]> Git Repo - linux.git/blame - drivers/net/vxlan.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'char-misc-3.20-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
[linux.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
acbf74a7 45#include <net/udp_tunnel.h>
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
d342894c 52
53#define VXLAN_VERSION "0.1"
54
553675fb 55#define PORT_HASH_BITS 8
56#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 57#define VNI_HASH_BITS 10
58#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
59#define FDB_HASH_BITS 8
60#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
61#define FDB_AGE_DEFAULT 300 /* 5 min */
62#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
63
23c578bf 64/* UDP port for VXLAN traffic.
65 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 66 * for compatibility with early adopters.
23c578bf 67 */
9daaa397
SH		 68static unsigned short vxlan_port __read_mostly = 8472;
69module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 70MODULE_PARM_DESC(udp_port, "Destination UDP port");
71
72static bool log_ecn_error = true;
73module_param(log_ecn_error, bool, 0644);
74MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
75
60d9d4c6 76static int vxlan_net_id;
553675fb 77
afbd8bae
MR		 78static const u8 all_zeros_mac[ETH_ALEN];
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
e4c7ed41
CW		 87union vxlan_addr {
88 struct sockaddr_in sin;
89 struct sockaddr_in6 sin6;
90 struct sockaddr sa;
91};
92
6681712d 93struct vxlan_rdst {
e4c7ed41 94 union vxlan_addr remote_ip;
6681712d
DS		 95 __be16 remote_port;
96 u32 remote_vni;
97 u32 remote_ifindex;
3e61aa8f 98 struct list_head list;
bc7892ba 99 struct rcu_head rcu;
6681712d
DS		 100};
101
d342894c 102/* Forwarding table entry */
103struct vxlan_fdb {
104 struct hlist_node hlist; /* linked list of entries */
105 struct rcu_head rcu;
106 unsigned long updated; /* jiffies */
107 unsigned long used;
3e61aa8f 108 struct list_head remotes;
d342894c 109 u16 state; /* see ndm_state */
ae884082 110 u8 flags; /* see ndm_flags */
d342894c 111 u8 eth_addr[ETH_ALEN];
112};
113
d342894c 114/* Pseudo network device */
115struct vxlan_dev {
553675fb 116 struct hlist_node hlist; /* vni hash table */
117 struct list_head next; /* vxlan's per namespace list */
118 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 119 struct net_device *dev;
f01ec1c0 120 struct net *net; /* netns for packet i/o */
c7995c43 121 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 122 union vxlan_addr saddr; /* source address */
823aa873 123 __be16 dst_port;
05f47d69 124 __u16 port_min; /* source port range */
125 __u16 port_max;
d342894c 126 __u8 tos; /* TOS override */
127 __u8 ttl;
359a0ea9 128 u32 flags; /* VXLAN_F_* in vxlan.h */
d342894c 129
1c51a915 130 struct work_struct sock_work;
3fc2de2f 131 struct work_struct igmp_join;
132 struct work_struct igmp_leave;
1c51a915 133
d342894c 134 unsigned long age_interval;
135 struct timer_list age_timer;
136 spinlock_t hash_lock;
137 unsigned int addrcnt;
138 unsigned int addrmax;
d342894c 139
140 struct hlist_head fdb_head[FDB_HASH_SIZE];
141};
142
143/* salt for hash table */
144static u32 vxlan_salt __read_mostly;
758c57d1 145static struct workqueue_struct *vxlan_wq;
d342894c 146
1c51a915
SH		 147static void vxlan_sock_work(struct work_struct *work);
148
e4c7ed41
CW		 149#if IS_ENABLED(CONFIG_IPV6)
150static inline
151bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
152{
153 if (a->sa.sa_family != b->sa.sa_family)
154 return false;
155 if (a->sa.sa_family == AF_INET6)
156 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
157 else
158 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
159}
160
161static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
162{
163 if (ipa->sa.sa_family == AF_INET6)
164 return ipv6_addr_any(&ipa->sin6.sin6_addr);
165 else
166 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
167}
168
169static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
170{
171 if (ipa->sa.sa_family == AF_INET6)
172 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
173 else
174 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
175}
176
177static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
178{
179 if (nla_len(nla) >= sizeof(struct in6_addr)) {
180 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
181 ip->sa.sa_family = AF_INET6;
182 return 0;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
194{
195 if (ip->sa.sa_family == AF_INET6)
196 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
197 else
198 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
199}
200
201#else /* !CONFIG_IPV6 */
202
203static inline
204bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
205{
206 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
207}
208
209static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
210{
211 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
212}
213
214static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
215{
216 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
217}
218
219static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
220{
221 if (nla_len(nla) >= sizeof(struct in6_addr)) {
222 return -EAFNOSUPPORT;
223 } else if (nla_len(nla) >= sizeof(__be32)) {
224 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
225 ip->sa.sa_family = AF_INET;
226 return 0;
227 } else {
228 return -EAFNOSUPPORT;
229 }
230}
231
232static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
233 const union vxlan_addr *ip)
234{
235 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
236}
237#endif
238
553675fb 239/* Virtual Network hash table head */
240static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
241{
242 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
243}
244
245/* Socket hash table head */
246static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 247{
248 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
249
553675fb 250 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
251}
252
3e61aa8f
SH		 253/* First remote destination for a forwarding entry.
254 * Guaranteed to be non-NULL because remotes are never deleted.
255 */
5ca5461c 256static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 257{
5ca5461c 258 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
259}
260
261static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
262{
263 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 264}
265
ac5132d1
TG		 266/* Find VXLAN socket based on network namespace, address family and UDP port
267 * and enabled unshareable flags.
268 */
269static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
270 __be16 port, u32 flags)
553675fb 271{
272 struct vxlan_sock *vs;
af33c1ad
TH		 273
274 flags &= VXLAN_F_RCV_FLAGS;
553675fb 275
276 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 277 if (inet_sk(vs->sock->sk)->inet_sport == port &&
ac5132d1 278 inet_sk(vs->sock->sk)->sk.sk_family == family &&
af33c1ad 279 vs->flags == flags)
553675fb 280 return vs;
281 }
282 return NULL;
d342894c 283}
284
5cfccc5a 285static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 286{
287 struct vxlan_dev *vxlan;
d342894c 288
553675fb 289 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 290 if (vxlan->default_dst.remote_vni == id)
d342894c 291 return vxlan;
292 }
293
294 return NULL;
295}
296
5cfccc5a 297/* Look up VNI in a per net namespace table */
19ca9fc1 298static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 299 sa_family_t family, __be16 port,
300 u32 flags)
5cfccc5a
PS		 301{
302 struct vxlan_sock *vs;
303
ac5132d1 304 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 305 if (!vs)
306 return NULL;
307
308 return vxlan_vs_find_vni(vs, id);
309}
310
d342894c 311/* Fill in neighbour message in skbuff. */
312static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 313 const struct vxlan_fdb *fdb,
314 u32 portid, u32 seq, int type, unsigned int flags,
315 const struct vxlan_rdst *rdst)
d342894c 316{
317 unsigned long now = jiffies;
318 struct nda_cacheinfo ci;
319 struct nlmsghdr *nlh;
320 struct ndmsg *ndm;
e4f67add 321 bool send_ip, send_eth;
d342894c 322
323 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
324 if (nlh == NULL)
325 return -EMSGSIZE;
326
327 ndm = nlmsg_data(nlh);
328 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 329
330 send_eth = send_ip = true;
331
332 if (type == RTM_GETNEIGH) {
333 ndm->ndm_family = AF_INET;
e4c7ed41 334 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 335 send_eth = !is_zero_ether_addr(fdb->eth_addr);
336 } else
337 ndm->ndm_family = AF_BRIDGE;
d342894c 338 ndm->ndm_state = fdb->state;
339 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 340 ndm->ndm_flags = fdb->flags;
545469f7 341 ndm->ndm_type = RTN_UNICAST;
d342894c 342
193523bf 343 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b
ND		 344 nla_put_s32(skb, NDA_LINK_NETNSID,
345 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 346 goto nla_put_failure;
347
e4f67add 348 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 349 goto nla_put_failure;
350
e4c7ed41 351 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 352 goto nla_put_failure;
353
823aa873 354 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 355 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
356 goto nla_put_failure;
c7995c43 357 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 358 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 359 goto nla_put_failure;
360 if (rdst->remote_ifindex &&
361 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 362 goto nla_put_failure;
363
364 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
365 ci.ndm_confirmed = 0;
366 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
367 ci.ndm_refcnt = 0;
368
369 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
370 goto nla_put_failure;
371
053c095a
JB		 372 nlmsg_end(skb, nlh);
373 return 0;
d342894c 374
375nla_put_failure:
376 nlmsg_cancel(skb, nlh);
377 return -EMSGSIZE;
378}
379
380static inline size_t vxlan_nlmsg_size(void)
381{
382 return NLMSG_ALIGN(sizeof(struct ndmsg))
383 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 384 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 385 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 386 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
387 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 388 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 389 + nla_total_size(sizeof(struct nda_cacheinfo));
390}
391
9e4b93f9
ND		 392static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
393 struct vxlan_rdst *rd, int type)
d342894c 394{
395 struct net *net = dev_net(vxlan->dev);
396 struct sk_buff *skb;
397 int err = -ENOBUFS;
398
399 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
400 if (skb == NULL)
401 goto errout;
402
9e4b93f9 403 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 404 if (err < 0) {
405 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
406 WARN_ON(err == -EMSGSIZE);
407 kfree_skb(skb);
408 goto errout;
409 }
410
411 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
412 return;
413errout:
414 if (err < 0)
415 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
416}
417
e4c7ed41 418static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 419{
420 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 421 struct vxlan_fdb f = {
422 .state = NUD_STALE,
423 };
424 struct vxlan_rdst remote = {
e4c7ed41 425 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 426 .remote_vni = VXLAN_N_VID,
427 };
3e61aa8f 428
9e4b93f9 429 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 430}
431
432static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
433{
bb3fd687
SH		 434 struct vxlan_fdb f = {
435 .state = NUD_STALE,
436 };
9e4b93f9 437 struct vxlan_rdst remote = { };
e4f67add 438
e4f67add
DS		 439 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
440
9e4b93f9 441 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 442}
443
d342894c 444/* Hash Ethernet address */
445static u32 eth_hash(const unsigned char *addr)
446{
447 u64 value = get_unaligned((u64 *)addr);
448
449 /* only want 6 bytes */
450#ifdef __BIG_ENDIAN
d342894c 451 value >>= 16;
321fb991 452#else
453 value <<= 16;
d342894c 454#endif
455 return hash_64(value, FDB_HASH_BITS);
456}
457
458/* Hash chain to use given mac address */
459static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
460 const u8 *mac)
461{
462 return &vxlan->fdb_head[eth_hash(mac)];
463}
464
465/* Look up Ethernet address in forwarding table */
014be2c8 466static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 467 const u8 *mac)
d342894c 468{
469 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
470 struct vxlan_fdb *f;
d342894c 471
b67bfe0d 472 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 473 if (ether_addr_equal(mac, f->eth_addr))
d342894c 474 return f;
475 }
476
477 return NULL;
478}
479
014be2c8
SS		 480static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
481 const u8 *mac)
482{
483 struct vxlan_fdb *f;
484
485 f = __vxlan_find_mac(vxlan, mac);
486 if (f)
487 f->used = jiffies;
488
489 return f;
490}
491
a5e7c10a
MR		 492/* caller should hold vxlan->hash_lock */
493static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 494 union vxlan_addr *ip, __be16 port,
a5e7c10a 495 __u32 vni, __u32 ifindex)
6681712d 496{
3e61aa8f 497 struct vxlan_rdst *rd;
6681712d 498
3e61aa8f 499 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 500 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 501 rd->remote_port == port &&
502 rd->remote_vni == vni &&
503 rd->remote_ifindex == ifindex)
a5e7c10a 504 return rd;
6681712d 505 }
3e61aa8f 506
a5e7c10a
MR		 507 return NULL;
508}
509
906dc186
TR		 510/* Replace destination of unicast mac */
511static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 512 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 513{
514 struct vxlan_rdst *rd;
515
516 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
517 if (rd)
518 return 0;
519
520 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
521 if (!rd)
522 return 0;
e4c7ed41 523 rd->remote_ip = *ip;
906dc186
TR		 524 rd->remote_port = port;
525 rd->remote_vni = vni;
526 rd->remote_ifindex = ifindex;
527 return 1;
528}
529
a5e7c10a
MR		 530/* Add/update destinations for multicast */
531static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 532 union vxlan_addr *ip, __be16 port, __u32 vni,
533 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 534{
535 struct vxlan_rdst *rd;
536
537 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
538 if (rd)
539 return 0;
540
6681712d
DS		 541 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
542 if (rd == NULL)
543 return -ENOBUFS;
e4c7ed41 544 rd->remote_ip = *ip;
6681712d
DS		 545 rd->remote_port = port;
546 rd->remote_vni = vni;
547 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 548
549 list_add_tail_rcu(&rd->list, &f->remotes);
550
9e4b93f9 551 *rdp = rd;
6681712d
DS		 552 return 1;
553}
554
dfd8645e
TH		 555static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
556 unsigned int off,
557 struct vxlanhdr *vh, size_t hdrlen,
558 u32 data)
559{
560 size_t start, offset, plen;
dfd8645e
TH		 561
562 if (skb->remcsum_offload)
563 return vh;
564
565 if (!NAPI_GRO_CB(skb)->csum_valid)
566 return NULL;
567
568 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
569 offset = start + ((data & VXLAN_RCO_UDP) ?
570 offsetof(struct udphdr, check) :
571 offsetof(struct tcphdr, check));
572
573 plen = hdrlen + offset + sizeof(u16);
574
575 /* Pull checksum that will be written */
576 if (skb_gro_header_hard(skb, off + plen)) {
577 vh = skb_gro_header_slow(skb, off + plen, off);
578 if (!vh)
579 return NULL;
580 }
581
dcdc8994 582 skb_gro_remcsum_process(skb, (void *)vh + hdrlen, start, offset);
dfd8645e
TH		 583
584 skb->remcsum_offload = 1;
585
586 return vh;
587}
588
a2b12f3c
TH		 589static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
590 struct sk_buff *skb,
591 struct udp_offload *uoff)
dc01e7d3
OG		 592{
593 struct sk_buff *p, **pp = NULL;
594 struct vxlanhdr *vh, *vh2;
9b174d88 595 unsigned int hlen, off_vx;
dc01e7d3 596 int flush = 1;
dfd8645e
TH		 597 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
598 udp_offloads);
599 u32 flags;
dc01e7d3
OG		 600
601 off_vx = skb_gro_offset(skb);
602 hlen = off_vx + sizeof(*vh);
603 vh = skb_gro_header_fast(skb, off_vx);
604 if (skb_gro_header_hard(skb, hlen)) {
605 vh = skb_gro_header_slow(skb, hlen, off_vx);
606 if (unlikely(!vh))
607 goto out;
608 }
dc01e7d3 609
dfd8645e
TH		 610 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
611 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
612
613 flags = ntohl(vh->vx_flags);
614
615 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
616 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
617 ntohl(vh->vx_vni));
618
619 if (!vh)
620 goto out;
621 }
622
dc01e7d3
OG		 623 flush = 0;
624
625 for (p = *head; p; p = p->next) {
626 if (!NAPI_GRO_CB(p)->same_flow)
627 continue;
628
629 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 630 if (vh->vx_flags != vh2->vx_flags ||
631 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 632 NAPI_GRO_CB(p)->same_flow = 0;
633 continue;
634 }
dc01e7d3
OG		 635 }
636
9b174d88 637 pp = eth_gro_receive(head, skb);
dc01e7d3 638
dc01e7d3
OG		 639out:
640 NAPI_GRO_CB(skb)->flush |= flush;
641
642 return pp;
643}
644
a2b12f3c
TH		 645static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
646 struct udp_offload *uoff)
dc01e7d3 647{
cfdf1e1b
JG		 648 udp_tunnel_gro_complete(skb, nhoff);
649
9b174d88 650 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 651}
652
53cf5275 653/* Notify netdevs that UDP port started listening */
dc01e7d3 654static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 655{
656 struct net_device *dev;
dc01e7d3 657 struct sock *sk = vs->sock->sk;
53cf5275
JG		 658 struct net *net = sock_net(sk);
659 sa_family_t sa_family = sk->sk_family;
35e42379 660 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 661 int err;
662
663 if (sa_family == AF_INET) {
664 err = udp_add_offload(&vs->udp_offloads);
665 if (err)
666 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
667 }
53cf5275
JG		 668
669 rcu_read_lock();
670 for_each_netdev_rcu(net, dev) {
671 if (dev->netdev_ops->ndo_add_vxlan_port)
672 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
673 port);
674 }
675 rcu_read_unlock();
676}
677
678/* Notify netdevs that UDP port is no more listening */
dc01e7d3 679static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 680{
681 struct net_device *dev;
dc01e7d3 682 struct sock *sk = vs->sock->sk;
53cf5275
JG		 683 struct net *net = sock_net(sk);
684 sa_family_t sa_family = sk->sk_family;
35e42379 685 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 686
687 rcu_read_lock();
688 for_each_netdev_rcu(net, dev) {
689 if (dev->netdev_ops->ndo_del_vxlan_port)
690 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
691 port);
692 }
693 rcu_read_unlock();
dc01e7d3
OG		 694
695 if (sa_family == AF_INET)
696 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 697}
698
d342894c 699/* Add new entry to forwarding table -- assumes lock held */
700static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 701 const u8 *mac, union vxlan_addr *ip,
6681712d 702 __u16 state, __u16 flags,
73cf3317 703 __be16 port, __u32 vni, __u32 ifindex,
ae884082 704 __u8 ndm_flags)
d342894c 705{
9e4b93f9 706 struct vxlan_rdst *rd = NULL;
d342894c 707 struct vxlan_fdb *f;
708 int notify = 0;
709
014be2c8 710 f = __vxlan_find_mac(vxlan, mac);
d342894c 711 if (f) {
712 if (flags & NLM_F_EXCL) {
713 netdev_dbg(vxlan->dev,
714 "lost race to create %pM\n", mac);
715 return -EEXIST;
716 }
717 if (f->state != state) {
718 f->state = state;
719 f->updated = jiffies;
720 notify = 1;
721 }
ae884082
DS		 722 if (f->flags != ndm_flags) {
723 f->flags = ndm_flags;
724 f->updated = jiffies;
725 notify = 1;
726 }
906dc186
TR		 727 if ((flags & NLM_F_REPLACE)) {
728 /* Only change unicasts */
729 if (!(is_multicast_ether_addr(f->eth_addr) ||
730 is_zero_ether_addr(f->eth_addr))) {
731 int rc = vxlan_fdb_replace(f, ip, port, vni,
732 ifindex);
733
734 if (rc < 0)
735 return rc;
736 notify |= rc;
737 } else
738 return -EOPNOTSUPP;
739 }
6681712d 740 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 741 (is_multicast_ether_addr(f->eth_addr) ||
742 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 743 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
744 &rd);
6681712d
DS		 745
746 if (rc < 0)
747 return rc;
748 notify |= rc;
749 }
d342894c 750 } else {
751 if (!(flags & NLM_F_CREATE))
752 return -ENOENT;
753
754 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
755 return -ENOSPC;
756
906dc186
TR		 757 /* Disallow replace to add a multicast entry */
758 if ((flags & NLM_F_REPLACE) &&
759 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
760 return -EOPNOTSUPP;
761
e4c7ed41 762 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 763 f = kmalloc(sizeof(*f), GFP_ATOMIC);
764 if (!f)
765 return -ENOMEM;
766
767 notify = 1;
d342894c 768 f->state = state;
ae884082 769 f->flags = ndm_flags;
d342894c 770 f->updated = f->used = jiffies;
3e61aa8f 771 INIT_LIST_HEAD(&f->remotes);
d342894c 772 memcpy(f->eth_addr, mac, ETH_ALEN);
773
9e4b93f9 774 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 775
d342894c 776 ++vxlan->addrcnt;
777 hlist_add_head_rcu(&f->hlist,
778 vxlan_fdb_head(vxlan, mac));
779 }
780
9e4b93f9
ND		 781 if (notify) {
782 if (rd == NULL)
783 rd = first_remote_rtnl(f);
784 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
785 }
d342894c 786
787 return 0;
788}
789
6706c82e 790static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 791{
792 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 793 struct vxlan_rdst *rd, *nd;
6681712d 794
3e61aa8f 795 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 796 kfree(rd);
6681712d
DS		 797 kfree(f);
798}
799
d342894c 800static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
801{
802 netdev_dbg(vxlan->dev,
803 "delete %pM\n", f->eth_addr);
804
805 --vxlan->addrcnt;
9e4b93f9 806 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 807
808 hlist_del_rcu(&f->hlist);
6681712d 809 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 810}
811
f0b074be 812static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 813 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 814{
6681712d 815 struct net *net = dev_net(vxlan->dev);
e4c7ed41 816 int err;
d342894c 817
f0b074be 818 if (tb[NDA_DST]) {
e4c7ed41
CW		 819 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
820 if (err)
821 return err;
f0b074be 822 } else {
e4c7ed41
CW		 823 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
824 if (remote->sa.sa_family == AF_INET) {
825 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
826 ip->sa.sa_family = AF_INET;
827#if IS_ENABLED(CONFIG_IPV6)
828 } else {
829 ip->sin6.sin6_addr = in6addr_any;
830 ip->sa.sa_family = AF_INET6;
831#endif
832 }
f0b074be 833 }
d342894c 834
6681712d 835 if (tb[NDA_PORT]) {
73cf3317 836 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 837 return -EINVAL;
f0b074be
MR		 838 *port = nla_get_be16(tb[NDA_PORT]);
839 } else {
840 *port = vxlan->dst_port;
841 }
6681712d
DS		 842
843 if (tb[NDA_VNI]) {
844 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
845 return -EINVAL;
f0b074be
MR		 846 *vni = nla_get_u32(tb[NDA_VNI]);
847 } else {
848 *vni = vxlan->default_dst.remote_vni;
849 }
6681712d
DS		 850
851 if (tb[NDA_IFINDEX]) {
5abb0029 852 struct net_device *tdev;
6681712d
DS		 853
854 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
855 return -EINVAL;
f0b074be 856 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 857 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 858 if (!tdev)
6681712d 859 return -EADDRNOTAVAIL;
f0b074be
MR		 860 } else {
861 *ifindex = 0;
862 }
863
864 return 0;
865}
866
867/* Add static entry (via netlink) */
868static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
869 struct net_device *dev,
f6f6424b 870 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 871{
872 struct vxlan_dev *vxlan = netdev_priv(dev);
873 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 874 union vxlan_addr ip;
f0b074be
MR		 875 __be16 port;
876 u32 vni, ifindex;
877 int err;
878
879 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
880 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
881 ndm->ndm_state);
882 return -EINVAL;
883 }
884
885 if (tb[NDA_DST] == NULL)
886 return -EINVAL;
887
888 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
889 if (err)
890 return err;
6681712d 891
5933a7bb
MR		 892 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
893 return -EAFNOSUPPORT;
894
d342894c 895 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 896 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 897 port, vni, ifindex, ndm->ndm_flags);
d342894c 898 spin_unlock_bh(&vxlan->hash_lock);
899
900 return err;
901}
902
903/* Delete entry (via netlink) */
1690be63
VY		 904static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
905 struct net_device *dev,
f6f6424b 906 const unsigned char *addr, u16 vid)
d342894c 907{
908 struct vxlan_dev *vxlan = netdev_priv(dev);
909 struct vxlan_fdb *f;
bc7892ba 910 struct vxlan_rdst *rd = NULL;
e4c7ed41 911 union vxlan_addr ip;
bc7892ba
MR		 912 __be16 port;
913 u32 vni, ifindex;
914 int err;
915
916 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
917 if (err)
918 return err;
919
920 err = -ENOENT;
d342894c 921
922 spin_lock_bh(&vxlan->hash_lock);
923 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 924 if (!f)
925 goto out;
926
e4c7ed41
CW		 927 if (!vxlan_addr_any(&ip)) {
928 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 929 if (!rd)
930 goto out;
931 }
932
933 err = 0;
934
935 /* remove a destination if it's not the only one on the list,
936 * otherwise destroy the fdb entry
937 */
938 if (rd && !list_is_singular(&f->remotes)) {
939 list_del_rcu(&rd->list);
9e4b93f9 940 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 941 kfree_rcu(rd, rcu);
bc7892ba 942 goto out;
d342894c 943 }
bc7892ba
MR		 944
945 vxlan_fdb_destroy(vxlan, f);
946
947out:
d342894c 948 spin_unlock_bh(&vxlan->hash_lock);
949
950 return err;
951}
952
953/* Dump forwarding table */
954static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 955 struct net_device *dev,
956 struct net_device *filter_dev, int idx)
d342894c 957{
958 struct vxlan_dev *vxlan = netdev_priv(dev);
959 unsigned int h;
960
961 for (h = 0; h < FDB_HASH_SIZE; ++h) {
962 struct vxlan_fdb *f;
d342894c 963 int err;
964
b67bfe0d 965 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 966 struct vxlan_rdst *rd;
6681712d 967
3e61aa8f
SH		 968 if (idx < cb->args[0])
969 goto skip;
970
971 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 972 err = vxlan_fdb_info(skb, vxlan, f,
973 NETLINK_CB(cb->skb).portid,
974 cb->nlh->nlmsg_seq,
975 RTM_NEWNEIGH,
976 NLM_F_MULTI, rd);
977 if (err < 0)
3e61aa8f 978 goto out;
6681712d 979 }
3e61aa8f
SH		 980skip:
981 ++idx;
d342894c 982 }
983 }
3e61aa8f 984out:
d342894c 985 return idx;
986}
987
988/* Watch incoming packets to learn mapping between Ethernet address
989 * and Tunnel endpoint.
26a41ae6 990 * Return true if packet is bogus and should be droppped.
d342894c 991 */
26a41ae6 992static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 993 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 994{
995 struct vxlan_dev *vxlan = netdev_priv(dev);
996 struct vxlan_fdb *f;
d342894c 997
998 f = vxlan_find_mac(vxlan, src_mac);
999 if (likely(f)) {
5ca5461c 1000 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1001
e4c7ed41 1002 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 1003 return false;
1004
1005 /* Don't migrate static entries, drop packets */
eb064c3b 1006 if (f->state & NUD_NOARP)
26a41ae6 1007 return true;
d342894c 1008
1009 if (net_ratelimit())
1010 netdev_info(dev,
e4c7ed41 1011 "%pM migrated from %pIS to %pIS\n",
a4870f79 1012 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1013
e4c7ed41 1014 rdst->remote_ip = *src_ip;
d342894c 1015 f->updated = jiffies;
9e4b93f9 1016 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1017 } else {
1018 /* learned new entry */
1019 spin_lock(&vxlan->hash_lock);
3bf74b1a 1020
1021 /* close off race between vxlan_flush and incoming packets */
1022 if (netif_running(dev))
1023 vxlan_fdb_create(vxlan, src_mac, src_ip,
1024 NUD_REACHABLE,
1025 NLM_F_EXCL|NLM_F_CREATE,
1026 vxlan->dst_port,
1027 vxlan->default_dst.remote_vni,
1028 0, NTF_SELF);
d342894c 1029 spin_unlock(&vxlan->hash_lock);
1030 }
26a41ae6 1031
1032 return false;
d342894c 1033}
1034
d342894c 1035/* See if multicast group is already in use by other ID */
95ab0991 1036static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1037{
553675fb 1038 struct vxlan_dev *vxlan;
d342894c 1039
95ab0991
G		 1040 /* The vxlan_sock is only used by dev, leaving group has
1041 * no effect on other vxlan devices.
1042 */
1043 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1044 return false;
1045
553675fb 1046 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1047 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1048 continue;
d342894c 1049
95ab0991
G		 1050 if (vxlan->vn_sock != dev->vn_sock)
1051 continue;
1052
1053 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1054 &dev->default_dst.remote_ip))
1055 continue;
1056
1057 if (vxlan->default_dst.remote_ifindex !=
1058 dev->default_dst.remote_ifindex)
1059 continue;
1060
1061 return true;
553675fb 1062 }
d342894c 1063
1064 return false;
1065}
1066
7c47cedf 1067static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1068{
7c47cedf
SH		 1069 atomic_inc(&vs->refcnt);
1070}
d342894c 1071
012a5729 1072void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1073{
53cf5275
JG		 1074 struct sock *sk = vs->sock->sk;
1075 struct net *net = sock_net(sk);
1076 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1077
7c47cedf
SH		 1078 if (!atomic_dec_and_test(&vs->refcnt))
1079 return;
d342894c 1080
1c51a915 1081 spin_lock(&vn->sock_lock);
7c47cedf 1082 hlist_del_rcu(&vs->hlist);
dc01e7d3 1083 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1084 spin_unlock(&vn->sock_lock);
1085
7c47cedf 1086 queue_work(vxlan_wq, &vs->del_work);
d342894c 1087}
012a5729 1088EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1089
3fc2de2f 1090/* Callback to update multicast group membership when first VNI on
1091 * multicast asddress is brought up
1092 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1093 */
3fc2de2f 1094static void vxlan_igmp_join(struct work_struct *work)
d342894c 1095{
3fc2de2f 1096 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1097 struct vxlan_sock *vs = vxlan->vn_sock;
1098 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1099 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1100 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1101
d342894c 1102 lock_sock(sk);
e4c7ed41
CW		 1103 if (ip->sa.sa_family == AF_INET) {
1104 struct ip_mreqn mreq = {
1105 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1106 .imr_ifindex = ifindex,
1107 };
1108
1109 ip_mc_join_group(sk, &mreq);
1110#if IS_ENABLED(CONFIG_IPV6)
1111 } else {
1112 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1113 &ip->sin6.sin6_addr);
1114#endif
1115 }
3fc2de2f 1116 release_sock(sk);
1117
012a5729 1118 vxlan_sock_release(vs);
3fc2de2f 1119 dev_put(vxlan->dev);
1120}
1121
1122/* Inverse of vxlan_igmp_join when last VNI is brought down */
1123static void vxlan_igmp_leave(struct work_struct *work)
1124{
1125 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1126 struct vxlan_sock *vs = vxlan->vn_sock;
1127 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1128 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1129 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1130
1131 lock_sock(sk);
e4c7ed41
CW		 1132 if (ip->sa.sa_family == AF_INET) {
1133 struct ip_mreqn mreq = {
1134 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1135 .imr_ifindex = ifindex,
1136 };
1137
1138 ip_mc_leave_group(sk, &mreq);
1139#if IS_ENABLED(CONFIG_IPV6)
1140 } else {
1141 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1142 &ip->sin6.sin6_addr);
1143#endif
1144 }
1145
d342894c 1146 release_sock(sk);
d342894c 1147
012a5729 1148 vxlan_sock_release(vs);
7c47cedf 1149 dev_put(vxlan->dev);
d342894c 1150}
1151
dfd8645e
TH		 1152static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
1153 size_t hdrlen, u32 data)
1154{
1155 size_t start, offset, plen;
dfd8645e
TH		 1156
1157 if (skb->remcsum_offload) {
1158 /* Already processed in GRO path */
1159 skb->remcsum_offload = 0;
1160 return vh;
1161 }
1162
1163 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1164 offset = start + ((data & VXLAN_RCO_UDP) ?
1165 offsetof(struct udphdr, check) :
1166 offsetof(struct tcphdr, check));
1167
1168 plen = hdrlen + offset + sizeof(u16);
1169
1170 if (!pskb_may_pull(skb, plen))
1171 return NULL;
1172
1173 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1174
dcdc8994 1175 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset);
dfd8645e
TH		 1176
1177 return vh;
1178}
1179
d342894c 1180/* Callback from net/ipv4/udp.c to receive packets */
1181static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1182{
5cfccc5a 1183 struct vxlan_sock *vs;
d342894c 1184 struct vxlanhdr *vxh;
3bf39475 1185 u32 flags, vni;
3511494c 1186 struct vxlan_metadata md = {0};
d342894c 1187
d342894c 1188 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1189 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1190 goto error;
1191
7ce04758 1192 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1193 flags = ntohl(vxh->vx_flags);
1194 vni = ntohl(vxh->vx_vni);
1195
1196 if (flags & VXLAN_HF_VNI) {
1197 flags &= ~VXLAN_HF_VNI;
1198 } else {
1199 /* VNI flag always required to be set */
1200 goto bad_flags;
d342894c 1201 }
1202
5cfccc5a
PS		 1203 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1204 goto drop;
dfd8645e 1205 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1206
559835ea 1207 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1208 if (!vs)
d342894c 1209 goto drop;
d342894c 1210
dfd8645e
TH		 1211 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
1212 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni);
1213 if (!vxh)
1214 goto drop;
1215
1216 flags &= ~VXLAN_HF_RCO;
1217 vni &= VXLAN_VID_MASK;
1218 }
1219
3511494c
TG		 1220 /* For backwards compatibility, only allow reserved fields to be
1221 * used by VXLAN extensions if explicitly requested.
1222 */
1223 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1224 struct vxlanhdr_gbp *gbp;
1225
1226 gbp = (struct vxlanhdr_gbp *)vxh;
1227 md.gbp = ntohs(gbp->policy_id);
1228
1229 if (gbp->dont_learn)
1230 md.gbp |= VXLAN_GBP_DONT_LEARN;
1231
1232 if (gbp->policy_applied)
1233 md.gbp |= VXLAN_GBP_POLICY_APPLIED;
1234
1235 flags &= ~VXLAN_GBP_USED_BITS;
1236 }
1237
dfd8645e 1238 if (flags || (vni & ~VXLAN_VID_MASK)) {
3bf39475
TH		 1239 /* If there are any unprocessed flags remaining treat
1240 * this as a malformed packet. This behavior diverges from
1241 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1242 * in reserved fields are to be ignored. The approach here
1243 * maintains compatbility with previous stack code, and also
1244 * is more robust and provides a little more security in
1245 * adding extensions to VXLAN.
1246 */
1247
1248 goto bad_flags;
1249 }
1250
3511494c
TG		 1251 md.vni = vxh->vx_vni;
1252 vs->rcv(vs, skb, &md);
5cfccc5a
PS		 1253 return 0;
1254
1255drop:
1256 /* Consume bad packet */
1257 kfree_skb(skb);
1258 return 0;
1259
3bf39475
TH		 1260bad_flags:
1261 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1262 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1263
5cfccc5a
PS		 1264error:
1265 /* Return non vxlan pkt */
1266 return 1;
1267}
1268
3511494c
TG		 1269static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1270 struct vxlan_metadata *md)
5cfccc5a 1271{
e4c7ed41
CW		 1272 struct iphdr *oip = NULL;
1273 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1274 struct vxlan_dev *vxlan;
8f84985f 1275 struct pcpu_sw_netstats *stats;
e4c7ed41 1276 union vxlan_addr saddr;
5cfccc5a 1277 __u32 vni;
e4c7ed41
CW		 1278 int err = 0;
1279 union vxlan_addr *remote_ip;
5cfccc5a 1280
3511494c 1281 vni = ntohl(md->vni) >> 8;
5cfccc5a
PS		 1282 /* Is this VNI defined? */
1283 vxlan = vxlan_vs_find_vni(vs, vni);
1284 if (!vxlan)
d342894c 1285 goto drop;
d342894c 1286
e4c7ed41 1287 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1288 skb_reset_mac_header(skb);
f01ec1c0 1289 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1290 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1291 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1292
1293 /* Ignore packet loops (and multicast echo) */
7367d0b5 1294 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1295 goto drop;
1296
7ce04758 1297 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1298 if (remote_ip->sa.sa_family == AF_INET) {
1299 oip = ip_hdr(skb);
1300 saddr.sin.sin_addr.s_addr = oip->saddr;
1301 saddr.sa.sa_family = AF_INET;
1302#if IS_ENABLED(CONFIG_IPV6)
1303 } else {
1304 oip6 = ipv6_hdr(skb);
1305 saddr.sin6.sin6_addr = oip6->saddr;
1306 saddr.sa.sa_family = AF_INET6;
1307#endif
1308 }
1309
26a41ae6 1310 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1311 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1312 goto drop;
d342894c 1313
d342894c 1314 skb_reset_network_header(skb);
3511494c 1315 skb->mark = md->gbp;
0afb1666 1316
e4c7ed41
CW		 1317 if (oip6)
1318 err = IP6_ECN_decapsulate(oip6, skb);
1319 if (oip)
1320 err = IP_ECN_decapsulate(oip, skb);
1321
d342894c 1322 if (unlikely(err)) {
e4c7ed41
CW		 1323 if (log_ecn_error) {
1324 if (oip6)
1325 net_info_ratelimited("non-ECT from %pI6\n",
1326 &oip6->saddr);
1327 if (oip)
1328 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1329 &oip->saddr, oip->tos);
1330 }
d342894c 1331 if (err > 1) {
1332 ++vxlan->dev->stats.rx_frame_errors;
1333 ++vxlan->dev->stats.rx_errors;
1334 goto drop;
1335 }
1336 }
1337
e8171045 1338 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1339 u64_stats_update_begin(&stats->syncp);
1340 stats->rx_packets++;
1341 stats->rx_bytes += skb->len;
1342 u64_stats_update_end(&stats->syncp);
1343
1344 netif_rx(skb);
1345
5cfccc5a 1346 return;
d342894c 1347drop:
1348 /* Consume bad packet */
1349 kfree_skb(skb);
d342894c 1350}
1351
e4f67add
DS		 1352static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1353{
1354 struct vxlan_dev *vxlan = netdev_priv(dev);
1355 struct arphdr *parp;
1356 u8 *arpptr, *sha;
1357 __be32 sip, tip;
1358 struct neighbour *n;
1359
1360 if (dev->flags & IFF_NOARP)
1361 goto out;
1362
1363 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1364 dev->stats.tx_dropped++;
1365 goto out;
1366 }
1367 parp = arp_hdr(skb);
1368
1369 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1370 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1371 parp->ar_pro != htons(ETH_P_IP) ||
1372 parp->ar_op != htons(ARPOP_REQUEST) ||
1373 parp->ar_hln != dev->addr_len ||
1374 parp->ar_pln != 4)
1375 goto out;
1376 arpptr = (u8 *)parp + sizeof(struct arphdr);
1377 sha = arpptr;
1378 arpptr += dev->addr_len; /* sha */
1379 memcpy(&sip, arpptr, sizeof(sip));
1380 arpptr += sizeof(sip);
1381 arpptr += dev->addr_len; /* tha */
1382 memcpy(&tip, arpptr, sizeof(tip));
1383
1384 if (ipv4_is_loopback(tip) ||
1385 ipv4_is_multicast(tip))
1386 goto out;
1387
1388 n = neigh_lookup(&arp_tbl, &tip, dev);
1389
1390 if (n) {
e4f67add
DS		 1391 struct vxlan_fdb *f;
1392 struct sk_buff *reply;
1393
1394 if (!(n->nud_state & NUD_CONNECTED)) {
1395 neigh_release(n);
1396 goto out;
1397 }
1398
1399 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1400 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1401 /* bridge-local neighbor */
1402 neigh_release(n);
1403 goto out;
1404 }
1405
1406 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1407 n->ha, sha);
1408
1409 neigh_release(n);
1410
7346135d
DS		 1411 if (reply == NULL)
1412 goto out;
1413
e4f67add
DS		 1414 skb_reset_mac_header(reply);
1415 __skb_pull(reply, skb_network_offset(reply));
1416 reply->ip_summed = CHECKSUM_UNNECESSARY;
1417 reply->pkt_type = PACKET_HOST;
1418
1419 if (netif_rx_ni(reply) == NET_RX_DROP)
1420 dev->stats.rx_dropped++;
e4c7ed41
CW		 1421 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1422 union vxlan_addr ipa = {
1423 .sin.sin_addr.s_addr = tip,
a45e92a5 1424 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1425 };
1426
1427 vxlan_ip_miss(dev, &ipa);
1428 }
e4f67add
DS		 1429out:
1430 consume_skb(skb);
1431 return NETDEV_TX_OK;
1432}
1433
f564f45c 1434#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1435static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1436 struct neighbour *n, bool isrouter)
1437{
1438 struct net_device *dev = request->dev;
1439 struct sk_buff *reply;
1440 struct nd_msg *ns, *na;
1441 struct ipv6hdr *pip6;
1442 u8 *daddr;
1443 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1444 int ns_olen;
1445 int i, len;
1446
1447 if (dev == NULL)
1448 return NULL;
1449
1450 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1451 sizeof(*na) + na_olen + dev->needed_tailroom;
1452 reply = alloc_skb(len, GFP_ATOMIC);
1453 if (reply == NULL)
1454 return NULL;
1455
1456 reply->protocol = htons(ETH_P_IPV6);
1457 reply->dev = dev;
1458 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1459 skb_push(reply, sizeof(struct ethhdr));
1460 skb_set_mac_header(reply, 0);
1461
1462 ns = (struct nd_msg *)skb_transport_header(request);
1463
1464 daddr = eth_hdr(request)->h_source;
1465 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1466 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1467 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1468 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1469 break;
1470 }
1471 }
1472
1473 /* Ethernet header */
1474 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1475 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1476 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1477 reply->protocol = htons(ETH_P_IPV6);
1478
1479 skb_pull(reply, sizeof(struct ethhdr));
1480 skb_set_network_header(reply, 0);
1481 skb_put(reply, sizeof(struct ipv6hdr));
1482
1483 /* IPv6 header */
1484
1485 pip6 = ipv6_hdr(reply);
1486 memset(pip6, 0, sizeof(struct ipv6hdr));
1487 pip6->version = 6;
1488 pip6->priority = ipv6_hdr(request)->priority;
1489 pip6->nexthdr = IPPROTO_ICMPV6;
1490 pip6->hop_limit = 255;
1491 pip6->daddr = ipv6_hdr(request)->saddr;
1492 pip6->saddr = *(struct in6_addr *)n->primary_key;
1493
1494 skb_pull(reply, sizeof(struct ipv6hdr));
1495 skb_set_transport_header(reply, 0);
1496
1497 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1498
1499 /* Neighbor Advertisement */
1500 memset(na, 0, sizeof(*na)+na_olen);
1501 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1502 na->icmph.icmp6_router = isrouter;
1503 na->icmph.icmp6_override = 1;
1504 na->icmph.icmp6_solicited = 1;
1505 na->target = ns->target;
1506 ether_addr_copy(&na->opt[2], n->ha);
1507 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1508 na->opt[1] = na_olen >> 3;
1509
1510 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1511 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1512 csum_partial(na, sizeof(*na)+na_olen, 0));
1513
1514 pip6->payload_len = htons(sizeof(*na)+na_olen);
1515
1516 skb_push(reply, sizeof(struct ipv6hdr));
1517
1518 reply->ip_summed = CHECKSUM_UNNECESSARY;
1519
1520 return reply;
1521}
1522
f564f45c
CW		 1523static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1524{
1525 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1526 struct nd_msg *msg;
f564f45c
CW		 1527 const struct ipv6hdr *iphdr;
1528 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1529 struct neighbour *n;
1530 struct inet6_dev *in6_dev;
f564f45c
CW		 1531
1532 in6_dev = __in6_dev_get(dev);
1533 if (!in6_dev)
1534 goto out;
1535
f564f45c
CW		 1536 iphdr = ipv6_hdr(skb);
1537 saddr = &iphdr->saddr;
1538 daddr = &iphdr->daddr;
1539
f564f45c
CW		 1540 msg = (struct nd_msg *)skb_transport_header(skb);
1541 if (msg->icmph.icmp6_code != 0 ||
1542 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1543 goto out;
1544
4b29dba9
DS		 1545 if (ipv6_addr_loopback(daddr) ||
1546 ipv6_addr_is_multicast(&msg->target))
1547 goto out;
1548
1549 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1550
1551 if (n) {
1552 struct vxlan_fdb *f;
4b29dba9 1553 struct sk_buff *reply;
f564f45c
CW		 1554
1555 if (!(n->nud_state & NUD_CONNECTED)) {
1556 neigh_release(n);
1557 goto out;
1558 }
1559
1560 f = vxlan_find_mac(vxlan, n->ha);
1561 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1562 /* bridge-local neighbor */
1563 neigh_release(n);
1564 goto out;
1565 }
1566
4b29dba9
DS		 1567 reply = vxlan_na_create(skb, n,
1568 !!(f ? f->flags & NTF_ROUTER : 0));
1569
f564f45c 1570 neigh_release(n);
4b29dba9
DS		 1571
1572 if (reply == NULL)
1573 goto out;
1574
1575 if (netif_rx_ni(reply) == NET_RX_DROP)
1576 dev->stats.rx_dropped++;
1577
f564f45c 1578 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1579 union vxlan_addr ipa = {
1580 .sin6.sin6_addr = msg->target,
a45e92a5 1581 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1582 };
1583
f564f45c
CW		 1584 vxlan_ip_miss(dev, &ipa);
1585 }
1586
1587out:
1588 consume_skb(skb);
1589 return NETDEV_TX_OK;
1590}
1591#endif
1592
e4f67add
DS		 1593static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1594{
1595 struct vxlan_dev *vxlan = netdev_priv(dev);
1596 struct neighbour *n;
e4f67add
DS		 1597
1598 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1599 return false;
1600
1601 n = NULL;
1602 switch (ntohs(eth_hdr(skb)->h_proto)) {
1603 case ETH_P_IP:
e15a00aa
CW		 1604 {
1605 struct iphdr *pip;
1606
e4f67add
DS		 1607 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1608 return false;
1609 pip = ip_hdr(skb);
1610 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1611 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1612 union vxlan_addr ipa = {
1613 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1614 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1615 };
1616
1617 vxlan_ip_miss(dev, &ipa);
1618 return false;
1619 }
1620
e4f67add 1621 break;
e15a00aa
CW		 1622 }
1623#if IS_ENABLED(CONFIG_IPV6)
1624 case ETH_P_IPV6:
1625 {
1626 struct ipv6hdr *pip6;
1627
1628 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1629 return false;
1630 pip6 = ipv6_hdr(skb);
1631 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1632 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1633 union vxlan_addr ipa = {
1634 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1635 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1636 };
1637
1638 vxlan_ip_miss(dev, &ipa);
1639 return false;
1640 }
1641
1642 break;
1643 }
1644#endif
e4f67add
DS		 1645 default:
1646 return false;
1647 }
1648
1649 if (n) {
1650 bool diff;
1651
7367d0b5 1652 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1653 if (diff) {
1654 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1655 dev->addr_len);
1656 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1657 }
1658 neigh_release(n);
1659 return diff;
e4c7ed41
CW		 1660 }
1661
e4f67add
DS		 1662 return false;
1663}
1664
af33c1ad 1665static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1666 struct vxlan_metadata *md)
1667{
1668 struct vxlanhdr_gbp *gbp;
1669
db79a621
TG		 1670 if (!md->gbp)
1671 return;
1672
3511494c
TG		 1673 gbp = (struct vxlanhdr_gbp *)vxh;
1674 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1675
1676 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1677 gbp->dont_learn = 1;
1678
1679 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1680 gbp->policy_applied = 1;
1681
1682 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1683}
1684
e4c7ed41 1685#if IS_ENABLED(CONFIG_IPV6)
af33c1ad 1686static int vxlan6_xmit_skb(struct dst_entry *dst, struct sk_buff *skb,
e4c7ed41
CW		 1687 struct net_device *dev, struct in6_addr *saddr,
1688 struct in6_addr *daddr, __u8 prio, __u8 ttl,
3511494c 1689 __be16 src_port, __be16 dst_port,
af33c1ad 1690 struct vxlan_metadata *md, bool xnet, u32 vxflags)
e4c7ed41 1691{
e4c7ed41 1692 struct vxlanhdr *vxh;
e4c7ed41
CW		 1693 int min_headroom;
1694 int err;
af33c1ad 1695 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
dfd8645e
TH		 1696 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1697 u16 hdrlen = sizeof(struct vxlanhdr);
1698
af33c1ad 1699 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1700 skb->ip_summed == CHECKSUM_PARTIAL) {
1701 int csum_start = skb_checksum_start_offset(skb);
1702
1703 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1704 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1705 (skb->csum_offset == offsetof(struct udphdr, check) ||
1706 skb->csum_offset == offsetof(struct tcphdr, check))) {
1707 udp_sum = false;
1708 type |= SKB_GSO_TUNNEL_REMCSUM;
1709 }
1710 }
e4c7ed41 1711
dfd8645e 1712 skb = iptunnel_handle_offloads(skb, udp_sum, type);
74f47278
PS		 1713 if (IS_ERR(skb)) {
1714 err = -EINVAL;
1715 goto err;
1716 }
e4c7ed41 1717
f01ec1c0 1718 skb_scrub_packet(skb, xnet);
963a88b3 1719
e4c7ed41
CW		 1720 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1721 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1722 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1723
1724 /* Need space for new headers (invalidates iph ptr) */
1725 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1726 if (unlikely(err)) {
1727 kfree_skb(skb);
1728 goto err;
1729 }
e4c7ed41 1730
5968250c 1731 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1732 if (WARN_ON(!skb)) {
1733 err = -ENOMEM;
1734 goto err;
1735 }
e4c7ed41
CW		 1736
1737 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1738 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1739 vxh->vx_vni = md->vni;
e4c7ed41 1740
dfd8645e
TH		 1741 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1742 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1743 VXLAN_RCO_SHIFT;
1744
1745 if (skb->csum_offset == offsetof(struct udphdr, check))
1746 data |= VXLAN_RCO_UDP;
1747
1748 vxh->vx_vni |= htonl(data);
1749 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1750
1751 if (!skb_is_gso(skb)) {
1752 skb->ip_summed = CHECKSUM_NONE;
1753 skb->encapsulation = 0;
1754 }
1755 }
1756
af33c1ad
TH		 1757 if (vxflags & VXLAN_F_GBP)
1758 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1759
996c9fd1
TH		 1760 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1761
d998f8ef
TH		 1762 udp_tunnel6_xmit_skb(dst, skb, dev, saddr, daddr, prio,
1763 ttl, src_port, dst_port,
af33c1ad 1764 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
e4c7ed41 1765 return 0;
74f47278
PS		 1766err:
1767 dst_release(dst);
1768 return err;
e4c7ed41
CW		 1769}
1770#endif
1771
af33c1ad 1772int vxlan_xmit_skb(struct rtable *rt, struct sk_buff *skb,
49560532 1773 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
3511494c 1774 __be16 src_port, __be16 dst_port,
af33c1ad 1775 struct vxlan_metadata *md, bool xnet, u32 vxflags)
49560532
PS		 1776{
1777 struct vxlanhdr *vxh;
649c5b8b 1778 int min_headroom;
49560532 1779 int err;
af33c1ad 1780 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
dfd8645e
TH		 1781 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1782 u16 hdrlen = sizeof(struct vxlanhdr);
1783
af33c1ad 1784 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1785 skb->ip_summed == CHECKSUM_PARTIAL) {
1786 int csum_start = skb_checksum_start_offset(skb);
1787
1788 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1789 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1790 (skb->csum_offset == offsetof(struct udphdr, check) ||
1791 skb->csum_offset == offsetof(struct tcphdr, check))) {
1792 udp_sum = false;
1793 type |= SKB_GSO_TUNNEL_REMCSUM;
1794 }
1795 }
49560532 1796
dfd8645e 1797 skb = iptunnel_handle_offloads(skb, udp_sum, type);
359a0ea9 1798 if (IS_ERR(skb))
74f47278 1799 return PTR_ERR(skb);
49560532 1800
649c5b8b 1801 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1802 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1803 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1804
1805 /* Need space for new headers (invalidates iph ptr) */
1806 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1807 if (unlikely(err)) {
1808 kfree_skb(skb);
649c5b8b 1809 return err;
74f47278 1810 }
649c5b8b 1811
5968250c
JP		 1812 skb = vlan_hwaccel_push_inside(skb);
1813 if (WARN_ON(!skb))
1814 return -ENOMEM;
1eaa8178 1815
49560532 1816 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1817 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1818 vxh->vx_vni = md->vni;
49560532 1819
dfd8645e
TH		 1820 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1821 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1822 VXLAN_RCO_SHIFT;
1823
1824 if (skb->csum_offset == offsetof(struct udphdr, check))
1825 data |= VXLAN_RCO_UDP;
1826
1827 vxh->vx_vni |= htonl(data);
1828 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1829
1830 if (!skb_is_gso(skb)) {
1831 skb->ip_summed = CHECKSUM_NONE;
1832 skb->encapsulation = 0;
1833 }
1834 }
1835
af33c1ad
TH		 1836 if (vxflags & VXLAN_F_GBP)
1837 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1838
996c9fd1
TH		 1839 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1840
d998f8ef
TH		 1841 return udp_tunnel_xmit_skb(rt, skb, src, dst, tos,
1842 ttl, df, src_port, dst_port, xnet,
af33c1ad 1843 !(vxflags & VXLAN_F_UDP_CSUM));
49560532
PS		 1844}
1845EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1846
9dcc71e1
SS		 1847/* Bypass encapsulation if the destination is local */
1848static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1849 struct vxlan_dev *dst_vxlan)
1850{
8f84985f 1851 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1852 union vxlan_addr loopback;
1853 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1854 struct net_device *dev = skb->dev;
1855 int len = skb->len;
9dcc71e1 1856
8f84985f
LR		 1857 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1858 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1859 skb->pkt_type = PACKET_HOST;
1860 skb->encapsulation = 0;
1861 skb->dev = dst_vxlan->dev;
1862 __skb_pull(skb, skb_network_offset(skb));
1863
e4c7ed41
CW		 1864 if (remote_ip->sa.sa_family == AF_INET) {
1865 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1866 loopback.sa.sa_family = AF_INET;
1867#if IS_ENABLED(CONFIG_IPV6)
1868 } else {
1869 loopback.sin6.sin6_addr = in6addr_loopback;
1870 loopback.sa.sa_family = AF_INET6;
1871#endif
1872 }
1873
9dcc71e1 1874 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1875 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1876
1877 u64_stats_update_begin(&tx_stats->syncp);
1878 tx_stats->tx_packets++;
ce6502a8 1879 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1880 u64_stats_update_end(&tx_stats->syncp);
1881
1882 if (netif_rx(skb) == NET_RX_SUCCESS) {
1883 u64_stats_update_begin(&rx_stats->syncp);
1884 rx_stats->rx_packets++;
ce6502a8 1885 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1886 u64_stats_update_end(&rx_stats->syncp);
1887 } else {
ce6502a8 1888 dev->stats.rx_dropped++;
9dcc71e1
SS		 1889 }
1890}
1891
4ad16930
SH		 1892static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1893 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1894{
1895 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1896 struct rtable *rt = NULL;
d342894c 1897 const struct iphdr *old_iph;
d342894c 1898 struct flowi4 fl4;
e4c7ed41 1899 union vxlan_addr *dst;
3511494c 1900 struct vxlan_metadata md;
e4c7ed41 1901 __be16 src_port = 0, dst_port;
234f5b73 1902 u32 vni;
d342894c 1903 __be16 df = 0;
1904 __u8 tos, ttl;
0e6fbc5b 1905 int err;
d342894c 1906
823aa873 1907 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1908 vni = rdst->remote_vni;
e4c7ed41 1909 dst = &rdst->remote_ip;
e4f67add 1910
e4c7ed41 1911 if (vxlan_addr_any(dst)) {
e4f67add 1912 if (did_rsc) {
e4f67add 1913 /* short-circuited back to local bridge */
9dcc71e1 1914 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1915 return;
e4f67add 1916 }
ef59febe 1917 goto drop;
e4f67add 1918 }
ef59febe 1919
d342894c 1920 old_iph = ip_hdr(skb);
1921
d342894c 1922 ttl = vxlan->ttl;
e4c7ed41 1923 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1924 ttl = 1;
1925
1926 tos = vxlan->tos;
1927 if (tos == 1)
206aaafc 1928 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1929
535fb8d0
TH		 1930 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1931 vxlan->port_max, true);
d342894c 1932
e4c7ed41
CW		 1933 if (dst->sa.sa_family == AF_INET) {
1934 memset(&fl4, 0, sizeof(fl4));
1935 fl4.flowi4_oif = rdst->remote_ifindex;
1936 fl4.flowi4_tos = RT_TOS(tos);
1937 fl4.daddr = dst->sin.sin_addr.s_addr;
1938 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1939
f01ec1c0 1940 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1941 if (IS_ERR(rt)) {
1942 netdev_dbg(dev, "no route to %pI4\n",
1943 &dst->sin.sin_addr.s_addr);
1944 dev->stats.tx_carrier_errors++;
1945 goto tx_error;
1946 }
d342894c 1947
e4c7ed41
CW		 1948 if (rt->dst.dev == dev) {
1949 netdev_dbg(dev, "circular route to %pI4\n",
1950 &dst->sin.sin_addr.s_addr);
1951 dev->stats.collisions++;
fffc15a5 1952 goto rt_tx_error;
e4c7ed41
CW		 1953 }
1954
1955 /* Bypass encapsulation if the destination is local */
1956 if (rt->rt_flags & RTCF_LOCAL &&
1957 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1958 struct vxlan_dev *dst_vxlan;
1959
1960 ip_rt_put(rt);
19ca9fc1 1961 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1962 dst->sa.sa_family, dst_port,
1963 vxlan->flags);
e4c7ed41
CW		 1964 if (!dst_vxlan)
1965 goto tx_error;
1966 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1967 return;
1968 }
1969
1970 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1971 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
3511494c
TG		 1972 md.vni = htonl(vni << 8);
1973 md.gbp = skb->mark;
d342894c 1974
af33c1ad
TH		 1975 err = vxlan_xmit_skb(rt, skb, fl4.saddr,
1976 dst->sin.sin_addr.s_addr, tos, ttl, df,
1977 src_port, dst_port, &md,
1978 !net_eq(vxlan->net, dev_net(vxlan->dev)),
1979 vxlan->flags);
74f47278
PS		 1980 if (err < 0) {
1981 /* skb is already freed. */
1982 skb = NULL;
e4c7ed41 1983 goto rt_tx_error;
74f47278
PS		 1984 }
1985
e4c7ed41
CW		 1986 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1987#if IS_ENABLED(CONFIG_IPV6)
1988 } else {
1989 struct sock *sk = vxlan->vn_sock->sock->sk;
1990 struct dst_entry *ndst;
1991 struct flowi6 fl6;
1992 u32 flags;
1993
1994 memset(&fl6, 0, sizeof(fl6));
1995 fl6.flowi6_oif = rdst->remote_ifindex;
1996 fl6.daddr = dst->sin6.sin6_addr;
1997 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1998 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1999
2000 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
2001 netdev_dbg(dev, "no route to %pI6\n",
2002 &dst->sin6.sin6_addr);
2003 dev->stats.tx_carrier_errors++;
9dcc71e1 2004 goto tx_error;
e4c7ed41 2005 }
d342894c 2006
e4c7ed41
CW		 2007 if (ndst->dev == dev) {
2008 netdev_dbg(dev, "circular route to %pI6\n",
2009 &dst->sin6.sin6_addr);
2010 dst_release(ndst);
2011 dev->stats.collisions++;
2012 goto tx_error;
2013 }
0e6fbc5b 2014
e4c7ed41
CW		 2015 /* Bypass encapsulation if the destination is local */
2016 flags = ((struct rt6_info *)ndst)->rt6i_flags;
2017 if (flags & RTF_LOCAL &&
2018 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2019 struct vxlan_dev *dst_vxlan;
2020
2021 dst_release(ndst);
19ca9fc1 2022 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2023 dst->sa.sa_family, dst_port,
2024 vxlan->flags);
e4c7ed41
CW		 2025 if (!dst_vxlan)
2026 goto tx_error;
2027 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2028 return;
2029 }
49560532 2030
e4c7ed41 2031 ttl = ttl ? : ip6_dst_hoplimit(ndst);
3511494c
TG		 2032 md.vni = htonl(vni << 8);
2033 md.gbp = skb->mark;
e4c7ed41 2034
af33c1ad
TH		 2035 err = vxlan6_xmit_skb(ndst, skb, dev, &fl6.saddr, &fl6.daddr,
2036 0, ttl, src_port, dst_port, &md,
2037 !net_eq(vxlan->net, dev_net(vxlan->dev)),
2038 vxlan->flags);
e4c7ed41
CW		 2039#endif
2040 }
0e6fbc5b 2041
4ad16930 2042 return;
d342894c 2043
2044drop:
2045 dev->stats.tx_dropped++;
2046 goto tx_free;
2047
49560532
PS		 2048rt_tx_error:
2049 ip_rt_put(rt);
d342894c 2050tx_error:
2051 dev->stats.tx_errors++;
2052tx_free:
2053 dev_kfree_skb(skb);
d342894c 2054}
2055
6681712d
DS		 2056/* Transmit local packets over Vxlan
2057 *
2058 * Outer IP header inherits ECN and DF from inner header.
2059 * Outer UDP destination is the VXLAN assigned port.
2060 * source port is based on hash of flow
2061 */
2062static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2063{
2064 struct vxlan_dev *vxlan = netdev_priv(dev);
2065 struct ethhdr *eth;
2066 bool did_rsc = false;
8f646c92 2067 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2068 struct vxlan_fdb *f;
6681712d
DS		 2069
2070 skb_reset_mac_header(skb);
2071 eth = eth_hdr(skb);
2072
f564f45c
CW		 2073 if ((vxlan->flags & VXLAN_F_PROXY)) {
2074 if (ntohs(eth->h_proto) == ETH_P_ARP)
2075 return arp_reduce(dev, skb);
2076#if IS_ENABLED(CONFIG_IPV6)
2077 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2078 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2079 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2080 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2081 struct nd_msg *msg;
2082
2083 msg = (struct nd_msg *)skb_transport_header(skb);
2084 if (msg->icmph.icmp6_code == 0 &&
2085 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2086 return neigh_reduce(dev, skb);
2087 }
7a9f526f 2088 eth = eth_hdr(skb);
f564f45c
CW		 2089#endif
2090 }
6681712d
DS		 2091
2092 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2093 did_rsc = false;
2094
2095 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2096 (ntohs(eth->h_proto) == ETH_P_IP ||
2097 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2098 did_rsc = route_shortcircuit(dev, skb);
2099 if (did_rsc)
2100 f = vxlan_find_mac(vxlan, eth->h_dest);
2101 }
2102
6681712d 2103 if (f == NULL) {
afbd8bae
MR		 2104 f = vxlan_find_mac(vxlan, all_zeros_mac);
2105 if (f == NULL) {
2106 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2107 !is_multicast_ether_addr(eth->h_dest))
2108 vxlan_fdb_miss(vxlan, eth->h_dest);
2109
2110 dev->stats.tx_dropped++;
8f646c92 2111 kfree_skb(skb);
afbd8bae
MR		 2112 return NETDEV_TX_OK;
2113 }
2114 }
6681712d 2115
afbd8bae
MR		 2116 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2117 struct sk_buff *skb1;
6681712d 2118
8f646c92
ED		 2119 if (!fdst) {
2120 fdst = rdst;
2121 continue;
2122 }
afbd8bae
MR		 2123 skb1 = skb_clone(skb, GFP_ATOMIC);
2124 if (skb1)
2125 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2126 }
2127
8f646c92
ED		 2128 if (fdst)
2129 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2130 else
2131 kfree_skb(skb);
4ad16930 2132 return NETDEV_TX_OK;
6681712d
DS		 2133}
2134
d342894c 2135/* Walk the forwarding table and purge stale entries */
2136static void vxlan_cleanup(unsigned long arg)
2137{
2138 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2139 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2140 unsigned int h;
2141
2142 if (!netif_running(vxlan->dev))
2143 return;
2144
2145 spin_lock_bh(&vxlan->hash_lock);
2146 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2147 struct hlist_node *p, *n;
2148 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2149 struct vxlan_fdb *f
2150 = container_of(p, struct vxlan_fdb, hlist);
2151 unsigned long timeout;
2152
3c172868 2153 if (f->state & NUD_PERMANENT)
d342894c 2154 continue;
2155
2156 timeout = f->used + vxlan->age_interval * HZ;
2157 if (time_before_eq(timeout, jiffies)) {
2158 netdev_dbg(vxlan->dev,
2159 "garbage collect %pM\n",
2160 f->eth_addr);
2161 f->state = NUD_STALE;
2162 vxlan_fdb_destroy(vxlan, f);
2163 } else if (time_before(timeout, next_timer))
2164 next_timer = timeout;
2165 }
2166 }
2167 spin_unlock_bh(&vxlan->hash_lock);
2168
2169 mod_timer(&vxlan->age_timer, next_timer);
2170}
2171
9c2e24e1
PS		 2172static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2173{
2174 __u32 vni = vxlan->default_dst.remote_vni;
2175
2176 vxlan->vn_sock = vs;
2177 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2178}
2179
d342894c 2180/* Setup stats when device is created */
2181static int vxlan_init(struct net_device *dev)
2182{
1c51a915 2183 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2184 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2185 struct vxlan_sock *vs;
19ca9fc1 2186 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
1c51a915 2187
1c213bd2 2188 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2189 if (!dev->tstats)
d342894c 2190 return -ENOMEM;
2191
1c51a915 2192 spin_lock(&vn->sock_lock);
19ca9fc1 2193 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
ac5132d1 2194 vxlan->dst_port, vxlan->flags);
00c83b01 2195 if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
1c51a915 2196 /* If we have a socket with same port already, reuse it */
9c2e24e1 2197 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 2198 } else {
2199 /* otherwise make new socket outside of RTNL */
2200 dev_hold(dev);
2201 queue_work(vxlan_wq, &vxlan->sock_work);
2202 }
2203 spin_unlock(&vn->sock_lock);
2204
d342894c 2205 return 0;
2206}
2207
ba609e9b 2208static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2209{
2210 struct vxlan_fdb *f;
2211
2212 spin_lock_bh(&vxlan->hash_lock);
2213 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2214 if (f)
2215 vxlan_fdb_destroy(vxlan, f);
2216 spin_unlock_bh(&vxlan->hash_lock);
2217}
2218
ebf4063e
SH		 2219static void vxlan_uninit(struct net_device *dev)
2220{
2221 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2222 struct vxlan_sock *vs = vxlan->vn_sock;
2223
ba609e9b 2224 vxlan_fdb_delete_default(vxlan);
afbd8bae 2225
ebf4063e 2226 if (vs)
012a5729 2227 vxlan_sock_release(vs);
ebf4063e
SH		 2228 free_percpu(dev->tstats);
2229}
2230
d342894c 2231/* Start ageing timer and join group when device is brought up */
2232static int vxlan_open(struct net_device *dev)
2233{
2234 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2235 struct vxlan_sock *vs = vxlan->vn_sock;
2236
2237 /* socket hasn't been created */
2238 if (!vs)
2239 return -ENOTCONN;
d342894c 2240
79d4a94f 2241 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2242 vxlan_sock_hold(vs);
7c47cedf 2243 dev_hold(dev);
3fc2de2f 2244 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2245 }
2246
2247 if (vxlan->age_interval)
2248 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2249
2250 return 0;
2251}
2252
2253/* Purge the forwarding table */
2254static void vxlan_flush(struct vxlan_dev *vxlan)
2255{
31fec5aa 2256 unsigned int h;
d342894c 2257
2258 spin_lock_bh(&vxlan->hash_lock);
2259 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2260 struct hlist_node *p, *n;
2261 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2262 struct vxlan_fdb *f
2263 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2264 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2265 if (!is_zero_ether_addr(f->eth_addr))
2266 vxlan_fdb_destroy(vxlan, f);
d342894c 2267 }
2268 }
2269 spin_unlock_bh(&vxlan->hash_lock);
2270}
2271
2272/* Cleanup timer and forwarding table on shutdown */
2273static int vxlan_stop(struct net_device *dev)
2274{
2275 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2276 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2277 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2278
e4c7ed41 2279 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2280 !vxlan_group_used(vn, vxlan)) {
1c51a915 2281 vxlan_sock_hold(vs);
7c47cedf 2282 dev_hold(dev);
3fc2de2f 2283 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2284 }
d342894c 2285
2286 del_timer_sync(&vxlan->age_timer);
2287
2288 vxlan_flush(vxlan);
2289
2290 return 0;
2291}
2292
d342894c 2293/* Stub, nothing needs to be done. */
2294static void vxlan_set_multicast_list(struct net_device *dev)
2295{
2296}
2297
345010b5
DB		 2298static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2299{
2300 struct vxlan_dev *vxlan = netdev_priv(dev);
2301 struct vxlan_rdst *dst = &vxlan->default_dst;
2302 struct net_device *lowerdev;
2303 int max_mtu;
2304
f01ec1c0 2305 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2306 if (lowerdev == NULL)
2307 return eth_change_mtu(dev, new_mtu);
2308
2309 if (dst->remote_ip.sa.sa_family == AF_INET6)
2310 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2311 else
2312 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2313
2314 if (new_mtu < 68 || new_mtu > max_mtu)
2315 return -EINVAL;
2316
2317 dev->mtu = new_mtu;
2318 return 0;
2319}
2320
d342894c 2321static const struct net_device_ops vxlan_netdev_ops = {
2322 .ndo_init = vxlan_init,
ebf4063e 2323 .ndo_uninit = vxlan_uninit,
d342894c 2324 .ndo_open = vxlan_open,
2325 .ndo_stop = vxlan_stop,
2326 .ndo_start_xmit = vxlan_xmit,
e8171045 2327 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2328 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2329 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2330 .ndo_validate_addr = eth_validate_addr,
2331 .ndo_set_mac_address = eth_mac_addr,
2332 .ndo_fdb_add = vxlan_fdb_add,
2333 .ndo_fdb_del = vxlan_fdb_delete,
2334 .ndo_fdb_dump = vxlan_fdb_dump,
2335};
2336
2337/* Info for udev, that this is a virtual tunnel endpoint */
2338static struct device_type vxlan_type = {
2339 .name = "vxlan",
2340};
2341
53cf5275 2342/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2343 * supply the listening VXLAN udp ports. Callers are expected
2344 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2345 */
2346void vxlan_get_rx_port(struct net_device *dev)
2347{
2348 struct vxlan_sock *vs;
2349 struct net *net = dev_net(dev);
2350 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2351 sa_family_t sa_family;
35e42379
JG		 2352 __be16 port;
2353 unsigned int i;
53cf5275
JG		 2354
2355 spin_lock(&vn->sock_lock);
2356 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2357 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2358 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2359 sa_family = vs->sock->sk->sk_family;
2360 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2361 port);
2362 }
2363 }
2364 spin_unlock(&vn->sock_lock);
2365}
2366EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2367
d342894c 2368/* Initialize the device structure. */
2369static void vxlan_setup(struct net_device *dev)
2370{
2371 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2372 unsigned int h;
d342894c 2373
2374 eth_hw_addr_random(dev);
2375 ether_setup(dev);
e4c7ed41 2376 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2377 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2378 else
2853af6a 2379 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2380
2381 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2382 dev->destructor = free_netdev;
d342894c 2383 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2384
2385 dev->tx_queue_len = 0;
2386 dev->features |= NETIF_F_LLTX;
d6727fe3 2387 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2388 dev->features |= NETIF_F_RXCSUM;
05c0db08 2389 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2390
1eaa8178
PS		 2391 dev->vlan_features = dev->features;
2392 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2393 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2394 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2395 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2396 netif_keep_dst(dev);
6602d007 2397 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2398
553675fb 2399 INIT_LIST_HEAD(&vxlan->next);
d342894c 2400 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2401 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2402 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2403 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2404
2405 init_timer_deferrable(&vxlan->age_timer);
2406 vxlan->age_timer.function = vxlan_cleanup;
2407 vxlan->age_timer.data = (unsigned long) vxlan;
2408
823aa873 2409 vxlan->dst_port = htons(vxlan_port);
05f47d69 2410
d342894c 2411 vxlan->dev = dev;
2412
2413 for (h = 0; h < FDB_HASH_SIZE; ++h)
2414 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2415}
2416
2417static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2418 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2419 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2420 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2421 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2422 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2423 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2424 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2425 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2426 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2427 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2428 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2429 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2430 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2431 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2432 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2433 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2434 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2435 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2436 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2437 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2438 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2439 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2440 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
d342894c 2441};
2442
2443static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2444{
2445 if (tb[IFLA_ADDRESS]) {
2446 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2447 pr_debug("invalid link address (not ethernet)\n");
2448 return -EINVAL;
2449 }
2450
2451 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2452 pr_debug("invalid all zero ethernet address\n");
2453 return -EADDRNOTAVAIL;
2454 }
2455 }
2456
2457 if (!data)
2458 return -EINVAL;
2459
2460 if (data[IFLA_VXLAN_ID]) {
2461 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2462 if (id >= VXLAN_VID_MASK)
2463 return -ERANGE;
2464 }
2465
05f47d69 2466 if (data[IFLA_VXLAN_PORT_RANGE]) {
2467 const struct ifla_vxlan_port_range *p
2468 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2469
2470 if (ntohs(p->high) < ntohs(p->low)) {
2471 pr_debug("port range %u .. %u not valid\n",
2472 ntohs(p->low), ntohs(p->high));
2473 return -EINVAL;
2474 }
2475 }
2476
d342894c 2477 return 0;
2478}
2479
1b13c97f
YB		 2480static void vxlan_get_drvinfo(struct net_device *netdev,
2481 struct ethtool_drvinfo *drvinfo)
2482{
2483 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2484 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2485}
2486
2487static const struct ethtool_ops vxlan_ethtool_ops = {
2488 .get_drvinfo = vxlan_get_drvinfo,
2489 .get_link = ethtool_op_get_link,
2490};
2491
553675fb 2492static void vxlan_del_work(struct work_struct *work)
2493{
2494 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2495 udp_tunnel_sock_release(vs->sock);
553675fb 2496 kfree_rcu(vs, rcu);
2497}
2498
3ee64f39
TH		 2499static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2500 __be16 port, u32 flags)
553675fb 2501{
e4c7ed41 2502 struct socket *sock;
3ee64f39
TH		 2503 struct udp_port_cfg udp_conf;
2504 int err;
e4c7ed41 2505
3ee64f39 2506 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2507
3ee64f39
TH		 2508 if (ipv6) {
2509 udp_conf.family = AF_INET6;
3ee64f39 2510 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2511 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2512 } else {
2513 udp_conf.family = AF_INET;
2514 udp_conf.local_ip.s_addr = INADDR_ANY;
e4c7ed41 2515 }
e4c7ed41 2516
3ee64f39 2517 udp_conf.local_udp_port = port;
553675fb 2518
3ee64f39
TH		 2519 /* Open UDP socket */
2520 err = udp_sock_create(net, &udp_conf, &sock);
2521 if (err < 0)
2522 return ERR_PTR(err);
e4c7ed41 2523
39deb2c7 2524 return sock;
e4c7ed41
CW		 2525}
2526
2527/* Create new listen socket if needed */
2528static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2529 vxlan_rcv_t *rcv, void *data,
2530 u32 flags)
e4c7ed41
CW		 2531{
2532 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2533 struct vxlan_sock *vs;
2534 struct socket *sock;
e4c7ed41 2535 unsigned int h;
359a0ea9 2536 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2537 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2538
dc01e7d3 2539 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2540 if (!vs)
2541 return ERR_PTR(-ENOMEM);
2542
2543 for (h = 0; h < VNI_HASH_SIZE; ++h)
2544 INIT_HLIST_HEAD(&vs->vni_list[h]);
2545
2546 INIT_WORK(&vs->del_work, vxlan_del_work);
2547
3ee64f39 2548 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2549 if (IS_ERR(sock)) {
553675fb 2550 kfree(vs);
e50fddc8 2551 return ERR_CAST(sock);
553675fb 2552 }
e4c7ed41
CW		 2553
2554 vs->sock = sock;
9c2e24e1 2555 atomic_set(&vs->refcnt, 1);
5cfccc5a 2556 vs->rcv = rcv;
012a5729 2557 vs->data = data;
af33c1ad 2558 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2559
dc01e7d3
OG		 2560 /* Initialize the vxlan udp offloads structure */
2561 vs->udp_offloads.port = port;
2562 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2563 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2564
9c2e24e1
PS		 2565 spin_lock(&vn->sock_lock);
2566 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2567 vxlan_notify_add_rx_port(vs);
9c2e24e1 2568 spin_unlock(&vn->sock_lock);
553675fb 2569
2570 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2571 tunnel_cfg.sk_user_data = vs;
2572 tunnel_cfg.encap_type = 1;
2573 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2574 tunnel_cfg.encap_destroy = NULL;
2575
2576 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2577
9c2e24e1
PS		 2578 return vs;
2579}
2580
012a5729
PS		 2581struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2582 vxlan_rcv_t *rcv, void *data,
359a0ea9 2583 bool no_share, u32 flags)
9c2e24e1
PS		 2584{
2585 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2586 struct vxlan_sock *vs;
19ca9fc1 2587 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2588
359a0ea9 2589 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2590 if (!IS_ERR(vs))
2591 return vs;
553675fb 2592
012a5729
PS		 2593 if (no_share) /* Return error if sharing is not allowed. */
2594 return vs;
2595
9c2e24e1 2596 spin_lock(&vn->sock_lock);
ac5132d1 2597 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port, flags);
00c83b01
ML		 2598 if (vs && ((vs->rcv != rcv) ||
2599 !atomic_add_unless(&vs->refcnt, 1, 0)))
5cfccc5a 2600 vs = ERR_PTR(-EBUSY);
5cfccc5a
PS		 2601 spin_unlock(&vn->sock_lock);
2602
2603 if (!vs)
9c2e24e1
PS		 2604 vs = ERR_PTR(-EINVAL);
2605
553675fb 2606 return vs;
2607}
012a5729 2608EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2609
1c51a915
SH		 2610/* Scheduled at device creation to bind to a socket */
2611static void vxlan_sock_work(struct work_struct *work)
2612{
9c2e24e1 2613 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2614 struct net *net = vxlan->net;
1c51a915 2615 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2616 __be16 port = vxlan->dst_port;
2617 struct vxlan_sock *nvs;
1c51a915 2618
359a0ea9 2619 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2620 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2621 if (!IS_ERR(nvs))
2622 vxlan_vs_add_dev(nvs, vxlan);
2623 spin_unlock(&vn->sock_lock);
2624
2625 dev_put(vxlan->dev);
1c51a915
SH		 2626}
2627
33564bbb 2628static int vxlan_newlink(struct net *src_net, struct net_device *dev,
d342894c 2629 struct nlattr *tb[], struct nlattr *data[])
2630{
33564bbb 2631 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
d342894c 2632 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2633 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2634 __u32 vni;
2635 int err;
e4c7ed41 2636 bool use_ipv6 = false;
d342894c 2637
2638 if (!data[IFLA_VXLAN_ID])
2639 return -EINVAL;
2640
33564bbb 2641 vxlan->net = src_net;
f01ec1c0 2642
d342894c 2643 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2644 dst->remote_vni = vni;
d342894c 2645
5933a7bb
MR		 2646 /* Unless IPv6 is explicitly requested, assume IPv4 */
2647 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2648 if (data[IFLA_VXLAN_GROUP]) {
2649 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2650 } else if (data[IFLA_VXLAN_GROUP6]) {
2651 if (!IS_ENABLED(CONFIG_IPV6))
2652 return -EPFNOSUPPORT;
2653
2654 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2655 sizeof(struct in6_addr));
2656 dst->remote_ip.sa.sa_family = AF_INET6;
2657 use_ipv6 = true;
2658 }
d342894c 2659
e4c7ed41
CW		 2660 if (data[IFLA_VXLAN_LOCAL]) {
2661 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2662 vxlan->saddr.sa.sa_family = AF_INET;
2663 } else if (data[IFLA_VXLAN_LOCAL6]) {
2664 if (!IS_ENABLED(CONFIG_IPV6))
2665 return -EPFNOSUPPORT;
2666
2667 /* TODO: respect scope id */
2668 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2669 sizeof(struct in6_addr));
2670 vxlan->saddr.sa.sa_family = AF_INET6;
2671 use_ipv6 = true;
2672 }
d342894c 2673
34e02aa1 2674 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2675 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2676 struct net_device *lowerdev
33564bbb 2677 = __dev_get_by_index(src_net, dst->remote_ifindex);
34e02aa1 2678
2679 if (!lowerdev) {
c7995c43 2680 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2681 return -ENODEV;
2682 }
d342894c 2683
e4c7ed41
CW		 2684#if IS_ENABLED(CONFIG_IPV6)
2685 if (use_ipv6) {
2686 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2687 if (idev && idev->cnf.disable_ipv6) {
2688 pr_info("IPv6 is disabled via sysctl\n");
2689 return -EPERM;
2690 }
2691 vxlan->flags |= VXLAN_F_IPV6;
2692 }
2693#endif
2694
34e02aa1 2695 if (!tb[IFLA_MTU])
e4c7ed41 2696 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2697
2853af6a 2698 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2699 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2700 } else if (use_ipv6)
2701 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2702
2703 if (data[IFLA_VXLAN_TOS])
2704 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2705
afb97186
VB		 2706 if (data[IFLA_VXLAN_TTL])
2707 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2708
d342894c 2709 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2710 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2711
2712 if (data[IFLA_VXLAN_AGEING])
2713 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2714 else
2715 vxlan->age_interval = FDB_AGE_DEFAULT;
2716
e4f67add
DS		 2717 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2718 vxlan->flags |= VXLAN_F_PROXY;
2719
2720 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2721 vxlan->flags |= VXLAN_F_RSC;
2722
2723 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2724 vxlan->flags |= VXLAN_F_L2MISS;
2725
2726 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2727 vxlan->flags |= VXLAN_F_L3MISS;
2728
d342894c 2729 if (data[IFLA_VXLAN_LIMIT])
2730 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2731
05f47d69 2732 if (data[IFLA_VXLAN_PORT_RANGE]) {
2733 const struct ifla_vxlan_port_range *p
2734 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2735 vxlan->port_min = ntohs(p->low);
2736 vxlan->port_max = ntohs(p->high);
2737 }
2738
823aa873 2739 if (data[IFLA_VXLAN_PORT])
2740 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2741
359a0ea9
TH		 2742 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2743 vxlan->flags |= VXLAN_F_UDP_CSUM;
2744
2745 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2746 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2747 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2748
2749 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2750 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2751 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2752
dfd8645e
TH		 2753 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2754 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2755 vxlan->flags |= VXLAN_F_REMCSUM_TX;
2756
2757 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2758 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2759 vxlan->flags |= VXLAN_F_REMCSUM_RX;
2760
3511494c
TG		 2761 if (data[IFLA_VXLAN_GBP])
2762 vxlan->flags |= VXLAN_F_GBP;
2763
33564bbb 2764 if (vxlan_find_vni(src_net, vni, use_ipv6 ? AF_INET6 : AF_INET,
ac5132d1 2765 vxlan->dst_port, vxlan->flags)) {
553675fb 2766 pr_info("duplicate VNI %u\n", vni);
2767 return -EEXIST;
2768 }
2769
7ad24ea4 2770 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2771
2936b6ab
SS		 2772 /* create an fdb entry for a valid default destination */
2773 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2774 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2775 &vxlan->default_dst.remote_ip,
2776 NUD_REACHABLE|NUD_PERMANENT,
2777 NLM_F_EXCL|NLM_F_CREATE,
2778 vxlan->dst_port,
2779 vxlan->default_dst.remote_vni,
2780 vxlan->default_dst.remote_ifindex,
2781 NTF_SELF);
2782 if (err)
2783 return err;
2784 }
d342894c 2785
afbd8bae
MR		 2786 err = register_netdevice(dev);
2787 if (err) {
ba609e9b 2788 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2789 return err;
2790 }
2791
553675fb 2792 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2793
2794 return 0;
d342894c 2795}
2796
2797static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2798{
2799 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2800 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2801
fe5c3561 2802 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2803 if (!hlist_unhashed(&vxlan->hlist))
2804 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2805 spin_unlock(&vn->sock_lock);
2806
553675fb 2807 list_del(&vxlan->next);
d342894c 2808 unregister_netdevice_queue(dev, head);
2809}
2810
2811static size_t vxlan_get_size(const struct net_device *dev)
2812{
2813
2814 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2815 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2816 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2817 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2818 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2819 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2820 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2821 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2822 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2823 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2824 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2825 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2826 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2827 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2828 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2829 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2830 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2831 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2832 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2833 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2834 0;
2835}
2836
2837static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2838{
2839 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2840 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2841 struct ifla_vxlan_port_range ports = {
2842 .low = htons(vxlan->port_min),
2843 .high = htons(vxlan->port_max),
2844 };
d342894c 2845
c7995c43 2846 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2847 goto nla_put_failure;
2848
e4c7ed41
CW		 2849 if (!vxlan_addr_any(&dst->remote_ip)) {
2850 if (dst->remote_ip.sa.sa_family == AF_INET) {
2851 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2852 dst->remote_ip.sin.sin_addr.s_addr))
2853 goto nla_put_failure;
2854#if IS_ENABLED(CONFIG_IPV6)
2855 } else {
2856 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2857 &dst->remote_ip.sin6.sin6_addr))
2858 goto nla_put_failure;
2859#endif
2860 }
2861 }
d342894c 2862
c7995c43 2863 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2864 goto nla_put_failure;
2865
e4c7ed41
CW		 2866 if (!vxlan_addr_any(&vxlan->saddr)) {
2867 if (vxlan->saddr.sa.sa_family == AF_INET) {
2868 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2869 vxlan->saddr.sin.sin_addr.s_addr))
2870 goto nla_put_failure;
2871#if IS_ENABLED(CONFIG_IPV6)
2872 } else {
2873 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2874 &vxlan->saddr.sin6.sin6_addr))
2875 goto nla_put_failure;
2876#endif
2877 }
2878 }
d342894c 2879
2880 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2881 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2882 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2883 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2884 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2885 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2886 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2887 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2888 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2889 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2890 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2891 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2892 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2893 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2894 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2895 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2896 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2897 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2898 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 2899 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2900 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2901 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2902 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2903 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 2904 goto nla_put_failure;
2905
05f47d69 2906 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2907 goto nla_put_failure;
2908
3511494c
TG		 2909 if (vxlan->flags & VXLAN_F_GBP &&
2910 nla_put_flag(skb, IFLA_VXLAN_GBP))
2911 goto nla_put_failure;
2912
d342894c 2913 return 0;
2914
2915nla_put_failure:
2916 return -EMSGSIZE;
2917}
2918
1728d4fa
ND		 2919static struct net *vxlan_get_link_net(const struct net_device *dev)
2920{
2921 struct vxlan_dev *vxlan = netdev_priv(dev);
2922
2923 return vxlan->net;
2924}
2925
d342894c 2926static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2927 .kind = "vxlan",
2928 .maxtype = IFLA_VXLAN_MAX,
2929 .policy = vxlan_policy,
2930 .priv_size = sizeof(struct vxlan_dev),
2931 .setup = vxlan_setup,
2932 .validate = vxlan_validate,
2933 .newlink = vxlan_newlink,
2934 .dellink = vxlan_dellink,
2935 .get_size = vxlan_get_size,
2936 .fill_info = vxlan_fill_info,
1728d4fa 2937 .get_link_net = vxlan_get_link_net,
d342894c 2938};
2939
acaf4e70
DB		 2940static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2941 struct net_device *dev)
2942{
2943 struct vxlan_dev *vxlan, *next;
2944 LIST_HEAD(list_kill);
2945
2946 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2947 struct vxlan_rdst *dst = &vxlan->default_dst;
2948
2949 /* In case we created vxlan device with carrier
2950 * and we loose the carrier due to module unload
2951 * we also need to remove vxlan device. In other
2952 * cases, it's not necessary and remote_ifindex
2953 * is 0 here, so no matches.
2954 */
2955 if (dst->remote_ifindex == dev->ifindex)
2956 vxlan_dellink(vxlan->dev, &list_kill);
2957 }
2958
2959 unregister_netdevice_many(&list_kill);
2960}
2961
2962static int vxlan_lowerdev_event(struct notifier_block *unused,
2963 unsigned long event, void *ptr)
2964{
2965 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2966 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2967
783c1463 2968 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2969 vxlan_handle_lowerdev_unregister(vn, dev);
2970
2971 return NOTIFY_DONE;
2972}
2973
2974static struct notifier_block vxlan_notifier_block __read_mostly = {
2975 .notifier_call = vxlan_lowerdev_event,
2976};
2977
d342894c 2978static __net_init int vxlan_init_net(struct net *net)
2979{
2980 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2981 unsigned int h;
d342894c 2982
553675fb 2983 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2984 spin_lock_init(&vn->sock_lock);
d342894c 2985
553675fb 2986 for (h = 0; h < PORT_HASH_SIZE; ++h)
2987 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2988
2989 return 0;
2990}
2991
f01ec1c0
ND		 2992static void __net_exit vxlan_exit_net(struct net *net)
2993{
2994 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2995 struct vxlan_dev *vxlan, *next;
2996 struct net_device *dev, *aux;
2997 LIST_HEAD(list);
2998
2999 rtnl_lock();
3000 for_each_netdev_safe(net, dev, aux)
3001 if (dev->rtnl_link_ops == &vxlan_link_ops)
3002 unregister_netdevice_queue(dev, &list);
3003
3004 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3005 /* If vxlan->dev is in the same netns, it has already been added
3006 * to the list by the previous loop.
3007 */
3008 if (!net_eq(dev_net(vxlan->dev), net))
3009 unregister_netdevice_queue(dev, &list);
3010 }
3011
3012 unregister_netdevice_many(&list);
3013 rtnl_unlock();
3014}
3015
d342894c 3016static struct pernet_operations vxlan_net_ops = {
3017 .init = vxlan_init_net,
f01ec1c0 3018 .exit = vxlan_exit_net,
d342894c 3019 .id = &vxlan_net_id,
3020 .size = sizeof(struct vxlan_net),
3021};
3022
3023static int __init vxlan_init_module(void)
3024{
3025 int rc;
3026
758c57d1
SH		 3027 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3028 if (!vxlan_wq)
3029 return -ENOMEM;
3030
d342894c 3031 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3032
783c1463 3033 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3034 if (rc)
3035 goto out1;
3036
acaf4e70 3037 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3038 if (rc)
3039 goto out2;
3040
acaf4e70
DB		 3041 rc = rtnl_link_register(&vxlan_link_ops);
3042 if (rc)
3043 goto out3;
d342894c 3044
acaf4e70
DB		 3045 return 0;
3046out3:
3047 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3048out2:
783c1463 3049 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3050out1:
758c57d1 3051 destroy_workqueue(vxlan_wq);
d342894c 3052 return rc;
3053}
7332a13b 3054late_initcall(vxlan_init_module);
d342894c 3055
3056static void __exit vxlan_cleanup_module(void)
3057{
b7153984 3058 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3059 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3060 destroy_workqueue(vxlan_wq);
783c1463
DB		 3061 unregister_pernet_subsys(&vxlan_net_ops);
3062 /* rcu_barrier() is called by netns */
d342894c 3063}
3064module_exit(vxlan_cleanup_module);
3065
3066MODULE_LICENSE("GPL");
3067MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3068MODULE_AUTHOR("Stephen Hemminger <[email protected]>");
ead5139a 3069MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3070MODULE_ALIAS_RTNL_LINK("vxlan");
Empty description
This page took 0.807412 seconds and 4 git commands to generate.