]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - drivers/net/vxlan.c
Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux
[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 for (p = *head; p; p = p->next) {
595 if (!NAPI_GRO_CB(p)->same_flow)
596 continue;
597
598 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG
599 if (vh->vx_flags != vh2->vx_flags ||
600 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG
601 NAPI_GRO_CB(p)->same_flow = 0;
602 continue;
603 }
dc01e7d3
OG
604 }
605
9b174d88 606 pp = eth_gro_receive(head, skb);
c194cf93 607 flush = 0;
dc01e7d3 608
dc01e7d3 609out:
26c4f7da 610 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG
611 NAPI_GRO_CB(skb)->flush |= flush;
612
613 return pp;
614}
615
a2b12f3c
TH
616static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
617 struct udp_offload *uoff)
dc01e7d3 618{
cfdf1e1b
JG
619 udp_tunnel_gro_complete(skb, nhoff);
620
9b174d88 621 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG
622}
623
53cf5275 624/* Notify netdevs that UDP port started listening */
dc01e7d3 625static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG
626{
627 struct net_device *dev;
dc01e7d3 628 struct sock *sk = vs->sock->sk;
53cf5275 629 struct net *net = sock_net(sk);
705cc62f 630 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 631 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG
632 int err;
633
634 if (sa_family == AF_INET) {
787d7ac3 635 err = udp_add_offload(net, &vs->udp_offloads);
dc01e7d3
OG
636 if (err)
637 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
638 }
53cf5275
JG
639
640 rcu_read_lock();
641 for_each_netdev_rcu(net, dev) {
642 if (dev->netdev_ops->ndo_add_vxlan_port)
643 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
644 port);
645 }
646 rcu_read_unlock();
647}
648
649/* Notify netdevs that UDP port is no more listening */
dc01e7d3 650static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG
651{
652 struct net_device *dev;
dc01e7d3 653 struct sock *sk = vs->sock->sk;
53cf5275 654 struct net *net = sock_net(sk);
705cc62f 655 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 656 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG
657
658 rcu_read_lock();
659 for_each_netdev_rcu(net, dev) {
660 if (dev->netdev_ops->ndo_del_vxlan_port)
661 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
662 port);
663 }
664 rcu_read_unlock();
dc01e7d3
OG
665
666 if (sa_family == AF_INET)
667 udp_del_offload(&vs->udp_offloads);
53cf5275
JG
668}
669
d342894c 670/* Add new entry to forwarding table -- assumes lock held */
671static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 672 const u8 *mac, union vxlan_addr *ip,
6681712d 673 __u16 state, __u16 flags,
54bfd872 674 __be16 port, __be32 vni, __u32 ifindex,
ae884082 675 __u8 ndm_flags)
d342894c 676{
9e4b93f9 677 struct vxlan_rdst *rd = NULL;
d342894c 678 struct vxlan_fdb *f;
679 int notify = 0;
680
014be2c8 681 f = __vxlan_find_mac(vxlan, mac);
d342894c 682 if (f) {
683 if (flags & NLM_F_EXCL) {
684 netdev_dbg(vxlan->dev,
685 "lost race to create %pM\n", mac);
686 return -EEXIST;
687 }
688 if (f->state != state) {
689 f->state = state;
690 f->updated = jiffies;
691 notify = 1;
692 }
ae884082
DS
693 if (f->flags != ndm_flags) {
694 f->flags = ndm_flags;
695 f->updated = jiffies;
696 notify = 1;
697 }
906dc186
TR
698 if ((flags & NLM_F_REPLACE)) {
699 /* Only change unicasts */
700 if (!(is_multicast_ether_addr(f->eth_addr) ||
701 is_zero_ether_addr(f->eth_addr))) {
60840429 702 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 703 ifindex);
906dc186
TR
704 } else
705 return -EOPNOTSUPP;
706 }
6681712d 707 if ((flags & NLM_F_APPEND) &&
58e4c767
MR
708 (is_multicast_ether_addr(f->eth_addr) ||
709 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND
710 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
711 &rd);
6681712d
DS
712
713 if (rc < 0)
714 return rc;
715 notify |= rc;
716 }
d342894c 717 } else {
718 if (!(flags & NLM_F_CREATE))
719 return -ENOENT;
720
0dfbdf41
TG
721 if (vxlan->cfg.addrmax &&
722 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 723 return -ENOSPC;
724
906dc186
TR
725 /* Disallow replace to add a multicast entry */
726 if ((flags & NLM_F_REPLACE) &&
727 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
728 return -EOPNOTSUPP;
729
e4c7ed41 730 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 731 f = kmalloc(sizeof(*f), GFP_ATOMIC);
732 if (!f)
733 return -ENOMEM;
734
735 notify = 1;
d342894c 736 f->state = state;
ae884082 737 f->flags = ndm_flags;
d342894c 738 f->updated = f->used = jiffies;
3e61aa8f 739 INIT_LIST_HEAD(&f->remotes);
d342894c 740 memcpy(f->eth_addr, mac, ETH_ALEN);
741
9e4b93f9 742 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 743
d342894c 744 ++vxlan->addrcnt;
745 hlist_add_head_rcu(&f->hlist,
746 vxlan_fdb_head(vxlan, mac));
747 }
748
9e4b93f9
ND
749 if (notify) {
750 if (rd == NULL)
751 rd = first_remote_rtnl(f);
752 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
753 }
d342894c 754
755 return 0;
756}
757
6706c82e 758static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS
759{
760 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 761 struct vxlan_rdst *rd, *nd;
6681712d 762
0c1d70af
PA
763 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
764 dst_cache_destroy(&rd->dst_cache);
6681712d 765 kfree(rd);
0c1d70af 766 }
6681712d
DS
767 kfree(f);
768}
769
d342894c 770static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
771{
772 netdev_dbg(vxlan->dev,
773 "delete %pM\n", f->eth_addr);
774
775 --vxlan->addrcnt;
9e4b93f9 776 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 777
778 hlist_del_rcu(&f->hlist);
6681712d 779 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 780}
781
f0b074be 782static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB
783 union vxlan_addr *ip, __be16 *port, __be32 *vni,
784 u32 *ifindex)
d342894c 785{
6681712d 786 struct net *net = dev_net(vxlan->dev);
e4c7ed41 787 int err;
d342894c 788
f0b074be 789 if (tb[NDA_DST]) {
e4c7ed41
CW
790 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
791 if (err)
792 return err;
f0b074be 793 } else {
e4c7ed41
CW
794 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
795 if (remote->sa.sa_family == AF_INET) {
796 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
797 ip->sa.sa_family = AF_INET;
798#if IS_ENABLED(CONFIG_IPV6)
799 } else {
800 ip->sin6.sin6_addr = in6addr_any;
801 ip->sa.sa_family = AF_INET6;
802#endif
803 }
f0b074be 804 }
d342894c 805
6681712d 806 if (tb[NDA_PORT]) {
73cf3317 807 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 808 return -EINVAL;
f0b074be
MR
809 *port = nla_get_be16(tb[NDA_PORT]);
810 } else {
0dfbdf41 811 *port = vxlan->cfg.dst_port;
f0b074be 812 }
6681712d
DS
813
814 if (tb[NDA_VNI]) {
815 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
816 return -EINVAL;
54bfd872 817 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR
818 } else {
819 *vni = vxlan->default_dst.remote_vni;
820 }
6681712d
DS
821
822 if (tb[NDA_IFINDEX]) {
5abb0029 823 struct net_device *tdev;
6681712d
DS
824
825 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
826 return -EINVAL;
f0b074be 827 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 828 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 829 if (!tdev)
6681712d 830 return -EADDRNOTAVAIL;
f0b074be
MR
831 } else {
832 *ifindex = 0;
833 }
834
835 return 0;
836}
837
838/* Add static entry (via netlink) */
839static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
840 struct net_device *dev,
f6f6424b 841 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR
842{
843 struct vxlan_dev *vxlan = netdev_priv(dev);
844 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 845 union vxlan_addr ip;
f0b074be 846 __be16 port;
54bfd872
JB
847 __be32 vni;
848 u32 ifindex;
f0b074be
MR
849 int err;
850
851 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
852 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
853 ndm->ndm_state);
854 return -EINVAL;
855 }
856
857 if (tb[NDA_DST] == NULL)
858 return -EINVAL;
859
860 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
861 if (err)
862 return err;
6681712d 863
5933a7bb
MR
864 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
865 return -EAFNOSUPPORT;
866
d342894c 867 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 868 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 869 port, vni, ifindex, ndm->ndm_flags);
d342894c 870 spin_unlock_bh(&vxlan->hash_lock);
871
872 return err;
873}
874
875/* Delete entry (via netlink) */
1690be63
VY
876static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
877 struct net_device *dev,
f6f6424b 878 const unsigned char *addr, u16 vid)
d342894c 879{
880 struct vxlan_dev *vxlan = netdev_priv(dev);
881 struct vxlan_fdb *f;
bc7892ba 882 struct vxlan_rdst *rd = NULL;
e4c7ed41 883 union vxlan_addr ip;
bc7892ba 884 __be16 port;
54bfd872
JB
885 __be32 vni;
886 u32 ifindex;
bc7892ba
MR
887 int err;
888
889 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
890 if (err)
891 return err;
892
893 err = -ENOENT;
d342894c 894
895 spin_lock_bh(&vxlan->hash_lock);
896 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR
897 if (!f)
898 goto out;
899
e4c7ed41
CW
900 if (!vxlan_addr_any(&ip)) {
901 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR
902 if (!rd)
903 goto out;
904 }
905
906 err = 0;
907
908 /* remove a destination if it's not the only one on the list,
909 * otherwise destroy the fdb entry
910 */
911 if (rd && !list_is_singular(&f->remotes)) {
912 list_del_rcu(&rd->list);
9e4b93f9 913 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 914 kfree_rcu(rd, rcu);
bc7892ba 915 goto out;
d342894c 916 }
bc7892ba
MR
917
918 vxlan_fdb_destroy(vxlan, f);
919
920out:
d342894c 921 spin_unlock_bh(&vxlan->hash_lock);
922
923 return err;
924}
925
926/* Dump forwarding table */
927static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
928 struct net_device *dev,
929 struct net_device *filter_dev, int idx)
d342894c 930{
931 struct vxlan_dev *vxlan = netdev_priv(dev);
932 unsigned int h;
933
934 for (h = 0; h < FDB_HASH_SIZE; ++h) {
935 struct vxlan_fdb *f;
d342894c 936 int err;
937
b67bfe0d 938 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 939 struct vxlan_rdst *rd;
6681712d 940
3e61aa8f 941 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW
942 if (idx < cb->args[0])
943 goto skip;
944
6681712d
DS
945 err = vxlan_fdb_info(skb, vxlan, f,
946 NETLINK_CB(cb->skb).portid,
947 cb->nlh->nlmsg_seq,
948 RTM_NEWNEIGH,
949 NLM_F_MULTI, rd);
472681d5
MM
950 if (err < 0) {
951 cb->args[1] = err;
3e61aa8f 952 goto out;
472681d5 953 }
3e61aa8f 954skip:
07a51cd3
AW
955 ++idx;
956 }
d342894c 957 }
958 }
3e61aa8f 959out:
d342894c 960 return idx;
961}
962
963/* Watch incoming packets to learn mapping between Ethernet address
964 * and Tunnel endpoint.
c4b49512 965 * Return true if packet is bogus and should be dropped.
d342894c 966 */
26a41ae6 967static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 968 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 969{
970 struct vxlan_dev *vxlan = netdev_priv(dev);
971 struct vxlan_fdb *f;
d342894c 972
973 f = vxlan_find_mac(vxlan, src_mac);
974 if (likely(f)) {
5ca5461c 975 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 976
e4c7ed41 977 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 978 return false;
979
980 /* Don't migrate static entries, drop packets */
eb064c3b 981 if (f->state & NUD_NOARP)
26a41ae6 982 return true;
d342894c 983
984 if (net_ratelimit())
985 netdev_info(dev,
e4c7ed41 986 "%pM migrated from %pIS to %pIS\n",
a4870f79 987 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 988
e4c7ed41 989 rdst->remote_ip = *src_ip;
d342894c 990 f->updated = jiffies;
9e4b93f9 991 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 992 } else {
993 /* learned new entry */
994 spin_lock(&vxlan->hash_lock);
3bf74b1a 995
996 /* close off race between vxlan_flush and incoming packets */
997 if (netif_running(dev))
998 vxlan_fdb_create(vxlan, src_mac, src_ip,
999 NUD_REACHABLE,
1000 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1001 vxlan->cfg.dst_port,
3bf74b1a 1002 vxlan->default_dst.remote_vni,
1003 0, NTF_SELF);
d342894c 1004 spin_unlock(&vxlan->hash_lock);
1005 }
26a41ae6 1006
1007 return false;
d342894c 1008}
1009
d342894c 1010/* See if multicast group is already in use by other ID */
95ab0991 1011static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1012{
553675fb 1013 struct vxlan_dev *vxlan;
b1be00a6 1014 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1015
95ab0991
G
1016 /* The vxlan_sock is only used by dev, leaving group has
1017 * no effect on other vxlan devices.
1018 */
b1be00a6
JB
1019 if (family == AF_INET && dev->vn4_sock &&
1020 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1021 return false;
b1be00a6
JB
1022#if IS_ENABLED(CONFIG_IPV6)
1023 if (family == AF_INET6 && dev->vn6_sock &&
1024 atomic_read(&dev->vn6_sock->refcnt) == 1)
1025 return false;
1026#endif
95ab0991 1027
553675fb 1028 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1029 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1030 continue;
d342894c 1031
b1be00a6 1032 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1033 continue;
b1be00a6
JB
1034#if IS_ENABLED(CONFIG_IPV6)
1035 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1036 continue;
1037#endif
95ab0991
G
1038
1039 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1040 &dev->default_dst.remote_ip))
1041 continue;
1042
1043 if (vxlan->default_dst.remote_ifindex !=
1044 dev->default_dst.remote_ifindex)
1045 continue;
1046
1047 return true;
553675fb 1048 }
d342894c 1049
1050 return false;
1051}
1052
b1be00a6 1053static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1054{
b1be00a6 1055 struct vxlan_net *vn;
012a5729 1056
b1be00a6
JB
1057 if (!vs)
1058 return;
7c47cedf
SH
1059 if (!atomic_dec_and_test(&vs->refcnt))
1060 return;
d342894c 1061
b1be00a6 1062 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1063 spin_lock(&vn->sock_lock);
7c47cedf 1064 hlist_del_rcu(&vs->hlist);
dc01e7d3 1065 vxlan_notify_del_rx_port(vs);
1c51a915
SH
1066 spin_unlock(&vn->sock_lock);
1067
7c47cedf 1068 queue_work(vxlan_wq, &vs->del_work);
d342894c 1069}
1070
b1be00a6
JB
1071static void vxlan_sock_release(struct vxlan_dev *vxlan)
1072{
1073 __vxlan_sock_release(vxlan->vn4_sock);
1074#if IS_ENABLED(CONFIG_IPV6)
1075 __vxlan_sock_release(vxlan->vn6_sock);
1076#endif
1077}
1078
56ef9c90 1079/* Update multicast group membership when first VNI on
c4b49512 1080 * multicast address is brought up
7c47cedf 1081 */
56ef9c90 1082static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1083{
b1be00a6 1084 struct sock *sk;
e4c7ed41
CW
1085 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1086 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1087 int ret = -EINVAL;
d342894c 1088
e4c7ed41
CW
1089 if (ip->sa.sa_family == AF_INET) {
1090 struct ip_mreqn mreq = {
1091 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1092 .imr_ifindex = ifindex,
1093 };
1094
b1be00a6
JB
1095 sk = vxlan->vn4_sock->sock->sk;
1096 lock_sock(sk);
56ef9c90 1097 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1098 release_sock(sk);
e4c7ed41
CW
1099#if IS_ENABLED(CONFIG_IPV6)
1100 } else {
b1be00a6
JB
1101 sk = vxlan->vn6_sock->sock->sk;
1102 lock_sock(sk);
56ef9c90
MRL
1103 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1104 &ip->sin6.sin6_addr);
b1be00a6 1105 release_sock(sk);
e4c7ed41
CW
1106#endif
1107 }
3fc2de2f 1108
56ef9c90 1109 return ret;
3fc2de2f 1110}
1111
1112/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1113static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1114{
b1be00a6 1115 struct sock *sk;
e4c7ed41
CW
1116 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1117 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1118 int ret = -EINVAL;
3fc2de2f 1119
e4c7ed41
CW
1120 if (ip->sa.sa_family == AF_INET) {
1121 struct ip_mreqn mreq = {
1122 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1123 .imr_ifindex = ifindex,
1124 };
1125
b1be00a6
JB
1126 sk = vxlan->vn4_sock->sock->sk;
1127 lock_sock(sk);
56ef9c90 1128 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1129 release_sock(sk);
e4c7ed41
CW
1130#if IS_ENABLED(CONFIG_IPV6)
1131 } else {
b1be00a6
JB
1132 sk = vxlan->vn6_sock->sock->sk;
1133 lock_sock(sk);
56ef9c90
MRL
1134 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1135 &ip->sin6.sin6_addr);
b1be00a6 1136 release_sock(sk);
e4c7ed41
CW
1137#endif
1138 }
d342894c 1139
56ef9c90 1140 return ret;
d342894c 1141}
1142
f14ecebb
JB
1143static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1144 struct sk_buff *skb, u32 vxflags)
dfd8645e
TH
1145{
1146 size_t start, offset, plen;
dfd8645e 1147
f14ecebb
JB
1148 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1149 goto out;
b7fe10e5 1150
f14ecebb
JB
1151 start = vxlan_rco_start(unparsed->vx_vni);
1152 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1153
be5cfeab 1154 plen = sizeof(struct vxlanhdr) + offset + sizeof(u16);
dfd8645e
TH
1155
1156 if (!pskb_may_pull(skb, plen))
be5cfeab 1157 return false;
dfd8645e 1158
be5cfeab
JB
1159 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1160 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB
1161out:
1162 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1163 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1164 return true;
dfd8645e
TH
1165}
1166
f14ecebb 1167static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1168 struct sk_buff *skb, u32 vxflags,
10a5af23 1169 struct vxlan_metadata *md)
3288af08 1170{
f14ecebb 1171 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1172 struct metadata_dst *tun_dst;
f14ecebb
JB
1173
1174 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1175 goto out;
3288af08 1176
3288af08
JB
1177 md->gbp = ntohs(gbp->policy_id);
1178
10a5af23 1179 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1180 if (tun_dst) {
3288af08 1181 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM
1182 tun_dst->u.tun_info.options_len = sizeof(*md);
1183 }
3288af08
JB
1184 if (gbp->dont_learn)
1185 md->gbp |= VXLAN_GBP_DONT_LEARN;
1186
1187 if (gbp->policy_applied)
1188 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1189
64f87d36
JB
1190 /* In flow-based mode, GBP is carried in dst_metadata */
1191 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1192 skb->mark = md->gbp;
f14ecebb
JB
1193out:
1194 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB
1195}
1196
1ab016e2
JB
1197static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1198 struct vxlan_sock *vs,
1199 struct sk_buff *skb)
614732ea 1200{
614732ea 1201 union vxlan_addr saddr;
614732ea 1202
614732ea 1203 skb_reset_mac_header(skb);
614732ea
TG
1204 skb->protocol = eth_type_trans(skb, vxlan->dev);
1205 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1206
1207 /* Ignore packet loops (and multicast echo) */
1208 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1209 return false;
614732ea 1210
760c6805 1211 /* Get address from the outer IP header */
ce212d0f 1212 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1213 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG
1214 saddr.sa.sa_family = AF_INET;
1215#if IS_ENABLED(CONFIG_IPV6)
1216 } else {
1ab016e2 1217 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG
1218 saddr.sa.sa_family = AF_INET6;
1219#endif
1220 }
1221
1ab016e2
JB
1222 if ((vxlan->flags & VXLAN_F_LEARN) &&
1223 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1224 return false;
1225
1226 return true;
1227}
1228
760c6805
JB
1229static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1230 struct sk_buff *skb)
1231{
1232 int err = 0;
1233
1234 if (vxlan_get_sk_family(vs) == AF_INET)
1235 err = IP_ECN_decapsulate(oiph, skb);
1236#if IS_ENABLED(CONFIG_IPV6)
1237 else
1238 err = IP6_ECN_decapsulate(oiph, skb);
1239#endif
1240
1241 if (unlikely(err) && log_ecn_error) {
1242 if (vxlan_get_sk_family(vs) == AF_INET)
1243 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1244 &((struct iphdr *)oiph)->saddr,
1245 ((struct iphdr *)oiph)->tos);
1246 else
1247 net_info_ratelimited("non-ECT from %pI6\n",
1248 &((struct ipv6hdr *)oiph)->saddr);
1249 }
1250 return err <= 1;
1251}
1252
d342894c 1253/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1254static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1255{
f2d1968e 1256 struct pcpu_sw_netstats *stats;
c9e78efb 1257 struct vxlan_dev *vxlan;
5cfccc5a 1258 struct vxlan_sock *vs;
f14ecebb 1259 struct vxlanhdr unparsed;
ee122c79
TG
1260 struct vxlan_metadata _md;
1261 struct vxlan_metadata *md = &_md;
f2d1968e 1262 void *oiph;
d342894c 1263
d342894c 1264 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1265 if (!pskb_may_pull(skb, VXLAN_HLEN))
288b01c8 1266 return 1;
d342894c 1267
f14ecebb 1268 unparsed = *vxlan_hdr(skb);
288b01c8
JB
1269 /* VNI flag always required to be set */
1270 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1271 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1272 ntohl(vxlan_hdr(skb)->vx_flags),
1273 ntohl(vxlan_hdr(skb)->vx_vni));
1274 /* Return non vxlan pkt */
1275 return 1;
d342894c 1276 }
288b01c8
JB
1277 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1278 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1279
559835ea 1280 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1281 if (!vs)
d342894c 1282 goto drop;
d342894c 1283
c9e78efb
JB
1284 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1285 if (!vxlan)
1286 goto drop;
1287
7f290c94
JB
1288 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB),
1289 !net_eq(vxlan->net, dev_net(vxlan->dev))))
c9e78efb
JB
1290 goto drop;
1291
ee122c79 1292 if (vxlan_collect_metadata(vs)) {
07dabf20 1293 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
10a5af23 1294 struct metadata_dst *tun_dst;
07dabf20 1295
c29a70d2 1296 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
07dabf20 1297 vxlan_vni_to_tun_id(vni), sizeof(*md));
c29a70d2 1298
ee122c79
TG
1299 if (!tun_dst)
1300 goto drop;
1301
0f1b7354 1302 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB
1303
1304 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG
1305 } else {
1306 memset(md, 0, sizeof(*md));
1307 }
1308
3511494c
TG
1309 /* For backwards compatibility, only allow reserved fields to be
1310 * used by VXLAN extensions if explicitly requested.
1311 */
f14ecebb
JB
1312 if (vs->flags & VXLAN_F_REMCSUM_RX)
1313 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1314 goto drop;
1315 if (vs->flags & VXLAN_F_GBP)
10a5af23 1316 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
3511494c 1317
f14ecebb 1318 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH
1319 /* If there are any unprocessed flags remaining treat
1320 * this as a malformed packet. This behavior diverges from
1321 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1322 * in reserved fields are to be ignored. The approach here
c4b49512 1323 * maintains compatibility with previous stack code, and also
3bf39475
TH
1324 * is more robust and provides a little more security in
1325 * adding extensions to VXLAN.
1326 */
288b01c8 1327 goto drop;
3bf39475
TH
1328 }
1329
f2d1968e
JB
1330 if (!vxlan_set_mac(vxlan, vs, skb))
1331 goto drop;
1332
f2d1968e
JB
1333 oiph = skb_network_header(skb);
1334 skb_reset_network_header(skb);
1335
1336 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1337 ++vxlan->dev->stats.rx_frame_errors;
1338 ++vxlan->dev->stats.rx_errors;
1339 goto drop;
1340 }
1341
1342 stats = this_cpu_ptr(vxlan->dev->tstats);
1343 u64_stats_update_begin(&stats->syncp);
1344 stats->rx_packets++;
1345 stats->rx_bytes += skb->len;
1346 u64_stats_update_end(&stats->syncp);
1347
1348 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS
1349 return 0;
1350
1351drop:
288b01c8
JB
1352 /* Consume bad packet */
1353 kfree_skb(skb);
1354 return 0;
5cfccc5a
PS
1355}
1356
e4f67add
DS
1357static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1358{
1359 struct vxlan_dev *vxlan = netdev_priv(dev);
1360 struct arphdr *parp;
1361 u8 *arpptr, *sha;
1362 __be32 sip, tip;
1363 struct neighbour *n;
1364
1365 if (dev->flags & IFF_NOARP)
1366 goto out;
1367
1368 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1369 dev->stats.tx_dropped++;
1370 goto out;
1371 }
1372 parp = arp_hdr(skb);
1373
1374 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1375 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1376 parp->ar_pro != htons(ETH_P_IP) ||
1377 parp->ar_op != htons(ARPOP_REQUEST) ||
1378 parp->ar_hln != dev->addr_len ||
1379 parp->ar_pln != 4)
1380 goto out;
1381 arpptr = (u8 *)parp + sizeof(struct arphdr);
1382 sha = arpptr;
1383 arpptr += dev->addr_len; /* sha */
1384 memcpy(&sip, arpptr, sizeof(sip));
1385 arpptr += sizeof(sip);
1386 arpptr += dev->addr_len; /* tha */
1387 memcpy(&tip, arpptr, sizeof(tip));
1388
1389 if (ipv4_is_loopback(tip) ||
1390 ipv4_is_multicast(tip))
1391 goto out;
1392
1393 n = neigh_lookup(&arp_tbl, &tip, dev);
1394
1395 if (n) {
e4f67add
DS
1396 struct vxlan_fdb *f;
1397 struct sk_buff *reply;
1398
1399 if (!(n->nud_state & NUD_CONNECTED)) {
1400 neigh_release(n);
1401 goto out;
1402 }
1403
1404 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1405 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS
1406 /* bridge-local neighbor */
1407 neigh_release(n);
1408 goto out;
1409 }
1410
1411 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1412 n->ha, sha);
1413
1414 neigh_release(n);
1415
7346135d
DS
1416 if (reply == NULL)
1417 goto out;
1418
e4f67add
DS
1419 skb_reset_mac_header(reply);
1420 __skb_pull(reply, skb_network_offset(reply));
1421 reply->ip_summed = CHECKSUM_UNNECESSARY;
1422 reply->pkt_type = PACKET_HOST;
1423
1424 if (netif_rx_ni(reply) == NET_RX_DROP)
1425 dev->stats.rx_dropped++;
e4c7ed41
CW
1426 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1427 union vxlan_addr ipa = {
1428 .sin.sin_addr.s_addr = tip,
a45e92a5 1429 .sin.sin_family = AF_INET,
e4c7ed41
CW
1430 };
1431
1432 vxlan_ip_miss(dev, &ipa);
1433 }
e4f67add
DS
1434out:
1435 consume_skb(skb);
1436 return NETDEV_TX_OK;
1437}
1438
f564f45c 1439#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS
1440static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1441 struct neighbour *n, bool isrouter)
1442{
1443 struct net_device *dev = request->dev;
1444 struct sk_buff *reply;
1445 struct nd_msg *ns, *na;
1446 struct ipv6hdr *pip6;
1447 u8 *daddr;
1448 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1449 int ns_olen;
1450 int i, len;
1451
1452 if (dev == NULL)
1453 return NULL;
1454
1455 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1456 sizeof(*na) + na_olen + dev->needed_tailroom;
1457 reply = alloc_skb(len, GFP_ATOMIC);
1458 if (reply == NULL)
1459 return NULL;
1460
1461 reply->protocol = htons(ETH_P_IPV6);
1462 reply->dev = dev;
1463 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1464 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1465 skb_reset_mac_header(reply);
4b29dba9
DS
1466
1467 ns = (struct nd_msg *)skb_transport_header(request);
1468
1469 daddr = eth_hdr(request)->h_source;
1470 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1471 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1472 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1473 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1474 break;
1475 }
1476 }
1477
1478 /* Ethernet header */
1479 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1480 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1481 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1482 reply->protocol = htons(ETH_P_IPV6);
1483
1484 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1485 skb_reset_network_header(reply);
4b29dba9
DS
1486 skb_put(reply, sizeof(struct ipv6hdr));
1487
1488 /* IPv6 header */
1489
1490 pip6 = ipv6_hdr(reply);
1491 memset(pip6, 0, sizeof(struct ipv6hdr));
1492 pip6->version = 6;
1493 pip6->priority = ipv6_hdr(request)->priority;
1494 pip6->nexthdr = IPPROTO_ICMPV6;
1495 pip6->hop_limit = 255;
1496 pip6->daddr = ipv6_hdr(request)->saddr;
1497 pip6->saddr = *(struct in6_addr *)n->primary_key;
1498
1499 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1500 skb_reset_transport_header(reply);
4b29dba9
DS
1501
1502 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1503
1504 /* Neighbor Advertisement */
1505 memset(na, 0, sizeof(*na)+na_olen);
1506 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1507 na->icmph.icmp6_router = isrouter;
1508 na->icmph.icmp6_override = 1;
1509 na->icmph.icmp6_solicited = 1;
1510 na->target = ns->target;
1511 ether_addr_copy(&na->opt[2], n->ha);
1512 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1513 na->opt[1] = na_olen >> 3;
1514
1515 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1516 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1517 csum_partial(na, sizeof(*na)+na_olen, 0));
1518
1519 pip6->payload_len = htons(sizeof(*na)+na_olen);
1520
1521 skb_push(reply, sizeof(struct ipv6hdr));
1522
1523 reply->ip_summed = CHECKSUM_UNNECESSARY;
1524
1525 return reply;
1526}
1527
f564f45c
CW
1528static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1529{
1530 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1531 struct nd_msg *msg;
f564f45c
CW
1532 const struct ipv6hdr *iphdr;
1533 const struct in6_addr *saddr, *daddr;
4b29dba9
DS
1534 struct neighbour *n;
1535 struct inet6_dev *in6_dev;
f564f45c
CW
1536
1537 in6_dev = __in6_dev_get(dev);
1538 if (!in6_dev)
1539 goto out;
1540
f564f45c
CW
1541 iphdr = ipv6_hdr(skb);
1542 saddr = &iphdr->saddr;
1543 daddr = &iphdr->daddr;
1544
f564f45c
CW
1545 msg = (struct nd_msg *)skb_transport_header(skb);
1546 if (msg->icmph.icmp6_code != 0 ||
1547 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1548 goto out;
1549
4b29dba9
DS
1550 if (ipv6_addr_loopback(daddr) ||
1551 ipv6_addr_is_multicast(&msg->target))
1552 goto out;
1553
1554 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW
1555
1556 if (n) {
1557 struct vxlan_fdb *f;
4b29dba9 1558 struct sk_buff *reply;
f564f45c
CW
1559
1560 if (!(n->nud_state & NUD_CONNECTED)) {
1561 neigh_release(n);
1562 goto out;
1563 }
1564
1565 f = vxlan_find_mac(vxlan, n->ha);
1566 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1567 /* bridge-local neighbor */
1568 neigh_release(n);
1569 goto out;
1570 }
1571
4b29dba9
DS
1572 reply = vxlan_na_create(skb, n,
1573 !!(f ? f->flags & NTF_ROUTER : 0));
1574
f564f45c 1575 neigh_release(n);
4b29dba9
DS
1576
1577 if (reply == NULL)
1578 goto out;
1579
1580 if (netif_rx_ni(reply) == NET_RX_DROP)
1581 dev->stats.rx_dropped++;
1582
f564f45c 1583 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS
1584 union vxlan_addr ipa = {
1585 .sin6.sin6_addr = msg->target,
a45e92a5 1586 .sin6.sin6_family = AF_INET6,
4b29dba9
DS
1587 };
1588
f564f45c
CW
1589 vxlan_ip_miss(dev, &ipa);
1590 }
1591
1592out:
1593 consume_skb(skb);
1594 return NETDEV_TX_OK;
1595}
1596#endif
1597
e4f67add
DS
1598static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1599{
1600 struct vxlan_dev *vxlan = netdev_priv(dev);
1601 struct neighbour *n;
e4f67add
DS
1602
1603 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1604 return false;
1605
1606 n = NULL;
1607 switch (ntohs(eth_hdr(skb)->h_proto)) {
1608 case ETH_P_IP:
e15a00aa
CW
1609 {
1610 struct iphdr *pip;
1611
e4f67add
DS
1612 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1613 return false;
1614 pip = ip_hdr(skb);
1615 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW
1616 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1617 union vxlan_addr ipa = {
1618 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1619 .sin.sin_family = AF_INET,
e4c7ed41
CW
1620 };
1621
1622 vxlan_ip_miss(dev, &ipa);
1623 return false;
1624 }
1625
e4f67add 1626 break;
e15a00aa
CW
1627 }
1628#if IS_ENABLED(CONFIG_IPV6)
1629 case ETH_P_IPV6:
1630 {
1631 struct ipv6hdr *pip6;
1632
1633 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1634 return false;
1635 pip6 = ipv6_hdr(skb);
1636 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1637 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1638 union vxlan_addr ipa = {
1639 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1640 .sin6.sin6_family = AF_INET6,
e15a00aa
CW
1641 };
1642
1643 vxlan_ip_miss(dev, &ipa);
1644 return false;
1645 }
1646
1647 break;
1648 }
1649#endif
e4f67add
DS
1650 default:
1651 return false;
1652 }
1653
1654 if (n) {
1655 bool diff;
1656
7367d0b5 1657 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS
1658 if (diff) {
1659 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1660 dev->addr_len);
1661 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1662 }
1663 neigh_release(n);
1664 return diff;
e4c7ed41
CW
1665 }
1666
e4f67add
DS
1667 return false;
1668}
1669
af33c1ad 1670static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG
1671 struct vxlan_metadata *md)
1672{
1673 struct vxlanhdr_gbp *gbp;
1674
db79a621
TG
1675 if (!md->gbp)
1676 return;
1677
3511494c 1678 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1679 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG
1680
1681 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1682 gbp->dont_learn = 1;
1683
1684 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1685 gbp->policy_applied = 1;
1686
1687 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1688}
1689
f491e56d
JB
1690static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1691 int iphdr_len, __be32 vni,
1692 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1693 bool udp_sum)
e4c7ed41 1694{
e4c7ed41 1695 struct vxlanhdr *vxh;
e4c7ed41
CW
1696 int min_headroom;
1697 int err;
dfd8645e 1698 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
dfd8645e 1699
af33c1ad 1700 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH
1701 skb->ip_summed == CHECKSUM_PARTIAL) {
1702 int csum_start = skb_checksum_start_offset(skb);
1703
1704 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1705 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1706 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1707 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1708 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1709 }
e4c7ed41 1710
e4c7ed41 1711 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1712 + VXLAN_HLEN + iphdr_len
df8a39de 1713 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS
1714
1715 /* Need space for new headers (invalidates iph ptr) */
1716 err = skb_cow_head(skb, min_headroom);
74f47278
PS
1717 if (unlikely(err)) {
1718 kfree_skb(skb);
649c5b8b 1719 return err;
74f47278 1720 }
649c5b8b 1721
5968250c
JP
1722 skb = vlan_hwaccel_push_inside(skb);
1723 if (WARN_ON(!skb))
1724 return -ENOMEM;
1eaa8178 1725
6fa79666 1726 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG
1727 if (IS_ERR(skb))
1728 return PTR_ERR(skb);
1729
49560532 1730 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB
1731 vxh->vx_flags = VXLAN_HF_VNI;
1732 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1733
dfd8645e 1734 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1735 unsigned int start;
dfd8645e 1736
54bfd872
JB
1737 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1738 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1739 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH
1740
1741 if (!skb_is_gso(skb)) {
1742 skb->ip_summed = CHECKSUM_NONE;
1743 skb->encapsulation = 0;
1744 }
1745 }
1746
af33c1ad
TH
1747 if (vxflags & VXLAN_F_GBP)
1748 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1749
996c9fd1 1750 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
039f5062 1751 return 0;
49560532 1752}
49560532 1753
1a8496ba
JB
1754static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1755 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA
1756 __be32 daddr, __be32 *saddr,
1757 struct dst_cache *dst_cache,
db3c6139 1758 const struct ip_tunnel_info *info)
1a8496ba 1759{
db3c6139 1760 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB
1761 struct rtable *rt = NULL;
1762 struct flowi4 fl4;
1763
db3c6139
DB
1764 if (tos && !info)
1765 use_cache = false;
1766 if (use_cache) {
0c1d70af
PA
1767 rt = dst_cache_get_ip4(dst_cache, saddr);
1768 if (rt)
1769 return rt;
1770 }
1771
1a8496ba
JB
1772 memset(&fl4, 0, sizeof(fl4));
1773 fl4.flowi4_oif = oif;
1774 fl4.flowi4_tos = RT_TOS(tos);
1775 fl4.flowi4_mark = skb->mark;
1776 fl4.flowi4_proto = IPPROTO_UDP;
1777 fl4.daddr = daddr;
1778 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1779
1780 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1781 if (!IS_ERR(rt)) {
1a8496ba 1782 *saddr = fl4.saddr;
0c1d70af
PA
1783 if (use_cache)
1784 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1785 }
1a8496ba
JB
1786 return rt;
1787}
1788
e5d4b29f
JB
1789#if IS_ENABLED(CONFIG_IPV6)
1790static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1400615d 1791 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1792 __be32 label,
e5d4b29f 1793 const struct in6_addr *daddr,
0c1d70af 1794 struct in6_addr *saddr,
db3c6139
DB
1795 struct dst_cache *dst_cache,
1796 const struct ip_tunnel_info *info)
e5d4b29f 1797{
db3c6139 1798 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB
1799 struct dst_entry *ndst;
1800 struct flowi6 fl6;
1801 int err;
1802
1400615d
DB
1803 if (tos && !info)
1804 use_cache = false;
db3c6139 1805 if (use_cache) {
0c1d70af
PA
1806 ndst = dst_cache_get_ip6(dst_cache, saddr);
1807 if (ndst)
1808 return ndst;
1809 }
1810
e5d4b29f
JB
1811 memset(&fl6, 0, sizeof(fl6));
1812 fl6.flowi6_oif = oif;
1400615d 1813 fl6.flowi6_tos = RT_TOS(tos);
e5d4b29f
JB
1814 fl6.daddr = *daddr;
1815 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
e7f70af1 1816 fl6.flowlabel = label;
e5d4b29f
JB
1817 fl6.flowi6_mark = skb->mark;
1818 fl6.flowi6_proto = IPPROTO_UDP;
1819
1820 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1821 vxlan->vn6_sock->sock->sk,
1822 &ndst, &fl6);
1823 if (err < 0)
1824 return ERR_PTR(err);
1825
1826 *saddr = fl6.saddr;
db3c6139 1827 if (use_cache)
0c1d70af 1828 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB
1829 return ndst;
1830}
1831#endif
1832
9dcc71e1
SS
1833/* Bypass encapsulation if the destination is local */
1834static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1835 struct vxlan_dev *dst_vxlan)
1836{
8f84985f 1837 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW
1838 union vxlan_addr loopback;
1839 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR
1840 struct net_device *dev = skb->dev;
1841 int len = skb->len;
9dcc71e1 1842
8f84985f
LR
1843 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1844 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS
1845 skb->pkt_type = PACKET_HOST;
1846 skb->encapsulation = 0;
1847 skb->dev = dst_vxlan->dev;
1848 __skb_pull(skb, skb_network_offset(skb));
1849
e4c7ed41
CW
1850 if (remote_ip->sa.sa_family == AF_INET) {
1851 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1852 loopback.sa.sa_family = AF_INET;
1853#if IS_ENABLED(CONFIG_IPV6)
1854 } else {
1855 loopback.sin6.sin6_addr = in6addr_loopback;
1856 loopback.sa.sa_family = AF_INET6;
1857#endif
1858 }
1859
9dcc71e1 1860 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1861 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS
1862
1863 u64_stats_update_begin(&tx_stats->syncp);
1864 tx_stats->tx_packets++;
ce6502a8 1865 tx_stats->tx_bytes += len;
9dcc71e1
SS
1866 u64_stats_update_end(&tx_stats->syncp);
1867
1868 if (netif_rx(skb) == NET_RX_SUCCESS) {
1869 u64_stats_update_begin(&rx_stats->syncp);
1870 rx_stats->rx_packets++;
ce6502a8 1871 rx_stats->rx_bytes += len;
9dcc71e1
SS
1872 u64_stats_update_end(&rx_stats->syncp);
1873 } else {
ce6502a8 1874 dev->stats.rx_dropped++;
9dcc71e1
SS
1875 }
1876}
1877
4ad16930
SH
1878static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1879 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1880{
d71785ff 1881 struct dst_cache *dst_cache;
3093fbe7 1882 struct ip_tunnel_info *info;
d342894c 1883 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1884 struct sock *sk;
e4c7ed41 1885 struct rtable *rt = NULL;
d342894c 1886 const struct iphdr *old_iph;
e4c7ed41 1887 union vxlan_addr *dst;
ee122c79
TG
1888 union vxlan_addr remote_ip;
1889 struct vxlan_metadata _md;
1890 struct vxlan_metadata *md = &_md;
e4c7ed41 1891 __be16 src_port = 0, dst_port;
e7f70af1 1892 __be32 vni, label;
d342894c 1893 __be16 df = 0;
1894 __u8 tos, ttl;
0e6fbc5b 1895 int err;
ee122c79 1896 u32 flags = vxlan->flags;
b4ed5cad 1897 bool udp_sum = false;
f491e56d 1898 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1899
61adedf3 1900 info = skb_tunnel_info(skb);
3093fbe7 1901
ee122c79 1902 if (rdst) {
0dfbdf41 1903 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG
1904 vni = rdst->remote_vni;
1905 dst = &rdst->remote_ip;
d71785ff 1906 dst_cache = &rdst->dst_cache;
ee122c79
TG
1907 } else {
1908 if (!info) {
1909 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1910 dev->name);
1911 goto drop;
1912 }
0dfbdf41 1913 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
54bfd872 1914 vni = vxlan_tun_id_to_vni(info->key.tun_id);
b1be00a6
JB
1915 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1916 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB
1917 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1918 else
1919 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1920 dst = &remote_ip;
d71785ff 1921 dst_cache = &info->dst_cache;
ee122c79 1922 }
e4f67add 1923
e4c7ed41 1924 if (vxlan_addr_any(dst)) {
e4f67add 1925 if (did_rsc) {
e4f67add 1926 /* short-circuited back to local bridge */
9dcc71e1 1927 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1928 return;
e4f67add 1929 }
ef59febe 1930 goto drop;
e4f67add 1931 }
ef59febe 1932
d342894c 1933 old_iph = ip_hdr(skb);
1934
0dfbdf41 1935 ttl = vxlan->cfg.ttl;
e4c7ed41 1936 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1937 ttl = 1;
1938
0dfbdf41 1939 tos = vxlan->cfg.tos;
d342894c 1940 if (tos == 1)
206aaafc 1941 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1942
e7f70af1 1943 label = vxlan->cfg.label;
0dfbdf41
TG
1944 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1945 vxlan->cfg.port_max, true);
d342894c 1946
a725e514 1947 if (info) {
a725e514
JB
1948 ttl = info->key.ttl;
1949 tos = info->key.tos;
e7f70af1 1950 label = info->key.label;
b4ed5cad 1951 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB
1952
1953 if (info->options_len)
4c222798 1954 md = ip_tunnel_info_opts(info);
a725e514
JB
1955 } else {
1956 md->gbp = skb->mark;
1957 }
1958
e4c7ed41 1959 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB
1960 __be32 saddr;
1961
b1be00a6
JB
1962 if (!vxlan->vn4_sock)
1963 goto drop;
1964 sk = vxlan->vn4_sock->sock->sk;
1965
1a8496ba
JB
1966 rt = vxlan_get_route(vxlan, skb,
1967 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 1968 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 1969 dst_cache, info);
e4c7ed41
CW
1970 if (IS_ERR(rt)) {
1971 netdev_dbg(dev, "no route to %pI4\n",
1972 &dst->sin.sin_addr.s_addr);
1973 dev->stats.tx_carrier_errors++;
1974 goto tx_error;
1975 }
d342894c 1976
e4c7ed41
CW
1977 if (rt->dst.dev == dev) {
1978 netdev_dbg(dev, "circular route to %pI4\n",
1979 &dst->sin.sin_addr.s_addr);
1980 dev->stats.collisions++;
fffc15a5 1981 goto rt_tx_error;
e4c7ed41
CW
1982 }
1983
1984 /* Bypass encapsulation if the destination is local */
1985 if (rt->rt_flags & RTCF_LOCAL &&
1986 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1987 struct vxlan_dev *dst_vxlan;
1988
1989 ip_rt_put(rt);
19ca9fc1 1990 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG
1991 dst->sa.sa_family, dst_port,
1992 vxlan->flags);
e4c7ed41
CW
1993 if (!dst_vxlan)
1994 goto tx_error;
1995 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1996 return;
1997 }
1998
6ceb31ca
AD
1999 if (!info)
2000 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2001 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2002 df = htons(IP_DF);
2003
e4c7ed41
CW
2004 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2005 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d 2006 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
54bfd872 2007 vni, md, flags, udp_sum);
f491e56d
JB
2008 if (err < 0)
2009 goto xmit_tx_error;
2010
2011 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2012 dst->sin.sin_addr.s_addr, tos, ttl, df,
2013 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW
2014#if IS_ENABLED(CONFIG_IPV6)
2015 } else {
e4c7ed41 2016 struct dst_entry *ndst;
e5d4b29f 2017 struct in6_addr saddr;
6f264e47 2018 u32 rt6i_flags;
e4c7ed41 2019
b1be00a6
JB
2020 if (!vxlan->vn6_sock)
2021 goto drop;
2022 sk = vxlan->vn6_sock->sock->sk;
2023
e5d4b29f 2024 ndst = vxlan6_get_route(vxlan, skb,
1400615d 2025 rdst ? rdst->remote_ifindex : 0, tos,
e7f70af1 2026 label, &dst->sin6.sin6_addr, &saddr,
db3c6139 2027 dst_cache, info);
e5d4b29f 2028 if (IS_ERR(ndst)) {
e4c7ed41
CW
2029 netdev_dbg(dev, "no route to %pI6\n",
2030 &dst->sin6.sin6_addr);
2031 dev->stats.tx_carrier_errors++;
9dcc71e1 2032 goto tx_error;
e4c7ed41 2033 }
d342894c 2034
e4c7ed41
CW
2035 if (ndst->dev == dev) {
2036 netdev_dbg(dev, "circular route to %pI6\n",
2037 &dst->sin6.sin6_addr);
2038 dst_release(ndst);
2039 dev->stats.collisions++;
2040 goto tx_error;
2041 }
0e6fbc5b 2042
e4c7ed41 2043 /* Bypass encapsulation if the destination is local */
6f264e47
JB
2044 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2045 if (rt6i_flags & RTF_LOCAL &&
2046 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW
2047 struct vxlan_dev *dst_vxlan;
2048
2049 dst_release(ndst);
19ca9fc1 2050 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG
2051 dst->sa.sa_family, dst_port,
2052 vxlan->flags);
e4c7ed41
CW
2053 if (!dst_vxlan)
2054 goto tx_error;
2055 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2056 return;
2057 }
49560532 2058
b4ed5cad
JB
2059 if (!info)
2060 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2061
1400615d 2062 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2063 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB
2064 skb_scrub_packet(skb, xnet);
2065 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2066 vni, md, flags, udp_sum);
f491e56d
JB
2067 if (err < 0) {
2068 dst_release(ndst);
2069 return;
2070 }
2071 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
e7f70af1
DB
2072 &saddr, &dst->sin6.sin6_addr, tos, ttl,
2073 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW
2074#endif
2075 }
0e6fbc5b 2076
4ad16930 2077 return;
d342894c 2078
2079drop:
2080 dev->stats.tx_dropped++;
2081 goto tx_free;
2082
f491e56d
JB
2083xmit_tx_error:
2084 /* skb is already freed. */
2085 skb = NULL;
49560532
PS
2086rt_tx_error:
2087 ip_rt_put(rt);
d342894c 2088tx_error:
2089 dev->stats.tx_errors++;
2090tx_free:
2091 dev_kfree_skb(skb);
d342894c 2092}
2093
6681712d
DS
2094/* Transmit local packets over Vxlan
2095 *
2096 * Outer IP header inherits ECN and DF from inner header.
2097 * Outer UDP destination is the VXLAN assigned port.
2098 * source port is based on hash of flow
2099 */
2100static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2101{
2102 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2103 const struct ip_tunnel_info *info;
6681712d
DS
2104 struct ethhdr *eth;
2105 bool did_rsc = false;
8f646c92 2106 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2107 struct vxlan_fdb *f;
6681712d 2108
61adedf3 2109 info = skb_tunnel_info(skb);
3093fbe7 2110
6681712d
DS
2111 skb_reset_mac_header(skb);
2112 eth = eth_hdr(skb);
2113
f564f45c
CW
2114 if ((vxlan->flags & VXLAN_F_PROXY)) {
2115 if (ntohs(eth->h_proto) == ETH_P_ARP)
2116 return arp_reduce(dev, skb);
2117#if IS_ENABLED(CONFIG_IPV6)
2118 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR
2119 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2120 + sizeof(struct nd_msg)) &&
f564f45c
CW
2121 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2122 struct nd_msg *msg;
2123
2124 msg = (struct nd_msg *)skb_transport_header(skb);
2125 if (msg->icmph.icmp6_code == 0 &&
2126 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2127 return neigh_reduce(dev, skb);
2128 }
7a9f526f 2129 eth = eth_hdr(skb);
f564f45c
CW
2130#endif
2131 }
6681712d 2132
f468a729
JB
2133 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2134 if (info && info->mode & IP_TUNNEL_INFO_TX)
2135 vxlan_xmit_one(skb, dev, NULL, false);
2136 else
2137 kfree_skb(skb);
ee122c79
TG
2138 return NETDEV_TX_OK;
2139 }
2140
6681712d 2141 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS
2142 did_rsc = false;
2143
2144 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW
2145 (ntohs(eth->h_proto) == ETH_P_IP ||
2146 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS
2147 did_rsc = route_shortcircuit(dev, skb);
2148 if (did_rsc)
2149 f = vxlan_find_mac(vxlan, eth->h_dest);
2150 }
2151
6681712d 2152 if (f == NULL) {
afbd8bae
MR
2153 f = vxlan_find_mac(vxlan, all_zeros_mac);
2154 if (f == NULL) {
2155 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2156 !is_multicast_ether_addr(eth->h_dest))
2157 vxlan_fdb_miss(vxlan, eth->h_dest);
2158
2159 dev->stats.tx_dropped++;
8f646c92 2160 kfree_skb(skb);
afbd8bae
MR
2161 return NETDEV_TX_OK;
2162 }
2163 }
6681712d 2164
afbd8bae
MR
2165 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2166 struct sk_buff *skb1;
6681712d 2167
8f646c92
ED
2168 if (!fdst) {
2169 fdst = rdst;
2170 continue;
2171 }
afbd8bae
MR
2172 skb1 = skb_clone(skb, GFP_ATOMIC);
2173 if (skb1)
2174 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS
2175 }
2176
8f646c92
ED
2177 if (fdst)
2178 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2179 else
2180 kfree_skb(skb);
4ad16930 2181 return NETDEV_TX_OK;
6681712d
DS
2182}
2183
d342894c 2184/* Walk the forwarding table and purge stale entries */
2185static void vxlan_cleanup(unsigned long arg)
2186{
2187 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2188 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2189 unsigned int h;
2190
2191 if (!netif_running(vxlan->dev))
2192 return;
2193
d342894c 2194 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2195 struct hlist_node *p, *n;
14e1d0fa
SD
2196
2197 spin_lock_bh(&vxlan->hash_lock);
d342894c 2198 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2199 struct vxlan_fdb *f
2200 = container_of(p, struct vxlan_fdb, hlist);
2201 unsigned long timeout;
2202
3c172868 2203 if (f->state & NUD_PERMANENT)
d342894c 2204 continue;
2205
0dfbdf41 2206 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2207 if (time_before_eq(timeout, jiffies)) {
2208 netdev_dbg(vxlan->dev,
2209 "garbage collect %pM\n",
2210 f->eth_addr);
2211 f->state = NUD_STALE;
2212 vxlan_fdb_destroy(vxlan, f);
2213 } else if (time_before(timeout, next_timer))
2214 next_timer = timeout;
2215 }
14e1d0fa 2216 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2217 }
d342894c 2218
2219 mod_timer(&vxlan->age_timer, next_timer);
2220}
2221
9c2e24e1
PS
2222static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2223{
56ef9c90 2224 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2225 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2226
56ef9c90 2227 spin_lock(&vn->sock_lock);
9c2e24e1 2228 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2229 spin_unlock(&vn->sock_lock);
9c2e24e1
PS
2230}
2231
d342894c 2232/* Setup stats when device is created */
2233static int vxlan_init(struct net_device *dev)
2234{
1c213bd2 2235 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2236 if (!dev->tstats)
d342894c 2237 return -ENOMEM;
2238
2239 return 0;
2240}
2241
ba609e9b 2242static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR
2243{
2244 struct vxlan_fdb *f;
2245
2246 spin_lock_bh(&vxlan->hash_lock);
2247 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2248 if (f)
2249 vxlan_fdb_destroy(vxlan, f);
2250 spin_unlock_bh(&vxlan->hash_lock);
2251}
2252
ebf4063e
SH
2253static void vxlan_uninit(struct net_device *dev)
2254{
2255 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2256
ba609e9b 2257 vxlan_fdb_delete_default(vxlan);
afbd8bae 2258
ebf4063e
SH
2259 free_percpu(dev->tstats);
2260}
2261
d342894c 2262/* Start ageing timer and join group when device is brought up */
2263static int vxlan_open(struct net_device *dev)
2264{
2265 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2266 int ret;
56ef9c90 2267
205f356d
JB
2268 ret = vxlan_sock_add(vxlan);
2269 if (ret < 0)
2270 return ret;
d342894c 2271
79d4a94f 2272 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2273 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL
2274 if (ret == -EADDRINUSE)
2275 ret = 0;
56ef9c90 2276 if (ret) {
205f356d 2277 vxlan_sock_release(vxlan);
56ef9c90
MRL
2278 return ret;
2279 }
d342894c 2280 }
2281
0dfbdf41 2282 if (vxlan->cfg.age_interval)
d342894c 2283 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2284
56ef9c90 2285 return ret;
d342894c 2286}
2287
2288/* Purge the forwarding table */
2289static void vxlan_flush(struct vxlan_dev *vxlan)
2290{
31fec5aa 2291 unsigned int h;
d342894c 2292
2293 spin_lock_bh(&vxlan->hash_lock);
2294 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2295 struct hlist_node *p, *n;
2296 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2297 struct vxlan_fdb *f
2298 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR
2299 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2300 if (!is_zero_ether_addr(f->eth_addr))
2301 vxlan_fdb_destroy(vxlan, f);
d342894c 2302 }
2303 }
2304 spin_unlock_bh(&vxlan->hash_lock);
2305}
2306
2307/* Cleanup timer and forwarding table on shutdown */
2308static int vxlan_stop(struct net_device *dev)
2309{
2310 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2311 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2312 int ret = 0;
d342894c 2313
24c0e683 2314 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2315 !vxlan_group_used(vn, vxlan))
56ef9c90 2316 ret = vxlan_igmp_leave(vxlan);
d342894c 2317
2318 del_timer_sync(&vxlan->age_timer);
2319
2320 vxlan_flush(vxlan);
205f356d 2321 vxlan_sock_release(vxlan);
d342894c 2322
56ef9c90 2323 return ret;
d342894c 2324}
2325
d342894c 2326/* Stub, nothing needs to be done. */
2327static void vxlan_set_multicast_list(struct net_device *dev)
2328{
2329}
2330
72564b59
DW
2331static int __vxlan_change_mtu(struct net_device *dev,
2332 struct net_device *lowerdev,
2333 struct vxlan_rdst *dst, int new_mtu, bool strict)
345010b5 2334{
72564b59 2335 int max_mtu = IP_MAX_MTU;
345010b5 2336
72564b59
DW
2337 if (lowerdev)
2338 max_mtu = lowerdev->mtu;
345010b5
DB
2339
2340 if (dst->remote_ip.sa.sa_family == AF_INET6)
72564b59 2341 max_mtu -= VXLAN6_HEADROOM;
345010b5 2342 else
72564b59 2343 max_mtu -= VXLAN_HEADROOM;
345010b5 2344
72564b59 2345 if (new_mtu < 68)
345010b5
DB
2346 return -EINVAL;
2347
72564b59
DW
2348 if (new_mtu > max_mtu) {
2349 if (strict)
2350 return -EINVAL;
2351
2352 new_mtu = max_mtu;
2353 }
2354
345010b5
DB
2355 dev->mtu = new_mtu;
2356 return 0;
2357}
2358
72564b59
DW
2359static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2360{
2361 struct vxlan_dev *vxlan = netdev_priv(dev);
2362 struct vxlan_rdst *dst = &vxlan->default_dst;
2363 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2364 dst->remote_ifindex);
2365 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2366}
2367
fc4099f1
PS
2368static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2369{
2370 struct vxlan_dev *vxlan = netdev_priv(dev);
2371 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2372 __be16 sport, dport;
2373
2374 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2375 vxlan->cfg.port_max, true);
2376 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2377
239e944f 2378 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB
2379 struct rtable *rt;
2380
239e944f
JB
2381 if (!vxlan->vn4_sock)
2382 return -EINVAL;
1a8496ba
JB
2383 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2384 info->key.u.ipv4.dst,
0c1d70af 2385 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB
2386 if (IS_ERR(rt))
2387 return PTR_ERR(rt);
2388 ip_rt_put(rt);
239e944f
JB
2389 } else {
2390#if IS_ENABLED(CONFIG_IPV6)
2391 struct dst_entry *ndst;
2392
2393 if (!vxlan->vn6_sock)
2394 return -EINVAL;
1400615d 2395 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
e7f70af1 2396 info->key.label, &info->key.u.ipv6.dst,
db3c6139 2397 &info->key.u.ipv6.src, NULL, info);
239e944f
JB
2398 if (IS_ERR(ndst))
2399 return PTR_ERR(ndst);
2400 dst_release(ndst);
239e944f
JB
2401#else /* !CONFIG_IPV6 */
2402 return -EPFNOSUPPORT;
2403#endif
2404 }
1a8496ba
JB
2405 info->key.tp_src = sport;
2406 info->key.tp_dst = dport;
239e944f 2407 return 0;
fc4099f1
PS
2408}
2409
d342894c 2410static const struct net_device_ops vxlan_netdev_ops = {
2411 .ndo_init = vxlan_init,
ebf4063e 2412 .ndo_uninit = vxlan_uninit,
d342894c 2413 .ndo_open = vxlan_open,
2414 .ndo_stop = vxlan_stop,
2415 .ndo_start_xmit = vxlan_xmit,
e8171045 2416 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2417 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2418 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2419 .ndo_validate_addr = eth_validate_addr,
2420 .ndo_set_mac_address = eth_mac_addr,
2421 .ndo_fdb_add = vxlan_fdb_add,
2422 .ndo_fdb_del = vxlan_fdb_delete,
2423 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2424 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2425};
2426
2427/* Info for udev, that this is a virtual tunnel endpoint */
2428static struct device_type vxlan_type = {
2429 .name = "vxlan",
2430};
2431
53cf5275 2432/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG
2433 * supply the listening VXLAN udp ports. Callers are expected
2434 * to implement the ndo_add_vxlan_port.
53cf5275
JG
2435 */
2436void vxlan_get_rx_port(struct net_device *dev)
2437{
2438 struct vxlan_sock *vs;
2439 struct net *net = dev_net(dev);
2440 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2441 sa_family_t sa_family;
35e42379
JG
2442 __be16 port;
2443 unsigned int i;
53cf5275
JG
2444
2445 spin_lock(&vn->sock_lock);
2446 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG
2447 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2448 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2449 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG
2450 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2451 port);
2452 }
2453 }
2454 spin_unlock(&vn->sock_lock);
2455}
2456EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2457
d342894c 2458/* Initialize the device structure. */
2459static void vxlan_setup(struct net_device *dev)
2460{
2461 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2462 unsigned int h;
d342894c 2463
2464 eth_hw_addr_random(dev);
2465 ether_setup(dev);
2466
2467 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2468 dev->destructor = free_netdev;
d342894c 2469 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2470
d342894c 2471 dev->features |= NETIF_F_LLTX;
d6727fe3 2472 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2473 dev->features |= NETIF_F_RXCSUM;
05c0db08 2474 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2475
1eaa8178
PS
2476 dev->vlan_features = dev->features;
2477 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2478 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2479 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2480 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2481 netif_keep_dst(dev);
82a0f6b4 2482 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
22dba393 2483 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2484
553675fb 2485 INIT_LIST_HEAD(&vxlan->next);
d342894c 2486 spin_lock_init(&vxlan->hash_lock);
2487
2488 init_timer_deferrable(&vxlan->age_timer);
2489 vxlan->age_timer.function = vxlan_cleanup;
2490 vxlan->age_timer.data = (unsigned long) vxlan;
2491
0dfbdf41 2492 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2493
d342894c 2494 vxlan->dev = dev;
2495
58ce31cc
TH
2496 gro_cells_init(&vxlan->gro_cells, dev);
2497
d342894c 2498 for (h = 0; h < FDB_HASH_SIZE; ++h)
2499 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2500}
2501
2502static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2503 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2504 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2505 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2506 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2507 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2508 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2509 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2510 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2511 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2512 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2513 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2514 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2515 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS
2516 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2517 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2518 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2519 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2520 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2521 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH
2522 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2523 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2524 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH
2525 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2526 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2527 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2528 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2529};
2530
2531static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2532{
2533 if (tb[IFLA_ADDRESS]) {
2534 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2535 pr_debug("invalid link address (not ethernet)\n");
2536 return -EINVAL;
2537 }
2538
2539 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2540 pr_debug("invalid all zero ethernet address\n");
2541 return -EADDRNOTAVAIL;
2542 }
2543 }
2544
2545 if (!data)
2546 return -EINVAL;
2547
2548 if (data[IFLA_VXLAN_ID]) {
2549 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2550 if (id >= VXLAN_VID_MASK)
2551 return -ERANGE;
2552 }
2553
05f47d69 2554 if (data[IFLA_VXLAN_PORT_RANGE]) {
2555 const struct ifla_vxlan_port_range *p
2556 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2557
2558 if (ntohs(p->high) < ntohs(p->low)) {
2559 pr_debug("port range %u .. %u not valid\n",
2560 ntohs(p->low), ntohs(p->high));
2561 return -EINVAL;
2562 }
2563 }
2564
d342894c 2565 return 0;
2566}
2567
1b13c97f
YB
2568static void vxlan_get_drvinfo(struct net_device *netdev,
2569 struct ethtool_drvinfo *drvinfo)
2570{
2571 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2572 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2573}
2574
2575static const struct ethtool_ops vxlan_ethtool_ops = {
2576 .get_drvinfo = vxlan_get_drvinfo,
2577 .get_link = ethtool_op_get_link,
2578};
2579
553675fb 2580static void vxlan_del_work(struct work_struct *work)
2581{
2582 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2583 udp_tunnel_sock_release(vs->sock);
553675fb 2584 kfree_rcu(vs, rcu);
2585}
2586
3ee64f39
TH
2587static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2588 __be16 port, u32 flags)
553675fb 2589{
e4c7ed41 2590 struct socket *sock;
3ee64f39
TH
2591 struct udp_port_cfg udp_conf;
2592 int err;
e4c7ed41 2593
3ee64f39 2594 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2595
3ee64f39
TH
2596 if (ipv6) {
2597 udp_conf.family = AF_INET6;
3ee64f39 2598 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2599 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2600 udp_conf.ipv6_v6only = 1;
3ee64f39
TH
2601 } else {
2602 udp_conf.family = AF_INET;
e4c7ed41 2603 }
e4c7ed41 2604
3ee64f39 2605 udp_conf.local_udp_port = port;
553675fb 2606
3ee64f39
TH
2607 /* Open UDP socket */
2608 err = udp_sock_create(net, &udp_conf, &sock);
2609 if (err < 0)
2610 return ERR_PTR(err);
e4c7ed41 2611
39deb2c7 2612 return sock;
e4c7ed41
CW
2613}
2614
2615/* Create new listen socket if needed */
b1be00a6
JB
2616static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2617 __be16 port, u32 flags)
e4c7ed41
CW
2618{
2619 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2620 struct vxlan_sock *vs;
2621 struct socket *sock;
e4c7ed41 2622 unsigned int h;
acbf74a7 2623 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2624
dc01e7d3 2625 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW
2626 if (!vs)
2627 return ERR_PTR(-ENOMEM);
2628
2629 for (h = 0; h < VNI_HASH_SIZE; ++h)
2630 INIT_HLIST_HEAD(&vs->vni_list[h]);
2631
2632 INIT_WORK(&vs->del_work, vxlan_del_work);
2633
3ee64f39 2634 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2635 if (IS_ERR(sock)) {
56ef9c90
MRL
2636 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2637 PTR_ERR(sock));
553675fb 2638 kfree(vs);
e50fddc8 2639 return ERR_CAST(sock);
553675fb 2640 }
e4c7ed41
CW
2641
2642 vs->sock = sock;
9c2e24e1 2643 atomic_set(&vs->refcnt, 1);
af33c1ad 2644 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2645
dc01e7d3
OG
2646 /* Initialize the vxlan udp offloads structure */
2647 vs->udp_offloads.port = port;
2648 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2649 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2650
9c2e24e1
PS
2651 spin_lock(&vn->sock_lock);
2652 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2653 vxlan_notify_add_rx_port(vs);
9c2e24e1 2654 spin_unlock(&vn->sock_lock);
553675fb 2655
2656 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ
2657 tunnel_cfg.sk_user_data = vs;
2658 tunnel_cfg.encap_type = 1;
f2d1968e 2659 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7
AZ
2660 tunnel_cfg.encap_destroy = NULL;
2661
2662 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2663
9c2e24e1
PS
2664 return vs;
2665}
2666
b1be00a6 2667static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2668{
205f356d
JB
2669 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2670 struct vxlan_sock *vs = NULL;
9c2e24e1 2671
205f356d 2672 if (!vxlan->cfg.no_share) {
56ef9c90 2673 spin_lock(&vn->sock_lock);
205f356d
JB
2674 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2675 vxlan->cfg.dst_port, vxlan->flags);
2676 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2677 spin_unlock(&vn->sock_lock);
205f356d 2678 return -EBUSY;
56ef9c90
MRL
2679 }
2680 spin_unlock(&vn->sock_lock);
2681 }
205f356d 2682 if (!vs)
b1be00a6
JB
2683 vs = vxlan_socket_create(vxlan->net, ipv6,
2684 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB
2685 if (IS_ERR(vs))
2686 return PTR_ERR(vs);
b1be00a6
JB
2687#if IS_ENABLED(CONFIG_IPV6)
2688 if (ipv6)
2689 vxlan->vn6_sock = vs;
2690 else
2691#endif
2692 vxlan->vn4_sock = vs;
205f356d
JB
2693 vxlan_vs_add_dev(vs, vxlan);
2694 return 0;
553675fb 2695}
2696
b1be00a6
JB
2697static int vxlan_sock_add(struct vxlan_dev *vxlan)
2698{
2699 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2700 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2701 int ret = 0;
2702
2703 vxlan->vn4_sock = NULL;
2704#if IS_ENABLED(CONFIG_IPV6)
2705 vxlan->vn6_sock = NULL;
2706 if (ipv6 || metadata)
2707 ret = __vxlan_sock_add(vxlan, true);
2708#endif
2709 if (!ret && (!ipv6 || metadata))
2710 ret = __vxlan_sock_add(vxlan, false);
2711 if (ret < 0)
2712 vxlan_sock_release(vxlan);
2713 return ret;
2714}
2715
0dfbdf41
TG
2716static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2717 struct vxlan_config *conf)
d342894c 2718{
33564bbb 2719 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2720 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2721 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2722 unsigned short needed_headroom = ETH_HLEN;
d342894c 2723 int err;
e4c7ed41 2724 bool use_ipv6 = false;
0dfbdf41 2725 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2726 struct net_device *lowerdev = NULL;
d342894c 2727
33564bbb 2728 vxlan->net = src_net;
f01ec1c0 2729
0dfbdf41 2730 dst->remote_vni = conf->vni;
d342894c 2731
0dfbdf41 2732 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2733
0dfbdf41
TG
2734 /* Unless IPv6 is explicitly requested, assume IPv4 */
2735 if (!dst->remote_ip.sa.sa_family)
2736 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2737
0dfbdf41 2738 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB
2739 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2740 if (!IS_ENABLED(CONFIG_IPV6))
2741 return -EPFNOSUPPORT;
e4c7ed41 2742 use_ipv6 = true;
b1be00a6 2743 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2744 }
d342894c 2745
e7f70af1
DB
2746 if (conf->label && !use_ipv6) {
2747 pr_info("label only supported in use with IPv6\n");
2748 return -EINVAL;
2749 }
2750
0dfbdf41 2751 if (conf->remote_ifindex) {
7e059158 2752 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2753 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2754
2755 if (!lowerdev) {
c7995c43 2756 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2757 return -ENODEV;
2758 }
d342894c 2759
e4c7ed41
CW
2760#if IS_ENABLED(CONFIG_IPV6)
2761 if (use_ipv6) {
2762 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2763 if (idev && idev->cnf.disable_ipv6) {
2764 pr_info("IPv6 is disabled via sysctl\n");
2765 return -EPERM;
2766 }
e4c7ed41
CW
2767 }
2768#endif
2769
0dfbdf41 2770 if (!conf->mtu)
e4c7ed41 2771 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2772
b1be00a6 2773 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2774 }
d342894c 2775
7e059158
DW
2776 if (conf->mtu) {
2777 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2778 if (err)
2779 return err;
2780 }
2781
b1be00a6
JB
2782 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2783 needed_headroom += VXLAN6_HEADROOM;
2784 else
2785 needed_headroom += VXLAN_HEADROOM;
2786 dev->needed_headroom = needed_headroom;
2787
0dfbdf41
TG
2788 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2789 if (!vxlan->cfg.dst_port)
2790 vxlan->cfg.dst_port = default_port;
2791 vxlan->flags |= conf->flags;
2792
2793 if (!vxlan->cfg.age_interval)
2794 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2795
07b9b37c
ND
2796 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2797 if (tmp->cfg.vni == conf->vni &&
2798 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2799 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2800 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2801 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2802 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2803 return -EEXIST;
07b9b37c 2804 }
0dfbdf41
TG
2805
2806 dev->ethtool_ops = &vxlan_ethtool_ops;
2807
2808 /* create an fdb entry for a valid default destination */
2809 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2810 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2811 &vxlan->default_dst.remote_ip,
2812 NUD_REACHABLE|NUD_PERMANENT,
2813 NLM_F_EXCL|NLM_F_CREATE,
2814 vxlan->cfg.dst_port,
2815 vxlan->default_dst.remote_vni,
2816 vxlan->default_dst.remote_ifindex,
2817 NTF_SELF);
2818 if (err)
2819 return err;
2820 }
2821
2822 err = register_netdevice(dev);
2823 if (err) {
2824 vxlan_fdb_delete_default(vxlan);
2825 return err;
2826 }
2827
2828 list_add(&vxlan->next, &vn->vxlan_list);
2829
2830 return 0;
2831}
2832
2833struct net_device *vxlan_dev_create(struct net *net, const char *name,
2834 u8 name_assign_type, struct vxlan_config *conf)
2835{
2836 struct nlattr *tb[IFLA_MAX+1];
2837 struct net_device *dev;
2838 int err;
2839
2840 memset(&tb, 0, sizeof(tb));
2841
2842 dev = rtnl_create_link(net, name, name_assign_type,
2843 &vxlan_link_ops, tb);
2844 if (IS_ERR(dev))
2845 return dev;
2846
2847 err = vxlan_dev_configure(net, dev, conf);
2848 if (err < 0) {
2849 free_netdev(dev);
2850 return ERR_PTR(err);
2851 }
2852
2853 return dev;
2854}
2855EXPORT_SYMBOL_GPL(vxlan_dev_create);
2856
2857static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2858 struct nlattr *tb[], struct nlattr *data[])
2859{
2860 struct vxlan_config conf;
2861 int err;
2862
0dfbdf41 2863 memset(&conf, 0, sizeof(conf));
e277de5f
JG
2864
2865 if (data[IFLA_VXLAN_ID])
54bfd872 2866 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG
2867
2868 if (data[IFLA_VXLAN_GROUP]) {
2869 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2870 } else if (data[IFLA_VXLAN_GROUP6]) {
2871 if (!IS_ENABLED(CONFIG_IPV6))
2872 return -EPFNOSUPPORT;
2873
2874 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2875 conf.remote_ip.sa.sa_family = AF_INET6;
2876 }
2877
2878 if (data[IFLA_VXLAN_LOCAL]) {
2879 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2880 conf.saddr.sa.sa_family = AF_INET;
2881 } else if (data[IFLA_VXLAN_LOCAL6]) {
2882 if (!IS_ENABLED(CONFIG_IPV6))
2883 return -EPFNOSUPPORT;
2884
2885 /* TODO: respect scope id */
2886 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2887 conf.saddr.sa.sa_family = AF_INET6;
2888 }
2889
2890 if (data[IFLA_VXLAN_LINK])
2891 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2892
d342894c 2893 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2894 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2895
afb97186 2896 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2897 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2898
e7f70af1
DB
2899 if (data[IFLA_VXLAN_LABEL])
2900 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
2901 IPV6_FLOWLABEL_MASK;
2902
d342894c 2903 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2904 conf.flags |= VXLAN_F_LEARN;
d342894c 2905
2906 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2907 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2908
e4f67add 2909 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2910 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS
2911
2912 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2913 conf.flags |= VXLAN_F_RSC;
e4f67add
DS
2914
2915 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2916 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS
2917
2918 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2919 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2920
d342894c 2921 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2922 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2923
f8a9b1bc
AS
2924 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2925 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2926 conf.flags |= VXLAN_F_COLLECT_METADATA;
2927
05f47d69 2928 if (data[IFLA_VXLAN_PORT_RANGE]) {
2929 const struct ifla_vxlan_port_range *p
2930 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG
2931 conf.port_min = ntohs(p->low);
2932 conf.port_max = ntohs(p->high);
05f47d69 2933 }
2934
823aa873 2935 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2936 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2937
6ceb31ca
AD
2938 if (data[IFLA_VXLAN_UDP_CSUM] &&
2939 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2940 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH
2941
2942 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2943 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2944 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH
2945
2946 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2947 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2948 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2949
dfd8645e
TH
2950 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2951 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2952 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH
2953
2954 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2955 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2956 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2957
3511494c 2958 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2959 conf.flags |= VXLAN_F_GBP;
3511494c 2960
0ace2ca8 2961 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2962 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2963
0dfbdf41
TG
2964 err = vxlan_dev_configure(src_net, dev, &conf);
2965 switch (err) {
2966 case -ENODEV:
2967 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2968 break;
1b13c97f 2969
0dfbdf41
TG
2970 case -EPERM:
2971 pr_info("IPv6 is disabled via sysctl\n");
2972 break;
d342894c 2973
0dfbdf41 2974 case -EEXIST:
54bfd872 2975 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
0dfbdf41 2976 break;
afbd8bae
MR
2977 }
2978
0dfbdf41 2979 return err;
d342894c 2980}
2981
2982static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2983{
2984 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2985 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2986
fe5c3561 2987 spin_lock(&vn->sock_lock);
9c2e24e1
PS
2988 if (!hlist_unhashed(&vxlan->hlist))
2989 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2990 spin_unlock(&vn->sock_lock);
2991
58ce31cc 2992 gro_cells_destroy(&vxlan->gro_cells);
553675fb 2993 list_del(&vxlan->next);
d342894c 2994 unregister_netdevice_queue(dev, head);
2995}
2996
2997static size_t vxlan_get_size(const struct net_device *dev)
2998{
2999
3000 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3001 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3002 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3003 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3005 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3006 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3007 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS
3008 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3009 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3010 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3011 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3012 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3013 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3014 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3015 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH
3016 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3017 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3018 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3019 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH
3020 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3021 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3022 0;
3023}
3024
3025static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3026{
3027 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3028 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3029 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG
3030 .low = htons(vxlan->cfg.port_min),
3031 .high = htons(vxlan->cfg.port_max),
05f47d69 3032 };
d342894c 3033
54bfd872 3034 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3035 goto nla_put_failure;
3036
e4c7ed41
CW
3037 if (!vxlan_addr_any(&dst->remote_ip)) {
3038 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB
3039 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3040 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW
3041 goto nla_put_failure;
3042#if IS_ENABLED(CONFIG_IPV6)
3043 } else {
930345ea
JB
3044 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3045 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW
3046 goto nla_put_failure;
3047#endif
3048 }
3049 }
d342894c 3050
c7995c43 3051 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3052 goto nla_put_failure;
3053
0dfbdf41
TG
3054 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3055 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3056 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3057 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW
3058 goto nla_put_failure;
3059#if IS_ENABLED(CONFIG_IPV6)
3060 } else {
930345ea 3061 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3062 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW
3063 goto nla_put_failure;
3064#endif
3065 }
3066 }
d342894c 3067
0dfbdf41
TG
3068 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3069 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3070 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS
3071 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3072 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3073 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3074 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3075 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3076 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3077 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3078 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3079 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS
3080 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3081 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG
3082 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3083 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3084 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3085 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3086 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH
3087 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3088 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3089 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH
3090 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3091 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3092 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3093 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3094 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3095 goto nla_put_failure;
3096
05f47d69 3097 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3098 goto nla_put_failure;
3099
3511494c
TG
3100 if (vxlan->flags & VXLAN_F_GBP &&
3101 nla_put_flag(skb, IFLA_VXLAN_GBP))
3102 goto nla_put_failure;
3103
0ace2ca8
TH
3104 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3105 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3106 goto nla_put_failure;
3107
d342894c 3108 return 0;
3109
3110nla_put_failure:
3111 return -EMSGSIZE;
3112}
3113
1728d4fa
ND
3114static struct net *vxlan_get_link_net(const struct net_device *dev)
3115{
3116 struct vxlan_dev *vxlan = netdev_priv(dev);
3117
3118 return vxlan->net;
3119}
3120
d342894c 3121static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3122 .kind = "vxlan",
3123 .maxtype = IFLA_VXLAN_MAX,
3124 .policy = vxlan_policy,
3125 .priv_size = sizeof(struct vxlan_dev),
3126 .setup = vxlan_setup,
3127 .validate = vxlan_validate,
3128 .newlink = vxlan_newlink,
3129 .dellink = vxlan_dellink,
3130 .get_size = vxlan_get_size,
3131 .fill_info = vxlan_fill_info,
1728d4fa 3132 .get_link_net = vxlan_get_link_net,
d342894c 3133};
3134
acaf4e70
DB
3135static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3136 struct net_device *dev)
3137{
3138 struct vxlan_dev *vxlan, *next;
3139 LIST_HEAD(list_kill);
3140
3141 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3142 struct vxlan_rdst *dst = &vxlan->default_dst;
3143
3144 /* In case we created vxlan device with carrier
3145 * and we loose the carrier due to module unload
3146 * we also need to remove vxlan device. In other
3147 * cases, it's not necessary and remote_ifindex
3148 * is 0 here, so no matches.
3149 */
3150 if (dst->remote_ifindex == dev->ifindex)
3151 vxlan_dellink(vxlan->dev, &list_kill);
3152 }
3153
3154 unregister_netdevice_many(&list_kill);
3155}
3156
3157static int vxlan_lowerdev_event(struct notifier_block *unused,
3158 unsigned long event, void *ptr)
3159{
3160 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3161 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3162
783c1463 3163 if (event == NETDEV_UNREGISTER)
acaf4e70
DB
3164 vxlan_handle_lowerdev_unregister(vn, dev);
3165
3166 return NOTIFY_DONE;
3167}
3168
3169static struct notifier_block vxlan_notifier_block __read_mostly = {
3170 .notifier_call = vxlan_lowerdev_event,
3171};
3172
d342894c 3173static __net_init int vxlan_init_net(struct net *net)
3174{
3175 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3176 unsigned int h;
d342894c 3177
553675fb 3178 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3179 spin_lock_init(&vn->sock_lock);
d342894c 3180
553675fb 3181 for (h = 0; h < PORT_HASH_SIZE; ++h)
3182 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3183
3184 return 0;
3185}
3186
f01ec1c0
ND
3187static void __net_exit vxlan_exit_net(struct net *net)
3188{
3189 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3190 struct vxlan_dev *vxlan, *next;
3191 struct net_device *dev, *aux;
3192 LIST_HEAD(list);
3193
3194 rtnl_lock();
3195 for_each_netdev_safe(net, dev, aux)
3196 if (dev->rtnl_link_ops == &vxlan_link_ops)
3197 unregister_netdevice_queue(dev, &list);
3198
3199 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3200 /* If vxlan->dev is in the same netns, it has already been added
3201 * to the list by the previous loop.
3202 */
58ce31cc
TH
3203 if (!net_eq(dev_net(vxlan->dev), net)) {
3204 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3205 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3206 }
f01ec1c0
ND
3207 }
3208
3209 unregister_netdevice_many(&list);
3210 rtnl_unlock();
3211}
3212
d342894c 3213static struct pernet_operations vxlan_net_ops = {
3214 .init = vxlan_init_net,
f01ec1c0 3215 .exit = vxlan_exit_net,
d342894c 3216 .id = &vxlan_net_id,
3217 .size = sizeof(struct vxlan_net),
3218};
3219
3220static int __init vxlan_init_module(void)
3221{
3222 int rc;
3223
758c57d1
SH
3224 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3225 if (!vxlan_wq)
3226 return -ENOMEM;
3227
d342894c 3228 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3229
783c1463 3230 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3231 if (rc)
3232 goto out1;
3233
acaf4e70 3234 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3235 if (rc)
3236 goto out2;
3237
acaf4e70
DB
3238 rc = rtnl_link_register(&vxlan_link_ops);
3239 if (rc)
3240 goto out3;
d342894c 3241
acaf4e70
DB
3242 return 0;
3243out3:
3244 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3245out2:
783c1463 3246 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3247out1:
758c57d1 3248 destroy_workqueue(vxlan_wq);
d342894c 3249 return rc;
3250}
7332a13b 3251late_initcall(vxlan_init_module);
d342894c 3252
3253static void __exit vxlan_cleanup_module(void)
3254{
b7153984 3255 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3256 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3257 destroy_workqueue(vxlan_wq);
783c1463
DB
3258 unregister_pernet_subsys(&vxlan_net_ops);
3259 /* rcu_barrier() is called by netns */
d342894c 3260}
3261module_exit(vxlan_cleanup_module);
3262
3263MODULE_LICENSE("GPL");
3264MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3265MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3266MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3267MODULE_ALIAS_RTNL_LINK("vxlan");