c7ed41 2064 struct dst_entry *ndst;
e5d4b29f 2065 struct in6_addr saddr;
6f264e47 2066 u32 rt6i_flags;
e4c7ed41 2067
b1be00a6
JB		 2068 if (!vxlan->vn6_sock)
2069 goto drop;
2070 sk = vxlan->vn6_sock->sock->sk;
2071
e5d4b29f
JB		 2072 ndst = vxlan6_get_route(vxlan, skb,
2073 rdst ? rdst->remote_ifindex : 0,
2074 &dst->sin6.sin6_addr, &saddr);
2075 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2076 netdev_dbg(dev, "no route to %pI6\n",
2077 &dst->sin6.sin6_addr);
2078 dev->stats.tx_carrier_errors++;
9dcc71e1 2079 goto tx_error;
e4c7ed41 2080 }
d342894c 2081
e4c7ed41
CW		 2082 if (ndst->dev == dev) {
2083 netdev_dbg(dev, "circular route to %pI6\n",
2084 &dst->sin6.sin6_addr);
2085 dst_release(ndst);
2086 dev->stats.collisions++;
2087 goto tx_error;
2088 }
0e6fbc5b 2089
e4c7ed41 2090 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2091 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2092 if (rt6i_flags & RTF_LOCAL &&
2093 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2094 struct vxlan_dev *dst_vxlan;
2095
2096 dst_release(ndst);
19ca9fc1 2097 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2098 dst->sa.sa_family, dst_port,
2099 vxlan->flags);
e4c7ed41
CW		 2100 if (!dst_vxlan)
2101 goto tx_error;
2102 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2103 return;
2104 }
49560532 2105
35e2d115
JG		 2106 if (info) {
2107 if (info->key.tun_flags & TUNNEL_CSUM)
2108 flags &= ~VXLAN_F_UDP_ZERO_CSUM6_TX;
2109 else
2110 flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2111 }
2112
e4c7ed41 2113 ttl = ttl ? : ip6_dst_hoplimit(ndst);
e5d4b29f 2114 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &saddr, &dst->sin6.sin6_addr,
614732ea 2115 0, ttl, src_port, dst_port, htonl(vni << 8), md,
af33c1ad 2116 !net_eq(vxlan->net, dev_net(vxlan->dev)),
a725e514 2117 flags);
e4c7ed41
CW		 2118#endif
2119 }
0e6fbc5b 2120
4ad16930 2121 return;
d342894c 2122
2123drop:
2124 dev->stats.tx_dropped++;
2125 goto tx_free;
2126
49560532
PS		 2127rt_tx_error:
2128 ip_rt_put(rt);
d342894c 2129tx_error:
2130 dev->stats.tx_errors++;
2131tx_free:
2132 dev_kfree_skb(skb);
d342894c 2133}
2134
6681712d
DS		 2135/* Transmit local packets over Vxlan
2136 *
2137 * Outer IP header inherits ECN and DF from inner header.
2138 * Outer UDP destination is the VXLAN assigned port.
2139 * source port is based on hash of flow
2140 */
2141static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2142{
2143 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2144 const struct ip_tunnel_info *info;
6681712d
DS		 2145 struct ethhdr *eth;
2146 bool did_rsc = false;
8f646c92 2147 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2148 struct vxlan_fdb *f;
6681712d 2149
61adedf3 2150 info = skb_tunnel_info(skb);
3093fbe7 2151
6681712d
DS		 2152 skb_reset_mac_header(skb);
2153 eth = eth_hdr(skb);
2154
f564f45c
CW		 2155 if ((vxlan->flags & VXLAN_F_PROXY)) {
2156 if (ntohs(eth->h_proto) == ETH_P_ARP)
2157 return arp_reduce(dev, skb);
2158#if IS_ENABLED(CONFIG_IPV6)
2159 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2160 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2161 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2162 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2163 struct nd_msg *msg;
2164
2165 msg = (struct nd_msg *)skb_transport_header(skb);
2166 if (msg->icmph.icmp6_code == 0 &&
2167 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2168 return neigh_reduce(dev, skb);
2169 }
7a9f526f 2170 eth = eth_hdr(skb);
f564f45c
CW		 2171#endif
2172 }
6681712d 2173
f468a729
JB		 2174 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2175 if (info && info->mode & IP_TUNNEL_INFO_TX)
2176 vxlan_xmit_one(skb, dev, NULL, false);
2177 else
2178 kfree_skb(skb);
ee122c79
TG		 2179 return NETDEV_TX_OK;
2180 }
2181
6681712d 2182 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2183 did_rsc = false;
2184
2185 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2186 (ntohs(eth->h_proto) == ETH_P_IP ||
2187 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2188 did_rsc = route_shortcircuit(dev, skb);
2189 if (did_rsc)
2190 f = vxlan_find_mac(vxlan, eth->h_dest);
2191 }
2192
6681712d 2193 if (f == NULL) {
afbd8bae
MR		 2194 f = vxlan_find_mac(vxlan, all_zeros_mac);
2195 if (f == NULL) {
2196 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2197 !is_multicast_ether_addr(eth->h_dest))
2198 vxlan_fdb_miss(vxlan, eth->h_dest);
2199
2200 dev->stats.tx_dropped++;
8f646c92 2201 kfree_skb(skb);
afbd8bae
MR		 2202 return NETDEV_TX_OK;
2203 }
2204 }
6681712d 2205
afbd8bae
MR		 2206 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2207 struct sk_buff *skb1;
6681712d 2208
8f646c92
ED		 2209 if (!fdst) {
2210 fdst = rdst;
2211 continue;
2212 }
afbd8bae
MR		 2213 skb1 = skb_clone(skb, GFP_ATOMIC);
2214 if (skb1)
2215 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2216 }
2217
8f646c92
ED		 2218 if (fdst)
2219 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2220 else
2221 kfree_skb(skb);
4ad16930 2222 return NETDEV_TX_OK;
6681712d
DS		 2223}
2224
d342894c 2225/* Walk the forwarding table and purge stale entries */
2226static void vxlan_cleanup(unsigned long arg)
2227{
2228 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2229 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2230 unsigned int h;
2231
2232 if (!netif_running(vxlan->dev))
2233 return;
2234
d342894c 2235 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2236 struct hlist_node *p, *n;
14e1d0fa
SD		 2237
2238 spin_lock_bh(&vxlan->hash_lock);
d342894c 2239 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2240 struct vxlan_fdb *f
2241 = container_of(p, struct vxlan_fdb, hlist);
2242 unsigned long timeout;
2243
3c172868 2244 if (f->state & NUD_PERMANENT)
d342894c 2245 continue;
2246
0dfbdf41 2247 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2248 if (time_before_eq(timeout, jiffies)) {
2249 netdev_dbg(vxlan->dev,
2250 "garbage collect %pM\n",
2251 f->eth_addr);
2252 f->state = NUD_STALE;
2253 vxlan_fdb_destroy(vxlan, f);
2254 } else if (time_before(timeout, next_timer))
2255 next_timer = timeout;
2256 }
14e1d0fa 2257 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2258 }
d342894c 2259
2260 mod_timer(&vxlan->age_timer, next_timer);
2261}
2262
9c2e24e1
PS		 2263static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2264{
56ef9c90 2265 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
9c2e24e1
PS		 2266 __u32 vni = vxlan->default_dst.remote_vni;
2267
56ef9c90 2268 spin_lock(&vn->sock_lock);
9c2e24e1 2269 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2270 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2271}
2272
d342894c 2273/* Setup stats when device is created */
2274static int vxlan_init(struct net_device *dev)
2275{
1c213bd2 2276 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2277 if (!dev->tstats)
d342894c 2278 return -ENOMEM;
2279
2280 return 0;
2281}
2282
ba609e9b 2283static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2284{
2285 struct vxlan_fdb *f;
2286
2287 spin_lock_bh(&vxlan->hash_lock);
2288 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2289 if (f)
2290 vxlan_fdb_destroy(vxlan, f);
2291 spin_unlock_bh(&vxlan->hash_lock);
2292}
2293
ebf4063e
SH		 2294static void vxlan_uninit(struct net_device *dev)
2295{
2296 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2297
ba609e9b 2298 vxlan_fdb_delete_default(vxlan);
afbd8bae 2299
ebf4063e
SH		 2300 free_percpu(dev->tstats);
2301}
2302
d342894c 2303/* Start ageing timer and join group when device is brought up */
2304static int vxlan_open(struct net_device *dev)
2305{
2306 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2307 int ret;
56ef9c90 2308
205f356d
JB		 2309 ret = vxlan_sock_add(vxlan);
2310 if (ret < 0)
2311 return ret;
d342894c 2312
79d4a94f 2313 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2314 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2315 if (ret == -EADDRINUSE)
2316 ret = 0;
56ef9c90 2317 if (ret) {
205f356d 2318 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2319 return ret;
2320 }
d342894c 2321 }
2322
0dfbdf41 2323 if (vxlan->cfg.age_interval)
d342894c 2324 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2325
56ef9c90 2326 return ret;
d342894c 2327}
2328
2329/* Purge the forwarding table */
2330static void vxlan_flush(struct vxlan_dev *vxlan)
2331{
31fec5aa 2332 unsigned int h;
d342894c 2333
2334 spin_lock_bh(&vxlan->hash_lock);
2335 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2336 struct hlist_node *p, *n;
2337 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2338 struct vxlan_fdb *f
2339 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2340 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2341 if (!is_zero_ether_addr(f->eth_addr))
2342 vxlan_fdb_destroy(vxlan, f);
d342894c 2343 }
2344 }
2345 spin_unlock_bh(&vxlan->hash_lock);
2346}
2347
2348/* Cleanup timer and forwarding table on shutdown */
2349static int vxlan_stop(struct net_device *dev)
2350{
2351 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2352 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2353 int ret = 0;
d342894c 2354
24c0e683 2355 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2356 !vxlan_group_used(vn, vxlan))
56ef9c90 2357 ret = vxlan_igmp_leave(vxlan);
d342894c 2358
2359 del_timer_sync(&vxlan->age_timer);
2360
2361 vxlan_flush(vxlan);
205f356d 2362 vxlan_sock_release(vxlan);
d342894c 2363
56ef9c90 2364 return ret;
d342894c 2365}
2366
d342894c 2367/* Stub, nothing needs to be done. */
2368static void vxlan_set_multicast_list(struct net_device *dev)
2369{
2370}
2371
72564b59
DW		 2372static int __vxlan_change_mtu(struct net_device *dev,
2373 struct net_device *lowerdev,
2374 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2375{
72564b59 2376 int max_mtu = IP_MAX_MTU;
345010b5 2377
72564b59
DW		 2378 if (lowerdev)
2379 max_mtu = lowerdev->mtu;
345010b5
DB		 2380
2381 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2382 max_mtu -= VXLAN6_HEADROOM;
345010b5 2383 else
72564b59 2384 max_mtu -= VXLAN_HEADROOM;
345010b5 2385
72564b59 2386 if (new_mtu < 68)
345010b5
DB		 2387 return -EINVAL;
2388
72564b59
DW		 2389 if (new_mtu > max_mtu) {
2390 if (strict)
2391 return -EINVAL;
2392
2393 new_mtu = max_mtu;
2394 }
2395
345010b5
DB		 2396 dev->mtu = new_mtu;
2397 return 0;
2398}
2399
72564b59
DW		 2400static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2401{
2402 struct vxlan_dev *vxlan = netdev_priv(dev);
2403 struct vxlan_rdst *dst = &vxlan->default_dst;
2404 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2405 dst->remote_ifindex);
2406 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2407}
2408
fc4099f1
PS		 2409static int egress_ipv4_tun_info(struct net_device *dev, struct sk_buff *skb,
2410 struct ip_tunnel_info *info,
2411 __be16 sport, __be16 dport)
2412{
2413 struct vxlan_dev *vxlan = netdev_priv(dev);
2414 struct rtable *rt;
2415 struct flowi4 fl4;
2416
2417 memset(&fl4, 0, sizeof(fl4));
2418 fl4.flowi4_tos = RT_TOS(info->key.tos);
2419 fl4.flowi4_mark = skb->mark;
2420 fl4.flowi4_proto = IPPROTO_UDP;
2421 fl4.daddr = info->key.u.ipv4.dst;
2422
2423 rt = ip_route_output_key(vxlan->net, &fl4);
2424 if (IS_ERR(rt))
2425 return PTR_ERR(rt);
2426 ip_rt_put(rt);
2427
2428 info->key.u.ipv4.src = fl4.saddr;
2429 info->key.tp_src = sport;
2430 info->key.tp_dst = dport;
2431 return 0;
2432}
2433
2434static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2435{
2436 struct vxlan_dev *vxlan = netdev_priv(dev);
2437 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2438 __be16 sport, dport;
2439
2440 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2441 vxlan->cfg.port_max, true);
2442 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2443
239e944f
JB		 2444 if (ip_tunnel_info_af(info) == AF_INET) {
2445 if (!vxlan->vn4_sock)
2446 return -EINVAL;
fc4099f1 2447 return egress_ipv4_tun_info(dev, skb, info, sport, dport);
239e944f
JB		 2448 } else {
2449#if IS_ENABLED(CONFIG_IPV6)
2450 struct dst_entry *ndst;
2451
2452 if (!vxlan->vn6_sock)
2453 return -EINVAL;
2454 ndst = vxlan6_get_route(vxlan, skb, 0,
2455 &info->key.u.ipv6.dst,
2456 &info->key.u.ipv6.src);
2457 if (IS_ERR(ndst))
2458 return PTR_ERR(ndst);
2459 dst_release(ndst);
2460
2461 info->key.tp_src = sport;
2462 info->key.tp_dst = dport;
2463#else /* !CONFIG_IPV6 */
2464 return -EPFNOSUPPORT;
2465#endif
2466 }
2467 return 0;
fc4099f1
PS		 2468}
2469
d342894c 2470static const struct net_device_ops vxlan_netdev_ops = {
2471 .ndo_init = vxlan_init,
ebf4063e 2472 .ndo_uninit = vxlan_uninit,
d342894c 2473 .ndo_open = vxlan_open,
2474 .ndo_stop = vxlan_stop,
2475 .ndo_start_xmit = vxlan_xmit,
e8171045 2476 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2477 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2478 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2479 .ndo_validate_addr = eth_validate_addr,
2480 .ndo_set_mac_address = eth_mac_addr,
2481 .ndo_fdb_add = vxlan_fdb_add,
2482 .ndo_fdb_del = vxlan_fdb_delete,
2483 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2484 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2485};
2486
2487/* Info for udev, that this is a virtual tunnel endpoint */
2488static struct device_type vxlan_type = {
2489 .name = "vxlan",
2490};
2491
53cf5275 2492/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2493 * supply the listening VXLAN udp ports. Callers are expected
2494 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2495 */
2496void vxlan_get_rx_port(struct net_device *dev)
2497{
2498 struct vxlan_sock *vs;
2499 struct net *net = dev_net(dev);
2500 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2501 sa_family_t sa_family;
35e42379
JG		 2502 __be16 port;
2503 unsigned int i;
53cf5275
JG		 2504
2505 spin_lock(&vn->sock_lock);
2506 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2507 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2508 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2509 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2510 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2511 port);
2512 }
2513 }
2514 spin_unlock(&vn->sock_lock);
2515}
2516EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2517
d342894c 2518/* Initialize the device structure. */
2519static void vxlan_setup(struct net_device *dev)
2520{
2521 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2522 unsigned int h;
d342894c 2523
2524 eth_hw_addr_random(dev);
2525 ether_setup(dev);
2526
2527 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2528 dev->destructor = free_netdev;
d342894c 2529 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2530
d342894c 2531 dev->features |= NETIF_F_LLTX;
d6727fe3 2532 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2533 dev->features |= NETIF_F_RXCSUM;
05c0db08 2534 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2535
1eaa8178
PS		 2536 dev->vlan_features = dev->features;
2537 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2538 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2539 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2540 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2541 netif_keep_dst(dev);
82a0f6b4 2542 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
22dba393 2543 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2544
553675fb 2545 INIT_LIST_HEAD(&vxlan->next);
d342894c 2546 spin_lock_init(&vxlan->hash_lock);
2547
2548 init_timer_deferrable(&vxlan->age_timer);
2549 vxlan->age_timer.function = vxlan_cleanup;
2550 vxlan->age_timer.data = (unsigned long) vxlan;
2551
0dfbdf41 2552 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2553
d342894c 2554 vxlan->dev = dev;
2555
58ce31cc
TH		 2556 gro_cells_init(&vxlan->gro_cells, dev);
2557
d342894c 2558 for (h = 0; h < FDB_HASH_SIZE; ++h)
2559 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2560}
2561
2562static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2563 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2564 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2565 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2566 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2567 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2568 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2569 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2570 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2571 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2572 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2573 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2574 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2575 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2576 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2577 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2578 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2579 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2580 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2581 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2582 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2583 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2584 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2585 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2586 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2587 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2588};
2589
2590static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2591{
2592 if (tb[IFLA_ADDRESS]) {
2593 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2594 pr_debug("invalid link address (not ethernet)\n");
2595 return -EINVAL;
2596 }
2597
2598 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2599 pr_debug("invalid all zero ethernet address\n");
2600 return -EADDRNOTAVAIL;
2601 }
2602 }
2603
2604 if (!data)
2605 return -EINVAL;
2606
2607 if (data[IFLA_VXLAN_ID]) {
2608 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2609 if (id >= VXLAN_VID_MASK)
2610 return -ERANGE;
2611 }
2612
05f47d69 2613 if (data[IFLA_VXLAN_PORT_RANGE]) {
2614 const struct ifla_vxlan_port_range *p
2615 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2616
2617 if (ntohs(p->high) < ntohs(p->low)) {
2618 pr_debug("port range %u .. %u not valid\n",
2619 ntohs(p->low), ntohs(p->high));
2620 return -EINVAL;
2621 }
2622 }
2623
d342894c 2624 return 0;
2625}
2626
1b13c97f
YB		 2627static void vxlan_get_drvinfo(struct net_device *netdev,
2628 struct ethtool_drvinfo *drvinfo)
2629{
2630 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2631 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2632}
2633
2634static const struct ethtool_ops vxlan_ethtool_ops = {
2635 .get_drvinfo = vxlan_get_drvinfo,
2636 .get_link = ethtool_op_get_link,
2637};
2638
553675fb 2639static void vxlan_del_work(struct work_struct *work)
2640{
2641 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2642 udp_tunnel_sock_release(vs->sock);
553675fb 2643 kfree_rcu(vs, rcu);
2644}
2645
3ee64f39
TH		 2646static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2647 __be16 port, u32 flags)
553675fb 2648{
e4c7ed41 2649 struct socket *sock;
3ee64f39
TH		 2650 struct udp_port_cfg udp_conf;
2651 int err;
e4c7ed41 2652
3ee64f39 2653 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2654
3ee64f39
TH		 2655 if (ipv6) {
2656 udp_conf.family = AF_INET6;
3ee64f39 2657 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2658 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2659 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2660 } else {
2661 udp_conf.family = AF_INET;
e4c7ed41 2662 }
e4c7ed41 2663
3ee64f39 2664 udp_conf.local_udp_port = port;
553675fb 2665
3ee64f39
TH		 2666 /* Open UDP socket */
2667 err = udp_sock_create(net, &udp_conf, &sock);
2668 if (err < 0)
2669 return ERR_PTR(err);
e4c7ed41 2670
39deb2c7 2671 return sock;
e4c7ed41
CW		 2672}
2673
2674/* Create new listen socket if needed */
b1be00a6
JB		 2675static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2676 __be16 port, u32 flags)
e4c7ed41
CW		 2677{
2678 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2679 struct vxlan_sock *vs;
2680 struct socket *sock;
e4c7ed41 2681 unsigned int h;
acbf74a7 2682 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2683
dc01e7d3 2684 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2685 if (!vs)
2686 return ERR_PTR(-ENOMEM);
2687
2688 for (h = 0; h < VNI_HASH_SIZE; ++h)
2689 INIT_HLIST_HEAD(&vs->vni_list[h]);
2690
2691 INIT_WORK(&vs->del_work, vxlan_del_work);
2692
3ee64f39 2693 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2694 if (IS_ERR(sock)) {
56ef9c90
MRL		 2695 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2696 PTR_ERR(sock));
553675fb 2697 kfree(vs);
e50fddc8 2698 return ERR_CAST(sock);
553675fb 2699 }
e4c7ed41
CW		 2700
2701 vs->sock = sock;
9c2e24e1 2702 atomic_set(&vs->refcnt, 1);
af33c1ad 2703 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2704
dc01e7d3
OG		 2705 /* Initialize the vxlan udp offloads structure */
2706 vs->udp_offloads.port = port;
2707 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2708 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2709
9c2e24e1
PS		 2710 spin_lock(&vn->sock_lock);
2711 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2712 vxlan_notify_add_rx_port(vs);
9c2e24e1 2713 spin_unlock(&vn->sock_lock);
553675fb 2714
2715 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2716 tunnel_cfg.sk_user_data = vs;
2717 tunnel_cfg.encap_type = 1;
2718 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2719 tunnel_cfg.encap_destroy = NULL;
2720
2721 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2722
9c2e24e1
PS		 2723 return vs;
2724}
2725
b1be00a6 2726static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2727{
205f356d
JB		 2728 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2729 struct vxlan_sock *vs = NULL;
9c2e24e1 2730
205f356d 2731 if (!vxlan->cfg.no_share) {
56ef9c90 2732 spin_lock(&vn->sock_lock);
205f356d
JB		 2733 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2734 vxlan->cfg.dst_port, vxlan->flags);
2735 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2736 spin_unlock(&vn->sock_lock);
205f356d 2737 return -EBUSY;
56ef9c90
MRL		 2738 }
2739 spin_unlock(&vn->sock_lock);
2740 }
205f356d 2741 if (!vs)
b1be00a6
JB		 2742 vs = vxlan_socket_create(vxlan->net, ipv6,
2743 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2744 if (IS_ERR(vs))
2745 return PTR_ERR(vs);
b1be00a6
JB		 2746#if IS_ENABLED(CONFIG_IPV6)
2747 if (ipv6)
2748 vxlan->vn6_sock = vs;
2749 else
2750#endif
2751 vxlan->vn4_sock = vs;
205f356d
JB		 2752 vxlan_vs_add_dev(vs, vxlan);
2753 return 0;
553675fb 2754}
2755
b1be00a6
JB		 2756static int vxlan_sock_add(struct vxlan_dev *vxlan)
2757{
2758 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2759 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2760 int ret = 0;
2761
2762 vxlan->vn4_sock = NULL;
2763#if IS_ENABLED(CONFIG_IPV6)
2764 vxlan->vn6_sock = NULL;
2765 if (ipv6 || metadata)
2766 ret = __vxlan_sock_add(vxlan, true);
2767#endif
2768 if (!ret && (!ipv6 || metadata))
2769 ret = __vxlan_sock_add(vxlan, false);
2770 if (ret < 0)
2771 vxlan_sock_release(vxlan);
2772 return ret;
2773}
2774
0dfbdf41
TG		 2775static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2776 struct vxlan_config *conf)
d342894c 2777{
33564bbb 2778 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2779 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2780 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2781 unsigned short needed_headroom = ETH_HLEN;
d342894c 2782 int err;
e4c7ed41 2783 bool use_ipv6 = false;
0dfbdf41 2784 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2785 struct net_device *lowerdev = NULL;
d342894c 2786
33564bbb 2787 vxlan->net = src_net;
f01ec1c0 2788
0dfbdf41 2789 dst->remote_vni = conf->vni;
d342894c 2790
0dfbdf41 2791 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2792
0dfbdf41
TG		 2793 /* Unless IPv6 is explicitly requested, assume IPv4 */
2794 if (!dst->remote_ip.sa.sa_family)
2795 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2796
0dfbdf41 2797 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2798 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2799 if (!IS_ENABLED(CONFIG_IPV6))
2800 return -EPFNOSUPPORT;
e4c7ed41 2801 use_ipv6 = true;
b1be00a6 2802 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2803 }
d342894c 2804
0dfbdf41 2805 if (conf->remote_ifindex) {
7e059158 2806 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2807 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2808
2809 if (!lowerdev) {
c7995c43 2810 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2811 return -ENODEV;
2812 }
d342894c 2813
e4c7ed41
CW		 2814#if IS_ENABLED(CONFIG_IPV6)
2815 if (use_ipv6) {
2816 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2817 if (idev && idev->cnf.disable_ipv6) {
2818 pr_info("IPv6 is disabled via sysctl\n");
2819 return -EPERM;
2820 }
e4c7ed41
CW		 2821 }
2822#endif
2823
0dfbdf41 2824 if (!conf->mtu)
e4c7ed41 2825 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2826
b1be00a6 2827 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2828 }
d342894c 2829
7e059158
DW		 2830 if (conf->mtu) {
2831 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2832 if (err)
2833 return err;
2834 }
2835
b1be00a6
JB		 2836 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2837 needed_headroom += VXLAN6_HEADROOM;
2838 else
2839 needed_headroom += VXLAN_HEADROOM;
2840 dev->needed_headroom = needed_headroom;
2841
0dfbdf41
TG		 2842 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2843 if (!vxlan->cfg.dst_port)
2844 vxlan->cfg.dst_port = default_port;
2845 vxlan->flags |= conf->flags;
2846
2847 if (!vxlan->cfg.age_interval)
2848 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2849
07b9b37c
ND		 2850 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2851 if (tmp->cfg.vni == conf->vni &&
2852 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2853 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2854 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2855 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2856 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2857 return -EEXIST;
07b9b37c 2858 }
0dfbdf41
TG		 2859
2860 dev->ethtool_ops = &vxlan_ethtool_ops;
2861
2862 /* create an fdb entry for a valid default destination */
2863 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2864 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2865 &vxlan->default_dst.remote_ip,
2866 NUD_REACHABLE|NUD_PERMANENT,
2867 NLM_F_EXCL|NLM_F_CREATE,
2868 vxlan->cfg.dst_port,
2869 vxlan->default_dst.remote_vni,
2870 vxlan->default_dst.remote_ifindex,
2871 NTF_SELF);
2872 if (err)
2873 return err;
2874 }
2875
2876 err = register_netdevice(dev);
2877 if (err) {
2878 vxlan_fdb_delete_default(vxlan);
2879 return err;
2880 }
2881
2882 list_add(&vxlan->next, &vn->vxlan_list);
2883
2884 return 0;
2885}
2886
2887struct net_device *vxlan_dev_create(struct net *net, const char *name,
2888 u8 name_assign_type, struct vxlan_config *conf)
2889{
2890 struct nlattr *tb[IFLA_MAX+1];
2891 struct net_device *dev;
2892 int err;
2893
2894 memset(&tb, 0, sizeof(tb));
2895
2896 dev = rtnl_create_link(net, name, name_assign_type,
2897 &vxlan_link_ops, tb);
2898 if (IS_ERR(dev))
2899 return dev;
2900
2901 err = vxlan_dev_configure(net, dev, conf);
2902 if (err < 0) {
2903 free_netdev(dev);
2904 return ERR_PTR(err);
2905 }
2906
2907 return dev;
2908}
2909EXPORT_SYMBOL_GPL(vxlan_dev_create);
2910
2911static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2912 struct nlattr *tb[], struct nlattr *data[])
2913{
2914 struct vxlan_config conf;
2915 int err;
2916
0dfbdf41 2917 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2918
2919 if (data[IFLA_VXLAN_ID])
2920 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
0dfbdf41
TG		 2921
2922 if (data[IFLA_VXLAN_GROUP]) {
2923 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2924 } else if (data[IFLA_VXLAN_GROUP6]) {
2925 if (!IS_ENABLED(CONFIG_IPV6))
2926 return -EPFNOSUPPORT;
2927
2928 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2929 conf.remote_ip.sa.sa_family = AF_INET6;
2930 }
2931
2932 if (data[IFLA_VXLAN_LOCAL]) {
2933 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2934 conf.saddr.sa.sa_family = AF_INET;
2935 } else if (data[IFLA_VXLAN_LOCAL6]) {
2936 if (!IS_ENABLED(CONFIG_IPV6))
2937 return -EPFNOSUPPORT;
2938
2939 /* TODO: respect scope id */
2940 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2941 conf.saddr.sa.sa_family = AF_INET6;
2942 }
2943
2944 if (data[IFLA_VXLAN_LINK])
2945 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2946
d342894c 2947 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2948 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2949
afb97186 2950 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2951 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2952
d342894c 2953 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2954 conf.flags |= VXLAN_F_LEARN;
d342894c 2955
2956 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2957 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2958
e4f67add 2959 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2960 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2961
2962 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2963 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2964
2965 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2966 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2967
2968 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2969 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2970
d342894c 2971 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2972 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2973
f8a9b1bc
AS		 2974 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2975 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2976 conf.flags |= VXLAN_F_COLLECT_METADATA;
2977
05f47d69 2978 if (data[IFLA_VXLAN_PORT_RANGE]) {
2979 const struct ifla_vxlan_port_range *p
2980 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2981 conf.port_min = ntohs(p->low);
2982 conf.port_max = ntohs(p->high);
05f47d69 2983 }
2984
823aa873 2985 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2986 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2987
359a0ea9 2988 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
0dfbdf41 2989 conf.flags |= VXLAN_F_UDP_CSUM;
359a0ea9
TH		 2990
2991 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2992 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2993 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2994
2995 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2996 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2997 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2998
dfd8645e
TH		 2999 if (data[IFLA_VXLAN_REMCSUM_TX] &&
3000 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 3001 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 3002
3003 if (data[IFLA_VXLAN_REMCSUM_RX] &&
3004 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 3005 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 3006
3511494c 3007 if (data[IFLA_VXLAN_GBP])
0dfbdf41 3008 conf.flags |= VXLAN_F_GBP;
3511494c 3009
0ace2ca8 3010 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 3011 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 3012
0dfbdf41
TG		 3013 err = vxlan_dev_configure(src_net, dev, &conf);
3014 switch (err) {
3015 case -ENODEV:
3016 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
3017 break;
1b13c97f 3018
0dfbdf41
TG		 3019 case -EPERM:
3020 pr_info("IPv6 is disabled via sysctl\n");
3021 break;
d342894c 3022
0dfbdf41
TG		 3023 case -EEXIST:
3024 pr_info("duplicate VNI %u\n", conf.vni);
3025 break;
afbd8bae
MR		 3026 }
3027
0dfbdf41 3028 return err;
d342894c 3029}
3030
3031static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3032{
3033 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3034 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 3035
fe5c3561 3036 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 3037 if (!hlist_unhashed(&vxlan->hlist))
3038 hlist_del_rcu(&vxlan->hlist);
fe5c3561 3039 spin_unlock(&vn->sock_lock);
3040
58ce31cc 3041 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3042 list_del(&vxlan->next);
d342894c 3043 unregister_netdevice_queue(dev, head);
3044}
3045
3046static size_t vxlan_get_size(const struct net_device *dev)
3047{
3048
3049 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3050 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3051 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3052 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3053 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3054 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3055 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3056 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3057 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3058 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3059 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3060 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3061 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3062 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3063 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3064 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3065 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3066 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3067 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3068 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3069 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3070 0;
3071}
3072
3073static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3074{
3075 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3076 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3077 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3078 .low = htons(vxlan->cfg.port_min),
3079 .high = htons(vxlan->cfg.port_max),
05f47d69 3080 };
d342894c 3081
c7995c43 3082 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 3083 goto nla_put_failure;
3084
e4c7ed41
CW		 3085 if (!vxlan_addr_any(&dst->remote_ip)) {
3086 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3087 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3088 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3089 goto nla_put_failure;
3090#if IS_ENABLED(CONFIG_IPV6)
3091 } else {
930345ea
JB		 3092 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3093 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3094 goto nla_put_failure;
3095#endif
3096 }
3097 }
d342894c 3098
c7995c43 3099 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3100 goto nla_put_failure;
3101
0dfbdf41
TG		 3102 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3103 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3104 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3105 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3106 goto nla_put_failure;
3107#if IS_ENABLED(CONFIG_IPV6)
3108 } else {
930345ea 3109 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3110 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3111 goto nla_put_failure;
3112#endif
3113 }
3114 }
d342894c 3115
0dfbdf41
TG		 3116 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3117 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 3118 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3119 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3120 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3121 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3122 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3123 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3124 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3125 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3126 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3127 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3128 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3129 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3130 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3131 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9
TH		 3132 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3133 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3134 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3135 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3136 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3137 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3138 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3139 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3140 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3141 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3142 goto nla_put_failure;
3143
05f47d69 3144 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3145 goto nla_put_failure;
3146
3511494c
TG		 3147 if (vxlan->flags & VXLAN_F_GBP &&
3148 nla_put_flag(skb, IFLA_VXLAN_GBP))
3149 goto nla_put_failure;
3150
0ace2ca8
TH		 3151 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3152 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3153 goto nla_put_failure;
3154
d342894c 3155 return 0;
3156
3157nla_put_failure:
3158 return -EMSGSIZE;
3159}
3160
1728d4fa
ND		 3161static struct net *vxlan_get_link_net(const struct net_device *dev)
3162{
3163 struct vxlan_dev *vxlan = netdev_priv(dev);
3164
3165 return vxlan->net;
3166}
3167
d342894c 3168static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3169 .kind = "vxlan",
3170 .maxtype = IFLA_VXLAN_MAX,
3171 .policy = vxlan_policy,
3172 .priv_size = sizeof(struct vxlan_dev),
3173 .setup = vxlan_setup,
3174 .validate = vxlan_validate,
3175 .newlink = vxlan_newlink,
3176 .dellink = vxlan_dellink,
3177 .get_size = vxlan_get_size,
3178 .fill_info = vxlan_fill_info,
1728d4fa 3179 .get_link_net = vxlan_get_link_net,
d342894c 3180};
3181
acaf4e70
DB		 3182static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3183 struct net_device *dev)
3184{
3185 struct vxlan_dev *vxlan, *next;
3186 LIST_HEAD(list_kill);
3187
3188 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3189 struct vxlan_rdst *dst = &vxlan->default_dst;
3190
3191 /* In case we created vxlan device with carrier
3192 * and we loose the carrier due to module unload
3193 * we also need to remove vxlan device. In other
3194 * cases, it's not necessary and remote_ifindex
3195 * is 0 here, so no matches.
3196 */
3197 if (dst->remote_ifindex == dev->ifindex)
3198 vxlan_dellink(vxlan->dev, &list_kill);
3199 }
3200
3201 unregister_netdevice_many(&list_kill);
3202}
3203
3204static int vxlan_lowerdev_event(struct notifier_block *unused,
3205 unsigned long event, void *ptr)
3206{
3207 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3208 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3209
783c1463 3210 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3211 vxlan_handle_lowerdev_unregister(vn, dev);
3212
3213 return NOTIFY_DONE;
3214}
3215
3216static struct notifier_block vxlan_notifier_block __read_mostly = {
3217 .notifier_call = vxlan_lowerdev_event,
3218};
3219
d342894c 3220static __net_init int vxlan_init_net(struct net *net)
3221{
3222 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3223 unsigned int h;
d342894c 3224
553675fb 3225 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3226 spin_lock_init(&vn->sock_lock);
d342894c 3227
553675fb 3228 for (h = 0; h < PORT_HASH_SIZE; ++h)
3229 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3230
3231 return 0;
3232}
3233
f01ec1c0
ND		 3234static void __net_exit vxlan_exit_net(struct net *net)
3235{
3236 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3237 struct vxlan_dev *vxlan, *next;
3238 struct net_device *dev, *aux;
3239 LIST_HEAD(list);
3240
3241 rtnl_lock();
3242 for_each_netdev_safe(net, dev, aux)
3243 if (dev->rtnl_link_ops == &vxlan_link_ops)
3244 unregister_netdevice_queue(dev, &list);
3245
3246 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3247 /* If vxlan->dev is in the same netns, it has already been added
3248 * to the list by the previous loop.
3249 */
58ce31cc
TH		 3250 if (!net_eq(dev_net(vxlan->dev), net)) {
3251 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3252 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3253 }
f01ec1c0
ND		 3254 }
3255
3256 unregister_netdevice_many(&list);
3257 rtnl_unlock();
3258}
3259
d342894c 3260static struct pernet_operations vxlan_net_ops = {
3261 .init = vxlan_init_net,
f01ec1c0 3262 .exit = vxlan_exit_net,
d342894c 3263 .id = &vxlan_net_id,
3264 .size = sizeof(struct vxlan_net),
3265};
3266
3267static int __init vxlan_init_module(void)
3268{
3269 int rc;
3270
758c57d1
SH		 3271 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3272 if (!vxlan_wq)
3273 return -ENOMEM;
3274
d342894c 3275 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3276
783c1463 3277 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3278 if (rc)
3279 goto out1;
3280
acaf4e70 3281 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3282 if (rc)
3283 goto out2;
3284
acaf4e70
DB		 3285 rc = rtnl_link_register(&vxlan_link_ops);
3286 if (rc)
3287 goto out3;
d342894c 3288
acaf4e70
DB		 3289 return 0;
3290out3:
3291 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3292out2:
783c1463 3293 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3294out1:
758c57d1 3295 destroy_workqueue(vxlan_wq);
d342894c 3296 return rc;
3297}
7332a13b 3298late_initcall(vxlan_init_module);
d342894c 3299
3300static void __exit vxlan_cleanup_module(void)
3301{
b7153984 3302 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3303 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3304 destroy_workqueue(vxlan_wq);
783c1463
DB		 3305 unregister_pernet_subsys(&vxlan_net_ops);
3306 /* rcu_barrier() is called by netns */
d342894c 3307}
3308module_exit(vxlan_cleanup_module);
3309
3310MODULE_LICENSE("GPL");
3311MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3312MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3313MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3314MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)