xref: /linux/drivers/net/vxlan/vxlan_core.c (revision d209706f562ee4fa81bdf24cf6b679c3222aa06c)
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 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
281 }
282 
283 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
284 			    const struct vxlan_fdb *fdb,
285 			    const struct vxlan_rdst *rd,
286 			    struct netlink_ext_ack *extack,
287 			    struct switchdev_notifier_vxlan_fdb_info *fdb_info)
288 {
289 	fdb_info->info.dev = vxlan->dev;
290 	fdb_info->info.extack = extack;
291 	fdb_info->remote_ip = rd->remote_ip;
292 	fdb_info->remote_port = rd->remote_port;
293 	fdb_info->remote_vni = rd->remote_vni;
294 	fdb_info->remote_ifindex = rd->remote_ifindex;
295 	memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
296 	fdb_info->vni = fdb->vni;
297 	fdb_info->offloaded = rd->offloaded;
298 	fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
299 }
300 
301 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
302 					      struct vxlan_fdb *fdb,
303 					      struct vxlan_rdst *rd,
304 					      bool adding,
305 					      struct netlink_ext_ack *extack)
306 {
307 	struct switchdev_notifier_vxlan_fdb_info info;
308 	enum switchdev_notifier_type notifier_type;
309 	int ret;
310 
311 	if (WARN_ON(!rd))
312 		return 0;
313 
314 	notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
315 			       : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
316 	vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
317 	ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
318 				       &info.info, extack);
319 	return notifier_to_errno(ret);
320 }
321 
322 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
323 			    struct vxlan_rdst *rd, int type, bool swdev_notify,
324 			    struct netlink_ext_ack *extack)
325 {
326 	int err;
327 
328 	if (swdev_notify && rd) {
329 		switch (type) {
330 		case RTM_NEWNEIGH:
331 			err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
332 								 true, extack);
333 			if (err)
334 				return err;
335 			break;
336 		case RTM_DELNEIGH:
337 			vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
338 							   false, extack);
339 			break;
340 		}
341 	}
342 
343 	__vxlan_fdb_notify(vxlan, fdb, rd, type);
344 	return 0;
345 }
346 
347 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
348 {
349 	struct vxlan_dev *vxlan = netdev_priv(dev);
350 	struct vxlan_fdb f = {
351 		.state = NUD_STALE,
352 	};
353 	struct vxlan_rdst remote = {
354 		.remote_ip = *ipa, /* goes to NDA_DST */
355 		.remote_vni = cpu_to_be32(VXLAN_N_VID),
356 	};
357 
358 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
359 }
360 
361 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
362 {
363 	struct vxlan_fdb f = {
364 		.state = NUD_STALE,
365 	};
366 	struct vxlan_rdst remote = { };
367 
368 	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
369 
370 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
371 }
372 
373 /* Hash Ethernet address */
374 static u32 eth_hash(const unsigned char *addr)
375 {
376 	u64 value = get_unaligned((u64 *)addr);
377 
378 	/* only want 6 bytes */
379 #ifdef __BIG_ENDIAN
380 	value >>= 16;
381 #else
382 	value <<= 16;
383 #endif
384 	return hash_64(value, FDB_HASH_BITS);
385 }
386 
387 u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
388 {
389 	/* use 1 byte of OUI and 3 bytes of NIC */
390 	u32 key = get_unaligned((u32 *)(addr + 2));
391 
392 	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
393 }
394 
395 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
396 {
397 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
398 		return eth_vni_hash(mac, vni);
399 	else
400 		return eth_hash(mac);
401 }
402 
403 /* Hash chain to use given mac address */
404 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
405 						const u8 *mac, __be32 vni)
406 {
407 	return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
408 }
409 
410 /* Look up Ethernet address in forwarding table */
411 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
412 					  const u8 *mac, __be32 vni)
413 {
414 	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
415 	struct vxlan_fdb *f;
416 
417 	hlist_for_each_entry_rcu(f, head, hlist) {
418 		if (ether_addr_equal(mac, f->eth_addr)) {
419 			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
420 				if (vni == f->vni)
421 					return f;
422 			} else {
423 				return f;
424 			}
425 		}
426 	}
427 
428 	return NULL;
429 }
430 
431 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
432 					const u8 *mac, __be32 vni)
433 {
434 	struct vxlan_fdb *f;
435 
436 	f = __vxlan_find_mac(vxlan, mac, vni);
437 	if (f && f->used != jiffies)
438 		f->used = jiffies;
439 
440 	return f;
441 }
442 
443 /* caller should hold vxlan->hash_lock */
444 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
445 					      union vxlan_addr *ip, __be16 port,
446 					      __be32 vni, __u32 ifindex)
447 {
448 	struct vxlan_rdst *rd;
449 
450 	list_for_each_entry(rd, &f->remotes, list) {
451 		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
452 		    rd->remote_port == port &&
453 		    rd->remote_vni == vni &&
454 		    rd->remote_ifindex == ifindex)
455 			return rd;
456 	}
457 
458 	return NULL;
459 }
460 
461 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
462 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info)
463 {
464 	struct vxlan_dev *vxlan = netdev_priv(dev);
465 	u8 eth_addr[ETH_ALEN + 2] = { 0 };
466 	struct vxlan_rdst *rdst;
467 	struct vxlan_fdb *f;
468 	int rc = 0;
469 
470 	if (is_multicast_ether_addr(mac) ||
471 	    is_zero_ether_addr(mac))
472 		return -EINVAL;
473 
474 	ether_addr_copy(eth_addr, mac);
475 
476 	rcu_read_lock();
477 
478 	f = __vxlan_find_mac(vxlan, eth_addr, vni);
479 	if (!f) {
480 		rc = -ENOENT;
481 		goto out;
482 	}
483 
484 	rdst = first_remote_rcu(f);
485 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
486 
487 out:
488 	rcu_read_unlock();
489 	return rc;
490 }
491 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
492 
493 static int vxlan_fdb_notify_one(struct notifier_block *nb,
494 				const struct vxlan_dev *vxlan,
495 				const struct vxlan_fdb *f,
496 				const struct vxlan_rdst *rdst,
497 				struct netlink_ext_ack *extack)
498 {
499 	struct switchdev_notifier_vxlan_fdb_info fdb_info;
500 	int rc;
501 
502 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
503 	rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
504 			       &fdb_info);
505 	return notifier_to_errno(rc);
506 }
507 
508 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
509 		     struct notifier_block *nb,
510 		     struct netlink_ext_ack *extack)
511 {
512 	struct vxlan_dev *vxlan;
513 	struct vxlan_rdst *rdst;
514 	struct vxlan_fdb *f;
515 	unsigned int h;
516 	int rc = 0;
517 
518 	if (!netif_is_vxlan(dev))
519 		return -EINVAL;
520 	vxlan = netdev_priv(dev);
521 
522 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
523 		spin_lock_bh(&vxlan->hash_lock[h]);
524 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
525 			if (f->vni == vni) {
526 				list_for_each_entry(rdst, &f->remotes, list) {
527 					rc = vxlan_fdb_notify_one(nb, vxlan,
528 								  f, rdst,
529 								  extack);
530 					if (rc)
531 						goto unlock;
532 				}
533 			}
534 		}
535 		spin_unlock_bh(&vxlan->hash_lock[h]);
536 	}
537 	return 0;
538 
539 unlock:
540 	spin_unlock_bh(&vxlan->hash_lock[h]);
541 	return rc;
542 }
543 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
544 
545 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
546 {
547 	struct vxlan_dev *vxlan;
548 	struct vxlan_rdst *rdst;
549 	struct vxlan_fdb *f;
550 	unsigned int h;
551 
552 	if (!netif_is_vxlan(dev))
553 		return;
554 	vxlan = netdev_priv(dev);
555 
556 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
557 		spin_lock_bh(&vxlan->hash_lock[h]);
558 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
559 			if (f->vni == vni)
560 				list_for_each_entry(rdst, &f->remotes, list)
561 					rdst->offloaded = false;
562 		spin_unlock_bh(&vxlan->hash_lock[h]);
563 	}
564 
565 }
566 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
567 
568 /* Replace destination of unicast mac */
569 static int vxlan_fdb_replace(struct vxlan_fdb *f,
570 			     union vxlan_addr *ip, __be16 port, __be32 vni,
571 			     __u32 ifindex, struct vxlan_rdst *oldrd)
572 {
573 	struct vxlan_rdst *rd;
574 
575 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
576 	if (rd)
577 		return 0;
578 
579 	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
580 	if (!rd)
581 		return 0;
582 
583 	*oldrd = *rd;
584 	dst_cache_reset(&rd->dst_cache);
585 	rd->remote_ip = *ip;
586 	rd->remote_port = port;
587 	rd->remote_vni = vni;
588 	rd->remote_ifindex = ifindex;
589 	rd->offloaded = false;
590 	return 1;
591 }
592 
593 /* Add/update destinations for multicast */
594 static int vxlan_fdb_append(struct vxlan_fdb *f,
595 			    union vxlan_addr *ip, __be16 port, __be32 vni,
596 			    __u32 ifindex, struct vxlan_rdst **rdp)
597 {
598 	struct vxlan_rdst *rd;
599 
600 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
601 	if (rd)
602 		return 0;
603 
604 	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
605 	if (rd == NULL)
606 		return -ENOMEM;
607 
608 	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
609 		kfree(rd);
610 		return -ENOMEM;
611 	}
612 
613 	rd->remote_ip = *ip;
614 	rd->remote_port = port;
615 	rd->offloaded = false;
616 	rd->remote_vni = vni;
617 	rd->remote_ifindex = ifindex;
618 
619 	list_add_tail_rcu(&rd->list, &f->remotes);
620 
621 	*rdp = rd;
622 	return 1;
623 }
624 
625 static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol)
626 {
627 	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr;
628 
629 	/* Need to have Next Protocol set for interfaces in GPE mode. */
630 	if (!gpe->np_applied)
631 		return false;
632 	/* "The initial version is 0. If a receiver does not support the
633 	 * version indicated it MUST drop the packet.
634 	 */
635 	if (gpe->version != 0)
636 		return false;
637 	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
638 	 * processing MUST occur." However, we don't implement OAM
639 	 * processing, thus drop the packet.
640 	 */
641 	if (gpe->oam_flag)
642 		return false;
643 
644 	*protocol = tun_p_to_eth_p(gpe->next_protocol);
645 	if (!*protocol)
646 		return false;
647 
648 	return true;
649 }
650 
651 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
652 					  unsigned int off,
653 					  struct vxlanhdr *vh, size_t hdrlen,
654 					  __be32 vni_field,
655 					  struct gro_remcsum *grc,
656 					  bool nopartial)
657 {
658 	size_t start, offset;
659 
660 	if (skb->remcsum_offload)
661 		return vh;
662 
663 	if (!NAPI_GRO_CB(skb)->csum_valid)
664 		return NULL;
665 
666 	start = vxlan_rco_start(vni_field);
667 	offset = start + vxlan_rco_offset(vni_field);
668 
669 	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
670 				     start, offset, grc, nopartial);
671 
672 	skb->remcsum_offload = 1;
673 
674 	return vh;
675 }
676 
677 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
678 						  struct list_head *head,
679 						  struct sk_buff *skb,
680 						  struct gro_remcsum *grc)
681 {
682 	struct sk_buff *p;
683 	struct vxlanhdr *vh, *vh2;
684 	unsigned int hlen, off_vx;
685 	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
686 	__be32 flags;
687 
688 	skb_gro_remcsum_init(grc);
689 
690 	off_vx = skb_gro_offset(skb);
691 	hlen = off_vx + sizeof(*vh);
692 	vh = skb_gro_header(skb, hlen, off_vx);
693 	if (unlikely(!vh))
694 		return NULL;
695 
696 	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
697 
698 	flags = vh->vx_flags;
699 
700 	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
701 		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
702 				       vh->vx_vni, grc,
703 				       !!(vs->flags &
704 					  VXLAN_F_REMCSUM_NOPARTIAL));
705 
706 		if (!vh)
707 			return NULL;
708 	}
709 
710 	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
711 
712 	list_for_each_entry(p, head, list) {
713 		if (!NAPI_GRO_CB(p)->same_flow)
714 			continue;
715 
716 		vh2 = (struct vxlanhdr *)(p->data + off_vx);
717 		if (vh->vx_flags != vh2->vx_flags ||
718 		    vh->vx_vni != vh2->vx_vni) {
719 			NAPI_GRO_CB(p)->same_flow = 0;
720 			continue;
721 		}
722 	}
723 
724 	return vh;
725 }
726 
727 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
728 					 struct list_head *head,
729 					 struct sk_buff *skb)
730 {
731 	struct sk_buff *pp = NULL;
732 	struct gro_remcsum grc;
733 	int flush = 1;
734 
735 	if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
736 		pp = call_gro_receive(eth_gro_receive, head, skb);
737 		flush = 0;
738 	}
739 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
740 	return pp;
741 }
742 
743 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
744 					     struct list_head *head,
745 					     struct sk_buff *skb)
746 {
747 	const struct packet_offload *ptype;
748 	struct sk_buff *pp = NULL;
749 	struct gro_remcsum grc;
750 	struct vxlanhdr *vh;
751 	__be16 protocol;
752 	int flush = 1;
753 
754 	vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
755 	if (vh) {
756 		if (!vxlan_parse_gpe_proto(vh, &protocol))
757 			goto out;
758 		ptype = gro_find_receive_by_type(protocol);
759 		if (!ptype)
760 			goto out;
761 		pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
762 		flush = 0;
763 	}
764 out:
765 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
766 	return pp;
767 }
768 
769 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
770 {
771 	/* Sets 'skb->inner_mac_header' since we are always called with
772 	 * 'skb->encapsulation' set.
773 	 */
774 	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
775 }
776 
777 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
778 {
779 	struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
780 	const struct packet_offload *ptype;
781 	int err = -ENOSYS;
782 	__be16 protocol;
783 
784 	if (!vxlan_parse_gpe_proto(vh, &protocol))
785 		return err;
786 	ptype = gro_find_complete_by_type(protocol);
787 	if (ptype)
788 		err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
789 	return err;
790 }
791 
792 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
793 					 __u16 state, __be32 src_vni,
794 					 __u16 ndm_flags)
795 {
796 	struct vxlan_fdb *f;
797 
798 	f = kmalloc(sizeof(*f), GFP_ATOMIC);
799 	if (!f)
800 		return NULL;
801 	f->state = state;
802 	f->flags = ndm_flags;
803 	f->updated = f->used = jiffies;
804 	f->vni = src_vni;
805 	f->nh = NULL;
806 	RCU_INIT_POINTER(f->vdev, vxlan);
807 	INIT_LIST_HEAD(&f->nh_list);
808 	INIT_LIST_HEAD(&f->remotes);
809 	memcpy(f->eth_addr, mac, ETH_ALEN);
810 
811 	return f;
812 }
813 
814 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
815 			     __be32 src_vni, struct vxlan_fdb *f)
816 {
817 	++vxlan->addrcnt;
818 	hlist_add_head_rcu(&f->hlist,
819 			   vxlan_fdb_head(vxlan, mac, src_vni));
820 }
821 
822 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
823 			       u32 nhid, struct netlink_ext_ack *extack)
824 {
825 	struct nexthop *old_nh = rtnl_dereference(fdb->nh);
826 	struct nexthop *nh;
827 	int err = -EINVAL;
828 
829 	if (old_nh && old_nh->id == nhid)
830 		return 0;
831 
832 	nh = nexthop_find_by_id(vxlan->net, nhid);
833 	if (!nh) {
834 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
835 		goto err_inval;
836 	}
837 
838 	if (!nexthop_get(nh)) {
839 		NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
840 		nh = NULL;
841 		goto err_inval;
842 	}
843 	if (!nexthop_is_fdb(nh)) {
844 		NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
845 		goto err_inval;
846 	}
847 
848 	if (!nexthop_is_multipath(nh)) {
849 		NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
850 		goto err_inval;
851 	}
852 
853 	/* check nexthop group family */
854 	switch (vxlan->default_dst.remote_ip.sa.sa_family) {
855 	case AF_INET:
856 		if (!nexthop_has_v4(nh)) {
857 			err = -EAFNOSUPPORT;
858 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
859 			goto err_inval;
860 		}
861 		break;
862 	case AF_INET6:
863 		if (nexthop_has_v4(nh)) {
864 			err = -EAFNOSUPPORT;
865 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
866 			goto err_inval;
867 		}
868 	}
869 
870 	if (old_nh) {
871 		list_del_rcu(&fdb->nh_list);
872 		nexthop_put(old_nh);
873 	}
874 	rcu_assign_pointer(fdb->nh, nh);
875 	list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
876 	return 1;
877 
878 err_inval:
879 	if (nh)
880 		nexthop_put(nh);
881 	return err;
882 }
883 
884 int vxlan_fdb_create(struct vxlan_dev *vxlan,
885 		     const u8 *mac, union vxlan_addr *ip,
886 		     __u16 state, __be16 port, __be32 src_vni,
887 		     __be32 vni, __u32 ifindex, __u16 ndm_flags,
888 		     u32 nhid, struct vxlan_fdb **fdb,
889 		     struct netlink_ext_ack *extack)
890 {
891 	struct vxlan_rdst *rd = NULL;
892 	struct vxlan_fdb *f;
893 	int rc;
894 
895 	if (vxlan->cfg.addrmax &&
896 	    vxlan->addrcnt >= vxlan->cfg.addrmax)
897 		return -ENOSPC;
898 
899 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
900 	f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
901 	if (!f)
902 		return -ENOMEM;
903 
904 	if (nhid)
905 		rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
906 	else
907 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
908 	if (rc < 0)
909 		goto errout;
910 
911 	*fdb = f;
912 
913 	return 0;
914 
915 errout:
916 	kfree(f);
917 	return rc;
918 }
919 
920 static void __vxlan_fdb_free(struct vxlan_fdb *f)
921 {
922 	struct vxlan_rdst *rd, *nd;
923 	struct nexthop *nh;
924 
925 	nh = rcu_dereference_raw(f->nh);
926 	if (nh) {
927 		rcu_assign_pointer(f->nh, NULL);
928 		rcu_assign_pointer(f->vdev, NULL);
929 		nexthop_put(nh);
930 	}
931 
932 	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
933 		dst_cache_destroy(&rd->dst_cache);
934 		kfree(rd);
935 	}
936 	kfree(f);
937 }
938 
939 static void vxlan_fdb_free(struct rcu_head *head)
940 {
941 	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
942 
943 	__vxlan_fdb_free(f);
944 }
945 
946 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
947 			      bool do_notify, bool swdev_notify)
948 {
949 	struct vxlan_rdst *rd;
950 
951 	netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
952 
953 	--vxlan->addrcnt;
954 	if (do_notify) {
955 		if (rcu_access_pointer(f->nh))
956 			vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
957 					 swdev_notify, NULL);
958 		else
959 			list_for_each_entry(rd, &f->remotes, list)
960 				vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
961 						 swdev_notify, NULL);
962 	}
963 
964 	hlist_del_rcu(&f->hlist);
965 	list_del_rcu(&f->nh_list);
966 	call_rcu(&f->rcu, vxlan_fdb_free);
967 }
968 
969 static void vxlan_dst_free(struct rcu_head *head)
970 {
971 	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
972 
973 	dst_cache_destroy(&rd->dst_cache);
974 	kfree(rd);
975 }
976 
977 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
978 				     union vxlan_addr *ip,
979 				     __u16 state, __u16 flags,
980 				     __be16 port, __be32 vni,
981 				     __u32 ifindex, __u16 ndm_flags,
982 				     struct vxlan_fdb *f, u32 nhid,
983 				     bool swdev_notify,
984 				     struct netlink_ext_ack *extack)
985 {
986 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
987 	struct vxlan_rdst *rd = NULL;
988 	struct vxlan_rdst oldrd;
989 	int notify = 0;
990 	int rc = 0;
991 	int err;
992 
993 	if (nhid && !rcu_access_pointer(f->nh)) {
994 		NL_SET_ERR_MSG(extack,
995 			       "Cannot replace an existing non nexthop fdb with a nexthop");
996 		return -EOPNOTSUPP;
997 	}
998 
999 	if (nhid && (flags & NLM_F_APPEND)) {
1000 		NL_SET_ERR_MSG(extack,
1001 			       "Cannot append to a nexthop fdb");
1002 		return -EOPNOTSUPP;
1003 	}
1004 
1005 	/* Do not allow an externally learned entry to take over an entry added
1006 	 * by the user.
1007 	 */
1008 	if (!(fdb_flags & NTF_EXT_LEARNED) ||
1009 	    !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1010 		if (f->state != state) {
1011 			f->state = state;
1012 			f->updated = jiffies;
1013 			notify = 1;
1014 		}
1015 		if (f->flags != fdb_flags) {
1016 			f->flags = fdb_flags;
1017 			f->updated = jiffies;
1018 			notify = 1;
1019 		}
1020 	}
1021 
1022 	if ((flags & NLM_F_REPLACE)) {
1023 		/* Only change unicasts */
1024 		if (!(is_multicast_ether_addr(f->eth_addr) ||
1025 		      is_zero_ether_addr(f->eth_addr))) {
1026 			if (nhid) {
1027 				rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1028 				if (rc < 0)
1029 					return rc;
1030 			} else {
1031 				rc = vxlan_fdb_replace(f, ip, port, vni,
1032 						       ifindex, &oldrd);
1033 			}
1034 			notify |= rc;
1035 		} else {
1036 			NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1037 			return -EOPNOTSUPP;
1038 		}
1039 	}
1040 	if ((flags & NLM_F_APPEND) &&
1041 	    (is_multicast_ether_addr(f->eth_addr) ||
1042 	     is_zero_ether_addr(f->eth_addr))) {
1043 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1044 
1045 		if (rc < 0)
1046 			return rc;
1047 		notify |= rc;
1048 	}
1049 
1050 	if (ndm_flags & NTF_USE)
1051 		f->used = jiffies;
1052 
1053 	if (notify) {
1054 		if (rd == NULL)
1055 			rd = first_remote_rtnl(f);
1056 
1057 		err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1058 				       swdev_notify, extack);
1059 		if (err)
1060 			goto err_notify;
1061 	}
1062 
1063 	return 0;
1064 
1065 err_notify:
1066 	if (nhid)
1067 		return err;
1068 	if ((flags & NLM_F_REPLACE) && rc)
1069 		*rd = oldrd;
1070 	else if ((flags & NLM_F_APPEND) && rc) {
1071 		list_del_rcu(&rd->list);
1072 		call_rcu(&rd->rcu, vxlan_dst_free);
1073 	}
1074 	return err;
1075 }
1076 
1077 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1078 				   const u8 *mac, union vxlan_addr *ip,
1079 				   __u16 state, __u16 flags,
1080 				   __be16 port, __be32 src_vni, __be32 vni,
1081 				   __u32 ifindex, __u16 ndm_flags, u32 nhid,
1082 				   bool swdev_notify,
1083 				   struct netlink_ext_ack *extack)
1084 {
1085 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1086 	struct vxlan_fdb *f;
1087 	int rc;
1088 
1089 	/* Disallow replace to add a multicast entry */
1090 	if ((flags & NLM_F_REPLACE) &&
1091 	    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1092 		return -EOPNOTSUPP;
1093 
1094 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1095 	rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1096 			      vni, ifindex, fdb_flags, nhid, &f, extack);
1097 	if (rc < 0)
1098 		return rc;
1099 
1100 	vxlan_fdb_insert(vxlan, mac, src_vni, f);
1101 	rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1102 			      swdev_notify, extack);
1103 	if (rc)
1104 		goto err_notify;
1105 
1106 	return 0;
1107 
1108 err_notify:
1109 	vxlan_fdb_destroy(vxlan, f, false, false);
1110 	return rc;
1111 }
1112 
1113 /* Add new entry to forwarding table -- assumes lock held */
1114 int vxlan_fdb_update(struct vxlan_dev *vxlan,
1115 		     const u8 *mac, union vxlan_addr *ip,
1116 		     __u16 state, __u16 flags,
1117 		     __be16 port, __be32 src_vni, __be32 vni,
1118 		     __u32 ifindex, __u16 ndm_flags, u32 nhid,
1119 		     bool swdev_notify,
1120 		     struct netlink_ext_ack *extack)
1121 {
1122 	struct vxlan_fdb *f;
1123 
1124 	f = __vxlan_find_mac(vxlan, mac, src_vni);
1125 	if (f) {
1126 		if (flags & NLM_F_EXCL) {
1127 			netdev_dbg(vxlan->dev,
1128 				   "lost race to create %pM\n", mac);
1129 			return -EEXIST;
1130 		}
1131 
1132 		return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1133 						 vni, ifindex, ndm_flags, f,
1134 						 nhid, swdev_notify, extack);
1135 	} else {
1136 		if (!(flags & NLM_F_CREATE))
1137 			return -ENOENT;
1138 
1139 		return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1140 					       port, src_vni, vni, ifindex,
1141 					       ndm_flags, nhid, swdev_notify,
1142 					       extack);
1143 	}
1144 }
1145 
1146 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1147 				  struct vxlan_rdst *rd, bool swdev_notify)
1148 {
1149 	list_del_rcu(&rd->list);
1150 	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1151 	call_rcu(&rd->rcu, vxlan_dst_free);
1152 }
1153 
1154 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1155 			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1156 			   __be32 *vni, u32 *ifindex, u32 *nhid,
1157 			   struct netlink_ext_ack *extack)
1158 {
1159 	struct net *net = dev_net(vxlan->dev);
1160 	int err;
1161 
1162 	if (tb[NDA_NH_ID] &&
1163 	    (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1164 		NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1165 		return -EINVAL;
1166 	}
1167 
1168 	if (tb[NDA_DST]) {
1169 		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1170 		if (err) {
1171 			NL_SET_ERR_MSG(extack, "Unsupported address family");
1172 			return err;
1173 		}
1174 	} else {
1175 		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1176 
1177 		if (remote->sa.sa_family == AF_INET) {
1178 			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1179 			ip->sa.sa_family = AF_INET;
1180 #if IS_ENABLED(CONFIG_IPV6)
1181 		} else {
1182 			ip->sin6.sin6_addr = in6addr_any;
1183 			ip->sa.sa_family = AF_INET6;
1184 #endif
1185 		}
1186 	}
1187 
1188 	if (tb[NDA_PORT]) {
1189 		if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1190 			NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1191 			return -EINVAL;
1192 		}
1193 		*port = nla_get_be16(tb[NDA_PORT]);
1194 	} else {
1195 		*port = vxlan->cfg.dst_port;
1196 	}
1197 
1198 	if (tb[NDA_VNI]) {
1199 		if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1200 			NL_SET_ERR_MSG(extack, "Invalid vni");
1201 			return -EINVAL;
1202 		}
1203 		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1204 	} else {
1205 		*vni = vxlan->default_dst.remote_vni;
1206 	}
1207 
1208 	if (tb[NDA_SRC_VNI]) {
1209 		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1210 			NL_SET_ERR_MSG(extack, "Invalid src vni");
1211 			return -EINVAL;
1212 		}
1213 		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1214 	} else {
1215 		*src_vni = vxlan->default_dst.remote_vni;
1216 	}
1217 
1218 	if (tb[NDA_IFINDEX]) {
1219 		struct net_device *tdev;
1220 
1221 		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1222 			NL_SET_ERR_MSG(extack, "Invalid ifindex");
1223 			return -EINVAL;
1224 		}
1225 		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1226 		tdev = __dev_get_by_index(net, *ifindex);
1227 		if (!tdev) {
1228 			NL_SET_ERR_MSG(extack, "Device not found");
1229 			return -EADDRNOTAVAIL;
1230 		}
1231 	} else {
1232 		*ifindex = 0;
1233 	}
1234 
1235 	if (tb[NDA_NH_ID])
1236 		*nhid = nla_get_u32(tb[NDA_NH_ID]);
1237 	else
1238 		*nhid = 0;
1239 
1240 	return 0;
1241 }
1242 
1243 /* Add static entry (via netlink) */
1244 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1245 			 struct net_device *dev,
1246 			 const unsigned char *addr, u16 vid, u16 flags,
1247 			 struct netlink_ext_ack *extack)
1248 {
1249 	struct vxlan_dev *vxlan = netdev_priv(dev);
1250 	/* struct net *net = dev_net(vxlan->dev); */
1251 	union vxlan_addr ip;
1252 	__be16 port;
1253 	__be32 src_vni, vni;
1254 	u32 ifindex, nhid;
1255 	u32 hash_index;
1256 	int err;
1257 
1258 	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1259 		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1260 			ndm->ndm_state);
1261 		return -EINVAL;
1262 	}
1263 
1264 	if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1265 		return -EINVAL;
1266 
1267 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1268 			      &nhid, extack);
1269 	if (err)
1270 		return err;
1271 
1272 	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1273 		return -EAFNOSUPPORT;
1274 
1275 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1276 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1277 	err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1278 			       port, src_vni, vni, ifindex,
1279 			       ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1280 			       nhid, true, extack);
1281 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1282 
1283 	return err;
1284 }
1285 
1286 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1287 		       const unsigned char *addr, union vxlan_addr ip,
1288 		       __be16 port, __be32 src_vni, __be32 vni,
1289 		       u32 ifindex, bool swdev_notify)
1290 {
1291 	struct vxlan_rdst *rd = NULL;
1292 	struct vxlan_fdb *f;
1293 	int err = -ENOENT;
1294 
1295 	f = vxlan_find_mac(vxlan, addr, src_vni);
1296 	if (!f)
1297 		return err;
1298 
1299 	if (!vxlan_addr_any(&ip)) {
1300 		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1301 		if (!rd)
1302 			goto out;
1303 	}
1304 
1305 	/* remove a destination if it's not the only one on the list,
1306 	 * otherwise destroy the fdb entry
1307 	 */
1308 	if (rd && !list_is_singular(&f->remotes)) {
1309 		vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1310 		goto out;
1311 	}
1312 
1313 	vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1314 
1315 out:
1316 	return 0;
1317 }
1318 
1319 /* Delete entry (via netlink) */
1320 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1321 			    struct net_device *dev,
1322 			    const unsigned char *addr, u16 vid,
1323 			    struct netlink_ext_ack *extack)
1324 {
1325 	struct vxlan_dev *vxlan = netdev_priv(dev);
1326 	union vxlan_addr ip;
1327 	__be32 src_vni, vni;
1328 	u32 ifindex, nhid;
1329 	u32 hash_index;
1330 	__be16 port;
1331 	int err;
1332 
1333 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1334 			      &nhid, extack);
1335 	if (err)
1336 		return err;
1337 
1338 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1339 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1340 	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1341 				 true);
1342 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1343 
1344 	return err;
1345 }
1346 
1347 /* Dump forwarding table */
1348 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1349 			  struct net_device *dev,
1350 			  struct net_device *filter_dev, int *idx)
1351 {
1352 	struct vxlan_dev *vxlan = netdev_priv(dev);
1353 	unsigned int h;
1354 	int err = 0;
1355 
1356 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
1357 		struct vxlan_fdb *f;
1358 
1359 		rcu_read_lock();
1360 		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1361 			struct vxlan_rdst *rd;
1362 
1363 			if (rcu_access_pointer(f->nh)) {
1364 				if (*idx < cb->args[2])
1365 					goto skip_nh;
1366 				err = vxlan_fdb_info(skb, vxlan, f,
1367 						     NETLINK_CB(cb->skb).portid,
1368 						     cb->nlh->nlmsg_seq,
1369 						     RTM_NEWNEIGH,
1370 						     NLM_F_MULTI, NULL);
1371 				if (err < 0) {
1372 					rcu_read_unlock();
1373 					goto out;
1374 				}
1375 skip_nh:
1376 				*idx += 1;
1377 				continue;
1378 			}
1379 
1380 			list_for_each_entry_rcu(rd, &f->remotes, list) {
1381 				if (*idx < cb->args[2])
1382 					goto skip;
1383 
1384 				err = vxlan_fdb_info(skb, vxlan, f,
1385 						     NETLINK_CB(cb->skb).portid,
1386 						     cb->nlh->nlmsg_seq,
1387 						     RTM_NEWNEIGH,
1388 						     NLM_F_MULTI, rd);
1389 				if (err < 0) {
1390 					rcu_read_unlock();
1391 					goto out;
1392 				}
1393 skip:
1394 				*idx += 1;
1395 			}
1396 		}
1397 		rcu_read_unlock();
1398 	}
1399 out:
1400 	return err;
1401 }
1402 
1403 static int vxlan_fdb_get(struct sk_buff *skb,
1404 			 struct nlattr *tb[],
1405 			 struct net_device *dev,
1406 			 const unsigned char *addr,
1407 			 u16 vid, u32 portid, u32 seq,
1408 			 struct netlink_ext_ack *extack)
1409 {
1410 	struct vxlan_dev *vxlan = netdev_priv(dev);
1411 	struct vxlan_fdb *f;
1412 	__be32 vni;
1413 	int err;
1414 
1415 	if (tb[NDA_VNI])
1416 		vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1417 	else
1418 		vni = vxlan->default_dst.remote_vni;
1419 
1420 	rcu_read_lock();
1421 
1422 	f = __vxlan_find_mac(vxlan, addr, vni);
1423 	if (!f) {
1424 		NL_SET_ERR_MSG(extack, "Fdb entry not found");
1425 		err = -ENOENT;
1426 		goto errout;
1427 	}
1428 
1429 	err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1430 			     RTM_NEWNEIGH, 0, first_remote_rcu(f));
1431 errout:
1432 	rcu_read_unlock();
1433 	return err;
1434 }
1435 
1436 /* Watch incoming packets to learn mapping between Ethernet address
1437  * and Tunnel endpoint.
1438  * Return true if packet is bogus and should be dropped.
1439  */
1440 static enum skb_drop_reason vxlan_snoop(struct net_device *dev,
1441 					union vxlan_addr *src_ip,
1442 					const u8 *src_mac, u32 src_ifindex,
1443 					__be32 vni)
1444 {
1445 	struct vxlan_dev *vxlan = netdev_priv(dev);
1446 	struct vxlan_fdb *f;
1447 	u32 ifindex = 0;
1448 
1449 	/* Ignore packets from invalid src-address */
1450 	if (!is_valid_ether_addr(src_mac))
1451 		return SKB_DROP_REASON_MAC_INVALID_SOURCE;
1452 
1453 #if IS_ENABLED(CONFIG_IPV6)
1454 	if (src_ip->sa.sa_family == AF_INET6 &&
1455 	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1456 		ifindex = src_ifindex;
1457 #endif
1458 
1459 	f = vxlan_find_mac(vxlan, src_mac, vni);
1460 	if (likely(f)) {
1461 		struct vxlan_rdst *rdst = first_remote_rcu(f);
1462 
1463 		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1464 			   rdst->remote_ifindex == ifindex))
1465 			return SKB_NOT_DROPPED_YET;
1466 
1467 		/* Don't migrate static entries, drop packets */
1468 		if (f->state & (NUD_PERMANENT | NUD_NOARP))
1469 			return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1470 
1471 		/* Don't override an fdb with nexthop with a learnt entry */
1472 		if (rcu_access_pointer(f->nh))
1473 			return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1474 
1475 		if (net_ratelimit())
1476 			netdev_info(dev,
1477 				    "%pM migrated from %pIS to %pIS\n",
1478 				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1479 
1480 		rdst->remote_ip = *src_ip;
1481 		f->updated = jiffies;
1482 		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1483 	} else {
1484 		u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1485 
1486 		/* learned new entry */
1487 		spin_lock(&vxlan->hash_lock[hash_index]);
1488 
1489 		/* close off race between vxlan_flush and incoming packets */
1490 		if (netif_running(dev))
1491 			vxlan_fdb_update(vxlan, src_mac, src_ip,
1492 					 NUD_REACHABLE,
1493 					 NLM_F_EXCL|NLM_F_CREATE,
1494 					 vxlan->cfg.dst_port,
1495 					 vni,
1496 					 vxlan->default_dst.remote_vni,
1497 					 ifindex, NTF_SELF, 0, true, NULL);
1498 		spin_unlock(&vxlan->hash_lock[hash_index]);
1499 	}
1500 
1501 	return SKB_NOT_DROPPED_YET;
1502 }
1503 
1504 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1505 {
1506 	struct vxlan_net *vn;
1507 
1508 	if (!vs)
1509 		return false;
1510 	if (!refcount_dec_and_test(&vs->refcnt))
1511 		return false;
1512 
1513 	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1514 	spin_lock(&vn->sock_lock);
1515 	hlist_del_rcu(&vs->hlist);
1516 	udp_tunnel_notify_del_rx_port(vs->sock,
1517 				      (vs->flags & VXLAN_F_GPE) ?
1518 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1519 				      UDP_TUNNEL_TYPE_VXLAN);
1520 	spin_unlock(&vn->sock_lock);
1521 
1522 	return true;
1523 }
1524 
1525 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1526 {
1527 	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1528 #if IS_ENABLED(CONFIG_IPV6)
1529 	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1530 
1531 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1532 #endif
1533 
1534 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1535 	synchronize_net();
1536 
1537 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1538 		vxlan_vs_del_vnigrp(vxlan);
1539 	else
1540 		vxlan_vs_del_dev(vxlan);
1541 
1542 	if (__vxlan_sock_release_prep(sock4)) {
1543 		udp_tunnel_sock_release(sock4->sock);
1544 		kfree(sock4);
1545 	}
1546 
1547 #if IS_ENABLED(CONFIG_IPV6)
1548 	if (__vxlan_sock_release_prep(sock6)) {
1549 		udp_tunnel_sock_release(sock6->sock);
1550 		kfree(sock6);
1551 	}
1552 #endif
1553 }
1554 
1555 static enum skb_drop_reason vxlan_remcsum(struct vxlanhdr *unparsed,
1556 					  struct sk_buff *skb,
1557 					  u32 vxflags)
1558 {
1559 	enum skb_drop_reason reason;
1560 	size_t start, offset;
1561 
1562 	if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1563 		goto out;
1564 
1565 	start = vxlan_rco_start(unparsed->vx_vni);
1566 	offset = start + vxlan_rco_offset(unparsed->vx_vni);
1567 
1568 	reason = pskb_may_pull_reason(skb, offset + sizeof(u16));
1569 	if (reason)
1570 		return reason;
1571 
1572 	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1573 			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1574 out:
1575 	unparsed->vx_flags &= ~VXLAN_HF_RCO;
1576 	unparsed->vx_vni &= VXLAN_VNI_MASK;
1577 
1578 	return SKB_NOT_DROPPED_YET;
1579 }
1580 
1581 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1582 				struct sk_buff *skb, u32 vxflags,
1583 				struct vxlan_metadata *md)
1584 {
1585 	struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1586 	struct metadata_dst *tun_dst;
1587 
1588 	if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1589 		goto out;
1590 
1591 	md->gbp = ntohs(gbp->policy_id);
1592 
1593 	tun_dst = (struct metadata_dst *)skb_dst(skb);
1594 	if (tun_dst) {
1595 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT,
1596 			  tun_dst->u.tun_info.key.tun_flags);
1597 		tun_dst->u.tun_info.options_len = sizeof(*md);
1598 	}
1599 	if (gbp->dont_learn)
1600 		md->gbp |= VXLAN_GBP_DONT_LEARN;
1601 
1602 	if (gbp->policy_applied)
1603 		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1604 
1605 	/* In flow-based mode, GBP is carried in dst_metadata */
1606 	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1607 		skb->mark = md->gbp;
1608 out:
1609 	unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1610 }
1611 
1612 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan,
1613 					  struct vxlan_sock *vs,
1614 					  struct sk_buff *skb, __be32 vni)
1615 {
1616 	union vxlan_addr saddr;
1617 	u32 ifindex = skb->dev->ifindex;
1618 
1619 	skb_reset_mac_header(skb);
1620 	skb->protocol = eth_type_trans(skb, vxlan->dev);
1621 	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1622 
1623 	/* Ignore packet loops (and multicast echo) */
1624 	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1625 		return SKB_DROP_REASON_LOCAL_MAC;
1626 
1627 	/* Get address from the outer IP header */
1628 	if (vxlan_get_sk_family(vs) == AF_INET) {
1629 		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1630 		saddr.sa.sa_family = AF_INET;
1631 #if IS_ENABLED(CONFIG_IPV6)
1632 	} else {
1633 		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1634 		saddr.sa.sa_family = AF_INET6;
1635 #endif
1636 	}
1637 
1638 	if (!(vxlan->cfg.flags & VXLAN_F_LEARN))
1639 		return SKB_NOT_DROPPED_YET;
1640 
1641 	return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source,
1642 			   ifindex, vni);
1643 }
1644 
1645 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1646 				  struct sk_buff *skb)
1647 {
1648 	int err = 0;
1649 
1650 	if (vxlan_get_sk_family(vs) == AF_INET)
1651 		err = IP_ECN_decapsulate(oiph, skb);
1652 #if IS_ENABLED(CONFIG_IPV6)
1653 	else
1654 		err = IP6_ECN_decapsulate(oiph, skb);
1655 #endif
1656 
1657 	if (unlikely(err) && log_ecn_error) {
1658 		if (vxlan_get_sk_family(vs) == AF_INET)
1659 			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1660 					     &((struct iphdr *)oiph)->saddr,
1661 					     ((struct iphdr *)oiph)->tos);
1662 		else
1663 			net_info_ratelimited("non-ECT from %pI6\n",
1664 					     &((struct ipv6hdr *)oiph)->saddr);
1665 	}
1666 	return err <= 1;
1667 }
1668 
1669 /* Callback from net/ipv4/udp.c to receive packets */
1670 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1671 {
1672 	struct vxlan_vni_node *vninode = NULL;
1673 	struct vxlan_dev *vxlan;
1674 	struct vxlan_sock *vs;
1675 	struct vxlanhdr unparsed;
1676 	struct vxlan_metadata _md;
1677 	struct vxlan_metadata *md = &_md;
1678 	__be16 protocol = htons(ETH_P_TEB);
1679 	enum skb_drop_reason reason;
1680 	bool raw_proto = false;
1681 	void *oiph;
1682 	__be32 vni = 0;
1683 	int nh;
1684 
1685 	/* Need UDP and VXLAN header to be present */
1686 	reason = pskb_may_pull_reason(skb, VXLAN_HLEN);
1687 	if (reason)
1688 		goto drop;
1689 
1690 	unparsed = *vxlan_hdr(skb);
1691 	/* VNI flag always required to be set */
1692 	if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1693 		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1694 			   ntohl(vxlan_hdr(skb)->vx_flags),
1695 			   ntohl(vxlan_hdr(skb)->vx_vni));
1696 		reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1697 		/* Return non vxlan pkt */
1698 		goto drop;
1699 	}
1700 	unparsed.vx_flags &= ~VXLAN_HF_VNI;
1701 	unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1702 
1703 	vs = rcu_dereference_sk_user_data(sk);
1704 	if (!vs)
1705 		goto drop;
1706 
1707 	vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1708 
1709 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1710 	if (!vxlan) {
1711 		reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND;
1712 		goto drop;
1713 	}
1714 
1715 	/* For backwards compatibility, only allow reserved fields to be
1716 	 * used by VXLAN extensions if explicitly requested.
1717 	 */
1718 	if (vs->flags & VXLAN_F_GPE) {
1719 		if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1720 			goto drop;
1721 		unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1722 		raw_proto = true;
1723 	}
1724 
1725 	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1726 				   !net_eq(vxlan->net, dev_net(vxlan->dev)))) {
1727 		reason = SKB_DROP_REASON_NOMEM;
1728 		goto drop;
1729 	}
1730 
1731 	if (vs->flags & VXLAN_F_REMCSUM_RX) {
1732 		reason = vxlan_remcsum(&unparsed, skb, vs->flags);
1733 		if (unlikely(reason))
1734 			goto drop;
1735 	}
1736 
1737 	if (vxlan_collect_metadata(vs)) {
1738 		IP_TUNNEL_DECLARE_FLAGS(flags) = { };
1739 		struct metadata_dst *tun_dst;
1740 
1741 		__set_bit(IP_TUNNEL_KEY_BIT, flags);
1742 		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags,
1743 					 key32_to_tunnel_id(vni), sizeof(*md));
1744 
1745 		if (!tun_dst) {
1746 			reason = SKB_DROP_REASON_NOMEM;
1747 			goto drop;
1748 		}
1749 
1750 		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1751 
1752 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1753 	} else {
1754 		memset(md, 0, sizeof(*md));
1755 	}
1756 
1757 	if (vs->flags & VXLAN_F_GBP)
1758 		vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1759 	/* Note that GBP and GPE can never be active together. This is
1760 	 * ensured in vxlan_dev_configure.
1761 	 */
1762 
1763 	if (unparsed.vx_flags || unparsed.vx_vni) {
1764 		/* If there are any unprocessed flags remaining treat
1765 		 * this as a malformed packet. This behavior diverges from
1766 		 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1767 		 * in reserved fields are to be ignored. The approach here
1768 		 * maintains compatibility with previous stack code, and also
1769 		 * is more robust and provides a little more security in
1770 		 * adding extensions to VXLAN.
1771 		 */
1772 		reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1773 		goto drop;
1774 	}
1775 
1776 	if (!raw_proto) {
1777 		reason = vxlan_set_mac(vxlan, vs, skb, vni);
1778 		if (reason)
1779 			goto drop;
1780 	} else {
1781 		skb_reset_mac_header(skb);
1782 		skb->dev = vxlan->dev;
1783 		skb->pkt_type = PACKET_HOST;
1784 	}
1785 
1786 	/* Save offset of outer header relative to skb->head,
1787 	 * because we are going to reset the network header to the inner header
1788 	 * and might change skb->head.
1789 	 */
1790 	nh = skb_network_header(skb) - skb->head;
1791 
1792 	skb_reset_network_header(skb);
1793 
1794 	reason = pskb_inet_may_pull_reason(skb);
1795 	if (reason) {
1796 		DEV_STATS_INC(vxlan->dev, rx_length_errors);
1797 		DEV_STATS_INC(vxlan->dev, rx_errors);
1798 		vxlan_vnifilter_count(vxlan, vni, vninode,
1799 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1800 		goto drop;
1801 	}
1802 
1803 	/* Get the outer header. */
1804 	oiph = skb->head + nh;
1805 
1806 	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1807 		reason = SKB_DROP_REASON_IP_TUNNEL_ECN;
1808 		DEV_STATS_INC(vxlan->dev, rx_frame_errors);
1809 		DEV_STATS_INC(vxlan->dev, rx_errors);
1810 		vxlan_vnifilter_count(vxlan, vni, vninode,
1811 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1812 		goto drop;
1813 	}
1814 
1815 	rcu_read_lock();
1816 
1817 	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1818 		rcu_read_unlock();
1819 		dev_core_stats_rx_dropped_inc(vxlan->dev);
1820 		vxlan_vnifilter_count(vxlan, vni, vninode,
1821 				      VXLAN_VNI_STATS_RX_DROPS, 0);
1822 		reason = SKB_DROP_REASON_DEV_READY;
1823 		goto drop;
1824 	}
1825 
1826 	dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1827 	vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1828 	gro_cells_receive(&vxlan->gro_cells, skb);
1829 
1830 	rcu_read_unlock();
1831 
1832 	return 0;
1833 
1834 drop:
1835 	reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED;
1836 	/* Consume bad packet */
1837 	kfree_skb_reason(skb, reason);
1838 	return 0;
1839 }
1840 
1841 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1842 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1843 {
1844 	struct vxlan_dev *vxlan;
1845 	struct vxlan_sock *vs;
1846 	struct vxlanhdr *hdr;
1847 	__be32 vni;
1848 
1849 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1850 		return -EINVAL;
1851 
1852 	hdr = vxlan_hdr(skb);
1853 
1854 	if (!(hdr->vx_flags & VXLAN_HF_VNI))
1855 		return -EINVAL;
1856 
1857 	vs = rcu_dereference_sk_user_data(sk);
1858 	if (!vs)
1859 		return -ENOENT;
1860 
1861 	vni = vxlan_vni(hdr->vx_vni);
1862 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1863 	if (!vxlan)
1864 		return -ENOENT;
1865 
1866 	return 0;
1867 }
1868 
1869 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1870 {
1871 	struct vxlan_dev *vxlan = netdev_priv(dev);
1872 	struct arphdr *parp;
1873 	u8 *arpptr, *sha;
1874 	__be32 sip, tip;
1875 	struct neighbour *n;
1876 
1877 	if (dev->flags & IFF_NOARP)
1878 		goto out;
1879 
1880 	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1881 		dev_core_stats_tx_dropped_inc(dev);
1882 		vxlan_vnifilter_count(vxlan, vni, NULL,
1883 				      VXLAN_VNI_STATS_TX_DROPS, 0);
1884 		goto out;
1885 	}
1886 	parp = arp_hdr(skb);
1887 
1888 	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1889 	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1890 	    parp->ar_pro != htons(ETH_P_IP) ||
1891 	    parp->ar_op != htons(ARPOP_REQUEST) ||
1892 	    parp->ar_hln != dev->addr_len ||
1893 	    parp->ar_pln != 4)
1894 		goto out;
1895 	arpptr = (u8 *)parp + sizeof(struct arphdr);
1896 	sha = arpptr;
1897 	arpptr += dev->addr_len;	/* sha */
1898 	memcpy(&sip, arpptr, sizeof(sip));
1899 	arpptr += sizeof(sip);
1900 	arpptr += dev->addr_len;	/* tha */
1901 	memcpy(&tip, arpptr, sizeof(tip));
1902 
1903 	if (ipv4_is_loopback(tip) ||
1904 	    ipv4_is_multicast(tip))
1905 		goto out;
1906 
1907 	n = neigh_lookup(&arp_tbl, &tip, dev);
1908 
1909 	if (n) {
1910 		struct vxlan_fdb *f;
1911 		struct sk_buff	*reply;
1912 
1913 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1914 			neigh_release(n);
1915 			goto out;
1916 		}
1917 
1918 		f = vxlan_find_mac(vxlan, n->ha, vni);
1919 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1920 			/* bridge-local neighbor */
1921 			neigh_release(n);
1922 			goto out;
1923 		}
1924 
1925 		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1926 				n->ha, sha);
1927 
1928 		neigh_release(n);
1929 
1930 		if (reply == NULL)
1931 			goto out;
1932 
1933 		skb_reset_mac_header(reply);
1934 		__skb_pull(reply, skb_network_offset(reply));
1935 		reply->ip_summed = CHECKSUM_UNNECESSARY;
1936 		reply->pkt_type = PACKET_HOST;
1937 
1938 		if (netif_rx(reply) == NET_RX_DROP) {
1939 			dev_core_stats_rx_dropped_inc(dev);
1940 			vxlan_vnifilter_count(vxlan, vni, NULL,
1941 					      VXLAN_VNI_STATS_RX_DROPS, 0);
1942 		}
1943 
1944 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1945 		union vxlan_addr ipa = {
1946 			.sin.sin_addr.s_addr = tip,
1947 			.sin.sin_family = AF_INET,
1948 		};
1949 
1950 		vxlan_ip_miss(dev, &ipa);
1951 	}
1952 out:
1953 	consume_skb(skb);
1954 	return NETDEV_TX_OK;
1955 }
1956 
1957 #if IS_ENABLED(CONFIG_IPV6)
1958 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1959 	struct neighbour *n, bool isrouter)
1960 {
1961 	struct net_device *dev = request->dev;
1962 	struct sk_buff *reply;
1963 	struct nd_msg *ns, *na;
1964 	struct ipv6hdr *pip6;
1965 	u8 *daddr;
1966 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1967 	int ns_olen;
1968 	int i, len;
1969 
1970 	if (dev == NULL || !pskb_may_pull(request, request->len))
1971 		return NULL;
1972 
1973 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1974 		sizeof(*na) + na_olen + dev->needed_tailroom;
1975 	reply = alloc_skb(len, GFP_ATOMIC);
1976 	if (reply == NULL)
1977 		return NULL;
1978 
1979 	reply->protocol = htons(ETH_P_IPV6);
1980 	reply->dev = dev;
1981 	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1982 	skb_push(reply, sizeof(struct ethhdr));
1983 	skb_reset_mac_header(reply);
1984 
1985 	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1986 
1987 	daddr = eth_hdr(request)->h_source;
1988 	ns_olen = request->len - skb_network_offset(request) -
1989 		sizeof(struct ipv6hdr) - sizeof(*ns);
1990 	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1991 		if (!ns->opt[i + 1]) {
1992 			kfree_skb(reply);
1993 			return NULL;
1994 		}
1995 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1996 			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1997 			break;
1998 		}
1999 	}
2000 
2001 	/* Ethernet header */
2002 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2003 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2004 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2005 	reply->protocol = htons(ETH_P_IPV6);
2006 
2007 	skb_pull(reply, sizeof(struct ethhdr));
2008 	skb_reset_network_header(reply);
2009 	skb_put(reply, sizeof(struct ipv6hdr));
2010 
2011 	/* IPv6 header */
2012 
2013 	pip6 = ipv6_hdr(reply);
2014 	memset(pip6, 0, sizeof(struct ipv6hdr));
2015 	pip6->version = 6;
2016 	pip6->priority = ipv6_hdr(request)->priority;
2017 	pip6->nexthdr = IPPROTO_ICMPV6;
2018 	pip6->hop_limit = 255;
2019 	pip6->daddr = ipv6_hdr(request)->saddr;
2020 	pip6->saddr = *(struct in6_addr *)n->primary_key;
2021 
2022 	skb_pull(reply, sizeof(struct ipv6hdr));
2023 	skb_reset_transport_header(reply);
2024 
2025 	/* Neighbor Advertisement */
2026 	na = skb_put_zero(reply, sizeof(*na) + na_olen);
2027 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2028 	na->icmph.icmp6_router = isrouter;
2029 	na->icmph.icmp6_override = 1;
2030 	na->icmph.icmp6_solicited = 1;
2031 	na->target = ns->target;
2032 	ether_addr_copy(&na->opt[2], n->ha);
2033 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2034 	na->opt[1] = na_olen >> 3;
2035 
2036 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2037 		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2038 		csum_partial(na, sizeof(*na)+na_olen, 0));
2039 
2040 	pip6->payload_len = htons(sizeof(*na)+na_olen);
2041 
2042 	skb_push(reply, sizeof(struct ipv6hdr));
2043 
2044 	reply->ip_summed = CHECKSUM_UNNECESSARY;
2045 
2046 	return reply;
2047 }
2048 
2049 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2050 {
2051 	struct vxlan_dev *vxlan = netdev_priv(dev);
2052 	const struct in6_addr *daddr;
2053 	const struct ipv6hdr *iphdr;
2054 	struct inet6_dev *in6_dev;
2055 	struct neighbour *n;
2056 	struct nd_msg *msg;
2057 
2058 	rcu_read_lock();
2059 	in6_dev = __in6_dev_get(dev);
2060 	if (!in6_dev)
2061 		goto out;
2062 
2063 	iphdr = ipv6_hdr(skb);
2064 	daddr = &iphdr->daddr;
2065 	msg = (struct nd_msg *)(iphdr + 1);
2066 
2067 	if (ipv6_addr_loopback(daddr) ||
2068 	    ipv6_addr_is_multicast(&msg->target))
2069 		goto out;
2070 
2071 	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2072 
2073 	if (n) {
2074 		struct vxlan_fdb *f;
2075 		struct sk_buff *reply;
2076 
2077 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2078 			neigh_release(n);
2079 			goto out;
2080 		}
2081 
2082 		f = vxlan_find_mac(vxlan, n->ha, vni);
2083 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2084 			/* bridge-local neighbor */
2085 			neigh_release(n);
2086 			goto out;
2087 		}
2088 
2089 		reply = vxlan_na_create(skb, n,
2090 					!!(f ? f->flags & NTF_ROUTER : 0));
2091 
2092 		neigh_release(n);
2093 
2094 		if (reply == NULL)
2095 			goto out;
2096 
2097 		if (netif_rx(reply) == NET_RX_DROP) {
2098 			dev_core_stats_rx_dropped_inc(dev);
2099 			vxlan_vnifilter_count(vxlan, vni, NULL,
2100 					      VXLAN_VNI_STATS_RX_DROPS, 0);
2101 		}
2102 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2103 		union vxlan_addr ipa = {
2104 			.sin6.sin6_addr = msg->target,
2105 			.sin6.sin6_family = AF_INET6,
2106 		};
2107 
2108 		vxlan_ip_miss(dev, &ipa);
2109 	}
2110 
2111 out:
2112 	rcu_read_unlock();
2113 	consume_skb(skb);
2114 	return NETDEV_TX_OK;
2115 }
2116 #endif
2117 
2118 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2119 {
2120 	struct vxlan_dev *vxlan = netdev_priv(dev);
2121 	struct neighbour *n;
2122 
2123 	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2124 		return false;
2125 
2126 	n = NULL;
2127 	switch (ntohs(eth_hdr(skb)->h_proto)) {
2128 	case ETH_P_IP:
2129 	{
2130 		struct iphdr *pip;
2131 
2132 		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2133 			return false;
2134 		pip = ip_hdr(skb);
2135 		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2136 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2137 			union vxlan_addr ipa = {
2138 				.sin.sin_addr.s_addr = pip->daddr,
2139 				.sin.sin_family = AF_INET,
2140 			};
2141 
2142 			vxlan_ip_miss(dev, &ipa);
2143 			return false;
2144 		}
2145 
2146 		break;
2147 	}
2148 #if IS_ENABLED(CONFIG_IPV6)
2149 	case ETH_P_IPV6:
2150 	{
2151 		struct ipv6hdr *pip6;
2152 
2153 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2154 			return false;
2155 		pip6 = ipv6_hdr(skb);
2156 		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2157 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2158 			union vxlan_addr ipa = {
2159 				.sin6.sin6_addr = pip6->daddr,
2160 				.sin6.sin6_family = AF_INET6,
2161 			};
2162 
2163 			vxlan_ip_miss(dev, &ipa);
2164 			return false;
2165 		}
2166 
2167 		break;
2168 	}
2169 #endif
2170 	default:
2171 		return false;
2172 	}
2173 
2174 	if (n) {
2175 		bool diff;
2176 
2177 		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2178 		if (diff) {
2179 			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2180 				dev->addr_len);
2181 			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2182 		}
2183 		neigh_release(n);
2184 		return diff;
2185 	}
2186 
2187 	return false;
2188 }
2189 
2190 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2191 {
2192 	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2193 
2194 	gpe->np_applied = 1;
2195 	gpe->next_protocol = tun_p_from_eth_p(protocol);
2196 	if (!gpe->next_protocol)
2197 		return -EPFNOSUPPORT;
2198 	return 0;
2199 }
2200 
2201 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2202 			   int iphdr_len, __be32 vni,
2203 			   struct vxlan_metadata *md, u32 vxflags,
2204 			   bool udp_sum)
2205 {
2206 	struct vxlanhdr *vxh;
2207 	int min_headroom;
2208 	int err;
2209 	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2210 	__be16 inner_protocol = htons(ETH_P_TEB);
2211 
2212 	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2213 	    skb->ip_summed == CHECKSUM_PARTIAL) {
2214 		int csum_start = skb_checksum_start_offset(skb);
2215 
2216 		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2217 		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2218 		    (skb->csum_offset == offsetof(struct udphdr, check) ||
2219 		     skb->csum_offset == offsetof(struct tcphdr, check)))
2220 			type |= SKB_GSO_TUNNEL_REMCSUM;
2221 	}
2222 
2223 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2224 			+ VXLAN_HLEN + iphdr_len;
2225 
2226 	/* Need space for new headers (invalidates iph ptr) */
2227 	err = skb_cow_head(skb, min_headroom);
2228 	if (unlikely(err))
2229 		return err;
2230 
2231 	err = iptunnel_handle_offloads(skb, type);
2232 	if (err)
2233 		return err;
2234 
2235 	vxh = __skb_push(skb, sizeof(*vxh));
2236 	vxh->vx_flags = VXLAN_HF_VNI;
2237 	vxh->vx_vni = vxlan_vni_field(vni);
2238 
2239 	if (type & SKB_GSO_TUNNEL_REMCSUM) {
2240 		unsigned int start;
2241 
2242 		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2243 		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2244 		vxh->vx_flags |= VXLAN_HF_RCO;
2245 
2246 		if (!skb_is_gso(skb)) {
2247 			skb->ip_summed = CHECKSUM_NONE;
2248 			skb->encapsulation = 0;
2249 		}
2250 	}
2251 
2252 	if (vxflags & VXLAN_F_GBP)
2253 		vxlan_build_gbp_hdr(vxh, md);
2254 	if (vxflags & VXLAN_F_GPE) {
2255 		err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2256 		if (err < 0)
2257 			return err;
2258 		inner_protocol = skb->protocol;
2259 	}
2260 
2261 	skb_set_inner_protocol(skb, inner_protocol);
2262 	return 0;
2263 }
2264 
2265 /* Bypass encapsulation if the destination is local */
2266 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2267 			       struct vxlan_dev *dst_vxlan, __be32 vni,
2268 			       bool snoop)
2269 {
2270 	union vxlan_addr loopback;
2271 	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2272 	struct net_device *dev;
2273 	int len = skb->len;
2274 
2275 	skb->pkt_type = PACKET_HOST;
2276 	skb->encapsulation = 0;
2277 	skb->dev = dst_vxlan->dev;
2278 	__skb_pull(skb, skb_network_offset(skb));
2279 
2280 	if (remote_ip->sa.sa_family == AF_INET) {
2281 		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2282 		loopback.sa.sa_family =  AF_INET;
2283 #if IS_ENABLED(CONFIG_IPV6)
2284 	} else {
2285 		loopback.sin6.sin6_addr = in6addr_loopback;
2286 		loopback.sa.sa_family =  AF_INET6;
2287 #endif
2288 	}
2289 
2290 	rcu_read_lock();
2291 	dev = skb->dev;
2292 	if (unlikely(!(dev->flags & IFF_UP))) {
2293 		kfree_skb(skb);
2294 		goto drop;
2295 	}
2296 
2297 	if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2298 		vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2299 
2300 	dev_sw_netstats_tx_add(src_vxlan->dev, 1, len);
2301 	vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2302 
2303 	if (__netif_rx(skb) == NET_RX_SUCCESS) {
2304 		dev_sw_netstats_rx_add(dst_vxlan->dev, len);
2305 		vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2306 				      len);
2307 	} else {
2308 drop:
2309 		dev_core_stats_rx_dropped_inc(dev);
2310 		vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2311 				      VXLAN_VNI_STATS_RX_DROPS, 0);
2312 	}
2313 	rcu_read_unlock();
2314 }
2315 
2316 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2317 				 struct vxlan_dev *vxlan,
2318 				 int addr_family,
2319 				 __be16 dst_port, int dst_ifindex, __be32 vni,
2320 				 struct dst_entry *dst,
2321 				 u32 rt_flags)
2322 {
2323 #if IS_ENABLED(CONFIG_IPV6)
2324 	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2325 	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2326 	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2327 	 */
2328 	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2329 #endif
2330 	/* Bypass encapsulation if the destination is local */
2331 	if (rt_flags & RTCF_LOCAL &&
2332 	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2333 	    vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2334 		struct vxlan_dev *dst_vxlan;
2335 
2336 		dst_release(dst);
2337 		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2338 					   addr_family, dst_port,
2339 					   vxlan->cfg.flags);
2340 		if (!dst_vxlan) {
2341 			DEV_STATS_INC(dev, tx_errors);
2342 			vxlan_vnifilter_count(vxlan, vni, NULL,
2343 					      VXLAN_VNI_STATS_TX_ERRORS, 0);
2344 			kfree_skb(skb);
2345 
2346 			return -ENOENT;
2347 		}
2348 		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2349 		return 1;
2350 	}
2351 
2352 	return 0;
2353 }
2354 
2355 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2356 		    __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2357 {
2358 	struct dst_cache *dst_cache;
2359 	struct ip_tunnel_info *info;
2360 	struct ip_tunnel_key *pkey;
2361 	struct ip_tunnel_key key;
2362 	struct vxlan_dev *vxlan = netdev_priv(dev);
2363 	const struct iphdr *old_iph;
2364 	struct vxlan_metadata _md;
2365 	struct vxlan_metadata *md = &_md;
2366 	unsigned int pkt_len = skb->len;
2367 	__be16 src_port = 0, dst_port;
2368 	struct dst_entry *ndst = NULL;
2369 	int addr_family;
2370 	__u8 tos, ttl;
2371 	int ifindex;
2372 	int err;
2373 	u32 flags = vxlan->cfg.flags;
2374 	bool use_cache;
2375 	bool udp_sum = false;
2376 	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2377 	bool no_eth_encap;
2378 	__be32 vni = 0;
2379 
2380 	no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB);
2381 	if (skb_vlan_inet_prepare(skb, no_eth_encap))
2382 		goto drop;
2383 
2384 	old_iph = ip_hdr(skb);
2385 
2386 	info = skb_tunnel_info(skb);
2387 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
2388 
2389 	if (rdst) {
2390 		memset(&key, 0, sizeof(key));
2391 		pkey = &key;
2392 
2393 		if (vxlan_addr_any(&rdst->remote_ip)) {
2394 			if (did_rsc) {
2395 				/* short-circuited back to local bridge */
2396 				vxlan_encap_bypass(skb, vxlan, vxlan,
2397 						   default_vni, true);
2398 				return;
2399 			}
2400 			goto drop;
2401 		}
2402 
2403 		addr_family = vxlan->cfg.saddr.sa.sa_family;
2404 		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2405 		vni = (rdst->remote_vni) ? : default_vni;
2406 		ifindex = rdst->remote_ifindex;
2407 
2408 		if (addr_family == AF_INET) {
2409 			key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr;
2410 			key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr;
2411 		} else {
2412 			key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr;
2413 			key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr;
2414 		}
2415 
2416 		dst_cache = &rdst->dst_cache;
2417 		md->gbp = skb->mark;
2418 		if (flags & VXLAN_F_TTL_INHERIT) {
2419 			ttl = ip_tunnel_get_ttl(old_iph, skb);
2420 		} else {
2421 			ttl = vxlan->cfg.ttl;
2422 			if (!ttl && vxlan_addr_multicast(&rdst->remote_ip))
2423 				ttl = 1;
2424 		}
2425 		tos = vxlan->cfg.tos;
2426 		if (tos == 1)
2427 			tos = ip_tunnel_get_dsfield(old_iph, skb);
2428 		if (tos && !info)
2429 			use_cache = false;
2430 
2431 		if (addr_family == AF_INET)
2432 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2433 		else
2434 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2435 #if IS_ENABLED(CONFIG_IPV6)
2436 		switch (vxlan->cfg.label_policy) {
2437 		case VXLAN_LABEL_FIXED:
2438 			key.label = vxlan->cfg.label;
2439 			break;
2440 		case VXLAN_LABEL_INHERIT:
2441 			key.label = ip_tunnel_get_flowlabel(old_iph, skb);
2442 			break;
2443 		default:
2444 			DEBUG_NET_WARN_ON_ONCE(1);
2445 			goto drop;
2446 		}
2447 #endif
2448 	} else {
2449 		if (!info) {
2450 			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2451 				  dev->name);
2452 			goto drop;
2453 		}
2454 		pkey = &info->key;
2455 		addr_family = ip_tunnel_info_af(info);
2456 		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2457 		vni = tunnel_id_to_key32(info->key.tun_id);
2458 		ifindex = 0;
2459 		dst_cache = &info->dst_cache;
2460 		if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
2461 			if (info->options_len < sizeof(*md))
2462 				goto drop;
2463 			md = ip_tunnel_info_opts(info);
2464 		}
2465 		ttl = info->key.ttl;
2466 		tos = info->key.tos;
2467 		udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
2468 	}
2469 	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2470 				     vxlan->cfg.port_max, true);
2471 
2472 	rcu_read_lock();
2473 	if (addr_family == AF_INET) {
2474 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2475 		struct rtable *rt;
2476 		__be16 df = 0;
2477 		__be32 saddr;
2478 
2479 		if (!ifindex)
2480 			ifindex = sock4->sock->sk->sk_bound_dev_if;
2481 
2482 		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex,
2483 					   &saddr, pkey, src_port, dst_port,
2484 					   tos, use_cache ? dst_cache : NULL);
2485 		if (IS_ERR(rt)) {
2486 			err = PTR_ERR(rt);
2487 			goto tx_error;
2488 		}
2489 
2490 		if (!info) {
2491 			/* Bypass encapsulation if the destination is local */
2492 			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET,
2493 						    dst_port, ifindex, vni,
2494 						    &rt->dst, rt->rt_flags);
2495 			if (err)
2496 				goto out_unlock;
2497 
2498 			if (vxlan->cfg.df == VXLAN_DF_SET) {
2499 				df = htons(IP_DF);
2500 			} else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2501 				struct ethhdr *eth = eth_hdr(skb);
2502 
2503 				if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2504 				    (ntohs(eth->h_proto) == ETH_P_IP &&
2505 				     old_iph->frag_off & htons(IP_DF)))
2506 					df = htons(IP_DF);
2507 			}
2508 		} else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT,
2509 				    info->key.tun_flags)) {
2510 			df = htons(IP_DF);
2511 		}
2512 
2513 		ndst = &rt->dst;
2514 		err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2515 					    netif_is_any_bridge_port(dev));
2516 		if (err < 0) {
2517 			goto tx_error;
2518 		} else if (err) {
2519 			if (info) {
2520 				struct ip_tunnel_info *unclone;
2521 
2522 				unclone = skb_tunnel_info_unclone(skb);
2523 				if (unlikely(!unclone))
2524 					goto tx_error;
2525 
2526 				unclone->key.u.ipv4.src = pkey->u.ipv4.dst;
2527 				unclone->key.u.ipv4.dst = saddr;
2528 			}
2529 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2530 			dst_release(ndst);
2531 			goto out_unlock;
2532 		}
2533 
2534 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2535 		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2536 		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2537 				      vni, md, flags, udp_sum);
2538 		if (err < 0)
2539 			goto tx_error;
2540 
2541 		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr,
2542 				    pkey->u.ipv4.dst, tos, ttl, df,
2543 				    src_port, dst_port, xnet, !udp_sum);
2544 #if IS_ENABLED(CONFIG_IPV6)
2545 	} else {
2546 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2547 		struct in6_addr saddr;
2548 
2549 		if (!ifindex)
2550 			ifindex = sock6->sock->sk->sk_bound_dev_if;
2551 
2552 		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
2553 					      ifindex, &saddr, pkey,
2554 					      src_port, dst_port, tos,
2555 					      use_cache ? dst_cache : NULL);
2556 		if (IS_ERR(ndst)) {
2557 			err = PTR_ERR(ndst);
2558 			ndst = NULL;
2559 			goto tx_error;
2560 		}
2561 
2562 		if (!info) {
2563 			u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags;
2564 
2565 			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6,
2566 						    dst_port, ifindex, vni,
2567 						    ndst, rt6i_flags);
2568 			if (err)
2569 				goto out_unlock;
2570 		}
2571 
2572 		err = skb_tunnel_check_pmtu(skb, ndst,
2573 					    vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2574 					    netif_is_any_bridge_port(dev));
2575 		if (err < 0) {
2576 			goto tx_error;
2577 		} else if (err) {
2578 			if (info) {
2579 				struct ip_tunnel_info *unclone;
2580 
2581 				unclone = skb_tunnel_info_unclone(skb);
2582 				if (unlikely(!unclone))
2583 					goto tx_error;
2584 
2585 				unclone->key.u.ipv6.src = pkey->u.ipv6.dst;
2586 				unclone->key.u.ipv6.dst = saddr;
2587 			}
2588 
2589 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2590 			dst_release(ndst);
2591 			goto out_unlock;
2592 		}
2593 
2594 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2595 		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2596 		skb_scrub_packet(skb, xnet);
2597 		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2598 				      vni, md, flags, udp_sum);
2599 		if (err < 0)
2600 			goto tx_error;
2601 
2602 		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2603 				     &saddr, &pkey->u.ipv6.dst, tos, ttl,
2604 				     pkey->label, src_port, dst_port, !udp_sum);
2605 #endif
2606 	}
2607 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2608 out_unlock:
2609 	rcu_read_unlock();
2610 	return;
2611 
2612 drop:
2613 	dev_core_stats_tx_dropped_inc(dev);
2614 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2615 	dev_kfree_skb(skb);
2616 	return;
2617 
2618 tx_error:
2619 	rcu_read_unlock();
2620 	if (err == -ELOOP)
2621 		DEV_STATS_INC(dev, collisions);
2622 	else if (err == -ENETUNREACH)
2623 		DEV_STATS_INC(dev, tx_carrier_errors);
2624 	dst_release(ndst);
2625 	DEV_STATS_INC(dev, tx_errors);
2626 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2627 	kfree_skb(skb);
2628 }
2629 
2630 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2631 			  struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2632 {
2633 	struct vxlan_rdst nh_rdst;
2634 	struct nexthop *nh;
2635 	bool do_xmit;
2636 	u32 hash;
2637 
2638 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2639 	hash = skb_get_hash(skb);
2640 
2641 	rcu_read_lock();
2642 	nh = rcu_dereference(f->nh);
2643 	if (!nh) {
2644 		rcu_read_unlock();
2645 		goto drop;
2646 	}
2647 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2648 	rcu_read_unlock();
2649 
2650 	if (likely(do_xmit))
2651 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2652 	else
2653 		goto drop;
2654 
2655 	return;
2656 
2657 drop:
2658 	dev_core_stats_tx_dropped_inc(dev);
2659 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2660 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2661 	dev_kfree_skb(skb);
2662 }
2663 
2664 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2665 				   u32 nhid, __be32 vni)
2666 {
2667 	struct vxlan_dev *vxlan = netdev_priv(dev);
2668 	struct vxlan_rdst nh_rdst;
2669 	struct nexthop *nh;
2670 	bool do_xmit;
2671 	u32 hash;
2672 
2673 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2674 	hash = skb_get_hash(skb);
2675 
2676 	rcu_read_lock();
2677 	nh = nexthop_find_by_id(dev_net(dev), nhid);
2678 	if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2679 		rcu_read_unlock();
2680 		goto drop;
2681 	}
2682 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2683 	rcu_read_unlock();
2684 
2685 	if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2686 		goto drop;
2687 
2688 	if (likely(do_xmit))
2689 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2690 	else
2691 		goto drop;
2692 
2693 	return NETDEV_TX_OK;
2694 
2695 drop:
2696 	dev_core_stats_tx_dropped_inc(dev);
2697 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2698 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2699 	dev_kfree_skb(skb);
2700 	return NETDEV_TX_OK;
2701 }
2702 
2703 /* Transmit local packets over Vxlan
2704  *
2705  * Outer IP header inherits ECN and DF from inner header.
2706  * Outer UDP destination is the VXLAN assigned port.
2707  *           source port is based on hash of flow
2708  */
2709 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2710 {
2711 	struct vxlan_dev *vxlan = netdev_priv(dev);
2712 	struct vxlan_rdst *rdst, *fdst = NULL;
2713 	const struct ip_tunnel_info *info;
2714 	struct vxlan_fdb *f;
2715 	struct ethhdr *eth;
2716 	__be32 vni = 0;
2717 	u32 nhid = 0;
2718 	bool did_rsc;
2719 
2720 	info = skb_tunnel_info(skb);
2721 
2722 	skb_reset_mac_header(skb);
2723 
2724 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2725 		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2726 		    info->mode & IP_TUNNEL_INFO_TX) {
2727 			vni = tunnel_id_to_key32(info->key.tun_id);
2728 			nhid = info->key.nhid;
2729 		} else {
2730 			if (info && info->mode & IP_TUNNEL_INFO_TX)
2731 				vxlan_xmit_one(skb, dev, vni, NULL, false);
2732 			else
2733 				kfree_skb(skb);
2734 			return NETDEV_TX_OK;
2735 		}
2736 	}
2737 
2738 	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2739 		eth = eth_hdr(skb);
2740 		if (ntohs(eth->h_proto) == ETH_P_ARP)
2741 			return arp_reduce(dev, skb, vni);
2742 #if IS_ENABLED(CONFIG_IPV6)
2743 		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2744 			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2745 					    sizeof(struct nd_msg)) &&
2746 			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2747 			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2748 
2749 			if (m->icmph.icmp6_code == 0 &&
2750 			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2751 				return neigh_reduce(dev, skb, vni);
2752 		}
2753 #endif
2754 	}
2755 
2756 	if (nhid)
2757 		return vxlan_xmit_nhid(skb, dev, nhid, vni);
2758 
2759 	if (vxlan->cfg.flags & VXLAN_F_MDB) {
2760 		struct vxlan_mdb_entry *mdb_entry;
2761 
2762 		rcu_read_lock();
2763 		mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2764 		if (mdb_entry) {
2765 			netdev_tx_t ret;
2766 
2767 			ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2768 			rcu_read_unlock();
2769 			return ret;
2770 		}
2771 		rcu_read_unlock();
2772 	}
2773 
2774 	eth = eth_hdr(skb);
2775 	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2776 	did_rsc = false;
2777 
2778 	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2779 	    (ntohs(eth->h_proto) == ETH_P_IP ||
2780 	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2781 		did_rsc = route_shortcircuit(dev, skb);
2782 		if (did_rsc)
2783 			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2784 	}
2785 
2786 	if (f == NULL) {
2787 		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2788 		if (f == NULL) {
2789 			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2790 			    !is_multicast_ether_addr(eth->h_dest))
2791 				vxlan_fdb_miss(vxlan, eth->h_dest);
2792 
2793 			dev_core_stats_tx_dropped_inc(dev);
2794 			vxlan_vnifilter_count(vxlan, vni, NULL,
2795 					      VXLAN_VNI_STATS_TX_DROPS, 0);
2796 			kfree_skb(skb);
2797 			return NETDEV_TX_OK;
2798 		}
2799 	}
2800 
2801 	if (rcu_access_pointer(f->nh)) {
2802 		vxlan_xmit_nh(skb, dev, f,
2803 			      (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2804 	} else {
2805 		list_for_each_entry_rcu(rdst, &f->remotes, list) {
2806 			struct sk_buff *skb1;
2807 
2808 			if (!fdst) {
2809 				fdst = rdst;
2810 				continue;
2811 			}
2812 			skb1 = skb_clone(skb, GFP_ATOMIC);
2813 			if (skb1)
2814 				vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2815 		}
2816 		if (fdst)
2817 			vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2818 		else
2819 			kfree_skb(skb);
2820 	}
2821 
2822 	return NETDEV_TX_OK;
2823 }
2824 
2825 /* Walk the forwarding table and purge stale entries */
2826 static void vxlan_cleanup(struct timer_list *t)
2827 {
2828 	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2829 	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2830 	unsigned int h;
2831 
2832 	if (!netif_running(vxlan->dev))
2833 		return;
2834 
2835 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2836 		struct hlist_node *p, *n;
2837 
2838 		spin_lock(&vxlan->hash_lock[h]);
2839 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2840 			struct vxlan_fdb *f
2841 				= container_of(p, struct vxlan_fdb, hlist);
2842 			unsigned long timeout;
2843 
2844 			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2845 				continue;
2846 
2847 			if (f->flags & NTF_EXT_LEARNED)
2848 				continue;
2849 
2850 			timeout = f->used + vxlan->cfg.age_interval * HZ;
2851 			if (time_before_eq(timeout, jiffies)) {
2852 				netdev_dbg(vxlan->dev,
2853 					   "garbage collect %pM\n",
2854 					   f->eth_addr);
2855 				f->state = NUD_STALE;
2856 				vxlan_fdb_destroy(vxlan, f, true, true);
2857 			} else if (time_before(timeout, next_timer))
2858 				next_timer = timeout;
2859 		}
2860 		spin_unlock(&vxlan->hash_lock[h]);
2861 	}
2862 
2863 	mod_timer(&vxlan->age_timer, next_timer);
2864 }
2865 
2866 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2867 {
2868 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2869 
2870 	spin_lock(&vn->sock_lock);
2871 	hlist_del_init_rcu(&vxlan->hlist4.hlist);
2872 #if IS_ENABLED(CONFIG_IPV6)
2873 	hlist_del_init_rcu(&vxlan->hlist6.hlist);
2874 #endif
2875 	spin_unlock(&vn->sock_lock);
2876 }
2877 
2878 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2879 			     struct vxlan_dev_node *node)
2880 {
2881 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2882 	__be32 vni = vxlan->default_dst.remote_vni;
2883 
2884 	node->vxlan = vxlan;
2885 	spin_lock(&vn->sock_lock);
2886 	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2887 	spin_unlock(&vn->sock_lock);
2888 }
2889 
2890 /* Setup stats when device is created */
2891 static int vxlan_init(struct net_device *dev)
2892 {
2893 	struct vxlan_dev *vxlan = netdev_priv(dev);
2894 	int err;
2895 
2896 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2897 		vxlan_vnigroup_init(vxlan);
2898 
2899 	err = gro_cells_init(&vxlan->gro_cells, dev);
2900 	if (err)
2901 		goto err_vnigroup_uninit;
2902 
2903 	err = vxlan_mdb_init(vxlan);
2904 	if (err)
2905 		goto err_gro_cells_destroy;
2906 
2907 	netdev_lockdep_set_classes(dev);
2908 	return 0;
2909 
2910 err_gro_cells_destroy:
2911 	gro_cells_destroy(&vxlan->gro_cells);
2912 err_vnigroup_uninit:
2913 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2914 		vxlan_vnigroup_uninit(vxlan);
2915 	return err;
2916 }
2917 
2918 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2919 {
2920 	struct vxlan_fdb *f;
2921 	u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2922 
2923 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
2924 	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2925 	if (f)
2926 		vxlan_fdb_destroy(vxlan, f, true, true);
2927 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2928 }
2929 
2930 static void vxlan_uninit(struct net_device *dev)
2931 {
2932 	struct vxlan_dev *vxlan = netdev_priv(dev);
2933 
2934 	vxlan_mdb_fini(vxlan);
2935 
2936 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2937 		vxlan_vnigroup_uninit(vxlan);
2938 
2939 	gro_cells_destroy(&vxlan->gro_cells);
2940 
2941 	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2942 }
2943 
2944 /* Start ageing timer and join group when device is brought up */
2945 static int vxlan_open(struct net_device *dev)
2946 {
2947 	struct vxlan_dev *vxlan = netdev_priv(dev);
2948 	int ret;
2949 
2950 	ret = vxlan_sock_add(vxlan);
2951 	if (ret < 0)
2952 		return ret;
2953 
2954 	ret = vxlan_multicast_join(vxlan);
2955 	if (ret) {
2956 		vxlan_sock_release(vxlan);
2957 		return ret;
2958 	}
2959 
2960 	if (vxlan->cfg.age_interval)
2961 		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2962 
2963 	return ret;
2964 }
2965 
2966 struct vxlan_fdb_flush_desc {
2967 	bool				ignore_default_entry;
2968 	unsigned long                   state;
2969 	unsigned long			state_mask;
2970 	unsigned long                   flags;
2971 	unsigned long			flags_mask;
2972 	__be32				src_vni;
2973 	u32				nhid;
2974 	__be32				vni;
2975 	__be16				port;
2976 	union vxlan_addr		dst_ip;
2977 };
2978 
2979 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f,
2980 				       const struct vxlan_dev *vxlan)
2981 {
2982 	return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni;
2983 }
2984 
2985 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid)
2986 {
2987 	struct nexthop *nh = rtnl_dereference(f->nh);
2988 
2989 	return nh && nh->id == nhid;
2990 }
2991 
2992 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f,
2993 				    const struct vxlan_dev *vxlan,
2994 				    const struct vxlan_fdb_flush_desc *desc)
2995 {
2996 	if (desc->state_mask && (f->state & desc->state_mask) != desc->state)
2997 		return false;
2998 
2999 	if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags)
3000 		return false;
3001 
3002 	if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan))
3003 		return false;
3004 
3005 	if (desc->src_vni && f->vni != desc->src_vni)
3006 		return false;
3007 
3008 	if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid))
3009 		return false;
3010 
3011 	return true;
3012 }
3013 
3014 static bool
3015 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc)
3016 {
3017 	return desc->vni || desc->port || desc->dst_ip.sa.sa_family;
3018 }
3019 
3020 static bool
3021 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc,
3022 			       const struct vxlan_rdst *rd)
3023 {
3024 	if (desc->vni && rd->remote_vni != desc->vni)
3025 		return false;
3026 
3027 	if (desc->port && rd->remote_port != desc->port)
3028 		return false;
3029 
3030 	if (desc->dst_ip.sa.sa_family &&
3031 	    !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip))
3032 		return false;
3033 
3034 	return true;
3035 }
3036 
3037 static void
3038 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan,
3039 			      const struct vxlan_fdb_flush_desc *desc,
3040 			      bool *p_destroy_fdb)
3041 {
3042 	bool remotes_flushed = false;
3043 	struct vxlan_rdst *rd, *tmp;
3044 
3045 	list_for_each_entry_safe(rd, tmp, &f->remotes, list) {
3046 		if (!vxlan_fdb_flush_remote_matches(desc, rd))
3047 			continue;
3048 
3049 		vxlan_fdb_dst_destroy(vxlan, f, rd, true);
3050 		remotes_flushed = true;
3051 	}
3052 
3053 	*p_destroy_fdb = remotes_flushed && list_empty(&f->remotes);
3054 }
3055 
3056 /* Purge the forwarding table */
3057 static void vxlan_flush(struct vxlan_dev *vxlan,
3058 			const struct vxlan_fdb_flush_desc *desc)
3059 {
3060 	bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc);
3061 	unsigned int h;
3062 
3063 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3064 		struct hlist_node *p, *n;
3065 
3066 		spin_lock_bh(&vxlan->hash_lock[h]);
3067 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3068 			struct vxlan_fdb *f
3069 				= container_of(p, struct vxlan_fdb, hlist);
3070 
3071 			if (!vxlan_fdb_flush_matches(f, vxlan, desc))
3072 				continue;
3073 
3074 			if (match_remotes) {
3075 				bool destroy_fdb = false;
3076 
3077 				vxlan_fdb_flush_match_remotes(f, vxlan, desc,
3078 							      &destroy_fdb);
3079 
3080 				if (!destroy_fdb)
3081 					continue;
3082 			}
3083 
3084 			vxlan_fdb_destroy(vxlan, f, true, true);
3085 		}
3086 		spin_unlock_bh(&vxlan->hash_lock[h]);
3087 	}
3088 }
3089 
3090 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = {
3091 	[NDA_SRC_VNI]   = { .type = NLA_U32 },
3092 	[NDA_NH_ID]	= { .type = NLA_U32 },
3093 	[NDA_VNI]	= { .type = NLA_U32 },
3094 	[NDA_PORT]	= { .type = NLA_U16 },
3095 	[NDA_DST]	= NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr),
3096 					   sizeof(struct in6_addr)),
3097 	[NDA_NDM_STATE_MASK]	= { .type = NLA_U16 },
3098 	[NDA_NDM_FLAGS_MASK]	= { .type = NLA_U8 },
3099 };
3100 
3101 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF)
3102 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP)
3103 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \
3104 					   NTF_ROUTER)
3105 
3106 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev,
3107 				 struct netlink_ext_ack *extack)
3108 {
3109 	struct vxlan_dev *vxlan = netdev_priv(dev);
3110 	struct vxlan_fdb_flush_desc desc = {};
3111 	struct ndmsg *ndm = nlmsg_data(nlh);
3112 	struct nlattr *tb[NDA_MAX + 1];
3113 	u8 ndm_flags;
3114 	int err;
3115 
3116 	ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS;
3117 
3118 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy,
3119 			  extack);
3120 	if (err)
3121 		return err;
3122 
3123 	if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) {
3124 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set");
3125 		return -EINVAL;
3126 	}
3127 	if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) {
3128 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set");
3129 		return -EINVAL;
3130 	}
3131 
3132 	desc.state = ndm->ndm_state;
3133 	desc.flags = ndm_flags;
3134 
3135 	if (tb[NDA_NDM_STATE_MASK])
3136 		desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]);
3137 
3138 	if (tb[NDA_NDM_FLAGS_MASK])
3139 		desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]);
3140 
3141 	if (tb[NDA_SRC_VNI])
3142 		desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
3143 
3144 	if (tb[NDA_NH_ID])
3145 		desc.nhid = nla_get_u32(tb[NDA_NH_ID]);
3146 
3147 	if (tb[NDA_VNI])
3148 		desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
3149 
3150 	if (tb[NDA_PORT])
3151 		desc.port = nla_get_be16(tb[NDA_PORT]);
3152 
3153 	if (tb[NDA_DST]) {
3154 		union vxlan_addr ip;
3155 
3156 		err = vxlan_nla_get_addr(&ip, tb[NDA_DST]);
3157 		if (err) {
3158 			NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST],
3159 					    "Unsupported address family");
3160 			return err;
3161 		}
3162 		desc.dst_ip = ip;
3163 	}
3164 
3165 	vxlan_flush(vxlan, &desc);
3166 
3167 	return 0;
3168 }
3169 
3170 /* Cleanup timer and forwarding table on shutdown */
3171 static int vxlan_stop(struct net_device *dev)
3172 {
3173 	struct vxlan_dev *vxlan = netdev_priv(dev);
3174 	struct vxlan_fdb_flush_desc desc = {
3175 		/* Default entry is deleted at vxlan_uninit. */
3176 		.ignore_default_entry = true,
3177 		.state = 0,
3178 		.state_mask = NUD_PERMANENT | NUD_NOARP,
3179 	};
3180 
3181 	vxlan_multicast_leave(vxlan);
3182 
3183 	del_timer_sync(&vxlan->age_timer);
3184 
3185 	vxlan_flush(vxlan, &desc);
3186 	vxlan_sock_release(vxlan);
3187 
3188 	return 0;
3189 }
3190 
3191 /* Stub, nothing needs to be done. */
3192 static void vxlan_set_multicast_list(struct net_device *dev)
3193 {
3194 }
3195 
3196 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3197 {
3198 	struct vxlan_dev *vxlan = netdev_priv(dev);
3199 	struct vxlan_rdst *dst = &vxlan->default_dst;
3200 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3201 							 dst->remote_ifindex);
3202 
3203 	/* This check is different than dev->max_mtu, because it looks at
3204 	 * the lowerdev->mtu, rather than the static dev->max_mtu
3205 	 */
3206 	if (lowerdev) {
3207 		int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3208 		if (new_mtu > max_mtu)
3209 			return -EINVAL;
3210 	}
3211 
3212 	WRITE_ONCE(dev->mtu, new_mtu);
3213 	return 0;
3214 }
3215 
3216 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3217 {
3218 	struct vxlan_dev *vxlan = netdev_priv(dev);
3219 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
3220 	__be16 sport, dport;
3221 
3222 	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3223 				  vxlan->cfg.port_max, true);
3224 	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3225 
3226 	if (ip_tunnel_info_af(info) == AF_INET) {
3227 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3228 		struct rtable *rt;
3229 
3230 		if (!sock4)
3231 			return -EIO;
3232 
3233 		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0,
3234 					   &info->key.u.ipv4.src,
3235 					   &info->key,
3236 					   sport, dport, info->key.tos,
3237 					   &info->dst_cache);
3238 		if (IS_ERR(rt))
3239 			return PTR_ERR(rt);
3240 		ip_rt_put(rt);
3241 	} else {
3242 #if IS_ENABLED(CONFIG_IPV6)
3243 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3244 		struct dst_entry *ndst;
3245 
3246 		if (!sock6)
3247 			return -EIO;
3248 
3249 		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
3250 					      0, &info->key.u.ipv6.src,
3251 					      &info->key,
3252 					      sport, dport, info->key.tos,
3253 					      &info->dst_cache);
3254 		if (IS_ERR(ndst))
3255 			return PTR_ERR(ndst);
3256 		dst_release(ndst);
3257 #else /* !CONFIG_IPV6 */
3258 		return -EPFNOSUPPORT;
3259 #endif
3260 	}
3261 	info->key.tp_src = sport;
3262 	info->key.tp_dst = dport;
3263 	return 0;
3264 }
3265 
3266 static const struct net_device_ops vxlan_netdev_ether_ops = {
3267 	.ndo_init		= vxlan_init,
3268 	.ndo_uninit		= vxlan_uninit,
3269 	.ndo_open		= vxlan_open,
3270 	.ndo_stop		= vxlan_stop,
3271 	.ndo_start_xmit		= vxlan_xmit,
3272 	.ndo_set_rx_mode	= vxlan_set_multicast_list,
3273 	.ndo_change_mtu		= vxlan_change_mtu,
3274 	.ndo_validate_addr	= eth_validate_addr,
3275 	.ndo_set_mac_address	= eth_mac_addr,
3276 	.ndo_fdb_add		= vxlan_fdb_add,
3277 	.ndo_fdb_del		= vxlan_fdb_delete,
3278 	.ndo_fdb_del_bulk	= vxlan_fdb_delete_bulk,
3279 	.ndo_fdb_dump		= vxlan_fdb_dump,
3280 	.ndo_fdb_get		= vxlan_fdb_get,
3281 	.ndo_mdb_add		= vxlan_mdb_add,
3282 	.ndo_mdb_del		= vxlan_mdb_del,
3283 	.ndo_mdb_del_bulk	= vxlan_mdb_del_bulk,
3284 	.ndo_mdb_dump		= vxlan_mdb_dump,
3285 	.ndo_mdb_get		= vxlan_mdb_get,
3286 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3287 };
3288 
3289 static const struct net_device_ops vxlan_netdev_raw_ops = {
3290 	.ndo_init		= vxlan_init,
3291 	.ndo_uninit		= vxlan_uninit,
3292 	.ndo_open		= vxlan_open,
3293 	.ndo_stop		= vxlan_stop,
3294 	.ndo_start_xmit		= vxlan_xmit,
3295 	.ndo_change_mtu		= vxlan_change_mtu,
3296 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3297 };
3298 
3299 /* Info for udev, that this is a virtual tunnel endpoint */
3300 static const struct device_type vxlan_type = {
3301 	.name = "vxlan",
3302 };
3303 
3304 /* Calls the ndo_udp_tunnel_add of the caller in order to
3305  * supply the listening VXLAN udp ports. Callers are expected
3306  * to implement the ndo_udp_tunnel_add.
3307  */
3308 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3309 {
3310 	struct vxlan_sock *vs;
3311 	struct net *net = dev_net(dev);
3312 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3313 	unsigned int i;
3314 
3315 	spin_lock(&vn->sock_lock);
3316 	for (i = 0; i < PORT_HASH_SIZE; ++i) {
3317 		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3318 			unsigned short type;
3319 
3320 			if (vs->flags & VXLAN_F_GPE)
3321 				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3322 			else
3323 				type = UDP_TUNNEL_TYPE_VXLAN;
3324 
3325 			if (push)
3326 				udp_tunnel_push_rx_port(dev, vs->sock, type);
3327 			else
3328 				udp_tunnel_drop_rx_port(dev, vs->sock, type);
3329 		}
3330 	}
3331 	spin_unlock(&vn->sock_lock);
3332 }
3333 
3334 /* Initialize the device structure. */
3335 static void vxlan_setup(struct net_device *dev)
3336 {
3337 	struct vxlan_dev *vxlan = netdev_priv(dev);
3338 	unsigned int h;
3339 
3340 	eth_hw_addr_random(dev);
3341 	ether_setup(dev);
3342 
3343 	dev->needs_free_netdev = true;
3344 	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3345 
3346 	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3347 	dev->features   |= NETIF_F_RXCSUM;
3348 	dev->features   |= NETIF_F_GSO_SOFTWARE;
3349 
3350 	dev->vlan_features = dev->features;
3351 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3352 	dev->hw_features |= NETIF_F_RXCSUM;
3353 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3354 	netif_keep_dst(dev);
3355 	dev->priv_flags |= IFF_NO_QUEUE;
3356 	dev->change_proto_down = true;
3357 	dev->lltx = true;
3358 
3359 	/* MTU range: 68 - 65535 */
3360 	dev->min_mtu = ETH_MIN_MTU;
3361 	dev->max_mtu = ETH_MAX_MTU;
3362 
3363 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
3364 	INIT_LIST_HEAD(&vxlan->next);
3365 
3366 	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3367 
3368 	vxlan->dev = dev;
3369 
3370 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3371 		spin_lock_init(&vxlan->hash_lock[h]);
3372 		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3373 	}
3374 }
3375 
3376 static void vxlan_ether_setup(struct net_device *dev)
3377 {
3378 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3379 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3380 	dev->netdev_ops = &vxlan_netdev_ether_ops;
3381 }
3382 
3383 static void vxlan_raw_setup(struct net_device *dev)
3384 {
3385 	dev->header_ops = NULL;
3386 	dev->type = ARPHRD_NONE;
3387 	dev->hard_header_len = 0;
3388 	dev->addr_len = 0;
3389 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3390 	dev->netdev_ops = &vxlan_netdev_raw_ops;
3391 }
3392 
3393 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3394 	[IFLA_VXLAN_UNSPEC]     = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3395 	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
3396 	[IFLA_VXLAN_GROUP]	= { .len = sizeof_field(struct iphdr, daddr) },
3397 	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
3398 	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
3399 	[IFLA_VXLAN_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
3400 	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
3401 	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
3402 	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
3403 	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
3404 	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
3405 	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
3406 	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
3407 	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3408 	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
3409 	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
3410 	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
3411 	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
3412 	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
3413 	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
3414 	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
3415 	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
3416 	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
3417 	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
3418 	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
3419 	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
3420 	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
3421 	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
3422 	[IFLA_VXLAN_TTL_INHERIT]	= { .type = NLA_FLAG },
3423 	[IFLA_VXLAN_DF]		= { .type = NLA_U8 },
3424 	[IFLA_VXLAN_VNIFILTER]	= { .type = NLA_U8 },
3425 	[IFLA_VXLAN_LOCALBYPASS]	= NLA_POLICY_MAX(NLA_U8, 1),
3426 	[IFLA_VXLAN_LABEL_POLICY]       = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX),
3427 };
3428 
3429 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3430 			  struct netlink_ext_ack *extack)
3431 {
3432 	if (tb[IFLA_ADDRESS]) {
3433 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3434 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3435 					    "Provided link layer address is not Ethernet");
3436 			return -EINVAL;
3437 		}
3438 
3439 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3440 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3441 					    "Provided Ethernet address is not unicast");
3442 			return -EADDRNOTAVAIL;
3443 		}
3444 	}
3445 
3446 	if (tb[IFLA_MTU]) {
3447 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3448 
3449 		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3450 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3451 					    "MTU must be between 68 and 65535");
3452 			return -EINVAL;
3453 		}
3454 	}
3455 
3456 	if (!data) {
3457 		NL_SET_ERR_MSG(extack,
3458 			       "Required attributes not provided to perform the operation");
3459 		return -EINVAL;
3460 	}
3461 
3462 	if (data[IFLA_VXLAN_ID]) {
3463 		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3464 
3465 		if (id >= VXLAN_N_VID) {
3466 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3467 					    "VXLAN ID must be lower than 16777216");
3468 			return -ERANGE;
3469 		}
3470 	}
3471 
3472 	if (data[IFLA_VXLAN_PORT_RANGE]) {
3473 		const struct ifla_vxlan_port_range *p
3474 			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3475 
3476 		if (ntohs(p->high) < ntohs(p->low)) {
3477 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3478 					    "Invalid source port range");
3479 			return -EINVAL;
3480 		}
3481 	}
3482 
3483 	if (data[IFLA_VXLAN_DF]) {
3484 		enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3485 
3486 		if (df < 0 || df > VXLAN_DF_MAX) {
3487 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3488 					    "Invalid DF attribute");
3489 			return -EINVAL;
3490 		}
3491 	}
3492 
3493 	return 0;
3494 }
3495 
3496 static void vxlan_get_drvinfo(struct net_device *netdev,
3497 			      struct ethtool_drvinfo *drvinfo)
3498 {
3499 	strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3500 	strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3501 }
3502 
3503 static int vxlan_get_link_ksettings(struct net_device *dev,
3504 				    struct ethtool_link_ksettings *cmd)
3505 {
3506 	struct vxlan_dev *vxlan = netdev_priv(dev);
3507 	struct vxlan_rdst *dst = &vxlan->default_dst;
3508 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3509 							 dst->remote_ifindex);
3510 
3511 	if (!lowerdev) {
3512 		cmd->base.duplex = DUPLEX_UNKNOWN;
3513 		cmd->base.port = PORT_OTHER;
3514 		cmd->base.speed = SPEED_UNKNOWN;
3515 
3516 		return 0;
3517 	}
3518 
3519 	return __ethtool_get_link_ksettings(lowerdev, cmd);
3520 }
3521 
3522 static const struct ethtool_ops vxlan_ethtool_ops = {
3523 	.get_drvinfo		= vxlan_get_drvinfo,
3524 	.get_link		= ethtool_op_get_link,
3525 	.get_link_ksettings	= vxlan_get_link_ksettings,
3526 };
3527 
3528 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3529 					__be16 port, u32 flags, int ifindex)
3530 {
3531 	struct socket *sock;
3532 	struct udp_port_cfg udp_conf;
3533 	int err;
3534 
3535 	memset(&udp_conf, 0, sizeof(udp_conf));
3536 
3537 	if (ipv6) {
3538 		udp_conf.family = AF_INET6;
3539 		udp_conf.use_udp6_rx_checksums =
3540 		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3541 		udp_conf.ipv6_v6only = 1;
3542 	} else {
3543 		udp_conf.family = AF_INET;
3544 	}
3545 
3546 	udp_conf.local_udp_port = port;
3547 	udp_conf.bind_ifindex = ifindex;
3548 
3549 	/* Open UDP socket */
3550 	err = udp_sock_create(net, &udp_conf, &sock);
3551 	if (err < 0)
3552 		return ERR_PTR(err);
3553 
3554 	udp_allow_gso(sock->sk);
3555 	return sock;
3556 }
3557 
3558 /* Create new listen socket if needed */
3559 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3560 					      __be16 port, u32 flags,
3561 					      int ifindex)
3562 {
3563 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3564 	struct vxlan_sock *vs;
3565 	struct socket *sock;
3566 	unsigned int h;
3567 	struct udp_tunnel_sock_cfg tunnel_cfg;
3568 
3569 	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3570 	if (!vs)
3571 		return ERR_PTR(-ENOMEM);
3572 
3573 	for (h = 0; h < VNI_HASH_SIZE; ++h)
3574 		INIT_HLIST_HEAD(&vs->vni_list[h]);
3575 
3576 	sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3577 	if (IS_ERR(sock)) {
3578 		kfree(vs);
3579 		return ERR_CAST(sock);
3580 	}
3581 
3582 	vs->sock = sock;
3583 	refcount_set(&vs->refcnt, 1);
3584 	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3585 
3586 	spin_lock(&vn->sock_lock);
3587 	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3588 	udp_tunnel_notify_add_rx_port(sock,
3589 				      (vs->flags & VXLAN_F_GPE) ?
3590 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
3591 				      UDP_TUNNEL_TYPE_VXLAN);
3592 	spin_unlock(&vn->sock_lock);
3593 
3594 	/* Mark socket as an encapsulation socket. */
3595 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3596 	tunnel_cfg.sk_user_data = vs;
3597 	tunnel_cfg.encap_type = 1;
3598 	tunnel_cfg.encap_rcv = vxlan_rcv;
3599 	tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3600 	tunnel_cfg.encap_destroy = NULL;
3601 	if (vs->flags & VXLAN_F_GPE) {
3602 		tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3603 		tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3604 	} else {
3605 		tunnel_cfg.gro_receive = vxlan_gro_receive;
3606 		tunnel_cfg.gro_complete = vxlan_gro_complete;
3607 	}
3608 
3609 	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3610 
3611 	return vs;
3612 }
3613 
3614 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3615 {
3616 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3617 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3618 	struct vxlan_sock *vs = NULL;
3619 	struct vxlan_dev_node *node;
3620 	int l3mdev_index = 0;
3621 
3622 	if (vxlan->cfg.remote_ifindex)
3623 		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3624 			vxlan->net, vxlan->cfg.remote_ifindex);
3625 
3626 	if (!vxlan->cfg.no_share) {
3627 		spin_lock(&vn->sock_lock);
3628 		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3629 				     vxlan->cfg.dst_port, vxlan->cfg.flags,
3630 				     l3mdev_index);
3631 		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3632 			spin_unlock(&vn->sock_lock);
3633 			return -EBUSY;
3634 		}
3635 		spin_unlock(&vn->sock_lock);
3636 	}
3637 	if (!vs)
3638 		vs = vxlan_socket_create(vxlan->net, ipv6,
3639 					 vxlan->cfg.dst_port, vxlan->cfg.flags,
3640 					 l3mdev_index);
3641 	if (IS_ERR(vs))
3642 		return PTR_ERR(vs);
3643 #if IS_ENABLED(CONFIG_IPV6)
3644 	if (ipv6) {
3645 		rcu_assign_pointer(vxlan->vn6_sock, vs);
3646 		node = &vxlan->hlist6;
3647 	} else
3648 #endif
3649 	{
3650 		rcu_assign_pointer(vxlan->vn4_sock, vs);
3651 		node = &vxlan->hlist4;
3652 	}
3653 
3654 	if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3655 		vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3656 	else
3657 		vxlan_vs_add_dev(vs, vxlan, node);
3658 
3659 	return 0;
3660 }
3661 
3662 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3663 {
3664 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3665 	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3666 	bool ipv4 = !ipv6 || metadata;
3667 	int ret = 0;
3668 
3669 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3670 #if IS_ENABLED(CONFIG_IPV6)
3671 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3672 	if (ipv6) {
3673 		ret = __vxlan_sock_add(vxlan, true);
3674 		if (ret < 0 && ret != -EAFNOSUPPORT)
3675 			ipv4 = false;
3676 	}
3677 #endif
3678 	if (ipv4)
3679 		ret = __vxlan_sock_add(vxlan, false);
3680 	if (ret < 0)
3681 		vxlan_sock_release(vxlan);
3682 	return ret;
3683 }
3684 
3685 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3686 		     struct vxlan_config *conf, __be32 vni)
3687 {
3688 	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3689 	struct vxlan_dev *tmp;
3690 
3691 	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3692 		if (tmp == vxlan)
3693 			continue;
3694 		if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3695 			if (!vxlan_vnifilter_lookup(tmp, vni))
3696 				continue;
3697 		} else if (tmp->cfg.vni != vni) {
3698 			continue;
3699 		}
3700 		if (tmp->cfg.dst_port != conf->dst_port)
3701 			continue;
3702 		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3703 		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3704 			continue;
3705 
3706 		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3707 		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3708 			continue;
3709 
3710 		return -EEXIST;
3711 	}
3712 
3713 	return 0;
3714 }
3715 
3716 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3717 				 struct net_device **lower,
3718 				 struct vxlan_dev *old,
3719 				 struct netlink_ext_ack *extack)
3720 {
3721 	bool use_ipv6 = false;
3722 
3723 	if (conf->flags & VXLAN_F_GPE) {
3724 		/* For now, allow GPE only together with
3725 		 * COLLECT_METADATA. This can be relaxed later; in such
3726 		 * case, the other side of the PtP link will have to be
3727 		 * provided.
3728 		 */
3729 		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3730 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3731 			NL_SET_ERR_MSG(extack,
3732 				       "VXLAN GPE does not support this combination of attributes");
3733 			return -EINVAL;
3734 		}
3735 	}
3736 
3737 	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3738 		/* Unless IPv6 is explicitly requested, assume IPv4 */
3739 		conf->remote_ip.sa.sa_family = AF_INET;
3740 		conf->saddr.sa.sa_family = AF_INET;
3741 	} else if (!conf->remote_ip.sa.sa_family) {
3742 		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3743 	} else if (!conf->saddr.sa.sa_family) {
3744 		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3745 	}
3746 
3747 	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3748 		NL_SET_ERR_MSG(extack,
3749 			       "Local and remote address must be from the same family");
3750 		return -EINVAL;
3751 	}
3752 
3753 	if (vxlan_addr_multicast(&conf->saddr)) {
3754 		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3755 		return -EINVAL;
3756 	}
3757 
3758 	if (conf->saddr.sa.sa_family == AF_INET6) {
3759 		if (!IS_ENABLED(CONFIG_IPV6)) {
3760 			NL_SET_ERR_MSG(extack,
3761 				       "IPv6 support not enabled in the kernel");
3762 			return -EPFNOSUPPORT;
3763 		}
3764 		use_ipv6 = true;
3765 		conf->flags |= VXLAN_F_IPV6;
3766 
3767 		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3768 			int local_type =
3769 				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3770 			int remote_type =
3771 				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3772 
3773 			if (local_type & IPV6_ADDR_LINKLOCAL) {
3774 				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3775 				    (remote_type != IPV6_ADDR_ANY)) {
3776 					NL_SET_ERR_MSG(extack,
3777 						       "Invalid combination of local and remote address scopes");
3778 					return -EINVAL;
3779 				}
3780 
3781 				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3782 			} else {
3783 				if (remote_type ==
3784 				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3785 					NL_SET_ERR_MSG(extack,
3786 						       "Invalid combination of local and remote address scopes");
3787 					return -EINVAL;
3788 				}
3789 
3790 				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3791 			}
3792 		}
3793 	}
3794 
3795 	if (conf->label && !use_ipv6) {
3796 		NL_SET_ERR_MSG(extack,
3797 			       "Label attribute only applies to IPv6 VXLAN devices");
3798 		return -EINVAL;
3799 	}
3800 
3801 	if (conf->label_policy && !use_ipv6) {
3802 		NL_SET_ERR_MSG(extack,
3803 			       "Label policy only applies to IPv6 VXLAN devices");
3804 		return -EINVAL;
3805 	}
3806 
3807 	if (conf->remote_ifindex) {
3808 		struct net_device *lowerdev;
3809 
3810 		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3811 		if (!lowerdev) {
3812 			NL_SET_ERR_MSG(extack,
3813 				       "Invalid local interface, device not found");
3814 			return -ENODEV;
3815 		}
3816 
3817 #if IS_ENABLED(CONFIG_IPV6)
3818 		if (use_ipv6) {
3819 			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3820 
3821 			if (idev && idev->cnf.disable_ipv6) {
3822 				NL_SET_ERR_MSG(extack,
3823 					       "IPv6 support disabled by administrator");
3824 				return -EPERM;
3825 			}
3826 		}
3827 #endif
3828 
3829 		*lower = lowerdev;
3830 	} else {
3831 		if (vxlan_addr_multicast(&conf->remote_ip)) {
3832 			NL_SET_ERR_MSG(extack,
3833 				       "Local interface required for multicast remote destination");
3834 
3835 			return -EINVAL;
3836 		}
3837 
3838 #if IS_ENABLED(CONFIG_IPV6)
3839 		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3840 			NL_SET_ERR_MSG(extack,
3841 				       "Local interface required for link-local local/remote addresses");
3842 			return -EINVAL;
3843 		}
3844 #endif
3845 
3846 		*lower = NULL;
3847 	}
3848 
3849 	if (!conf->dst_port) {
3850 		if (conf->flags & VXLAN_F_GPE)
3851 			conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3852 		else
3853 			conf->dst_port = htons(vxlan_port);
3854 	}
3855 
3856 	if (!conf->age_interval)
3857 		conf->age_interval = FDB_AGE_DEFAULT;
3858 
3859 	if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3860 		NL_SET_ERR_MSG(extack,
3861 			       "A VXLAN device with the specified VNI already exists");
3862 		return -EEXIST;
3863 	}
3864 
3865 	return 0;
3866 }
3867 
3868 static void vxlan_config_apply(struct net_device *dev,
3869 			       struct vxlan_config *conf,
3870 			       struct net_device *lowerdev,
3871 			       struct net *src_net,
3872 			       bool changelink)
3873 {
3874 	struct vxlan_dev *vxlan = netdev_priv(dev);
3875 	struct vxlan_rdst *dst = &vxlan->default_dst;
3876 	unsigned short needed_headroom = ETH_HLEN;
3877 	int max_mtu = ETH_MAX_MTU;
3878 	u32 flags = conf->flags;
3879 
3880 	if (!changelink) {
3881 		if (flags & VXLAN_F_GPE)
3882 			vxlan_raw_setup(dev);
3883 		else
3884 			vxlan_ether_setup(dev);
3885 
3886 		if (conf->mtu)
3887 			dev->mtu = conf->mtu;
3888 
3889 		vxlan->net = src_net;
3890 	}
3891 
3892 	dst->remote_vni = conf->vni;
3893 
3894 	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3895 
3896 	if (lowerdev) {
3897 		dst->remote_ifindex = conf->remote_ifindex;
3898 
3899 		netif_inherit_tso_max(dev, lowerdev);
3900 
3901 		needed_headroom = lowerdev->hard_header_len;
3902 		needed_headroom += lowerdev->needed_headroom;
3903 
3904 		dev->needed_tailroom = lowerdev->needed_tailroom;
3905 
3906 		max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3907 		if (max_mtu < ETH_MIN_MTU)
3908 			max_mtu = ETH_MIN_MTU;
3909 
3910 		if (!changelink && !conf->mtu)
3911 			dev->mtu = max_mtu;
3912 	}
3913 
3914 	if (dev->mtu > max_mtu)
3915 		dev->mtu = max_mtu;
3916 
3917 	if (flags & VXLAN_F_COLLECT_METADATA)
3918 		flags |= VXLAN_F_IPV6;
3919 	needed_headroom += vxlan_headroom(flags);
3920 	dev->needed_headroom = needed_headroom;
3921 
3922 	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3923 }
3924 
3925 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3926 			       struct vxlan_config *conf, bool changelink,
3927 			       struct netlink_ext_ack *extack)
3928 {
3929 	struct vxlan_dev *vxlan = netdev_priv(dev);
3930 	struct net_device *lowerdev;
3931 	int ret;
3932 
3933 	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3934 	if (ret)
3935 		return ret;
3936 
3937 	vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3938 
3939 	return 0;
3940 }
3941 
3942 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3943 			      struct vxlan_config *conf,
3944 			      struct netlink_ext_ack *extack)
3945 {
3946 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3947 	struct vxlan_dev *vxlan = netdev_priv(dev);
3948 	struct net_device *remote_dev = NULL;
3949 	struct vxlan_fdb *f = NULL;
3950 	bool unregister = false;
3951 	struct vxlan_rdst *dst;
3952 	int err;
3953 
3954 	dst = &vxlan->default_dst;
3955 	err = vxlan_dev_configure(net, dev, conf, false, extack);
3956 	if (err)
3957 		return err;
3958 
3959 	dev->ethtool_ops = &vxlan_ethtool_ops;
3960 
3961 	/* create an fdb entry for a valid default destination */
3962 	if (!vxlan_addr_any(&dst->remote_ip)) {
3963 		err = vxlan_fdb_create(vxlan, all_zeros_mac,
3964 				       &dst->remote_ip,
3965 				       NUD_REACHABLE | NUD_PERMANENT,
3966 				       vxlan->cfg.dst_port,
3967 				       dst->remote_vni,
3968 				       dst->remote_vni,
3969 				       dst->remote_ifindex,
3970 				       NTF_SELF, 0, &f, extack);
3971 		if (err)
3972 			return err;
3973 	}
3974 
3975 	err = register_netdevice(dev);
3976 	if (err)
3977 		goto errout;
3978 	unregister = true;
3979 
3980 	if (dst->remote_ifindex) {
3981 		remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3982 		if (!remote_dev) {
3983 			err = -ENODEV;
3984 			goto errout;
3985 		}
3986 
3987 		err = netdev_upper_dev_link(remote_dev, dev, extack);
3988 		if (err)
3989 			goto errout;
3990 	}
3991 
3992 	err = rtnl_configure_link(dev, NULL, 0, NULL);
3993 	if (err < 0)
3994 		goto unlink;
3995 
3996 	if (f) {
3997 		vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3998 
3999 		/* notify default fdb entry */
4000 		err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4001 				       RTM_NEWNEIGH, true, extack);
4002 		if (err) {
4003 			vxlan_fdb_destroy(vxlan, f, false, false);
4004 			if (remote_dev)
4005 				netdev_upper_dev_unlink(remote_dev, dev);
4006 			goto unregister;
4007 		}
4008 	}
4009 
4010 	list_add(&vxlan->next, &vn->vxlan_list);
4011 	if (remote_dev)
4012 		dst->remote_dev = remote_dev;
4013 	return 0;
4014 unlink:
4015 	if (remote_dev)
4016 		netdev_upper_dev_unlink(remote_dev, dev);
4017 errout:
4018 	/* unregister_netdevice() destroys the default FDB entry with deletion
4019 	 * notification. But the addition notification was not sent yet, so
4020 	 * destroy the entry by hand here.
4021 	 */
4022 	if (f)
4023 		__vxlan_fdb_free(f);
4024 unregister:
4025 	if (unregister)
4026 		unregister_netdevice(dev);
4027 	return err;
4028 }
4029 
4030 /* Set/clear flags based on attribute */
4031 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
4032 			  int attrtype, unsigned long mask, bool changelink,
4033 			  bool changelink_supported,
4034 			  struct netlink_ext_ack *extack)
4035 {
4036 	unsigned long flags;
4037 
4038 	if (!tb[attrtype])
4039 		return 0;
4040 
4041 	if (changelink && !changelink_supported) {
4042 		vxlan_flag_attr_error(attrtype, extack);
4043 		return -EOPNOTSUPP;
4044 	}
4045 
4046 	if (vxlan_policy[attrtype].type == NLA_FLAG)
4047 		flags = conf->flags | mask;
4048 	else if (nla_get_u8(tb[attrtype]))
4049 		flags = conf->flags | mask;
4050 	else
4051 		flags = conf->flags & ~mask;
4052 
4053 	conf->flags = flags;
4054 
4055 	return 0;
4056 }
4057 
4058 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4059 			 struct net_device *dev, struct vxlan_config *conf,
4060 			 bool changelink, struct netlink_ext_ack *extack)
4061 {
4062 	struct vxlan_dev *vxlan = netdev_priv(dev);
4063 	int err = 0;
4064 
4065 	memset(conf, 0, sizeof(*conf));
4066 
4067 	/* if changelink operation, start with old existing cfg */
4068 	if (changelink)
4069 		memcpy(conf, &vxlan->cfg, sizeof(*conf));
4070 
4071 	if (data[IFLA_VXLAN_ID]) {
4072 		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4073 
4074 		if (changelink && (vni != conf->vni)) {
4075 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
4076 			return -EOPNOTSUPP;
4077 		}
4078 		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4079 	}
4080 
4081 	if (data[IFLA_VXLAN_GROUP]) {
4082 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4083 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4084 			return -EOPNOTSUPP;
4085 		}
4086 
4087 		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4088 		conf->remote_ip.sa.sa_family = AF_INET;
4089 	} else if (data[IFLA_VXLAN_GROUP6]) {
4090 		if (!IS_ENABLED(CONFIG_IPV6)) {
4091 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4092 			return -EPFNOSUPPORT;
4093 		}
4094 
4095 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4096 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4097 			return -EOPNOTSUPP;
4098 		}
4099 
4100 		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4101 		conf->remote_ip.sa.sa_family = AF_INET6;
4102 	}
4103 
4104 	if (data[IFLA_VXLAN_LOCAL]) {
4105 		if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4106 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4107 			return -EOPNOTSUPP;
4108 		}
4109 
4110 		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4111 		conf->saddr.sa.sa_family = AF_INET;
4112 	} else if (data[IFLA_VXLAN_LOCAL6]) {
4113 		if (!IS_ENABLED(CONFIG_IPV6)) {
4114 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4115 			return -EPFNOSUPPORT;
4116 		}
4117 
4118 		if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4119 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4120 			return -EOPNOTSUPP;
4121 		}
4122 
4123 		/* TODO: respect scope id */
4124 		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4125 		conf->saddr.sa.sa_family = AF_INET6;
4126 	}
4127 
4128 	if (data[IFLA_VXLAN_LINK])
4129 		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4130 
4131 	if (data[IFLA_VXLAN_TOS])
4132 		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4133 
4134 	if (data[IFLA_VXLAN_TTL])
4135 		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4136 
4137 	if (data[IFLA_VXLAN_TTL_INHERIT]) {
4138 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4139 				    VXLAN_F_TTL_INHERIT, changelink, false,
4140 				    extack);
4141 		if (err)
4142 			return err;
4143 
4144 	}
4145 
4146 	if (data[IFLA_VXLAN_LABEL])
4147 		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4148 			     IPV6_FLOWLABEL_MASK;
4149 	if (data[IFLA_VXLAN_LABEL_POLICY])
4150 		conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]);
4151 
4152 	if (data[IFLA_VXLAN_LEARNING]) {
4153 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4154 				    VXLAN_F_LEARN, changelink, true,
4155 				    extack);
4156 		if (err)
4157 			return err;
4158 	} else if (!changelink) {
4159 		/* default to learn on a new device */
4160 		conf->flags |= VXLAN_F_LEARN;
4161 	}
4162 
4163 	if (data[IFLA_VXLAN_AGEING])
4164 		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4165 
4166 	if (data[IFLA_VXLAN_PROXY]) {
4167 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4168 				    VXLAN_F_PROXY, changelink, false,
4169 				    extack);
4170 		if (err)
4171 			return err;
4172 	}
4173 
4174 	if (data[IFLA_VXLAN_RSC]) {
4175 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4176 				    VXLAN_F_RSC, changelink, false,
4177 				    extack);
4178 		if (err)
4179 			return err;
4180 	}
4181 
4182 	if (data[IFLA_VXLAN_L2MISS]) {
4183 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4184 				    VXLAN_F_L2MISS, changelink, false,
4185 				    extack);
4186 		if (err)
4187 			return err;
4188 	}
4189 
4190 	if (data[IFLA_VXLAN_L3MISS]) {
4191 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4192 				    VXLAN_F_L3MISS, changelink, false,
4193 				    extack);
4194 		if (err)
4195 			return err;
4196 	}
4197 
4198 	if (data[IFLA_VXLAN_LIMIT]) {
4199 		if (changelink) {
4200 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4201 					    "Cannot change limit");
4202 			return -EOPNOTSUPP;
4203 		}
4204 		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4205 	}
4206 
4207 	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4208 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4209 				    VXLAN_F_COLLECT_METADATA, changelink, false,
4210 				    extack);
4211 		if (err)
4212 			return err;
4213 	}
4214 
4215 	if (data[IFLA_VXLAN_PORT_RANGE]) {
4216 		if (!changelink) {
4217 			const struct ifla_vxlan_port_range *p
4218 				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4219 			conf->port_min = ntohs(p->low);
4220 			conf->port_max = ntohs(p->high);
4221 		} else {
4222 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4223 					    "Cannot change port range");
4224 			return -EOPNOTSUPP;
4225 		}
4226 	}
4227 
4228 	if (data[IFLA_VXLAN_PORT]) {
4229 		if (changelink) {
4230 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4231 					    "Cannot change port");
4232 			return -EOPNOTSUPP;
4233 		}
4234 		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4235 	}
4236 
4237 	if (data[IFLA_VXLAN_UDP_CSUM]) {
4238 		if (changelink) {
4239 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4240 					    "Cannot change UDP_CSUM flag");
4241 			return -EOPNOTSUPP;
4242 		}
4243 		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4244 			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4245 	}
4246 
4247 	if (data[IFLA_VXLAN_LOCALBYPASS]) {
4248 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4249 				    VXLAN_F_LOCALBYPASS, changelink,
4250 				    true, extack);
4251 		if (err)
4252 			return err;
4253 	} else if (!changelink) {
4254 		/* default to local bypass on a new device */
4255 		conf->flags |= VXLAN_F_LOCALBYPASS;
4256 	}
4257 
4258 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4259 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4260 				    VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4261 				    false, extack);
4262 		if (err)
4263 			return err;
4264 	}
4265 
4266 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4267 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4268 				    VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4269 				    false, extack);
4270 		if (err)
4271 			return err;
4272 	}
4273 
4274 	if (data[IFLA_VXLAN_REMCSUM_TX]) {
4275 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4276 				    VXLAN_F_REMCSUM_TX, changelink, false,
4277 				    extack);
4278 		if (err)
4279 			return err;
4280 	}
4281 
4282 	if (data[IFLA_VXLAN_REMCSUM_RX]) {
4283 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4284 				    VXLAN_F_REMCSUM_RX, changelink, false,
4285 				    extack);
4286 		if (err)
4287 			return err;
4288 	}
4289 
4290 	if (data[IFLA_VXLAN_GBP]) {
4291 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4292 				    VXLAN_F_GBP, changelink, false, extack);
4293 		if (err)
4294 			return err;
4295 	}
4296 
4297 	if (data[IFLA_VXLAN_GPE]) {
4298 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4299 				    VXLAN_F_GPE, changelink, false,
4300 				    extack);
4301 		if (err)
4302 			return err;
4303 	}
4304 
4305 	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4306 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4307 				    VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4308 				    false, extack);
4309 		if (err)
4310 			return err;
4311 	}
4312 
4313 	if (tb[IFLA_MTU]) {
4314 		if (changelink) {
4315 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4316 					    "Cannot change mtu");
4317 			return -EOPNOTSUPP;
4318 		}
4319 		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4320 	}
4321 
4322 	if (data[IFLA_VXLAN_DF])
4323 		conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4324 
4325 	if (data[IFLA_VXLAN_VNIFILTER]) {
4326 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4327 				    VXLAN_F_VNIFILTER, changelink, false,
4328 				    extack);
4329 		if (err)
4330 			return err;
4331 
4332 		if ((conf->flags & VXLAN_F_VNIFILTER) &&
4333 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4334 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4335 					    "vxlan vnifilter only valid in collect metadata mode");
4336 			return -EINVAL;
4337 		}
4338 	}
4339 
4340 	return 0;
4341 }
4342 
4343 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4344 			 struct nlattr *tb[], struct nlattr *data[],
4345 			 struct netlink_ext_ack *extack)
4346 {
4347 	struct vxlan_config conf;
4348 	int err;
4349 
4350 	err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4351 	if (err)
4352 		return err;
4353 
4354 	return __vxlan_dev_create(src_net, dev, &conf, extack);
4355 }
4356 
4357 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4358 			    struct nlattr *data[],
4359 			    struct netlink_ext_ack *extack)
4360 {
4361 	struct vxlan_dev *vxlan = netdev_priv(dev);
4362 	struct net_device *lowerdev;
4363 	struct vxlan_config conf;
4364 	struct vxlan_rdst *dst;
4365 	int err;
4366 
4367 	dst = &vxlan->default_dst;
4368 	err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4369 	if (err)
4370 		return err;
4371 
4372 	err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4373 				    vxlan, extack);
4374 	if (err)
4375 		return err;
4376 
4377 	if (dst->remote_dev == lowerdev)
4378 		lowerdev = NULL;
4379 
4380 	err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4381 					     extack);
4382 	if (err)
4383 		return err;
4384 
4385 	/* handle default dst entry */
4386 	if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4387 		u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4388 
4389 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4390 		if (!vxlan_addr_any(&conf.remote_ip)) {
4391 			err = vxlan_fdb_update(vxlan, all_zeros_mac,
4392 					       &conf.remote_ip,
4393 					       NUD_REACHABLE | NUD_PERMANENT,
4394 					       NLM_F_APPEND | NLM_F_CREATE,
4395 					       vxlan->cfg.dst_port,
4396 					       conf.vni, conf.vni,
4397 					       conf.remote_ifindex,
4398 					       NTF_SELF, 0, true, extack);
4399 			if (err) {
4400 				spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4401 				netdev_adjacent_change_abort(dst->remote_dev,
4402 							     lowerdev, dev);
4403 				return err;
4404 			}
4405 		}
4406 		if (!vxlan_addr_any(&dst->remote_ip))
4407 			__vxlan_fdb_delete(vxlan, all_zeros_mac,
4408 					   dst->remote_ip,
4409 					   vxlan->cfg.dst_port,
4410 					   dst->remote_vni,
4411 					   dst->remote_vni,
4412 					   dst->remote_ifindex,
4413 					   true);
4414 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4415 
4416 		/* If vni filtering device, also update fdb entries of
4417 		 * all vnis that were using default remote ip
4418 		 */
4419 		if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4420 			err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4421 							 &conf.remote_ip, extack);
4422 			if (err) {
4423 				netdev_adjacent_change_abort(dst->remote_dev,
4424 							     lowerdev, dev);
4425 				return err;
4426 			}
4427 		}
4428 	}
4429 
4430 	if (conf.age_interval != vxlan->cfg.age_interval)
4431 		mod_timer(&vxlan->age_timer, jiffies);
4432 
4433 	netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4434 	if (lowerdev && lowerdev != dst->remote_dev)
4435 		dst->remote_dev = lowerdev;
4436 	vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4437 	return 0;
4438 }
4439 
4440 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4441 {
4442 	struct vxlan_dev *vxlan = netdev_priv(dev);
4443 	struct vxlan_fdb_flush_desc desc = {
4444 		/* Default entry is deleted at vxlan_uninit. */
4445 		.ignore_default_entry = true,
4446 	};
4447 
4448 	vxlan_flush(vxlan, &desc);
4449 
4450 	list_del(&vxlan->next);
4451 	unregister_netdevice_queue(dev, head);
4452 	if (vxlan->default_dst.remote_dev)
4453 		netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4454 }
4455 
4456 static size_t vxlan_get_size(const struct net_device *dev)
4457 {
4458 	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
4459 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4460 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
4461 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4462 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
4463 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL_INHERIT */
4464 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
4465 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_DF */
4466 		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4467 		nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_LABEL_POLICY */
4468 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
4469 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
4470 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
4471 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
4472 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
4473 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
4474 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
4475 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
4476 		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4477 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4478 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4479 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4480 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4481 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4482 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4483 		/* IFLA_VXLAN_PORT_RANGE */
4484 		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4485 		nla_total_size(0) + /* IFLA_VXLAN_GBP */
4486 		nla_total_size(0) + /* IFLA_VXLAN_GPE */
4487 		nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */
4488 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */
4489 		0;
4490 }
4491 
4492 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4493 {
4494 	const struct vxlan_dev *vxlan = netdev_priv(dev);
4495 	const struct vxlan_rdst *dst = &vxlan->default_dst;
4496 	struct ifla_vxlan_port_range ports = {
4497 		.low =  htons(vxlan->cfg.port_min),
4498 		.high = htons(vxlan->cfg.port_max),
4499 	};
4500 
4501 	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4502 		goto nla_put_failure;
4503 
4504 	if (!vxlan_addr_any(&dst->remote_ip)) {
4505 		if (dst->remote_ip.sa.sa_family == AF_INET) {
4506 			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4507 					    dst->remote_ip.sin.sin_addr.s_addr))
4508 				goto nla_put_failure;
4509 #if IS_ENABLED(CONFIG_IPV6)
4510 		} else {
4511 			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4512 					     &dst->remote_ip.sin6.sin6_addr))
4513 				goto nla_put_failure;
4514 #endif
4515 		}
4516 	}
4517 
4518 	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4519 		goto nla_put_failure;
4520 
4521 	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4522 		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4523 			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4524 					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
4525 				goto nla_put_failure;
4526 #if IS_ENABLED(CONFIG_IPV6)
4527 		} else {
4528 			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4529 					     &vxlan->cfg.saddr.sin6.sin6_addr))
4530 				goto nla_put_failure;
4531 #endif
4532 		}
4533 	}
4534 
4535 	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4536 	    nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4537 		       !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4538 	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4539 	    nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4540 	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4541 	    nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) ||
4542 	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4543 		       !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4544 	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
4545 		       !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4546 	    nla_put_u8(skb, IFLA_VXLAN_RSC,
4547 		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4548 	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4549 		       !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4550 	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4551 		       !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4552 	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4553 		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4554 	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4555 	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4556 	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4557 	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4558 		       !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4559 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4560 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4561 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4562 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4563 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4564 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4565 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4566 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4567 	    nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4568 		       !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4569 		goto nla_put_failure;
4570 
4571 	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4572 		goto nla_put_failure;
4573 
4574 	if (vxlan->cfg.flags & VXLAN_F_GBP &&
4575 	    nla_put_flag(skb, IFLA_VXLAN_GBP))
4576 		goto nla_put_failure;
4577 
4578 	if (vxlan->cfg.flags & VXLAN_F_GPE &&
4579 	    nla_put_flag(skb, IFLA_VXLAN_GPE))
4580 		goto nla_put_failure;
4581 
4582 	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4583 	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4584 		goto nla_put_failure;
4585 
4586 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4587 	    nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4588 		       !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4589 		goto nla_put_failure;
4590 
4591 	return 0;
4592 
4593 nla_put_failure:
4594 	return -EMSGSIZE;
4595 }
4596 
4597 static struct net *vxlan_get_link_net(const struct net_device *dev)
4598 {
4599 	struct vxlan_dev *vxlan = netdev_priv(dev);
4600 
4601 	return READ_ONCE(vxlan->net);
4602 }
4603 
4604 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4605 	.kind		= "vxlan",
4606 	.maxtype	= IFLA_VXLAN_MAX,
4607 	.policy		= vxlan_policy,
4608 	.priv_size	= sizeof(struct vxlan_dev),
4609 	.setup		= vxlan_setup,
4610 	.validate	= vxlan_validate,
4611 	.newlink	= vxlan_newlink,
4612 	.changelink	= vxlan_changelink,
4613 	.dellink	= vxlan_dellink,
4614 	.get_size	= vxlan_get_size,
4615 	.fill_info	= vxlan_fill_info,
4616 	.get_link_net	= vxlan_get_link_net,
4617 };
4618 
4619 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4620 				    u8 name_assign_type,
4621 				    struct vxlan_config *conf)
4622 {
4623 	struct nlattr *tb[IFLA_MAX + 1];
4624 	struct net_device *dev;
4625 	int err;
4626 
4627 	memset(&tb, 0, sizeof(tb));
4628 
4629 	dev = rtnl_create_link(net, name, name_assign_type,
4630 			       &vxlan_link_ops, tb, NULL);
4631 	if (IS_ERR(dev))
4632 		return dev;
4633 
4634 	err = __vxlan_dev_create(net, dev, conf, NULL);
4635 	if (err < 0) {
4636 		free_netdev(dev);
4637 		return ERR_PTR(err);
4638 	}
4639 
4640 	err = rtnl_configure_link(dev, NULL, 0, NULL);
4641 	if (err < 0) {
4642 		LIST_HEAD(list_kill);
4643 
4644 		vxlan_dellink(dev, &list_kill);
4645 		unregister_netdevice_many(&list_kill);
4646 		return ERR_PTR(err);
4647 	}
4648 
4649 	return dev;
4650 }
4651 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4652 
4653 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4654 					     struct net_device *dev)
4655 {
4656 	struct vxlan_dev *vxlan, *next;
4657 	LIST_HEAD(list_kill);
4658 
4659 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4660 		struct vxlan_rdst *dst = &vxlan->default_dst;
4661 
4662 		/* In case we created vxlan device with carrier
4663 		 * and we loose the carrier due to module unload
4664 		 * we also need to remove vxlan device. In other
4665 		 * cases, it's not necessary and remote_ifindex
4666 		 * is 0 here, so no matches.
4667 		 */
4668 		if (dst->remote_ifindex == dev->ifindex)
4669 			vxlan_dellink(vxlan->dev, &list_kill);
4670 	}
4671 
4672 	unregister_netdevice_many(&list_kill);
4673 }
4674 
4675 static int vxlan_netdevice_event(struct notifier_block *unused,
4676 				 unsigned long event, void *ptr)
4677 {
4678 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4679 	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4680 
4681 	if (event == NETDEV_UNREGISTER)
4682 		vxlan_handle_lowerdev_unregister(vn, dev);
4683 	else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4684 		vxlan_offload_rx_ports(dev, true);
4685 	else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4686 		vxlan_offload_rx_ports(dev, false);
4687 
4688 	return NOTIFY_DONE;
4689 }
4690 
4691 static struct notifier_block vxlan_notifier_block __read_mostly = {
4692 	.notifier_call = vxlan_netdevice_event,
4693 };
4694 
4695 static void
4696 vxlan_fdb_offloaded_set(struct net_device *dev,
4697 			struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4698 {
4699 	struct vxlan_dev *vxlan = netdev_priv(dev);
4700 	struct vxlan_rdst *rdst;
4701 	struct vxlan_fdb *f;
4702 	u32 hash_index;
4703 
4704 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4705 
4706 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4707 
4708 	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4709 	if (!f)
4710 		goto out;
4711 
4712 	rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4713 				   fdb_info->remote_port,
4714 				   fdb_info->remote_vni,
4715 				   fdb_info->remote_ifindex);
4716 	if (!rdst)
4717 		goto out;
4718 
4719 	rdst->offloaded = fdb_info->offloaded;
4720 
4721 out:
4722 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4723 }
4724 
4725 static int
4726 vxlan_fdb_external_learn_add(struct net_device *dev,
4727 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4728 {
4729 	struct vxlan_dev *vxlan = netdev_priv(dev);
4730 	struct netlink_ext_ack *extack;
4731 	u32 hash_index;
4732 	int err;
4733 
4734 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4735 	extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4736 
4737 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4738 	err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4739 			       NUD_REACHABLE,
4740 			       NLM_F_CREATE | NLM_F_REPLACE,
4741 			       fdb_info->remote_port,
4742 			       fdb_info->vni,
4743 			       fdb_info->remote_vni,
4744 			       fdb_info->remote_ifindex,
4745 			       NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4746 			       0, false, extack);
4747 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4748 
4749 	return err;
4750 }
4751 
4752 static int
4753 vxlan_fdb_external_learn_del(struct net_device *dev,
4754 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4755 {
4756 	struct vxlan_dev *vxlan = netdev_priv(dev);
4757 	struct vxlan_fdb *f;
4758 	u32 hash_index;
4759 	int err = 0;
4760 
4761 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4762 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4763 
4764 	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4765 	if (!f)
4766 		err = -ENOENT;
4767 	else if (f->flags & NTF_EXT_LEARNED)
4768 		err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4769 					 fdb_info->remote_ip,
4770 					 fdb_info->remote_port,
4771 					 fdb_info->vni,
4772 					 fdb_info->remote_vni,
4773 					 fdb_info->remote_ifindex,
4774 					 false);
4775 
4776 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4777 
4778 	return err;
4779 }
4780 
4781 static int vxlan_switchdev_event(struct notifier_block *unused,
4782 				 unsigned long event, void *ptr)
4783 {
4784 	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4785 	struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4786 	int err = 0;
4787 
4788 	switch (event) {
4789 	case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4790 		vxlan_fdb_offloaded_set(dev, ptr);
4791 		break;
4792 	case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4793 		fdb_info = ptr;
4794 		err = vxlan_fdb_external_learn_add(dev, fdb_info);
4795 		if (err) {
4796 			err = notifier_from_errno(err);
4797 			break;
4798 		}
4799 		fdb_info->offloaded = true;
4800 		vxlan_fdb_offloaded_set(dev, fdb_info);
4801 		break;
4802 	case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4803 		fdb_info = ptr;
4804 		err = vxlan_fdb_external_learn_del(dev, fdb_info);
4805 		if (err) {
4806 			err = notifier_from_errno(err);
4807 			break;
4808 		}
4809 		fdb_info->offloaded = false;
4810 		vxlan_fdb_offloaded_set(dev, fdb_info);
4811 		break;
4812 	}
4813 
4814 	return err;
4815 }
4816 
4817 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4818 	.notifier_call = vxlan_switchdev_event,
4819 };
4820 
4821 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4822 {
4823 	struct vxlan_fdb *fdb;
4824 	struct vxlan_dev *vxlan;
4825 	u32 hash_index;
4826 
4827 	rcu_read_lock();
4828 	list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4829 		vxlan = rcu_dereference(fdb->vdev);
4830 		WARN_ON(!vxlan);
4831 		hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4832 					    vxlan->default_dst.remote_vni);
4833 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4834 		if (!hlist_unhashed(&fdb->hlist))
4835 			vxlan_fdb_destroy(vxlan, fdb, false, false);
4836 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4837 	}
4838 	rcu_read_unlock();
4839 }
4840 
4841 static int vxlan_nexthop_event(struct notifier_block *nb,
4842 			       unsigned long event, void *ptr)
4843 {
4844 	struct nh_notifier_info *info = ptr;
4845 	struct nexthop *nh;
4846 
4847 	if (event != NEXTHOP_EVENT_DEL)
4848 		return NOTIFY_DONE;
4849 
4850 	nh = nexthop_find_by_id(info->net, info->id);
4851 	if (!nh)
4852 		return NOTIFY_DONE;
4853 
4854 	vxlan_fdb_nh_flush(nh);
4855 
4856 	return NOTIFY_DONE;
4857 }
4858 
4859 static __net_init int vxlan_init_net(struct net *net)
4860 {
4861 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4862 	unsigned int h;
4863 
4864 	INIT_LIST_HEAD(&vn->vxlan_list);
4865 	spin_lock_init(&vn->sock_lock);
4866 	vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4867 
4868 	for (h = 0; h < PORT_HASH_SIZE; ++h)
4869 		INIT_HLIST_HEAD(&vn->sock_list[h]);
4870 
4871 	return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4872 					 NULL);
4873 }
4874 
4875 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn,
4876 					     struct list_head *dev_to_kill)
4877 {
4878 	struct vxlan_dev *vxlan, *next;
4879 
4880 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next)
4881 		vxlan_dellink(vxlan->dev, dev_to_kill);
4882 }
4883 
4884 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list,
4885 					     struct list_head *dev_to_kill)
4886 {
4887 	struct net *net;
4888 
4889 	ASSERT_RTNL();
4890 	list_for_each_entry(net, net_list, exit_list) {
4891 		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4892 
4893 		__unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4894 
4895 		vxlan_destroy_tunnels(vn, dev_to_kill);
4896 	}
4897 }
4898 
4899 static void __net_exit vxlan_exit_net(struct net *net)
4900 {
4901 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4902 	unsigned int h;
4903 
4904 	for (h = 0; h < PORT_HASH_SIZE; ++h)
4905 		WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4906 }
4907 
4908 static struct pernet_operations vxlan_net_ops = {
4909 	.init = vxlan_init_net,
4910 	.exit_batch_rtnl = vxlan_exit_batch_rtnl,
4911 	.exit = vxlan_exit_net,
4912 	.id   = &vxlan_net_id,
4913 	.size = sizeof(struct vxlan_net),
4914 };
4915 
4916 static int __init vxlan_init_module(void)
4917 {
4918 	int rc;
4919 
4920 	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4921 
4922 	rc = register_pernet_subsys(&vxlan_net_ops);
4923 	if (rc)
4924 		goto out1;
4925 
4926 	rc = register_netdevice_notifier(&vxlan_notifier_block);
4927 	if (rc)
4928 		goto out2;
4929 
4930 	rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4931 	if (rc)
4932 		goto out3;
4933 
4934 	rc = rtnl_link_register(&vxlan_link_ops);
4935 	if (rc)
4936 		goto out4;
4937 
4938 	rc = vxlan_vnifilter_init();
4939 	if (rc)
4940 		goto out5;
4941 
4942 	return 0;
4943 out5:
4944 	rtnl_link_unregister(&vxlan_link_ops);
4945 out4:
4946 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4947 out3:
4948 	unregister_netdevice_notifier(&vxlan_notifier_block);
4949 out2:
4950 	unregister_pernet_subsys(&vxlan_net_ops);
4951 out1:
4952 	return rc;
4953 }
4954 late_initcall(vxlan_init_module);
4955 
4956 static void __exit vxlan_cleanup_module(void)
4957 {
4958 	vxlan_vnifilter_uninit();
4959 	rtnl_link_unregister(&vxlan_link_ops);
4960 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4961 	unregister_netdevice_notifier(&vxlan_notifier_block);
4962 	unregister_pernet_subsys(&vxlan_net_ops);
4963 	/* rcu_barrier() is called by netns */
4964 }
4965 module_exit(vxlan_cleanup_module);
4966 
4967 MODULE_LICENSE("GPL");
4968 MODULE_VERSION(VXLAN_VERSION);
4969 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4970 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4971 MODULE_ALIAS_RTNL_LINK("vxlan");
4972