projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: Group Policy extension
[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>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
acbf74a7 45#include <net/udp_tunnel.h>
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
d342894c 52
53#define VXLAN_VERSION "0.1"
54
553675fb 55#define PORT_HASH_BITS 8
56#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 57#define VNI_HASH_BITS 10
58#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
59#define FDB_HASH_BITS 8
60#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
61#define FDB_AGE_DEFAULT 300 /* 5 min */
62#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
63
23c578bf 64/* UDP port for VXLAN traffic.
65 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 66 * for compatibility with early adopters.
23c578bf 67 */
9daaa397
SH		 68static unsigned short vxlan_port __read_mostly = 8472;
69module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 70MODULE_PARM_DESC(udp_port, "Destination UDP port");
71
72static bool log_ecn_error = true;
73module_param(log_ecn_error, bool, 0644);
74MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
75
60d9d4c6 76static int vxlan_net_id;
553675fb 77
afbd8bae
MR		 78static const u8 all_zeros_mac[ETH_ALEN];
79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
e4c7ed41
CW		 87union vxlan_addr {
88 struct sockaddr_in sin;
89 struct sockaddr_in6 sin6;
90 struct sockaddr sa;
91};
92
6681712d 93struct vxlan_rdst {
e4c7ed41 94 union vxlan_addr remote_ip;
6681712d
DS		 95 __be16 remote_port;
96 u32 remote_vni;
97 u32 remote_ifindex;
3e61aa8f 98 struct list_head list;
bc7892ba 99 struct rcu_head rcu;
6681712d
DS		 100};
101
d342894c 102/* Forwarding table entry */
103struct vxlan_fdb {
104 struct hlist_node hlist; /* linked list of entries */
105 struct rcu_head rcu;
106 unsigned long updated; /* jiffies */
107 unsigned long used;
3e61aa8f 108 struct list_head remotes;
d342894c 109 u16 state; /* see ndm_state */
ae884082 110 u8 flags; /* see ndm_flags */
d342894c 111 u8 eth_addr[ETH_ALEN];
112};
113
d342894c 114/* Pseudo network device */
115struct vxlan_dev {
553675fb 116 struct hlist_node hlist; /* vni hash table */
117 struct list_head next; /* vxlan's per namespace list */
118 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 119 struct net_device *dev;
f01ec1c0 120 struct net *net; /* netns for packet i/o */
c7995c43 121 struct vxlan_rdst default_dst; /* default destination */
e4c7ed41 122 union vxlan_addr saddr; /* source address */
823aa873 123 __be16 dst_port;
05f47d69 124 __u16 port_min; /* source port range */
125 __u16 port_max;
d342894c 126 __u8 tos; /* TOS override */
127 __u8 ttl;
359a0ea9 128 u32 flags; /* VXLAN_F_* in vxlan.h */
d342894c 129
1c51a915 130 struct work_struct sock_work;
3fc2de2f 131 struct work_struct igmp_join;
132 struct work_struct igmp_leave;
1c51a915 133
d342894c 134 unsigned long age_interval;
135 struct timer_list age_timer;
136 spinlock_t hash_lock;
137 unsigned int addrcnt;
138 unsigned int addrmax;
d342894c 139
140 struct hlist_head fdb_head[FDB_HASH_SIZE];
141};
142
143/* salt for hash table */
144static u32 vxlan_salt __read_mostly;
758c57d1 145static struct workqueue_struct *vxlan_wq;
d342894c 146
1c51a915
SH		 147static void vxlan_sock_work(struct work_struct *work);
148
e4c7ed41
CW		 149#if IS_ENABLED(CONFIG_IPV6)
150static inline
151bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
152{
153 if (a->sa.sa_family != b->sa.sa_family)
154 return false;
155 if (a->sa.sa_family == AF_INET6)
156 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
157 else
158 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
159}
160
161static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
162{
163 if (ipa->sa.sa_family == AF_INET6)
164 return ipv6_addr_any(&ipa->sin6.sin6_addr);
165 else
166 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
167}
168
169static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
170{
171 if (ipa->sa.sa_family == AF_INET6)
172 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
173 else
174 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
175}
176
177static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
178{
179 if (nla_len(nla) >= sizeof(struct in6_addr)) {
180 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
181 ip->sa.sa_family = AF_INET6;
182 return 0;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
194{
195 if (ip->sa.sa_family == AF_INET6)
196 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
197 else
198 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
199}
200
201#else /* !CONFIG_IPV6 */
202
203static inline
204bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
205{
206 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
207}
208
209static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
210{
211 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
212}
213
214static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
215{
216 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
217}
218
219static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
220{
221 if (nla_len(nla) >= sizeof(struct in6_addr)) {
222 return -EAFNOSUPPORT;
223 } else if (nla_len(nla) >= sizeof(__be32)) {
224 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
225 ip->sa.sa_family = AF_INET;
226 return 0;
227 } else {
228 return -EAFNOSUPPORT;
229 }
230}
231
232static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
233 const union vxlan_addr *ip)
234{
235 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
236}
237#endif
238
553675fb 239/* Virtual Network hash table head */
240static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
241{
242 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
243}
244
245/* Socket hash table head */
246static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 247{
248 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
249
553675fb 250 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
251}
252
3e61aa8f
SH		 253/* First remote destination for a forwarding entry.
254 * Guaranteed to be non-NULL because remotes are never deleted.
255 */
5ca5461c 256static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 257{
5ca5461c 258 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
259}
260
261static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
262{
263 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 264}
265
19ca9fc1
ML		 266/* Find VXLAN socket based on network namespace, address family and UDP port */
267static struct vxlan_sock *vxlan_find_sock(struct net *net,
268 sa_family_t family, __be16 port)
553675fb 269{
270 struct vxlan_sock *vs;
271
272 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1
ML		 273 if (inet_sk(vs->sock->sk)->inet_sport == port &&
274 inet_sk(vs->sock->sk)->sk.sk_family == family)
553675fb 275 return vs;
276 }
277 return NULL;
d342894c 278}
279
5cfccc5a 280static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 281{
282 struct vxlan_dev *vxlan;
d342894c 283
553675fb 284 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 285 if (vxlan->default_dst.remote_vni == id)
d342894c 286 return vxlan;
287 }
288
289 return NULL;
290}
291
5cfccc5a 292/* Look up VNI in a per net namespace table */
19ca9fc1
ML		 293static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
294 sa_family_t family, __be16 port)
5cfccc5a
PS		 295{
296 struct vxlan_sock *vs;
297
19ca9fc1 298 vs = vxlan_find_sock(net, family, port);
5cfccc5a
PS		 299 if (!vs)
300 return NULL;
301
302 return vxlan_vs_find_vni(vs, id);
303}
304
d342894c 305/* Fill in neighbour message in skbuff. */
306static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 307 const struct vxlan_fdb *fdb,
308 u32 portid, u32 seq, int type, unsigned int flags,
309 const struct vxlan_rdst *rdst)
d342894c 310{
311 unsigned long now = jiffies;
312 struct nda_cacheinfo ci;
313 struct nlmsghdr *nlh;
314 struct ndmsg *ndm;
e4f67add 315 bool send_ip, send_eth;
d342894c 316
317 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
318 if (nlh == NULL)
319 return -EMSGSIZE;
320
321 ndm = nlmsg_data(nlh);
322 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 323
324 send_eth = send_ip = true;
325
326 if (type == RTM_GETNEIGH) {
327 ndm->ndm_family = AF_INET;
e4c7ed41 328 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 329 send_eth = !is_zero_ether_addr(fdb->eth_addr);
330 } else
331 ndm->ndm_family = AF_BRIDGE;
d342894c 332 ndm->ndm_state = fdb->state;
333 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 334 ndm->ndm_flags = fdb->flags;
545469f7 335 ndm->ndm_type = RTN_UNICAST;
d342894c 336
e4f67add 337 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 338 goto nla_put_failure;
339
e4c7ed41 340 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 341 goto nla_put_failure;
342
823aa873 343 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 344 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
345 goto nla_put_failure;
c7995c43 346 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 347 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 348 goto nla_put_failure;
349 if (rdst->remote_ifindex &&
350 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 351 goto nla_put_failure;
352
353 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
354 ci.ndm_confirmed = 0;
355 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
356 ci.ndm_refcnt = 0;
357
358 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
359 goto nla_put_failure;
360
361 return nlmsg_end(skb, nlh);
362
363nla_put_failure:
364 nlmsg_cancel(skb, nlh);
365 return -EMSGSIZE;
366}
367
368static inline size_t vxlan_nlmsg_size(void)
369{
370 return NLMSG_ALIGN(sizeof(struct ndmsg))
371 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 372 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 373 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 374 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
375 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 376 + nla_total_size(sizeof(struct nda_cacheinfo));
377}
378
9e4b93f9
ND		 379static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
380 struct vxlan_rdst *rd, int type)
d342894c 381{
382 struct net *net = dev_net(vxlan->dev);
383 struct sk_buff *skb;
384 int err = -ENOBUFS;
385
386 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
387 if (skb == NULL)
388 goto errout;
389
9e4b93f9 390 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 391 if (err < 0) {
392 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
393 WARN_ON(err == -EMSGSIZE);
394 kfree_skb(skb);
395 goto errout;
396 }
397
398 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
399 return;
400errout:
401 if (err < 0)
402 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
403}
404
e4c7ed41 405static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 406{
407 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 408 struct vxlan_fdb f = {
409 .state = NUD_STALE,
410 };
411 struct vxlan_rdst remote = {
e4c7ed41 412 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 413 .remote_vni = VXLAN_N_VID,
414 };
3e61aa8f 415
9e4b93f9 416 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 417}
418
419static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
420{
bb3fd687
SH		 421 struct vxlan_fdb f = {
422 .state = NUD_STALE,
423 };
9e4b93f9 424 struct vxlan_rdst remote = { };
e4f67add 425
e4f67add
DS		 426 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
427
9e4b93f9 428 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 429}
430
d342894c 431/* Hash Ethernet address */
432static u32 eth_hash(const unsigned char *addr)
433{
434 u64 value = get_unaligned((u64 *)addr);
435
436 /* only want 6 bytes */
437#ifdef __BIG_ENDIAN
d342894c 438 value >>= 16;
321fb991 439#else
440 value <<= 16;
d342894c 441#endif
442 return hash_64(value, FDB_HASH_BITS);
443}
444
445/* Hash chain to use given mac address */
446static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
447 const u8 *mac)
448{
449 return &vxlan->fdb_head[eth_hash(mac)];
450}
451
452/* Look up Ethernet address in forwarding table */
014be2c8 453static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 454 const u8 *mac)
d342894c 455{
456 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
457 struct vxlan_fdb *f;
d342894c 458
b67bfe0d 459 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 460 if (ether_addr_equal(mac, f->eth_addr))
d342894c 461 return f;
462 }
463
464 return NULL;
465}
466
014be2c8
SS		 467static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
468 const u8 *mac)
469{
470 struct vxlan_fdb *f;
471
472 f = __vxlan_find_mac(vxlan, mac);
473 if (f)
474 f->used = jiffies;
475
476 return f;
477}
478
a5e7c10a
MR		 479/* caller should hold vxlan->hash_lock */
480static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 481 union vxlan_addr *ip, __be16 port,
a5e7c10a 482 __u32 vni, __u32 ifindex)
6681712d 483{
3e61aa8f 484 struct vxlan_rdst *rd;
6681712d 485
3e61aa8f 486 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 487 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 488 rd->remote_port == port &&
489 rd->remote_vni == vni &&
490 rd->remote_ifindex == ifindex)
a5e7c10a 491 return rd;
6681712d 492 }
3e61aa8f 493
a5e7c10a
MR		 494 return NULL;
495}
496
906dc186
TR		 497/* Replace destination of unicast mac */
498static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 499 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 500{
501 struct vxlan_rdst *rd;
502
503 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
504 if (rd)
505 return 0;
506
507 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
508 if (!rd)
509 return 0;
e4c7ed41 510 rd->remote_ip = *ip;
906dc186
TR		 511 rd->remote_port = port;
512 rd->remote_vni = vni;
513 rd->remote_ifindex = ifindex;
514 return 1;
515}
516
a5e7c10a
MR		 517/* Add/update destinations for multicast */
518static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 519 union vxlan_addr *ip, __be16 port, __u32 vni,
520 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 521{
522 struct vxlan_rdst *rd;
523
524 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
525 if (rd)
526 return 0;
527
6681712d
DS		 528 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
529 if (rd == NULL)
530 return -ENOBUFS;
e4c7ed41 531 rd->remote_ip = *ip;
6681712d
DS		 532 rd->remote_port = port;
533 rd->remote_vni = vni;
534 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 535
536 list_add_tail_rcu(&rd->list, &f->remotes);
537
9e4b93f9 538 *rdp = rd;
6681712d
DS		 539 return 1;
540}
541
dfd8645e
TH		 542static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
543 unsigned int off,
544 struct vxlanhdr *vh, size_t hdrlen,
545 u32 data)
546{
547 size_t start, offset, plen;
548 __wsum delta;
549
550 if (skb->remcsum_offload)
551 return vh;
552
553 if (!NAPI_GRO_CB(skb)->csum_valid)
554 return NULL;
555
556 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
557 offset = start + ((data & VXLAN_RCO_UDP) ?
558 offsetof(struct udphdr, check) :
559 offsetof(struct tcphdr, check));
560
561 plen = hdrlen + offset + sizeof(u16);
562
563 /* Pull checksum that will be written */
564 if (skb_gro_header_hard(skb, off + plen)) {
565 vh = skb_gro_header_slow(skb, off + plen, off);
566 if (!vh)
567 return NULL;
568 }
569
570 delta = remcsum_adjust((void *)vh + hdrlen,
571 NAPI_GRO_CB(skb)->csum, start, offset);
572
573 /* Adjust skb->csum since we changed the packet */
574 skb->csum = csum_add(skb->csum, delta);
575 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
576
577 skb->remcsum_offload = 1;
578
579 return vh;
580}
581
a2b12f3c
TH		 582static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
583 struct sk_buff *skb,
584 struct udp_offload *uoff)
dc01e7d3
OG		 585{
586 struct sk_buff *p, **pp = NULL;
587 struct vxlanhdr *vh, *vh2;
9b174d88 588 unsigned int hlen, off_vx;
dc01e7d3 589 int flush = 1;
dfd8645e
TH		 590 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
591 udp_offloads);
592 u32 flags;
dc01e7d3
OG		 593
594 off_vx = skb_gro_offset(skb);
595 hlen = off_vx + sizeof(*vh);
596 vh = skb_gro_header_fast(skb, off_vx);
597 if (skb_gro_header_hard(skb, hlen)) {
598 vh = skb_gro_header_slow(skb, hlen, off_vx);
599 if (unlikely(!vh))
600 goto out;
601 }
dc01e7d3 602
dfd8645e
TH		 603 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
604 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
605
606 flags = ntohl(vh->vx_flags);
607
608 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
609 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
610 ntohl(vh->vx_vni));
611
612 if (!vh)
613 goto out;
614 }
615
dc01e7d3
OG		 616 flush = 0;
617
618 for (p = *head; p; p = p->next) {
619 if (!NAPI_GRO_CB(p)->same_flow)
620 continue;
621
622 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 623 if (vh->vx_flags != vh2->vx_flags ||
624 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 625 NAPI_GRO_CB(p)->same_flow = 0;
626 continue;
627 }
dc01e7d3
OG		 628 }
629
9b174d88 630 pp = eth_gro_receive(head, skb);
dc01e7d3 631
dc01e7d3
OG		 632out:
633 NAPI_GRO_CB(skb)->flush |= flush;
634
635 return pp;
636}
637
a2b12f3c
TH		 638static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
639 struct udp_offload *uoff)
dc01e7d3 640{
cfdf1e1b
JG		 641 udp_tunnel_gro_complete(skb, nhoff);
642
9b174d88 643 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 644}
645
53cf5275 646/* Notify netdevs that UDP port started listening */
dc01e7d3 647static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 648{
649 struct net_device *dev;
dc01e7d3 650 struct sock *sk = vs->sock->sk;
53cf5275
JG		 651 struct net *net = sock_net(sk);
652 sa_family_t sa_family = sk->sk_family;
35e42379 653 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 654 int err;
655
656 if (sa_family == AF_INET) {
657 err = udp_add_offload(&vs->udp_offloads);
658 if (err)
659 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
660 }
53cf5275
JG		 661
662 rcu_read_lock();
663 for_each_netdev_rcu(net, dev) {
664 if (dev->netdev_ops->ndo_add_vxlan_port)
665 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
666 port);
667 }
668 rcu_read_unlock();
669}
670
671/* Notify netdevs that UDP port is no more listening */
dc01e7d3 672static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 673{
674 struct net_device *dev;
dc01e7d3 675 struct sock *sk = vs->sock->sk;
53cf5275
JG		 676 struct net *net = sock_net(sk);
677 sa_family_t sa_family = sk->sk_family;
35e42379 678 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 679
680 rcu_read_lock();
681 for_each_netdev_rcu(net, dev) {
682 if (dev->netdev_ops->ndo_del_vxlan_port)
683 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
684 port);
685 }
686 rcu_read_unlock();
dc01e7d3
OG		 687
688 if (sa_family == AF_INET)
689 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 690}
691
d342894c 692/* Add new entry to forwarding table -- assumes lock held */
693static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 694 const u8 *mac, union vxlan_addr *ip,
6681712d 695 __u16 state, __u16 flags,
73cf3317 696 __be16 port, __u32 vni, __u32 ifindex,
ae884082 697 __u8 ndm_flags)
d342894c 698{
9e4b93f9 699 struct vxlan_rdst *rd = NULL;
d342894c 700 struct vxlan_fdb *f;
701 int notify = 0;
702
014be2c8 703 f = __vxlan_find_mac(vxlan, mac);
d342894c 704 if (f) {
705 if (flags & NLM_F_EXCL) {
706 netdev_dbg(vxlan->dev,
707 "lost race to create %pM\n", mac);
708 return -EEXIST;
709 }
710 if (f->state != state) {
711 f->state = state;
712 f->updated = jiffies;
713 notify = 1;
714 }
ae884082
DS		 715 if (f->flags != ndm_flags) {
716 f->flags = ndm_flags;
717 f->updated = jiffies;
718 notify = 1;
719 }
906dc186
TR		 720 if ((flags & NLM_F_REPLACE)) {
721 /* Only change unicasts */
722 if (!(is_multicast_ether_addr(f->eth_addr) ||
723 is_zero_ether_addr(f->eth_addr))) {
724 int rc = vxlan_fdb_replace(f, ip, port, vni,
725 ifindex);
726
727 if (rc < 0)
728 return rc;
729 notify |= rc;
730 } else
731 return -EOPNOTSUPP;
732 }
6681712d 733 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 734 (is_multicast_ether_addr(f->eth_addr) ||
735 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 736 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
737 &rd);
6681712d
DS		 738
739 if (rc < 0)
740 return rc;
741 notify |= rc;
742 }
d342894c 743 } else {
744 if (!(flags & NLM_F_CREATE))
745 return -ENOENT;
746
747 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
748 return -ENOSPC;
749
906dc186
TR		 750 /* Disallow replace to add a multicast entry */
751 if ((flags & NLM_F_REPLACE) &&
752 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
753 return -EOPNOTSUPP;
754
e4c7ed41 755 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 756 f = kmalloc(sizeof(*f), GFP_ATOMIC);
757 if (!f)
758 return -ENOMEM;
759
760 notify = 1;
d342894c 761 f->state = state;
ae884082 762 f->flags = ndm_flags;
d342894c 763 f->updated = f->used = jiffies;
3e61aa8f 764 INIT_LIST_HEAD(&f->remotes);
d342894c 765 memcpy(f->eth_addr, mac, ETH_ALEN);
766
9e4b93f9 767 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 768
d342894c 769 ++vxlan->addrcnt;
770 hlist_add_head_rcu(&f->hlist,
771 vxlan_fdb_head(vxlan, mac));
772 }
773
9e4b93f9
ND		 774 if (notify) {
775 if (rd == NULL)
776 rd = first_remote_rtnl(f);
777 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
778 }
d342894c 779
780 return 0;
781}
782
6706c82e 783static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 784{
785 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 786 struct vxlan_rdst *rd, *nd;
6681712d 787
3e61aa8f 788 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 789 kfree(rd);
6681712d
DS		 790 kfree(f);
791}
792
d342894c 793static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
794{
795 netdev_dbg(vxlan->dev,
796 "delete %pM\n", f->eth_addr);
797
798 --vxlan->addrcnt;
9e4b93f9 799 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 800
801 hlist_del_rcu(&f->hlist);
6681712d 802 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 803}
804
f0b074be 805static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 806 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 807{
6681712d 808 struct net *net = dev_net(vxlan->dev);
e4c7ed41 809 int err;
d342894c 810
f0b074be 811 if (tb[NDA_DST]) {
e4c7ed41
CW		 812 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
813 if (err)
814 return err;
f0b074be 815 } else {
e4c7ed41
CW		 816 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
817 if (remote->sa.sa_family == AF_INET) {
818 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
819 ip->sa.sa_family = AF_INET;
820#if IS_ENABLED(CONFIG_IPV6)
821 } else {
822 ip->sin6.sin6_addr = in6addr_any;
823 ip->sa.sa_family = AF_INET6;
824#endif
825 }
f0b074be 826 }
d342894c 827
6681712d 828 if (tb[NDA_PORT]) {
73cf3317 829 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 830 return -EINVAL;
f0b074be
MR		 831 *port = nla_get_be16(tb[NDA_PORT]);
832 } else {
833 *port = vxlan->dst_port;
834 }
6681712d
DS		 835
836 if (tb[NDA_VNI]) {
837 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
838 return -EINVAL;
f0b074be
MR		 839 *vni = nla_get_u32(tb[NDA_VNI]);
840 } else {
841 *vni = vxlan->default_dst.remote_vni;
842 }
6681712d
DS		 843
844 if (tb[NDA_IFINDEX]) {
5abb0029 845 struct net_device *tdev;
6681712d
DS		 846
847 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
848 return -EINVAL;
f0b074be 849 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 850 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 851 if (!tdev)
6681712d 852 return -EADDRNOTAVAIL;
f0b074be
MR		 853 } else {
854 *ifindex = 0;
855 }
856
857 return 0;
858}
859
860/* Add static entry (via netlink) */
861static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
862 struct net_device *dev,
f6f6424b 863 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 864{
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 867 union vxlan_addr ip;
f0b074be
MR		 868 __be16 port;
869 u32 vni, ifindex;
870 int err;
871
872 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
873 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
874 ndm->ndm_state);
875 return -EINVAL;
876 }
877
878 if (tb[NDA_DST] == NULL)
879 return -EINVAL;
880
881 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
882 if (err)
883 return err;
6681712d 884
5933a7bb
MR		 885 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
886 return -EAFNOSUPPORT;
887
d342894c 888 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 889 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 890 port, vni, ifindex, ndm->ndm_flags);
d342894c 891 spin_unlock_bh(&vxlan->hash_lock);
892
893 return err;
894}
895
896/* Delete entry (via netlink) */
1690be63
VY		 897static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
898 struct net_device *dev,
f6f6424b 899 const unsigned char *addr, u16 vid)
d342894c 900{
901 struct vxlan_dev *vxlan = netdev_priv(dev);
902 struct vxlan_fdb *f;
bc7892ba 903 struct vxlan_rdst *rd = NULL;
e4c7ed41 904 union vxlan_addr ip;
bc7892ba
MR		 905 __be16 port;
906 u32 vni, ifindex;
907 int err;
908
909 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
910 if (err)
911 return err;
912
913 err = -ENOENT;
d342894c 914
915 spin_lock_bh(&vxlan->hash_lock);
916 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 917 if (!f)
918 goto out;
919
e4c7ed41
CW		 920 if (!vxlan_addr_any(&ip)) {
921 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 922 if (!rd)
923 goto out;
924 }
925
926 err = 0;
927
928 /* remove a destination if it's not the only one on the list,
929 * otherwise destroy the fdb entry
930 */
931 if (rd && !list_is_singular(&f->remotes)) {
932 list_del_rcu(&rd->list);
9e4b93f9 933 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 934 kfree_rcu(rd, rcu);
bc7892ba 935 goto out;
d342894c 936 }
bc7892ba
MR		 937
938 vxlan_fdb_destroy(vxlan, f);
939
940out:
d342894c 941 spin_unlock_bh(&vxlan->hash_lock);
942
943 return err;
944}
945
946/* Dump forwarding table */
947static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 948 struct net_device *dev,
949 struct net_device *filter_dev, int idx)
d342894c 950{
951 struct vxlan_dev *vxlan = netdev_priv(dev);
952 unsigned int h;
953
954 for (h = 0; h < FDB_HASH_SIZE; ++h) {
955 struct vxlan_fdb *f;
d342894c 956 int err;
957
b67bfe0d 958 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 959 struct vxlan_rdst *rd;
6681712d 960
3e61aa8f
SH		 961 if (idx < cb->args[0])
962 goto skip;
963
964 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 965 err = vxlan_fdb_info(skb, vxlan, f,
966 NETLINK_CB(cb->skb).portid,
967 cb->nlh->nlmsg_seq,
968 RTM_NEWNEIGH,
969 NLM_F_MULTI, rd);
970 if (err < 0)
3e61aa8f 971 goto out;
6681712d 972 }
3e61aa8f
SH		 973skip:
974 ++idx;
d342894c 975 }
976 }
3e61aa8f 977out:
d342894c 978 return idx;
979}
980
981/* Watch incoming packets to learn mapping between Ethernet address
982 * and Tunnel endpoint.
26a41ae6 983 * Return true if packet is bogus and should be droppped.
d342894c 984 */
26a41ae6 985static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 986 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 987{
988 struct vxlan_dev *vxlan = netdev_priv(dev);
989 struct vxlan_fdb *f;
d342894c 990
991 f = vxlan_find_mac(vxlan, src_mac);
992 if (likely(f)) {
5ca5461c 993 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 994
e4c7ed41 995 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 996 return false;
997
998 /* Don't migrate static entries, drop packets */
eb064c3b 999 if (f->state & NUD_NOARP)
26a41ae6 1000 return true;
d342894c 1001
1002 if (net_ratelimit())
1003 netdev_info(dev,
e4c7ed41 1004 "%pM migrated from %pIS to %pIS\n",
3e61aa8f 1005 src_mac, &rdst->remote_ip, &src_ip);
d342894c 1006
e4c7ed41 1007 rdst->remote_ip = *src_ip;
d342894c 1008 f->updated = jiffies;
9e4b93f9 1009 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1010 } else {
1011 /* learned new entry */
1012 spin_lock(&vxlan->hash_lock);
3bf74b1a 1013
1014 /* close off race between vxlan_flush and incoming packets */
1015 if (netif_running(dev))
1016 vxlan_fdb_create(vxlan, src_mac, src_ip,
1017 NUD_REACHABLE,
1018 NLM_F_EXCL|NLM_F_CREATE,
1019 vxlan->dst_port,
1020 vxlan->default_dst.remote_vni,
1021 0, NTF_SELF);
d342894c 1022 spin_unlock(&vxlan->hash_lock);
1023 }
26a41ae6 1024
1025 return false;
d342894c 1026}
1027
d342894c 1028/* See if multicast group is already in use by other ID */
95ab0991 1029static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1030{
553675fb 1031 struct vxlan_dev *vxlan;
d342894c 1032
95ab0991
G		 1033 /* The vxlan_sock is only used by dev, leaving group has
1034 * no effect on other vxlan devices.
1035 */
1036 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1037 return false;
1038
553675fb 1039 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1040 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1041 continue;
d342894c 1042
95ab0991
G		 1043 if (vxlan->vn_sock != dev->vn_sock)
1044 continue;
1045
1046 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1047 &dev->default_dst.remote_ip))
1048 continue;
1049
1050 if (vxlan->default_dst.remote_ifindex !=
1051 dev->default_dst.remote_ifindex)
1052 continue;
1053
1054 return true;
553675fb 1055 }
d342894c 1056
1057 return false;
1058}
1059
7c47cedf 1060static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 1061{
7c47cedf
SH		 1062 atomic_inc(&vs->refcnt);
1063}
d342894c 1064
012a5729 1065void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1066{
53cf5275
JG		 1067 struct sock *sk = vs->sock->sk;
1068 struct net *net = sock_net(sk);
1069 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
012a5729 1070
7c47cedf
SH		 1071 if (!atomic_dec_and_test(&vs->refcnt))
1072 return;
d342894c 1073
1c51a915 1074 spin_lock(&vn->sock_lock);
7c47cedf 1075 hlist_del_rcu(&vs->hlist);
dc01e7d3 1076 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1077 spin_unlock(&vn->sock_lock);
1078
7c47cedf 1079 queue_work(vxlan_wq, &vs->del_work);
d342894c 1080}
012a5729 1081EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 1082
3fc2de2f 1083/* Callback to update multicast group membership when first VNI on
1084 * multicast asddress is brought up
1085 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 1086 */
3fc2de2f 1087static void vxlan_igmp_join(struct work_struct *work)
d342894c 1088{
3fc2de2f 1089 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 1090 struct vxlan_sock *vs = vxlan->vn_sock;
1091 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1092 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1093 int ifindex = vxlan->default_dst.remote_ifindex;
d342894c 1094
d342894c 1095 lock_sock(sk);
e4c7ed41
CW		 1096 if (ip->sa.sa_family == AF_INET) {
1097 struct ip_mreqn mreq = {
1098 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1099 .imr_ifindex = ifindex,
1100 };
1101
1102 ip_mc_join_group(sk, &mreq);
1103#if IS_ENABLED(CONFIG_IPV6)
1104 } else {
1105 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1106 &ip->sin6.sin6_addr);
1107#endif
1108 }
3fc2de2f 1109 release_sock(sk);
1110
012a5729 1111 vxlan_sock_release(vs);
3fc2de2f 1112 dev_put(vxlan->dev);
1113}
1114
1115/* Inverse of vxlan_igmp_join when last VNI is brought down */
1116static void vxlan_igmp_leave(struct work_struct *work)
1117{
1118 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 1119 struct vxlan_sock *vs = vxlan->vn_sock;
1120 struct sock *sk = vs->sock->sk;
e4c7ed41
CW		 1121 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1122 int ifindex = vxlan->default_dst.remote_ifindex;
3fc2de2f 1123
1124 lock_sock(sk);
e4c7ed41
CW		 1125 if (ip->sa.sa_family == AF_INET) {
1126 struct ip_mreqn mreq = {
1127 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1128 .imr_ifindex = ifindex,
1129 };
1130
1131 ip_mc_leave_group(sk, &mreq);
1132#if IS_ENABLED(CONFIG_IPV6)
1133 } else {
1134 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1135 &ip->sin6.sin6_addr);
1136#endif
1137 }
1138
d342894c 1139 release_sock(sk);
d342894c 1140
012a5729 1141 vxlan_sock_release(vs);
7c47cedf 1142 dev_put(vxlan->dev);
d342894c 1143}
1144
dfd8645e
TH		 1145static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
1146 size_t hdrlen, u32 data)
1147{
1148 size_t start, offset, plen;
1149 __wsum delta;
1150
1151 if (skb->remcsum_offload) {
1152 /* Already processed in GRO path */
1153 skb->remcsum_offload = 0;
1154 return vh;
1155 }
1156
1157 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1158 offset = start + ((data & VXLAN_RCO_UDP) ?
1159 offsetof(struct udphdr, check) :
1160 offsetof(struct tcphdr, check));
1161
1162 plen = hdrlen + offset + sizeof(u16);
1163
1164 if (!pskb_may_pull(skb, plen))
1165 return NULL;
1166
1167 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1168
1169 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE))
1170 __skb_checksum_complete(skb);
1171
1172 delta = remcsum_adjust((void *)vh + hdrlen,
1173 skb->csum, start, offset);
1174
1175 /* Adjust skb->csum since we changed the packet */
1176 skb->csum = csum_add(skb->csum, delta);
1177
1178 return vh;
1179}
1180
d342894c 1181/* Callback from net/ipv4/udp.c to receive packets */
1182static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1183{
5cfccc5a 1184 struct vxlan_sock *vs;
d342894c 1185 struct vxlanhdr *vxh;
3bf39475 1186 u32 flags, vni;
3511494c 1187 struct vxlan_metadata md = {0};
d342894c 1188
d342894c 1189 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1190 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1191 goto error;
1192
7ce04758 1193 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1194 flags = ntohl(vxh->vx_flags);
1195 vni = ntohl(vxh->vx_vni);
1196
1197 if (flags & VXLAN_HF_VNI) {
1198 flags &= ~VXLAN_HF_VNI;
1199 } else {
1200 /* VNI flag always required to be set */
1201 goto bad_flags;
d342894c 1202 }
1203
5cfccc5a
PS		 1204 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1205 goto drop;
dfd8645e 1206 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1207
559835ea 1208 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1209 if (!vs)
d342894c 1210 goto drop;
d342894c 1211
dfd8645e
TH		 1212 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
1213 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni);
1214 if (!vxh)
1215 goto drop;
1216
1217 flags &= ~VXLAN_HF_RCO;
1218 vni &= VXLAN_VID_MASK;
1219 }
1220
3511494c
TG		 1221 /* For backwards compatibility, only allow reserved fields to be
1222 * used by VXLAN extensions if explicitly requested.
1223 */
1224 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1225 struct vxlanhdr_gbp *gbp;
1226
1227 gbp = (struct vxlanhdr_gbp *)vxh;
1228 md.gbp = ntohs(gbp->policy_id);
1229
1230 if (gbp->dont_learn)
1231 md.gbp |= VXLAN_GBP_DONT_LEARN;
1232
1233 if (gbp->policy_applied)
1234 md.gbp |= VXLAN_GBP_POLICY_APPLIED;
1235
1236 flags &= ~VXLAN_GBP_USED_BITS;
1237 }
1238
dfd8645e 1239 if (flags || (vni & ~VXLAN_VID_MASK)) {
3bf39475
TH		 1240 /* If there are any unprocessed flags remaining treat
1241 * this as a malformed packet. This behavior diverges from
1242 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1243 * in reserved fields are to be ignored. The approach here
1244 * maintains compatbility with previous stack code, and also
1245 * is more robust and provides a little more security in
1246 * adding extensions to VXLAN.
1247 */
1248
1249 goto bad_flags;
1250 }
1251
3511494c
TG		 1252 md.vni = vxh->vx_vni;
1253 vs->rcv(vs, skb, &md);
5cfccc5a
PS		 1254 return 0;
1255
1256drop:
1257 /* Consume bad packet */
1258 kfree_skb(skb);
1259 return 0;
1260
3bf39475
TH		 1261bad_flags:
1262 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1263 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1264
5cfccc5a
PS		 1265error:
1266 /* Return non vxlan pkt */
1267 return 1;
1268}
1269
3511494c
TG		 1270static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1271 struct vxlan_metadata *md)
5cfccc5a 1272{
e4c7ed41
CW		 1273 struct iphdr *oip = NULL;
1274 struct ipv6hdr *oip6 = NULL;
5cfccc5a 1275 struct vxlan_dev *vxlan;
8f84985f 1276 struct pcpu_sw_netstats *stats;
e4c7ed41 1277 union vxlan_addr saddr;
5cfccc5a 1278 __u32 vni;
e4c7ed41
CW		 1279 int err = 0;
1280 union vxlan_addr *remote_ip;
5cfccc5a 1281
3511494c 1282 vni = ntohl(md->vni) >> 8;
5cfccc5a
PS		 1283 /* Is this VNI defined? */
1284 vxlan = vxlan_vs_find_vni(vs, vni);
1285 if (!vxlan)
d342894c 1286 goto drop;
d342894c 1287
e4c7ed41 1288 remote_ip = &vxlan->default_dst.remote_ip;
e4f67add 1289 skb_reset_mac_header(skb);
f01ec1c0 1290 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
d342894c 1291 skb->protocol = eth_type_trans(skb, vxlan->dev);
f79b064c 1292 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
d342894c 1293
1294 /* Ignore packet loops (and multicast echo) */
7367d0b5 1295 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
d342894c 1296 goto drop;
1297
7ce04758 1298 /* Re-examine inner Ethernet packet */
e4c7ed41
CW		 1299 if (remote_ip->sa.sa_family == AF_INET) {
1300 oip = ip_hdr(skb);
1301 saddr.sin.sin_addr.s_addr = oip->saddr;
1302 saddr.sa.sa_family = AF_INET;
1303#if IS_ENABLED(CONFIG_IPV6)
1304 } else {
1305 oip6 = ipv6_hdr(skb);
1306 saddr.sin6.sin6_addr = oip6->saddr;
1307 saddr.sa.sa_family = AF_INET6;
1308#endif
1309 }
1310
26a41ae6 1311 if ((vxlan->flags & VXLAN_F_LEARN) &&
e4c7ed41 1312 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
26a41ae6 1313 goto drop;
d342894c 1314
d342894c 1315 skb_reset_network_header(skb);
3511494c 1316 skb->mark = md->gbp;
0afb1666 1317
e4c7ed41
CW		 1318 if (oip6)
1319 err = IP6_ECN_decapsulate(oip6, skb);
1320 if (oip)
1321 err = IP_ECN_decapsulate(oip, skb);
1322
d342894c 1323 if (unlikely(err)) {
e4c7ed41
CW		 1324 if (log_ecn_error) {
1325 if (oip6)
1326 net_info_ratelimited("non-ECT from %pI6\n",
1327 &oip6->saddr);
1328 if (oip)
1329 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1330 &oip->saddr, oip->tos);
1331 }
d342894c 1332 if (err > 1) {
1333 ++vxlan->dev->stats.rx_frame_errors;
1334 ++vxlan->dev->stats.rx_errors;
1335 goto drop;
1336 }
1337 }
1338
e8171045 1339 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 1340 u64_stats_update_begin(&stats->syncp);
1341 stats->rx_packets++;
1342 stats->rx_bytes += skb->len;
1343 u64_stats_update_end(&stats->syncp);
1344
1345 netif_rx(skb);
1346
5cfccc5a 1347 return;
d342894c 1348drop:
1349 /* Consume bad packet */
1350 kfree_skb(skb);
d342894c 1351}
1352
e4f67add
DS		 1353static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1354{
1355 struct vxlan_dev *vxlan = netdev_priv(dev);
1356 struct arphdr *parp;
1357 u8 *arpptr, *sha;
1358 __be32 sip, tip;
1359 struct neighbour *n;
1360
1361 if (dev->flags & IFF_NOARP)
1362 goto out;
1363
1364 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1365 dev->stats.tx_dropped++;
1366 goto out;
1367 }
1368 parp = arp_hdr(skb);
1369
1370 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1371 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1372 parp->ar_pro != htons(ETH_P_IP) ||
1373 parp->ar_op != htons(ARPOP_REQUEST) ||
1374 parp->ar_hln != dev->addr_len ||
1375 parp->ar_pln != 4)
1376 goto out;
1377 arpptr = (u8 *)parp + sizeof(struct arphdr);
1378 sha = arpptr;
1379 arpptr += dev->addr_len; /* sha */
1380 memcpy(&sip, arpptr, sizeof(sip));
1381 arpptr += sizeof(sip);
1382 arpptr += dev->addr_len; /* tha */
1383 memcpy(&tip, arpptr, sizeof(tip));
1384
1385 if (ipv4_is_loopback(tip) ||
1386 ipv4_is_multicast(tip))
1387 goto out;
1388
1389 n = neigh_lookup(&arp_tbl, &tip, dev);
1390
1391 if (n) {
e4f67add
DS		 1392 struct vxlan_fdb *f;
1393 struct sk_buff *reply;
1394
1395 if (!(n->nud_state & NUD_CONNECTED)) {
1396 neigh_release(n);
1397 goto out;
1398 }
1399
1400 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1401 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1402 /* bridge-local neighbor */
1403 neigh_release(n);
1404 goto out;
1405 }
1406
1407 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1408 n->ha, sha);
1409
1410 neigh_release(n);
1411
7346135d
DS		 1412 if (reply == NULL)
1413 goto out;
1414
e4f67add
DS		 1415 skb_reset_mac_header(reply);
1416 __skb_pull(reply, skb_network_offset(reply));
1417 reply->ip_summed = CHECKSUM_UNNECESSARY;
1418 reply->pkt_type = PACKET_HOST;
1419
1420 if (netif_rx_ni(reply) == NET_RX_DROP)
1421 dev->stats.rx_dropped++;
e4c7ed41
CW		 1422 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1423 union vxlan_addr ipa = {
1424 .sin.sin_addr.s_addr = tip,
a45e92a5 1425 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1426 };
1427
1428 vxlan_ip_miss(dev, &ipa);
1429 }
e4f67add
DS		 1430out:
1431 consume_skb(skb);
1432 return NETDEV_TX_OK;
1433}
1434
f564f45c 1435#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1436static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1437 struct neighbour *n, bool isrouter)
1438{
1439 struct net_device *dev = request->dev;
1440 struct sk_buff *reply;
1441 struct nd_msg *ns, *na;
1442 struct ipv6hdr *pip6;
1443 u8 *daddr;
1444 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1445 int ns_olen;
1446 int i, len;
1447
1448 if (dev == NULL)
1449 return NULL;
1450
1451 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1452 sizeof(*na) + na_olen + dev->needed_tailroom;
1453 reply = alloc_skb(len, GFP_ATOMIC);
1454 if (reply == NULL)
1455 return NULL;
1456
1457 reply->protocol = htons(ETH_P_IPV6);
1458 reply->dev = dev;
1459 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1460 skb_push(reply, sizeof(struct ethhdr));
1461 skb_set_mac_header(reply, 0);
1462
1463 ns = (struct nd_msg *)skb_transport_header(request);
1464
1465 daddr = eth_hdr(request)->h_source;
1466 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1467 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1468 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1469 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1470 break;
1471 }
1472 }
1473
1474 /* Ethernet header */
1475 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1476 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1477 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1478 reply->protocol = htons(ETH_P_IPV6);
1479
1480 skb_pull(reply, sizeof(struct ethhdr));
1481 skb_set_network_header(reply, 0);
1482 skb_put(reply, sizeof(struct ipv6hdr));
1483
1484 /* IPv6 header */
1485
1486 pip6 = ipv6_hdr(reply);
1487 memset(pip6, 0, sizeof(struct ipv6hdr));
1488 pip6->version = 6;
1489 pip6->priority = ipv6_hdr(request)->priority;
1490 pip6->nexthdr = IPPROTO_ICMPV6;
1491 pip6->hop_limit = 255;
1492 pip6->daddr = ipv6_hdr(request)->saddr;
1493 pip6->saddr = *(struct in6_addr *)n->primary_key;
1494
1495 skb_pull(reply, sizeof(struct ipv6hdr));
1496 skb_set_transport_header(reply, 0);
1497
1498 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1499
1500 /* Neighbor Advertisement */
1501 memset(na, 0, sizeof(*na)+na_olen);
1502 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1503 na->icmph.icmp6_router = isrouter;
1504 na->icmph.icmp6_override = 1;
1505 na->icmph.icmp6_solicited = 1;
1506 na->target = ns->target;
1507 ether_addr_copy(&na->opt[2], n->ha);
1508 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1509 na->opt[1] = na_olen >> 3;
1510
1511 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1512 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1513 csum_partial(na, sizeof(*na)+na_olen, 0));
1514
1515 pip6->payload_len = htons(sizeof(*na)+na_olen);
1516
1517 skb_push(reply, sizeof(struct ipv6hdr));
1518
1519 reply->ip_summed = CHECKSUM_UNNECESSARY;
1520
1521 return reply;
1522}
1523
f564f45c
CW		 1524static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1525{
1526 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1527 struct nd_msg *msg;
f564f45c
CW		 1528 const struct ipv6hdr *iphdr;
1529 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1530 struct neighbour *n;
1531 struct inet6_dev *in6_dev;
f564f45c
CW		 1532
1533 in6_dev = __in6_dev_get(dev);
1534 if (!in6_dev)
1535 goto out;
1536
f564f45c
CW		 1537 iphdr = ipv6_hdr(skb);
1538 saddr = &iphdr->saddr;
1539 daddr = &iphdr->daddr;
1540
f564f45c
CW		 1541 msg = (struct nd_msg *)skb_transport_header(skb);
1542 if (msg->icmph.icmp6_code != 0 ||
1543 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1544 goto out;
1545
4b29dba9
DS		 1546 if (ipv6_addr_loopback(daddr) ||
1547 ipv6_addr_is_multicast(&msg->target))
1548 goto out;
1549
1550 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1551
1552 if (n) {
1553 struct vxlan_fdb *f;
4b29dba9 1554 struct sk_buff *reply;
f564f45c
CW		 1555
1556 if (!(n->nud_state & NUD_CONNECTED)) {
1557 neigh_release(n);
1558 goto out;
1559 }
1560
1561 f = vxlan_find_mac(vxlan, n->ha);
1562 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1563 /* bridge-local neighbor */
1564 neigh_release(n);
1565 goto out;
1566 }
1567
4b29dba9
DS		 1568 reply = vxlan_na_create(skb, n,
1569 !!(f ? f->flags & NTF_ROUTER : 0));
1570
f564f45c 1571 neigh_release(n);
4b29dba9
DS		 1572
1573 if (reply == NULL)
1574 goto out;
1575
1576 if (netif_rx_ni(reply) == NET_RX_DROP)
1577 dev->stats.rx_dropped++;
1578
f564f45c 1579 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1580 union vxlan_addr ipa = {
1581 .sin6.sin6_addr = msg->target,
a45e92a5 1582 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1583 };
1584
f564f45c
CW		 1585 vxlan_ip_miss(dev, &ipa);
1586 }
1587
1588out:
1589 consume_skb(skb);
1590 return NETDEV_TX_OK;
1591}
1592#endif
1593
e4f67add
DS		 1594static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1595{
1596 struct vxlan_dev *vxlan = netdev_priv(dev);
1597 struct neighbour *n;
e4f67add
DS		 1598
1599 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1600 return false;
1601
1602 n = NULL;
1603 switch (ntohs(eth_hdr(skb)->h_proto)) {
1604 case ETH_P_IP:
e15a00aa
CW		 1605 {
1606 struct iphdr *pip;
1607
e4f67add
DS		 1608 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1609 return false;
1610 pip = ip_hdr(skb);
1611 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1612 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1613 union vxlan_addr ipa = {
1614 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1615 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1616 };
1617
1618 vxlan_ip_miss(dev, &ipa);
1619 return false;
1620 }
1621
e4f67add 1622 break;
e15a00aa
CW		 1623 }
1624#if IS_ENABLED(CONFIG_IPV6)
1625 case ETH_P_IPV6:
1626 {
1627 struct ipv6hdr *pip6;
1628
1629 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1630 return false;
1631 pip6 = ipv6_hdr(skb);
1632 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1633 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1634 union vxlan_addr ipa = {
1635 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1636 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1637 };
1638
1639 vxlan_ip_miss(dev, &ipa);
1640 return false;
1641 }
1642
1643 break;
1644 }
1645#endif
e4f67add
DS		 1646 default:
1647 return false;
1648 }
1649
1650 if (n) {
1651 bool diff;
1652
7367d0b5 1653 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1654 if (diff) {
1655 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1656 dev->addr_len);
1657 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1658 }
1659 neigh_release(n);
1660 return diff;
e4c7ed41
CW		 1661 }
1662
e4f67add
DS		 1663 return false;
1664}
1665
3511494c
TG		 1666static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, struct vxlan_sock *vs,
1667 struct vxlan_metadata *md)
1668{
1669 struct vxlanhdr_gbp *gbp;
1670
1671 gbp = (struct vxlanhdr_gbp *)vxh;
1672 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1673
1674 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1675 gbp->dont_learn = 1;
1676
1677 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1678 gbp->policy_applied = 1;
1679
1680 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1681}
1682
e4c7ed41 1683#if IS_ENABLED(CONFIG_IPV6)
11796187 1684static int vxlan6_xmit_skb(struct vxlan_sock *vs,
e4c7ed41
CW		 1685 struct dst_entry *dst, struct sk_buff *skb,
1686 struct net_device *dev, struct in6_addr *saddr,
1687 struct in6_addr *daddr, __u8 prio, __u8 ttl,
3511494c
TG		 1688 __be16 src_port, __be16 dst_port,
1689 struct vxlan_metadata *md, bool xnet)
e4c7ed41 1690{
e4c7ed41 1691 struct vxlanhdr *vxh;
e4c7ed41
CW		 1692 int min_headroom;
1693 int err;
acbf74a7 1694 bool udp_sum = !udp_get_no_check6_tx(vs->sock->sk);
dfd8645e
TH		 1695 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1696 u16 hdrlen = sizeof(struct vxlanhdr);
1697
1698 if ((vs->flags & VXLAN_F_REMCSUM_TX) &&
1699 skb->ip_summed == CHECKSUM_PARTIAL) {
1700 int csum_start = skb_checksum_start_offset(skb);
1701
1702 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1703 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1704 (skb->csum_offset == offsetof(struct udphdr, check) ||
1705 skb->csum_offset == offsetof(struct tcphdr, check))) {
1706 udp_sum = false;
1707 type |= SKB_GSO_TUNNEL_REMCSUM;
1708 }
1709 }
e4c7ed41 1710
dfd8645e 1711 skb = iptunnel_handle_offloads(skb, udp_sum, type);
74f47278
PS		 1712 if (IS_ERR(skb)) {
1713 err = -EINVAL;
1714 goto err;
1715 }
e4c7ed41 1716
f01ec1c0 1717 skb_scrub_packet(skb, xnet);
963a88b3 1718
e4c7ed41
CW		 1719 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1720 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1721 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1722
1723 /* Need space for new headers (invalidates iph ptr) */
1724 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1725 if (unlikely(err)) {
1726 kfree_skb(skb);
1727 goto err;
1728 }
e4c7ed41 1729
5968250c 1730 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1731 if (WARN_ON(!skb)) {
1732 err = -ENOMEM;
1733 goto err;
1734 }
e4c7ed41
CW		 1735
1736 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1737 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1738 vxh->vx_vni = md->vni;
e4c7ed41 1739
dfd8645e
TH		 1740 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1741 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1742 VXLAN_RCO_SHIFT;
1743
1744 if (skb->csum_offset == offsetof(struct udphdr, check))
1745 data |= VXLAN_RCO_UDP;
1746
1747 vxh->vx_vni |= htonl(data);
1748 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1749
1750 if (!skb_is_gso(skb)) {
1751 skb->ip_summed = CHECKSUM_NONE;
1752 skb->encapsulation = 0;
1753 }
1754 }
1755
3511494c
TG		 1756 if (vs->flags & VXLAN_F_GBP)
1757 vxlan_build_gbp_hdr(vxh, vs, md);
1758
996c9fd1
TH		 1759 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1760
acbf74a7
AZ		 1761 udp_tunnel6_xmit_skb(vs->sock, dst, skb, dev, saddr, daddr, prio,
1762 ttl, src_port, dst_port);
e4c7ed41 1763 return 0;
74f47278
PS		 1764err:
1765 dst_release(dst);
1766 return err;
e4c7ed41
CW		 1767}
1768#endif
1769
11796187 1770int vxlan_xmit_skb(struct vxlan_sock *vs,
49560532
PS		 1771 struct rtable *rt, struct sk_buff *skb,
1772 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
3511494c
TG		 1773 __be16 src_port, __be16 dst_port,
1774 struct vxlan_metadata *md, bool xnet)
49560532
PS		 1775{
1776 struct vxlanhdr *vxh;
649c5b8b 1777 int min_headroom;
49560532 1778 int err;
acbf74a7 1779 bool udp_sum = !vs->sock->sk->sk_no_check_tx;
dfd8645e
TH		 1780 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1781 u16 hdrlen = sizeof(struct vxlanhdr);
1782
1783 if ((vs->flags & VXLAN_F_REMCSUM_TX) &&
1784 skb->ip_summed == CHECKSUM_PARTIAL) {
1785 int csum_start = skb_checksum_start_offset(skb);
1786
1787 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1788 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1789 (skb->csum_offset == offsetof(struct udphdr, check) ||
1790 skb->csum_offset == offsetof(struct tcphdr, check))) {
1791 udp_sum = false;
1792 type |= SKB_GSO_TUNNEL_REMCSUM;
1793 }
1794 }
49560532 1795
dfd8645e 1796 skb = iptunnel_handle_offloads(skb, udp_sum, type);
359a0ea9 1797 if (IS_ERR(skb))
74f47278 1798 return PTR_ERR(skb);
49560532 1799
649c5b8b 1800 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1801 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1802 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1803
1804 /* Need space for new headers (invalidates iph ptr) */
1805 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1806 if (unlikely(err)) {
1807 kfree_skb(skb);
649c5b8b 1808 return err;
74f47278 1809 }
649c5b8b 1810
5968250c
JP		 1811 skb = vlan_hwaccel_push_inside(skb);
1812 if (WARN_ON(!skb))
1813 return -ENOMEM;
1eaa8178 1814
49560532 1815 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1816 vxh->vx_flags = htonl(VXLAN_HF_VNI);
3511494c 1817 vxh->vx_vni = md->vni;
49560532 1818
dfd8645e
TH		 1819 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1820 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1821 VXLAN_RCO_SHIFT;
1822
1823 if (skb->csum_offset == offsetof(struct udphdr, check))
1824 data |= VXLAN_RCO_UDP;
1825
1826 vxh->vx_vni |= htonl(data);
1827 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1828
1829 if (!skb_is_gso(skb)) {
1830 skb->ip_summed = CHECKSUM_NONE;
1831 skb->encapsulation = 0;
1832 }
1833 }
1834
3511494c
TG		 1835 if (vs->flags & VXLAN_F_GBP)
1836 vxlan_build_gbp_hdr(vxh, vs, md);
1837
996c9fd1
TH		 1838 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1839
acbf74a7
AZ		 1840 return udp_tunnel_xmit_skb(vs->sock, rt, skb, src, dst, tos,
1841 ttl, df, src_port, dst_port, xnet);
49560532
PS		 1842}
1843EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1844
9dcc71e1
SS		 1845/* Bypass encapsulation if the destination is local */
1846static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1847 struct vxlan_dev *dst_vxlan)
1848{
8f84985f 1849 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1850 union vxlan_addr loopback;
1851 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1852 struct net_device *dev = skb->dev;
1853 int len = skb->len;
9dcc71e1 1854
8f84985f
LR		 1855 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1856 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1857 skb->pkt_type = PACKET_HOST;
1858 skb->encapsulation = 0;
1859 skb->dev = dst_vxlan->dev;
1860 __skb_pull(skb, skb_network_offset(skb));
1861
e4c7ed41
CW		 1862 if (remote_ip->sa.sa_family == AF_INET) {
1863 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1864 loopback.sa.sa_family = AF_INET;
1865#if IS_ENABLED(CONFIG_IPV6)
1866 } else {
1867 loopback.sin6.sin6_addr = in6addr_loopback;
1868 loopback.sa.sa_family = AF_INET6;
1869#endif
1870 }
1871
9dcc71e1 1872 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1873 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1874
1875 u64_stats_update_begin(&tx_stats->syncp);
1876 tx_stats->tx_packets++;
ce6502a8 1877 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1878 u64_stats_update_end(&tx_stats->syncp);
1879
1880 if (netif_rx(skb) == NET_RX_SUCCESS) {
1881 u64_stats_update_begin(&rx_stats->syncp);
1882 rx_stats->rx_packets++;
ce6502a8 1883 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1884 u64_stats_update_end(&rx_stats->syncp);
1885 } else {
ce6502a8 1886 dev->stats.rx_dropped++;
9dcc71e1
SS		 1887 }
1888}
1889
4ad16930
SH		 1890static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1891 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1892{
1893 struct vxlan_dev *vxlan = netdev_priv(dev);
e4c7ed41 1894 struct rtable *rt = NULL;
d342894c 1895 const struct iphdr *old_iph;
d342894c 1896 struct flowi4 fl4;
e4c7ed41 1897 union vxlan_addr *dst;
3511494c 1898 struct vxlan_metadata md;
e4c7ed41 1899 __be16 src_port = 0, dst_port;
234f5b73 1900 u32 vni;
d342894c 1901 __be16 df = 0;
1902 __u8 tos, ttl;
0e6fbc5b 1903 int err;
d342894c 1904
823aa873 1905 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d 1906 vni = rdst->remote_vni;
e4c7ed41 1907 dst = &rdst->remote_ip;
e4f67add 1908
e4c7ed41 1909 if (vxlan_addr_any(dst)) {
e4f67add 1910 if (did_rsc) {
e4f67add 1911 /* short-circuited back to local bridge */
9dcc71e1 1912 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1913 return;
e4f67add 1914 }
ef59febe 1915 goto drop;
e4f67add 1916 }
ef59febe 1917
d342894c 1918 old_iph = ip_hdr(skb);
1919
d342894c 1920 ttl = vxlan->ttl;
e4c7ed41 1921 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1922 ttl = 1;
1923
1924 tos = vxlan->tos;
1925 if (tos == 1)
206aaafc 1926 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1927
535fb8d0
TH		 1928 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1929 vxlan->port_max, true);
d342894c 1930
e4c7ed41
CW		 1931 if (dst->sa.sa_family == AF_INET) {
1932 memset(&fl4, 0, sizeof(fl4));
1933 fl4.flowi4_oif = rdst->remote_ifindex;
1934 fl4.flowi4_tos = RT_TOS(tos);
1935 fl4.daddr = dst->sin.sin_addr.s_addr;
1936 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1937
f01ec1c0 1938 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 1939 if (IS_ERR(rt)) {
1940 netdev_dbg(dev, "no route to %pI4\n",
1941 &dst->sin.sin_addr.s_addr);
1942 dev->stats.tx_carrier_errors++;
1943 goto tx_error;
1944 }
d342894c 1945
e4c7ed41
CW		 1946 if (rt->dst.dev == dev) {
1947 netdev_dbg(dev, "circular route to %pI4\n",
1948 &dst->sin.sin_addr.s_addr);
1949 dev->stats.collisions++;
fffc15a5 1950 goto rt_tx_error;
e4c7ed41
CW		 1951 }
1952
1953 /* Bypass encapsulation if the destination is local */
1954 if (rt->rt_flags & RTCF_LOCAL &&
1955 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1956 struct vxlan_dev *dst_vxlan;
1957
1958 ip_rt_put(rt);
19ca9fc1
ML		 1959 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1960 dst->sa.sa_family, dst_port);
e4c7ed41
CW		 1961 if (!dst_vxlan)
1962 goto tx_error;
1963 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1964 return;
1965 }
1966
1967 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1968 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
3511494c
TG		 1969 md.vni = htonl(vni << 8);
1970 md.gbp = skb->mark;
d342894c 1971
3c4d1dae
AZ		 1972 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1973 fl4.saddr, dst->sin.sin_addr.s_addr,
3511494c 1974 tos, ttl, df, src_port, dst_port, &md,
3c4d1dae 1975 !net_eq(vxlan->net, dev_net(vxlan->dev)));
74f47278
PS		 1976 if (err < 0) {
1977 /* skb is already freed. */
1978 skb = NULL;
e4c7ed41 1979 goto rt_tx_error;
74f47278
PS		 1980 }
1981
e4c7ed41
CW		 1982 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1983#if IS_ENABLED(CONFIG_IPV6)
1984 } else {
1985 struct sock *sk = vxlan->vn_sock->sock->sk;
1986 struct dst_entry *ndst;
1987 struct flowi6 fl6;
1988 u32 flags;
1989
1990 memset(&fl6, 0, sizeof(fl6));
1991 fl6.flowi6_oif = rdst->remote_ifindex;
1992 fl6.daddr = dst->sin6.sin6_addr;
1993 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
8c1bb79f 1994 fl6.flowi6_proto = IPPROTO_UDP;
e4c7ed41
CW		 1995
1996 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1997 netdev_dbg(dev, "no route to %pI6\n",
1998 &dst->sin6.sin6_addr);
1999 dev->stats.tx_carrier_errors++;
9dcc71e1 2000 goto tx_error;
e4c7ed41 2001 }
d342894c 2002
e4c7ed41
CW		 2003 if (ndst->dev == dev) {
2004 netdev_dbg(dev, "circular route to %pI6\n",
2005 &dst->sin6.sin6_addr);
2006 dst_release(ndst);
2007 dev->stats.collisions++;
2008 goto tx_error;
2009 }
0e6fbc5b 2010
e4c7ed41
CW		 2011 /* Bypass encapsulation if the destination is local */
2012 flags = ((struct rt6_info *)ndst)->rt6i_flags;
2013 if (flags & RTF_LOCAL &&
2014 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2015 struct vxlan_dev *dst_vxlan;
2016
2017 dst_release(ndst);
19ca9fc1
ML		 2018 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2019 dst->sa.sa_family, dst_port);
e4c7ed41
CW		 2020 if (!dst_vxlan)
2021 goto tx_error;
2022 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2023 return;
2024 }
49560532 2025
e4c7ed41 2026 ttl = ttl ? : ip6_dst_hoplimit(ndst);
3511494c
TG		 2027 md.vni = htonl(vni << 8);
2028 md.gbp = skb->mark;
e4c7ed41 2029
11796187 2030 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
e4c7ed41 2031 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
3511494c 2032 src_port, dst_port, &md,
f01ec1c0 2033 !net_eq(vxlan->net, dev_net(vxlan->dev)));
e4c7ed41
CW		 2034#endif
2035 }
0e6fbc5b 2036
4ad16930 2037 return;
d342894c 2038
2039drop:
2040 dev->stats.tx_dropped++;
2041 goto tx_free;
2042
49560532
PS		 2043rt_tx_error:
2044 ip_rt_put(rt);
d342894c 2045tx_error:
2046 dev->stats.tx_errors++;
2047tx_free:
2048 dev_kfree_skb(skb);
d342894c 2049}
2050
6681712d
DS		 2051/* Transmit local packets over Vxlan
2052 *
2053 * Outer IP header inherits ECN and DF from inner header.
2054 * Outer UDP destination is the VXLAN assigned port.
2055 * source port is based on hash of flow
2056 */
2057static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2058{
2059 struct vxlan_dev *vxlan = netdev_priv(dev);
2060 struct ethhdr *eth;
2061 bool did_rsc = false;
8f646c92 2062 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2063 struct vxlan_fdb *f;
6681712d
DS		 2064
2065 skb_reset_mac_header(skb);
2066 eth = eth_hdr(skb);
2067
f564f45c
CW		 2068 if ((vxlan->flags & VXLAN_F_PROXY)) {
2069 if (ntohs(eth->h_proto) == ETH_P_ARP)
2070 return arp_reduce(dev, skb);
2071#if IS_ENABLED(CONFIG_IPV6)
2072 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2073 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2074 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2075 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2076 struct nd_msg *msg;
2077
2078 msg = (struct nd_msg *)skb_transport_header(skb);
2079 if (msg->icmph.icmp6_code == 0 &&
2080 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2081 return neigh_reduce(dev, skb);
2082 }
7a9f526f 2083 eth = eth_hdr(skb);
f564f45c
CW		 2084#endif
2085 }
6681712d
DS		 2086
2087 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2088 did_rsc = false;
2089
2090 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2091 (ntohs(eth->h_proto) == ETH_P_IP ||
2092 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2093 did_rsc = route_shortcircuit(dev, skb);
2094 if (did_rsc)
2095 f = vxlan_find_mac(vxlan, eth->h_dest);
2096 }
2097
6681712d 2098 if (f == NULL) {
afbd8bae
MR		 2099 f = vxlan_find_mac(vxlan, all_zeros_mac);
2100 if (f == NULL) {
2101 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2102 !is_multicast_ether_addr(eth->h_dest))
2103 vxlan_fdb_miss(vxlan, eth->h_dest);
2104
2105 dev->stats.tx_dropped++;
8f646c92 2106 kfree_skb(skb);
afbd8bae
MR		 2107 return NETDEV_TX_OK;
2108 }
2109 }
6681712d 2110
afbd8bae
MR		 2111 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2112 struct sk_buff *skb1;
6681712d 2113
8f646c92
ED		 2114 if (!fdst) {
2115 fdst = rdst;
2116 continue;
2117 }
afbd8bae
MR		 2118 skb1 = skb_clone(skb, GFP_ATOMIC);
2119 if (skb1)
2120 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2121 }
2122
8f646c92
ED		 2123 if (fdst)
2124 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2125 else
2126 kfree_skb(skb);
4ad16930 2127 return NETDEV_TX_OK;
6681712d
DS		 2128}
2129
d342894c 2130/* Walk the forwarding table and purge stale entries */
2131static void vxlan_cleanup(unsigned long arg)
2132{
2133 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2134 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2135 unsigned int h;
2136
2137 if (!netif_running(vxlan->dev))
2138 return;
2139
2140 spin_lock_bh(&vxlan->hash_lock);
2141 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2142 struct hlist_node *p, *n;
2143 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2144 struct vxlan_fdb *f
2145 = container_of(p, struct vxlan_fdb, hlist);
2146 unsigned long timeout;
2147
3c172868 2148 if (f->state & NUD_PERMANENT)
d342894c 2149 continue;
2150
2151 timeout = f->used + vxlan->age_interval * HZ;
2152 if (time_before_eq(timeout, jiffies)) {
2153 netdev_dbg(vxlan->dev,
2154 "garbage collect %pM\n",
2155 f->eth_addr);
2156 f->state = NUD_STALE;
2157 vxlan_fdb_destroy(vxlan, f);
2158 } else if (time_before(timeout, next_timer))
2159 next_timer = timeout;
2160 }
2161 }
2162 spin_unlock_bh(&vxlan->hash_lock);
2163
2164 mod_timer(&vxlan->age_timer, next_timer);
2165}
2166
9c2e24e1
PS		 2167static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2168{
2169 __u32 vni = vxlan->default_dst.remote_vni;
2170
2171 vxlan->vn_sock = vs;
2172 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2173}
2174
d342894c 2175/* Setup stats when device is created */
2176static int vxlan_init(struct net_device *dev)
2177{
1c51a915 2178 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2179 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2180 struct vxlan_sock *vs;
19ca9fc1 2181 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
1c51a915 2182
1c213bd2 2183 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2184 if (!dev->tstats)
d342894c 2185 return -ENOMEM;
2186
1c51a915 2187 spin_lock(&vn->sock_lock);
19ca9fc1
ML		 2188 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2189 vxlan->dst_port);
00c83b01 2190 if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
1c51a915 2191 /* If we have a socket with same port already, reuse it */
9c2e24e1 2192 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 2193 } else {
2194 /* otherwise make new socket outside of RTNL */
2195 dev_hold(dev);
2196 queue_work(vxlan_wq, &vxlan->sock_work);
2197 }
2198 spin_unlock(&vn->sock_lock);
2199
d342894c 2200 return 0;
2201}
2202
ba609e9b 2203static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2204{
2205 struct vxlan_fdb *f;
2206
2207 spin_lock_bh(&vxlan->hash_lock);
2208 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2209 if (f)
2210 vxlan_fdb_destroy(vxlan, f);
2211 spin_unlock_bh(&vxlan->hash_lock);
2212}
2213
ebf4063e
SH		 2214static void vxlan_uninit(struct net_device *dev)
2215{
2216 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 2217 struct vxlan_sock *vs = vxlan->vn_sock;
2218
ba609e9b 2219 vxlan_fdb_delete_default(vxlan);
afbd8bae 2220
ebf4063e 2221 if (vs)
012a5729 2222 vxlan_sock_release(vs);
ebf4063e
SH		 2223 free_percpu(dev->tstats);
2224}
2225
d342894c 2226/* Start ageing timer and join group when device is brought up */
2227static int vxlan_open(struct net_device *dev)
2228{
2229 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 2230 struct vxlan_sock *vs = vxlan->vn_sock;
2231
2232 /* socket hasn't been created */
2233 if (!vs)
2234 return -ENOTCONN;
d342894c 2235
79d4a94f 2236 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1c51a915 2237 vxlan_sock_hold(vs);
7c47cedf 2238 dev_hold(dev);
3fc2de2f 2239 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 2240 }
2241
2242 if (vxlan->age_interval)
2243 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2244
2245 return 0;
2246}
2247
2248/* Purge the forwarding table */
2249static void vxlan_flush(struct vxlan_dev *vxlan)
2250{
31fec5aa 2251 unsigned int h;
d342894c 2252
2253 spin_lock_bh(&vxlan->hash_lock);
2254 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2255 struct hlist_node *p, *n;
2256 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2257 struct vxlan_fdb *f
2258 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2259 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2260 if (!is_zero_ether_addr(f->eth_addr))
2261 vxlan_fdb_destroy(vxlan, f);
d342894c 2262 }
2263 }
2264 spin_unlock_bh(&vxlan->hash_lock);
2265}
2266
2267/* Cleanup timer and forwarding table on shutdown */
2268static int vxlan_stop(struct net_device *dev)
2269{
2270 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2271 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1c51a915 2272 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 2273
e4c7ed41 2274 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
95ab0991 2275 !vxlan_group_used(vn, vxlan)) {
1c51a915 2276 vxlan_sock_hold(vs);
7c47cedf 2277 dev_hold(dev);
3fc2de2f 2278 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 2279 }
d342894c 2280
2281 del_timer_sync(&vxlan->age_timer);
2282
2283 vxlan_flush(vxlan);
2284
2285 return 0;
2286}
2287
d342894c 2288/* Stub, nothing needs to be done. */
2289static void vxlan_set_multicast_list(struct net_device *dev)
2290{
2291}
2292
345010b5
DB		 2293static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2294{
2295 struct vxlan_dev *vxlan = netdev_priv(dev);
2296 struct vxlan_rdst *dst = &vxlan->default_dst;
2297 struct net_device *lowerdev;
2298 int max_mtu;
2299
f01ec1c0 2300 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2301 if (lowerdev == NULL)
2302 return eth_change_mtu(dev, new_mtu);
2303
2304 if (dst->remote_ip.sa.sa_family == AF_INET6)
2305 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2306 else
2307 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2308
2309 if (new_mtu < 68 || new_mtu > max_mtu)
2310 return -EINVAL;
2311
2312 dev->mtu = new_mtu;
2313 return 0;
2314}
2315
d342894c 2316static const struct net_device_ops vxlan_netdev_ops = {
2317 .ndo_init = vxlan_init,
ebf4063e 2318 .ndo_uninit = vxlan_uninit,
d342894c 2319 .ndo_open = vxlan_open,
2320 .ndo_stop = vxlan_stop,
2321 .ndo_start_xmit = vxlan_xmit,
e8171045 2322 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2323 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2324 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2325 .ndo_validate_addr = eth_validate_addr,
2326 .ndo_set_mac_address = eth_mac_addr,
2327 .ndo_fdb_add = vxlan_fdb_add,
2328 .ndo_fdb_del = vxlan_fdb_delete,
2329 .ndo_fdb_dump = vxlan_fdb_dump,
2330};
2331
2332/* Info for udev, that this is a virtual tunnel endpoint */
2333static struct device_type vxlan_type = {
2334 .name = "vxlan",
2335};
2336
53cf5275 2337/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2338 * supply the listening VXLAN udp ports. Callers are expected
2339 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2340 */
2341void vxlan_get_rx_port(struct net_device *dev)
2342{
2343 struct vxlan_sock *vs;
2344 struct net *net = dev_net(dev);
2345 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2346 sa_family_t sa_family;
35e42379
JG		 2347 __be16 port;
2348 unsigned int i;
53cf5275
JG		 2349
2350 spin_lock(&vn->sock_lock);
2351 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2352 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2353 port = inet_sk(vs->sock->sk)->inet_sport;
53cf5275
JG		 2354 sa_family = vs->sock->sk->sk_family;
2355 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2356 port);
2357 }
2358 }
2359 spin_unlock(&vn->sock_lock);
2360}
2361EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2362
d342894c 2363/* Initialize the device structure. */
2364static void vxlan_setup(struct net_device *dev)
2365{
2366 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2367 unsigned int h;
d342894c 2368
2369 eth_hw_addr_random(dev);
2370 ether_setup(dev);
e4c7ed41 2371 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2853af6a 2372 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
e4c7ed41 2373 else
2853af6a 2374 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
d342894c 2375
2376 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2377 dev->destructor = free_netdev;
d342894c 2378 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2379
2380 dev->tx_queue_len = 0;
2381 dev->features |= NETIF_F_LLTX;
d6727fe3 2382 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2383 dev->features |= NETIF_F_RXCSUM;
05c0db08 2384 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2385
1eaa8178
PS		 2386 dev->vlan_features = dev->features;
2387 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2388 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2389 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2390 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2391 netif_keep_dst(dev);
6602d007 2392 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 2393
553675fb 2394 INIT_LIST_HEAD(&vxlan->next);
d342894c 2395 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 2396 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2397 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 2398 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 2399
2400 init_timer_deferrable(&vxlan->age_timer);
2401 vxlan->age_timer.function = vxlan_cleanup;
2402 vxlan->age_timer.data = (unsigned long) vxlan;
2403
823aa873 2404 vxlan->dst_port = htons(vxlan_port);
05f47d69 2405
d342894c 2406 vxlan->dev = dev;
2407
2408 for (h = 0; h < FDB_HASH_SIZE; ++h)
2409 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2410}
2411
2412static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2413 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2414 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2415 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2416 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2417 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2418 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2419 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2420 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2421 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2422 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2423 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2424 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2425 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2426 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2427 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2428 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 2429 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2430 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2431 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2432 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2433 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2434 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2435 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
d342894c 2436};
2437
2438static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2439{
2440 if (tb[IFLA_ADDRESS]) {
2441 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2442 pr_debug("invalid link address (not ethernet)\n");
2443 return -EINVAL;
2444 }
2445
2446 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2447 pr_debug("invalid all zero ethernet address\n");
2448 return -EADDRNOTAVAIL;
2449 }
2450 }
2451
2452 if (!data)
2453 return -EINVAL;
2454
2455 if (data[IFLA_VXLAN_ID]) {
2456 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2457 if (id >= VXLAN_VID_MASK)
2458 return -ERANGE;
2459 }
2460
05f47d69 2461 if (data[IFLA_VXLAN_PORT_RANGE]) {
2462 const struct ifla_vxlan_port_range *p
2463 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2464
2465 if (ntohs(p->high) < ntohs(p->low)) {
2466 pr_debug("port range %u .. %u not valid\n",
2467 ntohs(p->low), ntohs(p->high));
2468 return -EINVAL;
2469 }
2470 }
2471
d342894c 2472 return 0;
2473}
2474
1b13c97f
YB		 2475static void vxlan_get_drvinfo(struct net_device *netdev,
2476 struct ethtool_drvinfo *drvinfo)
2477{
2478 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2479 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2480}
2481
2482static const struct ethtool_ops vxlan_ethtool_ops = {
2483 .get_drvinfo = vxlan_get_drvinfo,
2484 .get_link = ethtool_op_get_link,
2485};
2486
553675fb 2487static void vxlan_del_work(struct work_struct *work)
2488{
2489 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2490 udp_tunnel_sock_release(vs->sock);
553675fb 2491 kfree_rcu(vs, rcu);
2492}
2493
3ee64f39
TH		 2494static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2495 __be16 port, u32 flags)
553675fb 2496{
e4c7ed41 2497 struct socket *sock;
3ee64f39
TH		 2498 struct udp_port_cfg udp_conf;
2499 int err;
e4c7ed41 2500
3ee64f39 2501 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2502
3ee64f39
TH		 2503 if (ipv6) {
2504 udp_conf.family = AF_INET6;
2505 udp_conf.use_udp6_tx_checksums =
3dc2b6a8 2506 !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
3ee64f39 2507 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2508 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3ee64f39
TH		 2509 } else {
2510 udp_conf.family = AF_INET;
2511 udp_conf.local_ip.s_addr = INADDR_ANY;
2512 udp_conf.use_udp_checksums =
2513 !!(flags & VXLAN_F_UDP_CSUM);
e4c7ed41 2514 }
e4c7ed41 2515
3ee64f39 2516 udp_conf.local_udp_port = port;
553675fb 2517
3ee64f39
TH		 2518 /* Open UDP socket */
2519 err = udp_sock_create(net, &udp_conf, &sock);
2520 if (err < 0)
2521 return ERR_PTR(err);
e4c7ed41 2522
39deb2c7 2523 return sock;
e4c7ed41
CW		 2524}
2525
2526/* Create new listen socket if needed */
2527static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
359a0ea9
TH		 2528 vxlan_rcv_t *rcv, void *data,
2529 u32 flags)
e4c7ed41
CW		 2530{
2531 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2532 struct vxlan_sock *vs;
2533 struct socket *sock;
e4c7ed41 2534 unsigned int h;
359a0ea9 2535 bool ipv6 = !!(flags & VXLAN_F_IPV6);
acbf74a7 2536 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2537
dc01e7d3 2538 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2539 if (!vs)
2540 return ERR_PTR(-ENOMEM);
2541
2542 for (h = 0; h < VNI_HASH_SIZE; ++h)
2543 INIT_HLIST_HEAD(&vs->vni_list[h]);
2544
2545 INIT_WORK(&vs->del_work, vxlan_del_work);
2546
3ee64f39 2547 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2548 if (IS_ERR(sock)) {
553675fb 2549 kfree(vs);
e50fddc8 2550 return ERR_CAST(sock);
553675fb 2551 }
e4c7ed41
CW		 2552
2553 vs->sock = sock;
9c2e24e1 2554 atomic_set(&vs->refcnt, 1);
5cfccc5a 2555 vs->rcv = rcv;
012a5729 2556 vs->data = data;
dfd8645e 2557 vs->flags = flags;
553675fb 2558
dc01e7d3
OG		 2559 /* Initialize the vxlan udp offloads structure */
2560 vs->udp_offloads.port = port;
2561 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2562 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2563
9c2e24e1
PS		 2564 spin_lock(&vn->sock_lock);
2565 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2566 vxlan_notify_add_rx_port(vs);
9c2e24e1 2567 spin_unlock(&vn->sock_lock);
553675fb 2568
2569 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2570 tunnel_cfg.sk_user_data = vs;
2571 tunnel_cfg.encap_type = 1;
2572 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2573 tunnel_cfg.encap_destroy = NULL;
2574
2575 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2576
9c2e24e1
PS		 2577 return vs;
2578}
2579
012a5729
PS		 2580struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2581 vxlan_rcv_t *rcv, void *data,
359a0ea9 2582 bool no_share, u32 flags)
9c2e24e1
PS		 2583{
2584 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2585 struct vxlan_sock *vs;
19ca9fc1 2586 bool ipv6 = flags & VXLAN_F_IPV6;
9c2e24e1 2587
359a0ea9 2588 vs = vxlan_socket_create(net, port, rcv, data, flags);
9c2e24e1
PS		 2589 if (!IS_ERR(vs))
2590 return vs;
553675fb 2591
012a5729
PS		 2592 if (no_share) /* Return error if sharing is not allowed. */
2593 return vs;
2594
9c2e24e1 2595 spin_lock(&vn->sock_lock);
19ca9fc1 2596 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port);
00c83b01
ML		 2597 if (vs && ((vs->rcv != rcv) ||
2598 !atomic_add_unless(&vs->refcnt, 1, 0)))
5cfccc5a 2599 vs = ERR_PTR(-EBUSY);
5cfccc5a
PS		 2600 spin_unlock(&vn->sock_lock);
2601
2602 if (!vs)
9c2e24e1
PS		 2603 vs = ERR_PTR(-EINVAL);
2604
553675fb 2605 return vs;
2606}
012a5729 2607EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 2608
1c51a915
SH		 2609/* Scheduled at device creation to bind to a socket */
2610static void vxlan_sock_work(struct work_struct *work)
2611{
9c2e24e1 2612 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
f01ec1c0 2613 struct net *net = vxlan->net;
1c51a915 2614 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 2615 __be16 port = vxlan->dst_port;
2616 struct vxlan_sock *nvs;
1c51a915 2617
359a0ea9 2618 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
1c51a915 2619 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2620 if (!IS_ERR(nvs))
2621 vxlan_vs_add_dev(nvs, vxlan);
2622 spin_unlock(&vn->sock_lock);
2623
2624 dev_put(vxlan->dev);
1c51a915
SH		 2625}
2626
d342894c 2627static int vxlan_newlink(struct net *net, struct net_device *dev,
2628 struct nlattr *tb[], struct nlattr *data[])
2629{
553675fb 2630 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 2631 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2632 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 2633 __u32 vni;
2634 int err;
e4c7ed41 2635 bool use_ipv6 = false;
d342894c 2636
2637 if (!data[IFLA_VXLAN_ID])
2638 return -EINVAL;
2639
f01ec1c0
ND		 2640 vxlan->net = dev_net(dev);
2641
d342894c 2642 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 2643 dst->remote_vni = vni;
d342894c 2644
5933a7bb
MR		 2645 /* Unless IPv6 is explicitly requested, assume IPv4 */
2646 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41
CW		 2647 if (data[IFLA_VXLAN_GROUP]) {
2648 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
e4c7ed41
CW		 2649 } else if (data[IFLA_VXLAN_GROUP6]) {
2650 if (!IS_ENABLED(CONFIG_IPV6))
2651 return -EPFNOSUPPORT;
2652
2653 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2654 sizeof(struct in6_addr));
2655 dst->remote_ip.sa.sa_family = AF_INET6;
2656 use_ipv6 = true;
2657 }
d342894c 2658
e4c7ed41
CW		 2659 if (data[IFLA_VXLAN_LOCAL]) {
2660 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2661 vxlan->saddr.sa.sa_family = AF_INET;
2662 } else if (data[IFLA_VXLAN_LOCAL6]) {
2663 if (!IS_ENABLED(CONFIG_IPV6))
2664 return -EPFNOSUPPORT;
2665
2666 /* TODO: respect scope id */
2667 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2668 sizeof(struct in6_addr));
2669 vxlan->saddr.sa.sa_family = AF_INET6;
2670 use_ipv6 = true;
2671 }
d342894c 2672
34e02aa1 2673 if (data[IFLA_VXLAN_LINK] &&
c7995c43 2674 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 2675 struct net_device *lowerdev
c7995c43 2676 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 2677
2678 if (!lowerdev) {
c7995c43 2679 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2680 return -ENODEV;
2681 }
d342894c 2682
e4c7ed41
CW		 2683#if IS_ENABLED(CONFIG_IPV6)
2684 if (use_ipv6) {
2685 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2686 if (idev && idev->cnf.disable_ipv6) {
2687 pr_info("IPv6 is disabled via sysctl\n");
2688 return -EPERM;
2689 }
2690 vxlan->flags |= VXLAN_F_IPV6;
2691 }
2692#endif
2693
34e02aa1 2694 if (!tb[IFLA_MTU])
e4c7ed41 2695 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2696
2853af6a 2697 dev->needed_headroom = lowerdev->hard_header_len +
e4c7ed41 2698 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
7bda701e 2699 } else if (use_ipv6)
2700 vxlan->flags |= VXLAN_F_IPV6;
d342894c 2701
2702 if (data[IFLA_VXLAN_TOS])
2703 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2704
afb97186
VB		 2705 if (data[IFLA_VXLAN_TTL])
2706 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2707
d342894c 2708 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 2709 vxlan->flags |= VXLAN_F_LEARN;
d342894c 2710
2711 if (data[IFLA_VXLAN_AGEING])
2712 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2713 else
2714 vxlan->age_interval = FDB_AGE_DEFAULT;
2715
e4f67add
DS		 2716 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2717 vxlan->flags |= VXLAN_F_PROXY;
2718
2719 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2720 vxlan->flags |= VXLAN_F_RSC;
2721
2722 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2723 vxlan->flags |= VXLAN_F_L2MISS;
2724
2725 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2726 vxlan->flags |= VXLAN_F_L3MISS;
2727
d342894c 2728 if (data[IFLA_VXLAN_LIMIT])
2729 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2730
05f47d69 2731 if (data[IFLA_VXLAN_PORT_RANGE]) {
2732 const struct ifla_vxlan_port_range *p
2733 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2734 vxlan->port_min = ntohs(p->low);
2735 vxlan->port_max = ntohs(p->high);
2736 }
2737
823aa873 2738 if (data[IFLA_VXLAN_PORT])
2739 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2740
359a0ea9
TH		 2741 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2742 vxlan->flags |= VXLAN_F_UDP_CSUM;
2743
2744 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2745 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2746 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2747
2748 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2749 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2750 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2751
dfd8645e
TH		 2752 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2753 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2754 vxlan->flags |= VXLAN_F_REMCSUM_TX;
2755
2756 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2757 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2758 vxlan->flags |= VXLAN_F_REMCSUM_RX;
2759
3511494c
TG		 2760 if (data[IFLA_VXLAN_GBP])
2761 vxlan->flags |= VXLAN_F_GBP;
2762
19ca9fc1
ML		 2763 if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
2764 vxlan->dst_port)) {
553675fb 2765 pr_info("duplicate VNI %u\n", vni);
2766 return -EEXIST;
2767 }
2768
7ad24ea4 2769 dev->ethtool_ops = &vxlan_ethtool_ops;
1b13c97f 2770
2936b6ab
SS		 2771 /* create an fdb entry for a valid default destination */
2772 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2773 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2774 &vxlan->default_dst.remote_ip,
2775 NUD_REACHABLE|NUD_PERMANENT,
2776 NLM_F_EXCL|NLM_F_CREATE,
2777 vxlan->dst_port,
2778 vxlan->default_dst.remote_vni,
2779 vxlan->default_dst.remote_ifindex,
2780 NTF_SELF);
2781 if (err)
2782 return err;
2783 }
d342894c 2784
afbd8bae
MR		 2785 err = register_netdevice(dev);
2786 if (err) {
ba609e9b 2787 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 2788 return err;
2789 }
2790
553675fb 2791 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 2792
2793 return 0;
d342894c 2794}
2795
2796static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2797{
2798 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2799 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2800
fe5c3561 2801 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2802 if (!hlist_unhashed(&vxlan->hlist))
2803 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2804 spin_unlock(&vn->sock_lock);
2805
553675fb 2806 list_del(&vxlan->next);
d342894c 2807 unregister_netdevice_queue(dev, head);
2808}
2809
2810static size_t vxlan_get_size(const struct net_device *dev)
2811{
2812
2813 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2814 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2815 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2816 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2817 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2818 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2819 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2820 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2821 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2822 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2823 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 2824 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2825 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2826 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2827 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2828 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2829 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2830 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2831 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2832 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2833 0;
2834}
2835
2836static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2837{
2838 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2839 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2840 struct ifla_vxlan_port_range ports = {
2841 .low = htons(vxlan->port_min),
2842 .high = htons(vxlan->port_max),
2843 };
d342894c 2844
c7995c43 2845 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 2846 goto nla_put_failure;
2847
e4c7ed41
CW		 2848 if (!vxlan_addr_any(&dst->remote_ip)) {
2849 if (dst->remote_ip.sa.sa_family == AF_INET) {
2850 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2851 dst->remote_ip.sin.sin_addr.s_addr))
2852 goto nla_put_failure;
2853#if IS_ENABLED(CONFIG_IPV6)
2854 } else {
2855 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2856 &dst->remote_ip.sin6.sin6_addr))
2857 goto nla_put_failure;
2858#endif
2859 }
2860 }
d342894c 2861
c7995c43 2862 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 2863 goto nla_put_failure;
2864
e4c7ed41
CW		 2865 if (!vxlan_addr_any(&vxlan->saddr)) {
2866 if (vxlan->saddr.sa.sa_family == AF_INET) {
2867 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2868 vxlan->saddr.sin.sin_addr.s_addr))
2869 goto nla_put_failure;
2870#if IS_ENABLED(CONFIG_IPV6)
2871 } else {
2872 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2873 &vxlan->saddr.sin6.sin6_addr))
2874 goto nla_put_failure;
2875#endif
2876 }
2877 }
d342894c 2878
2879 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2880 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 2881 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2882 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2883 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2884 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2885 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2886 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2887 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2888 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2889 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 2890 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 2891 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
359a0ea9
TH		 2892 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2893 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2894 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2895 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2896 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2897 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 2898 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2899 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2900 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2901 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2902 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 2903 goto nla_put_failure;
2904
05f47d69 2905 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2906 goto nla_put_failure;
2907
3511494c
TG		 2908 if (vxlan->flags & VXLAN_F_GBP &&
2909 nla_put_flag(skb, IFLA_VXLAN_GBP))
2910 goto nla_put_failure;
2911
d342894c 2912 return 0;
2913
2914nla_put_failure:
2915 return -EMSGSIZE;
2916}
2917
2918static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2919 .kind = "vxlan",
2920 .maxtype = IFLA_VXLAN_MAX,
2921 .policy = vxlan_policy,
2922 .priv_size = sizeof(struct vxlan_dev),
2923 .setup = vxlan_setup,
2924 .validate = vxlan_validate,
2925 .newlink = vxlan_newlink,
2926 .dellink = vxlan_dellink,
2927 .get_size = vxlan_get_size,
2928 .fill_info = vxlan_fill_info,
2929};
2930
acaf4e70
DB		 2931static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2932 struct net_device *dev)
2933{
2934 struct vxlan_dev *vxlan, *next;
2935 LIST_HEAD(list_kill);
2936
2937 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2938 struct vxlan_rdst *dst = &vxlan->default_dst;
2939
2940 /* In case we created vxlan device with carrier
2941 * and we loose the carrier due to module unload
2942 * we also need to remove vxlan device. In other
2943 * cases, it's not necessary and remote_ifindex
2944 * is 0 here, so no matches.
2945 */
2946 if (dst->remote_ifindex == dev->ifindex)
2947 vxlan_dellink(vxlan->dev, &list_kill);
2948 }
2949
2950 unregister_netdevice_many(&list_kill);
2951}
2952
2953static int vxlan_lowerdev_event(struct notifier_block *unused,
2954 unsigned long event, void *ptr)
2955{
2956 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 2957 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 2958
783c1463 2959 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 2960 vxlan_handle_lowerdev_unregister(vn, dev);
2961
2962 return NOTIFY_DONE;
2963}
2964
2965static struct notifier_block vxlan_notifier_block __read_mostly = {
2966 .notifier_call = vxlan_lowerdev_event,
2967};
2968
d342894c 2969static __net_init int vxlan_init_net(struct net *net)
2970{
2971 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 2972 unsigned int h;
d342894c 2973
553675fb 2974 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 2975 spin_lock_init(&vn->sock_lock);
d342894c 2976
553675fb 2977 for (h = 0; h < PORT_HASH_SIZE; ++h)
2978 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 2979
2980 return 0;
2981}
2982
f01ec1c0
ND		 2983static void __net_exit vxlan_exit_net(struct net *net)
2984{
2985 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2986 struct vxlan_dev *vxlan, *next;
2987 struct net_device *dev, *aux;
2988 LIST_HEAD(list);
2989
2990 rtnl_lock();
2991 for_each_netdev_safe(net, dev, aux)
2992 if (dev->rtnl_link_ops == &vxlan_link_ops)
2993 unregister_netdevice_queue(dev, &list);
2994
2995 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2996 /* If vxlan->dev is in the same netns, it has already been added
2997 * to the list by the previous loop.
2998 */
2999 if (!net_eq(dev_net(vxlan->dev), net))
3000 unregister_netdevice_queue(dev, &list);
3001 }
3002
3003 unregister_netdevice_many(&list);
3004 rtnl_unlock();
3005}
3006
d342894c 3007static struct pernet_operations vxlan_net_ops = {
3008 .init = vxlan_init_net,
f01ec1c0 3009 .exit = vxlan_exit_net,
d342894c 3010 .id = &vxlan_net_id,
3011 .size = sizeof(struct vxlan_net),
3012};
3013
3014static int __init vxlan_init_module(void)
3015{
3016 int rc;
3017
758c57d1
SH		 3018 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3019 if (!vxlan_wq)
3020 return -ENOMEM;
3021
d342894c 3022 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3023
783c1463 3024 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3025 if (rc)
3026 goto out1;
3027
acaf4e70 3028 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3029 if (rc)
3030 goto out2;
3031
acaf4e70
DB		 3032 rc = rtnl_link_register(&vxlan_link_ops);
3033 if (rc)
3034 goto out3;
d342894c 3035
acaf4e70
DB		 3036 return 0;
3037out3:
3038 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3039out2:
783c1463 3040 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3041out1:
758c57d1 3042 destroy_workqueue(vxlan_wq);
d342894c 3043 return rc;
3044}
7332a13b 3045late_initcall(vxlan_init_module);
d342894c 3046
3047static void __exit vxlan_cleanup_module(void)
3048{
b7153984 3049 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3050 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3051 destroy_workqueue(vxlan_wq);
783c1463
DB		 3052 unregister_pernet_subsys(&vxlan_net_ops);
3053 /* rcu_barrier() is called by netns */
d342894c 3054}
3055module_exit(vxlan_cleanup_module);
3056
3057MODULE_LICENSE("GPL");
3058MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3059MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3060MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3061MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.3833 seconds and 5 git commands to generate.