]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
udp: Add udp_sock_create for UDP tunnels to open listener socket
[mirror_ubuntu-jammy-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
36#include <net/rtnetlink.h>
37#include <net/route.h>
38#include <net/dsfield.h>
39#include <net/inet_ecn.h>
40#include <net/net_namespace.h>
41#include <net/netns/generic.h>
012a5729 42#include <net/vxlan.h>
dc01e7d3 43#include <net/protocol.h>
e4c7ed41
CW		 44#if IS_ENABLED(CONFIG_IPV6)
45#include <net/ipv6.h>
46#include <net/addrconf.h>
47#include <net/ip6_tunnel.h>
660d98ca 48#include <net/ip6_checksum.h>
e4c7ed41 49#endif
d342894c 50
51#define VXLAN_VERSION "0.1"
52
553675fb 53#define PORT_HASH_BITS 8
54#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 55#define VNI_HASH_BITS 10
56#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
57#define FDB_HASH_BITS 8
58#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
59#define FDB_AGE_DEFAULT 300 /* 5 min */
60#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61
62#define VXLAN_N_VID (1u << 24)
63#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
7ce04758 64#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
d342894c 65
66#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
67
68/* VXLAN protocol header */
69struct vxlanhdr {
70 __be32 vx_flags;
71 __be32 vx_vni;
72};
73
23c578bf 74/* UDP port for VXLAN traffic.
75 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 76 * for compatibility with early adopters.
23c578bf 77 */
9daaa397
SH		 78static unsigned short vxlan_port __read_mostly = 8472;
79module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 80MODULE_PARM_DESC(udp_port, "Destination UDP port");
81
82static bool log_ecn_error = true;
83module_param(log_ecn_error, bool, 0644);
84MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
85
60d9d4c6 86static int vxlan_net_id;
553675fb 87
afbd8bae
MR		 88static const u8 all_zeros_mac[ETH_ALEN];
89
553675fb 90/* per-network namespace private data for this module */
91struct vxlan_net {
92 struct list_head vxlan_list;
93 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 94 spinlock_t sock_lock;
553675fb 95};
96
e4c7ed41
CW		 97union vxlan_addr {
98 struct sockaddr_in sin;
99 struct sockaddr_in6 sin6;
100 struct sockaddr sa;
101};
102
6681712d 103struct vxlan_rdst {
e4c7ed41 104 union vxlan_addr remote_ip;
6681712d
DS		 105 __be16 remote_port;
106 u32 remote_vni;
107 u32 remote_ifindex;
3e61aa8f 108 struct list_head list;
bc7892ba 109 struct rcu_head rcu;
6681712d
DS		 110};
111
d342894c 112/* Forwarding table entry */
113struct vxlan_fdb {
114 struct hlist_node hlist; /* linked list of entries */
115 struct rcu_head rcu;
116 unsigned long updated; /* jiffies */
117 unsigned long used;
3e61aa8f 118 struct list_head remotes;
d342894c 119 u16 state; /* see ndm_state */
ae884082 120 u8 flags; /* see ndm_flags */
d342894c 121 u8 eth_addr[ETH_ALEN];
122};
123
d342894c 124/* Pseudo network device */
125struct vxlan_dev {
553675fb 126 struct hlist_node hlist; /* vni hash table */
127 struct list_head next; /* vxlan's per namespace list */
128 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 129 struct net_device *dev;
f01ec1c0 130 struct net *net; /* netns for packet i/o */
c7995c43 131 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 132 union vxlan_addr saddr; /* source address */
823aa873 133 __be16 dst_port;
05f47d69 134 __u16 port_min; /* source port range */
135 __u16 port_max;
d342894c 136 __u8 tos; /* TOS override */
137 __u8 ttl;
359a0ea9 138 u32 flags; /* VXLAN_F_* in vxlan.h */
d342894c 139
1c51a915 140 struct work_struct sock_work;
3fc2de2f 141 struct work_struct igmp_join;
142 struct work_struct igmp_leave;
1c51a915 143
d342894c 144 unsigned long age_interval;
145 struct timer_list age_timer;
146 spinlock_t hash_lock;
147 unsigned int addrcnt;
148 unsigned int addrmax;
d342894c 149
150 struct hlist_head fdb_head[FDB_HASH_SIZE];
151};
152
153/* salt for hash table */
154static u32 vxlan_salt __read_mostly;
758c57d1 155static struct workqueue_struct *vxlan_wq;
d342894c 156
1c51a915
SH		 157static void vxlan_sock_work(struct work_struct *work);
158
e4c7ed41
CW		 159#if IS_ENABLED(CONFIG_IPV6)
160static inline
161bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
162{
163 if (a->sa.sa_family != b->sa.sa_family)
164 return false;
165 if (a->sa.sa_family == AF_INET6)
166 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
167 else
168 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
169}
170
171static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
172{
173 if (ipa->sa.sa_family == AF_INET6)
174 return ipv6_addr_any(&ipa->sin6.sin6_addr);
175 else
176 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
177}
178
179static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
180{
181 if (ipa->sa.sa_family == AF_INET6)
182 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
183 else
184 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
185}
186
187static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
188{
189 if (nla_len(nla) >= sizeof(struct in6_addr)) {
190 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
191 ip->sa.sa_family = AF_INET6;
192 return 0;
193 } else if (nla_len(nla) >= sizeof(__be32)) {
194 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
195 ip->sa.sa_family = AF_INET;
196 return 0;
197 } else {
198 return -EAFNOSUPPORT;
199 }
200}
201
202static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
203 const union vxlan_addr *ip)
204{
205 if (ip->sa.sa_family == AF_INET6)
206 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
207 else
208 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
209}
210
211#else /* !CONFIG_IPV6 */
212
213static inline
214bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
215{
216 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
217}
218
219static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
220{
221 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
222}
223
224static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
225{
226 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
227}
228
229static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
230{
231 if (nla_len(nla) >= sizeof(struct in6_addr)) {
232 return -EAFNOSUPPORT;
233 } else if (nla_len(nla) >= sizeof(__be32)) {
234 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
235 ip->sa.sa_family = AF_INET;
236 return 0;
237 } else {
238 return -EAFNOSUPPORT;
239 }
240}
241
242static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
243 const union vxlan_addr *ip)
244{
245 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
246}
247#endif
248
553675fb 249/* Virtual Network hash table head */
250static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
251{
252 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
253}
254
255/* Socket hash table head */
256static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 257{
258 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
259
553675fb 260 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
261}
262
3e61aa8f
SH		 263/* First remote destination for a forwarding entry.
264 * Guaranteed to be non-NULL because remotes are never deleted.
265 */
5ca5461c 266static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 267{
5ca5461c 268 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
269}
270
271static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
272{
273 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 274}
275
553675fb 276/* Find VXLAN socket based on network namespace and UDP port */
9c2e24e1 277static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
553675fb 278{
279 struct vxlan_sock *vs;
280
281 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
282 if (inet_sk(vs->sock->sk)->inet_sport == port)
283 return vs;
284 }
285 return NULL;
d342894c 286}
287
5cfccc5a 288static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 289{
290 struct vxlan_dev *vxlan;
d342894c 291
553675fb 292 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 293 if (vxlan->default_dst.remote_vni == id)
d342894c 294 return vxlan;
295 }
296
297 return NULL;
298}
299
5cfccc5a
PS		 300/* Look up VNI in a per net namespace table */
301static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
302{
303 struct vxlan_sock *vs;
304
305 vs = vxlan_find_sock(net, port);
306 if (!vs)
307 return NULL;
308
309 return vxlan_vs_find_vni(vs, id);
310}
311
d342894c 312/* Fill in neighbour message in skbuff. */
313static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 314 const struct vxlan_fdb *fdb,
315 u32 portid, u32 seq, int type, unsigned int flags,
316 const struct vxlan_rdst *rdst)
d342894c 317{
318 unsigned long now = jiffies;
319 struct nda_cacheinfo ci;
320 struct nlmsghdr *nlh;
321 struct ndmsg *ndm;
e4f67add 322 bool send_ip, send_eth;
d342894c 323
324 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
325 if (nlh == NULL)
326 return -EMSGSIZE;
327
328 ndm = nlmsg_data(nlh);
329 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 330
331 send_eth = send_ip = true;
332
333 if (type == RTM_GETNEIGH) {
334 ndm->ndm_family = AF_INET;
e4c7ed41 335 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 336 send_eth = !is_zero_ether_addr(fdb->eth_addr);
337 } else
338 ndm->ndm_family = AF_BRIDGE;
d342894c 339 ndm->ndm_state = fdb->state;
340 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 341 ndm->ndm_flags = fdb->flags;
d342894c 342 ndm->ndm_type = NDA_DST;
343
e4f67add 344 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 345 goto nla_put_failure;
346
e4c7ed41 347 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 348 goto nla_put_failure;
349
823aa873 350 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 351 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
352 goto nla_put_failure;
c7995c43 353 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 354 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 355 goto nla_put_failure;
356 if (rdst->remote_ifindex &&
357 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 358 goto nla_put_failure;
359
360 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
361 ci.ndm_confirmed = 0;
362 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
363 ci.ndm_refcnt = 0;
364
365 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
366 goto nla_put_failure;
367
368 return nlmsg_end(skb, nlh);
369
370nla_put_failure:
371 nlmsg_cancel(skb, nlh);
372 return -EMSGSIZE;
373}
374
375static inline size_t vxlan_nlmsg_size(void)
376{
377 return NLMSG_ALIGN(sizeof(struct ndmsg))
378 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 379 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 380 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 381 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
382 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
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
dc01e7d3
OG		 549static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb)
550{
551 struct sk_buff *p, **pp = NULL;
552 struct vxlanhdr *vh, *vh2;
553 struct ethhdr *eh, *eh2;
554 unsigned int hlen, off_vx, off_eth;
555 const struct packet_offload *ptype;
556 __be16 type;
557 int flush = 1;
558
559 off_vx = skb_gro_offset(skb);
560 hlen = off_vx + sizeof(*vh);
561 vh = skb_gro_header_fast(skb, off_vx);
562 if (skb_gro_header_hard(skb, hlen)) {
563 vh = skb_gro_header_slow(skb, hlen, off_vx);
564 if (unlikely(!vh))
565 goto out;
566 }
567 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
6bae1d4c 568 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
dc01e7d3
OG		 569
570 off_eth = skb_gro_offset(skb);
571 hlen = off_eth + sizeof(*eh);
572 eh = skb_gro_header_fast(skb, off_eth);
573 if (skb_gro_header_hard(skb, hlen)) {
574 eh = skb_gro_header_slow(skb, hlen, off_eth);
575 if (unlikely(!eh))
576 goto out;
577 }
578
579 flush = 0;
580
581 for (p = *head; p; p = p->next) {
582 if (!NAPI_GRO_CB(p)->same_flow)
583 continue;
584
585 vh2 = (struct vxlanhdr *)(p->data + off_vx);
586 eh2 = (struct ethhdr *)(p->data + off_eth);
587 if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) {
588 NAPI_GRO_CB(p)->same_flow = 0;
589 continue;
590 }
dc01e7d3
OG		 591 }
592
dc01e7d3
OG		 593 type = eh->h_proto;
594
595 rcu_read_lock();
596 ptype = gro_find_receive_by_type(type);
597 if (ptype == NULL) {
598 flush = 1;
599 goto out_unlock;
600 }
601
602 skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */
6bae1d4c 603 skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
dc01e7d3
OG		 604 pp = ptype->callbacks.gro_receive(head, skb);
605
606out_unlock:
607 rcu_read_unlock();
608out:
609 NAPI_GRO_CB(skb)->flush |= flush;
610
611 return pp;
612}
613
614static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
615{
616 struct ethhdr *eh;
617 struct packet_offload *ptype;
618 __be16 type;
619 int vxlan_len = sizeof(struct vxlanhdr) + sizeof(struct ethhdr);
620 int err = -ENOSYS;
621
622 eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr));
623 type = eh->h_proto;
624
625 rcu_read_lock();
626 ptype = gro_find_complete_by_type(type);
627 if (ptype != NULL)
628 err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len);
629
630 rcu_read_unlock();
631 return err;
632}
633
53cf5275 634/* Notify netdevs that UDP port started listening */
dc01e7d3 635static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 636{
637 struct net_device *dev;
dc01e7d3 638 struct sock *sk = vs->sock->sk;
53cf5275
JG		 639 struct net *net = sock_net(sk);
640 sa_family_t sa_family = sk->sk_family;
35e42379 641 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 642 int err;
643
644 if (sa_family == AF_INET) {
645 err = udp_add_offload(&vs->udp_offloads);
646 if (err)
647 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
648 }
53cf5275
JG		 649
650 rcu_read_lock();
651 for_each_netdev_rcu(net, dev) {
652 if (dev->netdev_ops->ndo_add_vxlan_port)
653 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
654 port);
655 }
656 rcu_read_unlock();
657}
658
659/* Notify netdevs that UDP port is no more listening */
dc01e7d3 660static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 661{
662 struct net_device *dev;
dc01e7d3 663 struct sock *sk = vs->sock->sk;
53cf5275
JG		 664 struct net *net = sock_net(sk);
665 sa_family_t sa_family = sk->sk_family;
35e42379 666 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 667
668 rcu_read_lock();
669 for_each_netdev_rcu(net, dev) {
670 if (dev->netdev_ops->ndo_del_vxlan_port)
671 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
672 port);
673 }
674 rcu_read_unlock();
dc01e7d3
OG		 675
676 if (sa_family == AF_INET)
677 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 678}
679
d342894c 680/* Add new entry to forwarding table -- assumes lock held */
681static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 682 const u8 *mac, union vxlan_addr *ip,
6681712d 683 __u16 state, __u16 flags,
73cf3317 684 __be16 port, __u32 vni, __u32 ifindex,
ae884082 685 __u8 ndm_flags)
d342894c 686{
9e4b93f9 687 struct vxlan_rdst *rd = NULL;
d342894c 688 struct vxlan_fdb *f;
689 int notify = 0;
690
014be2c8 691 f = __vxlan_find_mac(vxlan, mac);
d342894c 692 if (f) {
693 if (flags & NLM_F_EXCL) {
694 netdev_dbg(vxlan->dev,
695 "lost race to create %pM\n", mac);
696 return -EEXIST;
697 }
698 if (f->state != state) {
699 f->state = state;
700 f->updated = jiffies;
701 notify = 1;
702 }
ae884082
DS		 703 if (f->flags != ndm_flags) {
704 f->flags = ndm_flags;
705 f->updated = jiffies;
706 notify = 1;
707 }
906dc186
TR		 708 if ((flags & NLM_F_REPLACE)) {
709 /* Only change unicasts */
710 if (!(is_multicast_ether_addr(f->eth_addr) ||
711 is_zero_ether_addr(f->eth_addr))) {
712 int rc = vxlan_fdb_replace(f, ip, port, vni,
713 ifindex);
714
715 if (rc < 0)
716 return rc;
717 notify |= rc;
718 } else
719 return -EOPNOTSUPP;
720 }
6681712d 721 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 722 (is_multicast_ether_addr(f->eth_addr) ||
723 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 724 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
725 &rd);
6681712d
DS		 726
727 if (rc < 0)
728 return rc;
729 notify |= rc;
730 }
d342894c 731 } else {
732 if (!(flags & NLM_F_CREATE))
733 return -ENOENT;
734
735 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
736 return -ENOSPC;
737
906dc186
TR		 738 /* Disallow replace to add a multicast entry */
739 if ((flags & NLM_F_REPLACE) &&
740 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
741 return -EOPNOTSUPP;
742
e4c7ed41 743 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 744 f = kmalloc(sizeof(*f), GFP_ATOMIC);
745 if (!f)
746 return -ENOMEM;
747
748 notify = 1;
d342894c 749 f->state = state;
ae884082 750 f->flags = ndm_flags;
d342894c 751 f->updated = f->used = jiffies;
3e61aa8f 752 INIT_LIST_HEAD(&f->remotes);
d342894c 753 memcpy(f->eth_addr, mac, ETH_ALEN);
754
9e4b93f9 755 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 756
d342894c 757 ++vxlan->addrcnt;
758 hlist_add_head_rcu(&f->hlist,
759 vxlan_fdb_head(vxlan, mac));
760 }
761
9e4b93f9
ND		 762 if (notify) {
763 if (rd == NULL)
764 rd = first_remote_rtnl(f);
765 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
766 }
d342894c 767
768 return 0;
769}
770
6706c82e 771static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 772{
773 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 774 struct vxlan_rdst *rd, *nd;
6681712d 775
3e61aa8f 776 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 777 kfree(rd);
6681712d
DS		 778 kfree(f);
779}
780
d342894c 781static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
782{
783 netdev_dbg(vxlan->dev,
784 "delete %pM\n", f->eth_addr);
785
786 --vxlan->addrcnt;
9e4b93f9 787 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 788
789 hlist_del_rcu(&f->hlist);
6681712d 790 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 791}
792
f0b074be 793static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 794 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 795{
6681712d 796 struct net *net = dev_net(vxlan->dev);
e4c7ed41 797 int err;
d342894c 798
f0b074be 799 if (tb[NDA_DST]) {
e4c7ed41
CW		 800 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
801 if (err)
802 return err;
f0b074be 803 } else {
e4c7ed41
CW		 804 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
805 if (remote->sa.sa_family == AF_INET) {
806 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
807 ip->sa.sa_family = AF_INET;
808#if IS_ENABLED(CONFIG_IPV6)
809 } else {
810 ip->sin6.sin6_addr = in6addr_any;
811 ip->sa.sa_family = AF_INET6;
812#endif
813 }
f0b074be 814 }
d342894c 815
6681712d 816 if (tb[NDA_PORT]) {
73cf3317 817 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 818 return -EINVAL;
f0b074be
MR		 819 *port = nla_get_be16(tb[NDA_PORT]);
820 } else {
821 *port = vxlan->dst_port;
822 }
6681712d
DS		 823
824 if (tb[NDA_VNI]) {
825 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
826 return -EINVAL;
f0b074be
MR		 827 *vni = nla_get_u32(tb[NDA_VNI]);
828 } else {
829 *vni = vxlan->default_dst.remote_vni;
830 }
6681712d
DS		 831
832 if (tb[NDA_IFINDEX]) {
5abb0029 833 struct net_device *tdev;
6681712d
DS		 834
835 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
836 return -EINVAL;
f0b074be 837 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 838 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 839 if (!tdev)
6681712d 840 return -EADDRNOTAVAIL;
f0b074be
MR		 841 } else {
842 *ifindex = 0;
843 }
844
845 return 0;
846}
847
848/* Add static entry (via netlink) */
849static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
850 struct net_device *dev,
851 const unsigned char *addr, u16 flags)
852{
853 struct vxlan_dev *vxlan = netdev_priv(dev);
854 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 855 union vxlan_addr ip;
f0b074be
MR		 856 __be16 port;
857 u32 vni, ifindex;
858 int err;
859
860 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
861 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
862 ndm->ndm_state);
863 return -EINVAL;
864 }
865
866 if (tb[NDA_DST] == NULL)
867 return -EINVAL;
868
869 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
870 if (err)
871 return err;
6681712d 872
5933a7bb
MR		 873 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
874 return -EAFNOSUPPORT;
875
d342894c 876 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 877 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 878 port, vni, ifindex, ndm->ndm_flags);
d342894c 879 spin_unlock_bh(&vxlan->hash_lock);
880
881 return err;
882}
883
884/* Delete entry (via netlink) */
1690be63
VY		 885static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
886 struct net_device *dev,
d342894c 887 const unsigned char *addr)
888{
889 struct vxlan_dev *vxlan = netdev_priv(dev);
890 struct vxlan_fdb *f;
bc7892ba 891 struct vxlan_rdst *rd = NULL;
e4c7ed41 892 union vxlan_addr ip;
bc7892ba
MR		 893 __be16 port;
894 u32 vni, ifindex;
895 int err;
896
897 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
898 if (err)
899 return err;
900
901 err = -ENOENT;
d342894c 902
903 spin_lock_bh(&vxlan->hash_lock);
904 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 905 if (!f)
906 goto out;
907
e4c7ed41
CW		 908 if (!vxlan_addr_any(&ip)) {
909 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 910 if (!rd)
911 goto out;
912 }
913
914 err = 0;
915
916 /* remove a destination if it's not the only one on the list,
917 * otherwise destroy the fdb entry
918 */
919 if (rd && !list_is_singular(&f->remotes)) {
920 list_del_rcu(&rd->list);
9e4b93f9 921 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 922 kfree_rcu(rd, rcu);
bc7892ba 923 goto out;
d342894c 924 }
bc7892ba
MR		 925
926 vxlan_fdb_destroy(vxlan, f);
927
928out:
d342894c 929 spin_unlock_bh(&vxlan->hash_lock);
930
931 return err;
932}
933
934/* Dump forwarding table */
935static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 936 struct net_device *dev,
937 struct net_device *filter_dev, int idx)
d342894c 938{
939 struct vxlan_dev *vxlan = netdev_priv(dev);
940 unsigned int h;
941
942 for (h = 0; h < FDB_HASH_SIZE; ++h) {
943 struct vxlan_fdb *f;
d342894c 944 int err;
945
b67bfe0d 946 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 947 struct vxlan_rdst *rd;
6681712d 948
3e61aa8f
SH		 949 if (idx < cb->args[0])
950 goto skip;
951
952 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 953 err = vxlan_fdb_info(skb, vxlan, f,
954 NETLINK_CB(cb->skb).portid,
955 cb->nlh->nlmsg_seq,
956 RTM_NEWNEIGH,
957 NLM_F_MULTI, rd);
958 if (err < 0)
3e61aa8f 959 goto out;
6681712d 960 }
3e61aa8f
SH		 961skip:
962 ++idx;
d342894c 963 }
964 }
3e61aa8f 965out:
d342894c 966 return idx;
967}
968
969/* Watch incoming packets to learn mapping between Ethernet address
970 * and Tunnel endpoint.
26a41ae6 971 * Return true if packet is bogus and should be droppped.
d342894c 972 */
26a41ae6 973static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 974 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 975{
976 struct vxlan_dev *vxlan = netdev_priv(dev);
977 struct vxlan_fdb *f;
d342894c 978
979 f = vxlan_find_mac(vxlan, src_mac);
980 if (likely(f)) {
5ca5461c 981 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 982
e4c7ed41 983 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 984 return false;
985
986 /* Don't migrate static entries, drop packets */
eb064c3b 987 if (f->state & NUD_NOARP)
26a41ae6 988 return true;
d342894c 989
990 if (net_ratelimit())
991 netdev_info(dev,
e4c7ed41 992 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 993 src_mac, &rdst->remote_ip, &src_ip);
d342894c 994
e4c7ed41 995 rdst->remote_ip = *src_ip;
d342894c 996 f->updated = jiffies;
9e4b93f9 997 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 998 } else {
999 /* learned new entry */
1000 spin_lock(&vxlan->hash_lock);
3bf74b1a 1001
1002 /* close off race between vxlan_flush and incoming packets */
1003 if (netif_running(dev))
1004 vxlan_fdb_create(vxlan, src_mac, src_ip,
1005 NUD_REACHABLE,
1006 NLM_F_EXCL|NLM_F_CREATE,
1007 vxlan->dst_port,
1008 vxlan->default_dst.remote_vni,
1009 0, NTF_SELF);
d342894c 1010 spin_unlock(&vxlan->hash_lock);
1011 }
26a41ae6 1012
1013 return false;
d342894c 1014}
1015
d342894c 1016/* See if multicast group is already in use by other ID */
95ab0991 1017static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1018{
553675fb 1019 struct vxlan_dev *vxlan;
d342894c 1020
95ab0991
G		 1021 /* The vxlan_sock is only used by dev, leaving group has
1022 * no effect on other vxlan devices.
1023 */
1024 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1025 return false;
1026
553675fb 1027 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1028 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1029 continue;
d342894c 1030
95ab0991
G		 1031 if (vxlan->vn_sock != dev->vn_sock)
1032 continue;
1033
1034 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1035 &dev->default_dst.remote_ip))
1036 continue;
1037
1038 if (vxlan->default_dst.remote_ifindex !=
1039 dev->default_dst.remote_ifindex)
1040 continue;
1041
1042 return true;
553675fb 1043 }
d342894c 1044
1045 return false;
1046}
1047
7c47cedf 1048static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1049{
7c47cedf
SH		 1050 atomic_inc(&vs->refcnt);
1051}
d342894c 1052
012a5729 1053void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1054{
53cf5275
JG		 1055 struct sock *sk = vs->sock->sk;
1056 struct net *net = sock_net(sk);
1057 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1058
7c47cedf
SH		 1059 if (!atomic_dec_and_test(&vs->refcnt))
1060 return;
d342894c 1061
1c51a915 1062 spin_lock(&vn->sock_lock);
7c47cedf 1063 hlist_del_rcu(&vs->hlist);
559835ea 1064 rcu_assign_sk_user_data(vs->sock->sk, NULL);
dc01e7d3 1065 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1066 spin_unlock(&vn->sock_lock);
1067
7c47cedf 1068 queue_work(vxlan_wq, &vs->del_work);
d342894c 1069}
012a5729 1070EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1071
3fc2de2f 1072/* Callback to update multicast group membership when first VNI on
1073 * multicast asddress is brought up
1074 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1075 */
3fc2de2f 1076static void vxlan_igmp_join(struct work_struct *work)
d342894c 1077{
3fc2de2f 1078 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1079 struct vxlan_sock *vs = vxlan->vn_sock;
1080 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1081 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1082 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1083
d342894c 1084 lock_sock(sk);
e4c7ed41
CW		 1085 if (ip->sa.sa_family == AF_INET) {
1086 struct ip_mreqn mreq = {
1087 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1088 .imr_ifindex = ifindex,
1089 };
1090
1091 ip_mc_join_group(sk, &mreq);
1092#if IS_ENABLED(CONFIG_IPV6)
1093 } else {
1094 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1095 &ip->sin6.sin6_addr);
1096#endif
1097 }
3fc2de2f 1098 release_sock(sk);
1099
012a5729 1100 vxlan_sock_release(vs);
3fc2de2f 1101 dev_put(vxlan->dev);
1102}
1103
1104/* Inverse of vxlan_igmp_join when last VNI is brought down */
1105static void vxlan_igmp_leave(struct work_struct *work)
1106{
1107 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1108 struct vxlan_sock *vs = vxlan->vn_sock;
1109 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1110 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1111 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1112
1113 lock_sock(sk);
e4c7ed41
CW		 1114 if (ip->sa.sa_family == AF_INET) {
1115 struct ip_mreqn mreq = {
1116 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1117 .imr_ifindex = ifindex,
1118 };
1119
1120 ip_mc_leave_group(sk, &mreq);
1121#if IS_ENABLED(CONFIG_IPV6)
1122 } else {
1123 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1124 &ip->sin6.sin6_addr);
1125#endif
1126 }
1127
d342894c 1128 release_sock(sk);
d342894c 1129
012a5729 1130 vxlan_sock_release(vs);
7c47cedf 1131 dev_put(vxlan->dev);
d342894c 1132}
1133
1134/* Callback from net/ipv4/udp.c to receive packets */
1135static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1136{
5cfccc5a 1137 struct vxlan_sock *vs;
d342894c 1138 struct vxlanhdr *vxh;
d342894c 1139
d342894c 1140 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1141 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1142 goto error;
1143
7ce04758
PS		 1144 /* Return packets with reserved bits set */
1145 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 1146 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1147 (vxh->vx_vni & htonl(0xff))) {
1148 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1149 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1150 goto error;
1151 }
1152
5cfccc5a
PS		 1153 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1154 goto drop;
1155
559835ea 1156 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1157 if (!vs)
d342894c 1158 goto drop;
d342894c 1159
f79b064c 1160 skb_pop_rcv_encapsulation(skb);
d0bc6555 1161
5cfccc5a
PS		 1162 vs->rcv(vs, skb, vxh->vx_vni);
1163 return 0;
1164
1165drop:
1166 /* Consume bad packet */
1167 kfree_skb(skb);
1168 return 0;
1169
1170error:
1171 /* Return non vxlan pkt */
1172 return 1;
1173}
1174
1175static void vxlan_rcv(struct vxlan_sock *vs,
1176 struct sk_buff *skb, __be32 vx_vni)
1177{
e4c7ed41
CW		 1178 struct iphdr *oip = NULL;
1179 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1180 struct vxlan_dev *vxlan;
8f84985f 1181 struct pcpu_sw_netstats *stats;
e4c7ed41 1182 union vxlan_addr saddr;
5cfccc5a 1183 __u32 vni;
e4c7ed41
CW		 1184 int err = 0;
1185 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1186
1187 vni = ntohl(vx_vni) >> 8;
1188 /* Is this VNI defined? */
1189 vxlan = vxlan_vs_find_vni(vs, vni);
1190 if (!vxlan)
d342894c 1191 goto drop;
d342894c 1192
e4c7ed41 1193 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1194 skb_reset_mac_header(skb);
f01ec1c0 1195 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1196 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1197 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1198
1199 /* Ignore packet loops (and multicast echo) */
7367d0b5 1200 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1201 goto drop;
1202
7ce04758 1203 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1204 if (remote_ip->sa.sa_family == AF_INET) {
1205 oip = ip_hdr(skb);
1206 saddr.sin.sin_addr.s_addr = oip->saddr;
1207 saddr.sa.sa_family = AF_INET;
1208#if IS_ENABLED(CONFIG_IPV6)
1209 } else {
1210 oip6 = ipv6_hdr(skb);
1211 saddr.sin6.sin6_addr = oip6->saddr;
1212 saddr.sa.sa_family = AF_INET6;
1213#endif
1214 }
1215
26a41ae6 1216 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1217 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1218 goto drop;
d342894c 1219
d342894c 1220 skb_reset_network_header(skb);
0afb1666 1221
e4c7ed41
CW		 1222 if (oip6)
1223 err = IP6_ECN_decapsulate(oip6, skb);
1224 if (oip)
1225 err = IP_ECN_decapsulate(oip, skb);
1226
d342894c 1227 if (unlikely(err)) {
e4c7ed41
CW		 1228 if (log_ecn_error) {
1229 if (oip6)
1230 net_info_ratelimited("non-ECT from %pI6\n",
1231 &oip6->saddr);
1232 if (oip)
1233 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1234 &oip->saddr, oip->tos);
1235 }
d342894c 1236 if (err > 1) {
1237 ++vxlan->dev->stats.rx_frame_errors;
1238 ++vxlan->dev->stats.rx_errors;
1239 goto drop;
1240 }
1241 }
1242
e8171045 1243 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1244 u64_stats_update_begin(&stats->syncp);
1245 stats->rx_packets++;
1246 stats->rx_bytes += skb->len;
1247 u64_stats_update_end(&stats->syncp);
1248
1249 netif_rx(skb);
1250
5cfccc5a 1251 return;
d342894c 1252drop:
1253 /* Consume bad packet */
1254 kfree_skb(skb);
d342894c 1255}
1256
e4f67add
DS		 1257static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1258{
1259 struct vxlan_dev *vxlan = netdev_priv(dev);
1260 struct arphdr *parp;
1261 u8 *arpptr, *sha;
1262 __be32 sip, tip;
1263 struct neighbour *n;
1264
1265 if (dev->flags & IFF_NOARP)
1266 goto out;
1267
1268 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1269 dev->stats.tx_dropped++;
1270 goto out;
1271 }
1272 parp = arp_hdr(skb);
1273
1274 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1275 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1276 parp->ar_pro != htons(ETH_P_IP) ||
1277 parp->ar_op != htons(ARPOP_REQUEST) ||
1278 parp->ar_hln != dev->addr_len ||
1279 parp->ar_pln != 4)
1280 goto out;
1281 arpptr = (u8 *)parp + sizeof(struct arphdr);
1282 sha = arpptr;
1283 arpptr += dev->addr_len; /* sha */
1284 memcpy(&sip, arpptr, sizeof(sip));
1285 arpptr += sizeof(sip);
1286 arpptr += dev->addr_len; /* tha */
1287 memcpy(&tip, arpptr, sizeof(tip));
1288
1289 if (ipv4_is_loopback(tip) ||
1290 ipv4_is_multicast(tip))
1291 goto out;
1292
1293 n = neigh_lookup(&arp_tbl, &tip, dev);
1294
1295 if (n) {
e4f67add
DS		 1296 struct vxlan_fdb *f;
1297 struct sk_buff *reply;
1298
1299 if (!(n->nud_state & NUD_CONNECTED)) {
1300 neigh_release(n);
1301 goto out;
1302 }
1303
1304 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1305 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1306 /* bridge-local neighbor */
1307 neigh_release(n);
1308 goto out;
1309 }
1310
1311 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1312 n->ha, sha);
1313
1314 neigh_release(n);
1315
7346135d
DS		 1316 if (reply == NULL)
1317 goto out;
1318
e4f67add
DS		 1319 skb_reset_mac_header(reply);
1320 __skb_pull(reply, skb_network_offset(reply));
1321 reply->ip_summed = CHECKSUM_UNNECESSARY;
1322 reply->pkt_type = PACKET_HOST;
1323
1324 if (netif_rx_ni(reply) == NET_RX_DROP)
1325 dev->stats.rx_dropped++;
e4c7ed41
CW		 1326 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1327 union vxlan_addr ipa = {
1328 .sin.sin_addr.s_addr = tip,
1329 .sa.sa_family = AF_INET,
1330 };
1331
1332 vxlan_ip_miss(dev, &ipa);
1333 }
e4f67add
DS		 1334out:
1335 consume_skb(skb);
1336 return NETDEV_TX_OK;
1337}
1338
f564f45c 1339#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1340
1341static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1342 struct neighbour *n, bool isrouter)
1343{
1344 struct net_device *dev = request->dev;
1345 struct sk_buff *reply;
1346 struct nd_msg *ns, *na;
1347 struct ipv6hdr *pip6;
1348 u8 *daddr;
1349 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1350 int ns_olen;
1351 int i, len;
1352
1353 if (dev == NULL)
1354 return NULL;
1355
1356 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1357 sizeof(*na) + na_olen + dev->needed_tailroom;
1358 reply = alloc_skb(len, GFP_ATOMIC);
1359 if (reply == NULL)
1360 return NULL;
1361
1362 reply->protocol = htons(ETH_P_IPV6);
1363 reply->dev = dev;
1364 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1365 skb_push(reply, sizeof(struct ethhdr));
1366 skb_set_mac_header(reply, 0);
1367
1368 ns = (struct nd_msg *)skb_transport_header(request);
1369
1370 daddr = eth_hdr(request)->h_source;
1371 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1372 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1373 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1374 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1375 break;
1376 }
1377 }
1378
1379 /* Ethernet header */
1380 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1381 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1382 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1383 reply->protocol = htons(ETH_P_IPV6);
1384
1385 skb_pull(reply, sizeof(struct ethhdr));
1386 skb_set_network_header(reply, 0);
1387 skb_put(reply, sizeof(struct ipv6hdr));
1388
1389 /* IPv6 header */
1390
1391 pip6 = ipv6_hdr(reply);
1392 memset(pip6, 0, sizeof(struct ipv6hdr));
1393 pip6->version = 6;
1394 pip6->priority = ipv6_hdr(request)->priority;
1395 pip6->nexthdr = IPPROTO_ICMPV6;
1396 pip6->hop_limit = 255;
1397 pip6->daddr = ipv6_hdr(request)->saddr;
1398 pip6->saddr = *(struct in6_addr *)n->primary_key;
1399
1400 skb_pull(reply, sizeof(struct ipv6hdr));
1401 skb_set_transport_header(reply, 0);
1402
1403 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1404
1405 /* Neighbor Advertisement */
1406 memset(na, 0, sizeof(*na)+na_olen);
1407 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1408 na->icmph.icmp6_router = isrouter;
1409 na->icmph.icmp6_override = 1;
1410 na->icmph.icmp6_solicited = 1;
1411 na->target = ns->target;
1412 ether_addr_copy(&na->opt[2], n->ha);
1413 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1414 na->opt[1] = na_olen >> 3;
1415
1416 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1417 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1418 csum_partial(na, sizeof(*na)+na_olen, 0));
1419
1420 pip6->payload_len = htons(sizeof(*na)+na_olen);
1421
1422 skb_push(reply, sizeof(struct ipv6hdr));
1423
1424 reply->ip_summed = CHECKSUM_UNNECESSARY;
1425
1426 return reply;
1427}
1428
f564f45c
CW		 1429static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1430{
1431 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1432 struct nd_msg *msg;
f564f45c
CW		 1433 const struct ipv6hdr *iphdr;
1434 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1435 struct neighbour *n;
1436 struct inet6_dev *in6_dev;
f564f45c
CW		 1437
1438 in6_dev = __in6_dev_get(dev);
1439 if (!in6_dev)
1440 goto out;
1441
1442 if (!pskb_may_pull(skb, skb->len))
1443 goto out;
1444
1445 iphdr = ipv6_hdr(skb);
1446 saddr = &iphdr->saddr;
1447 daddr = &iphdr->daddr;
1448
f564f45c
CW		 1449 msg = (struct nd_msg *)skb_transport_header(skb);
1450 if (msg->icmph.icmp6_code != 0 ||
1451 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1452 goto out;
1453
4b29dba9
DS		 1454 if (ipv6_addr_loopback(daddr) ||
1455 ipv6_addr_is_multicast(&msg->target))
1456 goto out;
1457
1458 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1459
1460 if (n) {
1461 struct vxlan_fdb *f;
4b29dba9 1462 struct sk_buff *reply;
f564f45c
CW		 1463
1464 if (!(n->nud_state & NUD_CONNECTED)) {
1465 neigh_release(n);
1466 goto out;
1467 }
1468
1469 f = vxlan_find_mac(vxlan, n->ha);
1470 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1471 /* bridge-local neighbor */
1472 neigh_release(n);
1473 goto out;
1474 }
1475
4b29dba9
DS		 1476 reply = vxlan_na_create(skb, n,
1477 !!(f ? f->flags & NTF_ROUTER : 0));
1478
f564f45c 1479 neigh_release(n);
4b29dba9
DS		 1480
1481 if (reply == NULL)
1482 goto out;
1483
1484 if (netif_rx_ni(reply) == NET_RX_DROP)
1485 dev->stats.rx_dropped++;
1486
f564f45c 1487 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1488 union vxlan_addr ipa = {
1489 .sin6.sin6_addr = msg->target,
1490 .sa.sa_family = AF_INET6,
1491 };
1492
f564f45c
CW		 1493 vxlan_ip_miss(dev, &ipa);
1494 }
1495
1496out:
1497 consume_skb(skb);
1498 return NETDEV_TX_OK;
1499}
1500#endif
1501
e4f67add
DS		 1502static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1503{
1504 struct vxlan_dev *vxlan = netdev_priv(dev);
1505 struct neighbour *n;
e4f67add
DS		 1506
1507 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1508 return false;
1509
1510 n = NULL;
1511 switch (ntohs(eth_hdr(skb)->h_proto)) {
1512 case ETH_P_IP:
e15a00aa
CW		 1513 {
1514 struct iphdr *pip;
1515
e4f67add
DS		 1516 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1517 return false;
1518 pip = ip_hdr(skb);
1519 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1520 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1521 union vxlan_addr ipa = {
1522 .sin.sin_addr.s_addr = pip->daddr,
1523 .sa.sa_family = AF_INET,
1524 };
1525
1526 vxlan_ip_miss(dev, &ipa);
1527 return false;
1528 }
1529
e4f67add 1530 break;
e15a00aa
CW		 1531 }
1532#if IS_ENABLED(CONFIG_IPV6)
1533 case ETH_P_IPV6:
1534 {
1535 struct ipv6hdr *pip6;
1536
1537 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1538 return false;
1539 pip6 = ipv6_hdr(skb);
1540 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1541 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1542 union vxlan_addr ipa = {
1543 .sin6.sin6_addr = pip6->daddr,
1544 .sa.sa_family = AF_INET6,
1545 };
1546
1547 vxlan_ip_miss(dev, &ipa);
1548 return false;
1549 }
1550
1551 break;
1552 }
1553#endif
e4f67add
DS		 1554 default:
1555 return false;
1556 }
1557
1558 if (n) {
1559 bool diff;
1560
7367d0b5 1561 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1562 if (diff) {
1563 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1564 dev->addr_len);
1565 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1566 }
1567 neigh_release(n);
1568 return diff;
e4c7ed41
CW		 1569 }
1570
e4f67add
DS		 1571 return false;
1572}
1573
359a0ea9
TH		 1574static inline struct sk_buff *vxlan_handle_offloads(struct sk_buff *skb,
1575 bool udp_csum)
05c0db08 1576{
359a0ea9
TH		 1577 int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1578 return iptunnel_handle_offloads(skb, udp_csum, type);
05c0db08
PS		 1579}
1580
e4c7ed41 1581#if IS_ENABLED(CONFIG_IPV6)
11796187 1582static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1583 struct dst_entry *dst, struct sk_buff *skb,
1584 struct net_device *dev, struct in6_addr *saddr,
1585 struct in6_addr *daddr, __u8 prio, __u8 ttl,
f01ec1c0
ND		 1586 __be16 src_port, __be16 dst_port, __be32 vni,
1587 bool xnet)
e4c7ed41
CW		 1588{
1589 struct ipv6hdr *ip6h;
1590 struct vxlanhdr *vxh;
1591 struct udphdr *uh;
1592 int min_headroom;
1593 int err;
1594
359a0ea9
TH		 1595 skb = vxlan_handle_offloads(skb, !udp_get_no_check6_tx(vs->sock->sk));
1596 if (IS_ERR(skb))
1597 return -EINVAL;
e4c7ed41 1598
f01ec1c0 1599 skb_scrub_packet(skb, xnet);
963a88b3 1600
e4c7ed41
CW		 1601 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1602 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1603 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1604
1605 /* Need space for new headers (invalidates iph ptr) */
1606 err = skb_cow_head(skb, min_headroom);
1607 if (unlikely(err))
1608 return err;
1609
1610 if (vlan_tx_tag_present(skb)) {
1611 if (WARN_ON(!__vlan_put_tag(skb,
1612 skb->vlan_proto,
1613 vlan_tx_tag_get(skb))))
1614 return -ENOMEM;
1615
1616 skb->vlan_tci = 0;
1617 }
1618
1619 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1620 vxh->vx_flags = htonl(VXLAN_FLAGS);
1621 vxh->vx_vni = vni;
1622
1623 __skb_push(skb, sizeof(*uh));
1624 skb_reset_transport_header(skb);
1625 uh = udp_hdr(skb);
1626
1627 uh->dest = dst_port;
1628 uh->source = src_port;
1629
1630 uh->len = htons(skb->len);
e4c7ed41
CW		 1631
1632 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1633 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1634 IPSKB_REROUTED);
e4c7ed41
CW		 1635 skb_dst_set(skb, dst);
1636
359a0ea9
TH		 1637 udp6_set_csum(udp_get_no_check6_tx(vs->sock->sk), skb,
1638 saddr, daddr, skb->len);
e4c7ed41
CW		 1639
1640 __skb_push(skb, sizeof(*ip6h));
1641 skb_reset_network_header(skb);
1642 ip6h = ipv6_hdr(skb);
1643 ip6h->version = 6;
1644 ip6h->priority = prio;
1645 ip6h->flow_lbl[0] = 0;
1646 ip6h->flow_lbl[1] = 0;
1647 ip6h->flow_lbl[2] = 0;
1648 ip6h->payload_len = htons(skb->len);
1649 ip6h->nexthdr = IPPROTO_UDP;
1650 ip6h->hop_limit = ttl;
1651 ip6h->daddr = *daddr;
1652 ip6h->saddr = *saddr;
1653
e4c7ed41
CW		 1654 ip6tunnel_xmit(skb, dev);
1655 return 0;
1656}
1657#endif
1658
11796187 1659int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1660 struct rtable *rt, struct sk_buff *skb,
1661 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
f01ec1c0 1662 __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
49560532
PS		 1663{
1664 struct vxlanhdr *vxh;
1665 struct udphdr *uh;
649c5b8b 1666 int min_headroom;
49560532
PS		 1667 int err;
1668
359a0ea9
TH		 1669 skb = vxlan_handle_offloads(skb, !vs->sock->sk->sk_no_check_tx);
1670 if (IS_ERR(skb))
1671 return -EINVAL;
49560532 1672
649c5b8b 1673 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1674 + VXLAN_HLEN + sizeof(struct iphdr)
1675 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1676
1677 /* Need space for new headers (invalidates iph ptr) */
1678 err = skb_cow_head(skb, min_headroom);
1679 if (unlikely(err))
1680 return err;
1681
1eaa8178
PS		 1682 if (vlan_tx_tag_present(skb)) {
1683 if (WARN_ON(!__vlan_put_tag(skb,
1684 skb->vlan_proto,
1685 vlan_tx_tag_get(skb))))
1686 return -ENOMEM;
1687
1688 skb->vlan_tci = 0;
1689 }
1690
49560532
PS		 1691 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1692 vxh->vx_flags = htonl(VXLAN_FLAGS);
1693 vxh->vx_vni = vni;
1694
1695 __skb_push(skb, sizeof(*uh));
1696 skb_reset_transport_header(skb);
1697 uh = udp_hdr(skb);
1698
1699 uh->dest = dst_port;
1700 uh->source = src_port;
1701
1702 uh->len = htons(skb->len);
49560532 1703
359a0ea9
TH		 1704 udp_set_csum(vs->sock->sk->sk_no_check_tx, skb,
1705 src, dst, skb->len);
49560532 1706
aad88724 1707 return iptunnel_xmit(vs->sock->sk, rt, skb, src, dst, IPPROTO_UDP,
f01ec1c0 1708 tos, ttl, df, xnet);
49560532
PS		 1709}
1710EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1711
9dcc71e1
SS		 1712/* Bypass encapsulation if the destination is local */
1713static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1714 struct vxlan_dev *dst_vxlan)
1715{
8f84985f 1716 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1717 union vxlan_addr loopback;
1718 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1 1719
8f84985f
LR		 1720 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1721 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1722 skb->pkt_type = PACKET_HOST;
1723 skb->encapsulation = 0;
1724 skb->dev = dst_vxlan->dev;
1725 __skb_pull(skb, skb_network_offset(skb));
1726
e4c7ed41
CW		 1727 if (remote_ip->sa.sa_family == AF_INET) {
1728 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1729 loopback.sa.sa_family = AF_INET;
1730#if IS_ENABLED(CONFIG_IPV6)
1731 } else {
1732 loopback.sin6.sin6_addr = in6addr_loopback;
1733 loopback.sa.sa_family = AF_INET6;
1734#endif
1735 }
1736
9dcc71e1 1737 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1738 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1739
1740 u64_stats_update_begin(&tx_stats->syncp);
1741 tx_stats->tx_packets++;
1742 tx_stats->tx_bytes += skb->len;
1743 u64_stats_update_end(&tx_stats->syncp);
1744
1745 if (netif_rx(skb) == NET_RX_SUCCESS) {
1746 u64_stats_update_begin(&rx_stats->syncp);
1747 rx_stats->rx_packets++;
1748 rx_stats->rx_bytes += skb->len;
1749 u64_stats_update_end(&rx_stats->syncp);
1750 } else {
1751 skb->dev->stats.rx_dropped++;
1752 }
1753}
1754
4ad16930
SH		 1755static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1756 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1757{
1758 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1759 struct rtable *rt = NULL;
d342894c 1760 const struct iphdr *old_iph;
d342894c 1761 struct flowi4 fl4;
e4c7ed41
CW		 1762 union vxlan_addr *dst;
1763 __be16 src_port = 0, dst_port;
234f5b73 1764 u32 vni;
d342894c 1765 __be16 df = 0;
1766 __u8 tos, ttl;
0e6fbc5b 1767 int err;
d342894c 1768
823aa873 1769 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1770 vni = rdst->remote_vni;
e4c7ed41 1771 dst = &rdst->remote_ip;
e4f67add 1772
e4c7ed41 1773 if (vxlan_addr_any(dst)) {
e4f67add 1774 if (did_rsc) {
e4f67add 1775 /* short-circuited back to local bridge */
9dcc71e1 1776 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1777 return;
e4f67add 1778 }
ef59febe 1779 goto drop;
e4f67add 1780 }
ef59febe 1781
d342894c 1782 old_iph = ip_hdr(skb);
1783
d342894c 1784 ttl = vxlan->ttl;
e4c7ed41 1785 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1786 ttl = 1;
1787
1788 tos = vxlan->tos;
1789 if (tos == 1)
206aaafc 1790 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1791
535fb8d0
TH		 1792 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1793 vxlan->port_max, true);
d342894c 1794
e4c7ed41
CW		 1795 if (dst->sa.sa_family == AF_INET) {
1796 memset(&fl4, 0, sizeof(fl4));
1797 fl4.flowi4_oif = rdst->remote_ifindex;
1798 fl4.flowi4_tos = RT_TOS(tos);
1799 fl4.daddr = dst->sin.sin_addr.s_addr;
1800 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1801
f01ec1c0 1802 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1803 if (IS_ERR(rt)) {
1804 netdev_dbg(dev, "no route to %pI4\n",
1805 &dst->sin.sin_addr.s_addr);
1806 dev->stats.tx_carrier_errors++;
1807 goto tx_error;
1808 }
d342894c 1809
e4c7ed41
CW		 1810 if (rt->dst.dev == dev) {
1811 netdev_dbg(dev, "circular route to %pI4\n",
1812 &dst->sin.sin_addr.s_addr);
1813 dev->stats.collisions++;
fffc15a5 1814 goto rt_tx_error;
e4c7ed41
CW		 1815 }
1816
1817 /* Bypass encapsulation if the destination is local */
1818 if (rt->rt_flags & RTCF_LOCAL &&
1819 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1820 struct vxlan_dev *dst_vxlan;
1821
1822 ip_rt_put(rt);
f01ec1c0 1823 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
e4c7ed41
CW		 1824 if (!dst_vxlan)
1825 goto tx_error;
1826 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1827 return;
1828 }
1829
1830 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1831 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1832
11796187 1833 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
e4c7ed41
CW		 1834 fl4.saddr, dst->sin.sin_addr.s_addr,
1835 tos, ttl, df, src_port, dst_port,
f01ec1c0
ND		 1836 htonl(vni << 8),
1837 !net_eq(vxlan->net, dev_net(vxlan->dev)));
9dcc71e1 1838
e4c7ed41
CW		 1839 if (err < 0)
1840 goto rt_tx_error;
1841 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1842#if IS_ENABLED(CONFIG_IPV6)
1843 } else {
1844 struct sock *sk = vxlan->vn_sock->sock->sk;
1845 struct dst_entry *ndst;
1846 struct flowi6 fl6;
1847 u32 flags;
1848
1849 memset(&fl6, 0, sizeof(fl6));
1850 fl6.flowi6_oif = rdst->remote_ifindex;
1851 fl6.daddr = dst->sin6.sin6_addr;
1852 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1853 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1854
1855 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1856 netdev_dbg(dev, "no route to %pI6\n",
1857 &dst->sin6.sin6_addr);
1858 dev->stats.tx_carrier_errors++;
9dcc71e1 1859 goto tx_error;
e4c7ed41 1860 }
d342894c 1861
e4c7ed41
CW		 1862 if (ndst->dev == dev) {
1863 netdev_dbg(dev, "circular route to %pI6\n",
1864 &dst->sin6.sin6_addr);
1865 dst_release(ndst);
1866 dev->stats.collisions++;
1867 goto tx_error;
1868 }
0e6fbc5b 1869
e4c7ed41
CW		 1870 /* Bypass encapsulation if the destination is local */
1871 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1872 if (flags & RTF_LOCAL &&
1873 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1874 struct vxlan_dev *dst_vxlan;
1875
1876 dst_release(ndst);
f01ec1c0 1877 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
e4c7ed41
CW		 1878 if (!dst_vxlan)
1879 goto tx_error;
1880 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1881 return;
1882 }
49560532 1883
e4c7ed41
CW		 1884 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1885
11796187 1886 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41 1887 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
f01ec1c0
ND		 1888 src_port, dst_port, htonl(vni << 8),
1889 !net_eq(vxlan->net, dev_net(vxlan->dev)));
e4c7ed41
CW		 1890#endif
1891 }
0e6fbc5b 1892
4ad16930 1893 return;
d342894c 1894
1895drop:
1896 dev->stats.tx_dropped++;
1897 goto tx_free;
1898
49560532
PS		 1899rt_tx_error:
1900 ip_rt_put(rt);
d342894c 1901tx_error:
1902 dev->stats.tx_errors++;
1903tx_free:
1904 dev_kfree_skb(skb);
d342894c 1905}
1906
6681712d
DS		 1907/* Transmit local packets over Vxlan
1908 *
1909 * Outer IP header inherits ECN and DF from inner header.
1910 * Outer UDP destination is the VXLAN assigned port.
1911 * source port is based on hash of flow
1912 */
1913static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1914{
1915 struct vxlan_dev *vxlan = netdev_priv(dev);
1916 struct ethhdr *eth;
1917 bool did_rsc = false;
8f646c92 1918 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 1919 struct vxlan_fdb *f;
6681712d
DS		 1920
1921 skb_reset_mac_header(skb);
1922 eth = eth_hdr(skb);
1923
f564f45c
CW		 1924 if ((vxlan->flags & VXLAN_F_PROXY)) {
1925 if (ntohs(eth->h_proto) == ETH_P_ARP)
1926 return arp_reduce(dev, skb);
1927#if IS_ENABLED(CONFIG_IPV6)
1928 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1929 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1930 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1931 struct nd_msg *msg;
1932
1933 msg = (struct nd_msg *)skb_transport_header(skb);
1934 if (msg->icmph.icmp6_code == 0 &&
1935 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1936 return neigh_reduce(dev, skb);
1937 }
1938#endif
1939 }
6681712d
DS		 1940
1941 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1942 did_rsc = false;
1943
1944 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1945 (ntohs(eth->h_proto) == ETH_P_IP ||
1946 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1947 did_rsc = route_shortcircuit(dev, skb);
1948 if (did_rsc)
1949 f = vxlan_find_mac(vxlan, eth->h_dest);
1950 }
1951
6681712d 1952 if (f == NULL) {
afbd8bae
MR		 1953 f = vxlan_find_mac(vxlan, all_zeros_mac);
1954 if (f == NULL) {
1955 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1956 !is_multicast_ether_addr(eth->h_dest))
1957 vxlan_fdb_miss(vxlan, eth->h_dest);
1958
1959 dev->stats.tx_dropped++;
8f646c92 1960 kfree_skb(skb);
afbd8bae
MR		 1961 return NETDEV_TX_OK;
1962 }
1963 }
6681712d 1964
afbd8bae
MR		 1965 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1966 struct sk_buff *skb1;
6681712d 1967
8f646c92
ED		 1968 if (!fdst) {
1969 fdst = rdst;
1970 continue;
1971 }
afbd8bae
MR		 1972 skb1 = skb_clone(skb, GFP_ATOMIC);
1973 if (skb1)
1974 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1975 }
1976
8f646c92
ED		 1977 if (fdst)
1978 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1979 else
1980 kfree_skb(skb);
4ad16930 1981 return NETDEV_TX_OK;
6681712d
DS		 1982}
1983
d342894c 1984/* Walk the forwarding table and purge stale entries */
1985static void vxlan_cleanup(unsigned long arg)
1986{
1987 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1988 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1989 unsigned int h;
1990
1991 if (!netif_running(vxlan->dev))
1992 return;
1993
1994 spin_lock_bh(&vxlan->hash_lock);
1995 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1996 struct hlist_node *p, *n;
1997 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1998 struct vxlan_fdb *f
1999 = container_of(p, struct vxlan_fdb, hlist);
2000 unsigned long timeout;
2001
3c172868 2002 if (f->state & NUD_PERMANENT)
d342894c 2003 continue;
2004
2005 timeout = f->used + vxlan->age_interval * HZ;
2006 if (time_before_eq(timeout, jiffies)) {
2007 netdev_dbg(vxlan->dev,
2008 "garbage collect %pM\n",
2009 f->eth_addr);
2010 f->state = NUD_STALE;
2011 vxlan_fdb_destroy(vxlan, f);
2012 } else if (time_before(timeout, next_timer))
2013 next_timer = timeout;
2014 }
2015 }
2016 spin_unlock_bh(&vxlan->hash_lock);
2017
2018 mod_timer(&vxlan->age_timer, next_timer);
2019}
2020
9c2e24e1
PS		 2021static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2022{
2023 __u32 vni = vxlan->default_dst.remote_vni;
2024
2025 vxlan->vn_sock = vs;
2026 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2027}
2028
d342894c 2029/* Setup stats when device is created */
2030static int vxlan_init(struct net_device *dev)
2031{
1c51a915 2032 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2033 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2034 struct vxlan_sock *vs;
1c51a915 2035
1c213bd2 2036 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2037 if (!dev->tstats)
d342894c 2038 return -ENOMEM;
2039
1c51a915 2040 spin_lock(&vn->sock_lock);
f01ec1c0 2041 vs = vxlan_find_sock(vxlan->net, vxlan->dst_port);
1c51a915
SH		 2042 if (vs) {
2043 /* If we have a socket with same port already, reuse it */
2044 atomic_inc(&vs->refcnt);
9c2e24e1 2045 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 2046 } else {
2047 /* otherwise make new socket outside of RTNL */
2048 dev_hold(dev);
2049 queue_work(vxlan_wq, &vxlan->sock_work);
2050 }
2051 spin_unlock(&vn->sock_lock);
2052
d342894c 2053 return 0;
2054}
2055
ba609e9b 2056static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2057{
2058 struct vxlan_fdb *f;
2059
2060 spin_lock_bh(&vxlan->hash_lock);
2061 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2062 if (f)
2063 vxlan_fdb_destroy(vxlan, f);
2064 spin_unlock_bh(&vxlan->hash_lock);
2065}
2066
ebf4063e
SH		 2067static void vxlan_uninit(struct net_device *dev)
2068{
2069 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2070 struct vxlan_sock *vs = vxlan->vn_sock;
2071
ba609e9b 2072 vxlan_fdb_delete_default(vxlan);
afbd8bae 2073
ebf4063e 2074 if (vs)
012a5729 2075 vxlan_sock_release(vs);
ebf4063e
SH		 2076 free_percpu(dev->tstats);
2077}
2078
d342894c 2079/* Start ageing timer and join group when device is brought up */
2080static int vxlan_open(struct net_device *dev)
2081{
2082 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2083 struct vxlan_sock *vs = vxlan->vn_sock;
2084
2085 /* socket hasn't been created */
2086 if (!vs)
2087 return -ENOTCONN;
d342894c 2088
79d4a94f 2089 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2090 vxlan_sock_hold(vs);
7c47cedf 2091 dev_hold(dev);
3fc2de2f 2092 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2093 }
2094
2095 if (vxlan->age_interval)
2096 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2097
2098 return 0;
2099}
2100
2101/* Purge the forwarding table */
2102static void vxlan_flush(struct vxlan_dev *vxlan)
2103{
31fec5aa 2104 unsigned int h;
d342894c 2105
2106 spin_lock_bh(&vxlan->hash_lock);
2107 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2108 struct hlist_node *p, *n;
2109 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2110 struct vxlan_fdb *f
2111 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2112 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2113 if (!is_zero_ether_addr(f->eth_addr))
2114 vxlan_fdb_destroy(vxlan, f);
d342894c 2115 }
2116 }
2117 spin_unlock_bh(&vxlan->hash_lock);
2118}
2119
2120/* Cleanup timer and forwarding table on shutdown */
2121static int vxlan_stop(struct net_device *dev)
2122{
2123 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2124 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2125 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2126
e4c7ed41 2127 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2128 !vxlan_group_used(vn, vxlan)) {
1c51a915 2129 vxlan_sock_hold(vs);
7c47cedf 2130 dev_hold(dev);
3fc2de2f 2131 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2132 }
d342894c 2133
2134 del_timer_sync(&vxlan->age_timer);
2135
2136 vxlan_flush(vxlan);
2137
2138 return 0;
2139}
2140
d342894c 2141/* Stub, nothing needs to be done. */
2142static void vxlan_set_multicast_list(struct net_device *dev)
2143{
2144}
2145
345010b5
DB		 2146static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2147{
2148 struct vxlan_dev *vxlan = netdev_priv(dev);
2149 struct vxlan_rdst *dst = &vxlan->default_dst;
2150 struct net_device *lowerdev;
2151 int max_mtu;
2152
f01ec1c0 2153 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2154 if (lowerdev == NULL)
2155 return eth_change_mtu(dev, new_mtu);
2156
2157 if (dst->remote_ip.sa.sa_family == AF_INET6)
2158 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2159 else
2160 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2161
2162 if (new_mtu < 68 || new_mtu > max_mtu)
2163 return -EINVAL;
2164
2165 dev->mtu = new_mtu;
2166 return 0;
2167}
2168
d342894c 2169static const struct net_device_ops vxlan_netdev_ops = {
2170 .ndo_init = vxlan_init,
ebf4063e 2171 .ndo_uninit = vxlan_uninit,
d342894c 2172 .ndo_open = vxlan_open,
2173 .ndo_stop = vxlan_stop,
2174 .ndo_start_xmit = vxlan_xmit,
e8171045 2175 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2176 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2177 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2178 .ndo_validate_addr = eth_validate_addr,
2179 .ndo_set_mac_address = eth_mac_addr,
2180 .ndo_fdb_add = vxlan_fdb_add,
2181 .ndo_fdb_del = vxlan_fdb_delete,
2182 .ndo_fdb_dump = vxlan_fdb_dump,
2183};
2184
2185/* Info for udev, that this is a virtual tunnel endpoint */
2186static struct device_type vxlan_type = {
2187 .name = "vxlan",
2188};
2189
53cf5275 2190/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2191 * supply the listening VXLAN udp ports. Callers are expected
2192 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2193 */
2194void vxlan_get_rx_port(struct net_device *dev)
2195{
2196 struct vxlan_sock *vs;
2197 struct net *net = dev_net(dev);
2198 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2199 sa_family_t sa_family;
35e42379
JG		 2200 __be16 port;
2201 unsigned int i;
53cf5275
JG		 2202
2203 spin_lock(&vn->sock_lock);
2204 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2205 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2206 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2207 sa_family = vs->sock->sk->sk_family;
2208 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2209 port);
2210 }
2211 }
2212 spin_unlock(&vn->sock_lock);
2213}
2214EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2215
d342894c 2216/* Initialize the device structure. */
2217static void vxlan_setup(struct net_device *dev)
2218{
2219 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2220 unsigned int h;
d342894c 2221
2222 eth_hw_addr_random(dev);
2223 ether_setup(dev);
e4c7ed41 2224 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2225 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2226 else
2853af6a 2227 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2228
2229 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2230 dev->destructor = free_netdev;
d342894c 2231 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2232
2233 dev->tx_queue_len = 0;
2234 dev->features |= NETIF_F_LLTX;
d6727fe3 2235 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2236 dev->features |= NETIF_F_RXCSUM;
05c0db08 2237 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2238
1eaa8178
PS		 2239 dev->vlan_features = dev->features;
2240 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2241 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2242 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2243 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2244 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2245 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2246
553675fb 2247 INIT_LIST_HEAD(&vxlan->next);
d342894c 2248 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2249 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2250 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2251 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2252
2253 init_timer_deferrable(&vxlan->age_timer);
2254 vxlan->age_timer.function = vxlan_cleanup;
2255 vxlan->age_timer.data = (unsigned long) vxlan;
2256
823aa873 2257 vxlan->dst_port = htons(vxlan_port);
05f47d69 2258
d342894c 2259 vxlan->dev = dev;
2260
2261 for (h = 0; h < FDB_HASH_SIZE; ++h)
2262 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2263}
2264
2265static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2266 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2267 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2268 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2269 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2270 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2271 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2272 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2273 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2274 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2275 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2276 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2277 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2278 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2279 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2280 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2281 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2282 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2283};
2284
2285static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2286{
2287 if (tb[IFLA_ADDRESS]) {
2288 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2289 pr_debug("invalid link address (not ethernet)\n");
2290 return -EINVAL;
2291 }
2292
2293 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2294 pr_debug("invalid all zero ethernet address\n");
2295 return -EADDRNOTAVAIL;
2296 }
2297 }
2298
2299 if (!data)
2300 return -EINVAL;
2301
2302 if (data[IFLA_VXLAN_ID]) {
2303 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2304 if (id >= VXLAN_VID_MASK)
2305 return -ERANGE;
2306 }
2307
05f47d69 2308 if (data[IFLA_VXLAN_PORT_RANGE]) {
2309 const struct ifla_vxlan_port_range *p
2310 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2311
2312 if (ntohs(p->high) < ntohs(p->low)) {
2313 pr_debug("port range %u .. %u not valid\n",
2314 ntohs(p->low), ntohs(p->high));
2315 return -EINVAL;
2316 }
2317 }
2318
d342894c 2319 return 0;
2320}
2321
1b13c97f
YB		 2322static void vxlan_get_drvinfo(struct net_device *netdev,
2323 struct ethtool_drvinfo *drvinfo)
2324{
2325 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2326 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2327}
2328
2329static const struct ethtool_ops vxlan_ethtool_ops = {
2330 .get_drvinfo = vxlan_get_drvinfo,
2331 .get_link = ethtool_op_get_link,
2332};
2333
553675fb 2334static void vxlan_del_work(struct work_struct *work)
2335{
2336 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2337
2338 sk_release_kernel(vs->sock->sk);
2339 kfree_rcu(vs, rcu);
2340}
2341
e4c7ed41
CW		 2342#if IS_ENABLED(CONFIG_IPV6)
2343/* Create UDP socket for encapsulation receive. AF_INET6 socket
2344 * could be used for both IPv4 and IPv6 communications, but
2345 * users may set bindv6only=1.
2346 */
359a0ea9 2347static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
553675fb 2348{
553675fb 2349 struct sock *sk;
e4c7ed41
CW		 2350 struct socket *sock;
2351 struct sockaddr_in6 vxlan_addr = {
2352 .sin6_family = AF_INET6,
2353 .sin6_port = port,
2354 };
2355 int rc, val = 1;
2356
2357 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2358 if (rc < 0) {
2359 pr_debug("UDPv6 socket create failed\n");
39deb2c7 2360 return ERR_PTR(rc);
e4c7ed41
CW		 2361 }
2362
2363 /* Put in proper namespace */
2364 sk = sock->sk;
2365 sk_change_net(sk, net);
2366
2367 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2368 (char *)&val, sizeof(val));
2369 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2370 sizeof(struct sockaddr_in6));
2371 if (rc < 0) {
2372 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2373 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2374 sk_release_kernel(sk);
39deb2c7 2375 return ERR_PTR(rc);
e4c7ed41
CW		 2376 }
2377 /* At this point, IPv6 module should have been loaded in
2378 * sock_create_kern().
2379 */
2380 BUG_ON(!ipv6_stub);
2381
e4c7ed41
CW		 2382 /* Disable multicast loopback */
2383 inet_sk(sk)->mc_loop = 0;
359a0ea9
TH		 2384
2385 if (flags & VXLAN_F_UDP_ZERO_CSUM6_TX)
2386 udp_set_no_check6_tx(sk, true);
2387
2388 if (flags & VXLAN_F_UDP_ZERO_CSUM6_RX)
2389 udp_set_no_check6_rx(sk, true);
2390
39deb2c7 2391 return sock;
e4c7ed41
CW		 2392}
2393
2394#else
2395
359a0ea9 2396static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
e4c7ed41 2397{
39deb2c7 2398 return ERR_PTR(-EPFNOSUPPORT);
e4c7ed41
CW		 2399}
2400#endif
2401
359a0ea9 2402static struct socket *create_v4_sock(struct net *net, __be16 port, u32 flags)
e4c7ed41
CW		 2403{
2404 struct sock *sk;
2405 struct socket *sock;
553675fb 2406 struct sockaddr_in vxlan_addr = {
2407 .sin_family = AF_INET,
2408 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2409 .sin_port = port,
553675fb 2410 };
2411 int rc;
553675fb 2412
2413 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2414 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2415 if (rc < 0) {
2416 pr_debug("UDP socket create failed\n");
39deb2c7 2417 return ERR_PTR(rc);
553675fb 2418 }
2419
2420 /* Put in proper namespace */
e4c7ed41 2421 sk = sock->sk;
553675fb 2422 sk_change_net(sk, net);
2423
e4c7ed41 2424 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2425 sizeof(vxlan_addr));
2426 if (rc < 0) {
2427 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2428 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2429 sk_release_kernel(sk);
39deb2c7 2430 return ERR_PTR(rc);
e4c7ed41
CW		 2431 }
2432
e4c7ed41
CW		 2433 /* Disable multicast loopback */
2434 inet_sk(sk)->mc_loop = 0;
359a0ea9
TH		 2435
2436 if (!(flags & VXLAN_F_UDP_CSUM))
2437 sock->sk->sk_no_check_tx = 1;
2438
39deb2c7 2439 return sock;
e4c7ed41
CW		 2440}
2441
2442/* Create new listen socket if needed */
2443static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2444 vxlan_rcv_t *rcv, void *data,
2445 u32 flags)
e4c7ed41
CW		 2446{
2447 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2448 struct vxlan_sock *vs;
2449 struct socket *sock;
2450 struct sock *sk;
e4c7ed41 2451 unsigned int h;
359a0ea9 2452 bool ipv6 = !!(flags & VXLAN_F_IPV6);
e4c7ed41 2453
dc01e7d3 2454 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2455 if (!vs)
2456 return ERR_PTR(-ENOMEM);
2457
2458 for (h = 0; h < VNI_HASH_SIZE; ++h)
2459 INIT_HLIST_HEAD(&vs->vni_list[h]);
2460
2461 INIT_WORK(&vs->del_work, vxlan_del_work);
2462
2463 if (ipv6)
359a0ea9 2464 sock = create_v6_sock(net, port, flags);
e4c7ed41 2465 else
359a0ea9 2466 sock = create_v4_sock(net, port, flags);
39deb2c7 2467 if (IS_ERR(sock)) {
553675fb 2468 kfree(vs);
e50fddc8 2469 return ERR_CAST(sock);
553675fb 2470 }
e4c7ed41
CW		 2471
2472 vs->sock = sock;
2473 sk = sock->sk;
9c2e24e1 2474 atomic_set(&vs->refcnt, 1);
5cfccc5a 2475 vs->rcv = rcv;
012a5729 2476 vs->data = data;
559835ea 2477 rcu_assign_sk_user_data(vs->sock->sk, vs);
553675fb 2478
dc01e7d3
OG		 2479 /* Initialize the vxlan udp offloads structure */
2480 vs->udp_offloads.port = port;
2481 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2482 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2483
9c2e24e1
PS		 2484 spin_lock(&vn->sock_lock);
2485 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2486 vxlan_notify_add_rx_port(vs);
9c2e24e1 2487 spin_unlock(&vn->sock_lock);
553675fb 2488
2489 /* Mark socket as an encapsulation socket. */
2490 udp_sk(sk)->encap_type = 1;
2491 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2492#if IS_ENABLED(CONFIG_IPV6)
2493 if (ipv6)
2494 ipv6_stub->udpv6_encap_enable();
2495 else
2496#endif
2497 udp_encap_enable();
2498
9c2e24e1
PS		 2499 return vs;
2500}
2501
012a5729
PS		 2502struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2503 vxlan_rcv_t *rcv, void *data,
359a0ea9 2504 bool no_share, u32 flags)
9c2e24e1
PS		 2505{
2506 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2507 struct vxlan_sock *vs;
2508
359a0ea9 2509 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2510 if (!IS_ERR(vs))
2511 return vs;
553675fb 2512
012a5729
PS		 2513 if (no_share) /* Return error if sharing is not allowed. */
2514 return vs;
2515
9c2e24e1
PS		 2516 spin_lock(&vn->sock_lock);
2517 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2518 if (vs) {
2519 if (vs->rcv == rcv)
2520 atomic_inc(&vs->refcnt);
2521 else
2522 vs = ERR_PTR(-EBUSY);
2523 }
2524 spin_unlock(&vn->sock_lock);
2525
2526 if (!vs)
9c2e24e1
PS		 2527 vs = ERR_PTR(-EINVAL);
2528
553675fb 2529 return vs;
2530}
012a5729 2531EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2532
1c51a915
SH		 2533/* Scheduled at device creation to bind to a socket */
2534static void vxlan_sock_work(struct work_struct *work)
2535{
9c2e24e1 2536 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2537 struct net *net = vxlan->net;
1c51a915 2538 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2539 __be16 port = vxlan->dst_port;
2540 struct vxlan_sock *nvs;
1c51a915 2541
359a0ea9 2542 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2543 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2544 if (!IS_ERR(nvs))
2545 vxlan_vs_add_dev(nvs, vxlan);
2546 spin_unlock(&vn->sock_lock);
2547
2548 dev_put(vxlan->dev);
1c51a915
SH		 2549}
2550
d342894c 2551static int vxlan_newlink(struct net *net, struct net_device *dev,
2552 struct nlattr *tb[], struct nlattr *data[])
2553{
553675fb 2554 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2555 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2556 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2557 __u32 vni;
2558 int err;
e4c7ed41 2559 bool use_ipv6 = false;
d342894c 2560
2561 if (!data[IFLA_VXLAN_ID])
2562 return -EINVAL;
2563
f01ec1c0
ND		 2564 vxlan->net = dev_net(dev);
2565
d342894c 2566 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2567 dst->remote_vni = vni;
d342894c 2568
5933a7bb
MR		 2569 /* Unless IPv6 is explicitly requested, assume IPv4 */
2570 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2571 if (data[IFLA_VXLAN_GROUP]) {
2572 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2573 } else if (data[IFLA_VXLAN_GROUP6]) {
2574 if (!IS_ENABLED(CONFIG_IPV6))
2575 return -EPFNOSUPPORT;
2576
2577 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2578 sizeof(struct in6_addr));
2579 dst->remote_ip.sa.sa_family = AF_INET6;
2580 use_ipv6 = true;
2581 }
d342894c 2582
e4c7ed41
CW		 2583 if (data[IFLA_VXLAN_LOCAL]) {
2584 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2585 vxlan->saddr.sa.sa_family = AF_INET;
2586 } else if (data[IFLA_VXLAN_LOCAL6]) {
2587 if (!IS_ENABLED(CONFIG_IPV6))
2588 return -EPFNOSUPPORT;
2589
2590 /* TODO: respect scope id */
2591 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2592 sizeof(struct in6_addr));
2593 vxlan->saddr.sa.sa_family = AF_INET6;
2594 use_ipv6 = true;
2595 }
d342894c 2596
34e02aa1 2597 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2598 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2599 struct net_device *lowerdev
c7995c43 2600 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2601
2602 if (!lowerdev) {
c7995c43 2603 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2604 return -ENODEV;
2605 }
d342894c 2606
e4c7ed41
CW		 2607#if IS_ENABLED(CONFIG_IPV6)
2608 if (use_ipv6) {
2609 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2610 if (idev && idev->cnf.disable_ipv6) {
2611 pr_info("IPv6 is disabled via sysctl\n");
2612 return -EPERM;
2613 }
2614 vxlan->flags |= VXLAN_F_IPV6;
2615 }
2616#endif
2617
34e02aa1 2618 if (!tb[IFLA_MTU])
e4c7ed41 2619 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2620
2853af6a 2621 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2622 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2623 } else if (use_ipv6)
2624 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2625
2626 if (data[IFLA_VXLAN_TOS])
2627 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2628
afb97186
VB		 2629 if (data[IFLA_VXLAN_TTL])
2630 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2631
d342894c 2632 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2633 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2634
2635 if (data[IFLA_VXLAN_AGEING])
2636 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2637 else
2638 vxlan->age_interval = FDB_AGE_DEFAULT;
2639
e4f67add
DS		 2640 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2641 vxlan->flags |= VXLAN_F_PROXY;
2642
2643 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2644 vxlan->flags |= VXLAN_F_RSC;
2645
2646 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2647 vxlan->flags |= VXLAN_F_L2MISS;
2648
2649 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2650 vxlan->flags |= VXLAN_F_L3MISS;
2651
d342894c 2652 if (data[IFLA_VXLAN_LIMIT])
2653 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2654
05f47d69 2655 if (data[IFLA_VXLAN_PORT_RANGE]) {
2656 const struct ifla_vxlan_port_range *p
2657 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2658 vxlan->port_min = ntohs(p->low);
2659 vxlan->port_max = ntohs(p->high);
2660 }
2661
823aa873 2662 if (data[IFLA_VXLAN_PORT])
2663 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2664
359a0ea9
TH		 2665 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2666 vxlan->flags |= VXLAN_F_UDP_CSUM;
2667
2668 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2669 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2670 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2671
2672 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2673 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2674 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2675
553675fb 2676 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2677 pr_info("duplicate VNI %u\n", vni);
2678 return -EEXIST;
2679 }
2680
7ad24ea4 2681 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2682
2936b6ab
SS		 2683 /* create an fdb entry for a valid default destination */
2684 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2685 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2686 &vxlan->default_dst.remote_ip,
2687 NUD_REACHABLE|NUD_PERMANENT,
2688 NLM_F_EXCL|NLM_F_CREATE,
2689 vxlan->dst_port,
2690 vxlan->default_dst.remote_vni,
2691 vxlan->default_dst.remote_ifindex,
2692 NTF_SELF);
2693 if (err)
2694 return err;
2695 }
d342894c 2696
afbd8bae
MR		 2697 err = register_netdevice(dev);
2698 if (err) {
ba609e9b 2699 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2700 return err;
2701 }
2702
553675fb 2703 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2704
2705 return 0;
d342894c 2706}
2707
2708static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2709{
2710 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2711 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2712
fe5c3561 2713 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2714 if (!hlist_unhashed(&vxlan->hlist))
2715 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2716 spin_unlock(&vn->sock_lock);
2717
553675fb 2718 list_del(&vxlan->next);
d342894c 2719 unregister_netdevice_queue(dev, head);
2720}
2721
2722static size_t vxlan_get_size(const struct net_device *dev)
2723{
2724
2725 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2726 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2727 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2728 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2729 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2730 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2731 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2732 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2733 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2734 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2735 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2736 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2737 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2738 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2739 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2740 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2741 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2742 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
d342894c 2743 0;
2744}
2745
2746static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2747{
2748 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2749 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2750 struct ifla_vxlan_port_range ports = {
2751 .low = htons(vxlan->port_min),
2752 .high = htons(vxlan->port_max),
2753 };
d342894c 2754
c7995c43 2755 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2756 goto nla_put_failure;
2757
e4c7ed41
CW		 2758 if (!vxlan_addr_any(&dst->remote_ip)) {
2759 if (dst->remote_ip.sa.sa_family == AF_INET) {
2760 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2761 dst->remote_ip.sin.sin_addr.s_addr))
2762 goto nla_put_failure;
2763#if IS_ENABLED(CONFIG_IPV6)
2764 } else {
2765 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2766 &dst->remote_ip.sin6.sin6_addr))
2767 goto nla_put_failure;
2768#endif
2769 }
2770 }
d342894c 2771
c7995c43 2772 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2773 goto nla_put_failure;
2774
e4c7ed41
CW		 2775 if (!vxlan_addr_any(&vxlan->saddr)) {
2776 if (vxlan->saddr.sa.sa_family == AF_INET) {
2777 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2778 vxlan->saddr.sin.sin_addr.s_addr))
2779 goto nla_put_failure;
2780#if IS_ENABLED(CONFIG_IPV6)
2781 } else {
2782 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2783 &vxlan->saddr.sin6.sin6_addr))
2784 goto nla_put_failure;
2785#endif
2786 }
2787 }
d342894c 2788
2789 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2790 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2791 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2792 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2793 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2794 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2795 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2796 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2797 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2798 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2799 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2800 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2801 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2802 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2803 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2804 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2805 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2806 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2807 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2808 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
d342894c 2809 goto nla_put_failure;
2810
05f47d69 2811 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2812 goto nla_put_failure;
2813
d342894c 2814 return 0;
2815
2816nla_put_failure:
2817 return -EMSGSIZE;
2818}
2819
2820static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2821 .kind = "vxlan",
2822 .maxtype = IFLA_VXLAN_MAX,
2823 .policy = vxlan_policy,
2824 .priv_size = sizeof(struct vxlan_dev),
2825 .setup = vxlan_setup,
2826 .validate = vxlan_validate,
2827 .newlink = vxlan_newlink,
2828 .dellink = vxlan_dellink,
2829 .get_size = vxlan_get_size,
2830 .fill_info = vxlan_fill_info,
2831};
2832
acaf4e70
DB		 2833static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2834 struct net_device *dev)
2835{
2836 struct vxlan_dev *vxlan, *next;
2837 LIST_HEAD(list_kill);
2838
2839 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2840 struct vxlan_rdst *dst = &vxlan->default_dst;
2841
2842 /* In case we created vxlan device with carrier
2843 * and we loose the carrier due to module unload
2844 * we also need to remove vxlan device. In other
2845 * cases, it's not necessary and remote_ifindex
2846 * is 0 here, so no matches.
2847 */
2848 if (dst->remote_ifindex == dev->ifindex)
2849 vxlan_dellink(vxlan->dev, &list_kill);
2850 }
2851
2852 unregister_netdevice_many(&list_kill);
2853}
2854
2855static int vxlan_lowerdev_event(struct notifier_block *unused,
2856 unsigned long event, void *ptr)
2857{
2858 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2859 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2860
783c1463 2861 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2862 vxlan_handle_lowerdev_unregister(vn, dev);
2863
2864 return NOTIFY_DONE;
2865}
2866
2867static struct notifier_block vxlan_notifier_block __read_mostly = {
2868 .notifier_call = vxlan_lowerdev_event,
2869};
2870
d342894c 2871static __net_init int vxlan_init_net(struct net *net)
2872{
2873 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2874 unsigned int h;
d342894c 2875
553675fb 2876 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2877 spin_lock_init(&vn->sock_lock);
d342894c 2878
553675fb 2879 for (h = 0; h < PORT_HASH_SIZE; ++h)
2880 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2881
2882 return 0;
2883}
2884
f01ec1c0
ND		 2885static void __net_exit vxlan_exit_net(struct net *net)
2886{
2887 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2888 struct vxlan_dev *vxlan, *next;
2889 struct net_device *dev, *aux;
2890 LIST_HEAD(list);
2891
2892 rtnl_lock();
2893 for_each_netdev_safe(net, dev, aux)
2894 if (dev->rtnl_link_ops == &vxlan_link_ops)
2895 unregister_netdevice_queue(dev, &list);
2896
2897 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2898 /* If vxlan->dev is in the same netns, it has already been added
2899 * to the list by the previous loop.
2900 */
2901 if (!net_eq(dev_net(vxlan->dev), net))
2902 unregister_netdevice_queue(dev, &list);
2903 }
2904
2905 unregister_netdevice_many(&list);
2906 rtnl_unlock();
2907}
2908
d342894c 2909static struct pernet_operations vxlan_net_ops = {
2910 .init = vxlan_init_net,
f01ec1c0 2911 .exit = vxlan_exit_net,
d342894c 2912 .id = &vxlan_net_id,
2913 .size = sizeof(struct vxlan_net),
2914};
2915
2916static int __init vxlan_init_module(void)
2917{
2918 int rc;
2919
758c57d1
SH		 2920 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2921 if (!vxlan_wq)
2922 return -ENOMEM;
2923
d342894c 2924 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2925
783c1463 2926 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 2927 if (rc)
2928 goto out1;
2929
acaf4e70 2930 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 2931 if (rc)
2932 goto out2;
2933
acaf4e70
DB		 2934 rc = rtnl_link_register(&vxlan_link_ops);
2935 if (rc)
2936 goto out3;
d342894c 2937
acaf4e70
DB		 2938 return 0;
2939out3:
2940 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 2941out2:
783c1463 2942 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 2943out1:
758c57d1 2944 destroy_workqueue(vxlan_wq);
d342894c 2945 return rc;
2946}
7332a13b 2947late_initcall(vxlan_init_module);
d342894c 2948
2949static void __exit vxlan_cleanup_module(void)
2950{
b7153984 2951 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 2952 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 2953 destroy_workqueue(vxlan_wq);
783c1463
DB		 2954 unregister_pernet_subsys(&vxlan_net_ops);
2955 /* rcu_barrier() is called by netns */
d342894c 2956}
2957module_exit(vxlan_cleanup_module);
2958
2959MODULE_LICENSE("GPL");
2960MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2961MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 2962MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 2963MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror