xref: /linux/drivers/net/vxlan/vxlan_core.c (revision 2dcd22023cabc1b15b36af3c10924fcfee79b341)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * VXLAN: Virtual eXtensible Local Area Network
4  *
5  * Copyright (c) 2012-2013 Vyatta Inc.
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
18 #include <net/arp.h>
19 #include <net/ndisc.h>
20 #include <net/gro.h>
21 #include <net/ipv6_stubs.h>
22 #include <net/ip.h>
23 #include <net/icmp.h>
24 #include <net/rtnetlink.h>
25 #include <net/inet_ecn.h>
26 #include <net/net_namespace.h>
27 #include <net/netns/generic.h>
28 #include <net/tun_proto.h>
29 #include <net/vxlan.h>
30 #include <net/nexthop.h>
31 
32 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_tunnel.h>
34 #include <net/ip6_checksum.h>
35 #endif
36 
37 #include "vxlan_private.h"
38 
39 #define VXLAN_VERSION	"0.1"
40 
41 #define FDB_AGE_DEFAULT 300 /* 5 min */
42 #define FDB_AGE_INTERVAL (10 * HZ)	/* rescan interval */
43 
44 /* UDP port for VXLAN traffic.
45  * The IANA assigned port is 4789, but the Linux default is 8472
46  * for compatibility with early adopters.
47  */
48 static unsigned short vxlan_port __read_mostly = 8472;
49 module_param_named(udp_port, vxlan_port, ushort, 0444);
50 MODULE_PARM_DESC(udp_port, "Destination UDP port");
51 
52 static bool log_ecn_error = true;
53 module_param(log_ecn_error, bool, 0644);
54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55 
56 unsigned int vxlan_net_id;
57 
58 const u8 all_zeros_mac[ETH_ALEN + 2];
59 static struct rtnl_link_ops vxlan_link_ops;
60 
61 static int vxlan_sock_add(struct vxlan_dev *vxlan);
62 
63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64 
65 /* salt for hash table */
66 static u32 vxlan_salt __read_mostly;
67 
68 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
69 {
70 	return vs->flags & VXLAN_F_COLLECT_METADATA ||
71 	       ip_tunnel_collect_metadata();
72 }
73 
74 /* Find VXLAN socket based on network namespace, address family, UDP port,
75  * enabled unshareable flags and socket device binding (see l3mdev with
76  * non-default VRF).
77  */
78 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
79 					  __be16 port, u32 flags, int ifindex)
80 {
81 	struct vxlan_sock *vs;
82 
83 	flags &= VXLAN_F_RCV_FLAGS;
84 
85 	hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
86 		if (inet_sk(vs->sock->sk)->inet_sport == port &&
87 		    vxlan_get_sk_family(vs) == family &&
88 		    vs->flags == flags &&
89 		    vs->sock->sk->sk_bound_dev_if == ifindex)
90 			return vs;
91 	}
92 	return NULL;
93 }
94 
95 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
96 					   int ifindex, __be32 vni,
97 					   struct vxlan_vni_node **vninode)
98 {
99 	struct vxlan_vni_node *vnode;
100 	struct vxlan_dev_node *node;
101 
102 	/* For flow based devices, map all packets to VNI 0 */
103 	if (vs->flags & VXLAN_F_COLLECT_METADATA &&
104 	    !(vs->flags & VXLAN_F_VNIFILTER))
105 		vni = 0;
106 
107 	hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
108 		if (!node->vxlan)
109 			continue;
110 		vnode = NULL;
111 		if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
112 			vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
113 			if (!vnode)
114 				continue;
115 		} else if (node->vxlan->default_dst.remote_vni != vni) {
116 			continue;
117 		}
118 
119 		if (IS_ENABLED(CONFIG_IPV6)) {
120 			const struct vxlan_config *cfg = &node->vxlan->cfg;
121 
122 			if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
123 			    cfg->remote_ifindex != ifindex)
124 				continue;
125 		}
126 
127 		if (vninode)
128 			*vninode = vnode;
129 		return node->vxlan;
130 	}
131 
132 	return NULL;
133 }
134 
135 /* Look up VNI in a per net namespace table */
136 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
137 					__be32 vni, sa_family_t family,
138 					__be16 port, u32 flags)
139 {
140 	struct vxlan_sock *vs;
141 
142 	vs = vxlan_find_sock(net, family, port, flags, ifindex);
143 	if (!vs)
144 		return NULL;
145 
146 	return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
147 }
148 
149 /* Fill in neighbour message in skbuff. */
150 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
151 			  const struct vxlan_fdb *fdb,
152 			  u32 portid, u32 seq, int type, unsigned int flags,
153 			  const struct vxlan_rdst *rdst)
154 {
155 	unsigned long now = jiffies;
156 	struct nda_cacheinfo ci;
157 	bool send_ip, send_eth;
158 	struct nlmsghdr *nlh;
159 	struct nexthop *nh;
160 	struct ndmsg *ndm;
161 	int nh_family;
162 	u32 nh_id;
163 
164 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
165 	if (nlh == NULL)
166 		return -EMSGSIZE;
167 
168 	ndm = nlmsg_data(nlh);
169 	memset(ndm, 0, sizeof(*ndm));
170 
171 	send_eth = send_ip = true;
172 
173 	rcu_read_lock();
174 	nh = rcu_dereference(fdb->nh);
175 	if (nh) {
176 		nh_family = nexthop_get_family(nh);
177 		nh_id = nh->id;
178 	}
179 	rcu_read_unlock();
180 
181 	if (type == RTM_GETNEIGH) {
182 		if (rdst) {
183 			send_ip = !vxlan_addr_any(&rdst->remote_ip);
184 			ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
185 		} else if (nh) {
186 			ndm->ndm_family = nh_family;
187 		}
188 		send_eth = !is_zero_ether_addr(fdb->eth_addr);
189 	} else
190 		ndm->ndm_family	= AF_BRIDGE;
191 	ndm->ndm_state = fdb->state;
192 	ndm->ndm_ifindex = vxlan->dev->ifindex;
193 	ndm->ndm_flags = fdb->flags;
194 	if (rdst && rdst->offloaded)
195 		ndm->ndm_flags |= NTF_OFFLOADED;
196 	ndm->ndm_type = RTN_UNICAST;
197 
198 	if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
199 	    nla_put_s32(skb, NDA_LINK_NETNSID,
200 			peernet2id(dev_net(vxlan->dev), vxlan->net)))
201 		goto nla_put_failure;
202 
203 	if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
204 		goto nla_put_failure;
205 	if (nh) {
206 		if (nla_put_u32(skb, NDA_NH_ID, nh_id))
207 			goto nla_put_failure;
208 	} else if (rdst) {
209 		if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
210 						  &rdst->remote_ip))
211 			goto nla_put_failure;
212 
213 		if (rdst->remote_port &&
214 		    rdst->remote_port != vxlan->cfg.dst_port &&
215 		    nla_put_be16(skb, NDA_PORT, rdst->remote_port))
216 			goto nla_put_failure;
217 		if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
218 		    nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
219 			goto nla_put_failure;
220 		if (rdst->remote_ifindex &&
221 		    nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
222 			goto nla_put_failure;
223 	}
224 
225 	if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
226 	    nla_put_u32(skb, NDA_SRC_VNI,
227 			be32_to_cpu(fdb->vni)))
228 		goto nla_put_failure;
229 
230 	ci.ndm_used	 = jiffies_to_clock_t(now - fdb->used);
231 	ci.ndm_confirmed = 0;
232 	ci.ndm_updated	 = jiffies_to_clock_t(now - fdb->updated);
233 	ci.ndm_refcnt	 = 0;
234 
235 	if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
236 		goto nla_put_failure;
237 
238 	nlmsg_end(skb, nlh);
239 	return 0;
240 
241 nla_put_failure:
242 	nlmsg_cancel(skb, nlh);
243 	return -EMSGSIZE;
244 }
245 
246 static inline size_t vxlan_nlmsg_size(void)
247 {
248 	return NLMSG_ALIGN(sizeof(struct ndmsg))
249 		+ nla_total_size(ETH_ALEN) /* NDA_LLADDR */
250 		+ nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
251 		+ nla_total_size(sizeof(__be16)) /* NDA_PORT */
252 		+ nla_total_size(sizeof(__be32)) /* NDA_VNI */
253 		+ nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
254 		+ nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
255 		+ nla_total_size(sizeof(struct nda_cacheinfo));
256 }
257 
258 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
259 			       struct vxlan_rdst *rd, int type)
260 {
261 	struct net *net = dev_net(vxlan->dev);
262 	struct sk_buff *skb;
263 	int err = -ENOBUFS;
264 
265 	skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
266 	if (skb == NULL)
267 		goto errout;
268 
269 	err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
270 	if (err < 0) {
271 		/* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
272 		WARN_ON(err == -EMSGSIZE);
273 		kfree_skb(skb);
274 		goto errout;
275 	}
276 
277 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
278 	return;
279 errout:
280 	if (err < 0)
281 		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
282 }
283 
284 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
285 			    const struct vxlan_fdb *fdb,
286 			    const struct vxlan_rdst *rd,
287 			    struct netlink_ext_ack *extack,
288 			    struct switchdev_notifier_vxlan_fdb_info *fdb_info)
289 {
290 	fdb_info->info.dev = vxlan->dev;
291 	fdb_info->info.extack = extack;
292 	fdb_info->remote_ip = rd->remote_ip;
293 	fdb_info->remote_port = rd->remote_port;
294 	fdb_info->remote_vni = rd->remote_vni;
295 	fdb_info->remote_ifindex = rd->remote_ifindex;
296 	memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
297 	fdb_info->vni = fdb->vni;
298 	fdb_info->offloaded = rd->offloaded;
299 	fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
300 }
301 
302 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
303 					      struct vxlan_fdb *fdb,
304 					      struct vxlan_rdst *rd,
305 					      bool adding,
306 					      struct netlink_ext_ack *extack)
307 {
308 	struct switchdev_notifier_vxlan_fdb_info info;
309 	enum switchdev_notifier_type notifier_type;
310 	int ret;
311 
312 	if (WARN_ON(!rd))
313 		return 0;
314 
315 	notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
316 			       : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
317 	vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
318 	ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
319 				       &info.info, extack);
320 	return notifier_to_errno(ret);
321 }
322 
323 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
324 			    struct vxlan_rdst *rd, int type, bool swdev_notify,
325 			    struct netlink_ext_ack *extack)
326 {
327 	int err;
328 
329 	if (swdev_notify && rd) {
330 		switch (type) {
331 		case RTM_NEWNEIGH:
332 			err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
333 								 true, extack);
334 			if (err)
335 				return err;
336 			break;
337 		case RTM_DELNEIGH:
338 			vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
339 							   false, extack);
340 			break;
341 		}
342 	}
343 
344 	__vxlan_fdb_notify(vxlan, fdb, rd, type);
345 	return 0;
346 }
347 
348 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
349 {
350 	struct vxlan_dev *vxlan = netdev_priv(dev);
351 	struct vxlan_fdb f = {
352 		.state = NUD_STALE,
353 	};
354 	struct vxlan_rdst remote = {
355 		.remote_ip = *ipa, /* goes to NDA_DST */
356 		.remote_vni = cpu_to_be32(VXLAN_N_VID),
357 	};
358 
359 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
360 }
361 
362 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
363 {
364 	struct vxlan_fdb f = {
365 		.state = NUD_STALE,
366 	};
367 	struct vxlan_rdst remote = { };
368 
369 	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
370 
371 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
372 }
373 
374 /* Hash Ethernet address */
375 static u32 eth_hash(const unsigned char *addr)
376 {
377 	u64 value = get_unaligned((u64 *)addr);
378 
379 	/* only want 6 bytes */
380 #ifdef __BIG_ENDIAN
381 	value >>= 16;
382 #else
383 	value <<= 16;
384 #endif
385 	return hash_64(value, FDB_HASH_BITS);
386 }
387 
388 u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
389 {
390 	/* use 1 byte of OUI and 3 bytes of NIC */
391 	u32 key = get_unaligned((u32 *)(addr + 2));
392 
393 	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
394 }
395 
396 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
397 {
398 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
399 		return eth_vni_hash(mac, vni);
400 	else
401 		return eth_hash(mac);
402 }
403 
404 /* Hash chain to use given mac address */
405 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
406 						const u8 *mac, __be32 vni)
407 {
408 	return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
409 }
410 
411 /* Look up Ethernet address in forwarding table */
412 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
413 					  const u8 *mac, __be32 vni)
414 {
415 	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
416 	struct vxlan_fdb *f;
417 
418 	hlist_for_each_entry_rcu(f, head, hlist) {
419 		if (ether_addr_equal(mac, f->eth_addr)) {
420 			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
421 				if (vni == f->vni)
422 					return f;
423 			} else {
424 				return f;
425 			}
426 		}
427 	}
428 
429 	return NULL;
430 }
431 
432 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
433 					const u8 *mac, __be32 vni)
434 {
435 	struct vxlan_fdb *f;
436 
437 	f = __vxlan_find_mac(vxlan, mac, vni);
438 	if (f && f->used != jiffies)
439 		f->used = jiffies;
440 
441 	return f;
442 }
443 
444 /* caller should hold vxlan->hash_lock */
445 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
446 					      union vxlan_addr *ip, __be16 port,
447 					      __be32 vni, __u32 ifindex)
448 {
449 	struct vxlan_rdst *rd;
450 
451 	list_for_each_entry(rd, &f->remotes, list) {
452 		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
453 		    rd->remote_port == port &&
454 		    rd->remote_vni == vni &&
455 		    rd->remote_ifindex == ifindex)
456 			return rd;
457 	}
458 
459 	return NULL;
460 }
461 
462 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
463 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info)
464 {
465 	struct vxlan_dev *vxlan = netdev_priv(dev);
466 	u8 eth_addr[ETH_ALEN + 2] = { 0 };
467 	struct vxlan_rdst *rdst;
468 	struct vxlan_fdb *f;
469 	int rc = 0;
470 
471 	if (is_multicast_ether_addr(mac) ||
472 	    is_zero_ether_addr(mac))
473 		return -EINVAL;
474 
475 	ether_addr_copy(eth_addr, mac);
476 
477 	rcu_read_lock();
478 
479 	f = __vxlan_find_mac(vxlan, eth_addr, vni);
480 	if (!f) {
481 		rc = -ENOENT;
482 		goto out;
483 	}
484 
485 	rdst = first_remote_rcu(f);
486 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
487 
488 out:
489 	rcu_read_unlock();
490 	return rc;
491 }
492 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
493 
494 static int vxlan_fdb_notify_one(struct notifier_block *nb,
495 				const struct vxlan_dev *vxlan,
496 				const struct vxlan_fdb *f,
497 				const struct vxlan_rdst *rdst,
498 				struct netlink_ext_ack *extack)
499 {
500 	struct switchdev_notifier_vxlan_fdb_info fdb_info;
501 	int rc;
502 
503 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
504 	rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
505 			       &fdb_info);
506 	return notifier_to_errno(rc);
507 }
508 
509 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
510 		     struct notifier_block *nb,
511 		     struct netlink_ext_ack *extack)
512 {
513 	struct vxlan_dev *vxlan;
514 	struct vxlan_rdst *rdst;
515 	struct vxlan_fdb *f;
516 	unsigned int h;
517 	int rc = 0;
518 
519 	if (!netif_is_vxlan(dev))
520 		return -EINVAL;
521 	vxlan = netdev_priv(dev);
522 
523 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
524 		spin_lock_bh(&vxlan->hash_lock[h]);
525 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
526 			if (f->vni == vni) {
527 				list_for_each_entry(rdst, &f->remotes, list) {
528 					rc = vxlan_fdb_notify_one(nb, vxlan,
529 								  f, rdst,
530 								  extack);
531 					if (rc)
532 						goto unlock;
533 				}
534 			}
535 		}
536 		spin_unlock_bh(&vxlan->hash_lock[h]);
537 	}
538 	return 0;
539 
540 unlock:
541 	spin_unlock_bh(&vxlan->hash_lock[h]);
542 	return rc;
543 }
544 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
545 
546 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
547 {
548 	struct vxlan_dev *vxlan;
549 	struct vxlan_rdst *rdst;
550 	struct vxlan_fdb *f;
551 	unsigned int h;
552 
553 	if (!netif_is_vxlan(dev))
554 		return;
555 	vxlan = netdev_priv(dev);
556 
557 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
558 		spin_lock_bh(&vxlan->hash_lock[h]);
559 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
560 			if (f->vni == vni)
561 				list_for_each_entry(rdst, &f->remotes, list)
562 					rdst->offloaded = false;
563 		spin_unlock_bh(&vxlan->hash_lock[h]);
564 	}
565 
566 }
567 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
568 
569 /* Replace destination of unicast mac */
570 static int vxlan_fdb_replace(struct vxlan_fdb *f,
571 			     union vxlan_addr *ip, __be16 port, __be32 vni,
572 			     __u32 ifindex, struct vxlan_rdst *oldrd)
573 {
574 	struct vxlan_rdst *rd;
575 
576 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
577 	if (rd)
578 		return 0;
579 
580 	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
581 	if (!rd)
582 		return 0;
583 
584 	*oldrd = *rd;
585 	dst_cache_reset(&rd->dst_cache);
586 	rd->remote_ip = *ip;
587 	rd->remote_port = port;
588 	rd->remote_vni = vni;
589 	rd->remote_ifindex = ifindex;
590 	rd->offloaded = false;
591 	return 1;
592 }
593 
594 /* Add/update destinations for multicast */
595 static int vxlan_fdb_append(struct vxlan_fdb *f,
596 			    union vxlan_addr *ip, __be16 port, __be32 vni,
597 			    __u32 ifindex, struct vxlan_rdst **rdp)
598 {
599 	struct vxlan_rdst *rd;
600 
601 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
602 	if (rd)
603 		return 0;
604 
605 	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
606 	if (rd == NULL)
607 		return -ENOMEM;
608 
609 	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
610 		kfree(rd);
611 		return -ENOMEM;
612 	}
613 
614 	rd->remote_ip = *ip;
615 	rd->remote_port = port;
616 	rd->offloaded = false;
617 	rd->remote_vni = vni;
618 	rd->remote_ifindex = ifindex;
619 
620 	list_add_tail_rcu(&rd->list, &f->remotes);
621 
622 	*rdp = rd;
623 	return 1;
624 }
625 
626 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol)
627 {
628 	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr;
629 
630 	/* Need to have Next Protocol set for interfaces in GPE mode. */
631 	if (!gpe->np_applied)
632 		return false;
633 	/* "The initial version is 0. If a receiver does not support the
634 	 * version indicated it MUST drop the packet.
635 	 */
636 	if (gpe->version != 0)
637 		return false;
638 	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
639 	 * processing MUST occur." However, we don't implement OAM
640 	 * processing, thus drop the packet.
641 	 */
642 	if (gpe->oam_flag)
643 		return false;
644 
645 	*protocol = tun_p_to_eth_p(gpe->next_protocol);
646 	if (!*protocol)
647 		return false;
648 
649 	return true;
650 }
651 
652 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
653 					  unsigned int off,
654 					  struct vxlanhdr *vh, size_t hdrlen,
655 					  __be32 vni_field,
656 					  struct gro_remcsum *grc,
657 					  bool nopartial)
658 {
659 	size_t start, offset;
660 
661 	if (skb->remcsum_offload)
662 		return vh;
663 
664 	if (!NAPI_GRO_CB(skb)->csum_valid)
665 		return NULL;
666 
667 	start = vxlan_rco_start(vni_field);
668 	offset = start + vxlan_rco_offset(vni_field);
669 
670 	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
671 				     start, offset, grc, nopartial);
672 
673 	skb->remcsum_offload = 1;
674 
675 	return vh;
676 }
677 
678 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
679 						  struct list_head *head,
680 						  struct sk_buff *skb,
681 						  struct gro_remcsum *grc)
682 {
683 	struct sk_buff *p;
684 	struct vxlanhdr *vh, *vh2;
685 	unsigned int hlen, off_vx;
686 	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
687 	__be32 flags;
688 
689 	skb_gro_remcsum_init(grc);
690 
691 	off_vx = skb_gro_offset(skb);
692 	hlen = off_vx + sizeof(*vh);
693 	vh = skb_gro_header(skb, hlen, off_vx);
694 	if (unlikely(!vh))
695 		return NULL;
696 
697 	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
698 
699 	flags = vh->vx_flags;
700 
701 	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
702 		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
703 				       vh->vx_vni, grc,
704 				       !!(vs->flags &
705 					  VXLAN_F_REMCSUM_NOPARTIAL));
706 
707 		if (!vh)
708 			return NULL;
709 	}
710 
711 	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
712 
713 	list_for_each_entry(p, head, list) {
714 		if (!NAPI_GRO_CB(p)->same_flow)
715 			continue;
716 
717 		vh2 = (struct vxlanhdr *)(p->data + off_vx);
718 		if (vh->vx_flags != vh2->vx_flags ||
719 		    vh->vx_vni != vh2->vx_vni) {
720 			NAPI_GRO_CB(p)->same_flow = 0;
721 			continue;
722 		}
723 	}
724 
725 	return vh;
726 }
727 
728 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
729 					 struct list_head *head,
730 					 struct sk_buff *skb)
731 {
732 	struct sk_buff *pp = NULL;
733 	struct gro_remcsum grc;
734 	int flush = 1;
735 
736 	if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
737 		pp = call_gro_receive(eth_gro_receive, head, skb);
738 		flush = 0;
739 	}
740 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
741 	return pp;
742 }
743 
744 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
745 					     struct list_head *head,
746 					     struct sk_buff *skb)
747 {
748 	const struct packet_offload *ptype;
749 	struct sk_buff *pp = NULL;
750 	struct gro_remcsum grc;
751 	struct vxlanhdr *vh;
752 	__be16 protocol;
753 	int flush = 1;
754 
755 	vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
756 	if (vh) {
757 		if (!vxlan_parse_gpe_proto(vh, &protocol))
758 			goto out;
759 		ptype = gro_find_receive_by_type(protocol);
760 		if (!ptype)
761 			goto out;
762 		pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
763 		flush = 0;
764 	}
765 out:
766 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
767 	return pp;
768 }
769 
770 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
771 {
772 	/* Sets 'skb->inner_mac_header' since we are always called with
773 	 * 'skb->encapsulation' set.
774 	 */
775 	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
776 }
777 
778 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
779 {
780 	struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
781 	const struct packet_offload *ptype;
782 	int err = -ENOSYS;
783 	__be16 protocol;
784 
785 	if (!vxlan_parse_gpe_proto(vh, &protocol))
786 		return err;
787 	ptype = gro_find_complete_by_type(protocol);
788 	if (ptype)
789 		err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
790 	return err;
791 }
792 
793 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
794 					 __u16 state, __be32 src_vni,
795 					 __u16 ndm_flags)
796 {
797 	struct vxlan_fdb *f;
798 
799 	f = kmalloc(sizeof(*f), GFP_ATOMIC);
800 	if (!f)
801 		return NULL;
802 	f->state = state;
803 	f->flags = ndm_flags;
804 	f->updated = f->used = jiffies;
805 	f->vni = src_vni;
806 	f->nh = NULL;
807 	RCU_INIT_POINTER(f->vdev, vxlan);
808 	INIT_LIST_HEAD(&f->nh_list);
809 	INIT_LIST_HEAD(&f->remotes);
810 	memcpy(f->eth_addr, mac, ETH_ALEN);
811 
812 	return f;
813 }
814 
815 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
816 			     __be32 src_vni, struct vxlan_fdb *f)
817 {
818 	++vxlan->addrcnt;
819 	hlist_add_head_rcu(&f->hlist,
820 			   vxlan_fdb_head(vxlan, mac, src_vni));
821 }
822 
823 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
824 			       u32 nhid, struct netlink_ext_ack *extack)
825 {
826 	struct nexthop *old_nh = rtnl_dereference(fdb->nh);
827 	struct nexthop *nh;
828 	int err = -EINVAL;
829 
830 	if (old_nh && old_nh->id == nhid)
831 		return 0;
832 
833 	nh = nexthop_find_by_id(vxlan->net, nhid);
834 	if (!nh) {
835 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
836 		goto err_inval;
837 	}
838 
839 	if (!nexthop_get(nh)) {
840 		NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
841 		nh = NULL;
842 		goto err_inval;
843 	}
844 	if (!nexthop_is_fdb(nh)) {
845 		NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
846 		goto err_inval;
847 	}
848 
849 	if (!nexthop_is_multipath(nh)) {
850 		NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
851 		goto err_inval;
852 	}
853 
854 	/* check nexthop group family */
855 	switch (vxlan->default_dst.remote_ip.sa.sa_family) {
856 	case AF_INET:
857 		if (!nexthop_has_v4(nh)) {
858 			err = -EAFNOSUPPORT;
859 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
860 			goto err_inval;
861 		}
862 		break;
863 	case AF_INET6:
864 		if (nexthop_has_v4(nh)) {
865 			err = -EAFNOSUPPORT;
866 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
867 			goto err_inval;
868 		}
869 	}
870 
871 	if (old_nh) {
872 		list_del_rcu(&fdb->nh_list);
873 		nexthop_put(old_nh);
874 	}
875 	rcu_assign_pointer(fdb->nh, nh);
876 	list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
877 	return 1;
878 
879 err_inval:
880 	if (nh)
881 		nexthop_put(nh);
882 	return err;
883 }
884 
885 int vxlan_fdb_create(struct vxlan_dev *vxlan,
886 		     const u8 *mac, union vxlan_addr *ip,
887 		     __u16 state, __be16 port, __be32 src_vni,
888 		     __be32 vni, __u32 ifindex, __u16 ndm_flags,
889 		     u32 nhid, struct vxlan_fdb **fdb,
890 		     struct netlink_ext_ack *extack)
891 {
892 	struct vxlan_rdst *rd = NULL;
893 	struct vxlan_fdb *f;
894 	int rc;
895 
896 	if (vxlan->cfg.addrmax &&
897 	    vxlan->addrcnt >= vxlan->cfg.addrmax)
898 		return -ENOSPC;
899 
900 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
901 	f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
902 	if (!f)
903 		return -ENOMEM;
904 
905 	if (nhid)
906 		rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
907 	else
908 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
909 	if (rc < 0)
910 		goto errout;
911 
912 	*fdb = f;
913 
914 	return 0;
915 
916 errout:
917 	kfree(f);
918 	return rc;
919 }
920 
921 static void __vxlan_fdb_free(struct vxlan_fdb *f)
922 {
923 	struct vxlan_rdst *rd, *nd;
924 	struct nexthop *nh;
925 
926 	nh = rcu_dereference_raw(f->nh);
927 	if (nh) {
928 		rcu_assign_pointer(f->nh, NULL);
929 		rcu_assign_pointer(f->vdev, NULL);
930 		nexthop_put(nh);
931 	}
932 
933 	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
934 		dst_cache_destroy(&rd->dst_cache);
935 		kfree(rd);
936 	}
937 	kfree(f);
938 }
939 
940 static void vxlan_fdb_free(struct rcu_head *head)
941 {
942 	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
943 
944 	__vxlan_fdb_free(f);
945 }
946 
947 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
948 			      bool do_notify, bool swdev_notify)
949 {
950 	struct vxlan_rdst *rd;
951 
952 	netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
953 
954 	--vxlan->addrcnt;
955 	if (do_notify) {
956 		if (rcu_access_pointer(f->nh))
957 			vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
958 					 swdev_notify, NULL);
959 		else
960 			list_for_each_entry(rd, &f->remotes, list)
961 				vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
962 						 swdev_notify, NULL);
963 	}
964 
965 	hlist_del_rcu(&f->hlist);
966 	list_del_rcu(&f->nh_list);
967 	call_rcu(&f->rcu, vxlan_fdb_free);
968 }
969 
970 static void vxlan_dst_free(struct rcu_head *head)
971 {
972 	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
973 
974 	dst_cache_destroy(&rd->dst_cache);
975 	kfree(rd);
976 }
977 
978 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
979 				     union vxlan_addr *ip,
980 				     __u16 state, __u16 flags,
981 				     __be16 port, __be32 vni,
982 				     __u32 ifindex, __u16 ndm_flags,
983 				     struct vxlan_fdb *f, u32 nhid,
984 				     bool swdev_notify,
985 				     struct netlink_ext_ack *extack)
986 {
987 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
988 	struct vxlan_rdst *rd = NULL;
989 	struct vxlan_rdst oldrd;
990 	int notify = 0;
991 	int rc = 0;
992 	int err;
993 
994 	if (nhid && !rcu_access_pointer(f->nh)) {
995 		NL_SET_ERR_MSG(extack,
996 			       "Cannot replace an existing non nexthop fdb with a nexthop");
997 		return -EOPNOTSUPP;
998 	}
999 
1000 	if (nhid && (flags & NLM_F_APPEND)) {
1001 		NL_SET_ERR_MSG(extack,
1002 			       "Cannot append to a nexthop fdb");
1003 		return -EOPNOTSUPP;
1004 	}
1005 
1006 	/* Do not allow an externally learned entry to take over an entry added
1007 	 * by the user.
1008 	 */
1009 	if (!(fdb_flags & NTF_EXT_LEARNED) ||
1010 	    !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1011 		if (f->state != state) {
1012 			f->state = state;
1013 			f->updated = jiffies;
1014 			notify = 1;
1015 		}
1016 		if (f->flags != fdb_flags) {
1017 			f->flags = fdb_flags;
1018 			f->updated = jiffies;
1019 			notify = 1;
1020 		}
1021 	}
1022 
1023 	if ((flags & NLM_F_REPLACE)) {
1024 		/* Only change unicasts */
1025 		if (!(is_multicast_ether_addr(f->eth_addr) ||
1026 		      is_zero_ether_addr(f->eth_addr))) {
1027 			if (nhid) {
1028 				rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1029 				if (rc < 0)
1030 					return rc;
1031 			} else {
1032 				rc = vxlan_fdb_replace(f, ip, port, vni,
1033 						       ifindex, &oldrd);
1034 			}
1035 			notify |= rc;
1036 		} else {
1037 			NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1038 			return -EOPNOTSUPP;
1039 		}
1040 	}
1041 	if ((flags & NLM_F_APPEND) &&
1042 	    (is_multicast_ether_addr(f->eth_addr) ||
1043 	     is_zero_ether_addr(f->eth_addr))) {
1044 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1045 
1046 		if (rc < 0)
1047 			return rc;
1048 		notify |= rc;
1049 	}
1050 
1051 	if (ndm_flags & NTF_USE)
1052 		f->used = jiffies;
1053 
1054 	if (notify) {
1055 		if (rd == NULL)
1056 			rd = first_remote_rtnl(f);
1057 
1058 		err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1059 				       swdev_notify, extack);
1060 		if (err)
1061 			goto err_notify;
1062 	}
1063 
1064 	return 0;
1065 
1066 err_notify:
1067 	if (nhid)
1068 		return err;
1069 	if ((flags & NLM_F_REPLACE) && rc)
1070 		*rd = oldrd;
1071 	else if ((flags & NLM_F_APPEND) && rc) {
1072 		list_del_rcu(&rd->list);
1073 		call_rcu(&rd->rcu, vxlan_dst_free);
1074 	}
1075 	return err;
1076 }
1077 
1078 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1079 				   const u8 *mac, union vxlan_addr *ip,
1080 				   __u16 state, __u16 flags,
1081 				   __be16 port, __be32 src_vni, __be32 vni,
1082 				   __u32 ifindex, __u16 ndm_flags, u32 nhid,
1083 				   bool swdev_notify,
1084 				   struct netlink_ext_ack *extack)
1085 {
1086 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1087 	struct vxlan_fdb *f;
1088 	int rc;
1089 
1090 	/* Disallow replace to add a multicast entry */
1091 	if ((flags & NLM_F_REPLACE) &&
1092 	    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1093 		return -EOPNOTSUPP;
1094 
1095 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1096 	rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1097 			      vni, ifindex, fdb_flags, nhid, &f, extack);
1098 	if (rc < 0)
1099 		return rc;
1100 
1101 	vxlan_fdb_insert(vxlan, mac, src_vni, f);
1102 	rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1103 			      swdev_notify, extack);
1104 	if (rc)
1105 		goto err_notify;
1106 
1107 	return 0;
1108 
1109 err_notify:
1110 	vxlan_fdb_destroy(vxlan, f, false, false);
1111 	return rc;
1112 }
1113 
1114 /* Add new entry to forwarding table -- assumes lock held */
1115 int vxlan_fdb_update(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,
1119 		     __u32 ifindex, __u16 ndm_flags, u32 nhid,
1120 		     bool swdev_notify,
1121 		     struct netlink_ext_ack *extack)
1122 {
1123 	struct vxlan_fdb *f;
1124 
1125 	f = __vxlan_find_mac(vxlan, mac, src_vni);
1126 	if (f) {
1127 		if (flags & NLM_F_EXCL) {
1128 			netdev_dbg(vxlan->dev,
1129 				   "lost race to create %pM\n", mac);
1130 			return -EEXIST;
1131 		}
1132 
1133 		return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1134 						 vni, ifindex, ndm_flags, f,
1135 						 nhid, swdev_notify, extack);
1136 	} else {
1137 		if (!(flags & NLM_F_CREATE))
1138 			return -ENOENT;
1139 
1140 		return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1141 					       port, src_vni, vni, ifindex,
1142 					       ndm_flags, nhid, swdev_notify,
1143 					       extack);
1144 	}
1145 }
1146 
1147 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1148 				  struct vxlan_rdst *rd, bool swdev_notify)
1149 {
1150 	list_del_rcu(&rd->list);
1151 	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1152 	call_rcu(&rd->rcu, vxlan_dst_free);
1153 }
1154 
1155 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1156 			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1157 			   __be32 *vni, u32 *ifindex, u32 *nhid,
1158 			   struct netlink_ext_ack *extack)
1159 {
1160 	struct net *net = dev_net(vxlan->dev);
1161 	int err;
1162 
1163 	if (tb[NDA_NH_ID] &&
1164 	    (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1165 		NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1166 		return -EINVAL;
1167 	}
1168 
1169 	if (tb[NDA_DST]) {
1170 		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1171 		if (err) {
1172 			NL_SET_ERR_MSG(extack, "Unsupported address family");
1173 			return err;
1174 		}
1175 	} else {
1176 		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1177 
1178 		if (remote->sa.sa_family == AF_INET) {
1179 			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1180 			ip->sa.sa_family = AF_INET;
1181 #if IS_ENABLED(CONFIG_IPV6)
1182 		} else {
1183 			ip->sin6.sin6_addr = in6addr_any;
1184 			ip->sa.sa_family = AF_INET6;
1185 #endif
1186 		}
1187 	}
1188 
1189 	if (tb[NDA_PORT]) {
1190 		if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1191 			NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1192 			return -EINVAL;
1193 		}
1194 		*port = nla_get_be16(tb[NDA_PORT]);
1195 	} else {
1196 		*port = vxlan->cfg.dst_port;
1197 	}
1198 
1199 	if (tb[NDA_VNI]) {
1200 		if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1201 			NL_SET_ERR_MSG(extack, "Invalid vni");
1202 			return -EINVAL;
1203 		}
1204 		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1205 	} else {
1206 		*vni = vxlan->default_dst.remote_vni;
1207 	}
1208 
1209 	if (tb[NDA_SRC_VNI]) {
1210 		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1211 			NL_SET_ERR_MSG(extack, "Invalid src vni");
1212 			return -EINVAL;
1213 		}
1214 		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1215 	} else {
1216 		*src_vni = vxlan->default_dst.remote_vni;
1217 	}
1218 
1219 	if (tb[NDA_IFINDEX]) {
1220 		struct net_device *tdev;
1221 
1222 		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1223 			NL_SET_ERR_MSG(extack, "Invalid ifindex");
1224 			return -EINVAL;
1225 		}
1226 		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1227 		tdev = __dev_get_by_index(net, *ifindex);
1228 		if (!tdev) {
1229 			NL_SET_ERR_MSG(extack, "Device not found");
1230 			return -EADDRNOTAVAIL;
1231 		}
1232 	} else {
1233 		*ifindex = 0;
1234 	}
1235 
1236 	if (tb[NDA_NH_ID])
1237 		*nhid = nla_get_u32(tb[NDA_NH_ID]);
1238 	else
1239 		*nhid = 0;
1240 
1241 	return 0;
1242 }
1243 
1244 /* Add static entry (via netlink) */
1245 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1246 			 struct net_device *dev,
1247 			 const unsigned char *addr, u16 vid, u16 flags,
1248 			 struct netlink_ext_ack *extack)
1249 {
1250 	struct vxlan_dev *vxlan = netdev_priv(dev);
1251 	/* struct net *net = dev_net(vxlan->dev); */
1252 	union vxlan_addr ip;
1253 	__be16 port;
1254 	__be32 src_vni, vni;
1255 	u32 ifindex, nhid;
1256 	u32 hash_index;
1257 	int err;
1258 
1259 	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1260 		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1261 			ndm->ndm_state);
1262 		return -EINVAL;
1263 	}
1264 
1265 	if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1266 		return -EINVAL;
1267 
1268 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1269 			      &nhid, extack);
1270 	if (err)
1271 		return err;
1272 
1273 	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1274 		return -EAFNOSUPPORT;
1275 
1276 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1277 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1278 	err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1279 			       port, src_vni, vni, ifindex,
1280 			       ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1281 			       nhid, true, extack);
1282 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1283 
1284 	return err;
1285 }
1286 
1287 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1288 		       const unsigned char *addr, union vxlan_addr ip,
1289 		       __be16 port, __be32 src_vni, __be32 vni,
1290 		       u32 ifindex, bool swdev_notify)
1291 {
1292 	struct vxlan_rdst *rd = NULL;
1293 	struct vxlan_fdb *f;
1294 	int err = -ENOENT;
1295 
1296 	f = vxlan_find_mac(vxlan, addr, src_vni);
1297 	if (!f)
1298 		return err;
1299 
1300 	if (!vxlan_addr_any(&ip)) {
1301 		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1302 		if (!rd)
1303 			goto out;
1304 	}
1305 
1306 	/* remove a destination if it's not the only one on the list,
1307 	 * otherwise destroy the fdb entry
1308 	 */
1309 	if (rd && !list_is_singular(&f->remotes)) {
1310 		vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1311 		goto out;
1312 	}
1313 
1314 	vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1315 
1316 out:
1317 	return 0;
1318 }
1319 
1320 /* Delete entry (via netlink) */
1321 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1322 			    struct net_device *dev,
1323 			    const unsigned char *addr, u16 vid,
1324 			    struct netlink_ext_ack *extack)
1325 {
1326 	struct vxlan_dev *vxlan = netdev_priv(dev);
1327 	union vxlan_addr ip;
1328 	__be32 src_vni, vni;
1329 	u32 ifindex, nhid;
1330 	u32 hash_index;
1331 	__be16 port;
1332 	int err;
1333 
1334 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1335 			      &nhid, extack);
1336 	if (err)
1337 		return err;
1338 
1339 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1340 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1341 	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1342 				 true);
1343 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1344 
1345 	return err;
1346 }
1347 
1348 /* Dump forwarding table */
1349 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1350 			  struct net_device *dev,
1351 			  struct net_device *filter_dev, int *idx)
1352 {
1353 	struct vxlan_dev *vxlan = netdev_priv(dev);
1354 	unsigned int h;
1355 	int err = 0;
1356 
1357 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
1358 		struct vxlan_fdb *f;
1359 
1360 		rcu_read_lock();
1361 		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1362 			struct vxlan_rdst *rd;
1363 
1364 			if (rcu_access_pointer(f->nh)) {
1365 				if (*idx < cb->args[2])
1366 					goto skip_nh;
1367 				err = vxlan_fdb_info(skb, vxlan, f,
1368 						     NETLINK_CB(cb->skb).portid,
1369 						     cb->nlh->nlmsg_seq,
1370 						     RTM_NEWNEIGH,
1371 						     NLM_F_MULTI, NULL);
1372 				if (err < 0) {
1373 					rcu_read_unlock();
1374 					goto out;
1375 				}
1376 skip_nh:
1377 				*idx += 1;
1378 				continue;
1379 			}
1380 
1381 			list_for_each_entry_rcu(rd, &f->remotes, list) {
1382 				if (*idx < cb->args[2])
1383 					goto skip;
1384 
1385 				err = vxlan_fdb_info(skb, vxlan, f,
1386 						     NETLINK_CB(cb->skb).portid,
1387 						     cb->nlh->nlmsg_seq,
1388 						     RTM_NEWNEIGH,
1389 						     NLM_F_MULTI, rd);
1390 				if (err < 0) {
1391 					rcu_read_unlock();
1392 					goto out;
1393 				}
1394 skip:
1395 				*idx += 1;
1396 			}
1397 		}
1398 		rcu_read_unlock();
1399 	}
1400 out:
1401 	return err;
1402 }
1403 
1404 static int vxlan_fdb_get(struct sk_buff *skb,
1405 			 struct nlattr *tb[],
1406 			 struct net_device *dev,
1407 			 const unsigned char *addr,
1408 			 u16 vid, u32 portid, u32 seq,
1409 			 struct netlink_ext_ack *extack)
1410 {
1411 	struct vxlan_dev *vxlan = netdev_priv(dev);
1412 	struct vxlan_fdb *f;
1413 	__be32 vni;
1414 	int err;
1415 
1416 	if (tb[NDA_VNI])
1417 		vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1418 	else
1419 		vni = vxlan->default_dst.remote_vni;
1420 
1421 	rcu_read_lock();
1422 
1423 	f = __vxlan_find_mac(vxlan, addr, vni);
1424 	if (!f) {
1425 		NL_SET_ERR_MSG(extack, "Fdb entry not found");
1426 		err = -ENOENT;
1427 		goto errout;
1428 	}
1429 
1430 	err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1431 			     RTM_NEWNEIGH, 0, first_remote_rcu(f));
1432 errout:
1433 	rcu_read_unlock();
1434 	return err;
1435 }
1436 
1437 /* Watch incoming packets to learn mapping between Ethernet address
1438  * and Tunnel endpoint.
1439  * Return true if packet is bogus and should be dropped.
1440  */
1441 static bool vxlan_snoop(struct net_device *dev,
1442 			union vxlan_addr *src_ip, const u8 *src_mac,
1443 			u32 src_ifindex, __be32 vni)
1444 {
1445 	struct vxlan_dev *vxlan = netdev_priv(dev);
1446 	struct vxlan_fdb *f;
1447 	u32 ifindex = 0;
1448 
1449 #if IS_ENABLED(CONFIG_IPV6)
1450 	if (src_ip->sa.sa_family == AF_INET6 &&
1451 	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1452 		ifindex = src_ifindex;
1453 #endif
1454 
1455 	f = vxlan_find_mac(vxlan, src_mac, vni);
1456 	if (likely(f)) {
1457 		struct vxlan_rdst *rdst = first_remote_rcu(f);
1458 
1459 		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1460 			   rdst->remote_ifindex == ifindex))
1461 			return false;
1462 
1463 		/* Don't migrate static entries, drop packets */
1464 		if (f->state & (NUD_PERMANENT | NUD_NOARP))
1465 			return true;
1466 
1467 		/* Don't override an fdb with nexthop with a learnt entry */
1468 		if (rcu_access_pointer(f->nh))
1469 			return true;
1470 
1471 		if (net_ratelimit())
1472 			netdev_info(dev,
1473 				    "%pM migrated from %pIS to %pIS\n",
1474 				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1475 
1476 		rdst->remote_ip = *src_ip;
1477 		f->updated = jiffies;
1478 		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1479 	} else {
1480 		u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1481 
1482 		/* learned new entry */
1483 		spin_lock(&vxlan->hash_lock[hash_index]);
1484 
1485 		/* close off race between vxlan_flush and incoming packets */
1486 		if (netif_running(dev))
1487 			vxlan_fdb_update(vxlan, src_mac, src_ip,
1488 					 NUD_REACHABLE,
1489 					 NLM_F_EXCL|NLM_F_CREATE,
1490 					 vxlan->cfg.dst_port,
1491 					 vni,
1492 					 vxlan->default_dst.remote_vni,
1493 					 ifindex, NTF_SELF, 0, true, NULL);
1494 		spin_unlock(&vxlan->hash_lock[hash_index]);
1495 	}
1496 
1497 	return false;
1498 }
1499 
1500 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1501 {
1502 	struct vxlan_net *vn;
1503 
1504 	if (!vs)
1505 		return false;
1506 	if (!refcount_dec_and_test(&vs->refcnt))
1507 		return false;
1508 
1509 	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1510 	spin_lock(&vn->sock_lock);
1511 	hlist_del_rcu(&vs->hlist);
1512 	udp_tunnel_notify_del_rx_port(vs->sock,
1513 				      (vs->flags & VXLAN_F_GPE) ?
1514 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1515 				      UDP_TUNNEL_TYPE_VXLAN);
1516 	spin_unlock(&vn->sock_lock);
1517 
1518 	return true;
1519 }
1520 
1521 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1522 {
1523 	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1524 #if IS_ENABLED(CONFIG_IPV6)
1525 	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1526 
1527 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1528 #endif
1529 
1530 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1531 	synchronize_net();
1532 
1533 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1534 		vxlan_vs_del_vnigrp(vxlan);
1535 	else
1536 		vxlan_vs_del_dev(vxlan);
1537 
1538 	if (__vxlan_sock_release_prep(sock4)) {
1539 		udp_tunnel_sock_release(sock4->sock);
1540 		kfree(sock4);
1541 	}
1542 
1543 #if IS_ENABLED(CONFIG_IPV6)
1544 	if (__vxlan_sock_release_prep(sock6)) {
1545 		udp_tunnel_sock_release(sock6->sock);
1546 		kfree(sock6);
1547 	}
1548 #endif
1549 }
1550 
1551 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1552 			  struct sk_buff *skb, u32 vxflags)
1553 {
1554 	size_t start, offset;
1555 
1556 	if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1557 		goto out;
1558 
1559 	start = vxlan_rco_start(unparsed->vx_vni);
1560 	offset = start + vxlan_rco_offset(unparsed->vx_vni);
1561 
1562 	if (!pskb_may_pull(skb, offset + sizeof(u16)))
1563 		return false;
1564 
1565 	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1566 			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1567 out:
1568 	unparsed->vx_flags &= ~VXLAN_HF_RCO;
1569 	unparsed->vx_vni &= VXLAN_VNI_MASK;
1570 	return true;
1571 }
1572 
1573 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1574 				struct sk_buff *skb, u32 vxflags,
1575 				struct vxlan_metadata *md)
1576 {
1577 	struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1578 	struct metadata_dst *tun_dst;
1579 
1580 	if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1581 		goto out;
1582 
1583 	md->gbp = ntohs(gbp->policy_id);
1584 
1585 	tun_dst = (struct metadata_dst *)skb_dst(skb);
1586 	if (tun_dst) {
1587 		tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1588 		tun_dst->u.tun_info.options_len = sizeof(*md);
1589 	}
1590 	if (gbp->dont_learn)
1591 		md->gbp |= VXLAN_GBP_DONT_LEARN;
1592 
1593 	if (gbp->policy_applied)
1594 		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1595 
1596 	/* In flow-based mode, GBP is carried in dst_metadata */
1597 	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1598 		skb->mark = md->gbp;
1599 out:
1600 	unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1601 }
1602 
1603 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1604 			  struct vxlan_sock *vs,
1605 			  struct sk_buff *skb, __be32 vni)
1606 {
1607 	union vxlan_addr saddr;
1608 	u32 ifindex = skb->dev->ifindex;
1609 
1610 	skb_reset_mac_header(skb);
1611 	skb->protocol = eth_type_trans(skb, vxlan->dev);
1612 	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1613 
1614 	/* Ignore packet loops (and multicast echo) */
1615 	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1616 		return false;
1617 
1618 	/* Get address from the outer IP header */
1619 	if (vxlan_get_sk_family(vs) == AF_INET) {
1620 		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1621 		saddr.sa.sa_family = AF_INET;
1622 #if IS_ENABLED(CONFIG_IPV6)
1623 	} else {
1624 		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1625 		saddr.sa.sa_family = AF_INET6;
1626 #endif
1627 	}
1628 
1629 	if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1630 	    vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1631 		return false;
1632 
1633 	return true;
1634 }
1635 
1636 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1637 				  struct sk_buff *skb)
1638 {
1639 	int err = 0;
1640 
1641 	if (vxlan_get_sk_family(vs) == AF_INET)
1642 		err = IP_ECN_decapsulate(oiph, skb);
1643 #if IS_ENABLED(CONFIG_IPV6)
1644 	else
1645 		err = IP6_ECN_decapsulate(oiph, skb);
1646 #endif
1647 
1648 	if (unlikely(err) && log_ecn_error) {
1649 		if (vxlan_get_sk_family(vs) == AF_INET)
1650 			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1651 					     &((struct iphdr *)oiph)->saddr,
1652 					     ((struct iphdr *)oiph)->tos);
1653 		else
1654 			net_info_ratelimited("non-ECT from %pI6\n",
1655 					     &((struct ipv6hdr *)oiph)->saddr);
1656 	}
1657 	return err <= 1;
1658 }
1659 
1660 /* Callback from net/ipv4/udp.c to receive packets */
1661 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1662 {
1663 	struct vxlan_vni_node *vninode = NULL;
1664 	struct vxlan_dev *vxlan;
1665 	struct vxlan_sock *vs;
1666 	struct vxlanhdr unparsed;
1667 	struct vxlan_metadata _md;
1668 	struct vxlan_metadata *md = &_md;
1669 	__be16 protocol = htons(ETH_P_TEB);
1670 	bool raw_proto = false;
1671 	void *oiph;
1672 	__be32 vni = 0;
1673 
1674 	/* Need UDP and VXLAN header to be present */
1675 	if (!pskb_may_pull(skb, VXLAN_HLEN))
1676 		goto drop;
1677 
1678 	unparsed = *vxlan_hdr(skb);
1679 	/* VNI flag always required to be set */
1680 	if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1681 		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1682 			   ntohl(vxlan_hdr(skb)->vx_flags),
1683 			   ntohl(vxlan_hdr(skb)->vx_vni));
1684 		/* Return non vxlan pkt */
1685 		goto drop;
1686 	}
1687 	unparsed.vx_flags &= ~VXLAN_HF_VNI;
1688 	unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1689 
1690 	vs = rcu_dereference_sk_user_data(sk);
1691 	if (!vs)
1692 		goto drop;
1693 
1694 	vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1695 
1696 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1697 	if (!vxlan)
1698 		goto drop;
1699 
1700 	/* For backwards compatibility, only allow reserved fields to be
1701 	 * used by VXLAN extensions if explicitly requested.
1702 	 */
1703 	if (vs->flags & VXLAN_F_GPE) {
1704 		if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1705 			goto drop;
1706 		unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1707 		raw_proto = true;
1708 	}
1709 
1710 	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1711 				   !net_eq(vxlan->net, dev_net(vxlan->dev))))
1712 		goto drop;
1713 
1714 	if (vs->flags & VXLAN_F_REMCSUM_RX)
1715 		if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1716 			goto drop;
1717 
1718 	if (vxlan_collect_metadata(vs)) {
1719 		struct metadata_dst *tun_dst;
1720 
1721 		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1722 					 key32_to_tunnel_id(vni), sizeof(*md));
1723 
1724 		if (!tun_dst)
1725 			goto drop;
1726 
1727 		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1728 
1729 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1730 	} else {
1731 		memset(md, 0, sizeof(*md));
1732 	}
1733 
1734 	if (vs->flags & VXLAN_F_GBP)
1735 		vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1736 	/* Note that GBP and GPE can never be active together. This is
1737 	 * ensured in vxlan_dev_configure.
1738 	 */
1739 
1740 	if (unparsed.vx_flags || unparsed.vx_vni) {
1741 		/* If there are any unprocessed flags remaining treat
1742 		 * this as a malformed packet. This behavior diverges from
1743 		 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1744 		 * in reserved fields are to be ignored. The approach here
1745 		 * maintains compatibility with previous stack code, and also
1746 		 * is more robust and provides a little more security in
1747 		 * adding extensions to VXLAN.
1748 		 */
1749 		goto drop;
1750 	}
1751 
1752 	if (!raw_proto) {
1753 		if (!vxlan_set_mac(vxlan, vs, skb, vni))
1754 			goto drop;
1755 	} else {
1756 		skb_reset_mac_header(skb);
1757 		skb->dev = vxlan->dev;
1758 		skb->pkt_type = PACKET_HOST;
1759 	}
1760 
1761 	oiph = skb_network_header(skb);
1762 	skb_reset_network_header(skb);
1763 
1764 	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1765 		++vxlan->dev->stats.rx_frame_errors;
1766 		++vxlan->dev->stats.rx_errors;
1767 		vxlan_vnifilter_count(vxlan, vni, vninode,
1768 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1769 		goto drop;
1770 	}
1771 
1772 	rcu_read_lock();
1773 
1774 	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1775 		rcu_read_unlock();
1776 		dev_core_stats_rx_dropped_inc(vxlan->dev);
1777 		vxlan_vnifilter_count(vxlan, vni, vninode,
1778 				      VXLAN_VNI_STATS_RX_DROPS, 0);
1779 		goto drop;
1780 	}
1781 
1782 	dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1783 	vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1784 	gro_cells_receive(&vxlan->gro_cells, skb);
1785 
1786 	rcu_read_unlock();
1787 
1788 	return 0;
1789 
1790 drop:
1791 	/* Consume bad packet */
1792 	kfree_skb(skb);
1793 	return 0;
1794 }
1795 
1796 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1797 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1798 {
1799 	struct vxlan_dev *vxlan;
1800 	struct vxlan_sock *vs;
1801 	struct vxlanhdr *hdr;
1802 	__be32 vni;
1803 
1804 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1805 		return -EINVAL;
1806 
1807 	hdr = vxlan_hdr(skb);
1808 
1809 	if (!(hdr->vx_flags & VXLAN_HF_VNI))
1810 		return -EINVAL;
1811 
1812 	vs = rcu_dereference_sk_user_data(sk);
1813 	if (!vs)
1814 		return -ENOENT;
1815 
1816 	vni = vxlan_vni(hdr->vx_vni);
1817 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1818 	if (!vxlan)
1819 		return -ENOENT;
1820 
1821 	return 0;
1822 }
1823 
1824 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1825 {
1826 	struct vxlan_dev *vxlan = netdev_priv(dev);
1827 	struct arphdr *parp;
1828 	u8 *arpptr, *sha;
1829 	__be32 sip, tip;
1830 	struct neighbour *n;
1831 
1832 	if (dev->flags & IFF_NOARP)
1833 		goto out;
1834 
1835 	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1836 		dev->stats.tx_dropped++;
1837 		goto out;
1838 	}
1839 	parp = arp_hdr(skb);
1840 
1841 	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1842 	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1843 	    parp->ar_pro != htons(ETH_P_IP) ||
1844 	    parp->ar_op != htons(ARPOP_REQUEST) ||
1845 	    parp->ar_hln != dev->addr_len ||
1846 	    parp->ar_pln != 4)
1847 		goto out;
1848 	arpptr = (u8 *)parp + sizeof(struct arphdr);
1849 	sha = arpptr;
1850 	arpptr += dev->addr_len;	/* sha */
1851 	memcpy(&sip, arpptr, sizeof(sip));
1852 	arpptr += sizeof(sip);
1853 	arpptr += dev->addr_len;	/* tha */
1854 	memcpy(&tip, arpptr, sizeof(tip));
1855 
1856 	if (ipv4_is_loopback(tip) ||
1857 	    ipv4_is_multicast(tip))
1858 		goto out;
1859 
1860 	n = neigh_lookup(&arp_tbl, &tip, dev);
1861 
1862 	if (n) {
1863 		struct vxlan_fdb *f;
1864 		struct sk_buff	*reply;
1865 
1866 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1867 			neigh_release(n);
1868 			goto out;
1869 		}
1870 
1871 		f = vxlan_find_mac(vxlan, n->ha, vni);
1872 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1873 			/* bridge-local neighbor */
1874 			neigh_release(n);
1875 			goto out;
1876 		}
1877 
1878 		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1879 				n->ha, sha);
1880 
1881 		neigh_release(n);
1882 
1883 		if (reply == NULL)
1884 			goto out;
1885 
1886 		skb_reset_mac_header(reply);
1887 		__skb_pull(reply, skb_network_offset(reply));
1888 		reply->ip_summed = CHECKSUM_UNNECESSARY;
1889 		reply->pkt_type = PACKET_HOST;
1890 
1891 		if (netif_rx(reply) == NET_RX_DROP) {
1892 			dev->stats.rx_dropped++;
1893 			vxlan_vnifilter_count(vxlan, vni, NULL,
1894 					      VXLAN_VNI_STATS_RX_DROPS, 0);
1895 		}
1896 
1897 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1898 		union vxlan_addr ipa = {
1899 			.sin.sin_addr.s_addr = tip,
1900 			.sin.sin_family = AF_INET,
1901 		};
1902 
1903 		vxlan_ip_miss(dev, &ipa);
1904 	}
1905 out:
1906 	consume_skb(skb);
1907 	return NETDEV_TX_OK;
1908 }
1909 
1910 #if IS_ENABLED(CONFIG_IPV6)
1911 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1912 	struct neighbour *n, bool isrouter)
1913 {
1914 	struct net_device *dev = request->dev;
1915 	struct sk_buff *reply;
1916 	struct nd_msg *ns, *na;
1917 	struct ipv6hdr *pip6;
1918 	u8 *daddr;
1919 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1920 	int ns_olen;
1921 	int i, len;
1922 
1923 	if (dev == NULL || !pskb_may_pull(request, request->len))
1924 		return NULL;
1925 
1926 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1927 		sizeof(*na) + na_olen + dev->needed_tailroom;
1928 	reply = alloc_skb(len, GFP_ATOMIC);
1929 	if (reply == NULL)
1930 		return NULL;
1931 
1932 	reply->protocol = htons(ETH_P_IPV6);
1933 	reply->dev = dev;
1934 	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1935 	skb_push(reply, sizeof(struct ethhdr));
1936 	skb_reset_mac_header(reply);
1937 
1938 	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1939 
1940 	daddr = eth_hdr(request)->h_source;
1941 	ns_olen = request->len - skb_network_offset(request) -
1942 		sizeof(struct ipv6hdr) - sizeof(*ns);
1943 	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1944 		if (!ns->opt[i + 1]) {
1945 			kfree_skb(reply);
1946 			return NULL;
1947 		}
1948 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1949 			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1950 			break;
1951 		}
1952 	}
1953 
1954 	/* Ethernet header */
1955 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1956 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1957 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1958 	reply->protocol = htons(ETH_P_IPV6);
1959 
1960 	skb_pull(reply, sizeof(struct ethhdr));
1961 	skb_reset_network_header(reply);
1962 	skb_put(reply, sizeof(struct ipv6hdr));
1963 
1964 	/* IPv6 header */
1965 
1966 	pip6 = ipv6_hdr(reply);
1967 	memset(pip6, 0, sizeof(struct ipv6hdr));
1968 	pip6->version = 6;
1969 	pip6->priority = ipv6_hdr(request)->priority;
1970 	pip6->nexthdr = IPPROTO_ICMPV6;
1971 	pip6->hop_limit = 255;
1972 	pip6->daddr = ipv6_hdr(request)->saddr;
1973 	pip6->saddr = *(struct in6_addr *)n->primary_key;
1974 
1975 	skb_pull(reply, sizeof(struct ipv6hdr));
1976 	skb_reset_transport_header(reply);
1977 
1978 	/* Neighbor Advertisement */
1979 	na = skb_put_zero(reply, sizeof(*na) + na_olen);
1980 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1981 	na->icmph.icmp6_router = isrouter;
1982 	na->icmph.icmp6_override = 1;
1983 	na->icmph.icmp6_solicited = 1;
1984 	na->target = ns->target;
1985 	ether_addr_copy(&na->opt[2], n->ha);
1986 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1987 	na->opt[1] = na_olen >> 3;
1988 
1989 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1990 		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1991 		csum_partial(na, sizeof(*na)+na_olen, 0));
1992 
1993 	pip6->payload_len = htons(sizeof(*na)+na_olen);
1994 
1995 	skb_push(reply, sizeof(struct ipv6hdr));
1996 
1997 	reply->ip_summed = CHECKSUM_UNNECESSARY;
1998 
1999 	return reply;
2000 }
2001 
2002 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2003 {
2004 	struct vxlan_dev *vxlan = netdev_priv(dev);
2005 	const struct in6_addr *daddr;
2006 	const struct ipv6hdr *iphdr;
2007 	struct inet6_dev *in6_dev;
2008 	struct neighbour *n;
2009 	struct nd_msg *msg;
2010 
2011 	rcu_read_lock();
2012 	in6_dev = __in6_dev_get(dev);
2013 	if (!in6_dev)
2014 		goto out;
2015 
2016 	iphdr = ipv6_hdr(skb);
2017 	daddr = &iphdr->daddr;
2018 	msg = (struct nd_msg *)(iphdr + 1);
2019 
2020 	if (ipv6_addr_loopback(daddr) ||
2021 	    ipv6_addr_is_multicast(&msg->target))
2022 		goto out;
2023 
2024 	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2025 
2026 	if (n) {
2027 		struct vxlan_fdb *f;
2028 		struct sk_buff *reply;
2029 
2030 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2031 			neigh_release(n);
2032 			goto out;
2033 		}
2034 
2035 		f = vxlan_find_mac(vxlan, n->ha, vni);
2036 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2037 			/* bridge-local neighbor */
2038 			neigh_release(n);
2039 			goto out;
2040 		}
2041 
2042 		reply = vxlan_na_create(skb, n,
2043 					!!(f ? f->flags & NTF_ROUTER : 0));
2044 
2045 		neigh_release(n);
2046 
2047 		if (reply == NULL)
2048 			goto out;
2049 
2050 		if (netif_rx(reply) == NET_RX_DROP) {
2051 			dev->stats.rx_dropped++;
2052 			vxlan_vnifilter_count(vxlan, vni, NULL,
2053 					      VXLAN_VNI_STATS_RX_DROPS, 0);
2054 		}
2055 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2056 		union vxlan_addr ipa = {
2057 			.sin6.sin6_addr = msg->target,
2058 			.sin6.sin6_family = AF_INET6,
2059 		};
2060 
2061 		vxlan_ip_miss(dev, &ipa);
2062 	}
2063 
2064 out:
2065 	rcu_read_unlock();
2066 	consume_skb(skb);
2067 	return NETDEV_TX_OK;
2068 }
2069 #endif
2070 
2071 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2072 {
2073 	struct vxlan_dev *vxlan = netdev_priv(dev);
2074 	struct neighbour *n;
2075 
2076 	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2077 		return false;
2078 
2079 	n = NULL;
2080 	switch (ntohs(eth_hdr(skb)->h_proto)) {
2081 	case ETH_P_IP:
2082 	{
2083 		struct iphdr *pip;
2084 
2085 		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2086 			return false;
2087 		pip = ip_hdr(skb);
2088 		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2089 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2090 			union vxlan_addr ipa = {
2091 				.sin.sin_addr.s_addr = pip->daddr,
2092 				.sin.sin_family = AF_INET,
2093 			};
2094 
2095 			vxlan_ip_miss(dev, &ipa);
2096 			return false;
2097 		}
2098 
2099 		break;
2100 	}
2101 #if IS_ENABLED(CONFIG_IPV6)
2102 	case ETH_P_IPV6:
2103 	{
2104 		struct ipv6hdr *pip6;
2105 
2106 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2107 			return false;
2108 		pip6 = ipv6_hdr(skb);
2109 		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2110 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2111 			union vxlan_addr ipa = {
2112 				.sin6.sin6_addr = pip6->daddr,
2113 				.sin6.sin6_family = AF_INET6,
2114 			};
2115 
2116 			vxlan_ip_miss(dev, &ipa);
2117 			return false;
2118 		}
2119 
2120 		break;
2121 	}
2122 #endif
2123 	default:
2124 		return false;
2125 	}
2126 
2127 	if (n) {
2128 		bool diff;
2129 
2130 		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2131 		if (diff) {
2132 			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2133 				dev->addr_len);
2134 			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2135 		}
2136 		neigh_release(n);
2137 		return diff;
2138 	}
2139 
2140 	return false;
2141 }
2142 
2143 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2144 {
2145 	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2146 
2147 	gpe->np_applied = 1;
2148 	gpe->next_protocol = tun_p_from_eth_p(protocol);
2149 	if (!gpe->next_protocol)
2150 		return -EPFNOSUPPORT;
2151 	return 0;
2152 }
2153 
2154 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2155 			   int iphdr_len, __be32 vni,
2156 			   struct vxlan_metadata *md, u32 vxflags,
2157 			   bool udp_sum)
2158 {
2159 	struct vxlanhdr *vxh;
2160 	int min_headroom;
2161 	int err;
2162 	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2163 	__be16 inner_protocol = htons(ETH_P_TEB);
2164 
2165 	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2166 	    skb->ip_summed == CHECKSUM_PARTIAL) {
2167 		int csum_start = skb_checksum_start_offset(skb);
2168 
2169 		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2170 		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2171 		    (skb->csum_offset == offsetof(struct udphdr, check) ||
2172 		     skb->csum_offset == offsetof(struct tcphdr, check)))
2173 			type |= SKB_GSO_TUNNEL_REMCSUM;
2174 	}
2175 
2176 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2177 			+ VXLAN_HLEN + iphdr_len;
2178 
2179 	/* Need space for new headers (invalidates iph ptr) */
2180 	err = skb_cow_head(skb, min_headroom);
2181 	if (unlikely(err))
2182 		return err;
2183 
2184 	err = iptunnel_handle_offloads(skb, type);
2185 	if (err)
2186 		return err;
2187 
2188 	vxh = __skb_push(skb, sizeof(*vxh));
2189 	vxh->vx_flags = VXLAN_HF_VNI;
2190 	vxh->vx_vni = vxlan_vni_field(vni);
2191 
2192 	if (type & SKB_GSO_TUNNEL_REMCSUM) {
2193 		unsigned int start;
2194 
2195 		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2196 		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2197 		vxh->vx_flags |= VXLAN_HF_RCO;
2198 
2199 		if (!skb_is_gso(skb)) {
2200 			skb->ip_summed = CHECKSUM_NONE;
2201 			skb->encapsulation = 0;
2202 		}
2203 	}
2204 
2205 	if (vxflags & VXLAN_F_GBP)
2206 		vxlan_build_gbp_hdr(vxh, md);
2207 	if (vxflags & VXLAN_F_GPE) {
2208 		err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2209 		if (err < 0)
2210 			return err;
2211 		inner_protocol = skb->protocol;
2212 	}
2213 
2214 	skb_set_inner_protocol(skb, inner_protocol);
2215 	return 0;
2216 }
2217 
2218 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2219 				      struct vxlan_sock *sock4,
2220 				      struct sk_buff *skb, int oif, u8 tos,
2221 				      __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2222 				      __u8 flow_flags, struct dst_cache *dst_cache,
2223 				      const struct ip_tunnel_info *info)
2224 {
2225 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2226 	struct rtable *rt = NULL;
2227 	struct flowi4 fl4;
2228 
2229 	if (!sock4)
2230 		return ERR_PTR(-EIO);
2231 
2232 	if (tos && !info)
2233 		use_cache = false;
2234 	if (use_cache) {
2235 		rt = dst_cache_get_ip4(dst_cache, saddr);
2236 		if (rt)
2237 			return rt;
2238 	}
2239 
2240 	memset(&fl4, 0, sizeof(fl4));
2241 	fl4.flowi4_oif = oif;
2242 	fl4.flowi4_tos = RT_TOS(tos);
2243 	fl4.flowi4_mark = skb->mark;
2244 	fl4.flowi4_proto = IPPROTO_UDP;
2245 	fl4.daddr = daddr;
2246 	fl4.saddr = *saddr;
2247 	fl4.fl4_dport = dport;
2248 	fl4.fl4_sport = sport;
2249 	fl4.flowi4_flags = flow_flags;
2250 
2251 	rt = ip_route_output_key(vxlan->net, &fl4);
2252 	if (!IS_ERR(rt)) {
2253 		if (rt->dst.dev == dev) {
2254 			netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2255 			ip_rt_put(rt);
2256 			return ERR_PTR(-ELOOP);
2257 		}
2258 
2259 		*saddr = fl4.saddr;
2260 		if (use_cache)
2261 			dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2262 	} else {
2263 		netdev_dbg(dev, "no route to %pI4\n", &daddr);
2264 		return ERR_PTR(-ENETUNREACH);
2265 	}
2266 	return rt;
2267 }
2268 
2269 #if IS_ENABLED(CONFIG_IPV6)
2270 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2271 					  struct net_device *dev,
2272 					  struct vxlan_sock *sock6,
2273 					  struct sk_buff *skb, int oif, u8 tos,
2274 					  __be32 label,
2275 					  const struct in6_addr *daddr,
2276 					  struct in6_addr *saddr,
2277 					  __be16 dport, __be16 sport,
2278 					  struct dst_cache *dst_cache,
2279 					  const struct ip_tunnel_info *info)
2280 {
2281 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2282 	struct dst_entry *ndst;
2283 	struct flowi6 fl6;
2284 
2285 	if (!sock6)
2286 		return ERR_PTR(-EIO);
2287 
2288 	if (tos && !info)
2289 		use_cache = false;
2290 	if (use_cache) {
2291 		ndst = dst_cache_get_ip6(dst_cache, saddr);
2292 		if (ndst)
2293 			return ndst;
2294 	}
2295 
2296 	memset(&fl6, 0, sizeof(fl6));
2297 	fl6.flowi6_oif = oif;
2298 	fl6.daddr = *daddr;
2299 	fl6.saddr = *saddr;
2300 	fl6.flowlabel = ip6_make_flowinfo(tos, label);
2301 	fl6.flowi6_mark = skb->mark;
2302 	fl6.flowi6_proto = IPPROTO_UDP;
2303 	fl6.fl6_dport = dport;
2304 	fl6.fl6_sport = sport;
2305 
2306 	ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2307 					       &fl6, NULL);
2308 	if (IS_ERR(ndst)) {
2309 		netdev_dbg(dev, "no route to %pI6\n", daddr);
2310 		return ERR_PTR(-ENETUNREACH);
2311 	}
2312 
2313 	if (unlikely(ndst->dev == dev)) {
2314 		netdev_dbg(dev, "circular route to %pI6\n", daddr);
2315 		dst_release(ndst);
2316 		return ERR_PTR(-ELOOP);
2317 	}
2318 
2319 	*saddr = fl6.saddr;
2320 	if (use_cache)
2321 		dst_cache_set_ip6(dst_cache, ndst, saddr);
2322 	return ndst;
2323 }
2324 #endif
2325 
2326 /* Bypass encapsulation if the destination is local */
2327 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2328 			       struct vxlan_dev *dst_vxlan, __be32 vni,
2329 			       bool snoop)
2330 {
2331 	union vxlan_addr loopback;
2332 	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2333 	struct net_device *dev;
2334 	int len = skb->len;
2335 
2336 	skb->pkt_type = PACKET_HOST;
2337 	skb->encapsulation = 0;
2338 	skb->dev = dst_vxlan->dev;
2339 	__skb_pull(skb, skb_network_offset(skb));
2340 
2341 	if (remote_ip->sa.sa_family == AF_INET) {
2342 		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2343 		loopback.sa.sa_family =  AF_INET;
2344 #if IS_ENABLED(CONFIG_IPV6)
2345 	} else {
2346 		loopback.sin6.sin6_addr = in6addr_loopback;
2347 		loopback.sa.sa_family =  AF_INET6;
2348 #endif
2349 	}
2350 
2351 	rcu_read_lock();
2352 	dev = skb->dev;
2353 	if (unlikely(!(dev->flags & IFF_UP))) {
2354 		kfree_skb(skb);
2355 		goto drop;
2356 	}
2357 
2358 	if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2359 		vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2360 
2361 	dev_sw_netstats_tx_add(src_vxlan->dev, 1, len);
2362 	vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2363 
2364 	if (__netif_rx(skb) == NET_RX_SUCCESS) {
2365 		dev_sw_netstats_rx_add(dst_vxlan->dev, len);
2366 		vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2367 				      len);
2368 	} else {
2369 drop:
2370 		dev->stats.rx_dropped++;
2371 		vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2372 				      VXLAN_VNI_STATS_RX_DROPS, 0);
2373 	}
2374 	rcu_read_unlock();
2375 }
2376 
2377 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2378 				 struct vxlan_dev *vxlan,
2379 				 union vxlan_addr *daddr,
2380 				 __be16 dst_port, int dst_ifindex, __be32 vni,
2381 				 struct dst_entry *dst,
2382 				 u32 rt_flags)
2383 {
2384 #if IS_ENABLED(CONFIG_IPV6)
2385 	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2386 	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2387 	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2388 	 */
2389 	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2390 #endif
2391 	/* Bypass encapsulation if the destination is local */
2392 	if (rt_flags & RTCF_LOCAL &&
2393 	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2394 	    vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2395 		struct vxlan_dev *dst_vxlan;
2396 
2397 		dst_release(dst);
2398 		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2399 					   daddr->sa.sa_family, dst_port,
2400 					   vxlan->cfg.flags);
2401 		if (!dst_vxlan) {
2402 			dev->stats.tx_errors++;
2403 			vxlan_vnifilter_count(vxlan, vni, NULL,
2404 					      VXLAN_VNI_STATS_TX_ERRORS, 0);
2405 			kfree_skb(skb);
2406 
2407 			return -ENOENT;
2408 		}
2409 		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2410 		return 1;
2411 	}
2412 
2413 	return 0;
2414 }
2415 
2416 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2417 		    __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2418 {
2419 	struct dst_cache *dst_cache;
2420 	struct ip_tunnel_info *info;
2421 	struct vxlan_dev *vxlan = netdev_priv(dev);
2422 	const struct iphdr *old_iph = ip_hdr(skb);
2423 	union vxlan_addr *dst;
2424 	union vxlan_addr remote_ip, local_ip;
2425 	struct vxlan_metadata _md;
2426 	struct vxlan_metadata *md = &_md;
2427 	unsigned int pkt_len = skb->len;
2428 	__be16 src_port = 0, dst_port;
2429 	struct dst_entry *ndst = NULL;
2430 	__u8 tos, ttl, flow_flags = 0;
2431 	int ifindex;
2432 	int err;
2433 	u32 flags = vxlan->cfg.flags;
2434 	bool udp_sum = false;
2435 	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2436 	__be32 vni = 0;
2437 #if IS_ENABLED(CONFIG_IPV6)
2438 	__be32 label;
2439 #endif
2440 
2441 	info = skb_tunnel_info(skb);
2442 
2443 	if (rdst) {
2444 		dst = &rdst->remote_ip;
2445 		if (vxlan_addr_any(dst)) {
2446 			if (did_rsc) {
2447 				/* short-circuited back to local bridge */
2448 				vxlan_encap_bypass(skb, vxlan, vxlan,
2449 						   default_vni, true);
2450 				return;
2451 			}
2452 			goto drop;
2453 		}
2454 
2455 		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2456 		vni = (rdst->remote_vni) ? : default_vni;
2457 		ifindex = rdst->remote_ifindex;
2458 		local_ip = vxlan->cfg.saddr;
2459 		dst_cache = &rdst->dst_cache;
2460 		md->gbp = skb->mark;
2461 		if (flags & VXLAN_F_TTL_INHERIT) {
2462 			ttl = ip_tunnel_get_ttl(old_iph, skb);
2463 		} else {
2464 			ttl = vxlan->cfg.ttl;
2465 			if (!ttl && vxlan_addr_multicast(dst))
2466 				ttl = 1;
2467 		}
2468 
2469 		tos = vxlan->cfg.tos;
2470 		if (tos == 1)
2471 			tos = ip_tunnel_get_dsfield(old_iph, skb);
2472 
2473 		if (dst->sa.sa_family == AF_INET)
2474 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2475 		else
2476 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2477 #if IS_ENABLED(CONFIG_IPV6)
2478 		label = vxlan->cfg.label;
2479 #endif
2480 	} else {
2481 		if (!info) {
2482 			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2483 				  dev->name);
2484 			goto drop;
2485 		}
2486 		remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2487 		if (remote_ip.sa.sa_family == AF_INET) {
2488 			remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2489 			local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2490 		} else {
2491 			remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2492 			local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2493 		}
2494 		dst = &remote_ip;
2495 		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2496 		flow_flags = info->key.flow_flags;
2497 		vni = tunnel_id_to_key32(info->key.tun_id);
2498 		ifindex = 0;
2499 		dst_cache = &info->dst_cache;
2500 		if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2501 			if (info->options_len < sizeof(*md))
2502 				goto drop;
2503 			md = ip_tunnel_info_opts(info);
2504 		}
2505 		ttl = info->key.ttl;
2506 		tos = info->key.tos;
2507 #if IS_ENABLED(CONFIG_IPV6)
2508 		label = info->key.label;
2509 #endif
2510 		udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2511 	}
2512 	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2513 				     vxlan->cfg.port_max, true);
2514 
2515 	rcu_read_lock();
2516 	if (dst->sa.sa_family == AF_INET) {
2517 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2518 		struct rtable *rt;
2519 		__be16 df = 0;
2520 
2521 		if (!ifindex)
2522 			ifindex = sock4->sock->sk->sk_bound_dev_if;
2523 
2524 		rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2525 				     dst->sin.sin_addr.s_addr,
2526 				     &local_ip.sin.sin_addr.s_addr,
2527 				     dst_port, src_port, flow_flags,
2528 				     dst_cache, info);
2529 		if (IS_ERR(rt)) {
2530 			err = PTR_ERR(rt);
2531 			goto tx_error;
2532 		}
2533 
2534 		if (!info) {
2535 			/* Bypass encapsulation if the destination is local */
2536 			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2537 						    dst_port, ifindex, vni,
2538 						    &rt->dst, rt->rt_flags);
2539 			if (err)
2540 				goto out_unlock;
2541 
2542 			if (vxlan->cfg.df == VXLAN_DF_SET) {
2543 				df = htons(IP_DF);
2544 			} else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2545 				struct ethhdr *eth = eth_hdr(skb);
2546 
2547 				if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2548 				    (ntohs(eth->h_proto) == ETH_P_IP &&
2549 				     old_iph->frag_off & htons(IP_DF)))
2550 					df = htons(IP_DF);
2551 			}
2552 		} else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2553 			df = htons(IP_DF);
2554 		}
2555 
2556 		ndst = &rt->dst;
2557 		err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2558 					    netif_is_any_bridge_port(dev));
2559 		if (err < 0) {
2560 			goto tx_error;
2561 		} else if (err) {
2562 			if (info) {
2563 				struct ip_tunnel_info *unclone;
2564 				struct in_addr src, dst;
2565 
2566 				unclone = skb_tunnel_info_unclone(skb);
2567 				if (unlikely(!unclone))
2568 					goto tx_error;
2569 
2570 				src = remote_ip.sin.sin_addr;
2571 				dst = local_ip.sin.sin_addr;
2572 				unclone->key.u.ipv4.src = src.s_addr;
2573 				unclone->key.u.ipv4.dst = dst.s_addr;
2574 			}
2575 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2576 			dst_release(ndst);
2577 			goto out_unlock;
2578 		}
2579 
2580 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2581 		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2582 		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2583 				      vni, md, flags, udp_sum);
2584 		if (err < 0)
2585 			goto tx_error;
2586 
2587 		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2588 				    dst->sin.sin_addr.s_addr, tos, ttl, df,
2589 				    src_port, dst_port, xnet, !udp_sum);
2590 #if IS_ENABLED(CONFIG_IPV6)
2591 	} else {
2592 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2593 
2594 		if (!ifindex)
2595 			ifindex = sock6->sock->sk->sk_bound_dev_if;
2596 
2597 		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2598 					label, &dst->sin6.sin6_addr,
2599 					&local_ip.sin6.sin6_addr,
2600 					dst_port, src_port,
2601 					dst_cache, info);
2602 		if (IS_ERR(ndst)) {
2603 			err = PTR_ERR(ndst);
2604 			ndst = NULL;
2605 			goto tx_error;
2606 		}
2607 
2608 		if (!info) {
2609 			u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2610 
2611 			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2612 						    dst_port, ifindex, vni,
2613 						    ndst, rt6i_flags);
2614 			if (err)
2615 				goto out_unlock;
2616 		}
2617 
2618 		err = skb_tunnel_check_pmtu(skb, ndst,
2619 					    vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2620 					    netif_is_any_bridge_port(dev));
2621 		if (err < 0) {
2622 			goto tx_error;
2623 		} else if (err) {
2624 			if (info) {
2625 				struct ip_tunnel_info *unclone;
2626 				struct in6_addr src, dst;
2627 
2628 				unclone = skb_tunnel_info_unclone(skb);
2629 				if (unlikely(!unclone))
2630 					goto tx_error;
2631 
2632 				src = remote_ip.sin6.sin6_addr;
2633 				dst = local_ip.sin6.sin6_addr;
2634 				unclone->key.u.ipv6.src = src;
2635 				unclone->key.u.ipv6.dst = dst;
2636 			}
2637 
2638 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2639 			dst_release(ndst);
2640 			goto out_unlock;
2641 		}
2642 
2643 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2644 		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2645 		skb_scrub_packet(skb, xnet);
2646 		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2647 				      vni, md, flags, udp_sum);
2648 		if (err < 0)
2649 			goto tx_error;
2650 
2651 		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2652 				     &local_ip.sin6.sin6_addr,
2653 				     &dst->sin6.sin6_addr, tos, ttl,
2654 				     label, src_port, dst_port, !udp_sum);
2655 #endif
2656 	}
2657 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2658 out_unlock:
2659 	rcu_read_unlock();
2660 	return;
2661 
2662 drop:
2663 	dev->stats.tx_dropped++;
2664 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2665 	dev_kfree_skb(skb);
2666 	return;
2667 
2668 tx_error:
2669 	rcu_read_unlock();
2670 	if (err == -ELOOP)
2671 		dev->stats.collisions++;
2672 	else if (err == -ENETUNREACH)
2673 		dev->stats.tx_carrier_errors++;
2674 	dst_release(ndst);
2675 	dev->stats.tx_errors++;
2676 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2677 	kfree_skb(skb);
2678 }
2679 
2680 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2681 			  struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2682 {
2683 	struct vxlan_rdst nh_rdst;
2684 	struct nexthop *nh;
2685 	bool do_xmit;
2686 	u32 hash;
2687 
2688 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2689 	hash = skb_get_hash(skb);
2690 
2691 	rcu_read_lock();
2692 	nh = rcu_dereference(f->nh);
2693 	if (!nh) {
2694 		rcu_read_unlock();
2695 		goto drop;
2696 	}
2697 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2698 	rcu_read_unlock();
2699 
2700 	if (likely(do_xmit))
2701 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2702 	else
2703 		goto drop;
2704 
2705 	return;
2706 
2707 drop:
2708 	dev->stats.tx_dropped++;
2709 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2710 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2711 	dev_kfree_skb(skb);
2712 }
2713 
2714 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2715 				   u32 nhid, __be32 vni)
2716 {
2717 	struct vxlan_dev *vxlan = netdev_priv(dev);
2718 	struct vxlan_rdst nh_rdst;
2719 	struct nexthop *nh;
2720 	bool do_xmit;
2721 	u32 hash;
2722 
2723 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2724 	hash = skb_get_hash(skb);
2725 
2726 	rcu_read_lock();
2727 	nh = nexthop_find_by_id(dev_net(dev), nhid);
2728 	if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2729 		rcu_read_unlock();
2730 		goto drop;
2731 	}
2732 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2733 	rcu_read_unlock();
2734 
2735 	if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2736 		goto drop;
2737 
2738 	if (likely(do_xmit))
2739 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2740 	else
2741 		goto drop;
2742 
2743 	return NETDEV_TX_OK;
2744 
2745 drop:
2746 	dev->stats.tx_dropped++;
2747 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2748 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2749 	dev_kfree_skb(skb);
2750 	return NETDEV_TX_OK;
2751 }
2752 
2753 /* Transmit local packets over Vxlan
2754  *
2755  * Outer IP header inherits ECN and DF from inner header.
2756  * Outer UDP destination is the VXLAN assigned port.
2757  *           source port is based on hash of flow
2758  */
2759 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2760 {
2761 	struct vxlan_dev *vxlan = netdev_priv(dev);
2762 	struct vxlan_rdst *rdst, *fdst = NULL;
2763 	const struct ip_tunnel_info *info;
2764 	bool did_rsc = false;
2765 	struct vxlan_fdb *f;
2766 	struct ethhdr *eth;
2767 	__be32 vni = 0;
2768 	u32 nhid = 0;
2769 
2770 	info = skb_tunnel_info(skb);
2771 
2772 	skb_reset_mac_header(skb);
2773 
2774 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2775 		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2776 		    info->mode & IP_TUNNEL_INFO_TX) {
2777 			vni = tunnel_id_to_key32(info->key.tun_id);
2778 			nhid = info->key.nhid;
2779 		} else {
2780 			if (info && info->mode & IP_TUNNEL_INFO_TX)
2781 				vxlan_xmit_one(skb, dev, vni, NULL, false);
2782 			else
2783 				kfree_skb(skb);
2784 			return NETDEV_TX_OK;
2785 		}
2786 	}
2787 
2788 	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2789 		eth = eth_hdr(skb);
2790 		if (ntohs(eth->h_proto) == ETH_P_ARP)
2791 			return arp_reduce(dev, skb, vni);
2792 #if IS_ENABLED(CONFIG_IPV6)
2793 		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2794 			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2795 					    sizeof(struct nd_msg)) &&
2796 			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2797 			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2798 
2799 			if (m->icmph.icmp6_code == 0 &&
2800 			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2801 				return neigh_reduce(dev, skb, vni);
2802 		}
2803 #endif
2804 	}
2805 
2806 	if (nhid)
2807 		return vxlan_xmit_nhid(skb, dev, nhid, vni);
2808 
2809 	if (vxlan->cfg.flags & VXLAN_F_MDB) {
2810 		struct vxlan_mdb_entry *mdb_entry;
2811 
2812 		rcu_read_lock();
2813 		mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2814 		if (mdb_entry) {
2815 			netdev_tx_t ret;
2816 
2817 			ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2818 			rcu_read_unlock();
2819 			return ret;
2820 		}
2821 		rcu_read_unlock();
2822 	}
2823 
2824 	eth = eth_hdr(skb);
2825 	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2826 	did_rsc = false;
2827 
2828 	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2829 	    (ntohs(eth->h_proto) == ETH_P_IP ||
2830 	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2831 		did_rsc = route_shortcircuit(dev, skb);
2832 		if (did_rsc)
2833 			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2834 	}
2835 
2836 	if (f == NULL) {
2837 		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2838 		if (f == NULL) {
2839 			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2840 			    !is_multicast_ether_addr(eth->h_dest))
2841 				vxlan_fdb_miss(vxlan, eth->h_dest);
2842 
2843 			dev->stats.tx_dropped++;
2844 			vxlan_vnifilter_count(vxlan, vni, NULL,
2845 					      VXLAN_VNI_STATS_TX_DROPS, 0);
2846 			kfree_skb(skb);
2847 			return NETDEV_TX_OK;
2848 		}
2849 	}
2850 
2851 	if (rcu_access_pointer(f->nh)) {
2852 		vxlan_xmit_nh(skb, dev, f,
2853 			      (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2854 	} else {
2855 		list_for_each_entry_rcu(rdst, &f->remotes, list) {
2856 			struct sk_buff *skb1;
2857 
2858 			if (!fdst) {
2859 				fdst = rdst;
2860 				continue;
2861 			}
2862 			skb1 = skb_clone(skb, GFP_ATOMIC);
2863 			if (skb1)
2864 				vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2865 		}
2866 		if (fdst)
2867 			vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2868 		else
2869 			kfree_skb(skb);
2870 	}
2871 
2872 	return NETDEV_TX_OK;
2873 }
2874 
2875 /* Walk the forwarding table and purge stale entries */
2876 static void vxlan_cleanup(struct timer_list *t)
2877 {
2878 	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2879 	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2880 	unsigned int h;
2881 
2882 	if (!netif_running(vxlan->dev))
2883 		return;
2884 
2885 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2886 		struct hlist_node *p, *n;
2887 
2888 		spin_lock(&vxlan->hash_lock[h]);
2889 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2890 			struct vxlan_fdb *f
2891 				= container_of(p, struct vxlan_fdb, hlist);
2892 			unsigned long timeout;
2893 
2894 			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2895 				continue;
2896 
2897 			if (f->flags & NTF_EXT_LEARNED)
2898 				continue;
2899 
2900 			timeout = f->used + vxlan->cfg.age_interval * HZ;
2901 			if (time_before_eq(timeout, jiffies)) {
2902 				netdev_dbg(vxlan->dev,
2903 					   "garbage collect %pM\n",
2904 					   f->eth_addr);
2905 				f->state = NUD_STALE;
2906 				vxlan_fdb_destroy(vxlan, f, true, true);
2907 			} else if (time_before(timeout, next_timer))
2908 				next_timer = timeout;
2909 		}
2910 		spin_unlock(&vxlan->hash_lock[h]);
2911 	}
2912 
2913 	mod_timer(&vxlan->age_timer, next_timer);
2914 }
2915 
2916 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2917 {
2918 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2919 
2920 	spin_lock(&vn->sock_lock);
2921 	hlist_del_init_rcu(&vxlan->hlist4.hlist);
2922 #if IS_ENABLED(CONFIG_IPV6)
2923 	hlist_del_init_rcu(&vxlan->hlist6.hlist);
2924 #endif
2925 	spin_unlock(&vn->sock_lock);
2926 }
2927 
2928 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2929 			     struct vxlan_dev_node *node)
2930 {
2931 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2932 	__be32 vni = vxlan->default_dst.remote_vni;
2933 
2934 	node->vxlan = vxlan;
2935 	spin_lock(&vn->sock_lock);
2936 	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2937 	spin_unlock(&vn->sock_lock);
2938 }
2939 
2940 /* Setup stats when device is created */
2941 static int vxlan_init(struct net_device *dev)
2942 {
2943 	struct vxlan_dev *vxlan = netdev_priv(dev);
2944 	int err;
2945 
2946 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2947 		vxlan_vnigroup_init(vxlan);
2948 
2949 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2950 	if (!dev->tstats) {
2951 		err = -ENOMEM;
2952 		goto err_vnigroup_uninit;
2953 	}
2954 
2955 	err = gro_cells_init(&vxlan->gro_cells, dev);
2956 	if (err)
2957 		goto err_free_percpu;
2958 
2959 	err = vxlan_mdb_init(vxlan);
2960 	if (err)
2961 		goto err_gro_cells_destroy;
2962 
2963 	return 0;
2964 
2965 err_gro_cells_destroy:
2966 	gro_cells_destroy(&vxlan->gro_cells);
2967 err_free_percpu:
2968 	free_percpu(dev->tstats);
2969 err_vnigroup_uninit:
2970 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2971 		vxlan_vnigroup_uninit(vxlan);
2972 	return err;
2973 }
2974 
2975 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2976 {
2977 	struct vxlan_fdb *f;
2978 	u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2979 
2980 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
2981 	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2982 	if (f)
2983 		vxlan_fdb_destroy(vxlan, f, true, true);
2984 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2985 }
2986 
2987 static void vxlan_uninit(struct net_device *dev)
2988 {
2989 	struct vxlan_dev *vxlan = netdev_priv(dev);
2990 
2991 	vxlan_mdb_fini(vxlan);
2992 
2993 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2994 		vxlan_vnigroup_uninit(vxlan);
2995 
2996 	gro_cells_destroy(&vxlan->gro_cells);
2997 
2998 	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2999 
3000 	free_percpu(dev->tstats);
3001 }
3002 
3003 /* Start ageing timer and join group when device is brought up */
3004 static int vxlan_open(struct net_device *dev)
3005 {
3006 	struct vxlan_dev *vxlan = netdev_priv(dev);
3007 	int ret;
3008 
3009 	ret = vxlan_sock_add(vxlan);
3010 	if (ret < 0)
3011 		return ret;
3012 
3013 	ret = vxlan_multicast_join(vxlan);
3014 	if (ret) {
3015 		vxlan_sock_release(vxlan);
3016 		return ret;
3017 	}
3018 
3019 	if (vxlan->cfg.age_interval)
3020 		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3021 
3022 	return ret;
3023 }
3024 
3025 struct vxlan_fdb_flush_desc {
3026 	bool				ignore_default_entry;
3027 	unsigned long                   state;
3028 	unsigned long			state_mask;
3029 	unsigned long                   flags;
3030 	unsigned long			flags_mask;
3031 	__be32				src_vni;
3032 	u32				nhid;
3033 	__be32				vni;
3034 	__be16				port;
3035 	union vxlan_addr		dst_ip;
3036 };
3037 
3038 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f,
3039 				       const struct vxlan_dev *vxlan)
3040 {
3041 	return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni;
3042 }
3043 
3044 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid)
3045 {
3046 	struct nexthop *nh = rtnl_dereference(f->nh);
3047 
3048 	return nh && nh->id == nhid;
3049 }
3050 
3051 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f,
3052 				    const struct vxlan_dev *vxlan,
3053 				    const struct vxlan_fdb_flush_desc *desc)
3054 {
3055 	if (desc->state_mask && (f->state & desc->state_mask) != desc->state)
3056 		return false;
3057 
3058 	if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags)
3059 		return false;
3060 
3061 	if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan))
3062 		return false;
3063 
3064 	if (desc->src_vni && f->vni != desc->src_vni)
3065 		return false;
3066 
3067 	if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid))
3068 		return false;
3069 
3070 	return true;
3071 }
3072 
3073 static bool
3074 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc)
3075 {
3076 	return desc->vni || desc->port || desc->dst_ip.sa.sa_family;
3077 }
3078 
3079 static bool
3080 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc,
3081 			       const struct vxlan_rdst *rd)
3082 {
3083 	if (desc->vni && rd->remote_vni != desc->vni)
3084 		return false;
3085 
3086 	if (desc->port && rd->remote_port != desc->port)
3087 		return false;
3088 
3089 	if (desc->dst_ip.sa.sa_family &&
3090 	    !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip))
3091 		return false;
3092 
3093 	return true;
3094 }
3095 
3096 static void
3097 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan,
3098 			      const struct vxlan_fdb_flush_desc *desc,
3099 			      bool *p_destroy_fdb)
3100 {
3101 	bool remotes_flushed = false;
3102 	struct vxlan_rdst *rd, *tmp;
3103 
3104 	list_for_each_entry_safe(rd, tmp, &f->remotes, list) {
3105 		if (!vxlan_fdb_flush_remote_matches(desc, rd))
3106 			continue;
3107 
3108 		vxlan_fdb_dst_destroy(vxlan, f, rd, true);
3109 		remotes_flushed = true;
3110 	}
3111 
3112 	*p_destroy_fdb = remotes_flushed && list_empty(&f->remotes);
3113 }
3114 
3115 /* Purge the forwarding table */
3116 static void vxlan_flush(struct vxlan_dev *vxlan,
3117 			const struct vxlan_fdb_flush_desc *desc)
3118 {
3119 	bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc);
3120 	unsigned int h;
3121 
3122 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3123 		struct hlist_node *p, *n;
3124 
3125 		spin_lock_bh(&vxlan->hash_lock[h]);
3126 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3127 			struct vxlan_fdb *f
3128 				= container_of(p, struct vxlan_fdb, hlist);
3129 
3130 			if (!vxlan_fdb_flush_matches(f, vxlan, desc))
3131 				continue;
3132 
3133 			if (match_remotes) {
3134 				bool destroy_fdb = false;
3135 
3136 				vxlan_fdb_flush_match_remotes(f, vxlan, desc,
3137 							      &destroy_fdb);
3138 
3139 				if (!destroy_fdb)
3140 					continue;
3141 			}
3142 
3143 			vxlan_fdb_destroy(vxlan, f, true, true);
3144 		}
3145 		spin_unlock_bh(&vxlan->hash_lock[h]);
3146 	}
3147 }
3148 
3149 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = {
3150 	[NDA_SRC_VNI]   = { .type = NLA_U32 },
3151 	[NDA_NH_ID]	= { .type = NLA_U32 },
3152 	[NDA_VNI]	= { .type = NLA_U32 },
3153 	[NDA_PORT]	= { .type = NLA_U16 },
3154 	[NDA_DST]	= NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr),
3155 					   sizeof(struct in6_addr)),
3156 	[NDA_NDM_STATE_MASK]	= { .type = NLA_U16 },
3157 	[NDA_NDM_FLAGS_MASK]	= { .type = NLA_U8 },
3158 };
3159 
3160 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF)
3161 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP)
3162 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \
3163 					   NTF_ROUTER)
3164 
3165 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev,
3166 				 struct netlink_ext_ack *extack)
3167 {
3168 	struct vxlan_dev *vxlan = netdev_priv(dev);
3169 	struct vxlan_fdb_flush_desc desc = {};
3170 	struct ndmsg *ndm = nlmsg_data(nlh);
3171 	struct nlattr *tb[NDA_MAX + 1];
3172 	u8 ndm_flags;
3173 	int err;
3174 
3175 	ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS;
3176 
3177 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy,
3178 			  extack);
3179 	if (err)
3180 		return err;
3181 
3182 	if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) {
3183 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set");
3184 		return -EINVAL;
3185 	}
3186 	if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) {
3187 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set");
3188 		return -EINVAL;
3189 	}
3190 
3191 	desc.state = ndm->ndm_state;
3192 	desc.flags = ndm_flags;
3193 
3194 	if (tb[NDA_NDM_STATE_MASK])
3195 		desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]);
3196 
3197 	if (tb[NDA_NDM_FLAGS_MASK])
3198 		desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]);
3199 
3200 	if (tb[NDA_SRC_VNI])
3201 		desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
3202 
3203 	if (tb[NDA_NH_ID])
3204 		desc.nhid = nla_get_u32(tb[NDA_NH_ID]);
3205 
3206 	if (tb[NDA_VNI])
3207 		desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
3208 
3209 	if (tb[NDA_PORT])
3210 		desc.port = nla_get_be16(tb[NDA_PORT]);
3211 
3212 	if (tb[NDA_DST]) {
3213 		union vxlan_addr ip;
3214 
3215 		err = vxlan_nla_get_addr(&ip, tb[NDA_DST]);
3216 		if (err) {
3217 			NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST],
3218 					    "Unsupported address family");
3219 			return err;
3220 		}
3221 		desc.dst_ip = ip;
3222 	}
3223 
3224 	vxlan_flush(vxlan, &desc);
3225 
3226 	return 0;
3227 }
3228 
3229 /* Cleanup timer and forwarding table on shutdown */
3230 static int vxlan_stop(struct net_device *dev)
3231 {
3232 	struct vxlan_dev *vxlan = netdev_priv(dev);
3233 	struct vxlan_fdb_flush_desc desc = {
3234 		/* Default entry is deleted at vxlan_uninit. */
3235 		.ignore_default_entry = true,
3236 		.state = 0,
3237 		.state_mask = NUD_PERMANENT | NUD_NOARP,
3238 	};
3239 
3240 	vxlan_multicast_leave(vxlan);
3241 
3242 	del_timer_sync(&vxlan->age_timer);
3243 
3244 	vxlan_flush(vxlan, &desc);
3245 	vxlan_sock_release(vxlan);
3246 
3247 	return 0;
3248 }
3249 
3250 /* Stub, nothing needs to be done. */
3251 static void vxlan_set_multicast_list(struct net_device *dev)
3252 {
3253 }
3254 
3255 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3256 {
3257 	struct vxlan_dev *vxlan = netdev_priv(dev);
3258 	struct vxlan_rdst *dst = &vxlan->default_dst;
3259 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3260 							 dst->remote_ifindex);
3261 
3262 	/* This check is different than dev->max_mtu, because it looks at
3263 	 * the lowerdev->mtu, rather than the static dev->max_mtu
3264 	 */
3265 	if (lowerdev) {
3266 		int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3267 		if (new_mtu > max_mtu)
3268 			return -EINVAL;
3269 	}
3270 
3271 	dev->mtu = new_mtu;
3272 	return 0;
3273 }
3274 
3275 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3276 {
3277 	struct vxlan_dev *vxlan = netdev_priv(dev);
3278 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
3279 	__be16 sport, dport;
3280 
3281 	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3282 				  vxlan->cfg.port_max, true);
3283 	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3284 
3285 	if (ip_tunnel_info_af(info) == AF_INET) {
3286 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3287 		struct rtable *rt;
3288 
3289 		rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3290 				     info->key.u.ipv4.dst,
3291 				     &info->key.u.ipv4.src, dport, sport,
3292 				     info->key.flow_flags, &info->dst_cache,
3293 				     info);
3294 		if (IS_ERR(rt))
3295 			return PTR_ERR(rt);
3296 		ip_rt_put(rt);
3297 	} else {
3298 #if IS_ENABLED(CONFIG_IPV6)
3299 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3300 		struct dst_entry *ndst;
3301 
3302 		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3303 					info->key.label, &info->key.u.ipv6.dst,
3304 					&info->key.u.ipv6.src, dport, sport,
3305 					&info->dst_cache, info);
3306 		if (IS_ERR(ndst))
3307 			return PTR_ERR(ndst);
3308 		dst_release(ndst);
3309 #else /* !CONFIG_IPV6 */
3310 		return -EPFNOSUPPORT;
3311 #endif
3312 	}
3313 	info->key.tp_src = sport;
3314 	info->key.tp_dst = dport;
3315 	return 0;
3316 }
3317 
3318 static const struct net_device_ops vxlan_netdev_ether_ops = {
3319 	.ndo_init		= vxlan_init,
3320 	.ndo_uninit		= vxlan_uninit,
3321 	.ndo_open		= vxlan_open,
3322 	.ndo_stop		= vxlan_stop,
3323 	.ndo_start_xmit		= vxlan_xmit,
3324 	.ndo_get_stats64	= dev_get_tstats64,
3325 	.ndo_set_rx_mode	= vxlan_set_multicast_list,
3326 	.ndo_change_mtu		= vxlan_change_mtu,
3327 	.ndo_validate_addr	= eth_validate_addr,
3328 	.ndo_set_mac_address	= eth_mac_addr,
3329 	.ndo_fdb_add		= vxlan_fdb_add,
3330 	.ndo_fdb_del		= vxlan_fdb_delete,
3331 	.ndo_fdb_del_bulk	= vxlan_fdb_delete_bulk,
3332 	.ndo_fdb_dump		= vxlan_fdb_dump,
3333 	.ndo_fdb_get		= vxlan_fdb_get,
3334 	.ndo_mdb_add		= vxlan_mdb_add,
3335 	.ndo_mdb_del		= vxlan_mdb_del,
3336 	.ndo_mdb_dump		= vxlan_mdb_dump,
3337 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3338 };
3339 
3340 static const struct net_device_ops vxlan_netdev_raw_ops = {
3341 	.ndo_init		= vxlan_init,
3342 	.ndo_uninit		= vxlan_uninit,
3343 	.ndo_open		= vxlan_open,
3344 	.ndo_stop		= vxlan_stop,
3345 	.ndo_start_xmit		= vxlan_xmit,
3346 	.ndo_get_stats64	= dev_get_tstats64,
3347 	.ndo_change_mtu		= vxlan_change_mtu,
3348 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3349 };
3350 
3351 /* Info for udev, that this is a virtual tunnel endpoint */
3352 static struct device_type vxlan_type = {
3353 	.name = "vxlan",
3354 };
3355 
3356 /* Calls the ndo_udp_tunnel_add of the caller in order to
3357  * supply the listening VXLAN udp ports. Callers are expected
3358  * to implement the ndo_udp_tunnel_add.
3359  */
3360 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3361 {
3362 	struct vxlan_sock *vs;
3363 	struct net *net = dev_net(dev);
3364 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3365 	unsigned int i;
3366 
3367 	spin_lock(&vn->sock_lock);
3368 	for (i = 0; i < PORT_HASH_SIZE; ++i) {
3369 		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3370 			unsigned short type;
3371 
3372 			if (vs->flags & VXLAN_F_GPE)
3373 				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3374 			else
3375 				type = UDP_TUNNEL_TYPE_VXLAN;
3376 
3377 			if (push)
3378 				udp_tunnel_push_rx_port(dev, vs->sock, type);
3379 			else
3380 				udp_tunnel_drop_rx_port(dev, vs->sock, type);
3381 		}
3382 	}
3383 	spin_unlock(&vn->sock_lock);
3384 }
3385 
3386 /* Initialize the device structure. */
3387 static void vxlan_setup(struct net_device *dev)
3388 {
3389 	struct vxlan_dev *vxlan = netdev_priv(dev);
3390 	unsigned int h;
3391 
3392 	eth_hw_addr_random(dev);
3393 	ether_setup(dev);
3394 
3395 	dev->needs_free_netdev = true;
3396 	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3397 
3398 	dev->features	|= NETIF_F_LLTX;
3399 	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3400 	dev->features   |= NETIF_F_RXCSUM;
3401 	dev->features   |= NETIF_F_GSO_SOFTWARE;
3402 
3403 	dev->vlan_features = dev->features;
3404 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3405 	dev->hw_features |= NETIF_F_RXCSUM;
3406 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3407 	netif_keep_dst(dev);
3408 	dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN;
3409 
3410 	/* MTU range: 68 - 65535 */
3411 	dev->min_mtu = ETH_MIN_MTU;
3412 	dev->max_mtu = ETH_MAX_MTU;
3413 
3414 	INIT_LIST_HEAD(&vxlan->next);
3415 
3416 	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3417 
3418 	vxlan->dev = dev;
3419 
3420 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3421 		spin_lock_init(&vxlan->hash_lock[h]);
3422 		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3423 	}
3424 }
3425 
3426 static void vxlan_ether_setup(struct net_device *dev)
3427 {
3428 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3429 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3430 	dev->netdev_ops = &vxlan_netdev_ether_ops;
3431 }
3432 
3433 static void vxlan_raw_setup(struct net_device *dev)
3434 {
3435 	dev->header_ops = NULL;
3436 	dev->type = ARPHRD_NONE;
3437 	dev->hard_header_len = 0;
3438 	dev->addr_len = 0;
3439 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3440 	dev->netdev_ops = &vxlan_netdev_raw_ops;
3441 }
3442 
3443 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3444 	[IFLA_VXLAN_UNSPEC]     = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3445 	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
3446 	[IFLA_VXLAN_GROUP]	= { .len = sizeof_field(struct iphdr, daddr) },
3447 	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
3448 	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
3449 	[IFLA_VXLAN_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
3450 	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
3451 	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
3452 	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
3453 	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
3454 	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
3455 	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
3456 	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
3457 	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3458 	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
3459 	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
3460 	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
3461 	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
3462 	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
3463 	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
3464 	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
3465 	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
3466 	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
3467 	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
3468 	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
3469 	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
3470 	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
3471 	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
3472 	[IFLA_VXLAN_TTL_INHERIT]	= { .type = NLA_FLAG },
3473 	[IFLA_VXLAN_DF]		= { .type = NLA_U8 },
3474 	[IFLA_VXLAN_VNIFILTER]	= { .type = NLA_U8 },
3475 	[IFLA_VXLAN_LOCALBYPASS]	= NLA_POLICY_MAX(NLA_U8, 1),
3476 };
3477 
3478 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3479 			  struct netlink_ext_ack *extack)
3480 {
3481 	if (tb[IFLA_ADDRESS]) {
3482 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3483 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3484 					    "Provided link layer address is not Ethernet");
3485 			return -EINVAL;
3486 		}
3487 
3488 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3489 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3490 					    "Provided Ethernet address is not unicast");
3491 			return -EADDRNOTAVAIL;
3492 		}
3493 	}
3494 
3495 	if (tb[IFLA_MTU]) {
3496 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3497 
3498 		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3499 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3500 					    "MTU must be between 68 and 65535");
3501 			return -EINVAL;
3502 		}
3503 	}
3504 
3505 	if (!data) {
3506 		NL_SET_ERR_MSG(extack,
3507 			       "Required attributes not provided to perform the operation");
3508 		return -EINVAL;
3509 	}
3510 
3511 	if (data[IFLA_VXLAN_ID]) {
3512 		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3513 
3514 		if (id >= VXLAN_N_VID) {
3515 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3516 					    "VXLAN ID must be lower than 16777216");
3517 			return -ERANGE;
3518 		}
3519 	}
3520 
3521 	if (data[IFLA_VXLAN_PORT_RANGE]) {
3522 		const struct ifla_vxlan_port_range *p
3523 			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3524 
3525 		if (ntohs(p->high) < ntohs(p->low)) {
3526 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3527 					    "Invalid source port range");
3528 			return -EINVAL;
3529 		}
3530 	}
3531 
3532 	if (data[IFLA_VXLAN_DF]) {
3533 		enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3534 
3535 		if (df < 0 || df > VXLAN_DF_MAX) {
3536 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3537 					    "Invalid DF attribute");
3538 			return -EINVAL;
3539 		}
3540 	}
3541 
3542 	return 0;
3543 }
3544 
3545 static void vxlan_get_drvinfo(struct net_device *netdev,
3546 			      struct ethtool_drvinfo *drvinfo)
3547 {
3548 	strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3549 	strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3550 }
3551 
3552 static int vxlan_get_link_ksettings(struct net_device *dev,
3553 				    struct ethtool_link_ksettings *cmd)
3554 {
3555 	struct vxlan_dev *vxlan = netdev_priv(dev);
3556 	struct vxlan_rdst *dst = &vxlan->default_dst;
3557 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3558 							 dst->remote_ifindex);
3559 
3560 	if (!lowerdev) {
3561 		cmd->base.duplex = DUPLEX_UNKNOWN;
3562 		cmd->base.port = PORT_OTHER;
3563 		cmd->base.speed = SPEED_UNKNOWN;
3564 
3565 		return 0;
3566 	}
3567 
3568 	return __ethtool_get_link_ksettings(lowerdev, cmd);
3569 }
3570 
3571 static const struct ethtool_ops vxlan_ethtool_ops = {
3572 	.get_drvinfo		= vxlan_get_drvinfo,
3573 	.get_link		= ethtool_op_get_link,
3574 	.get_link_ksettings	= vxlan_get_link_ksettings,
3575 };
3576 
3577 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3578 					__be16 port, u32 flags, int ifindex)
3579 {
3580 	struct socket *sock;
3581 	struct udp_port_cfg udp_conf;
3582 	int err;
3583 
3584 	memset(&udp_conf, 0, sizeof(udp_conf));
3585 
3586 	if (ipv6) {
3587 		udp_conf.family = AF_INET6;
3588 		udp_conf.use_udp6_rx_checksums =
3589 		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3590 		udp_conf.ipv6_v6only = 1;
3591 	} else {
3592 		udp_conf.family = AF_INET;
3593 	}
3594 
3595 	udp_conf.local_udp_port = port;
3596 	udp_conf.bind_ifindex = ifindex;
3597 
3598 	/* Open UDP socket */
3599 	err = udp_sock_create(net, &udp_conf, &sock);
3600 	if (err < 0)
3601 		return ERR_PTR(err);
3602 
3603 	udp_allow_gso(sock->sk);
3604 	return sock;
3605 }
3606 
3607 /* Create new listen socket if needed */
3608 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3609 					      __be16 port, u32 flags,
3610 					      int ifindex)
3611 {
3612 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3613 	struct vxlan_sock *vs;
3614 	struct socket *sock;
3615 	unsigned int h;
3616 	struct udp_tunnel_sock_cfg tunnel_cfg;
3617 
3618 	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3619 	if (!vs)
3620 		return ERR_PTR(-ENOMEM);
3621 
3622 	for (h = 0; h < VNI_HASH_SIZE; ++h)
3623 		INIT_HLIST_HEAD(&vs->vni_list[h]);
3624 
3625 	sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3626 	if (IS_ERR(sock)) {
3627 		kfree(vs);
3628 		return ERR_CAST(sock);
3629 	}
3630 
3631 	vs->sock = sock;
3632 	refcount_set(&vs->refcnt, 1);
3633 	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3634 
3635 	spin_lock(&vn->sock_lock);
3636 	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3637 	udp_tunnel_notify_add_rx_port(sock,
3638 				      (vs->flags & VXLAN_F_GPE) ?
3639 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
3640 				      UDP_TUNNEL_TYPE_VXLAN);
3641 	spin_unlock(&vn->sock_lock);
3642 
3643 	/* Mark socket as an encapsulation socket. */
3644 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3645 	tunnel_cfg.sk_user_data = vs;
3646 	tunnel_cfg.encap_type = 1;
3647 	tunnel_cfg.encap_rcv = vxlan_rcv;
3648 	tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3649 	tunnel_cfg.encap_destroy = NULL;
3650 	if (vs->flags & VXLAN_F_GPE) {
3651 		tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3652 		tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3653 	} else {
3654 		tunnel_cfg.gro_receive = vxlan_gro_receive;
3655 		tunnel_cfg.gro_complete = vxlan_gro_complete;
3656 	}
3657 
3658 	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3659 
3660 	return vs;
3661 }
3662 
3663 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3664 {
3665 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3666 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3667 	struct vxlan_sock *vs = NULL;
3668 	struct vxlan_dev_node *node;
3669 	int l3mdev_index = 0;
3670 
3671 	if (vxlan->cfg.remote_ifindex)
3672 		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3673 			vxlan->net, vxlan->cfg.remote_ifindex);
3674 
3675 	if (!vxlan->cfg.no_share) {
3676 		spin_lock(&vn->sock_lock);
3677 		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3678 				     vxlan->cfg.dst_port, vxlan->cfg.flags,
3679 				     l3mdev_index);
3680 		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3681 			spin_unlock(&vn->sock_lock);
3682 			return -EBUSY;
3683 		}
3684 		spin_unlock(&vn->sock_lock);
3685 	}
3686 	if (!vs)
3687 		vs = vxlan_socket_create(vxlan->net, ipv6,
3688 					 vxlan->cfg.dst_port, vxlan->cfg.flags,
3689 					 l3mdev_index);
3690 	if (IS_ERR(vs))
3691 		return PTR_ERR(vs);
3692 #if IS_ENABLED(CONFIG_IPV6)
3693 	if (ipv6) {
3694 		rcu_assign_pointer(vxlan->vn6_sock, vs);
3695 		node = &vxlan->hlist6;
3696 	} else
3697 #endif
3698 	{
3699 		rcu_assign_pointer(vxlan->vn4_sock, vs);
3700 		node = &vxlan->hlist4;
3701 	}
3702 
3703 	if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3704 		vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3705 	else
3706 		vxlan_vs_add_dev(vs, vxlan, node);
3707 
3708 	return 0;
3709 }
3710 
3711 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3712 {
3713 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3714 	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3715 	bool ipv4 = !ipv6 || metadata;
3716 	int ret = 0;
3717 
3718 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3719 #if IS_ENABLED(CONFIG_IPV6)
3720 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3721 	if (ipv6) {
3722 		ret = __vxlan_sock_add(vxlan, true);
3723 		if (ret < 0 && ret != -EAFNOSUPPORT)
3724 			ipv4 = false;
3725 	}
3726 #endif
3727 	if (ipv4)
3728 		ret = __vxlan_sock_add(vxlan, false);
3729 	if (ret < 0)
3730 		vxlan_sock_release(vxlan);
3731 	return ret;
3732 }
3733 
3734 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3735 		     struct vxlan_config *conf, __be32 vni)
3736 {
3737 	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3738 	struct vxlan_dev *tmp;
3739 
3740 	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3741 		if (tmp == vxlan)
3742 			continue;
3743 		if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3744 			if (!vxlan_vnifilter_lookup(tmp, vni))
3745 				continue;
3746 		} else if (tmp->cfg.vni != vni) {
3747 			continue;
3748 		}
3749 		if (tmp->cfg.dst_port != conf->dst_port)
3750 			continue;
3751 		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3752 		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3753 			continue;
3754 
3755 		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3756 		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3757 			continue;
3758 
3759 		return -EEXIST;
3760 	}
3761 
3762 	return 0;
3763 }
3764 
3765 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3766 				 struct net_device **lower,
3767 				 struct vxlan_dev *old,
3768 				 struct netlink_ext_ack *extack)
3769 {
3770 	bool use_ipv6 = false;
3771 
3772 	if (conf->flags & VXLAN_F_GPE) {
3773 		/* For now, allow GPE only together with
3774 		 * COLLECT_METADATA. This can be relaxed later; in such
3775 		 * case, the other side of the PtP link will have to be
3776 		 * provided.
3777 		 */
3778 		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3779 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3780 			NL_SET_ERR_MSG(extack,
3781 				       "VXLAN GPE does not support this combination of attributes");
3782 			return -EINVAL;
3783 		}
3784 	}
3785 
3786 	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3787 		/* Unless IPv6 is explicitly requested, assume IPv4 */
3788 		conf->remote_ip.sa.sa_family = AF_INET;
3789 		conf->saddr.sa.sa_family = AF_INET;
3790 	} else if (!conf->remote_ip.sa.sa_family) {
3791 		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3792 	} else if (!conf->saddr.sa.sa_family) {
3793 		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3794 	}
3795 
3796 	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3797 		NL_SET_ERR_MSG(extack,
3798 			       "Local and remote address must be from the same family");
3799 		return -EINVAL;
3800 	}
3801 
3802 	if (vxlan_addr_multicast(&conf->saddr)) {
3803 		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3804 		return -EINVAL;
3805 	}
3806 
3807 	if (conf->saddr.sa.sa_family == AF_INET6) {
3808 		if (!IS_ENABLED(CONFIG_IPV6)) {
3809 			NL_SET_ERR_MSG(extack,
3810 				       "IPv6 support not enabled in the kernel");
3811 			return -EPFNOSUPPORT;
3812 		}
3813 		use_ipv6 = true;
3814 		conf->flags |= VXLAN_F_IPV6;
3815 
3816 		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3817 			int local_type =
3818 				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3819 			int remote_type =
3820 				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3821 
3822 			if (local_type & IPV6_ADDR_LINKLOCAL) {
3823 				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3824 				    (remote_type != IPV6_ADDR_ANY)) {
3825 					NL_SET_ERR_MSG(extack,
3826 						       "Invalid combination of local and remote address scopes");
3827 					return -EINVAL;
3828 				}
3829 
3830 				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3831 			} else {
3832 				if (remote_type ==
3833 				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3834 					NL_SET_ERR_MSG(extack,
3835 						       "Invalid combination of local and remote address scopes");
3836 					return -EINVAL;
3837 				}
3838 
3839 				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3840 			}
3841 		}
3842 	}
3843 
3844 	if (conf->label && !use_ipv6) {
3845 		NL_SET_ERR_MSG(extack,
3846 			       "Label attribute only applies to IPv6 VXLAN devices");
3847 		return -EINVAL;
3848 	}
3849 
3850 	if (conf->remote_ifindex) {
3851 		struct net_device *lowerdev;
3852 
3853 		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3854 		if (!lowerdev) {
3855 			NL_SET_ERR_MSG(extack,
3856 				       "Invalid local interface, device not found");
3857 			return -ENODEV;
3858 		}
3859 
3860 #if IS_ENABLED(CONFIG_IPV6)
3861 		if (use_ipv6) {
3862 			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3863 
3864 			if (idev && idev->cnf.disable_ipv6) {
3865 				NL_SET_ERR_MSG(extack,
3866 					       "IPv6 support disabled by administrator");
3867 				return -EPERM;
3868 			}
3869 		}
3870 #endif
3871 
3872 		*lower = lowerdev;
3873 	} else {
3874 		if (vxlan_addr_multicast(&conf->remote_ip)) {
3875 			NL_SET_ERR_MSG(extack,
3876 				       "Local interface required for multicast remote destination");
3877 
3878 			return -EINVAL;
3879 		}
3880 
3881 #if IS_ENABLED(CONFIG_IPV6)
3882 		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3883 			NL_SET_ERR_MSG(extack,
3884 				       "Local interface required for link-local local/remote addresses");
3885 			return -EINVAL;
3886 		}
3887 #endif
3888 
3889 		*lower = NULL;
3890 	}
3891 
3892 	if (!conf->dst_port) {
3893 		if (conf->flags & VXLAN_F_GPE)
3894 			conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3895 		else
3896 			conf->dst_port = htons(vxlan_port);
3897 	}
3898 
3899 	if (!conf->age_interval)
3900 		conf->age_interval = FDB_AGE_DEFAULT;
3901 
3902 	if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3903 		NL_SET_ERR_MSG(extack,
3904 			       "A VXLAN device with the specified VNI already exists");
3905 		return -EEXIST;
3906 	}
3907 
3908 	return 0;
3909 }
3910 
3911 static void vxlan_config_apply(struct net_device *dev,
3912 			       struct vxlan_config *conf,
3913 			       struct net_device *lowerdev,
3914 			       struct net *src_net,
3915 			       bool changelink)
3916 {
3917 	struct vxlan_dev *vxlan = netdev_priv(dev);
3918 	struct vxlan_rdst *dst = &vxlan->default_dst;
3919 	unsigned short needed_headroom = ETH_HLEN;
3920 	int max_mtu = ETH_MAX_MTU;
3921 	u32 flags = conf->flags;
3922 
3923 	if (!changelink) {
3924 		if (flags & VXLAN_F_GPE)
3925 			vxlan_raw_setup(dev);
3926 		else
3927 			vxlan_ether_setup(dev);
3928 
3929 		if (conf->mtu)
3930 			dev->mtu = conf->mtu;
3931 
3932 		vxlan->net = src_net;
3933 	}
3934 
3935 	dst->remote_vni = conf->vni;
3936 
3937 	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3938 
3939 	if (lowerdev) {
3940 		dst->remote_ifindex = conf->remote_ifindex;
3941 
3942 		netif_inherit_tso_max(dev, lowerdev);
3943 
3944 		needed_headroom = lowerdev->hard_header_len;
3945 		needed_headroom += lowerdev->needed_headroom;
3946 
3947 		dev->needed_tailroom = lowerdev->needed_tailroom;
3948 
3949 		max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3950 		if (max_mtu < ETH_MIN_MTU)
3951 			max_mtu = ETH_MIN_MTU;
3952 
3953 		if (!changelink && !conf->mtu)
3954 			dev->mtu = max_mtu;
3955 	}
3956 
3957 	if (dev->mtu > max_mtu)
3958 		dev->mtu = max_mtu;
3959 
3960 	if (flags & VXLAN_F_COLLECT_METADATA)
3961 		flags |= VXLAN_F_IPV6;
3962 	needed_headroom += vxlan_headroom(flags);
3963 	dev->needed_headroom = needed_headroom;
3964 
3965 	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3966 }
3967 
3968 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3969 			       struct vxlan_config *conf, bool changelink,
3970 			       struct netlink_ext_ack *extack)
3971 {
3972 	struct vxlan_dev *vxlan = netdev_priv(dev);
3973 	struct net_device *lowerdev;
3974 	int ret;
3975 
3976 	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3977 	if (ret)
3978 		return ret;
3979 
3980 	vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3981 
3982 	return 0;
3983 }
3984 
3985 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3986 			      struct vxlan_config *conf,
3987 			      struct netlink_ext_ack *extack)
3988 {
3989 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3990 	struct vxlan_dev *vxlan = netdev_priv(dev);
3991 	struct net_device *remote_dev = NULL;
3992 	struct vxlan_fdb *f = NULL;
3993 	bool unregister = false;
3994 	struct vxlan_rdst *dst;
3995 	int err;
3996 
3997 	dst = &vxlan->default_dst;
3998 	err = vxlan_dev_configure(net, dev, conf, false, extack);
3999 	if (err)
4000 		return err;
4001 
4002 	dev->ethtool_ops = &vxlan_ethtool_ops;
4003 
4004 	/* create an fdb entry for a valid default destination */
4005 	if (!vxlan_addr_any(&dst->remote_ip)) {
4006 		err = vxlan_fdb_create(vxlan, all_zeros_mac,
4007 				       &dst->remote_ip,
4008 				       NUD_REACHABLE | NUD_PERMANENT,
4009 				       vxlan->cfg.dst_port,
4010 				       dst->remote_vni,
4011 				       dst->remote_vni,
4012 				       dst->remote_ifindex,
4013 				       NTF_SELF, 0, &f, extack);
4014 		if (err)
4015 			return err;
4016 	}
4017 
4018 	err = register_netdevice(dev);
4019 	if (err)
4020 		goto errout;
4021 	unregister = true;
4022 
4023 	if (dst->remote_ifindex) {
4024 		remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
4025 		if (!remote_dev) {
4026 			err = -ENODEV;
4027 			goto errout;
4028 		}
4029 
4030 		err = netdev_upper_dev_link(remote_dev, dev, extack);
4031 		if (err)
4032 			goto errout;
4033 	}
4034 
4035 	err = rtnl_configure_link(dev, NULL, 0, NULL);
4036 	if (err < 0)
4037 		goto unlink;
4038 
4039 	if (f) {
4040 		vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
4041 
4042 		/* notify default fdb entry */
4043 		err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4044 				       RTM_NEWNEIGH, true, extack);
4045 		if (err) {
4046 			vxlan_fdb_destroy(vxlan, f, false, false);
4047 			if (remote_dev)
4048 				netdev_upper_dev_unlink(remote_dev, dev);
4049 			goto unregister;
4050 		}
4051 	}
4052 
4053 	list_add(&vxlan->next, &vn->vxlan_list);
4054 	if (remote_dev)
4055 		dst->remote_dev = remote_dev;
4056 	return 0;
4057 unlink:
4058 	if (remote_dev)
4059 		netdev_upper_dev_unlink(remote_dev, dev);
4060 errout:
4061 	/* unregister_netdevice() destroys the default FDB entry with deletion
4062 	 * notification. But the addition notification was not sent yet, so
4063 	 * destroy the entry by hand here.
4064 	 */
4065 	if (f)
4066 		__vxlan_fdb_free(f);
4067 unregister:
4068 	if (unregister)
4069 		unregister_netdevice(dev);
4070 	return err;
4071 }
4072 
4073 /* Set/clear flags based on attribute */
4074 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
4075 			  int attrtype, unsigned long mask, bool changelink,
4076 			  bool changelink_supported,
4077 			  struct netlink_ext_ack *extack)
4078 {
4079 	unsigned long flags;
4080 
4081 	if (!tb[attrtype])
4082 		return 0;
4083 
4084 	if (changelink && !changelink_supported) {
4085 		vxlan_flag_attr_error(attrtype, extack);
4086 		return -EOPNOTSUPP;
4087 	}
4088 
4089 	if (vxlan_policy[attrtype].type == NLA_FLAG)
4090 		flags = conf->flags | mask;
4091 	else if (nla_get_u8(tb[attrtype]))
4092 		flags = conf->flags | mask;
4093 	else
4094 		flags = conf->flags & ~mask;
4095 
4096 	conf->flags = flags;
4097 
4098 	return 0;
4099 }
4100 
4101 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4102 			 struct net_device *dev, struct vxlan_config *conf,
4103 			 bool changelink, struct netlink_ext_ack *extack)
4104 {
4105 	struct vxlan_dev *vxlan = netdev_priv(dev);
4106 	int err = 0;
4107 
4108 	memset(conf, 0, sizeof(*conf));
4109 
4110 	/* if changelink operation, start with old existing cfg */
4111 	if (changelink)
4112 		memcpy(conf, &vxlan->cfg, sizeof(*conf));
4113 
4114 	if (data[IFLA_VXLAN_ID]) {
4115 		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4116 
4117 		if (changelink && (vni != conf->vni)) {
4118 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
4119 			return -EOPNOTSUPP;
4120 		}
4121 		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4122 	}
4123 
4124 	if (data[IFLA_VXLAN_GROUP]) {
4125 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4126 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4127 			return -EOPNOTSUPP;
4128 		}
4129 
4130 		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4131 		conf->remote_ip.sa.sa_family = AF_INET;
4132 	} else if (data[IFLA_VXLAN_GROUP6]) {
4133 		if (!IS_ENABLED(CONFIG_IPV6)) {
4134 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4135 			return -EPFNOSUPPORT;
4136 		}
4137 
4138 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4139 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4140 			return -EOPNOTSUPP;
4141 		}
4142 
4143 		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4144 		conf->remote_ip.sa.sa_family = AF_INET6;
4145 	}
4146 
4147 	if (data[IFLA_VXLAN_LOCAL]) {
4148 		if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4149 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4150 			return -EOPNOTSUPP;
4151 		}
4152 
4153 		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4154 		conf->saddr.sa.sa_family = AF_INET;
4155 	} else if (data[IFLA_VXLAN_LOCAL6]) {
4156 		if (!IS_ENABLED(CONFIG_IPV6)) {
4157 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4158 			return -EPFNOSUPPORT;
4159 		}
4160 
4161 		if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4162 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4163 			return -EOPNOTSUPP;
4164 		}
4165 
4166 		/* TODO: respect scope id */
4167 		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4168 		conf->saddr.sa.sa_family = AF_INET6;
4169 	}
4170 
4171 	if (data[IFLA_VXLAN_LINK])
4172 		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4173 
4174 	if (data[IFLA_VXLAN_TOS])
4175 		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4176 
4177 	if (data[IFLA_VXLAN_TTL])
4178 		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4179 
4180 	if (data[IFLA_VXLAN_TTL_INHERIT]) {
4181 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4182 				    VXLAN_F_TTL_INHERIT, changelink, false,
4183 				    extack);
4184 		if (err)
4185 			return err;
4186 
4187 	}
4188 
4189 	if (data[IFLA_VXLAN_LABEL])
4190 		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4191 			     IPV6_FLOWLABEL_MASK;
4192 
4193 	if (data[IFLA_VXLAN_LEARNING]) {
4194 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4195 				    VXLAN_F_LEARN, changelink, true,
4196 				    extack);
4197 		if (err)
4198 			return err;
4199 	} else if (!changelink) {
4200 		/* default to learn on a new device */
4201 		conf->flags |= VXLAN_F_LEARN;
4202 	}
4203 
4204 	if (data[IFLA_VXLAN_AGEING])
4205 		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4206 
4207 	if (data[IFLA_VXLAN_PROXY]) {
4208 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4209 				    VXLAN_F_PROXY, changelink, false,
4210 				    extack);
4211 		if (err)
4212 			return err;
4213 	}
4214 
4215 	if (data[IFLA_VXLAN_RSC]) {
4216 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4217 				    VXLAN_F_RSC, changelink, false,
4218 				    extack);
4219 		if (err)
4220 			return err;
4221 	}
4222 
4223 	if (data[IFLA_VXLAN_L2MISS]) {
4224 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4225 				    VXLAN_F_L2MISS, changelink, false,
4226 				    extack);
4227 		if (err)
4228 			return err;
4229 	}
4230 
4231 	if (data[IFLA_VXLAN_L3MISS]) {
4232 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4233 				    VXLAN_F_L3MISS, changelink, false,
4234 				    extack);
4235 		if (err)
4236 			return err;
4237 	}
4238 
4239 	if (data[IFLA_VXLAN_LIMIT]) {
4240 		if (changelink) {
4241 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4242 					    "Cannot change limit");
4243 			return -EOPNOTSUPP;
4244 		}
4245 		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4246 	}
4247 
4248 	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4249 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4250 				    VXLAN_F_COLLECT_METADATA, changelink, false,
4251 				    extack);
4252 		if (err)
4253 			return err;
4254 	}
4255 
4256 	if (data[IFLA_VXLAN_PORT_RANGE]) {
4257 		if (!changelink) {
4258 			const struct ifla_vxlan_port_range *p
4259 				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4260 			conf->port_min = ntohs(p->low);
4261 			conf->port_max = ntohs(p->high);
4262 		} else {
4263 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4264 					    "Cannot change port range");
4265 			return -EOPNOTSUPP;
4266 		}
4267 	}
4268 
4269 	if (data[IFLA_VXLAN_PORT]) {
4270 		if (changelink) {
4271 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4272 					    "Cannot change port");
4273 			return -EOPNOTSUPP;
4274 		}
4275 		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4276 	}
4277 
4278 	if (data[IFLA_VXLAN_UDP_CSUM]) {
4279 		if (changelink) {
4280 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4281 					    "Cannot change UDP_CSUM flag");
4282 			return -EOPNOTSUPP;
4283 		}
4284 		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4285 			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4286 	}
4287 
4288 	if (data[IFLA_VXLAN_LOCALBYPASS]) {
4289 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4290 				    VXLAN_F_LOCALBYPASS, changelink,
4291 				    true, extack);
4292 		if (err)
4293 			return err;
4294 	} else if (!changelink) {
4295 		/* default to local bypass on a new device */
4296 		conf->flags |= VXLAN_F_LOCALBYPASS;
4297 	}
4298 
4299 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4300 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4301 				    VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4302 				    false, extack);
4303 		if (err)
4304 			return err;
4305 	}
4306 
4307 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4308 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4309 				    VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4310 				    false, extack);
4311 		if (err)
4312 			return err;
4313 	}
4314 
4315 	if (data[IFLA_VXLAN_REMCSUM_TX]) {
4316 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4317 				    VXLAN_F_REMCSUM_TX, changelink, false,
4318 				    extack);
4319 		if (err)
4320 			return err;
4321 	}
4322 
4323 	if (data[IFLA_VXLAN_REMCSUM_RX]) {
4324 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4325 				    VXLAN_F_REMCSUM_RX, changelink, false,
4326 				    extack);
4327 		if (err)
4328 			return err;
4329 	}
4330 
4331 	if (data[IFLA_VXLAN_GBP]) {
4332 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4333 				    VXLAN_F_GBP, changelink, false, extack);
4334 		if (err)
4335 			return err;
4336 	}
4337 
4338 	if (data[IFLA_VXLAN_GPE]) {
4339 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4340 				    VXLAN_F_GPE, changelink, false,
4341 				    extack);
4342 		if (err)
4343 			return err;
4344 	}
4345 
4346 	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4347 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4348 				    VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4349 				    false, extack);
4350 		if (err)
4351 			return err;
4352 	}
4353 
4354 	if (tb[IFLA_MTU]) {
4355 		if (changelink) {
4356 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4357 					    "Cannot change mtu");
4358 			return -EOPNOTSUPP;
4359 		}
4360 		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4361 	}
4362 
4363 	if (data[IFLA_VXLAN_DF])
4364 		conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4365 
4366 	if (data[IFLA_VXLAN_VNIFILTER]) {
4367 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4368 				    VXLAN_F_VNIFILTER, changelink, false,
4369 				    extack);
4370 		if (err)
4371 			return err;
4372 
4373 		if ((conf->flags & VXLAN_F_VNIFILTER) &&
4374 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4375 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4376 					    "vxlan vnifilter only valid in collect metadata mode");
4377 			return -EINVAL;
4378 		}
4379 	}
4380 
4381 	return 0;
4382 }
4383 
4384 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4385 			 struct nlattr *tb[], struct nlattr *data[],
4386 			 struct netlink_ext_ack *extack)
4387 {
4388 	struct vxlan_config conf;
4389 	int err;
4390 
4391 	err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4392 	if (err)
4393 		return err;
4394 
4395 	return __vxlan_dev_create(src_net, dev, &conf, extack);
4396 }
4397 
4398 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4399 			    struct nlattr *data[],
4400 			    struct netlink_ext_ack *extack)
4401 {
4402 	struct vxlan_dev *vxlan = netdev_priv(dev);
4403 	struct net_device *lowerdev;
4404 	struct vxlan_config conf;
4405 	struct vxlan_rdst *dst;
4406 	int err;
4407 
4408 	dst = &vxlan->default_dst;
4409 	err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4410 	if (err)
4411 		return err;
4412 
4413 	err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4414 				    vxlan, extack);
4415 	if (err)
4416 		return err;
4417 
4418 	if (dst->remote_dev == lowerdev)
4419 		lowerdev = NULL;
4420 
4421 	err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4422 					     extack);
4423 	if (err)
4424 		return err;
4425 
4426 	/* handle default dst entry */
4427 	if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4428 		u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4429 
4430 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4431 		if (!vxlan_addr_any(&conf.remote_ip)) {
4432 			err = vxlan_fdb_update(vxlan, all_zeros_mac,
4433 					       &conf.remote_ip,
4434 					       NUD_REACHABLE | NUD_PERMANENT,
4435 					       NLM_F_APPEND | NLM_F_CREATE,
4436 					       vxlan->cfg.dst_port,
4437 					       conf.vni, conf.vni,
4438 					       conf.remote_ifindex,
4439 					       NTF_SELF, 0, true, extack);
4440 			if (err) {
4441 				spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4442 				netdev_adjacent_change_abort(dst->remote_dev,
4443 							     lowerdev, dev);
4444 				return err;
4445 			}
4446 		}
4447 		if (!vxlan_addr_any(&dst->remote_ip))
4448 			__vxlan_fdb_delete(vxlan, all_zeros_mac,
4449 					   dst->remote_ip,
4450 					   vxlan->cfg.dst_port,
4451 					   dst->remote_vni,
4452 					   dst->remote_vni,
4453 					   dst->remote_ifindex,
4454 					   true);
4455 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4456 
4457 		/* If vni filtering device, also update fdb entries of
4458 		 * all vnis that were using default remote ip
4459 		 */
4460 		if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4461 			err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4462 							 &conf.remote_ip, extack);
4463 			if (err) {
4464 				netdev_adjacent_change_abort(dst->remote_dev,
4465 							     lowerdev, dev);
4466 				return err;
4467 			}
4468 		}
4469 	}
4470 
4471 	if (conf.age_interval != vxlan->cfg.age_interval)
4472 		mod_timer(&vxlan->age_timer, jiffies);
4473 
4474 	netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4475 	if (lowerdev && lowerdev != dst->remote_dev)
4476 		dst->remote_dev = lowerdev;
4477 	vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4478 	return 0;
4479 }
4480 
4481 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4482 {
4483 	struct vxlan_dev *vxlan = netdev_priv(dev);
4484 	struct vxlan_fdb_flush_desc desc = {
4485 		/* Default entry is deleted at vxlan_uninit. */
4486 		.ignore_default_entry = true,
4487 	};
4488 
4489 	vxlan_flush(vxlan, &desc);
4490 
4491 	list_del(&vxlan->next);
4492 	unregister_netdevice_queue(dev, head);
4493 	if (vxlan->default_dst.remote_dev)
4494 		netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4495 }
4496 
4497 static size_t vxlan_get_size(const struct net_device *dev)
4498 {
4499 
4500 	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
4501 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4502 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
4503 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4504 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
4505 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL_INHERIT */
4506 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
4507 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_DF */
4508 		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4509 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
4510 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
4511 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
4512 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
4513 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
4514 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
4515 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
4516 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
4517 		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4518 		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4519 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4520 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4521 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4522 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4523 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4524 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4525 		nla_total_size(0) + /* IFLA_VXLAN_GBP */
4526 		nla_total_size(0) + /* IFLA_VXLAN_GPE */
4527 		nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */
4528 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */
4529 		0;
4530 }
4531 
4532 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4533 {
4534 	const struct vxlan_dev *vxlan = netdev_priv(dev);
4535 	const struct vxlan_rdst *dst = &vxlan->default_dst;
4536 	struct ifla_vxlan_port_range ports = {
4537 		.low =  htons(vxlan->cfg.port_min),
4538 		.high = htons(vxlan->cfg.port_max),
4539 	};
4540 
4541 	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4542 		goto nla_put_failure;
4543 
4544 	if (!vxlan_addr_any(&dst->remote_ip)) {
4545 		if (dst->remote_ip.sa.sa_family == AF_INET) {
4546 			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4547 					    dst->remote_ip.sin.sin_addr.s_addr))
4548 				goto nla_put_failure;
4549 #if IS_ENABLED(CONFIG_IPV6)
4550 		} else {
4551 			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4552 					     &dst->remote_ip.sin6.sin6_addr))
4553 				goto nla_put_failure;
4554 #endif
4555 		}
4556 	}
4557 
4558 	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4559 		goto nla_put_failure;
4560 
4561 	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4562 		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4563 			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4564 					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
4565 				goto nla_put_failure;
4566 #if IS_ENABLED(CONFIG_IPV6)
4567 		} else {
4568 			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4569 					     &vxlan->cfg.saddr.sin6.sin6_addr))
4570 				goto nla_put_failure;
4571 #endif
4572 		}
4573 	}
4574 
4575 	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4576 	    nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4577 		       !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4578 	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4579 	    nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4580 	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4581 	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4582 		       !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4583 	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
4584 		       !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4585 	    nla_put_u8(skb, IFLA_VXLAN_RSC,
4586 		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4587 	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4588 		       !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4589 	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4590 		       !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4591 	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4592 		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4593 	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4594 	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4595 	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4596 	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4597 		       !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4598 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4599 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4600 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4601 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4602 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4603 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4604 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4605 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4606 	    nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4607 		       !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4608 		goto nla_put_failure;
4609 
4610 	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4611 		goto nla_put_failure;
4612 
4613 	if (vxlan->cfg.flags & VXLAN_F_GBP &&
4614 	    nla_put_flag(skb, IFLA_VXLAN_GBP))
4615 		goto nla_put_failure;
4616 
4617 	if (vxlan->cfg.flags & VXLAN_F_GPE &&
4618 	    nla_put_flag(skb, IFLA_VXLAN_GPE))
4619 		goto nla_put_failure;
4620 
4621 	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4622 	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4623 		goto nla_put_failure;
4624 
4625 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4626 	    nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4627 		       !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4628 		goto nla_put_failure;
4629 
4630 	return 0;
4631 
4632 nla_put_failure:
4633 	return -EMSGSIZE;
4634 }
4635 
4636 static struct net *vxlan_get_link_net(const struct net_device *dev)
4637 {
4638 	struct vxlan_dev *vxlan = netdev_priv(dev);
4639 
4640 	return vxlan->net;
4641 }
4642 
4643 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4644 	.kind		= "vxlan",
4645 	.maxtype	= IFLA_VXLAN_MAX,
4646 	.policy		= vxlan_policy,
4647 	.priv_size	= sizeof(struct vxlan_dev),
4648 	.setup		= vxlan_setup,
4649 	.validate	= vxlan_validate,
4650 	.newlink	= vxlan_newlink,
4651 	.changelink	= vxlan_changelink,
4652 	.dellink	= vxlan_dellink,
4653 	.get_size	= vxlan_get_size,
4654 	.fill_info	= vxlan_fill_info,
4655 	.get_link_net	= vxlan_get_link_net,
4656 };
4657 
4658 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4659 				    u8 name_assign_type,
4660 				    struct vxlan_config *conf)
4661 {
4662 	struct nlattr *tb[IFLA_MAX + 1];
4663 	struct net_device *dev;
4664 	int err;
4665 
4666 	memset(&tb, 0, sizeof(tb));
4667 
4668 	dev = rtnl_create_link(net, name, name_assign_type,
4669 			       &vxlan_link_ops, tb, NULL);
4670 	if (IS_ERR(dev))
4671 		return dev;
4672 
4673 	err = __vxlan_dev_create(net, dev, conf, NULL);
4674 	if (err < 0) {
4675 		free_netdev(dev);
4676 		return ERR_PTR(err);
4677 	}
4678 
4679 	err = rtnl_configure_link(dev, NULL, 0, NULL);
4680 	if (err < 0) {
4681 		LIST_HEAD(list_kill);
4682 
4683 		vxlan_dellink(dev, &list_kill);
4684 		unregister_netdevice_many(&list_kill);
4685 		return ERR_PTR(err);
4686 	}
4687 
4688 	return dev;
4689 }
4690 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4691 
4692 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4693 					     struct net_device *dev)
4694 {
4695 	struct vxlan_dev *vxlan, *next;
4696 	LIST_HEAD(list_kill);
4697 
4698 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4699 		struct vxlan_rdst *dst = &vxlan->default_dst;
4700 
4701 		/* In case we created vxlan device with carrier
4702 		 * and we loose the carrier due to module unload
4703 		 * we also need to remove vxlan device. In other
4704 		 * cases, it's not necessary and remote_ifindex
4705 		 * is 0 here, so no matches.
4706 		 */
4707 		if (dst->remote_ifindex == dev->ifindex)
4708 			vxlan_dellink(vxlan->dev, &list_kill);
4709 	}
4710 
4711 	unregister_netdevice_many(&list_kill);
4712 }
4713 
4714 static int vxlan_netdevice_event(struct notifier_block *unused,
4715 				 unsigned long event, void *ptr)
4716 {
4717 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4718 	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4719 
4720 	if (event == NETDEV_UNREGISTER)
4721 		vxlan_handle_lowerdev_unregister(vn, dev);
4722 	else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4723 		vxlan_offload_rx_ports(dev, true);
4724 	else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4725 		vxlan_offload_rx_ports(dev, false);
4726 
4727 	return NOTIFY_DONE;
4728 }
4729 
4730 static struct notifier_block vxlan_notifier_block __read_mostly = {
4731 	.notifier_call = vxlan_netdevice_event,
4732 };
4733 
4734 static void
4735 vxlan_fdb_offloaded_set(struct net_device *dev,
4736 			struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4737 {
4738 	struct vxlan_dev *vxlan = netdev_priv(dev);
4739 	struct vxlan_rdst *rdst;
4740 	struct vxlan_fdb *f;
4741 	u32 hash_index;
4742 
4743 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4744 
4745 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4746 
4747 	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4748 	if (!f)
4749 		goto out;
4750 
4751 	rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4752 				   fdb_info->remote_port,
4753 				   fdb_info->remote_vni,
4754 				   fdb_info->remote_ifindex);
4755 	if (!rdst)
4756 		goto out;
4757 
4758 	rdst->offloaded = fdb_info->offloaded;
4759 
4760 out:
4761 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4762 }
4763 
4764 static int
4765 vxlan_fdb_external_learn_add(struct net_device *dev,
4766 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4767 {
4768 	struct vxlan_dev *vxlan = netdev_priv(dev);
4769 	struct netlink_ext_ack *extack;
4770 	u32 hash_index;
4771 	int err;
4772 
4773 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4774 	extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4775 
4776 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4777 	err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4778 			       NUD_REACHABLE,
4779 			       NLM_F_CREATE | NLM_F_REPLACE,
4780 			       fdb_info->remote_port,
4781 			       fdb_info->vni,
4782 			       fdb_info->remote_vni,
4783 			       fdb_info->remote_ifindex,
4784 			       NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4785 			       0, false, extack);
4786 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4787 
4788 	return err;
4789 }
4790 
4791 static int
4792 vxlan_fdb_external_learn_del(struct net_device *dev,
4793 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4794 {
4795 	struct vxlan_dev *vxlan = netdev_priv(dev);
4796 	struct vxlan_fdb *f;
4797 	u32 hash_index;
4798 	int err = 0;
4799 
4800 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4801 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4802 
4803 	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4804 	if (!f)
4805 		err = -ENOENT;
4806 	else if (f->flags & NTF_EXT_LEARNED)
4807 		err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4808 					 fdb_info->remote_ip,
4809 					 fdb_info->remote_port,
4810 					 fdb_info->vni,
4811 					 fdb_info->remote_vni,
4812 					 fdb_info->remote_ifindex,
4813 					 false);
4814 
4815 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4816 
4817 	return err;
4818 }
4819 
4820 static int vxlan_switchdev_event(struct notifier_block *unused,
4821 				 unsigned long event, void *ptr)
4822 {
4823 	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4824 	struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4825 	int err = 0;
4826 
4827 	switch (event) {
4828 	case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4829 		vxlan_fdb_offloaded_set(dev, ptr);
4830 		break;
4831 	case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4832 		fdb_info = ptr;
4833 		err = vxlan_fdb_external_learn_add(dev, fdb_info);
4834 		if (err) {
4835 			err = notifier_from_errno(err);
4836 			break;
4837 		}
4838 		fdb_info->offloaded = true;
4839 		vxlan_fdb_offloaded_set(dev, fdb_info);
4840 		break;
4841 	case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4842 		fdb_info = ptr;
4843 		err = vxlan_fdb_external_learn_del(dev, fdb_info);
4844 		if (err) {
4845 			err = notifier_from_errno(err);
4846 			break;
4847 		}
4848 		fdb_info->offloaded = false;
4849 		vxlan_fdb_offloaded_set(dev, fdb_info);
4850 		break;
4851 	}
4852 
4853 	return err;
4854 }
4855 
4856 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4857 	.notifier_call = vxlan_switchdev_event,
4858 };
4859 
4860 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4861 {
4862 	struct vxlan_fdb *fdb;
4863 	struct vxlan_dev *vxlan;
4864 	u32 hash_index;
4865 
4866 	rcu_read_lock();
4867 	list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4868 		vxlan = rcu_dereference(fdb->vdev);
4869 		WARN_ON(!vxlan);
4870 		hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4871 					    vxlan->default_dst.remote_vni);
4872 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4873 		if (!hlist_unhashed(&fdb->hlist))
4874 			vxlan_fdb_destroy(vxlan, fdb, false, false);
4875 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4876 	}
4877 	rcu_read_unlock();
4878 }
4879 
4880 static int vxlan_nexthop_event(struct notifier_block *nb,
4881 			       unsigned long event, void *ptr)
4882 {
4883 	struct nh_notifier_info *info = ptr;
4884 	struct nexthop *nh;
4885 
4886 	if (event != NEXTHOP_EVENT_DEL)
4887 		return NOTIFY_DONE;
4888 
4889 	nh = nexthop_find_by_id(info->net, info->id);
4890 	if (!nh)
4891 		return NOTIFY_DONE;
4892 
4893 	vxlan_fdb_nh_flush(nh);
4894 
4895 	return NOTIFY_DONE;
4896 }
4897 
4898 static __net_init int vxlan_init_net(struct net *net)
4899 {
4900 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4901 	unsigned int h;
4902 
4903 	INIT_LIST_HEAD(&vn->vxlan_list);
4904 	spin_lock_init(&vn->sock_lock);
4905 	vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4906 
4907 	for (h = 0; h < PORT_HASH_SIZE; ++h)
4908 		INIT_HLIST_HEAD(&vn->sock_list[h]);
4909 
4910 	return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4911 					 NULL);
4912 }
4913 
4914 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4915 {
4916 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4917 	struct vxlan_dev *vxlan, *next;
4918 	struct net_device *dev, *aux;
4919 
4920 	for_each_netdev_safe(net, dev, aux)
4921 		if (dev->rtnl_link_ops == &vxlan_link_ops)
4922 			unregister_netdevice_queue(dev, head);
4923 
4924 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4925 		/* If vxlan->dev is in the same netns, it has already been added
4926 		 * to the list by the previous loop.
4927 		 */
4928 		if (!net_eq(dev_net(vxlan->dev), net))
4929 			unregister_netdevice_queue(vxlan->dev, head);
4930 	}
4931 
4932 }
4933 
4934 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4935 {
4936 	struct net *net;
4937 	LIST_HEAD(list);
4938 	unsigned int h;
4939 
4940 	list_for_each_entry(net, net_list, exit_list) {
4941 		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4942 
4943 		unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4944 	}
4945 	rtnl_lock();
4946 	list_for_each_entry(net, net_list, exit_list)
4947 		vxlan_destroy_tunnels(net, &list);
4948 
4949 	unregister_netdevice_many(&list);
4950 	rtnl_unlock();
4951 
4952 	list_for_each_entry(net, net_list, exit_list) {
4953 		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4954 
4955 		for (h = 0; h < PORT_HASH_SIZE; ++h)
4956 			WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4957 	}
4958 }
4959 
4960 static struct pernet_operations vxlan_net_ops = {
4961 	.init = vxlan_init_net,
4962 	.exit_batch = vxlan_exit_batch_net,
4963 	.id   = &vxlan_net_id,
4964 	.size = sizeof(struct vxlan_net),
4965 };
4966 
4967 static int __init vxlan_init_module(void)
4968 {
4969 	int rc;
4970 
4971 	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4972 
4973 	rc = register_pernet_subsys(&vxlan_net_ops);
4974 	if (rc)
4975 		goto out1;
4976 
4977 	rc = register_netdevice_notifier(&vxlan_notifier_block);
4978 	if (rc)
4979 		goto out2;
4980 
4981 	rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4982 	if (rc)
4983 		goto out3;
4984 
4985 	rc = rtnl_link_register(&vxlan_link_ops);
4986 	if (rc)
4987 		goto out4;
4988 
4989 	vxlan_vnifilter_init();
4990 
4991 	return 0;
4992 out4:
4993 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4994 out3:
4995 	unregister_netdevice_notifier(&vxlan_notifier_block);
4996 out2:
4997 	unregister_pernet_subsys(&vxlan_net_ops);
4998 out1:
4999 	return rc;
5000 }
5001 late_initcall(vxlan_init_module);
5002 
5003 static void __exit vxlan_cleanup_module(void)
5004 {
5005 	vxlan_vnifilter_uninit();
5006 	rtnl_link_unregister(&vxlan_link_ops);
5007 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
5008 	unregister_netdevice_notifier(&vxlan_notifier_block);
5009 	unregister_pernet_subsys(&vxlan_net_ops);
5010 	/* rcu_barrier() is called by netns */
5011 }
5012 module_exit(vxlan_cleanup_module);
5013 
5014 MODULE_LICENSE("GPL");
5015 MODULE_VERSION(VXLAN_VERSION);
5016 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
5017 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
5018 MODULE_ALIAS_RTNL_LINK("vxlan");
5019