projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bpf: add helpers to access tunnel metadata
[deliverable/linux.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
afbd8bae
MR		 76static const u8 all_zeros_mac[ETH_ALEN];
77
614732ea
TG		 78static struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
79 bool no_share, u32 flags);
80
553675fb 81/* per-network namespace private data for this module */
82struct vxlan_net {
83 struct list_head vxlan_list;
84 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 85 spinlock_t sock_lock;
553675fb 86};
87
d342894c 88/* Forwarding table entry */
89struct vxlan_fdb {
90 struct hlist_node hlist; /* linked list of entries */
91 struct rcu_head rcu;
92 unsigned long updated; /* jiffies */
93 unsigned long used;
3e61aa8f 94 struct list_head remotes;
7177a3b0 95 u8 eth_addr[ETH_ALEN];
d342894c 96 u16 state; /* see ndm_state */
ae884082 97 u8 flags; /* see ndm_flags */
d342894c 98};
99
d342894c 100/* salt for hash table */
101static u32 vxlan_salt __read_mostly;
758c57d1 102static struct workqueue_struct *vxlan_wq;
d342894c 103
ee122c79
TG		 104static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
105{
e7030878
TG		 106 return vs->flags & VXLAN_F_COLLECT_METADATA ||
107 ip_tunnel_collect_metadata();
ee122c79
TG		 108}
109
e4c7ed41
CW		 110#if IS_ENABLED(CONFIG_IPV6)
111static inline
112bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
113{
f0ef3126
JB		 114 if (a->sa.sa_family != b->sa.sa_family)
115 return false;
116 if (a->sa.sa_family == AF_INET6)
117 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
118 else
119 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 120}
121
122static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
123{
f0ef3126
JB		 124 if (ipa->sa.sa_family == AF_INET6)
125 return ipv6_addr_any(&ipa->sin6.sin6_addr);
126 else
127 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 128}
129
130static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
131{
f0ef3126
JB		 132 if (ipa->sa.sa_family == AF_INET6)
133 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
134 else
135 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 136}
137
138static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
139{
f0ef3126 140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 141 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 142 ip->sa.sa_family = AF_INET6;
143 return 0;
144 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 145 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 146 ip->sa.sa_family = AF_INET;
147 return 0;
148 } else {
149 return -EAFNOSUPPORT;
150 }
e4c7ed41
CW		 151}
152
153static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 154 const union vxlan_addr *ip)
e4c7ed41 155{
f0ef3126 156 if (ip->sa.sa_family == AF_INET6)
930345ea 157 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 158 else
930345ea 159 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 160}
161
162#else /* !CONFIG_IPV6 */
163
164static inline
165bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
166{
f0ef3126 167 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 168}
169
170static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
171{
f0ef3126 172 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 173}
174
175static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
176{
f0ef3126 177 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 178}
179
180static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
181{
f0ef3126
JB		 182 if (nla_len(nla) >= sizeof(struct in6_addr)) {
183 return -EAFNOSUPPORT;
184 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 185 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 186 ip->sa.sa_family = AF_INET;
187 return 0;
188 } else {
189 return -EAFNOSUPPORT;
190 }
e4c7ed41
CW		 191}
192
193static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 194 const union vxlan_addr *ip)
e4c7ed41 195{
930345ea 196 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 197}
198#endif
199
553675fb 200/* Virtual Network hash table head */
201static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
202{
203 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
204}
205
206/* Socket hash table head */
207static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 208{
209 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
210
553675fb 211 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
212}
213
3e61aa8f
SH		 214/* First remote destination for a forwarding entry.
215 * Guaranteed to be non-NULL because remotes are never deleted.
216 */
5ca5461c 217static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 218{
5ca5461c 219 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
220}
221
222static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
223{
224 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 225}
226
ac5132d1
TG		 227/* Find VXLAN socket based on network namespace, address family and UDP port
228 * and enabled unshareable flags.
229 */
230static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
231 __be16 port, u32 flags)
553675fb 232{
233 struct vxlan_sock *vs;
af33c1ad
TH		 234
235 flags &= VXLAN_F_RCV_FLAGS;
553675fb 236
237 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 238 if (inet_sk(vs->sock->sk)->inet_sport == port &&
ac5132d1 239 inet_sk(vs->sock->sk)->sk.sk_family == family &&
af33c1ad 240 vs->flags == flags)
553675fb 241 return vs;
242 }
243 return NULL;
d342894c 244}
245
5cfccc5a 246static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 247{
248 struct vxlan_dev *vxlan;
d342894c 249
553675fb 250 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 251 if (vxlan->default_dst.remote_vni == id)
d342894c 252 return vxlan;
253 }
254
255 return NULL;
256}
257
5cfccc5a 258/* Look up VNI in a per net namespace table */
19ca9fc1 259static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 260 sa_family_t family, __be16 port,
261 u32 flags)
5cfccc5a
PS		 262{
263 struct vxlan_sock *vs;
264
ac5132d1 265 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 266 if (!vs)
267 return NULL;
268
269 return vxlan_vs_find_vni(vs, id);
270}
271
d342894c 272/* Fill in neighbour message in skbuff. */
273static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 274 const struct vxlan_fdb *fdb,
275 u32 portid, u32 seq, int type, unsigned int flags,
276 const struct vxlan_rdst *rdst)
d342894c 277{
278 unsigned long now = jiffies;
279 struct nda_cacheinfo ci;
280 struct nlmsghdr *nlh;
281 struct ndmsg *ndm;
e4f67add 282 bool send_ip, send_eth;
d342894c 283
284 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
285 if (nlh == NULL)
286 return -EMSGSIZE;
287
288 ndm = nlmsg_data(nlh);
289 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 290
291 send_eth = send_ip = true;
292
293 if (type == RTM_GETNEIGH) {
294 ndm->ndm_family = AF_INET;
e4c7ed41 295 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 296 send_eth = !is_zero_ether_addr(fdb->eth_addr);
297 } else
298 ndm->ndm_family = AF_BRIDGE;
d342894c 299 ndm->ndm_state = fdb->state;
300 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 301 ndm->ndm_flags = fdb->flags;
545469f7 302 ndm->ndm_type = RTN_UNICAST;
d342894c 303
193523bf 304 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 305 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 306 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 307 goto nla_put_failure;
308
e4f67add 309 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 310 goto nla_put_failure;
311
e4c7ed41 312 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 313 goto nla_put_failure;
314
0dfbdf41 315 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 316 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
317 goto nla_put_failure;
c7995c43 318 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 319 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 320 goto nla_put_failure;
321 if (rdst->remote_ifindex &&
322 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 323 goto nla_put_failure;
324
325 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
326 ci.ndm_confirmed = 0;
327 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
328 ci.ndm_refcnt = 0;
329
330 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
331 goto nla_put_failure;
332
053c095a
JB		 333 nlmsg_end(skb, nlh);
334 return 0;
d342894c 335
336nla_put_failure:
337 nlmsg_cancel(skb, nlh);
338 return -EMSGSIZE;
339}
340
341static inline size_t vxlan_nlmsg_size(void)
342{
343 return NLMSG_ALIGN(sizeof(struct ndmsg))
344 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 345 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 346 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 347 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
348 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 349 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 350 + nla_total_size(sizeof(struct nda_cacheinfo));
351}
352
9e4b93f9
ND		 353static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
354 struct vxlan_rdst *rd, int type)
d342894c 355{
356 struct net *net = dev_net(vxlan->dev);
357 struct sk_buff *skb;
358 int err = -ENOBUFS;
359
360 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
361 if (skb == NULL)
362 goto errout;
363
9e4b93f9 364 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 365 if (err < 0) {
366 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
367 WARN_ON(err == -EMSGSIZE);
368 kfree_skb(skb);
369 goto errout;
370 }
371
372 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
373 return;
374errout:
375 if (err < 0)
376 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
377}
378
e4c7ed41 379static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 380{
381 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 382 struct vxlan_fdb f = {
383 .state = NUD_STALE,
384 };
385 struct vxlan_rdst remote = {
e4c7ed41 386 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 387 .remote_vni = VXLAN_N_VID,
388 };
3e61aa8f 389
9e4b93f9 390 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 391}
392
393static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
394{
bb3fd687
SH		 395 struct vxlan_fdb f = {
396 .state = NUD_STALE,
397 };
9e4b93f9 398 struct vxlan_rdst remote = { };
e4f67add 399
e4f67add
DS		 400 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
401
9e4b93f9 402 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 403}
404
d342894c 405/* Hash Ethernet address */
406static u32 eth_hash(const unsigned char *addr)
407{
408 u64 value = get_unaligned((u64 *)addr);
409
410 /* only want 6 bytes */
411#ifdef __BIG_ENDIAN
d342894c 412 value >>= 16;
321fb991 413#else
414 value <<= 16;
d342894c 415#endif
416 return hash_64(value, FDB_HASH_BITS);
417}
418
419/* Hash chain to use given mac address */
420static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
421 const u8 *mac)
422{
423 return &vxlan->fdb_head[eth_hash(mac)];
424}
425
426/* Look up Ethernet address in forwarding table */
014be2c8 427static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 428 const u8 *mac)
d342894c 429{
430 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
431 struct vxlan_fdb *f;
d342894c 432
b67bfe0d 433 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 434 if (ether_addr_equal(mac, f->eth_addr))
d342894c 435 return f;
436 }
437
438 return NULL;
439}
440
014be2c8
SS		 441static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
442 const u8 *mac)
443{
444 struct vxlan_fdb *f;
445
446 f = __vxlan_find_mac(vxlan, mac);
447 if (f)
448 f->used = jiffies;
449
450 return f;
451}
452
a5e7c10a
MR		 453/* caller should hold vxlan->hash_lock */
454static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 455 union vxlan_addr *ip, __be16 port,
a5e7c10a 456 __u32 vni, __u32 ifindex)
6681712d 457{
3e61aa8f 458 struct vxlan_rdst *rd;
6681712d 459
3e61aa8f 460 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 461 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 462 rd->remote_port == port &&
463 rd->remote_vni == vni &&
464 rd->remote_ifindex == ifindex)
a5e7c10a 465 return rd;
6681712d 466 }
3e61aa8f 467
a5e7c10a
MR		 468 return NULL;
469}
470
906dc186
TR		 471/* Replace destination of unicast mac */
472static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 473 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 474{
475 struct vxlan_rdst *rd;
476
477 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
478 if (rd)
479 return 0;
480
481 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
482 if (!rd)
483 return 0;
e4c7ed41 484 rd->remote_ip = *ip;
906dc186
TR		 485 rd->remote_port = port;
486 rd->remote_vni = vni;
487 rd->remote_ifindex = ifindex;
488 return 1;
489}
490
a5e7c10a
MR		 491/* Add/update destinations for multicast */
492static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 493 union vxlan_addr *ip, __be16 port, __u32 vni,
494 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 495{
496 struct vxlan_rdst *rd;
497
498 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
499 if (rd)
500 return 0;
501
6681712d
DS		 502 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
503 if (rd == NULL)
504 return -ENOBUFS;
e4c7ed41 505 rd->remote_ip = *ip;
6681712d
DS		 506 rd->remote_port = port;
507 rd->remote_vni = vni;
508 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 509
510 list_add_tail_rcu(&rd->list, &f->remotes);
511
9e4b93f9 512 *rdp = rd;
6681712d
DS		 513 return 1;
514}
515
dfd8645e
TH		 516static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
517 unsigned int off,
518 struct vxlanhdr *vh, size_t hdrlen,
0ace2ca8
TH		 519 u32 data, struct gro_remcsum *grc,
520 bool nopartial)
dfd8645e
TH		 521{
522 size_t start, offset, plen;
dfd8645e
TH		 523
524 if (skb->remcsum_offload)
26c4f7da 525 return NULL;
dfd8645e
TH		 526
527 if (!NAPI_GRO_CB(skb)->csum_valid)
528 return NULL;
529
530 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
531 offset = start + ((data & VXLAN_RCO_UDP) ?
532 offsetof(struct udphdr, check) :
533 offsetof(struct tcphdr, check));
534
535 plen = hdrlen + offset + sizeof(u16);
536
537 /* Pull checksum that will be written */
538 if (skb_gro_header_hard(skb, off + plen)) {
539 vh = skb_gro_header_slow(skb, off + plen, off);
540 if (!vh)
541 return NULL;
542 }
543
26c4f7da 544 skb_gro_remcsum_process(skb, (void *)vh + hdrlen,
0ace2ca8 545 start, offset, grc, nopartial);
dfd8645e
TH		 546
547 skb->remcsum_offload = 1;
548
549 return vh;
550}
551
a2b12f3c
TH		 552static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
553 struct sk_buff *skb,
554 struct udp_offload *uoff)
dc01e7d3
OG		 555{
556 struct sk_buff *p, **pp = NULL;
557 struct vxlanhdr *vh, *vh2;
9b174d88 558 unsigned int hlen, off_vx;
dc01e7d3 559 int flush = 1;
dfd8645e
TH		 560 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
561 udp_offloads);
562 u32 flags;
26c4f7da
TH		 563 struct gro_remcsum grc;
564
565 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 566
567 off_vx = skb_gro_offset(skb);
568 hlen = off_vx + sizeof(*vh);
569 vh = skb_gro_header_fast(skb, off_vx);
570 if (skb_gro_header_hard(skb, hlen)) {
571 vh = skb_gro_header_slow(skb, hlen, off_vx);
572 if (unlikely(!vh))
573 goto out;
574 }
dc01e7d3 575
dfd8645e
TH		 576 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
577 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
578
579 flags = ntohl(vh->vx_flags);
580
581 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
582 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
0ace2ca8
TH		 583 ntohl(vh->vx_vni), &grc,
584 !!(vs->flags &
585 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 586
587 if (!vh)
588 goto out;
589 }
590
dc01e7d3
OG		 591 flush = 0;
592
593 for (p = *head; p; p = p->next) {
594 if (!NAPI_GRO_CB(p)->same_flow)
595 continue;
596
597 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 598 if (vh->vx_flags != vh2->vx_flags ||
599 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 600 NAPI_GRO_CB(p)->same_flow = 0;
601 continue;
602 }
dc01e7d3
OG		 603 }
604
9b174d88 605 pp = eth_gro_receive(head, skb);
dc01e7d3 606
dc01e7d3 607out:
26c4f7da 608 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 609 NAPI_GRO_CB(skb)->flush |= flush;
610
611 return pp;
612}
613
a2b12f3c
TH		 614static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
615 struct udp_offload *uoff)
dc01e7d3 616{
cfdf1e1b
JG		 617 udp_tunnel_gro_complete(skb, nhoff);
618
9b174d88 619 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 620}
621
53cf5275 622/* Notify netdevs that UDP port started listening */
dc01e7d3 623static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 624{
625 struct net_device *dev;
dc01e7d3 626 struct sock *sk = vs->sock->sk;
53cf5275
JG		 627 struct net *net = sock_net(sk);
628 sa_family_t sa_family = sk->sk_family;
35e42379 629 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 630 int err;
631
632 if (sa_family == AF_INET) {
633 err = udp_add_offload(&vs->udp_offloads);
634 if (err)
635 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
636 }
53cf5275
JG		 637
638 rcu_read_lock();
639 for_each_netdev_rcu(net, dev) {
640 if (dev->netdev_ops->ndo_add_vxlan_port)
641 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
642 port);
643 }
644 rcu_read_unlock();
645}
646
647/* Notify netdevs that UDP port is no more listening */
dc01e7d3 648static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 649{
650 struct net_device *dev;
dc01e7d3 651 struct sock *sk = vs->sock->sk;
53cf5275
JG		 652 struct net *net = sock_net(sk);
653 sa_family_t sa_family = sk->sk_family;
35e42379 654 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 655
656 rcu_read_lock();
657 for_each_netdev_rcu(net, dev) {
658 if (dev->netdev_ops->ndo_del_vxlan_port)
659 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
660 port);
661 }
662 rcu_read_unlock();
dc01e7d3
OG		 663
664 if (sa_family == AF_INET)
665 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 666}
667
d342894c 668/* Add new entry to forwarding table -- assumes lock held */
669static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 670 const u8 *mac, union vxlan_addr *ip,
6681712d 671 __u16 state, __u16 flags,
73cf3317 672 __be16 port, __u32 vni, __u32 ifindex,
ae884082 673 __u8 ndm_flags)
d342894c 674{
9e4b93f9 675 struct vxlan_rdst *rd = NULL;
d342894c 676 struct vxlan_fdb *f;
677 int notify = 0;
678
014be2c8 679 f = __vxlan_find_mac(vxlan, mac);
d342894c 680 if (f) {
681 if (flags & NLM_F_EXCL) {
682 netdev_dbg(vxlan->dev,
683 "lost race to create %pM\n", mac);
684 return -EEXIST;
685 }
686 if (f->state != state) {
687 f->state = state;
688 f->updated = jiffies;
689 notify = 1;
690 }
ae884082
DS		 691 if (f->flags != ndm_flags) {
692 f->flags = ndm_flags;
693 f->updated = jiffies;
694 notify = 1;
695 }
906dc186
TR		 696 if ((flags & NLM_F_REPLACE)) {
697 /* Only change unicasts */
698 if (!(is_multicast_ether_addr(f->eth_addr) ||
699 is_zero_ether_addr(f->eth_addr))) {
60840429 700 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 701 ifindex);
906dc186
TR		 702 } else
703 return -EOPNOTSUPP;
704 }
6681712d 705 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 706 (is_multicast_ether_addr(f->eth_addr) ||
707 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 708 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
709 &rd);
6681712d
DS		 710
711 if (rc < 0)
712 return rc;
713 notify |= rc;
714 }
d342894c 715 } else {
716 if (!(flags & NLM_F_CREATE))
717 return -ENOENT;
718
0dfbdf41
TG		 719 if (vxlan->cfg.addrmax &&
720 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 721 return -ENOSPC;
722
906dc186
TR		 723 /* Disallow replace to add a multicast entry */
724 if ((flags & NLM_F_REPLACE) &&
725 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
726 return -EOPNOTSUPP;
727
e4c7ed41 728 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 729 f = kmalloc(sizeof(*f), GFP_ATOMIC);
730 if (!f)
731 return -ENOMEM;
732
733 notify = 1;
d342894c 734 f->state = state;
ae884082 735 f->flags = ndm_flags;
d342894c 736 f->updated = f->used = jiffies;
3e61aa8f 737 INIT_LIST_HEAD(&f->remotes);
d342894c 738 memcpy(f->eth_addr, mac, ETH_ALEN);
739
9e4b93f9 740 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 741
d342894c 742 ++vxlan->addrcnt;
743 hlist_add_head_rcu(&f->hlist,
744 vxlan_fdb_head(vxlan, mac));
745 }
746
9e4b93f9
ND		 747 if (notify) {
748 if (rd == NULL)
749 rd = first_remote_rtnl(f);
750 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
751 }
d342894c 752
753 return 0;
754}
755
6706c82e 756static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 757{
758 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 759 struct vxlan_rdst *rd, *nd;
6681712d 760
3e61aa8f 761 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 762 kfree(rd);
6681712d
DS		 763 kfree(f);
764}
765
d342894c 766static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
767{
768 netdev_dbg(vxlan->dev,
769 "delete %pM\n", f->eth_addr);
770
771 --vxlan->addrcnt;
9e4b93f9 772 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 773
774 hlist_del_rcu(&f->hlist);
6681712d 775 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 776}
777
f0b074be 778static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 779 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 780{
6681712d 781 struct net *net = dev_net(vxlan->dev);
e4c7ed41 782 int err;
d342894c 783
f0b074be 784 if (tb[NDA_DST]) {
e4c7ed41
CW		 785 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
786 if (err)
787 return err;
f0b074be 788 } else {
e4c7ed41
CW		 789 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
790 if (remote->sa.sa_family == AF_INET) {
791 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
792 ip->sa.sa_family = AF_INET;
793#if IS_ENABLED(CONFIG_IPV6)
794 } else {
795 ip->sin6.sin6_addr = in6addr_any;
796 ip->sa.sa_family = AF_INET6;
797#endif
798 }
f0b074be 799 }
d342894c 800
6681712d 801 if (tb[NDA_PORT]) {
73cf3317 802 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 803 return -EINVAL;
f0b074be
MR		 804 *port = nla_get_be16(tb[NDA_PORT]);
805 } else {
0dfbdf41 806 *port = vxlan->cfg.dst_port;
f0b074be 807 }
6681712d
DS		 808
809 if (tb[NDA_VNI]) {
810 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
811 return -EINVAL;
f0b074be
MR		 812 *vni = nla_get_u32(tb[NDA_VNI]);
813 } else {
814 *vni = vxlan->default_dst.remote_vni;
815 }
6681712d
DS		 816
817 if (tb[NDA_IFINDEX]) {
5abb0029 818 struct net_device *tdev;
6681712d
DS		 819
820 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
821 return -EINVAL;
f0b074be 822 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 823 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 824 if (!tdev)
6681712d 825 return -EADDRNOTAVAIL;
f0b074be
MR		 826 } else {
827 *ifindex = 0;
828 }
829
830 return 0;
831}
832
833/* Add static entry (via netlink) */
834static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
835 struct net_device *dev,
f6f6424b 836 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 837{
838 struct vxlan_dev *vxlan = netdev_priv(dev);
839 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 840 union vxlan_addr ip;
f0b074be
MR		 841 __be16 port;
842 u32 vni, ifindex;
843 int err;
844
845 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
846 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
847 ndm->ndm_state);
848 return -EINVAL;
849 }
850
851 if (tb[NDA_DST] == NULL)
852 return -EINVAL;
853
854 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
855 if (err)
856 return err;
6681712d 857
5933a7bb
MR		 858 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
859 return -EAFNOSUPPORT;
860
d342894c 861 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 862 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 863 port, vni, ifindex, ndm->ndm_flags);
d342894c 864 spin_unlock_bh(&vxlan->hash_lock);
865
866 return err;
867}
868
869/* Delete entry (via netlink) */
1690be63
VY		 870static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
871 struct net_device *dev,
f6f6424b 872 const unsigned char *addr, u16 vid)
d342894c 873{
874 struct vxlan_dev *vxlan = netdev_priv(dev);
875 struct vxlan_fdb *f;
bc7892ba 876 struct vxlan_rdst *rd = NULL;
e4c7ed41 877 union vxlan_addr ip;
bc7892ba
MR		 878 __be16 port;
879 u32 vni, ifindex;
880 int err;
881
882 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
883 if (err)
884 return err;
885
886 err = -ENOENT;
d342894c 887
888 spin_lock_bh(&vxlan->hash_lock);
889 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 890 if (!f)
891 goto out;
892
e4c7ed41
CW		 893 if (!vxlan_addr_any(&ip)) {
894 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 895 if (!rd)
896 goto out;
897 }
898
899 err = 0;
900
901 /* remove a destination if it's not the only one on the list,
902 * otherwise destroy the fdb entry
903 */
904 if (rd && !list_is_singular(&f->remotes)) {
905 list_del_rcu(&rd->list);
9e4b93f9 906 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 907 kfree_rcu(rd, rcu);
bc7892ba 908 goto out;
d342894c 909 }
bc7892ba
MR		 910
911 vxlan_fdb_destroy(vxlan, f);
912
913out:
d342894c 914 spin_unlock_bh(&vxlan->hash_lock);
915
916 return err;
917}
918
919/* Dump forwarding table */
920static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 921 struct net_device *dev,
922 struct net_device *filter_dev, int idx)
d342894c 923{
924 struct vxlan_dev *vxlan = netdev_priv(dev);
925 unsigned int h;
926
927 for (h = 0; h < FDB_HASH_SIZE; ++h) {
928 struct vxlan_fdb *f;
d342894c 929 int err;
930
b67bfe0d 931 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 932 struct vxlan_rdst *rd;
6681712d 933
3e61aa8f
SH		 934 if (idx < cb->args[0])
935 goto skip;
936
937 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 938 err = vxlan_fdb_info(skb, vxlan, f,
939 NETLINK_CB(cb->skb).portid,
940 cb->nlh->nlmsg_seq,
941 RTM_NEWNEIGH,
942 NLM_F_MULTI, rd);
943 if (err < 0)
3e61aa8f 944 goto out;
6681712d 945 }
3e61aa8f
SH		 946skip:
947 ++idx;
d342894c 948 }
949 }
3e61aa8f 950out:
d342894c 951 return idx;
952}
953
954/* Watch incoming packets to learn mapping between Ethernet address
955 * and Tunnel endpoint.
c4b49512 956 * Return true if packet is bogus and should be dropped.
d342894c 957 */
26a41ae6 958static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 959 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 960{
961 struct vxlan_dev *vxlan = netdev_priv(dev);
962 struct vxlan_fdb *f;
d342894c 963
964 f = vxlan_find_mac(vxlan, src_mac);
965 if (likely(f)) {
5ca5461c 966 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 967
e4c7ed41 968 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 969 return false;
970
971 /* Don't migrate static entries, drop packets */
eb064c3b 972 if (f->state & NUD_NOARP)
26a41ae6 973 return true;
d342894c 974
975 if (net_ratelimit())
976 netdev_info(dev,
e4c7ed41 977 "%pM migrated from %pIS to %pIS\n",
a4870f79 978 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 979
e4c7ed41 980 rdst->remote_ip = *src_ip;
d342894c 981 f->updated = jiffies;
9e4b93f9 982 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 983 } else {
984 /* learned new entry */
985 spin_lock(&vxlan->hash_lock);
3bf74b1a 986
987 /* close off race between vxlan_flush and incoming packets */
988 if (netif_running(dev))
989 vxlan_fdb_create(vxlan, src_mac, src_ip,
990 NUD_REACHABLE,
991 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 992 vxlan->cfg.dst_port,
3bf74b1a 993 vxlan->default_dst.remote_vni,
994 0, NTF_SELF);
d342894c 995 spin_unlock(&vxlan->hash_lock);
996 }
26a41ae6 997
998 return false;
d342894c 999}
1000
d342894c 1001/* See if multicast group is already in use by other ID */
95ab0991 1002static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1003{
553675fb 1004 struct vxlan_dev *vxlan;
d342894c 1005
95ab0991
G		 1006 /* The vxlan_sock is only used by dev, leaving group has
1007 * no effect on other vxlan devices.
1008 */
1009 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1010 return false;
1011
553675fb 1012 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1013 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1014 continue;
d342894c 1015
95ab0991
G		 1016 if (vxlan->vn_sock != dev->vn_sock)
1017 continue;
1018
1019 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1020 &dev->default_dst.remote_ip))
1021 continue;
1022
1023 if (vxlan->default_dst.remote_ifindex !=
1024 dev->default_dst.remote_ifindex)
1025 continue;
1026
1027 return true;
553675fb 1028 }
d342894c 1029
1030 return false;
1031}
1032
614732ea 1033static void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1034{
53cf5275
JG		 1035 struct sock *sk = vs->sock->sk;
1036 struct net *net = sock_net(sk);
1037 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1038
7c47cedf
SH		 1039 if (!atomic_dec_and_test(&vs->refcnt))
1040 return;
d342894c 1041
1c51a915 1042 spin_lock(&vn->sock_lock);
7c47cedf 1043 hlist_del_rcu(&vs->hlist);
dc01e7d3 1044 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1045 spin_unlock(&vn->sock_lock);
1046
7c47cedf 1047 queue_work(vxlan_wq, &vs->del_work);
d342894c 1048}
1049
56ef9c90 1050/* Update multicast group membership when first VNI on
c4b49512 1051 * multicast address is brought up
7c47cedf 1052 */
56ef9c90 1053static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1054{
7c47cedf
SH		 1055 struct vxlan_sock *vs = vxlan->vn_sock;
1056 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1057 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1058 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1059 int ret = -EINVAL;
d342894c 1060
d342894c 1061 lock_sock(sk);
e4c7ed41
CW		 1062 if (ip->sa.sa_family == AF_INET) {
1063 struct ip_mreqn mreq = {
1064 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1065 .imr_ifindex = ifindex,
1066 };
1067
56ef9c90 1068 ret = ip_mc_join_group(sk, &mreq);
e4c7ed41
CW		 1069#if IS_ENABLED(CONFIG_IPV6)
1070 } else {
56ef9c90
MRL		 1071 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1072 &ip->sin6.sin6_addr);
e4c7ed41
CW		 1073#endif
1074 }
3fc2de2f 1075 release_sock(sk);
1076
56ef9c90 1077 return ret;
3fc2de2f 1078}
1079
1080/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1081static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1082{
3fc2de2f 1083 struct vxlan_sock *vs = vxlan->vn_sock;
1084 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1085 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1086 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1087 int ret = -EINVAL;
3fc2de2f 1088
1089 lock_sock(sk);
e4c7ed41
CW		 1090 if (ip->sa.sa_family == AF_INET) {
1091 struct ip_mreqn mreq = {
1092 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1093 .imr_ifindex = ifindex,
1094 };
1095
56ef9c90 1096 ret = ip_mc_leave_group(sk, &mreq);
e4c7ed41
CW		 1097#if IS_ENABLED(CONFIG_IPV6)
1098 } else {
56ef9c90
MRL		 1099 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1100 &ip->sin6.sin6_addr);
e4c7ed41
CW		 1101#endif
1102 }
d342894c 1103 release_sock(sk);
d342894c 1104
56ef9c90 1105 return ret;
d342894c 1106}
1107
dfd8645e 1108static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
0ace2ca8 1109 size_t hdrlen, u32 data, bool nopartial)
dfd8645e
TH		 1110{
1111 size_t start, offset, plen;
dfd8645e 1112
dfd8645e
TH		 1113 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1114 offset = start + ((data & VXLAN_RCO_UDP) ?
1115 offsetof(struct udphdr, check) :
1116 offsetof(struct tcphdr, check));
1117
1118 plen = hdrlen + offset + sizeof(u16);
1119
1120 if (!pskb_may_pull(skb, plen))
1121 return NULL;
1122
1123 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1124
0ace2ca8
TH		 1125 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1126 nopartial);
dfd8645e
TH		 1127
1128 return vh;
1129}
1130
614732ea
TG		 1131static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1132 struct vxlan_metadata *md, u32 vni,
1133 struct metadata_dst *tun_dst)
1134{
1135 struct iphdr *oip = NULL;
1136 struct ipv6hdr *oip6 = NULL;
1137 struct vxlan_dev *vxlan;
1138 struct pcpu_sw_netstats *stats;
1139 union vxlan_addr saddr;
1140 int err = 0;
1141 union vxlan_addr *remote_ip;
1142
1143 /* For flow based devices, map all packets to VNI 0 */
1144 if (vs->flags & VXLAN_F_FLOW_BASED)
1145 vni = 0;
1146
1147 /* Is this VNI defined? */
1148 vxlan = vxlan_vs_find_vni(vs, vni);
1149 if (!vxlan)
1150 goto drop;
1151
1152 remote_ip = &vxlan->default_dst.remote_ip;
1153 skb_reset_mac_header(skb);
1154 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1155 skb->protocol = eth_type_trans(skb, vxlan->dev);
1156 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1157
1158 /* Ignore packet loops (and multicast echo) */
1159 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1160 goto drop;
1161
1162 /* Re-examine inner Ethernet packet */
1163 if (remote_ip->sa.sa_family == AF_INET) {
1164 oip = ip_hdr(skb);
1165 saddr.sin.sin_addr.s_addr = oip->saddr;
1166 saddr.sa.sa_family = AF_INET;
1167#if IS_ENABLED(CONFIG_IPV6)
1168 } else {
1169 oip6 = ipv6_hdr(skb);
1170 saddr.sin6.sin6_addr = oip6->saddr;
1171 saddr.sa.sa_family = AF_INET6;
1172#endif
1173 }
1174
1175 if (tun_dst) {
1176 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1177 tun_dst = NULL;
1178 }
1179
1180 if ((vxlan->flags & VXLAN_F_LEARN) &&
1181 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1182 goto drop;
1183
1184 skb_reset_network_header(skb);
1185 /* In flow-based mode, GBP is carried in dst_metadata */
1186 if (!(vs->flags & VXLAN_F_FLOW_BASED))
1187 skb->mark = md->gbp;
1188
1189 if (oip6)
1190 err = IP6_ECN_decapsulate(oip6, skb);
1191 if (oip)
1192 err = IP_ECN_decapsulate(oip, skb);
1193
1194 if (unlikely(err)) {
1195 if (log_ecn_error) {
1196 if (oip6)
1197 net_info_ratelimited("non-ECT from %pI6\n",
1198 &oip6->saddr);
1199 if (oip)
1200 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1201 &oip->saddr, oip->tos);
1202 }
1203 if (err > 1) {
1204 ++vxlan->dev->stats.rx_frame_errors;
1205 ++vxlan->dev->stats.rx_errors;
1206 goto drop;
1207 }
1208 }
1209
1210 stats = this_cpu_ptr(vxlan->dev->tstats);
1211 u64_stats_update_begin(&stats->syncp);
1212 stats->rx_packets++;
1213 stats->rx_bytes += skb->len;
1214 u64_stats_update_end(&stats->syncp);
1215
1216 netif_rx(skb);
1217
1218 return;
1219drop:
1220 if (tun_dst)
1221 dst_release((struct dst_entry *)tun_dst);
1222
1223 /* Consume bad packet */
1224 kfree_skb(skb);
1225}
1226
d342894c 1227/* Callback from net/ipv4/udp.c to receive packets */
1228static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1229{
ee122c79
TG		 1230 struct metadata_dst *tun_dst = NULL;
1231 struct ip_tunnel_info *info;
5cfccc5a 1232 struct vxlan_sock *vs;
d342894c 1233 struct vxlanhdr *vxh;
3bf39475 1234 u32 flags, vni;
ee122c79
TG		 1235 struct vxlan_metadata _md;
1236 struct vxlan_metadata *md = &_md;
d342894c 1237
d342894c 1238 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1239 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1240 goto error;
1241
7ce04758 1242 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1243 flags = ntohl(vxh->vx_flags);
1244 vni = ntohl(vxh->vx_vni);
1245
1246 if (flags & VXLAN_HF_VNI) {
1247 flags &= ~VXLAN_HF_VNI;
1248 } else {
1249 /* VNI flag always required to be set */
1250 goto bad_flags;
d342894c 1251 }
1252
5cfccc5a
PS		 1253 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1254 goto drop;
dfd8645e 1255 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1256
559835ea 1257 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1258 if (!vs)
d342894c 1259 goto drop;
d342894c 1260
dfd8645e 1261 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
0ace2ca8
TH		 1262 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1263 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 1264 if (!vxh)
1265 goto drop;
1266
1267 flags &= ~VXLAN_HF_RCO;
40fb70f3 1268 vni &= VXLAN_VNI_MASK;
dfd8645e
TH		 1269 }
1270
ee122c79
TG		 1271 if (vxlan_collect_metadata(vs)) {
1272 const struct iphdr *iph = ip_hdr(skb);
1273
1274 tun_dst = metadata_dst_alloc(sizeof(*md), GFP_ATOMIC);
1275 if (!tun_dst)
1276 goto drop;
1277
1278 info = &tun_dst->u.tun_info;
1279 info->key.ipv4_src = iph->saddr;
1280 info->key.ipv4_dst = iph->daddr;
1281 info->key.ipv4_tos = iph->tos;
1282 info->key.ipv4_ttl = iph->ttl;
1283 info->key.tp_src = udp_hdr(skb)->source;
1284 info->key.tp_dst = udp_hdr(skb)->dest;
1285
1286 info->mode = IP_TUNNEL_INFO_RX;
1287 info->key.tun_flags = TUNNEL_KEY;
1288 info->key.tun_id = cpu_to_be64(vni >> 8);
1289 if (udp_hdr(skb)->check != 0)
1290 info->key.tun_flags |= TUNNEL_CSUM;
1291
1292 md = ip_tunnel_info_opts(info, sizeof(*md));
ee122c79
TG		 1293 } else {
1294 memset(md, 0, sizeof(*md));
1295 }
1296
3511494c
TG		 1297 /* For backwards compatibility, only allow reserved fields to be
1298 * used by VXLAN extensions if explicitly requested.
1299 */
1300 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1301 struct vxlanhdr_gbp *gbp;
1302
1303 gbp = (struct vxlanhdr_gbp *)vxh;
ee122c79
TG		 1304 md->gbp = ntohs(gbp->policy_id);
1305
1306 if (tun_dst)
1307 info->key.tun_flags |= TUNNEL_VXLAN_OPT;
3511494c
TG		 1308
1309 if (gbp->dont_learn)
ee122c79 1310 md->gbp |= VXLAN_GBP_DONT_LEARN;
3511494c
TG		 1311
1312 if (gbp->policy_applied)
ee122c79 1313 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
3511494c
TG		 1314
1315 flags &= ~VXLAN_GBP_USED_BITS;
1316 }
1317
40fb70f3 1318 if (flags || vni & ~VXLAN_VNI_MASK) {
3bf39475
TH		 1319 /* If there are any unprocessed flags remaining treat
1320 * this as a malformed packet. This behavior diverges from
1321 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1322 * in reserved fields are to be ignored. The approach here
c4b49512 1323 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1324 * is more robust and provides a little more security in
1325 * adding extensions to VXLAN.
1326 */
1327
1328 goto bad_flags;
1329 }
1330
614732ea 1331 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
5cfccc5a
PS		 1332 return 0;
1333
1334drop:
1335 /* Consume bad packet */
1336 kfree_skb(skb);
1337 return 0;
1338
3bf39475
TH		 1339bad_flags:
1340 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1341 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1342
5cfccc5a 1343error:
ee122c79
TG		 1344 if (tun_dst)
1345 dst_release((struct dst_entry *)tun_dst);
1346
5cfccc5a
PS		 1347 /* Return non vxlan pkt */
1348 return 1;
1349}
1350
e4f67add
DS		 1351static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1352{
1353 struct vxlan_dev *vxlan = netdev_priv(dev);
1354 struct arphdr *parp;
1355 u8 *arpptr, *sha;
1356 __be32 sip, tip;
1357 struct neighbour *n;
1358
1359 if (dev->flags & IFF_NOARP)
1360 goto out;
1361
1362 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1363 dev->stats.tx_dropped++;
1364 goto out;
1365 }
1366 parp = arp_hdr(skb);
1367
1368 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1369 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1370 parp->ar_pro != htons(ETH_P_IP) ||
1371 parp->ar_op != htons(ARPOP_REQUEST) ||
1372 parp->ar_hln != dev->addr_len ||
1373 parp->ar_pln != 4)
1374 goto out;
1375 arpptr = (u8 *)parp + sizeof(struct arphdr);
1376 sha = arpptr;
1377 arpptr += dev->addr_len; /* sha */
1378 memcpy(&sip, arpptr, sizeof(sip));
1379 arpptr += sizeof(sip);
1380 arpptr += dev->addr_len; /* tha */
1381 memcpy(&tip, arpptr, sizeof(tip));
1382
1383 if (ipv4_is_loopback(tip) ||
1384 ipv4_is_multicast(tip))
1385 goto out;
1386
1387 n = neigh_lookup(&arp_tbl, &tip, dev);
1388
1389 if (n) {
e4f67add
DS		 1390 struct vxlan_fdb *f;
1391 struct sk_buff *reply;
1392
1393 if (!(n->nud_state & NUD_CONNECTED)) {
1394 neigh_release(n);
1395 goto out;
1396 }
1397
1398 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1399 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1400 /* bridge-local neighbor */
1401 neigh_release(n);
1402 goto out;
1403 }
1404
1405 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1406 n->ha, sha);
1407
1408 neigh_release(n);
1409
7346135d
DS		 1410 if (reply == NULL)
1411 goto out;
1412
e4f67add
DS		 1413 skb_reset_mac_header(reply);
1414 __skb_pull(reply, skb_network_offset(reply));
1415 reply->ip_summed = CHECKSUM_UNNECESSARY;
1416 reply->pkt_type = PACKET_HOST;
1417
1418 if (netif_rx_ni(reply) == NET_RX_DROP)
1419 dev->stats.rx_dropped++;
e4c7ed41
CW		 1420 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1421 union vxlan_addr ipa = {
1422 .sin.sin_addr.s_addr = tip,
a45e92a5 1423 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1424 };
1425
1426 vxlan_ip_miss(dev, &ipa);
1427 }
e4f67add
DS		 1428out:
1429 consume_skb(skb);
1430 return NETDEV_TX_OK;
1431}
1432
f564f45c 1433#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1434static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1435 struct neighbour *n, bool isrouter)
1436{
1437 struct net_device *dev = request->dev;
1438 struct sk_buff *reply;
1439 struct nd_msg *ns, *na;
1440 struct ipv6hdr *pip6;
1441 u8 *daddr;
1442 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1443 int ns_olen;
1444 int i, len;
1445
1446 if (dev == NULL)
1447 return NULL;
1448
1449 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1450 sizeof(*na) + na_olen + dev->needed_tailroom;
1451 reply = alloc_skb(len, GFP_ATOMIC);
1452 if (reply == NULL)
1453 return NULL;
1454
1455 reply->protocol = htons(ETH_P_IPV6);
1456 reply->dev = dev;
1457 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1458 skb_push(reply, sizeof(struct ethhdr));
1459 skb_set_mac_header(reply, 0);
1460
1461 ns = (struct nd_msg *)skb_transport_header(request);
1462
1463 daddr = eth_hdr(request)->h_source;
1464 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1465 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1466 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1467 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1468 break;
1469 }
1470 }
1471
1472 /* Ethernet header */
1473 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1474 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1475 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1476 reply->protocol = htons(ETH_P_IPV6);
1477
1478 skb_pull(reply, sizeof(struct ethhdr));
1479 skb_set_network_header(reply, 0);
1480 skb_put(reply, sizeof(struct ipv6hdr));
1481
1482 /* IPv6 header */
1483
1484 pip6 = ipv6_hdr(reply);
1485 memset(pip6, 0, sizeof(struct ipv6hdr));
1486 pip6->version = 6;
1487 pip6->priority = ipv6_hdr(request)->priority;
1488 pip6->nexthdr = IPPROTO_ICMPV6;
1489 pip6->hop_limit = 255;
1490 pip6->daddr = ipv6_hdr(request)->saddr;
1491 pip6->saddr = *(struct in6_addr *)n->primary_key;
1492
1493 skb_pull(reply, sizeof(struct ipv6hdr));
1494 skb_set_transport_header(reply, 0);
1495
1496 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1497
1498 /* Neighbor Advertisement */
1499 memset(na, 0, sizeof(*na)+na_olen);
1500 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1501 na->icmph.icmp6_router = isrouter;
1502 na->icmph.icmp6_override = 1;
1503 na->icmph.icmp6_solicited = 1;
1504 na->target = ns->target;
1505 ether_addr_copy(&na->opt[2], n->ha);
1506 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1507 na->opt[1] = na_olen >> 3;
1508
1509 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1510 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1511 csum_partial(na, sizeof(*na)+na_olen, 0));
1512
1513 pip6->payload_len = htons(sizeof(*na)+na_olen);
1514
1515 skb_push(reply, sizeof(struct ipv6hdr));
1516
1517 reply->ip_summed = CHECKSUM_UNNECESSARY;
1518
1519 return reply;
1520}
1521
f564f45c
CW		 1522static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1523{
1524 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1525 struct nd_msg *msg;
f564f45c
CW		 1526 const struct ipv6hdr *iphdr;
1527 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1528 struct neighbour *n;
1529 struct inet6_dev *in6_dev;
f564f45c
CW		 1530
1531 in6_dev = __in6_dev_get(dev);
1532 if (!in6_dev)
1533 goto out;
1534
f564f45c
CW		 1535 iphdr = ipv6_hdr(skb);
1536 saddr = &iphdr->saddr;
1537 daddr = &iphdr->daddr;
1538
f564f45c
CW		 1539 msg = (struct nd_msg *)skb_transport_header(skb);
1540 if (msg->icmph.icmp6_code != 0 ||
1541 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1542 goto out;
1543
4b29dba9
DS		 1544 if (ipv6_addr_loopback(daddr) ||
1545 ipv6_addr_is_multicast(&msg->target))
1546 goto out;
1547
1548 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1549
1550 if (n) {
1551 struct vxlan_fdb *f;
4b29dba9 1552 struct sk_buff *reply;
f564f45c
CW		 1553
1554 if (!(n->nud_state & NUD_CONNECTED)) {
1555 neigh_release(n);
1556 goto out;
1557 }
1558
1559 f = vxlan_find_mac(vxlan, n->ha);
1560 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1561 /* bridge-local neighbor */
1562 neigh_release(n);
1563 goto out;
1564 }
1565
4b29dba9
DS		 1566 reply = vxlan_na_create(skb, n,
1567 !!(f ? f->flags & NTF_ROUTER : 0));
1568
f564f45c 1569 neigh_release(n);
4b29dba9
DS		 1570
1571 if (reply == NULL)
1572 goto out;
1573
1574 if (netif_rx_ni(reply) == NET_RX_DROP)
1575 dev->stats.rx_dropped++;
1576
f564f45c 1577 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1578 union vxlan_addr ipa = {
1579 .sin6.sin6_addr = msg->target,
a45e92a5 1580 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1581 };
1582
f564f45c
CW		 1583 vxlan_ip_miss(dev, &ipa);
1584 }
1585
1586out:
1587 consume_skb(skb);
1588 return NETDEV_TX_OK;
1589}
1590#endif
1591
e4f67add
DS		 1592static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1593{
1594 struct vxlan_dev *vxlan = netdev_priv(dev);
1595 struct neighbour *n;
e4f67add
DS		 1596
1597 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1598 return false;
1599
1600 n = NULL;
1601 switch (ntohs(eth_hdr(skb)->h_proto)) {
1602 case ETH_P_IP:
e15a00aa
CW		 1603 {
1604 struct iphdr *pip;
1605
e4f67add
DS		 1606 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1607 return false;
1608 pip = ip_hdr(skb);
1609 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1610 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1611 union vxlan_addr ipa = {
1612 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1613 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1614 };
1615
1616 vxlan_ip_miss(dev, &ipa);
1617 return false;
1618 }
1619
e4f67add 1620 break;
e15a00aa
CW		 1621 }
1622#if IS_ENABLED(CONFIG_IPV6)
1623 case ETH_P_IPV6:
1624 {
1625 struct ipv6hdr *pip6;
1626
1627 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1628 return false;
1629 pip6 = ipv6_hdr(skb);
1630 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1631 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1632 union vxlan_addr ipa = {
1633 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1634 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1635 };
1636
1637 vxlan_ip_miss(dev, &ipa);
1638 return false;
1639 }
1640
1641 break;
1642 }
1643#endif
e4f67add
DS		 1644 default:
1645 return false;
1646 }
1647
1648 if (n) {
1649 bool diff;
1650
7367d0b5 1651 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1652 if (diff) {
1653 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1654 dev->addr_len);
1655 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1656 }
1657 neigh_release(n);
1658 return diff;
e4c7ed41
CW		 1659 }
1660
e4f67add
DS		 1661 return false;
1662}
1663
af33c1ad 1664static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1665 struct vxlan_metadata *md)
1666{
1667 struct vxlanhdr_gbp *gbp;
1668
db79a621
TG		 1669 if (!md->gbp)
1670 return;
1671
3511494c
TG		 1672 gbp = (struct vxlanhdr_gbp *)vxh;
1673 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1674
1675 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1676 gbp->dont_learn = 1;
1677
1678 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1679 gbp->policy_applied = 1;
1680
1681 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1682}
1683
e4c7ed41 1684#if IS_ENABLED(CONFIG_IPV6)
79b16aad
DM		 1685static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1686 struct sk_buff *skb,
e4c7ed41
CW		 1687 struct net_device *dev, struct in6_addr *saddr,
1688 struct in6_addr *daddr, __u8 prio, __u8 ttl,
6b6948dd 1689 __be16 src_port, __be16 dst_port, __be32 vni,
af33c1ad 1690 struct vxlan_metadata *md, bool xnet, u32 vxflags)
e4c7ed41 1691{
e4c7ed41 1692 struct vxlanhdr *vxh;
e4c7ed41
CW		 1693 int min_headroom;
1694 int err;
af33c1ad 1695 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
dfd8645e
TH		 1696 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1697 u16 hdrlen = sizeof(struct vxlanhdr);
1698
af33c1ad 1699 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1700 skb->ip_summed == CHECKSUM_PARTIAL) {
1701 int csum_start = skb_checksum_start_offset(skb);
1702
1703 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1704 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1705 (skb->csum_offset == offsetof(struct udphdr, check) ||
1706 skb->csum_offset == offsetof(struct tcphdr, check))) {
1707 udp_sum = false;
1708 type |= SKB_GSO_TUNNEL_REMCSUM;
1709 }
1710 }
e4c7ed41 1711
f01ec1c0 1712 skb_scrub_packet(skb, xnet);
963a88b3 1713
e4c7ed41
CW		 1714 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1715 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1716 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1717
1718 /* Need space for new headers (invalidates iph ptr) */
1719 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1720 if (unlikely(err)) {
1721 kfree_skb(skb);
1722 goto err;
1723 }
e4c7ed41 1724
5968250c 1725 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1726 if (WARN_ON(!skb)) {
1727 err = -ENOMEM;
1728 goto err;
1729 }
e4c7ed41 1730
b736a623
JG		 1731 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1732 if (IS_ERR(skb)) {
1733 err = -EINVAL;
1734 goto err;
1735 }
1736
e4c7ed41 1737 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1738 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1739 vxh->vx_vni = vni;
e4c7ed41 1740
dfd8645e
TH		 1741 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1742 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1743 VXLAN_RCO_SHIFT;
1744
1745 if (skb->csum_offset == offsetof(struct udphdr, check))
1746 data |= VXLAN_RCO_UDP;
1747
1748 vxh->vx_vni |= htonl(data);
1749 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1750
1751 if (!skb_is_gso(skb)) {
1752 skb->ip_summed = CHECKSUM_NONE;
1753 skb->encapsulation = 0;
1754 }
1755 }
1756
af33c1ad
TH		 1757 if (vxflags & VXLAN_F_GBP)
1758 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1759
996c9fd1
TH		 1760 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1761
79b16aad 1762 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
d998f8ef 1763 ttl, src_port, dst_port,
af33c1ad 1764 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
e4c7ed41 1765 return 0;
74f47278
PS		 1766err:
1767 dst_release(dst);
1768 return err;
e4c7ed41
CW		 1769}
1770#endif
1771
614732ea
TG		 1772static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1773 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
6b6948dd 1774 __be16 src_port, __be16 dst_port, __be32 vni,
614732ea 1775 struct vxlan_metadata *md, bool xnet, u32 vxflags)
49560532
PS		 1776{
1777 struct vxlanhdr *vxh;
649c5b8b 1778 int min_headroom;
49560532 1779 int err;
af33c1ad 1780 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
dfd8645e
TH		 1781 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1782 u16 hdrlen = sizeof(struct vxlanhdr);
1783
af33c1ad 1784 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1785 skb->ip_summed == CHECKSUM_PARTIAL) {
1786 int csum_start = skb_checksum_start_offset(skb);
1787
1788 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1789 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1790 (skb->csum_offset == offsetof(struct udphdr, check) ||
1791 skb->csum_offset == offsetof(struct tcphdr, check))) {
1792 udp_sum = false;
1793 type |= SKB_GSO_TUNNEL_REMCSUM;
1794 }
1795 }
49560532 1796
649c5b8b 1797 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1798 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1799 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1800
1801 /* Need space for new headers (invalidates iph ptr) */
1802 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1803 if (unlikely(err)) {
1804 kfree_skb(skb);
649c5b8b 1805 return err;
74f47278 1806 }
649c5b8b 1807
5968250c
JP		 1808 skb = vlan_hwaccel_push_inside(skb);
1809 if (WARN_ON(!skb))
1810 return -ENOMEM;
1eaa8178 1811
b736a623
JG		 1812 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1813 if (IS_ERR(skb))
1814 return PTR_ERR(skb);
1815
49560532 1816 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1817 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1818 vxh->vx_vni = vni;
49560532 1819
dfd8645e
TH		 1820 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1821 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1822 VXLAN_RCO_SHIFT;
1823
1824 if (skb->csum_offset == offsetof(struct udphdr, check))
1825 data |= VXLAN_RCO_UDP;
1826
1827 vxh->vx_vni |= htonl(data);
1828 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1829
1830 if (!skb_is_gso(skb)) {
1831 skb->ip_summed = CHECKSUM_NONE;
1832 skb->encapsulation = 0;
1833 }
1834 }
1835
af33c1ad
TH		 1836 if (vxflags & VXLAN_F_GBP)
1837 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1838
996c9fd1
TH		 1839 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1840
79b16aad 1841 return udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos,
d998f8ef 1842 ttl, df, src_port, dst_port, xnet,
af33c1ad 1843 !(vxflags & VXLAN_F_UDP_CSUM));
49560532 1844}
49560532 1845
9dcc71e1
SS		 1846/* Bypass encapsulation if the destination is local */
1847static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1848 struct vxlan_dev *dst_vxlan)
1849{
8f84985f 1850 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1851 union vxlan_addr loopback;
1852 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1853 struct net_device *dev = skb->dev;
1854 int len = skb->len;
9dcc71e1 1855
8f84985f
LR		 1856 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1857 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1858 skb->pkt_type = PACKET_HOST;
1859 skb->encapsulation = 0;
1860 skb->dev = dst_vxlan->dev;
1861 __skb_pull(skb, skb_network_offset(skb));
1862
e4c7ed41
CW		 1863 if (remote_ip->sa.sa_family == AF_INET) {
1864 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1865 loopback.sa.sa_family = AF_INET;
1866#if IS_ENABLED(CONFIG_IPV6)
1867 } else {
1868 loopback.sin6.sin6_addr = in6addr_loopback;
1869 loopback.sa.sa_family = AF_INET6;
1870#endif
1871 }
1872
9dcc71e1 1873 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1874 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1875
1876 u64_stats_update_begin(&tx_stats->syncp);
1877 tx_stats->tx_packets++;
ce6502a8 1878 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1879 u64_stats_update_end(&tx_stats->syncp);
1880
1881 if (netif_rx(skb) == NET_RX_SUCCESS) {
1882 u64_stats_update_begin(&rx_stats->syncp);
1883 rx_stats->rx_packets++;
ce6502a8 1884 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1885 u64_stats_update_end(&rx_stats->syncp);
1886 } else {
ce6502a8 1887 dev->stats.rx_dropped++;
9dcc71e1
SS		 1888 }
1889}
1890
4ad16930
SH		 1891static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1892 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1893{
3093fbe7 1894 struct ip_tunnel_info *info;
d342894c 1895 struct vxlan_dev *vxlan = netdev_priv(dev);
79b16aad 1896 struct sock *sk = vxlan->vn_sock->sock->sk;
e4c7ed41 1897 struct rtable *rt = NULL;
d342894c 1898 const struct iphdr *old_iph;
d342894c 1899 struct flowi4 fl4;
e4c7ed41 1900 union vxlan_addr *dst;
ee122c79
TG		 1901 union vxlan_addr remote_ip;
1902 struct vxlan_metadata _md;
1903 struct vxlan_metadata *md = &_md;
e4c7ed41 1904 __be16 src_port = 0, dst_port;
234f5b73 1905 u32 vni;
d342894c 1906 __be16 df = 0;
1907 __u8 tos, ttl;
0e6fbc5b 1908 int err;
ee122c79 1909 u32 flags = vxlan->flags;
d342894c 1910
3093fbe7
TG		 1911 /* FIXME: Support IPv6 */
1912 info = skb_tunnel_info(skb, AF_INET);
1913
ee122c79 1914 if (rdst) {
0dfbdf41 1915 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1916 vni = rdst->remote_vni;
1917 dst = &rdst->remote_ip;
1918 } else {
1919 if (!info) {
1920 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1921 dev->name);
1922 goto drop;
1923 }
1924
0dfbdf41 1925 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
ee122c79
TG		 1926 vni = be64_to_cpu(info->key.tun_id);
1927 remote_ip.sin.sin_family = AF_INET;
1928 remote_ip.sin.sin_addr.s_addr = info->key.ipv4_dst;
1929 dst = &remote_ip;
1930 }
e4f67add 1931
e4c7ed41 1932 if (vxlan_addr_any(dst)) {
e4f67add 1933 if (did_rsc) {
e4f67add 1934 /* short-circuited back to local bridge */
9dcc71e1 1935 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1936 return;
e4f67add 1937 }
ef59febe 1938 goto drop;
e4f67add 1939 }
ef59febe 1940
d342894c 1941 old_iph = ip_hdr(skb);
1942
0dfbdf41 1943 ttl = vxlan->cfg.ttl;
e4c7ed41 1944 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1945 ttl = 1;
1946
0dfbdf41 1947 tos = vxlan->cfg.tos;
d342894c 1948 if (tos == 1)
206aaafc 1949 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1950
0dfbdf41
TG		 1951 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1952 vxlan->cfg.port_max, true);
d342894c 1953
e4c7ed41 1954 if (dst->sa.sa_family == AF_INET) {
ee122c79
TG		 1955 if (info) {
1956 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1957 df = htons(IP_DF);
1958 if (info->key.tun_flags & TUNNEL_CSUM)
1959 flags |= VXLAN_F_UDP_CSUM;
1960 else
1961 flags &= ~VXLAN_F_UDP_CSUM;
1962
1963 ttl = info->key.ipv4_ttl;
1964 tos = info->key.ipv4_tos;
1965
1966 if (info->options_len)
1967 md = ip_tunnel_info_opts(info, sizeof(*md));
1968 } else {
1969 md->gbp = skb->mark;
1970 }
1971
e4c7ed41 1972 memset(&fl4, 0, sizeof(fl4));
ee122c79 1973 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
e4c7ed41 1974 fl4.flowi4_tos = RT_TOS(tos);
239fb791
TG		 1975 fl4.flowi4_mark = skb->mark;
1976 fl4.flowi4_proto = IPPROTO_UDP;
e4c7ed41 1977 fl4.daddr = dst->sin.sin_addr.s_addr;
0dfbdf41 1978 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
e4c7ed41 1979
f01ec1c0 1980 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1981 if (IS_ERR(rt)) {
1982 netdev_dbg(dev, "no route to %pI4\n",
1983 &dst->sin.sin_addr.s_addr);
1984 dev->stats.tx_carrier_errors++;
1985 goto tx_error;
1986 }
d342894c 1987
e4c7ed41
CW		 1988 if (rt->dst.dev == dev) {
1989 netdev_dbg(dev, "circular route to %pI4\n",
1990 &dst->sin.sin_addr.s_addr);
1991 dev->stats.collisions++;
fffc15a5 1992 goto rt_tx_error;
e4c7ed41
CW		 1993 }
1994
1995 /* Bypass encapsulation if the destination is local */
1996 if (rt->rt_flags & RTCF_LOCAL &&
1997 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1998 struct vxlan_dev *dst_vxlan;
1999
2000 ip_rt_put(rt);
19ca9fc1 2001 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2002 dst->sa.sa_family, dst_port,
2003 vxlan->flags);
e4c7ed41
CW		 2004 if (!dst_vxlan)
2005 goto tx_error;
2006 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2007 return;
2008 }
2009
2010 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2011 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
79b16aad 2012 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
af33c1ad 2013 dst->sin.sin_addr.s_addr, tos, ttl, df,
614732ea 2014 src_port, dst_port, htonl(vni << 8), md,
af33c1ad 2015 !net_eq(vxlan->net, dev_net(vxlan->dev)),
ee122c79 2016 flags);
74f47278
PS		 2017 if (err < 0) {
2018 /* skb is already freed. */
2019 skb = NULL;
e4c7ed41 2020 goto rt_tx_error;
74f47278
PS		 2021 }
2022
e4c7ed41
CW		 2023 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
2024#if IS_ENABLED(CONFIG_IPV6)
2025 } else {
e4c7ed41
CW		 2026 struct dst_entry *ndst;
2027 struct flowi6 fl6;
2028 u32 flags;
2029
2030 memset(&fl6, 0, sizeof(fl6));
ee122c79 2031 fl6.flowi6_oif = rdst ? rdst->remote_ifindex : 0;
e4c7ed41 2032 fl6.daddr = dst->sin6.sin6_addr;
0dfbdf41 2033 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
239fb791 2034 fl6.flowi6_mark = skb->mark;
8c1bb79f 2035 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 2036
2037 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
2038 netdev_dbg(dev, "no route to %pI6\n",
2039 &dst->sin6.sin6_addr);
2040 dev->stats.tx_carrier_errors++;
9dcc71e1 2041 goto tx_error;
e4c7ed41 2042 }
d342894c 2043
e4c7ed41
CW		 2044 if (ndst->dev == dev) {
2045 netdev_dbg(dev, "circular route to %pI6\n",
2046 &dst->sin6.sin6_addr);
2047 dst_release(ndst);
2048 dev->stats.collisions++;
2049 goto tx_error;
2050 }
0e6fbc5b 2051
e4c7ed41
CW		 2052 /* Bypass encapsulation if the destination is local */
2053 flags = ((struct rt6_info *)ndst)->rt6i_flags;
2054 if (flags & RTF_LOCAL &&
2055 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2056 struct vxlan_dev *dst_vxlan;
2057
2058 dst_release(ndst);
19ca9fc1 2059 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2060 dst->sa.sa_family, dst_port,
2061 vxlan->flags);
e4c7ed41
CW		 2062 if (!dst_vxlan)
2063 goto tx_error;
2064 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2065 return;
2066 }
49560532 2067
e4c7ed41 2068 ttl = ttl ? : ip6_dst_hoplimit(ndst);
ee122c79 2069 md->gbp = skb->mark;
e4c7ed41 2070
79b16aad 2071 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &fl6.saddr, &fl6.daddr,
614732ea 2072 0, ttl, src_port, dst_port, htonl(vni << 8), md,
af33c1ad
TH		 2073 !net_eq(vxlan->net, dev_net(vxlan->dev)),
2074 vxlan->flags);
e4c7ed41
CW		 2075#endif
2076 }
0e6fbc5b 2077
4ad16930 2078 return;
d342894c 2079
2080drop:
2081 dev->stats.tx_dropped++;
2082 goto tx_free;
2083
49560532
PS		 2084rt_tx_error:
2085 ip_rt_put(rt);
d342894c 2086tx_error:
2087 dev->stats.tx_errors++;
2088tx_free:
2089 dev_kfree_skb(skb);
d342894c 2090}
2091
6681712d
DS		 2092/* Transmit local packets over Vxlan
2093 *
2094 * Outer IP header inherits ECN and DF from inner header.
2095 * Outer UDP destination is the VXLAN assigned port.
2096 * source port is based on hash of flow
2097 */
2098static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2099{
2100 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2101 const struct ip_tunnel_info *info;
6681712d
DS		 2102 struct ethhdr *eth;
2103 bool did_rsc = false;
8f646c92 2104 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2105 struct vxlan_fdb *f;
6681712d 2106
3093fbe7
TG		 2107 /* FIXME: Support IPv6 */
2108 info = skb_tunnel_info(skb, AF_INET);
2109
6681712d
DS		 2110 skb_reset_mac_header(skb);
2111 eth = eth_hdr(skb);
2112
f564f45c
CW		 2113 if ((vxlan->flags & VXLAN_F_PROXY)) {
2114 if (ntohs(eth->h_proto) == ETH_P_ARP)
2115 return arp_reduce(dev, skb);
2116#if IS_ENABLED(CONFIG_IPV6)
2117 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2118 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2119 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2120 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2121 struct nd_msg *msg;
2122
2123 msg = (struct nd_msg *)skb_transport_header(skb);
2124 if (msg->icmph.icmp6_code == 0 &&
2125 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2126 return neigh_reduce(dev, skb);
2127 }
7a9f526f 2128 eth = eth_hdr(skb);
f564f45c
CW		 2129#endif
2130 }
6681712d 2131
ee122c79
TG		 2132 if (vxlan->flags & VXLAN_F_FLOW_BASED &&
2133 info && info->mode == IP_TUNNEL_INFO_TX) {
2134 vxlan_xmit_one(skb, dev, NULL, false);
2135 return NETDEV_TX_OK;
2136 }
2137
6681712d 2138 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2139 did_rsc = false;
2140
2141 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2142 (ntohs(eth->h_proto) == ETH_P_IP ||
2143 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2144 did_rsc = route_shortcircuit(dev, skb);
2145 if (did_rsc)
2146 f = vxlan_find_mac(vxlan, eth->h_dest);
2147 }
2148
6681712d 2149 if (f == NULL) {
afbd8bae
MR		 2150 f = vxlan_find_mac(vxlan, all_zeros_mac);
2151 if (f == NULL) {
2152 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2153 !is_multicast_ether_addr(eth->h_dest))
2154 vxlan_fdb_miss(vxlan, eth->h_dest);
2155
2156 dev->stats.tx_dropped++;
8f646c92 2157 kfree_skb(skb);
afbd8bae
MR		 2158 return NETDEV_TX_OK;
2159 }
2160 }
6681712d 2161
afbd8bae
MR		 2162 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2163 struct sk_buff *skb1;
6681712d 2164
8f646c92
ED		 2165 if (!fdst) {
2166 fdst = rdst;
2167 continue;
2168 }
afbd8bae
MR		 2169 skb1 = skb_clone(skb, GFP_ATOMIC);
2170 if (skb1)
2171 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2172 }
2173
8f646c92
ED		 2174 if (fdst)
2175 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2176 else
2177 kfree_skb(skb);
4ad16930 2178 return NETDEV_TX_OK;
6681712d
DS		 2179}
2180
d342894c 2181/* Walk the forwarding table and purge stale entries */
2182static void vxlan_cleanup(unsigned long arg)
2183{
2184 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2185 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2186 unsigned int h;
2187
2188 if (!netif_running(vxlan->dev))
2189 return;
2190
d342894c 2191 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2192 struct hlist_node *p, *n;
14e1d0fa
SD		 2193
2194 spin_lock_bh(&vxlan->hash_lock);
d342894c 2195 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2196 struct vxlan_fdb *f
2197 = container_of(p, struct vxlan_fdb, hlist);
2198 unsigned long timeout;
2199
3c172868 2200 if (f->state & NUD_PERMANENT)
d342894c 2201 continue;
2202
0dfbdf41 2203 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2204 if (time_before_eq(timeout, jiffies)) {
2205 netdev_dbg(vxlan->dev,
2206 "garbage collect %pM\n",
2207 f->eth_addr);
2208 f->state = NUD_STALE;
2209 vxlan_fdb_destroy(vxlan, f);
2210 } else if (time_before(timeout, next_timer))
2211 next_timer = timeout;
2212 }
14e1d0fa 2213 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2214 }
d342894c 2215
2216 mod_timer(&vxlan->age_timer, next_timer);
2217}
2218
9c2e24e1
PS		 2219static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2220{
56ef9c90 2221 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
9c2e24e1
PS		 2222 __u32 vni = vxlan->default_dst.remote_vni;
2223
2224 vxlan->vn_sock = vs;
56ef9c90 2225 spin_lock(&vn->sock_lock);
9c2e24e1 2226 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2227 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2228}
2229
d342894c 2230/* Setup stats when device is created */
2231static int vxlan_init(struct net_device *dev)
2232{
1c213bd2 2233 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2234 if (!dev->tstats)
d342894c 2235 return -ENOMEM;
2236
2237 return 0;
2238}
2239
ba609e9b 2240static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2241{
2242 struct vxlan_fdb *f;
2243
2244 spin_lock_bh(&vxlan->hash_lock);
2245 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2246 if (f)
2247 vxlan_fdb_destroy(vxlan, f);
2248 spin_unlock_bh(&vxlan->hash_lock);
2249}
2250
ebf4063e
SH		 2251static void vxlan_uninit(struct net_device *dev)
2252{
2253 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2254
ba609e9b 2255 vxlan_fdb_delete_default(vxlan);
afbd8bae 2256
ebf4063e
SH		 2257 free_percpu(dev->tstats);
2258}
2259
d342894c 2260/* Start ageing timer and join group when device is brought up */
2261static int vxlan_open(struct net_device *dev)
2262{
2263 struct vxlan_dev *vxlan = netdev_priv(dev);
56ef9c90
MRL		 2264 struct vxlan_sock *vs;
2265 int ret = 0;
1c51a915 2266
614732ea
TG		 2267 vs = vxlan_sock_add(vxlan->net, vxlan->cfg.dst_port,
2268 vxlan->cfg.no_share, vxlan->flags);
56ef9c90
MRL		 2269 if (IS_ERR(vs))
2270 return PTR_ERR(vs);
2271
2272 vxlan_vs_add_dev(vs, vxlan);
d342894c 2273
79d4a94f 2274 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90
MRL		 2275 ret = vxlan_igmp_join(vxlan);
2276 if (ret) {
2277 vxlan_sock_release(vs);
2278 return ret;
2279 }
d342894c 2280 }
2281
0dfbdf41 2282 if (vxlan->cfg.age_interval)
d342894c 2283 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2284
56ef9c90 2285 return ret;
d342894c 2286}
2287
2288/* Purge the forwarding table */
2289static void vxlan_flush(struct vxlan_dev *vxlan)
2290{
31fec5aa 2291 unsigned int h;
d342894c 2292
2293 spin_lock_bh(&vxlan->hash_lock);
2294 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2295 struct hlist_node *p, *n;
2296 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2297 struct vxlan_fdb *f
2298 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2299 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2300 if (!is_zero_ether_addr(f->eth_addr))
2301 vxlan_fdb_destroy(vxlan, f);
d342894c 2302 }
2303 }
2304 spin_unlock_bh(&vxlan->hash_lock);
2305}
2306
2307/* Cleanup timer and forwarding table on shutdown */
2308static int vxlan_stop(struct net_device *dev)
2309{
2310 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2311 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2312 struct vxlan_sock *vs = vxlan->vn_sock;
56ef9c90 2313 int ret = 0;
d342894c 2314
24c0e683 2315 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2316 !vxlan_group_used(vn, vxlan))
56ef9c90 2317 ret = vxlan_igmp_leave(vxlan);
d342894c 2318
2319 del_timer_sync(&vxlan->age_timer);
2320
2321 vxlan_flush(vxlan);
56ef9c90 2322 vxlan_sock_release(vs);
d342894c 2323
56ef9c90 2324 return ret;
d342894c 2325}
2326
d342894c 2327/* Stub, nothing needs to be done. */
2328static void vxlan_set_multicast_list(struct net_device *dev)
2329{
2330}
2331
345010b5
DB		 2332static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2333{
2334 struct vxlan_dev *vxlan = netdev_priv(dev);
2335 struct vxlan_rdst *dst = &vxlan->default_dst;
2336 struct net_device *lowerdev;
2337 int max_mtu;
2338
f01ec1c0 2339 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2340 if (lowerdev == NULL)
2341 return eth_change_mtu(dev, new_mtu);
2342
2343 if (dst->remote_ip.sa.sa_family == AF_INET6)
2344 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2345 else
2346 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2347
2348 if (new_mtu < 68 || new_mtu > max_mtu)
2349 return -EINVAL;
2350
2351 dev->mtu = new_mtu;
2352 return 0;
2353}
2354
d342894c 2355static const struct net_device_ops vxlan_netdev_ops = {
2356 .ndo_init = vxlan_init,
ebf4063e 2357 .ndo_uninit = vxlan_uninit,
d342894c 2358 .ndo_open = vxlan_open,
2359 .ndo_stop = vxlan_stop,
2360 .ndo_start_xmit = vxlan_xmit,
e8171045 2361 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2362 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2363 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2364 .ndo_validate_addr = eth_validate_addr,
2365 .ndo_set_mac_address = eth_mac_addr,
2366 .ndo_fdb_add = vxlan_fdb_add,
2367 .ndo_fdb_del = vxlan_fdb_delete,
2368 .ndo_fdb_dump = vxlan_fdb_dump,
2369};
2370
2371/* Info for udev, that this is a virtual tunnel endpoint */
2372static struct device_type vxlan_type = {
2373 .name = "vxlan",
2374};
2375
53cf5275 2376/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2377 * supply the listening VXLAN udp ports. Callers are expected
2378 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2379 */
2380void vxlan_get_rx_port(struct net_device *dev)
2381{
2382 struct vxlan_sock *vs;
2383 struct net *net = dev_net(dev);
2384 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2385 sa_family_t sa_family;
35e42379
JG		 2386 __be16 port;
2387 unsigned int i;
53cf5275
JG		 2388
2389 spin_lock(&vn->sock_lock);
2390 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2391 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2392 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2393 sa_family = vs->sock->sk->sk_family;
2394 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2395 port);
2396 }
2397 }
2398 spin_unlock(&vn->sock_lock);
2399}
2400EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2401
d342894c 2402/* Initialize the device structure. */
2403static void vxlan_setup(struct net_device *dev)
2404{
2405 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2406 unsigned int h;
d342894c 2407
2408 eth_hw_addr_random(dev);
2409 ether_setup(dev);
e4c7ed41 2410 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2411 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2412 else
2853af6a 2413 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2414
2415 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2416 dev->destructor = free_netdev;
d342894c 2417 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2418
2419 dev->tx_queue_len = 0;
2420 dev->features |= NETIF_F_LLTX;
d6727fe3 2421 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2422 dev->features |= NETIF_F_RXCSUM;
05c0db08 2423 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2424
1eaa8178
PS		 2425 dev->vlan_features = dev->features;
2426 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2427 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2428 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2429 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2430 netif_keep_dst(dev);
6602d007 2431 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2432
553675fb 2433 INIT_LIST_HEAD(&vxlan->next);
d342894c 2434 spin_lock_init(&vxlan->hash_lock);
2435
2436 init_timer_deferrable(&vxlan->age_timer);
2437 vxlan->age_timer.function = vxlan_cleanup;
2438 vxlan->age_timer.data = (unsigned long) vxlan;
2439
0dfbdf41 2440 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2441
d342894c 2442 vxlan->dev = dev;
2443
2444 for (h = 0; h < FDB_HASH_SIZE; ++h)
2445 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2446}
2447
2448static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2449 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2450 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2451 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2452 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2453 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2454 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2455 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2456 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2457 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2458 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2459 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2460 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2461 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2462 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2463 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2464 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
ee122c79 2465 [IFLA_VXLAN_FLOWBASED] = { .type = NLA_U8 },
823aa873 2466 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2467 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2468 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2469 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2470 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2471 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2472 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2473 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2474};
2475
2476static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2477{
2478 if (tb[IFLA_ADDRESS]) {
2479 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2480 pr_debug("invalid link address (not ethernet)\n");
2481 return -EINVAL;
2482 }
2483
2484 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2485 pr_debug("invalid all zero ethernet address\n");
2486 return -EADDRNOTAVAIL;
2487 }
2488 }
2489
2490 if (!data)
2491 return -EINVAL;
2492
2493 if (data[IFLA_VXLAN_ID]) {
2494 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2495 if (id >= VXLAN_VID_MASK)
2496 return -ERANGE;
2497 }
2498
05f47d69 2499 if (data[IFLA_VXLAN_PORT_RANGE]) {
2500 const struct ifla_vxlan_port_range *p
2501 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2502
2503 if (ntohs(p->high) < ntohs(p->low)) {
2504 pr_debug("port range %u .. %u not valid\n",
2505 ntohs(p->low), ntohs(p->high));
2506 return -EINVAL;
2507 }
2508 }
2509
d342894c 2510 return 0;
2511}
2512
1b13c97f
YB		 2513static void vxlan_get_drvinfo(struct net_device *netdev,
2514 struct ethtool_drvinfo *drvinfo)
2515{
2516 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2517 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2518}
2519
2520static const struct ethtool_ops vxlan_ethtool_ops = {
2521 .get_drvinfo = vxlan_get_drvinfo,
2522 .get_link = ethtool_op_get_link,
2523};
2524
553675fb 2525static void vxlan_del_work(struct work_struct *work)
2526{
2527 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2528 udp_tunnel_sock_release(vs->sock);
553675fb 2529 kfree_rcu(vs, rcu);
2530}
2531
3ee64f39
TH		 2532static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2533 __be16 port, u32 flags)
553675fb 2534{
e4c7ed41 2535 struct socket *sock;
3ee64f39
TH		 2536 struct udp_port_cfg udp_conf;
2537 int err;
e4c7ed41 2538
3ee64f39 2539 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2540
3ee64f39
TH		 2541 if (ipv6) {
2542 udp_conf.family = AF_INET6;
3ee64f39 2543 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2544 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2545 } else {
2546 udp_conf.family = AF_INET;
e4c7ed41 2547 }
e4c7ed41 2548
3ee64f39 2549 udp_conf.local_udp_port = port;
553675fb 2550
3ee64f39
TH		 2551 /* Open UDP socket */
2552 err = udp_sock_create(net, &udp_conf, &sock);
2553 if (err < 0)
2554 return ERR_PTR(err);
e4c7ed41 2555
39deb2c7 2556 return sock;
e4c7ed41
CW		 2557}
2558
2559/* Create new listen socket if needed */
2560static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9 2561 u32 flags)
e4c7ed41
CW		 2562{
2563 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2564 struct vxlan_sock *vs;
2565 struct socket *sock;
e4c7ed41 2566 unsigned int h;
359a0ea9 2567 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2568 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2569
dc01e7d3 2570 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2571 if (!vs)
2572 return ERR_PTR(-ENOMEM);
2573
2574 for (h = 0; h < VNI_HASH_SIZE; ++h)
2575 INIT_HLIST_HEAD(&vs->vni_list[h]);
2576
2577 INIT_WORK(&vs->del_work, vxlan_del_work);
2578
3ee64f39 2579 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2580 if (IS_ERR(sock)) {
56ef9c90
MRL		 2581 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2582 PTR_ERR(sock));
553675fb 2583 kfree(vs);
e50fddc8 2584 return ERR_CAST(sock);
553675fb 2585 }
e4c7ed41
CW		 2586
2587 vs->sock = sock;
9c2e24e1 2588 atomic_set(&vs->refcnt, 1);
af33c1ad 2589 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2590
dc01e7d3
OG		 2591 /* Initialize the vxlan udp offloads structure */
2592 vs->udp_offloads.port = port;
2593 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2594 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2595
9c2e24e1
PS		 2596 spin_lock(&vn->sock_lock);
2597 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2598 vxlan_notify_add_rx_port(vs);
9c2e24e1 2599 spin_unlock(&vn->sock_lock);
553675fb 2600
2601 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2602 tunnel_cfg.sk_user_data = vs;
2603 tunnel_cfg.encap_type = 1;
2604 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2605 tunnel_cfg.encap_destroy = NULL;
2606
2607 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2608
9c2e24e1
PS		 2609 return vs;
2610}
2611
614732ea
TG		 2612static struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2613 bool no_share, u32 flags)
9c2e24e1
PS		 2614{
2615 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2616 struct vxlan_sock *vs;
19ca9fc1 2617 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2618
56ef9c90
MRL		 2619 if (!no_share) {
2620 spin_lock(&vn->sock_lock);
2621 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port,
2622 flags);
614732ea 2623 if (vs) {
56ef9c90
MRL		 2624 if (!atomic_add_unless(&vs->refcnt, 1, 0))
2625 vs = ERR_PTR(-EBUSY);
2626 spin_unlock(&vn->sock_lock);
2627 return vs;
2628 }
2629 spin_unlock(&vn->sock_lock);
2630 }
9c2e24e1 2631
614732ea 2632 return vxlan_socket_create(net, port, flags);
553675fb 2633}
2634
0dfbdf41
TG		 2635static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2636 struct vxlan_config *conf)
d342894c 2637{
33564bbb 2638 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
d342894c 2639 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2640 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2641 int err;
e4c7ed41 2642 bool use_ipv6 = false;
0dfbdf41 2643 __be16 default_port = vxlan->cfg.dst_port;
d342894c 2644
33564bbb 2645 vxlan->net = src_net;
f01ec1c0 2646
0dfbdf41 2647 dst->remote_vni = conf->vni;
d342894c 2648
0dfbdf41 2649 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2650
0dfbdf41
TG		 2651 /* Unless IPv6 is explicitly requested, assume IPv4 */
2652 if (!dst->remote_ip.sa.sa_family)
2653 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2654
0dfbdf41
TG		 2655 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
2656 vxlan->cfg.saddr.sa.sa_family == AF_INET6)
e4c7ed41 2657 use_ipv6 = true;
d342894c 2658
0dfbdf41 2659 if (conf->remote_ifindex) {
34e02aa1 2660 struct net_device *lowerdev
0dfbdf41
TG		 2661 = __dev_get_by_index(src_net, conf->remote_ifindex);
2662
2663 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2664
2665 if (!lowerdev) {
c7995c43 2666 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2667 return -ENODEV;
2668 }
d342894c 2669
e4c7ed41
CW		 2670#if IS_ENABLED(CONFIG_IPV6)
2671 if (use_ipv6) {
2672 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2673 if (idev && idev->cnf.disable_ipv6) {
2674 pr_info("IPv6 is disabled via sysctl\n");
2675 return -EPERM;
2676 }
2677 vxlan->flags |= VXLAN_F_IPV6;
2678 }
2679#endif
2680
0dfbdf41 2681 if (!conf->mtu)
e4c7ed41 2682 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2683
2853af6a 2684 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2685 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2686 } else if (use_ipv6)
2687 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2688
0dfbdf41
TG		 2689 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2690 if (!vxlan->cfg.dst_port)
2691 vxlan->cfg.dst_port = default_port;
2692 vxlan->flags |= conf->flags;
2693
2694 if (!vxlan->cfg.age_interval)
2695 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2696
2697 if (vxlan_find_vni(src_net, conf->vni, use_ipv6 ? AF_INET6 : AF_INET,
2698 vxlan->cfg.dst_port, vxlan->flags))
2699 return -EEXIST;
2700
2701 dev->ethtool_ops = &vxlan_ethtool_ops;
2702
2703 /* create an fdb entry for a valid default destination */
2704 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2705 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2706 &vxlan->default_dst.remote_ip,
2707 NUD_REACHABLE|NUD_PERMANENT,
2708 NLM_F_EXCL|NLM_F_CREATE,
2709 vxlan->cfg.dst_port,
2710 vxlan->default_dst.remote_vni,
2711 vxlan->default_dst.remote_ifindex,
2712 NTF_SELF);
2713 if (err)
2714 return err;
2715 }
2716
2717 err = register_netdevice(dev);
2718 if (err) {
2719 vxlan_fdb_delete_default(vxlan);
2720 return err;
2721 }
2722
2723 list_add(&vxlan->next, &vn->vxlan_list);
2724
2725 return 0;
2726}
2727
2728struct net_device *vxlan_dev_create(struct net *net, const char *name,
2729 u8 name_assign_type, struct vxlan_config *conf)
2730{
2731 struct nlattr *tb[IFLA_MAX+1];
2732 struct net_device *dev;
2733 int err;
2734
2735 memset(&tb, 0, sizeof(tb));
2736
2737 dev = rtnl_create_link(net, name, name_assign_type,
2738 &vxlan_link_ops, tb);
2739 if (IS_ERR(dev))
2740 return dev;
2741
2742 err = vxlan_dev_configure(net, dev, conf);
2743 if (err < 0) {
2744 free_netdev(dev);
2745 return ERR_PTR(err);
2746 }
2747
2748 return dev;
2749}
2750EXPORT_SYMBOL_GPL(vxlan_dev_create);
2751
2752static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2753 struct nlattr *tb[], struct nlattr *data[])
2754{
2755 struct vxlan_config conf;
2756 int err;
2757
2758 if (!data[IFLA_VXLAN_ID])
2759 return -EINVAL;
2760
2761 memset(&conf, 0, sizeof(conf));
2762 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2763
2764 if (data[IFLA_VXLAN_GROUP]) {
2765 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2766 } else if (data[IFLA_VXLAN_GROUP6]) {
2767 if (!IS_ENABLED(CONFIG_IPV6))
2768 return -EPFNOSUPPORT;
2769
2770 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2771 conf.remote_ip.sa.sa_family = AF_INET6;
2772 }
2773
2774 if (data[IFLA_VXLAN_LOCAL]) {
2775 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2776 conf.saddr.sa.sa_family = AF_INET;
2777 } else if (data[IFLA_VXLAN_LOCAL6]) {
2778 if (!IS_ENABLED(CONFIG_IPV6))
2779 return -EPFNOSUPPORT;
2780
2781 /* TODO: respect scope id */
2782 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2783 conf.saddr.sa.sa_family = AF_INET6;
2784 }
2785
2786 if (data[IFLA_VXLAN_LINK])
2787 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2788
d342894c 2789 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2790 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2791
afb97186 2792 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2793 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2794
d342894c 2795 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2796 conf.flags |= VXLAN_F_LEARN;
d342894c 2797
2798 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2799 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2800
e4f67add 2801 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2802 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2803
2804 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2805 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2806
2807 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2808 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2809
2810 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2811 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2812
d342894c 2813 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2814 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2815
ee122c79
TG		 2816 if (data[IFLA_VXLAN_FLOWBASED] &&
2817 nla_get_u8(data[IFLA_VXLAN_FLOWBASED]))
0dfbdf41 2818 conf.flags |= VXLAN_F_FLOW_BASED;
ee122c79 2819
05f47d69 2820 if (data[IFLA_VXLAN_PORT_RANGE]) {
2821 const struct ifla_vxlan_port_range *p
2822 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2823 conf.port_min = ntohs(p->low);
2824 conf.port_max = ntohs(p->high);
05f47d69 2825 }
2826
823aa873 2827 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2828 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2829
359a0ea9 2830 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
0dfbdf41 2831 conf.flags |= VXLAN_F_UDP_CSUM;
359a0ea9
TH		 2832
2833 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2834 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2835 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2836
2837 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2838 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2839 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2840
dfd8645e
TH		 2841 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2842 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2843 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 2844
2845 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2846 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2847 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2848
3511494c 2849 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2850 conf.flags |= VXLAN_F_GBP;
3511494c 2851
0ace2ca8 2852 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2853 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2854
0dfbdf41
TG		 2855 err = vxlan_dev_configure(src_net, dev, &conf);
2856 switch (err) {
2857 case -ENODEV:
2858 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2859 break;
1b13c97f 2860
0dfbdf41
TG		 2861 case -EPERM:
2862 pr_info("IPv6 is disabled via sysctl\n");
2863 break;
d342894c 2864
0dfbdf41
TG		 2865 case -EEXIST:
2866 pr_info("duplicate VNI %u\n", conf.vni);
2867 break;
afbd8bae
MR		 2868 }
2869
0dfbdf41 2870 return err;
d342894c 2871}
2872
2873static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2874{
2875 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2876 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2877
fe5c3561 2878 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2879 if (!hlist_unhashed(&vxlan->hlist))
2880 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2881 spin_unlock(&vn->sock_lock);
2882
553675fb 2883 list_del(&vxlan->next);
d342894c 2884 unregister_netdevice_queue(dev, head);
2885}
2886
2887static size_t vxlan_get_size(const struct net_device *dev)
2888{
2889
2890 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2891 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2892 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2893 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2894 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2895 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2896 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2897 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2898 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2899 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2900 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
ee122c79 2901 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_FLOWBASED */
d342894c 2902 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2903 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2904 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2905 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2906 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2907 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2908 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2909 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2910 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2911 0;
2912}
2913
2914static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2915{
2916 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2917 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2918 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 2919 .low = htons(vxlan->cfg.port_min),
2920 .high = htons(vxlan->cfg.port_max),
05f47d69 2921 };
d342894c 2922
c7995c43 2923 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2924 goto nla_put_failure;
2925
e4c7ed41
CW		 2926 if (!vxlan_addr_any(&dst->remote_ip)) {
2927 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 2928 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
2929 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 2930 goto nla_put_failure;
2931#if IS_ENABLED(CONFIG_IPV6)
2932 } else {
930345ea
JB		 2933 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
2934 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 2935 goto nla_put_failure;
2936#endif
2937 }
2938 }
d342894c 2939
c7995c43 2940 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2941 goto nla_put_failure;
2942
0dfbdf41
TG		 2943 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
2944 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 2945 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 2946 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 2947 goto nla_put_failure;
2948#if IS_ENABLED(CONFIG_IPV6)
2949 } else {
930345ea 2950 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 2951 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 2952 goto nla_put_failure;
2953#endif
2954 }
2955 }
d342894c 2956
0dfbdf41
TG		 2957 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
2958 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 2959 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2960 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2961 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2962 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2963 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2964 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2965 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2966 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2967 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
ee122c79
TG		 2968 nla_put_u8(skb, IFLA_VXLAN_FLOWBASED,
2969 !!(vxlan->flags & VXLAN_F_FLOW_BASED)) ||
0dfbdf41
TG		 2970 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
2971 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
2972 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9
TH		 2973 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2974 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2975 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2976 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2977 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 2978 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2979 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2980 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2981 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2982 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 2983 goto nla_put_failure;
2984
05f47d69 2985 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2986 goto nla_put_failure;
2987
3511494c
TG		 2988 if (vxlan->flags & VXLAN_F_GBP &&
2989 nla_put_flag(skb, IFLA_VXLAN_GBP))
2990 goto nla_put_failure;
2991
0ace2ca8
TH		 2992 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
2993 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
2994 goto nla_put_failure;
2995
d342894c 2996 return 0;
2997
2998nla_put_failure:
2999 return -EMSGSIZE;
3000}
3001
1728d4fa
ND		 3002static struct net *vxlan_get_link_net(const struct net_device *dev)
3003{
3004 struct vxlan_dev *vxlan = netdev_priv(dev);
3005
3006 return vxlan->net;
3007}
3008
d342894c 3009static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3010 .kind = "vxlan",
3011 .maxtype = IFLA_VXLAN_MAX,
3012 .policy = vxlan_policy,
3013 .priv_size = sizeof(struct vxlan_dev),
3014 .setup = vxlan_setup,
3015 .validate = vxlan_validate,
3016 .newlink = vxlan_newlink,
3017 .dellink = vxlan_dellink,
3018 .get_size = vxlan_get_size,
3019 .fill_info = vxlan_fill_info,
1728d4fa 3020 .get_link_net = vxlan_get_link_net,
d342894c 3021};
3022
acaf4e70
DB		 3023static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3024 struct net_device *dev)
3025{
3026 struct vxlan_dev *vxlan, *next;
3027 LIST_HEAD(list_kill);
3028
3029 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3030 struct vxlan_rdst *dst = &vxlan->default_dst;
3031
3032 /* In case we created vxlan device with carrier
3033 * and we loose the carrier due to module unload
3034 * we also need to remove vxlan device. In other
3035 * cases, it's not necessary and remote_ifindex
3036 * is 0 here, so no matches.
3037 */
3038 if (dst->remote_ifindex == dev->ifindex)
3039 vxlan_dellink(vxlan->dev, &list_kill);
3040 }
3041
3042 unregister_netdevice_many(&list_kill);
3043}
3044
3045static int vxlan_lowerdev_event(struct notifier_block *unused,
3046 unsigned long event, void *ptr)
3047{
3048 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3049 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3050
783c1463 3051 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3052 vxlan_handle_lowerdev_unregister(vn, dev);
3053
3054 return NOTIFY_DONE;
3055}
3056
3057static struct notifier_block vxlan_notifier_block __read_mostly = {
3058 .notifier_call = vxlan_lowerdev_event,
3059};
3060
d342894c 3061static __net_init int vxlan_init_net(struct net *net)
3062{
3063 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3064 unsigned int h;
d342894c 3065
553675fb 3066 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3067 spin_lock_init(&vn->sock_lock);
d342894c 3068
553675fb 3069 for (h = 0; h < PORT_HASH_SIZE; ++h)
3070 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3071
3072 return 0;
3073}
3074
f01ec1c0
ND		 3075static void __net_exit vxlan_exit_net(struct net *net)
3076{
3077 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3078 struct vxlan_dev *vxlan, *next;
3079 struct net_device *dev, *aux;
3080 LIST_HEAD(list);
3081
3082 rtnl_lock();
3083 for_each_netdev_safe(net, dev, aux)
3084 if (dev->rtnl_link_ops == &vxlan_link_ops)
3085 unregister_netdevice_queue(dev, &list);
3086
3087 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3088 /* If vxlan->dev is in the same netns, it has already been added
3089 * to the list by the previous loop.
3090 */
3091 if (!net_eq(dev_net(vxlan->dev), net))
13c3ed6a 3092 unregister_netdevice_queue(vxlan->dev, &list);
f01ec1c0
ND		 3093 }
3094
3095 unregister_netdevice_many(&list);
3096 rtnl_unlock();
3097}
3098
d342894c 3099static struct pernet_operations vxlan_net_ops = {
3100 .init = vxlan_init_net,
f01ec1c0 3101 .exit = vxlan_exit_net,
d342894c 3102 .id = &vxlan_net_id,
3103 .size = sizeof(struct vxlan_net),
3104};
3105
3106static int __init vxlan_init_module(void)
3107{
3108 int rc;
3109
758c57d1
SH		 3110 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3111 if (!vxlan_wq)
3112 return -ENOMEM;
3113
d342894c 3114 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3115
783c1463 3116 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3117 if (rc)
3118 goto out1;
3119
acaf4e70 3120 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3121 if (rc)
3122 goto out2;
3123
acaf4e70
DB		 3124 rc = rtnl_link_register(&vxlan_link_ops);
3125 if (rc)
3126 goto out3;
d342894c 3127
acaf4e70
DB		 3128 return 0;
3129out3:
3130 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3131out2:
783c1463 3132 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3133out1:
758c57d1 3134 destroy_workqueue(vxlan_wq);
d342894c 3135 return rc;
3136}
7332a13b 3137late_initcall(vxlan_init_module);
d342894c 3138
3139static void __exit vxlan_cleanup_module(void)
3140{
b7153984 3141 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3142 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3143 destroy_workqueue(vxlan_wq);
783c1463
DB		 3144 unregister_pernet_subsys(&vxlan_net_ops);
3145 /* rcu_barrier() is called by netns */
d342894c 3146}
3147module_exit(vxlan_cleanup_module);
3148
3149MODULE_LICENSE("GPL");
3150MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3151MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3152MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3153MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.614325 seconds and 5 git commands to generate.