projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: Save software checksum complete
[deliverable/linux.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
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 */
568
569 off_eth = skb_gro_offset(skb);
570 hlen = off_eth + sizeof(*eh);
571 eh = skb_gro_header_fast(skb, off_eth);
572 if (skb_gro_header_hard(skb, hlen)) {
573 eh = skb_gro_header_slow(skb, hlen, off_eth);
574 if (unlikely(!eh))
575 goto out;
576 }
577
578 flush = 0;
579
580 for (p = *head; p; p = p->next) {
581 if (!NAPI_GRO_CB(p)->same_flow)
582 continue;
583
584 vh2 = (struct vxlanhdr *)(p->data + off_vx);
585 eh2 = (struct ethhdr *)(p->data + off_eth);
586 if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) {
587 NAPI_GRO_CB(p)->same_flow = 0;
588 continue;
589 }
dc01e7d3
OG		 590 }
591
dc01e7d3
OG		 592 type = eh->h_proto;
593
594 rcu_read_lock();
595 ptype = gro_find_receive_by_type(type);
596 if (ptype == NULL) {
597 flush = 1;
598 goto out_unlock;
599 }
600
601 skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */
602 pp = ptype->callbacks.gro_receive(head, skb);
603
604out_unlock:
605 rcu_read_unlock();
606out:
607 NAPI_GRO_CB(skb)->flush |= flush;
608
609 return pp;
610}
611
612static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
613{
614 struct ethhdr *eh;
615 struct packet_offload *ptype;
616 __be16 type;
617 int vxlan_len = sizeof(struct vxlanhdr) + sizeof(struct ethhdr);
618 int err = -ENOSYS;
619
620 eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr));
621 type = eh->h_proto;
622
623 rcu_read_lock();
624 ptype = gro_find_complete_by_type(type);
625 if (ptype != NULL)
626 err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len);
627
628 rcu_read_unlock();
629 return err;
630}
631
53cf5275 632/* Notify netdevs that UDP port started listening */
dc01e7d3 633static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 634{
635 struct net_device *dev;
dc01e7d3 636 struct sock *sk = vs->sock->sk;
53cf5275
JG		 637 struct net *net = sock_net(sk);
638 sa_family_t sa_family = sk->sk_family;
35e42379 639 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 640 int err;
641
642 if (sa_family == AF_INET) {
643 err = udp_add_offload(&vs->udp_offloads);
644 if (err)
645 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
646 }
53cf5275
JG		 647
648 rcu_read_lock();
649 for_each_netdev_rcu(net, dev) {
650 if (dev->netdev_ops->ndo_add_vxlan_port)
651 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
652 port);
653 }
654 rcu_read_unlock();
655}
656
657/* Notify netdevs that UDP port is no more listening */
dc01e7d3 658static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 659{
660 struct net_device *dev;
dc01e7d3 661 struct sock *sk = vs->sock->sk;
53cf5275
JG		 662 struct net *net = sock_net(sk);
663 sa_family_t sa_family = sk->sk_family;
35e42379 664 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 665
666 rcu_read_lock();
667 for_each_netdev_rcu(net, dev) {
668 if (dev->netdev_ops->ndo_del_vxlan_port)
669 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
670 port);
671 }
672 rcu_read_unlock();
dc01e7d3
OG		 673
674 if (sa_family == AF_INET)
675 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 676}
677
d342894c 678/* Add new entry to forwarding table -- assumes lock held */
679static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 680 const u8 *mac, union vxlan_addr *ip,
6681712d 681 __u16 state, __u16 flags,
73cf3317 682 __be16 port, __u32 vni, __u32 ifindex,
ae884082 683 __u8 ndm_flags)
d342894c 684{
9e4b93f9 685 struct vxlan_rdst *rd = NULL;
d342894c 686 struct vxlan_fdb *f;
687 int notify = 0;
688
014be2c8 689 f = __vxlan_find_mac(vxlan, mac);
d342894c 690 if (f) {
691 if (flags & NLM_F_EXCL) {
692 netdev_dbg(vxlan->dev,
693 "lost race to create %pM\n", mac);
694 return -EEXIST;
695 }
696 if (f->state != state) {
697 f->state = state;
698 f->updated = jiffies;
699 notify = 1;
700 }
ae884082
DS		 701 if (f->flags != ndm_flags) {
702 f->flags = ndm_flags;
703 f->updated = jiffies;
704 notify = 1;
705 }
906dc186
TR		 706 if ((flags & NLM_F_REPLACE)) {
707 /* Only change unicasts */
708 if (!(is_multicast_ether_addr(f->eth_addr) ||
709 is_zero_ether_addr(f->eth_addr))) {
710 int rc = vxlan_fdb_replace(f, ip, port, vni,
711 ifindex);
712
713 if (rc < 0)
714 return rc;
715 notify |= rc;
716 } else
717 return -EOPNOTSUPP;
718 }
6681712d 719 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 720 (is_multicast_ether_addr(f->eth_addr) ||
721 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 722 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
723 &rd);
6681712d
DS		 724
725 if (rc < 0)
726 return rc;
727 notify |= rc;
728 }
d342894c 729 } else {
730 if (!(flags & NLM_F_CREATE))
731 return -ENOENT;
732
733 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
734 return -ENOSPC;
735
906dc186
TR		 736 /* Disallow replace to add a multicast entry */
737 if ((flags & NLM_F_REPLACE) &&
738 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
739 return -EOPNOTSUPP;
740
e4c7ed41 741 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 742 f = kmalloc(sizeof(*f), GFP_ATOMIC);
743 if (!f)
744 return -ENOMEM;
745
746 notify = 1;
d342894c 747 f->state = state;
ae884082 748 f->flags = ndm_flags;
d342894c 749 f->updated = f->used = jiffies;
3e61aa8f 750 INIT_LIST_HEAD(&f->remotes);
d342894c 751 memcpy(f->eth_addr, mac, ETH_ALEN);
752
9e4b93f9 753 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 754
d342894c 755 ++vxlan->addrcnt;
756 hlist_add_head_rcu(&f->hlist,
757 vxlan_fdb_head(vxlan, mac));
758 }
759
9e4b93f9
ND		 760 if (notify) {
761 if (rd == NULL)
762 rd = first_remote_rtnl(f);
763 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
764 }
d342894c 765
766 return 0;
767}
768
6706c82e 769static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 770{
771 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 772 struct vxlan_rdst *rd, *nd;
6681712d 773
3e61aa8f 774 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 775 kfree(rd);
6681712d
DS		 776 kfree(f);
777}
778
d342894c 779static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
780{
781 netdev_dbg(vxlan->dev,
782 "delete %pM\n", f->eth_addr);
783
784 --vxlan->addrcnt;
9e4b93f9 785 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 786
787 hlist_del_rcu(&f->hlist);
6681712d 788 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 789}
790
f0b074be 791static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 792 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 793{
6681712d 794 struct net *net = dev_net(vxlan->dev);
e4c7ed41 795 int err;
d342894c 796
f0b074be 797 if (tb[NDA_DST]) {
e4c7ed41
CW		 798 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
799 if (err)
800 return err;
f0b074be 801 } else {
e4c7ed41
CW		 802 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
803 if (remote->sa.sa_family == AF_INET) {
804 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
805 ip->sa.sa_family = AF_INET;
806#if IS_ENABLED(CONFIG_IPV6)
807 } else {
808 ip->sin6.sin6_addr = in6addr_any;
809 ip->sa.sa_family = AF_INET6;
810#endif
811 }
f0b074be 812 }
d342894c 813
6681712d 814 if (tb[NDA_PORT]) {
73cf3317 815 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 816 return -EINVAL;
f0b074be
MR		 817 *port = nla_get_be16(tb[NDA_PORT]);
818 } else {
819 *port = vxlan->dst_port;
820 }
6681712d
DS		 821
822 if (tb[NDA_VNI]) {
823 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
824 return -EINVAL;
f0b074be
MR		 825 *vni = nla_get_u32(tb[NDA_VNI]);
826 } else {
827 *vni = vxlan->default_dst.remote_vni;
828 }
6681712d
DS		 829
830 if (tb[NDA_IFINDEX]) {
5abb0029 831 struct net_device *tdev;
6681712d
DS		 832
833 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
834 return -EINVAL;
f0b074be 835 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 836 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 837 if (!tdev)
6681712d 838 return -EADDRNOTAVAIL;
f0b074be
MR		 839 } else {
840 *ifindex = 0;
841 }
842
843 return 0;
844}
845
846/* Add static entry (via netlink) */
847static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
848 struct net_device *dev,
849 const unsigned char *addr, u16 flags)
850{
851 struct vxlan_dev *vxlan = netdev_priv(dev);
852 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 853 union vxlan_addr ip;
f0b074be
MR		 854 __be16 port;
855 u32 vni, ifindex;
856 int err;
857
858 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
859 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
860 ndm->ndm_state);
861 return -EINVAL;
862 }
863
864 if (tb[NDA_DST] == NULL)
865 return -EINVAL;
866
867 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
868 if (err)
869 return err;
6681712d 870
5933a7bb
MR		 871 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
872 return -EAFNOSUPPORT;
873
d342894c 874 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 875 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 876 port, vni, ifindex, ndm->ndm_flags);
d342894c 877 spin_unlock_bh(&vxlan->hash_lock);
878
879 return err;
880}
881
882/* Delete entry (via netlink) */
1690be63
VY		 883static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
884 struct net_device *dev,
d342894c 885 const unsigned char *addr)
886{
887 struct vxlan_dev *vxlan = netdev_priv(dev);
888 struct vxlan_fdb *f;
bc7892ba 889 struct vxlan_rdst *rd = NULL;
e4c7ed41 890 union vxlan_addr ip;
bc7892ba
MR		 891 __be16 port;
892 u32 vni, ifindex;
893 int err;
894
895 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
896 if (err)
897 return err;
898
899 err = -ENOENT;
d342894c 900
901 spin_lock_bh(&vxlan->hash_lock);
902 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 903 if (!f)
904 goto out;
905
e4c7ed41
CW		 906 if (!vxlan_addr_any(&ip)) {
907 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 908 if (!rd)
909 goto out;
910 }
911
912 err = 0;
913
914 /* remove a destination if it's not the only one on the list,
915 * otherwise destroy the fdb entry
916 */
917 if (rd && !list_is_singular(&f->remotes)) {
918 list_del_rcu(&rd->list);
9e4b93f9 919 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 920 kfree_rcu(rd, rcu);
bc7892ba 921 goto out;
d342894c 922 }
bc7892ba
MR		 923
924 vxlan_fdb_destroy(vxlan, f);
925
926out:
d342894c 927 spin_unlock_bh(&vxlan->hash_lock);
928
929 return err;
930}
931
932/* Dump forwarding table */
933static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
934 struct net_device *dev, int idx)
935{
936 struct vxlan_dev *vxlan = netdev_priv(dev);
937 unsigned int h;
938
939 for (h = 0; h < FDB_HASH_SIZE; ++h) {
940 struct vxlan_fdb *f;
d342894c 941 int err;
942
b67bfe0d 943 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 944 struct vxlan_rdst *rd;
6681712d 945
3e61aa8f
SH		 946 if (idx < cb->args[0])
947 goto skip;
948
949 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 950 err = vxlan_fdb_info(skb, vxlan, f,
951 NETLINK_CB(cb->skb).portid,
952 cb->nlh->nlmsg_seq,
953 RTM_NEWNEIGH,
954 NLM_F_MULTI, rd);
955 if (err < 0)
3e61aa8f 956 goto out;
6681712d 957 }
3e61aa8f
SH		 958skip:
959 ++idx;
d342894c 960 }
961 }
3e61aa8f 962out:
d342894c 963 return idx;
964}
965
966/* Watch incoming packets to learn mapping between Ethernet address
967 * and Tunnel endpoint.
26a41ae6 968 * Return true if packet is bogus and should be droppped.
d342894c 969 */
26a41ae6 970static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 971 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 972{
973 struct vxlan_dev *vxlan = netdev_priv(dev);
974 struct vxlan_fdb *f;
d342894c 975
976 f = vxlan_find_mac(vxlan, src_mac);
977 if (likely(f)) {
5ca5461c 978 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 979
e4c7ed41 980 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 981 return false;
982
983 /* Don't migrate static entries, drop packets */
eb064c3b 984 if (f->state & NUD_NOARP)
26a41ae6 985 return true;
d342894c 986
987 if (net_ratelimit())
988 netdev_info(dev,
e4c7ed41 989 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 990 src_mac, &rdst->remote_ip, &src_ip);
d342894c 991
e4c7ed41 992 rdst->remote_ip = *src_ip;
d342894c 993 f->updated = jiffies;
9e4b93f9 994 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 995 } else {
996 /* learned new entry */
997 spin_lock(&vxlan->hash_lock);
3bf74b1a 998
999 /* close off race between vxlan_flush and incoming packets */
1000 if (netif_running(dev))
1001 vxlan_fdb_create(vxlan, src_mac, src_ip,
1002 NUD_REACHABLE,
1003 NLM_F_EXCL|NLM_F_CREATE,
1004 vxlan->dst_port,
1005 vxlan->default_dst.remote_vni,
1006 0, NTF_SELF);
d342894c 1007 spin_unlock(&vxlan->hash_lock);
1008 }
26a41ae6 1009
1010 return false;
d342894c 1011}
1012
d342894c 1013/* See if multicast group is already in use by other ID */
95ab0991 1014static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1015{
553675fb 1016 struct vxlan_dev *vxlan;
d342894c 1017
95ab0991
G		 1018 /* The vxlan_sock is only used by dev, leaving group has
1019 * no effect on other vxlan devices.
1020 */
1021 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1022 return false;
1023
553675fb 1024 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1025 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1026 continue;
d342894c 1027
95ab0991
G		 1028 if (vxlan->vn_sock != dev->vn_sock)
1029 continue;
1030
1031 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1032 &dev->default_dst.remote_ip))
1033 continue;
1034
1035 if (vxlan->default_dst.remote_ifindex !=
1036 dev->default_dst.remote_ifindex)
1037 continue;
1038
1039 return true;
553675fb 1040 }
d342894c 1041
1042 return false;
1043}
1044
7c47cedf 1045static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1046{
7c47cedf
SH		 1047 atomic_inc(&vs->refcnt);
1048}
d342894c 1049
012a5729 1050void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1051{
53cf5275
JG		 1052 struct sock *sk = vs->sock->sk;
1053 struct net *net = sock_net(sk);
1054 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1055
7c47cedf
SH		 1056 if (!atomic_dec_and_test(&vs->refcnt))
1057 return;
d342894c 1058
1c51a915 1059 spin_lock(&vn->sock_lock);
7c47cedf 1060 hlist_del_rcu(&vs->hlist);
559835ea 1061 rcu_assign_sk_user_data(vs->sock->sk, NULL);
dc01e7d3 1062 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1063 spin_unlock(&vn->sock_lock);
1064
7c47cedf 1065 queue_work(vxlan_wq, &vs->del_work);
d342894c 1066}
012a5729 1067EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1068
3fc2de2f 1069/* Callback to update multicast group membership when first VNI on
1070 * multicast asddress is brought up
1071 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1072 */
3fc2de2f 1073static void vxlan_igmp_join(struct work_struct *work)
d342894c 1074{
3fc2de2f 1075 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1076 struct vxlan_sock *vs = vxlan->vn_sock;
1077 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1078 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1079 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1080
d342894c 1081 lock_sock(sk);
e4c7ed41
CW		 1082 if (ip->sa.sa_family == AF_INET) {
1083 struct ip_mreqn mreq = {
1084 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1085 .imr_ifindex = ifindex,
1086 };
1087
1088 ip_mc_join_group(sk, &mreq);
1089#if IS_ENABLED(CONFIG_IPV6)
1090 } else {
1091 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1092 &ip->sin6.sin6_addr);
1093#endif
1094 }
3fc2de2f 1095 release_sock(sk);
1096
012a5729 1097 vxlan_sock_release(vs);
3fc2de2f 1098 dev_put(vxlan->dev);
1099}
1100
1101/* Inverse of vxlan_igmp_join when last VNI is brought down */
1102static void vxlan_igmp_leave(struct work_struct *work)
1103{
1104 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1105 struct vxlan_sock *vs = vxlan->vn_sock;
1106 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1107 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1108 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1109
1110 lock_sock(sk);
e4c7ed41
CW		 1111 if (ip->sa.sa_family == AF_INET) {
1112 struct ip_mreqn mreq = {
1113 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1114 .imr_ifindex = ifindex,
1115 };
1116
1117 ip_mc_leave_group(sk, &mreq);
1118#if IS_ENABLED(CONFIG_IPV6)
1119 } else {
1120 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1121 &ip->sin6.sin6_addr);
1122#endif
1123 }
1124
d342894c 1125 release_sock(sk);
d342894c 1126
012a5729 1127 vxlan_sock_release(vs);
7c47cedf 1128 dev_put(vxlan->dev);
d342894c 1129}
1130
1131/* Callback from net/ipv4/udp.c to receive packets */
1132static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1133{
5cfccc5a 1134 struct vxlan_sock *vs;
d342894c 1135 struct vxlanhdr *vxh;
d342894c 1136
d342894c 1137 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1138 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1139 goto error;
1140
7ce04758
PS		 1141 /* Return packets with reserved bits set */
1142 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 1143 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1144 (vxh->vx_vni & htonl(0xff))) {
1145 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1146 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1147 goto error;
1148 }
1149
5cfccc5a
PS		 1150 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1151 goto drop;
1152
559835ea 1153 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1154 if (!vs)
d342894c 1155 goto drop;
d342894c 1156
d0bc6555
OG		 1157 /* If the NIC driver gave us an encapsulated packet
1158 * with the encapsulation mark, the device checksummed it
1159 * for us. Otherwise force the upper layers to verify it.
1160 */
1161 if ((skb->ip_summed != CHECKSUM_UNNECESSARY && skb->ip_summed != CHECKSUM_PARTIAL) ||
1162 !skb->encapsulation)
1163 skb->ip_summed = CHECKSUM_NONE;
1164
1165 skb->encapsulation = 0;
1166
5cfccc5a
PS		 1167 vs->rcv(vs, skb, vxh->vx_vni);
1168 return 0;
1169
1170drop:
1171 /* Consume bad packet */
1172 kfree_skb(skb);
1173 return 0;
1174
1175error:
1176 /* Return non vxlan pkt */
1177 return 1;
1178}
1179
1180static void vxlan_rcv(struct vxlan_sock *vs,
1181 struct sk_buff *skb, __be32 vx_vni)
1182{
e4c7ed41
CW		 1183 struct iphdr *oip = NULL;
1184 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1185 struct vxlan_dev *vxlan;
8f84985f 1186 struct pcpu_sw_netstats *stats;
e4c7ed41 1187 union vxlan_addr saddr;
5cfccc5a 1188 __u32 vni;
e4c7ed41
CW		 1189 int err = 0;
1190 union vxlan_addr *remote_ip;
5cfccc5a
PS		 1191
1192 vni = ntohl(vx_vni) >> 8;
1193 /* Is this VNI defined? */
1194 vxlan = vxlan_vs_find_vni(vs, vni);
1195 if (!vxlan)
d342894c 1196 goto drop;
d342894c 1197
e4c7ed41 1198 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1199 skb_reset_mac_header(skb);
f01ec1c0 1200 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1201 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 1202
1203 /* Ignore packet loops (and multicast echo) */
7367d0b5 1204 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1205 goto drop;
1206
7ce04758 1207 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1208 if (remote_ip->sa.sa_family == AF_INET) {
1209 oip = ip_hdr(skb);
1210 saddr.sin.sin_addr.s_addr = oip->saddr;
1211 saddr.sa.sa_family = AF_INET;
1212#if IS_ENABLED(CONFIG_IPV6)
1213 } else {
1214 oip6 = ipv6_hdr(skb);
1215 saddr.sin6.sin6_addr = oip6->saddr;
1216 saddr.sa.sa_family = AF_INET6;
1217#endif
1218 }
1219
26a41ae6 1220 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1221 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1222 goto drop;
d342894c 1223
d342894c 1224 skb_reset_network_header(skb);
0afb1666 1225
e4c7ed41
CW		 1226 if (oip6)
1227 err = IP6_ECN_decapsulate(oip6, skb);
1228 if (oip)
1229 err = IP_ECN_decapsulate(oip, skb);
1230
d342894c 1231 if (unlikely(err)) {
e4c7ed41
CW		 1232 if (log_ecn_error) {
1233 if (oip6)
1234 net_info_ratelimited("non-ECT from %pI6\n",
1235 &oip6->saddr);
1236 if (oip)
1237 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1238 &oip->saddr, oip->tos);
1239 }
d342894c 1240 if (err > 1) {
1241 ++vxlan->dev->stats.rx_frame_errors;
1242 ++vxlan->dev->stats.rx_errors;
1243 goto drop;
1244 }
1245 }
1246
e8171045 1247 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1248 u64_stats_update_begin(&stats->syncp);
1249 stats->rx_packets++;
1250 stats->rx_bytes += skb->len;
1251 u64_stats_update_end(&stats->syncp);
1252
1253 netif_rx(skb);
1254
5cfccc5a 1255 return;
d342894c 1256drop:
1257 /* Consume bad packet */
1258 kfree_skb(skb);
d342894c 1259}
1260
e4f67add
DS		 1261static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1262{
1263 struct vxlan_dev *vxlan = netdev_priv(dev);
1264 struct arphdr *parp;
1265 u8 *arpptr, *sha;
1266 __be32 sip, tip;
1267 struct neighbour *n;
1268
1269 if (dev->flags & IFF_NOARP)
1270 goto out;
1271
1272 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1273 dev->stats.tx_dropped++;
1274 goto out;
1275 }
1276 parp = arp_hdr(skb);
1277
1278 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1279 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1280 parp->ar_pro != htons(ETH_P_IP) ||
1281 parp->ar_op != htons(ARPOP_REQUEST) ||
1282 parp->ar_hln != dev->addr_len ||
1283 parp->ar_pln != 4)
1284 goto out;
1285 arpptr = (u8 *)parp + sizeof(struct arphdr);
1286 sha = arpptr;
1287 arpptr += dev->addr_len; /* sha */
1288 memcpy(&sip, arpptr, sizeof(sip));
1289 arpptr += sizeof(sip);
1290 arpptr += dev->addr_len; /* tha */
1291 memcpy(&tip, arpptr, sizeof(tip));
1292
1293 if (ipv4_is_loopback(tip) ||
1294 ipv4_is_multicast(tip))
1295 goto out;
1296
1297 n = neigh_lookup(&arp_tbl, &tip, dev);
1298
1299 if (n) {
e4f67add
DS		 1300 struct vxlan_fdb *f;
1301 struct sk_buff *reply;
1302
1303 if (!(n->nud_state & NUD_CONNECTED)) {
1304 neigh_release(n);
1305 goto out;
1306 }
1307
1308 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1309 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1310 /* bridge-local neighbor */
1311 neigh_release(n);
1312 goto out;
1313 }
1314
1315 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1316 n->ha, sha);
1317
1318 neigh_release(n);
1319
7346135d
DS		 1320 if (reply == NULL)
1321 goto out;
1322
e4f67add
DS		 1323 skb_reset_mac_header(reply);
1324 __skb_pull(reply, skb_network_offset(reply));
1325 reply->ip_summed = CHECKSUM_UNNECESSARY;
1326 reply->pkt_type = PACKET_HOST;
1327
1328 if (netif_rx_ni(reply) == NET_RX_DROP)
1329 dev->stats.rx_dropped++;
e4c7ed41
CW		 1330 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1331 union vxlan_addr ipa = {
1332 .sin.sin_addr.s_addr = tip,
1333 .sa.sa_family = AF_INET,
1334 };
1335
1336 vxlan_ip_miss(dev, &ipa);
1337 }
e4f67add
DS		 1338out:
1339 consume_skb(skb);
1340 return NETDEV_TX_OK;
1341}
1342
f564f45c 1343#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1344
1345static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1346 struct neighbour *n, bool isrouter)
1347{
1348 struct net_device *dev = request->dev;
1349 struct sk_buff *reply;
1350 struct nd_msg *ns, *na;
1351 struct ipv6hdr *pip6;
1352 u8 *daddr;
1353 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1354 int ns_olen;
1355 int i, len;
1356
1357 if (dev == NULL)
1358 return NULL;
1359
1360 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1361 sizeof(*na) + na_olen + dev->needed_tailroom;
1362 reply = alloc_skb(len, GFP_ATOMIC);
1363 if (reply == NULL)
1364 return NULL;
1365
1366 reply->protocol = htons(ETH_P_IPV6);
1367 reply->dev = dev;
1368 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1369 skb_push(reply, sizeof(struct ethhdr));
1370 skb_set_mac_header(reply, 0);
1371
1372 ns = (struct nd_msg *)skb_transport_header(request);
1373
1374 daddr = eth_hdr(request)->h_source;
1375 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1376 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1377 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1378 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1379 break;
1380 }
1381 }
1382
1383 /* Ethernet header */
1384 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1385 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1386 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1387 reply->protocol = htons(ETH_P_IPV6);
1388
1389 skb_pull(reply, sizeof(struct ethhdr));
1390 skb_set_network_header(reply, 0);
1391 skb_put(reply, sizeof(struct ipv6hdr));
1392
1393 /* IPv6 header */
1394
1395 pip6 = ipv6_hdr(reply);
1396 memset(pip6, 0, sizeof(struct ipv6hdr));
1397 pip6->version = 6;
1398 pip6->priority = ipv6_hdr(request)->priority;
1399 pip6->nexthdr = IPPROTO_ICMPV6;
1400 pip6->hop_limit = 255;
1401 pip6->daddr = ipv6_hdr(request)->saddr;
1402 pip6->saddr = *(struct in6_addr *)n->primary_key;
1403
1404 skb_pull(reply, sizeof(struct ipv6hdr));
1405 skb_set_transport_header(reply, 0);
1406
1407 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1408
1409 /* Neighbor Advertisement */
1410 memset(na, 0, sizeof(*na)+na_olen);
1411 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1412 na->icmph.icmp6_router = isrouter;
1413 na->icmph.icmp6_override = 1;
1414 na->icmph.icmp6_solicited = 1;
1415 na->target = ns->target;
1416 ether_addr_copy(&na->opt[2], n->ha);
1417 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1418 na->opt[1] = na_olen >> 3;
1419
1420 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1421 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1422 csum_partial(na, sizeof(*na)+na_olen, 0));
1423
1424 pip6->payload_len = htons(sizeof(*na)+na_olen);
1425
1426 skb_push(reply, sizeof(struct ipv6hdr));
1427
1428 reply->ip_summed = CHECKSUM_UNNECESSARY;
1429
1430 return reply;
1431}
1432
f564f45c
CW		 1433static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1434{
1435 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1436 struct nd_msg *msg;
f564f45c
CW		 1437 const struct ipv6hdr *iphdr;
1438 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1439 struct neighbour *n;
1440 struct inet6_dev *in6_dev;
f564f45c
CW		 1441
1442 in6_dev = __in6_dev_get(dev);
1443 if (!in6_dev)
1444 goto out;
1445
1446 if (!pskb_may_pull(skb, skb->len))
1447 goto out;
1448
1449 iphdr = ipv6_hdr(skb);
1450 saddr = &iphdr->saddr;
1451 daddr = &iphdr->daddr;
1452
f564f45c
CW		 1453 msg = (struct nd_msg *)skb_transport_header(skb);
1454 if (msg->icmph.icmp6_code != 0 ||
1455 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1456 goto out;
1457
4b29dba9
DS		 1458 if (ipv6_addr_loopback(daddr) ||
1459 ipv6_addr_is_multicast(&msg->target))
1460 goto out;
1461
1462 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1463
1464 if (n) {
1465 struct vxlan_fdb *f;
4b29dba9 1466 struct sk_buff *reply;
f564f45c
CW		 1467
1468 if (!(n->nud_state & NUD_CONNECTED)) {
1469 neigh_release(n);
1470 goto out;
1471 }
1472
1473 f = vxlan_find_mac(vxlan, n->ha);
1474 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1475 /* bridge-local neighbor */
1476 neigh_release(n);
1477 goto out;
1478 }
1479
4b29dba9
DS		 1480 reply = vxlan_na_create(skb, n,
1481 !!(f ? f->flags & NTF_ROUTER : 0));
1482
f564f45c 1483 neigh_release(n);
4b29dba9
DS		 1484
1485 if (reply == NULL)
1486 goto out;
1487
1488 if (netif_rx_ni(reply) == NET_RX_DROP)
1489 dev->stats.rx_dropped++;
1490
f564f45c 1491 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1492 union vxlan_addr ipa = {
1493 .sin6.sin6_addr = msg->target,
1494 .sa.sa_family = AF_INET6,
1495 };
1496
f564f45c
CW		 1497 vxlan_ip_miss(dev, &ipa);
1498 }
1499
1500out:
1501 consume_skb(skb);
1502 return NETDEV_TX_OK;
1503}
1504#endif
1505
e4f67add
DS		 1506static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1507{
1508 struct vxlan_dev *vxlan = netdev_priv(dev);
1509 struct neighbour *n;
e4f67add
DS		 1510
1511 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1512 return false;
1513
1514 n = NULL;
1515 switch (ntohs(eth_hdr(skb)->h_proto)) {
1516 case ETH_P_IP:
e15a00aa
CW		 1517 {
1518 struct iphdr *pip;
1519
e4f67add
DS		 1520 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1521 return false;
1522 pip = ip_hdr(skb);
1523 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1524 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1525 union vxlan_addr ipa = {
1526 .sin.sin_addr.s_addr = pip->daddr,
1527 .sa.sa_family = AF_INET,
1528 };
1529
1530 vxlan_ip_miss(dev, &ipa);
1531 return false;
1532 }
1533
e4f67add 1534 break;
e15a00aa
CW		 1535 }
1536#if IS_ENABLED(CONFIG_IPV6)
1537 case ETH_P_IPV6:
1538 {
1539 struct ipv6hdr *pip6;
1540
1541 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1542 return false;
1543 pip6 = ipv6_hdr(skb);
1544 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1545 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1546 union vxlan_addr ipa = {
1547 .sin6.sin6_addr = pip6->daddr,
1548 .sa.sa_family = AF_INET6,
1549 };
1550
1551 vxlan_ip_miss(dev, &ipa);
1552 return false;
1553 }
1554
1555 break;
1556 }
1557#endif
e4f67add
DS		 1558 default:
1559 return false;
1560 }
1561
1562 if (n) {
1563 bool diff;
1564
7367d0b5 1565 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1566 if (diff) {
1567 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1568 dev->addr_len);
1569 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1570 }
1571 neigh_release(n);
1572 return diff;
e4c7ed41
CW		 1573 }
1574
e4f67add
DS		 1575 return false;
1576}
1577
05f47d69 1578/* Compute source port for outgoing packet
1579 * first choice to use L4 flow hash since it will spread
1580 * better and maybe available from hardware
1581 * secondary choice is to use jhash on the Ethernet header
1582 */
49560532 1583__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
05f47d69 1584{
49560532 1585 unsigned int range = (port_max - port_min) + 1;
05f47d69 1586 u32 hash;
1587
3958afa1 1588 hash = skb_get_hash(skb);
05f47d69 1589 if (!hash)
1590 hash = jhash(skb->data, 2 * ETH_ALEN,
1591 (__force u32) skb->protocol);
1592
49560532 1593 return htons((((u64) hash * range) >> 32) + port_min);
05f47d69 1594}
49560532 1595EXPORT_SYMBOL_GPL(vxlan_src_port);
05f47d69 1596
359a0ea9
TH		 1597static inline struct sk_buff *vxlan_handle_offloads(struct sk_buff *skb,
1598 bool udp_csum)
05c0db08 1599{
359a0ea9
TH		 1600 int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1601 return iptunnel_handle_offloads(skb, udp_csum, type);
05c0db08
PS		 1602}
1603
e4c7ed41 1604#if IS_ENABLED(CONFIG_IPV6)
11796187 1605static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1606 struct dst_entry *dst, struct sk_buff *skb,
1607 struct net_device *dev, struct in6_addr *saddr,
1608 struct in6_addr *daddr, __u8 prio, __u8 ttl,
f01ec1c0
ND		 1609 __be16 src_port, __be16 dst_port, __be32 vni,
1610 bool xnet)
e4c7ed41
CW		 1611{
1612 struct ipv6hdr *ip6h;
1613 struct vxlanhdr *vxh;
1614 struct udphdr *uh;
1615 int min_headroom;
1616 int err;
1617
359a0ea9
TH		 1618 skb = vxlan_handle_offloads(skb, !udp_get_no_check6_tx(vs->sock->sk));
1619 if (IS_ERR(skb))
1620 return -EINVAL;
e4c7ed41 1621
f01ec1c0 1622 skb_scrub_packet(skb, xnet);
963a88b3 1623
e4c7ed41
CW		 1624 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1625 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1626 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1627
1628 /* Need space for new headers (invalidates iph ptr) */
1629 err = skb_cow_head(skb, min_headroom);
1630 if (unlikely(err))
1631 return err;
1632
1633 if (vlan_tx_tag_present(skb)) {
1634 if (WARN_ON(!__vlan_put_tag(skb,
1635 skb->vlan_proto,
1636 vlan_tx_tag_get(skb))))
1637 return -ENOMEM;
1638
1639 skb->vlan_tci = 0;
1640 }
1641
1642 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1643 vxh->vx_flags = htonl(VXLAN_FLAGS);
1644 vxh->vx_vni = vni;
1645
1646 __skb_push(skb, sizeof(*uh));
1647 skb_reset_transport_header(skb);
1648 uh = udp_hdr(skb);
1649
1650 uh->dest = dst_port;
1651 uh->source = src_port;
1652
1653 uh->len = htons(skb->len);
e4c7ed41
CW		 1654
1655 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1656 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1657 IPSKB_REROUTED);
e4c7ed41
CW		 1658 skb_dst_set(skb, dst);
1659
359a0ea9
TH		 1660 udp6_set_csum(udp_get_no_check6_tx(vs->sock->sk), skb,
1661 saddr, daddr, skb->len);
e4c7ed41
CW		 1662
1663 __skb_push(skb, sizeof(*ip6h));
1664 skb_reset_network_header(skb);
1665 ip6h = ipv6_hdr(skb);
1666 ip6h->version = 6;
1667 ip6h->priority = prio;
1668 ip6h->flow_lbl[0] = 0;
1669 ip6h->flow_lbl[1] = 0;
1670 ip6h->flow_lbl[2] = 0;
1671 ip6h->payload_len = htons(skb->len);
1672 ip6h->nexthdr = IPPROTO_UDP;
1673 ip6h->hop_limit = ttl;
1674 ip6h->daddr = *daddr;
1675 ip6h->saddr = *saddr;
1676
e4c7ed41
CW		 1677 ip6tunnel_xmit(skb, dev);
1678 return 0;
1679}
1680#endif
1681
11796187 1682int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1683 struct rtable *rt, struct sk_buff *skb,
1684 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
f01ec1c0 1685 __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
49560532
PS		 1686{
1687 struct vxlanhdr *vxh;
1688 struct udphdr *uh;
649c5b8b 1689 int min_headroom;
49560532
PS		 1690 int err;
1691
359a0ea9
TH		 1692 skb = vxlan_handle_offloads(skb, !vs->sock->sk->sk_no_check_tx);
1693 if (IS_ERR(skb))
1694 return -EINVAL;
49560532 1695
649c5b8b 1696 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178
PS		 1697 + VXLAN_HLEN + sizeof(struct iphdr)
1698 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1699
1700 /* Need space for new headers (invalidates iph ptr) */
1701 err = skb_cow_head(skb, min_headroom);
1702 if (unlikely(err))
1703 return err;
1704
1eaa8178
PS		 1705 if (vlan_tx_tag_present(skb)) {
1706 if (WARN_ON(!__vlan_put_tag(skb,
1707 skb->vlan_proto,
1708 vlan_tx_tag_get(skb))))
1709 return -ENOMEM;
1710
1711 skb->vlan_tci = 0;
1712 }
1713
49560532
PS		 1714 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1715 vxh->vx_flags = htonl(VXLAN_FLAGS);
1716 vxh->vx_vni = vni;
1717
1718 __skb_push(skb, sizeof(*uh));
1719 skb_reset_transport_header(skb);
1720 uh = udp_hdr(skb);
1721
1722 uh->dest = dst_port;
1723 uh->source = src_port;
1724
1725 uh->len = htons(skb->len);
49560532 1726
359a0ea9
TH		 1727 udp_set_csum(vs->sock->sk->sk_no_check_tx, skb,
1728 src, dst, skb->len);
49560532 1729
aad88724 1730 return iptunnel_xmit(vs->sock->sk, rt, skb, src, dst, IPPROTO_UDP,
f01ec1c0 1731 tos, ttl, df, xnet);
49560532
PS		 1732}
1733EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1734
9dcc71e1
SS		 1735/* Bypass encapsulation if the destination is local */
1736static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1737 struct vxlan_dev *dst_vxlan)
1738{
8f84985f 1739 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1740 union vxlan_addr loopback;
1741 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
9dcc71e1 1742
8f84985f
LR		 1743 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1744 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1745 skb->pkt_type = PACKET_HOST;
1746 skb->encapsulation = 0;
1747 skb->dev = dst_vxlan->dev;
1748 __skb_pull(skb, skb_network_offset(skb));
1749
e4c7ed41
CW		 1750 if (remote_ip->sa.sa_family == AF_INET) {
1751 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1752 loopback.sa.sa_family = AF_INET;
1753#if IS_ENABLED(CONFIG_IPV6)
1754 } else {
1755 loopback.sin6.sin6_addr = in6addr_loopback;
1756 loopback.sa.sa_family = AF_INET6;
1757#endif
1758 }
1759
9dcc71e1 1760 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1761 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1762
1763 u64_stats_update_begin(&tx_stats->syncp);
1764 tx_stats->tx_packets++;
1765 tx_stats->tx_bytes += skb->len;
1766 u64_stats_update_end(&tx_stats->syncp);
1767
1768 if (netif_rx(skb) == NET_RX_SUCCESS) {
1769 u64_stats_update_begin(&rx_stats->syncp);
1770 rx_stats->rx_packets++;
1771 rx_stats->rx_bytes += skb->len;
1772 u64_stats_update_end(&rx_stats->syncp);
1773 } else {
1774 skb->dev->stats.rx_dropped++;
1775 }
1776}
1777
4ad16930
SH		 1778static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1779 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1780{
1781 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1782 struct rtable *rt = NULL;
d342894c 1783 const struct iphdr *old_iph;
d342894c 1784 struct flowi4 fl4;
e4c7ed41
CW		 1785 union vxlan_addr *dst;
1786 __be16 src_port = 0, dst_port;
234f5b73 1787 u32 vni;
d342894c 1788 __be16 df = 0;
1789 __u8 tos, ttl;
0e6fbc5b 1790 int err;
d342894c 1791
823aa873 1792 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1793 vni = rdst->remote_vni;
e4c7ed41 1794 dst = &rdst->remote_ip;
e4f67add 1795
e4c7ed41 1796 if (vxlan_addr_any(dst)) {
e4f67add 1797 if (did_rsc) {
e4f67add 1798 /* short-circuited back to local bridge */
9dcc71e1 1799 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1800 return;
e4f67add 1801 }
ef59febe 1802 goto drop;
e4f67add 1803 }
ef59febe 1804
d342894c 1805 old_iph = ip_hdr(skb);
1806
d342894c 1807 ttl = vxlan->ttl;
e4c7ed41 1808 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1809 ttl = 1;
1810
1811 tos = vxlan->tos;
1812 if (tos == 1)
206aaafc 1813 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1814
49560532 1815 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
d342894c 1816
e4c7ed41
CW		 1817 if (dst->sa.sa_family == AF_INET) {
1818 memset(&fl4, 0, sizeof(fl4));
1819 fl4.flowi4_oif = rdst->remote_ifindex;
1820 fl4.flowi4_tos = RT_TOS(tos);
1821 fl4.daddr = dst->sin.sin_addr.s_addr;
1822 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1823
f01ec1c0 1824 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1825 if (IS_ERR(rt)) {
1826 netdev_dbg(dev, "no route to %pI4\n",
1827 &dst->sin.sin_addr.s_addr);
1828 dev->stats.tx_carrier_errors++;
1829 goto tx_error;
1830 }
d342894c 1831
e4c7ed41
CW		 1832 if (rt->dst.dev == dev) {
1833 netdev_dbg(dev, "circular route to %pI4\n",
1834 &dst->sin.sin_addr.s_addr);
1835 dev->stats.collisions++;
fffc15a5 1836 goto rt_tx_error;
e4c7ed41
CW		 1837 }
1838
1839 /* Bypass encapsulation if the destination is local */
1840 if (rt->rt_flags & RTCF_LOCAL &&
1841 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1842 struct vxlan_dev *dst_vxlan;
1843
1844 ip_rt_put(rt);
f01ec1c0 1845 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
e4c7ed41
CW		 1846 if (!dst_vxlan)
1847 goto tx_error;
1848 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1849 return;
1850 }
1851
1852 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1853 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
d342894c 1854
11796187 1855 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
e4c7ed41
CW		 1856 fl4.saddr, dst->sin.sin_addr.s_addr,
1857 tos, ttl, df, src_port, dst_port,
f01ec1c0
ND		 1858 htonl(vni << 8),
1859 !net_eq(vxlan->net, dev_net(vxlan->dev)));
9dcc71e1 1860
e4c7ed41
CW		 1861 if (err < 0)
1862 goto rt_tx_error;
1863 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1864#if IS_ENABLED(CONFIG_IPV6)
1865 } else {
1866 struct sock *sk = vxlan->vn_sock->sock->sk;
1867 struct dst_entry *ndst;
1868 struct flowi6 fl6;
1869 u32 flags;
1870
1871 memset(&fl6, 0, sizeof(fl6));
1872 fl6.flowi6_oif = rdst->remote_ifindex;
1873 fl6.daddr = dst->sin6.sin6_addr;
1874 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1875 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1876
1877 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1878 netdev_dbg(dev, "no route to %pI6\n",
1879 &dst->sin6.sin6_addr);
1880 dev->stats.tx_carrier_errors++;
9dcc71e1 1881 goto tx_error;
e4c7ed41 1882 }
d342894c 1883
e4c7ed41
CW		 1884 if (ndst->dev == dev) {
1885 netdev_dbg(dev, "circular route to %pI6\n",
1886 &dst->sin6.sin6_addr);
1887 dst_release(ndst);
1888 dev->stats.collisions++;
1889 goto tx_error;
1890 }
0e6fbc5b 1891
e4c7ed41
CW		 1892 /* Bypass encapsulation if the destination is local */
1893 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1894 if (flags & RTF_LOCAL &&
1895 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1896 struct vxlan_dev *dst_vxlan;
1897
1898 dst_release(ndst);
f01ec1c0 1899 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
e4c7ed41
CW		 1900 if (!dst_vxlan)
1901 goto tx_error;
1902 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1903 return;
1904 }
49560532 1905
e4c7ed41
CW		 1906 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1907
11796187 1908 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41 1909 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
f01ec1c0
ND		 1910 src_port, dst_port, htonl(vni << 8),
1911 !net_eq(vxlan->net, dev_net(vxlan->dev)));
e4c7ed41
CW		 1912#endif
1913 }
0e6fbc5b 1914
4ad16930 1915 return;
d342894c 1916
1917drop:
1918 dev->stats.tx_dropped++;
1919 goto tx_free;
1920
49560532
PS		 1921rt_tx_error:
1922 ip_rt_put(rt);
d342894c 1923tx_error:
1924 dev->stats.tx_errors++;
1925tx_free:
1926 dev_kfree_skb(skb);
d342894c 1927}
1928
6681712d
DS		 1929/* Transmit local packets over Vxlan
1930 *
1931 * Outer IP header inherits ECN and DF from inner header.
1932 * Outer UDP destination is the VXLAN assigned port.
1933 * source port is based on hash of flow
1934 */
1935static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1936{
1937 struct vxlan_dev *vxlan = netdev_priv(dev);
1938 struct ethhdr *eth;
1939 bool did_rsc = false;
8f646c92 1940 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 1941 struct vxlan_fdb *f;
6681712d
DS		 1942
1943 skb_reset_mac_header(skb);
1944 eth = eth_hdr(skb);
1945
f564f45c
CW		 1946 if ((vxlan->flags & VXLAN_F_PROXY)) {
1947 if (ntohs(eth->h_proto) == ETH_P_ARP)
1948 return arp_reduce(dev, skb);
1949#if IS_ENABLED(CONFIG_IPV6)
1950 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1951 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1952 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1953 struct nd_msg *msg;
1954
1955 msg = (struct nd_msg *)skb_transport_header(skb);
1956 if (msg->icmph.icmp6_code == 0 &&
1957 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1958 return neigh_reduce(dev, skb);
1959 }
1960#endif
1961 }
6681712d
DS		 1962
1963 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1964 did_rsc = false;
1965
1966 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 1967 (ntohs(eth->h_proto) == ETH_P_IP ||
1968 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 1969 did_rsc = route_shortcircuit(dev, skb);
1970 if (did_rsc)
1971 f = vxlan_find_mac(vxlan, eth->h_dest);
1972 }
1973
6681712d 1974 if (f == NULL) {
afbd8bae
MR		 1975 f = vxlan_find_mac(vxlan, all_zeros_mac);
1976 if (f == NULL) {
1977 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1978 !is_multicast_ether_addr(eth->h_dest))
1979 vxlan_fdb_miss(vxlan, eth->h_dest);
1980
1981 dev->stats.tx_dropped++;
8f646c92 1982 kfree_skb(skb);
afbd8bae
MR		 1983 return NETDEV_TX_OK;
1984 }
1985 }
6681712d 1986
afbd8bae
MR		 1987 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1988 struct sk_buff *skb1;
6681712d 1989
8f646c92
ED		 1990 if (!fdst) {
1991 fdst = rdst;
1992 continue;
1993 }
afbd8bae
MR		 1994 skb1 = skb_clone(skb, GFP_ATOMIC);
1995 if (skb1)
1996 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1997 }
1998
8f646c92
ED		 1999 if (fdst)
2000 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2001 else
2002 kfree_skb(skb);
4ad16930 2003 return NETDEV_TX_OK;
6681712d
DS		 2004}
2005
d342894c 2006/* Walk the forwarding table and purge stale entries */
2007static void vxlan_cleanup(unsigned long arg)
2008{
2009 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2010 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2011 unsigned int h;
2012
2013 if (!netif_running(vxlan->dev))
2014 return;
2015
2016 spin_lock_bh(&vxlan->hash_lock);
2017 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2018 struct hlist_node *p, *n;
2019 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2020 struct vxlan_fdb *f
2021 = container_of(p, struct vxlan_fdb, hlist);
2022 unsigned long timeout;
2023
3c172868 2024 if (f->state & NUD_PERMANENT)
d342894c 2025 continue;
2026
2027 timeout = f->used + vxlan->age_interval * HZ;
2028 if (time_before_eq(timeout, jiffies)) {
2029 netdev_dbg(vxlan->dev,
2030 "garbage collect %pM\n",
2031 f->eth_addr);
2032 f->state = NUD_STALE;
2033 vxlan_fdb_destroy(vxlan, f);
2034 } else if (time_before(timeout, next_timer))
2035 next_timer = timeout;
2036 }
2037 }
2038 spin_unlock_bh(&vxlan->hash_lock);
2039
2040 mod_timer(&vxlan->age_timer, next_timer);
2041}
2042
9c2e24e1
PS		 2043static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2044{
2045 __u32 vni = vxlan->default_dst.remote_vni;
2046
2047 vxlan->vn_sock = vs;
2048 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2049}
2050
d342894c 2051/* Setup stats when device is created */
2052static int vxlan_init(struct net_device *dev)
2053{
1c51a915 2054 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2055 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2056 struct vxlan_sock *vs;
1c51a915 2057
1c213bd2 2058 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2059 if (!dev->tstats)
d342894c 2060 return -ENOMEM;
2061
1c51a915 2062 spin_lock(&vn->sock_lock);
f01ec1c0 2063 vs = vxlan_find_sock(vxlan->net, vxlan->dst_port);
1c51a915
SH		 2064 if (vs) {
2065 /* If we have a socket with same port already, reuse it */
2066 atomic_inc(&vs->refcnt);
9c2e24e1 2067 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 2068 } else {
2069 /* otherwise make new socket outside of RTNL */
2070 dev_hold(dev);
2071 queue_work(vxlan_wq, &vxlan->sock_work);
2072 }
2073 spin_unlock(&vn->sock_lock);
2074
d342894c 2075 return 0;
2076}
2077
ba609e9b 2078static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2079{
2080 struct vxlan_fdb *f;
2081
2082 spin_lock_bh(&vxlan->hash_lock);
2083 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2084 if (f)
2085 vxlan_fdb_destroy(vxlan, f);
2086 spin_unlock_bh(&vxlan->hash_lock);
2087}
2088
ebf4063e
SH		 2089static void vxlan_uninit(struct net_device *dev)
2090{
2091 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2092 struct vxlan_sock *vs = vxlan->vn_sock;
2093
ba609e9b 2094 vxlan_fdb_delete_default(vxlan);
afbd8bae 2095
ebf4063e 2096 if (vs)
012a5729 2097 vxlan_sock_release(vs);
ebf4063e
SH		 2098 free_percpu(dev->tstats);
2099}
2100
d342894c 2101/* Start ageing timer and join group when device is brought up */
2102static int vxlan_open(struct net_device *dev)
2103{
2104 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2105 struct vxlan_sock *vs = vxlan->vn_sock;
2106
2107 /* socket hasn't been created */
2108 if (!vs)
2109 return -ENOTCONN;
d342894c 2110
79d4a94f 2111 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2112 vxlan_sock_hold(vs);
7c47cedf 2113 dev_hold(dev);
3fc2de2f 2114 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2115 }
2116
2117 if (vxlan->age_interval)
2118 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2119
2120 return 0;
2121}
2122
2123/* Purge the forwarding table */
2124static void vxlan_flush(struct vxlan_dev *vxlan)
2125{
31fec5aa 2126 unsigned int h;
d342894c 2127
2128 spin_lock_bh(&vxlan->hash_lock);
2129 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2130 struct hlist_node *p, *n;
2131 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2132 struct vxlan_fdb *f
2133 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2134 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2135 if (!is_zero_ether_addr(f->eth_addr))
2136 vxlan_fdb_destroy(vxlan, f);
d342894c 2137 }
2138 }
2139 spin_unlock_bh(&vxlan->hash_lock);
2140}
2141
2142/* Cleanup timer and forwarding table on shutdown */
2143static int vxlan_stop(struct net_device *dev)
2144{
2145 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2146 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2147 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2148
e4c7ed41 2149 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2150 !vxlan_group_used(vn, vxlan)) {
1c51a915 2151 vxlan_sock_hold(vs);
7c47cedf 2152 dev_hold(dev);
3fc2de2f 2153 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2154 }
d342894c 2155
2156 del_timer_sync(&vxlan->age_timer);
2157
2158 vxlan_flush(vxlan);
2159
2160 return 0;
2161}
2162
d342894c 2163/* Stub, nothing needs to be done. */
2164static void vxlan_set_multicast_list(struct net_device *dev)
2165{
2166}
2167
345010b5
DB		 2168static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2169{
2170 struct vxlan_dev *vxlan = netdev_priv(dev);
2171 struct vxlan_rdst *dst = &vxlan->default_dst;
2172 struct net_device *lowerdev;
2173 int max_mtu;
2174
f01ec1c0 2175 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2176 if (lowerdev == NULL)
2177 return eth_change_mtu(dev, new_mtu);
2178
2179 if (dst->remote_ip.sa.sa_family == AF_INET6)
2180 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2181 else
2182 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2183
2184 if (new_mtu < 68 || new_mtu > max_mtu)
2185 return -EINVAL;
2186
2187 dev->mtu = new_mtu;
2188 return 0;
2189}
2190
d342894c 2191static const struct net_device_ops vxlan_netdev_ops = {
2192 .ndo_init = vxlan_init,
ebf4063e 2193 .ndo_uninit = vxlan_uninit,
d342894c 2194 .ndo_open = vxlan_open,
2195 .ndo_stop = vxlan_stop,
2196 .ndo_start_xmit = vxlan_xmit,
e8171045 2197 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2198 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2199 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2200 .ndo_validate_addr = eth_validate_addr,
2201 .ndo_set_mac_address = eth_mac_addr,
2202 .ndo_fdb_add = vxlan_fdb_add,
2203 .ndo_fdb_del = vxlan_fdb_delete,
2204 .ndo_fdb_dump = vxlan_fdb_dump,
2205};
2206
2207/* Info for udev, that this is a virtual tunnel endpoint */
2208static struct device_type vxlan_type = {
2209 .name = "vxlan",
2210};
2211
53cf5275 2212/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2213 * supply the listening VXLAN udp ports. Callers are expected
2214 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2215 */
2216void vxlan_get_rx_port(struct net_device *dev)
2217{
2218 struct vxlan_sock *vs;
2219 struct net *net = dev_net(dev);
2220 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2221 sa_family_t sa_family;
35e42379
JG		 2222 __be16 port;
2223 unsigned int i;
53cf5275
JG		 2224
2225 spin_lock(&vn->sock_lock);
2226 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2227 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2228 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2229 sa_family = vs->sock->sk->sk_family;
2230 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2231 port);
2232 }
2233 }
2234 spin_unlock(&vn->sock_lock);
2235}
2236EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2237
d342894c 2238/* Initialize the device structure. */
2239static void vxlan_setup(struct net_device *dev)
2240{
2241 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2242 unsigned int h;
05f47d69 2243 int low, high;
d342894c 2244
2245 eth_hw_addr_random(dev);
2246 ether_setup(dev);
e4c7ed41
CW		 2247 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2248 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
2249 else
2250 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2251
2252 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2253 dev->destructor = free_netdev;
d342894c 2254 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2255
2256 dev->tx_queue_len = 0;
2257 dev->features |= NETIF_F_LLTX;
d6727fe3 2258 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2259 dev->features |= NETIF_F_RXCSUM;
05c0db08 2260 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2261
1eaa8178
PS		 2262 dev->vlan_features = dev->features;
2263 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2264 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2265 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2266 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
d342894c 2267 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 2268 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2269
553675fb 2270 INIT_LIST_HEAD(&vxlan->next);
d342894c 2271 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2272 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2273 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2274 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2275
2276 init_timer_deferrable(&vxlan->age_timer);
2277 vxlan->age_timer.function = vxlan_cleanup;
2278 vxlan->age_timer.data = (unsigned long) vxlan;
2279
0bbf87d8 2280 inet_get_local_port_range(dev_net(dev), &low, &high);
05f47d69 2281 vxlan->port_min = low;
2282 vxlan->port_max = high;
823aa873 2283 vxlan->dst_port = htons(vxlan_port);
05f47d69 2284
d342894c 2285 vxlan->dev = dev;
2286
2287 for (h = 0; h < FDB_HASH_SIZE; ++h)
2288 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2289}
2290
2291static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2292 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2293 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2294 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2295 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2296 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2297 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2298 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2299 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2300 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2301 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2302 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2303 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2304 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2305 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2306 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2307 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2308 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 2309};
2310
2311static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2312{
2313 if (tb[IFLA_ADDRESS]) {
2314 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2315 pr_debug("invalid link address (not ethernet)\n");
2316 return -EINVAL;
2317 }
2318
2319 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2320 pr_debug("invalid all zero ethernet address\n");
2321 return -EADDRNOTAVAIL;
2322 }
2323 }
2324
2325 if (!data)
2326 return -EINVAL;
2327
2328 if (data[IFLA_VXLAN_ID]) {
2329 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2330 if (id >= VXLAN_VID_MASK)
2331 return -ERANGE;
2332 }
2333
05f47d69 2334 if (data[IFLA_VXLAN_PORT_RANGE]) {
2335 const struct ifla_vxlan_port_range *p
2336 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2337
2338 if (ntohs(p->high) < ntohs(p->low)) {
2339 pr_debug("port range %u .. %u not valid\n",
2340 ntohs(p->low), ntohs(p->high));
2341 return -EINVAL;
2342 }
2343 }
2344
d342894c 2345 return 0;
2346}
2347
1b13c97f
YB		 2348static void vxlan_get_drvinfo(struct net_device *netdev,
2349 struct ethtool_drvinfo *drvinfo)
2350{
2351 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2352 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2353}
2354
2355static const struct ethtool_ops vxlan_ethtool_ops = {
2356 .get_drvinfo = vxlan_get_drvinfo,
2357 .get_link = ethtool_op_get_link,
2358};
2359
553675fb 2360static void vxlan_del_work(struct work_struct *work)
2361{
2362 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2363
2364 sk_release_kernel(vs->sock->sk);
2365 kfree_rcu(vs, rcu);
2366}
2367
e4c7ed41
CW		 2368#if IS_ENABLED(CONFIG_IPV6)
2369/* Create UDP socket for encapsulation receive. AF_INET6 socket
2370 * could be used for both IPv4 and IPv6 communications, but
2371 * users may set bindv6only=1.
2372 */
359a0ea9 2373static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
553675fb 2374{
553675fb 2375 struct sock *sk;
e4c7ed41
CW		 2376 struct socket *sock;
2377 struct sockaddr_in6 vxlan_addr = {
2378 .sin6_family = AF_INET6,
2379 .sin6_port = port,
2380 };
2381 int rc, val = 1;
2382
2383 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2384 if (rc < 0) {
2385 pr_debug("UDPv6 socket create failed\n");
39deb2c7 2386 return ERR_PTR(rc);
e4c7ed41
CW		 2387 }
2388
2389 /* Put in proper namespace */
2390 sk = sock->sk;
2391 sk_change_net(sk, net);
2392
2393 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2394 (char *)&val, sizeof(val));
2395 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2396 sizeof(struct sockaddr_in6));
2397 if (rc < 0) {
2398 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2399 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2400 sk_release_kernel(sk);
39deb2c7 2401 return ERR_PTR(rc);
e4c7ed41
CW		 2402 }
2403 /* At this point, IPv6 module should have been loaded in
2404 * sock_create_kern().
2405 */
2406 BUG_ON(!ipv6_stub);
2407
e4c7ed41
CW		 2408 /* Disable multicast loopback */
2409 inet_sk(sk)->mc_loop = 0;
359a0ea9
TH		 2410
2411 if (flags & VXLAN_F_UDP_ZERO_CSUM6_TX)
2412 udp_set_no_check6_tx(sk, true);
2413
2414 if (flags & VXLAN_F_UDP_ZERO_CSUM6_RX)
2415 udp_set_no_check6_rx(sk, true);
2416
39deb2c7 2417 return sock;
e4c7ed41
CW		 2418}
2419
2420#else
2421
359a0ea9 2422static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
e4c7ed41 2423{
39deb2c7 2424 return ERR_PTR(-EPFNOSUPPORT);
e4c7ed41
CW		 2425}
2426#endif
2427
359a0ea9 2428static struct socket *create_v4_sock(struct net *net, __be16 port, u32 flags)
e4c7ed41
CW		 2429{
2430 struct sock *sk;
2431 struct socket *sock;
553675fb 2432 struct sockaddr_in vxlan_addr = {
2433 .sin_family = AF_INET,
2434 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 2435 .sin_port = port,
553675fb 2436 };
2437 int rc;
553675fb 2438
2439 /* Create UDP socket for encapsulation receive. */
e4c7ed41 2440 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
553675fb 2441 if (rc < 0) {
2442 pr_debug("UDP socket create failed\n");
39deb2c7 2443 return ERR_PTR(rc);
553675fb 2444 }
2445
2446 /* Put in proper namespace */
e4c7ed41 2447 sk = sock->sk;
553675fb 2448 sk_change_net(sk, net);
2449
e4c7ed41 2450 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
553675fb 2451 sizeof(vxlan_addr));
2452 if (rc < 0) {
2453 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2454 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2455 sk_release_kernel(sk);
39deb2c7 2456 return ERR_PTR(rc);
e4c7ed41
CW		 2457 }
2458
e4c7ed41
CW		 2459 /* Disable multicast loopback */
2460 inet_sk(sk)->mc_loop = 0;
359a0ea9
TH		 2461
2462 if (!(flags & VXLAN_F_UDP_CSUM))
2463 sock->sk->sk_no_check_tx = 1;
2464
39deb2c7 2465 return sock;
e4c7ed41
CW		 2466}
2467
2468/* Create new listen socket if needed */
2469static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2470 vxlan_rcv_t *rcv, void *data,
2471 u32 flags)
e4c7ed41
CW		 2472{
2473 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2474 struct vxlan_sock *vs;
2475 struct socket *sock;
2476 struct sock *sk;
e4c7ed41 2477 unsigned int h;
359a0ea9 2478 bool ipv6 = !!(flags & VXLAN_F_IPV6);
e4c7ed41 2479
dc01e7d3 2480 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2481 if (!vs)
2482 return ERR_PTR(-ENOMEM);
2483
2484 for (h = 0; h < VNI_HASH_SIZE; ++h)
2485 INIT_HLIST_HEAD(&vs->vni_list[h]);
2486
2487 INIT_WORK(&vs->del_work, vxlan_del_work);
2488
2489 if (ipv6)
359a0ea9 2490 sock = create_v6_sock(net, port, flags);
e4c7ed41 2491 else
359a0ea9 2492 sock = create_v4_sock(net, port, flags);
39deb2c7 2493 if (IS_ERR(sock)) {
553675fb 2494 kfree(vs);
e50fddc8 2495 return ERR_CAST(sock);
553675fb 2496 }
e4c7ed41
CW		 2497
2498 vs->sock = sock;
2499 sk = sock->sk;
9c2e24e1 2500 atomic_set(&vs->refcnt, 1);
5cfccc5a 2501 vs->rcv = rcv;
012a5729 2502 vs->data = data;
559835ea 2503 rcu_assign_sk_user_data(vs->sock->sk, vs);
553675fb 2504
dc01e7d3
OG		 2505 /* Initialize the vxlan udp offloads structure */
2506 vs->udp_offloads.port = port;
2507 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2508 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2509
9c2e24e1
PS		 2510 spin_lock(&vn->sock_lock);
2511 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2512 vxlan_notify_add_rx_port(vs);
9c2e24e1 2513 spin_unlock(&vn->sock_lock);
553675fb 2514
2515 /* Mark socket as an encapsulation socket. */
2516 udp_sk(sk)->encap_type = 1;
2517 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
e4c7ed41
CW		 2518#if IS_ENABLED(CONFIG_IPV6)
2519 if (ipv6)
2520 ipv6_stub->udpv6_encap_enable();
2521 else
2522#endif
2523 udp_encap_enable();
2524
9c2e24e1
PS		 2525 return vs;
2526}
2527
012a5729
PS		 2528struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2529 vxlan_rcv_t *rcv, void *data,
359a0ea9 2530 bool no_share, u32 flags)
9c2e24e1
PS		 2531{
2532 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2533 struct vxlan_sock *vs;
2534
359a0ea9 2535 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2536 if (!IS_ERR(vs))
2537 return vs;
553675fb 2538
012a5729
PS		 2539 if (no_share) /* Return error if sharing is not allowed. */
2540 return vs;
2541
9c2e24e1
PS		 2542 spin_lock(&vn->sock_lock);
2543 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 2544 if (vs) {
2545 if (vs->rcv == rcv)
2546 atomic_inc(&vs->refcnt);
2547 else
2548 vs = ERR_PTR(-EBUSY);
2549 }
2550 spin_unlock(&vn->sock_lock);
2551
2552 if (!vs)
9c2e24e1
PS		 2553 vs = ERR_PTR(-EINVAL);
2554
553675fb 2555 return vs;
2556}
012a5729 2557EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2558
1c51a915
SH		 2559/* Scheduled at device creation to bind to a socket */
2560static void vxlan_sock_work(struct work_struct *work)
2561{
9c2e24e1 2562 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2563 struct net *net = vxlan->net;
1c51a915 2564 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2565 __be16 port = vxlan->dst_port;
2566 struct vxlan_sock *nvs;
1c51a915 2567
359a0ea9 2568 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2569 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2570 if (!IS_ERR(nvs))
2571 vxlan_vs_add_dev(nvs, vxlan);
2572 spin_unlock(&vn->sock_lock);
2573
2574 dev_put(vxlan->dev);
1c51a915
SH		 2575}
2576
d342894c 2577static int vxlan_newlink(struct net *net, struct net_device *dev,
2578 struct nlattr *tb[], struct nlattr *data[])
2579{
553675fb 2580 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2581 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2582 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2583 __u32 vni;
2584 int err;
e4c7ed41 2585 bool use_ipv6 = false;
d342894c 2586
2587 if (!data[IFLA_VXLAN_ID])
2588 return -EINVAL;
2589
f01ec1c0
ND		 2590 vxlan->net = dev_net(dev);
2591
d342894c 2592 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2593 dst->remote_vni = vni;
d342894c 2594
5933a7bb
MR		 2595 /* Unless IPv6 is explicitly requested, assume IPv4 */
2596 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2597 if (data[IFLA_VXLAN_GROUP]) {
2598 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2599 } else if (data[IFLA_VXLAN_GROUP6]) {
2600 if (!IS_ENABLED(CONFIG_IPV6))
2601 return -EPFNOSUPPORT;
2602
2603 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2604 sizeof(struct in6_addr));
2605 dst->remote_ip.sa.sa_family = AF_INET6;
2606 use_ipv6 = true;
2607 }
d342894c 2608
e4c7ed41
CW		 2609 if (data[IFLA_VXLAN_LOCAL]) {
2610 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2611 vxlan->saddr.sa.sa_family = AF_INET;
2612 } else if (data[IFLA_VXLAN_LOCAL6]) {
2613 if (!IS_ENABLED(CONFIG_IPV6))
2614 return -EPFNOSUPPORT;
2615
2616 /* TODO: respect scope id */
2617 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2618 sizeof(struct in6_addr));
2619 vxlan->saddr.sa.sa_family = AF_INET6;
2620 use_ipv6 = true;
2621 }
d342894c 2622
34e02aa1 2623 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2624 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2625 struct net_device *lowerdev
c7995c43 2626 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2627
2628 if (!lowerdev) {
c7995c43 2629 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2630 return -ENODEV;
2631 }
d342894c 2632
e4c7ed41
CW		 2633#if IS_ENABLED(CONFIG_IPV6)
2634 if (use_ipv6) {
2635 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2636 if (idev && idev->cnf.disable_ipv6) {
2637 pr_info("IPv6 is disabled via sysctl\n");
2638 return -EPERM;
2639 }
2640 vxlan->flags |= VXLAN_F_IPV6;
2641 }
2642#endif
2643
34e02aa1 2644 if (!tb[IFLA_MTU])
e4c7ed41 2645 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4
AD		 2646
2647 /* update header length based on lower device */
2648 dev->hard_header_len = lowerdev->hard_header_len +
e4c7ed41 2649 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2650 } else if (use_ipv6)
2651 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2652
2653 if (data[IFLA_VXLAN_TOS])
2654 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2655
afb97186
VB		 2656 if (data[IFLA_VXLAN_TTL])
2657 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2658
d342894c 2659 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2660 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2661
2662 if (data[IFLA_VXLAN_AGEING])
2663 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2664 else
2665 vxlan->age_interval = FDB_AGE_DEFAULT;
2666
e4f67add
DS		 2667 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2668 vxlan->flags |= VXLAN_F_PROXY;
2669
2670 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2671 vxlan->flags |= VXLAN_F_RSC;
2672
2673 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2674 vxlan->flags |= VXLAN_F_L2MISS;
2675
2676 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2677 vxlan->flags |= VXLAN_F_L3MISS;
2678
d342894c 2679 if (data[IFLA_VXLAN_LIMIT])
2680 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2681
05f47d69 2682 if (data[IFLA_VXLAN_PORT_RANGE]) {
2683 const struct ifla_vxlan_port_range *p
2684 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2685 vxlan->port_min = ntohs(p->low);
2686 vxlan->port_max = ntohs(p->high);
2687 }
2688
823aa873 2689 if (data[IFLA_VXLAN_PORT])
2690 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2691
359a0ea9
TH		 2692 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2693 vxlan->flags |= VXLAN_F_UDP_CSUM;
2694
2695 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2696 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2697 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2698
2699 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2700 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2701 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2702
553675fb 2703 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2704 pr_info("duplicate VNI %u\n", vni);
2705 return -EEXIST;
2706 }
2707
7ad24ea4 2708 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2709
2936b6ab
SS		 2710 /* create an fdb entry for a valid default destination */
2711 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2712 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2713 &vxlan->default_dst.remote_ip,
2714 NUD_REACHABLE|NUD_PERMANENT,
2715 NLM_F_EXCL|NLM_F_CREATE,
2716 vxlan->dst_port,
2717 vxlan->default_dst.remote_vni,
2718 vxlan->default_dst.remote_ifindex,
2719 NTF_SELF);
2720 if (err)
2721 return err;
2722 }
d342894c 2723
afbd8bae
MR		 2724 err = register_netdevice(dev);
2725 if (err) {
ba609e9b 2726 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2727 return err;
2728 }
2729
553675fb 2730 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2731
2732 return 0;
d342894c 2733}
2734
2735static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2736{
2737 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2738 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2739
fe5c3561 2740 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2741 if (!hlist_unhashed(&vxlan->hlist))
2742 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2743 spin_unlock(&vn->sock_lock);
2744
553675fb 2745 list_del(&vxlan->next);
d342894c 2746 unregister_netdevice_queue(dev, head);
2747}
2748
2749static size_t vxlan_get_size(const struct net_device *dev)
2750{
2751
2752 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2753 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2754 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2755 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2756 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2757 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2758 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2759 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2760 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2761 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2762 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2763 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2764 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2765 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2766 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2767 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2768 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2769 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
d342894c 2770 0;
2771}
2772
2773static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2774{
2775 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2776 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2777 struct ifla_vxlan_port_range ports = {
2778 .low = htons(vxlan->port_min),
2779 .high = htons(vxlan->port_max),
2780 };
d342894c 2781
c7995c43 2782 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2783 goto nla_put_failure;
2784
e4c7ed41
CW		 2785 if (!vxlan_addr_any(&dst->remote_ip)) {
2786 if (dst->remote_ip.sa.sa_family == AF_INET) {
2787 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2788 dst->remote_ip.sin.sin_addr.s_addr))
2789 goto nla_put_failure;
2790#if IS_ENABLED(CONFIG_IPV6)
2791 } else {
2792 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2793 &dst->remote_ip.sin6.sin6_addr))
2794 goto nla_put_failure;
2795#endif
2796 }
2797 }
d342894c 2798
c7995c43 2799 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2800 goto nla_put_failure;
2801
e4c7ed41
CW		 2802 if (!vxlan_addr_any(&vxlan->saddr)) {
2803 if (vxlan->saddr.sa.sa_family == AF_INET) {
2804 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2805 vxlan->saddr.sin.sin_addr.s_addr))
2806 goto nla_put_failure;
2807#if IS_ENABLED(CONFIG_IPV6)
2808 } else {
2809 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2810 &vxlan->saddr.sin6.sin6_addr))
2811 goto nla_put_failure;
2812#endif
2813 }
2814 }
d342894c 2815
2816 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2817 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2818 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2819 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2820 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2821 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2822 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2823 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2824 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2825 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2826 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2827 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2828 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2829 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2830 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2831 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2832 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2833 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2834 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2835 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
d342894c 2836 goto nla_put_failure;
2837
05f47d69 2838 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2839 goto nla_put_failure;
2840
d342894c 2841 return 0;
2842
2843nla_put_failure:
2844 return -EMSGSIZE;
2845}
2846
2847static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2848 .kind = "vxlan",
2849 .maxtype = IFLA_VXLAN_MAX,
2850 .policy = vxlan_policy,
2851 .priv_size = sizeof(struct vxlan_dev),
2852 .setup = vxlan_setup,
2853 .validate = vxlan_validate,
2854 .newlink = vxlan_newlink,
2855 .dellink = vxlan_dellink,
2856 .get_size = vxlan_get_size,
2857 .fill_info = vxlan_fill_info,
2858};
2859
acaf4e70
DB		 2860static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2861 struct net_device *dev)
2862{
2863 struct vxlan_dev *vxlan, *next;
2864 LIST_HEAD(list_kill);
2865
2866 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2867 struct vxlan_rdst *dst = &vxlan->default_dst;
2868
2869 /* In case we created vxlan device with carrier
2870 * and we loose the carrier due to module unload
2871 * we also need to remove vxlan device. In other
2872 * cases, it's not necessary and remote_ifindex
2873 * is 0 here, so no matches.
2874 */
2875 if (dst->remote_ifindex == dev->ifindex)
2876 vxlan_dellink(vxlan->dev, &list_kill);
2877 }
2878
2879 unregister_netdevice_many(&list_kill);
2880}
2881
2882static int vxlan_lowerdev_event(struct notifier_block *unused,
2883 unsigned long event, void *ptr)
2884{
2885 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2886 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2887
783c1463 2888 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2889 vxlan_handle_lowerdev_unregister(vn, dev);
2890
2891 return NOTIFY_DONE;
2892}
2893
2894static struct notifier_block vxlan_notifier_block __read_mostly = {
2895 .notifier_call = vxlan_lowerdev_event,
2896};
2897
d342894c 2898static __net_init int vxlan_init_net(struct net *net)
2899{
2900 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2901 unsigned int h;
d342894c 2902
553675fb 2903 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2904 spin_lock_init(&vn->sock_lock);
d342894c 2905
553675fb 2906 for (h = 0; h < PORT_HASH_SIZE; ++h)
2907 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2908
2909 return 0;
2910}
2911
f01ec1c0
ND		 2912static void __net_exit vxlan_exit_net(struct net *net)
2913{
2914 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2915 struct vxlan_dev *vxlan, *next;
2916 struct net_device *dev, *aux;
2917 LIST_HEAD(list);
2918
2919 rtnl_lock();
2920 for_each_netdev_safe(net, dev, aux)
2921 if (dev->rtnl_link_ops == &vxlan_link_ops)
2922 unregister_netdevice_queue(dev, &list);
2923
2924 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2925 /* If vxlan->dev is in the same netns, it has already been added
2926 * to the list by the previous loop.
2927 */
2928 if (!net_eq(dev_net(vxlan->dev), net))
2929 unregister_netdevice_queue(dev, &list);
2930 }
2931
2932 unregister_netdevice_many(&list);
2933 rtnl_unlock();
2934}
2935
d342894c 2936static struct pernet_operations vxlan_net_ops = {
2937 .init = vxlan_init_net,
f01ec1c0 2938 .exit = vxlan_exit_net,
d342894c 2939 .id = &vxlan_net_id,
2940 .size = sizeof(struct vxlan_net),
2941};
2942
2943static int __init vxlan_init_module(void)
2944{
2945 int rc;
2946
758c57d1
SH		 2947 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2948 if (!vxlan_wq)
2949 return -ENOMEM;
2950
d342894c 2951 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2952
783c1463 2953 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 2954 if (rc)
2955 goto out1;
2956
acaf4e70 2957 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 2958 if (rc)
2959 goto out2;
2960
acaf4e70
DB		 2961 rc = rtnl_link_register(&vxlan_link_ops);
2962 if (rc)
2963 goto out3;
d342894c 2964
acaf4e70
DB		 2965 return 0;
2966out3:
2967 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 2968out2:
783c1463 2969 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 2970out1:
758c57d1 2971 destroy_workqueue(vxlan_wq);
d342894c 2972 return rc;
2973}
7332a13b 2974late_initcall(vxlan_init_module);
d342894c 2975
2976static void __exit vxlan_cleanup_module(void)
2977{
b7153984 2978 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 2979 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 2980 destroy_workqueue(vxlan_wq);
783c1463
DB		 2981 unregister_pernet_subsys(&vxlan_net_ops);
2982 /* rcu_barrier() is called by netns */
d342894c 2983}
2984module_exit(vxlan_cleanup_module);
2985
2986MODULE_LICENSE("GPL");
2987MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2988MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 2989MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 2990MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.348132 seconds and 5 git commands to generate.