projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
xen-netback: making the bandwidth limiter runtime settable
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
acbf74a7 45#include <net/udp_tunnel.h>
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
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
130 unsigned long age_interval;
131 struct timer_list age_timer;
132 spinlock_t hash_lock;
133 unsigned int addrcnt;
134 unsigned int addrmax;
d342894c 135
136 struct hlist_head fdb_head[FDB_HASH_SIZE];
137};
138
139/* salt for hash table */
140static u32 vxlan_salt __read_mostly;
758c57d1 141static struct workqueue_struct *vxlan_wq;
d342894c 142
e4c7ed41
CW		 143#if IS_ENABLED(CONFIG_IPV6)
144static inline
145bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
146{
147 if (a->sa.sa_family != b->sa.sa_family)
148 return false;
149 if (a->sa.sa_family == AF_INET6)
150 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
151 else
152 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
153}
154
155static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
156{
157 if (ipa->sa.sa_family == AF_INET6)
158 return ipv6_addr_any(&ipa->sin6.sin6_addr);
159 else
160 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
161}
162
163static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
164{
165 if (ipa->sa.sa_family == AF_INET6)
166 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
167 else
168 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
169}
170
171static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
172{
173 if (nla_len(nla) >= sizeof(struct in6_addr)) {
174 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
175 ip->sa.sa_family = AF_INET6;
176 return 0;
177 } else if (nla_len(nla) >= sizeof(__be32)) {
178 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
179 ip->sa.sa_family = AF_INET;
180 return 0;
181 } else {
182 return -EAFNOSUPPORT;
183 }
184}
185
186static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
187 const union vxlan_addr *ip)
188{
189 if (ip->sa.sa_family == AF_INET6)
190 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
191 else
192 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
193}
194
195#else /* !CONFIG_IPV6 */
196
197static inline
198bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
199{
200 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
201}
202
203static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
204{
205 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
206}
207
208static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
209{
210 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
211}
212
213static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
214{
215 if (nla_len(nla) >= sizeof(struct in6_addr)) {
216 return -EAFNOSUPPORT;
217 } else if (nla_len(nla) >= sizeof(__be32)) {
218 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
219 ip->sa.sa_family = AF_INET;
220 return 0;
221 } else {
222 return -EAFNOSUPPORT;
223 }
224}
225
226static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
227 const union vxlan_addr *ip)
228{
229 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
230}
231#endif
232
553675fb 233/* Virtual Network hash table head */
234static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
235{
236 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
237}
238
239/* Socket hash table head */
240static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 241{
242 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
243
553675fb 244 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
245}
246
3e61aa8f
SH		 247/* First remote destination for a forwarding entry.
248 * Guaranteed to be non-NULL because remotes are never deleted.
249 */
5ca5461c 250static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 251{
5ca5461c 252 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
253}
254
255static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
256{
257 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 258}
259
ac5132d1
TG		 260/* Find VXLAN socket based on network namespace, address family and UDP port
261 * and enabled unshareable flags.
262 */
263static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
264 __be16 port, u32 flags)
553675fb 265{
266 struct vxlan_sock *vs;
af33c1ad
TH		 267
268 flags &= VXLAN_F_RCV_FLAGS;
553675fb 269
270 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 271 if (inet_sk(vs->sock->sk)->inet_sport == port &&
ac5132d1 272 inet_sk(vs->sock->sk)->sk.sk_family == family &&
af33c1ad 273 vs->flags == flags)
553675fb 274 return vs;
275 }
276 return NULL;
d342894c 277}
278
5cfccc5a 279static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 280{
281 struct vxlan_dev *vxlan;
d342894c 282
553675fb 283 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 284 if (vxlan->default_dst.remote_vni == id)
d342894c 285 return vxlan;
286 }
287
288 return NULL;
289}
290
5cfccc5a 291/* Look up VNI in a per net namespace table */
19ca9fc1 292static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 293 sa_family_t family, __be16 port,
294 u32 flags)
5cfccc5a
PS		 295{
296 struct vxlan_sock *vs;
297
ac5132d1 298 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 299 if (!vs)
300 return NULL;
301
302 return vxlan_vs_find_vni(vs, id);
303}
304
d342894c 305/* Fill in neighbour message in skbuff. */
306static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 307 const struct vxlan_fdb *fdb,
308 u32 portid, u32 seq, int type, unsigned int flags,
309 const struct vxlan_rdst *rdst)
d342894c 310{
311 unsigned long now = jiffies;
312 struct nda_cacheinfo ci;
313 struct nlmsghdr *nlh;
314 struct ndmsg *ndm;
e4f67add 315 bool send_ip, send_eth;
d342894c 316
317 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
318 if (nlh == NULL)
319 return -EMSGSIZE;
320
321 ndm = nlmsg_data(nlh);
322 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 323
324 send_eth = send_ip = true;
325
326 if (type == RTM_GETNEIGH) {
327 ndm->ndm_family = AF_INET;
e4c7ed41 328 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 329 send_eth = !is_zero_ether_addr(fdb->eth_addr);
330 } else
331 ndm->ndm_family = AF_BRIDGE;
d342894c 332 ndm->ndm_state = fdb->state;
333 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 334 ndm->ndm_flags = fdb->flags;
545469f7 335 ndm->ndm_type = RTN_UNICAST;
d342894c 336
193523bf 337 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b
ND		 338 nla_put_s32(skb, NDA_LINK_NETNSID,
339 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 340 goto nla_put_failure;
341
e4f67add 342 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 343 goto nla_put_failure;
344
e4c7ed41 345 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 346 goto nla_put_failure;
347
823aa873 348 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 349 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
350 goto nla_put_failure;
c7995c43 351 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 352 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 353 goto nla_put_failure;
354 if (rdst->remote_ifindex &&
355 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 356 goto nla_put_failure;
357
358 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
359 ci.ndm_confirmed = 0;
360 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
361 ci.ndm_refcnt = 0;
362
363 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
364 goto nla_put_failure;
365
053c095a
JB		 366 nlmsg_end(skb, nlh);
367 return 0;
d342894c 368
369nla_put_failure:
370 nlmsg_cancel(skb, nlh);
371 return -EMSGSIZE;
372}
373
374static inline size_t vxlan_nlmsg_size(void)
375{
376 return NLMSG_ALIGN(sizeof(struct ndmsg))
377 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 378 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 379 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 380 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
381 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 382 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 383 + nla_total_size(sizeof(struct nda_cacheinfo));
384}
385
9e4b93f9
ND		 386static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
387 struct vxlan_rdst *rd, int type)
d342894c 388{
389 struct net *net = dev_net(vxlan->dev);
390 struct sk_buff *skb;
391 int err = -ENOBUFS;
392
393 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
394 if (skb == NULL)
395 goto errout;
396
9e4b93f9 397 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 398 if (err < 0) {
399 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
400 WARN_ON(err == -EMSGSIZE);
401 kfree_skb(skb);
402 goto errout;
403 }
404
405 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
406 return;
407errout:
408 if (err < 0)
409 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
410}
411
e4c7ed41 412static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 413{
414 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 415 struct vxlan_fdb f = {
416 .state = NUD_STALE,
417 };
418 struct vxlan_rdst remote = {
e4c7ed41 419 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 420 .remote_vni = VXLAN_N_VID,
421 };
3e61aa8f 422
9e4b93f9 423 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 424}
425
426static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
427{
bb3fd687
SH		 428 struct vxlan_fdb f = {
429 .state = NUD_STALE,
430 };
9e4b93f9 431 struct vxlan_rdst remote = { };
e4f67add 432
e4f67add
DS		 433 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
434
9e4b93f9 435 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 436}
437
d342894c 438/* Hash Ethernet address */
439static u32 eth_hash(const unsigned char *addr)
440{
441 u64 value = get_unaligned((u64 *)addr);
442
443 /* only want 6 bytes */
444#ifdef __BIG_ENDIAN
d342894c 445 value >>= 16;
321fb991 446#else
447 value <<= 16;
d342894c 448#endif
449 return hash_64(value, FDB_HASH_BITS);
450}
451
452/* Hash chain to use given mac address */
453static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
454 const u8 *mac)
455{
456 return &vxlan->fdb_head[eth_hash(mac)];
457}
458
459/* Look up Ethernet address in forwarding table */
014be2c8 460static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 461 const u8 *mac)
d342894c 462{
463 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
464 struct vxlan_fdb *f;
d342894c 465
b67bfe0d 466 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 467 if (ether_addr_equal(mac, f->eth_addr))
d342894c 468 return f;
469 }
470
471 return NULL;
472}
473
014be2c8
SS		 474static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
475 const u8 *mac)
476{
477 struct vxlan_fdb *f;
478
479 f = __vxlan_find_mac(vxlan, mac);
480 if (f)
481 f->used = jiffies;
482
483 return f;
484}
485
a5e7c10a
MR		 486/* caller should hold vxlan->hash_lock */
487static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 488 union vxlan_addr *ip, __be16 port,
a5e7c10a 489 __u32 vni, __u32 ifindex)
6681712d 490{
3e61aa8f 491 struct vxlan_rdst *rd;
6681712d 492
3e61aa8f 493 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 494 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 495 rd->remote_port == port &&
496 rd->remote_vni == vni &&
497 rd->remote_ifindex == ifindex)
a5e7c10a 498 return rd;
6681712d 499 }
3e61aa8f 500
a5e7c10a
MR		 501 return NULL;
502}
503
906dc186
TR		 504/* Replace destination of unicast mac */
505static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 506 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 507{
508 struct vxlan_rdst *rd;
509
510 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
511 if (rd)
512 return 0;
513
514 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
515 if (!rd)
516 return 0;
e4c7ed41 517 rd->remote_ip = *ip;
906dc186
TR		 518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
521 return 1;
522}
523
a5e7c10a
MR		 524/* Add/update destinations for multicast */
525static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 526 union vxlan_addr *ip, __be16 port, __u32 vni,
527 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 528{
529 struct vxlan_rdst *rd;
530
531 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
532 if (rd)
533 return 0;
534
6681712d
DS		 535 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
536 if (rd == NULL)
537 return -ENOBUFS;
e4c7ed41 538 rd->remote_ip = *ip;
6681712d
DS		 539 rd->remote_port = port;
540 rd->remote_vni = vni;
541 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 542
543 list_add_tail_rcu(&rd->list, &f->remotes);
544
9e4b93f9 545 *rdp = rd;
6681712d
DS		 546 return 1;
547}
548
dfd8645e
TH		 549static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
550 unsigned int off,
551 struct vxlanhdr *vh, size_t hdrlen,
0ace2ca8
TH		 552 u32 data, struct gro_remcsum *grc,
553 bool nopartial)
dfd8645e
TH		 554{
555 size_t start, offset, plen;
dfd8645e
TH		 556
557 if (skb->remcsum_offload)
26c4f7da 558 return NULL;
dfd8645e
TH		 559
560 if (!NAPI_GRO_CB(skb)->csum_valid)
561 return NULL;
562
563 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
564 offset = start + ((data & VXLAN_RCO_UDP) ?
565 offsetof(struct udphdr, check) :
566 offsetof(struct tcphdr, check));
567
568 plen = hdrlen + offset + sizeof(u16);
569
570 /* Pull checksum that will be written */
571 if (skb_gro_header_hard(skb, off + plen)) {
572 vh = skb_gro_header_slow(skb, off + plen, off);
573 if (!vh)
574 return NULL;
575 }
576
26c4f7da 577 skb_gro_remcsum_process(skb, (void *)vh + hdrlen,
0ace2ca8 578 start, offset, grc, nopartial);
dfd8645e
TH		 579
580 skb->remcsum_offload = 1;
581
582 return vh;
583}
584
a2b12f3c
TH		 585static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
586 struct sk_buff *skb,
587 struct udp_offload *uoff)
dc01e7d3
OG		 588{
589 struct sk_buff *p, **pp = NULL;
590 struct vxlanhdr *vh, *vh2;
9b174d88 591 unsigned int hlen, off_vx;
dc01e7d3 592 int flush = 1;
dfd8645e
TH		 593 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
594 udp_offloads);
595 u32 flags;
26c4f7da
TH		 596 struct gro_remcsum grc;
597
598 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 599
600 off_vx = skb_gro_offset(skb);
601 hlen = off_vx + sizeof(*vh);
602 vh = skb_gro_header_fast(skb, off_vx);
603 if (skb_gro_header_hard(skb, hlen)) {
604 vh = skb_gro_header_slow(skb, hlen, off_vx);
605 if (unlikely(!vh))
606 goto out;
607 }
dc01e7d3 608
dfd8645e
TH		 609 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
610 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
611
612 flags = ntohl(vh->vx_flags);
613
614 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
615 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
0ace2ca8
TH		 616 ntohl(vh->vx_vni), &grc,
617 !!(vs->flags &
618 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 619
620 if (!vh)
621 goto out;
622 }
623
dc01e7d3
OG		 624 flush = 0;
625
626 for (p = *head; p; p = p->next) {
627 if (!NAPI_GRO_CB(p)->same_flow)
628 continue;
629
630 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 631 if (vh->vx_flags != vh2->vx_flags ||
632 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 633 NAPI_GRO_CB(p)->same_flow = 0;
634 continue;
635 }
dc01e7d3
OG		 636 }
637
9b174d88 638 pp = eth_gro_receive(head, skb);
dc01e7d3 639
dc01e7d3 640out:
26c4f7da 641 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 642 NAPI_GRO_CB(skb)->flush |= flush;
643
644 return pp;
645}
646
a2b12f3c
TH		 647static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
648 struct udp_offload *uoff)
dc01e7d3 649{
cfdf1e1b
JG		 650 udp_tunnel_gro_complete(skb, nhoff);
651
9b174d88 652 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 653}
654
53cf5275 655/* Notify netdevs that UDP port started listening */
dc01e7d3 656static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 657{
658 struct net_device *dev;
dc01e7d3 659 struct sock *sk = vs->sock->sk;
53cf5275
JG		 660 struct net *net = sock_net(sk);
661 sa_family_t sa_family = sk->sk_family;
35e42379 662 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 663 int err;
664
665 if (sa_family == AF_INET) {
666 err = udp_add_offload(&vs->udp_offloads);
667 if (err)
668 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
669 }
53cf5275
JG		 670
671 rcu_read_lock();
672 for_each_netdev_rcu(net, dev) {
673 if (dev->netdev_ops->ndo_add_vxlan_port)
674 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
675 port);
676 }
677 rcu_read_unlock();
678}
679
680/* Notify netdevs that UDP port is no more listening */
dc01e7d3 681static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 682{
683 struct net_device *dev;
dc01e7d3 684 struct sock *sk = vs->sock->sk;
53cf5275
JG		 685 struct net *net = sock_net(sk);
686 sa_family_t sa_family = sk->sk_family;
35e42379 687 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 688
689 rcu_read_lock();
690 for_each_netdev_rcu(net, dev) {
691 if (dev->netdev_ops->ndo_del_vxlan_port)
692 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
693 port);
694 }
695 rcu_read_unlock();
dc01e7d3
OG		 696
697 if (sa_family == AF_INET)
698 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 699}
700
d342894c 701/* Add new entry to forwarding table -- assumes lock held */
702static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 703 const u8 *mac, union vxlan_addr *ip,
6681712d 704 __u16 state, __u16 flags,
73cf3317 705 __be16 port, __u32 vni, __u32 ifindex,
ae884082 706 __u8 ndm_flags)
d342894c 707{
9e4b93f9 708 struct vxlan_rdst *rd = NULL;
d342894c 709 struct vxlan_fdb *f;
710 int notify = 0;
711
014be2c8 712 f = __vxlan_find_mac(vxlan, mac);
d342894c 713 if (f) {
714 if (flags & NLM_F_EXCL) {
715 netdev_dbg(vxlan->dev,
716 "lost race to create %pM\n", mac);
717 return -EEXIST;
718 }
719 if (f->state != state) {
720 f->state = state;
721 f->updated = jiffies;
722 notify = 1;
723 }
ae884082
DS		 724 if (f->flags != ndm_flags) {
725 f->flags = ndm_flags;
726 f->updated = jiffies;
727 notify = 1;
728 }
906dc186
TR		 729 if ((flags & NLM_F_REPLACE)) {
730 /* Only change unicasts */
731 if (!(is_multicast_ether_addr(f->eth_addr) ||
732 is_zero_ether_addr(f->eth_addr))) {
733 int rc = vxlan_fdb_replace(f, ip, port, vni,
734 ifindex);
735
736 if (rc < 0)
737 return rc;
738 notify |= rc;
739 } else
740 return -EOPNOTSUPP;
741 }
6681712d 742 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 743 (is_multicast_ether_addr(f->eth_addr) ||
744 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 745 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
746 &rd);
6681712d
DS		 747
748 if (rc < 0)
749 return rc;
750 notify |= rc;
751 }
d342894c 752 } else {
753 if (!(flags & NLM_F_CREATE))
754 return -ENOENT;
755
756 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
757 return -ENOSPC;
758
906dc186
TR		 759 /* Disallow replace to add a multicast entry */
760 if ((flags & NLM_F_REPLACE) &&
761 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
762 return -EOPNOTSUPP;
763
e4c7ed41 764 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 765 f = kmalloc(sizeof(*f), GFP_ATOMIC);
766 if (!f)
767 return -ENOMEM;
768
769 notify = 1;
d342894c 770 f->state = state;
ae884082 771 f->flags = ndm_flags;
d342894c 772 f->updated = f->used = jiffies;
3e61aa8f 773 INIT_LIST_HEAD(&f->remotes);
d342894c 774 memcpy(f->eth_addr, mac, ETH_ALEN);
775
9e4b93f9 776 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 777
d342894c 778 ++vxlan->addrcnt;
779 hlist_add_head_rcu(&f->hlist,
780 vxlan_fdb_head(vxlan, mac));
781 }
782
9e4b93f9
ND		 783 if (notify) {
784 if (rd == NULL)
785 rd = first_remote_rtnl(f);
786 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
787 }
d342894c 788
789 return 0;
790}
791
6706c82e 792static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 793{
794 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 795 struct vxlan_rdst *rd, *nd;
6681712d 796
3e61aa8f 797 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 798 kfree(rd);
6681712d
DS		 799 kfree(f);
800}
801
d342894c 802static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
803{
804 netdev_dbg(vxlan->dev,
805 "delete %pM\n", f->eth_addr);
806
807 --vxlan->addrcnt;
9e4b93f9 808 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 809
810 hlist_del_rcu(&f->hlist);
6681712d 811 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 812}
813
f0b074be 814static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 815 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 816{
6681712d 817 struct net *net = dev_net(vxlan->dev);
e4c7ed41 818 int err;
d342894c 819
f0b074be 820 if (tb[NDA_DST]) {
e4c7ed41
CW		 821 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
822 if (err)
823 return err;
f0b074be 824 } else {
e4c7ed41
CW		 825 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
826 if (remote->sa.sa_family == AF_INET) {
827 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
828 ip->sa.sa_family = AF_INET;
829#if IS_ENABLED(CONFIG_IPV6)
830 } else {
831 ip->sin6.sin6_addr = in6addr_any;
832 ip->sa.sa_family = AF_INET6;
833#endif
834 }
f0b074be 835 }
d342894c 836
6681712d 837 if (tb[NDA_PORT]) {
73cf3317 838 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 839 return -EINVAL;
f0b074be
MR		 840 *port = nla_get_be16(tb[NDA_PORT]);
841 } else {
842 *port = vxlan->dst_port;
843 }
6681712d
DS		 844
845 if (tb[NDA_VNI]) {
846 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
847 return -EINVAL;
f0b074be
MR		 848 *vni = nla_get_u32(tb[NDA_VNI]);
849 } else {
850 *vni = vxlan->default_dst.remote_vni;
851 }
6681712d
DS		 852
853 if (tb[NDA_IFINDEX]) {
5abb0029 854 struct net_device *tdev;
6681712d
DS		 855
856 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
857 return -EINVAL;
f0b074be 858 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 859 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 860 if (!tdev)
6681712d 861 return -EADDRNOTAVAIL;
f0b074be
MR		 862 } else {
863 *ifindex = 0;
864 }
865
866 return 0;
867}
868
869/* Add static entry (via netlink) */
870static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
871 struct net_device *dev,
f6f6424b 872 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 873{
874 struct vxlan_dev *vxlan = netdev_priv(dev);
875 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 876 union vxlan_addr ip;
f0b074be
MR		 877 __be16 port;
878 u32 vni, ifindex;
879 int err;
880
881 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
882 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
883 ndm->ndm_state);
884 return -EINVAL;
885 }
886
887 if (tb[NDA_DST] == NULL)
888 return -EINVAL;
889
890 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
891 if (err)
892 return err;
6681712d 893
5933a7bb
MR		 894 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
895 return -EAFNOSUPPORT;
896
d342894c 897 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 898 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 899 port, vni, ifindex, ndm->ndm_flags);
d342894c 900 spin_unlock_bh(&vxlan->hash_lock);
901
902 return err;
903}
904
905/* Delete entry (via netlink) */
1690be63
VY		 906static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
907 struct net_device *dev,
f6f6424b 908 const unsigned char *addr, u16 vid)
d342894c 909{
910 struct vxlan_dev *vxlan = netdev_priv(dev);
911 struct vxlan_fdb *f;
bc7892ba 912 struct vxlan_rdst *rd = NULL;
e4c7ed41 913 union vxlan_addr ip;
bc7892ba
MR		 914 __be16 port;
915 u32 vni, ifindex;
916 int err;
917
918 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
919 if (err)
920 return err;
921
922 err = -ENOENT;
d342894c 923
924 spin_lock_bh(&vxlan->hash_lock);
925 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 926 if (!f)
927 goto out;
928
e4c7ed41
CW		 929 if (!vxlan_addr_any(&ip)) {
930 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 931 if (!rd)
932 goto out;
933 }
934
935 err = 0;
936
937 /* remove a destination if it's not the only one on the list,
938 * otherwise destroy the fdb entry
939 */
940 if (rd && !list_is_singular(&f->remotes)) {
941 list_del_rcu(&rd->list);
9e4b93f9 942 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 943 kfree_rcu(rd, rcu);
bc7892ba 944 goto out;
d342894c 945 }
bc7892ba
MR		 946
947 vxlan_fdb_destroy(vxlan, f);
948
949out:
d342894c 950 spin_unlock_bh(&vxlan->hash_lock);
951
952 return err;
953}
954
955/* Dump forwarding table */
956static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 957 struct net_device *dev,
958 struct net_device *filter_dev, int idx)
d342894c 959{
960 struct vxlan_dev *vxlan = netdev_priv(dev);
961 unsigned int h;
962
963 for (h = 0; h < FDB_HASH_SIZE; ++h) {
964 struct vxlan_fdb *f;
d342894c 965 int err;
966
b67bfe0d 967 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 968 struct vxlan_rdst *rd;
6681712d 969
3e61aa8f
SH		 970 if (idx < cb->args[0])
971 goto skip;
972
973 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 974 err = vxlan_fdb_info(skb, vxlan, f,
975 NETLINK_CB(cb->skb).portid,
976 cb->nlh->nlmsg_seq,
977 RTM_NEWNEIGH,
978 NLM_F_MULTI, rd);
979 if (err < 0)
3e61aa8f 980 goto out;
6681712d 981 }
3e61aa8f
SH		 982skip:
983 ++idx;
d342894c 984 }
985 }
3e61aa8f 986out:
d342894c 987 return idx;
988}
989
990/* Watch incoming packets to learn mapping between Ethernet address
991 * and Tunnel endpoint.
26a41ae6 992 * Return true if packet is bogus and should be droppped.
d342894c 993 */
26a41ae6 994static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 995 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 996{
997 struct vxlan_dev *vxlan = netdev_priv(dev);
998 struct vxlan_fdb *f;
d342894c 999
1000 f = vxlan_find_mac(vxlan, src_mac);
1001 if (likely(f)) {
5ca5461c 1002 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1003
e4c7ed41 1004 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 1005 return false;
1006
1007 /* Don't migrate static entries, drop packets */
eb064c3b 1008 if (f->state & NUD_NOARP)
26a41ae6 1009 return true;
d342894c 1010
1011 if (net_ratelimit())
1012 netdev_info(dev,
e4c7ed41 1013 "%pM migrated from %pIS to %pIS\n",
a4870f79 1014 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1015
e4c7ed41 1016 rdst->remote_ip = *src_ip;
d342894c 1017 f->updated = jiffies;
9e4b93f9 1018 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1019 } else {
1020 /* learned new entry */
1021 spin_lock(&vxlan->hash_lock);
3bf74b1a 1022
1023 /* close off race between vxlan_flush and incoming packets */
1024 if (netif_running(dev))
1025 vxlan_fdb_create(vxlan, src_mac, src_ip,
1026 NUD_REACHABLE,
1027 NLM_F_EXCL|NLM_F_CREATE,
1028 vxlan->dst_port,
1029 vxlan->default_dst.remote_vni,
1030 0, NTF_SELF);
d342894c 1031 spin_unlock(&vxlan->hash_lock);
1032 }
26a41ae6 1033
1034 return false;
d342894c 1035}
1036
d342894c 1037/* See if multicast group is already in use by other ID */
95ab0991 1038static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1039{
553675fb 1040 struct vxlan_dev *vxlan;
d342894c 1041
95ab0991
G		 1042 /* The vxlan_sock is only used by dev, leaving group has
1043 * no effect on other vxlan devices.
1044 */
1045 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1046 return false;
1047
553675fb 1048 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1049 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1050 continue;
d342894c 1051
95ab0991
G		 1052 if (vxlan->vn_sock != dev->vn_sock)
1053 continue;
1054
1055 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1056 &dev->default_dst.remote_ip))
1057 continue;
1058
1059 if (vxlan->default_dst.remote_ifindex !=
1060 dev->default_dst.remote_ifindex)
1061 continue;
1062
1063 return true;
553675fb 1064 }
d342894c 1065
1066 return false;
1067}
1068
012a5729 1069void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1070{
53cf5275
JG		 1071 struct sock *sk = vs->sock->sk;
1072 struct net *net = sock_net(sk);
1073 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1074
7c47cedf
SH		 1075 if (!atomic_dec_and_test(&vs->refcnt))
1076 return;
d342894c 1077
1c51a915 1078 spin_lock(&vn->sock_lock);
7c47cedf 1079 hlist_del_rcu(&vs->hlist);
dc01e7d3 1080 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1081 spin_unlock(&vn->sock_lock);
1082
7c47cedf 1083 queue_work(vxlan_wq, &vs->del_work);
d342894c 1084}
012a5729 1085EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1086
56ef9c90 1087/* Update multicast group membership when first VNI on
3fc2de2f 1088 * multicast asddress is brought up
7c47cedf 1089 */
56ef9c90 1090static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1091{
7c47cedf
SH		 1092 struct vxlan_sock *vs = vxlan->vn_sock;
1093 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1094 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1095 int ifindex = vxlan->default_dst.remote_ifindex;
56ef9c90 1096 int ret;
d342894c 1097
d342894c 1098 lock_sock(sk);
e4c7ed41
CW		 1099 if (ip->sa.sa_family == AF_INET) {
1100 struct ip_mreqn mreq = {
1101 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1102 .imr_ifindex = ifindex,
1103 };
1104
56ef9c90 1105 ret = ip_mc_join_group(sk, &mreq);
e4c7ed41
CW		 1106#if IS_ENABLED(CONFIG_IPV6)
1107 } else {
56ef9c90
MRL		 1108 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1109 &ip->sin6.sin6_addr);
e4c7ed41
CW		 1110#endif
1111 }
3fc2de2f 1112 release_sock(sk);
1113
56ef9c90 1114 return ret;
3fc2de2f 1115}
1116
1117/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1118static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1119{
3fc2de2f 1120 struct vxlan_sock *vs = vxlan->vn_sock;
1121 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1122 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1123 int ifindex = vxlan->default_dst.remote_ifindex;
56ef9c90 1124 int ret;
3fc2de2f 1125
1126 lock_sock(sk);
e4c7ed41
CW		 1127 if (ip->sa.sa_family == AF_INET) {
1128 struct ip_mreqn mreq = {
1129 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1130 .imr_ifindex = ifindex,
1131 };
1132
56ef9c90 1133 ret = ip_mc_leave_group(sk, &mreq);
e4c7ed41
CW		 1134#if IS_ENABLED(CONFIG_IPV6)
1135 } else {
56ef9c90
MRL		 1136 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1137 &ip->sin6.sin6_addr);
e4c7ed41
CW		 1138#endif
1139 }
d342894c 1140 release_sock(sk);
d342894c 1141
56ef9c90 1142 return ret;
d342894c 1143}
1144
dfd8645e 1145static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
0ace2ca8 1146 size_t hdrlen, u32 data, bool nopartial)
dfd8645e
TH		 1147{
1148 size_t start, offset, plen;
dfd8645e 1149
dfd8645e
TH		 1150 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1151 offset = start + ((data & VXLAN_RCO_UDP) ?
1152 offsetof(struct udphdr, check) :
1153 offsetof(struct tcphdr, check));
1154
1155 plen = hdrlen + offset + sizeof(u16);
1156
1157 if (!pskb_may_pull(skb, plen))
1158 return NULL;
1159
1160 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1161
0ace2ca8
TH		 1162 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1163 nopartial);
dfd8645e
TH		 1164
1165 return vh;
1166}
1167
d342894c 1168/* Callback from net/ipv4/udp.c to receive packets */
1169static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1170{
5cfccc5a 1171 struct vxlan_sock *vs;
d342894c 1172 struct vxlanhdr *vxh;
3bf39475 1173 u32 flags, vni;
3511494c 1174 struct vxlan_metadata md = {0};
d342894c 1175
d342894c 1176 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1177 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1178 goto error;
1179
7ce04758 1180 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1181 flags = ntohl(vxh->vx_flags);
1182 vni = ntohl(vxh->vx_vni);
1183
1184 if (flags & VXLAN_HF_VNI) {
1185 flags &= ~VXLAN_HF_VNI;
1186 } else {
1187 /* VNI flag always required to be set */
1188 goto bad_flags;
d342894c 1189 }
1190
5cfccc5a
PS		 1191 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1192 goto drop;
dfd8645e 1193 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1194
559835ea 1195 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1196 if (!vs)
d342894c 1197 goto drop;
d342894c 1198
dfd8645e 1199 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
0ace2ca8
TH		 1200 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1201 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 1202 if (!vxh)
1203 goto drop;
1204
1205 flags &= ~VXLAN_HF_RCO;
1206 vni &= VXLAN_VID_MASK;
1207 }
1208
3511494c
TG		 1209 /* For backwards compatibility, only allow reserved fields to be
1210 * used by VXLAN extensions if explicitly requested.
1211 */
1212 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1213 struct vxlanhdr_gbp *gbp;
1214
1215 gbp = (struct vxlanhdr_gbp *)vxh;
1216 md.gbp = ntohs(gbp->policy_id);
1217
1218 if (gbp->dont_learn)
1219 md.gbp |= VXLAN_GBP_DONT_LEARN;
1220
1221 if (gbp->policy_applied)
1222 md.gbp |= VXLAN_GBP_POLICY_APPLIED;
1223
1224 flags &= ~VXLAN_GBP_USED_BITS;
1225 }
1226
dfd8645e 1227 if (flags || (vni & ~VXLAN_VID_MASK)) {
3bf39475
TH		 1228 /* If there are any unprocessed flags remaining treat
1229 * this as a malformed packet. This behavior diverges from
1230 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1231 * in reserved fields are to be ignored. The approach here
1232 * maintains compatbility with previous stack code, and also
1233 * is more robust and provides a little more security in
1234 * adding extensions to VXLAN.
1235 */
1236
1237 goto bad_flags;
1238 }
1239
3511494c
TG		 1240 md.vni = vxh->vx_vni;
1241 vs->rcv(vs, skb, &md);
5cfccc5a
PS		 1242 return 0;
1243
1244drop:
1245 /* Consume bad packet */
1246 kfree_skb(skb);
1247 return 0;
1248
3bf39475
TH		 1249bad_flags:
1250 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1251 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1252
5cfccc5a
PS		 1253error:
1254 /* Return non vxlan pkt */
1255 return 1;
1256}
1257
3511494c
TG		 1258static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1259 struct vxlan_metadata *md)
5cfccc5a 1260{
e4c7ed41
CW		 1261 struct iphdr *oip = NULL;
1262 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1263 struct vxlan_dev *vxlan;
8f84985f 1264 struct pcpu_sw_netstats *stats;
e4c7ed41 1265 union vxlan_addr saddr;
5cfccc5a 1266 __u32 vni;
e4c7ed41
CW		 1267 int err = 0;
1268 union vxlan_addr *remote_ip;
5cfccc5a 1269
3511494c 1270 vni = ntohl(md->vni) >> 8;
5cfccc5a
PS		 1271 /* Is this VNI defined? */
1272 vxlan = vxlan_vs_find_vni(vs, vni);
1273 if (!vxlan)
d342894c 1274 goto drop;
d342894c 1275
e4c7ed41 1276 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1277 skb_reset_mac_header(skb);
f01ec1c0 1278 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1279 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1280 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1281
1282 /* Ignore packet loops (and multicast echo) */
7367d0b5 1283 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1284 goto drop;
1285
7ce04758 1286 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1287 if (remote_ip->sa.sa_family == AF_INET) {
1288 oip = ip_hdr(skb);
1289 saddr.sin.sin_addr.s_addr = oip->saddr;
1290 saddr.sa.sa_family = AF_INET;
1291#if IS_ENABLED(CONFIG_IPV6)
1292 } else {
1293 oip6 = ipv6_hdr(skb);
1294 saddr.sin6.sin6_addr = oip6->saddr;
1295 saddr.sa.sa_family = AF_INET6;
1296#endif
1297 }
1298
26a41ae6 1299 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1300 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1301 goto drop;
d342894c 1302
d342894c 1303 skb_reset_network_header(skb);
3511494c 1304 skb->mark = md->gbp;
0afb1666 1305
e4c7ed41
CW		 1306 if (oip6)
1307 err = IP6_ECN_decapsulate(oip6, skb);
1308 if (oip)
1309 err = IP_ECN_decapsulate(oip, skb);
1310
d342894c 1311 if (unlikely(err)) {
e4c7ed41
CW		 1312 if (log_ecn_error) {
1313 if (oip6)
1314 net_info_ratelimited("non-ECT from %pI6\n",
1315 &oip6->saddr);
1316 if (oip)
1317 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1318 &oip->saddr, oip->tos);
1319 }
d342894c 1320 if (err > 1) {
1321 ++vxlan->dev->stats.rx_frame_errors;
1322 ++vxlan->dev->stats.rx_errors;
1323 goto drop;
1324 }
1325 }
1326
e8171045 1327 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1328 u64_stats_update_begin(&stats->syncp);
1329 stats->rx_packets++;
1330 stats->rx_bytes += skb->len;
1331 u64_stats_update_end(&stats->syncp);
1332
1333 netif_rx(skb);
1334
5cfccc5a 1335 return;
d342894c 1336drop:
1337 /* Consume bad packet */
1338 kfree_skb(skb);
d342894c 1339}
1340
e4f67add
DS		 1341static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1342{
1343 struct vxlan_dev *vxlan = netdev_priv(dev);
1344 struct arphdr *parp;
1345 u8 *arpptr, *sha;
1346 __be32 sip, tip;
1347 struct neighbour *n;
1348
1349 if (dev->flags & IFF_NOARP)
1350 goto out;
1351
1352 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1353 dev->stats.tx_dropped++;
1354 goto out;
1355 }
1356 parp = arp_hdr(skb);
1357
1358 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1359 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1360 parp->ar_pro != htons(ETH_P_IP) ||
1361 parp->ar_op != htons(ARPOP_REQUEST) ||
1362 parp->ar_hln != dev->addr_len ||
1363 parp->ar_pln != 4)
1364 goto out;
1365 arpptr = (u8 *)parp + sizeof(struct arphdr);
1366 sha = arpptr;
1367 arpptr += dev->addr_len; /* sha */
1368 memcpy(&sip, arpptr, sizeof(sip));
1369 arpptr += sizeof(sip);
1370 arpptr += dev->addr_len; /* tha */
1371 memcpy(&tip, arpptr, sizeof(tip));
1372
1373 if (ipv4_is_loopback(tip) ||
1374 ipv4_is_multicast(tip))
1375 goto out;
1376
1377 n = neigh_lookup(&arp_tbl, &tip, dev);
1378
1379 if (n) {
e4f67add
DS		 1380 struct vxlan_fdb *f;
1381 struct sk_buff *reply;
1382
1383 if (!(n->nud_state & NUD_CONNECTED)) {
1384 neigh_release(n);
1385 goto out;
1386 }
1387
1388 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1389 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1390 /* bridge-local neighbor */
1391 neigh_release(n);
1392 goto out;
1393 }
1394
1395 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1396 n->ha, sha);
1397
1398 neigh_release(n);
1399
7346135d
DS		 1400 if (reply == NULL)
1401 goto out;
1402
e4f67add
DS		 1403 skb_reset_mac_header(reply);
1404 __skb_pull(reply, skb_network_offset(reply));
1405 reply->ip_summed = CHECKSUM_UNNECESSARY;
1406 reply->pkt_type = PACKET_HOST;
1407
1408 if (netif_rx_ni(reply) == NET_RX_DROP)
1409 dev->stats.rx_dropped++;
e4c7ed41
CW		 1410 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1411 union vxlan_addr ipa = {
1412 .sin.sin_addr.s_addr = tip,
a45e92a5 1413 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1414 };
1415
1416 vxlan_ip_miss(dev, &ipa);
1417 }
e4f67add
DS		 1418out:
1419 consume_skb(skb);
1420 return NETDEV_TX_OK;
1421}
1422
f564f45c 1423#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1424static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1425 struct neighbour *n, bool isrouter)
1426{
1427 struct net_device *dev = request->dev;
1428 struct sk_buff *reply;
1429 struct nd_msg *ns, *na;
1430 struct ipv6hdr *pip6;
1431 u8 *daddr;
1432 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1433 int ns_olen;
1434 int i, len;
1435
1436 if (dev == NULL)
1437 return NULL;
1438
1439 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1440 sizeof(*na) + na_olen + dev->needed_tailroom;
1441 reply = alloc_skb(len, GFP_ATOMIC);
1442 if (reply == NULL)
1443 return NULL;
1444
1445 reply->protocol = htons(ETH_P_IPV6);
1446 reply->dev = dev;
1447 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1448 skb_push(reply, sizeof(struct ethhdr));
1449 skb_set_mac_header(reply, 0);
1450
1451 ns = (struct nd_msg *)skb_transport_header(request);
1452
1453 daddr = eth_hdr(request)->h_source;
1454 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1455 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1456 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1457 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1458 break;
1459 }
1460 }
1461
1462 /* Ethernet header */
1463 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1464 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1465 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1466 reply->protocol = htons(ETH_P_IPV6);
1467
1468 skb_pull(reply, sizeof(struct ethhdr));
1469 skb_set_network_header(reply, 0);
1470 skb_put(reply, sizeof(struct ipv6hdr));
1471
1472 /* IPv6 header */
1473
1474 pip6 = ipv6_hdr(reply);
1475 memset(pip6, 0, sizeof(struct ipv6hdr));
1476 pip6->version = 6;
1477 pip6->priority = ipv6_hdr(request)->priority;
1478 pip6->nexthdr = IPPROTO_ICMPV6;
1479 pip6->hop_limit = 255;
1480 pip6->daddr = ipv6_hdr(request)->saddr;
1481 pip6->saddr = *(struct in6_addr *)n->primary_key;
1482
1483 skb_pull(reply, sizeof(struct ipv6hdr));
1484 skb_set_transport_header(reply, 0);
1485
1486 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1487
1488 /* Neighbor Advertisement */
1489 memset(na, 0, sizeof(*na)+na_olen);
1490 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1491 na->icmph.icmp6_router = isrouter;
1492 na->icmph.icmp6_override = 1;
1493 na->icmph.icmp6_solicited = 1;
1494 na->target = ns->target;
1495 ether_addr_copy(&na->opt[2], n->ha);
1496 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1497 na->opt[1] = na_olen >> 3;
1498
1499 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1500 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1501 csum_partial(na, sizeof(*na)+na_olen, 0));
1502
1503 pip6->payload_len = htons(sizeof(*na)+na_olen);
1504
1505 skb_push(reply, sizeof(struct ipv6hdr));
1506
1507 reply->ip_summed = CHECKSUM_UNNECESSARY;
1508
1509 return reply;
1510}
1511
f564f45c
CW		 1512static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1513{
1514 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1515 struct nd_msg *msg;
f564f45c
CW		 1516 const struct ipv6hdr *iphdr;
1517 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1518 struct neighbour *n;
1519 struct inet6_dev *in6_dev;
f564f45c
CW		 1520
1521 in6_dev = __in6_dev_get(dev);
1522 if (!in6_dev)
1523 goto out;
1524
f564f45c
CW		 1525 iphdr = ipv6_hdr(skb);
1526 saddr = &iphdr->saddr;
1527 daddr = &iphdr->daddr;
1528
f564f45c
CW		 1529 msg = (struct nd_msg *)skb_transport_header(skb);
1530 if (msg->icmph.icmp6_code != 0 ||
1531 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1532 goto out;
1533
4b29dba9
DS		 1534 if (ipv6_addr_loopback(daddr) ||
1535 ipv6_addr_is_multicast(&msg->target))
1536 goto out;
1537
1538 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1539
1540 if (n) {
1541 struct vxlan_fdb *f;
4b29dba9 1542 struct sk_buff *reply;
f564f45c
CW		 1543
1544 if (!(n->nud_state & NUD_CONNECTED)) {
1545 neigh_release(n);
1546 goto out;
1547 }
1548
1549 f = vxlan_find_mac(vxlan, n->ha);
1550 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1551 /* bridge-local neighbor */
1552 neigh_release(n);
1553 goto out;
1554 }
1555
4b29dba9
DS		 1556 reply = vxlan_na_create(skb, n,
1557 !!(f ? f->flags & NTF_ROUTER : 0));
1558
f564f45c 1559 neigh_release(n);
4b29dba9
DS		 1560
1561 if (reply == NULL)
1562 goto out;
1563
1564 if (netif_rx_ni(reply) == NET_RX_DROP)
1565 dev->stats.rx_dropped++;
1566
f564f45c 1567 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1568 union vxlan_addr ipa = {
1569 .sin6.sin6_addr = msg->target,
a45e92a5 1570 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1571 };
1572
f564f45c
CW		 1573 vxlan_ip_miss(dev, &ipa);
1574 }
1575
1576out:
1577 consume_skb(skb);
1578 return NETDEV_TX_OK;
1579}
1580#endif
1581
e4f67add
DS		 1582static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1583{
1584 struct vxlan_dev *vxlan = netdev_priv(dev);
1585 struct neighbour *n;
e4f67add
DS		 1586
1587 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1588 return false;
1589
1590 n = NULL;
1591 switch (ntohs(eth_hdr(skb)->h_proto)) {
1592 case ETH_P_IP:
e15a00aa
CW		 1593 {
1594 struct iphdr *pip;
1595
e4f67add
DS		 1596 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1597 return false;
1598 pip = ip_hdr(skb);
1599 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1600 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1601 union vxlan_addr ipa = {
1602 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1603 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1604 };
1605
1606 vxlan_ip_miss(dev, &ipa);
1607 return false;
1608 }
1609
e4f67add 1610 break;
e15a00aa
CW		 1611 }
1612#if IS_ENABLED(CONFIG_IPV6)
1613 case ETH_P_IPV6:
1614 {
1615 struct ipv6hdr *pip6;
1616
1617 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1618 return false;
1619 pip6 = ipv6_hdr(skb);
1620 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1621 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1622 union vxlan_addr ipa = {
1623 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1624 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1625 };
1626
1627 vxlan_ip_miss(dev, &ipa);
1628 return false;
1629 }
1630
1631 break;
1632 }
1633#endif
e4f67add
DS		 1634 default:
1635 return false;
1636 }
1637
1638 if (n) {
1639 bool diff;
1640
7367d0b5 1641 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1642 if (diff) {
1643 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1644 dev->addr_len);
1645 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1646 }
1647 neigh_release(n);
1648 return diff;
e4c7ed41
CW		 1649 }
1650
e4f67add
DS		 1651 return false;
1652}
1653
af33c1ad 1654static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1655 struct vxlan_metadata *md)
1656{
1657 struct vxlanhdr_gbp *gbp;
1658
db79a621
TG		 1659 if (!md->gbp)
1660 return;
1661
3511494c
TG		 1662 gbp = (struct vxlanhdr_gbp *)vxh;
1663 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1664
1665 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1666 gbp->dont_learn = 1;
1667
1668 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1669 gbp->policy_applied = 1;
1670
1671 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1672}
1673
e4c7ed41 1674#if IS_ENABLED(CONFIG_IPV6)
af33c1ad 1675static int vxlan6_xmit_skb(struct dst_entry *dst, struct sk_buff *skb,
e4c7ed41
CW		 1676 struct net_device *dev, struct in6_addr *saddr,
1677 struct in6_addr *daddr, __u8 prio, __u8 ttl,
3511494c 1678 __be16 src_port, __be16 dst_port,
af33c1ad 1679 struct vxlan_metadata *md, bool xnet, u32 vxflags)
e4c7ed41 1680{
e4c7ed41 1681 struct vxlanhdr *vxh;
e4c7ed41
CW		 1682 int min_headroom;
1683 int err;
af33c1ad 1684 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
dfd8645e
TH		 1685 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1686 u16 hdrlen = sizeof(struct vxlanhdr);
1687
af33c1ad 1688 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1689 skb->ip_summed == CHECKSUM_PARTIAL) {
1690 int csum_start = skb_checksum_start_offset(skb);
1691
1692 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1693 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1694 (skb->csum_offset == offsetof(struct udphdr, check) ||
1695 skb->csum_offset == offsetof(struct tcphdr, check))) {
1696 udp_sum = false;
1697 type |= SKB_GSO_TUNNEL_REMCSUM;
1698 }
1699 }
e4c7ed41 1700
dfd8645e 1701 skb = iptunnel_handle_offloads(skb, udp_sum, type);
74f47278
PS		 1702 if (IS_ERR(skb)) {
1703 err = -EINVAL;
1704 goto err;
1705 }
e4c7ed41 1706
f01ec1c0 1707 skb_scrub_packet(skb, xnet);
963a88b3 1708
e4c7ed41
CW		 1709 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1710 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1711 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1712
1713 /* Need space for new headers (invalidates iph ptr) */
1714 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1715 if (unlikely(err)) {
1716 kfree_skb(skb);
1717 goto err;
1718 }
e4c7ed41 1719
5968250c 1720 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1721 if (WARN_ON(!skb)) {
1722 err = -ENOMEM;
1723 goto err;
1724 }
e4c7ed41
CW		 1725
1726 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1727 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1728 vxh->vx_vni = md->vni;
e4c7ed41 1729
dfd8645e
TH		 1730 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1731 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1732 VXLAN_RCO_SHIFT;
1733
1734 if (skb->csum_offset == offsetof(struct udphdr, check))
1735 data |= VXLAN_RCO_UDP;
1736
1737 vxh->vx_vni |= htonl(data);
1738 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1739
1740 if (!skb_is_gso(skb)) {
1741 skb->ip_summed = CHECKSUM_NONE;
1742 skb->encapsulation = 0;
1743 }
1744 }
1745
af33c1ad
TH		 1746 if (vxflags & VXLAN_F_GBP)
1747 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1748
996c9fd1
TH		 1749 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1750
d998f8ef
TH		 1751 udp_tunnel6_xmit_skb(dst, skb, dev, saddr, daddr, prio,
1752 ttl, src_port, dst_port,
af33c1ad 1753 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
e4c7ed41 1754 return 0;
74f47278
PS		 1755err:
1756 dst_release(dst);
1757 return err;
e4c7ed41
CW		 1758}
1759#endif
1760
af33c1ad 1761int vxlan_xmit_skb(struct rtable *rt, struct sk_buff *skb,
49560532 1762 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
3511494c 1763 __be16 src_port, __be16 dst_port,
af33c1ad 1764 struct vxlan_metadata *md, bool xnet, u32 vxflags)
49560532
PS		 1765{
1766 struct vxlanhdr *vxh;
649c5b8b 1767 int min_headroom;
49560532 1768 int err;
af33c1ad 1769 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
dfd8645e
TH		 1770 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1771 u16 hdrlen = sizeof(struct vxlanhdr);
1772
af33c1ad 1773 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1774 skb->ip_summed == CHECKSUM_PARTIAL) {
1775 int csum_start = skb_checksum_start_offset(skb);
1776
1777 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1778 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1779 (skb->csum_offset == offsetof(struct udphdr, check) ||
1780 skb->csum_offset == offsetof(struct tcphdr, check))) {
1781 udp_sum = false;
1782 type |= SKB_GSO_TUNNEL_REMCSUM;
1783 }
1784 }
49560532 1785
dfd8645e 1786 skb = iptunnel_handle_offloads(skb, udp_sum, type);
359a0ea9 1787 if (IS_ERR(skb))
74f47278 1788 return PTR_ERR(skb);
49560532 1789
649c5b8b 1790 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1791 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1792 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1793
1794 /* Need space for new headers (invalidates iph ptr) */
1795 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1796 if (unlikely(err)) {
1797 kfree_skb(skb);
649c5b8b 1798 return err;
74f47278 1799 }
649c5b8b 1800
5968250c
JP		 1801 skb = vlan_hwaccel_push_inside(skb);
1802 if (WARN_ON(!skb))
1803 return -ENOMEM;
1eaa8178 1804
49560532 1805 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1806 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1807 vxh->vx_vni = md->vni;
49560532 1808
dfd8645e
TH		 1809 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1810 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1811 VXLAN_RCO_SHIFT;
1812
1813 if (skb->csum_offset == offsetof(struct udphdr, check))
1814 data |= VXLAN_RCO_UDP;
1815
1816 vxh->vx_vni |= htonl(data);
1817 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1818
1819 if (!skb_is_gso(skb)) {
1820 skb->ip_summed = CHECKSUM_NONE;
1821 skb->encapsulation = 0;
1822 }
1823 }
1824
af33c1ad
TH		 1825 if (vxflags & VXLAN_F_GBP)
1826 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1827
996c9fd1
TH		 1828 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1829
d998f8ef
TH		 1830 return udp_tunnel_xmit_skb(rt, skb, src, dst, tos,
1831 ttl, df, src_port, dst_port, xnet,
af33c1ad 1832 !(vxflags & VXLAN_F_UDP_CSUM));
49560532
PS		 1833}
1834EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1835
9dcc71e1
SS		 1836/* Bypass encapsulation if the destination is local */
1837static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1838 struct vxlan_dev *dst_vxlan)
1839{
8f84985f 1840 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1841 union vxlan_addr loopback;
1842 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1843 struct net_device *dev = skb->dev;
1844 int len = skb->len;
9dcc71e1 1845
8f84985f
LR		 1846 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1847 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1848 skb->pkt_type = PACKET_HOST;
1849 skb->encapsulation = 0;
1850 skb->dev = dst_vxlan->dev;
1851 __skb_pull(skb, skb_network_offset(skb));
1852
e4c7ed41
CW		 1853 if (remote_ip->sa.sa_family == AF_INET) {
1854 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1855 loopback.sa.sa_family = AF_INET;
1856#if IS_ENABLED(CONFIG_IPV6)
1857 } else {
1858 loopback.sin6.sin6_addr = in6addr_loopback;
1859 loopback.sa.sa_family = AF_INET6;
1860#endif
1861 }
1862
9dcc71e1 1863 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1864 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1865
1866 u64_stats_update_begin(&tx_stats->syncp);
1867 tx_stats->tx_packets++;
ce6502a8 1868 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1869 u64_stats_update_end(&tx_stats->syncp);
1870
1871 if (netif_rx(skb) == NET_RX_SUCCESS) {
1872 u64_stats_update_begin(&rx_stats->syncp);
1873 rx_stats->rx_packets++;
ce6502a8 1874 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1875 u64_stats_update_end(&rx_stats->syncp);
1876 } else {
ce6502a8 1877 dev->stats.rx_dropped++;
9dcc71e1
SS		 1878 }
1879}
1880
4ad16930
SH		 1881static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1882 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1883{
1884 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1885 struct rtable *rt = NULL;
d342894c 1886 const struct iphdr *old_iph;
d342894c 1887 struct flowi4 fl4;
e4c7ed41 1888 union vxlan_addr *dst;
3511494c 1889 struct vxlan_metadata md;
e4c7ed41 1890 __be16 src_port = 0, dst_port;
234f5b73 1891 u32 vni;
d342894c 1892 __be16 df = 0;
1893 __u8 tos, ttl;
0e6fbc5b 1894 int err;
d342894c 1895
823aa873 1896 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1897 vni = rdst->remote_vni;
e4c7ed41 1898 dst = &rdst->remote_ip;
e4f67add 1899
e4c7ed41 1900 if (vxlan_addr_any(dst)) {
e4f67add 1901 if (did_rsc) {
e4f67add 1902 /* short-circuited back to local bridge */
9dcc71e1 1903 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1904 return;
e4f67add 1905 }
ef59febe 1906 goto drop;
e4f67add 1907 }
ef59febe 1908
d342894c 1909 old_iph = ip_hdr(skb);
1910
d342894c 1911 ttl = vxlan->ttl;
e4c7ed41 1912 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1913 ttl = 1;
1914
1915 tos = vxlan->tos;
1916 if (tos == 1)
206aaafc 1917 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1918
535fb8d0
TH		 1919 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1920 vxlan->port_max, true);
d342894c 1921
e4c7ed41
CW		 1922 if (dst->sa.sa_family == AF_INET) {
1923 memset(&fl4, 0, sizeof(fl4));
1924 fl4.flowi4_oif = rdst->remote_ifindex;
1925 fl4.flowi4_tos = RT_TOS(tos);
1926 fl4.daddr = dst->sin.sin_addr.s_addr;
1927 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1928
f01ec1c0 1929 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1930 if (IS_ERR(rt)) {
1931 netdev_dbg(dev, "no route to %pI4\n",
1932 &dst->sin.sin_addr.s_addr);
1933 dev->stats.tx_carrier_errors++;
1934 goto tx_error;
1935 }
d342894c 1936
e4c7ed41
CW		 1937 if (rt->dst.dev == dev) {
1938 netdev_dbg(dev, "circular route to %pI4\n",
1939 &dst->sin.sin_addr.s_addr);
1940 dev->stats.collisions++;
fffc15a5 1941 goto rt_tx_error;
e4c7ed41
CW		 1942 }
1943
1944 /* Bypass encapsulation if the destination is local */
1945 if (rt->rt_flags & RTCF_LOCAL &&
1946 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1947 struct vxlan_dev *dst_vxlan;
1948
1949 ip_rt_put(rt);
19ca9fc1 1950 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1951 dst->sa.sa_family, dst_port,
1952 vxlan->flags);
e4c7ed41
CW		 1953 if (!dst_vxlan)
1954 goto tx_error;
1955 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1956 return;
1957 }
1958
1959 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1960 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
3511494c
TG		 1961 md.vni = htonl(vni << 8);
1962 md.gbp = skb->mark;
d342894c 1963
af33c1ad
TH		 1964 err = vxlan_xmit_skb(rt, skb, fl4.saddr,
1965 dst->sin.sin_addr.s_addr, tos, ttl, df,
1966 src_port, dst_port, &md,
1967 !net_eq(vxlan->net, dev_net(vxlan->dev)),
1968 vxlan->flags);
74f47278
PS		 1969 if (err < 0) {
1970 /* skb is already freed. */
1971 skb = NULL;
e4c7ed41 1972 goto rt_tx_error;
74f47278
PS		 1973 }
1974
e4c7ed41
CW		 1975 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1976#if IS_ENABLED(CONFIG_IPV6)
1977 } else {
1978 struct sock *sk = vxlan->vn_sock->sock->sk;
1979 struct dst_entry *ndst;
1980 struct flowi6 fl6;
1981 u32 flags;
1982
1983 memset(&fl6, 0, sizeof(fl6));
1984 fl6.flowi6_oif = rdst->remote_ifindex;
1985 fl6.daddr = dst->sin6.sin6_addr;
1986 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1987 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1988
1989 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1990 netdev_dbg(dev, "no route to %pI6\n",
1991 &dst->sin6.sin6_addr);
1992 dev->stats.tx_carrier_errors++;
9dcc71e1 1993 goto tx_error;
e4c7ed41 1994 }
d342894c 1995
e4c7ed41
CW		 1996 if (ndst->dev == dev) {
1997 netdev_dbg(dev, "circular route to %pI6\n",
1998 &dst->sin6.sin6_addr);
1999 dst_release(ndst);
2000 dev->stats.collisions++;
2001 goto tx_error;
2002 }
0e6fbc5b 2003
e4c7ed41
CW		 2004 /* Bypass encapsulation if the destination is local */
2005 flags = ((struct rt6_info *)ndst)->rt6i_flags;
2006 if (flags & RTF_LOCAL &&
2007 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2008 struct vxlan_dev *dst_vxlan;
2009
2010 dst_release(ndst);
19ca9fc1 2011 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2012 dst->sa.sa_family, dst_port,
2013 vxlan->flags);
e4c7ed41
CW		 2014 if (!dst_vxlan)
2015 goto tx_error;
2016 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2017 return;
2018 }
49560532 2019
e4c7ed41 2020 ttl = ttl ? : ip6_dst_hoplimit(ndst);
3511494c
TG		 2021 md.vni = htonl(vni << 8);
2022 md.gbp = skb->mark;
e4c7ed41 2023
af33c1ad
TH		 2024 err = vxlan6_xmit_skb(ndst, skb, dev, &fl6.saddr, &fl6.daddr,
2025 0, ttl, src_port, dst_port, &md,
2026 !net_eq(vxlan->net, dev_net(vxlan->dev)),
2027 vxlan->flags);
e4c7ed41
CW		 2028#endif
2029 }
0e6fbc5b 2030
4ad16930 2031 return;
d342894c 2032
2033drop:
2034 dev->stats.tx_dropped++;
2035 goto tx_free;
2036
49560532
PS		 2037rt_tx_error:
2038 ip_rt_put(rt);
d342894c 2039tx_error:
2040 dev->stats.tx_errors++;
2041tx_free:
2042 dev_kfree_skb(skb);
d342894c 2043}
2044
6681712d
DS		 2045/* Transmit local packets over Vxlan
2046 *
2047 * Outer IP header inherits ECN and DF from inner header.
2048 * Outer UDP destination is the VXLAN assigned port.
2049 * source port is based on hash of flow
2050 */
2051static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2052{
2053 struct vxlan_dev *vxlan = netdev_priv(dev);
2054 struct ethhdr *eth;
2055 bool did_rsc = false;
8f646c92 2056 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2057 struct vxlan_fdb *f;
6681712d
DS		 2058
2059 skb_reset_mac_header(skb);
2060 eth = eth_hdr(skb);
2061
f564f45c
CW		 2062 if ((vxlan->flags & VXLAN_F_PROXY)) {
2063 if (ntohs(eth->h_proto) == ETH_P_ARP)
2064 return arp_reduce(dev, skb);
2065#if IS_ENABLED(CONFIG_IPV6)
2066 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2067 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2068 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2069 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2070 struct nd_msg *msg;
2071
2072 msg = (struct nd_msg *)skb_transport_header(skb);
2073 if (msg->icmph.icmp6_code == 0 &&
2074 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2075 return neigh_reduce(dev, skb);
2076 }
7a9f526f 2077 eth = eth_hdr(skb);
f564f45c
CW		 2078#endif
2079 }
6681712d
DS		 2080
2081 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2082 did_rsc = false;
2083
2084 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2085 (ntohs(eth->h_proto) == ETH_P_IP ||
2086 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2087 did_rsc = route_shortcircuit(dev, skb);
2088 if (did_rsc)
2089 f = vxlan_find_mac(vxlan, eth->h_dest);
2090 }
2091
6681712d 2092 if (f == NULL) {
afbd8bae
MR		 2093 f = vxlan_find_mac(vxlan, all_zeros_mac);
2094 if (f == NULL) {
2095 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2096 !is_multicast_ether_addr(eth->h_dest))
2097 vxlan_fdb_miss(vxlan, eth->h_dest);
2098
2099 dev->stats.tx_dropped++;
8f646c92 2100 kfree_skb(skb);
afbd8bae
MR		 2101 return NETDEV_TX_OK;
2102 }
2103 }
6681712d 2104
afbd8bae
MR		 2105 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2106 struct sk_buff *skb1;
6681712d 2107
8f646c92
ED		 2108 if (!fdst) {
2109 fdst = rdst;
2110 continue;
2111 }
afbd8bae
MR		 2112 skb1 = skb_clone(skb, GFP_ATOMIC);
2113 if (skb1)
2114 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2115 }
2116
8f646c92
ED		 2117 if (fdst)
2118 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2119 else
2120 kfree_skb(skb);
4ad16930 2121 return NETDEV_TX_OK;
6681712d
DS		 2122}
2123
d342894c 2124/* Walk the forwarding table and purge stale entries */
2125static void vxlan_cleanup(unsigned long arg)
2126{
2127 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2128 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2129 unsigned int h;
2130
2131 if (!netif_running(vxlan->dev))
2132 return;
2133
2134 spin_lock_bh(&vxlan->hash_lock);
2135 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2136 struct hlist_node *p, *n;
2137 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2138 struct vxlan_fdb *f
2139 = container_of(p, struct vxlan_fdb, hlist);
2140 unsigned long timeout;
2141
3c172868 2142 if (f->state & NUD_PERMANENT)
d342894c 2143 continue;
2144
2145 timeout = f->used + vxlan->age_interval * HZ;
2146 if (time_before_eq(timeout, jiffies)) {
2147 netdev_dbg(vxlan->dev,
2148 "garbage collect %pM\n",
2149 f->eth_addr);
2150 f->state = NUD_STALE;
2151 vxlan_fdb_destroy(vxlan, f);
2152 } else if (time_before(timeout, next_timer))
2153 next_timer = timeout;
2154 }
2155 }
2156 spin_unlock_bh(&vxlan->hash_lock);
2157
2158 mod_timer(&vxlan->age_timer, next_timer);
2159}
2160
9c2e24e1
PS		 2161static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2162{
56ef9c90 2163 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
9c2e24e1
PS		 2164 __u32 vni = vxlan->default_dst.remote_vni;
2165
2166 vxlan->vn_sock = vs;
56ef9c90 2167 spin_lock(&vn->sock_lock);
9c2e24e1 2168 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2169 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2170}
2171
d342894c 2172/* Setup stats when device is created */
2173static int vxlan_init(struct net_device *dev)
2174{
1c213bd2 2175 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2176 if (!dev->tstats)
d342894c 2177 return -ENOMEM;
2178
2179 return 0;
2180}
2181
ba609e9b 2182static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2183{
2184 struct vxlan_fdb *f;
2185
2186 spin_lock_bh(&vxlan->hash_lock);
2187 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2188 if (f)
2189 vxlan_fdb_destroy(vxlan, f);
2190 spin_unlock_bh(&vxlan->hash_lock);
2191}
2192
ebf4063e
SH		 2193static void vxlan_uninit(struct net_device *dev)
2194{
2195 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2196
ba609e9b 2197 vxlan_fdb_delete_default(vxlan);
afbd8bae 2198
ebf4063e
SH		 2199 free_percpu(dev->tstats);
2200}
2201
d342894c 2202/* Start ageing timer and join group when device is brought up */
2203static int vxlan_open(struct net_device *dev)
2204{
2205 struct vxlan_dev *vxlan = netdev_priv(dev);
56ef9c90
MRL		 2206 struct vxlan_sock *vs;
2207 int ret = 0;
1c51a915 2208
56ef9c90
MRL		 2209 vs = vxlan_sock_add(vxlan->net, vxlan->dst_port, vxlan_rcv, NULL,
2210 false, vxlan->flags);
2211 if (IS_ERR(vs))
2212 return PTR_ERR(vs);
2213
2214 vxlan_vs_add_dev(vs, vxlan);
d342894c 2215
79d4a94f 2216 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90
MRL		 2217 ret = vxlan_igmp_join(vxlan);
2218 if (ret) {
2219 vxlan_sock_release(vs);
2220 return ret;
2221 }
d342894c 2222 }
2223
2224 if (vxlan->age_interval)
2225 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2226
56ef9c90 2227 return ret;
d342894c 2228}
2229
2230/* Purge the forwarding table */
2231static void vxlan_flush(struct vxlan_dev *vxlan)
2232{
31fec5aa 2233 unsigned int h;
d342894c 2234
2235 spin_lock_bh(&vxlan->hash_lock);
2236 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2237 struct hlist_node *p, *n;
2238 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2239 struct vxlan_fdb *f
2240 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2241 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2242 if (!is_zero_ether_addr(f->eth_addr))
2243 vxlan_fdb_destroy(vxlan, f);
d342894c 2244 }
2245 }
2246 spin_unlock_bh(&vxlan->hash_lock);
2247}
2248
2249/* Cleanup timer and forwarding table on shutdown */
2250static int vxlan_stop(struct net_device *dev)
2251{
2252 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2253 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2254 struct vxlan_sock *vs = vxlan->vn_sock;
56ef9c90 2255 int ret = 0;
d342894c 2256
e4c7ed41 2257 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2258 !vxlan_group_used(vn, vxlan)) {
56ef9c90
MRL		 2259 ret = vxlan_igmp_leave(vxlan);
2260 if (ret)
2261 return ret;
7c47cedf 2262 }
d342894c 2263
2264 del_timer_sync(&vxlan->age_timer);
2265
2266 vxlan_flush(vxlan);
56ef9c90 2267 vxlan_sock_release(vs);
d342894c 2268
56ef9c90 2269 return ret;
d342894c 2270}
2271
d342894c 2272/* Stub, nothing needs to be done. */
2273static void vxlan_set_multicast_list(struct net_device *dev)
2274{
2275}
2276
345010b5
DB		 2277static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2278{
2279 struct vxlan_dev *vxlan = netdev_priv(dev);
2280 struct vxlan_rdst *dst = &vxlan->default_dst;
2281 struct net_device *lowerdev;
2282 int max_mtu;
2283
f01ec1c0 2284 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2285 if (lowerdev == NULL)
2286 return eth_change_mtu(dev, new_mtu);
2287
2288 if (dst->remote_ip.sa.sa_family == AF_INET6)
2289 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2290 else
2291 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2292
2293 if (new_mtu < 68 || new_mtu > max_mtu)
2294 return -EINVAL;
2295
2296 dev->mtu = new_mtu;
2297 return 0;
2298}
2299
d342894c 2300static const struct net_device_ops vxlan_netdev_ops = {
2301 .ndo_init = vxlan_init,
ebf4063e 2302 .ndo_uninit = vxlan_uninit,
d342894c 2303 .ndo_open = vxlan_open,
2304 .ndo_stop = vxlan_stop,
2305 .ndo_start_xmit = vxlan_xmit,
e8171045 2306 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2307 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2308 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2309 .ndo_validate_addr = eth_validate_addr,
2310 .ndo_set_mac_address = eth_mac_addr,
2311 .ndo_fdb_add = vxlan_fdb_add,
2312 .ndo_fdb_del = vxlan_fdb_delete,
2313 .ndo_fdb_dump = vxlan_fdb_dump,
2314};
2315
2316/* Info for udev, that this is a virtual tunnel endpoint */
2317static struct device_type vxlan_type = {
2318 .name = "vxlan",
2319};
2320
53cf5275 2321/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2322 * supply the listening VXLAN udp ports. Callers are expected
2323 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2324 */
2325void vxlan_get_rx_port(struct net_device *dev)
2326{
2327 struct vxlan_sock *vs;
2328 struct net *net = dev_net(dev);
2329 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2330 sa_family_t sa_family;
35e42379
JG		 2331 __be16 port;
2332 unsigned int i;
53cf5275
JG		 2333
2334 spin_lock(&vn->sock_lock);
2335 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2336 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2337 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2338 sa_family = vs->sock->sk->sk_family;
2339 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2340 port);
2341 }
2342 }
2343 spin_unlock(&vn->sock_lock);
2344}
2345EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2346
d342894c 2347/* Initialize the device structure. */
2348static void vxlan_setup(struct net_device *dev)
2349{
2350 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2351 unsigned int h;
d342894c 2352
2353 eth_hw_addr_random(dev);
2354 ether_setup(dev);
e4c7ed41 2355 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2356 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2357 else
2853af6a 2358 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2359
2360 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2361 dev->destructor = free_netdev;
d342894c 2362 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2363
2364 dev->tx_queue_len = 0;
2365 dev->features |= NETIF_F_LLTX;
d6727fe3 2366 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2367 dev->features |= NETIF_F_RXCSUM;
05c0db08 2368 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2369
1eaa8178
PS		 2370 dev->vlan_features = dev->features;
2371 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2372 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2373 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2374 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2375 netif_keep_dst(dev);
6602d007 2376 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2377
553675fb 2378 INIT_LIST_HEAD(&vxlan->next);
d342894c 2379 spin_lock_init(&vxlan->hash_lock);
2380
2381 init_timer_deferrable(&vxlan->age_timer);
2382 vxlan->age_timer.function = vxlan_cleanup;
2383 vxlan->age_timer.data = (unsigned long) vxlan;
2384
823aa873 2385 vxlan->dst_port = htons(vxlan_port);
05f47d69 2386
d342894c 2387 vxlan->dev = dev;
2388
2389 for (h = 0; h < FDB_HASH_SIZE; ++h)
2390 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2391}
2392
2393static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2394 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2395 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2396 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2397 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2398 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2399 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2400 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2401 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2402 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2403 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2404 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2405 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2406 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2407 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2408 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2409 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2410 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2411 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2412 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2413 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2414 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2415 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2416 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2417 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2418};
2419
2420static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2421{
2422 if (tb[IFLA_ADDRESS]) {
2423 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2424 pr_debug("invalid link address (not ethernet)\n");
2425 return -EINVAL;
2426 }
2427
2428 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2429 pr_debug("invalid all zero ethernet address\n");
2430 return -EADDRNOTAVAIL;
2431 }
2432 }
2433
2434 if (!data)
2435 return -EINVAL;
2436
2437 if (data[IFLA_VXLAN_ID]) {
2438 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2439 if (id >= VXLAN_VID_MASK)
2440 return -ERANGE;
2441 }
2442
05f47d69 2443 if (data[IFLA_VXLAN_PORT_RANGE]) {
2444 const struct ifla_vxlan_port_range *p
2445 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2446
2447 if (ntohs(p->high) < ntohs(p->low)) {
2448 pr_debug("port range %u .. %u not valid\n",
2449 ntohs(p->low), ntohs(p->high));
2450 return -EINVAL;
2451 }
2452 }
2453
d342894c 2454 return 0;
2455}
2456
1b13c97f
YB		 2457static void vxlan_get_drvinfo(struct net_device *netdev,
2458 struct ethtool_drvinfo *drvinfo)
2459{
2460 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2461 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2462}
2463
2464static const struct ethtool_ops vxlan_ethtool_ops = {
2465 .get_drvinfo = vxlan_get_drvinfo,
2466 .get_link = ethtool_op_get_link,
2467};
2468
553675fb 2469static void vxlan_del_work(struct work_struct *work)
2470{
2471 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2472 udp_tunnel_sock_release(vs->sock);
553675fb 2473 kfree_rcu(vs, rcu);
2474}
2475
3ee64f39
TH		 2476static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2477 __be16 port, u32 flags)
553675fb 2478{
e4c7ed41 2479 struct socket *sock;
3ee64f39
TH		 2480 struct udp_port_cfg udp_conf;
2481 int err;
e4c7ed41 2482
3ee64f39 2483 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2484
3ee64f39
TH		 2485 if (ipv6) {
2486 udp_conf.family = AF_INET6;
3ee64f39 2487 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2488 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2489 } else {
2490 udp_conf.family = AF_INET;
e4c7ed41 2491 }
e4c7ed41 2492
3ee64f39 2493 udp_conf.local_udp_port = port;
553675fb 2494
3ee64f39
TH		 2495 /* Open UDP socket */
2496 err = udp_sock_create(net, &udp_conf, &sock);
2497 if (err < 0)
2498 return ERR_PTR(err);
e4c7ed41 2499
39deb2c7 2500 return sock;
e4c7ed41
CW		 2501}
2502
2503/* Create new listen socket if needed */
2504static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2505 vxlan_rcv_t *rcv, void *data,
2506 u32 flags)
e4c7ed41
CW		 2507{
2508 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2509 struct vxlan_sock *vs;
2510 struct socket *sock;
e4c7ed41 2511 unsigned int h;
359a0ea9 2512 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2513 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2514
dc01e7d3 2515 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2516 if (!vs)
2517 return ERR_PTR(-ENOMEM);
2518
2519 for (h = 0; h < VNI_HASH_SIZE; ++h)
2520 INIT_HLIST_HEAD(&vs->vni_list[h]);
2521
2522 INIT_WORK(&vs->del_work, vxlan_del_work);
2523
3ee64f39 2524 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2525 if (IS_ERR(sock)) {
56ef9c90
MRL		 2526 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2527 PTR_ERR(sock));
553675fb 2528 kfree(vs);
e50fddc8 2529 return ERR_CAST(sock);
553675fb 2530 }
e4c7ed41
CW		 2531
2532 vs->sock = sock;
9c2e24e1 2533 atomic_set(&vs->refcnt, 1);
5cfccc5a 2534 vs->rcv = rcv;
012a5729 2535 vs->data = data;
af33c1ad 2536 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2537
dc01e7d3
OG		 2538 /* Initialize the vxlan udp offloads structure */
2539 vs->udp_offloads.port = port;
2540 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2541 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2542
9c2e24e1
PS		 2543 spin_lock(&vn->sock_lock);
2544 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2545 vxlan_notify_add_rx_port(vs);
9c2e24e1 2546 spin_unlock(&vn->sock_lock);
553675fb 2547
2548 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2549 tunnel_cfg.sk_user_data = vs;
2550 tunnel_cfg.encap_type = 1;
2551 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2552 tunnel_cfg.encap_destroy = NULL;
2553
2554 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2555
9c2e24e1
PS		 2556 return vs;
2557}
2558
012a5729
PS		 2559struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2560 vxlan_rcv_t *rcv, void *data,
359a0ea9 2561 bool no_share, u32 flags)
9c2e24e1
PS		 2562{
2563 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2564 struct vxlan_sock *vs;
19ca9fc1 2565 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2566
56ef9c90
MRL		 2567 if (!no_share) {
2568 spin_lock(&vn->sock_lock);
2569 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port,
2570 flags);
2571 if (vs && vs->rcv == rcv) {
2572 if (!atomic_add_unless(&vs->refcnt, 1, 0))
2573 vs = ERR_PTR(-EBUSY);
2574 spin_unlock(&vn->sock_lock);
2575 return vs;
2576 }
2577 spin_unlock(&vn->sock_lock);
2578 }
9c2e24e1 2579
56ef9c90 2580 return vxlan_socket_create(net, port, rcv, data, flags);
553675fb 2581}
012a5729 2582EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2583
33564bbb 2584static int vxlan_newlink(struct net *src_net, struct net_device *dev,
d342894c 2585 struct nlattr *tb[], struct nlattr *data[])
2586{
33564bbb 2587 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
d342894c 2588 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2589 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2590 __u32 vni;
2591 int err;
e4c7ed41 2592 bool use_ipv6 = false;
d342894c 2593
2594 if (!data[IFLA_VXLAN_ID])
2595 return -EINVAL;
2596
33564bbb 2597 vxlan->net = src_net;
f01ec1c0 2598
d342894c 2599 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2600 dst->remote_vni = vni;
d342894c 2601
5933a7bb
MR		 2602 /* Unless IPv6 is explicitly requested, assume IPv4 */
2603 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2604 if (data[IFLA_VXLAN_GROUP]) {
2605 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2606 } else if (data[IFLA_VXLAN_GROUP6]) {
2607 if (!IS_ENABLED(CONFIG_IPV6))
2608 return -EPFNOSUPPORT;
2609
2610 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2611 sizeof(struct in6_addr));
2612 dst->remote_ip.sa.sa_family = AF_INET6;
2613 use_ipv6 = true;
2614 }
d342894c 2615
e4c7ed41
CW		 2616 if (data[IFLA_VXLAN_LOCAL]) {
2617 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2618 vxlan->saddr.sa.sa_family = AF_INET;
2619 } else if (data[IFLA_VXLAN_LOCAL6]) {
2620 if (!IS_ENABLED(CONFIG_IPV6))
2621 return -EPFNOSUPPORT;
2622
2623 /* TODO: respect scope id */
2624 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2625 sizeof(struct in6_addr));
2626 vxlan->saddr.sa.sa_family = AF_INET6;
2627 use_ipv6 = true;
2628 }
d342894c 2629
34e02aa1 2630 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2631 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2632 struct net_device *lowerdev
33564bbb 2633 = __dev_get_by_index(src_net, dst->remote_ifindex);
34e02aa1 2634
2635 if (!lowerdev) {
c7995c43 2636 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2637 return -ENODEV;
2638 }
d342894c 2639
e4c7ed41
CW		 2640#if IS_ENABLED(CONFIG_IPV6)
2641 if (use_ipv6) {
2642 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2643 if (idev && idev->cnf.disable_ipv6) {
2644 pr_info("IPv6 is disabled via sysctl\n");
2645 return -EPERM;
2646 }
2647 vxlan->flags |= VXLAN_F_IPV6;
2648 }
2649#endif
2650
34e02aa1 2651 if (!tb[IFLA_MTU])
e4c7ed41 2652 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2653
2853af6a 2654 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2655 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2656 } else if (use_ipv6)
2657 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2658
2659 if (data[IFLA_VXLAN_TOS])
2660 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2661
afb97186
VB		 2662 if (data[IFLA_VXLAN_TTL])
2663 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2664
d342894c 2665 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2666 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2667
2668 if (data[IFLA_VXLAN_AGEING])
2669 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2670 else
2671 vxlan->age_interval = FDB_AGE_DEFAULT;
2672
e4f67add
DS		 2673 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2674 vxlan->flags |= VXLAN_F_PROXY;
2675
2676 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2677 vxlan->flags |= VXLAN_F_RSC;
2678
2679 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2680 vxlan->flags |= VXLAN_F_L2MISS;
2681
2682 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2683 vxlan->flags |= VXLAN_F_L3MISS;
2684
d342894c 2685 if (data[IFLA_VXLAN_LIMIT])
2686 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2687
05f47d69 2688 if (data[IFLA_VXLAN_PORT_RANGE]) {
2689 const struct ifla_vxlan_port_range *p
2690 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2691 vxlan->port_min = ntohs(p->low);
2692 vxlan->port_max = ntohs(p->high);
2693 }
2694
823aa873 2695 if (data[IFLA_VXLAN_PORT])
2696 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2697
359a0ea9
TH		 2698 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2699 vxlan->flags |= VXLAN_F_UDP_CSUM;
2700
2701 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2702 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2703 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2704
2705 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2706 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2707 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2708
dfd8645e
TH		 2709 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2710 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2711 vxlan->flags |= VXLAN_F_REMCSUM_TX;
2712
2713 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2714 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2715 vxlan->flags |= VXLAN_F_REMCSUM_RX;
2716
3511494c
TG		 2717 if (data[IFLA_VXLAN_GBP])
2718 vxlan->flags |= VXLAN_F_GBP;
2719
0ace2ca8
TH		 2720 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
2721 vxlan->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
2722
33564bbb 2723 if (vxlan_find_vni(src_net, vni, use_ipv6 ? AF_INET6 : AF_INET,
ac5132d1 2724 vxlan->dst_port, vxlan->flags)) {
553675fb 2725 pr_info("duplicate VNI %u\n", vni);
2726 return -EEXIST;
2727 }
2728
7ad24ea4 2729 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2730
2936b6ab
SS		 2731 /* create an fdb entry for a valid default destination */
2732 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2733 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2734 &vxlan->default_dst.remote_ip,
2735 NUD_REACHABLE|NUD_PERMANENT,
2736 NLM_F_EXCL|NLM_F_CREATE,
2737 vxlan->dst_port,
2738 vxlan->default_dst.remote_vni,
2739 vxlan->default_dst.remote_ifindex,
2740 NTF_SELF);
2741 if (err)
2742 return err;
2743 }
d342894c 2744
afbd8bae
MR		 2745 err = register_netdevice(dev);
2746 if (err) {
ba609e9b 2747 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2748 return err;
2749 }
2750
553675fb 2751 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2752
2753 return 0;
d342894c 2754}
2755
2756static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2757{
2758 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2759 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2760
fe5c3561 2761 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2762 if (!hlist_unhashed(&vxlan->hlist))
2763 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2764 spin_unlock(&vn->sock_lock);
2765
553675fb 2766 list_del(&vxlan->next);
d342894c 2767 unregister_netdevice_queue(dev, head);
2768}
2769
2770static size_t vxlan_get_size(const struct net_device *dev)
2771{
2772
2773 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2774 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2775 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2776 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2777 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2778 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2779 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2780 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2781 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2782 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2783 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2784 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2785 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2786 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2787 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2788 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2789 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2790 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2791 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2792 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2793 0;
2794}
2795
2796static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2797{
2798 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2799 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2800 struct ifla_vxlan_port_range ports = {
2801 .low = htons(vxlan->port_min),
2802 .high = htons(vxlan->port_max),
2803 };
d342894c 2804
c7995c43 2805 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2806 goto nla_put_failure;
2807
e4c7ed41
CW		 2808 if (!vxlan_addr_any(&dst->remote_ip)) {
2809 if (dst->remote_ip.sa.sa_family == AF_INET) {
2810 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2811 dst->remote_ip.sin.sin_addr.s_addr))
2812 goto nla_put_failure;
2813#if IS_ENABLED(CONFIG_IPV6)
2814 } else {
2815 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2816 &dst->remote_ip.sin6.sin6_addr))
2817 goto nla_put_failure;
2818#endif
2819 }
2820 }
d342894c 2821
c7995c43 2822 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2823 goto nla_put_failure;
2824
e4c7ed41
CW		 2825 if (!vxlan_addr_any(&vxlan->saddr)) {
2826 if (vxlan->saddr.sa.sa_family == AF_INET) {
2827 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2828 vxlan->saddr.sin.sin_addr.s_addr))
2829 goto nla_put_failure;
2830#if IS_ENABLED(CONFIG_IPV6)
2831 } else {
2832 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2833 &vxlan->saddr.sin6.sin6_addr))
2834 goto nla_put_failure;
2835#endif
2836 }
2837 }
d342894c 2838
2839 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2840 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2841 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2842 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2843 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2844 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2845 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2846 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2847 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2848 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2849 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2850 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2851 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2852 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2853 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2854 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2855 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2856 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2857 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 2858 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2859 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2860 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2861 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2862 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 2863 goto nla_put_failure;
2864
05f47d69 2865 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2866 goto nla_put_failure;
2867
3511494c
TG		 2868 if (vxlan->flags & VXLAN_F_GBP &&
2869 nla_put_flag(skb, IFLA_VXLAN_GBP))
2870 goto nla_put_failure;
2871
0ace2ca8
TH		 2872 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
2873 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
2874 goto nla_put_failure;
2875
d342894c 2876 return 0;
2877
2878nla_put_failure:
2879 return -EMSGSIZE;
2880}
2881
1728d4fa
ND		 2882static struct net *vxlan_get_link_net(const struct net_device *dev)
2883{
2884 struct vxlan_dev *vxlan = netdev_priv(dev);
2885
2886 return vxlan->net;
2887}
2888
d342894c 2889static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2890 .kind = "vxlan",
2891 .maxtype = IFLA_VXLAN_MAX,
2892 .policy = vxlan_policy,
2893 .priv_size = sizeof(struct vxlan_dev),
2894 .setup = vxlan_setup,
2895 .validate = vxlan_validate,
2896 .newlink = vxlan_newlink,
2897 .dellink = vxlan_dellink,
2898 .get_size = vxlan_get_size,
2899 .fill_info = vxlan_fill_info,
1728d4fa 2900 .get_link_net = vxlan_get_link_net,
d342894c 2901};
2902
acaf4e70
DB		 2903static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2904 struct net_device *dev)
2905{
2906 struct vxlan_dev *vxlan, *next;
2907 LIST_HEAD(list_kill);
2908
2909 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2910 struct vxlan_rdst *dst = &vxlan->default_dst;
2911
2912 /* In case we created vxlan device with carrier
2913 * and we loose the carrier due to module unload
2914 * we also need to remove vxlan device. In other
2915 * cases, it's not necessary and remote_ifindex
2916 * is 0 here, so no matches.
2917 */
2918 if (dst->remote_ifindex == dev->ifindex)
2919 vxlan_dellink(vxlan->dev, &list_kill);
2920 }
2921
2922 unregister_netdevice_many(&list_kill);
2923}
2924
2925static int vxlan_lowerdev_event(struct notifier_block *unused,
2926 unsigned long event, void *ptr)
2927{
2928 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2929 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2930
783c1463 2931 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2932 vxlan_handle_lowerdev_unregister(vn, dev);
2933
2934 return NOTIFY_DONE;
2935}
2936
2937static struct notifier_block vxlan_notifier_block __read_mostly = {
2938 .notifier_call = vxlan_lowerdev_event,
2939};
2940
d342894c 2941static __net_init int vxlan_init_net(struct net *net)
2942{
2943 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2944 unsigned int h;
d342894c 2945
553675fb 2946 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2947 spin_lock_init(&vn->sock_lock);
d342894c 2948
553675fb 2949 for (h = 0; h < PORT_HASH_SIZE; ++h)
2950 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2951
2952 return 0;
2953}
2954
f01ec1c0
ND		 2955static void __net_exit vxlan_exit_net(struct net *net)
2956{
2957 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2958 struct vxlan_dev *vxlan, *next;
2959 struct net_device *dev, *aux;
2960 LIST_HEAD(list);
2961
2962 rtnl_lock();
2963 for_each_netdev_safe(net, dev, aux)
2964 if (dev->rtnl_link_ops == &vxlan_link_ops)
2965 unregister_netdevice_queue(dev, &list);
2966
2967 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2968 /* If vxlan->dev is in the same netns, it has already been added
2969 * to the list by the previous loop.
2970 */
2971 if (!net_eq(dev_net(vxlan->dev), net))
2972 unregister_netdevice_queue(dev, &list);
2973 }
2974
2975 unregister_netdevice_many(&list);
2976 rtnl_unlock();
2977}
2978
d342894c 2979static struct pernet_operations vxlan_net_ops = {
2980 .init = vxlan_init_net,
f01ec1c0 2981 .exit = vxlan_exit_net,
d342894c 2982 .id = &vxlan_net_id,
2983 .size = sizeof(struct vxlan_net),
2984};
2985
2986static int __init vxlan_init_module(void)
2987{
2988 int rc;
2989
758c57d1
SH		 2990 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2991 if (!vxlan_wq)
2992 return -ENOMEM;
2993
d342894c 2994 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2995
783c1463 2996 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 2997 if (rc)
2998 goto out1;
2999
acaf4e70 3000 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3001 if (rc)
3002 goto out2;
3003
acaf4e70
DB		 3004 rc = rtnl_link_register(&vxlan_link_ops);
3005 if (rc)
3006 goto out3;
d342894c 3007
acaf4e70
DB		 3008 return 0;
3009out3:
3010 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3011out2:
783c1463 3012 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3013out1:
758c57d1 3014 destroy_workqueue(vxlan_wq);
d342894c 3015 return rc;
3016}
7332a13b 3017late_initcall(vxlan_init_module);
d342894c 3018
3019static void __exit vxlan_cleanup_module(void)
3020{
b7153984 3021 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3022 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3023 destroy_workqueue(vxlan_wq);
783c1463
DB		 3024 unregister_pernet_subsys(&vxlan_net_ops);
3025 /* rcu_barrier() is called by netns */
d342894c 3026}
3027module_exit(vxlan_cleanup_module);
3028
3029MODULE_LICENSE("GPL");
3030MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3031MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3032MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3033MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)