]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: consolidate GBP handling even more
[mirror_ubuntu-artful-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
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>
40d29af0 45
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
ee122c79 52#include <net/dst_metadata.h>
d342894c 53
54#define VXLAN_VERSION "0.1"
55
553675fb 56#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
23c578bf 61/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 63 * for compatibility with early adopters.
23c578bf 64 */
9daaa397
SH		 65static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 67MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
60d9d4c6 73static int vxlan_net_id;
0dfbdf41 74static struct rtnl_link_ops vxlan_link_ops;
553675fb 75
7256eac1 76static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 77
205f356d 78static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
d342894c 87/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
3e61aa8f 93 struct list_head remotes;
7177a3b0 94 u8 eth_addr[ETH_ALEN];
d342894c 95 u16 state; /* see ndm_state */
ae884082 96 u8 flags; /* see ndm_flags */
d342894c 97};
98
d342894c 99/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
758c57d1 101static struct workqueue_struct *vxlan_wq;
d342894c 102
ee122c79
TG		 103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
e7030878
TG		 105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
ee122c79
TG		 107}
108
e4c7ed41
CW		 109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
f0ef3126
JB		 113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
f0ef3126
JB		 123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
f0ef3126
JB		 131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
f0ef3126 139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
e4c7ed41
CW		 150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 153 const union vxlan_addr *ip)
e4c7ed41 154{
f0ef3126 155 if (ip->sa.sa_family == AF_INET6)
930345ea 156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 157 else
930345ea 158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
f0ef3126 166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
f0ef3126 171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
f0ef3126 176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
f0ef3126
JB		 181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
e4c7ed41
CW		 190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 193 const union vxlan_addr *ip)
e4c7ed41 194{
930345ea 195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 196}
197#endif
198
553675fb 199/* Virtual Network hash table head */
54bfd872 200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 201{
54bfd872 202 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
553675fb 210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
3e61aa8f
SH		 213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
5ca5461c 216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 217{
5ca5461c 218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 224}
225
ac5132d1
TG		 226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
553675fb 231{
232 struct vxlan_sock *vs;
af33c1ad
TH		 233
234 flags &= VXLAN_F_RCV_FLAGS;
553675fb 235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 238 vxlan_get_sk_family(vs) == family &&
af33c1ad 239 vs->flags == flags)
553675fb 240 return vs;
241 }
242 return NULL;
d342894c 243}
244
54bfd872 245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 246{
247 struct vxlan_dev *vxlan;
d342894c 248
b9167b2e
JB		 249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs->flags & VXLAN_F_COLLECT_METADATA)
251 vni = 0;
252
54bfd872
JB		 253 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
254 if (vxlan->default_dst.remote_vni == vni)
d342894c 255 return vxlan;
256 }
257
258 return NULL;
259}
260
5cfccc5a 261/* Look up VNI in a per net namespace table */
54bfd872 262static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 263 sa_family_t family, __be16 port,
264 u32 flags)
5cfccc5a
PS		 265{
266 struct vxlan_sock *vs;
267
ac5132d1 268 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 269 if (!vs)
270 return NULL;
271
54bfd872 272 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 273}
274
d342894c 275/* Fill in neighbour message in skbuff. */
276static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 277 const struct vxlan_fdb *fdb,
278 u32 portid, u32 seq, int type, unsigned int flags,
279 const struct vxlan_rdst *rdst)
d342894c 280{
281 unsigned long now = jiffies;
282 struct nda_cacheinfo ci;
283 struct nlmsghdr *nlh;
284 struct ndmsg *ndm;
e4f67add 285 bool send_ip, send_eth;
d342894c 286
287 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
288 if (nlh == NULL)
289 return -EMSGSIZE;
290
291 ndm = nlmsg_data(nlh);
292 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 293
294 send_eth = send_ip = true;
295
296 if (type == RTM_GETNEIGH) {
297 ndm->ndm_family = AF_INET;
e4c7ed41 298 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 299 send_eth = !is_zero_ether_addr(fdb->eth_addr);
300 } else
301 ndm->ndm_family = AF_BRIDGE;
d342894c 302 ndm->ndm_state = fdb->state;
303 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 304 ndm->ndm_flags = fdb->flags;
545469f7 305 ndm->ndm_type = RTN_UNICAST;
d342894c 306
193523bf 307 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 308 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 309 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 310 goto nla_put_failure;
311
e4f67add 312 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 313 goto nla_put_failure;
314
e4c7ed41 315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 316 goto nla_put_failure;
317
0dfbdf41 318 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 319 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
320 goto nla_put_failure;
c7995c43 321 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 322 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d
DS		 323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
053c095a
JB		 336 nlmsg_end(skb, nlh);
337 return 0;
d342894c 338
339nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342}
343
344static inline size_t vxlan_nlmsg_size(void)
345{
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 353 + nla_total_size(sizeof(struct nda_cacheinfo));
354}
355
9e4b93f9
ND		 356static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
d342894c 358{
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
9e4b93f9 367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380}
381
e4c7ed41 382static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 383{
384 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
e4c7ed41 389 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 391 };
3e61aa8f 392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 394}
395
396static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397{
bb3fd687
SH		 398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
9e4b93f9 401 struct vxlan_rdst remote = { };
e4f67add 402
e4f67add
DS		 403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
9e4b93f9 405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 406}
407
d342894c 408/* Hash Ethernet address */
409static u32 eth_hash(const unsigned char *addr)
410{
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414#ifdef __BIG_ENDIAN
d342894c 415 value >>= 16;
321fb991 416#else
417 value <<= 16;
d342894c 418#endif
419 return hash_64(value, FDB_HASH_BITS);
420}
421
422/* Hash chain to use given mac address */
423static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
424 const u8 *mac)
425{
426 return &vxlan->fdb_head[eth_hash(mac)];
427}
428
429/* Look up Ethernet address in forwarding table */
014be2c8 430static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 431 const u8 *mac)
d342894c 432{
433 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
434 struct vxlan_fdb *f;
d342894c 435
b67bfe0d 436 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 437 if (ether_addr_equal(mac, f->eth_addr))
d342894c 438 return f;
439 }
440
441 return NULL;
442}
443
014be2c8
SS		 444static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
445 const u8 *mac)
446{
447 struct vxlan_fdb *f;
448
449 f = __vxlan_find_mac(vxlan, mac);
450 if (f)
451 f->used = jiffies;
452
453 return f;
454}
455
a5e7c10a
MR		 456/* caller should hold vxlan->hash_lock */
457static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 458 union vxlan_addr *ip, __be16 port,
54bfd872 459 __be32 vni, __u32 ifindex)
6681712d 460{
3e61aa8f 461 struct vxlan_rdst *rd;
6681712d 462
3e61aa8f 463 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 464 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 465 rd->remote_port == port &&
466 rd->remote_vni == vni &&
467 rd->remote_ifindex == ifindex)
a5e7c10a 468 return rd;
6681712d 469 }
3e61aa8f 470
a5e7c10a
MR		 471 return NULL;
472}
473
906dc186
TR		 474/* Replace destination of unicast mac */
475static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 476 union vxlan_addr *ip, __be16 port, __be32 vni,
477 __u32 ifindex)
906dc186
TR		 478{
479 struct vxlan_rdst *rd;
480
481 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
482 if (rd)
483 return 0;
484
485 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
486 if (!rd)
487 return 0;
0c1d70af
PA		 488
489 dst_cache_reset(&rd->dst_cache);
e4c7ed41 490 rd->remote_ip = *ip;
906dc186
TR		 491 rd->remote_port = port;
492 rd->remote_vni = vni;
493 rd->remote_ifindex = ifindex;
494 return 1;
495}
496
a5e7c10a
MR		 497/* Add/update destinations for multicast */
498static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 499 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 500 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 501{
502 struct vxlan_rdst *rd;
503
504 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
505 if (rd)
506 return 0;
507
6681712d
DS		 508 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
509 if (rd == NULL)
510 return -ENOBUFS;
0c1d70af
PA		 511
512 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
513 kfree(rd);
514 return -ENOBUFS;
515 }
516
e4c7ed41 517 rd->remote_ip = *ip;
6681712d
DS		 518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 521
522 list_add_tail_rcu(&rd->list, &f->remotes);
523
9e4b93f9 524 *rdp = rd;
6681712d
DS		 525 return 1;
526}
527
dfd8645e
TH		 528static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
529 unsigned int off,
530 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 531 __be32 vni_field,
532 struct gro_remcsum *grc,
0ace2ca8 533 bool nopartial)
dfd8645e 534{
b7fe10e5 535 size_t start, offset;
dfd8645e
TH		 536
537 if (skb->remcsum_offload)
b7fe10e5 538 return vh;
dfd8645e
TH		 539
540 if (!NAPI_GRO_CB(skb)->csum_valid)
541 return NULL;
542
54bfd872
JB		 543 start = vxlan_rco_start(vni_field);
544 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 545
b7fe10e5
TH		 546 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
547 start, offset, grc, nopartial);
dfd8645e
TH		 548
549 skb->remcsum_offload = 1;
550
551 return vh;
552}
553
a2b12f3c
TH		 554static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
555 struct sk_buff *skb,
556 struct udp_offload *uoff)
dc01e7d3
OG		 557{
558 struct sk_buff *p, **pp = NULL;
559 struct vxlanhdr *vh, *vh2;
9b174d88 560 unsigned int hlen, off_vx;
dc01e7d3 561 int flush = 1;
dfd8645e
TH		 562 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
563 udp_offloads);
54bfd872 564 __be32 flags;
26c4f7da
TH		 565 struct gro_remcsum grc;
566
567 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 568
569 off_vx = skb_gro_offset(skb);
570 hlen = off_vx + sizeof(*vh);
571 vh = skb_gro_header_fast(skb, off_vx);
572 if (skb_gro_header_hard(skb, hlen)) {
573 vh = skb_gro_header_slow(skb, hlen, off_vx);
574 if (unlikely(!vh))
575 goto out;
576 }
dc01e7d3 577
dfd8645e
TH		 578 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
579
54bfd872 580 flags = vh->vx_flags;
dfd8645e
TH		 581
582 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
583 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 584 vh->vx_vni, &grc,
0ace2ca8
TH		 585 !!(vs->flags &
586 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 587
588 if (!vh)
589 goto out;
590 }
591
b7fe10e5
TH		 592 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
593
dc01e7d3
OG		 594 flush = 0;
595
596 for (p = *head; p; p = p->next) {
597 if (!NAPI_GRO_CB(p)->same_flow)
598 continue;
599
600 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 601 if (vh->vx_flags != vh2->vx_flags ||
602 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 603 NAPI_GRO_CB(p)->same_flow = 0;
604 continue;
605 }
dc01e7d3
OG		 606 }
607
9b174d88 608 pp = eth_gro_receive(head, skb);
dc01e7d3 609
dc01e7d3 610out:
26c4f7da 611 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 612 NAPI_GRO_CB(skb)->flush |= flush;
613
614 return pp;
615}
616
a2b12f3c
TH		 617static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
618 struct udp_offload *uoff)
dc01e7d3 619{
cfdf1e1b
JG		 620 udp_tunnel_gro_complete(skb, nhoff);
621
9b174d88 622 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 623}
624
53cf5275 625/* Notify netdevs that UDP port started listening */
dc01e7d3 626static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 627{
628 struct net_device *dev;
dc01e7d3 629 struct sock *sk = vs->sock->sk;
53cf5275 630 struct net *net = sock_net(sk);
705cc62f 631 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 632 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 633 int err;
634
635 if (sa_family == AF_INET) {
787d7ac3 636 err = udp_add_offload(net, &vs->udp_offloads);
dc01e7d3
OG		 637 if (err)
638 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
639 }
53cf5275
JG		 640
641 rcu_read_lock();
642 for_each_netdev_rcu(net, dev) {
643 if (dev->netdev_ops->ndo_add_vxlan_port)
644 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
645 port);
646 }
647 rcu_read_unlock();
648}
649
650/* Notify netdevs that UDP port is no more listening */
dc01e7d3 651static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 652{
653 struct net_device *dev;
dc01e7d3 654 struct sock *sk = vs->sock->sk;
53cf5275 655 struct net *net = sock_net(sk);
705cc62f 656 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 657 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 658
659 rcu_read_lock();
660 for_each_netdev_rcu(net, dev) {
661 if (dev->netdev_ops->ndo_del_vxlan_port)
662 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
663 port);
664 }
665 rcu_read_unlock();
dc01e7d3
OG		 666
667 if (sa_family == AF_INET)
668 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 669}
670
d342894c 671/* Add new entry to forwarding table -- assumes lock held */
672static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 673 const u8 *mac, union vxlan_addr *ip,
6681712d 674 __u16 state, __u16 flags,
54bfd872 675 __be16 port, __be32 vni, __u32 ifindex,
ae884082 676 __u8 ndm_flags)
d342894c 677{
9e4b93f9 678 struct vxlan_rdst *rd = NULL;
d342894c 679 struct vxlan_fdb *f;
680 int notify = 0;
681
014be2c8 682 f = __vxlan_find_mac(vxlan, mac);
d342894c 683 if (f) {
684 if (flags & NLM_F_EXCL) {
685 netdev_dbg(vxlan->dev,
686 "lost race to create %pM\n", mac);
687 return -EEXIST;
688 }
689 if (f->state != state) {
690 f->state = state;
691 f->updated = jiffies;
692 notify = 1;
693 }
ae884082
DS		 694 if (f->flags != ndm_flags) {
695 f->flags = ndm_flags;
696 f->updated = jiffies;
697 notify = 1;
698 }
906dc186
TR		 699 if ((flags & NLM_F_REPLACE)) {
700 /* Only change unicasts */
701 if (!(is_multicast_ether_addr(f->eth_addr) ||
702 is_zero_ether_addr(f->eth_addr))) {
60840429 703 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 704 ifindex);
906dc186
TR		 705 } else
706 return -EOPNOTSUPP;
707 }
6681712d 708 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 709 (is_multicast_ether_addr(f->eth_addr) ||
710 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 711 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
712 &rd);
6681712d
DS		 713
714 if (rc < 0)
715 return rc;
716 notify |= rc;
717 }
d342894c 718 } else {
719 if (!(flags & NLM_F_CREATE))
720 return -ENOENT;
721
0dfbdf41
TG		 722 if (vxlan->cfg.addrmax &&
723 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 724 return -ENOSPC;
725
906dc186
TR		 726 /* Disallow replace to add a multicast entry */
727 if ((flags & NLM_F_REPLACE) &&
728 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
729 return -EOPNOTSUPP;
730
e4c7ed41 731 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 732 f = kmalloc(sizeof(*f), GFP_ATOMIC);
733 if (!f)
734 return -ENOMEM;
735
736 notify = 1;
d342894c 737 f->state = state;
ae884082 738 f->flags = ndm_flags;
d342894c 739 f->updated = f->used = jiffies;
3e61aa8f 740 INIT_LIST_HEAD(&f->remotes);
d342894c 741 memcpy(f->eth_addr, mac, ETH_ALEN);
742
9e4b93f9 743 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 744
d342894c 745 ++vxlan->addrcnt;
746 hlist_add_head_rcu(&f->hlist,
747 vxlan_fdb_head(vxlan, mac));
748 }
749
9e4b93f9
ND		 750 if (notify) {
751 if (rd == NULL)
752 rd = first_remote_rtnl(f);
753 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
754 }
d342894c 755
756 return 0;
757}
758
6706c82e 759static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 760{
761 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 762 struct vxlan_rdst *rd, *nd;
6681712d 763
0c1d70af
PA		 764 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
765 dst_cache_destroy(&rd->dst_cache);
6681712d 766 kfree(rd);
0c1d70af 767 }
6681712d
DS		 768 kfree(f);
769}
770
d342894c 771static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
772{
773 netdev_dbg(vxlan->dev,
774 "delete %pM\n", f->eth_addr);
775
776 --vxlan->addrcnt;
9e4b93f9 777 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 778
779 hlist_del_rcu(&f->hlist);
6681712d 780 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 781}
782
f0b074be 783static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB		 784 union vxlan_addr *ip, __be16 *port, __be32 *vni,
785 u32 *ifindex)
d342894c 786{
6681712d 787 struct net *net = dev_net(vxlan->dev);
e4c7ed41 788 int err;
d342894c 789
f0b074be 790 if (tb[NDA_DST]) {
e4c7ed41
CW		 791 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
792 if (err)
793 return err;
f0b074be 794 } else {
e4c7ed41
CW		 795 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
796 if (remote->sa.sa_family == AF_INET) {
797 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
798 ip->sa.sa_family = AF_INET;
799#if IS_ENABLED(CONFIG_IPV6)
800 } else {
801 ip->sin6.sin6_addr = in6addr_any;
802 ip->sa.sa_family = AF_INET6;
803#endif
804 }
f0b074be 805 }
d342894c 806
6681712d 807 if (tb[NDA_PORT]) {
73cf3317 808 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 809 return -EINVAL;
f0b074be
MR		 810 *port = nla_get_be16(tb[NDA_PORT]);
811 } else {
0dfbdf41 812 *port = vxlan->cfg.dst_port;
f0b074be 813 }
6681712d
DS		 814
815 if (tb[NDA_VNI]) {
816 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
817 return -EINVAL;
54bfd872 818 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 819 } else {
820 *vni = vxlan->default_dst.remote_vni;
821 }
6681712d
DS		 822
823 if (tb[NDA_IFINDEX]) {
5abb0029 824 struct net_device *tdev;
6681712d
DS		 825
826 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
827 return -EINVAL;
f0b074be 828 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 829 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 830 if (!tdev)
6681712d 831 return -EADDRNOTAVAIL;
f0b074be
MR		 832 } else {
833 *ifindex = 0;
834 }
835
836 return 0;
837}
838
839/* Add static entry (via netlink) */
840static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
841 struct net_device *dev,
f6f6424b 842 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 843{
844 struct vxlan_dev *vxlan = netdev_priv(dev);
845 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 846 union vxlan_addr ip;
f0b074be 847 __be16 port;
54bfd872
JB		 848 __be32 vni;
849 u32 ifindex;
f0b074be
MR		 850 int err;
851
852 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
853 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
854 ndm->ndm_state);
855 return -EINVAL;
856 }
857
858 if (tb[NDA_DST] == NULL)
859 return -EINVAL;
860
861 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
862 if (err)
863 return err;
6681712d 864
5933a7bb
MR		 865 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
866 return -EAFNOSUPPORT;
867
d342894c 868 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 869 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 870 port, vni, ifindex, ndm->ndm_flags);
d342894c 871 spin_unlock_bh(&vxlan->hash_lock);
872
873 return err;
874}
875
876/* Delete entry (via netlink) */
1690be63
VY		 877static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
878 struct net_device *dev,
f6f6424b 879 const unsigned char *addr, u16 vid)
d342894c 880{
881 struct vxlan_dev *vxlan = netdev_priv(dev);
882 struct vxlan_fdb *f;
bc7892ba 883 struct vxlan_rdst *rd = NULL;
e4c7ed41 884 union vxlan_addr ip;
bc7892ba 885 __be16 port;
54bfd872
JB		 886 __be32 vni;
887 u32 ifindex;
bc7892ba
MR		 888 int err;
889
890 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
891 if (err)
892 return err;
893
894 err = -ENOENT;
d342894c 895
896 spin_lock_bh(&vxlan->hash_lock);
897 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 898 if (!f)
899 goto out;
900
e4c7ed41
CW		 901 if (!vxlan_addr_any(&ip)) {
902 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 903 if (!rd)
904 goto out;
905 }
906
907 err = 0;
908
909 /* remove a destination if it's not the only one on the list,
910 * otherwise destroy the fdb entry
911 */
912 if (rd && !list_is_singular(&f->remotes)) {
913 list_del_rcu(&rd->list);
9e4b93f9 914 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 915 kfree_rcu(rd, rcu);
bc7892ba 916 goto out;
d342894c 917 }
bc7892ba
MR		 918
919 vxlan_fdb_destroy(vxlan, f);
920
921out:
d342894c 922 spin_unlock_bh(&vxlan->hash_lock);
923
924 return err;
925}
926
927/* Dump forwarding table */
928static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 929 struct net_device *dev,
930 struct net_device *filter_dev, int idx)
d342894c 931{
932 struct vxlan_dev *vxlan = netdev_priv(dev);
933 unsigned int h;
934
935 for (h = 0; h < FDB_HASH_SIZE; ++h) {
936 struct vxlan_fdb *f;
d342894c 937 int err;
938
b67bfe0d 939 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 940 struct vxlan_rdst *rd;
6681712d 941
3e61aa8f 942 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 943 if (idx < cb->args[0])
944 goto skip;
945
6681712d
DS		 946 err = vxlan_fdb_info(skb, vxlan, f,
947 NETLINK_CB(cb->skb).portid,
948 cb->nlh->nlmsg_seq,
949 RTM_NEWNEIGH,
950 NLM_F_MULTI, rd);
951 if (err < 0)
3e61aa8f 952 goto out;
3e61aa8f 953skip:
07a51cd3
AW		 954 ++idx;
955 }
d342894c 956 }
957 }
3e61aa8f 958out:
d342894c 959 return idx;
960}
961
962/* Watch incoming packets to learn mapping between Ethernet address
963 * and Tunnel endpoint.
c4b49512 964 * Return true if packet is bogus and should be dropped.
d342894c 965 */
26a41ae6 966static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 967 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 968{
969 struct vxlan_dev *vxlan = netdev_priv(dev);
970 struct vxlan_fdb *f;
d342894c 971
972 f = vxlan_find_mac(vxlan, src_mac);
973 if (likely(f)) {
5ca5461c 974 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 975
e4c7ed41 976 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 977 return false;
978
979 /* Don't migrate static entries, drop packets */
eb064c3b 980 if (f->state & NUD_NOARP)
26a41ae6 981 return true;
d342894c 982
983 if (net_ratelimit())
984 netdev_info(dev,
e4c7ed41 985 "%pM migrated from %pIS to %pIS\n",
a4870f79 986 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 987
e4c7ed41 988 rdst->remote_ip = *src_ip;
d342894c 989 f->updated = jiffies;
9e4b93f9 990 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 991 } else {
992 /* learned new entry */
993 spin_lock(&vxlan->hash_lock);
3bf74b1a 994
995 /* close off race between vxlan_flush and incoming packets */
996 if (netif_running(dev))
997 vxlan_fdb_create(vxlan, src_mac, src_ip,
998 NUD_REACHABLE,
999 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1000 vxlan->cfg.dst_port,
3bf74b1a 1001 vxlan->default_dst.remote_vni,
1002 0, NTF_SELF);
d342894c 1003 spin_unlock(&vxlan->hash_lock);
1004 }
26a41ae6 1005
1006 return false;
d342894c 1007}
1008
d342894c 1009/* See if multicast group is already in use by other ID */
95ab0991 1010static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1011{
553675fb 1012 struct vxlan_dev *vxlan;
b1be00a6 1013 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1014
95ab0991
G		 1015 /* The vxlan_sock is only used by dev, leaving group has
1016 * no effect on other vxlan devices.
1017 */
b1be00a6
JB		 1018 if (family == AF_INET && dev->vn4_sock &&
1019 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1020 return false;
b1be00a6
JB		 1021#if IS_ENABLED(CONFIG_IPV6)
1022 if (family == AF_INET6 && dev->vn6_sock &&
1023 atomic_read(&dev->vn6_sock->refcnt) == 1)
1024 return false;
1025#endif
95ab0991 1026
553675fb 1027 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1028 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1029 continue;
d342894c 1030
b1be00a6 1031 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1032 continue;
b1be00a6
JB		 1033#if IS_ENABLED(CONFIG_IPV6)
1034 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1035 continue;
1036#endif
95ab0991
G		 1037
1038 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1039 &dev->default_dst.remote_ip))
1040 continue;
1041
1042 if (vxlan->default_dst.remote_ifindex !=
1043 dev->default_dst.remote_ifindex)
1044 continue;
1045
1046 return true;
553675fb 1047 }
d342894c 1048
1049 return false;
1050}
1051
b1be00a6 1052static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1053{
b1be00a6 1054 struct vxlan_net *vn;
012a5729 1055
b1be00a6
JB		 1056 if (!vs)
1057 return;
7c47cedf
SH		 1058 if (!atomic_dec_and_test(&vs->refcnt))
1059 return;
d342894c 1060
b1be00a6 1061 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1062 spin_lock(&vn->sock_lock);
7c47cedf 1063 hlist_del_rcu(&vs->hlist);
dc01e7d3 1064 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1065 spin_unlock(&vn->sock_lock);
1066
7c47cedf 1067 queue_work(vxlan_wq, &vs->del_work);
d342894c 1068}
1069
b1be00a6
JB		 1070static void vxlan_sock_release(struct vxlan_dev *vxlan)
1071{
1072 __vxlan_sock_release(vxlan->vn4_sock);
1073#if IS_ENABLED(CONFIG_IPV6)
1074 __vxlan_sock_release(vxlan->vn6_sock);
1075#endif
1076}
1077
56ef9c90 1078/* Update multicast group membership when first VNI on
c4b49512 1079 * multicast address is brought up
7c47cedf 1080 */
56ef9c90 1081static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1082{
b1be00a6 1083 struct sock *sk;
e4c7ed41
CW		 1084 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1085 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1086 int ret = -EINVAL;
d342894c 1087
e4c7ed41
CW		 1088 if (ip->sa.sa_family == AF_INET) {
1089 struct ip_mreqn mreq = {
1090 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1091 .imr_ifindex = ifindex,
1092 };
1093
b1be00a6
JB		 1094 sk = vxlan->vn4_sock->sock->sk;
1095 lock_sock(sk);
56ef9c90 1096 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1097 release_sock(sk);
e4c7ed41
CW		 1098#if IS_ENABLED(CONFIG_IPV6)
1099 } else {
b1be00a6
JB		 1100 sk = vxlan->vn6_sock->sock->sk;
1101 lock_sock(sk);
56ef9c90
MRL		 1102 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1103 &ip->sin6.sin6_addr);
b1be00a6 1104 release_sock(sk);
e4c7ed41
CW		 1105#endif
1106 }
3fc2de2f 1107
56ef9c90 1108 return ret;
3fc2de2f 1109}
1110
1111/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1112static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1113{
b1be00a6 1114 struct sock *sk;
e4c7ed41
CW		 1115 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1116 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1117 int ret = -EINVAL;
3fc2de2f 1118
e4c7ed41
CW		 1119 if (ip->sa.sa_family == AF_INET) {
1120 struct ip_mreqn mreq = {
1121 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1122 .imr_ifindex = ifindex,
1123 };
1124
b1be00a6
JB		 1125 sk = vxlan->vn4_sock->sock->sk;
1126 lock_sock(sk);
56ef9c90 1127 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1128 release_sock(sk);
e4c7ed41
CW		 1129#if IS_ENABLED(CONFIG_IPV6)
1130 } else {
b1be00a6
JB		 1131 sk = vxlan->vn6_sock->sock->sk;
1132 lock_sock(sk);
56ef9c90
MRL		 1133 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1134 &ip->sin6.sin6_addr);
b1be00a6 1135 release_sock(sk);
e4c7ed41
CW		 1136#endif
1137 }
d342894c 1138
56ef9c90 1139 return ret;
d342894c 1140}
1141
f14ecebb
JB		 1142static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1143 struct sk_buff *skb, u32 vxflags)
dfd8645e
TH		 1144{
1145 size_t start, offset, plen;
dfd8645e 1146
f14ecebb
JB		 1147 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1148 goto out;
b7fe10e5 1149
f14ecebb
JB		 1150 start = vxlan_rco_start(unparsed->vx_vni);
1151 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1152
be5cfeab 1153 plen = sizeof(struct vxlanhdr) + offset + sizeof(u16);
dfd8645e
TH		 1154
1155 if (!pskb_may_pull(skb, plen))
be5cfeab 1156 return false;
dfd8645e 1157
be5cfeab
JB		 1158 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1159 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1160out:
1161 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1162 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1163 return true;
dfd8645e
TH		 1164}
1165
f14ecebb 1166static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1167 struct sk_buff *skb, u32 vxflags,
f14ecebb 1168 struct vxlan_metadata *md,
3288af08
JB		 1169 struct metadata_dst *tun_dst)
1170{
f14ecebb
JB		 1171 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1172
1173 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1174 goto out;
3288af08 1175
3288af08
JB		 1176 md->gbp = ntohs(gbp->policy_id);
1177
1178 if (tun_dst)
1179 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1180
1181 if (gbp->dont_learn)
1182 md->gbp |= VXLAN_GBP_DONT_LEARN;
1183
1184 if (gbp->policy_applied)
1185 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1186
64f87d36
JB		 1187 /* In flow-based mode, GBP is carried in dst_metadata */
1188 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1189 skb->mark = md->gbp;
f14ecebb
JB		 1190out:
1191 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1192}
1193
c9e78efb
JB		 1194static void vxlan_rcv(struct vxlan_dev *vxlan, struct vxlan_sock *vs,
1195 struct sk_buff *skb, struct vxlan_metadata *md,
614732ea
TG		 1196 struct metadata_dst *tun_dst)
1197{
1198 struct iphdr *oip = NULL;
1199 struct ipv6hdr *oip6 = NULL;
614732ea
TG		 1200 struct pcpu_sw_netstats *stats;
1201 union vxlan_addr saddr;
1202 int err = 0;
614732ea 1203
614732ea 1204 skb_reset_mac_header(skb);
614732ea
TG		 1205 skb->protocol = eth_type_trans(skb, vxlan->dev);
1206 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1207
1208 /* Ignore packet loops (and multicast echo) */
1209 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1210 goto drop;
1211
ce212d0f
JB		 1212 /* Get data from the outer IP header */
1213 if (vxlan_get_sk_family(vs) == AF_INET) {
614732ea
TG		 1214 oip = ip_hdr(skb);
1215 saddr.sin.sin_addr.s_addr = oip->saddr;
1216 saddr.sa.sa_family = AF_INET;
1217#if IS_ENABLED(CONFIG_IPV6)
1218 } else {
1219 oip6 = ipv6_hdr(skb);
1220 saddr.sin6.sin6_addr = oip6->saddr;
1221 saddr.sa.sa_family = AF_INET6;
1222#endif
1223 }
1224
1225 if (tun_dst) {
1226 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1227 tun_dst = NULL;
1228 }
1229
1230 if ((vxlan->flags & VXLAN_F_LEARN) &&
1231 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1232 goto drop;
1233
1234 skb_reset_network_header(skb);
614732ea
TG		 1235
1236 if (oip6)
1237 err = IP6_ECN_decapsulate(oip6, skb);
1238 if (oip)
1239 err = IP_ECN_decapsulate(oip, skb);
1240
1241 if (unlikely(err)) {
1242 if (log_ecn_error) {
1243 if (oip6)
1244 net_info_ratelimited("non-ECT from %pI6\n",
1245 &oip6->saddr);
1246 if (oip)
1247 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1248 &oip->saddr, oip->tos);
1249 }
1250 if (err > 1) {
1251 ++vxlan->dev->stats.rx_frame_errors;
1252 ++vxlan->dev->stats.rx_errors;
1253 goto drop;
1254 }
1255 }
1256
1257 stats = this_cpu_ptr(vxlan->dev->tstats);
1258 u64_stats_update_begin(&stats->syncp);
1259 stats->rx_packets++;
1260 stats->rx_bytes += skb->len;
1261 u64_stats_update_end(&stats->syncp);
1262
58ce31cc 1263 gro_cells_receive(&vxlan->gro_cells, skb);
614732ea
TG		 1264
1265 return;
1266drop:
1267 if (tun_dst)
1268 dst_release((struct dst_entry *)tun_dst);
1269
1270 /* Consume bad packet */
1271 kfree_skb(skb);
1272}
1273
d342894c 1274/* Callback from net/ipv4/udp.c to receive packets */
1275static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1276{
ee122c79 1277 struct metadata_dst *tun_dst = NULL;
c9e78efb 1278 struct vxlan_dev *vxlan;
5cfccc5a 1279 struct vxlan_sock *vs;
f14ecebb 1280 struct vxlanhdr unparsed;
ee122c79
TG		 1281 struct vxlan_metadata _md;
1282 struct vxlan_metadata *md = &_md;
d342894c 1283
d342894c 1284 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1285 if (!pskb_may_pull(skb, VXLAN_HLEN))
288b01c8 1286 return 1;
d342894c 1287
f14ecebb 1288 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1289 /* VNI flag always required to be set */
1290 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1291 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1292 ntohl(vxlan_hdr(skb)->vx_flags),
1293 ntohl(vxlan_hdr(skb)->vx_vni));
1294 /* Return non vxlan pkt */
1295 return 1;
d342894c 1296 }
288b01c8
JB		 1297 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1298 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1299
559835ea 1300 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1301 if (!vs)
d342894c 1302 goto drop;
d342894c 1303
c9e78efb
JB		 1304 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1305 if (!vxlan)
1306 goto drop;
1307
7f290c94
JB		 1308 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB),
1309 !net_eq(vxlan->net, dev_net(vxlan->dev))))
c9e78efb
JB		 1310 goto drop;
1311
ee122c79 1312 if (vxlan_collect_metadata(vs)) {
07dabf20
JB		 1313 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1314
c29a70d2 1315 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
07dabf20 1316 vxlan_vni_to_tun_id(vni), sizeof(*md));
c29a70d2 1317
ee122c79
TG		 1318 if (!tun_dst)
1319 goto drop;
1320
0f1b7354 1321 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
ee122c79
TG		 1322 } else {
1323 memset(md, 0, sizeof(*md));
1324 }
1325
3511494c
TG		 1326 /* For backwards compatibility, only allow reserved fields to be
1327 * used by VXLAN extensions if explicitly requested.
1328 */
f14ecebb
JB		 1329 if (vs->flags & VXLAN_F_REMCSUM_RX)
1330 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1331 goto drop;
1332 if (vs->flags & VXLAN_F_GBP)
64f87d36 1333 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md, tun_dst);
3511494c 1334
f14ecebb 1335 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1336 /* If there are any unprocessed flags remaining treat
1337 * this as a malformed packet. This behavior diverges from
1338 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1339 * in reserved fields are to be ignored. The approach here
c4b49512 1340 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1341 * is more robust and provides a little more security in
1342 * adding extensions to VXLAN.
1343 */
288b01c8 1344 goto drop;
3bf39475
TH		 1345 }
1346
c9e78efb 1347 vxlan_rcv(vxlan, vs, skb, md, tun_dst);
5cfccc5a
PS		 1348 return 0;
1349
1350drop:
ee122c79
TG		 1351 if (tun_dst)
1352 dst_release((struct dst_entry *)tun_dst);
1353
288b01c8
JB		 1354 /* Consume bad packet */
1355 kfree_skb(skb);
1356 return 0;
5cfccc5a
PS		 1357}
1358
e4f67add
DS		 1359static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1360{
1361 struct vxlan_dev *vxlan = netdev_priv(dev);
1362 struct arphdr *parp;
1363 u8 *arpptr, *sha;
1364 __be32 sip, tip;
1365 struct neighbour *n;
1366
1367 if (dev->flags & IFF_NOARP)
1368 goto out;
1369
1370 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1371 dev->stats.tx_dropped++;
1372 goto out;
1373 }
1374 parp = arp_hdr(skb);
1375
1376 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1377 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1378 parp->ar_pro != htons(ETH_P_IP) ||
1379 parp->ar_op != htons(ARPOP_REQUEST) ||
1380 parp->ar_hln != dev->addr_len ||
1381 parp->ar_pln != 4)
1382 goto out;
1383 arpptr = (u8 *)parp + sizeof(struct arphdr);
1384 sha = arpptr;
1385 arpptr += dev->addr_len; /* sha */
1386 memcpy(&sip, arpptr, sizeof(sip));
1387 arpptr += sizeof(sip);
1388 arpptr += dev->addr_len; /* tha */
1389 memcpy(&tip, arpptr, sizeof(tip));
1390
1391 if (ipv4_is_loopback(tip) ||
1392 ipv4_is_multicast(tip))
1393 goto out;
1394
1395 n = neigh_lookup(&arp_tbl, &tip, dev);
1396
1397 if (n) {
e4f67add
DS		 1398 struct vxlan_fdb *f;
1399 struct sk_buff *reply;
1400
1401 if (!(n->nud_state & NUD_CONNECTED)) {
1402 neigh_release(n);
1403 goto out;
1404 }
1405
1406 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1407 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1408 /* bridge-local neighbor */
1409 neigh_release(n);
1410 goto out;
1411 }
1412
1413 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1414 n->ha, sha);
1415
1416 neigh_release(n);
1417
7346135d
DS		 1418 if (reply == NULL)
1419 goto out;
1420
e4f67add
DS		 1421 skb_reset_mac_header(reply);
1422 __skb_pull(reply, skb_network_offset(reply));
1423 reply->ip_summed = CHECKSUM_UNNECESSARY;
1424 reply->pkt_type = PACKET_HOST;
1425
1426 if (netif_rx_ni(reply) == NET_RX_DROP)
1427 dev->stats.rx_dropped++;
e4c7ed41
CW		 1428 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1429 union vxlan_addr ipa = {
1430 .sin.sin_addr.s_addr = tip,
a45e92a5 1431 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1432 };
1433
1434 vxlan_ip_miss(dev, &ipa);
1435 }
e4f67add
DS		 1436out:
1437 consume_skb(skb);
1438 return NETDEV_TX_OK;
1439}
1440
f564f45c 1441#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1442static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1443 struct neighbour *n, bool isrouter)
1444{
1445 struct net_device *dev = request->dev;
1446 struct sk_buff *reply;
1447 struct nd_msg *ns, *na;
1448 struct ipv6hdr *pip6;
1449 u8 *daddr;
1450 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1451 int ns_olen;
1452 int i, len;
1453
1454 if (dev == NULL)
1455 return NULL;
1456
1457 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1458 sizeof(*na) + na_olen + dev->needed_tailroom;
1459 reply = alloc_skb(len, GFP_ATOMIC);
1460 if (reply == NULL)
1461 return NULL;
1462
1463 reply->protocol = htons(ETH_P_IPV6);
1464 reply->dev = dev;
1465 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1466 skb_push(reply, sizeof(struct ethhdr));
1467 skb_set_mac_header(reply, 0);
1468
1469 ns = (struct nd_msg *)skb_transport_header(request);
1470
1471 daddr = eth_hdr(request)->h_source;
1472 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1473 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1474 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1475 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1476 break;
1477 }
1478 }
1479
1480 /* Ethernet header */
1481 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1482 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1483 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1484 reply->protocol = htons(ETH_P_IPV6);
1485
1486 skb_pull(reply, sizeof(struct ethhdr));
1487 skb_set_network_header(reply, 0);
1488 skb_put(reply, sizeof(struct ipv6hdr));
1489
1490 /* IPv6 header */
1491
1492 pip6 = ipv6_hdr(reply);
1493 memset(pip6, 0, sizeof(struct ipv6hdr));
1494 pip6->version = 6;
1495 pip6->priority = ipv6_hdr(request)->priority;
1496 pip6->nexthdr = IPPROTO_ICMPV6;
1497 pip6->hop_limit = 255;
1498 pip6->daddr = ipv6_hdr(request)->saddr;
1499 pip6->saddr = *(struct in6_addr *)n->primary_key;
1500
1501 skb_pull(reply, sizeof(struct ipv6hdr));
1502 skb_set_transport_header(reply, 0);
1503
1504 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1505
1506 /* Neighbor Advertisement */
1507 memset(na, 0, sizeof(*na)+na_olen);
1508 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1509 na->icmph.icmp6_router = isrouter;
1510 na->icmph.icmp6_override = 1;
1511 na->icmph.icmp6_solicited = 1;
1512 na->target = ns->target;
1513 ether_addr_copy(&na->opt[2], n->ha);
1514 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1515 na->opt[1] = na_olen >> 3;
1516
1517 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1518 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1519 csum_partial(na, sizeof(*na)+na_olen, 0));
1520
1521 pip6->payload_len = htons(sizeof(*na)+na_olen);
1522
1523 skb_push(reply, sizeof(struct ipv6hdr));
1524
1525 reply->ip_summed = CHECKSUM_UNNECESSARY;
1526
1527 return reply;
1528}
1529
f564f45c
CW		 1530static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1531{
1532 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1533 struct nd_msg *msg;
f564f45c
CW		 1534 const struct ipv6hdr *iphdr;
1535 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1536 struct neighbour *n;
1537 struct inet6_dev *in6_dev;
f564f45c
CW		 1538
1539 in6_dev = __in6_dev_get(dev);
1540 if (!in6_dev)
1541 goto out;
1542
f564f45c
CW		 1543 iphdr = ipv6_hdr(skb);
1544 saddr = &iphdr->saddr;
1545 daddr = &iphdr->daddr;
1546
f564f45c
CW		 1547 msg = (struct nd_msg *)skb_transport_header(skb);
1548 if (msg->icmph.icmp6_code != 0 ||
1549 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1550 goto out;
1551
4b29dba9
DS		 1552 if (ipv6_addr_loopback(daddr) ||
1553 ipv6_addr_is_multicast(&msg->target))
1554 goto out;
1555
1556 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1557
1558 if (n) {
1559 struct vxlan_fdb *f;
4b29dba9 1560 struct sk_buff *reply;
f564f45c
CW		 1561
1562 if (!(n->nud_state & NUD_CONNECTED)) {
1563 neigh_release(n);
1564 goto out;
1565 }
1566
1567 f = vxlan_find_mac(vxlan, n->ha);
1568 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1569 /* bridge-local neighbor */
1570 neigh_release(n);
1571 goto out;
1572 }
1573
4b29dba9
DS		 1574 reply = vxlan_na_create(skb, n,
1575 !!(f ? f->flags & NTF_ROUTER : 0));
1576
f564f45c 1577 neigh_release(n);
4b29dba9
DS		 1578
1579 if (reply == NULL)
1580 goto out;
1581
1582 if (netif_rx_ni(reply) == NET_RX_DROP)
1583 dev->stats.rx_dropped++;
1584
f564f45c 1585 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1586 union vxlan_addr ipa = {
1587 .sin6.sin6_addr = msg->target,
a45e92a5 1588 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1589 };
1590
f564f45c
CW		 1591 vxlan_ip_miss(dev, &ipa);
1592 }
1593
1594out:
1595 consume_skb(skb);
1596 return NETDEV_TX_OK;
1597}
1598#endif
1599
e4f67add
DS		 1600static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1601{
1602 struct vxlan_dev *vxlan = netdev_priv(dev);
1603 struct neighbour *n;
e4f67add
DS		 1604
1605 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1606 return false;
1607
1608 n = NULL;
1609 switch (ntohs(eth_hdr(skb)->h_proto)) {
1610 case ETH_P_IP:
e15a00aa
CW		 1611 {
1612 struct iphdr *pip;
1613
e4f67add
DS		 1614 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1615 return false;
1616 pip = ip_hdr(skb);
1617 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1618 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1619 union vxlan_addr ipa = {
1620 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1621 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1622 };
1623
1624 vxlan_ip_miss(dev, &ipa);
1625 return false;
1626 }
1627
e4f67add 1628 break;
e15a00aa
CW		 1629 }
1630#if IS_ENABLED(CONFIG_IPV6)
1631 case ETH_P_IPV6:
1632 {
1633 struct ipv6hdr *pip6;
1634
1635 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1636 return false;
1637 pip6 = ipv6_hdr(skb);
1638 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1639 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1640 union vxlan_addr ipa = {
1641 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1642 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1643 };
1644
1645 vxlan_ip_miss(dev, &ipa);
1646 return false;
1647 }
1648
1649 break;
1650 }
1651#endif
e4f67add
DS		 1652 default:
1653 return false;
1654 }
1655
1656 if (n) {
1657 bool diff;
1658
7367d0b5 1659 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1660 if (diff) {
1661 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1662 dev->addr_len);
1663 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1664 }
1665 neigh_release(n);
1666 return diff;
e4c7ed41
CW		 1667 }
1668
e4f67add
DS		 1669 return false;
1670}
1671
af33c1ad 1672static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1673 struct vxlan_metadata *md)
1674{
1675 struct vxlanhdr_gbp *gbp;
1676
db79a621
TG		 1677 if (!md->gbp)
1678 return;
1679
3511494c 1680 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1681 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1682
1683 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1684 gbp->dont_learn = 1;
1685
1686 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1687 gbp->policy_applied = 1;
1688
1689 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1690}
1691
f491e56d
JB		 1692static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1693 int iphdr_len, __be32 vni,
1694 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1695 bool udp_sum)
e4c7ed41 1696{
e4c7ed41 1697 struct vxlanhdr *vxh;
e4c7ed41
CW		 1698 int min_headroom;
1699 int err;
dfd8645e 1700 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
dfd8645e 1701
af33c1ad 1702 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1703 skb->ip_summed == CHECKSUM_PARTIAL) {
1704 int csum_start = skb_checksum_start_offset(skb);
1705
1706 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1707 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1708 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1709 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1710 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1711 }
e4c7ed41 1712
e4c7ed41 1713 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1714 + VXLAN_HLEN + iphdr_len
df8a39de 1715 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1716
1717 /* Need space for new headers (invalidates iph ptr) */
1718 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1719 if (unlikely(err)) {
1720 kfree_skb(skb);
649c5b8b 1721 return err;
74f47278 1722 }
649c5b8b 1723
5968250c
JP		 1724 skb = vlan_hwaccel_push_inside(skb);
1725 if (WARN_ON(!skb))
1726 return -ENOMEM;
1eaa8178 1727
6fa79666 1728 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG		 1729 if (IS_ERR(skb))
1730 return PTR_ERR(skb);
1731
49560532 1732 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1733 vxh->vx_flags = VXLAN_HF_VNI;
1734 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1735
dfd8645e 1736 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1737 unsigned int start;
dfd8645e 1738
54bfd872
JB		 1739 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1740 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1741 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1742
1743 if (!skb_is_gso(skb)) {
1744 skb->ip_summed = CHECKSUM_NONE;
1745 skb->encapsulation = 0;
1746 }
1747 }
1748
af33c1ad
TH		 1749 if (vxflags & VXLAN_F_GBP)
1750 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1751
996c9fd1 1752 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
039f5062 1753 return 0;
49560532 1754}
49560532 1755
1a8496ba
JB		 1756static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1757 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1758 __be32 daddr, __be32 *saddr,
1759 struct dst_cache *dst_cache,
1760 struct ip_tunnel_info *info)
1a8496ba
JB		 1761{
1762 struct rtable *rt = NULL;
0c1d70af 1763 bool use_cache = false;
1a8496ba
JB		 1764 struct flowi4 fl4;
1765
0c1d70af
PA		 1766 /* when the ip_tunnel_info is availble, the tos used for lookup is
1767 * packet independent, so we can use the cache
1768 */
d71785ff 1769 if (!skb->mark && (!tos || info)) {
0c1d70af
PA		 1770 use_cache = true;
1771 rt = dst_cache_get_ip4(dst_cache, saddr);
1772 if (rt)
1773 return rt;
1774 }
1775
1a8496ba
JB		 1776 memset(&fl4, 0, sizeof(fl4));
1777 fl4.flowi4_oif = oif;
1778 fl4.flowi4_tos = RT_TOS(tos);
1779 fl4.flowi4_mark = skb->mark;
1780 fl4.flowi4_proto = IPPROTO_UDP;
1781 fl4.daddr = daddr;
1782 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1783
1784 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1785 if (!IS_ERR(rt)) {
1a8496ba 1786 *saddr = fl4.saddr;
0c1d70af
PA		 1787 if (use_cache)
1788 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1789 }
1a8496ba
JB		 1790 return rt;
1791}
1792
e5d4b29f
JB		 1793#if IS_ENABLED(CONFIG_IPV6)
1794static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1795 struct sk_buff *skb, int oif,
1796 const struct in6_addr *daddr,
0c1d70af
PA		 1797 struct in6_addr *saddr,
1798 struct dst_cache *dst_cache)
e5d4b29f
JB		 1799{
1800 struct dst_entry *ndst;
1801 struct flowi6 fl6;
1802 int err;
1803
d71785ff 1804 if (!skb->mark) {
0c1d70af
PA		 1805 ndst = dst_cache_get_ip6(dst_cache, saddr);
1806 if (ndst)
1807 return ndst;
1808 }
1809
e5d4b29f
JB		 1810 memset(&fl6, 0, sizeof(fl6));
1811 fl6.flowi6_oif = oif;
1812 fl6.daddr = *daddr;
1813 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1814 fl6.flowi6_mark = skb->mark;
1815 fl6.flowi6_proto = IPPROTO_UDP;
1816
1817 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1818 vxlan->vn6_sock->sock->sk,
1819 &ndst, &fl6);
1820 if (err < 0)
1821 return ERR_PTR(err);
1822
1823 *saddr = fl6.saddr;
d71785ff 1824 if (!skb->mark)
0c1d70af 1825 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1826 return ndst;
1827}
1828#endif
1829
9dcc71e1
SS		 1830/* Bypass encapsulation if the destination is local */
1831static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1832 struct vxlan_dev *dst_vxlan)
1833{
8f84985f 1834 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1835 union vxlan_addr loopback;
1836 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1837 struct net_device *dev = skb->dev;
1838 int len = skb->len;
9dcc71e1 1839
8f84985f
LR		 1840 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1841 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1842 skb->pkt_type = PACKET_HOST;
1843 skb->encapsulation = 0;
1844 skb->dev = dst_vxlan->dev;
1845 __skb_pull(skb, skb_network_offset(skb));
1846
e4c7ed41
CW		 1847 if (remote_ip->sa.sa_family == AF_INET) {
1848 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1849 loopback.sa.sa_family = AF_INET;
1850#if IS_ENABLED(CONFIG_IPV6)
1851 } else {
1852 loopback.sin6.sin6_addr = in6addr_loopback;
1853 loopback.sa.sa_family = AF_INET6;
1854#endif
1855 }
1856
9dcc71e1 1857 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1858 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1859
1860 u64_stats_update_begin(&tx_stats->syncp);
1861 tx_stats->tx_packets++;
ce6502a8 1862 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1863 u64_stats_update_end(&tx_stats->syncp);
1864
1865 if (netif_rx(skb) == NET_RX_SUCCESS) {
1866 u64_stats_update_begin(&rx_stats->syncp);
1867 rx_stats->rx_packets++;
ce6502a8 1868 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1869 u64_stats_update_end(&rx_stats->syncp);
1870 } else {
ce6502a8 1871 dev->stats.rx_dropped++;
9dcc71e1
SS		 1872 }
1873}
1874
4ad16930
SH		 1875static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1876 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1877{
d71785ff 1878 struct dst_cache *dst_cache;
3093fbe7 1879 struct ip_tunnel_info *info;
d342894c 1880 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1881 struct sock *sk;
e4c7ed41 1882 struct rtable *rt = NULL;
d342894c 1883 const struct iphdr *old_iph;
e4c7ed41 1884 union vxlan_addr *dst;
ee122c79
TG		 1885 union vxlan_addr remote_ip;
1886 struct vxlan_metadata _md;
1887 struct vxlan_metadata *md = &_md;
e4c7ed41 1888 __be16 src_port = 0, dst_port;
54bfd872 1889 __be32 vni;
d342894c 1890 __be16 df = 0;
1891 __u8 tos, ttl;
0e6fbc5b 1892 int err;
ee122c79 1893 u32 flags = vxlan->flags;
b4ed5cad 1894 bool udp_sum = false;
f491e56d 1895 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1896
61adedf3 1897 info = skb_tunnel_info(skb);
3093fbe7 1898
ee122c79 1899 if (rdst) {
0dfbdf41 1900 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1901 vni = rdst->remote_vni;
1902 dst = &rdst->remote_ip;
d71785ff 1903 dst_cache = &rdst->dst_cache;
ee122c79
TG		 1904 } else {
1905 if (!info) {
1906 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1907 dev->name);
1908 goto drop;
1909 }
0dfbdf41 1910 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
54bfd872 1911 vni = vxlan_tun_id_to_vni(info->key.tun_id);
b1be00a6
JB		 1912 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1913 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1914 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1915 else
1916 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1917 dst = &remote_ip;
d71785ff 1918 dst_cache = &info->dst_cache;
ee122c79 1919 }
e4f67add 1920
e4c7ed41 1921 if (vxlan_addr_any(dst)) {
e4f67add 1922 if (did_rsc) {
e4f67add 1923 /* short-circuited back to local bridge */
9dcc71e1 1924 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1925 return;
e4f67add 1926 }
ef59febe 1927 goto drop;
e4f67add 1928 }
ef59febe 1929
d342894c 1930 old_iph = ip_hdr(skb);
1931
0dfbdf41 1932 ttl = vxlan->cfg.ttl;
e4c7ed41 1933 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1934 ttl = 1;
1935
0dfbdf41 1936 tos = vxlan->cfg.tos;
d342894c 1937 if (tos == 1)
206aaafc 1938 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1939
0dfbdf41
TG		 1940 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1941 vxlan->cfg.port_max, true);
d342894c 1942
a725e514 1943 if (info) {
a725e514
JB		 1944 ttl = info->key.ttl;
1945 tos = info->key.tos;
b4ed5cad 1946 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB		 1947
1948 if (info->options_len)
4c222798 1949 md = ip_tunnel_info_opts(info);
a725e514
JB		 1950 } else {
1951 md->gbp = skb->mark;
1952 }
1953
e4c7ed41 1954 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB		 1955 __be32 saddr;
1956
b1be00a6
JB		 1957 if (!vxlan->vn4_sock)
1958 goto drop;
1959 sk = vxlan->vn4_sock->sock->sk;
1960
1a8496ba
JB		 1961 rt = vxlan_get_route(vxlan, skb,
1962 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 1963 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 1964 dst_cache, info);
e4c7ed41
CW		 1965 if (IS_ERR(rt)) {
1966 netdev_dbg(dev, "no route to %pI4\n",
1967 &dst->sin.sin_addr.s_addr);
1968 dev->stats.tx_carrier_errors++;
1969 goto tx_error;
1970 }
d342894c 1971
e4c7ed41
CW		 1972 if (rt->dst.dev == dev) {
1973 netdev_dbg(dev, "circular route to %pI4\n",
1974 &dst->sin.sin_addr.s_addr);
1975 dev->stats.collisions++;
fffc15a5 1976 goto rt_tx_error;
e4c7ed41
CW		 1977 }
1978
1979 /* Bypass encapsulation if the destination is local */
1980 if (rt->rt_flags & RTCF_LOCAL &&
1981 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1982 struct vxlan_dev *dst_vxlan;
1983
1984 ip_rt_put(rt);
19ca9fc1 1985 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1986 dst->sa.sa_family, dst_port,
1987 vxlan->flags);
e4c7ed41
CW		 1988 if (!dst_vxlan)
1989 goto tx_error;
1990 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1991 return;
1992 }
1993
6ceb31ca
AD		 1994 if (!info)
1995 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
1996 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1997 df = htons(IP_DF);
1998
e4c7ed41
CW		 1999 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2000 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d 2001 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
54bfd872 2002 vni, md, flags, udp_sum);
f491e56d
JB		 2003 if (err < 0)
2004 goto xmit_tx_error;
2005
2006 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2007 dst->sin.sin_addr.s_addr, tos, ttl, df,
2008 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2009#if IS_ENABLED(CONFIG_IPV6)
2010 } else {
e4c7ed41 2011 struct dst_entry *ndst;
e5d4b29f 2012 struct in6_addr saddr;
6f264e47 2013 u32 rt6i_flags;
e4c7ed41 2014
b1be00a6
JB		 2015 if (!vxlan->vn6_sock)
2016 goto drop;
2017 sk = vxlan->vn6_sock->sock->sk;
2018
e5d4b29f
JB		 2019 ndst = vxlan6_get_route(vxlan, skb,
2020 rdst ? rdst->remote_ifindex : 0,
0c1d70af 2021 &dst->sin6.sin6_addr, &saddr,
d71785ff 2022 dst_cache);
e5d4b29f 2023 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2024 netdev_dbg(dev, "no route to %pI6\n",
2025 &dst->sin6.sin6_addr);
2026 dev->stats.tx_carrier_errors++;
9dcc71e1 2027 goto tx_error;
e4c7ed41 2028 }
d342894c 2029
e4c7ed41
CW		 2030 if (ndst->dev == dev) {
2031 netdev_dbg(dev, "circular route to %pI6\n",
2032 &dst->sin6.sin6_addr);
2033 dst_release(ndst);
2034 dev->stats.collisions++;
2035 goto tx_error;
2036 }
0e6fbc5b 2037
e4c7ed41 2038 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2039 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2040 if (rt6i_flags & RTF_LOCAL &&
2041 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2042 struct vxlan_dev *dst_vxlan;
2043
2044 dst_release(ndst);
19ca9fc1 2045 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2046 dst->sa.sa_family, dst_port,
2047 vxlan->flags);
e4c7ed41
CW		 2048 if (!dst_vxlan)
2049 goto tx_error;
2050 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2051 return;
2052 }
49560532 2053
b4ed5cad
JB		 2054 if (!info)
2055 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2056
e4c7ed41 2057 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2058 skb_scrub_packet(skb, xnet);
2059 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2060 vni, md, flags, udp_sum);
f491e56d
JB		 2061 if (err < 0) {
2062 dst_release(ndst);
2063 return;
2064 }
2065 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
2066 &saddr, &dst->sin6.sin6_addr,
2067 0, ttl, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2068#endif
2069 }
0e6fbc5b 2070
4ad16930 2071 return;
d342894c 2072
2073drop:
2074 dev->stats.tx_dropped++;
2075 goto tx_free;
2076
f491e56d
JB		 2077xmit_tx_error:
2078 /* skb is already freed. */
2079 skb = NULL;
49560532
PS		 2080rt_tx_error:
2081 ip_rt_put(rt);
d342894c 2082tx_error:
2083 dev->stats.tx_errors++;
2084tx_free:
2085 dev_kfree_skb(skb);
d342894c 2086}
2087
6681712d
DS		 2088/* Transmit local packets over Vxlan
2089 *
2090 * Outer IP header inherits ECN and DF from inner header.
2091 * Outer UDP destination is the VXLAN assigned port.
2092 * source port is based on hash of flow
2093 */
2094static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2095{
2096 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2097 const struct ip_tunnel_info *info;
6681712d
DS		 2098 struct ethhdr *eth;
2099 bool did_rsc = false;
8f646c92 2100 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2101 struct vxlan_fdb *f;
6681712d 2102
61adedf3 2103 info = skb_tunnel_info(skb);
3093fbe7 2104
6681712d
DS		 2105 skb_reset_mac_header(skb);
2106 eth = eth_hdr(skb);
2107
f564f45c
CW		 2108 if ((vxlan->flags & VXLAN_F_PROXY)) {
2109 if (ntohs(eth->h_proto) == ETH_P_ARP)
2110 return arp_reduce(dev, skb);
2111#if IS_ENABLED(CONFIG_IPV6)
2112 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2113 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2114 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2115 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2116 struct nd_msg *msg;
2117
2118 msg = (struct nd_msg *)skb_transport_header(skb);
2119 if (msg->icmph.icmp6_code == 0 &&
2120 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2121 return neigh_reduce(dev, skb);
2122 }
7a9f526f 2123 eth = eth_hdr(skb);
f564f45c
CW		 2124#endif
2125 }
6681712d 2126
f468a729
JB		 2127 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2128 if (info && info->mode & IP_TUNNEL_INFO_TX)
2129 vxlan_xmit_one(skb, dev, NULL, false);
2130 else
2131 kfree_skb(skb);
ee122c79
TG		 2132 return NETDEV_TX_OK;
2133 }
2134
6681712d 2135 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2136 did_rsc = false;
2137
2138 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2139 (ntohs(eth->h_proto) == ETH_P_IP ||
2140 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2141 did_rsc = route_shortcircuit(dev, skb);
2142 if (did_rsc)
2143 f = vxlan_find_mac(vxlan, eth->h_dest);
2144 }
2145
6681712d 2146 if (f == NULL) {
afbd8bae
MR		 2147 f = vxlan_find_mac(vxlan, all_zeros_mac);
2148 if (f == NULL) {
2149 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2150 !is_multicast_ether_addr(eth->h_dest))
2151 vxlan_fdb_miss(vxlan, eth->h_dest);
2152
2153 dev->stats.tx_dropped++;
8f646c92 2154 kfree_skb(skb);
afbd8bae
MR		 2155 return NETDEV_TX_OK;
2156 }
2157 }
6681712d 2158
afbd8bae
MR		 2159 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2160 struct sk_buff *skb1;
6681712d 2161
8f646c92
ED		 2162 if (!fdst) {
2163 fdst = rdst;
2164 continue;
2165 }
afbd8bae
MR		 2166 skb1 = skb_clone(skb, GFP_ATOMIC);
2167 if (skb1)
2168 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2169 }
2170
8f646c92
ED		 2171 if (fdst)
2172 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2173 else
2174 kfree_skb(skb);
4ad16930 2175 return NETDEV_TX_OK;
6681712d
DS		 2176}
2177
d342894c 2178/* Walk the forwarding table and purge stale entries */
2179static void vxlan_cleanup(unsigned long arg)
2180{
2181 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2182 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2183 unsigned int h;
2184
2185 if (!netif_running(vxlan->dev))
2186 return;
2187
d342894c 2188 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2189 struct hlist_node *p, *n;
14e1d0fa
SD		 2190
2191 spin_lock_bh(&vxlan->hash_lock);
d342894c 2192 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2193 struct vxlan_fdb *f
2194 = container_of(p, struct vxlan_fdb, hlist);
2195 unsigned long timeout;
2196
3c172868 2197 if (f->state & NUD_PERMANENT)
d342894c 2198 continue;
2199
0dfbdf41 2200 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2201 if (time_before_eq(timeout, jiffies)) {
2202 netdev_dbg(vxlan->dev,
2203 "garbage collect %pM\n",
2204 f->eth_addr);
2205 f->state = NUD_STALE;
2206 vxlan_fdb_destroy(vxlan, f);
2207 } else if (time_before(timeout, next_timer))
2208 next_timer = timeout;
2209 }
14e1d0fa 2210 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2211 }
d342894c 2212
2213 mod_timer(&vxlan->age_timer, next_timer);
2214}
2215
9c2e24e1
PS		 2216static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2217{
56ef9c90 2218 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2219 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2220
56ef9c90 2221 spin_lock(&vn->sock_lock);
9c2e24e1 2222 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2223 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2224}
2225
d342894c 2226/* Setup stats when device is created */
2227static int vxlan_init(struct net_device *dev)
2228{
1c213bd2 2229 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2230 if (!dev->tstats)
d342894c 2231 return -ENOMEM;
2232
2233 return 0;
2234}
2235
ba609e9b 2236static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2237{
2238 struct vxlan_fdb *f;
2239
2240 spin_lock_bh(&vxlan->hash_lock);
2241 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2242 if (f)
2243 vxlan_fdb_destroy(vxlan, f);
2244 spin_unlock_bh(&vxlan->hash_lock);
2245}
2246
ebf4063e
SH		 2247static void vxlan_uninit(struct net_device *dev)
2248{
2249 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2250
ba609e9b 2251 vxlan_fdb_delete_default(vxlan);
afbd8bae 2252
ebf4063e
SH		 2253 free_percpu(dev->tstats);
2254}
2255
d342894c 2256/* Start ageing timer and join group when device is brought up */
2257static int vxlan_open(struct net_device *dev)
2258{
2259 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2260 int ret;
56ef9c90 2261
205f356d
JB		 2262 ret = vxlan_sock_add(vxlan);
2263 if (ret < 0)
2264 return ret;
d342894c 2265
79d4a94f 2266 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2267 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2268 if (ret == -EADDRINUSE)
2269 ret = 0;
56ef9c90 2270 if (ret) {
205f356d 2271 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2272 return ret;
2273 }
d342894c 2274 }
2275
0dfbdf41 2276 if (vxlan->cfg.age_interval)
d342894c 2277 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2278
56ef9c90 2279 return ret;
d342894c 2280}
2281
2282/* Purge the forwarding table */
2283static void vxlan_flush(struct vxlan_dev *vxlan)
2284{
31fec5aa 2285 unsigned int h;
d342894c 2286
2287 spin_lock_bh(&vxlan->hash_lock);
2288 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2289 struct hlist_node *p, *n;
2290 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2291 struct vxlan_fdb *f
2292 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2293 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2294 if (!is_zero_ether_addr(f->eth_addr))
2295 vxlan_fdb_destroy(vxlan, f);
d342894c 2296 }
2297 }
2298 spin_unlock_bh(&vxlan->hash_lock);
2299}
2300
2301/* Cleanup timer and forwarding table on shutdown */
2302static int vxlan_stop(struct net_device *dev)
2303{
2304 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2305 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2306 int ret = 0;
d342894c 2307
24c0e683 2308 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2309 !vxlan_group_used(vn, vxlan))
56ef9c90 2310 ret = vxlan_igmp_leave(vxlan);
d342894c 2311
2312 del_timer_sync(&vxlan->age_timer);
2313
2314 vxlan_flush(vxlan);
205f356d 2315 vxlan_sock_release(vxlan);
d342894c 2316
56ef9c90 2317 return ret;
d342894c 2318}
2319
d342894c 2320/* Stub, nothing needs to be done. */
2321static void vxlan_set_multicast_list(struct net_device *dev)
2322{
2323}
2324
72564b59
DW		 2325static int __vxlan_change_mtu(struct net_device *dev,
2326 struct net_device *lowerdev,
2327 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2328{
72564b59 2329 int max_mtu = IP_MAX_MTU;
345010b5 2330
72564b59
DW		 2331 if (lowerdev)
2332 max_mtu = lowerdev->mtu;
345010b5
DB		 2333
2334 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2335 max_mtu -= VXLAN6_HEADROOM;
345010b5 2336 else
72564b59 2337 max_mtu -= VXLAN_HEADROOM;
345010b5 2338
72564b59 2339 if (new_mtu < 68)
345010b5
DB		 2340 return -EINVAL;
2341
72564b59
DW		 2342 if (new_mtu > max_mtu) {
2343 if (strict)
2344 return -EINVAL;
2345
2346 new_mtu = max_mtu;
2347 }
2348
345010b5
DB		 2349 dev->mtu = new_mtu;
2350 return 0;
2351}
2352
72564b59
DW		 2353static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2354{
2355 struct vxlan_dev *vxlan = netdev_priv(dev);
2356 struct vxlan_rdst *dst = &vxlan->default_dst;
2357 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2358 dst->remote_ifindex);
2359 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2360}
2361
fc4099f1
PS		 2362static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2363{
2364 struct vxlan_dev *vxlan = netdev_priv(dev);
2365 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2366 __be16 sport, dport;
2367
2368 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2369 vxlan->cfg.port_max, true);
2370 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2371
239e944f 2372 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB		 2373 struct rtable *rt;
2374
239e944f
JB		 2375 if (!vxlan->vn4_sock)
2376 return -EINVAL;
1a8496ba
JB		 2377 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2378 info->key.u.ipv4.dst,
0c1d70af 2379 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2380 if (IS_ERR(rt))
2381 return PTR_ERR(rt);
2382 ip_rt_put(rt);
239e944f
JB		 2383 } else {
2384#if IS_ENABLED(CONFIG_IPV6)
2385 struct dst_entry *ndst;
2386
2387 if (!vxlan->vn6_sock)
2388 return -EINVAL;
2389 ndst = vxlan6_get_route(vxlan, skb, 0,
2390 &info->key.u.ipv6.dst,
0c1d70af 2391 &info->key.u.ipv6.src, NULL);
239e944f
JB		 2392 if (IS_ERR(ndst))
2393 return PTR_ERR(ndst);
2394 dst_release(ndst);
239e944f
JB		 2395#else /* !CONFIG_IPV6 */
2396 return -EPFNOSUPPORT;
2397#endif
2398 }
1a8496ba
JB		 2399 info->key.tp_src = sport;
2400 info->key.tp_dst = dport;
239e944f 2401 return 0;
fc4099f1
PS		 2402}
2403
d342894c 2404static const struct net_device_ops vxlan_netdev_ops = {
2405 .ndo_init = vxlan_init,
ebf4063e 2406 .ndo_uninit = vxlan_uninit,
d342894c 2407 .ndo_open = vxlan_open,
2408 .ndo_stop = vxlan_stop,
2409 .ndo_start_xmit = vxlan_xmit,
e8171045 2410 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2411 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2412 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2413 .ndo_validate_addr = eth_validate_addr,
2414 .ndo_set_mac_address = eth_mac_addr,
2415 .ndo_fdb_add = vxlan_fdb_add,
2416 .ndo_fdb_del = vxlan_fdb_delete,
2417 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2418 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2419};
2420
2421/* Info for udev, that this is a virtual tunnel endpoint */
2422static struct device_type vxlan_type = {
2423 .name = "vxlan",
2424};
2425
53cf5275 2426/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2427 * supply the listening VXLAN udp ports. Callers are expected
2428 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2429 */
2430void vxlan_get_rx_port(struct net_device *dev)
2431{
2432 struct vxlan_sock *vs;
2433 struct net *net = dev_net(dev);
2434 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2435 sa_family_t sa_family;
35e42379
JG		 2436 __be16 port;
2437 unsigned int i;
53cf5275
JG		 2438
2439 spin_lock(&vn->sock_lock);
2440 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2441 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2442 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2443 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2444 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2445 port);
2446 }
2447 }
2448 spin_unlock(&vn->sock_lock);
2449}
2450EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2451
d342894c 2452/* Initialize the device structure. */
2453static void vxlan_setup(struct net_device *dev)
2454{
2455 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2456 unsigned int h;
d342894c 2457
2458 eth_hw_addr_random(dev);
2459 ether_setup(dev);
2460
2461 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2462 dev->destructor = free_netdev;
d342894c 2463 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2464
d342894c 2465 dev->features |= NETIF_F_LLTX;
d6727fe3 2466 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2467 dev->features |= NETIF_F_RXCSUM;
05c0db08 2468 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2469
1eaa8178
PS		 2470 dev->vlan_features = dev->features;
2471 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2472 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2473 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2474 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2475 netif_keep_dst(dev);
82a0f6b4 2476 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
22dba393 2477 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2478
553675fb 2479 INIT_LIST_HEAD(&vxlan->next);
d342894c 2480 spin_lock_init(&vxlan->hash_lock);
2481
2482 init_timer_deferrable(&vxlan->age_timer);
2483 vxlan->age_timer.function = vxlan_cleanup;
2484 vxlan->age_timer.data = (unsigned long) vxlan;
2485
0dfbdf41 2486 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2487
d342894c 2488 vxlan->dev = dev;
2489
58ce31cc
TH		 2490 gro_cells_init(&vxlan->gro_cells, dev);
2491
d342894c 2492 for (h = 0; h < FDB_HASH_SIZE; ++h)
2493 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2494}
2495
2496static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2497 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2498 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2499 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2500 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2501 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2502 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2503 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2504 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2505 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2506 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2507 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2508 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2509 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2510 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2511 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2512 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2513 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2514 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2515 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2516 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2517 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2518 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2519 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2520 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2521 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2522};
2523
2524static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2525{
2526 if (tb[IFLA_ADDRESS]) {
2527 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2528 pr_debug("invalid link address (not ethernet)\n");
2529 return -EINVAL;
2530 }
2531
2532 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2533 pr_debug("invalid all zero ethernet address\n");
2534 return -EADDRNOTAVAIL;
2535 }
2536 }
2537
2538 if (!data)
2539 return -EINVAL;
2540
2541 if (data[IFLA_VXLAN_ID]) {
2542 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2543 if (id >= VXLAN_VID_MASK)
2544 return -ERANGE;
2545 }
2546
05f47d69 2547 if (data[IFLA_VXLAN_PORT_RANGE]) {
2548 const struct ifla_vxlan_port_range *p
2549 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2550
2551 if (ntohs(p->high) < ntohs(p->low)) {
2552 pr_debug("port range %u .. %u not valid\n",
2553 ntohs(p->low), ntohs(p->high));
2554 return -EINVAL;
2555 }
2556 }
2557
d342894c 2558 return 0;
2559}
2560
1b13c97f
YB		 2561static void vxlan_get_drvinfo(struct net_device *netdev,
2562 struct ethtool_drvinfo *drvinfo)
2563{
2564 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2565 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2566}
2567
2568static const struct ethtool_ops vxlan_ethtool_ops = {
2569 .get_drvinfo = vxlan_get_drvinfo,
2570 .get_link = ethtool_op_get_link,
2571};
2572
553675fb 2573static void vxlan_del_work(struct work_struct *work)
2574{
2575 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2576 udp_tunnel_sock_release(vs->sock);
553675fb 2577 kfree_rcu(vs, rcu);
2578}
2579
3ee64f39
TH		 2580static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2581 __be16 port, u32 flags)
553675fb 2582{
e4c7ed41 2583 struct socket *sock;
3ee64f39
TH		 2584 struct udp_port_cfg udp_conf;
2585 int err;
e4c7ed41 2586
3ee64f39 2587 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2588
3ee64f39
TH		 2589 if (ipv6) {
2590 udp_conf.family = AF_INET6;
3ee64f39 2591 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2592 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2593 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2594 } else {
2595 udp_conf.family = AF_INET;
e4c7ed41 2596 }
e4c7ed41 2597
3ee64f39 2598 udp_conf.local_udp_port = port;
553675fb 2599
3ee64f39
TH		 2600 /* Open UDP socket */
2601 err = udp_sock_create(net, &udp_conf, &sock);
2602 if (err < 0)
2603 return ERR_PTR(err);
e4c7ed41 2604
39deb2c7 2605 return sock;
e4c7ed41
CW		 2606}
2607
2608/* Create new listen socket if needed */
b1be00a6
JB		 2609static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2610 __be16 port, u32 flags)
e4c7ed41
CW		 2611{
2612 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2613 struct vxlan_sock *vs;
2614 struct socket *sock;
e4c7ed41 2615 unsigned int h;
acbf74a7 2616 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2617
dc01e7d3 2618 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2619 if (!vs)
2620 return ERR_PTR(-ENOMEM);
2621
2622 for (h = 0; h < VNI_HASH_SIZE; ++h)
2623 INIT_HLIST_HEAD(&vs->vni_list[h]);
2624
2625 INIT_WORK(&vs->del_work, vxlan_del_work);
2626
3ee64f39 2627 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2628 if (IS_ERR(sock)) {
56ef9c90
MRL		 2629 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2630 PTR_ERR(sock));
553675fb 2631 kfree(vs);
e50fddc8 2632 return ERR_CAST(sock);
553675fb 2633 }
e4c7ed41
CW		 2634
2635 vs->sock = sock;
9c2e24e1 2636 atomic_set(&vs->refcnt, 1);
af33c1ad 2637 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2638
dc01e7d3
OG		 2639 /* Initialize the vxlan udp offloads structure */
2640 vs->udp_offloads.port = port;
2641 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2642 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2643
9c2e24e1
PS		 2644 spin_lock(&vn->sock_lock);
2645 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2646 vxlan_notify_add_rx_port(vs);
9c2e24e1 2647 spin_unlock(&vn->sock_lock);
553675fb 2648
2649 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2650 tunnel_cfg.sk_user_data = vs;
2651 tunnel_cfg.encap_type = 1;
2652 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2653 tunnel_cfg.encap_destroy = NULL;
2654
2655 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2656
9c2e24e1
PS		 2657 return vs;
2658}
2659
b1be00a6 2660static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2661{
205f356d
JB		 2662 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2663 struct vxlan_sock *vs = NULL;
9c2e24e1 2664
205f356d 2665 if (!vxlan->cfg.no_share) {
56ef9c90 2666 spin_lock(&vn->sock_lock);
205f356d
JB		 2667 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2668 vxlan->cfg.dst_port, vxlan->flags);
2669 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2670 spin_unlock(&vn->sock_lock);
205f356d 2671 return -EBUSY;
56ef9c90
MRL		 2672 }
2673 spin_unlock(&vn->sock_lock);
2674 }
205f356d 2675 if (!vs)
b1be00a6
JB		 2676 vs = vxlan_socket_create(vxlan->net, ipv6,
2677 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2678 if (IS_ERR(vs))
2679 return PTR_ERR(vs);
b1be00a6
JB		 2680#if IS_ENABLED(CONFIG_IPV6)
2681 if (ipv6)
2682 vxlan->vn6_sock = vs;
2683 else
2684#endif
2685 vxlan->vn4_sock = vs;
205f356d
JB		 2686 vxlan_vs_add_dev(vs, vxlan);
2687 return 0;
553675fb 2688}
2689
b1be00a6
JB		 2690static int vxlan_sock_add(struct vxlan_dev *vxlan)
2691{
2692 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2693 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2694 int ret = 0;
2695
2696 vxlan->vn4_sock = NULL;
2697#if IS_ENABLED(CONFIG_IPV6)
2698 vxlan->vn6_sock = NULL;
2699 if (ipv6 || metadata)
2700 ret = __vxlan_sock_add(vxlan, true);
2701#endif
2702 if (!ret && (!ipv6 || metadata))
2703 ret = __vxlan_sock_add(vxlan, false);
2704 if (ret < 0)
2705 vxlan_sock_release(vxlan);
2706 return ret;
2707}
2708
0dfbdf41
TG		 2709static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2710 struct vxlan_config *conf)
d342894c 2711{
33564bbb 2712 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2713 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2714 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2715 unsigned short needed_headroom = ETH_HLEN;
d342894c 2716 int err;
e4c7ed41 2717 bool use_ipv6 = false;
0dfbdf41 2718 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2719 struct net_device *lowerdev = NULL;
d342894c 2720
33564bbb 2721 vxlan->net = src_net;
f01ec1c0 2722
0dfbdf41 2723 dst->remote_vni = conf->vni;
d342894c 2724
0dfbdf41 2725 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2726
0dfbdf41
TG		 2727 /* Unless IPv6 is explicitly requested, assume IPv4 */
2728 if (!dst->remote_ip.sa.sa_family)
2729 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2730
0dfbdf41 2731 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2732 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2733 if (!IS_ENABLED(CONFIG_IPV6))
2734 return -EPFNOSUPPORT;
e4c7ed41 2735 use_ipv6 = true;
b1be00a6 2736 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2737 }
d342894c 2738
0dfbdf41 2739 if (conf->remote_ifindex) {
7e059158 2740 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2741 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2742
2743 if (!lowerdev) {
c7995c43 2744 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2745 return -ENODEV;
2746 }
d342894c 2747
e4c7ed41
CW		 2748#if IS_ENABLED(CONFIG_IPV6)
2749 if (use_ipv6) {
2750 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2751 if (idev && idev->cnf.disable_ipv6) {
2752 pr_info("IPv6 is disabled via sysctl\n");
2753 return -EPERM;
2754 }
e4c7ed41
CW		 2755 }
2756#endif
2757
0dfbdf41 2758 if (!conf->mtu)
e4c7ed41 2759 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2760
b1be00a6 2761 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2762 }
d342894c 2763
7e059158
DW		 2764 if (conf->mtu) {
2765 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2766 if (err)
2767 return err;
2768 }
2769
b1be00a6
JB		 2770 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2771 needed_headroom += VXLAN6_HEADROOM;
2772 else
2773 needed_headroom += VXLAN_HEADROOM;
2774 dev->needed_headroom = needed_headroom;
2775
0dfbdf41
TG		 2776 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2777 if (!vxlan->cfg.dst_port)
2778 vxlan->cfg.dst_port = default_port;
2779 vxlan->flags |= conf->flags;
2780
2781 if (!vxlan->cfg.age_interval)
2782 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2783
07b9b37c
ND		 2784 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2785 if (tmp->cfg.vni == conf->vni &&
2786 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2787 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2788 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2789 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2790 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2791 return -EEXIST;
07b9b37c 2792 }
0dfbdf41
TG		 2793
2794 dev->ethtool_ops = &vxlan_ethtool_ops;
2795
2796 /* create an fdb entry for a valid default destination */
2797 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2798 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2799 &vxlan->default_dst.remote_ip,
2800 NUD_REACHABLE|NUD_PERMANENT,
2801 NLM_F_EXCL|NLM_F_CREATE,
2802 vxlan->cfg.dst_port,
2803 vxlan->default_dst.remote_vni,
2804 vxlan->default_dst.remote_ifindex,
2805 NTF_SELF);
2806 if (err)
2807 return err;
2808 }
2809
2810 err = register_netdevice(dev);
2811 if (err) {
2812 vxlan_fdb_delete_default(vxlan);
2813 return err;
2814 }
2815
2816 list_add(&vxlan->next, &vn->vxlan_list);
2817
2818 return 0;
2819}
2820
2821struct net_device *vxlan_dev_create(struct net *net, const char *name,
2822 u8 name_assign_type, struct vxlan_config *conf)
2823{
2824 struct nlattr *tb[IFLA_MAX+1];
2825 struct net_device *dev;
2826 int err;
2827
2828 memset(&tb, 0, sizeof(tb));
2829
2830 dev = rtnl_create_link(net, name, name_assign_type,
2831 &vxlan_link_ops, tb);
2832 if (IS_ERR(dev))
2833 return dev;
2834
2835 err = vxlan_dev_configure(net, dev, conf);
2836 if (err < 0) {
2837 free_netdev(dev);
2838 return ERR_PTR(err);
2839 }
2840
2841 return dev;
2842}
2843EXPORT_SYMBOL_GPL(vxlan_dev_create);
2844
2845static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2846 struct nlattr *tb[], struct nlattr *data[])
2847{
2848 struct vxlan_config conf;
2849 int err;
2850
0dfbdf41 2851 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2852
2853 if (data[IFLA_VXLAN_ID])
54bfd872 2854 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG		 2855
2856 if (data[IFLA_VXLAN_GROUP]) {
2857 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2858 } else if (data[IFLA_VXLAN_GROUP6]) {
2859 if (!IS_ENABLED(CONFIG_IPV6))
2860 return -EPFNOSUPPORT;
2861
2862 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2863 conf.remote_ip.sa.sa_family = AF_INET6;
2864 }
2865
2866 if (data[IFLA_VXLAN_LOCAL]) {
2867 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2868 conf.saddr.sa.sa_family = AF_INET;
2869 } else if (data[IFLA_VXLAN_LOCAL6]) {
2870 if (!IS_ENABLED(CONFIG_IPV6))
2871 return -EPFNOSUPPORT;
2872
2873 /* TODO: respect scope id */
2874 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2875 conf.saddr.sa.sa_family = AF_INET6;
2876 }
2877
2878 if (data[IFLA_VXLAN_LINK])
2879 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2880
d342894c 2881 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2882 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2883
afb97186 2884 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2885 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2886
d342894c 2887 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2888 conf.flags |= VXLAN_F_LEARN;
d342894c 2889
2890 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2891 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2892
e4f67add 2893 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2894 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2895
2896 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2897 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2898
2899 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2900 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2901
2902 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2903 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2904
d342894c 2905 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2906 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2907
f8a9b1bc
AS		 2908 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2909 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2910 conf.flags |= VXLAN_F_COLLECT_METADATA;
2911
05f47d69 2912 if (data[IFLA_VXLAN_PORT_RANGE]) {
2913 const struct ifla_vxlan_port_range *p
2914 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2915 conf.port_min = ntohs(p->low);
2916 conf.port_max = ntohs(p->high);
05f47d69 2917 }
2918
823aa873 2919 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2920 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2921
6ceb31ca
AD		 2922 if (data[IFLA_VXLAN_UDP_CSUM] &&
2923 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2924 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH		 2925
2926 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2927 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2928 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2929
2930 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2931 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2932 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2933
dfd8645e
TH		 2934 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2935 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2936 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 2937
2938 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2939 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2940 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2941
3511494c 2942 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2943 conf.flags |= VXLAN_F_GBP;
3511494c 2944
0ace2ca8 2945 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2946 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2947
0dfbdf41
TG		 2948 err = vxlan_dev_configure(src_net, dev, &conf);
2949 switch (err) {
2950 case -ENODEV:
2951 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2952 break;
1b13c97f 2953
0dfbdf41
TG		 2954 case -EPERM:
2955 pr_info("IPv6 is disabled via sysctl\n");
2956 break;
d342894c 2957
0dfbdf41 2958 case -EEXIST:
54bfd872 2959 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
0dfbdf41 2960 break;
afbd8bae
MR		 2961 }
2962
0dfbdf41 2963 return err;
d342894c 2964}
2965
2966static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2967{
2968 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2969 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2970
fe5c3561 2971 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2972 if (!hlist_unhashed(&vxlan->hlist))
2973 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2974 spin_unlock(&vn->sock_lock);
2975
58ce31cc 2976 gro_cells_destroy(&vxlan->gro_cells);
553675fb 2977 list_del(&vxlan->next);
d342894c 2978 unregister_netdevice_queue(dev, head);
2979}
2980
2981static size_t vxlan_get_size(const struct net_device *dev)
2982{
2983
2984 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2985 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2986 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2987 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2988 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2989 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2990 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2991 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2992 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2993 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2994 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 2995 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 2996 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2997 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2998 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2999 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3000 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3001 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3002 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3003 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3005 0;
3006}
3007
3008static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3009{
3010 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3011 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3012 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3013 .low = htons(vxlan->cfg.port_min),
3014 .high = htons(vxlan->cfg.port_max),
05f47d69 3015 };
d342894c 3016
54bfd872 3017 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3018 goto nla_put_failure;
3019
e4c7ed41
CW		 3020 if (!vxlan_addr_any(&dst->remote_ip)) {
3021 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3022 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3023 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3024 goto nla_put_failure;
3025#if IS_ENABLED(CONFIG_IPV6)
3026 } else {
930345ea
JB		 3027 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3028 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3029 goto nla_put_failure;
3030#endif
3031 }
3032 }
d342894c 3033
c7995c43 3034 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3035 goto nla_put_failure;
3036
0dfbdf41
TG		 3037 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3038 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3039 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3040 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3041 goto nla_put_failure;
3042#if IS_ENABLED(CONFIG_IPV6)
3043 } else {
930345ea 3044 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3045 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3046 goto nla_put_failure;
3047#endif
3048 }
3049 }
d342894c 3050
0dfbdf41
TG		 3051 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3052 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 3053 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3054 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3055 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3056 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3057 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3058 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3059 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3060 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3061 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3062 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3063 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3064 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3065 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3066 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3067 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3068 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3069 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3070 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3071 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3072 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3073 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3074 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3075 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3076 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3077 goto nla_put_failure;
3078
05f47d69 3079 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3080 goto nla_put_failure;
3081
3511494c
TG		 3082 if (vxlan->flags & VXLAN_F_GBP &&
3083 nla_put_flag(skb, IFLA_VXLAN_GBP))
3084 goto nla_put_failure;
3085
0ace2ca8
TH		 3086 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3087 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3088 goto nla_put_failure;
3089
d342894c 3090 return 0;
3091
3092nla_put_failure:
3093 return -EMSGSIZE;
3094}
3095
1728d4fa
ND		 3096static struct net *vxlan_get_link_net(const struct net_device *dev)
3097{
3098 struct vxlan_dev *vxlan = netdev_priv(dev);
3099
3100 return vxlan->net;
3101}
3102
d342894c 3103static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3104 .kind = "vxlan",
3105 .maxtype = IFLA_VXLAN_MAX,
3106 .policy = vxlan_policy,
3107 .priv_size = sizeof(struct vxlan_dev),
3108 .setup = vxlan_setup,
3109 .validate = vxlan_validate,
3110 .newlink = vxlan_newlink,
3111 .dellink = vxlan_dellink,
3112 .get_size = vxlan_get_size,
3113 .fill_info = vxlan_fill_info,
1728d4fa 3114 .get_link_net = vxlan_get_link_net,
d342894c 3115};
3116
acaf4e70
DB		 3117static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3118 struct net_device *dev)
3119{
3120 struct vxlan_dev *vxlan, *next;
3121 LIST_HEAD(list_kill);
3122
3123 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3124 struct vxlan_rdst *dst = &vxlan->default_dst;
3125
3126 /* In case we created vxlan device with carrier
3127 * and we loose the carrier due to module unload
3128 * we also need to remove vxlan device. In other
3129 * cases, it's not necessary and remote_ifindex
3130 * is 0 here, so no matches.
3131 */
3132 if (dst->remote_ifindex == dev->ifindex)
3133 vxlan_dellink(vxlan->dev, &list_kill);
3134 }
3135
3136 unregister_netdevice_many(&list_kill);
3137}
3138
3139static int vxlan_lowerdev_event(struct notifier_block *unused,
3140 unsigned long event, void *ptr)
3141{
3142 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3143 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3144
783c1463 3145 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3146 vxlan_handle_lowerdev_unregister(vn, dev);
3147
3148 return NOTIFY_DONE;
3149}
3150
3151static struct notifier_block vxlan_notifier_block __read_mostly = {
3152 .notifier_call = vxlan_lowerdev_event,
3153};
3154
d342894c 3155static __net_init int vxlan_init_net(struct net *net)
3156{
3157 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3158 unsigned int h;
d342894c 3159
553675fb 3160 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3161 spin_lock_init(&vn->sock_lock);
d342894c 3162
553675fb 3163 for (h = 0; h < PORT_HASH_SIZE; ++h)
3164 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3165
3166 return 0;
3167}
3168
f01ec1c0
ND		 3169static void __net_exit vxlan_exit_net(struct net *net)
3170{
3171 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3172 struct vxlan_dev *vxlan, *next;
3173 struct net_device *dev, *aux;
3174 LIST_HEAD(list);
3175
3176 rtnl_lock();
3177 for_each_netdev_safe(net, dev, aux)
3178 if (dev->rtnl_link_ops == &vxlan_link_ops)
3179 unregister_netdevice_queue(dev, &list);
3180
3181 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3182 /* If vxlan->dev is in the same netns, it has already been added
3183 * to the list by the previous loop.
3184 */
58ce31cc
TH		 3185 if (!net_eq(dev_net(vxlan->dev), net)) {
3186 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3187 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3188 }
f01ec1c0
ND		 3189 }
3190
3191 unregister_netdevice_many(&list);
3192 rtnl_unlock();
3193}
3194
d342894c 3195static struct pernet_operations vxlan_net_ops = {
3196 .init = vxlan_init_net,
f01ec1c0 3197 .exit = vxlan_exit_net,
d342894c 3198 .id = &vxlan_net_id,
3199 .size = sizeof(struct vxlan_net),
3200};
3201
3202static int __init vxlan_init_module(void)
3203{
3204 int rc;
3205
758c57d1
SH		 3206 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3207 if (!vxlan_wq)
3208 return -ENOMEM;
3209
d342894c 3210 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3211
783c1463 3212 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3213 if (rc)
3214 goto out1;
3215
acaf4e70 3216 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3217 if (rc)
3218 goto out2;
3219
acaf4e70
DB		 3220 rc = rtnl_link_register(&vxlan_link_ops);
3221 if (rc)
3222 goto out3;
d342894c 3223
acaf4e70
DB		 3224 return 0;
3225out3:
3226 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3227out2:
783c1463 3228 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3229out1:
758c57d1 3230 destroy_workqueue(vxlan_wq);
d342894c 3231 return rc;
3232}
7332a13b 3233late_initcall(vxlan_init_module);
d342894c 3234
3235static void __exit vxlan_cleanup_module(void)
3236{
b7153984 3237 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3238 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3239 destroy_workqueue(vxlan_wq);
783c1463
DB		 3240 unregister_pernet_subsys(&vxlan_net_ops);
3241 /* rcu_barrier() is called by netns */
d342894c 3242}
3243module_exit(vxlan_cleanup_module);
3244
3245MODULE_LICENSE("GPL");
3246MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3247MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3248MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3249MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Artful kernel mirror