projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
udp: enable busy polling for all sockets
[linux-2.6-block.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
d342894c 14#include <linux/module.h>
15#include <linux/errno.h>
16#include <linux/slab.h>
d342894c 17#include <linux/udp.h>
18#include <linux/igmp.h>
d342894c 19#include <linux/if_ether.h>
1b13c97f 20#include <linux/ethtool.h>
e4f67add
DS		 21#include <net/arp.h>
22#include <net/ndisc.h>
d342894c 23#include <net/ip.h>
24#include <net/icmp.h>
d342894c 25#include <net/rtnetlink.h>
d342894c 26#include <net/inet_ecn.h>
27#include <net/net_namespace.h>
28#include <net/netns/generic.h>
012a5729 29#include <net/vxlan.h>
40d29af0 30
e4c7ed41 31#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 32#include <net/ip6_tunnel.h>
660d98ca 33#include <net/ip6_checksum.h>
e4c7ed41 34#endif
d342894c 35
36#define VXLAN_VERSION "0.1"
37
553675fb 38#define PORT_HASH_BITS 8
39#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 40#define FDB_AGE_DEFAULT 300 /* 5 min */
41#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42
23c578bf 43/* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 45 * for compatibility with early adopters.
23c578bf 46 */
9daaa397
SH		 47static unsigned short vxlan_port __read_mostly = 8472;
48module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 49MODULE_PARM_DESC(udp_port, "Destination UDP port");
50
51static bool log_ecn_error = true;
52module_param(log_ecn_error, bool, 0644);
53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54
60d9d4c6 55static int vxlan_net_id;
0dfbdf41 56static struct rtnl_link_ops vxlan_link_ops;
553675fb 57
7256eac1 58static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 59
205f356d 60static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 61
553675fb 62/* per-network namespace private data for this module */
63struct vxlan_net {
64 struct list_head vxlan_list;
65 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 66 spinlock_t sock_lock;
553675fb 67};
68
d342894c 69/* Forwarding table entry */
70struct vxlan_fdb {
71 struct hlist_node hlist; /* linked list of entries */
72 struct rcu_head rcu;
73 unsigned long updated; /* jiffies */
74 unsigned long used;
3e61aa8f 75 struct list_head remotes;
7177a3b0 76 u8 eth_addr[ETH_ALEN];
d342894c 77 u16 state; /* see ndm_state */
ae884082 78 u8 flags; /* see ndm_flags */
d342894c 79};
80
d342894c 81/* salt for hash table */
82static u32 vxlan_salt __read_mostly;
83
ee122c79
TG		 84static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
85{
e7030878
TG		 86 return vs->flags & VXLAN_F_COLLECT_METADATA ||
87 ip_tunnel_collect_metadata();
ee122c79
TG		 88}
89
e4c7ed41
CW		 90#if IS_ENABLED(CONFIG_IPV6)
91static inline
92bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
93{
f0ef3126
JB		 94 if (a->sa.sa_family != b->sa.sa_family)
95 return false;
96 if (a->sa.sa_family == AF_INET6)
97 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
98 else
99 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 100}
101
102static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
103{
f0ef3126
JB		 104 if (ipa->sa.sa_family == AF_INET6)
105 return ipv6_addr_any(&ipa->sin6.sin6_addr);
106 else
107 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 108}
109
110static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
111{
f0ef3126
JB		 112 if (ipa->sa.sa_family == AF_INET6)
113 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
114 else
115 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 116}
117
118static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
119{
f0ef3126 120 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 121 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 122 ip->sa.sa_family = AF_INET6;
123 return 0;
124 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 125 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 126 ip->sa.sa_family = AF_INET;
127 return 0;
128 } else {
129 return -EAFNOSUPPORT;
130 }
e4c7ed41
CW		 131}
132
133static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 134 const union vxlan_addr *ip)
e4c7ed41 135{
f0ef3126 136 if (ip->sa.sa_family == AF_INET6)
930345ea 137 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 138 else
930345ea 139 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 140}
141
142#else /* !CONFIG_IPV6 */
143
144static inline
145bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
146{
f0ef3126 147 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 148}
149
150static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
151{
f0ef3126 152 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 153}
154
155static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
156{
f0ef3126 157 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 158}
159
160static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
161{
f0ef3126
JB		 162 if (nla_len(nla) >= sizeof(struct in6_addr)) {
163 return -EAFNOSUPPORT;
164 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 165 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 166 ip->sa.sa_family = AF_INET;
167 return 0;
168 } else {
169 return -EAFNOSUPPORT;
170 }
e4c7ed41
CW		 171}
172
173static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 174 const union vxlan_addr *ip)
e4c7ed41 175{
930345ea 176 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 177}
178#endif
179
553675fb 180/* Virtual Network hash table head */
54bfd872 181static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 182{
54bfd872 183 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 184}
185
186/* Socket hash table head */
187static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 188{
189 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
190
553675fb 191 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
192}
193
3e61aa8f
SH		 194/* First remote destination for a forwarding entry.
195 * Guaranteed to be non-NULL because remotes are never deleted.
196 */
5ca5461c 197static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 198{
5ca5461c 199 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
200}
201
202static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
203{
204 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 205}
206
ac5132d1
TG		 207/* Find VXLAN socket based on network namespace, address family and UDP port
208 * and enabled unshareable flags.
209 */
210static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
211 __be16 port, u32 flags)
553675fb 212{
213 struct vxlan_sock *vs;
af33c1ad
TH		 214
215 flags &= VXLAN_F_RCV_FLAGS;
553675fb 216
217 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 218 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 219 vxlan_get_sk_family(vs) == family &&
af33c1ad 220 vs->flags == flags)
553675fb 221 return vs;
222 }
223 return NULL;
d342894c 224}
225
54bfd872 226static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 227{
228 struct vxlan_dev *vxlan;
d342894c 229
b9167b2e
JB		 230 /* For flow based devices, map all packets to VNI 0 */
231 if (vs->flags & VXLAN_F_COLLECT_METADATA)
232 vni = 0;
233
54bfd872
JB		 234 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
235 if (vxlan->default_dst.remote_vni == vni)
d342894c 236 return vxlan;
237 }
238
239 return NULL;
240}
241
5cfccc5a 242/* Look up VNI in a per net namespace table */
54bfd872 243static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 244 sa_family_t family, __be16 port,
245 u32 flags)
5cfccc5a
PS		 246{
247 struct vxlan_sock *vs;
248
ac5132d1 249 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 250 if (!vs)
251 return NULL;
252
54bfd872 253 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 254}
255
d342894c 256/* Fill in neighbour message in skbuff. */
257static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 258 const struct vxlan_fdb *fdb,
259 u32 portid, u32 seq, int type, unsigned int flags,
260 const struct vxlan_rdst *rdst)
d342894c 261{
262 unsigned long now = jiffies;
263 struct nda_cacheinfo ci;
264 struct nlmsghdr *nlh;
265 struct ndmsg *ndm;
e4f67add 266 bool send_ip, send_eth;
d342894c 267
268 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
269 if (nlh == NULL)
270 return -EMSGSIZE;
271
272 ndm = nlmsg_data(nlh);
273 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 274
275 send_eth = send_ip = true;
276
277 if (type == RTM_GETNEIGH) {
278 ndm->ndm_family = AF_INET;
e4c7ed41 279 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 280 send_eth = !is_zero_ether_addr(fdb->eth_addr);
281 } else
282 ndm->ndm_family = AF_BRIDGE;
d342894c 283 ndm->ndm_state = fdb->state;
284 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 285 ndm->ndm_flags = fdb->flags;
545469f7 286 ndm->ndm_type = RTN_UNICAST;
d342894c 287
193523bf 288 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 289 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 290 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 291 goto nla_put_failure;
292
e4f67add 293 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 294 goto nla_put_failure;
295
e4c7ed41 296 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 297 goto nla_put_failure;
298
0dfbdf41 299 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 300 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
301 goto nla_put_failure;
c7995c43 302 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 303 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d
DS		 304 goto nla_put_failure;
305 if (rdst->remote_ifindex &&
306 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 307 goto nla_put_failure;
308
309 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
310 ci.ndm_confirmed = 0;
311 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
312 ci.ndm_refcnt = 0;
313
314 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
315 goto nla_put_failure;
316
053c095a
JB		 317 nlmsg_end(skb, nlh);
318 return 0;
d342894c 319
320nla_put_failure:
321 nlmsg_cancel(skb, nlh);
322 return -EMSGSIZE;
323}
324
325static inline size_t vxlan_nlmsg_size(void)
326{
327 return NLMSG_ALIGN(sizeof(struct ndmsg))
328 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 329 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 330 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 331 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
332 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 333 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 334 + nla_total_size(sizeof(struct nda_cacheinfo));
335}
336
9e4b93f9
ND		 337static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
338 struct vxlan_rdst *rd, int type)
d342894c 339{
340 struct net *net = dev_net(vxlan->dev);
341 struct sk_buff *skb;
342 int err = -ENOBUFS;
343
344 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
345 if (skb == NULL)
346 goto errout;
347
9e4b93f9 348 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 349 if (err < 0) {
350 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
351 WARN_ON(err == -EMSGSIZE);
352 kfree_skb(skb);
353 goto errout;
354 }
355
356 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
357 return;
358errout:
359 if (err < 0)
360 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
361}
362
e4c7ed41 363static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 364{
365 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 366 struct vxlan_fdb f = {
367 .state = NUD_STALE,
368 };
369 struct vxlan_rdst remote = {
e4c7ed41 370 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 371 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 372 };
3e61aa8f 373
9e4b93f9 374 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 375}
376
377static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
378{
bb3fd687
SH		 379 struct vxlan_fdb f = {
380 .state = NUD_STALE,
381 };
9e4b93f9 382 struct vxlan_rdst remote = { };
e4f67add 383
e4f67add
DS		 384 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
385
9e4b93f9 386 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 387}
388
d342894c 389/* Hash Ethernet address */
390static u32 eth_hash(const unsigned char *addr)
391{
392 u64 value = get_unaligned((u64 *)addr);
393
394 /* only want 6 bytes */
395#ifdef __BIG_ENDIAN
d342894c 396 value >>= 16;
321fb991 397#else
398 value <<= 16;
d342894c 399#endif
400 return hash_64(value, FDB_HASH_BITS);
401}
402
403/* Hash chain to use given mac address */
404static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
405 const u8 *mac)
406{
407 return &vxlan->fdb_head[eth_hash(mac)];
408}
409
410/* Look up Ethernet address in forwarding table */
014be2c8 411static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 412 const u8 *mac)
d342894c 413{
414 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
415 struct vxlan_fdb *f;
d342894c 416
b67bfe0d 417 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 418 if (ether_addr_equal(mac, f->eth_addr))
d342894c 419 return f;
420 }
421
422 return NULL;
423}
424
014be2c8
SS		 425static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
426 const u8 *mac)
427{
428 struct vxlan_fdb *f;
429
430 f = __vxlan_find_mac(vxlan, mac);
431 if (f)
432 f->used = jiffies;
433
434 return f;
435}
436
a5e7c10a
MR		 437/* caller should hold vxlan->hash_lock */
438static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 439 union vxlan_addr *ip, __be16 port,
54bfd872 440 __be32 vni, __u32 ifindex)
6681712d 441{
3e61aa8f 442 struct vxlan_rdst *rd;
6681712d 443
3e61aa8f 444 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 445 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 446 rd->remote_port == port &&
447 rd->remote_vni == vni &&
448 rd->remote_ifindex == ifindex)
a5e7c10a 449 return rd;
6681712d 450 }
3e61aa8f 451
a5e7c10a
MR		 452 return NULL;
453}
454
906dc186
TR		 455/* Replace destination of unicast mac */
456static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 457 union vxlan_addr *ip, __be16 port, __be32 vni,
458 __u32 ifindex)
906dc186
TR		 459{
460 struct vxlan_rdst *rd;
461
462 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
463 if (rd)
464 return 0;
465
466 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
467 if (!rd)
468 return 0;
0c1d70af
PA		 469
470 dst_cache_reset(&rd->dst_cache);
e4c7ed41 471 rd->remote_ip = *ip;
906dc186
TR		 472 rd->remote_port = port;
473 rd->remote_vni = vni;
474 rd->remote_ifindex = ifindex;
475 return 1;
476}
477
a5e7c10a
MR		 478/* Add/update destinations for multicast */
479static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 480 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 481 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 482{
483 struct vxlan_rdst *rd;
484
485 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
486 if (rd)
487 return 0;
488
6681712d
DS		 489 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
490 if (rd == NULL)
491 return -ENOBUFS;
0c1d70af
PA		 492
493 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
494 kfree(rd);
495 return -ENOBUFS;
496 }
497
e4c7ed41 498 rd->remote_ip = *ip;
6681712d
DS		 499 rd->remote_port = port;
500 rd->remote_vni = vni;
501 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 502
503 list_add_tail_rcu(&rd->list, &f->remotes);
504
9e4b93f9 505 *rdp = rd;
6681712d
DS		 506 return 1;
507}
508
dfd8645e
TH		 509static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
510 unsigned int off,
511 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 512 __be32 vni_field,
513 struct gro_remcsum *grc,
0ace2ca8 514 bool nopartial)
dfd8645e 515{
b7fe10e5 516 size_t start, offset;
dfd8645e
TH		 517
518 if (skb->remcsum_offload)
b7fe10e5 519 return vh;
dfd8645e
TH		 520
521 if (!NAPI_GRO_CB(skb)->csum_valid)
522 return NULL;
523
54bfd872
JB		 524 start = vxlan_rco_start(vni_field);
525 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 526
b7fe10e5
TH		 527 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
528 start, offset, grc, nopartial);
dfd8645e
TH		 529
530 skb->remcsum_offload = 1;
531
532 return vh;
533}
534
5602c48c
TH		 535static struct sk_buff **vxlan_gro_receive(struct sock *sk,
536 struct sk_buff **head,
537 struct sk_buff *skb)
dc01e7d3
OG		 538{
539 struct sk_buff *p, **pp = NULL;
540 struct vxlanhdr *vh, *vh2;
9b174d88 541 unsigned int hlen, off_vx;
dc01e7d3 542 int flush = 1;
5602c48c 543 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 544 __be32 flags;
26c4f7da
TH		 545 struct gro_remcsum grc;
546
547 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 548
549 off_vx = skb_gro_offset(skb);
550 hlen = off_vx + sizeof(*vh);
551 vh = skb_gro_header_fast(skb, off_vx);
552 if (skb_gro_header_hard(skb, hlen)) {
553 vh = skb_gro_header_slow(skb, hlen, off_vx);
554 if (unlikely(!vh))
555 goto out;
556 }
dc01e7d3 557
dfd8645e
TH		 558 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
559
54bfd872 560 flags = vh->vx_flags;
dfd8645e
TH		 561
562 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
563 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 564 vh->vx_vni, &grc,
0ace2ca8
TH		 565 !!(vs->flags &
566 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 567
568 if (!vh)
569 goto out;
570 }
571
b7fe10e5
TH		 572 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
573
dc01e7d3
OG		 574 for (p = *head; p; p = p->next) {
575 if (!NAPI_GRO_CB(p)->same_flow)
576 continue;
577
578 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 579 if (vh->vx_flags != vh2->vx_flags ||
580 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 581 NAPI_GRO_CB(p)->same_flow = 0;
582 continue;
583 }
dc01e7d3
OG		 584 }
585
fcd91dd4 586 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 587 flush = 0;
dc01e7d3 588
dc01e7d3 589out:
26c4f7da 590 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 591 NAPI_GRO_CB(skb)->flush |= flush;
592
593 return pp;
594}
595
5602c48c 596static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 597{
229740c6
JR		 598 /* Sets 'skb->inner_mac_header' since we are always called with
599 * 'skb->encapsulation' set.
600 */
9b174d88 601 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 602}
603
d342894c 604/* Add new entry to forwarding table -- assumes lock held */
605static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 606 const u8 *mac, union vxlan_addr *ip,
6681712d 607 __u16 state, __u16 flags,
54bfd872 608 __be16 port, __be32 vni, __u32 ifindex,
ae884082 609 __u8 ndm_flags)
d342894c 610{
9e4b93f9 611 struct vxlan_rdst *rd = NULL;
d342894c 612 struct vxlan_fdb *f;
613 int notify = 0;
614
014be2c8 615 f = __vxlan_find_mac(vxlan, mac);
d342894c 616 if (f) {
617 if (flags & NLM_F_EXCL) {
618 netdev_dbg(vxlan->dev,
619 "lost race to create %pM\n", mac);
620 return -EEXIST;
621 }
622 if (f->state != state) {
623 f->state = state;
624 f->updated = jiffies;
625 notify = 1;
626 }
ae884082
DS		 627 if (f->flags != ndm_flags) {
628 f->flags = ndm_flags;
629 f->updated = jiffies;
630 notify = 1;
631 }
906dc186
TR		 632 if ((flags & NLM_F_REPLACE)) {
633 /* Only change unicasts */
634 if (!(is_multicast_ether_addr(f->eth_addr) ||
635 is_zero_ether_addr(f->eth_addr))) {
60840429 636 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 637 ifindex);
906dc186
TR		 638 } else
639 return -EOPNOTSUPP;
640 }
6681712d 641 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 642 (is_multicast_ether_addr(f->eth_addr) ||
643 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 644 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
645 &rd);
6681712d
DS		 646
647 if (rc < 0)
648 return rc;
649 notify |= rc;
650 }
d342894c 651 } else {
652 if (!(flags & NLM_F_CREATE))
653 return -ENOENT;
654
0dfbdf41
TG		 655 if (vxlan->cfg.addrmax &&
656 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 657 return -ENOSPC;
658
906dc186
TR		 659 /* Disallow replace to add a multicast entry */
660 if ((flags & NLM_F_REPLACE) &&
661 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
662 return -EOPNOTSUPP;
663
e4c7ed41 664 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 665 f = kmalloc(sizeof(*f), GFP_ATOMIC);
666 if (!f)
667 return -ENOMEM;
668
669 notify = 1;
d342894c 670 f->state = state;
ae884082 671 f->flags = ndm_flags;
d342894c 672 f->updated = f->used = jiffies;
3e61aa8f 673 INIT_LIST_HEAD(&f->remotes);
d342894c 674 memcpy(f->eth_addr, mac, ETH_ALEN);
675
9e4b93f9 676 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 677
d342894c 678 ++vxlan->addrcnt;
679 hlist_add_head_rcu(&f->hlist,
680 vxlan_fdb_head(vxlan, mac));
681 }
682
9e4b93f9
ND		 683 if (notify) {
684 if (rd == NULL)
685 rd = first_remote_rtnl(f);
686 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
687 }
d342894c 688
689 return 0;
690}
691
6706c82e 692static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 693{
694 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 695 struct vxlan_rdst *rd, *nd;
6681712d 696
0c1d70af
PA		 697 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
698 dst_cache_destroy(&rd->dst_cache);
6681712d 699 kfree(rd);
0c1d70af 700 }
6681712d
DS		 701 kfree(f);
702}
703
d342894c 704static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
705{
706 netdev_dbg(vxlan->dev,
707 "delete %pM\n", f->eth_addr);
708
709 --vxlan->addrcnt;
9e4b93f9 710 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 711
712 hlist_del_rcu(&f->hlist);
6681712d 713 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 714}
715
f0b074be 716static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
54bfd872
JB		 717 union vxlan_addr *ip, __be16 *port, __be32 *vni,
718 u32 *ifindex)
d342894c 719{
6681712d 720 struct net *net = dev_net(vxlan->dev);
e4c7ed41 721 int err;
d342894c 722
f0b074be 723 if (tb[NDA_DST]) {
e4c7ed41
CW		 724 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
725 if (err)
726 return err;
f0b074be 727 } else {
e4c7ed41
CW		 728 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
729 if (remote->sa.sa_family == AF_INET) {
730 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
731 ip->sa.sa_family = AF_INET;
732#if IS_ENABLED(CONFIG_IPV6)
733 } else {
734 ip->sin6.sin6_addr = in6addr_any;
735 ip->sa.sa_family = AF_INET6;
736#endif
737 }
f0b074be 738 }
d342894c 739
6681712d 740 if (tb[NDA_PORT]) {
73cf3317 741 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 742 return -EINVAL;
f0b074be
MR		 743 *port = nla_get_be16(tb[NDA_PORT]);
744 } else {
0dfbdf41 745 *port = vxlan->cfg.dst_port;
f0b074be 746 }
6681712d
DS		 747
748 if (tb[NDA_VNI]) {
749 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
750 return -EINVAL;
54bfd872 751 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 752 } else {
753 *vni = vxlan->default_dst.remote_vni;
754 }
6681712d
DS		 755
756 if (tb[NDA_IFINDEX]) {
5abb0029 757 struct net_device *tdev;
6681712d
DS		 758
759 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
760 return -EINVAL;
f0b074be 761 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 762 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 763 if (!tdev)
6681712d 764 return -EADDRNOTAVAIL;
f0b074be
MR		 765 } else {
766 *ifindex = 0;
767 }
768
769 return 0;
770}
771
772/* Add static entry (via netlink) */
773static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
774 struct net_device *dev,
f6f6424b 775 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 776{
777 struct vxlan_dev *vxlan = netdev_priv(dev);
778 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 779 union vxlan_addr ip;
f0b074be 780 __be16 port;
54bfd872
JB		 781 __be32 vni;
782 u32 ifindex;
f0b074be
MR		 783 int err;
784
785 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
786 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
787 ndm->ndm_state);
788 return -EINVAL;
789 }
790
791 if (tb[NDA_DST] == NULL)
792 return -EINVAL;
793
794 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
795 if (err)
796 return err;
6681712d 797
5933a7bb
MR		 798 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
799 return -EAFNOSUPPORT;
800
d342894c 801 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 802 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 803 port, vni, ifindex, ndm->ndm_flags);
d342894c 804 spin_unlock_bh(&vxlan->hash_lock);
805
806 return err;
807}
808
809/* Delete entry (via netlink) */
1690be63
VY		 810static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
811 struct net_device *dev,
f6f6424b 812 const unsigned char *addr, u16 vid)
d342894c 813{
814 struct vxlan_dev *vxlan = netdev_priv(dev);
815 struct vxlan_fdb *f;
bc7892ba 816 struct vxlan_rdst *rd = NULL;
e4c7ed41 817 union vxlan_addr ip;
bc7892ba 818 __be16 port;
54bfd872
JB		 819 __be32 vni;
820 u32 ifindex;
bc7892ba
MR		 821 int err;
822
823 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
824 if (err)
825 return err;
826
827 err = -ENOENT;
d342894c 828
829 spin_lock_bh(&vxlan->hash_lock);
830 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 831 if (!f)
832 goto out;
833
e4c7ed41
CW		 834 if (!vxlan_addr_any(&ip)) {
835 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 836 if (!rd)
837 goto out;
838 }
839
840 err = 0;
841
842 /* remove a destination if it's not the only one on the list,
843 * otherwise destroy the fdb entry
844 */
845 if (rd && !list_is_singular(&f->remotes)) {
846 list_del_rcu(&rd->list);
9e4b93f9 847 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 848 kfree_rcu(rd, rcu);
bc7892ba 849 goto out;
d342894c 850 }
bc7892ba
MR		 851
852 vxlan_fdb_destroy(vxlan, f);
853
854out:
d342894c 855 spin_unlock_bh(&vxlan->hash_lock);
856
857 return err;
858}
859
860/* Dump forwarding table */
861static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 862 struct net_device *dev,
d297653d 863 struct net_device *filter_dev, int *idx)
d342894c 864{
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 unsigned int h;
d297653d 867 int err = 0;
d342894c 868
869 for (h = 0; h < FDB_HASH_SIZE; ++h) {
870 struct vxlan_fdb *f;
d342894c 871
b67bfe0d 872 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 873 struct vxlan_rdst *rd;
6681712d 874
3e61aa8f 875 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 876 if (*idx < cb->args[2])
07a51cd3
AW		 877 goto skip;
878
6681712d
DS		 879 err = vxlan_fdb_info(skb, vxlan, f,
880 NETLINK_CB(cb->skb).portid,
881 cb->nlh->nlmsg_seq,
882 RTM_NEWNEIGH,
883 NLM_F_MULTI, rd);
d297653d 884 if (err < 0)
3e61aa8f 885 goto out;
3e61aa8f 886skip:
d297653d 887 *idx += 1;
07a51cd3 888 }
d342894c 889 }
890 }
3e61aa8f 891out:
d297653d 892 return err;
d342894c 893}
894
895/* Watch incoming packets to learn mapping between Ethernet address
896 * and Tunnel endpoint.
c4b49512 897 * Return true if packet is bogus and should be dropped.
d342894c 898 */
26a41ae6 899static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 900 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 901{
902 struct vxlan_dev *vxlan = netdev_priv(dev);
903 struct vxlan_fdb *f;
d342894c 904
905 f = vxlan_find_mac(vxlan, src_mac);
906 if (likely(f)) {
5ca5461c 907 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 908
e4c7ed41 909 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 910 return false;
911
912 /* Don't migrate static entries, drop packets */
eb064c3b 913 if (f->state & NUD_NOARP)
26a41ae6 914 return true;
d342894c 915
916 if (net_ratelimit())
917 netdev_info(dev,
e4c7ed41 918 "%pM migrated from %pIS to %pIS\n",
a4870f79 919 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 920
e4c7ed41 921 rdst->remote_ip = *src_ip;
d342894c 922 f->updated = jiffies;
9e4b93f9 923 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 924 } else {
925 /* learned new entry */
926 spin_lock(&vxlan->hash_lock);
3bf74b1a 927
928 /* close off race between vxlan_flush and incoming packets */
929 if (netif_running(dev))
930 vxlan_fdb_create(vxlan, src_mac, src_ip,
931 NUD_REACHABLE,
932 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 933 vxlan->cfg.dst_port,
3bf74b1a 934 vxlan->default_dst.remote_vni,
935 0, NTF_SELF);
d342894c 936 spin_unlock(&vxlan->hash_lock);
937 }
26a41ae6 938
939 return false;
d342894c 940}
941
d342894c 942/* See if multicast group is already in use by other ID */
95ab0991 943static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 944{
553675fb 945 struct vxlan_dev *vxlan;
c6fcc4fc 946 struct vxlan_sock *sock4;
4053ab1b
AB		 947#if IS_ENABLED(CONFIG_IPV6)
948 struct vxlan_sock *sock6;
949#endif
b1be00a6 950 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 951
c6fcc4fc 952 sock4 = rtnl_dereference(dev->vn4_sock);
953
95ab0991
G		 954 /* The vxlan_sock is only used by dev, leaving group has
955 * no effect on other vxlan devices.
956 */
c6fcc4fc 957 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
95ab0991 958 return false;
b1be00a6 959#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 960 sock6 = rtnl_dereference(dev->vn6_sock);
961 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
b1be00a6
JB		 962 return false;
963#endif
95ab0991 964
553675fb 965 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 966 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 967 continue;
d342894c 968
c6fcc4fc 969 if (family == AF_INET &&
970 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 971 continue;
b1be00a6 972#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 973 if (family == AF_INET6 &&
974 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 975 continue;
976#endif
95ab0991
G		 977
978 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
979 &dev->default_dst.remote_ip))
980 continue;
981
982 if (vxlan->default_dst.remote_ifindex !=
983 dev->default_dst.remote_ifindex)
984 continue;
985
986 return true;
553675fb 987 }
d342894c 988
989 return false;
990}
991
544a773a 992static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 993{
b1be00a6 994 struct vxlan_net *vn;
012a5729 995
b1be00a6 996 if (!vs)
544a773a 997 return false;
7c47cedf 998 if (!atomic_dec_and_test(&vs->refcnt))
544a773a 999 return false;
d342894c 1000
b1be00a6 1001 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1002 spin_lock(&vn->sock_lock);
7c47cedf 1003 hlist_del_rcu(&vs->hlist);
e7b3db5e 1004 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1005 (vs->flags & VXLAN_F_GPE) ?
1006 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1007 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1008 spin_unlock(&vn->sock_lock);
1009
544a773a 1010 return true;
d342894c 1011}
1012
b1be00a6
JB		 1013static void vxlan_sock_release(struct vxlan_dev *vxlan)
1014{
c6fcc4fc 1015 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1016#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1017 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1018
1019 rcu_assign_pointer(vxlan->vn6_sock, NULL);
544a773a
HFS		 1020#endif
1021
c6fcc4fc 1022 rcu_assign_pointer(vxlan->vn4_sock, NULL);
544a773a
HFS		 1023 synchronize_net();
1024
c6fcc4fc 1025 if (__vxlan_sock_release_prep(sock4)) {
1026 udp_tunnel_sock_release(sock4->sock);
1027 kfree(sock4);
544a773a
HFS		 1028 }
1029
1030#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1031 if (__vxlan_sock_release_prep(sock6)) {
1032 udp_tunnel_sock_release(sock6->sock);
1033 kfree(sock6);
544a773a 1034 }
b1be00a6
JB		 1035#endif
1036}
1037
56ef9c90 1038/* Update multicast group membership when first VNI on
c4b49512 1039 * multicast address is brought up
7c47cedf 1040 */
56ef9c90 1041static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1042{
b1be00a6 1043 struct sock *sk;
e4c7ed41
CW		 1044 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1045 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1046 int ret = -EINVAL;
d342894c 1047
e4c7ed41 1048 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1049 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1050 struct ip_mreqn mreq = {
1051 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1052 .imr_ifindex = ifindex,
1053 };
1054
c6fcc4fc 1055 sk = sock4->sock->sk;
b1be00a6 1056 lock_sock(sk);
56ef9c90 1057 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1058 release_sock(sk);
e4c7ed41
CW		 1059#if IS_ENABLED(CONFIG_IPV6)
1060 } else {
c6fcc4fc 1061 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1062
1063 sk = sock6->sock->sk;
b1be00a6 1064 lock_sock(sk);
56ef9c90
MRL		 1065 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1066 &ip->sin6.sin6_addr);
b1be00a6 1067 release_sock(sk);
e4c7ed41
CW		 1068#endif
1069 }
3fc2de2f 1070
56ef9c90 1071 return ret;
3fc2de2f 1072}
1073
1074/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1075static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1076{
b1be00a6 1077 struct sock *sk;
e4c7ed41
CW		 1078 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1079 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1080 int ret = -EINVAL;
3fc2de2f 1081
e4c7ed41 1082 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1083 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1084 struct ip_mreqn mreq = {
1085 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1086 .imr_ifindex = ifindex,
1087 };
1088
c6fcc4fc 1089 sk = sock4->sock->sk;
b1be00a6 1090 lock_sock(sk);
56ef9c90 1091 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1092 release_sock(sk);
e4c7ed41
CW		 1093#if IS_ENABLED(CONFIG_IPV6)
1094 } else {
c6fcc4fc 1095 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1096
1097 sk = sock6->sock->sk;
b1be00a6 1098 lock_sock(sk);
56ef9c90
MRL		 1099 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1100 &ip->sin6.sin6_addr);
b1be00a6 1101 release_sock(sk);
e4c7ed41
CW		 1102#endif
1103 }
d342894c 1104
56ef9c90 1105 return ret;
d342894c 1106}
1107
f14ecebb
JB		 1108static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1109 struct sk_buff *skb, u32 vxflags)
dfd8645e 1110{
7d34fa75 1111 size_t start, offset;
dfd8645e 1112
f14ecebb
JB		 1113 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1114 goto out;
b7fe10e5 1115
f14ecebb
JB		 1116 start = vxlan_rco_start(unparsed->vx_vni);
1117 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1118
7d34fa75 1119 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1120 return false;
dfd8645e 1121
be5cfeab
JB		 1122 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1123 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1124out:
1125 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1126 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1127 return true;
dfd8645e
TH		 1128}
1129
f14ecebb 1130static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1131 struct sk_buff *skb, u32 vxflags,
10a5af23 1132 struct vxlan_metadata *md)
3288af08 1133{
f14ecebb 1134 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1135 struct metadata_dst *tun_dst;
f14ecebb
JB		 1136
1137 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1138 goto out;
3288af08 1139
3288af08
JB		 1140 md->gbp = ntohs(gbp->policy_id);
1141
10a5af23 1142 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1143 if (tun_dst) {
3288af08 1144 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1145 tun_dst->u.tun_info.options_len = sizeof(*md);
1146 }
3288af08
JB		 1147 if (gbp->dont_learn)
1148 md->gbp |= VXLAN_GBP_DONT_LEARN;
1149
1150 if (gbp->policy_applied)
1151 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1152
64f87d36
JB		 1153 /* In flow-based mode, GBP is carried in dst_metadata */
1154 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1155 skb->mark = md->gbp;
f14ecebb
JB		 1156out:
1157 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1158}
1159
e1e5314d 1160static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1161 __be16 *protocol,
e1e5314d
JB		 1162 struct sk_buff *skb, u32 vxflags)
1163{
1164 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1165
1166 /* Need to have Next Protocol set for interfaces in GPE mode. */
1167 if (!gpe->np_applied)
1168 return false;
1169 /* "The initial version is 0. If a receiver does not support the
1170 * version indicated it MUST drop the packet.
1171 */
1172 if (gpe->version != 0)
1173 return false;
1174 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1175 * processing MUST occur." However, we don't implement OAM
1176 * processing, thus drop the packet.
1177 */
1178 if (gpe->oam_flag)
1179 return false;
1180
1181 switch (gpe->next_protocol) {
1182 case VXLAN_GPE_NP_IPV4:
1183 *protocol = htons(ETH_P_IP);
1184 break;
1185 case VXLAN_GPE_NP_IPV6:
1186 *protocol = htons(ETH_P_IPV6);
1187 break;
1188 case VXLAN_GPE_NP_ETHERNET:
1189 *protocol = htons(ETH_P_TEB);
1190 break;
1191 default:
1192 return false;
1193 }
1194
1195 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1196 return true;
1197}
1198
1ab016e2
JB		 1199static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1200 struct vxlan_sock *vs,
1201 struct sk_buff *skb)
614732ea 1202{
614732ea 1203 union vxlan_addr saddr;
614732ea 1204
614732ea 1205 skb_reset_mac_header(skb);
614732ea
TG		 1206 skb->protocol = eth_type_trans(skb, vxlan->dev);
1207 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1208
1209 /* Ignore packet loops (and multicast echo) */
1210 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1211 return false;
614732ea 1212
760c6805 1213 /* Get address from the outer IP header */
ce212d0f 1214 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1215 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1216 saddr.sa.sa_family = AF_INET;
1217#if IS_ENABLED(CONFIG_IPV6)
1218 } else {
1ab016e2 1219 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1220 saddr.sa.sa_family = AF_INET6;
1221#endif
1222 }
1223
1ab016e2
JB		 1224 if ((vxlan->flags & VXLAN_F_LEARN) &&
1225 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1226 return false;
1227
1228 return true;
1229}
1230
760c6805
JB		 1231static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1232 struct sk_buff *skb)
1233{
1234 int err = 0;
1235
1236 if (vxlan_get_sk_family(vs) == AF_INET)
1237 err = IP_ECN_decapsulate(oiph, skb);
1238#if IS_ENABLED(CONFIG_IPV6)
1239 else
1240 err = IP6_ECN_decapsulate(oiph, skb);
1241#endif
1242
1243 if (unlikely(err) && log_ecn_error) {
1244 if (vxlan_get_sk_family(vs) == AF_INET)
1245 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1246 &((struct iphdr *)oiph)->saddr,
1247 ((struct iphdr *)oiph)->tos);
1248 else
1249 net_info_ratelimited("non-ECT from %pI6\n",
1250 &((struct ipv6hdr *)oiph)->saddr);
1251 }
1252 return err <= 1;
1253}
1254
d342894c 1255/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1256static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1257{
f2d1968e 1258 struct pcpu_sw_netstats *stats;
c9e78efb 1259 struct vxlan_dev *vxlan;
5cfccc5a 1260 struct vxlan_sock *vs;
f14ecebb 1261 struct vxlanhdr unparsed;
ee122c79
TG		 1262 struct vxlan_metadata _md;
1263 struct vxlan_metadata *md = &_md;
61618eea 1264 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1265 bool raw_proto = false;
f2d1968e 1266 void *oiph;
d342894c 1267
e1e5314d 1268 /* Need UDP and VXLAN header to be present */
7ce04758 1269 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1270 goto drop;
d342894c 1271
f14ecebb 1272 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1273 /* VNI flag always required to be set */
1274 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1275 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1276 ntohl(vxlan_hdr(skb)->vx_flags),
1277 ntohl(vxlan_hdr(skb)->vx_vni));
1278 /* Return non vxlan pkt */
e5aed006 1279 goto drop;
d342894c 1280 }
288b01c8
JB		 1281 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1282 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1283
559835ea 1284 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1285 if (!vs)
d342894c 1286 goto drop;
d342894c 1287
c9e78efb
JB		 1288 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1289 if (!vxlan)
1290 goto drop;
1291
e1e5314d
JB		 1292 /* For backwards compatibility, only allow reserved fields to be
1293 * used by VXLAN extensions if explicitly requested.
1294 */
1295 if (vs->flags & VXLAN_F_GPE) {
1296 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1297 goto drop;
1298 raw_proto = true;
1299 }
1300
1301 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1302 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1303 goto drop;
c9e78efb 1304
ee122c79 1305 if (vxlan_collect_metadata(vs)) {
07dabf20 1306 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
10a5af23 1307 struct metadata_dst *tun_dst;
07dabf20 1308
c29a70d2 1309 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1310 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1311
ee122c79
TG		 1312 if (!tun_dst)
1313 goto drop;
1314
0f1b7354 1315 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1316
1317 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1318 } else {
1319 memset(md, 0, sizeof(*md));
1320 }
1321
f14ecebb
JB		 1322 if (vs->flags & VXLAN_F_REMCSUM_RX)
1323 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1324 goto drop;
1325 if (vs->flags & VXLAN_F_GBP)
10a5af23 1326 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1327 /* Note that GBP and GPE can never be active together. This is
1328 * ensured in vxlan_dev_configure.
1329 */
3511494c 1330
f14ecebb 1331 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1332 /* If there are any unprocessed flags remaining treat
1333 * this as a malformed packet. This behavior diverges from
1334 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1335 * in reserved fields are to be ignored. The approach here
c4b49512 1336 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1337 * is more robust and provides a little more security in
1338 * adding extensions to VXLAN.
1339 */
288b01c8 1340 goto drop;
3bf39475
TH		 1341 }
1342
e1e5314d
JB		 1343 if (!raw_proto) {
1344 if (!vxlan_set_mac(vxlan, vs, skb))
1345 goto drop;
1346 } else {
8be0cfa4 1347 skb_reset_mac_header(skb);
e1e5314d
JB		 1348 skb->dev = vxlan->dev;
1349 skb->pkt_type = PACKET_HOST;
1350 }
f2d1968e 1351
f2d1968e
JB		 1352 oiph = skb_network_header(skb);
1353 skb_reset_network_header(skb);
1354
1355 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1356 ++vxlan->dev->stats.rx_frame_errors;
1357 ++vxlan->dev->stats.rx_errors;
1358 goto drop;
1359 }
1360
1361 stats = this_cpu_ptr(vxlan->dev->tstats);
1362 u64_stats_update_begin(&stats->syncp);
1363 stats->rx_packets++;
1364 stats->rx_bytes += skb->len;
1365 u64_stats_update_end(&stats->syncp);
1366
1367 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1368 return 0;
1369
1370drop:
288b01c8
JB		 1371 /* Consume bad packet */
1372 kfree_skb(skb);
1373 return 0;
5cfccc5a
PS		 1374}
1375
e4f67add
DS		 1376static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1377{
1378 struct vxlan_dev *vxlan = netdev_priv(dev);
1379 struct arphdr *parp;
1380 u8 *arpptr, *sha;
1381 __be32 sip, tip;
1382 struct neighbour *n;
1383
1384 if (dev->flags & IFF_NOARP)
1385 goto out;
1386
1387 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1388 dev->stats.tx_dropped++;
1389 goto out;
1390 }
1391 parp = arp_hdr(skb);
1392
1393 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1394 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1395 parp->ar_pro != htons(ETH_P_IP) ||
1396 parp->ar_op != htons(ARPOP_REQUEST) ||
1397 parp->ar_hln != dev->addr_len ||
1398 parp->ar_pln != 4)
1399 goto out;
1400 arpptr = (u8 *)parp + sizeof(struct arphdr);
1401 sha = arpptr;
1402 arpptr += dev->addr_len; /* sha */
1403 memcpy(&sip, arpptr, sizeof(sip));
1404 arpptr += sizeof(sip);
1405 arpptr += dev->addr_len; /* tha */
1406 memcpy(&tip, arpptr, sizeof(tip));
1407
1408 if (ipv4_is_loopback(tip) ||
1409 ipv4_is_multicast(tip))
1410 goto out;
1411
1412 n = neigh_lookup(&arp_tbl, &tip, dev);
1413
1414 if (n) {
e4f67add
DS		 1415 struct vxlan_fdb *f;
1416 struct sk_buff *reply;
1417
1418 if (!(n->nud_state & NUD_CONNECTED)) {
1419 neigh_release(n);
1420 goto out;
1421 }
1422
1423 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1424 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1425 /* bridge-local neighbor */
1426 neigh_release(n);
1427 goto out;
1428 }
1429
1430 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1431 n->ha, sha);
1432
1433 neigh_release(n);
1434
7346135d
DS		 1435 if (reply == NULL)
1436 goto out;
1437
e4f67add
DS		 1438 skb_reset_mac_header(reply);
1439 __skb_pull(reply, skb_network_offset(reply));
1440 reply->ip_summed = CHECKSUM_UNNECESSARY;
1441 reply->pkt_type = PACKET_HOST;
1442
1443 if (netif_rx_ni(reply) == NET_RX_DROP)
1444 dev->stats.rx_dropped++;
e4c7ed41
CW		 1445 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1446 union vxlan_addr ipa = {
1447 .sin.sin_addr.s_addr = tip,
a45e92a5 1448 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1449 };
1450
1451 vxlan_ip_miss(dev, &ipa);
1452 }
e4f67add
DS		 1453out:
1454 consume_skb(skb);
1455 return NETDEV_TX_OK;
1456}
1457
f564f45c 1458#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1459static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1460 struct neighbour *n, bool isrouter)
1461{
1462 struct net_device *dev = request->dev;
1463 struct sk_buff *reply;
1464 struct nd_msg *ns, *na;
1465 struct ipv6hdr *pip6;
1466 u8 *daddr;
1467 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1468 int ns_olen;
1469 int i, len;
1470
1471 if (dev == NULL)
1472 return NULL;
1473
1474 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1475 sizeof(*na) + na_olen + dev->needed_tailroom;
1476 reply = alloc_skb(len, GFP_ATOMIC);
1477 if (reply == NULL)
1478 return NULL;
1479
1480 reply->protocol = htons(ETH_P_IPV6);
1481 reply->dev = dev;
1482 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1483 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1484 skb_reset_mac_header(reply);
4b29dba9
DS		 1485
1486 ns = (struct nd_msg *)skb_transport_header(request);
1487
1488 daddr = eth_hdr(request)->h_source;
1489 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1490 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1491 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1492 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1493 break;
1494 }
1495 }
1496
1497 /* Ethernet header */
1498 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1499 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1500 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1501 reply->protocol = htons(ETH_P_IPV6);
1502
1503 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1504 skb_reset_network_header(reply);
4b29dba9
DS		 1505 skb_put(reply, sizeof(struct ipv6hdr));
1506
1507 /* IPv6 header */
1508
1509 pip6 = ipv6_hdr(reply);
1510 memset(pip6, 0, sizeof(struct ipv6hdr));
1511 pip6->version = 6;
1512 pip6->priority = ipv6_hdr(request)->priority;
1513 pip6->nexthdr = IPPROTO_ICMPV6;
1514 pip6->hop_limit = 255;
1515 pip6->daddr = ipv6_hdr(request)->saddr;
1516 pip6->saddr = *(struct in6_addr *)n->primary_key;
1517
1518 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1519 skb_reset_transport_header(reply);
4b29dba9
DS		 1520
1521 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1522
1523 /* Neighbor Advertisement */
1524 memset(na, 0, sizeof(*na)+na_olen);
1525 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1526 na->icmph.icmp6_router = isrouter;
1527 na->icmph.icmp6_override = 1;
1528 na->icmph.icmp6_solicited = 1;
1529 na->target = ns->target;
1530 ether_addr_copy(&na->opt[2], n->ha);
1531 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1532 na->opt[1] = na_olen >> 3;
1533
1534 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1535 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1536 csum_partial(na, sizeof(*na)+na_olen, 0));
1537
1538 pip6->payload_len = htons(sizeof(*na)+na_olen);
1539
1540 skb_push(reply, sizeof(struct ipv6hdr));
1541
1542 reply->ip_summed = CHECKSUM_UNNECESSARY;
1543
1544 return reply;
1545}
1546
f564f45c
CW		 1547static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1548{
1549 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1550 struct nd_msg *msg;
f564f45c
CW		 1551 const struct ipv6hdr *iphdr;
1552 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1553 struct neighbour *n;
1554 struct inet6_dev *in6_dev;
f564f45c
CW		 1555
1556 in6_dev = __in6_dev_get(dev);
1557 if (!in6_dev)
1558 goto out;
1559
f564f45c
CW		 1560 iphdr = ipv6_hdr(skb);
1561 saddr = &iphdr->saddr;
1562 daddr = &iphdr->daddr;
1563
f564f45c
CW		 1564 msg = (struct nd_msg *)skb_transport_header(skb);
1565 if (msg->icmph.icmp6_code != 0 ||
1566 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1567 goto out;
1568
4b29dba9
DS		 1569 if (ipv6_addr_loopback(daddr) ||
1570 ipv6_addr_is_multicast(&msg->target))
1571 goto out;
1572
1573 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1574
1575 if (n) {
1576 struct vxlan_fdb *f;
4b29dba9 1577 struct sk_buff *reply;
f564f45c
CW		 1578
1579 if (!(n->nud_state & NUD_CONNECTED)) {
1580 neigh_release(n);
1581 goto out;
1582 }
1583
1584 f = vxlan_find_mac(vxlan, n->ha);
1585 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1586 /* bridge-local neighbor */
1587 neigh_release(n);
1588 goto out;
1589 }
1590
4b29dba9
DS		 1591 reply = vxlan_na_create(skb, n,
1592 !!(f ? f->flags & NTF_ROUTER : 0));
1593
f564f45c 1594 neigh_release(n);
4b29dba9
DS		 1595
1596 if (reply == NULL)
1597 goto out;
1598
1599 if (netif_rx_ni(reply) == NET_RX_DROP)
1600 dev->stats.rx_dropped++;
1601
f564f45c 1602 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1603 union vxlan_addr ipa = {
1604 .sin6.sin6_addr = msg->target,
a45e92a5 1605 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1606 };
1607
f564f45c
CW		 1608 vxlan_ip_miss(dev, &ipa);
1609 }
1610
1611out:
1612 consume_skb(skb);
1613 return NETDEV_TX_OK;
1614}
1615#endif
1616
e4f67add
DS		 1617static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1618{
1619 struct vxlan_dev *vxlan = netdev_priv(dev);
1620 struct neighbour *n;
e4f67add
DS		 1621
1622 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1623 return false;
1624
1625 n = NULL;
1626 switch (ntohs(eth_hdr(skb)->h_proto)) {
1627 case ETH_P_IP:
e15a00aa
CW		 1628 {
1629 struct iphdr *pip;
1630
e4f67add
DS		 1631 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1632 return false;
1633 pip = ip_hdr(skb);
1634 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1635 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1636 union vxlan_addr ipa = {
1637 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1638 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1639 };
1640
1641 vxlan_ip_miss(dev, &ipa);
1642 return false;
1643 }
1644
e4f67add 1645 break;
e15a00aa
CW		 1646 }
1647#if IS_ENABLED(CONFIG_IPV6)
1648 case ETH_P_IPV6:
1649 {
1650 struct ipv6hdr *pip6;
1651
1652 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1653 return false;
1654 pip6 = ipv6_hdr(skb);
1655 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1656 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1657 union vxlan_addr ipa = {
1658 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1659 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1660 };
1661
1662 vxlan_ip_miss(dev, &ipa);
1663 return false;
1664 }
1665
1666 break;
1667 }
1668#endif
e4f67add
DS		 1669 default:
1670 return false;
1671 }
1672
1673 if (n) {
1674 bool diff;
1675
7367d0b5 1676 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1677 if (diff) {
1678 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1679 dev->addr_len);
1680 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1681 }
1682 neigh_release(n);
1683 return diff;
e4c7ed41
CW		 1684 }
1685
e4f67add
DS		 1686 return false;
1687}
1688
af33c1ad 1689static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1690 struct vxlan_metadata *md)
1691{
1692 struct vxlanhdr_gbp *gbp;
1693
db79a621
TG		 1694 if (!md->gbp)
1695 return;
1696
3511494c 1697 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1698 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1699
1700 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1701 gbp->dont_learn = 1;
1702
1703 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1704 gbp->policy_applied = 1;
1705
1706 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1707}
1708
e1e5314d
JB		 1709static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1710 __be16 protocol)
1711{
1712 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1713
1714 gpe->np_applied = 1;
1715
1716 switch (protocol) {
1717 case htons(ETH_P_IP):
1718 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1719 return 0;
1720 case htons(ETH_P_IPV6):
1721 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1722 return 0;
1723 case htons(ETH_P_TEB):
1724 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1725 return 0;
1726 }
1727 return -EPFNOSUPPORT;
1728}
1729
f491e56d
JB		 1730static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1731 int iphdr_len, __be32 vni,
1732 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1733 bool udp_sum)
e4c7ed41 1734{
e4c7ed41 1735 struct vxlanhdr *vxh;
e4c7ed41
CW		 1736 int min_headroom;
1737 int err;
dfd8645e 1738 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1739 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1740
af33c1ad 1741 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1742 skb->ip_summed == CHECKSUM_PARTIAL) {
1743 int csum_start = skb_checksum_start_offset(skb);
1744
1745 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1746 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1747 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1748 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1749 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1750 }
e4c7ed41 1751
e4c7ed41 1752 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1753 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1754
1755 /* Need space for new headers (invalidates iph ptr) */
1756 err = skb_cow_head(skb, min_headroom);
e1e5314d 1757 if (unlikely(err))
c46b7897 1758 return err;
649c5b8b 1759
aed069df
AD		 1760 err = iptunnel_handle_offloads(skb, type);
1761 if (err)
c46b7897 1762 return err;
b736a623 1763
49560532 1764 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1765 vxh->vx_flags = VXLAN_HF_VNI;
1766 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1767
dfd8645e 1768 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1769 unsigned int start;
dfd8645e 1770
54bfd872
JB		 1771 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1772 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1773 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1774
1775 if (!skb_is_gso(skb)) {
1776 skb->ip_summed = CHECKSUM_NONE;
1777 skb->encapsulation = 0;
1778 }
1779 }
1780
af33c1ad
TH		 1781 if (vxflags & VXLAN_F_GBP)
1782 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1783 if (vxflags & VXLAN_F_GPE) {
1784 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1785 if (err < 0)
c46b7897 1786 return err;
e1e5314d
JB		 1787 inner_protocol = skb->protocol;
1788 }
3511494c 1789
e1e5314d 1790 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1791 return 0;
49560532 1792}
49560532 1793
655c3de1 1794static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1795 struct vxlan_sock *sock4,
1a8496ba 1796 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1797 __be32 daddr, __be32 *saddr,
1798 struct dst_cache *dst_cache,
db3c6139 1799 const struct ip_tunnel_info *info)
1a8496ba 1800{
db3c6139 1801 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1802 struct rtable *rt = NULL;
1803 struct flowi4 fl4;
1804
655c3de1 1805 if (!sock4)
1806 return ERR_PTR(-EIO);
1807
db3c6139
DB		 1808 if (tos && !info)
1809 use_cache = false;
1810 if (use_cache) {
0c1d70af
PA		 1811 rt = dst_cache_get_ip4(dst_cache, saddr);
1812 if (rt)
1813 return rt;
1814 }
1815
1a8496ba
JB		 1816 memset(&fl4, 0, sizeof(fl4));
1817 fl4.flowi4_oif = oif;
1818 fl4.flowi4_tos = RT_TOS(tos);
1819 fl4.flowi4_mark = skb->mark;
1820 fl4.flowi4_proto = IPPROTO_UDP;
1821 fl4.daddr = daddr;
272d96a5 1822 fl4.saddr = *saddr;
1a8496ba
JB		 1823
1824 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1825 if (likely(!IS_ERR(rt))) {
1826 if (rt->dst.dev == dev) {
1827 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1828 ip_rt_put(rt);
1829 return ERR_PTR(-ELOOP);
1830 }
1831
1a8496ba 1832 *saddr = fl4.saddr;
0c1d70af
PA		 1833 if (use_cache)
1834 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1835 } else {
1836 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1837 return ERR_PTR(-ENETUNREACH);
0c1d70af 1838 }
1a8496ba
JB		 1839 return rt;
1840}
1841
e5d4b29f
JB		 1842#if IS_ENABLED(CONFIG_IPV6)
1843static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1844 struct net_device *dev,
03dc52a8 1845 struct vxlan_sock *sock6,
1400615d 1846 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1847 __be32 label,
e5d4b29f 1848 const struct in6_addr *daddr,
0c1d70af 1849 struct in6_addr *saddr,
db3c6139
DB		 1850 struct dst_cache *dst_cache,
1851 const struct ip_tunnel_info *info)
e5d4b29f 1852{
db3c6139 1853 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1854 struct dst_entry *ndst;
1855 struct flowi6 fl6;
1856 int err;
1857
c6fcc4fc 1858 if (!sock6)
1859 return ERR_PTR(-EIO);
1860
1400615d
DB		 1861 if (tos && !info)
1862 use_cache = false;
db3c6139 1863 if (use_cache) {
0c1d70af
PA		 1864 ndst = dst_cache_get_ip6(dst_cache, saddr);
1865 if (ndst)
1866 return ndst;
1867 }
1868
e5d4b29f
JB		 1869 memset(&fl6, 0, sizeof(fl6));
1870 fl6.flowi6_oif = oif;
1871 fl6.daddr = *daddr;
272d96a5 1872 fl6.saddr = *saddr;
eaa93bf4 1873 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1874 fl6.flowi6_mark = skb->mark;
1875 fl6.flowi6_proto = IPPROTO_UDP;
1876
1877 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1878 sock6->sock->sk,
e5d4b29f 1879 &ndst, &fl6);
655c3de1 1880 if (unlikely(err < 0)) {
1881 netdev_dbg(dev, "no route to %pI6\n", daddr);
1882 return ERR_PTR(-ENETUNREACH);
1883 }
1884
1885 if (unlikely(ndst->dev == dev)) {
1886 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1887 dst_release(ndst);
1888 return ERR_PTR(-ELOOP);
1889 }
e5d4b29f
JB		 1890
1891 *saddr = fl6.saddr;
db3c6139 1892 if (use_cache)
0c1d70af 1893 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1894 return ndst;
1895}
1896#endif
1897
9dcc71e1
SS		 1898/* Bypass encapsulation if the destination is local */
1899static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1900 struct vxlan_dev *dst_vxlan)
1901{
8f84985f 1902 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1903 union vxlan_addr loopback;
1904 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1905 struct net_device *dev = skb->dev;
1906 int len = skb->len;
9dcc71e1 1907
8f84985f
LR		 1908 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1909 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1910 skb->pkt_type = PACKET_HOST;
1911 skb->encapsulation = 0;
1912 skb->dev = dst_vxlan->dev;
1913 __skb_pull(skb, skb_network_offset(skb));
1914
e4c7ed41
CW		 1915 if (remote_ip->sa.sa_family == AF_INET) {
1916 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1917 loopback.sa.sa_family = AF_INET;
1918#if IS_ENABLED(CONFIG_IPV6)
1919 } else {
1920 loopback.sin6.sin6_addr = in6addr_loopback;
1921 loopback.sa.sa_family = AF_INET6;
1922#endif
1923 }
1924
9dcc71e1 1925 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1926 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1927
1928 u64_stats_update_begin(&tx_stats->syncp);
1929 tx_stats->tx_packets++;
ce6502a8 1930 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1931 u64_stats_update_end(&tx_stats->syncp);
1932
1933 if (netif_rx(skb) == NET_RX_SUCCESS) {
1934 u64_stats_update_begin(&rx_stats->syncp);
1935 rx_stats->rx_packets++;
ce6502a8 1936 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1937 u64_stats_update_end(&rx_stats->syncp);
1938 } else {
ce6502a8 1939 dev->stats.rx_dropped++;
9dcc71e1
SS		 1940 }
1941}
1942
fee1fad7 1943static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
1944 struct vxlan_dev *vxlan, union vxlan_addr *daddr,
1945 __be32 dst_port, __be32 vni, struct dst_entry *dst,
1946 u32 rt_flags)
1947{
1948#if IS_ENABLED(CONFIG_IPV6)
1949 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
1950 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
1951 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
1952 */
1953 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
1954#endif
1955 /* Bypass encapsulation if the destination is local */
1956 if (rt_flags & RTCF_LOCAL &&
1957 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1958 struct vxlan_dev *dst_vxlan;
1959
1960 dst_release(dst);
1961 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1962 daddr->sa.sa_family, dst_port,
1963 vxlan->flags);
1964 if (!dst_vxlan) {
1965 dev->stats.tx_errors++;
1966 kfree_skb(skb);
1967
1968 return -ENOENT;
1969 }
1970 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1971 return 1;
1972 }
1973
1974 return 0;
1975}
1976
4ad16930
SH		 1977static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1978 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1979{
d71785ff 1980 struct dst_cache *dst_cache;
3093fbe7 1981 struct ip_tunnel_info *info;
d342894c 1982 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 1983 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 1984 union vxlan_addr *dst;
272d96a5 1985 union vxlan_addr remote_ip, local_ip;
1986 union vxlan_addr *src;
ee122c79
TG		 1987 struct vxlan_metadata _md;
1988 struct vxlan_metadata *md = &_md;
e4c7ed41 1989 __be16 src_port = 0, dst_port;
655c3de1 1990 struct dst_entry *ndst = NULL;
e7f70af1 1991 __be32 vni, label;
d342894c 1992 __u8 tos, ttl;
0e6fbc5b 1993 int err;
ee122c79 1994 u32 flags = vxlan->flags;
b4ed5cad 1995 bool udp_sum = false;
f491e56d 1996 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1997
61adedf3 1998 info = skb_tunnel_info(skb);
3093fbe7 1999
ee122c79 2000 if (rdst) {
0770b53b 2001 dst = &rdst->remote_ip;
2002 if (vxlan_addr_any(dst)) {
2003 if (did_rsc) {
2004 /* short-circuited back to local bridge */
2005 vxlan_encap_bypass(skb, vxlan, vxlan);
2006 return;
2007 }
2008 goto drop;
2009 }
2010
0dfbdf41 2011 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79 2012 vni = rdst->remote_vni;
272d96a5 2013 src = &vxlan->cfg.saddr;
d71785ff 2014 dst_cache = &rdst->dst_cache;
0770b53b 2015 md->gbp = skb->mark;
2016 ttl = vxlan->cfg.ttl;
2017 if (!ttl && vxlan_addr_multicast(dst))
2018 ttl = 1;
2019
2020 tos = vxlan->cfg.tos;
2021 if (tos == 1)
2022 tos = ip_tunnel_get_dsfield(old_iph, skb);
2023
2024 if (dst->sa.sa_family == AF_INET)
2025 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2026 else
2027 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2028 label = vxlan->cfg.label;
ee122c79
TG		 2029 } else {
2030 if (!info) {
2031 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2032 dev->name);
2033 goto drop;
2034 }
b1be00a6 2035 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2036 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2037 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2038 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2039 } else {
a725e514 2040 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2041 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2042 }
ee122c79 2043 dst = &remote_ip;
0770b53b 2044 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2045 vni = tunnel_id_to_key32(info->key.tun_id);
272d96a5 2046 src = &local_ip;
d71785ff 2047 dst_cache = &info->dst_cache;
0770b53b 2048 if (info->options_len)
2049 md = ip_tunnel_info_opts(info);
a725e514
JB		 2050 ttl = info->key.ttl;
2051 tos = info->key.tos;
e7f70af1 2052 label = info->key.label;
b4ed5cad 2053 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2054 }
0770b53b 2055 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2056 vxlan->cfg.port_max, true);
a725e514 2057
e4c7ed41 2058 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2059 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2060 struct rtable *rt;
0770b53b 2061 __be16 df = 0;
c6fcc4fc 2062
655c3de1 2063 rt = vxlan_get_route(vxlan, dev, sock4, skb,
1a8496ba 2064 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2065 dst->sin.sin_addr.s_addr,
2066 &src->sin.sin_addr.s_addr,
d71785ff 2067 dst_cache, info);
8ebd115b
DM		 2068 if (IS_ERR(rt)) {
2069 err = PTR_ERR(rt);
c46b7897 2070 goto tx_error;
8ebd115b 2071 }
e4c7ed41
CW		 2072
2073 /* Bypass encapsulation if the destination is local */
fee1fad7 2074 if (!info) {
2075 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2076 dst_port, vni, &rt->dst,
2077 rt->rt_flags);
2078 if (err)
2079 return;
fee1fad7 2080 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2081 df = htons(IP_DF);
fee1fad7 2082 }
6ceb31ca 2083
c46b7897 2084 ndst = &rt->dst;
e4c7ed41
CW		 2085 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2086 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2087 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2088 vni, md, flags, udp_sum);
f491e56d 2089 if (err < 0)
c46b7897 2090 goto tx_error;
f491e56d 2091
0770b53b 2092 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, src->sin.sin_addr.s_addr,
f491e56d
JB		 2093 dst->sin.sin_addr.s_addr, tos, ttl, df,
2094 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2095#if IS_ENABLED(CONFIG_IPV6)
2096 } else {
c6fcc4fc 2097 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2098
655c3de1 2099 ndst = vxlan6_get_route(vxlan, dev, sock6, skb,
1400615d 2100 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2101 label, &dst->sin6.sin6_addr,
2102 &src->sin6.sin6_addr,
db3c6139 2103 dst_cache, info);
e5d4b29f 2104 if (IS_ERR(ndst)) {
8ebd115b 2105 err = PTR_ERR(ndst);
c46b7897 2106 ndst = NULL;
9dcc71e1 2107 goto tx_error;
e4c7ed41 2108 }
655c3de1 2109
fee1fad7 2110 if (!info) {
2111 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2112
fee1fad7 2113 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2114 dst_port, vni, ndst,
2115 rt6i_flags);
2116 if (err)
2117 return;
fee1fad7 2118 }
35e2d115 2119
1400615d 2120 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2121 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2122 skb_scrub_packet(skb, xnet);
2123 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2124 vni, md, flags, udp_sum);
c46b7897 2125 if (err < 0)
2126 goto tx_error;
2127
0770b53b 2128 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
272d96a5 2129 &src->sin6.sin6_addr,
2130 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2131 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2132#endif
2133 }
4ad16930 2134 return;
d342894c 2135
2136drop:
2137 dev->stats.tx_dropped++;
c46b7897 2138 dev_kfree_skb(skb);
2139 return;
d342894c 2140
2141tx_error:
655c3de1 2142 if (err == -ELOOP)
2143 dev->stats.collisions++;
2144 else if (err == -ENETUNREACH)
2145 dev->stats.tx_carrier_errors++;
c46b7897 2146 dst_release(ndst);
d342894c 2147 dev->stats.tx_errors++;
c46b7897 2148 kfree_skb(skb);
d342894c 2149}
2150
6681712d
DS		 2151/* Transmit local packets over Vxlan
2152 *
2153 * Outer IP header inherits ECN and DF from inner header.
2154 * Outer UDP destination is the VXLAN assigned port.
2155 * source port is based on hash of flow
2156 */
2157static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2158{
2159 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2160 const struct ip_tunnel_info *info;
6681712d
DS		 2161 struct ethhdr *eth;
2162 bool did_rsc = false;
8f646c92 2163 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2164 struct vxlan_fdb *f;
6681712d 2165
61adedf3 2166 info = skb_tunnel_info(skb);
3093fbe7 2167
6681712d 2168 skb_reset_mac_header(skb);
6681712d 2169
47e5d1b0
JB		 2170 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2171 if (info && info->mode & IP_TUNNEL_INFO_TX)
2172 vxlan_xmit_one(skb, dev, NULL, false);
2173 else
2174 kfree_skb(skb);
2175 return NETDEV_TX_OK;
2176 }
2177
2178 if (vxlan->flags & VXLAN_F_PROXY) {
2179 eth = eth_hdr(skb);
f564f45c
CW		 2180 if (ntohs(eth->h_proto) == ETH_P_ARP)
2181 return arp_reduce(dev, skb);
2182#if IS_ENABLED(CONFIG_IPV6)
2183 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2184 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2185 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2186 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2187 struct nd_msg *msg;
2188
2189 msg = (struct nd_msg *)skb_transport_header(skb);
2190 if (msg->icmph.icmp6_code == 0 &&
2191 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2192 return neigh_reduce(dev, skb);
2193 }
2194#endif
2195 }
6681712d 2196
47e5d1b0 2197 eth = eth_hdr(skb);
6681712d 2198 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2199 did_rsc = false;
2200
2201 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2202 (ntohs(eth->h_proto) == ETH_P_IP ||
2203 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2204 did_rsc = route_shortcircuit(dev, skb);
2205 if (did_rsc)
2206 f = vxlan_find_mac(vxlan, eth->h_dest);
2207 }
2208
6681712d 2209 if (f == NULL) {
afbd8bae
MR		 2210 f = vxlan_find_mac(vxlan, all_zeros_mac);
2211 if (f == NULL) {
2212 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2213 !is_multicast_ether_addr(eth->h_dest))
2214 vxlan_fdb_miss(vxlan, eth->h_dest);
2215
2216 dev->stats.tx_dropped++;
8f646c92 2217 kfree_skb(skb);
afbd8bae
MR		 2218 return NETDEV_TX_OK;
2219 }
2220 }
6681712d 2221
afbd8bae
MR		 2222 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2223 struct sk_buff *skb1;
6681712d 2224
8f646c92
ED		 2225 if (!fdst) {
2226 fdst = rdst;
2227 continue;
2228 }
afbd8bae
MR		 2229 skb1 = skb_clone(skb, GFP_ATOMIC);
2230 if (skb1)
2231 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2232 }
2233
8f646c92
ED		 2234 if (fdst)
2235 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2236 else
2237 kfree_skb(skb);
4ad16930 2238 return NETDEV_TX_OK;
6681712d
DS		 2239}
2240
d342894c 2241/* Walk the forwarding table and purge stale entries */
2242static void vxlan_cleanup(unsigned long arg)
2243{
2244 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2245 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2246 unsigned int h;
2247
2248 if (!netif_running(vxlan->dev))
2249 return;
2250
d342894c 2251 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2252 struct hlist_node *p, *n;
14e1d0fa
SD		 2253
2254 spin_lock_bh(&vxlan->hash_lock);
d342894c 2255 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2256 struct vxlan_fdb *f
2257 = container_of(p, struct vxlan_fdb, hlist);
2258 unsigned long timeout;
2259
3c172868 2260 if (f->state & NUD_PERMANENT)
d342894c 2261 continue;
2262
0dfbdf41 2263 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2264 if (time_before_eq(timeout, jiffies)) {
2265 netdev_dbg(vxlan->dev,
2266 "garbage collect %pM\n",
2267 f->eth_addr);
2268 f->state = NUD_STALE;
2269 vxlan_fdb_destroy(vxlan, f);
2270 } else if (time_before(timeout, next_timer))
2271 next_timer = timeout;
2272 }
14e1d0fa 2273 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2274 }
d342894c 2275
2276 mod_timer(&vxlan->age_timer, next_timer);
2277}
2278
9c2e24e1
PS		 2279static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2280{
56ef9c90 2281 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2282 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2283
56ef9c90 2284 spin_lock(&vn->sock_lock);
9c2e24e1 2285 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2286 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2287}
2288
d342894c 2289/* Setup stats when device is created */
2290static int vxlan_init(struct net_device *dev)
2291{
1c213bd2 2292 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2293 if (!dev->tstats)
d342894c 2294 return -ENOMEM;
2295
2296 return 0;
2297}
2298
ba609e9b 2299static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2300{
2301 struct vxlan_fdb *f;
2302
2303 spin_lock_bh(&vxlan->hash_lock);
2304 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2305 if (f)
2306 vxlan_fdb_destroy(vxlan, f);
2307 spin_unlock_bh(&vxlan->hash_lock);
2308}
2309
ebf4063e
SH		 2310static void vxlan_uninit(struct net_device *dev)
2311{
2312 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2313
ba609e9b 2314 vxlan_fdb_delete_default(vxlan);
afbd8bae 2315
ebf4063e
SH		 2316 free_percpu(dev->tstats);
2317}
2318
d342894c 2319/* Start ageing timer and join group when device is brought up */
2320static int vxlan_open(struct net_device *dev)
2321{
2322 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2323 int ret;
56ef9c90 2324
205f356d
JB		 2325 ret = vxlan_sock_add(vxlan);
2326 if (ret < 0)
2327 return ret;
d342894c 2328
79d4a94f 2329 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2330 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2331 if (ret == -EADDRINUSE)
2332 ret = 0;
56ef9c90 2333 if (ret) {
205f356d 2334 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2335 return ret;
2336 }
d342894c 2337 }
2338
0dfbdf41 2339 if (vxlan->cfg.age_interval)
d342894c 2340 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2341
56ef9c90 2342 return ret;
d342894c 2343}
2344
2345/* Purge the forwarding table */
2346static void vxlan_flush(struct vxlan_dev *vxlan)
2347{
31fec5aa 2348 unsigned int h;
d342894c 2349
2350 spin_lock_bh(&vxlan->hash_lock);
2351 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2352 struct hlist_node *p, *n;
2353 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2354 struct vxlan_fdb *f
2355 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2356 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2357 if (!is_zero_ether_addr(f->eth_addr))
2358 vxlan_fdb_destroy(vxlan, f);
d342894c 2359 }
2360 }
2361 spin_unlock_bh(&vxlan->hash_lock);
2362}
2363
2364/* Cleanup timer and forwarding table on shutdown */
2365static int vxlan_stop(struct net_device *dev)
2366{
2367 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2368 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2369 int ret = 0;
d342894c 2370
24c0e683 2371 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2372 !vxlan_group_used(vn, vxlan))
56ef9c90 2373 ret = vxlan_igmp_leave(vxlan);
d342894c 2374
2375 del_timer_sync(&vxlan->age_timer);
2376
2377 vxlan_flush(vxlan);
205f356d 2378 vxlan_sock_release(vxlan);
d342894c 2379
56ef9c90 2380 return ret;
d342894c 2381}
2382
d342894c 2383/* Stub, nothing needs to be done. */
2384static void vxlan_set_multicast_list(struct net_device *dev)
2385{
2386}
2387
91572088 2388static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2389{
91572088
JW		 2390 struct vxlan_dev *vxlan = netdev_priv(dev);
2391 struct vxlan_rdst *dst = &vxlan->default_dst;
2392 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2393 dst->remote_ifindex);
2394 bool use_ipv6 = false;
345010b5
DB		 2395
2396 if (dst->remote_ip.sa.sa_family == AF_INET6)
91572088 2397 use_ipv6 = true;
345010b5 2398
91572088
JW		 2399 /* This check is different than dev->max_mtu, because it looks at
2400 * the lowerdev->mtu, rather than the static dev->max_mtu
2401 */
2402 if (lowerdev) {
2403 int max_mtu = lowerdev->mtu -
2404 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2405 if (new_mtu > max_mtu)
72564b59 2406 return -EINVAL;
72564b59
DW		 2407 }
2408
345010b5
DB		 2409 dev->mtu = new_mtu;
2410 return 0;
2411}
2412
fc4099f1
PS		 2413static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2414{
2415 struct vxlan_dev *vxlan = netdev_priv(dev);
2416 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2417 __be16 sport, dport;
2418
2419 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2420 vxlan->cfg.port_max, true);
2421 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2422
239e944f 2423 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2424 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2425 struct rtable *rt;
2426
655c3de1 2427 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2428 info->key.u.ipv4.dst,
0c1d70af 2429 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2430 if (IS_ERR(rt))
2431 return PTR_ERR(rt);
2432 ip_rt_put(rt);
239e944f
JB		 2433 } else {
2434#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2435 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2436 struct dst_entry *ndst;
2437
655c3de1 2438 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2439 info->key.label, &info->key.u.ipv6.dst,
db3c6139 2440 &info->key.u.ipv6.src, NULL, info);
239e944f
JB		 2441 if (IS_ERR(ndst))
2442 return PTR_ERR(ndst);
2443 dst_release(ndst);
239e944f
JB		 2444#else /* !CONFIG_IPV6 */
2445 return -EPFNOSUPPORT;
2446#endif
2447 }
1a8496ba
JB		 2448 info->key.tp_src = sport;
2449 info->key.tp_dst = dport;
239e944f 2450 return 0;
fc4099f1
PS		 2451}
2452
0c867c9b 2453static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2454 .ndo_init = vxlan_init,
ebf4063e 2455 .ndo_uninit = vxlan_uninit,
d342894c 2456 .ndo_open = vxlan_open,
2457 .ndo_stop = vxlan_stop,
2458 .ndo_start_xmit = vxlan_xmit,
e8171045 2459 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2460 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2461 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2462 .ndo_validate_addr = eth_validate_addr,
2463 .ndo_set_mac_address = eth_mac_addr,
2464 .ndo_fdb_add = vxlan_fdb_add,
2465 .ndo_fdb_del = vxlan_fdb_delete,
2466 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2467 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2468};
2469
e1e5314d
JB		 2470static const struct net_device_ops vxlan_netdev_raw_ops = {
2471 .ndo_init = vxlan_init,
2472 .ndo_uninit = vxlan_uninit,
2473 .ndo_open = vxlan_open,
2474 .ndo_stop = vxlan_stop,
2475 .ndo_start_xmit = vxlan_xmit,
2476 .ndo_get_stats64 = ip_tunnel_get_stats64,
2477 .ndo_change_mtu = vxlan_change_mtu,
2478 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2479};
2480
d342894c 2481/* Info for udev, that this is a virtual tunnel endpoint */
2482static struct device_type vxlan_type = {
2483 .name = "vxlan",
2484};
2485
e5de25dc 2486/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2487 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2488 * to implement the ndo_udp_tunnel_add.
53cf5275 2489 */
b7aade15 2490static void vxlan_push_rx_ports(struct net_device *dev)
53cf5275
JG		 2491{
2492 struct vxlan_sock *vs;
2493 struct net *net = dev_net(dev);
2494 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2495 unsigned int i;
53cf5275
JG		 2496
2497 spin_lock(&vn->sock_lock);
2498 for (i = 0; i < PORT_HASH_SIZE; ++i) {
e7b3db5e
AD		 2499 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2500 udp_tunnel_push_rx_port(dev, vs->sock,
b9adcd69
AD		 2501 (vs->flags & VXLAN_F_GPE) ?
2502 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2503 UDP_TUNNEL_TYPE_VXLAN);
53cf5275
JG		 2504 }
2505 spin_unlock(&vn->sock_lock);
2506}
53cf5275 2507
d342894c 2508/* Initialize the device structure. */
2509static void vxlan_setup(struct net_device *dev)
2510{
2511 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2512 unsigned int h;
d342894c 2513
65226ef8
JB		 2514 eth_hw_addr_random(dev);
2515 ether_setup(dev);
2516
ebf4063e 2517 dev->destructor = free_netdev;
d342894c 2518 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2519
d342894c 2520 dev->features |= NETIF_F_LLTX;
d6727fe3 2521 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2522 dev->features |= NETIF_F_RXCSUM;
05c0db08 2523 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2524
1eaa8178 2525 dev->vlan_features = dev->features;
0afb1666 2526 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2527 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2528 netif_keep_dst(dev);
0c867c9b 2529 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2530
553675fb 2531 INIT_LIST_HEAD(&vxlan->next);
d342894c 2532 spin_lock_init(&vxlan->hash_lock);
2533
2534 init_timer_deferrable(&vxlan->age_timer);
2535 vxlan->age_timer.function = vxlan_cleanup;
2536 vxlan->age_timer.data = (unsigned long) vxlan;
2537
0dfbdf41 2538 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2539
d342894c 2540 vxlan->dev = dev;
2541
58ce31cc
TH		 2542 gro_cells_init(&vxlan->gro_cells, dev);
2543
d342894c 2544 for (h = 0; h < FDB_HASH_SIZE; ++h)
2545 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2546}
2547
0c867c9b
JB		 2548static void vxlan_ether_setup(struct net_device *dev)
2549{
0c867c9b
JB		 2550 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2551 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2552 dev->netdev_ops = &vxlan_netdev_ether_ops;
2553}
2554
e1e5314d
JB		 2555static void vxlan_raw_setup(struct net_device *dev)
2556{
65226ef8 2557 dev->header_ops = NULL;
e1e5314d
JB		 2558 dev->type = ARPHRD_NONE;
2559 dev->hard_header_len = 0;
2560 dev->addr_len = 0;
e1e5314d
JB		 2561 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2562 dev->netdev_ops = &vxlan_netdev_raw_ops;
2563}
2564
d342894c 2565static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2566 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2567 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2568 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2569 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2570 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2571 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2572 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2573 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2574 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2575 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2576 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2577 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2578 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2579 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2580 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2581 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2582 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2583 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2584 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2585 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2586 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2587 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2588 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2589 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2590 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2591 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2592 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2593};
2594
2595static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2596{
2597 if (tb[IFLA_ADDRESS]) {
2598 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2599 pr_debug("invalid link address (not ethernet)\n");
2600 return -EINVAL;
2601 }
2602
2603 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2604 pr_debug("invalid all zero ethernet address\n");
2605 return -EADDRNOTAVAIL;
2606 }
2607 }
2608
2609 if (!data)
2610 return -EINVAL;
2611
2612 if (data[IFLA_VXLAN_ID]) {
2613 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2614 if (id >= VXLAN_VID_MASK)
2615 return -ERANGE;
2616 }
2617
05f47d69 2618 if (data[IFLA_VXLAN_PORT_RANGE]) {
2619 const struct ifla_vxlan_port_range *p
2620 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2621
2622 if (ntohs(p->high) < ntohs(p->low)) {
2623 pr_debug("port range %u .. %u not valid\n",
2624 ntohs(p->low), ntohs(p->high));
2625 return -EINVAL;
2626 }
2627 }
2628
d342894c 2629 return 0;
2630}
2631
1b13c97f
YB		 2632static void vxlan_get_drvinfo(struct net_device *netdev,
2633 struct ethtool_drvinfo *drvinfo)
2634{
2635 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2636 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2637}
2638
2639static const struct ethtool_ops vxlan_ethtool_ops = {
2640 .get_drvinfo = vxlan_get_drvinfo,
2641 .get_link = ethtool_op_get_link,
2642};
2643
3ee64f39
TH		 2644static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2645 __be16 port, u32 flags)
553675fb 2646{
e4c7ed41 2647 struct socket *sock;
3ee64f39
TH		 2648 struct udp_port_cfg udp_conf;
2649 int err;
e4c7ed41 2650
3ee64f39 2651 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2652
3ee64f39
TH		 2653 if (ipv6) {
2654 udp_conf.family = AF_INET6;
3ee64f39 2655 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2656 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2657 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2658 } else {
2659 udp_conf.family = AF_INET;
e4c7ed41 2660 }
e4c7ed41 2661
3ee64f39 2662 udp_conf.local_udp_port = port;
553675fb 2663
3ee64f39
TH		 2664 /* Open UDP socket */
2665 err = udp_sock_create(net, &udp_conf, &sock);
2666 if (err < 0)
2667 return ERR_PTR(err);
e4c7ed41 2668
39deb2c7 2669 return sock;
e4c7ed41
CW		 2670}
2671
2672/* Create new listen socket if needed */
b1be00a6
JB		 2673static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2674 __be16 port, u32 flags)
e4c7ed41
CW		 2675{
2676 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2677 struct vxlan_sock *vs;
2678 struct socket *sock;
e4c7ed41 2679 unsigned int h;
acbf74a7 2680 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2681
dc01e7d3 2682 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2683 if (!vs)
2684 return ERR_PTR(-ENOMEM);
2685
2686 for (h = 0; h < VNI_HASH_SIZE; ++h)
2687 INIT_HLIST_HEAD(&vs->vni_list[h]);
2688
3ee64f39 2689 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2690 if (IS_ERR(sock)) {
56ef9c90
MRL		 2691 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2692 PTR_ERR(sock));
553675fb 2693 kfree(vs);
e50fddc8 2694 return ERR_CAST(sock);
553675fb 2695 }
e4c7ed41
CW		 2696
2697 vs->sock = sock;
9c2e24e1 2698 atomic_set(&vs->refcnt, 1);
af33c1ad 2699 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2700
9c2e24e1
PS		 2701 spin_lock(&vn->sock_lock);
2702 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2703 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2704 (vs->flags & VXLAN_F_GPE) ?
2705 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2706 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2707 spin_unlock(&vn->sock_lock);
553675fb 2708
2709 /* Mark socket as an encapsulation socket. */
5602c48c 2710 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2711 tunnel_cfg.sk_user_data = vs;
2712 tunnel_cfg.encap_type = 1;
f2d1968e 2713 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2714 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2715 tunnel_cfg.gro_receive = vxlan_gro_receive;
2716 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2717
2718 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2719
9c2e24e1
PS		 2720 return vs;
2721}
2722
b1be00a6 2723static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2724{
205f356d
JB		 2725 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2726 struct vxlan_sock *vs = NULL;
9c2e24e1 2727
205f356d 2728 if (!vxlan->cfg.no_share) {
56ef9c90 2729 spin_lock(&vn->sock_lock);
205f356d
JB		 2730 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2731 vxlan->cfg.dst_port, vxlan->flags);
2732 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2733 spin_unlock(&vn->sock_lock);
205f356d 2734 return -EBUSY;
56ef9c90
MRL		 2735 }
2736 spin_unlock(&vn->sock_lock);
2737 }
205f356d 2738 if (!vs)
b1be00a6
JB		 2739 vs = vxlan_socket_create(vxlan->net, ipv6,
2740 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2741 if (IS_ERR(vs))
2742 return PTR_ERR(vs);
b1be00a6
JB		 2743#if IS_ENABLED(CONFIG_IPV6)
2744 if (ipv6)
c6fcc4fc 2745 rcu_assign_pointer(vxlan->vn6_sock, vs);
b1be00a6
JB		 2746 else
2747#endif
c6fcc4fc 2748 rcu_assign_pointer(vxlan->vn4_sock, vs);
205f356d
JB		 2749 vxlan_vs_add_dev(vs, vxlan);
2750 return 0;
553675fb 2751}
2752
b1be00a6
JB		 2753static int vxlan_sock_add(struct vxlan_dev *vxlan)
2754{
2755 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2756 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2757 int ret = 0;
2758
c6fcc4fc 2759 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2760#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2761 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
b1be00a6
JB		 2762 if (ipv6 || metadata)
2763 ret = __vxlan_sock_add(vxlan, true);
2764#endif
2765 if (!ret && (!ipv6 || metadata))
2766 ret = __vxlan_sock_add(vxlan, false);
2767 if (ret < 0)
2768 vxlan_sock_release(vxlan);
2769 return ret;
2770}
2771
0dfbdf41
TG		 2772static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2773 struct vxlan_config *conf)
d342894c 2774{
33564bbb 2775 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2776 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2777 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2778 unsigned short needed_headroom = ETH_HLEN;
d342894c 2779 int err;
e4c7ed41 2780 bool use_ipv6 = false;
0dfbdf41 2781 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2782 struct net_device *lowerdev = NULL;
d342894c 2783
e1e5314d 2784 if (conf->flags & VXLAN_F_GPE) {
e1e5314d
JB		 2785 /* For now, allow GPE only together with COLLECT_METADATA.
2786 * This can be relaxed later; in such case, the other side
2787 * of the PtP link will have to be provided.
2788 */
3555621d
JB		 2789 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2790 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2791 pr_info("unsupported combination of extensions\n");
e1e5314d 2792 return -EINVAL;
3555621d 2793 }
e1e5314d
JB		 2794
2795 vxlan_raw_setup(dev);
2796 } else {
2797 vxlan_ether_setup(dev);
2798 }
0c867c9b 2799
91572088
JW		 2800 /* MTU range: 68 - 65535 */
2801 dev->min_mtu = ETH_MIN_MTU;
2802 dev->max_mtu = ETH_MAX_MTU;
2803
33564bbb 2804 vxlan->net = src_net;
f01ec1c0 2805
0dfbdf41 2806 dst->remote_vni = conf->vni;
d342894c 2807
0dfbdf41 2808 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2809
0dfbdf41
TG		 2810 /* Unless IPv6 is explicitly requested, assume IPv4 */
2811 if (!dst->remote_ip.sa.sa_family)
2812 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2813
0dfbdf41 2814 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2815 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2816 if (!IS_ENABLED(CONFIG_IPV6))
2817 return -EPFNOSUPPORT;
e4c7ed41 2818 use_ipv6 = true;
b1be00a6 2819 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2820 }
d342894c 2821
e7f70af1
DB		 2822 if (conf->label && !use_ipv6) {
2823 pr_info("label only supported in use with IPv6\n");
2824 return -EINVAL;
2825 }
2826
0dfbdf41 2827 if (conf->remote_ifindex) {
7e059158 2828 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2829 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2830
2831 if (!lowerdev) {
c7995c43 2832 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2833 return -ENODEV;
2834 }
d342894c 2835
e4c7ed41
CW		 2836#if IS_ENABLED(CONFIG_IPV6)
2837 if (use_ipv6) {
2838 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2839 if (idev && idev->cnf.disable_ipv6) {
2840 pr_info("IPv6 is disabled via sysctl\n");
2841 return -EPERM;
2842 }
e4c7ed41
CW		 2843 }
2844#endif
2845
0dfbdf41 2846 if (!conf->mtu)
91572088
JW		 2847 dev->mtu = lowerdev->mtu -
2848 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2849
b1be00a6 2850 needed_headroom = lowerdev->hard_header_len;
9b4cdd51
JB		 2851 } else if (vxlan_addr_multicast(&dst->remote_ip)) {
2852 pr_info("multicast destination requires interface to be specified\n");
2853 return -EINVAL;
9dc2ad10 2854 }
d342894c 2855
7e059158 2856 if (conf->mtu) {
91572088
JW		 2857 int max_mtu = ETH_MAX_MTU;
2858
2859 if (lowerdev)
2860 max_mtu = lowerdev->mtu;
2861
2862 max_mtu -= (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2863
2864 if (conf->mtu < dev->min_mtu || conf->mtu > dev->max_mtu)
2865 return -EINVAL;
2866
2867 dev->mtu = conf->mtu;
2868
2869 if (conf->mtu > max_mtu)
2870 dev->mtu = max_mtu;
7e059158
DW		 2871 }
2872
b1be00a6
JB		 2873 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2874 needed_headroom += VXLAN6_HEADROOM;
2875 else
2876 needed_headroom += VXLAN_HEADROOM;
2877 dev->needed_headroom = needed_headroom;
2878
0dfbdf41 2879 memcpy(&vxlan->cfg, conf, sizeof(*conf));
e1e5314d
JB		 2880 if (!vxlan->cfg.dst_port) {
2881 if (conf->flags & VXLAN_F_GPE)
2882 vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */
2883 else
2884 vxlan->cfg.dst_port = default_port;
2885 }
0dfbdf41
TG		 2886 vxlan->flags |= conf->flags;
2887
2888 if (!vxlan->cfg.age_interval)
2889 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2890
07b9b37c
ND		 2891 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2892 if (tmp->cfg.vni == conf->vni &&
2893 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2894 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2895 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2896 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
3555621d
JB		 2897 (vxlan->flags & VXLAN_F_RCV_FLAGS)) {
2898 pr_info("duplicate VNI %u\n", be32_to_cpu(conf->vni));
2899 return -EEXIST;
2900 }
07b9b37c 2901 }
0dfbdf41
TG		 2902
2903 dev->ethtool_ops = &vxlan_ethtool_ops;
2904
2905 /* create an fdb entry for a valid default destination */
2906 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2907 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2908 &vxlan->default_dst.remote_ip,
2909 NUD_REACHABLE|NUD_PERMANENT,
2910 NLM_F_EXCL|NLM_F_CREATE,
2911 vxlan->cfg.dst_port,
2912 vxlan->default_dst.remote_vni,
2913 vxlan->default_dst.remote_ifindex,
2914 NTF_SELF);
2915 if (err)
2916 return err;
2917 }
2918
2919 err = register_netdevice(dev);
2920 if (err) {
2921 vxlan_fdb_delete_default(vxlan);
2922 return err;
2923 }
2924
2925 list_add(&vxlan->next, &vn->vxlan_list);
2926
2927 return 0;
2928}
2929
0dfbdf41
TG		 2930static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2931 struct nlattr *tb[], struct nlattr *data[])
2932{
2933 struct vxlan_config conf;
0dfbdf41 2934
0dfbdf41 2935 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2936
2937 if (data[IFLA_VXLAN_ID])
54bfd872 2938 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
0dfbdf41
TG		 2939
2940 if (data[IFLA_VXLAN_GROUP]) {
2941 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2942 } else if (data[IFLA_VXLAN_GROUP6]) {
2943 if (!IS_ENABLED(CONFIG_IPV6))
2944 return -EPFNOSUPPORT;
2945
2946 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2947 conf.remote_ip.sa.sa_family = AF_INET6;
2948 }
2949
2950 if (data[IFLA_VXLAN_LOCAL]) {
2951 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2952 conf.saddr.sa.sa_family = AF_INET;
2953 } else if (data[IFLA_VXLAN_LOCAL6]) {
2954 if (!IS_ENABLED(CONFIG_IPV6))
2955 return -EPFNOSUPPORT;
2956
2957 /* TODO: respect scope id */
2958 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2959 conf.saddr.sa.sa_family = AF_INET6;
2960 }
2961
2962 if (data[IFLA_VXLAN_LINK])
2963 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2964
d342894c 2965 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2966 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2967
afb97186 2968 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2969 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2970
e7f70af1
DB		 2971 if (data[IFLA_VXLAN_LABEL])
2972 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
2973 IPV6_FLOWLABEL_MASK;
2974
d342894c 2975 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2976 conf.flags |= VXLAN_F_LEARN;
d342894c 2977
2978 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2979 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2980
e4f67add 2981 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2982 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2983
2984 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2985 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2986
2987 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2988 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2989
2990 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2991 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2992
d342894c 2993 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2994 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2995
f8a9b1bc
AS		 2996 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2997 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2998 conf.flags |= VXLAN_F_COLLECT_METADATA;
2999
05f47d69 3000 if (data[IFLA_VXLAN_PORT_RANGE]) {
3001 const struct ifla_vxlan_port_range *p
3002 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 3003 conf.port_min = ntohs(p->low);
3004 conf.port_max = ntohs(p->high);
05f47d69 3005 }
3006
823aa873 3007 if (data[IFLA_VXLAN_PORT])
0dfbdf41 3008 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 3009
6ceb31ca
AD		 3010 if (data[IFLA_VXLAN_UDP_CSUM] &&
3011 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3012 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
359a0ea9
TH		 3013
3014 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
3015 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 3016 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 3017
3018 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
3019 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 3020 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 3021
dfd8645e
TH		 3022 if (data[IFLA_VXLAN_REMCSUM_TX] &&
3023 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 3024 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 3025
3026 if (data[IFLA_VXLAN_REMCSUM_RX] &&
3027 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 3028 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 3029
3511494c 3030 if (data[IFLA_VXLAN_GBP])
0dfbdf41 3031 conf.flags |= VXLAN_F_GBP;
3511494c 3032
e1e5314d
JB		 3033 if (data[IFLA_VXLAN_GPE])
3034 conf.flags |= VXLAN_F_GPE;
3035
0ace2ca8 3036 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 3037 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 3038
ce577668
CH		 3039 if (tb[IFLA_MTU])
3040 conf.mtu = nla_get_u32(tb[IFLA_MTU]);
3041
3555621d 3042 return vxlan_dev_configure(src_net, dev, &conf);
d342894c 3043}
3044
3045static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3046{
3047 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3048 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 3049
fe5c3561 3050 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 3051 if (!hlist_unhashed(&vxlan->hlist))
3052 hlist_del_rcu(&vxlan->hlist);
fe5c3561 3053 spin_unlock(&vn->sock_lock);
3054
58ce31cc 3055 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3056 list_del(&vxlan->next);
d342894c 3057 unregister_netdevice_queue(dev, head);
3058}
3059
3060static size_t vxlan_get_size(const struct net_device *dev)
3061{
3062
3063 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3064 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3065 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3066 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3067 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3068 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3069 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3070 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3071 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3072 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3073 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3074 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3075 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3076 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3077 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3078 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3079 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3080 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3081 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3082 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3083 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3084 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3085 0;
3086}
3087
3088static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3089{
3090 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3091 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3092 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3093 .low = htons(vxlan->cfg.port_min),
3094 .high = htons(vxlan->cfg.port_max),
05f47d69 3095 };
d342894c 3096
54bfd872 3097 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3098 goto nla_put_failure;
3099
e4c7ed41
CW		 3100 if (!vxlan_addr_any(&dst->remote_ip)) {
3101 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3102 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3103 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3104 goto nla_put_failure;
3105#if IS_ENABLED(CONFIG_IPV6)
3106 } else {
930345ea
JB		 3107 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3108 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3109 goto nla_put_failure;
3110#endif
3111 }
3112 }
d342894c 3113
c7995c43 3114 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3115 goto nla_put_failure;
3116
0dfbdf41
TG		 3117 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3118 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3119 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3120 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3121 goto nla_put_failure;
3122#if IS_ENABLED(CONFIG_IPV6)
3123 } else {
930345ea 3124 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3125 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3126 goto nla_put_failure;
3127#endif
3128 }
3129 }
d342894c 3130
0dfbdf41
TG		 3131 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3132 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3133 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3134 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3135 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3136 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3137 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3138 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3139 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3140 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3141 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3142 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3143 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3144 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3145 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3146 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3147 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3148 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3149 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3150 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3151 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3152 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3153 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3154 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3155 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3156 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3157 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3158 goto nla_put_failure;
3159
05f47d69 3160 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3161 goto nla_put_failure;
3162
3511494c
TG		 3163 if (vxlan->flags & VXLAN_F_GBP &&
3164 nla_put_flag(skb, IFLA_VXLAN_GBP))
3165 goto nla_put_failure;
3166
e1e5314d
JB		 3167 if (vxlan->flags & VXLAN_F_GPE &&
3168 nla_put_flag(skb, IFLA_VXLAN_GPE))
3169 goto nla_put_failure;
3170
0ace2ca8
TH		 3171 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3172 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3173 goto nla_put_failure;
3174
d342894c 3175 return 0;
3176
3177nla_put_failure:
3178 return -EMSGSIZE;
3179}
3180
1728d4fa
ND		 3181static struct net *vxlan_get_link_net(const struct net_device *dev)
3182{
3183 struct vxlan_dev *vxlan = netdev_priv(dev);
3184
3185 return vxlan->net;
3186}
3187
d342894c 3188static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3189 .kind = "vxlan",
3190 .maxtype = IFLA_VXLAN_MAX,
3191 .policy = vxlan_policy,
3192 .priv_size = sizeof(struct vxlan_dev),
3193 .setup = vxlan_setup,
3194 .validate = vxlan_validate,
3195 .newlink = vxlan_newlink,
3196 .dellink = vxlan_dellink,
3197 .get_size = vxlan_get_size,
3198 .fill_info = vxlan_fill_info,
1728d4fa 3199 .get_link_net = vxlan_get_link_net,
d342894c 3200};
3201
cf5da330
ND		 3202struct net_device *vxlan_dev_create(struct net *net, const char *name,
3203 u8 name_assign_type,
3204 struct vxlan_config *conf)
3205{
3206 struct nlattr *tb[IFLA_MAX + 1];
3207 struct net_device *dev;
3208 int err;
3209
3210 memset(&tb, 0, sizeof(tb));
3211
3212 dev = rtnl_create_link(net, name, name_assign_type,
3213 &vxlan_link_ops, tb);
3214 if (IS_ERR(dev))
3215 return dev;
3216
3217 err = vxlan_dev_configure(net, dev, conf);
3218 if (err < 0) {
3219 free_netdev(dev);
3220 return ERR_PTR(err);
3221 }
3222
3223 err = rtnl_configure_link(dev, NULL);
3224 if (err < 0) {
3225 LIST_HEAD(list_kill);
3226
3227 vxlan_dellink(dev, &list_kill);
3228 unregister_netdevice_many(&list_kill);
3229 return ERR_PTR(err);
3230 }
3231
3232 return dev;
3233}
3234EXPORT_SYMBOL_GPL(vxlan_dev_create);
3235
acaf4e70
DB		 3236static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3237 struct net_device *dev)
3238{
3239 struct vxlan_dev *vxlan, *next;
3240 LIST_HEAD(list_kill);
3241
3242 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3243 struct vxlan_rdst *dst = &vxlan->default_dst;
3244
3245 /* In case we created vxlan device with carrier
3246 * and we loose the carrier due to module unload
3247 * we also need to remove vxlan device. In other
3248 * cases, it's not necessary and remote_ifindex
3249 * is 0 here, so no matches.
3250 */
3251 if (dst->remote_ifindex == dev->ifindex)
3252 vxlan_dellink(vxlan->dev, &list_kill);
3253 }
3254
3255 unregister_netdevice_many(&list_kill);
3256}
3257
b7aade15
HFS		 3258static int vxlan_netdevice_event(struct notifier_block *unused,
3259 unsigned long event, void *ptr)
acaf4e70
DB		 3260{
3261 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3262 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3263
783c1463 3264 if (event == NETDEV_UNREGISTER)
acaf4e70 3265 vxlan_handle_lowerdev_unregister(vn, dev);
7c46a640 3266 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
b7aade15 3267 vxlan_push_rx_ports(dev);
acaf4e70
DB		 3268
3269 return NOTIFY_DONE;
3270}
3271
3272static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3273 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3274};
3275
d342894c 3276static __net_init int vxlan_init_net(struct net *net)
3277{
3278 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3279 unsigned int h;
d342894c 3280
553675fb 3281 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3282 spin_lock_init(&vn->sock_lock);
d342894c 3283
553675fb 3284 for (h = 0; h < PORT_HASH_SIZE; ++h)
3285 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3286
3287 return 0;
3288}
3289
f01ec1c0
ND		 3290static void __net_exit vxlan_exit_net(struct net *net)
3291{
3292 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3293 struct vxlan_dev *vxlan, *next;
3294 struct net_device *dev, *aux;
3295 LIST_HEAD(list);
3296
3297 rtnl_lock();
3298 for_each_netdev_safe(net, dev, aux)
3299 if (dev->rtnl_link_ops == &vxlan_link_ops)
3300 unregister_netdevice_queue(dev, &list);
3301
3302 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3303 /* If vxlan->dev is in the same netns, it has already been added
3304 * to the list by the previous loop.
3305 */
58ce31cc
TH		 3306 if (!net_eq(dev_net(vxlan->dev), net)) {
3307 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3308 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3309 }
f01ec1c0
ND		 3310 }
3311
3312 unregister_netdevice_many(&list);
3313 rtnl_unlock();
3314}
3315
d342894c 3316static struct pernet_operations vxlan_net_ops = {
3317 .init = vxlan_init_net,
f01ec1c0 3318 .exit = vxlan_exit_net,
d342894c 3319 .id = &vxlan_net_id,
3320 .size = sizeof(struct vxlan_net),
3321};
3322
3323static int __init vxlan_init_module(void)
3324{
3325 int rc;
3326
3327 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3328
783c1463 3329 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3330 if (rc)
3331 goto out1;
3332
acaf4e70 3333 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3334 if (rc)
3335 goto out2;
3336
acaf4e70
DB		 3337 rc = rtnl_link_register(&vxlan_link_ops);
3338 if (rc)
3339 goto out3;
d342894c 3340
acaf4e70
DB		 3341 return 0;
3342out3:
3343 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3344out2:
783c1463 3345 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3346out1:
3347 return rc;
3348}
7332a13b 3349late_initcall(vxlan_init_module);
d342894c 3350
3351static void __exit vxlan_cleanup_module(void)
3352{
b7153984 3353 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3354 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3355 unregister_pernet_subsys(&vxlan_net_ops);
3356 /* rcu_barrier() is called by netns */
d342894c 3357}
3358module_exit(vxlan_cleanup_module);
3359
3360MODULE_LICENSE("GPL");
3361MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3362MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3363MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3364MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)