xref: /linux/net/ipv4/ip_tunnel_core.c (revision c715f13bb30f9f4d1bd8888667ef32e43b6fedc1)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (c) 2013 Nicira, Inc.
4  */
5 
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7 
8 #include <linux/types.h>
9 #include <linux/kernel.h>
10 #include <linux/skbuff.h>
11 #include <linux/netdevice.h>
12 #include <linux/in.h>
13 #include <linux/if_arp.h>
14 #include <linux/init.h>
15 #include <linux/in6.h>
16 #include <linux/inetdevice.h>
17 #include <linux/netfilter_ipv4.h>
18 #include <linux/etherdevice.h>
19 #include <linux/if_ether.h>
20 #include <linux/if_vlan.h>
21 #include <linux/static_key.h>
22 
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/protocol.h>
26 #include <net/ip_tunnels.h>
27 #include <net/ip6_tunnel.h>
28 #include <net/ip6_checksum.h>
29 #include <net/arp.h>
30 #include <net/checksum.h>
31 #include <net/dsfield.h>
32 #include <net/inet_ecn.h>
33 #include <net/xfrm.h>
34 #include <net/net_namespace.h>
35 #include <net/netns/generic.h>
36 #include <net/rtnetlink.h>
37 #include <net/dst_metadata.h>
38 #include <net/geneve.h>
39 #include <net/vxlan.h>
40 #include <net/erspan.h>
41 
42 const struct ip_tunnel_encap_ops __rcu *
43 		iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
44 EXPORT_SYMBOL(iptun_encaps);
45 
46 const struct ip6_tnl_encap_ops __rcu *
47 		ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
48 EXPORT_SYMBOL(ip6tun_encaps);
49 
50 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
51 		   __be32 src, __be32 dst, __u8 proto,
52 		   __u8 tos, __u8 ttl, __be16 df, bool xnet,
53 		   u16 ipcb_flags)
54 {
55 	int pkt_len = skb->len - skb_inner_network_offset(skb);
56 	struct net *net = dev_net(rt->dst.dev);
57 	struct net_device *dev = skb->dev;
58 	struct iphdr *iph;
59 	int err;
60 
61 	if (unlikely(dev_recursion_level() > IP_TUNNEL_RECURSION_LIMIT)) {
62 		if (dev) {
63 			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
64 					     dev->name);
65 			DEV_STATS_INC(dev, tx_errors);
66 		}
67 		ip_rt_put(rt);
68 		kfree_skb(skb);
69 		return;
70 	}
71 
72 	dev_xmit_recursion_inc();
73 
74 	skb_scrub_packet(skb, xnet);
75 
76 	skb_clear_hash_if_not_l4(skb);
77 	skb_dst_set(skb, &rt->dst);
78 	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
79 	IPCB(skb)->flags = ipcb_flags;
80 
81 	/* Push down and install the IP header. */
82 	skb_push(skb, sizeof(struct iphdr));
83 	skb_reset_network_header(skb);
84 
85 	iph = ip_hdr(skb);
86 
87 	iph->version	=	4;
88 	iph->ihl	=	sizeof(struct iphdr) >> 2;
89 	iph->frag_off	=	ip_mtu_locked(&rt->dst) ? 0 : df;
90 	iph->protocol	=	proto;
91 	iph->tos	=	tos;
92 	iph->daddr	=	dst;
93 	iph->saddr	=	src;
94 	iph->ttl	=	ttl;
95 	__ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
96 
97 	err = ip_local_out(net, sk, skb);
98 
99 	if (dev) {
100 		if (unlikely(net_xmit_eval(err)))
101 			pkt_len = 0;
102 		iptunnel_xmit_stats(dev, pkt_len);
103 	}
104 
105 	dev_xmit_recursion_dec();
106 }
107 EXPORT_SYMBOL_GPL(iptunnel_xmit);
108 
109 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
110 			   __be16 inner_proto, bool raw_proto, bool xnet)
111 {
112 	if (unlikely(!pskb_may_pull(skb, hdr_len)))
113 		return -ENOMEM;
114 
115 	skb_pull_rcsum(skb, hdr_len);
116 
117 	if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
118 		struct ethhdr *eh;
119 
120 		if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
121 			return -ENOMEM;
122 
123 		eh = (struct ethhdr *)skb->data;
124 		if (likely(eth_proto_is_802_3(eh->h_proto)))
125 			skb->protocol = eh->h_proto;
126 		else
127 			skb->protocol = htons(ETH_P_802_2);
128 
129 	} else {
130 		skb->protocol = inner_proto;
131 	}
132 
133 	skb_clear_hash_if_not_l4(skb);
134 	__vlan_hwaccel_clear_tag(skb);
135 	skb_set_queue_mapping(skb, 0);
136 	skb_scrub_packet(skb, xnet);
137 
138 	return iptunnel_pull_offloads(skb);
139 }
140 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
141 
142 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
143 					     gfp_t flags)
144 {
145 	IP_TUNNEL_DECLARE_FLAGS(tun_flags) = { };
146 	struct metadata_dst *res;
147 	struct ip_tunnel_info *dst, *src;
148 
149 	if (!md || md->type != METADATA_IP_TUNNEL ||
150 	    md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
151 		return NULL;
152 
153 	src = &md->u.tun_info;
154 	res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
155 	if (!res)
156 		return NULL;
157 
158 	dst = &res->u.tun_info;
159 	dst->key.tun_id = src->key.tun_id;
160 	if (src->mode & IP_TUNNEL_INFO_IPV6)
161 		memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
162 		       sizeof(struct in6_addr));
163 	else
164 		dst->key.u.ipv4.dst = src->key.u.ipv4.src;
165 	ip_tunnel_flags_copy(dst->key.tun_flags, src->key.tun_flags);
166 	dst->mode = src->mode | IP_TUNNEL_INFO_TX;
167 	ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
168 				src->options_len, tun_flags);
169 
170 	return res;
171 }
172 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
173 
174 int iptunnel_handle_offloads(struct sk_buff *skb,
175 			     int gso_type_mask)
176 {
177 	int err;
178 
179 	if (likely(!skb->encapsulation)) {
180 		skb_reset_inner_headers(skb);
181 		skb->encapsulation = 1;
182 	}
183 
184 	if (skb_is_gso(skb)) {
185 		err = skb_header_unclone(skb, GFP_ATOMIC);
186 		if (unlikely(err))
187 			return err;
188 		skb_shinfo(skb)->gso_type |= gso_type_mask;
189 		return 0;
190 	}
191 
192 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
193 		skb->ip_summed = CHECKSUM_NONE;
194 		/* We clear encapsulation here to prevent badly-written
195 		 * drivers potentially deciding to offload an inner checksum
196 		 * if we set CHECKSUM_PARTIAL on the outer header.
197 		 * This should go away when the drivers are all fixed.
198 		 */
199 		skb->encapsulation = 0;
200 	}
201 
202 	return 0;
203 }
204 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
205 
206 /**
207  * iptunnel_pmtud_build_icmp() - Build ICMP error message for PMTUD
208  * @skb:	Original packet with L2 header
209  * @mtu:	MTU value for ICMP error
210  *
211  * Return: length on success, negative error code if message couldn't be built.
212  */
213 static int iptunnel_pmtud_build_icmp(struct sk_buff *skb, int mtu)
214 {
215 	const struct iphdr *iph = ip_hdr(skb);
216 	struct icmphdr *icmph;
217 	struct iphdr *niph;
218 	struct ethhdr eh;
219 	int len, err;
220 
221 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct iphdr)))
222 		return -EINVAL;
223 
224 	if (skb_is_gso(skb))
225 		skb_gso_reset(skb);
226 
227 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
228 	pskb_pull(skb, ETH_HLEN);
229 	skb_reset_network_header(skb);
230 
231 	err = pskb_trim(skb, 576 - sizeof(*niph) - sizeof(*icmph));
232 	if (err)
233 		return err;
234 
235 	len = skb->len + sizeof(*icmph);
236 	err = skb_cow(skb, sizeof(*niph) + sizeof(*icmph) + ETH_HLEN);
237 	if (err)
238 		return err;
239 
240 	icmph = skb_push(skb, sizeof(*icmph));
241 	*icmph = (struct icmphdr) {
242 		.type			= ICMP_DEST_UNREACH,
243 		.code			= ICMP_FRAG_NEEDED,
244 		.checksum		= 0,
245 		.un.frag.__unused	= 0,
246 		.un.frag.mtu		= htons(mtu),
247 	};
248 	icmph->checksum = csum_fold(skb_checksum(skb, 0, len, 0));
249 	skb_reset_transport_header(skb);
250 
251 	niph = skb_push(skb, sizeof(*niph));
252 	*niph = (struct iphdr) {
253 		.ihl			= sizeof(*niph) / 4u,
254 		.version 		= 4,
255 		.tos 			= 0,
256 		.tot_len		= htons(len + sizeof(*niph)),
257 		.id			= 0,
258 		.frag_off		= htons(IP_DF),
259 		.ttl			= iph->ttl,
260 		.protocol		= IPPROTO_ICMP,
261 		.saddr			= iph->daddr,
262 		.daddr			= iph->saddr,
263 	};
264 	ip_send_check(niph);
265 	skb_reset_network_header(skb);
266 
267 	skb->ip_summed = CHECKSUM_NONE;
268 
269 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
270 	skb_reset_mac_header(skb);
271 
272 	return skb->len;
273 }
274 
275 /**
276  * iptunnel_pmtud_check_icmp() - Trigger ICMP reply if needed and allowed
277  * @skb:	Buffer being sent by encapsulation, L2 headers expected
278  * @mtu:	Network MTU for path
279  *
280  * Return: 0 for no ICMP reply, length if built, negative value on error.
281  */
282 static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
283 {
284 	const struct icmphdr *icmph = icmp_hdr(skb);
285 	const struct iphdr *iph = ip_hdr(skb);
286 
287 	if (mtu < 576 || iph->frag_off != htons(IP_DF))
288 		return 0;
289 
290 	if (ipv4_is_lbcast(iph->daddr)  || ipv4_is_multicast(iph->daddr) ||
291 	    ipv4_is_zeronet(iph->saddr) || ipv4_is_loopback(iph->saddr)  ||
292 	    ipv4_is_lbcast(iph->saddr)  || ipv4_is_multicast(iph->saddr))
293 		return 0;
294 
295 	if (iph->protocol == IPPROTO_ICMP && icmp_is_err(icmph->type))
296 		return 0;
297 
298 	return iptunnel_pmtud_build_icmp(skb, mtu);
299 }
300 
301 #if IS_ENABLED(CONFIG_IPV6)
302 /**
303  * iptunnel_pmtud_build_icmpv6() - Build ICMPv6 error message for PMTUD
304  * @skb:	Original packet with L2 header
305  * @mtu:	MTU value for ICMPv6 error
306  *
307  * Return: length on success, negative error code if message couldn't be built.
308  */
309 static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
310 {
311 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
312 	struct icmp6hdr *icmp6h;
313 	struct ipv6hdr *nip6h;
314 	struct ethhdr eh;
315 	int len, err;
316 	__wsum csum;
317 
318 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
319 		return -EINVAL;
320 
321 	if (skb_is_gso(skb))
322 		skb_gso_reset(skb);
323 
324 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
325 	pskb_pull(skb, ETH_HLEN);
326 	skb_reset_network_header(skb);
327 
328 	err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
329 	if (err)
330 		return err;
331 
332 	len = skb->len + sizeof(*icmp6h);
333 	err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
334 	if (err)
335 		return err;
336 
337 	icmp6h = skb_push(skb, sizeof(*icmp6h));
338 	*icmp6h = (struct icmp6hdr) {
339 		.icmp6_type		= ICMPV6_PKT_TOOBIG,
340 		.icmp6_code		= 0,
341 		.icmp6_cksum		= 0,
342 		.icmp6_mtu		= htonl(mtu),
343 	};
344 	skb_reset_transport_header(skb);
345 
346 	nip6h = skb_push(skb, sizeof(*nip6h));
347 	*nip6h = (struct ipv6hdr) {
348 		.priority		= 0,
349 		.version		= 6,
350 		.flow_lbl		= { 0 },
351 		.payload_len		= htons(len),
352 		.nexthdr		= IPPROTO_ICMPV6,
353 		.hop_limit		= ip6h->hop_limit,
354 		.saddr			= ip6h->daddr,
355 		.daddr			= ip6h->saddr,
356 	};
357 	skb_reset_network_header(skb);
358 
359 	csum = skb_checksum(skb, skb_transport_offset(skb), len, 0);
360 	icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
361 					      IPPROTO_ICMPV6, csum);
362 
363 	skb->ip_summed = CHECKSUM_NONE;
364 
365 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
366 	skb_reset_mac_header(skb);
367 
368 	return skb->len;
369 }
370 
371 /**
372  * iptunnel_pmtud_check_icmpv6() - Trigger ICMPv6 reply if needed and allowed
373  * @skb:	Buffer being sent by encapsulation, L2 headers expected
374  * @mtu:	Network MTU for path
375  *
376  * Return: 0 for no ICMPv6 reply, length if built, negative value on error.
377  */
378 static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
379 {
380 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
381 	int stype = ipv6_addr_type(&ip6h->saddr);
382 	u8 proto = ip6h->nexthdr;
383 	__be16 frag_off;
384 	int offset;
385 
386 	if (mtu < IPV6_MIN_MTU)
387 		return 0;
388 
389 	if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
390 	    stype == IPV6_ADDR_LOOPBACK)
391 		return 0;
392 
393 	offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
394 				  &frag_off);
395 	if (offset < 0 || (frag_off & htons(~0x7)))
396 		return 0;
397 
398 	if (proto == IPPROTO_ICMPV6) {
399 		struct icmp6hdr *icmp6h;
400 
401 		if (!pskb_may_pull(skb, skb_network_header(skb) +
402 					offset + 1 - skb->data))
403 			return 0;
404 
405 		icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
406 		if (icmpv6_is_err(icmp6h->icmp6_type) ||
407 		    icmp6h->icmp6_type == NDISC_REDIRECT)
408 			return 0;
409 	}
410 
411 	return iptunnel_pmtud_build_icmpv6(skb, mtu);
412 }
413 #endif /* IS_ENABLED(CONFIG_IPV6) */
414 
415 /**
416  * skb_tunnel_check_pmtu() - Check, update PMTU and trigger ICMP reply as needed
417  * @skb:	Buffer being sent by encapsulation, L2 headers expected
418  * @encap_dst:	Destination for tunnel encapsulation (outer IP)
419  * @headroom:	Encapsulation header size, bytes
420  * @reply:	Build matching ICMP or ICMPv6 message as a result
421  *
422  * L2 tunnel implementations that can carry IP and can be directly bridged
423  * (currently UDP tunnels) can't always rely on IP forwarding paths to handle
424  * PMTU discovery. In the bridged case, ICMP or ICMPv6 messages need to be built
425  * based on payload and sent back by the encapsulation itself.
426  *
427  * For routable interfaces, we just need to update the PMTU for the destination.
428  *
429  * Return: 0 if ICMP error not needed, length if built, negative value on error
430  */
431 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
432 			  int headroom, bool reply)
433 {
434 	u32 mtu = dst_mtu(encap_dst) - headroom;
435 
436 	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
437 	    (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
438 		return 0;
439 
440 	skb_dst_update_pmtu_no_confirm(skb, mtu);
441 
442 	if (!reply)
443 		return 0;
444 
445 	if (skb->protocol == htons(ETH_P_IP))
446 		return iptunnel_pmtud_check_icmp(skb, mtu);
447 
448 #if IS_ENABLED(CONFIG_IPV6)
449 	if (skb->protocol == htons(ETH_P_IPV6))
450 		return iptunnel_pmtud_check_icmpv6(skb, mtu);
451 #endif
452 	return 0;
453 }
454 EXPORT_SYMBOL(skb_tunnel_check_pmtu);
455 
456 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
457 	[LWTUNNEL_IP_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP_OPTS },
458 	[LWTUNNEL_IP_ID]	= { .type = NLA_U64 },
459 	[LWTUNNEL_IP_DST]	= { .type = NLA_U32 },
460 	[LWTUNNEL_IP_SRC]	= { .type = NLA_U32 },
461 	[LWTUNNEL_IP_TTL]	= { .type = NLA_U8 },
462 	[LWTUNNEL_IP_TOS]	= { .type = NLA_U8 },
463 	[LWTUNNEL_IP_FLAGS]	= { .type = NLA_U16 },
464 	[LWTUNNEL_IP_OPTS]	= { .type = NLA_NESTED },
465 };
466 
467 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
468 	[LWTUNNEL_IP_OPTS_GENEVE]	= { .type = NLA_NESTED },
469 	[LWTUNNEL_IP_OPTS_VXLAN]	= { .type = NLA_NESTED },
470 	[LWTUNNEL_IP_OPTS_ERSPAN]	= { .type = NLA_NESTED },
471 };
472 
473 static const struct nla_policy
474 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
475 	[LWTUNNEL_IP_OPT_GENEVE_CLASS]	= { .type = NLA_U16 },
476 	[LWTUNNEL_IP_OPT_GENEVE_TYPE]	= { .type = NLA_U8 },
477 	[LWTUNNEL_IP_OPT_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 127 },
478 };
479 
480 static const struct nla_policy
481 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
482 	[LWTUNNEL_IP_OPT_VXLAN_GBP]	= { .type = NLA_U32 },
483 };
484 
485 static const struct nla_policy
486 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
487 	[LWTUNNEL_IP_OPT_ERSPAN_VER]	= { .type = NLA_U8 },
488 	[LWTUNNEL_IP_OPT_ERSPAN_INDEX]	= { .type = NLA_U32 },
489 	[LWTUNNEL_IP_OPT_ERSPAN_DIR]	= { .type = NLA_U8 },
490 	[LWTUNNEL_IP_OPT_ERSPAN_HWID]	= { .type = NLA_U8 },
491 };
492 
493 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
494 				    struct ip_tunnel_info *info, int opts_len,
495 				    struct netlink_ext_ack *extack)
496 {
497 	struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
498 	int data_len, err;
499 
500 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
501 			       geneve_opt_policy, extack);
502 	if (err)
503 		return err;
504 
505 	if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
506 	    !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
507 	    !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
508 		return -EINVAL;
509 
510 	attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
511 	data_len = nla_len(attr);
512 	if (data_len % 4)
513 		return -EINVAL;
514 
515 	if (info) {
516 		struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
517 
518 		memcpy(opt->opt_data, nla_data(attr), data_len);
519 		opt->length = data_len / 4;
520 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
521 		opt->opt_class = nla_get_be16(attr);
522 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
523 		opt->type = nla_get_u8(attr);
524 		__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags);
525 	}
526 
527 	return sizeof(struct geneve_opt) + data_len;
528 }
529 
530 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
531 				   struct ip_tunnel_info *info, int opts_len,
532 				   struct netlink_ext_ack *extack)
533 {
534 	struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
535 	int err;
536 
537 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
538 			       vxlan_opt_policy, extack);
539 	if (err)
540 		return err;
541 
542 	if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
543 		return -EINVAL;
544 
545 	if (info) {
546 		struct vxlan_metadata *md =
547 			ip_tunnel_info_opts(info) + opts_len;
548 
549 		attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
550 		md->gbp = nla_get_u32(attr);
551 		md->gbp &= VXLAN_GBP_MASK;
552 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags);
553 	}
554 
555 	return sizeof(struct vxlan_metadata);
556 }
557 
558 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
559 				    struct ip_tunnel_info *info, int opts_len,
560 				    struct netlink_ext_ack *extack)
561 {
562 	struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
563 	int err;
564 	u8 ver;
565 
566 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
567 			       erspan_opt_policy, extack);
568 	if (err)
569 		return err;
570 
571 	if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
572 		return -EINVAL;
573 
574 	ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
575 	if (ver == 1) {
576 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
577 			return -EINVAL;
578 	} else if (ver == 2) {
579 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
580 		    !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
581 			return -EINVAL;
582 	} else {
583 		return -EINVAL;
584 	}
585 
586 	if (info) {
587 		struct erspan_metadata *md =
588 			ip_tunnel_info_opts(info) + opts_len;
589 
590 		md->version = ver;
591 		if (ver == 1) {
592 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
593 			md->u.index = nla_get_be32(attr);
594 		} else {
595 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
596 			md->u.md2.dir = nla_get_u8(attr);
597 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
598 			set_hwid(&md->u.md2, nla_get_u8(attr));
599 		}
600 
601 		__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags);
602 	}
603 
604 	return sizeof(struct erspan_metadata);
605 }
606 
607 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
608 			     struct netlink_ext_ack *extack)
609 {
610 	int err, rem, opt_len, opts_len = 0;
611 	struct nlattr *nla;
612 	u32 type = 0;
613 
614 	if (!attr)
615 		return 0;
616 
617 	err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
618 			   ip_opts_policy, extack);
619 	if (err)
620 		return err;
621 
622 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
623 		switch (nla_type(nla)) {
624 		case LWTUNNEL_IP_OPTS_GENEVE:
625 			if (type && type != IP_TUNNEL_GENEVE_OPT_BIT)
626 				return -EINVAL;
627 			opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
628 							   extack);
629 			if (opt_len < 0)
630 				return opt_len;
631 			opts_len += opt_len;
632 			if (opts_len > IP_TUNNEL_OPTS_MAX)
633 				return -EINVAL;
634 			type = IP_TUNNEL_GENEVE_OPT_BIT;
635 			break;
636 		case LWTUNNEL_IP_OPTS_VXLAN:
637 			if (type)
638 				return -EINVAL;
639 			opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
640 							  extack);
641 			if (opt_len < 0)
642 				return opt_len;
643 			opts_len += opt_len;
644 			type = IP_TUNNEL_VXLAN_OPT_BIT;
645 			break;
646 		case LWTUNNEL_IP_OPTS_ERSPAN:
647 			if (type)
648 				return -EINVAL;
649 			opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
650 							   extack);
651 			if (opt_len < 0)
652 				return opt_len;
653 			opts_len += opt_len;
654 			type = IP_TUNNEL_ERSPAN_OPT_BIT;
655 			break;
656 		default:
657 			return -EINVAL;
658 		}
659 	}
660 
661 	return opts_len;
662 }
663 
664 static int ip_tun_get_optlen(struct nlattr *attr,
665 			     struct netlink_ext_ack *extack)
666 {
667 	return ip_tun_parse_opts(attr, NULL, extack);
668 }
669 
670 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
671 			   struct netlink_ext_ack *extack)
672 {
673 	return ip_tun_parse_opts(attr, info, extack);
674 }
675 
676 static int ip_tun_build_state(struct net *net, struct nlattr *attr,
677 			      unsigned int family, const void *cfg,
678 			      struct lwtunnel_state **ts,
679 			      struct netlink_ext_ack *extack)
680 {
681 	struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
682 	struct lwtunnel_state *new_state;
683 	struct ip_tunnel_info *tun_info;
684 	int err, opt_len;
685 
686 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
687 					  ip_tun_policy, extack);
688 	if (err < 0)
689 		return err;
690 
691 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
692 	if (opt_len < 0)
693 		return opt_len;
694 
695 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
696 	if (!new_state)
697 		return -ENOMEM;
698 
699 	new_state->type = LWTUNNEL_ENCAP_IP;
700 
701 	tun_info = lwt_tun_info(new_state);
702 
703 	err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
704 	if (err < 0) {
705 		lwtstate_free(new_state);
706 		return err;
707 	}
708 
709 #ifdef CONFIG_DST_CACHE
710 	err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
711 	if (err) {
712 		lwtstate_free(new_state);
713 		return err;
714 	}
715 #endif
716 
717 	if (tb[LWTUNNEL_IP_ID])
718 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
719 
720 	if (tb[LWTUNNEL_IP_DST])
721 		tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
722 
723 	if (tb[LWTUNNEL_IP_SRC])
724 		tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
725 
726 	if (tb[LWTUNNEL_IP_TTL])
727 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
728 
729 	if (tb[LWTUNNEL_IP_TOS])
730 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
731 
732 	if (tb[LWTUNNEL_IP_FLAGS]) {
733 		IP_TUNNEL_DECLARE_FLAGS(flags);
734 
735 		ip_tunnel_flags_from_be16(flags,
736 					  nla_get_be16(tb[LWTUNNEL_IP_FLAGS]));
737 		ip_tunnel_clear_options_present(flags);
738 
739 		ip_tunnel_flags_or(tun_info->key.tun_flags,
740 				   tun_info->key.tun_flags, flags);
741 	}
742 
743 	tun_info->mode = IP_TUNNEL_INFO_TX;
744 	tun_info->options_len = opt_len;
745 
746 	*ts = new_state;
747 
748 	return 0;
749 }
750 
751 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
752 {
753 #ifdef CONFIG_DST_CACHE
754 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
755 
756 	dst_cache_destroy(&tun_info->dst_cache);
757 #endif
758 }
759 
760 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
761 					 struct ip_tunnel_info *tun_info)
762 {
763 	struct geneve_opt *opt;
764 	struct nlattr *nest;
765 	int offset = 0;
766 
767 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
768 	if (!nest)
769 		return -ENOMEM;
770 
771 	while (tun_info->options_len > offset) {
772 		opt = ip_tunnel_info_opts(tun_info) + offset;
773 		if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
774 				 opt->opt_class) ||
775 		    nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
776 		    nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
777 			    opt->opt_data)) {
778 			nla_nest_cancel(skb, nest);
779 			return -ENOMEM;
780 		}
781 		offset += sizeof(*opt) + opt->length * 4;
782 	}
783 
784 	nla_nest_end(skb, nest);
785 	return 0;
786 }
787 
788 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
789 					struct ip_tunnel_info *tun_info)
790 {
791 	struct vxlan_metadata *md;
792 	struct nlattr *nest;
793 
794 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
795 	if (!nest)
796 		return -ENOMEM;
797 
798 	md = ip_tunnel_info_opts(tun_info);
799 	if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
800 		nla_nest_cancel(skb, nest);
801 		return -ENOMEM;
802 	}
803 
804 	nla_nest_end(skb, nest);
805 	return 0;
806 }
807 
808 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
809 					 struct ip_tunnel_info *tun_info)
810 {
811 	struct erspan_metadata *md;
812 	struct nlattr *nest;
813 
814 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
815 	if (!nest)
816 		return -ENOMEM;
817 
818 	md = ip_tunnel_info_opts(tun_info);
819 	if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
820 		goto err;
821 
822 	if (md->version == 1 &&
823 	    nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
824 		goto err;
825 
826 	if (md->version == 2 &&
827 	    (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
828 	     nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
829 			get_hwid(&md->u.md2))))
830 		goto err;
831 
832 	nla_nest_end(skb, nest);
833 	return 0;
834 err:
835 	nla_nest_cancel(skb, nest);
836 	return -ENOMEM;
837 }
838 
839 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
840 				  struct ip_tunnel_info *tun_info)
841 {
842 	struct nlattr *nest;
843 	int err = 0;
844 
845 	if (!ip_tunnel_is_options_present(tun_info->key.tun_flags))
846 		return 0;
847 
848 	nest = nla_nest_start_noflag(skb, type);
849 	if (!nest)
850 		return -ENOMEM;
851 
852 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_info->key.tun_flags))
853 		err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
854 	else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_info->key.tun_flags))
855 		err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
856 	else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
857 		err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
858 
859 	if (err) {
860 		nla_nest_cancel(skb, nest);
861 		return err;
862 	}
863 
864 	nla_nest_end(skb, nest);
865 	return 0;
866 }
867 
868 static int ip_tun_fill_encap_info(struct sk_buff *skb,
869 				  struct lwtunnel_state *lwtstate)
870 {
871 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
872 
873 	if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
874 			 LWTUNNEL_IP_PAD) ||
875 	    nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
876 	    nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
877 	    nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
878 	    nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
879 	    nla_put_be16(skb, LWTUNNEL_IP_FLAGS,
880 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
881 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
882 		return -ENOMEM;
883 
884 	return 0;
885 }
886 
887 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
888 {
889 	int opt_len;
890 
891 	if (!ip_tunnel_is_options_present(info->key.tun_flags))
892 		return 0;
893 
894 	opt_len = nla_total_size(0);		/* LWTUNNEL_IP_OPTS */
895 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags)) {
896 		struct geneve_opt *opt;
897 		int offset = 0;
898 
899 		opt_len += nla_total_size(0);	/* LWTUNNEL_IP_OPTS_GENEVE */
900 		while (info->options_len > offset) {
901 			opt = ip_tunnel_info_opts(info) + offset;
902 			opt_len += nla_total_size(2)	/* OPT_GENEVE_CLASS */
903 				   + nla_total_size(1)	/* OPT_GENEVE_TYPE */
904 				   + nla_total_size(opt->length * 4);
905 							/* OPT_GENEVE_DATA */
906 			offset += sizeof(*opt) + opt->length * 4;
907 		}
908 	} else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
909 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_VXLAN */
910 			   + nla_total_size(4);	/* OPT_VXLAN_GBP */
911 	} else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags)) {
912 		struct erspan_metadata *md = ip_tunnel_info_opts(info);
913 
914 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_ERSPAN */
915 			   + nla_total_size(1)	/* OPT_ERSPAN_VER */
916 			   + (md->version == 1 ? nla_total_size(4)
917 						/* OPT_ERSPAN_INDEX (v1) */
918 					       : nla_total_size(1) +
919 						 nla_total_size(1));
920 						/* OPT_ERSPAN_DIR + HWID (v2) */
921 	}
922 
923 	return opt_len;
924 }
925 
926 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
927 {
928 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP_ID */
929 		+ nla_total_size(4)	/* LWTUNNEL_IP_DST */
930 		+ nla_total_size(4)	/* LWTUNNEL_IP_SRC */
931 		+ nla_total_size(1)	/* LWTUNNEL_IP_TOS */
932 		+ nla_total_size(1)	/* LWTUNNEL_IP_TTL */
933 		+ nla_total_size(2)	/* LWTUNNEL_IP_FLAGS */
934 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
935 					/* LWTUNNEL_IP_OPTS */
936 }
937 
938 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
939 {
940 	struct ip_tunnel_info *info_a = lwt_tun_info(a);
941 	struct ip_tunnel_info *info_b = lwt_tun_info(b);
942 
943 	return memcmp(info_a, info_b, sizeof(info_a->key)) ||
944 	       info_a->mode != info_b->mode ||
945 	       info_a->options_len != info_b->options_len ||
946 	       memcmp(ip_tunnel_info_opts(info_a),
947 		      ip_tunnel_info_opts(info_b), info_a->options_len);
948 }
949 
950 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
951 	.build_state = ip_tun_build_state,
952 	.destroy_state = ip_tun_destroy_state,
953 	.fill_encap = ip_tun_fill_encap_info,
954 	.get_encap_size = ip_tun_encap_nlsize,
955 	.cmp_encap = ip_tun_cmp_encap,
956 	.owner = THIS_MODULE,
957 };
958 
959 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
960 	[LWTUNNEL_IP6_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP6_OPTS },
961 	[LWTUNNEL_IP6_ID]		= { .type = NLA_U64 },
962 	[LWTUNNEL_IP6_DST]		= { .len = sizeof(struct in6_addr) },
963 	[LWTUNNEL_IP6_SRC]		= { .len = sizeof(struct in6_addr) },
964 	[LWTUNNEL_IP6_HOPLIMIT]		= { .type = NLA_U8 },
965 	[LWTUNNEL_IP6_TC]		= { .type = NLA_U8 },
966 	[LWTUNNEL_IP6_FLAGS]		= { .type = NLA_U16 },
967 	[LWTUNNEL_IP6_OPTS]		= { .type = NLA_NESTED },
968 };
969 
970 static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
971 			       unsigned int family, const void *cfg,
972 			       struct lwtunnel_state **ts,
973 			       struct netlink_ext_ack *extack)
974 {
975 	struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
976 	struct lwtunnel_state *new_state;
977 	struct ip_tunnel_info *tun_info;
978 	int err, opt_len;
979 
980 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
981 					  ip6_tun_policy, extack);
982 	if (err < 0)
983 		return err;
984 
985 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
986 	if (opt_len < 0)
987 		return opt_len;
988 
989 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
990 	if (!new_state)
991 		return -ENOMEM;
992 
993 	new_state->type = LWTUNNEL_ENCAP_IP6;
994 
995 	tun_info = lwt_tun_info(new_state);
996 
997 	err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
998 	if (err < 0) {
999 		lwtstate_free(new_state);
1000 		return err;
1001 	}
1002 
1003 	if (tb[LWTUNNEL_IP6_ID])
1004 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
1005 
1006 	if (tb[LWTUNNEL_IP6_DST])
1007 		tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
1008 
1009 	if (tb[LWTUNNEL_IP6_SRC])
1010 		tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
1011 
1012 	if (tb[LWTUNNEL_IP6_HOPLIMIT])
1013 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
1014 
1015 	if (tb[LWTUNNEL_IP6_TC])
1016 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
1017 
1018 	if (tb[LWTUNNEL_IP6_FLAGS]) {
1019 		IP_TUNNEL_DECLARE_FLAGS(flags);
1020 		__be16 data;
1021 
1022 		data = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
1023 		ip_tunnel_flags_from_be16(flags, data);
1024 		ip_tunnel_clear_options_present(flags);
1025 
1026 		ip_tunnel_flags_or(tun_info->key.tun_flags,
1027 				   tun_info->key.tun_flags, flags);
1028 	}
1029 
1030 	tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
1031 	tun_info->options_len = opt_len;
1032 
1033 	*ts = new_state;
1034 
1035 	return 0;
1036 }
1037 
1038 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1039 				   struct lwtunnel_state *lwtstate)
1040 {
1041 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1042 
1043 	if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1044 			 LWTUNNEL_IP6_PAD) ||
1045 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1046 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1047 	    nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1048 	    nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1049 	    nla_put_be16(skb, LWTUNNEL_IP6_FLAGS,
1050 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
1051 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1052 		return -ENOMEM;
1053 
1054 	return 0;
1055 }
1056 
1057 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1058 {
1059 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP6_ID */
1060 		+ nla_total_size(16)	/* LWTUNNEL_IP6_DST */
1061 		+ nla_total_size(16)	/* LWTUNNEL_IP6_SRC */
1062 		+ nla_total_size(1)	/* LWTUNNEL_IP6_HOPLIMIT */
1063 		+ nla_total_size(1)	/* LWTUNNEL_IP6_TC */
1064 		+ nla_total_size(2)	/* LWTUNNEL_IP6_FLAGS */
1065 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1066 					/* LWTUNNEL_IP6_OPTS */
1067 }
1068 
1069 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1070 	.build_state = ip6_tun_build_state,
1071 	.fill_encap = ip6_tun_fill_encap_info,
1072 	.get_encap_size = ip6_tun_encap_nlsize,
1073 	.cmp_encap = ip_tun_cmp_encap,
1074 	.owner = THIS_MODULE,
1075 };
1076 
1077 void __init ip_tunnel_core_init(void)
1078 {
1079 	/* If you land here, make sure whether increasing ip_tunnel_info's
1080 	 * options_len is a reasonable choice with its usage in front ends
1081 	 * (f.e., it's part of flow keys, etc).
1082 	 */
1083 	BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1084 
1085 	lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1086 	lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1087 }
1088 
1089 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1090 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1091 
1092 void ip_tunnel_need_metadata(void)
1093 {
1094 	static_branch_inc(&ip_tunnel_metadata_cnt);
1095 }
1096 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1097 
1098 void ip_tunnel_unneed_metadata(void)
1099 {
1100 	static_branch_dec(&ip_tunnel_metadata_cnt);
1101 }
1102 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1103 
1104 /* Returns either the correct skb->protocol value, or 0 if invalid. */
1105 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1106 {
1107 	if (skb_network_header(skb) >= skb->head &&
1108 	    (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1109 	    ip_hdr(skb)->version == 4)
1110 		return htons(ETH_P_IP);
1111 	if (skb_network_header(skb) >= skb->head &&
1112 	    (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1113 	    ipv6_hdr(skb)->version == 6)
1114 		return htons(ETH_P_IPV6);
1115 	return 0;
1116 }
1117 EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1118 
1119 const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1120 EXPORT_SYMBOL(ip_tunnel_header_ops);
1121 
1122 /* This function returns true when ENCAP attributes are present in the nl msg */
1123 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
1124 				   struct ip_tunnel_encap *encap)
1125 {
1126 	bool ret = false;
1127 
1128 	memset(encap, 0, sizeof(*encap));
1129 
1130 	if (!data)
1131 		return ret;
1132 
1133 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1134 		ret = true;
1135 		encap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1136 	}
1137 
1138 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1139 		ret = true;
1140 		encap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1141 	}
1142 
1143 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1144 		ret = true;
1145 		encap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1146 	}
1147 
1148 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1149 		ret = true;
1150 		encap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1151 	}
1152 
1153 	return ret;
1154 }
1155 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_encap_parms);
1156 
1157 void ip_tunnel_netlink_parms(struct nlattr *data[],
1158 			     struct ip_tunnel_parm_kern *parms)
1159 {
1160 	if (data[IFLA_IPTUN_LINK])
1161 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1162 
1163 	if (data[IFLA_IPTUN_LOCAL])
1164 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1165 
1166 	if (data[IFLA_IPTUN_REMOTE])
1167 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1168 
1169 	if (data[IFLA_IPTUN_TTL]) {
1170 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1171 		if (parms->iph.ttl)
1172 			parms->iph.frag_off = htons(IP_DF);
1173 	}
1174 
1175 	if (data[IFLA_IPTUN_TOS])
1176 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1177 
1178 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1179 		parms->iph.frag_off = htons(IP_DF);
1180 
1181 	if (data[IFLA_IPTUN_FLAGS]) {
1182 		__be16 flags;
1183 
1184 		flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1185 		ip_tunnel_flags_from_be16(parms->i_flags, flags);
1186 	}
1187 
1188 	if (data[IFLA_IPTUN_PROTO])
1189 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1190 }
1191 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_parms);
1192