xref: /linux/drivers/net/vxlan/vxlan_core.c (revision 1a9239bb4253f9076b5b4b2a1a4e8d7defd77a95)
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/netdev_lock.h>
29 #include <net/tun_proto.h>
30 #include <net/vxlan.h>
31 #include <net/nexthop.h>
32 
33 #if IS_ENABLED(CONFIG_IPV6)
34 #include <net/ip6_tunnel.h>
35 #include <net/ip6_checksum.h>
36 #endif
37 
38 #include "vxlan_private.h"
39 
40 #define VXLAN_VERSION	"0.1"
41 
42 #define FDB_AGE_DEFAULT 300 /* 5 min */
43 #define FDB_AGE_INTERVAL (10 * HZ)	/* rescan interval */
44 
45 /* UDP port for VXLAN traffic.
46  * The IANA assigned port is 4789, but the Linux default is 8472
47  * for compatibility with early adopters.
48  */
49 static unsigned short vxlan_port __read_mostly = 8472;
50 module_param_named(udp_port, vxlan_port, ushort, 0444);
51 MODULE_PARM_DESC(udp_port, "Destination UDP port");
52 
53 static bool log_ecn_error = true;
54 module_param(log_ecn_error, bool, 0644);
55 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
56 
57 unsigned int vxlan_net_id;
58 
59 const u8 all_zeros_mac[ETH_ALEN + 2];
60 static struct rtnl_link_ops vxlan_link_ops;
61 
62 static int vxlan_sock_add(struct vxlan_dev *vxlan);
63 
64 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
65 
66 /* salt for hash table */
67 static u32 vxlan_salt __read_mostly;
68 
vxlan_collect_metadata(struct vxlan_sock * vs)69 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
70 {
71 	return vs->flags & VXLAN_F_COLLECT_METADATA ||
72 	       ip_tunnel_collect_metadata();
73 }
74 
75 /* Find VXLAN socket based on network namespace, address family, UDP port,
76  * enabled unshareable flags and socket device binding (see l3mdev with
77  * non-default VRF).
78  */
vxlan_find_sock(struct net * net,sa_family_t family,__be16 port,u32 flags,int ifindex)79 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
80 					  __be16 port, u32 flags, int ifindex)
81 {
82 	struct vxlan_sock *vs;
83 
84 	flags &= VXLAN_F_RCV_FLAGS;
85 
86 	hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
87 		if (inet_sk(vs->sock->sk)->inet_sport == port &&
88 		    vxlan_get_sk_family(vs) == family &&
89 		    vs->flags == flags &&
90 		    vs->sock->sk->sk_bound_dev_if == ifindex)
91 			return vs;
92 	}
93 	return NULL;
94 }
95 
vxlan_vs_find_vni(struct vxlan_sock * vs,int ifindex,__be32 vni,struct vxlan_vni_node ** vninode)96 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
97 					   int ifindex, __be32 vni,
98 					   struct vxlan_vni_node **vninode)
99 {
100 	struct vxlan_vni_node *vnode;
101 	struct vxlan_dev_node *node;
102 
103 	/* For flow based devices, map all packets to VNI 0 */
104 	if (vs->flags & VXLAN_F_COLLECT_METADATA &&
105 	    !(vs->flags & VXLAN_F_VNIFILTER))
106 		vni = 0;
107 
108 	hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
109 		if (!node->vxlan)
110 			continue;
111 		vnode = NULL;
112 		if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
113 			vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
114 			if (!vnode)
115 				continue;
116 		} else if (node->vxlan->default_dst.remote_vni != vni) {
117 			continue;
118 		}
119 
120 		if (IS_ENABLED(CONFIG_IPV6)) {
121 			const struct vxlan_config *cfg = &node->vxlan->cfg;
122 
123 			if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
124 			    cfg->remote_ifindex != ifindex)
125 				continue;
126 		}
127 
128 		if (vninode)
129 			*vninode = vnode;
130 		return node->vxlan;
131 	}
132 
133 	return NULL;
134 }
135 
136 /* Look up VNI in a per net namespace table */
vxlan_find_vni(struct net * net,int ifindex,__be32 vni,sa_family_t family,__be16 port,u32 flags)137 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
138 					__be32 vni, sa_family_t family,
139 					__be16 port, u32 flags)
140 {
141 	struct vxlan_sock *vs;
142 
143 	vs = vxlan_find_sock(net, family, port, flags, ifindex);
144 	if (!vs)
145 		return NULL;
146 
147 	return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
148 }
149 
150 /* Fill in neighbour message in skbuff. */
vxlan_fdb_info(struct sk_buff * skb,struct vxlan_dev * vxlan,const struct vxlan_fdb * fdb,u32 portid,u32 seq,int type,unsigned int flags,const struct vxlan_rdst * rdst)151 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
152 			  const struct vxlan_fdb *fdb,
153 			  u32 portid, u32 seq, int type, unsigned int flags,
154 			  const struct vxlan_rdst *rdst)
155 {
156 	unsigned long now = jiffies;
157 	struct nda_cacheinfo ci;
158 	bool send_ip, send_eth;
159 	struct nlmsghdr *nlh;
160 	struct nexthop *nh;
161 	struct ndmsg *ndm;
162 	int nh_family;
163 	u32 nh_id;
164 
165 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
166 	if (nlh == NULL)
167 		return -EMSGSIZE;
168 
169 	ndm = nlmsg_data(nlh);
170 	memset(ndm, 0, sizeof(*ndm));
171 
172 	send_eth = send_ip = true;
173 
174 	rcu_read_lock();
175 	nh = rcu_dereference(fdb->nh);
176 	if (nh) {
177 		nh_family = nexthop_get_family(nh);
178 		nh_id = nh->id;
179 	}
180 	rcu_read_unlock();
181 
182 	if (type == RTM_GETNEIGH) {
183 		if (rdst) {
184 			send_ip = !vxlan_addr_any(&rdst->remote_ip);
185 			ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
186 		} else if (nh) {
187 			ndm->ndm_family = nh_family;
188 		}
189 		send_eth = !is_zero_ether_addr(fdb->eth_addr);
190 	} else
191 		ndm->ndm_family	= AF_BRIDGE;
192 	ndm->ndm_state = fdb->state;
193 	ndm->ndm_ifindex = vxlan->dev->ifindex;
194 	ndm->ndm_flags = fdb->flags;
195 	if (rdst && rdst->offloaded)
196 		ndm->ndm_flags |= NTF_OFFLOADED;
197 	ndm->ndm_type = RTN_UNICAST;
198 
199 	if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
200 	    nla_put_s32(skb, NDA_LINK_NETNSID,
201 			peernet2id(dev_net(vxlan->dev), vxlan->net)))
202 		goto nla_put_failure;
203 
204 	if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
205 		goto nla_put_failure;
206 	if (nh) {
207 		if (nla_put_u32(skb, NDA_NH_ID, nh_id))
208 			goto nla_put_failure;
209 	} else if (rdst) {
210 		if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
211 						  &rdst->remote_ip))
212 			goto nla_put_failure;
213 
214 		if (rdst->remote_port &&
215 		    rdst->remote_port != vxlan->cfg.dst_port &&
216 		    nla_put_be16(skb, NDA_PORT, rdst->remote_port))
217 			goto nla_put_failure;
218 		if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
219 		    nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
220 			goto nla_put_failure;
221 		if (rdst->remote_ifindex &&
222 		    nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
223 			goto nla_put_failure;
224 	}
225 
226 	if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
227 	    nla_put_u32(skb, NDA_SRC_VNI,
228 			be32_to_cpu(fdb->vni)))
229 		goto nla_put_failure;
230 
231 	ci.ndm_used	 = jiffies_to_clock_t(now - READ_ONCE(fdb->used));
232 	ci.ndm_confirmed = 0;
233 	ci.ndm_updated	 = jiffies_to_clock_t(now - READ_ONCE(fdb->updated));
234 	ci.ndm_refcnt	 = 0;
235 
236 	if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
237 		goto nla_put_failure;
238 
239 	nlmsg_end(skb, nlh);
240 	return 0;
241 
242 nla_put_failure:
243 	nlmsg_cancel(skb, nlh);
244 	return -EMSGSIZE;
245 }
246 
vxlan_nlmsg_size(void)247 static inline size_t vxlan_nlmsg_size(void)
248 {
249 	return NLMSG_ALIGN(sizeof(struct ndmsg))
250 		+ nla_total_size(ETH_ALEN) /* NDA_LLADDR */
251 		+ nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
252 		+ nla_total_size(sizeof(__be16)) /* NDA_PORT */
253 		+ nla_total_size(sizeof(__be32)) /* NDA_VNI */
254 		+ nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
255 		+ nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
256 		+ nla_total_size(sizeof(struct nda_cacheinfo));
257 }
258 
__vxlan_fdb_notify(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,int type)259 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
260 			       struct vxlan_rdst *rd, int type)
261 {
262 	struct net *net = dev_net(vxlan->dev);
263 	struct sk_buff *skb;
264 	int err = -ENOBUFS;
265 
266 	skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
267 	if (skb == NULL)
268 		goto errout;
269 
270 	err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
271 	if (err < 0) {
272 		/* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
273 		WARN_ON(err == -EMSGSIZE);
274 		kfree_skb(skb);
275 		goto errout;
276 	}
277 
278 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
279 	return;
280 errout:
281 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
282 }
283 
vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev * vxlan,const struct vxlan_fdb * fdb,const struct vxlan_rdst * rd,struct netlink_ext_ack * extack,struct switchdev_notifier_vxlan_fdb_info * fdb_info)284 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
285 			    const struct vxlan_fdb *fdb,
286 			    const struct vxlan_rdst *rd,
287 			    struct netlink_ext_ack *extack,
288 			    struct switchdev_notifier_vxlan_fdb_info *fdb_info)
289 {
290 	fdb_info->info.dev = vxlan->dev;
291 	fdb_info->info.extack = extack;
292 	fdb_info->remote_ip = rd->remote_ip;
293 	fdb_info->remote_port = rd->remote_port;
294 	fdb_info->remote_vni = rd->remote_vni;
295 	fdb_info->remote_ifindex = rd->remote_ifindex;
296 	memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
297 	fdb_info->vni = fdb->vni;
298 	fdb_info->offloaded = rd->offloaded;
299 	fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
300 }
301 
vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,bool adding,struct netlink_ext_ack * extack)302 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
303 					      struct vxlan_fdb *fdb,
304 					      struct vxlan_rdst *rd,
305 					      bool adding,
306 					      struct netlink_ext_ack *extack)
307 {
308 	struct switchdev_notifier_vxlan_fdb_info info;
309 	enum switchdev_notifier_type notifier_type;
310 	int ret;
311 
312 	if (WARN_ON(!rd))
313 		return 0;
314 
315 	notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
316 			       : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
317 	vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
318 	ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
319 				       &info.info, extack);
320 	return notifier_to_errno(ret);
321 }
322 
vxlan_fdb_notify(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,int type,bool swdev_notify,struct netlink_ext_ack * extack)323 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
324 			    struct vxlan_rdst *rd, int type, bool swdev_notify,
325 			    struct netlink_ext_ack *extack)
326 {
327 	int err;
328 
329 	if (swdev_notify && rd) {
330 		switch (type) {
331 		case RTM_NEWNEIGH:
332 			err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
333 								 true, extack);
334 			if (err)
335 				return err;
336 			break;
337 		case RTM_DELNEIGH:
338 			vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
339 							   false, extack);
340 			break;
341 		}
342 	}
343 
344 	__vxlan_fdb_notify(vxlan, fdb, rd, type);
345 	return 0;
346 }
347 
vxlan_ip_miss(struct net_device * dev,union vxlan_addr * ipa)348 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
349 {
350 	struct vxlan_dev *vxlan = netdev_priv(dev);
351 	struct vxlan_fdb f = {
352 		.state = NUD_STALE,
353 	};
354 	struct vxlan_rdst remote = {
355 		.remote_ip = *ipa, /* goes to NDA_DST */
356 		.remote_vni = cpu_to_be32(VXLAN_N_VID),
357 	};
358 
359 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
360 }
361 
vxlan_fdb_miss(struct vxlan_dev * vxlan,const u8 eth_addr[ETH_ALEN])362 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
363 {
364 	struct vxlan_fdb f = {
365 		.state = NUD_STALE,
366 	};
367 	struct vxlan_rdst remote = { };
368 
369 	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
370 
371 	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
372 }
373 
374 /* Hash Ethernet address */
eth_hash(const unsigned char * addr)375 static u32 eth_hash(const unsigned char *addr)
376 {
377 	u64 value = get_unaligned((u64 *)addr);
378 
379 	/* only want 6 bytes */
380 #ifdef __BIG_ENDIAN
381 	value >>= 16;
382 #else
383 	value <<= 16;
384 #endif
385 	return hash_64(value, FDB_HASH_BITS);
386 }
387 
eth_vni_hash(const unsigned char * addr,__be32 vni)388 u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
389 {
390 	/* use 1 byte of OUI and 3 bytes of NIC */
391 	u32 key = get_unaligned((u32 *)(addr + 2));
392 
393 	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
394 }
395 
fdb_head_index(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)396 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
397 {
398 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
399 		return eth_vni_hash(mac, vni);
400 	else
401 		return eth_hash(mac);
402 }
403 
404 /* Hash chain to use given mac address */
vxlan_fdb_head(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)405 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
406 						const u8 *mac, __be32 vni)
407 {
408 	return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
409 }
410 
411 /* Look up Ethernet address in forwarding table */
__vxlan_find_mac(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)412 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
413 					  const u8 *mac, __be32 vni)
414 {
415 	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
416 	struct vxlan_fdb *f;
417 
418 	hlist_for_each_entry_rcu(f, head, hlist) {
419 		if (ether_addr_equal(mac, f->eth_addr)) {
420 			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
421 				if (vni == f->vni)
422 					return f;
423 			} else {
424 				return f;
425 			}
426 		}
427 	}
428 
429 	return NULL;
430 }
431 
vxlan_find_mac(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)432 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
433 					const u8 *mac, __be32 vni)
434 {
435 	struct vxlan_fdb *f;
436 
437 	f = __vxlan_find_mac(vxlan, mac, vni);
438 	if (f) {
439 		unsigned long now = jiffies;
440 
441 		if (READ_ONCE(f->used) != now)
442 			WRITE_ONCE(f->used, now);
443 	}
444 
445 	return f;
446 }
447 
448 /* caller should hold vxlan->hash_lock */
vxlan_fdb_find_rdst(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex)449 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
450 					      union vxlan_addr *ip, __be16 port,
451 					      __be32 vni, __u32 ifindex)
452 {
453 	struct vxlan_rdst *rd;
454 
455 	list_for_each_entry(rd, &f->remotes, list) {
456 		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
457 		    rd->remote_port == port &&
458 		    rd->remote_vni == vni &&
459 		    rd->remote_ifindex == ifindex)
460 			return rd;
461 	}
462 
463 	return NULL;
464 }
465 
vxlan_fdb_find_uc(struct net_device * dev,const u8 * mac,__be32 vni,struct switchdev_notifier_vxlan_fdb_info * fdb_info)466 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
467 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info)
468 {
469 	struct vxlan_dev *vxlan = netdev_priv(dev);
470 	u8 eth_addr[ETH_ALEN + 2] = { 0 };
471 	struct vxlan_rdst *rdst;
472 	struct vxlan_fdb *f;
473 	int rc = 0;
474 
475 	if (is_multicast_ether_addr(mac) ||
476 	    is_zero_ether_addr(mac))
477 		return -EINVAL;
478 
479 	ether_addr_copy(eth_addr, mac);
480 
481 	rcu_read_lock();
482 
483 	f = __vxlan_find_mac(vxlan, eth_addr, vni);
484 	if (!f) {
485 		rc = -ENOENT;
486 		goto out;
487 	}
488 
489 	rdst = first_remote_rcu(f);
490 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
491 
492 out:
493 	rcu_read_unlock();
494 	return rc;
495 }
496 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
497 
vxlan_fdb_notify_one(struct notifier_block * nb,const struct vxlan_dev * vxlan,const struct vxlan_fdb * f,const struct vxlan_rdst * rdst,struct netlink_ext_ack * extack)498 static int vxlan_fdb_notify_one(struct notifier_block *nb,
499 				const struct vxlan_dev *vxlan,
500 				const struct vxlan_fdb *f,
501 				const struct vxlan_rdst *rdst,
502 				struct netlink_ext_ack *extack)
503 {
504 	struct switchdev_notifier_vxlan_fdb_info fdb_info;
505 	int rc;
506 
507 	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
508 	rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
509 			       &fdb_info);
510 	return notifier_to_errno(rc);
511 }
512 
vxlan_fdb_replay(const struct net_device * dev,__be32 vni,struct notifier_block * nb,struct netlink_ext_ack * extack)513 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
514 		     struct notifier_block *nb,
515 		     struct netlink_ext_ack *extack)
516 {
517 	struct vxlan_dev *vxlan;
518 	struct vxlan_rdst *rdst;
519 	struct vxlan_fdb *f;
520 	unsigned int h;
521 	int rc = 0;
522 
523 	if (!netif_is_vxlan(dev))
524 		return -EINVAL;
525 	vxlan = netdev_priv(dev);
526 
527 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
528 		spin_lock_bh(&vxlan->hash_lock[h]);
529 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
530 			if (f->vni == vni) {
531 				list_for_each_entry(rdst, &f->remotes, list) {
532 					rc = vxlan_fdb_notify_one(nb, vxlan,
533 								  f, rdst,
534 								  extack);
535 					if (rc)
536 						goto unlock;
537 				}
538 			}
539 		}
540 		spin_unlock_bh(&vxlan->hash_lock[h]);
541 	}
542 	return 0;
543 
544 unlock:
545 	spin_unlock_bh(&vxlan->hash_lock[h]);
546 	return rc;
547 }
548 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
549 
vxlan_fdb_clear_offload(const struct net_device * dev,__be32 vni)550 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
551 {
552 	struct vxlan_dev *vxlan;
553 	struct vxlan_rdst *rdst;
554 	struct vxlan_fdb *f;
555 	unsigned int h;
556 
557 	if (!netif_is_vxlan(dev))
558 		return;
559 	vxlan = netdev_priv(dev);
560 
561 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
562 		spin_lock_bh(&vxlan->hash_lock[h]);
563 		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
564 			if (f->vni == vni)
565 				list_for_each_entry(rdst, &f->remotes, list)
566 					rdst->offloaded = false;
567 		spin_unlock_bh(&vxlan->hash_lock[h]);
568 	}
569 
570 }
571 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
572 
573 /* Replace destination of unicast mac */
vxlan_fdb_replace(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex,struct vxlan_rdst * oldrd)574 static int vxlan_fdb_replace(struct vxlan_fdb *f,
575 			     union vxlan_addr *ip, __be16 port, __be32 vni,
576 			     __u32 ifindex, struct vxlan_rdst *oldrd)
577 {
578 	struct vxlan_rdst *rd;
579 
580 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
581 	if (rd)
582 		return 0;
583 
584 	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
585 	if (!rd)
586 		return 0;
587 
588 	*oldrd = *rd;
589 	dst_cache_reset(&rd->dst_cache);
590 	rd->remote_ip = *ip;
591 	rd->remote_port = port;
592 	rd->remote_vni = vni;
593 	rd->remote_ifindex = ifindex;
594 	rd->offloaded = false;
595 	return 1;
596 }
597 
598 /* Add/update destinations for multicast */
vxlan_fdb_append(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex,struct vxlan_rdst ** rdp)599 static int vxlan_fdb_append(struct vxlan_fdb *f,
600 			    union vxlan_addr *ip, __be16 port, __be32 vni,
601 			    __u32 ifindex, struct vxlan_rdst **rdp)
602 {
603 	struct vxlan_rdst *rd;
604 
605 	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
606 	if (rd)
607 		return 0;
608 
609 	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
610 	if (rd == NULL)
611 		return -ENOMEM;
612 
613 	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
614 		kfree(rd);
615 		return -ENOMEM;
616 	}
617 
618 	rd->remote_ip = *ip;
619 	rd->remote_port = port;
620 	rd->offloaded = false;
621 	rd->remote_vni = vni;
622 	rd->remote_ifindex = ifindex;
623 
624 	list_add_tail_rcu(&rd->list, &f->remotes);
625 
626 	*rdp = rd;
627 	return 1;
628 }
629 
vxlan_parse_gpe_proto(const struct vxlanhdr * hdr,__be16 * protocol)630 static bool vxlan_parse_gpe_proto(const struct vxlanhdr *hdr, __be16 *protocol)
631 {
632 	const struct vxlanhdr_gpe *gpe = (const struct vxlanhdr_gpe *)hdr;
633 
634 	/* Need to have Next Protocol set for interfaces in GPE mode. */
635 	if (!gpe->np_applied)
636 		return false;
637 	/* "The initial version is 0. If a receiver does not support the
638 	 * version indicated it MUST drop the packet.
639 	 */
640 	if (gpe->version != 0)
641 		return false;
642 	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
643 	 * processing MUST occur." However, we don't implement OAM
644 	 * processing, thus drop the packet.
645 	 */
646 	if (gpe->oam_flag)
647 		return false;
648 
649 	*protocol = tun_p_to_eth_p(gpe->next_protocol);
650 	if (!*protocol)
651 		return false;
652 
653 	return true;
654 }
655 
vxlan_gro_remcsum(struct sk_buff * skb,unsigned int off,struct vxlanhdr * vh,size_t hdrlen,__be32 vni_field,struct gro_remcsum * grc,bool nopartial)656 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
657 					  unsigned int off,
658 					  struct vxlanhdr *vh, size_t hdrlen,
659 					  __be32 vni_field,
660 					  struct gro_remcsum *grc,
661 					  bool nopartial)
662 {
663 	size_t start, offset;
664 
665 	if (skb->remcsum_offload)
666 		return vh;
667 
668 	if (!NAPI_GRO_CB(skb)->csum_valid)
669 		return NULL;
670 
671 	start = vxlan_rco_start(vni_field);
672 	offset = start + vxlan_rco_offset(vni_field);
673 
674 	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
675 				     start, offset, grc, nopartial);
676 
677 	skb->remcsum_offload = 1;
678 
679 	return vh;
680 }
681 
vxlan_gro_prepare_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb,struct gro_remcsum * grc)682 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
683 						  struct list_head *head,
684 						  struct sk_buff *skb,
685 						  struct gro_remcsum *grc)
686 {
687 	struct sk_buff *p;
688 	struct vxlanhdr *vh, *vh2;
689 	unsigned int hlen, off_vx;
690 	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
691 	__be32 flags;
692 
693 	skb_gro_remcsum_init(grc);
694 
695 	off_vx = skb_gro_offset(skb);
696 	hlen = off_vx + sizeof(*vh);
697 	vh = skb_gro_header(skb, hlen, off_vx);
698 	if (unlikely(!vh))
699 		return NULL;
700 
701 	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
702 
703 	flags = vh->vx_flags;
704 
705 	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
706 		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
707 				       vh->vx_vni, grc,
708 				       !!(vs->flags &
709 					  VXLAN_F_REMCSUM_NOPARTIAL));
710 
711 		if (!vh)
712 			return NULL;
713 	}
714 
715 	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
716 
717 	list_for_each_entry(p, head, list) {
718 		if (!NAPI_GRO_CB(p)->same_flow)
719 			continue;
720 
721 		vh2 = (struct vxlanhdr *)(p->data + off_vx);
722 		if (vh->vx_flags != vh2->vx_flags ||
723 		    vh->vx_vni != vh2->vx_vni) {
724 			NAPI_GRO_CB(p)->same_flow = 0;
725 			continue;
726 		}
727 	}
728 
729 	return vh;
730 }
731 
vxlan_gro_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb)732 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
733 					 struct list_head *head,
734 					 struct sk_buff *skb)
735 {
736 	struct sk_buff *pp = NULL;
737 	struct gro_remcsum grc;
738 	int flush = 1;
739 
740 	if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
741 		pp = call_gro_receive(eth_gro_receive, head, skb);
742 		flush = 0;
743 	}
744 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
745 	return pp;
746 }
747 
vxlan_gpe_gro_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb)748 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
749 					     struct list_head *head,
750 					     struct sk_buff *skb)
751 {
752 	const struct packet_offload *ptype;
753 	struct sk_buff *pp = NULL;
754 	struct gro_remcsum grc;
755 	struct vxlanhdr *vh;
756 	__be16 protocol;
757 	int flush = 1;
758 
759 	vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
760 	if (vh) {
761 		if (!vxlan_parse_gpe_proto(vh, &protocol))
762 			goto out;
763 		ptype = gro_find_receive_by_type(protocol);
764 		if (!ptype)
765 			goto out;
766 		pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
767 		flush = 0;
768 	}
769 out:
770 	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
771 	return pp;
772 }
773 
vxlan_gro_complete(struct sock * sk,struct sk_buff * skb,int nhoff)774 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
775 {
776 	/* Sets 'skb->inner_mac_header' since we are always called with
777 	 * 'skb->encapsulation' set.
778 	 */
779 	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
780 }
781 
vxlan_gpe_gro_complete(struct sock * sk,struct sk_buff * skb,int nhoff)782 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
783 {
784 	struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
785 	const struct packet_offload *ptype;
786 	int err = -ENOSYS;
787 	__be16 protocol;
788 
789 	if (!vxlan_parse_gpe_proto(vh, &protocol))
790 		return err;
791 	ptype = gro_find_complete_by_type(protocol);
792 	if (ptype)
793 		err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
794 	return err;
795 }
796 
vxlan_fdb_alloc(struct vxlan_dev * vxlan,const u8 * mac,__u16 state,__be32 src_vni,__u16 ndm_flags)797 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
798 					 __u16 state, __be32 src_vni,
799 					 __u16 ndm_flags)
800 {
801 	struct vxlan_fdb *f;
802 
803 	f = kmalloc(sizeof(*f), GFP_ATOMIC);
804 	if (!f)
805 		return NULL;
806 	f->state = state;
807 	f->flags = ndm_flags;
808 	f->updated = f->used = jiffies;
809 	f->vni = src_vni;
810 	f->nh = NULL;
811 	RCU_INIT_POINTER(f->vdev, vxlan);
812 	INIT_LIST_HEAD(&f->nh_list);
813 	INIT_LIST_HEAD(&f->remotes);
814 	memcpy(f->eth_addr, mac, ETH_ALEN);
815 
816 	return f;
817 }
818 
vxlan_fdb_insert(struct vxlan_dev * vxlan,const u8 * mac,__be32 src_vni,struct vxlan_fdb * f)819 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
820 			     __be32 src_vni, struct vxlan_fdb *f)
821 {
822 	++vxlan->addrcnt;
823 	hlist_add_head_rcu(&f->hlist,
824 			   vxlan_fdb_head(vxlan, mac, src_vni));
825 }
826 
vxlan_fdb_nh_update(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,u32 nhid,struct netlink_ext_ack * extack)827 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
828 			       u32 nhid, struct netlink_ext_ack *extack)
829 {
830 	struct nexthop *old_nh = rtnl_dereference(fdb->nh);
831 	struct nexthop *nh;
832 	int err = -EINVAL;
833 
834 	if (old_nh && old_nh->id == nhid)
835 		return 0;
836 
837 	nh = nexthop_find_by_id(vxlan->net, nhid);
838 	if (!nh) {
839 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
840 		goto err_inval;
841 	}
842 
843 	if (!nexthop_get(nh)) {
844 		NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
845 		nh = NULL;
846 		goto err_inval;
847 	}
848 	if (!nexthop_is_fdb(nh)) {
849 		NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
850 		goto err_inval;
851 	}
852 
853 	if (!nexthop_is_multipath(nh)) {
854 		NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
855 		goto err_inval;
856 	}
857 
858 	/* check nexthop group family */
859 	switch (vxlan->default_dst.remote_ip.sa.sa_family) {
860 	case AF_INET:
861 		if (!nexthop_has_v4(nh)) {
862 			err = -EAFNOSUPPORT;
863 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
864 			goto err_inval;
865 		}
866 		break;
867 	case AF_INET6:
868 		if (nexthop_has_v4(nh)) {
869 			err = -EAFNOSUPPORT;
870 			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
871 			goto err_inval;
872 		}
873 	}
874 
875 	if (old_nh) {
876 		list_del_rcu(&fdb->nh_list);
877 		nexthop_put(old_nh);
878 	}
879 	rcu_assign_pointer(fdb->nh, nh);
880 	list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
881 	return 1;
882 
883 err_inval:
884 	if (nh)
885 		nexthop_put(nh);
886 	return err;
887 }
888 
vxlan_fdb_create(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,struct vxlan_fdb ** fdb,struct netlink_ext_ack * extack)889 int vxlan_fdb_create(struct vxlan_dev *vxlan,
890 		     const u8 *mac, union vxlan_addr *ip,
891 		     __u16 state, __be16 port, __be32 src_vni,
892 		     __be32 vni, __u32 ifindex, __u16 ndm_flags,
893 		     u32 nhid, struct vxlan_fdb **fdb,
894 		     struct netlink_ext_ack *extack)
895 {
896 	struct vxlan_rdst *rd = NULL;
897 	struct vxlan_fdb *f;
898 	int rc;
899 
900 	if (vxlan->cfg.addrmax &&
901 	    vxlan->addrcnt >= vxlan->cfg.addrmax)
902 		return -ENOSPC;
903 
904 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
905 	f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
906 	if (!f)
907 		return -ENOMEM;
908 
909 	if (nhid)
910 		rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
911 	else
912 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
913 	if (rc < 0)
914 		goto errout;
915 
916 	*fdb = f;
917 
918 	return 0;
919 
920 errout:
921 	kfree(f);
922 	return rc;
923 }
924 
__vxlan_fdb_free(struct vxlan_fdb * f)925 static void __vxlan_fdb_free(struct vxlan_fdb *f)
926 {
927 	struct vxlan_rdst *rd, *nd;
928 	struct nexthop *nh;
929 
930 	nh = rcu_dereference_raw(f->nh);
931 	if (nh) {
932 		rcu_assign_pointer(f->nh, NULL);
933 		rcu_assign_pointer(f->vdev, NULL);
934 		nexthop_put(nh);
935 	}
936 
937 	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
938 		dst_cache_destroy(&rd->dst_cache);
939 		kfree(rd);
940 	}
941 	kfree(f);
942 }
943 
vxlan_fdb_free(struct rcu_head * head)944 static void vxlan_fdb_free(struct rcu_head *head)
945 {
946 	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
947 
948 	__vxlan_fdb_free(f);
949 }
950 
vxlan_fdb_destroy(struct vxlan_dev * vxlan,struct vxlan_fdb * f,bool do_notify,bool swdev_notify)951 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
952 			      bool do_notify, bool swdev_notify)
953 {
954 	struct vxlan_rdst *rd;
955 
956 	netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
957 
958 	--vxlan->addrcnt;
959 	if (do_notify) {
960 		if (rcu_access_pointer(f->nh))
961 			vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
962 					 swdev_notify, NULL);
963 		else
964 			list_for_each_entry(rd, &f->remotes, list)
965 				vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
966 						 swdev_notify, NULL);
967 	}
968 
969 	hlist_del_rcu(&f->hlist);
970 	list_del_rcu(&f->nh_list);
971 	call_rcu(&f->rcu, vxlan_fdb_free);
972 }
973 
vxlan_dst_free(struct rcu_head * head)974 static void vxlan_dst_free(struct rcu_head *head)
975 {
976 	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
977 
978 	dst_cache_destroy(&rd->dst_cache);
979 	kfree(rd);
980 }
981 
vxlan_fdb_update_existing(struct vxlan_dev * vxlan,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 vni,__u32 ifindex,__u16 ndm_flags,struct vxlan_fdb * f,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)982 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
983 				     union vxlan_addr *ip,
984 				     __u16 state, __u16 flags,
985 				     __be16 port, __be32 vni,
986 				     __u32 ifindex, __u16 ndm_flags,
987 				     struct vxlan_fdb *f, u32 nhid,
988 				     bool swdev_notify,
989 				     struct netlink_ext_ack *extack)
990 {
991 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
992 	struct vxlan_rdst *rd = NULL;
993 	struct vxlan_rdst oldrd;
994 	int notify = 0;
995 	int rc = 0;
996 	int err;
997 
998 	if (nhid && !rcu_access_pointer(f->nh)) {
999 		NL_SET_ERR_MSG(extack,
1000 			       "Cannot replace an existing non nexthop fdb with a nexthop");
1001 		return -EOPNOTSUPP;
1002 	}
1003 
1004 	if (nhid && (flags & NLM_F_APPEND)) {
1005 		NL_SET_ERR_MSG(extack,
1006 			       "Cannot append to a nexthop fdb");
1007 		return -EOPNOTSUPP;
1008 	}
1009 
1010 	/* Do not allow an externally learned entry to take over an entry added
1011 	 * by the user.
1012 	 */
1013 	if (!(fdb_flags & NTF_EXT_LEARNED) ||
1014 	    !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1015 		if (f->state != state) {
1016 			f->state = state;
1017 			notify = 1;
1018 		}
1019 		if (f->flags != fdb_flags) {
1020 			f->flags = fdb_flags;
1021 			notify = 1;
1022 		}
1023 	}
1024 
1025 	if ((flags & NLM_F_REPLACE)) {
1026 		/* Only change unicasts */
1027 		if (!(is_multicast_ether_addr(f->eth_addr) ||
1028 		      is_zero_ether_addr(f->eth_addr))) {
1029 			if (nhid) {
1030 				rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1031 				if (rc < 0)
1032 					return rc;
1033 			} else {
1034 				rc = vxlan_fdb_replace(f, ip, port, vni,
1035 						       ifindex, &oldrd);
1036 			}
1037 			notify |= rc;
1038 		} else {
1039 			NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1040 			return -EOPNOTSUPP;
1041 		}
1042 	}
1043 	if ((flags & NLM_F_APPEND) &&
1044 	    (is_multicast_ether_addr(f->eth_addr) ||
1045 	     is_zero_ether_addr(f->eth_addr))) {
1046 		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1047 
1048 		if (rc < 0)
1049 			return rc;
1050 		notify |= rc;
1051 	}
1052 
1053 	if (ndm_flags & NTF_USE)
1054 		WRITE_ONCE(f->updated, jiffies);
1055 
1056 	if (notify) {
1057 		if (rd == NULL)
1058 			rd = first_remote_rtnl(f);
1059 
1060 		WRITE_ONCE(f->updated, jiffies);
1061 		err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1062 				       swdev_notify, extack);
1063 		if (err)
1064 			goto err_notify;
1065 	}
1066 
1067 	return 0;
1068 
1069 err_notify:
1070 	if (nhid)
1071 		return err;
1072 	if ((flags & NLM_F_REPLACE) && rc)
1073 		*rd = oldrd;
1074 	else if ((flags & NLM_F_APPEND) && rc) {
1075 		list_del_rcu(&rd->list);
1076 		call_rcu(&rd->rcu, vxlan_dst_free);
1077 	}
1078 	return err;
1079 }
1080 
vxlan_fdb_update_create(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)1081 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1082 				   const u8 *mac, union vxlan_addr *ip,
1083 				   __u16 state, __u16 flags,
1084 				   __be16 port, __be32 src_vni, __be32 vni,
1085 				   __u32 ifindex, __u16 ndm_flags, u32 nhid,
1086 				   bool swdev_notify,
1087 				   struct netlink_ext_ack *extack)
1088 {
1089 	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1090 	struct vxlan_fdb *f;
1091 	int rc;
1092 
1093 	/* Disallow replace to add a multicast entry */
1094 	if ((flags & NLM_F_REPLACE) &&
1095 	    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1096 		return -EOPNOTSUPP;
1097 
1098 	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1099 	rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1100 			      vni, ifindex, fdb_flags, nhid, &f, extack);
1101 	if (rc < 0)
1102 		return rc;
1103 
1104 	vxlan_fdb_insert(vxlan, mac, src_vni, f);
1105 	rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1106 			      swdev_notify, extack);
1107 	if (rc)
1108 		goto err_notify;
1109 
1110 	return 0;
1111 
1112 err_notify:
1113 	vxlan_fdb_destroy(vxlan, f, false, false);
1114 	return rc;
1115 }
1116 
1117 /* Add new entry to forwarding table -- assumes lock held */
vxlan_fdb_update(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)1118 int vxlan_fdb_update(struct vxlan_dev *vxlan,
1119 		     const u8 *mac, union vxlan_addr *ip,
1120 		     __u16 state, __u16 flags,
1121 		     __be16 port, __be32 src_vni, __be32 vni,
1122 		     __u32 ifindex, __u16 ndm_flags, u32 nhid,
1123 		     bool swdev_notify,
1124 		     struct netlink_ext_ack *extack)
1125 {
1126 	struct vxlan_fdb *f;
1127 
1128 	f = __vxlan_find_mac(vxlan, mac, src_vni);
1129 	if (f) {
1130 		if (flags & NLM_F_EXCL) {
1131 			netdev_dbg(vxlan->dev,
1132 				   "lost race to create %pM\n", mac);
1133 			return -EEXIST;
1134 		}
1135 
1136 		return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1137 						 vni, ifindex, ndm_flags, f,
1138 						 nhid, swdev_notify, extack);
1139 	} else {
1140 		if (!(flags & NLM_F_CREATE))
1141 			return -ENOENT;
1142 
1143 		return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1144 					       port, src_vni, vni, ifindex,
1145 					       ndm_flags, nhid, swdev_notify,
1146 					       extack);
1147 	}
1148 }
1149 
vxlan_fdb_dst_destroy(struct vxlan_dev * vxlan,struct vxlan_fdb * f,struct vxlan_rdst * rd,bool swdev_notify)1150 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1151 				  struct vxlan_rdst *rd, bool swdev_notify)
1152 {
1153 	list_del_rcu(&rd->list);
1154 	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1155 	call_rcu(&rd->rcu, vxlan_dst_free);
1156 }
1157 
vxlan_fdb_parse(struct nlattr * tb[],struct vxlan_dev * vxlan,union vxlan_addr * ip,__be16 * port,__be32 * src_vni,__be32 * vni,u32 * ifindex,u32 * nhid,struct netlink_ext_ack * extack)1158 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1159 			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1160 			   __be32 *vni, u32 *ifindex, u32 *nhid,
1161 			   struct netlink_ext_ack *extack)
1162 {
1163 	struct net *net = dev_net(vxlan->dev);
1164 	int err;
1165 
1166 	if (tb[NDA_NH_ID] &&
1167 	    (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1168 		NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1169 		return -EINVAL;
1170 	}
1171 
1172 	if (tb[NDA_DST]) {
1173 		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1174 		if (err) {
1175 			NL_SET_ERR_MSG(extack, "Unsupported address family");
1176 			return err;
1177 		}
1178 	} else {
1179 		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1180 
1181 		if (remote->sa.sa_family == AF_INET) {
1182 			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1183 			ip->sa.sa_family = AF_INET;
1184 #if IS_ENABLED(CONFIG_IPV6)
1185 		} else {
1186 			ip->sin6.sin6_addr = in6addr_any;
1187 			ip->sa.sa_family = AF_INET6;
1188 #endif
1189 		}
1190 	}
1191 
1192 	if (tb[NDA_PORT]) {
1193 		if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1194 			NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1195 			return -EINVAL;
1196 		}
1197 		*port = nla_get_be16(tb[NDA_PORT]);
1198 	} else {
1199 		*port = vxlan->cfg.dst_port;
1200 	}
1201 
1202 	if (tb[NDA_VNI]) {
1203 		if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1204 			NL_SET_ERR_MSG(extack, "Invalid vni");
1205 			return -EINVAL;
1206 		}
1207 		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1208 	} else {
1209 		*vni = vxlan->default_dst.remote_vni;
1210 	}
1211 
1212 	if (tb[NDA_SRC_VNI]) {
1213 		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1214 			NL_SET_ERR_MSG(extack, "Invalid src vni");
1215 			return -EINVAL;
1216 		}
1217 		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1218 	} else {
1219 		*src_vni = vxlan->default_dst.remote_vni;
1220 	}
1221 
1222 	if (tb[NDA_IFINDEX]) {
1223 		struct net_device *tdev;
1224 
1225 		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1226 			NL_SET_ERR_MSG(extack, "Invalid ifindex");
1227 			return -EINVAL;
1228 		}
1229 		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1230 		tdev = __dev_get_by_index(net, *ifindex);
1231 		if (!tdev) {
1232 			NL_SET_ERR_MSG(extack, "Device not found");
1233 			return -EADDRNOTAVAIL;
1234 		}
1235 	} else {
1236 		*ifindex = 0;
1237 	}
1238 
1239 	*nhid = nla_get_u32_default(tb[NDA_NH_ID], 0);
1240 
1241 	return 0;
1242 }
1243 
1244 /* Add static entry (via netlink) */
vxlan_fdb_add(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,u16 flags,bool * notified,struct netlink_ext_ack * extack)1245 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1246 			 struct net_device *dev,
1247 			 const unsigned char *addr, u16 vid, u16 flags,
1248 			 bool *notified, struct netlink_ext_ack *extack)
1249 {
1250 	struct vxlan_dev *vxlan = netdev_priv(dev);
1251 	/* struct net *net = dev_net(vxlan->dev); */
1252 	union vxlan_addr ip;
1253 	__be16 port;
1254 	__be32 src_vni, vni;
1255 	u32 ifindex, nhid;
1256 	u32 hash_index;
1257 	int err;
1258 
1259 	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1260 		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1261 			ndm->ndm_state);
1262 		return -EINVAL;
1263 	}
1264 
1265 	if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1266 		return -EINVAL;
1267 
1268 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1269 			      &nhid, extack);
1270 	if (err)
1271 		return err;
1272 
1273 	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1274 		return -EAFNOSUPPORT;
1275 
1276 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1277 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1278 	err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1279 			       port, src_vni, vni, ifindex,
1280 			       ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1281 			       nhid, true, extack);
1282 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1283 
1284 	if (!err)
1285 		*notified = true;
1286 
1287 	return err;
1288 }
1289 
__vxlan_fdb_delete(struct vxlan_dev * vxlan,const unsigned char * addr,union vxlan_addr ip,__be16 port,__be32 src_vni,__be32 vni,u32 ifindex,bool swdev_notify)1290 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1291 		       const unsigned char *addr, union vxlan_addr ip,
1292 		       __be16 port, __be32 src_vni, __be32 vni,
1293 		       u32 ifindex, bool swdev_notify)
1294 {
1295 	struct vxlan_rdst *rd = NULL;
1296 	struct vxlan_fdb *f;
1297 	int err = -ENOENT;
1298 
1299 	f = __vxlan_find_mac(vxlan, addr, src_vni);
1300 	if (!f)
1301 		return err;
1302 
1303 	if (!vxlan_addr_any(&ip)) {
1304 		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1305 		if (!rd)
1306 			goto out;
1307 	}
1308 
1309 	/* remove a destination if it's not the only one on the list,
1310 	 * otherwise destroy the fdb entry
1311 	 */
1312 	if (rd && !list_is_singular(&f->remotes)) {
1313 		vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1314 		goto out;
1315 	}
1316 
1317 	vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1318 
1319 out:
1320 	return 0;
1321 }
1322 
1323 /* Delete entry (via netlink) */
vxlan_fdb_delete(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,bool * notified,struct netlink_ext_ack * extack)1324 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1325 			    struct net_device *dev,
1326 			    const unsigned char *addr, u16 vid, bool *notified,
1327 			    struct netlink_ext_ack *extack)
1328 {
1329 	struct vxlan_dev *vxlan = netdev_priv(dev);
1330 	union vxlan_addr ip;
1331 	__be32 src_vni, vni;
1332 	u32 ifindex, nhid;
1333 	u32 hash_index;
1334 	__be16 port;
1335 	int err;
1336 
1337 	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1338 			      &nhid, extack);
1339 	if (err)
1340 		return err;
1341 
1342 	hash_index = fdb_head_index(vxlan, addr, src_vni);
1343 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1344 	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1345 				 true);
1346 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1347 
1348 	if (!err)
1349 		*notified = true;
1350 
1351 	return err;
1352 }
1353 
1354 /* Dump forwarding table */
vxlan_fdb_dump(struct sk_buff * skb,struct netlink_callback * cb,struct net_device * dev,struct net_device * filter_dev,int * idx)1355 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1356 			  struct net_device *dev,
1357 			  struct net_device *filter_dev, int *idx)
1358 {
1359 	struct ndo_fdb_dump_context *ctx = (void *)cb->ctx;
1360 	struct vxlan_dev *vxlan = netdev_priv(dev);
1361 	unsigned int h;
1362 	int err = 0;
1363 
1364 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
1365 		struct vxlan_fdb *f;
1366 
1367 		rcu_read_lock();
1368 		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1369 			struct vxlan_rdst *rd;
1370 
1371 			if (rcu_access_pointer(f->nh)) {
1372 				if (*idx < ctx->fdb_idx)
1373 					goto skip_nh;
1374 				err = vxlan_fdb_info(skb, vxlan, f,
1375 						     NETLINK_CB(cb->skb).portid,
1376 						     cb->nlh->nlmsg_seq,
1377 						     RTM_NEWNEIGH,
1378 						     NLM_F_MULTI, NULL);
1379 				if (err < 0) {
1380 					rcu_read_unlock();
1381 					goto out;
1382 				}
1383 skip_nh:
1384 				*idx += 1;
1385 				continue;
1386 			}
1387 
1388 			list_for_each_entry_rcu(rd, &f->remotes, list) {
1389 				if (*idx < ctx->fdb_idx)
1390 					goto skip;
1391 
1392 				err = vxlan_fdb_info(skb, vxlan, f,
1393 						     NETLINK_CB(cb->skb).portid,
1394 						     cb->nlh->nlmsg_seq,
1395 						     RTM_NEWNEIGH,
1396 						     NLM_F_MULTI, rd);
1397 				if (err < 0) {
1398 					rcu_read_unlock();
1399 					goto out;
1400 				}
1401 skip:
1402 				*idx += 1;
1403 			}
1404 		}
1405 		rcu_read_unlock();
1406 	}
1407 out:
1408 	return err;
1409 }
1410 
vxlan_fdb_get(struct sk_buff * skb,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,u32 portid,u32 seq,struct netlink_ext_ack * extack)1411 static int vxlan_fdb_get(struct sk_buff *skb,
1412 			 struct nlattr *tb[],
1413 			 struct net_device *dev,
1414 			 const unsigned char *addr,
1415 			 u16 vid, u32 portid, u32 seq,
1416 			 struct netlink_ext_ack *extack)
1417 {
1418 	struct vxlan_dev *vxlan = netdev_priv(dev);
1419 	struct vxlan_fdb *f;
1420 	__be32 vni;
1421 	int err;
1422 
1423 	if (tb[NDA_VNI])
1424 		vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1425 	else
1426 		vni = vxlan->default_dst.remote_vni;
1427 
1428 	rcu_read_lock();
1429 
1430 	f = __vxlan_find_mac(vxlan, addr, vni);
1431 	if (!f) {
1432 		NL_SET_ERR_MSG(extack, "Fdb entry not found");
1433 		err = -ENOENT;
1434 		goto errout;
1435 	}
1436 
1437 	err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1438 			     RTM_NEWNEIGH, 0, first_remote_rcu(f));
1439 errout:
1440 	rcu_read_unlock();
1441 	return err;
1442 }
1443 
1444 /* Watch incoming packets to learn mapping between Ethernet address
1445  * and Tunnel endpoint.
1446  */
vxlan_snoop(struct net_device * dev,union vxlan_addr * src_ip,const u8 * src_mac,u32 src_ifindex,__be32 vni)1447 static enum skb_drop_reason vxlan_snoop(struct net_device *dev,
1448 					union vxlan_addr *src_ip,
1449 					const u8 *src_mac, u32 src_ifindex,
1450 					__be32 vni)
1451 {
1452 	struct vxlan_dev *vxlan = netdev_priv(dev);
1453 	struct vxlan_fdb *f;
1454 	u32 ifindex = 0;
1455 
1456 	/* Ignore packets from invalid src-address */
1457 	if (!is_valid_ether_addr(src_mac))
1458 		return SKB_DROP_REASON_MAC_INVALID_SOURCE;
1459 
1460 #if IS_ENABLED(CONFIG_IPV6)
1461 	if (src_ip->sa.sa_family == AF_INET6 &&
1462 	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1463 		ifindex = src_ifindex;
1464 #endif
1465 
1466 	f = __vxlan_find_mac(vxlan, src_mac, vni);
1467 	if (likely(f)) {
1468 		struct vxlan_rdst *rdst = first_remote_rcu(f);
1469 		unsigned long now = jiffies;
1470 
1471 		if (READ_ONCE(f->updated) != now)
1472 			WRITE_ONCE(f->updated, now);
1473 
1474 		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1475 			   rdst->remote_ifindex == ifindex))
1476 			return SKB_NOT_DROPPED_YET;
1477 
1478 		/* Don't migrate static entries, drop packets */
1479 		if (f->state & (NUD_PERMANENT | NUD_NOARP))
1480 			return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1481 
1482 		/* Don't override an fdb with nexthop with a learnt entry */
1483 		if (rcu_access_pointer(f->nh))
1484 			return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1485 
1486 		if (net_ratelimit())
1487 			netdev_info(dev,
1488 				    "%pM migrated from %pIS to %pIS\n",
1489 				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1490 
1491 		rdst->remote_ip = *src_ip;
1492 		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1493 	} else {
1494 		u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1495 
1496 		/* learned new entry */
1497 		spin_lock(&vxlan->hash_lock[hash_index]);
1498 
1499 		/* close off race between vxlan_flush and incoming packets */
1500 		if (netif_running(dev))
1501 			vxlan_fdb_update(vxlan, src_mac, src_ip,
1502 					 NUD_REACHABLE,
1503 					 NLM_F_EXCL|NLM_F_CREATE,
1504 					 vxlan->cfg.dst_port,
1505 					 vni,
1506 					 vxlan->default_dst.remote_vni,
1507 					 ifindex, NTF_SELF, 0, true, NULL);
1508 		spin_unlock(&vxlan->hash_lock[hash_index]);
1509 	}
1510 
1511 	return SKB_NOT_DROPPED_YET;
1512 }
1513 
__vxlan_sock_release_prep(struct vxlan_sock * vs)1514 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1515 {
1516 	struct vxlan_net *vn;
1517 
1518 	if (!vs)
1519 		return false;
1520 	if (!refcount_dec_and_test(&vs->refcnt))
1521 		return false;
1522 
1523 	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1524 	spin_lock(&vn->sock_lock);
1525 	hlist_del_rcu(&vs->hlist);
1526 	udp_tunnel_notify_del_rx_port(vs->sock,
1527 				      (vs->flags & VXLAN_F_GPE) ?
1528 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1529 				      UDP_TUNNEL_TYPE_VXLAN);
1530 	spin_unlock(&vn->sock_lock);
1531 
1532 	return true;
1533 }
1534 
vxlan_sock_release(struct vxlan_dev * vxlan)1535 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1536 {
1537 	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1538 #if IS_ENABLED(CONFIG_IPV6)
1539 	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1540 
1541 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1542 #endif
1543 
1544 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1545 	synchronize_net();
1546 
1547 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1548 		vxlan_vs_del_vnigrp(vxlan);
1549 	else
1550 		vxlan_vs_del_dev(vxlan);
1551 
1552 	if (__vxlan_sock_release_prep(sock4)) {
1553 		udp_tunnel_sock_release(sock4->sock);
1554 		kfree(sock4);
1555 	}
1556 
1557 #if IS_ENABLED(CONFIG_IPV6)
1558 	if (__vxlan_sock_release_prep(sock6)) {
1559 		udp_tunnel_sock_release(sock6->sock);
1560 		kfree(sock6);
1561 	}
1562 #endif
1563 }
1564 
vxlan_remcsum(struct sk_buff * skb,u32 vxflags)1565 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags)
1566 {
1567 	const struct vxlanhdr *vh = vxlan_hdr(skb);
1568 	enum skb_drop_reason reason;
1569 	size_t start, offset;
1570 
1571 	if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1572 		return SKB_NOT_DROPPED_YET;
1573 
1574 	start = vxlan_rco_start(vh->vx_vni);
1575 	offset = start + vxlan_rco_offset(vh->vx_vni);
1576 
1577 	reason = pskb_may_pull_reason(skb, offset + sizeof(u16));
1578 	if (reason)
1579 		return reason;
1580 
1581 	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1582 			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1583 	return SKB_NOT_DROPPED_YET;
1584 }
1585 
vxlan_parse_gbp_hdr(struct sk_buff * skb,u32 vxflags,struct vxlan_metadata * md)1586 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags,
1587 				struct vxlan_metadata *md)
1588 {
1589 	const struct vxlanhdr *vh = vxlan_hdr(skb);
1590 	const struct vxlanhdr_gbp *gbp;
1591 	struct metadata_dst *tun_dst;
1592 
1593 	gbp = (const struct vxlanhdr_gbp *)vh;
1594 
1595 	if (!(vh->vx_flags & VXLAN_HF_GBP))
1596 		return;
1597 
1598 	md->gbp = ntohs(gbp->policy_id);
1599 
1600 	tun_dst = (struct metadata_dst *)skb_dst(skb);
1601 	if (tun_dst) {
1602 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT,
1603 			  tun_dst->u.tun_info.key.tun_flags);
1604 		tun_dst->u.tun_info.options_len = sizeof(*md);
1605 	}
1606 	if (gbp->dont_learn)
1607 		md->gbp |= VXLAN_GBP_DONT_LEARN;
1608 
1609 	if (gbp->policy_applied)
1610 		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1611 
1612 	/* In flow-based mode, GBP is carried in dst_metadata */
1613 	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1614 		skb->mark = md->gbp;
1615 }
1616 
vxlan_set_mac(struct vxlan_dev * vxlan,struct vxlan_sock * vs,struct sk_buff * skb,__be32 vni)1617 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan,
1618 					  struct vxlan_sock *vs,
1619 					  struct sk_buff *skb, __be32 vni)
1620 {
1621 	union vxlan_addr saddr;
1622 	u32 ifindex = skb->dev->ifindex;
1623 
1624 	skb_reset_mac_header(skb);
1625 	skb->protocol = eth_type_trans(skb, vxlan->dev);
1626 	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1627 
1628 	/* Ignore packet loops (and multicast echo) */
1629 	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1630 		return SKB_DROP_REASON_LOCAL_MAC;
1631 
1632 	/* Get address from the outer IP header */
1633 	if (vxlan_get_sk_family(vs) == AF_INET) {
1634 		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1635 		saddr.sa.sa_family = AF_INET;
1636 #if IS_ENABLED(CONFIG_IPV6)
1637 	} else {
1638 		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1639 		saddr.sa.sa_family = AF_INET6;
1640 #endif
1641 	}
1642 
1643 	if (!(vxlan->cfg.flags & VXLAN_F_LEARN))
1644 		return SKB_NOT_DROPPED_YET;
1645 
1646 	return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source,
1647 			   ifindex, vni);
1648 }
1649 
vxlan_ecn_decapsulate(struct vxlan_sock * vs,void * oiph,struct sk_buff * skb)1650 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1651 				  struct sk_buff *skb)
1652 {
1653 	int err = 0;
1654 
1655 	if (vxlan_get_sk_family(vs) == AF_INET)
1656 		err = IP_ECN_decapsulate(oiph, skb);
1657 #if IS_ENABLED(CONFIG_IPV6)
1658 	else
1659 		err = IP6_ECN_decapsulate(oiph, skb);
1660 #endif
1661 
1662 	if (unlikely(err) && log_ecn_error) {
1663 		if (vxlan_get_sk_family(vs) == AF_INET)
1664 			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1665 					     &((struct iphdr *)oiph)->saddr,
1666 					     ((struct iphdr *)oiph)->tos);
1667 		else
1668 			net_info_ratelimited("non-ECT from %pI6\n",
1669 					     &((struct ipv6hdr *)oiph)->saddr);
1670 	}
1671 	return err <= 1;
1672 }
1673 
vxlan_rcv(struct sock * sk,struct sk_buff * skb)1674 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1675 {
1676 	struct vxlan_vni_node *vninode = NULL;
1677 	const struct vxlanhdr *vh;
1678 	struct vxlan_dev *vxlan;
1679 	struct vxlan_sock *vs;
1680 	struct vxlan_metadata _md;
1681 	struct vxlan_metadata *md = &_md;
1682 	__be16 protocol = htons(ETH_P_TEB);
1683 	enum skb_drop_reason reason;
1684 	bool raw_proto = false;
1685 	void *oiph;
1686 	__be32 vni = 0;
1687 	int nh;
1688 
1689 	/* Need UDP and VXLAN header to be present */
1690 	reason = pskb_may_pull_reason(skb, VXLAN_HLEN);
1691 	if (reason)
1692 		goto drop;
1693 
1694 	vh = vxlan_hdr(skb);
1695 	/* VNI flag always required to be set */
1696 	if (!(vh->vx_flags & VXLAN_HF_VNI)) {
1697 		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1698 			   ntohl(vh->vx_flags), ntohl(vh->vx_vni));
1699 		reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1700 		/* Return non vxlan pkt */
1701 		goto drop;
1702 	}
1703 
1704 	vs = rcu_dereference_sk_user_data(sk);
1705 	if (!vs)
1706 		goto drop;
1707 
1708 	vni = vxlan_vni(vh->vx_vni);
1709 
1710 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1711 	if (!vxlan) {
1712 		reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND;
1713 		goto drop;
1714 	}
1715 
1716 	if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags ||
1717 	    vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) {
1718 		/* If the header uses bits besides those enabled by the
1719 		 * netdevice configuration, treat this as a malformed packet.
1720 		 * This behavior diverges from VXLAN RFC (RFC7348) which
1721 		 * stipulates that bits in reserved in reserved fields are to be
1722 		 * ignored. The approach here maintains compatibility with
1723 		 * previous stack code, and also is more robust and provides a
1724 		 * little more security in adding extensions to VXLAN.
1725 		 */
1726 		reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1727 		DEV_STATS_INC(vxlan->dev, rx_frame_errors);
1728 		DEV_STATS_INC(vxlan->dev, rx_errors);
1729 		vxlan_vnifilter_count(vxlan, vni, vninode,
1730 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1731 		goto drop;
1732 	}
1733 
1734 	if (vxlan->cfg.flags & VXLAN_F_GPE) {
1735 		if (!vxlan_parse_gpe_proto(vh, &protocol))
1736 			goto drop;
1737 		raw_proto = true;
1738 	}
1739 
1740 	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1741 				   !net_eq(vxlan->net, dev_net(vxlan->dev)))) {
1742 		reason = SKB_DROP_REASON_NOMEM;
1743 		goto drop;
1744 	}
1745 
1746 	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) {
1747 		reason = vxlan_remcsum(skb, vxlan->cfg.flags);
1748 		if (unlikely(reason))
1749 			goto drop;
1750 	}
1751 
1752 	if (vxlan_collect_metadata(vs)) {
1753 		IP_TUNNEL_DECLARE_FLAGS(flags) = { };
1754 		struct metadata_dst *tun_dst;
1755 
1756 		__set_bit(IP_TUNNEL_KEY_BIT, flags);
1757 		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags,
1758 					 key32_to_tunnel_id(vni), sizeof(*md));
1759 
1760 		if (!tun_dst) {
1761 			reason = SKB_DROP_REASON_NOMEM;
1762 			goto drop;
1763 		}
1764 
1765 		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1766 
1767 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1768 	} else {
1769 		memset(md, 0, sizeof(*md));
1770 	}
1771 
1772 	if (vxlan->cfg.flags & VXLAN_F_GBP)
1773 		vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md);
1774 	/* Note that GBP and GPE can never be active together. This is
1775 	 * ensured in vxlan_dev_configure.
1776 	 */
1777 
1778 	if (!raw_proto) {
1779 		reason = vxlan_set_mac(vxlan, vs, skb, vni);
1780 		if (reason)
1781 			goto drop;
1782 	} else {
1783 		skb_reset_mac_header(skb);
1784 		skb->dev = vxlan->dev;
1785 		skb->pkt_type = PACKET_HOST;
1786 	}
1787 
1788 	/* Save offset of outer header relative to skb->head,
1789 	 * because we are going to reset the network header to the inner header
1790 	 * and might change skb->head.
1791 	 */
1792 	nh = skb_network_header(skb) - skb->head;
1793 
1794 	skb_reset_network_header(skb);
1795 
1796 	reason = pskb_inet_may_pull_reason(skb);
1797 	if (reason) {
1798 		DEV_STATS_INC(vxlan->dev, rx_length_errors);
1799 		DEV_STATS_INC(vxlan->dev, rx_errors);
1800 		vxlan_vnifilter_count(vxlan, vni, vninode,
1801 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1802 		goto drop;
1803 	}
1804 
1805 	/* Get the outer header. */
1806 	oiph = skb->head + nh;
1807 
1808 	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1809 		reason = SKB_DROP_REASON_IP_TUNNEL_ECN;
1810 		DEV_STATS_INC(vxlan->dev, rx_frame_errors);
1811 		DEV_STATS_INC(vxlan->dev, rx_errors);
1812 		vxlan_vnifilter_count(vxlan, vni, vninode,
1813 				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1814 		goto drop;
1815 	}
1816 
1817 	rcu_read_lock();
1818 
1819 	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1820 		rcu_read_unlock();
1821 		dev_dstats_rx_dropped(vxlan->dev);
1822 		vxlan_vnifilter_count(vxlan, vni, vninode,
1823 				      VXLAN_VNI_STATS_RX_DROPS, 0);
1824 		reason = SKB_DROP_REASON_DEV_READY;
1825 		goto drop;
1826 	}
1827 
1828 	dev_dstats_rx_add(vxlan->dev, skb->len);
1829 	vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1830 	gro_cells_receive(&vxlan->gro_cells, skb);
1831 
1832 	rcu_read_unlock();
1833 
1834 	return 0;
1835 
1836 drop:
1837 	reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED;
1838 	/* Consume bad packet */
1839 	kfree_skb_reason(skb, reason);
1840 	return 0;
1841 }
1842 
vxlan_err_lookup(struct sock * sk,struct sk_buff * skb)1843 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1844 {
1845 	struct vxlan_dev *vxlan;
1846 	struct vxlan_sock *vs;
1847 	struct vxlanhdr *hdr;
1848 	__be32 vni;
1849 
1850 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1851 		return -EINVAL;
1852 
1853 	hdr = vxlan_hdr(skb);
1854 
1855 	if (!(hdr->vx_flags & VXLAN_HF_VNI))
1856 		return -EINVAL;
1857 
1858 	vs = rcu_dereference_sk_user_data(sk);
1859 	if (!vs)
1860 		return -ENOENT;
1861 
1862 	vni = vxlan_vni(hdr->vx_vni);
1863 	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1864 	if (!vxlan)
1865 		return -ENOENT;
1866 
1867 	return 0;
1868 }
1869 
arp_reduce(struct net_device * dev,struct sk_buff * skb,__be32 vni)1870 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1871 {
1872 	struct vxlan_dev *vxlan = netdev_priv(dev);
1873 	struct arphdr *parp;
1874 	u8 *arpptr, *sha;
1875 	__be32 sip, tip;
1876 	struct neighbour *n;
1877 
1878 	if (dev->flags & IFF_NOARP)
1879 		goto out;
1880 
1881 	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1882 		dev_dstats_tx_dropped(dev);
1883 		vxlan_vnifilter_count(vxlan, vni, NULL,
1884 				      VXLAN_VNI_STATS_TX_DROPS, 0);
1885 		goto out;
1886 	}
1887 	parp = arp_hdr(skb);
1888 
1889 	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1890 	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1891 	    parp->ar_pro != htons(ETH_P_IP) ||
1892 	    parp->ar_op != htons(ARPOP_REQUEST) ||
1893 	    parp->ar_hln != dev->addr_len ||
1894 	    parp->ar_pln != 4)
1895 		goto out;
1896 	arpptr = (u8 *)parp + sizeof(struct arphdr);
1897 	sha = arpptr;
1898 	arpptr += dev->addr_len;	/* sha */
1899 	memcpy(&sip, arpptr, sizeof(sip));
1900 	arpptr += sizeof(sip);
1901 	arpptr += dev->addr_len;	/* tha */
1902 	memcpy(&tip, arpptr, sizeof(tip));
1903 
1904 	if (ipv4_is_loopback(tip) ||
1905 	    ipv4_is_multicast(tip))
1906 		goto out;
1907 
1908 	n = neigh_lookup(&arp_tbl, &tip, dev);
1909 
1910 	if (n) {
1911 		struct vxlan_fdb *f;
1912 		struct sk_buff	*reply;
1913 
1914 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1915 			neigh_release(n);
1916 			goto out;
1917 		}
1918 
1919 		f = vxlan_find_mac(vxlan, n->ha, vni);
1920 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1921 			/* bridge-local neighbor */
1922 			neigh_release(n);
1923 			goto out;
1924 		}
1925 
1926 		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1927 				n->ha, sha);
1928 
1929 		neigh_release(n);
1930 
1931 		if (reply == NULL)
1932 			goto out;
1933 
1934 		skb_reset_mac_header(reply);
1935 		__skb_pull(reply, skb_network_offset(reply));
1936 		reply->ip_summed = CHECKSUM_UNNECESSARY;
1937 		reply->pkt_type = PACKET_HOST;
1938 
1939 		if (netif_rx(reply) == NET_RX_DROP) {
1940 			dev_dstats_rx_dropped(dev);
1941 			vxlan_vnifilter_count(vxlan, vni, NULL,
1942 					      VXLAN_VNI_STATS_RX_DROPS, 0);
1943 		}
1944 
1945 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1946 		union vxlan_addr ipa = {
1947 			.sin.sin_addr.s_addr = tip,
1948 			.sin.sin_family = AF_INET,
1949 		};
1950 
1951 		vxlan_ip_miss(dev, &ipa);
1952 	}
1953 out:
1954 	consume_skb(skb);
1955 	return NETDEV_TX_OK;
1956 }
1957 
1958 #if IS_ENABLED(CONFIG_IPV6)
vxlan_na_create(struct sk_buff * request,struct neighbour * n,bool isrouter)1959 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1960 	struct neighbour *n, bool isrouter)
1961 {
1962 	struct net_device *dev = request->dev;
1963 	struct sk_buff *reply;
1964 	struct nd_msg *ns, *na;
1965 	struct ipv6hdr *pip6;
1966 	u8 *daddr;
1967 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1968 	int ns_olen;
1969 	int i, len;
1970 
1971 	if (dev == NULL || !pskb_may_pull(request, request->len))
1972 		return NULL;
1973 
1974 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1975 		sizeof(*na) + na_olen + dev->needed_tailroom;
1976 	reply = alloc_skb(len, GFP_ATOMIC);
1977 	if (reply == NULL)
1978 		return NULL;
1979 
1980 	reply->protocol = htons(ETH_P_IPV6);
1981 	reply->dev = dev;
1982 	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1983 	skb_push(reply, sizeof(struct ethhdr));
1984 	skb_reset_mac_header(reply);
1985 
1986 	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1987 
1988 	daddr = eth_hdr(request)->h_source;
1989 	ns_olen = request->len - skb_network_offset(request) -
1990 		sizeof(struct ipv6hdr) - sizeof(*ns);
1991 	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1992 		if (!ns->opt[i + 1]) {
1993 			kfree_skb(reply);
1994 			return NULL;
1995 		}
1996 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1997 			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1998 			break;
1999 		}
2000 	}
2001 
2002 	/* Ethernet header */
2003 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2004 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2005 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2006 	reply->protocol = htons(ETH_P_IPV6);
2007 
2008 	skb_pull(reply, sizeof(struct ethhdr));
2009 	skb_reset_network_header(reply);
2010 	skb_put(reply, sizeof(struct ipv6hdr));
2011 
2012 	/* IPv6 header */
2013 
2014 	pip6 = ipv6_hdr(reply);
2015 	memset(pip6, 0, sizeof(struct ipv6hdr));
2016 	pip6->version = 6;
2017 	pip6->priority = ipv6_hdr(request)->priority;
2018 	pip6->nexthdr = IPPROTO_ICMPV6;
2019 	pip6->hop_limit = 255;
2020 	pip6->daddr = ipv6_hdr(request)->saddr;
2021 	pip6->saddr = *(struct in6_addr *)n->primary_key;
2022 
2023 	skb_pull(reply, sizeof(struct ipv6hdr));
2024 	skb_reset_transport_header(reply);
2025 
2026 	/* Neighbor Advertisement */
2027 	na = skb_put_zero(reply, sizeof(*na) + na_olen);
2028 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2029 	na->icmph.icmp6_router = isrouter;
2030 	na->icmph.icmp6_override = 1;
2031 	na->icmph.icmp6_solicited = 1;
2032 	na->target = ns->target;
2033 	ether_addr_copy(&na->opt[2], n->ha);
2034 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2035 	na->opt[1] = na_olen >> 3;
2036 
2037 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2038 		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2039 		csum_partial(na, sizeof(*na)+na_olen, 0));
2040 
2041 	pip6->payload_len = htons(sizeof(*na)+na_olen);
2042 
2043 	skb_push(reply, sizeof(struct ipv6hdr));
2044 
2045 	reply->ip_summed = CHECKSUM_UNNECESSARY;
2046 
2047 	return reply;
2048 }
2049 
neigh_reduce(struct net_device * dev,struct sk_buff * skb,__be32 vni)2050 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2051 {
2052 	struct vxlan_dev *vxlan = netdev_priv(dev);
2053 	const struct in6_addr *daddr;
2054 	const struct ipv6hdr *iphdr;
2055 	struct inet6_dev *in6_dev;
2056 	struct neighbour *n;
2057 	struct nd_msg *msg;
2058 
2059 	rcu_read_lock();
2060 	in6_dev = __in6_dev_get(dev);
2061 	if (!in6_dev)
2062 		goto out;
2063 
2064 	iphdr = ipv6_hdr(skb);
2065 	daddr = &iphdr->daddr;
2066 	msg = (struct nd_msg *)(iphdr + 1);
2067 
2068 	if (ipv6_addr_loopback(daddr) ||
2069 	    ipv6_addr_is_multicast(&msg->target))
2070 		goto out;
2071 
2072 	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2073 
2074 	if (n) {
2075 		struct vxlan_fdb *f;
2076 		struct sk_buff *reply;
2077 
2078 		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2079 			neigh_release(n);
2080 			goto out;
2081 		}
2082 
2083 		f = vxlan_find_mac(vxlan, n->ha, vni);
2084 		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2085 			/* bridge-local neighbor */
2086 			neigh_release(n);
2087 			goto out;
2088 		}
2089 
2090 		reply = vxlan_na_create(skb, n,
2091 					!!(f ? f->flags & NTF_ROUTER : 0));
2092 
2093 		neigh_release(n);
2094 
2095 		if (reply == NULL)
2096 			goto out;
2097 
2098 		if (netif_rx(reply) == NET_RX_DROP) {
2099 			dev_dstats_rx_dropped(dev);
2100 			vxlan_vnifilter_count(vxlan, vni, NULL,
2101 					      VXLAN_VNI_STATS_RX_DROPS, 0);
2102 		}
2103 	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2104 		union vxlan_addr ipa = {
2105 			.sin6.sin6_addr = msg->target,
2106 			.sin6.sin6_family = AF_INET6,
2107 		};
2108 
2109 		vxlan_ip_miss(dev, &ipa);
2110 	}
2111 
2112 out:
2113 	rcu_read_unlock();
2114 	consume_skb(skb);
2115 	return NETDEV_TX_OK;
2116 }
2117 #endif
2118 
route_shortcircuit(struct net_device * dev,struct sk_buff * skb)2119 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2120 {
2121 	struct vxlan_dev *vxlan = netdev_priv(dev);
2122 	struct neighbour *n;
2123 
2124 	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2125 		return false;
2126 
2127 	n = NULL;
2128 	switch (ntohs(eth_hdr(skb)->h_proto)) {
2129 	case ETH_P_IP:
2130 	{
2131 		struct iphdr *pip;
2132 
2133 		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2134 			return false;
2135 		pip = ip_hdr(skb);
2136 		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2137 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2138 			union vxlan_addr ipa = {
2139 				.sin.sin_addr.s_addr = pip->daddr,
2140 				.sin.sin_family = AF_INET,
2141 			};
2142 
2143 			vxlan_ip_miss(dev, &ipa);
2144 			return false;
2145 		}
2146 
2147 		break;
2148 	}
2149 #if IS_ENABLED(CONFIG_IPV6)
2150 	case ETH_P_IPV6:
2151 	{
2152 		struct ipv6hdr *pip6;
2153 
2154 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2155 			return false;
2156 		pip6 = ipv6_hdr(skb);
2157 		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2158 		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2159 			union vxlan_addr ipa = {
2160 				.sin6.sin6_addr = pip6->daddr,
2161 				.sin6.sin6_family = AF_INET6,
2162 			};
2163 
2164 			vxlan_ip_miss(dev, &ipa);
2165 			return false;
2166 		}
2167 
2168 		break;
2169 	}
2170 #endif
2171 	default:
2172 		return false;
2173 	}
2174 
2175 	if (n) {
2176 		bool diff;
2177 
2178 		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2179 		if (diff) {
2180 			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2181 				dev->addr_len);
2182 			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2183 		}
2184 		neigh_release(n);
2185 		return diff;
2186 	}
2187 
2188 	return false;
2189 }
2190 
vxlan_build_gpe_hdr(struct vxlanhdr * vxh,__be16 protocol)2191 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2192 {
2193 	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2194 
2195 	gpe->np_applied = 1;
2196 	gpe->next_protocol = tun_p_from_eth_p(protocol);
2197 	if (!gpe->next_protocol)
2198 		return -EPFNOSUPPORT;
2199 	return 0;
2200 }
2201 
vxlan_build_skb(struct sk_buff * skb,struct dst_entry * dst,int iphdr_len,__be32 vni,struct vxlan_metadata * md,u32 vxflags,bool udp_sum)2202 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2203 			   int iphdr_len, __be32 vni,
2204 			   struct vxlan_metadata *md, u32 vxflags,
2205 			   bool udp_sum)
2206 {
2207 	struct vxlanhdr *vxh;
2208 	int min_headroom;
2209 	int err;
2210 	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2211 	__be16 inner_protocol = htons(ETH_P_TEB);
2212 
2213 	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2214 	    skb->ip_summed == CHECKSUM_PARTIAL) {
2215 		int csum_start = skb_checksum_start_offset(skb);
2216 
2217 		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2218 		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2219 		    (skb->csum_offset == offsetof(struct udphdr, check) ||
2220 		     skb->csum_offset == offsetof(struct tcphdr, check)))
2221 			type |= SKB_GSO_TUNNEL_REMCSUM;
2222 	}
2223 
2224 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2225 			+ VXLAN_HLEN + iphdr_len;
2226 
2227 	/* Need space for new headers (invalidates iph ptr) */
2228 	err = skb_cow_head(skb, min_headroom);
2229 	if (unlikely(err))
2230 		return err;
2231 
2232 	err = iptunnel_handle_offloads(skb, type);
2233 	if (err)
2234 		return err;
2235 
2236 	vxh = __skb_push(skb, sizeof(*vxh));
2237 	vxh->vx_flags = VXLAN_HF_VNI;
2238 	vxh->vx_vni = vxlan_vni_field(vni);
2239 
2240 	if (type & SKB_GSO_TUNNEL_REMCSUM) {
2241 		unsigned int start;
2242 
2243 		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2244 		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2245 		vxh->vx_flags |= VXLAN_HF_RCO;
2246 
2247 		if (!skb_is_gso(skb)) {
2248 			skb->ip_summed = CHECKSUM_NONE;
2249 			skb->encapsulation = 0;
2250 		}
2251 	}
2252 
2253 	if (vxflags & VXLAN_F_GBP)
2254 		vxlan_build_gbp_hdr(vxh, md);
2255 	if (vxflags & VXLAN_F_GPE) {
2256 		err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2257 		if (err < 0)
2258 			return err;
2259 		inner_protocol = skb->protocol;
2260 	}
2261 
2262 	skb_set_inner_protocol(skb, inner_protocol);
2263 	return 0;
2264 }
2265 
2266 /* Bypass encapsulation if the destination is local */
vxlan_encap_bypass(struct sk_buff * skb,struct vxlan_dev * src_vxlan,struct vxlan_dev * dst_vxlan,__be32 vni,bool snoop)2267 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2268 			       struct vxlan_dev *dst_vxlan, __be32 vni,
2269 			       bool snoop)
2270 {
2271 	union vxlan_addr loopback;
2272 	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2273 	unsigned int len = skb->len;
2274 	struct net_device *dev;
2275 
2276 	skb->pkt_type = PACKET_HOST;
2277 	skb->encapsulation = 0;
2278 	skb->dev = dst_vxlan->dev;
2279 	__skb_pull(skb, skb_network_offset(skb));
2280 
2281 	if (remote_ip->sa.sa_family == AF_INET) {
2282 		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2283 		loopback.sa.sa_family =  AF_INET;
2284 #if IS_ENABLED(CONFIG_IPV6)
2285 	} else {
2286 		loopback.sin6.sin6_addr = in6addr_loopback;
2287 		loopback.sa.sa_family =  AF_INET6;
2288 #endif
2289 	}
2290 
2291 	rcu_read_lock();
2292 	dev = skb->dev;
2293 	if (unlikely(!(dev->flags & IFF_UP))) {
2294 		kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY);
2295 		goto drop;
2296 	}
2297 
2298 	if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2299 		vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2300 
2301 	dev_dstats_tx_add(src_vxlan->dev, len);
2302 	vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2303 
2304 	if (__netif_rx(skb) == NET_RX_SUCCESS) {
2305 		dev_dstats_rx_add(dst_vxlan->dev, len);
2306 		vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2307 				      len);
2308 	} else {
2309 drop:
2310 		dev_dstats_rx_dropped(dev);
2311 		vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2312 				      VXLAN_VNI_STATS_RX_DROPS, 0);
2313 	}
2314 	rcu_read_unlock();
2315 }
2316 
encap_bypass_if_local(struct sk_buff * skb,struct net_device * dev,struct vxlan_dev * vxlan,int addr_family,__be16 dst_port,int dst_ifindex,__be32 vni,struct dst_entry * dst,u32 rt_flags)2317 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2318 				 struct vxlan_dev *vxlan,
2319 				 int addr_family,
2320 				 __be16 dst_port, int dst_ifindex, __be32 vni,
2321 				 struct dst_entry *dst,
2322 				 u32 rt_flags)
2323 {
2324 #if IS_ENABLED(CONFIG_IPV6)
2325 	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2326 	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2327 	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2328 	 */
2329 	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2330 #endif
2331 	/* Bypass encapsulation if the destination is local */
2332 	if (rt_flags & RTCF_LOCAL &&
2333 	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2334 	    vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2335 		struct vxlan_dev *dst_vxlan;
2336 
2337 		dst_release(dst);
2338 		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2339 					   addr_family, dst_port,
2340 					   vxlan->cfg.flags);
2341 		if (!dst_vxlan) {
2342 			DEV_STATS_INC(dev, tx_errors);
2343 			vxlan_vnifilter_count(vxlan, vni, NULL,
2344 					      VXLAN_VNI_STATS_TX_ERRORS, 0);
2345 			kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND);
2346 
2347 			return -ENOENT;
2348 		}
2349 		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2350 		return 1;
2351 	}
2352 
2353 	return 0;
2354 }
2355 
vxlan_xmit_one(struct sk_buff * skb,struct net_device * dev,__be32 default_vni,struct vxlan_rdst * rdst,bool did_rsc)2356 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2357 		    __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2358 {
2359 	struct dst_cache *dst_cache;
2360 	struct ip_tunnel_info *info;
2361 	struct ip_tunnel_key *pkey;
2362 	struct ip_tunnel_key key;
2363 	struct vxlan_dev *vxlan = netdev_priv(dev);
2364 	const struct iphdr *old_iph;
2365 	struct vxlan_metadata _md;
2366 	struct vxlan_metadata *md = &_md;
2367 	unsigned int pkt_len = skb->len;
2368 	__be16 src_port = 0, dst_port;
2369 	struct dst_entry *ndst = NULL;
2370 	int addr_family;
2371 	__u8 tos, ttl;
2372 	int ifindex;
2373 	int err;
2374 	u32 flags = vxlan->cfg.flags;
2375 	bool use_cache;
2376 	bool udp_sum = false;
2377 	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2378 	enum skb_drop_reason reason;
2379 	bool no_eth_encap;
2380 	__be32 vni = 0;
2381 
2382 	no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB);
2383 	reason = skb_vlan_inet_prepare(skb, no_eth_encap);
2384 	if (reason)
2385 		goto drop;
2386 
2387 	reason = SKB_DROP_REASON_NOT_SPECIFIED;
2388 	old_iph = ip_hdr(skb);
2389 
2390 	info = skb_tunnel_info(skb);
2391 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
2392 
2393 	if (rdst) {
2394 		memset(&key, 0, sizeof(key));
2395 		pkey = &key;
2396 
2397 		if (vxlan_addr_any(&rdst->remote_ip)) {
2398 			if (did_rsc) {
2399 				/* short-circuited back to local bridge */
2400 				vxlan_encap_bypass(skb, vxlan, vxlan,
2401 						   default_vni, true);
2402 				return;
2403 			}
2404 			goto drop;
2405 		}
2406 
2407 		addr_family = vxlan->cfg.saddr.sa.sa_family;
2408 		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2409 		vni = (rdst->remote_vni) ? : default_vni;
2410 		ifindex = rdst->remote_ifindex;
2411 
2412 		if (addr_family == AF_INET) {
2413 			key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr;
2414 			key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr;
2415 		} else {
2416 			key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr;
2417 			key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr;
2418 		}
2419 
2420 		dst_cache = &rdst->dst_cache;
2421 		md->gbp = skb->mark;
2422 		if (flags & VXLAN_F_TTL_INHERIT) {
2423 			ttl = ip_tunnel_get_ttl(old_iph, skb);
2424 		} else {
2425 			ttl = vxlan->cfg.ttl;
2426 			if (!ttl && vxlan_addr_multicast(&rdst->remote_ip))
2427 				ttl = 1;
2428 		}
2429 		tos = vxlan->cfg.tos;
2430 		if (tos == 1)
2431 			tos = ip_tunnel_get_dsfield(old_iph, skb);
2432 		if (tos && !info)
2433 			use_cache = false;
2434 
2435 		if (addr_family == AF_INET)
2436 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2437 		else
2438 			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2439 #if IS_ENABLED(CONFIG_IPV6)
2440 		switch (vxlan->cfg.label_policy) {
2441 		case VXLAN_LABEL_FIXED:
2442 			key.label = vxlan->cfg.label;
2443 			break;
2444 		case VXLAN_LABEL_INHERIT:
2445 			key.label = ip_tunnel_get_flowlabel(old_iph, skb);
2446 			break;
2447 		default:
2448 			DEBUG_NET_WARN_ON_ONCE(1);
2449 			goto drop;
2450 		}
2451 #endif
2452 	} else {
2453 		if (!info) {
2454 			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2455 				  dev->name);
2456 			goto drop;
2457 		}
2458 		pkey = &info->key;
2459 		addr_family = ip_tunnel_info_af(info);
2460 		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2461 		vni = tunnel_id_to_key32(info->key.tun_id);
2462 		ifindex = 0;
2463 		dst_cache = &info->dst_cache;
2464 		if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
2465 			if (info->options_len < sizeof(*md))
2466 				goto drop;
2467 			md = ip_tunnel_info_opts(info);
2468 		}
2469 		ttl = info->key.ttl;
2470 		tos = info->key.tos;
2471 		udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
2472 	}
2473 	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2474 				     vxlan->cfg.port_max, true);
2475 
2476 	rcu_read_lock();
2477 	if (addr_family == AF_INET) {
2478 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2479 		struct rtable *rt;
2480 		__be16 df = 0;
2481 		__be32 saddr;
2482 
2483 		if (!ifindex)
2484 			ifindex = sock4->sock->sk->sk_bound_dev_if;
2485 
2486 		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex,
2487 					   &saddr, pkey, src_port, dst_port,
2488 					   tos, use_cache ? dst_cache : NULL);
2489 		if (IS_ERR(rt)) {
2490 			err = PTR_ERR(rt);
2491 			reason = SKB_DROP_REASON_IP_OUTNOROUTES;
2492 			goto tx_error;
2493 		}
2494 
2495 		if (!info) {
2496 			/* Bypass encapsulation if the destination is local */
2497 			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET,
2498 						    dst_port, ifindex, vni,
2499 						    &rt->dst, rt->rt_flags);
2500 			if (err)
2501 				goto out_unlock;
2502 
2503 			if (vxlan->cfg.df == VXLAN_DF_SET) {
2504 				df = htons(IP_DF);
2505 			} else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2506 				struct ethhdr *eth = eth_hdr(skb);
2507 
2508 				if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2509 				    (ntohs(eth->h_proto) == ETH_P_IP &&
2510 				     old_iph->frag_off & htons(IP_DF)))
2511 					df = htons(IP_DF);
2512 			}
2513 		} else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT,
2514 				    info->key.tun_flags)) {
2515 			df = htons(IP_DF);
2516 		}
2517 
2518 		ndst = &rt->dst;
2519 		err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2520 					    netif_is_any_bridge_port(dev));
2521 		if (err < 0) {
2522 			goto tx_error;
2523 		} else if (err) {
2524 			if (info) {
2525 				struct ip_tunnel_info *unclone;
2526 
2527 				unclone = skb_tunnel_info_unclone(skb);
2528 				if (unlikely(!unclone))
2529 					goto tx_error;
2530 
2531 				unclone->key.u.ipv4.src = pkey->u.ipv4.dst;
2532 				unclone->key.u.ipv4.dst = saddr;
2533 			}
2534 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2535 			dst_release(ndst);
2536 			goto out_unlock;
2537 		}
2538 
2539 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2540 		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2541 		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2542 				      vni, md, flags, udp_sum);
2543 		if (err < 0) {
2544 			reason = SKB_DROP_REASON_NOMEM;
2545 			goto tx_error;
2546 		}
2547 
2548 		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr,
2549 				    pkey->u.ipv4.dst, tos, ttl, df,
2550 				    src_port, dst_port, xnet, !udp_sum);
2551 #if IS_ENABLED(CONFIG_IPV6)
2552 	} else {
2553 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2554 		struct in6_addr saddr;
2555 
2556 		if (!ifindex)
2557 			ifindex = sock6->sock->sk->sk_bound_dev_if;
2558 
2559 		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
2560 					      ifindex, &saddr, pkey,
2561 					      src_port, dst_port, tos,
2562 					      use_cache ? dst_cache : NULL);
2563 		if (IS_ERR(ndst)) {
2564 			err = PTR_ERR(ndst);
2565 			ndst = NULL;
2566 			reason = SKB_DROP_REASON_IP_OUTNOROUTES;
2567 			goto tx_error;
2568 		}
2569 
2570 		if (!info) {
2571 			u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags;
2572 
2573 			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6,
2574 						    dst_port, ifindex, vni,
2575 						    ndst, rt6i_flags);
2576 			if (err)
2577 				goto out_unlock;
2578 		}
2579 
2580 		err = skb_tunnel_check_pmtu(skb, ndst,
2581 					    vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2582 					    netif_is_any_bridge_port(dev));
2583 		if (err < 0) {
2584 			goto tx_error;
2585 		} else if (err) {
2586 			if (info) {
2587 				struct ip_tunnel_info *unclone;
2588 
2589 				unclone = skb_tunnel_info_unclone(skb);
2590 				if (unlikely(!unclone))
2591 					goto tx_error;
2592 
2593 				unclone->key.u.ipv6.src = pkey->u.ipv6.dst;
2594 				unclone->key.u.ipv6.dst = saddr;
2595 			}
2596 
2597 			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2598 			dst_release(ndst);
2599 			goto out_unlock;
2600 		}
2601 
2602 		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2603 		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2604 		skb_scrub_packet(skb, xnet);
2605 		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2606 				      vni, md, flags, udp_sum);
2607 		if (err < 0) {
2608 			reason = SKB_DROP_REASON_NOMEM;
2609 			goto tx_error;
2610 		}
2611 
2612 		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2613 				     &saddr, &pkey->u.ipv6.dst, tos, ttl,
2614 				     pkey->label, src_port, dst_port, !udp_sum);
2615 #endif
2616 	}
2617 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2618 out_unlock:
2619 	rcu_read_unlock();
2620 	return;
2621 
2622 drop:
2623 	dev_dstats_tx_dropped(dev);
2624 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2625 	kfree_skb_reason(skb, reason);
2626 	return;
2627 
2628 tx_error:
2629 	rcu_read_unlock();
2630 	if (err == -ELOOP)
2631 		DEV_STATS_INC(dev, collisions);
2632 	else if (err == -ENETUNREACH)
2633 		DEV_STATS_INC(dev, tx_carrier_errors);
2634 	dst_release(ndst);
2635 	DEV_STATS_INC(dev, tx_errors);
2636 	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2637 	kfree_skb_reason(skb, reason);
2638 }
2639 
vxlan_xmit_nh(struct sk_buff * skb,struct net_device * dev,struct vxlan_fdb * f,__be32 vni,bool did_rsc)2640 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2641 			  struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2642 {
2643 	struct vxlan_rdst nh_rdst;
2644 	struct nexthop *nh;
2645 	bool do_xmit;
2646 	u32 hash;
2647 
2648 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2649 	hash = skb_get_hash(skb);
2650 
2651 	rcu_read_lock();
2652 	nh = rcu_dereference(f->nh);
2653 	if (!nh) {
2654 		rcu_read_unlock();
2655 		goto drop;
2656 	}
2657 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2658 	rcu_read_unlock();
2659 
2660 	if (likely(do_xmit))
2661 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2662 	else
2663 		goto drop;
2664 
2665 	return;
2666 
2667 drop:
2668 	dev_dstats_tx_dropped(dev);
2669 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2670 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2671 	dev_kfree_skb(skb);
2672 }
2673 
vxlan_xmit_nhid(struct sk_buff * skb,struct net_device * dev,u32 nhid,__be32 vni)2674 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2675 				   u32 nhid, __be32 vni)
2676 {
2677 	struct vxlan_dev *vxlan = netdev_priv(dev);
2678 	struct vxlan_rdst nh_rdst;
2679 	struct nexthop *nh;
2680 	bool do_xmit;
2681 	u32 hash;
2682 
2683 	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2684 	hash = skb_get_hash(skb);
2685 
2686 	rcu_read_lock();
2687 	nh = nexthop_find_by_id(dev_net(dev), nhid);
2688 	if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2689 		rcu_read_unlock();
2690 		goto drop;
2691 	}
2692 	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2693 	rcu_read_unlock();
2694 
2695 	if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2696 		goto drop;
2697 
2698 	if (likely(do_xmit))
2699 		vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2700 	else
2701 		goto drop;
2702 
2703 	return NETDEV_TX_OK;
2704 
2705 drop:
2706 	dev_dstats_tx_dropped(dev);
2707 	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2708 			      VXLAN_VNI_STATS_TX_DROPS, 0);
2709 	dev_kfree_skb(skb);
2710 	return NETDEV_TX_OK;
2711 }
2712 
2713 /* Transmit local packets over Vxlan
2714  *
2715  * Outer IP header inherits ECN and DF from inner header.
2716  * Outer UDP destination is the VXLAN assigned port.
2717  *           source port is based on hash of flow
2718  */
vxlan_xmit(struct sk_buff * skb,struct net_device * dev)2719 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2720 {
2721 	struct vxlan_dev *vxlan = netdev_priv(dev);
2722 	struct vxlan_rdst *rdst, *fdst = NULL;
2723 	const struct ip_tunnel_info *info;
2724 	struct vxlan_fdb *f;
2725 	struct ethhdr *eth;
2726 	__be32 vni = 0;
2727 	u32 nhid = 0;
2728 	bool did_rsc;
2729 
2730 	info = skb_tunnel_info(skb);
2731 
2732 	skb_reset_mac_header(skb);
2733 
2734 	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2735 		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2736 		    info->mode & IP_TUNNEL_INFO_TX) {
2737 			vni = tunnel_id_to_key32(info->key.tun_id);
2738 			nhid = info->key.nhid;
2739 		} else {
2740 			if (info && info->mode & IP_TUNNEL_INFO_TX)
2741 				vxlan_xmit_one(skb, dev, vni, NULL, false);
2742 			else
2743 				kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO);
2744 			return NETDEV_TX_OK;
2745 		}
2746 	}
2747 
2748 	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2749 		eth = eth_hdr(skb);
2750 		if (ntohs(eth->h_proto) == ETH_P_ARP)
2751 			return arp_reduce(dev, skb, vni);
2752 #if IS_ENABLED(CONFIG_IPV6)
2753 		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2754 			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2755 					    sizeof(struct nd_msg)) &&
2756 			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2757 			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2758 
2759 			if (m->icmph.icmp6_code == 0 &&
2760 			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2761 				return neigh_reduce(dev, skb, vni);
2762 		}
2763 #endif
2764 	}
2765 
2766 	if (nhid)
2767 		return vxlan_xmit_nhid(skb, dev, nhid, vni);
2768 
2769 	if (vxlan->cfg.flags & VXLAN_F_MDB) {
2770 		struct vxlan_mdb_entry *mdb_entry;
2771 
2772 		rcu_read_lock();
2773 		mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2774 		if (mdb_entry) {
2775 			netdev_tx_t ret;
2776 
2777 			ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2778 			rcu_read_unlock();
2779 			return ret;
2780 		}
2781 		rcu_read_unlock();
2782 	}
2783 
2784 	eth = eth_hdr(skb);
2785 	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2786 	did_rsc = false;
2787 
2788 	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2789 	    (ntohs(eth->h_proto) == ETH_P_IP ||
2790 	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2791 		did_rsc = route_shortcircuit(dev, skb);
2792 		if (did_rsc)
2793 			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2794 	}
2795 
2796 	if (f == NULL) {
2797 		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2798 		if (f == NULL) {
2799 			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2800 			    !is_multicast_ether_addr(eth->h_dest))
2801 				vxlan_fdb_miss(vxlan, eth->h_dest);
2802 
2803 			dev_dstats_tx_dropped(dev);
2804 			vxlan_vnifilter_count(vxlan, vni, NULL,
2805 					      VXLAN_VNI_STATS_TX_DROPS, 0);
2806 			kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET);
2807 			return NETDEV_TX_OK;
2808 		}
2809 	}
2810 
2811 	if (rcu_access_pointer(f->nh)) {
2812 		vxlan_xmit_nh(skb, dev, f,
2813 			      (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2814 	} else {
2815 		list_for_each_entry_rcu(rdst, &f->remotes, list) {
2816 			struct sk_buff *skb1;
2817 
2818 			if (!fdst) {
2819 				fdst = rdst;
2820 				continue;
2821 			}
2822 			skb1 = skb_clone(skb, GFP_ATOMIC);
2823 			if (skb1)
2824 				vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2825 		}
2826 		if (fdst)
2827 			vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2828 		else
2829 			kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET);
2830 	}
2831 
2832 	return NETDEV_TX_OK;
2833 }
2834 
2835 /* Walk the forwarding table and purge stale entries */
vxlan_cleanup(struct timer_list * t)2836 static void vxlan_cleanup(struct timer_list *t)
2837 {
2838 	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2839 	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2840 	unsigned int h;
2841 
2842 	if (!netif_running(vxlan->dev))
2843 		return;
2844 
2845 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2846 		struct hlist_node *p, *n;
2847 
2848 		spin_lock(&vxlan->hash_lock[h]);
2849 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2850 			struct vxlan_fdb *f
2851 				= container_of(p, struct vxlan_fdb, hlist);
2852 			unsigned long timeout;
2853 
2854 			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2855 				continue;
2856 
2857 			if (f->flags & NTF_EXT_LEARNED)
2858 				continue;
2859 
2860 			timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ;
2861 			if (time_before_eq(timeout, jiffies)) {
2862 				netdev_dbg(vxlan->dev,
2863 					   "garbage collect %pM\n",
2864 					   f->eth_addr);
2865 				f->state = NUD_STALE;
2866 				vxlan_fdb_destroy(vxlan, f, true, true);
2867 			} else if (time_before(timeout, next_timer))
2868 				next_timer = timeout;
2869 		}
2870 		spin_unlock(&vxlan->hash_lock[h]);
2871 	}
2872 
2873 	mod_timer(&vxlan->age_timer, next_timer);
2874 }
2875 
vxlan_vs_del_dev(struct vxlan_dev * vxlan)2876 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2877 {
2878 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2879 
2880 	spin_lock(&vn->sock_lock);
2881 	hlist_del_init_rcu(&vxlan->hlist4.hlist);
2882 #if IS_ENABLED(CONFIG_IPV6)
2883 	hlist_del_init_rcu(&vxlan->hlist6.hlist);
2884 #endif
2885 	spin_unlock(&vn->sock_lock);
2886 }
2887 
vxlan_vs_add_dev(struct vxlan_sock * vs,struct vxlan_dev * vxlan,struct vxlan_dev_node * node)2888 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2889 			     struct vxlan_dev_node *node)
2890 {
2891 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2892 	__be32 vni = vxlan->default_dst.remote_vni;
2893 
2894 	node->vxlan = vxlan;
2895 	spin_lock(&vn->sock_lock);
2896 	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2897 	spin_unlock(&vn->sock_lock);
2898 }
2899 
2900 /* Setup stats when device is created */
vxlan_init(struct net_device * dev)2901 static int vxlan_init(struct net_device *dev)
2902 {
2903 	struct vxlan_dev *vxlan = netdev_priv(dev);
2904 	int err;
2905 
2906 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
2907 		err = vxlan_vnigroup_init(vxlan);
2908 		if (err)
2909 			return err;
2910 	}
2911 
2912 	err = gro_cells_init(&vxlan->gro_cells, dev);
2913 	if (err)
2914 		goto err_vnigroup_uninit;
2915 
2916 	err = vxlan_mdb_init(vxlan);
2917 	if (err)
2918 		goto err_gro_cells_destroy;
2919 
2920 	netdev_lockdep_set_classes(dev);
2921 	return 0;
2922 
2923 err_gro_cells_destroy:
2924 	gro_cells_destroy(&vxlan->gro_cells);
2925 err_vnigroup_uninit:
2926 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2927 		vxlan_vnigroup_uninit(vxlan);
2928 	return err;
2929 }
2930 
vxlan_fdb_delete_default(struct vxlan_dev * vxlan,__be32 vni)2931 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2932 {
2933 	struct vxlan_fdb *f;
2934 	u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2935 
2936 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
2937 	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2938 	if (f)
2939 		vxlan_fdb_destroy(vxlan, f, true, true);
2940 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2941 }
2942 
vxlan_uninit(struct net_device * dev)2943 static void vxlan_uninit(struct net_device *dev)
2944 {
2945 	struct vxlan_dev *vxlan = netdev_priv(dev);
2946 
2947 	vxlan_mdb_fini(vxlan);
2948 
2949 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2950 		vxlan_vnigroup_uninit(vxlan);
2951 
2952 	gro_cells_destroy(&vxlan->gro_cells);
2953 
2954 	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2955 }
2956 
2957 /* Start ageing timer and join group when device is brought up */
vxlan_open(struct net_device * dev)2958 static int vxlan_open(struct net_device *dev)
2959 {
2960 	struct vxlan_dev *vxlan = netdev_priv(dev);
2961 	int ret;
2962 
2963 	ret = vxlan_sock_add(vxlan);
2964 	if (ret < 0)
2965 		return ret;
2966 
2967 	ret = vxlan_multicast_join(vxlan);
2968 	if (ret) {
2969 		vxlan_sock_release(vxlan);
2970 		return ret;
2971 	}
2972 
2973 	if (vxlan->cfg.age_interval)
2974 		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2975 
2976 	return ret;
2977 }
2978 
2979 struct vxlan_fdb_flush_desc {
2980 	bool				ignore_default_entry;
2981 	unsigned long                   state;
2982 	unsigned long			state_mask;
2983 	unsigned long                   flags;
2984 	unsigned long			flags_mask;
2985 	__be32				src_vni;
2986 	u32				nhid;
2987 	__be32				vni;
2988 	__be16				port;
2989 	union vxlan_addr		dst_ip;
2990 };
2991 
vxlan_fdb_is_default_entry(const struct vxlan_fdb * f,const struct vxlan_dev * vxlan)2992 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f,
2993 				       const struct vxlan_dev *vxlan)
2994 {
2995 	return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni;
2996 }
2997 
vxlan_fdb_nhid_matches(const struct vxlan_fdb * f,u32 nhid)2998 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid)
2999 {
3000 	struct nexthop *nh = rtnl_dereference(f->nh);
3001 
3002 	return nh && nh->id == nhid;
3003 }
3004 
vxlan_fdb_flush_matches(const struct vxlan_fdb * f,const struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc)3005 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f,
3006 				    const struct vxlan_dev *vxlan,
3007 				    const struct vxlan_fdb_flush_desc *desc)
3008 {
3009 	if (desc->state_mask && (f->state & desc->state_mask) != desc->state)
3010 		return false;
3011 
3012 	if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags)
3013 		return false;
3014 
3015 	if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan))
3016 		return false;
3017 
3018 	if (desc->src_vni && f->vni != desc->src_vni)
3019 		return false;
3020 
3021 	if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid))
3022 		return false;
3023 
3024 	return true;
3025 }
3026 
3027 static bool
vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc * desc)3028 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc)
3029 {
3030 	return desc->vni || desc->port || desc->dst_ip.sa.sa_family;
3031 }
3032 
3033 static bool
vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc * desc,const struct vxlan_rdst * rd)3034 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc,
3035 			       const struct vxlan_rdst *rd)
3036 {
3037 	if (desc->vni && rd->remote_vni != desc->vni)
3038 		return false;
3039 
3040 	if (desc->port && rd->remote_port != desc->port)
3041 		return false;
3042 
3043 	if (desc->dst_ip.sa.sa_family &&
3044 	    !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip))
3045 		return false;
3046 
3047 	return true;
3048 }
3049 
3050 static void
vxlan_fdb_flush_match_remotes(struct vxlan_fdb * f,struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc,bool * p_destroy_fdb)3051 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan,
3052 			      const struct vxlan_fdb_flush_desc *desc,
3053 			      bool *p_destroy_fdb)
3054 {
3055 	bool remotes_flushed = false;
3056 	struct vxlan_rdst *rd, *tmp;
3057 
3058 	list_for_each_entry_safe(rd, tmp, &f->remotes, list) {
3059 		if (!vxlan_fdb_flush_remote_matches(desc, rd))
3060 			continue;
3061 
3062 		vxlan_fdb_dst_destroy(vxlan, f, rd, true);
3063 		remotes_flushed = true;
3064 	}
3065 
3066 	*p_destroy_fdb = remotes_flushed && list_empty(&f->remotes);
3067 }
3068 
3069 /* Purge the forwarding table */
vxlan_flush(struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc)3070 static void vxlan_flush(struct vxlan_dev *vxlan,
3071 			const struct vxlan_fdb_flush_desc *desc)
3072 {
3073 	bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc);
3074 	unsigned int h;
3075 
3076 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3077 		struct hlist_node *p, *n;
3078 
3079 		spin_lock_bh(&vxlan->hash_lock[h]);
3080 		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3081 			struct vxlan_fdb *f
3082 				= container_of(p, struct vxlan_fdb, hlist);
3083 
3084 			if (!vxlan_fdb_flush_matches(f, vxlan, desc))
3085 				continue;
3086 
3087 			if (match_remotes) {
3088 				bool destroy_fdb = false;
3089 
3090 				vxlan_fdb_flush_match_remotes(f, vxlan, desc,
3091 							      &destroy_fdb);
3092 
3093 				if (!destroy_fdb)
3094 					continue;
3095 			}
3096 
3097 			vxlan_fdb_destroy(vxlan, f, true, true);
3098 		}
3099 		spin_unlock_bh(&vxlan->hash_lock[h]);
3100 	}
3101 }
3102 
3103 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = {
3104 	[NDA_SRC_VNI]   = { .type = NLA_U32 },
3105 	[NDA_NH_ID]	= { .type = NLA_U32 },
3106 	[NDA_VNI]	= { .type = NLA_U32 },
3107 	[NDA_PORT]	= { .type = NLA_U16 },
3108 	[NDA_DST]	= NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr),
3109 					   sizeof(struct in6_addr)),
3110 	[NDA_NDM_STATE_MASK]	= { .type = NLA_U16 },
3111 	[NDA_NDM_FLAGS_MASK]	= { .type = NLA_U8 },
3112 };
3113 
3114 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF)
3115 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP)
3116 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \
3117 					   NTF_ROUTER)
3118 
vxlan_fdb_delete_bulk(struct nlmsghdr * nlh,struct net_device * dev,struct netlink_ext_ack * extack)3119 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev,
3120 				 struct netlink_ext_ack *extack)
3121 {
3122 	struct vxlan_dev *vxlan = netdev_priv(dev);
3123 	struct vxlan_fdb_flush_desc desc = {};
3124 	struct ndmsg *ndm = nlmsg_data(nlh);
3125 	struct nlattr *tb[NDA_MAX + 1];
3126 	u8 ndm_flags;
3127 	int err;
3128 
3129 	ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS;
3130 
3131 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy,
3132 			  extack);
3133 	if (err)
3134 		return err;
3135 
3136 	if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) {
3137 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set");
3138 		return -EINVAL;
3139 	}
3140 	if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) {
3141 		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set");
3142 		return -EINVAL;
3143 	}
3144 
3145 	desc.state = ndm->ndm_state;
3146 	desc.flags = ndm_flags;
3147 
3148 	if (tb[NDA_NDM_STATE_MASK])
3149 		desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]);
3150 
3151 	if (tb[NDA_NDM_FLAGS_MASK])
3152 		desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]);
3153 
3154 	if (tb[NDA_SRC_VNI])
3155 		desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
3156 
3157 	if (tb[NDA_NH_ID])
3158 		desc.nhid = nla_get_u32(tb[NDA_NH_ID]);
3159 
3160 	if (tb[NDA_VNI])
3161 		desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
3162 
3163 	if (tb[NDA_PORT])
3164 		desc.port = nla_get_be16(tb[NDA_PORT]);
3165 
3166 	if (tb[NDA_DST]) {
3167 		union vxlan_addr ip;
3168 
3169 		err = vxlan_nla_get_addr(&ip, tb[NDA_DST]);
3170 		if (err) {
3171 			NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST],
3172 					    "Unsupported address family");
3173 			return err;
3174 		}
3175 		desc.dst_ip = ip;
3176 	}
3177 
3178 	vxlan_flush(vxlan, &desc);
3179 
3180 	return 0;
3181 }
3182 
3183 /* Cleanup timer and forwarding table on shutdown */
vxlan_stop(struct net_device * dev)3184 static int vxlan_stop(struct net_device *dev)
3185 {
3186 	struct vxlan_dev *vxlan = netdev_priv(dev);
3187 	struct vxlan_fdb_flush_desc desc = {
3188 		/* Default entry is deleted at vxlan_uninit. */
3189 		.ignore_default_entry = true,
3190 		.state = 0,
3191 		.state_mask = NUD_PERMANENT | NUD_NOARP,
3192 	};
3193 
3194 	vxlan_multicast_leave(vxlan);
3195 
3196 	del_timer_sync(&vxlan->age_timer);
3197 
3198 	vxlan_flush(vxlan, &desc);
3199 	vxlan_sock_release(vxlan);
3200 
3201 	return 0;
3202 }
3203 
3204 /* Stub, nothing needs to be done. */
vxlan_set_multicast_list(struct net_device * dev)3205 static void vxlan_set_multicast_list(struct net_device *dev)
3206 {
3207 }
3208 
vxlan_change_mtu(struct net_device * dev,int new_mtu)3209 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3210 {
3211 	struct vxlan_dev *vxlan = netdev_priv(dev);
3212 	struct vxlan_rdst *dst = &vxlan->default_dst;
3213 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3214 							 dst->remote_ifindex);
3215 
3216 	/* This check is different than dev->max_mtu, because it looks at
3217 	 * the lowerdev->mtu, rather than the static dev->max_mtu
3218 	 */
3219 	if (lowerdev) {
3220 		int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3221 		if (new_mtu > max_mtu)
3222 			return -EINVAL;
3223 	}
3224 
3225 	WRITE_ONCE(dev->mtu, new_mtu);
3226 	return 0;
3227 }
3228 
vxlan_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)3229 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3230 {
3231 	struct vxlan_dev *vxlan = netdev_priv(dev);
3232 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
3233 	__be16 sport, dport;
3234 
3235 	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3236 				  vxlan->cfg.port_max, true);
3237 	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3238 
3239 	if (ip_tunnel_info_af(info) == AF_INET) {
3240 		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3241 		struct rtable *rt;
3242 
3243 		if (!sock4)
3244 			return -EIO;
3245 
3246 		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0,
3247 					   &info->key.u.ipv4.src,
3248 					   &info->key,
3249 					   sport, dport, info->key.tos,
3250 					   &info->dst_cache);
3251 		if (IS_ERR(rt))
3252 			return PTR_ERR(rt);
3253 		ip_rt_put(rt);
3254 	} else {
3255 #if IS_ENABLED(CONFIG_IPV6)
3256 		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3257 		struct dst_entry *ndst;
3258 
3259 		if (!sock6)
3260 			return -EIO;
3261 
3262 		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
3263 					      0, &info->key.u.ipv6.src,
3264 					      &info->key,
3265 					      sport, dport, info->key.tos,
3266 					      &info->dst_cache);
3267 		if (IS_ERR(ndst))
3268 			return PTR_ERR(ndst);
3269 		dst_release(ndst);
3270 #else /* !CONFIG_IPV6 */
3271 		return -EPFNOSUPPORT;
3272 #endif
3273 	}
3274 	info->key.tp_src = sport;
3275 	info->key.tp_dst = dport;
3276 	return 0;
3277 }
3278 
3279 static const struct net_device_ops vxlan_netdev_ether_ops = {
3280 	.ndo_init		= vxlan_init,
3281 	.ndo_uninit		= vxlan_uninit,
3282 	.ndo_open		= vxlan_open,
3283 	.ndo_stop		= vxlan_stop,
3284 	.ndo_start_xmit		= vxlan_xmit,
3285 	.ndo_set_rx_mode	= vxlan_set_multicast_list,
3286 	.ndo_change_mtu		= vxlan_change_mtu,
3287 	.ndo_validate_addr	= eth_validate_addr,
3288 	.ndo_set_mac_address	= eth_mac_addr,
3289 	.ndo_fdb_add		= vxlan_fdb_add,
3290 	.ndo_fdb_del		= vxlan_fdb_delete,
3291 	.ndo_fdb_del_bulk	= vxlan_fdb_delete_bulk,
3292 	.ndo_fdb_dump		= vxlan_fdb_dump,
3293 	.ndo_fdb_get		= vxlan_fdb_get,
3294 	.ndo_mdb_add		= vxlan_mdb_add,
3295 	.ndo_mdb_del		= vxlan_mdb_del,
3296 	.ndo_mdb_del_bulk	= vxlan_mdb_del_bulk,
3297 	.ndo_mdb_dump		= vxlan_mdb_dump,
3298 	.ndo_mdb_get		= vxlan_mdb_get,
3299 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3300 };
3301 
3302 static const struct net_device_ops vxlan_netdev_raw_ops = {
3303 	.ndo_init		= vxlan_init,
3304 	.ndo_uninit		= vxlan_uninit,
3305 	.ndo_open		= vxlan_open,
3306 	.ndo_stop		= vxlan_stop,
3307 	.ndo_start_xmit		= vxlan_xmit,
3308 	.ndo_change_mtu		= vxlan_change_mtu,
3309 	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3310 };
3311 
3312 /* Info for udev, that this is a virtual tunnel endpoint */
3313 static const struct device_type vxlan_type = {
3314 	.name = "vxlan",
3315 };
3316 
3317 /* Calls the ndo_udp_tunnel_add of the caller in order to
3318  * supply the listening VXLAN udp ports. Callers are expected
3319  * to implement the ndo_udp_tunnel_add.
3320  */
vxlan_offload_rx_ports(struct net_device * dev,bool push)3321 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3322 {
3323 	struct vxlan_sock *vs;
3324 	struct net *net = dev_net(dev);
3325 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3326 	unsigned int i;
3327 
3328 	spin_lock(&vn->sock_lock);
3329 	for (i = 0; i < PORT_HASH_SIZE; ++i) {
3330 		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3331 			unsigned short type;
3332 
3333 			if (vs->flags & VXLAN_F_GPE)
3334 				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3335 			else
3336 				type = UDP_TUNNEL_TYPE_VXLAN;
3337 
3338 			if (push)
3339 				udp_tunnel_push_rx_port(dev, vs->sock, type);
3340 			else
3341 				udp_tunnel_drop_rx_port(dev, vs->sock, type);
3342 		}
3343 	}
3344 	spin_unlock(&vn->sock_lock);
3345 }
3346 
3347 /* Initialize the device structure. */
vxlan_setup(struct net_device * dev)3348 static void vxlan_setup(struct net_device *dev)
3349 {
3350 	struct vxlan_dev *vxlan = netdev_priv(dev);
3351 	unsigned int h;
3352 
3353 	eth_hw_addr_random(dev);
3354 	ether_setup(dev);
3355 
3356 	dev->needs_free_netdev = true;
3357 	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3358 
3359 	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3360 	dev->features   |= NETIF_F_RXCSUM;
3361 	dev->features   |= NETIF_F_GSO_SOFTWARE;
3362 
3363 	dev->vlan_features = dev->features;
3364 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3365 	dev->hw_features |= NETIF_F_RXCSUM;
3366 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3367 	netif_keep_dst(dev);
3368 	dev->priv_flags |= IFF_NO_QUEUE;
3369 	dev->change_proto_down = true;
3370 	dev->lltx = true;
3371 
3372 	/* MTU range: 68 - 65535 */
3373 	dev->min_mtu = ETH_MIN_MTU;
3374 	dev->max_mtu = ETH_MAX_MTU;
3375 
3376 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS;
3377 	INIT_LIST_HEAD(&vxlan->next);
3378 
3379 	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3380 
3381 	vxlan->dev = dev;
3382 
3383 	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3384 		spin_lock_init(&vxlan->hash_lock[h]);
3385 		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3386 	}
3387 }
3388 
vxlan_ether_setup(struct net_device * dev)3389 static void vxlan_ether_setup(struct net_device *dev)
3390 {
3391 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3392 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3393 	dev->netdev_ops = &vxlan_netdev_ether_ops;
3394 }
3395 
vxlan_raw_setup(struct net_device * dev)3396 static void vxlan_raw_setup(struct net_device *dev)
3397 {
3398 	dev->header_ops = NULL;
3399 	dev->type = ARPHRD_NONE;
3400 	dev->hard_header_len = 0;
3401 	dev->addr_len = 0;
3402 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3403 	dev->netdev_ops = &vxlan_netdev_raw_ops;
3404 }
3405 
3406 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3407 	[IFLA_VXLAN_UNSPEC]     = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3408 	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
3409 	[IFLA_VXLAN_GROUP]	= { .len = sizeof_field(struct iphdr, daddr) },
3410 	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
3411 	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
3412 	[IFLA_VXLAN_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
3413 	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
3414 	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
3415 	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
3416 	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
3417 	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
3418 	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
3419 	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
3420 	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3421 	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
3422 	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
3423 	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
3424 	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
3425 	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
3426 	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
3427 	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
3428 	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
3429 	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
3430 	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
3431 	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
3432 	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
3433 	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
3434 	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
3435 	[IFLA_VXLAN_TTL_INHERIT]	= { .type = NLA_FLAG },
3436 	[IFLA_VXLAN_DF]		= { .type = NLA_U8 },
3437 	[IFLA_VXLAN_VNIFILTER]	= { .type = NLA_U8 },
3438 	[IFLA_VXLAN_LOCALBYPASS]	= NLA_POLICY_MAX(NLA_U8, 1),
3439 	[IFLA_VXLAN_LABEL_POLICY]       = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX),
3440 	[IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)),
3441 };
3442 
vxlan_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)3443 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3444 			  struct netlink_ext_ack *extack)
3445 {
3446 	if (tb[IFLA_ADDRESS]) {
3447 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3448 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3449 					    "Provided link layer address is not Ethernet");
3450 			return -EINVAL;
3451 		}
3452 
3453 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3454 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3455 					    "Provided Ethernet address is not unicast");
3456 			return -EADDRNOTAVAIL;
3457 		}
3458 	}
3459 
3460 	if (tb[IFLA_MTU]) {
3461 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3462 
3463 		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3464 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3465 					    "MTU must be between 68 and 65535");
3466 			return -EINVAL;
3467 		}
3468 	}
3469 
3470 	if (!data) {
3471 		NL_SET_ERR_MSG(extack,
3472 			       "Required attributes not provided to perform the operation");
3473 		return -EINVAL;
3474 	}
3475 
3476 	if (data[IFLA_VXLAN_ID]) {
3477 		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3478 
3479 		if (id >= VXLAN_N_VID) {
3480 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3481 					    "VXLAN ID must be lower than 16777216");
3482 			return -ERANGE;
3483 		}
3484 	}
3485 
3486 	if (data[IFLA_VXLAN_PORT_RANGE]) {
3487 		const struct ifla_vxlan_port_range *p
3488 			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3489 
3490 		if (ntohs(p->high) < ntohs(p->low)) {
3491 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3492 					    "Invalid source port range");
3493 			return -EINVAL;
3494 		}
3495 	}
3496 
3497 	if (data[IFLA_VXLAN_DF]) {
3498 		enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3499 
3500 		if (df < 0 || df > VXLAN_DF_MAX) {
3501 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3502 					    "Invalid DF attribute");
3503 			return -EINVAL;
3504 		}
3505 	}
3506 
3507 	return 0;
3508 }
3509 
vxlan_get_drvinfo(struct net_device * netdev,struct ethtool_drvinfo * drvinfo)3510 static void vxlan_get_drvinfo(struct net_device *netdev,
3511 			      struct ethtool_drvinfo *drvinfo)
3512 {
3513 	strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3514 	strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3515 }
3516 
vxlan_get_link_ksettings(struct net_device * dev,struct ethtool_link_ksettings * cmd)3517 static int vxlan_get_link_ksettings(struct net_device *dev,
3518 				    struct ethtool_link_ksettings *cmd)
3519 {
3520 	struct vxlan_dev *vxlan = netdev_priv(dev);
3521 	struct vxlan_rdst *dst = &vxlan->default_dst;
3522 	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3523 							 dst->remote_ifindex);
3524 
3525 	if (!lowerdev) {
3526 		cmd->base.duplex = DUPLEX_UNKNOWN;
3527 		cmd->base.port = PORT_OTHER;
3528 		cmd->base.speed = SPEED_UNKNOWN;
3529 
3530 		return 0;
3531 	}
3532 
3533 	return __ethtool_get_link_ksettings(lowerdev, cmd);
3534 }
3535 
3536 static const struct ethtool_ops vxlan_ethtool_ops = {
3537 	.get_drvinfo		= vxlan_get_drvinfo,
3538 	.get_link		= ethtool_op_get_link,
3539 	.get_link_ksettings	= vxlan_get_link_ksettings,
3540 };
3541 
vxlan_create_sock(struct net * net,bool ipv6,__be16 port,u32 flags,int ifindex)3542 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3543 					__be16 port, u32 flags, int ifindex)
3544 {
3545 	struct socket *sock;
3546 	struct udp_port_cfg udp_conf;
3547 	int err;
3548 
3549 	memset(&udp_conf, 0, sizeof(udp_conf));
3550 
3551 	if (ipv6) {
3552 		udp_conf.family = AF_INET6;
3553 		udp_conf.use_udp6_rx_checksums =
3554 		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3555 		udp_conf.ipv6_v6only = 1;
3556 	} else {
3557 		udp_conf.family = AF_INET;
3558 	}
3559 
3560 	udp_conf.local_udp_port = port;
3561 	udp_conf.bind_ifindex = ifindex;
3562 
3563 	/* Open UDP socket */
3564 	err = udp_sock_create(net, &udp_conf, &sock);
3565 	if (err < 0)
3566 		return ERR_PTR(err);
3567 
3568 	udp_allow_gso(sock->sk);
3569 	return sock;
3570 }
3571 
3572 /* Create new listen socket if needed */
vxlan_socket_create(struct net * net,bool ipv6,__be16 port,u32 flags,int ifindex)3573 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3574 					      __be16 port, u32 flags,
3575 					      int ifindex)
3576 {
3577 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3578 	struct vxlan_sock *vs;
3579 	struct socket *sock;
3580 	unsigned int h;
3581 	struct udp_tunnel_sock_cfg tunnel_cfg;
3582 
3583 	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3584 	if (!vs)
3585 		return ERR_PTR(-ENOMEM);
3586 
3587 	for (h = 0; h < VNI_HASH_SIZE; ++h)
3588 		INIT_HLIST_HEAD(&vs->vni_list[h]);
3589 
3590 	sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3591 	if (IS_ERR(sock)) {
3592 		kfree(vs);
3593 		return ERR_CAST(sock);
3594 	}
3595 
3596 	vs->sock = sock;
3597 	refcount_set(&vs->refcnt, 1);
3598 	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3599 
3600 	spin_lock(&vn->sock_lock);
3601 	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3602 	udp_tunnel_notify_add_rx_port(sock,
3603 				      (vs->flags & VXLAN_F_GPE) ?
3604 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
3605 				      UDP_TUNNEL_TYPE_VXLAN);
3606 	spin_unlock(&vn->sock_lock);
3607 
3608 	/* Mark socket as an encapsulation socket. */
3609 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3610 	tunnel_cfg.sk_user_data = vs;
3611 	tunnel_cfg.encap_type = 1;
3612 	tunnel_cfg.encap_rcv = vxlan_rcv;
3613 	tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3614 	tunnel_cfg.encap_destroy = NULL;
3615 	if (vs->flags & VXLAN_F_GPE) {
3616 		tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3617 		tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3618 	} else {
3619 		tunnel_cfg.gro_receive = vxlan_gro_receive;
3620 		tunnel_cfg.gro_complete = vxlan_gro_complete;
3621 	}
3622 
3623 	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3624 
3625 	return vs;
3626 }
3627 
__vxlan_sock_add(struct vxlan_dev * vxlan,bool ipv6)3628 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3629 {
3630 	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3631 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3632 	struct vxlan_sock *vs = NULL;
3633 	struct vxlan_dev_node *node;
3634 	int l3mdev_index = 0;
3635 
3636 	if (vxlan->cfg.remote_ifindex)
3637 		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3638 			vxlan->net, vxlan->cfg.remote_ifindex);
3639 
3640 	if (!vxlan->cfg.no_share) {
3641 		spin_lock(&vn->sock_lock);
3642 		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3643 				     vxlan->cfg.dst_port, vxlan->cfg.flags,
3644 				     l3mdev_index);
3645 		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3646 			spin_unlock(&vn->sock_lock);
3647 			return -EBUSY;
3648 		}
3649 		spin_unlock(&vn->sock_lock);
3650 	}
3651 	if (!vs)
3652 		vs = vxlan_socket_create(vxlan->net, ipv6,
3653 					 vxlan->cfg.dst_port, vxlan->cfg.flags,
3654 					 l3mdev_index);
3655 	if (IS_ERR(vs))
3656 		return PTR_ERR(vs);
3657 #if IS_ENABLED(CONFIG_IPV6)
3658 	if (ipv6) {
3659 		rcu_assign_pointer(vxlan->vn6_sock, vs);
3660 		node = &vxlan->hlist6;
3661 	} else
3662 #endif
3663 	{
3664 		rcu_assign_pointer(vxlan->vn4_sock, vs);
3665 		node = &vxlan->hlist4;
3666 	}
3667 
3668 	if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3669 		vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3670 	else
3671 		vxlan_vs_add_dev(vs, vxlan, node);
3672 
3673 	return 0;
3674 }
3675 
vxlan_sock_add(struct vxlan_dev * vxlan)3676 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3677 {
3678 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3679 	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3680 	bool ipv4 = !ipv6 || metadata;
3681 	int ret = 0;
3682 
3683 	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3684 #if IS_ENABLED(CONFIG_IPV6)
3685 	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3686 	if (ipv6) {
3687 		ret = __vxlan_sock_add(vxlan, true);
3688 		if (ret < 0 && ret != -EAFNOSUPPORT)
3689 			ipv4 = false;
3690 	}
3691 #endif
3692 	if (ipv4)
3693 		ret = __vxlan_sock_add(vxlan, false);
3694 	if (ret < 0)
3695 		vxlan_sock_release(vxlan);
3696 	return ret;
3697 }
3698 
vxlan_vni_in_use(struct net * src_net,struct vxlan_dev * vxlan,struct vxlan_config * conf,__be32 vni)3699 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3700 		     struct vxlan_config *conf, __be32 vni)
3701 {
3702 	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3703 	struct vxlan_dev *tmp;
3704 
3705 	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3706 		if (tmp == vxlan)
3707 			continue;
3708 		if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3709 			if (!vxlan_vnifilter_lookup(tmp, vni))
3710 				continue;
3711 		} else if (tmp->cfg.vni != vni) {
3712 			continue;
3713 		}
3714 		if (tmp->cfg.dst_port != conf->dst_port)
3715 			continue;
3716 		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3717 		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3718 			continue;
3719 
3720 		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3721 		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3722 			continue;
3723 
3724 		return -EEXIST;
3725 	}
3726 
3727 	return 0;
3728 }
3729 
vxlan_config_validate(struct net * src_net,struct vxlan_config * conf,struct net_device ** lower,struct vxlan_dev * old,struct netlink_ext_ack * extack)3730 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3731 				 struct net_device **lower,
3732 				 struct vxlan_dev *old,
3733 				 struct netlink_ext_ack *extack)
3734 {
3735 	bool use_ipv6 = false;
3736 
3737 	if (conf->flags & VXLAN_F_GPE) {
3738 		/* For now, allow GPE only together with
3739 		 * COLLECT_METADATA. This can be relaxed later; in such
3740 		 * case, the other side of the PtP link will have to be
3741 		 * provided.
3742 		 */
3743 		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3744 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3745 			NL_SET_ERR_MSG(extack,
3746 				       "VXLAN GPE does not support this combination of attributes");
3747 			return -EINVAL;
3748 		}
3749 	}
3750 
3751 	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3752 		/* Unless IPv6 is explicitly requested, assume IPv4 */
3753 		conf->remote_ip.sa.sa_family = AF_INET;
3754 		conf->saddr.sa.sa_family = AF_INET;
3755 	} else if (!conf->remote_ip.sa.sa_family) {
3756 		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3757 	} else if (!conf->saddr.sa.sa_family) {
3758 		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3759 	}
3760 
3761 	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3762 		NL_SET_ERR_MSG(extack,
3763 			       "Local and remote address must be from the same family");
3764 		return -EINVAL;
3765 	}
3766 
3767 	if (vxlan_addr_multicast(&conf->saddr)) {
3768 		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3769 		return -EINVAL;
3770 	}
3771 
3772 	if (conf->saddr.sa.sa_family == AF_INET6) {
3773 		if (!IS_ENABLED(CONFIG_IPV6)) {
3774 			NL_SET_ERR_MSG(extack,
3775 				       "IPv6 support not enabled in the kernel");
3776 			return -EPFNOSUPPORT;
3777 		}
3778 		use_ipv6 = true;
3779 		conf->flags |= VXLAN_F_IPV6;
3780 
3781 		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3782 			int local_type =
3783 				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3784 			int remote_type =
3785 				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3786 
3787 			if (local_type & IPV6_ADDR_LINKLOCAL) {
3788 				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3789 				    (remote_type != IPV6_ADDR_ANY)) {
3790 					NL_SET_ERR_MSG(extack,
3791 						       "Invalid combination of local and remote address scopes");
3792 					return -EINVAL;
3793 				}
3794 
3795 				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3796 			} else {
3797 				if (remote_type ==
3798 				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3799 					NL_SET_ERR_MSG(extack,
3800 						       "Invalid combination of local and remote address scopes");
3801 					return -EINVAL;
3802 				}
3803 
3804 				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3805 			}
3806 		}
3807 	}
3808 
3809 	if (conf->label && !use_ipv6) {
3810 		NL_SET_ERR_MSG(extack,
3811 			       "Label attribute only applies to IPv6 VXLAN devices");
3812 		return -EINVAL;
3813 	}
3814 
3815 	if (conf->label_policy && !use_ipv6) {
3816 		NL_SET_ERR_MSG(extack,
3817 			       "Label policy only applies to IPv6 VXLAN devices");
3818 		return -EINVAL;
3819 	}
3820 
3821 	if (conf->remote_ifindex) {
3822 		struct net_device *lowerdev;
3823 
3824 		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3825 		if (!lowerdev) {
3826 			NL_SET_ERR_MSG(extack,
3827 				       "Invalid local interface, device not found");
3828 			return -ENODEV;
3829 		}
3830 
3831 #if IS_ENABLED(CONFIG_IPV6)
3832 		if (use_ipv6) {
3833 			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3834 
3835 			if (idev && idev->cnf.disable_ipv6) {
3836 				NL_SET_ERR_MSG(extack,
3837 					       "IPv6 support disabled by administrator");
3838 				return -EPERM;
3839 			}
3840 		}
3841 #endif
3842 
3843 		*lower = lowerdev;
3844 	} else {
3845 		if (vxlan_addr_multicast(&conf->remote_ip)) {
3846 			NL_SET_ERR_MSG(extack,
3847 				       "Local interface required for multicast remote destination");
3848 
3849 			return -EINVAL;
3850 		}
3851 
3852 #if IS_ENABLED(CONFIG_IPV6)
3853 		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3854 			NL_SET_ERR_MSG(extack,
3855 				       "Local interface required for link-local local/remote addresses");
3856 			return -EINVAL;
3857 		}
3858 #endif
3859 
3860 		*lower = NULL;
3861 	}
3862 
3863 	if (!conf->dst_port) {
3864 		if (conf->flags & VXLAN_F_GPE)
3865 			conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3866 		else
3867 			conf->dst_port = htons(vxlan_port);
3868 	}
3869 
3870 	if (!conf->age_interval)
3871 		conf->age_interval = FDB_AGE_DEFAULT;
3872 
3873 	if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3874 		NL_SET_ERR_MSG(extack,
3875 			       "A VXLAN device with the specified VNI already exists");
3876 		return -EEXIST;
3877 	}
3878 
3879 	return 0;
3880 }
3881 
vxlan_config_apply(struct net_device * dev,struct vxlan_config * conf,struct net_device * lowerdev,struct net * src_net,bool changelink)3882 static void vxlan_config_apply(struct net_device *dev,
3883 			       struct vxlan_config *conf,
3884 			       struct net_device *lowerdev,
3885 			       struct net *src_net,
3886 			       bool changelink)
3887 {
3888 	struct vxlan_dev *vxlan = netdev_priv(dev);
3889 	struct vxlan_rdst *dst = &vxlan->default_dst;
3890 	unsigned short needed_headroom = ETH_HLEN;
3891 	int max_mtu = ETH_MAX_MTU;
3892 	u32 flags = conf->flags;
3893 
3894 	if (!changelink) {
3895 		if (flags & VXLAN_F_GPE)
3896 			vxlan_raw_setup(dev);
3897 		else
3898 			vxlan_ether_setup(dev);
3899 
3900 		if (conf->mtu)
3901 			dev->mtu = conf->mtu;
3902 
3903 		vxlan->net = src_net;
3904 	}
3905 
3906 	dst->remote_vni = conf->vni;
3907 
3908 	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3909 
3910 	if (lowerdev) {
3911 		dst->remote_ifindex = conf->remote_ifindex;
3912 
3913 		netif_inherit_tso_max(dev, lowerdev);
3914 
3915 		needed_headroom = lowerdev->hard_header_len;
3916 		needed_headroom += lowerdev->needed_headroom;
3917 
3918 		dev->needed_tailroom = lowerdev->needed_tailroom;
3919 
3920 		max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3921 		if (max_mtu < ETH_MIN_MTU)
3922 			max_mtu = ETH_MIN_MTU;
3923 
3924 		if (!changelink && !conf->mtu)
3925 			dev->mtu = max_mtu;
3926 	}
3927 
3928 	if (dev->mtu > max_mtu)
3929 		dev->mtu = max_mtu;
3930 
3931 	if (flags & VXLAN_F_COLLECT_METADATA)
3932 		flags |= VXLAN_F_IPV6;
3933 	needed_headroom += vxlan_headroom(flags);
3934 	dev->needed_headroom = needed_headroom;
3935 
3936 	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3937 }
3938 
vxlan_dev_configure(struct net * src_net,struct net_device * dev,struct vxlan_config * conf,struct netlink_ext_ack * extack)3939 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3940 			       struct vxlan_config *conf,
3941 			       struct netlink_ext_ack *extack)
3942 {
3943 	struct vxlan_dev *vxlan = netdev_priv(dev);
3944 	struct net_device *lowerdev;
3945 	int ret;
3946 
3947 	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3948 	if (ret)
3949 		return ret;
3950 
3951 	vxlan_config_apply(dev, conf, lowerdev, src_net, false);
3952 
3953 	return 0;
3954 }
3955 
__vxlan_dev_create(struct net * net,struct net_device * dev,struct vxlan_config * conf,struct netlink_ext_ack * extack)3956 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3957 			      struct vxlan_config *conf,
3958 			      struct netlink_ext_ack *extack)
3959 {
3960 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3961 	struct vxlan_dev *vxlan = netdev_priv(dev);
3962 	struct net_device *remote_dev = NULL;
3963 	struct vxlan_fdb *f = NULL;
3964 	bool unregister = false;
3965 	struct vxlan_rdst *dst;
3966 	int err;
3967 
3968 	dst = &vxlan->default_dst;
3969 	err = vxlan_dev_configure(net, dev, conf, extack);
3970 	if (err)
3971 		return err;
3972 
3973 	dev->ethtool_ops = &vxlan_ethtool_ops;
3974 
3975 	/* create an fdb entry for a valid default destination */
3976 	if (!vxlan_addr_any(&dst->remote_ip)) {
3977 		err = vxlan_fdb_create(vxlan, all_zeros_mac,
3978 				       &dst->remote_ip,
3979 				       NUD_REACHABLE | NUD_PERMANENT,
3980 				       vxlan->cfg.dst_port,
3981 				       dst->remote_vni,
3982 				       dst->remote_vni,
3983 				       dst->remote_ifindex,
3984 				       NTF_SELF, 0, &f, extack);
3985 		if (err)
3986 			return err;
3987 	}
3988 
3989 	err = register_netdevice(dev);
3990 	if (err)
3991 		goto errout;
3992 	unregister = true;
3993 
3994 	if (dst->remote_ifindex) {
3995 		remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3996 		if (!remote_dev) {
3997 			err = -ENODEV;
3998 			goto errout;
3999 		}
4000 
4001 		err = netdev_upper_dev_link(remote_dev, dev, extack);
4002 		if (err)
4003 			goto errout;
4004 	}
4005 
4006 	err = rtnl_configure_link(dev, NULL, 0, NULL);
4007 	if (err < 0)
4008 		goto unlink;
4009 
4010 	if (f) {
4011 		vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
4012 
4013 		/* notify default fdb entry */
4014 		err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4015 				       RTM_NEWNEIGH, true, extack);
4016 		if (err) {
4017 			vxlan_fdb_destroy(vxlan, f, false, false);
4018 			if (remote_dev)
4019 				netdev_upper_dev_unlink(remote_dev, dev);
4020 			goto unregister;
4021 		}
4022 	}
4023 
4024 	list_add(&vxlan->next, &vn->vxlan_list);
4025 	if (remote_dev)
4026 		dst->remote_dev = remote_dev;
4027 	return 0;
4028 unlink:
4029 	if (remote_dev)
4030 		netdev_upper_dev_unlink(remote_dev, dev);
4031 errout:
4032 	/* unregister_netdevice() destroys the default FDB entry with deletion
4033 	 * notification. But the addition notification was not sent yet, so
4034 	 * destroy the entry by hand here.
4035 	 */
4036 	if (f)
4037 		__vxlan_fdb_free(f);
4038 unregister:
4039 	if (unregister)
4040 		unregister_netdevice(dev);
4041 	return err;
4042 }
4043 
4044 /* Set/clear flags based on attribute */
vxlan_nl2flag(struct vxlan_config * conf,struct nlattr * tb[],int attrtype,unsigned long mask,bool changelink,bool changelink_supported,struct netlink_ext_ack * extack)4045 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
4046 			  int attrtype, unsigned long mask, bool changelink,
4047 			  bool changelink_supported,
4048 			  struct netlink_ext_ack *extack)
4049 {
4050 	unsigned long flags;
4051 
4052 	if (!tb[attrtype])
4053 		return 0;
4054 
4055 	if (changelink && !changelink_supported) {
4056 		vxlan_flag_attr_error(attrtype, extack);
4057 		return -EOPNOTSUPP;
4058 	}
4059 
4060 	if (vxlan_policy[attrtype].type == NLA_FLAG)
4061 		flags = conf->flags | mask;
4062 	else if (nla_get_u8(tb[attrtype]))
4063 		flags = conf->flags | mask;
4064 	else
4065 		flags = conf->flags & ~mask;
4066 
4067 	conf->flags = flags;
4068 
4069 	return 0;
4070 }
4071 
vxlan_nl2conf(struct nlattr * tb[],struct nlattr * data[],struct net_device * dev,struct vxlan_config * conf,bool changelink,struct netlink_ext_ack * extack)4072 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4073 			 struct net_device *dev, struct vxlan_config *conf,
4074 			 bool changelink, struct netlink_ext_ack *extack)
4075 {
4076 	struct vxlanhdr used_bits = {
4077 		.vx_flags = VXLAN_HF_VNI,
4078 		.vx_vni = VXLAN_VNI_MASK,
4079 	};
4080 	struct vxlan_dev *vxlan = netdev_priv(dev);
4081 	int err = 0;
4082 
4083 	memset(conf, 0, sizeof(*conf));
4084 
4085 	/* if changelink operation, start with old existing cfg */
4086 	if (changelink)
4087 		memcpy(conf, &vxlan->cfg, sizeof(*conf));
4088 
4089 	if (data[IFLA_VXLAN_ID]) {
4090 		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4091 
4092 		if (changelink && (vni != conf->vni)) {
4093 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
4094 			return -EOPNOTSUPP;
4095 		}
4096 		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4097 	}
4098 
4099 	if (data[IFLA_VXLAN_GROUP]) {
4100 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4101 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4102 			return -EOPNOTSUPP;
4103 		}
4104 
4105 		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4106 		conf->remote_ip.sa.sa_family = AF_INET;
4107 	} else if (data[IFLA_VXLAN_GROUP6]) {
4108 		if (!IS_ENABLED(CONFIG_IPV6)) {
4109 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4110 			return -EPFNOSUPPORT;
4111 		}
4112 
4113 		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4114 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4115 			return -EOPNOTSUPP;
4116 		}
4117 
4118 		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4119 		conf->remote_ip.sa.sa_family = AF_INET6;
4120 	}
4121 
4122 	if (data[IFLA_VXLAN_LOCAL]) {
4123 		if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4124 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4125 			return -EOPNOTSUPP;
4126 		}
4127 
4128 		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4129 		conf->saddr.sa.sa_family = AF_INET;
4130 	} else if (data[IFLA_VXLAN_LOCAL6]) {
4131 		if (!IS_ENABLED(CONFIG_IPV6)) {
4132 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4133 			return -EPFNOSUPPORT;
4134 		}
4135 
4136 		if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4137 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4138 			return -EOPNOTSUPP;
4139 		}
4140 
4141 		/* TODO: respect scope id */
4142 		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4143 		conf->saddr.sa.sa_family = AF_INET6;
4144 	}
4145 
4146 	if (data[IFLA_VXLAN_LINK])
4147 		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4148 
4149 	if (data[IFLA_VXLAN_TOS])
4150 		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4151 
4152 	if (data[IFLA_VXLAN_TTL])
4153 		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4154 
4155 	if (data[IFLA_VXLAN_TTL_INHERIT]) {
4156 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4157 				    VXLAN_F_TTL_INHERIT, changelink, false,
4158 				    extack);
4159 		if (err)
4160 			return err;
4161 
4162 	}
4163 
4164 	if (data[IFLA_VXLAN_LABEL])
4165 		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4166 			     IPV6_FLOWLABEL_MASK;
4167 	if (data[IFLA_VXLAN_LABEL_POLICY])
4168 		conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]);
4169 
4170 	if (data[IFLA_VXLAN_LEARNING]) {
4171 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4172 				    VXLAN_F_LEARN, changelink, true,
4173 				    extack);
4174 		if (err)
4175 			return err;
4176 	} else if (!changelink) {
4177 		/* default to learn on a new device */
4178 		conf->flags |= VXLAN_F_LEARN;
4179 	}
4180 
4181 	if (data[IFLA_VXLAN_AGEING])
4182 		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4183 
4184 	if (data[IFLA_VXLAN_PROXY]) {
4185 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4186 				    VXLAN_F_PROXY, changelink, false,
4187 				    extack);
4188 		if (err)
4189 			return err;
4190 	}
4191 
4192 	if (data[IFLA_VXLAN_RSC]) {
4193 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4194 				    VXLAN_F_RSC, changelink, false,
4195 				    extack);
4196 		if (err)
4197 			return err;
4198 	}
4199 
4200 	if (data[IFLA_VXLAN_L2MISS]) {
4201 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4202 				    VXLAN_F_L2MISS, changelink, false,
4203 				    extack);
4204 		if (err)
4205 			return err;
4206 	}
4207 
4208 	if (data[IFLA_VXLAN_L3MISS]) {
4209 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4210 				    VXLAN_F_L3MISS, changelink, false,
4211 				    extack);
4212 		if (err)
4213 			return err;
4214 	}
4215 
4216 	if (data[IFLA_VXLAN_LIMIT]) {
4217 		if (changelink) {
4218 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4219 					    "Cannot change limit");
4220 			return -EOPNOTSUPP;
4221 		}
4222 		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4223 	}
4224 
4225 	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4226 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4227 				    VXLAN_F_COLLECT_METADATA, changelink, false,
4228 				    extack);
4229 		if (err)
4230 			return err;
4231 	}
4232 
4233 	if (data[IFLA_VXLAN_PORT_RANGE]) {
4234 		if (!changelink) {
4235 			const struct ifla_vxlan_port_range *p
4236 				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4237 			conf->port_min = ntohs(p->low);
4238 			conf->port_max = ntohs(p->high);
4239 		} else {
4240 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4241 					    "Cannot change port range");
4242 			return -EOPNOTSUPP;
4243 		}
4244 	}
4245 
4246 	if (data[IFLA_VXLAN_PORT]) {
4247 		if (changelink) {
4248 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4249 					    "Cannot change port");
4250 			return -EOPNOTSUPP;
4251 		}
4252 		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4253 	}
4254 
4255 	if (data[IFLA_VXLAN_UDP_CSUM]) {
4256 		if (changelink) {
4257 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4258 					    "Cannot change UDP_CSUM flag");
4259 			return -EOPNOTSUPP;
4260 		}
4261 		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4262 			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4263 	}
4264 
4265 	if (data[IFLA_VXLAN_LOCALBYPASS]) {
4266 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4267 				    VXLAN_F_LOCALBYPASS, changelink,
4268 				    true, extack);
4269 		if (err)
4270 			return err;
4271 	} else if (!changelink) {
4272 		/* default to local bypass on a new device */
4273 		conf->flags |= VXLAN_F_LOCALBYPASS;
4274 	}
4275 
4276 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4277 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4278 				    VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4279 				    false, extack);
4280 		if (err)
4281 			return err;
4282 	}
4283 
4284 	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4285 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4286 				    VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4287 				    false, extack);
4288 		if (err)
4289 			return err;
4290 	}
4291 
4292 	if (data[IFLA_VXLAN_REMCSUM_TX]) {
4293 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4294 				    VXLAN_F_REMCSUM_TX, changelink, false,
4295 				    extack);
4296 		if (err)
4297 			return err;
4298 	}
4299 
4300 	if (data[IFLA_VXLAN_REMCSUM_RX]) {
4301 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4302 				    VXLAN_F_REMCSUM_RX, changelink, false,
4303 				    extack);
4304 		if (err)
4305 			return err;
4306 		used_bits.vx_flags |= VXLAN_HF_RCO;
4307 		used_bits.vx_vni |= ~VXLAN_VNI_MASK;
4308 	}
4309 
4310 	if (data[IFLA_VXLAN_GBP]) {
4311 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4312 				    VXLAN_F_GBP, changelink, false, extack);
4313 		if (err)
4314 			return err;
4315 		used_bits.vx_flags |= VXLAN_GBP_USED_BITS;
4316 	}
4317 
4318 	if (data[IFLA_VXLAN_GPE]) {
4319 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4320 				    VXLAN_F_GPE, changelink, false,
4321 				    extack);
4322 		if (err)
4323 			return err;
4324 
4325 		used_bits.vx_flags |= VXLAN_GPE_USED_BITS;
4326 	}
4327 
4328 	if (data[IFLA_VXLAN_RESERVED_BITS]) {
4329 		struct vxlanhdr reserved_bits;
4330 
4331 		if (changelink) {
4332 			NL_SET_ERR_MSG_ATTR(extack,
4333 					    data[IFLA_VXLAN_RESERVED_BITS],
4334 					    "Cannot change reserved_bits");
4335 			return -EOPNOTSUPP;
4336 		}
4337 
4338 		nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS],
4339 			   sizeof(reserved_bits));
4340 		if (used_bits.vx_flags & reserved_bits.vx_flags ||
4341 		    used_bits.vx_vni & reserved_bits.vx_vni) {
4342 			__be64 ub_be64, rb_be64;
4343 
4344 			memcpy(&ub_be64, &used_bits, sizeof(ub_be64));
4345 			memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64));
4346 
4347 			NL_SET_ERR_MSG_ATTR_FMT(extack,
4348 						data[IFLA_VXLAN_RESERVED_BITS],
4349 						"Used bits %#018llx cannot overlap reserved bits %#018llx",
4350 						be64_to_cpu(ub_be64),
4351 						be64_to_cpu(rb_be64));
4352 			return -EINVAL;
4353 		}
4354 
4355 		conf->reserved_bits = reserved_bits;
4356 	} else {
4357 		/* For backwards compatibility, only allow reserved fields to be
4358 		 * used by VXLAN extensions if explicitly requested.
4359 		 */
4360 		conf->reserved_bits = (struct vxlanhdr) {
4361 			.vx_flags = ~used_bits.vx_flags,
4362 			.vx_vni = ~used_bits.vx_vni,
4363 		};
4364 	}
4365 
4366 	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4367 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4368 				    VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4369 				    false, extack);
4370 		if (err)
4371 			return err;
4372 	}
4373 
4374 	if (tb[IFLA_MTU]) {
4375 		if (changelink) {
4376 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4377 					    "Cannot change mtu");
4378 			return -EOPNOTSUPP;
4379 		}
4380 		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4381 	}
4382 
4383 	if (data[IFLA_VXLAN_DF])
4384 		conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4385 
4386 	if (data[IFLA_VXLAN_VNIFILTER]) {
4387 		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4388 				    VXLAN_F_VNIFILTER, changelink, false,
4389 				    extack);
4390 		if (err)
4391 			return err;
4392 
4393 		if ((conf->flags & VXLAN_F_VNIFILTER) &&
4394 		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4395 			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4396 					    "vxlan vnifilter only valid in collect metadata mode");
4397 			return -EINVAL;
4398 		}
4399 	}
4400 
4401 	return 0;
4402 }
4403 
vxlan_newlink(struct net_device * dev,struct rtnl_newlink_params * params,struct netlink_ext_ack * extack)4404 static int vxlan_newlink(struct net_device *dev,
4405 			 struct rtnl_newlink_params *params,
4406 			 struct netlink_ext_ack *extack)
4407 {
4408 	struct net *link_net = rtnl_newlink_link_net(params);
4409 	struct nlattr **data = params->data;
4410 	struct nlattr **tb = params->tb;
4411 	struct vxlan_config conf;
4412 	int err;
4413 
4414 	err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4415 	if (err)
4416 		return err;
4417 
4418 	return __vxlan_dev_create(link_net, dev, &conf, extack);
4419 }
4420 
vxlan_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)4421 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4422 			    struct nlattr *data[],
4423 			    struct netlink_ext_ack *extack)
4424 {
4425 	struct vxlan_dev *vxlan = netdev_priv(dev);
4426 	bool rem_ip_changed, change_igmp;
4427 	struct net_device *lowerdev;
4428 	struct vxlan_config conf;
4429 	struct vxlan_rdst *dst;
4430 	int err;
4431 
4432 	dst = &vxlan->default_dst;
4433 	err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4434 	if (err)
4435 		return err;
4436 
4437 	err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4438 				    vxlan, extack);
4439 	if (err)
4440 		return err;
4441 
4442 	if (dst->remote_dev == lowerdev)
4443 		lowerdev = NULL;
4444 
4445 	err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4446 					     extack);
4447 	if (err)
4448 		return err;
4449 
4450 	rem_ip_changed = !vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip);
4451 	change_igmp = vxlan->dev->flags & IFF_UP &&
4452 		      (rem_ip_changed ||
4453 		       dst->remote_ifindex != conf.remote_ifindex);
4454 
4455 	/* handle default dst entry */
4456 	if (rem_ip_changed) {
4457 		u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4458 
4459 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4460 		if (!vxlan_addr_any(&conf.remote_ip)) {
4461 			err = vxlan_fdb_update(vxlan, all_zeros_mac,
4462 					       &conf.remote_ip,
4463 					       NUD_REACHABLE | NUD_PERMANENT,
4464 					       NLM_F_APPEND | NLM_F_CREATE,
4465 					       vxlan->cfg.dst_port,
4466 					       conf.vni, conf.vni,
4467 					       conf.remote_ifindex,
4468 					       NTF_SELF, 0, true, extack);
4469 			if (err) {
4470 				spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4471 				netdev_adjacent_change_abort(dst->remote_dev,
4472 							     lowerdev, dev);
4473 				return err;
4474 			}
4475 		}
4476 		if (!vxlan_addr_any(&dst->remote_ip))
4477 			__vxlan_fdb_delete(vxlan, all_zeros_mac,
4478 					   dst->remote_ip,
4479 					   vxlan->cfg.dst_port,
4480 					   dst->remote_vni,
4481 					   dst->remote_vni,
4482 					   dst->remote_ifindex,
4483 					   true);
4484 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4485 
4486 		/* If vni filtering device, also update fdb entries of
4487 		 * all vnis that were using default remote ip
4488 		 */
4489 		if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4490 			err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4491 							 &conf.remote_ip, extack);
4492 			if (err) {
4493 				netdev_adjacent_change_abort(dst->remote_dev,
4494 							     lowerdev, dev);
4495 				return err;
4496 			}
4497 		}
4498 	}
4499 
4500 	if (change_igmp && vxlan_addr_multicast(&dst->remote_ip))
4501 		err = vxlan_multicast_leave(vxlan);
4502 
4503 	if (conf.age_interval != vxlan->cfg.age_interval)
4504 		mod_timer(&vxlan->age_timer, jiffies);
4505 
4506 	netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4507 	if (lowerdev && lowerdev != dst->remote_dev)
4508 		dst->remote_dev = lowerdev;
4509 	vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4510 
4511 	if (!err && change_igmp &&
4512 	    vxlan_addr_multicast(&dst->remote_ip))
4513 		err = vxlan_multicast_join(vxlan);
4514 
4515 	return err;
4516 }
4517 
vxlan_dellink(struct net_device * dev,struct list_head * head)4518 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4519 {
4520 	struct vxlan_dev *vxlan = netdev_priv(dev);
4521 	struct vxlan_fdb_flush_desc desc = {
4522 		/* Default entry is deleted at vxlan_uninit. */
4523 		.ignore_default_entry = true,
4524 	};
4525 
4526 	vxlan_flush(vxlan, &desc);
4527 
4528 	list_del(&vxlan->next);
4529 	unregister_netdevice_queue(dev, head);
4530 	if (vxlan->default_dst.remote_dev)
4531 		netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4532 }
4533 
vxlan_get_size(const struct net_device * dev)4534 static size_t vxlan_get_size(const struct net_device *dev)
4535 {
4536 	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
4537 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4538 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
4539 		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4540 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
4541 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL_INHERIT */
4542 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
4543 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_DF */
4544 		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4545 		nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_LABEL_POLICY */
4546 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
4547 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
4548 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
4549 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
4550 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
4551 		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
4552 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
4553 		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
4554 		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4555 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4556 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4557 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4558 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4559 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4560 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4561 		/* IFLA_VXLAN_PORT_RANGE */
4562 		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4563 		nla_total_size(0) + /* IFLA_VXLAN_GBP */
4564 		nla_total_size(0) + /* IFLA_VXLAN_GPE */
4565 		nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */
4566 		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */
4567 		/* IFLA_VXLAN_RESERVED_BITS */
4568 		nla_total_size(sizeof(struct vxlanhdr)) +
4569 		0;
4570 }
4571 
vxlan_fill_info(struct sk_buff * skb,const struct net_device * dev)4572 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4573 {
4574 	const struct vxlan_dev *vxlan = netdev_priv(dev);
4575 	const struct vxlan_rdst *dst = &vxlan->default_dst;
4576 	struct ifla_vxlan_port_range ports = {
4577 		.low =  htons(vxlan->cfg.port_min),
4578 		.high = htons(vxlan->cfg.port_max),
4579 	};
4580 
4581 	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4582 		goto nla_put_failure;
4583 
4584 	if (!vxlan_addr_any(&dst->remote_ip)) {
4585 		if (dst->remote_ip.sa.sa_family == AF_INET) {
4586 			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4587 					    dst->remote_ip.sin.sin_addr.s_addr))
4588 				goto nla_put_failure;
4589 #if IS_ENABLED(CONFIG_IPV6)
4590 		} else {
4591 			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4592 					     &dst->remote_ip.sin6.sin6_addr))
4593 				goto nla_put_failure;
4594 #endif
4595 		}
4596 	}
4597 
4598 	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4599 		goto nla_put_failure;
4600 
4601 	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4602 		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4603 			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4604 					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
4605 				goto nla_put_failure;
4606 #if IS_ENABLED(CONFIG_IPV6)
4607 		} else {
4608 			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4609 					     &vxlan->cfg.saddr.sin6.sin6_addr))
4610 				goto nla_put_failure;
4611 #endif
4612 		}
4613 	}
4614 
4615 	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4616 	    nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4617 		       !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4618 	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4619 	    nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4620 	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4621 	    nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) ||
4622 	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4623 		       !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4624 	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
4625 		       !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4626 	    nla_put_u8(skb, IFLA_VXLAN_RSC,
4627 		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4628 	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4629 		       !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4630 	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4631 		       !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4632 	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4633 		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4634 	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4635 	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4636 	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4637 	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4638 		       !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4639 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4640 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4641 	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4642 		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4643 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4644 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4645 	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4646 		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4647 	    nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4648 		       !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4649 		goto nla_put_failure;
4650 
4651 	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4652 		goto nla_put_failure;
4653 
4654 	if (vxlan->cfg.flags & VXLAN_F_GBP &&
4655 	    nla_put_flag(skb, IFLA_VXLAN_GBP))
4656 		goto nla_put_failure;
4657 
4658 	if (vxlan->cfg.flags & VXLAN_F_GPE &&
4659 	    nla_put_flag(skb, IFLA_VXLAN_GPE))
4660 		goto nla_put_failure;
4661 
4662 	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4663 	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4664 		goto nla_put_failure;
4665 
4666 	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4667 	    nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4668 		       !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4669 		goto nla_put_failure;
4670 
4671 	if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS,
4672 		    sizeof(vxlan->cfg.reserved_bits),
4673 		    &vxlan->cfg.reserved_bits))
4674 		goto nla_put_failure;
4675 
4676 	return 0;
4677 
4678 nla_put_failure:
4679 	return -EMSGSIZE;
4680 }
4681 
vxlan_get_link_net(const struct net_device * dev)4682 static struct net *vxlan_get_link_net(const struct net_device *dev)
4683 {
4684 	struct vxlan_dev *vxlan = netdev_priv(dev);
4685 
4686 	return READ_ONCE(vxlan->net);
4687 }
4688 
4689 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4690 	.kind		= "vxlan",
4691 	.maxtype	= IFLA_VXLAN_MAX,
4692 	.policy		= vxlan_policy,
4693 	.priv_size	= sizeof(struct vxlan_dev),
4694 	.setup		= vxlan_setup,
4695 	.validate	= vxlan_validate,
4696 	.newlink	= vxlan_newlink,
4697 	.changelink	= vxlan_changelink,
4698 	.dellink	= vxlan_dellink,
4699 	.get_size	= vxlan_get_size,
4700 	.fill_info	= vxlan_fill_info,
4701 	.get_link_net	= vxlan_get_link_net,
4702 };
4703 
vxlan_dev_create(struct net * net,const char * name,u8 name_assign_type,struct vxlan_config * conf)4704 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4705 				    u8 name_assign_type,
4706 				    struct vxlan_config *conf)
4707 {
4708 	struct nlattr *tb[IFLA_MAX + 1];
4709 	struct net_device *dev;
4710 	int err;
4711 
4712 	memset(&tb, 0, sizeof(tb));
4713 
4714 	dev = rtnl_create_link(net, name, name_assign_type,
4715 			       &vxlan_link_ops, tb, NULL);
4716 	if (IS_ERR(dev))
4717 		return dev;
4718 
4719 	err = __vxlan_dev_create(net, dev, conf, NULL);
4720 	if (err < 0) {
4721 		free_netdev(dev);
4722 		return ERR_PTR(err);
4723 	}
4724 
4725 	err = rtnl_configure_link(dev, NULL, 0, NULL);
4726 	if (err < 0) {
4727 		LIST_HEAD(list_kill);
4728 
4729 		vxlan_dellink(dev, &list_kill);
4730 		unregister_netdevice_many(&list_kill);
4731 		return ERR_PTR(err);
4732 	}
4733 
4734 	return dev;
4735 }
4736 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4737 
vxlan_handle_lowerdev_unregister(struct vxlan_net * vn,struct net_device * dev)4738 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4739 					     struct net_device *dev)
4740 {
4741 	struct vxlan_dev *vxlan, *next;
4742 	LIST_HEAD(list_kill);
4743 
4744 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4745 		struct vxlan_rdst *dst = &vxlan->default_dst;
4746 
4747 		/* In case we created vxlan device with carrier
4748 		 * and we loose the carrier due to module unload
4749 		 * we also need to remove vxlan device. In other
4750 		 * cases, it's not necessary and remote_ifindex
4751 		 * is 0 here, so no matches.
4752 		 */
4753 		if (dst->remote_ifindex == dev->ifindex)
4754 			vxlan_dellink(vxlan->dev, &list_kill);
4755 	}
4756 
4757 	unregister_netdevice_many(&list_kill);
4758 }
4759 
vxlan_netdevice_event(struct notifier_block * unused,unsigned long event,void * ptr)4760 static int vxlan_netdevice_event(struct notifier_block *unused,
4761 				 unsigned long event, void *ptr)
4762 {
4763 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4764 	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4765 
4766 	if (event == NETDEV_UNREGISTER)
4767 		vxlan_handle_lowerdev_unregister(vn, dev);
4768 	else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4769 		vxlan_offload_rx_ports(dev, true);
4770 	else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4771 		vxlan_offload_rx_ports(dev, false);
4772 
4773 	return NOTIFY_DONE;
4774 }
4775 
4776 static struct notifier_block vxlan_notifier_block __read_mostly = {
4777 	.notifier_call = vxlan_netdevice_event,
4778 };
4779 
4780 static void
vxlan_fdb_offloaded_set(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4781 vxlan_fdb_offloaded_set(struct net_device *dev,
4782 			struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4783 {
4784 	struct vxlan_dev *vxlan = netdev_priv(dev);
4785 	struct vxlan_rdst *rdst;
4786 	struct vxlan_fdb *f;
4787 	u32 hash_index;
4788 
4789 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4790 
4791 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4792 
4793 	f = __vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4794 	if (!f)
4795 		goto out;
4796 
4797 	rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4798 				   fdb_info->remote_port,
4799 				   fdb_info->remote_vni,
4800 				   fdb_info->remote_ifindex);
4801 	if (!rdst)
4802 		goto out;
4803 
4804 	rdst->offloaded = fdb_info->offloaded;
4805 
4806 out:
4807 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4808 }
4809 
4810 static int
vxlan_fdb_external_learn_add(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4811 vxlan_fdb_external_learn_add(struct net_device *dev,
4812 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4813 {
4814 	struct vxlan_dev *vxlan = netdev_priv(dev);
4815 	struct netlink_ext_ack *extack;
4816 	u32 hash_index;
4817 	int err;
4818 
4819 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4820 	extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4821 
4822 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4823 	err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4824 			       NUD_REACHABLE,
4825 			       NLM_F_CREATE | NLM_F_REPLACE,
4826 			       fdb_info->remote_port,
4827 			       fdb_info->vni,
4828 			       fdb_info->remote_vni,
4829 			       fdb_info->remote_ifindex,
4830 			       NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4831 			       0, false, extack);
4832 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4833 
4834 	return err;
4835 }
4836 
4837 static int
vxlan_fdb_external_learn_del(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4838 vxlan_fdb_external_learn_del(struct net_device *dev,
4839 			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4840 {
4841 	struct vxlan_dev *vxlan = netdev_priv(dev);
4842 	struct vxlan_fdb *f;
4843 	u32 hash_index;
4844 	int err = 0;
4845 
4846 	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4847 	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4848 
4849 	f = __vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4850 	if (!f)
4851 		err = -ENOENT;
4852 	else if (f->flags & NTF_EXT_LEARNED)
4853 		err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4854 					 fdb_info->remote_ip,
4855 					 fdb_info->remote_port,
4856 					 fdb_info->vni,
4857 					 fdb_info->remote_vni,
4858 					 fdb_info->remote_ifindex,
4859 					 false);
4860 
4861 	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4862 
4863 	return err;
4864 }
4865 
vxlan_switchdev_event(struct notifier_block * unused,unsigned long event,void * ptr)4866 static int vxlan_switchdev_event(struct notifier_block *unused,
4867 				 unsigned long event, void *ptr)
4868 {
4869 	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4870 	struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4871 	int err = 0;
4872 
4873 	switch (event) {
4874 	case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4875 		vxlan_fdb_offloaded_set(dev, ptr);
4876 		break;
4877 	case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4878 		fdb_info = ptr;
4879 		err = vxlan_fdb_external_learn_add(dev, fdb_info);
4880 		if (err) {
4881 			err = notifier_from_errno(err);
4882 			break;
4883 		}
4884 		fdb_info->offloaded = true;
4885 		vxlan_fdb_offloaded_set(dev, fdb_info);
4886 		break;
4887 	case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4888 		fdb_info = ptr;
4889 		err = vxlan_fdb_external_learn_del(dev, fdb_info);
4890 		if (err) {
4891 			err = notifier_from_errno(err);
4892 			break;
4893 		}
4894 		fdb_info->offloaded = false;
4895 		vxlan_fdb_offloaded_set(dev, fdb_info);
4896 		break;
4897 	}
4898 
4899 	return err;
4900 }
4901 
4902 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4903 	.notifier_call = vxlan_switchdev_event,
4904 };
4905 
vxlan_fdb_nh_flush(struct nexthop * nh)4906 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4907 {
4908 	struct vxlan_fdb *fdb;
4909 	struct vxlan_dev *vxlan;
4910 	u32 hash_index;
4911 
4912 	rcu_read_lock();
4913 	list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4914 		vxlan = rcu_dereference(fdb->vdev);
4915 		WARN_ON(!vxlan);
4916 		hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4917 					    vxlan->default_dst.remote_vni);
4918 		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4919 		if (!hlist_unhashed(&fdb->hlist))
4920 			vxlan_fdb_destroy(vxlan, fdb, false, false);
4921 		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4922 	}
4923 	rcu_read_unlock();
4924 }
4925 
vxlan_nexthop_event(struct notifier_block * nb,unsigned long event,void * ptr)4926 static int vxlan_nexthop_event(struct notifier_block *nb,
4927 			       unsigned long event, void *ptr)
4928 {
4929 	struct nh_notifier_info *info = ptr;
4930 	struct nexthop *nh;
4931 
4932 	if (event != NEXTHOP_EVENT_DEL)
4933 		return NOTIFY_DONE;
4934 
4935 	nh = nexthop_find_by_id(info->net, info->id);
4936 	if (!nh)
4937 		return NOTIFY_DONE;
4938 
4939 	vxlan_fdb_nh_flush(nh);
4940 
4941 	return NOTIFY_DONE;
4942 }
4943 
vxlan_init_net(struct net * net)4944 static __net_init int vxlan_init_net(struct net *net)
4945 {
4946 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4947 	unsigned int h;
4948 
4949 	INIT_LIST_HEAD(&vn->vxlan_list);
4950 	spin_lock_init(&vn->sock_lock);
4951 	vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4952 
4953 	for (h = 0; h < PORT_HASH_SIZE; ++h)
4954 		INIT_HLIST_HEAD(&vn->sock_list[h]);
4955 
4956 	return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4957 					 NULL);
4958 }
4959 
vxlan_destroy_tunnels(struct vxlan_net * vn,struct list_head * dev_to_kill)4960 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn,
4961 					     struct list_head *dev_to_kill)
4962 {
4963 	struct vxlan_dev *vxlan, *next;
4964 
4965 	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next)
4966 		vxlan_dellink(vxlan->dev, dev_to_kill);
4967 }
4968 
vxlan_exit_batch_rtnl(struct list_head * net_list,struct list_head * dev_to_kill)4969 static void __net_exit vxlan_exit_batch_rtnl(struct list_head *net_list,
4970 					     struct list_head *dev_to_kill)
4971 {
4972 	struct net *net;
4973 
4974 	ASSERT_RTNL();
4975 	list_for_each_entry(net, net_list, exit_list) {
4976 		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4977 
4978 		__unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4979 
4980 		vxlan_destroy_tunnels(vn, dev_to_kill);
4981 	}
4982 }
4983 
vxlan_exit_net(struct net * net)4984 static void __net_exit vxlan_exit_net(struct net *net)
4985 {
4986 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4987 	unsigned int h;
4988 
4989 	for (h = 0; h < PORT_HASH_SIZE; ++h)
4990 		WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4991 }
4992 
4993 static struct pernet_operations vxlan_net_ops = {
4994 	.init = vxlan_init_net,
4995 	.exit_batch_rtnl = vxlan_exit_batch_rtnl,
4996 	.exit = vxlan_exit_net,
4997 	.id   = &vxlan_net_id,
4998 	.size = sizeof(struct vxlan_net),
4999 };
5000 
vxlan_init_module(void)5001 static int __init vxlan_init_module(void)
5002 {
5003 	int rc;
5004 
5005 	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
5006 
5007 	rc = register_pernet_subsys(&vxlan_net_ops);
5008 	if (rc)
5009 		goto out1;
5010 
5011 	rc = register_netdevice_notifier(&vxlan_notifier_block);
5012 	if (rc)
5013 		goto out2;
5014 
5015 	rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
5016 	if (rc)
5017 		goto out3;
5018 
5019 	rc = rtnl_link_register(&vxlan_link_ops);
5020 	if (rc)
5021 		goto out4;
5022 
5023 	rc = vxlan_vnifilter_init();
5024 	if (rc)
5025 		goto out5;
5026 
5027 	return 0;
5028 out5:
5029 	rtnl_link_unregister(&vxlan_link_ops);
5030 out4:
5031 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
5032 out3:
5033 	unregister_netdevice_notifier(&vxlan_notifier_block);
5034 out2:
5035 	unregister_pernet_subsys(&vxlan_net_ops);
5036 out1:
5037 	return rc;
5038 }
5039 late_initcall(vxlan_init_module);
5040 
vxlan_cleanup_module(void)5041 static void __exit vxlan_cleanup_module(void)
5042 {
5043 	vxlan_vnifilter_uninit();
5044 	rtnl_link_unregister(&vxlan_link_ops);
5045 	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
5046 	unregister_netdevice_notifier(&vxlan_notifier_block);
5047 	unregister_pernet_subsys(&vxlan_net_ops);
5048 	/* rcu_barrier() is called by netns */
5049 }
5050 module_exit(vxlan_cleanup_module);
5051 
5052 MODULE_LICENSE("GPL");
5053 MODULE_VERSION(VXLAN_VERSION);
5054 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
5055 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
5056 MODULE_ALIAS_RTNL_LINK("vxlan");
5057