]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ath11k: put hw to DBS using WMI_PDEV_SET_HW_MODE_CMDID
[mirror_ubuntu-hirsute-kernel.git] / drivers / net / vxlan.c
CommitLineData
d2912cb1 1// SPDX-License-Identifier: GPL-2.0-only
d342894c 2/*
eb5ce439 3 * VXLAN: Virtual eXtensible Local Area Network
d342894c 4 *
3b8df3c6 5 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/kernel.h>
d342894c 11#include <linux/module.h>
12#include <linux/errno.h>
13#include <linux/slab.h>
d342894c 14#include <linux/udp.h>
15#include <linux/igmp.h>
d342894c 16#include <linux/if_ether.h>
1b13c97f 17#include <linux/ethtool.h>
e4f67add
DS		 18#include <net/arp.h>
19#include <net/ndisc.h>
3616d08b 20#include <net/ipv6_stubs.h>
d342894c 21#include <net/ip.h>
22#include <net/icmp.h>
d342894c 23#include <net/rtnetlink.h>
d342894c 24#include <net/inet_ecn.h>
25#include <net/net_namespace.h>
26#include <net/netns/generic.h>
fa20e0e3 27#include <net/tun_proto.h>
012a5729 28#include <net/vxlan.h>
1274e1cc 29#include <net/nexthop.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
c7d03a00 55static unsigned 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
a53cb29b
MB		 62static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
63
553675fb 64/* per-network namespace private data for this module */
65struct vxlan_net {
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 68 spinlock_t sock_lock;
626d667b 69 struct notifier_block nexthop_notifier_block;
553675fb 70};
71
d342894c 72/* Forwarding table entry */
73struct vxlan_fdb {
74 struct hlist_node hlist; /* linked list of entries */
75 struct rcu_head rcu;
76 unsigned long updated; /* jiffies */
77 unsigned long used;
3e61aa8f 78 struct list_head remotes;
7177a3b0 79 u8 eth_addr[ETH_ALEN];
d342894c 80 u16 state; /* see ndm_state */
3ad7a4b1 81 __be32 vni;
45598c1c 82 u16 flags; /* see ndm_flags and below */
1274e1cc
RP		 83 struct list_head nh_list;
84 struct nexthop __rcu *nh;
79472fe8 85 struct vxlan_dev __rcu *vdev;
d342894c 86};
87
45598c1c
PM		 88#define NTF_VXLAN_ADDED_BY_USER 0x100
89
d342894c 90/* salt for hash table */
91static u32 vxlan_salt __read_mostly;
92
ee122c79
TG		 93static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
94{
e7030878
TG		 95 return vs->flags & VXLAN_F_COLLECT_METADATA ||
96 ip_tunnel_collect_metadata();
ee122c79
TG		 97}
98
e4c7ed41
CW		 99#if IS_ENABLED(CONFIG_IPV6)
100static inline
101bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
102{
f0ef3126
JB		 103 if (a->sa.sa_family != b->sa.sa_family)
104 return false;
105 if (a->sa.sa_family == AF_INET6)
106 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
107 else
108 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 109}
110
e4c7ed41
CW		 111static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
112{
f0ef3126 113 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 114 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 115 ip->sa.sa_family = AF_INET6;
116 return 0;
117 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 118 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 119 ip->sa.sa_family = AF_INET;
120 return 0;
121 } else {
122 return -EAFNOSUPPORT;
123 }
e4c7ed41
CW		 124}
125
126static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 127 const union vxlan_addr *ip)
e4c7ed41 128{
f0ef3126 129 if (ip->sa.sa_family == AF_INET6)
930345ea 130 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 131 else
930345ea 132 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 133}
134
135#else /* !CONFIG_IPV6 */
136
137static inline
138bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
139{
f0ef3126 140 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 141}
142
e4c7ed41
CW		 143static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
144{
f0ef3126
JB		 145 if (nla_len(nla) >= sizeof(struct in6_addr)) {
146 return -EAFNOSUPPORT;
147 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 148 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 149 ip->sa.sa_family = AF_INET;
150 return 0;
151 } else {
152 return -EAFNOSUPPORT;
153 }
e4c7ed41
CW		 154}
155
156static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 157 const union vxlan_addr *ip)
e4c7ed41 158{
930345ea 159 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 160}
161#endif
162
553675fb 163/* Virtual Network hash table head */
54bfd872 164static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 165{
54bfd872 166 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 167}
168
169/* Socket hash table head */
170static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 171{
172 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173
553675fb 174 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
175}
176
3e61aa8f
SH		 177/* First remote destination for a forwarding entry.
178 * Guaranteed to be non-NULL because remotes are never deleted.
179 */
5ca5461c 180static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 181{
1274e1cc
RP		 182 if (rcu_access_pointer(fdb->nh))
183 return NULL;
5ca5461c 184 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
185}
186
187static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
188{
1274e1cc
RP		 189 if (rcu_access_pointer(fdb->nh))
190 return NULL;
5ca5461c 191 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 192}
193
78ec710e
FF		 194/* Find VXLAN socket based on network namespace, address family, UDP port,
195 * enabled unshareable flags and socket device binding (see l3mdev with
196 * non-default VRF).
ac5132d1
TG		 197 */
198static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
aab8cc36 199 __be16 port, u32 flags, int ifindex)
553675fb 200{
201 struct vxlan_sock *vs;
af33c1ad
TH		 202
203 flags &= VXLAN_F_RCV_FLAGS;
553675fb 204
205 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 206 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 207 vxlan_get_sk_family(vs) == family &&
aab8cc36
AB		 208 vs->flags == flags &&
209 vs->sock->sk->sk_bound_dev_if == ifindex)
553675fb 210 return vs;
211 }
212 return NULL;
d342894c 213}
214
49f810f0
MS		 215static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
216 __be32 vni)
d342894c 217{
69e76661 218 struct vxlan_dev_node *node;
d342894c 219
b9167b2e
JB		 220 /* For flow based devices, map all packets to VNI 0 */
221 if (vs->flags & VXLAN_F_COLLECT_METADATA)
222 vni = 0;
223
69e76661
JB		 224 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
225 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS		 226 continue;
227
228 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 229 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS		 230
231 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
232 cfg->remote_ifindex != ifindex)
233 continue;
234 }
235
69e76661 236 return node->vxlan;
d342894c 237 }
238
239 return NULL;
240}
241
5cfccc5a 242/* Look up VNI in a per net namespace table */
49f810f0
MS		 243static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
244 __be32 vni, sa_family_t family,
245 __be16 port, u32 flags)
5cfccc5a
PS		 246{
247 struct vxlan_sock *vs;
248
aab8cc36 249 vs = vxlan_find_sock(net, family, port, flags, ifindex);
5cfccc5a
PS		 250 if (!vs)
251 return NULL;
252
49f810f0 253 return vxlan_vs_find_vni(vs, ifindex, 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;
1274e1cc 264 bool send_ip, send_eth;
d342894c 265 struct nlmsghdr *nlh;
1274e1cc 266 struct nexthop *nh;
d342894c 267 struct ndmsg *ndm;
06ec313e
IS		 268 int nh_family;
269 u32 nh_id;
d342894c 270
271 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
272 if (nlh == NULL)
273 return -EMSGSIZE;
274
275 ndm = nlmsg_data(nlh);
276 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 277
278 send_eth = send_ip = true;
279
06ec313e
IS		 280 rcu_read_lock();
281 nh = rcu_dereference(fdb->nh);
282 if (nh) {
283 nh_family = nexthop_get_family(nh);
284 nh_id = nh->id;
285 }
286 rcu_read_unlock();
287
e4f67add 288 if (type == RTM_GETNEIGH) {
1274e1cc
RP		 289 if (rdst) {
290 send_ip = !vxlan_addr_any(&rdst->remote_ip);
291 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
292 } else if (nh) {
06ec313e 293 ndm->ndm_family = nh_family;
1274e1cc 294 }
e4f67add
DS		 295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
d342894c 298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 300 ndm->ndm_flags = fdb->flags;
1274e1cc 301 if (rdst && rdst->offloaded)
0efe1173 302 ndm->ndm_flags |= NTF_OFFLOADED;
545469f7 303 ndm->ndm_type = RTN_UNICAST;
d342894c 304
193523bf 305 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 306 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 307 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 308 goto nla_put_failure;
309
e4f67add 310 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 311 goto nla_put_failure;
1274e1cc 312 if (nh) {
06ec313e 313 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
1274e1cc
RP		 314 goto nla_put_failure;
315 } else if (rdst) {
316 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
317 &rdst->remote_ip))
318 goto nla_put_failure;
319
320 if (rdst->remote_port &&
321 rdst->remote_port != vxlan->cfg.dst_port &&
322 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
323 goto nla_put_failure;
324 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
325 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
326 goto nla_put_failure;
327 if (rdst->remote_ifindex &&
328 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
329 goto nla_put_failure;
330 }
d342894c 331
dc5321d7 332 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 333 nla_put_u32(skb, NDA_SRC_VNI,
334 be32_to_cpu(fdb->vni)))
335 goto nla_put_failure;
d342894c 336
337 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
338 ci.ndm_confirmed = 0;
339 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
340 ci.ndm_refcnt = 0;
341
342 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
343 goto nla_put_failure;
344
053c095a
JB		 345 nlmsg_end(skb, nlh);
346 return 0;
d342894c 347
348nla_put_failure:
349 nlmsg_cancel(skb, nlh);
350 return -EMSGSIZE;
351}
352
353static inline size_t vxlan_nlmsg_size(void)
354{
355 return NLMSG_ALIGN(sizeof(struct ndmsg))
356 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 357 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 358 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 359 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
360 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 361 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 362 + nla_total_size(sizeof(struct nda_cacheinfo));
363}
364
9a997353
PM		 365static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
366 struct vxlan_rdst *rd, int type)
d342894c 367{
368 struct net *net = dev_net(vxlan->dev);
369 struct sk_buff *skb;
370 int err = -ENOBUFS;
371
372 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
373 if (skb == NULL)
374 goto errout;
375
9e4b93f9 376 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 377 if (err < 0) {
378 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
379 WARN_ON(err == -EMSGSIZE);
380 kfree_skb(skb);
381 goto errout;
382 }
383
384 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
385 return;
386errout:
387 if (err < 0)
388 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
389}
390
ff23b91c
PM		 391static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
392 const struct vxlan_fdb *fdb,
393 const struct vxlan_rdst *rd,
4c59b7d1 394 struct netlink_ext_ack *extack,
ff23b91c
PM		 395 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
396{
397 fdb_info->info.dev = vxlan->dev;
4c59b7d1 398 fdb_info->info.extack = extack;
ff23b91c
PM		 399 fdb_info->remote_ip = rd->remote_ip;
400 fdb_info->remote_port = rd->remote_port;
401 fdb_info->remote_vni = rd->remote_vni;
402 fdb_info->remote_ifindex = rd->remote_ifindex;
403 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
404 fdb_info->vni = fdb->vni;
405 fdb_info->offloaded = rd->offloaded;
406 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
407}
408
61f46fe8
PM		 409static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
410 struct vxlan_fdb *fdb,
411 struct vxlan_rdst *rd,
4c59b7d1
PM		 412 bool adding,
413 struct netlink_ext_ack *extack)
9a997353
PM		 414{
415 struct switchdev_notifier_vxlan_fdb_info info;
416 enum switchdev_notifier_type notifier_type;
61f46fe8 417 int ret;
9a997353
PM		 418
419 if (WARN_ON(!rd))
61f46fe8 420 return 0;
9a997353
PM		 421
422 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
423 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
4c59b7d1 424 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
61f46fe8 425 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
6685987c 426 &info.info, extack);
61f46fe8 427 return notifier_to_errno(ret);
9a997353
PM		 428}
429
61f46fe8 430static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
4c59b7d1
PM		 431 struct vxlan_rdst *rd, int type, bool swdev_notify,
432 struct netlink_ext_ack *extack)
9a997353 433{
61f46fe8
PM		 434 int err;
435
1274e1cc 436 if (swdev_notify && rd) {
0e6160f3
PM		 437 switch (type) {
438 case RTM_NEWNEIGH:
61f46fe8 439 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 440 true, extack);
61f46fe8
PM		 441 if (err)
442 return err;
0e6160f3
PM		 443 break;
444 case RTM_DELNEIGH:
445 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 446 false, extack);
0e6160f3
PM		 447 break;
448 }
9a997353
PM		 449 }
450
451 __vxlan_fdb_notify(vxlan, fdb, rd, type);
61f46fe8 452 return 0;
9a997353
PM		 453}
454
e4c7ed41 455static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 456{
457 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 458 struct vxlan_fdb f = {
459 .state = NUD_STALE,
460 };
461 struct vxlan_rdst remote = {
e4c7ed41 462 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 463 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 464 };
3e61aa8f 465
4c59b7d1 466 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 467}
468
469static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
470{
bb3fd687
SH		 471 struct vxlan_fdb f = {
472 .state = NUD_STALE,
473 };
9e4b93f9 474 struct vxlan_rdst remote = { };
e4f67add 475
e4f67add
DS		 476 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
477
4c59b7d1 478 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 479}
480
d342894c 481/* Hash Ethernet address */
482static u32 eth_hash(const unsigned char *addr)
483{
484 u64 value = get_unaligned((u64 *)addr);
485
486 /* only want 6 bytes */
487#ifdef __BIG_ENDIAN
d342894c 488 value >>= 16;
321fb991 489#else
490 value <<= 16;
d342894c 491#endif
492 return hash_64(value, FDB_HASH_BITS);
493}
494
3ad7a4b1
RP		 495static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
496{
497 /* use 1 byte of OUI and 3 bytes of NIC */
498 u32 key = get_unaligned((u32 *)(addr + 2));
499
500 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
501}
502
fe1e0713
L		 503static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
504{
505 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
506 return eth_vni_hash(mac, vni);
507 else
508 return eth_hash(mac);
509}
510
d342894c 511/* Hash chain to use given mac address */
512static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 513 const u8 *mac, __be32 vni)
d342894c 514{
fe1e0713 515 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
d342894c 516}
517
518/* Look up Ethernet address in forwarding table */
014be2c8 519static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 520 const u8 *mac, __be32 vni)
d342894c 521{
3ad7a4b1 522 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 523 struct vxlan_fdb *f;
d342894c 524
b67bfe0d 525 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 526 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 527 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 528 if (vni == f->vni)
529 return f;
530 } else {
531 return f;
532 }
533 }
d342894c 534 }
535
536 return NULL;
537}
538
014be2c8 539static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 540 const u8 *mac, __be32 vni)
014be2c8
SS		 541{
542 struct vxlan_fdb *f;
543
3ad7a4b1 544 f = __vxlan_find_mac(vxlan, mac, vni);
016f3d18 545 if (f && f->used != jiffies)
014be2c8
SS		 546 f->used = jiffies;
547
548 return f;
549}
550
a5e7c10a
MR		 551/* caller should hold vxlan->hash_lock */
552static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 553 union vxlan_addr *ip, __be16 port,
54bfd872 554 __be32 vni, __u32 ifindex)
6681712d 555{
3e61aa8f 556 struct vxlan_rdst *rd;
6681712d 557
3e61aa8f 558 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 559 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 560 rd->remote_port == port &&
561 rd->remote_vni == vni &&
562 rd->remote_ifindex == ifindex)
a5e7c10a 563 return rd;
6681712d 564 }
3e61aa8f 565
a5e7c10a
MR		 566 return NULL;
567}
568
1941f1d6
PM		 569int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
570 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
571{
572 struct vxlan_dev *vxlan = netdev_priv(dev);
573 u8 eth_addr[ETH_ALEN + 2] = { 0 };
574 struct vxlan_rdst *rdst;
575 struct vxlan_fdb *f;
576 int rc = 0;
577
578 if (is_multicast_ether_addr(mac) ||
579 is_zero_ether_addr(mac))
580 return -EINVAL;
581
582 ether_addr_copy(eth_addr, mac);
583
584 rcu_read_lock();
585
586 f = __vxlan_find_mac(vxlan, eth_addr, vni);
587 if (!f) {
588 rc = -ENOENT;
589 goto out;
590 }
591
592 rdst = first_remote_rcu(f);
4c59b7d1 593 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
1941f1d6
PM		 594
595out:
596 rcu_read_unlock();
597 return rc;
598}
599EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
600
4f89f5b5
PM		 601static int vxlan_fdb_notify_one(struct notifier_block *nb,
602 const struct vxlan_dev *vxlan,
603 const struct vxlan_fdb *f,
4c59b7d1
PM		 604 const struct vxlan_rdst *rdst,
605 struct netlink_ext_ack *extack)
4f89f5b5
PM		 606{
607 struct switchdev_notifier_vxlan_fdb_info fdb_info;
608 int rc;
609
4c59b7d1 610 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
4f89f5b5
PM		 611 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
612 &fdb_info);
613 return notifier_to_errno(rc);
614}
615
616int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4c59b7d1
PM		 617 struct notifier_block *nb,
618 struct netlink_ext_ack *extack)
4f89f5b5
PM		 619{
620 struct vxlan_dev *vxlan;
621 struct vxlan_rdst *rdst;
622 struct vxlan_fdb *f;
623 unsigned int h;
624 int rc = 0;
625
626 if (!netif_is_vxlan(dev))
627 return -EINVAL;
628 vxlan = netdev_priv(dev);
629
4f89f5b5 630 for (h = 0; h < FDB_HASH_SIZE; ++h) {
fe1e0713 631 spin_lock_bh(&vxlan->hash_lock[h]);
4f89f5b5
PM		 632 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
633 if (f->vni == vni) {
634 list_for_each_entry(rdst, &f->remotes, list) {
635 rc = vxlan_fdb_notify_one(nb, vxlan,
4c59b7d1
PM		 636 f, rdst,
637 extack);
4f89f5b5 638 if (rc)
fe1e0713 639 goto unlock;
4f89f5b5
PM		 640 }
641 }
642 }
fe1e0713 643 spin_unlock_bh(&vxlan->hash_lock[h]);
4f89f5b5 644 }
fe1e0713 645 return 0;
4f89f5b5 646
fe1e0713
L		 647unlock:
648 spin_unlock_bh(&vxlan->hash_lock[h]);
4f89f5b5
PM		 649 return rc;
650}
651EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
652
e5ff4b19
PM		 653void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
654{
655 struct vxlan_dev *vxlan;
656 struct vxlan_rdst *rdst;
657 struct vxlan_fdb *f;
658 unsigned int h;
659
660 if (!netif_is_vxlan(dev))
661 return;
662 vxlan = netdev_priv(dev);
663
e5ff4b19 664 for (h = 0; h < FDB_HASH_SIZE; ++h) {
fe1e0713 665 spin_lock_bh(&vxlan->hash_lock[h]);
e5ff4b19
PM		 666 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
667 if (f->vni == vni)
668 list_for_each_entry(rdst, &f->remotes, list)
669 rdst->offloaded = false;
fe1e0713 670 spin_unlock_bh(&vxlan->hash_lock[h]);
e5ff4b19 671 }
fe1e0713 672
e5ff4b19
PM		 673}
674EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
675
906dc186
TR		 676/* Replace destination of unicast mac */
677static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872 678 union vxlan_addr *ip, __be16 port, __be32 vni,
ccdfd4f7 679 __u32 ifindex, struct vxlan_rdst *oldrd)
906dc186
TR		 680{
681 struct vxlan_rdst *rd;
682
683 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
684 if (rd)
685 return 0;
686
687 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
688 if (!rd)
689 return 0;
0c1d70af 690
ccdfd4f7 691 *oldrd = *rd;
0c1d70af 692 dst_cache_reset(&rd->dst_cache);
e4c7ed41 693 rd->remote_ip = *ip;
906dc186
TR		 694 rd->remote_port = port;
695 rd->remote_vni = vni;
696 rd->remote_ifindex = ifindex;
6ad0b5a4 697 rd->offloaded = false;
906dc186
TR		 698 return 1;
699}
700
a5e7c10a
MR		 701/* Add/update destinations for multicast */
702static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 703 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 704 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 705{
706 struct vxlan_rdst *rd;
707
708 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
709 if (rd)
710 return 0;
711
6681712d
DS		 712 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
713 if (rd == NULL)
714 return -ENOBUFS;
0c1d70af
PA		 715
716 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
717 kfree(rd);
718 return -ENOBUFS;
719 }
720
e4c7ed41 721 rd->remote_ip = *ip;
6681712d 722 rd->remote_port = port;
0efe1173 723 rd->offloaded = false;
6681712d
DS		 724 rd->remote_vni = vni;
725 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 726
727 list_add_tail_rcu(&rd->list, &f->remotes);
728
9e4b93f9 729 *rdp = rd;
6681712d
DS		 730 return 1;
731}
732
dfd8645e
TH		 733static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
734 unsigned int off,
735 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 736 __be32 vni_field,
737 struct gro_remcsum *grc,
0ace2ca8 738 bool nopartial)
dfd8645e 739{
b7fe10e5 740 size_t start, offset;
dfd8645e
TH		 741
742 if (skb->remcsum_offload)
b7fe10e5 743 return vh;
dfd8645e
TH		 744
745 if (!NAPI_GRO_CB(skb)->csum_valid)
746 return NULL;
747
54bfd872
JB		 748 start = vxlan_rco_start(vni_field);
749 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 750
b7fe10e5
TH		 751 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
752 start, offset, grc, nopartial);
dfd8645e
TH		 753
754 skb->remcsum_offload = 1;
755
756 return vh;
757}
758
d4546c25
DM		 759static struct sk_buff *vxlan_gro_receive(struct sock *sk,
760 struct list_head *head,
761 struct sk_buff *skb)
dc01e7d3 762{
d4546c25
DM		 763 struct sk_buff *pp = NULL;
764 struct sk_buff *p;
dc01e7d3 765 struct vxlanhdr *vh, *vh2;
9b174d88 766 unsigned int hlen, off_vx;
dc01e7d3 767 int flush = 1;
5602c48c 768 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 769 __be32 flags;
26c4f7da
TH		 770 struct gro_remcsum grc;
771
772 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 773
774 off_vx = skb_gro_offset(skb);
775 hlen = off_vx + sizeof(*vh);
776 vh = skb_gro_header_fast(skb, off_vx);
777 if (skb_gro_header_hard(skb, hlen)) {
778 vh = skb_gro_header_slow(skb, hlen, off_vx);
779 if (unlikely(!vh))
780 goto out;
781 }
dc01e7d3 782
dfd8645e
TH		 783 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
784
54bfd872 785 flags = vh->vx_flags;
dfd8645e
TH		 786
787 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
788 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 789 vh->vx_vni, &grc,
0ace2ca8
TH		 790 !!(vs->flags &
791 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 792
793 if (!vh)
794 goto out;
795 }
796
b7fe10e5
TH		 797 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
798
d4546c25 799 list_for_each_entry(p, head, list) {
dc01e7d3
OG		 800 if (!NAPI_GRO_CB(p)->same_flow)
801 continue;
802
803 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 804 if (vh->vx_flags != vh2->vx_flags ||
805 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 806 NAPI_GRO_CB(p)->same_flow = 0;
807 continue;
808 }
dc01e7d3
OG		 809 }
810
fcd91dd4 811 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 812 flush = 0;
dc01e7d3 813
dc01e7d3 814out:
603d4cf8 815 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 816
817 return pp;
818}
819
5602c48c 820static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 821{
229740c6
JR		 822 /* Sets 'skb->inner_mac_header' since we are always called with
823 * 'skb->encapsulation' set.
824 */
9b174d88 825 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 826}
827
c7cdbe2e
RP		 828static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
829 __u16 state, __be32 src_vni,
830 __u16 ndm_flags)
25e20e73
RP		 831{
832 struct vxlan_fdb *f;
833
834 f = kmalloc(sizeof(*f), GFP_ATOMIC);
835 if (!f)
836 return NULL;
837 f->state = state;
838 f->flags = ndm_flags;
839 f->updated = f->used = jiffies;
840 f->vni = src_vni;
1274e1cc 841 f->nh = NULL;
79472fe8 842 RCU_INIT_POINTER(f->vdev, vxlan);
1274e1cc 843 INIT_LIST_HEAD(&f->nh_list);
25e20e73
RP		 844 INIT_LIST_HEAD(&f->remotes);
845 memcpy(f->eth_addr, mac, ETH_ALEN);
846
847 return f;
848}
849
7c31e54a
TY		 850static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
851 __be32 src_vni, struct vxlan_fdb *f)
852{
853 ++vxlan->addrcnt;
854 hlist_add_head_rcu(&f->hlist,
855 vxlan_fdb_head(vxlan, mac, src_vni));
856}
857
1274e1cc
RP		 858static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
859 u32 nhid, struct netlink_ext_ack *extack)
860{
861 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
1274e1cc
RP		 862 struct nexthop *nh;
863 int err = -EINVAL;
864
865 if (old_nh && old_nh->id == nhid)
866 return 0;
867
868 nh = nexthop_find_by_id(vxlan->net, nhid);
869 if (!nh) {
870 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
871 goto err_inval;
872 }
873
874 if (nh) {
875 if (!nexthop_get(nh)) {
876 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
877 nh = NULL;
878 goto err_inval;
879 }
ce9ac056 880 if (!nexthop_is_fdb(nh)) {
1274e1cc
RP		 881 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
882 goto err_inval;
883 }
884
50cb8769 885 if (!nexthop_is_multipath(nh)) {
1274e1cc
RP		 886 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
887 goto err_inval;
888 }
889
890 /* check nexthop group family */
1274e1cc
RP		 891 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
892 case AF_INET:
50cb8769 893 if (!nexthop_has_v4(nh)) {
1274e1cc
RP		 894 err = -EAFNOSUPPORT;
895 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
896 goto err_inval;
897 }
898 break;
899 case AF_INET6:
50cb8769 900 if (nexthop_has_v4(nh)) {
1274e1cc
RP		 901 err = -EAFNOSUPPORT;
902 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
903 goto err_inval;
904 }
905 }
906 }
907
908 if (old_nh) {
909 list_del_rcu(&fdb->nh_list);
910 nexthop_put(old_nh);
911 }
912 rcu_assign_pointer(fdb->nh, nh);
913 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
914 return 1;
915
916err_inval:
917 if (nh)
918 nexthop_put(nh);
919 return err;
920}
921
d342894c 922static int vxlan_fdb_create(struct vxlan_dev *vxlan,
25e20e73
RP		 923 const u8 *mac, union vxlan_addr *ip,
924 __u16 state, __be16 port, __be32 src_vni,
45598c1c 925 __be32 vni, __u32 ifindex, __u16 ndm_flags,
1274e1cc
RP		 926 u32 nhid, struct vxlan_fdb **fdb,
927 struct netlink_ext_ack *extack)
25e20e73
RP		 928{
929 struct vxlan_rdst *rd = NULL;
930 struct vxlan_fdb *f;
931 int rc;
932
933 if (vxlan->cfg.addrmax &&
934 vxlan->addrcnt >= vxlan->cfg.addrmax)
935 return -ENOSPC;
936
937 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
c7cdbe2e 938 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
25e20e73
RP		 939 if (!f)
940 return -ENOMEM;
941
1274e1cc
RP		 942 if (nhid)
943 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
944 else
945 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
946 if (rc < 0)
947 goto errout;
25e20e73 948
25e20e73
RP		 949 *fdb = f;
950
951 return 0;
1274e1cc
RP		 952
953errout:
954 kfree(f);
955 return rc;
25e20e73
RP		 956}
957
7c31e54a 958static void __vxlan_fdb_free(struct vxlan_fdb *f)
c2b200e0 959{
c2b200e0 960 struct vxlan_rdst *rd, *nd;
1274e1cc
RP		 961 struct nexthop *nh;
962
963 nh = rcu_dereference_raw(f->nh);
964 if (nh) {
965 rcu_assign_pointer(f->nh, NULL);
79472fe8 966 rcu_assign_pointer(f->vdev, NULL);
1274e1cc
RP		 967 nexthop_put(nh);
968 }
c2b200e0
PM		 969
970 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
971 dst_cache_destroy(&rd->dst_cache);
972 kfree(rd);
973 }
974 kfree(f);
975}
976
7c31e54a
TY		 977static void vxlan_fdb_free(struct rcu_head *head)
978{
979 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
980
981 __vxlan_fdb_free(f);
982}
983
c2b200e0
PM		 984static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
985 bool do_notify, bool swdev_notify)
986{
987 struct vxlan_rdst *rd;
988
989 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
990
991 --vxlan->addrcnt;
1274e1cc
RP		 992 if (do_notify) {
993 if (rcu_access_pointer(f->nh))
994 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
4c59b7d1 995 swdev_notify, NULL);
1274e1cc
RP		 996 else
997 list_for_each_entry(rd, &f->remotes, list)
998 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
999 swdev_notify, NULL);
1000 }
c2b200e0
PM		 1001
1002 hlist_del_rcu(&f->hlist);
79472fe8 1003 list_del_rcu(&f->nh_list);
c2b200e0
PM		 1004 call_rcu(&f->rcu, vxlan_fdb_free);
1005}
1006
fc4aa1ca
PM		 1007static void vxlan_dst_free(struct rcu_head *head)
1008{
1009 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1010
1011 dst_cache_destroy(&rd->dst_cache);
1012 kfree(rd);
1013}
1014
a76d1ca2
PM		 1015static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1016 union vxlan_addr *ip,
1017 __u16 state, __u16 flags,
1018 __be16 port, __be32 vni,
1019 __u32 ifindex, __u16 ndm_flags,
1274e1cc 1020 struct vxlan_fdb *f, u32 nhid,
4c59b7d1
PM		 1021 bool swdev_notify,
1022 struct netlink_ext_ack *extack)
d342894c 1023{
45598c1c 1024 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
9e4b93f9 1025 struct vxlan_rdst *rd = NULL;
ccdfd4f7 1026 struct vxlan_rdst oldrd;
d342894c 1027 int notify = 0;
61f46fe8
PM		 1028 int rc = 0;
1029 int err;
d342894c 1030
1274e1cc
RP		 1031 if (nhid && !rcu_access_pointer(f->nh)) {
1032 NL_SET_ERR_MSG(extack,
1033 "Cannot replace an existing non nexthop fdb with a nexthop");
1034 return -EOPNOTSUPP;
1035 }
1036
1037 if (nhid && (flags & NLM_F_APPEND)) {
1038 NL_SET_ERR_MSG(extack,
1039 "Cannot append to a nexthop fdb");
1040 return -EOPNOTSUPP;
1041 }
1042
a76d1ca2
PM		 1043 /* Do not allow an externally learned entry to take over an entry added
1044 * by the user.
1045 */
1046 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1047 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1048 if (f->state != state) {
1049 f->state = state;
1050 f->updated = jiffies;
1051 notify = 1;
ae884082 1052 }
a76d1ca2
PM		 1053 if (f->flags != fdb_flags) {
1054 f->flags = fdb_flags;
1055 f->updated = jiffies;
1056 notify = 1;
906dc186 1057 }
a76d1ca2 1058 }
6681712d 1059
a76d1ca2
PM		 1060 if ((flags & NLM_F_REPLACE)) {
1061 /* Only change unicasts */
1062 if (!(is_multicast_ether_addr(f->eth_addr) ||
1063 is_zero_ether_addr(f->eth_addr))) {
1274e1cc
RP		 1064 if (nhid) {
1065 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1066 if (rc < 0)
1067 return rc;
1068 } else {
1069 rc = vxlan_fdb_replace(f, ip, port, vni,
1070 ifindex, &oldrd);
1071 }
6681712d 1072 notify |= rc;
a76d1ca2 1073 } else {
1274e1cc 1074 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
906dc186 1075 return -EOPNOTSUPP;
a76d1ca2
PM		 1076 }
1077 }
1078 if ((flags & NLM_F_APPEND) &&
1079 (is_multicast_ether_addr(f->eth_addr) ||
1080 is_zero_ether_addr(f->eth_addr))) {
1081 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
906dc186 1082
25e20e73 1083 if (rc < 0)
17b46365 1084 return rc;
a76d1ca2 1085 notify |= rc;
d342894c 1086 }
1087
a76d1ca2
PM		 1088 if (ndm_flags & NTF_USE)
1089 f->used = jiffies;
1090
9e4b93f9
ND		 1091 if (notify) {
1092 if (rd == NULL)
1093 rd = first_remote_rtnl(f);
a76d1ca2 1094
61f46fe8 1095 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
4c59b7d1 1096 swdev_notify, extack);
61f46fe8
PM		 1097 if (err)
1098 goto err_notify;
9e4b93f9 1099 }
d342894c 1100
1101 return 0;
61f46fe8
PM		 1102
1103err_notify:
1274e1cc
RP		 1104 if (nhid)
1105 return err;
61f46fe8
PM		 1106 if ((flags & NLM_F_REPLACE) && rc)
1107 *rd = oldrd;
fc4aa1ca 1108 else if ((flags & NLM_F_APPEND) && rc) {
61f46fe8 1109 list_del_rcu(&rd->list);
fc4aa1ca
PM		 1110 call_rcu(&rd->rcu, vxlan_dst_free);
1111 }
61f46fe8 1112 return err;
d342894c 1113}
1114
a76d1ca2
PM		 1115static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1116 const u8 *mac, union vxlan_addr *ip,
1117 __u16 state, __u16 flags,
1118 __be16 port, __be32 src_vni, __be32 vni,
1274e1cc 1119 __u32 ifindex, __u16 ndm_flags, u32 nhid,
4c59b7d1
PM		 1120 bool swdev_notify,
1121 struct netlink_ext_ack *extack)
a76d1ca2
PM		 1122{
1123 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1124 struct vxlan_fdb *f;
1125 int rc;
1126
1127 /* Disallow replace to add a multicast entry */
1128 if ((flags & NLM_F_REPLACE) &&
1129 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1130 return -EOPNOTSUPP;
1131
1132 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1133 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1274e1cc 1134 vni, ifindex, fdb_flags, nhid, &f, extack);
a76d1ca2
PM		 1135 if (rc < 0)
1136 return rc;
1137
7c31e54a 1138 vxlan_fdb_insert(vxlan, mac, src_vni, f);
61f46fe8 1139 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
4c59b7d1 1140 swdev_notify, extack);
61f46fe8
PM		 1141 if (rc)
1142 goto err_notify;
1143
a76d1ca2 1144 return 0;
61f46fe8
PM		 1145
1146err_notify:
1147 vxlan_fdb_destroy(vxlan, f, false, false);
1148 return rc;
a76d1ca2
PM		 1149}
1150
1151/* Add new entry to forwarding table -- assumes lock held */
1152static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1153 const u8 *mac, union vxlan_addr *ip,
1154 __u16 state, __u16 flags,
1155 __be16 port, __be32 src_vni, __be32 vni,
1274e1cc 1156 __u32 ifindex, __u16 ndm_flags, u32 nhid,
4c59b7d1
PM		 1157 bool swdev_notify,
1158 struct netlink_ext_ack *extack)
a76d1ca2
PM		 1159{
1160 struct vxlan_fdb *f;
1161
1162 f = __vxlan_find_mac(vxlan, mac, src_vni);
1163 if (f) {
1164 if (flags & NLM_F_EXCL) {
1165 netdev_dbg(vxlan->dev,
1166 "lost race to create %pM\n", mac);
1167 return -EEXIST;
1168 }
1169
1170 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1171 vni, ifindex, ndm_flags, f,
1274e1cc 1172 nhid, swdev_notify, extack);
a76d1ca2
PM		 1173 } else {
1174 if (!(flags & NLM_F_CREATE))
1175 return -ENOENT;
1176
1177 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1178 port, src_vni, vni, ifindex,
1274e1cc
RP		 1179 ndm_flags, nhid, swdev_notify,
1180 extack);
a76d1ca2
PM		 1181 }
1182}
1183
35cf2845 1184static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
0e6160f3 1185 struct vxlan_rdst *rd, bool swdev_notify)
35cf2845
LR		 1186{
1187 list_del_rcu(&rd->list);
4c59b7d1 1188 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
35cf2845
LR		 1189 call_rcu(&rd->rcu, vxlan_dst_free);
1190}
1191
f0b074be 1192static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1 1193 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1274e1cc 1194 __be32 *vni, u32 *ifindex, u32 *nhid)
d342894c 1195{
6681712d 1196 struct net *net = dev_net(vxlan->dev);
e4c7ed41 1197 int err;
d342894c 1198
72b48682
RP		 1199 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1200 tb[NDA_PORT]))
1201 return -EINVAL;
1202
f0b074be 1203 if (tb[NDA_DST]) {
e4c7ed41
CW		 1204 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1205 if (err)
1206 return err;
f0b074be 1207 } else {
e4c7ed41 1208 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
98c81476 1209
e4c7ed41
CW		 1210 if (remote->sa.sa_family == AF_INET) {
1211 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1212 ip->sa.sa_family = AF_INET;
1213#if IS_ENABLED(CONFIG_IPV6)
1214 } else {
1215 ip->sin6.sin6_addr = in6addr_any;
1216 ip->sa.sa_family = AF_INET6;
1217#endif
1218 }
f0b074be 1219 }
d342894c 1220
6681712d 1221 if (tb[NDA_PORT]) {
73cf3317 1222 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 1223 return -EINVAL;
f0b074be
MR		 1224 *port = nla_get_be16(tb[NDA_PORT]);
1225 } else {
0dfbdf41 1226 *port = vxlan->cfg.dst_port;
f0b074be 1227 }
6681712d
DS		 1228
1229 if (tb[NDA_VNI]) {
1230 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1231 return -EINVAL;
54bfd872 1232 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 1233 } else {
1234 *vni = vxlan->default_dst.remote_vni;
1235 }
6681712d 1236
3ad7a4b1
RP		 1237 if (tb[NDA_SRC_VNI]) {
1238 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1239 return -EINVAL;
1240 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1241 } else {
1242 *src_vni = vxlan->default_dst.remote_vni;
1243 }
1244
6681712d 1245 if (tb[NDA_IFINDEX]) {
5abb0029 1246 struct net_device *tdev;
6681712d
DS		 1247
1248 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1249 return -EINVAL;
f0b074be 1250 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 1251 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 1252 if (!tdev)
6681712d 1253 return -EADDRNOTAVAIL;
f0b074be
MR		 1254 } else {
1255 *ifindex = 0;
1256 }
1257
1274e1cc
RP		 1258 if (tb[NDA_NH_ID])
1259 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1260 else
1261 *nhid = 0;
1262
f0b074be
MR		 1263 return 0;
1264}
1265
1266/* Add static entry (via netlink) */
1267static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1268 struct net_device *dev,
87b0984e
PM		 1269 const unsigned char *addr, u16 vid, u16 flags,
1270 struct netlink_ext_ack *extack)
f0b074be
MR		 1271{
1272 struct vxlan_dev *vxlan = netdev_priv(dev);
1273 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 1274 union vxlan_addr ip;
f0b074be 1275 __be16 port;
3ad7a4b1 1276 __be32 src_vni, vni;
1274e1cc 1277 u32 ifindex, nhid;
fe1e0713 1278 u32 hash_index;
f0b074be
MR		 1279 int err;
1280
1281 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1282 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1283 ndm->ndm_state);
1284 return -EINVAL;
1285 }
1286
1274e1cc 1287 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
f0b074be
MR		 1288 return -EINVAL;
1289
1274e1cc
RP		 1290 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1291 &nhid);
f0b074be
MR		 1292 if (err)
1293 return err;
6681712d 1294
5933a7bb
MR		 1295 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1296 return -EAFNOSUPPORT;
1297
fe1e0713
L		 1298 hash_index = fdb_head_index(vxlan, addr, src_vni);
1299 spin_lock_bh(&vxlan->hash_lock[hash_index]);
25e20e73 1300 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
45598c1c
PM		 1301 port, src_vni, vni, ifindex,
1302 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1274e1cc 1303 nhid, true, extack);
fe1e0713 1304 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
d342894c 1305
1306 return err;
1307}
1308
3ad7a4b1
RP		 1309static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1310 const unsigned char *addr, union vxlan_addr ip,
fc39c38b 1311 __be16 port, __be32 src_vni, __be32 vni,
0e6160f3 1312 u32 ifindex, bool swdev_notify)
d342894c 1313{
bc7892ba 1314 struct vxlan_rdst *rd = NULL;
1274e1cc 1315 struct vxlan_fdb *f;
3ad7a4b1 1316 int err = -ENOENT;
bc7892ba 1317
3ad7a4b1 1318 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 1319 if (!f)
3ad7a4b1 1320 return err;
bc7892ba 1321
e4c7ed41
CW		 1322 if (!vxlan_addr_any(&ip)) {
1323 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 1324 if (!rd)
1325 goto out;
1326 }
1327
bc7892ba
MR		 1328 /* remove a destination if it's not the only one on the list,
1329 * otherwise destroy the fdb entry
1330 */
1331 if (rd && !list_is_singular(&f->remotes)) {
0e6160f3 1332 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
bc7892ba 1333 goto out;
d342894c 1334 }
bc7892ba 1335
0e6160f3 1336 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
bc7892ba
MR		 1337
1338out:
3ad7a4b1
RP		 1339 return 0;
1340}
1341
1342/* Delete entry (via netlink) */
1343static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1344 struct net_device *dev,
1345 const unsigned char *addr, u16 vid)
1346{
1347 struct vxlan_dev *vxlan = netdev_priv(dev);
1348 union vxlan_addr ip;
1349 __be32 src_vni, vni;
1274e1cc 1350 u32 ifindex, nhid;
fe1e0713 1351 u32 hash_index;
1274e1cc 1352 __be16 port;
3ad7a4b1
RP		 1353 int err;
1354
1274e1cc
RP		 1355 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1356 &nhid);
3ad7a4b1
RP		 1357 if (err)
1358 return err;
1359
fe1e0713
L		 1360 hash_index = fdb_head_index(vxlan, addr, src_vni);
1361 spin_lock_bh(&vxlan->hash_lock[hash_index]);
0e6160f3
PM		 1362 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1363 true);
fe1e0713 1364 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
d342894c 1365
1366 return err;
1367}
1368
1369/* Dump forwarding table */
1370static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1371 struct net_device *dev,
d297653d 1372 struct net_device *filter_dev, int *idx)
d342894c 1373{
1374 struct vxlan_dev *vxlan = netdev_priv(dev);
1375 unsigned int h;
d297653d 1376 int err = 0;
d342894c 1377
1378 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1379 struct vxlan_fdb *f;
d342894c 1380
b5141915 1381 rcu_read_lock();
b67bfe0d 1382 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 1383 struct vxlan_rdst *rd;
6681712d 1384
1274e1cc 1385 if (rcu_access_pointer(f->nh)) {
b18e9834
RP		 1386 if (*idx < cb->args[2])
1387 goto skip_nh;
1274e1cc
RP		 1388 err = vxlan_fdb_info(skb, vxlan, f,
1389 NETLINK_CB(cb->skb).portid,
1390 cb->nlh->nlmsg_seq,
1391 RTM_NEWNEIGH,
1392 NLM_F_MULTI, NULL);
b5141915
IS		 1393 if (err < 0) {
1394 rcu_read_unlock();
1274e1cc 1395 goto out;
b5141915 1396 }
b18e9834
RP		 1397skip_nh:
1398 *idx += 1;
1274e1cc
RP		 1399 continue;
1400 }
1401
3e61aa8f 1402 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 1403 if (*idx < cb->args[2])
07a51cd3
AW		 1404 goto skip;
1405
6681712d
DS		 1406 err = vxlan_fdb_info(skb, vxlan, f,
1407 NETLINK_CB(cb->skb).portid,
1408 cb->nlh->nlmsg_seq,
1409 RTM_NEWNEIGH,
1410 NLM_F_MULTI, rd);
b5141915
IS		 1411 if (err < 0) {
1412 rcu_read_unlock();
3e61aa8f 1413 goto out;
b5141915 1414 }
3e61aa8f 1415skip:
d297653d 1416 *idx += 1;
07a51cd3 1417 }
d342894c 1418 }
b5141915 1419 rcu_read_unlock();
d342894c 1420 }
3e61aa8f 1421out:
d297653d 1422 return err;
d342894c 1423}
1424
474c3c89
RP		 1425static int vxlan_fdb_get(struct sk_buff *skb,
1426 struct nlattr *tb[],
1427 struct net_device *dev,
1428 const unsigned char *addr,
1429 u16 vid, u32 portid, u32 seq,
1430 struct netlink_ext_ack *extack)
1431{
1432 struct vxlan_dev *vxlan = netdev_priv(dev);
1433 struct vxlan_fdb *f;
1434 __be32 vni;
1435 int err;
1436
1437 if (tb[NDA_VNI])
1438 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1439 else
1440 vni = vxlan->default_dst.remote_vni;
1441
1442 rcu_read_lock();
1443
1444 f = __vxlan_find_mac(vxlan, addr, vni);
1445 if (!f) {
1446 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1447 err = -ENOENT;
1448 goto errout;
1449 }
1450
1451 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1452 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1453errout:
1454 rcu_read_unlock();
1455 return err;
1456}
1457
d342894c 1458/* Watch incoming packets to learn mapping between Ethernet address
1459 * and Tunnel endpoint.
c4b49512 1460 * Return true if packet is bogus and should be dropped.
d342894c 1461 */
26a41ae6 1462static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1463 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1464 u32 src_ifindex, __be32 vni)
d342894c 1465{
1466 struct vxlan_dev *vxlan = netdev_priv(dev);
1467 struct vxlan_fdb *f;
87613de9
MS		 1468 u32 ifindex = 0;
1469
1470#if IS_ENABLED(CONFIG_IPV6)
1471 if (src_ip->sa.sa_family == AF_INET6 &&
1472 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1473 ifindex = src_ifindex;
1474#endif
d342894c 1475
3ad7a4b1 1476 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1477 if (likely(f)) {
5ca5461c 1478 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1479
87613de9
MS		 1480 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1481 rdst->remote_ifindex == ifindex))
26a41ae6 1482 return false;
1483
1484 /* Don't migrate static entries, drop packets */
e0090a9e 1485 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1486 return true;
d342894c 1487
1274e1cc
RP		 1488 /* Don't override an fdb with nexthop with a learnt entry */
1489 if (rcu_access_pointer(f->nh))
1490 return true;
1491
d342894c 1492 if (net_ratelimit())
1493 netdev_info(dev,
e4c7ed41 1494 "%pM migrated from %pIS to %pIS\n",
a4870f79 1495 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1496
e4c7ed41 1497 rdst->remote_ip = *src_ip;
d342894c 1498 f->updated = jiffies;
4c59b7d1 1499 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
d342894c 1500 } else {
fe1e0713
L		 1501 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1502
d342894c 1503 /* learned new entry */
fe1e0713 1504 spin_lock(&vxlan->hash_lock[hash_index]);
3bf74b1a 1505
1506 /* close off race between vxlan_flush and incoming packets */
1507 if (netif_running(dev))
25e20e73 1508 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1509 NUD_REACHABLE,
1510 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1511 vxlan->cfg.dst_port,
3ad7a4b1 1512 vni,
3bf74b1a 1513 vxlan->default_dst.remote_vni,
1274e1cc 1514 ifindex, NTF_SELF, 0, true, NULL);
fe1e0713 1515 spin_unlock(&vxlan->hash_lock[hash_index]);
d342894c 1516 }
26a41ae6 1517
1518 return false;
d342894c 1519}
1520
d342894c 1521/* See if multicast group is already in use by other ID */
95ab0991 1522static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1523{
553675fb 1524 struct vxlan_dev *vxlan;
c6fcc4fc 1525 struct vxlan_sock *sock4;
4053ab1b
AB		 1526#if IS_ENABLED(CONFIG_IPV6)
1527 struct vxlan_sock *sock6;
1528#endif
b1be00a6 1529 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1530
c6fcc4fc 1531 sock4 = rtnl_dereference(dev->vn4_sock);
1532
95ab0991
G		 1533 /* The vxlan_sock is only used by dev, leaving group has
1534 * no effect on other vxlan devices.
1535 */
66af846f 1536 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1537 return false;
b1be00a6 1538#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1539 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1540 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1541 return false;
1542#endif
95ab0991 1543
553675fb 1544 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1545 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1546 continue;
d342894c 1547
c6fcc4fc 1548 if (family == AF_INET &&
1549 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1550 continue;
b1be00a6 1551#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1552 if (family == AF_INET6 &&
1553 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1554 continue;
1555#endif
95ab0991
G		 1556
1557 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1558 &dev->default_dst.remote_ip))
1559 continue;
1560
1561 if (vxlan->default_dst.remote_ifindex !=
1562 dev->default_dst.remote_ifindex)
1563 continue;
1564
1565 return true;
553675fb 1566 }
d342894c 1567
1568 return false;
1569}
1570
544a773a 1571static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1572{
b1be00a6 1573 struct vxlan_net *vn;
012a5729 1574
b1be00a6 1575 if (!vs)
544a773a 1576 return false;
66af846f 1577 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1578 return false;
d342894c 1579
b1be00a6 1580 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1581 spin_lock(&vn->sock_lock);
7c47cedf 1582 hlist_del_rcu(&vs->hlist);
e7b3db5e 1583 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1584 (vs->flags & VXLAN_F_GPE) ?
1585 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1586 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1587 spin_unlock(&vn->sock_lock);
1588
544a773a 1589 return true;
d342894c 1590}
1591
b1be00a6
JB		 1592static void vxlan_sock_release(struct vxlan_dev *vxlan)
1593{
c6fcc4fc 1594 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1595#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1596 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1597
57d88182 1598 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1599#endif
1600
57d88182 1601 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1602 synchronize_net();
1603
a53cb29b
MB		 1604 vxlan_vs_del_dev(vxlan);
1605
c6fcc4fc 1606 if (__vxlan_sock_release_prep(sock4)) {
1607 udp_tunnel_sock_release(sock4->sock);
1608 kfree(sock4);
544a773a
HFS		 1609 }
1610
1611#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1612 if (__vxlan_sock_release_prep(sock6)) {
1613 udp_tunnel_sock_release(sock6->sock);
1614 kfree(sock6);
544a773a 1615 }
b1be00a6
JB		 1616#endif
1617}
1618
56ef9c90 1619/* Update multicast group membership when first VNI on
c4b49512 1620 * multicast address is brought up
7c47cedf 1621 */
56ef9c90 1622static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1623{
b1be00a6 1624 struct sock *sk;
e4c7ed41
CW		 1625 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1626 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1627 int ret = -EINVAL;
d342894c 1628
e4c7ed41 1629 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1630 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1631 struct ip_mreqn mreq = {
1632 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1633 .imr_ifindex = ifindex,
1634 };
1635
c6fcc4fc 1636 sk = sock4->sock->sk;
b1be00a6 1637 lock_sock(sk);
56ef9c90 1638 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1639 release_sock(sk);
e4c7ed41
CW		 1640#if IS_ENABLED(CONFIG_IPV6)
1641 } else {
c6fcc4fc 1642 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1643
1644 sk = sock6->sock->sk;
b1be00a6 1645 lock_sock(sk);
56ef9c90
MRL		 1646 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1647 &ip->sin6.sin6_addr);
b1be00a6 1648 release_sock(sk);
e4c7ed41
CW		 1649#endif
1650 }
3fc2de2f 1651
56ef9c90 1652 return ret;
3fc2de2f 1653}
1654
1655/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1656static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1657{
b1be00a6 1658 struct sock *sk;
e4c7ed41
CW		 1659 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1660 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1661 int ret = -EINVAL;
3fc2de2f 1662
e4c7ed41 1663 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1664 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1665 struct ip_mreqn mreq = {
1666 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1667 .imr_ifindex = ifindex,
1668 };
1669
c6fcc4fc 1670 sk = sock4->sock->sk;
b1be00a6 1671 lock_sock(sk);
56ef9c90 1672 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1673 release_sock(sk);
e4c7ed41
CW		 1674#if IS_ENABLED(CONFIG_IPV6)
1675 } else {
c6fcc4fc 1676 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1677
1678 sk = sock6->sock->sk;
b1be00a6 1679 lock_sock(sk);
56ef9c90
MRL		 1680 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1681 &ip->sin6.sin6_addr);
b1be00a6 1682 release_sock(sk);
e4c7ed41
CW		 1683#endif
1684 }
d342894c 1685
56ef9c90 1686 return ret;
d342894c 1687}
1688
f14ecebb
JB		 1689static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1690 struct sk_buff *skb, u32 vxflags)
dfd8645e 1691{
7d34fa75 1692 size_t start, offset;
dfd8645e 1693
f14ecebb
JB		 1694 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1695 goto out;
b7fe10e5 1696
f14ecebb
JB		 1697 start = vxlan_rco_start(unparsed->vx_vni);
1698 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1699
7d34fa75 1700 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1701 return false;
dfd8645e 1702
be5cfeab
JB		 1703 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1704 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1705out:
1706 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1707 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1708 return true;
dfd8645e
TH		 1709}
1710
f14ecebb 1711static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1712 struct sk_buff *skb, u32 vxflags,
10a5af23 1713 struct vxlan_metadata *md)
3288af08 1714{
f14ecebb 1715 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1716 struct metadata_dst *tun_dst;
f14ecebb
JB		 1717
1718 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1719 goto out;
3288af08 1720
3288af08
JB		 1721 md->gbp = ntohs(gbp->policy_id);
1722
10a5af23 1723 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1724 if (tun_dst) {
3288af08 1725 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1726 tun_dst->u.tun_info.options_len = sizeof(*md);
1727 }
3288af08
JB		 1728 if (gbp->dont_learn)
1729 md->gbp |= VXLAN_GBP_DONT_LEARN;
1730
1731 if (gbp->policy_applied)
1732 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1733
64f87d36
JB		 1734 /* In flow-based mode, GBP is carried in dst_metadata */
1735 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1736 skb->mark = md->gbp;
f14ecebb
JB		 1737out:
1738 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1739}
1740
e1e5314d 1741static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1742 __be16 *protocol,
e1e5314d
JB		 1743 struct sk_buff *skb, u32 vxflags)
1744{
1745 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1746
1747 /* Need to have Next Protocol set for interfaces in GPE mode. */
1748 if (!gpe->np_applied)
1749 return false;
1750 /* "The initial version is 0. If a receiver does not support the
1751 * version indicated it MUST drop the packet.
1752 */
1753 if (gpe->version != 0)
1754 return false;
1755 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1756 * processing MUST occur." However, we don't implement OAM
1757 * processing, thus drop the packet.
1758 */
1759 if (gpe->oam_flag)
1760 return false;
1761
fa20e0e3
JB		 1762 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1763 if (!*protocol)
e1e5314d 1764 return false;
e1e5314d
JB		 1765
1766 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1767 return true;
1768}
1769
1ab016e2
JB		 1770static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1771 struct vxlan_sock *vs,
3ad7a4b1 1772 struct sk_buff *skb, __be32 vni)
614732ea 1773{
614732ea 1774 union vxlan_addr saddr;
87613de9 1775 u32 ifindex = skb->dev->ifindex;
614732ea 1776
614732ea 1777 skb_reset_mac_header(skb);
614732ea
TG		 1778 skb->protocol = eth_type_trans(skb, vxlan->dev);
1779 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1780
1781 /* Ignore packet loops (and multicast echo) */
1782 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1783 return false;
614732ea 1784
760c6805 1785 /* Get address from the outer IP header */
ce212d0f 1786 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1787 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1788 saddr.sa.sa_family = AF_INET;
1789#if IS_ENABLED(CONFIG_IPV6)
1790 } else {
1ab016e2 1791 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1792 saddr.sa.sa_family = AF_INET6;
1793#endif
1794 }
1795
dc5321d7 1796 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1797 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1798 return false;
1799
1800 return true;
1801}
1802
760c6805
JB		 1803static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1804 struct sk_buff *skb)
1805{
1806 int err = 0;
1807
1808 if (vxlan_get_sk_family(vs) == AF_INET)
1809 err = IP_ECN_decapsulate(oiph, skb);
1810#if IS_ENABLED(CONFIG_IPV6)
1811 else
1812 err = IP6_ECN_decapsulate(oiph, skb);
1813#endif
1814
1815 if (unlikely(err) && log_ecn_error) {
1816 if (vxlan_get_sk_family(vs) == AF_INET)
1817 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1818 &((struct iphdr *)oiph)->saddr,
1819 ((struct iphdr *)oiph)->tos);
1820 else
1821 net_info_ratelimited("non-ECT from %pI6\n",
1822 &((struct ipv6hdr *)oiph)->saddr);
1823 }
1824 return err <= 1;
1825}
1826
d342894c 1827/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1828static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1829{
c9e78efb 1830 struct vxlan_dev *vxlan;
5cfccc5a 1831 struct vxlan_sock *vs;
f14ecebb 1832 struct vxlanhdr unparsed;
ee122c79
TG		 1833 struct vxlan_metadata _md;
1834 struct vxlan_metadata *md = &_md;
61618eea 1835 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1836 bool raw_proto = false;
f2d1968e 1837 void *oiph;
3ad7a4b1 1838 __be32 vni = 0;
d342894c 1839
e1e5314d 1840 /* Need UDP and VXLAN header to be present */
7ce04758 1841 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1842 goto drop;
d342894c 1843
f14ecebb 1844 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1845 /* VNI flag always required to be set */
1846 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1847 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1848 ntohl(vxlan_hdr(skb)->vx_flags),
1849 ntohl(vxlan_hdr(skb)->vx_vni));
1850 /* Return non vxlan pkt */
e5aed006 1851 goto drop;
d342894c 1852 }
288b01c8
JB		 1853 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1854 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1855
559835ea 1856 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1857 if (!vs)
d342894c 1858 goto drop;
d342894c 1859
3ad7a4b1
RP		 1860 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1861
49f810f0 1862 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1863 if (!vxlan)
1864 goto drop;
1865
e1e5314d
JB		 1866 /* For backwards compatibility, only allow reserved fields to be
1867 * used by VXLAN extensions if explicitly requested.
1868 */
1869 if (vs->flags & VXLAN_F_GPE) {
1870 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1871 goto drop;
1872 raw_proto = true;
1873 }
1874
1875 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1876 !net_eq(vxlan->net, dev_net(vxlan->dev))))
98c81476 1877 goto drop;
c9e78efb 1878
2ae2904b 1879 if (vs->flags & VXLAN_F_REMCSUM_RX)
0189399c 1880 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
2ae2904b
FF		 1881 goto drop;
1882
ee122c79 1883 if (vxlan_collect_metadata(vs)) {
10a5af23 1884 struct metadata_dst *tun_dst;
07dabf20 1885
c29a70d2 1886 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1887 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1888
ee122c79
TG		 1889 if (!tun_dst)
1890 goto drop;
1891
0f1b7354 1892 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1893
1894 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1895 } else {
1896 memset(md, 0, sizeof(*md));
1897 }
1898
f14ecebb 1899 if (vs->flags & VXLAN_F_GBP)
10a5af23 1900 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1901 /* Note that GBP and GPE can never be active together. This is
1902 * ensured in vxlan_dev_configure.
1903 */
3511494c 1904
f14ecebb 1905 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1906 /* If there are any unprocessed flags remaining treat
1907 * this as a malformed packet. This behavior diverges from
1908 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1909 * in reserved fields are to be ignored. The approach here
c4b49512 1910 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1911 * is more robust and provides a little more security in
1912 * adding extensions to VXLAN.
1913 */
288b01c8 1914 goto drop;
3bf39475
TH		 1915 }
1916
e1e5314d 1917 if (!raw_proto) {
3ad7a4b1 1918 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1919 goto drop;
1920 } else {
8be0cfa4 1921 skb_reset_mac_header(skb);
e1e5314d
JB		 1922 skb->dev = vxlan->dev;
1923 skb->pkt_type = PACKET_HOST;
1924 }
f2d1968e 1925
f2d1968e
JB		 1926 oiph = skb_network_header(skb);
1927 skb_reset_network_header(skb);
1928
1929 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1930 ++vxlan->dev->stats.rx_frame_errors;
1931 ++vxlan->dev->stats.rx_errors;
1932 goto drop;
1933 }
1934
59cbf56f
ED		 1935 rcu_read_lock();
1936
1937 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1938 rcu_read_unlock();
1939 atomic_long_inc(&vxlan->dev->rx_dropped);
1940 goto drop;
1941 }
1942
1f8dda1d 1943 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
f2d1968e 1944 gro_cells_receive(&vxlan->gro_cells, skb);
59cbf56f
ED		 1945
1946 rcu_read_unlock();
1947
5cfccc5a
PS		 1948 return 0;
1949
1950drop:
288b01c8
JB		 1951 /* Consume bad packet */
1952 kfree_skb(skb);
1953 return 0;
5cfccc5a
PS		 1954}
1955
c3a43b9f
SB		 1956/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1957static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1958{
1959 struct vxlan_dev *vxlan;
1960 struct vxlan_sock *vs;
1961 struct vxlanhdr *hdr;
1962 __be32 vni;
1963
8399a693 1964 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
c3a43b9f
SB		 1965 return -EINVAL;
1966
1967 hdr = vxlan_hdr(skb);
1968
1969 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1970 return -EINVAL;
1971
1972 vs = rcu_dereference_sk_user_data(sk);
1973 if (!vs)
1974 return -ENOENT;
1975
1976 vni = vxlan_vni(hdr->vx_vni);
1977 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1978 if (!vxlan)
1979 return -ENOENT;
1980
1981 return 0;
1982}
1983
3ad7a4b1 1984static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1985{
1986 struct vxlan_dev *vxlan = netdev_priv(dev);
1987 struct arphdr *parp;
1988 u8 *arpptr, *sha;
1989 __be32 sip, tip;
1990 struct neighbour *n;
1991
1992 if (dev->flags & IFF_NOARP)
1993 goto out;
1994
1995 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1996 dev->stats.tx_dropped++;
1997 goto out;
1998 }
1999 parp = arp_hdr(skb);
2000
2001 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2002 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2003 parp->ar_pro != htons(ETH_P_IP) ||
2004 parp->ar_op != htons(ARPOP_REQUEST) ||
2005 parp->ar_hln != dev->addr_len ||
2006 parp->ar_pln != 4)
2007 goto out;
2008 arpptr = (u8 *)parp + sizeof(struct arphdr);
2009 sha = arpptr;
2010 arpptr += dev->addr_len; /* sha */
2011 memcpy(&sip, arpptr, sizeof(sip));
2012 arpptr += sizeof(sip);
2013 arpptr += dev->addr_len; /* tha */
2014 memcpy(&tip, arpptr, sizeof(tip));
2015
2016 if (ipv4_is_loopback(tip) ||
2017 ipv4_is_multicast(tip))
2018 goto out;
2019
2020 n = neigh_lookup(&arp_tbl, &tip, dev);
2021
2022 if (n) {
e4f67add
DS		 2023 struct vxlan_fdb *f;
2024 struct sk_buff *reply;
2025
2026 if (!(n->nud_state & NUD_CONNECTED)) {
2027 neigh_release(n);
2028 goto out;
2029 }
2030
3ad7a4b1 2031 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 2032 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 2033 /* bridge-local neighbor */
2034 neigh_release(n);
2035 goto out;
2036 }
2037
2038 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2039 n->ha, sha);
2040
2041 neigh_release(n);
2042
7346135d
DS		 2043 if (reply == NULL)
2044 goto out;
2045
e4f67add
DS		 2046 skb_reset_mac_header(reply);
2047 __skb_pull(reply, skb_network_offset(reply));
2048 reply->ip_summed = CHECKSUM_UNNECESSARY;
2049 reply->pkt_type = PACKET_HOST;
2050
2051 if (netif_rx_ni(reply) == NET_RX_DROP)
2052 dev->stats.rx_dropped++;
dc5321d7 2053 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 2054 union vxlan_addr ipa = {
2055 .sin.sin_addr.s_addr = tip,
a45e92a5 2056 .sin.sin_family = AF_INET,
e4c7ed41
CW		 2057 };
2058
2059 vxlan_ip_miss(dev, &ipa);
2060 }
e4f67add
DS		 2061out:
2062 consume_skb(skb);
2063 return NETDEV_TX_OK;
2064}
2065
f564f45c 2066#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 2067static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2068 struct neighbour *n, bool isrouter)
2069{
2070 struct net_device *dev = request->dev;
2071 struct sk_buff *reply;
2072 struct nd_msg *ns, *na;
2073 struct ipv6hdr *pip6;
2074 u8 *daddr;
2075 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2076 int ns_olen;
2077 int i, len;
2078
f1fb08f6 2079 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 2080 return NULL;
2081
2082 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2083 sizeof(*na) + na_olen + dev->needed_tailroom;
2084 reply = alloc_skb(len, GFP_ATOMIC);
2085 if (reply == NULL)
2086 return NULL;
2087
2088 reply->protocol = htons(ETH_P_IPV6);
2089 reply->dev = dev;
2090 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2091 skb_push(reply, sizeof(struct ethhdr));
6297b91c 2092 skb_reset_mac_header(reply);
4b29dba9 2093
f1fb08f6 2094 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 2095
2096 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 2097 ns_olen = request->len - skb_network_offset(request) -
2098 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9 2099 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
8066e6b4
IS		 2100 if (!ns->opt[i + 1]) {
2101 kfree_skb(reply);
2102 return NULL;
2103 }
4b29dba9
DS		 2104 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2105 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2106 break;
2107 }
2108 }
2109
2110 /* Ethernet header */
2111 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2112 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2113 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2114 reply->protocol = htons(ETH_P_IPV6);
2115
2116 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 2117 skb_reset_network_header(reply);
4b29dba9
DS		 2118 skb_put(reply, sizeof(struct ipv6hdr));
2119
2120 /* IPv6 header */
2121
2122 pip6 = ipv6_hdr(reply);
2123 memset(pip6, 0, sizeof(struct ipv6hdr));
2124 pip6->version = 6;
2125 pip6->priority = ipv6_hdr(request)->priority;
2126 pip6->nexthdr = IPPROTO_ICMPV6;
2127 pip6->hop_limit = 255;
2128 pip6->daddr = ipv6_hdr(request)->saddr;
2129 pip6->saddr = *(struct in6_addr *)n->primary_key;
2130
2131 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 2132 skb_reset_transport_header(reply);
4b29dba9 2133
4b29dba9 2134 /* Neighbor Advertisement */
b080db58 2135 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 2136 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2137 na->icmph.icmp6_router = isrouter;
2138 na->icmph.icmp6_override = 1;
2139 na->icmph.icmp6_solicited = 1;
2140 na->target = ns->target;
2141 ether_addr_copy(&na->opt[2], n->ha);
2142 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2143 na->opt[1] = na_olen >> 3;
2144
2145 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2146 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2147 csum_partial(na, sizeof(*na)+na_olen, 0));
2148
2149 pip6->payload_len = htons(sizeof(*na)+na_olen);
2150
2151 skb_push(reply, sizeof(struct ipv6hdr));
2152
2153 reply->ip_summed = CHECKSUM_UNNECESSARY;
2154
2155 return reply;
2156}
2157
3ad7a4b1 2158static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 2159{
2160 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 2161 const struct in6_addr *daddr;
8bff3685 2162 const struct ipv6hdr *iphdr;
4b29dba9 2163 struct inet6_dev *in6_dev;
8bff3685
XL		 2164 struct neighbour *n;
2165 struct nd_msg *msg;
f564f45c
CW		 2166
2167 in6_dev = __in6_dev_get(dev);
2168 if (!in6_dev)
2169 goto out;
2170
f564f45c 2171 iphdr = ipv6_hdr(skb);
f564f45c 2172 daddr = &iphdr->daddr;
f1fb08f6 2173 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 2174
4b29dba9
DS		 2175 if (ipv6_addr_loopback(daddr) ||
2176 ipv6_addr_is_multicast(&msg->target))
2177 goto out;
2178
2179 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 2180
2181 if (n) {
2182 struct vxlan_fdb *f;
4b29dba9 2183 struct sk_buff *reply;
f564f45c
CW		 2184
2185 if (!(n->nud_state & NUD_CONNECTED)) {
2186 neigh_release(n);
2187 goto out;
2188 }
2189
3ad7a4b1 2190 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 2191 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2192 /* bridge-local neighbor */
2193 neigh_release(n);
2194 goto out;
2195 }
2196
4b29dba9
DS		 2197 reply = vxlan_na_create(skb, n,
2198 !!(f ? f->flags & NTF_ROUTER : 0));
2199
f564f45c 2200 neigh_release(n);
4b29dba9
DS		 2201
2202 if (reply == NULL)
2203 goto out;
2204
2205 if (netif_rx_ni(reply) == NET_RX_DROP)
2206 dev->stats.rx_dropped++;
2207
dc5321d7 2208 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 2209 union vxlan_addr ipa = {
2210 .sin6.sin6_addr = msg->target,
a45e92a5 2211 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 2212 };
2213
f564f45c
CW		 2214 vxlan_ip_miss(dev, &ipa);
2215 }
2216
2217out:
2218 consume_skb(skb);
2219 return NETDEV_TX_OK;
2220}
2221#endif
2222
e4f67add
DS		 2223static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2224{
2225 struct vxlan_dev *vxlan = netdev_priv(dev);
2226 struct neighbour *n;
e4f67add
DS		 2227
2228 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2229 return false;
2230
2231 n = NULL;
2232 switch (ntohs(eth_hdr(skb)->h_proto)) {
2233 case ETH_P_IP:
e15a00aa
CW		 2234 {
2235 struct iphdr *pip;
2236
e4f67add
DS		 2237 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2238 return false;
2239 pip = ip_hdr(skb);
2240 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 2241 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 2242 union vxlan_addr ipa = {
2243 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 2244 .sin.sin_family = AF_INET,
e4c7ed41
CW		 2245 };
2246
2247 vxlan_ip_miss(dev, &ipa);
2248 return false;
2249 }
2250
e4f67add 2251 break;
e15a00aa
CW		 2252 }
2253#if IS_ENABLED(CONFIG_IPV6)
2254 case ETH_P_IPV6:
2255 {
2256 struct ipv6hdr *pip6;
2257
2258 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2259 return false;
2260 pip6 = ipv6_hdr(skb);
2261 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 2262 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 2263 union vxlan_addr ipa = {
2264 .sin6.sin6_addr = pip6->daddr,
a45e92a5 2265 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 2266 };
2267
2268 vxlan_ip_miss(dev, &ipa);
2269 return false;
2270 }
2271
2272 break;
2273 }
2274#endif
e4f67add
DS		 2275 default:
2276 return false;
2277 }
2278
2279 if (n) {
2280 bool diff;
2281
7367d0b5 2282 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 2283 if (diff) {
2284 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2285 dev->addr_len);
2286 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2287 }
2288 neigh_release(n);
2289 return diff;
e4c7ed41
CW		 2290 }
2291
e4f67add
DS		 2292 return false;
2293}
2294
af33c1ad 2295static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 2296 struct vxlan_metadata *md)
2297{
2298 struct vxlanhdr_gbp *gbp;
2299
db79a621
TG		 2300 if (!md->gbp)
2301 return;
2302
3511494c 2303 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 2304 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 2305
2306 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2307 gbp->dont_learn = 1;
2308
2309 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2310 gbp->policy_applied = 1;
2311
2312 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2313}
2314
e1e5314d
JB		 2315static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2316 __be16 protocol)
2317{
2318 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2319
2320 gpe->np_applied = 1;
fa20e0e3
JB		 2321 gpe->next_protocol = tun_p_from_eth_p(protocol);
2322 if (!gpe->next_protocol)
2323 return -EPFNOSUPPORT;
2324 return 0;
e1e5314d
JB		 2325}
2326
f491e56d
JB		 2327static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2328 int iphdr_len, __be32 vni,
2329 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 2330 bool udp_sum)
e4c7ed41 2331{
e4c7ed41 2332 struct vxlanhdr *vxh;
e4c7ed41
CW		 2333 int min_headroom;
2334 int err;
dfd8645e 2335 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 2336 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 2337
af33c1ad 2338 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 2339 skb->ip_summed == CHECKSUM_PARTIAL) {
2340 int csum_start = skb_checksum_start_offset(skb);
2341
2342 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2343 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2344 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 2345 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 2346 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 2347 }
e4c7ed41 2348
e4c7ed41 2349 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 2350 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 2351
2352 /* Need space for new headers (invalidates iph ptr) */
2353 err = skb_cow_head(skb, min_headroom);
e1e5314d 2354 if (unlikely(err))
c46b7897 2355 return err;
649c5b8b 2356
aed069df
AD		 2357 err = iptunnel_handle_offloads(skb, type);
2358 if (err)
c46b7897 2359 return err;
b736a623 2360
d58ff351 2361 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 2362 vxh->vx_flags = VXLAN_HF_VNI;
2363 vxh->vx_vni = vxlan_vni_field(vni);
49560532 2364
dfd8645e 2365 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 2366 unsigned int start;
dfd8645e 2367
54bfd872
JB		 2368 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2369 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2370 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 2371
2372 if (!skb_is_gso(skb)) {
2373 skb->ip_summed = CHECKSUM_NONE;
2374 skb->encapsulation = 0;
2375 }
2376 }
2377
af33c1ad
TH		 2378 if (vxflags & VXLAN_F_GBP)
2379 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 2380 if (vxflags & VXLAN_F_GPE) {
2381 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2382 if (err < 0)
c46b7897 2383 return err;
e1e5314d
JB		 2384 inner_protocol = skb->protocol;
2385 }
3511494c 2386
e1e5314d 2387 skb_set_inner_protocol(skb, inner_protocol);
039f5062 2388 return 0;
49560532 2389}
49560532 2390
655c3de1 2391static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2392 struct vxlan_sock *sock4,
1a8496ba 2393 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 2394 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 2395 struct dst_cache *dst_cache,
db3c6139 2396 const struct ip_tunnel_info *info)
1a8496ba 2397{
db3c6139 2398 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 2399 struct rtable *rt = NULL;
2400 struct flowi4 fl4;
2401
655c3de1 2402 if (!sock4)
2403 return ERR_PTR(-EIO);
2404
db3c6139
DB		 2405 if (tos && !info)
2406 use_cache = false;
2407 if (use_cache) {
0c1d70af
PA		 2408 rt = dst_cache_get_ip4(dst_cache, saddr);
2409 if (rt)
2410 return rt;
2411 }
2412
1a8496ba
JB		 2413 memset(&fl4, 0, sizeof(fl4));
2414 fl4.flowi4_oif = oif;
2415 fl4.flowi4_tos = RT_TOS(tos);
2416 fl4.flowi4_mark = skb->mark;
2417 fl4.flowi4_proto = IPPROTO_UDP;
2418 fl4.daddr = daddr;
272d96a5 2419 fl4.saddr = *saddr;
4ecb1d83
MP		 2420 fl4.fl4_dport = dport;
2421 fl4.fl4_sport = sport;
1a8496ba
JB		 2422
2423 rt = ip_route_output_key(vxlan->net, &fl4);
478db1f1 2424 if (!IS_ERR(rt)) {
655c3de1 2425 if (rt->dst.dev == dev) {
2426 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2427 ip_rt_put(rt);
2428 return ERR_PTR(-ELOOP);
2429 }
2430
1a8496ba 2431 *saddr = fl4.saddr;
0c1d70af
PA		 2432 if (use_cache)
2433 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 2434 } else {
2435 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2436 return ERR_PTR(-ENETUNREACH);
0c1d70af 2437 }
1a8496ba
JB		 2438 return rt;
2439}
2440
e5d4b29f
JB		 2441#if IS_ENABLED(CONFIG_IPV6)
2442static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 2443 struct net_device *dev,
03dc52a8 2444 struct vxlan_sock *sock6,
1400615d 2445 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 2446 __be32 label,
e5d4b29f 2447 const struct in6_addr *daddr,
0c1d70af 2448 struct in6_addr *saddr,
4ecb1d83 2449 __be16 dport, __be16 sport,
db3c6139
DB		 2450 struct dst_cache *dst_cache,
2451 const struct ip_tunnel_info *info)
e5d4b29f 2452{
db3c6139 2453 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 2454 struct dst_entry *ndst;
2455 struct flowi6 fl6;
e5d4b29f 2456
c6fcc4fc 2457 if (!sock6)
2458 return ERR_PTR(-EIO);
2459
1400615d
DB		 2460 if (tos && !info)
2461 use_cache = false;
db3c6139 2462 if (use_cache) {
0c1d70af
PA		 2463 ndst = dst_cache_get_ip6(dst_cache, saddr);
2464 if (ndst)
2465 return ndst;
2466 }
2467
e5d4b29f
JB		 2468 memset(&fl6, 0, sizeof(fl6));
2469 fl6.flowi6_oif = oif;
2470 fl6.daddr = *daddr;
272d96a5 2471 fl6.saddr = *saddr;
eaa93bf4 2472 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 2473 fl6.flowi6_mark = skb->mark;
2474 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 2475 fl6.fl6_dport = dport;
2476 fl6.fl6_sport = sport;
e5d4b29f 2477
6c8991f4
SD		 2478 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2479 &fl6, NULL);
2480 if (unlikely(IS_ERR(ndst))) {
655c3de1 2481 netdev_dbg(dev, "no route to %pI6\n", daddr);
2482 return ERR_PTR(-ENETUNREACH);
2483 }
2484
2485 if (unlikely(ndst->dev == dev)) {
2486 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2487 dst_release(ndst);
2488 return ERR_PTR(-ELOOP);
2489 }
e5d4b29f
JB		 2490
2491 *saddr = fl6.saddr;
db3c6139 2492 if (use_cache)
0c1d70af 2493 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 2494 return ndst;
2495}
2496#endif
2497
9dcc71e1
SS		 2498/* Bypass encapsulation if the destination is local */
2499static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
fc68c995
SB		 2500 struct vxlan_dev *dst_vxlan, __be32 vni,
2501 bool snoop)
9dcc71e1 2502{
8f84985f 2503 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2504 union vxlan_addr loopback;
2505 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
4179cb5a 2506 struct net_device *dev;
ce6502a8 2507 int len = skb->len;
9dcc71e1 2508
8f84985f
LR		 2509 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2510 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2511 skb->pkt_type = PACKET_HOST;
2512 skb->encapsulation = 0;
2513 skb->dev = dst_vxlan->dev;
2514 __skb_pull(skb, skb_network_offset(skb));
2515
e4c7ed41
CW		 2516 if (remote_ip->sa.sa_family == AF_INET) {
2517 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2518 loopback.sa.sa_family = AF_INET;
2519#if IS_ENABLED(CONFIG_IPV6)
2520 } else {
2521 loopback.sin6.sin6_addr = in6addr_loopback;
2522 loopback.sa.sa_family = AF_INET6;
2523#endif
2524 }
2525
4179cb5a
ED		 2526 rcu_read_lock();
2527 dev = skb->dev;
2528 if (unlikely(!(dev->flags & IFF_UP))) {
2529 kfree_skb(skb);
2530 goto drop;
2531 }
2532
fc68c995 2533 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
4179cb5a 2534 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
9dcc71e1
SS		 2535
2536 u64_stats_update_begin(&tx_stats->syncp);
2537 tx_stats->tx_packets++;
ce6502a8 2538 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2539 u64_stats_update_end(&tx_stats->syncp);
2540
2541 if (netif_rx(skb) == NET_RX_SUCCESS) {
2542 u64_stats_update_begin(&rx_stats->syncp);
2543 rx_stats->rx_packets++;
ce6502a8 2544 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2545 u64_stats_update_end(&rx_stats->syncp);
2546 } else {
4179cb5a 2547drop:
ce6502a8 2548 dev->stats.rx_dropped++;
9dcc71e1 2549 }
4179cb5a 2550 rcu_read_unlock();
9dcc71e1
SS		 2551}
2552
fee1fad7 2553static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2554 struct vxlan_dev *vxlan,
2555 union vxlan_addr *daddr,
2556 __be16 dst_port, int dst_ifindex, __be32 vni,
2557 struct dst_entry *dst,
fee1fad7 2558 u32 rt_flags)
2559{
2560#if IS_ENABLED(CONFIG_IPV6)
2561 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2562 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2563 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2564 */
2565 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2566#endif
2567 /* Bypass encapsulation if the destination is local */
2568 if (rt_flags & RTCF_LOCAL &&
2569 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2570 struct vxlan_dev *dst_vxlan;
2571
2572 dst_release(dst);
49f810f0 2573 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2574 daddr->sa.sa_family, dst_port,
dc5321d7 2575 vxlan->cfg.flags);
fee1fad7 2576 if (!dst_vxlan) {
2577 dev->stats.tx_errors++;
2578 kfree_skb(skb);
2579
2580 return -ENOENT;
2581 }
fc68c995 2582 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
fee1fad7 2583 return 1;
2584 }
2585
2586 return 0;
2587}
2588
4ad16930 2589static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2590 __be32 default_vni, struct vxlan_rdst *rdst,
2591 bool did_rsc)
d342894c 2592{
d71785ff 2593 struct dst_cache *dst_cache;
3093fbe7 2594 struct ip_tunnel_info *info;
d342894c 2595 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2596 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2597 union vxlan_addr *dst;
272d96a5 2598 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2599 struct vxlan_metadata _md;
2600 struct vxlan_metadata *md = &_md;
e4c7ed41 2601 __be16 src_port = 0, dst_port;
655c3de1 2602 struct dst_entry *ndst = NULL;
e7f70af1 2603 __be32 vni, label;
d342894c 2604 __u8 tos, ttl;
49f810f0 2605 int ifindex;
0e6fbc5b 2606 int err;
dc5321d7 2607 u32 flags = vxlan->cfg.flags;
b4ed5cad 2608 bool udp_sum = false;
f491e56d 2609 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2610
61adedf3 2611 info = skb_tunnel_info(skb);
3093fbe7 2612
ee122c79 2613 if (rdst) {
0770b53b 2614 dst = &rdst->remote_ip;
2615 if (vxlan_addr_any(dst)) {
2616 if (did_rsc) {
2617 /* short-circuited back to local bridge */
fc68c995
SB		 2618 vxlan_encap_bypass(skb, vxlan, vxlan,
2619 default_vni, true);
0770b53b 2620 return;
2621 }
2622 goto drop;
2623 }
2624
0dfbdf41 2625 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2626 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2627 ifindex = rdst->remote_ifindex;
1158632b 2628 local_ip = vxlan->cfg.saddr;
d71785ff 2629 dst_cache = &rdst->dst_cache;
0770b53b 2630 md->gbp = skb->mark;
72f6d71e
HL		 2631 if (flags & VXLAN_F_TTL_INHERIT) {
2632 ttl = ip_tunnel_get_ttl(old_iph, skb);
2633 } else {
2634 ttl = vxlan->cfg.ttl;
2635 if (!ttl && vxlan_addr_multicast(dst))
2636 ttl = 1;
2637 }
0770b53b 2638
2639 tos = vxlan->cfg.tos;
2640 if (tos == 1)
2641 tos = ip_tunnel_get_dsfield(old_iph, skb);
2642
2643 if (dst->sa.sa_family == AF_INET)
2644 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2645 else
2646 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2647 label = vxlan->cfg.label;
ee122c79 2648 } else {
435be28b
JK		 2649 if (!info) {
2650 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2651 dev->name);
2652 goto drop;
2653 }
b1be00a6 2654 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2655 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2656 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2657 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2658 } else {
a725e514 2659 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2660 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2661 }
ee122c79 2662 dst = &remote_ip;
0770b53b 2663 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2664 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2665 ifindex = 0;
d71785ff 2666 dst_cache = &info->dst_cache;
eadf52cf
XL		 2667 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2668 if (info->options_len < sizeof(*md))
2669 goto drop;
0770b53b 2670 md = ip_tunnel_info_opts(info);
eadf52cf 2671 }
a725e514
JB		 2672 ttl = info->key.ttl;
2673 tos = info->key.tos;
e7f70af1 2674 label = info->key.label;
b4ed5cad 2675 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2676 }
0770b53b 2677 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2678 vxlan->cfg.port_max, true);
a725e514 2679
56de859e 2680 rcu_read_lock();
e4c7ed41 2681 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2682 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2683 struct rtable *rt;
0770b53b 2684 __be16 df = 0;
c6fcc4fc 2685
aab8cc36
AB		 2686 if (!ifindex)
2687 ifindex = sock4->sock->sk->sk_bound_dev_if;
2688
49f810f0 2689 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2690 dst->sin.sin_addr.s_addr,
1158632b 2691 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2692 dst_port, src_port,
d71785ff 2693 dst_cache, info);
8ebd115b
DM		 2694 if (IS_ERR(rt)) {
2695 err = PTR_ERR(rt);
c46b7897 2696 goto tx_error;
8ebd115b 2697 }
e4c7ed41 2698
fee1fad7 2699 if (!info) {
b4d30697 2700 /* Bypass encapsulation if the destination is local */
fee1fad7 2701 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2702 dst_port, ifindex, vni,
2703 &rt->dst, rt->rt_flags);
fee1fad7 2704 if (err)
56de859e 2705 goto out_unlock;
b4d30697
SB		 2706
2707 if (vxlan->cfg.df == VXLAN_DF_SET) {
2708 df = htons(IP_DF);
2709 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2710 struct ethhdr *eth = eth_hdr(skb);
2711
2712 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2713 (ntohs(eth->h_proto) == ETH_P_IP &&
2714 old_iph->frag_off & htons(IP_DF)))
2715 df = htons(IP_DF);
2716 }
fee1fad7 2717 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2718 df = htons(IP_DF);
fee1fad7 2719 }
6ceb31ca 2720
c46b7897 2721 ndst = &rt->dst;
fc68c995
SB		 2722 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2723 netif_is_any_bridge_port(dev));
2724 if (err < 0) {
2725 goto tx_error;
2726 } else if (err) {
2727 if (info) {
2728 struct in_addr src, dst;
2729
2730 src = remote_ip.sin.sin_addr;
2731 dst = local_ip.sin.sin_addr;
2732 info->key.u.ipv4.src = src.s_addr;
2733 info->key.u.ipv4.dst = dst.s_addr;
2734 }
2735 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2736 dst_release(ndst);
2737 goto out_unlock;
2738 }
a93bf0ff 2739
a0dced17 2740 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2741 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2742 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2743 vni, md, flags, udp_sum);
f491e56d 2744 if (err < 0)
c46b7897 2745 goto tx_error;
f491e56d 2746
1158632b 2747 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2748 dst->sin.sin_addr.s_addr, tos, ttl, df,
2749 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2750#if IS_ENABLED(CONFIG_IPV6)
2751 } else {
c6fcc4fc 2752 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2753
aab8cc36
AB		 2754 if (!ifindex)
2755 ifindex = sock6->sock->sk->sk_bound_dev_if;
2756
49f810f0 2757 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2758 label, &dst->sin6.sin6_addr,
1158632b 2759 &local_ip.sin6.sin6_addr,
4ecb1d83 2760 dst_port, src_port,
db3c6139 2761 dst_cache, info);
e5d4b29f 2762 if (IS_ERR(ndst)) {
8ebd115b 2763 err = PTR_ERR(ndst);
c46b7897 2764 ndst = NULL;
9dcc71e1 2765 goto tx_error;
e4c7ed41 2766 }
655c3de1 2767
fee1fad7 2768 if (!info) {
2769 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2770
fee1fad7 2771 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2772 dst_port, ifindex, vni,
2773 ndst, rt6i_flags);
fee1fad7 2774 if (err)
56de859e 2775 goto out_unlock;
fee1fad7 2776 }
35e2d115 2777
fc68c995
SB		 2778 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2779 netif_is_any_bridge_port(dev));
2780 if (err < 0) {
2781 goto tx_error;
2782 } else if (err) {
2783 if (info) {
2784 struct in6_addr src, dst;
2785
2786 src = remote_ip.sin6.sin6_addr;
2787 dst = local_ip.sin6.sin6_addr;
2788 info->key.u.ipv6.src = src;
2789 info->key.u.ipv6.dst = dst;
2790 }
2791
2792 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2793 dst_release(ndst);
2794 goto out_unlock;
2795 }
a93bf0ff 2796
a0dced17 2797 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2798 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2799 skb_scrub_packet(skb, xnet);
2800 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2801 vni, md, flags, udp_sum);
c46b7897 2802 if (err < 0)
2803 goto tx_error;
2804
0770b53b 2805 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2806 &local_ip.sin6.sin6_addr,
272d96a5 2807 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2808 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2809#endif
2810 }
56de859e
JK		 2811out_unlock:
2812 rcu_read_unlock();
4ad16930 2813 return;
d342894c 2814
2815drop:
2816 dev->stats.tx_dropped++;
c46b7897 2817 dev_kfree_skb(skb);
2818 return;
d342894c 2819
2820tx_error:
56de859e 2821 rcu_read_unlock();
655c3de1 2822 if (err == -ELOOP)
2823 dev->stats.collisions++;
2824 else if (err == -ENETUNREACH)
2825 dev->stats.tx_carrier_errors++;
c46b7897 2826 dst_release(ndst);
d342894c 2827 dev->stats.tx_errors++;
c46b7897 2828 kfree_skb(skb);
d342894c 2829}
2830
1274e1cc
RP		 2831static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2832 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2833{
2834 struct vxlan_rdst nh_rdst;
2835 struct nexthop *nh;
2836 bool do_xmit;
2837 u32 hash;
2838
2839 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2840 hash = skb_get_hash(skb);
2841
2842 rcu_read_lock();
2843 nh = rcu_dereference(f->nh);
2844 if (!nh) {
2845 rcu_read_unlock();
2846 goto drop;
2847 }
2848 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2849 rcu_read_unlock();
2850
2851 if (likely(do_xmit))
2852 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2853 else
2854 goto drop;
2855
2856 return;
2857
2858drop:
2859 dev->stats.tx_dropped++;
2860 dev_kfree_skb(skb);
2861}
2862
6681712d
DS		 2863/* Transmit local packets over Vxlan
2864 *
2865 * Outer IP header inherits ECN and DF from inner header.
2866 * Outer UDP destination is the VXLAN assigned port.
2867 * source port is based on hash of flow
2868 */
2869static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2870{
2871 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2872 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2873 const struct ip_tunnel_info *info;
6681712d 2874 bool did_rsc = false;
6681712d 2875 struct vxlan_fdb *f;
8bff3685 2876 struct ethhdr *eth;
3ad7a4b1 2877 __be32 vni = 0;
6681712d 2878
61adedf3 2879 info = skb_tunnel_info(skb);
3093fbe7 2880
6681712d 2881 skb_reset_mac_header(skb);
6681712d 2882
dc5321d7 2883 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2884 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2885 info->mode & IP_TUNNEL_INFO_TX) {
2886 vni = tunnel_id_to_key32(info->key.tun_id);
2887 } else {
2888 if (info && info->mode & IP_TUNNEL_INFO_TX)
2889 vxlan_xmit_one(skb, dev, vni, NULL, false);
2890 else
2891 kfree_skb(skb);
2892 return NETDEV_TX_OK;
2893 }
47e5d1b0
JB		 2894 }
2895
dc5321d7 2896 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2897 eth = eth_hdr(skb);
f564f45c 2898 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2899 return arp_reduce(dev, skb, vni);
f564f45c 2900#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 2901 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2902 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2903 sizeof(struct nd_msg)) &&
2904 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2905 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2906
2907 if (m->icmph.icmp6_code == 0 &&
2908 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2909 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2910 }
2911#endif
2912 }
6681712d 2913
47e5d1b0 2914 eth = eth_hdr(skb);
3ad7a4b1 2915 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2916 did_rsc = false;
2917
dc5321d7 2918 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2919 (ntohs(eth->h_proto) == ETH_P_IP ||
2920 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2921 did_rsc = route_shortcircuit(dev, skb);
2922 if (did_rsc)
3ad7a4b1 2923 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2924 }
2925
6681712d 2926 if (f == NULL) {
3ad7a4b1 2927 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2928 if (f == NULL) {
dc5321d7 2929 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2930 !is_multicast_ether_addr(eth->h_dest))
2931 vxlan_fdb_miss(vxlan, eth->h_dest);
2932
2933 dev->stats.tx_dropped++;
8f646c92 2934 kfree_skb(skb);
afbd8bae
MR		 2935 return NETDEV_TX_OK;
2936 }
2937 }
6681712d 2938
1274e1cc
RP		 2939 if (rcu_access_pointer(f->nh)) {
2940 vxlan_xmit_nh(skb, dev, f,
2941 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2942 } else {
2943 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2944 struct sk_buff *skb1;
6681712d 2945
1274e1cc
RP		 2946 if (!fdst) {
2947 fdst = rdst;
2948 continue;
2949 }
2950 skb1 = skb_clone(skb, GFP_ATOMIC);
2951 if (skb1)
2952 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
8f646c92 2953 }
1274e1cc
RP		 2954 if (fdst)
2955 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2956 else
2957 kfree_skb(skb);
6681712d
DS		 2958 }
2959
4ad16930 2960 return NETDEV_TX_OK;
6681712d
DS		 2961}
2962
d342894c 2963/* Walk the forwarding table and purge stale entries */
df7e828c 2964static void vxlan_cleanup(struct timer_list *t)
d342894c 2965{
df7e828c 2966 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2967 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2968 unsigned int h;
2969
2970 if (!netif_running(vxlan->dev))
2971 return;
2972
d342894c 2973 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2974 struct hlist_node *p, *n;
14e1d0fa 2975
fe1e0713 2976 spin_lock(&vxlan->hash_lock[h]);
d342894c 2977 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2978 struct vxlan_fdb *f
2979 = container_of(p, struct vxlan_fdb, hlist);
2980 unsigned long timeout;
2981
efb5f68f 2982 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2983 continue;
2984
def499c9
RP		 2985 if (f->flags & NTF_EXT_LEARNED)
2986 continue;
2987
0dfbdf41 2988 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2989 if (time_before_eq(timeout, jiffies)) {
2990 netdev_dbg(vxlan->dev,
2991 "garbage collect %pM\n",
2992 f->eth_addr);
2993 f->state = NUD_STALE;
0e6160f3 2994 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 2995 } else if (time_before(timeout, next_timer))
2996 next_timer = timeout;
2997 }
fe1e0713 2998 spin_unlock(&vxlan->hash_lock[h]);
d342894c 2999 }
d342894c 3000
3001 mod_timer(&vxlan->age_timer, next_timer);
3002}
3003
a53cb29b
MB		 3004static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3005{
3006 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3007
3008 spin_lock(&vn->sock_lock);
69e76661
JB		 3009 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3010#if IS_ENABLED(CONFIG_IPV6)
3011 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3012#endif
a53cb29b
MB		 3013 spin_unlock(&vn->sock_lock);
3014}
3015
69e76661
JB		 3016static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3017 struct vxlan_dev_node *node)
9c2e24e1 3018{
56ef9c90 3019 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 3020 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 3021
69e76661 3022 node->vxlan = vxlan;
56ef9c90 3023 spin_lock(&vn->sock_lock);
69e76661 3024 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 3025 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 3026}
3027
d342894c 3028/* Setup stats when device is created */
3029static int vxlan_init(struct net_device *dev)
3030{
384d91c2
TY		 3031 struct vxlan_dev *vxlan = netdev_priv(dev);
3032 int err;
3033
1c213bd2 3034 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 3035 if (!dev->tstats)
d342894c 3036 return -ENOMEM;
3037
384d91c2
TY		 3038 err = gro_cells_init(&vxlan->gro_cells, dev);
3039 if (err) {
3040 free_percpu(dev->tstats);
3041 return err;
3042 }
3043
d342894c 3044 return 0;
3045}
3046
3ad7a4b1 3047static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 3048{
3049 struct vxlan_fdb *f;
fe1e0713 3050 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
afbd8bae 3051
fe1e0713 3052 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3ad7a4b1 3053 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 3054 if (f)
0e6160f3 3055 vxlan_fdb_destroy(vxlan, f, true, true);
fe1e0713 3056 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
afbd8bae
MR		 3057}
3058
ebf4063e
SH		 3059static void vxlan_uninit(struct net_device *dev)
3060{
3061 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 3062
ad6c9986
SB		 3063 gro_cells_destroy(&vxlan->gro_cells);
3064
3ad7a4b1 3065 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 3066
ebf4063e
SH		 3067 free_percpu(dev->tstats);
3068}
3069
d342894c 3070/* Start ageing timer and join group when device is brought up */
3071static int vxlan_open(struct net_device *dev)
3072{
3073 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 3074 int ret;
56ef9c90 3075
205f356d
JB		 3076 ret = vxlan_sock_add(vxlan);
3077 if (ret < 0)
3078 return ret;
d342894c 3079
79d4a94f 3080 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 3081 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 3082 if (ret == -EADDRINUSE)
3083 ret = 0;
56ef9c90 3084 if (ret) {
205f356d 3085 vxlan_sock_release(vxlan);
56ef9c90
MRL		 3086 return ret;
3087 }
d342894c 3088 }
3089
0dfbdf41 3090 if (vxlan->cfg.age_interval)
d342894c 3091 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3092
56ef9c90 3093 return ret;
d342894c 3094}
3095
3096/* Purge the forwarding table */
8b3f9337 3097static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 3098{
31fec5aa 3099 unsigned int h;
d342894c 3100
d342894c 3101 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3102 struct hlist_node *p, *n;
fe1e0713
L		 3103
3104 spin_lock_bh(&vxlan->hash_lock[h]);
d342894c 3105 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3106 struct vxlan_fdb *f
3107 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 3108 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3109 continue;
afbd8bae 3110 /* the all_zeros_mac entry is deleted at vxlan_uninit */
fda2ec62
TY		 3111 if (is_zero_ether_addr(f->eth_addr) &&
3112 f->vni == vxlan->cfg.vni)
3113 continue;
3114 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 3115 }
fe1e0713 3116 spin_unlock_bh(&vxlan->hash_lock[h]);
d342894c 3117 }
d342894c 3118}
3119
3120/* Cleanup timer and forwarding table on shutdown */
3121static int vxlan_stop(struct net_device *dev)
3122{
3123 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3124 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 3125 int ret = 0;
d342894c 3126
24c0e683 3127 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 3128 !vxlan_group_used(vn, vxlan))
56ef9c90 3129 ret = vxlan_igmp_leave(vxlan);
d342894c 3130
3131 del_timer_sync(&vxlan->age_timer);
3132
8b3f9337 3133 vxlan_flush(vxlan, false);
205f356d 3134 vxlan_sock_release(vxlan);
d342894c 3135
56ef9c90 3136 return ret;
d342894c 3137}
3138
d342894c 3139/* Stub, nothing needs to be done. */
3140static void vxlan_set_multicast_list(struct net_device *dev)
3141{
3142}
3143
91572088 3144static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 3145{
91572088
JW		 3146 struct vxlan_dev *vxlan = netdev_priv(dev);
3147 struct vxlan_rdst *dst = &vxlan->default_dst;
3148 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3149 dst->remote_ifindex);
ce44a4ae 3150 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 3151
91572088
JW		 3152 /* This check is different than dev->max_mtu, because it looks at
3153 * the lowerdev->mtu, rather than the static dev->max_mtu
3154 */
3155 if (lowerdev) {
3156 int max_mtu = lowerdev->mtu -
3157 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3158 if (new_mtu > max_mtu)
72564b59 3159 return -EINVAL;
72564b59
DW		 3160 }
3161
345010b5
DB		 3162 dev->mtu = new_mtu;
3163 return 0;
3164}
3165
fc4099f1
PS		 3166static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3167{
3168 struct vxlan_dev *vxlan = netdev_priv(dev);
3169 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3170 __be16 sport, dport;
3171
3172 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3173 vxlan->cfg.port_max, true);
3174 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3175
239e944f 3176 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 3177 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 3178 struct rtable *rt;
3179
655c3de1 3180 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 3181 info->key.u.ipv4.dst,
22f0708a
PA		 3182 &info->key.u.ipv4.src, dport, sport,
3183 &info->dst_cache, info);
1a8496ba
JB		 3184 if (IS_ERR(rt))
3185 return PTR_ERR(rt);
3186 ip_rt_put(rt);
239e944f
JB		 3187 } else {
3188#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 3189 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 3190 struct dst_entry *ndst;
3191
655c3de1 3192 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 3193 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 3194 &info->key.u.ipv6.src, dport, sport,
3195 &info->dst_cache, info);
239e944f
JB		 3196 if (IS_ERR(ndst))
3197 return PTR_ERR(ndst);
3198 dst_release(ndst);
239e944f
JB		 3199#else /* !CONFIG_IPV6 */
3200 return -EPFNOSUPPORT;
3201#endif
3202 }
1a8496ba
JB		 3203 info->key.tp_src = sport;
3204 info->key.tp_dst = dport;
239e944f 3205 return 0;
fc4099f1
PS		 3206}
3207
0c867c9b 3208static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 3209 .ndo_init = vxlan_init,
ebf4063e 3210 .ndo_uninit = vxlan_uninit,
d342894c 3211 .ndo_open = vxlan_open,
3212 .ndo_stop = vxlan_stop,
3213 .ndo_start_xmit = vxlan_xmit,
b220a4a7 3214 .ndo_get_stats64 = dev_get_tstats64,
d342894c 3215 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 3216 .ndo_change_mtu = vxlan_change_mtu,
d342894c 3217 .ndo_validate_addr = eth_validate_addr,
3218 .ndo_set_mac_address = eth_mac_addr,
3219 .ndo_fdb_add = vxlan_fdb_add,
3220 .ndo_fdb_del = vxlan_fdb_delete,
3221 .ndo_fdb_dump = vxlan_fdb_dump,
474c3c89 3222 .ndo_fdb_get = vxlan_fdb_get,
fc4099f1 3223 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
8f1af75d 3224 .ndo_change_proto_down = dev_change_proto_down_generic,
d342894c 3225};
3226
e1e5314d
JB		 3227static const struct net_device_ops vxlan_netdev_raw_ops = {
3228 .ndo_init = vxlan_init,
3229 .ndo_uninit = vxlan_uninit,
3230 .ndo_open = vxlan_open,
3231 .ndo_stop = vxlan_stop,
3232 .ndo_start_xmit = vxlan_xmit,
b220a4a7 3233 .ndo_get_stats64 = dev_get_tstats64,
e1e5314d
JB		 3234 .ndo_change_mtu = vxlan_change_mtu,
3235 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3236};
3237
d342894c 3238/* Info for udev, that this is a virtual tunnel endpoint */
3239static struct device_type vxlan_type = {
3240 .name = "vxlan",
3241};
3242
e5de25dc 3243/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 3244 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 3245 * to implement the ndo_udp_tunnel_add.
53cf5275 3246 */
2d2b13fc 3247static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 3248{
3249 struct vxlan_sock *vs;
3250 struct net *net = dev_net(dev);
3251 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 3252 unsigned int i;
53cf5275
JG		 3253
3254 spin_lock(&vn->sock_lock);
3255 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 3256 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3257 unsigned short type;
3258
3259 if (vs->flags & VXLAN_F_GPE)
3260 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3261 else
3262 type = UDP_TUNNEL_TYPE_VXLAN;
3263
3264 if (push)
3265 udp_tunnel_push_rx_port(dev, vs->sock, type);
3266 else
3267 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3268 }
53cf5275
JG		 3269 }
3270 spin_unlock(&vn->sock_lock);
3271}
53cf5275 3272
d342894c 3273/* Initialize the device structure. */
3274static void vxlan_setup(struct net_device *dev)
3275{
3276 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 3277 unsigned int h;
d342894c 3278
65226ef8
JB		 3279 eth_hw_addr_random(dev);
3280 ether_setup(dev);
3281
cf124db5 3282 dev->needs_free_netdev = true;
d342894c 3283 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3284
d342894c 3285 dev->features |= NETIF_F_LLTX;
d6727fe3 3286 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 3287 dev->features |= NETIF_F_RXCSUM;
05c0db08 3288 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 3289
1eaa8178 3290 dev->vlan_features = dev->features;
0afb1666 3291 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 3292 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 3293 netif_keep_dst(dev);
0c867c9b 3294 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 3295
a985343b
MS		 3296 /* MTU range: 68 - 65535 */
3297 dev->min_mtu = ETH_MIN_MTU;
3298 dev->max_mtu = ETH_MAX_MTU;
3299
553675fb 3300 INIT_LIST_HEAD(&vxlan->next);
d342894c 3301
df7e828c 3302 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 3303
3304 vxlan->dev = dev;
3305
fe1e0713
L		 3306 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3307 spin_lock_init(&vxlan->hash_lock[h]);
d342894c 3308 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
fe1e0713 3309 }
d342894c 3310}
3311
0c867c9b
JB		 3312static void vxlan_ether_setup(struct net_device *dev)
3313{
0c867c9b
JB		 3314 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3315 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3316 dev->netdev_ops = &vxlan_netdev_ether_ops;
3317}
3318
e1e5314d
JB		 3319static void vxlan_raw_setup(struct net_device *dev)
3320{
65226ef8 3321 dev->header_ops = NULL;
e1e5314d
JB		 3322 dev->type = ARPHRD_NONE;
3323 dev->hard_header_len = 0;
3324 dev->addr_len = 0;
e1e5314d
JB		 3325 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3326 dev->netdev_ops = &vxlan_netdev_raw_ops;
3327}
3328
d342894c 3329static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3330 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
c593642c 3331 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
e4c7ed41 3332 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 3333 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
c593642c 3334 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
e4c7ed41 3335 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 3336 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3337 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 3338 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 3339 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3340 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3341 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 3342 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 3343 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3344 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3345 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3346 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 3347 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 3348 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 3349 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3350 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3351 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 3352 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3353 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 3354 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 3355 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 3356 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
72f6d71e 3357 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
b4d30697 3358 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
d342894c 3359};
3360
a8b8a889
MS		 3361static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3362 struct netlink_ext_ack *extack)
d342894c 3363{
3364 if (tb[IFLA_ADDRESS]) {
3365 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 3366 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3367 "Provided link layer address is not Ethernet");
d342894c 3368 return -EINVAL;
3369 }
3370
3371 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 3372 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3373 "Provided Ethernet address is not unicast");
d342894c 3374 return -EADDRNOTAVAIL;
3375 }
3376 }
3377
a985343b 3378 if (tb[IFLA_MTU]) {
019b13ae 3379 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 3380
653ef6a3
GM		 3381 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3382 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3383 "MTU must be between 68 and 65535");
a985343b 3384 return -EINVAL;
653ef6a3 3385 }
a985343b
MS		 3386 }
3387
653ef6a3
GM		 3388 if (!data) {
3389 NL_SET_ERR_MSG(extack,
3390 "Required attributes not provided to perform the operation");
d342894c 3391 return -EINVAL;
653ef6a3 3392 }
d342894c 3393
3394 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 3395 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3396
653ef6a3 3397 if (id >= VXLAN_N_VID) {
cc8e7c69 3398 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
653ef6a3 3399 "VXLAN ID must be lower than 16777216");
d342894c 3400 return -ERANGE;
653ef6a3 3401 }
d342894c 3402 }
3403
05f47d69 3404 if (data[IFLA_VXLAN_PORT_RANGE]) {
3405 const struct ifla_vxlan_port_range *p
3406 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3407
3408 if (ntohs(p->high) < ntohs(p->low)) {
cc8e7c69 3409 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
653ef6a3 3410 "Invalid source port range");
05f47d69 3411 return -EINVAL;
3412 }
3413 }
3414
b4d30697
SB		 3415 if (data[IFLA_VXLAN_DF]) {
3416 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3417
3418 if (df < 0 || df > VXLAN_DF_MAX) {
cc8e7c69 3419 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
b4d30697
SB		 3420 "Invalid DF attribute");
3421 return -EINVAL;
3422 }
3423 }
3424
d342894c 3425 return 0;
3426}
3427
1b13c97f
YB		 3428static void vxlan_get_drvinfo(struct net_device *netdev,
3429 struct ethtool_drvinfo *drvinfo)
3430{
3431 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3432 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3433}
3434
36fe3a61
MS		 3435static int vxlan_get_link_ksettings(struct net_device *dev,
3436 struct ethtool_link_ksettings *cmd)
3437{
3438 struct vxlan_dev *vxlan = netdev_priv(dev);
3439 struct vxlan_rdst *dst = &vxlan->default_dst;
3440 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3441 dst->remote_ifindex);
3442
3443 if (!lowerdev) {
3444 cmd->base.duplex = DUPLEX_UNKNOWN;
3445 cmd->base.port = PORT_OTHER;
3446 cmd->base.speed = SPEED_UNKNOWN;
3447
3448 return 0;
3449 }
3450
3451 return __ethtool_get_link_ksettings(lowerdev, cmd);
3452}
3453
1b13c97f 3454static const struct ethtool_ops vxlan_ethtool_ops = {
36fe3a61
MS		 3455 .get_drvinfo = vxlan_get_drvinfo,
3456 .get_link = ethtool_op_get_link,
3457 .get_link_ksettings = vxlan_get_link_ksettings,
1b13c97f
YB		 3458};
3459
3ee64f39 3460static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
aab8cc36 3461 __be16 port, u32 flags, int ifindex)
553675fb 3462{
e4c7ed41 3463 struct socket *sock;
3ee64f39
TH		 3464 struct udp_port_cfg udp_conf;
3465 int err;
e4c7ed41 3466
3ee64f39 3467 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 3468
3ee64f39
TH		 3469 if (ipv6) {
3470 udp_conf.family = AF_INET6;
3ee64f39 3471 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 3472 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 3473 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 3474 } else {
3475 udp_conf.family = AF_INET;
e4c7ed41 3476 }
e4c7ed41 3477
3ee64f39 3478 udp_conf.local_udp_port = port;
aab8cc36 3479 udp_conf.bind_ifindex = ifindex;
553675fb 3480
3ee64f39
TH		 3481 /* Open UDP socket */
3482 err = udp_sock_create(net, &udp_conf, &sock);
3483 if (err < 0)
3484 return ERR_PTR(err);
e4c7ed41 3485
39deb2c7 3486 return sock;
e4c7ed41
CW		 3487}
3488
3489/* Create new listen socket if needed */
b1be00a6 3490static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
aab8cc36
AB		 3491 __be16 port, u32 flags,
3492 int ifindex)
e4c7ed41
CW		 3493{
3494 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3495 struct vxlan_sock *vs;
3496 struct socket *sock;
e4c7ed41 3497 unsigned int h;
acbf74a7 3498 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 3499
dc01e7d3 3500 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 3501 if (!vs)
3502 return ERR_PTR(-ENOMEM);
3503
3504 for (h = 0; h < VNI_HASH_SIZE; ++h)
3505 INIT_HLIST_HEAD(&vs->vni_list[h]);
3506
aab8cc36 3507 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
39deb2c7 3508 if (IS_ERR(sock)) {
553675fb 3509 kfree(vs);
e50fddc8 3510 return ERR_CAST(sock);
553675fb 3511 }
e4c7ed41
CW		 3512
3513 vs->sock = sock;
66af846f 3514 refcount_set(&vs->refcnt, 1);
af33c1ad 3515 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 3516
9c2e24e1
PS		 3517 spin_lock(&vn->sock_lock);
3518 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 3519 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 3520 (vs->flags & VXLAN_F_GPE) ?
3521 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 3522 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 3523 spin_unlock(&vn->sock_lock);
553675fb 3524
3525 /* Mark socket as an encapsulation socket. */
5602c48c 3526 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 3527 tunnel_cfg.sk_user_data = vs;
3528 tunnel_cfg.encap_type = 1;
f2d1968e 3529 tunnel_cfg.encap_rcv = vxlan_rcv;
c3a43b9f 3530 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
acbf74a7 3531 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 3532 tunnel_cfg.gro_receive = vxlan_gro_receive;
3533 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 3534
3535 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 3536
9c2e24e1
PS		 3537 return vs;
3538}
3539
b1be00a6 3540static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 3541{
205f356d
JB		 3542 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3543 struct vxlan_sock *vs = NULL;
69e76661 3544 struct vxlan_dev_node *node;
aab8cc36
AB		 3545 int l3mdev_index = 0;
3546
3547 if (vxlan->cfg.remote_ifindex)
3548 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3549 vxlan->net, vxlan->cfg.remote_ifindex);
9c2e24e1 3550
205f356d 3551 if (!vxlan->cfg.no_share) {
56ef9c90 3552 spin_lock(&vn->sock_lock);
205f356d 3553 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
aab8cc36
AB		 3554 vxlan->cfg.dst_port, vxlan->cfg.flags,
3555 l3mdev_index);
66af846f 3556 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 3557 spin_unlock(&vn->sock_lock);
205f356d 3558 return -EBUSY;
56ef9c90
MRL		 3559 }
3560 spin_unlock(&vn->sock_lock);
3561 }
205f356d 3562 if (!vs)
b1be00a6 3563 vs = vxlan_socket_create(vxlan->net, ipv6,
aab8cc36
AB		 3564 vxlan->cfg.dst_port, vxlan->cfg.flags,
3565 l3mdev_index);
205f356d
JB		 3566 if (IS_ERR(vs))
3567 return PTR_ERR(vs);
b1be00a6 3568#if IS_ENABLED(CONFIG_IPV6)
69e76661 3569 if (ipv6) {
c6fcc4fc 3570 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 3571 node = &vxlan->hlist6;
3572 } else
b1be00a6 3573#endif
69e76661 3574 {
c6fcc4fc 3575 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 3576 node = &vxlan->hlist4;
3577 }
3578 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 3579 return 0;
553675fb 3580}
3581
b1be00a6
JB		 3582static int vxlan_sock_add(struct vxlan_dev *vxlan)
3583{
dc5321d7
MS		 3584 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3585 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 3586 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 3587 int ret = 0;
3588
c6fcc4fc 3589 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 3590#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 3591 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 3592 if (ipv6) {
b1be00a6 3593 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 3594 if (ret < 0 && ret != -EAFNOSUPPORT)
3595 ipv4 = false;
3596 }
b1be00a6 3597#endif
d074bf96 3598 if (ipv4)
b1be00a6
JB		 3599 ret = __vxlan_sock_add(vxlan, false);
3600 if (ret < 0)
3601 vxlan_sock_release(vxlan);
3602 return ret;
3603}
3604
a985343b
MS		 3605static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3606 struct net_device **lower,
653ef6a3
GM		 3607 struct vxlan_dev *old,
3608 struct netlink_ext_ack *extack)
d342894c 3609{
33564bbb 3610 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 3611 struct vxlan_dev *tmp;
e4c7ed41 3612 bool use_ipv6 = false;
d342894c 3613
a985343b
MS		 3614 if (conf->flags & VXLAN_F_GPE) {
3615 /* For now, allow GPE only together with
3616 * COLLECT_METADATA. This can be relaxed later; in such
3617 * case, the other side of the PtP link will have to be
3618 * provided.
3619 */
3620 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3621 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 3622 NL_SET_ERR_MSG(extack,
3623 "VXLAN GPE does not support this combination of attributes");
a985343b 3624 return -EINVAL;
3555621d 3625 }
e1e5314d 3626 }
0c867c9b 3627
ce44a4ae
MS		 3628 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3629 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 3630 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 3631 conf->saddr.sa.sa_family = AF_INET;
3632 } else if (!conf->remote_ip.sa.sa_family) {
3633 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3634 } else if (!conf->saddr.sa.sa_family) {
3635 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3636 }
3637
653ef6a3
GM		 3638 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3639 NL_SET_ERR_MSG(extack,
3640 "Local and remote address must be from the same family");
ce44a4ae 3641 return -EINVAL;
653ef6a3 3642 }
e4c7ed41 3643
653ef6a3
GM		 3644 if (vxlan_addr_multicast(&conf->saddr)) {
3645 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 3646 return -EINVAL;
653ef6a3 3647 }
0f22a3c6 3648
ce44a4ae 3649 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3650 if (!IS_ENABLED(CONFIG_IPV6)) {
3651 NL_SET_ERR_MSG(extack,
3652 "IPv6 support not enabled in the kernel");
057ba29b 3653 return -EPFNOSUPPORT;
653ef6a3 3654 }
e4c7ed41 3655 use_ipv6 = true;
a985343b 3656 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3657
3658 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3659 int local_type =
3660 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3661 int remote_type =
3662 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3663
3664 if (local_type & IPV6_ADDR_LINKLOCAL) {
3665 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3666 (remote_type != IPV6_ADDR_ANY)) {
3667 NL_SET_ERR_MSG(extack,
3668 "Invalid combination of local and remote address scopes");
0f22a3c6 3669 return -EINVAL;
653ef6a3 3670 }
0f22a3c6
MS		 3671
3672 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3673 } else {
3674 if (remote_type ==
653ef6a3
GM		 3675 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3676 NL_SET_ERR_MSG(extack,
3677 "Invalid combination of local and remote address scopes");
0f22a3c6 3678 return -EINVAL;
653ef6a3 3679 }
0f22a3c6
MS		 3680
3681 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3682 }
3683 }
057ba29b 3684 }
d342894c 3685
653ef6a3
GM		 3686 if (conf->label && !use_ipv6) {
3687 NL_SET_ERR_MSG(extack,
3688 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3689 return -EINVAL;
653ef6a3 3690 }
e7f70af1 3691
a985343b
MS		 3692 if (conf->remote_ifindex) {
3693 struct net_device *lowerdev;
34e02aa1 3694
a985343b 3695 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3696 if (!lowerdev) {
3697 NL_SET_ERR_MSG(extack,
3698 "Invalid local interface, device not found");
34e02aa1 3699 return -ENODEV;
653ef6a3 3700 }
d342894c 3701
e4c7ed41
CW		 3702#if IS_ENABLED(CONFIG_IPV6)
3703 if (use_ipv6) {
3704 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3705 if (idev && idev->cnf.disable_ipv6) {
3706 NL_SET_ERR_MSG(extack,
3707 "IPv6 support disabled by administrator");
e4c7ed41 3708 return -EPERM;
653ef6a3 3709 }
e4c7ed41
CW		 3710 }
3711#endif
3712
a985343b
MS		 3713 *lower = lowerdev;
3714 } else {
653ef6a3
GM		 3715 if (vxlan_addr_multicast(&conf->remote_ip)) {
3716 NL_SET_ERR_MSG(extack,
3717 "Local interface required for multicast remote destination");
3718
a985343b 3719 return -EINVAL;
653ef6a3 3720 }
1ba56fb4 3721
0f22a3c6 3722#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3723 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3724 NL_SET_ERR_MSG(extack,
3725 "Local interface required for link-local local/remote addresses");
0f22a3c6 3726 return -EINVAL;
653ef6a3 3727 }
0f22a3c6
MS		 3728#endif
3729
a985343b 3730 *lower = NULL;
9dc2ad10 3731 }
d342894c 3732
a985343b
MS		 3733 if (!conf->dst_port) {
3734 if (conf->flags & VXLAN_F_GPE)
3735 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3736 else
3737 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3738 }
3739
a985343b
MS		 3740 if (!conf->age_interval)
3741 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3742
a985343b
MS		 3743 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3744 if (tmp == old)
3745 continue;
91572088 3746
49f810f0
MS		 3747 if (tmp->cfg.vni != conf->vni)
3748 continue;
3749 if (tmp->cfg.dst_port != conf->dst_port)
3750 continue;
3751 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3752 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3753 continue;
3754
3755 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3756 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3757 continue;
3758
653ef6a3
GM		 3759 NL_SET_ERR_MSG(extack,
3760 "A VXLAN device with the specified VNI already exists");
49f810f0 3761 return -EEXIST;
a985343b 3762 }
91572088 3763
a985343b
MS		 3764 return 0;
3765}
3766
3767static void vxlan_config_apply(struct net_device *dev,
3768 struct vxlan_config *conf,
889ce937
SD		 3769 struct net_device *lowerdev,
3770 struct net *src_net,
3771 bool changelink)
a985343b
MS		 3772{
3773 struct vxlan_dev *vxlan = netdev_priv(dev);
3774 struct vxlan_rdst *dst = &vxlan->default_dst;
3775 unsigned short needed_headroom = ETH_HLEN;
3776 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3777 int max_mtu = ETH_MAX_MTU;
3778
3779 if (!changelink) {
3780 if (conf->flags & VXLAN_F_GPE)
3781 vxlan_raw_setup(dev);
3782 else
3783 vxlan_ether_setup(dev);
3784
3785 if (conf->mtu)
3786 dev->mtu = conf->mtu;
889ce937
SD		 3787
3788 vxlan->net = src_net;
a985343b
MS		 3789 }
3790
3791 dst->remote_vni = conf->vni;
3792
3793 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3794
3795 if (lowerdev) {
3796 dst->remote_ifindex = conf->remote_ifindex;
3797
3798 dev->gso_max_size = lowerdev->gso_max_size;
3799 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3800
a985343b 3801 needed_headroom = lowerdev->hard_header_len;
0a35dc41 3802 needed_headroom += lowerdev->needed_headroom;
91572088 3803
a5e74021 3804 dev->needed_tailroom = lowerdev->needed_tailroom;
91572088 3805
a985343b
MS		 3806 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3807 VXLAN_HEADROOM);
f870c1ff
AK		 3808 if (max_mtu < ETH_MIN_MTU)
3809 max_mtu = ETH_MIN_MTU;
3810
3811 if (!changelink && !conf->mtu)
3812 dev->mtu = max_mtu;
7e059158
DW		 3813 }
3814
a985343b
MS		 3815 if (dev->mtu > max_mtu)
3816 dev->mtu = max_mtu;
3817
b1be00a6
JB		 3818 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3819 needed_headroom += VXLAN6_HEADROOM;
3820 else
3821 needed_headroom += VXLAN_HEADROOM;
3822 dev->needed_headroom = needed_headroom;
3823
0dfbdf41 3824 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3825}
0dfbdf41 3826
a985343b 3827static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 3828 struct vxlan_config *conf, bool changelink,
3829 struct netlink_ext_ack *extack)
a985343b
MS		 3830{
3831 struct vxlan_dev *vxlan = netdev_priv(dev);
3832 struct net_device *lowerdev;
3833 int ret;
0dfbdf41 3834
653ef6a3 3835 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 3836 if (ret)
3837 return ret;
8bcdc4f3 3838
889ce937 3839 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3840
0dfbdf41
TG		 3841 return 0;
3842}
3843
c80498e3 3844static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 3845 struct vxlan_config *conf,
3846 struct netlink_ext_ack *extack)
c80498e3
ND		 3847{
3848 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3849 struct vxlan_dev *vxlan = netdev_priv(dev);
0ce1822c 3850 struct net_device *remote_dev = NULL;
0241b836 3851 struct vxlan_fdb *f = NULL;
6db92468 3852 bool unregister = false;
0ce1822c 3853 struct vxlan_rdst *dst;
c80498e3
ND		 3854 int err;
3855
0ce1822c 3856 dst = &vxlan->default_dst;
653ef6a3 3857 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 3858 if (err)
3859 return err;
3860
3861 dev->ethtool_ops = &vxlan_ethtool_ops;
3862
3863 /* create an fdb entry for a valid default destination */
0ce1822c 3864 if (!vxlan_addr_any(&dst->remote_ip)) {
0241b836 3865 err = vxlan_fdb_create(vxlan, all_zeros_mac,
0ce1822c 3866 &dst->remote_ip,
c80498e3 3867 NUD_REACHABLE | NUD_PERMANENT,
c80498e3 3868 vxlan->cfg.dst_port,
0ce1822c
TY		 3869 dst->remote_vni,
3870 dst->remote_vni,
3871 dst->remote_ifindex,
1274e1cc 3872 NTF_SELF, 0, &f, extack);
c80498e3
ND		 3873 if (err)
3874 return err;
3875 }
3876
3877 err = register_netdevice(dev);
0241b836
RP		 3878 if (err)
3879 goto errout;
6db92468 3880 unregister = true;
0241b836 3881
0ce1822c
TY		 3882 if (dst->remote_ifindex) {
3883 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
832e0979
ZC		 3884 if (!remote_dev) {
3885 err = -ENODEV;
0ce1822c 3886 goto errout;
832e0979 3887 }
0ce1822c
TY		 3888
3889 err = netdev_upper_dev_link(remote_dev, dev, extack);
3890 if (err)
3891 goto errout;
3892 }
3893
0241b836 3894 err = rtnl_configure_link(dev, NULL);
2eabcb8a 3895 if (err < 0)
0ce1822c 3896 goto unlink;
c80498e3 3897
61f46fe8 3898 if (f) {
0ce1822c 3899 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
7c31e54a
TY		 3900
3901 /* notify default fdb entry */
61f46fe8 3902 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4c59b7d1 3903 RTM_NEWNEIGH, true, extack);
7c31e54a
TY		 3904 if (err) {
3905 vxlan_fdb_destroy(vxlan, f, false, false);
0ce1822c
TY		 3906 if (remote_dev)
3907 netdev_upper_dev_unlink(remote_dev, dev);
7c31e54a
TY		 3908 goto unregister;
3909 }
61f46fe8 3910 }
0241b836 3911
c80498e3 3912 list_add(&vxlan->next, &vn->vxlan_list);
0ce1822c
TY		 3913 if (remote_dev)
3914 dst->remote_dev = remote_dev;
c80498e3 3915 return 0;
0ce1822c
TY		 3916unlink:
3917 if (remote_dev)
3918 netdev_upper_dev_unlink(remote_dev, dev);
0241b836 3919errout:
6db92468
PM		 3920 /* unregister_netdevice() destroys the default FDB entry with deletion
3921 * notification. But the addition notification was not sent yet, so
3922 * destroy the entry by hand here.
3923 */
0241b836 3924 if (f)
7c31e54a
TY		 3925 __vxlan_fdb_free(f);
3926unregister:
6db92468
PM		 3927 if (unregister)
3928 unregister_netdevice(dev);
0241b836 3929 return err;
c80498e3
ND		 3930}
3931
70fb0828
RP		 3932/* Set/clear flags based on attribute */
3933static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3934 int attrtype, unsigned long mask, bool changelink,
3935 bool changelink_supported,
3936 struct netlink_ext_ack *extack)
3937{
3938 unsigned long flags;
3939
3940 if (!tb[attrtype])
3941 return 0;
3942
3943 if (changelink && !changelink_supported) {
3944 vxlan_flag_attr_error(attrtype, extack);
3945 return -EOPNOTSUPP;
3946 }
3947
3948 if (vxlan_policy[attrtype].type == NLA_FLAG)
3949 flags = conf->flags | mask;
3950 else if (nla_get_u8(tb[attrtype]))
3951 flags = conf->flags | mask;
3952 else
3953 flags = conf->flags & ~mask;
3954
3955 conf->flags = flags;
3956
3957 return 0;
3958}
3959
8bcdc4f3
RP		 3960static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3961 struct net_device *dev, struct vxlan_config *conf,
70fb0828 3962 bool changelink, struct netlink_ext_ack *extack)
0dfbdf41 3963{
8bcdc4f3 3964 struct vxlan_dev *vxlan = netdev_priv(dev);
70fb0828 3965 int err = 0;
0dfbdf41 3966
8bcdc4f3 3967 memset(conf, 0, sizeof(*conf));
e277de5f 3968
8bcdc4f3
RP		 3969 /* if changelink operation, start with old existing cfg */
3970 if (changelink)
3971 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3972
3973 if (data[IFLA_VXLAN_ID]) {
3974 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3975
70fb0828
RP		 3976 if (changelink && (vni != conf->vni)) {
3977 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
8bcdc4f3 3978 return -EOPNOTSUPP;
70fb0828 3979 }
8bcdc4f3
RP		 3980 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3981 }
0dfbdf41
TG		 3982
3983 if (data[IFLA_VXLAN_GROUP]) {
70fb0828
RP		 3984 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3985 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
ce44a4ae 3986 return -EOPNOTSUPP;
70fb0828 3987 }
ce44a4ae 3988
8bcdc4f3 3989 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3990 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41 3991 } else if (data[IFLA_VXLAN_GROUP6]) {
70fb0828
RP		 3992 if (!IS_ENABLED(CONFIG_IPV6)) {
3993 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
0dfbdf41 3994 return -EPFNOSUPPORT;
70fb0828 3995 }
0dfbdf41 3996
70fb0828
RP		 3997 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3998 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
ce44a4ae 3999 return -EOPNOTSUPP;
70fb0828 4000 }
ce44a4ae 4001
8bcdc4f3
RP		 4002 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4003 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 4004 }
4005
4006 if (data[IFLA_VXLAN_LOCAL]) {
70fb0828
RP		 4007 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4008 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
ce44a4ae 4009 return -EOPNOTSUPP;
70fb0828 4010 }
ce44a4ae 4011
8bcdc4f3
RP		 4012 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4013 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41 4014 } else if (data[IFLA_VXLAN_LOCAL6]) {
70fb0828
RP		 4015 if (!IS_ENABLED(CONFIG_IPV6)) {
4016 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
0dfbdf41 4017 return -EPFNOSUPPORT;
70fb0828 4018 }
0dfbdf41 4019
70fb0828
RP		 4020 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4021 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
ce44a4ae 4022 return -EOPNOTSUPP;
70fb0828 4023 }
ce44a4ae 4024
0dfbdf41 4025 /* TODO: respect scope id */
8bcdc4f3
RP		 4026 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4027 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 4028 }
4029
4030 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 4031 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 4032
d342894c 4033 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 4034 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 4035
afb97186 4036 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 4037 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 4038
72f6d71e 4039 if (data[IFLA_VXLAN_TTL_INHERIT]) {
70fb0828
RP		 4040 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4041 VXLAN_F_TTL_INHERIT, changelink, false,
4042 extack);
4043 if (err)
4044 return err;
4045
72f6d71e
HL		 4046 }
4047
e7f70af1 4048 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 4049 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 4050 IPV6_FLOWLABEL_MASK;
4051
8bcdc4f3 4052 if (data[IFLA_VXLAN_LEARNING]) {
70fb0828
RP		 4053 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4054 VXLAN_F_LEARN, changelink, true,
4055 extack);
4056 if (err)
4057 return err;
8bcdc4f3
RP		 4058 } else if (!changelink) {
4059 /* default to learn on a new device */
4060 conf->flags |= VXLAN_F_LEARN;
4061 }
d342894c 4062
40051c4d 4063 if (data[IFLA_VXLAN_AGEING])
8bcdc4f3 4064 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 4065
8bcdc4f3 4066 if (data[IFLA_VXLAN_PROXY]) {
70fb0828
RP		 4067 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4068 VXLAN_F_PROXY, changelink, false,
4069 extack);
4070 if (err)
4071 return err;
8bcdc4f3 4072 }
e4f67add 4073
8bcdc4f3 4074 if (data[IFLA_VXLAN_RSC]) {
70fb0828
RP		 4075 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4076 VXLAN_F_RSC, changelink, false,
4077 extack);
4078 if (err)
4079 return err;
8bcdc4f3 4080 }
e4f67add 4081
8bcdc4f3 4082 if (data[IFLA_VXLAN_L2MISS]) {
70fb0828
RP		 4083 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4084 VXLAN_F_L2MISS, changelink, false,
4085 extack);
4086 if (err)
4087 return err;
8bcdc4f3 4088 }
e4f67add 4089
8bcdc4f3 4090 if (data[IFLA_VXLAN_L3MISS]) {
70fb0828
RP		 4091 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4092 VXLAN_F_L3MISS, changelink, false,
4093 extack);
4094 if (err)
4095 return err;
8bcdc4f3 4096 }
e4f67add 4097
8bcdc4f3 4098 if (data[IFLA_VXLAN_LIMIT]) {
70fb0828
RP		 4099 if (changelink) {
4100 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4101 "Cannot change limit");
8bcdc4f3 4102 return -EOPNOTSUPP;
70fb0828 4103 }
8bcdc4f3
RP		 4104 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4105 }
d342894c 4106
8bcdc4f3 4107 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
70fb0828
RP		 4108 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4109 VXLAN_F_COLLECT_METADATA, changelink, false,
4110 extack);
4111 if (err)
4112 return err;
8bcdc4f3 4113 }
f8a9b1bc 4114
05f47d69 4115 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 4116 if (!changelink) {
4117 const struct ifla_vxlan_port_range *p
4118 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4119 conf->port_min = ntohs(p->low);
4120 conf->port_max = ntohs(p->high);
4121 } else {
70fb0828
RP		 4122 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4123 "Cannot change port range");
8bcdc4f3
RP		 4124 return -EOPNOTSUPP;
4125 }
05f47d69 4126 }
4127
8bcdc4f3 4128 if (data[IFLA_VXLAN_PORT]) {
70fb0828
RP		 4129 if (changelink) {
4130 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4131 "Cannot change port");
8bcdc4f3 4132 return -EOPNOTSUPP;
70fb0828 4133 }
8bcdc4f3
RP		 4134 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4135 }
823aa873 4136
8bcdc4f3 4137 if (data[IFLA_VXLAN_UDP_CSUM]) {
70fb0828
RP		 4138 if (changelink) {
4139 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4140 "Cannot change UDP_CSUM flag");
8bcdc4f3 4141 return -EOPNOTSUPP;
70fb0828 4142 }
8bcdc4f3
RP		 4143 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4144 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4145 }
359a0ea9 4146
8bcdc4f3 4147 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
70fb0828
RP		 4148 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4149 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4150 false, extack);
4151 if (err)
4152 return err;
8bcdc4f3 4153 }
359a0ea9 4154
8bcdc4f3 4155 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
70fb0828
RP		 4156 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4157 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4158 false, extack);
4159 if (err)
4160 return err;
8bcdc4f3 4161 }
359a0ea9 4162
8bcdc4f3 4163 if (data[IFLA_VXLAN_REMCSUM_TX]) {
70fb0828
RP		 4164 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4165 VXLAN_F_REMCSUM_TX, changelink, false,
4166 extack);
4167 if (err)
4168 return err;
8bcdc4f3 4169 }
dfd8645e 4170
8bcdc4f3 4171 if (data[IFLA_VXLAN_REMCSUM_RX]) {
70fb0828
RP		 4172 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4173 VXLAN_F_REMCSUM_RX, changelink, false,
4174 extack);
4175 if (err)
4176 return err;
8bcdc4f3
RP		 4177 }
4178
4179 if (data[IFLA_VXLAN_GBP]) {
70fb0828
RP		 4180 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4181 VXLAN_F_GBP, changelink, false, extack);
4182 if (err)
4183 return err;
8bcdc4f3
RP		 4184 }
4185
4186 if (data[IFLA_VXLAN_GPE]) {
70fb0828
RP		 4187 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4188 VXLAN_F_GPE, changelink, false,
4189 extack);
4190 if (err)
4191 return err;
8bcdc4f3
RP		 4192 }
4193
4194 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
70fb0828
RP		 4195 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4196 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4197 false, extack);
4198 if (err)
4199 return err;
8bcdc4f3
RP		 4200 }
4201
4202 if (tb[IFLA_MTU]) {
70fb0828
RP		 4203 if (changelink) {
4204 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4205 "Cannot change mtu");
8bcdc4f3 4206 return -EOPNOTSUPP;
70fb0828 4207 }
8bcdc4f3
RP		 4208 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4209 }
4210
b4d30697
SB		 4211 if (data[IFLA_VXLAN_DF])
4212 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4213
8bcdc4f3
RP		 4214 return 0;
4215}
4216
4217static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 4218 struct nlattr *tb[], struct nlattr *data[],
4219 struct netlink_ext_ack *extack)
8bcdc4f3 4220{
8bcdc4f3
RP		 4221 struct vxlan_config conf;
4222 int err;
4223
70fb0828 4224 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
8bcdc4f3
RP		 4225 if (err)
4226 return err;
4227
653ef6a3 4228 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 4229}
dfd8645e 4230
8bcdc4f3 4231static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 4232 struct nlattr *data[],
4233 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 4234{
4235 struct vxlan_dev *vxlan = netdev_priv(dev);
8db9427d 4236 struct net_device *lowerdev;
8bcdc4f3 4237 struct vxlan_config conf;
0ce1822c 4238 struct vxlan_rdst *dst;
8bcdc4f3
RP		 4239 int err;
4240
0ce1822c 4241 dst = &vxlan->default_dst;
70fb0828 4242 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
8bcdc4f3
RP		 4243 if (err)
4244 return err;
3511494c 4245
8db9427d
PM		 4246 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4247 vxlan, extack);
8bcdc4f3
RP		 4248 if (err)
4249 return err;
0ace2ca8 4250
c6761cf5
TY		 4251 if (dst->remote_dev == lowerdev)
4252 lowerdev = NULL;
4253
0ce1822c
TY		 4254 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4255 extack);
4256 if (err)
4257 return err;
4258
8bcdc4f3 4259 /* handle default dst entry */
038a5a99 4260 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
fe1e0713
L		 4261 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4262
4263 spin_lock_bh(&vxlan->hash_lock[hash_index]);
038a5a99 4264 if (!vxlan_addr_any(&conf.remote_ip)) {
ce5e098f 4265 err = vxlan_fdb_update(vxlan, all_zeros_mac,
038a5a99 4266 &conf.remote_ip,
8bcdc4f3 4267 NUD_REACHABLE | NUD_PERMANENT,
ce5e098f 4268 NLM_F_APPEND | NLM_F_CREATE,
8bcdc4f3 4269 vxlan->cfg.dst_port,
038a5a99
PM		 4270 conf.vni, conf.vni,
4271 conf.remote_ifindex,
1274e1cc 4272 NTF_SELF, 0, true, extack);
8bcdc4f3 4273 if (err) {
fe1e0713 4274 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
0ce1822c
TY		 4275 netdev_adjacent_change_abort(dst->remote_dev,
4276 lowerdev, dev);
8bcdc4f3
RP		 4277 return err;
4278 }
4279 }
1cdc98c2
PM		 4280 if (!vxlan_addr_any(&dst->remote_ip))
4281 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4282 dst->remote_ip,
4283 vxlan->cfg.dst_port,
4284 dst->remote_vni,
4285 dst->remote_vni,
4286 dst->remote_ifindex,
4287 true);
fe1e0713 4288 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
8bcdc4f3 4289 }
ce577668 4290
038a5a99
PM		 4291 if (conf.age_interval != vxlan->cfg.age_interval)
4292 mod_timer(&vxlan->age_timer, jiffies);
4293
0ce1822c 4294 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
845e0ebb 4295 if (lowerdev && lowerdev != dst->remote_dev)
0ce1822c 4296 dst->remote_dev = lowerdev;
038a5a99 4297 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
8bcdc4f3 4298 return 0;
d342894c 4299}
4300
4301static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4302{
4303 struct vxlan_dev *vxlan = netdev_priv(dev);
4304
8b3f9337
RP		 4305 vxlan_flush(vxlan, true);
4306
553675fb 4307 list_del(&vxlan->next);
d342894c 4308 unregister_netdevice_queue(dev, head);
0ce1822c
TY		 4309 if (vxlan->default_dst.remote_dev)
4310 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
d342894c 4311}
4312
4313static size_t vxlan_get_size(const struct net_device *dev)
4314{
4315
4316 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 4317 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 4318 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 4319 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 4320 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
8fd78069 4321 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
d342894c 4322 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
b4d30697 4323 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
e7f70af1 4324 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 4325 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 4326 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4327 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4328 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4329 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 4330 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 4331 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4332 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 4333 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 4334 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 4338 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4339 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 4340 0;
4341}
4342
4343static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4344{
4345 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 4346 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 4347 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 4348 .low = htons(vxlan->cfg.port_min),
4349 .high = htons(vxlan->cfg.port_max),
05f47d69 4350 };
d342894c 4351
54bfd872 4352 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 4353 goto nla_put_failure;
4354
e4c7ed41
CW		 4355 if (!vxlan_addr_any(&dst->remote_ip)) {
4356 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 4357 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4358 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 4359 goto nla_put_failure;
4360#if IS_ENABLED(CONFIG_IPV6)
4361 } else {
930345ea
JB		 4362 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4363 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 4364 goto nla_put_failure;
4365#endif
4366 }
4367 }
d342894c 4368
c7995c43 4369 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 4370 goto nla_put_failure;
4371
0dfbdf41
TG		 4372 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4373 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 4374 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 4375 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 4376 goto nla_put_failure;
4377#if IS_ENABLED(CONFIG_IPV6)
4378 } else {
930345ea 4379 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 4380 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 4381 goto nla_put_failure;
4382#endif
4383 }
4384 }
d342894c 4385
0dfbdf41 4386 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
8fd78069
HL		 4387 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4388 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
0dfbdf41 4389 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
b4d30697 4390 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
e7f70af1 4391 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 4392 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
98c81476 4393 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 4394 nla_put_u8(skb, IFLA_VXLAN_PROXY,
98c81476 4395 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
dc5321d7
MS		 4396 nla_put_u8(skb, IFLA_VXLAN_RSC,
4397 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 4398 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
98c81476 4399 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 4400 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
98c81476 4401 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 4402 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 4403 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 4404 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4405 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4406 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 4407 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
98c81476 4408 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 4409 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
98c81476 4410 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 4411 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
98c81476 4412 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 4413 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
98c81476 4414 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 4415 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
98c81476 4416 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 4417 goto nla_put_failure;
4418
05f47d69 4419 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4420 goto nla_put_failure;
4421
dc5321d7 4422 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 4423 nla_put_flag(skb, IFLA_VXLAN_GBP))
4424 goto nla_put_failure;
4425
dc5321d7 4426 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 4427 nla_put_flag(skb, IFLA_VXLAN_GPE))
4428 goto nla_put_failure;
4429
dc5321d7 4430 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 4431 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4432 goto nla_put_failure;
4433
d342894c 4434 return 0;
4435
4436nla_put_failure:
4437 return -EMSGSIZE;
4438}
4439
1728d4fa
ND		 4440static struct net *vxlan_get_link_net(const struct net_device *dev)
4441{
4442 struct vxlan_dev *vxlan = netdev_priv(dev);
4443
4444 return vxlan->net;
4445}
4446
d342894c 4447static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4448 .kind = "vxlan",
4449 .maxtype = IFLA_VXLAN_MAX,
4450 .policy = vxlan_policy,
4451 .priv_size = sizeof(struct vxlan_dev),
4452 .setup = vxlan_setup,
4453 .validate = vxlan_validate,
4454 .newlink = vxlan_newlink,
8bcdc4f3 4455 .changelink = vxlan_changelink,
d342894c 4456 .dellink = vxlan_dellink,
4457 .get_size = vxlan_get_size,
4458 .fill_info = vxlan_fill_info,
1728d4fa 4459 .get_link_net = vxlan_get_link_net,
d342894c 4460};
4461
cf5da330
ND		 4462struct net_device *vxlan_dev_create(struct net *net, const char *name,
4463 u8 name_assign_type,
4464 struct vxlan_config *conf)
4465{
4466 struct nlattr *tb[IFLA_MAX + 1];
4467 struct net_device *dev;
4468 int err;
4469
4470 memset(&tb, 0, sizeof(tb));
4471
4472 dev = rtnl_create_link(net, name, name_assign_type,
d0522f1c 4473 &vxlan_link_ops, tb, NULL);
cf5da330
ND		 4474 if (IS_ERR(dev))
4475 return dev;
4476
653ef6a3 4477 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 4478 if (err < 0) {
4479 free_netdev(dev);
4480 return ERR_PTR(err);
4481 }
4482
4483 err = rtnl_configure_link(dev, NULL);
4484 if (err < 0) {
4485 LIST_HEAD(list_kill);
4486
4487 vxlan_dellink(dev, &list_kill);
4488 unregister_netdevice_many(&list_kill);
4489 return ERR_PTR(err);
4490 }
4491
4492 return dev;
4493}
4494EXPORT_SYMBOL_GPL(vxlan_dev_create);
4495
acaf4e70
DB		 4496static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4497 struct net_device *dev)
4498{
4499 struct vxlan_dev *vxlan, *next;
4500 LIST_HEAD(list_kill);
4501
4502 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4503 struct vxlan_rdst *dst = &vxlan->default_dst;
4504
4505 /* In case we created vxlan device with carrier
4506 * and we loose the carrier due to module unload
4507 * we also need to remove vxlan device. In other
4508 * cases, it's not necessary and remote_ifindex
4509 * is 0 here, so no matches.
4510 */
4511 if (dst->remote_ifindex == dev->ifindex)
4512 vxlan_dellink(vxlan->dev, &list_kill);
4513 }
4514
4515 unregister_netdevice_many(&list_kill);
4516}
4517
b7aade15
HFS		 4518static int vxlan_netdevice_event(struct notifier_block *unused,
4519 unsigned long event, void *ptr)
acaf4e70
DB		 4520{
4521 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 4522 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 4523
04584957 4524 if (event == NETDEV_UNREGISTER) {
cc4e3835
JK		 4525 if (!dev->udp_tunnel_nic_info)
4526 vxlan_offload_rx_ports(dev, false);
acaf4e70 4527 vxlan_handle_lowerdev_unregister(vn, dev);
04584957 4528 } else if (event == NETDEV_REGISTER) {
cc4e3835
JK		 4529 if (!dev->udp_tunnel_nic_info)
4530 vxlan_offload_rx_ports(dev, true);
04584957
SD		 4531 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4532 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 4533 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 4534 }
acaf4e70
DB		 4535
4536 return NOTIFY_DONE;
4537}
4538
4539static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 4540 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 4541};
4542
0efe1173
PM		 4543static void
4544vxlan_fdb_offloaded_set(struct net_device *dev,
4545 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4546{
4547 struct vxlan_dev *vxlan = netdev_priv(dev);
4548 struct vxlan_rdst *rdst;
4549 struct vxlan_fdb *f;
fe1e0713
L		 4550 u32 hash_index;
4551
4552 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
0efe1173 4553
fe1e0713 4554 spin_lock_bh(&vxlan->hash_lock[hash_index]);
0efe1173
PM		 4555
4556 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4557 if (!f)
4558 goto out;
4559
4560 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4561 fdb_info->remote_port,
4562 fdb_info->remote_vni,
4563 fdb_info->remote_ifindex);
4564 if (!rdst)
4565 goto out;
4566
4567 rdst->offloaded = fdb_info->offloaded;
4568
4569out:
fe1e0713 4570 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
0efe1173
PM		 4571}
4572
5728ae0d
PM		 4573static int
4574vxlan_fdb_external_learn_add(struct net_device *dev,
4575 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4576{
4577 struct vxlan_dev *vxlan = netdev_priv(dev);
4c59b7d1 4578 struct netlink_ext_ack *extack;
fe1e0713 4579 u32 hash_index;
5728ae0d
PM		 4580 int err;
4581
fe1e0713 4582 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4c59b7d1
PM		 4583 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4584
fe1e0713 4585 spin_lock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4586 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4587 NUD_REACHABLE,
4588 NLM_F_CREATE | NLM_F_REPLACE,
4589 fdb_info->remote_port,
4590 fdb_info->vni,
4591 fdb_info->remote_vni,
4592 fdb_info->remote_ifindex,
4593 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
1274e1cc 4594 0, false, extack);
fe1e0713 4595 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4596
4597 return err;
4598}
4599
4600static int
4601vxlan_fdb_external_learn_del(struct net_device *dev,
4602 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4603{
4604 struct vxlan_dev *vxlan = netdev_priv(dev);
4605 struct vxlan_fdb *f;
fe1e0713 4606 u32 hash_index;
5728ae0d
PM		 4607 int err = 0;
4608
fe1e0713
L		 4609 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4610 spin_lock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4611
4612 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4613 if (!f)
4614 err = -ENOENT;
4615 else if (f->flags & NTF_EXT_LEARNED)
4616 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4617 fdb_info->remote_ip,
4618 fdb_info->remote_port,
4619 fdb_info->vni,
4620 fdb_info->remote_vni,
4621 fdb_info->remote_ifindex,
4622 false);
4623
fe1e0713 4624 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4625
4626 return err;
4627}
4628
0efe1173
PM		 4629static int vxlan_switchdev_event(struct notifier_block *unused,
4630 unsigned long event, void *ptr)
4631{
4632 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
5728ae0d
PM		 4633 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4634 int err = 0;
0efe1173
PM		 4635
4636 switch (event) {
4637 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4638 vxlan_fdb_offloaded_set(dev, ptr);
4639 break;
5728ae0d
PM		 4640 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4641 fdb_info = ptr;
4642 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4643 if (err) {
4644 err = notifier_from_errno(err);
4645 break;
4646 }
4647 fdb_info->offloaded = true;
4648 vxlan_fdb_offloaded_set(dev, fdb_info);
4649 break;
4650 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4651 fdb_info = ptr;
4652 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4653 if (err) {
4654 err = notifier_from_errno(err);
4655 break;
4656 }
4657 fdb_info->offloaded = false;
4658 vxlan_fdb_offloaded_set(dev, fdb_info);
4659 break;
0efe1173
PM		 4660 }
4661
5728ae0d 4662 return err;
0efe1173
PM		 4663}
4664
4665static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4666 .notifier_call = vxlan_switchdev_event,
4667};
4668
79472fe8
RP		 4669static void vxlan_fdb_nh_flush(struct nexthop *nh)
4670{
4671 struct vxlan_fdb *fdb;
4672 struct vxlan_dev *vxlan;
4673 u32 hash_index;
4674
4675 rcu_read_lock();
4676 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4677 vxlan = rcu_dereference(fdb->vdev);
4678 WARN_ON(!vxlan);
4679 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4680 vxlan->default_dst.remote_vni);
4681 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4682 if (!hlist_unhashed(&fdb->hlist))
4683 vxlan_fdb_destroy(vxlan, fdb, false, false);
4684 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4685 }
4686 rcu_read_unlock();
4687}
4688
c7cdbe2e
RP		 4689static int vxlan_nexthop_event(struct notifier_block *nb,
4690 unsigned long event, void *ptr)
4691{
1ec69d18
IS		 4692 struct nh_notifier_info *info = ptr;
4693 struct nexthop *nh;
4694
4695 if (event != NEXTHOP_EVENT_DEL)
4696 return NOTIFY_DONE;
c7cdbe2e 4697
1ec69d18
IS		 4698 nh = nexthop_find_by_id(info->net, info->id);
4699 if (!nh)
c7cdbe2e
RP		 4700 return NOTIFY_DONE;
4701
79472fe8 4702 vxlan_fdb_nh_flush(nh);
c7cdbe2e
RP		 4703
4704 return NOTIFY_DONE;
4705}
4706
d342894c 4707static __net_init int vxlan_init_net(struct net *net)
4708{
4709 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 4710 unsigned int h;
d342894c 4711
553675fb 4712 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 4713 spin_lock_init(&vn->sock_lock);
626d667b 4714 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
d342894c 4715
553675fb 4716 for (h = 0; h < PORT_HASH_SIZE; ++h)
4717 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 4718
ce7e9c8a
IS		 4719 return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4720 NULL);
d342894c 4721}
4722
57b61127 4723static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
f01ec1c0
ND		 4724{
4725 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4726 struct vxlan_dev *vxlan, *next;
4727 struct net_device *dev, *aux;
0e4ec5ac 4728 unsigned int h;
f01ec1c0 4729
f01ec1c0
ND		 4730 for_each_netdev_safe(net, dev, aux)
4731 if (dev->rtnl_link_ops == &vxlan_link_ops)
57b61127 4732 unregister_netdevice_queue(dev, head);
f01ec1c0
ND		 4733
4734 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4735 /* If vxlan->dev is in the same netns, it has already been added
4736 * to the list by the previous loop.
4737 */
cc4807bb 4738 if (!net_eq(dev_net(vxlan->dev), net))
57b61127 4739 unregister_netdevice_queue(vxlan->dev, head);
f01ec1c0
ND		 4740 }
4741
0e4ec5ac
VA		 4742 for (h = 0; h < PORT_HASH_SIZE; ++h)
4743 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND		 4744}
4745
57b61127
HY		 4746static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4747{
4748 struct net *net;
4749 LIST_HEAD(list);
4750
4751 rtnl_lock();
626d667b
IS		 4752 list_for_each_entry(net, net_list, exit_list) {
4753 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4754
4755 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4756 }
57b61127
HY		 4757 list_for_each_entry(net, net_list, exit_list)
4758 vxlan_destroy_tunnels(net, &list);
4759
4760 unregister_netdevice_many(&list);
4761 rtnl_unlock();
4762}
4763
d342894c 4764static struct pernet_operations vxlan_net_ops = {
4765 .init = vxlan_init_net,
57b61127 4766 .exit_batch = vxlan_exit_batch_net,
d342894c 4767 .id = &vxlan_net_id,
4768 .size = sizeof(struct vxlan_net),
4769};
4770
4771static int __init vxlan_init_module(void)
4772{
4773 int rc;
4774
4775 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4776
783c1463 4777 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 4778 if (rc)
4779 goto out1;
4780
acaf4e70 4781 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 4782 if (rc)
4783 goto out2;
4784
0efe1173 4785 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 4786 if (rc)
4787 goto out3;
d342894c 4788
0efe1173
PM		 4789 rc = rtnl_link_register(&vxlan_link_ops);
4790 if (rc)
4791 goto out4;
4792
acaf4e70 4793 return 0;
0efe1173
PM		 4794out4:
4795 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 4796out3:
4797 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 4798out2:
783c1463 4799 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 4800out1:
4801 return rc;
4802}
7332a13b 4803late_initcall(vxlan_init_module);
d342894c 4804
4805static void __exit vxlan_cleanup_module(void)
4806{
b7153984 4807 rtnl_link_unregister(&vxlan_link_ops);
0efe1173 4808 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70 4809 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 4810 unregister_pernet_subsys(&vxlan_net_ops);
4811 /* rcu_barrier() is called by netns */
d342894c 4812}
4813module_exit(vxlan_cleanup_module);
4814
4815MODULE_LICENSE("GPL");
4816MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 4817MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 4818MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 4819MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Hirsute Linux kernel mirror