xref: /linux/net/ipv4/ip_tunnel_core.c (revision b58d749633203d92c265317b45fccee555090352)
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 (net %llu), fix it urgently!\n",
64 					     dev->name, dev_net(dev)->net_cookie);
65 			DEV_STATS_INC(dev, tx_errors);
66 		}
67 		ip_rt_put(rt);
68 		kfree_skb_reason(skb, SKB_DROP_REASON_RECURSION_LIMIT);
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;
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 
230 	err = pskb_trim(skb, 576 - sizeof(*niph) - sizeof(*icmph));
231 	if (err)
232 		return err;
233 
234 	len = skb->len + sizeof(*icmph);
235 	err = skb_cow(skb, sizeof(*niph) + sizeof(*icmph) + ETH_HLEN);
236 	if (err)
237 		return err;
238 	iph = ip_hdr(skb);
239 	icmph = skb_push(skb, sizeof(*icmph));
240 	*icmph = (struct icmphdr) {
241 		.type			= ICMP_DEST_UNREACH,
242 		.code			= ICMP_FRAG_NEEDED,
243 		.checksum		= 0,
244 		.un.frag.__unused	= 0,
245 		.un.frag.mtu		= htons(mtu),
246 	};
247 	icmph->checksum = csum_fold(skb_checksum(skb, 0, len, 0));
248 	skb_reset_transport_header(skb);
249 
250 	niph = skb_push(skb, sizeof(*niph));
251 	*niph = (struct iphdr) {
252 		.ihl			= sizeof(*niph) / 4u,
253 		.version 		= 4,
254 		.tos 			= 0,
255 		.tot_len		= htons(len + sizeof(*niph)),
256 		.id			= 0,
257 		.frag_off		= htons(IP_DF),
258 		.ttl			= iph->ttl,
259 		.protocol		= IPPROTO_ICMP,
260 		.saddr			= iph->daddr,
261 		.daddr			= iph->saddr,
262 	};
263 	ip_send_check(niph);
264 	skb_reset_network_header(skb);
265 
266 	skb->ip_summed = CHECKSUM_NONE;
267 
268 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
269 	skb_reset_mac_header(skb);
270 
271 	if (skb_valid_dst(skb))
272 		skb_dst_drop(skb);
273 
274 	return skb->len;
275 }
276 
277 /**
278  * iptunnel_pmtud_check_icmp() - Trigger ICMP reply if needed and allowed
279  * @skb:	Buffer being sent by encapsulation, L2 headers expected
280  * @mtu:	Network MTU for path
281  *
282  * Return: 0 for no ICMP reply, length if built, negative value on error.
283  */
284 static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
285 {
286 	const struct iphdr *iph = ip_hdr(skb);
287 
288 	if (mtu < 576 || iph->frag_off != htons(IP_DF))
289 		return 0;
290 
291 	if (ipv4_is_lbcast(iph->daddr)  || ipv4_is_multicast(iph->daddr) ||
292 	    ipv4_is_zeronet(iph->saddr) || ipv4_is_loopback(iph->saddr)  ||
293 	    ipv4_is_lbcast(iph->saddr)  || ipv4_is_multicast(iph->saddr))
294 		return 0;
295 
296 	if (iph->protocol == IPPROTO_ICMP) {
297 		const struct icmphdr *icmph;
298 
299 		if (!pskb_network_may_pull(skb, iph->ihl * 4 +
300 						offsetofend(struct icmphdr, type)))
301 			return 0;
302 		iph = ip_hdr(skb);
303 		icmph = (void *)iph + iph->ihl * 4;
304 		if (icmp_is_err(icmph->type))
305 			return 0;
306 	}
307 	return iptunnel_pmtud_build_icmp(skb, mtu);
308 }
309 
310 #if IS_ENABLED(CONFIG_IPV6)
311 /**
312  * iptunnel_pmtud_build_icmpv6() - Build ICMPv6 error message for PMTUD
313  * @skb:	Original packet with L2 header
314  * @mtu:	MTU value for ICMPv6 error
315  *
316  * Return: length on success, negative error code if message couldn't be built.
317  */
318 static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
319 {
320 	const struct ipv6hdr *ip6h;
321 	struct icmp6hdr *icmp6h;
322 	struct ipv6hdr *nip6h;
323 	struct ethhdr eh;
324 	int len, err;
325 	__wsum csum;
326 
327 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
328 		return -EINVAL;
329 
330 	if (skb_is_gso(skb))
331 		skb_gso_reset(skb);
332 
333 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
334 	pskb_pull(skb, ETH_HLEN);
335 
336 	err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
337 	if (err)
338 		return err;
339 
340 	len = skb->len + sizeof(*icmp6h);
341 	err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
342 	if (err)
343 		return err;
344 
345 	ip6h = ipv6_hdr(skb);
346 	icmp6h = skb_push(skb, sizeof(*icmp6h));
347 	*icmp6h = (struct icmp6hdr) {
348 		.icmp6_type		= ICMPV6_PKT_TOOBIG,
349 		.icmp6_code		= 0,
350 		.icmp6_cksum		= 0,
351 		.icmp6_mtu		= htonl(mtu),
352 	};
353 	skb_reset_transport_header(skb);
354 
355 	nip6h = skb_push(skb, sizeof(*nip6h));
356 	*nip6h = (struct ipv6hdr) {
357 		.priority		= 0,
358 		.version		= 6,
359 		.flow_lbl		= { 0 },
360 		.payload_len		= htons(len),
361 		.nexthdr		= IPPROTO_ICMPV6,
362 		.hop_limit		= ip6h->hop_limit,
363 		.saddr			= ip6h->daddr,
364 		.daddr			= ip6h->saddr,
365 	};
366 	skb_reset_network_header(skb);
367 
368 	csum = skb_checksum(skb, skb_transport_offset(skb), len, 0);
369 	icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
370 					      IPPROTO_ICMPV6, csum);
371 
372 	skb->ip_summed = CHECKSUM_NONE;
373 
374 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
375 	skb_reset_mac_header(skb);
376 
377 	if (skb_valid_dst(skb))
378 		skb_dst_drop(skb);
379 
380 	return skb->len;
381 }
382 
383 /**
384  * iptunnel_pmtud_check_icmpv6() - Trigger ICMPv6 reply if needed and allowed
385  * @skb:	Buffer being sent by encapsulation, L2 headers expected
386  * @mtu:	Network MTU for path
387  *
388  * Return: 0 for no ICMPv6 reply, length if built, negative value on error.
389  */
390 static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
391 {
392 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
393 	int stype = ipv6_addr_type(&ip6h->saddr);
394 	u8 proto = ip6h->nexthdr;
395 	__be16 frag_off;
396 	int offset;
397 
398 	if (mtu < IPV6_MIN_MTU)
399 		return 0;
400 
401 	if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
402 	    stype == IPV6_ADDR_LOOPBACK)
403 		return 0;
404 
405 	offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
406 				  &frag_off);
407 	if (offset < 0 || (frag_off & htons(~0x7)))
408 		return 0;
409 
410 	if (proto == IPPROTO_ICMPV6) {
411 		struct icmp6hdr *icmp6h;
412 
413 		if (!pskb_may_pull(skb, skb_network_header(skb) +
414 					offset + 1 - skb->data))
415 			return 0;
416 
417 		icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
418 		if (icmpv6_is_err(icmp6h->icmp6_type) ||
419 		    icmp6h->icmp6_type == NDISC_REDIRECT)
420 			return 0;
421 	}
422 
423 	return iptunnel_pmtud_build_icmpv6(skb, mtu);
424 }
425 #endif /* IS_ENABLED(CONFIG_IPV6) */
426 
427 /**
428  * skb_tunnel_check_pmtu() - Check, update PMTU and trigger ICMP reply as needed
429  * @skb:	Buffer being sent by encapsulation, L2 headers expected
430  * @encap_dst:	Destination for tunnel encapsulation (outer IP)
431  * @headroom:	Encapsulation header size, bytes
432  * @reply:	Build matching ICMP or ICMPv6 message as a result
433  *
434  * L2 tunnel implementations that can carry IP and can be directly bridged
435  * (currently UDP tunnels) can't always rely on IP forwarding paths to handle
436  * PMTU discovery. In the bridged case, ICMP or ICMPv6 messages need to be built
437  * based on payload and sent back by the encapsulation itself.
438  *
439  * For routable interfaces, we just need to update the PMTU for the destination.
440  *
441  * Return: 0 if ICMP error not needed, length if built, negative value on error
442  */
443 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
444 			  int headroom, bool reply)
445 {
446 	u32 mtu = dst_mtu(encap_dst) - headroom;
447 
448 	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
449 	    (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
450 		return 0;
451 
452 	skb_dst_update_pmtu_no_confirm(skb, mtu);
453 
454 	if (!reply)
455 		return 0;
456 
457 	if (skb->protocol == htons(ETH_P_IP))
458 		return iptunnel_pmtud_check_icmp(skb, mtu);
459 
460 #if IS_ENABLED(CONFIG_IPV6)
461 	if (skb->protocol == htons(ETH_P_IPV6))
462 		return iptunnel_pmtud_check_icmpv6(skb, mtu);
463 #endif
464 	return 0;
465 }
466 EXPORT_SYMBOL(skb_tunnel_check_pmtu);
467 
468 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
469 	[LWTUNNEL_IP_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP_OPTS },
470 	[LWTUNNEL_IP_ID]	= { .type = NLA_U64 },
471 	[LWTUNNEL_IP_DST]	= { .type = NLA_U32 },
472 	[LWTUNNEL_IP_SRC]	= { .type = NLA_U32 },
473 	[LWTUNNEL_IP_TTL]	= { .type = NLA_U8 },
474 	[LWTUNNEL_IP_TOS]	= { .type = NLA_U8 },
475 	[LWTUNNEL_IP_FLAGS]	= { .type = NLA_U16 },
476 	[LWTUNNEL_IP_OPTS]	= { .type = NLA_NESTED },
477 };
478 
479 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
480 	[LWTUNNEL_IP_OPTS_GENEVE]	= { .type = NLA_NESTED },
481 	[LWTUNNEL_IP_OPTS_VXLAN]	= { .type = NLA_NESTED },
482 	[LWTUNNEL_IP_OPTS_ERSPAN]	= { .type = NLA_NESTED },
483 };
484 
485 static const struct nla_policy
486 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
487 	[LWTUNNEL_IP_OPT_GENEVE_CLASS]	= { .type = NLA_U16 },
488 	[LWTUNNEL_IP_OPT_GENEVE_TYPE]	= { .type = NLA_U8 },
489 	[LWTUNNEL_IP_OPT_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 127 },
490 };
491 
492 static const struct nla_policy
493 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
494 	[LWTUNNEL_IP_OPT_VXLAN_GBP]	= { .type = NLA_U32 },
495 };
496 
497 static const struct nla_policy
498 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
499 	[LWTUNNEL_IP_OPT_ERSPAN_VER]	= { .type = NLA_U8 },
500 	[LWTUNNEL_IP_OPT_ERSPAN_INDEX]	= { .type = NLA_U32 },
501 	[LWTUNNEL_IP_OPT_ERSPAN_DIR]	= { .type = NLA_U8 },
502 	[LWTUNNEL_IP_OPT_ERSPAN_HWID]	= { .type = NLA_U8 },
503 };
504 
505 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
506 				    struct ip_tunnel_info *info, int opts_len,
507 				    struct netlink_ext_ack *extack)
508 {
509 	struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
510 	int data_len, err;
511 
512 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
513 			       geneve_opt_policy, extack);
514 	if (err)
515 		return err;
516 
517 	if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
518 	    !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
519 	    !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
520 		return -EINVAL;
521 
522 	attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
523 	data_len = nla_len(attr);
524 	if (data_len % 4)
525 		return -EINVAL;
526 
527 	if (info) {
528 		struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
529 
530 		memcpy(opt->opt_data, nla_data(attr), data_len);
531 		opt->length = data_len / 4;
532 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
533 		opt->opt_class = nla_get_be16(attr);
534 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
535 		opt->type = nla_get_u8(attr);
536 		__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags);
537 	}
538 
539 	return sizeof(struct geneve_opt) + data_len;
540 }
541 
542 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
543 				   struct ip_tunnel_info *info, int opts_len,
544 				   struct netlink_ext_ack *extack)
545 {
546 	struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
547 	int err;
548 
549 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
550 			       vxlan_opt_policy, extack);
551 	if (err)
552 		return err;
553 
554 	if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
555 		return -EINVAL;
556 
557 	if (info) {
558 		struct vxlan_metadata *md =
559 			ip_tunnel_info_opts(info) + opts_len;
560 
561 		attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
562 		md->gbp = nla_get_u32(attr);
563 		md->gbp &= VXLAN_GBP_MASK;
564 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags);
565 	}
566 
567 	return sizeof(struct vxlan_metadata);
568 }
569 
570 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
571 				    struct ip_tunnel_info *info, int opts_len,
572 				    struct netlink_ext_ack *extack)
573 {
574 	struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
575 	int err;
576 	u8 ver;
577 
578 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
579 			       erspan_opt_policy, extack);
580 	if (err)
581 		return err;
582 
583 	if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
584 		return -EINVAL;
585 
586 	ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
587 	if (ver == 1) {
588 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
589 			return -EINVAL;
590 	} else if (ver == 2) {
591 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
592 		    !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
593 			return -EINVAL;
594 	} else {
595 		return -EINVAL;
596 	}
597 
598 	if (info) {
599 		struct erspan_metadata *md =
600 			ip_tunnel_info_opts(info) + opts_len;
601 
602 		md->version = ver;
603 		if (ver == 1) {
604 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
605 			md->u.index = nla_get_be32(attr);
606 		} else {
607 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
608 			md->u.md2.dir = nla_get_u8(attr);
609 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
610 			set_hwid(&md->u.md2, nla_get_u8(attr));
611 		}
612 
613 		__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags);
614 	}
615 
616 	return sizeof(struct erspan_metadata);
617 }
618 
619 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
620 			     struct netlink_ext_ack *extack)
621 {
622 	int err, rem, opt_len, opts_len = 0;
623 	struct nlattr *nla;
624 	u32 type = 0;
625 
626 	if (!attr)
627 		return 0;
628 
629 	err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
630 			   ip_opts_policy, extack);
631 	if (err)
632 		return err;
633 
634 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
635 		switch (nla_type(nla)) {
636 		case LWTUNNEL_IP_OPTS_GENEVE:
637 			if (type && type != IP_TUNNEL_GENEVE_OPT_BIT)
638 				return -EINVAL;
639 			opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
640 							   extack);
641 			if (opt_len < 0)
642 				return opt_len;
643 			opts_len += opt_len;
644 			if (opts_len > IP_TUNNEL_OPTS_MAX)
645 				return -EINVAL;
646 			type = IP_TUNNEL_GENEVE_OPT_BIT;
647 			break;
648 		case LWTUNNEL_IP_OPTS_VXLAN:
649 			if (type)
650 				return -EINVAL;
651 			opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
652 							  extack);
653 			if (opt_len < 0)
654 				return opt_len;
655 			opts_len += opt_len;
656 			type = IP_TUNNEL_VXLAN_OPT_BIT;
657 			break;
658 		case LWTUNNEL_IP_OPTS_ERSPAN:
659 			if (type)
660 				return -EINVAL;
661 			opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
662 							   extack);
663 			if (opt_len < 0)
664 				return opt_len;
665 			opts_len += opt_len;
666 			type = IP_TUNNEL_ERSPAN_OPT_BIT;
667 			break;
668 		default:
669 			return -EINVAL;
670 		}
671 	}
672 
673 	return opts_len;
674 }
675 
676 static int ip_tun_get_optlen(struct nlattr *attr,
677 			     struct netlink_ext_ack *extack)
678 {
679 	return ip_tun_parse_opts(attr, NULL, extack);
680 }
681 
682 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
683 			   struct netlink_ext_ack *extack)
684 {
685 	return ip_tun_parse_opts(attr, info, extack);
686 }
687 
688 static int ip_tun_build_state(struct net *net, struct nlattr *attr,
689 			      unsigned int family, const void *cfg,
690 			      struct lwtunnel_state **ts,
691 			      struct netlink_ext_ack *extack)
692 {
693 	struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
694 	struct lwtunnel_state *new_state;
695 	struct ip_tunnel_info *tun_info;
696 	int err, opt_len;
697 
698 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
699 					  ip_tun_policy, extack);
700 	if (err < 0)
701 		return err;
702 
703 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
704 	if (opt_len < 0)
705 		return opt_len;
706 
707 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
708 	if (!new_state)
709 		return -ENOMEM;
710 
711 	new_state->type = LWTUNNEL_ENCAP_IP;
712 
713 	tun_info = lwt_tun_info(new_state);
714 
715 	err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
716 	if (err < 0) {
717 		lwtstate_free(new_state);
718 		return err;
719 	}
720 
721 #ifdef CONFIG_DST_CACHE
722 	err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
723 	if (err) {
724 		lwtstate_free(new_state);
725 		return err;
726 	}
727 #endif
728 
729 	if (tb[LWTUNNEL_IP_ID])
730 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
731 
732 	if (tb[LWTUNNEL_IP_DST])
733 		tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
734 
735 	if (tb[LWTUNNEL_IP_SRC])
736 		tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
737 
738 	if (tb[LWTUNNEL_IP_TTL])
739 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
740 
741 	if (tb[LWTUNNEL_IP_TOS])
742 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
743 
744 	if (tb[LWTUNNEL_IP_FLAGS]) {
745 		IP_TUNNEL_DECLARE_FLAGS(flags);
746 
747 		ip_tunnel_flags_from_be16(flags,
748 					  nla_get_be16(tb[LWTUNNEL_IP_FLAGS]));
749 		ip_tunnel_clear_options_present(flags);
750 
751 		ip_tunnel_flags_or(tun_info->key.tun_flags,
752 				   tun_info->key.tun_flags, flags);
753 	}
754 
755 	tun_info->mode = IP_TUNNEL_INFO_TX;
756 	tun_info->options_len = opt_len;
757 
758 	*ts = new_state;
759 
760 	return 0;
761 }
762 
763 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
764 {
765 #ifdef CONFIG_DST_CACHE
766 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
767 
768 	dst_cache_destroy(&tun_info->dst_cache);
769 #endif
770 }
771 
772 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
773 					 struct ip_tunnel_info *tun_info)
774 {
775 	struct geneve_opt *opt;
776 	struct nlattr *nest;
777 	int offset = 0;
778 
779 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
780 	if (!nest)
781 		return -ENOMEM;
782 
783 	while (tun_info->options_len > offset) {
784 		opt = ip_tunnel_info_opts(tun_info) + offset;
785 		if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
786 				 opt->opt_class) ||
787 		    nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
788 		    nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
789 			    opt->opt_data)) {
790 			nla_nest_cancel(skb, nest);
791 			return -ENOMEM;
792 		}
793 		offset += sizeof(*opt) + opt->length * 4;
794 	}
795 
796 	nla_nest_end(skb, nest);
797 	return 0;
798 }
799 
800 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
801 					struct ip_tunnel_info *tun_info)
802 {
803 	struct vxlan_metadata *md;
804 	struct nlattr *nest;
805 
806 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
807 	if (!nest)
808 		return -ENOMEM;
809 
810 	md = ip_tunnel_info_opts(tun_info);
811 	if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
812 		nla_nest_cancel(skb, nest);
813 		return -ENOMEM;
814 	}
815 
816 	nla_nest_end(skb, nest);
817 	return 0;
818 }
819 
820 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
821 					 struct ip_tunnel_info *tun_info)
822 {
823 	struct erspan_metadata *md;
824 	struct nlattr *nest;
825 
826 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
827 	if (!nest)
828 		return -ENOMEM;
829 
830 	md = ip_tunnel_info_opts(tun_info);
831 	if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
832 		goto err;
833 
834 	if (md->version == 1 &&
835 	    nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
836 		goto err;
837 
838 	if (md->version == 2 &&
839 	    (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
840 	     nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
841 			get_hwid(&md->u.md2))))
842 		goto err;
843 
844 	nla_nest_end(skb, nest);
845 	return 0;
846 err:
847 	nla_nest_cancel(skb, nest);
848 	return -ENOMEM;
849 }
850 
851 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
852 				  struct ip_tunnel_info *tun_info)
853 {
854 	struct nlattr *nest;
855 	int err = 0;
856 
857 	if (!ip_tunnel_is_options_present(tun_info->key.tun_flags))
858 		return 0;
859 
860 	nest = nla_nest_start_noflag(skb, type);
861 	if (!nest)
862 		return -ENOMEM;
863 
864 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_info->key.tun_flags))
865 		err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
866 	else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_info->key.tun_flags))
867 		err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
868 	else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
869 		err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
870 
871 	if (err) {
872 		nla_nest_cancel(skb, nest);
873 		return err;
874 	}
875 
876 	nla_nest_end(skb, nest);
877 	return 0;
878 }
879 
880 static int ip_tun_fill_encap_info(struct sk_buff *skb,
881 				  struct lwtunnel_state *lwtstate)
882 {
883 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
884 
885 	if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
886 			 LWTUNNEL_IP_PAD) ||
887 	    nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
888 	    nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
889 	    nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
890 	    nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
891 	    nla_put_be16(skb, LWTUNNEL_IP_FLAGS,
892 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
893 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
894 		return -ENOMEM;
895 
896 	return 0;
897 }
898 
899 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
900 {
901 	int opt_len;
902 
903 	if (!ip_tunnel_is_options_present(info->key.tun_flags))
904 		return 0;
905 
906 	opt_len = nla_total_size(0);		/* LWTUNNEL_IP_OPTS */
907 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags)) {
908 		struct geneve_opt *opt;
909 		int offset = 0;
910 
911 		opt_len += nla_total_size(0);	/* LWTUNNEL_IP_OPTS_GENEVE */
912 		while (info->options_len > offset) {
913 			opt = ip_tunnel_info_opts(info) + offset;
914 			opt_len += nla_total_size(2)	/* OPT_GENEVE_CLASS */
915 				   + nla_total_size(1)	/* OPT_GENEVE_TYPE */
916 				   + nla_total_size(opt->length * 4);
917 							/* OPT_GENEVE_DATA */
918 			offset += sizeof(*opt) + opt->length * 4;
919 		}
920 	} else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
921 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_VXLAN */
922 			   + nla_total_size(4);	/* OPT_VXLAN_GBP */
923 	} else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags)) {
924 		struct erspan_metadata *md = ip_tunnel_info_opts(info);
925 
926 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_ERSPAN */
927 			   + nla_total_size(1)	/* OPT_ERSPAN_VER */
928 			   + (md->version == 1 ? nla_total_size(4)
929 						/* OPT_ERSPAN_INDEX (v1) */
930 					       : nla_total_size(1) +
931 						 nla_total_size(1));
932 						/* OPT_ERSPAN_DIR + HWID (v2) */
933 	}
934 
935 	return opt_len;
936 }
937 
938 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
939 {
940 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP_ID */
941 		+ nla_total_size(4)	/* LWTUNNEL_IP_DST */
942 		+ nla_total_size(4)	/* LWTUNNEL_IP_SRC */
943 		+ nla_total_size(1)	/* LWTUNNEL_IP_TOS */
944 		+ nla_total_size(1)	/* LWTUNNEL_IP_TTL */
945 		+ nla_total_size(2)	/* LWTUNNEL_IP_FLAGS */
946 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
947 					/* LWTUNNEL_IP_OPTS */
948 }
949 
950 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
951 {
952 	struct ip_tunnel_info *info_a = lwt_tun_info(a);
953 	struct ip_tunnel_info *info_b = lwt_tun_info(b);
954 
955 	return memcmp(info_a, info_b, sizeof(info_a->key)) ||
956 	       info_a->mode != info_b->mode ||
957 	       info_a->options_len != info_b->options_len ||
958 	       memcmp(ip_tunnel_info_opts(info_a),
959 		      ip_tunnel_info_opts(info_b), info_a->options_len);
960 }
961 
962 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
963 	.build_state = ip_tun_build_state,
964 	.destroy_state = ip_tun_destroy_state,
965 	.fill_encap = ip_tun_fill_encap_info,
966 	.get_encap_size = ip_tun_encap_nlsize,
967 	.cmp_encap = ip_tun_cmp_encap,
968 	.owner = THIS_MODULE,
969 };
970 
971 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
972 	[LWTUNNEL_IP6_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP6_OPTS },
973 	[LWTUNNEL_IP6_ID]		= { .type = NLA_U64 },
974 	[LWTUNNEL_IP6_DST]		= { .len = sizeof(struct in6_addr) },
975 	[LWTUNNEL_IP6_SRC]		= { .len = sizeof(struct in6_addr) },
976 	[LWTUNNEL_IP6_HOPLIMIT]		= { .type = NLA_U8 },
977 	[LWTUNNEL_IP6_TC]		= { .type = NLA_U8 },
978 	[LWTUNNEL_IP6_FLAGS]		= { .type = NLA_U16 },
979 	[LWTUNNEL_IP6_OPTS]		= { .type = NLA_NESTED },
980 };
981 
982 static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
983 			       unsigned int family, const void *cfg,
984 			       struct lwtunnel_state **ts,
985 			       struct netlink_ext_ack *extack)
986 {
987 	struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
988 	struct lwtunnel_state *new_state;
989 	struct ip_tunnel_info *tun_info;
990 	int err, opt_len;
991 
992 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
993 					  ip6_tun_policy, extack);
994 	if (err < 0)
995 		return err;
996 
997 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
998 	if (opt_len < 0)
999 		return opt_len;
1000 
1001 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
1002 	if (!new_state)
1003 		return -ENOMEM;
1004 
1005 	new_state->type = LWTUNNEL_ENCAP_IP6;
1006 
1007 	tun_info = lwt_tun_info(new_state);
1008 
1009 	err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
1010 	if (err < 0) {
1011 		lwtstate_free(new_state);
1012 		return err;
1013 	}
1014 
1015 	if (tb[LWTUNNEL_IP6_ID])
1016 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
1017 
1018 	if (tb[LWTUNNEL_IP6_DST])
1019 		tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
1020 
1021 	if (tb[LWTUNNEL_IP6_SRC])
1022 		tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
1023 
1024 	if (tb[LWTUNNEL_IP6_HOPLIMIT])
1025 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
1026 
1027 	if (tb[LWTUNNEL_IP6_TC])
1028 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
1029 
1030 	if (tb[LWTUNNEL_IP6_FLAGS]) {
1031 		IP_TUNNEL_DECLARE_FLAGS(flags);
1032 		__be16 data;
1033 
1034 		data = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
1035 		ip_tunnel_flags_from_be16(flags, data);
1036 		ip_tunnel_clear_options_present(flags);
1037 
1038 		ip_tunnel_flags_or(tun_info->key.tun_flags,
1039 				   tun_info->key.tun_flags, flags);
1040 	}
1041 
1042 	tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
1043 	tun_info->options_len = opt_len;
1044 
1045 	*ts = new_state;
1046 
1047 	return 0;
1048 }
1049 
1050 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1051 				   struct lwtunnel_state *lwtstate)
1052 {
1053 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1054 
1055 	if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1056 			 LWTUNNEL_IP6_PAD) ||
1057 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1058 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1059 	    nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1060 	    nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1061 	    nla_put_be16(skb, LWTUNNEL_IP6_FLAGS,
1062 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
1063 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1064 		return -ENOMEM;
1065 
1066 	return 0;
1067 }
1068 
1069 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1070 {
1071 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP6_ID */
1072 		+ nla_total_size(16)	/* LWTUNNEL_IP6_DST */
1073 		+ nla_total_size(16)	/* LWTUNNEL_IP6_SRC */
1074 		+ nla_total_size(1)	/* LWTUNNEL_IP6_HOPLIMIT */
1075 		+ nla_total_size(1)	/* LWTUNNEL_IP6_TC */
1076 		+ nla_total_size(2)	/* LWTUNNEL_IP6_FLAGS */
1077 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1078 					/* LWTUNNEL_IP6_OPTS */
1079 }
1080 
1081 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1082 	.build_state = ip6_tun_build_state,
1083 	.fill_encap = ip6_tun_fill_encap_info,
1084 	.get_encap_size = ip6_tun_encap_nlsize,
1085 	.cmp_encap = ip_tun_cmp_encap,
1086 	.owner = THIS_MODULE,
1087 };
1088 
1089 void __init ip_tunnel_core_init(void)
1090 {
1091 	/* If you land here, make sure whether increasing ip_tunnel_info's
1092 	 * options_len is a reasonable choice with its usage in front ends
1093 	 * (f.e., it's part of flow keys, etc).
1094 	 */
1095 	BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1096 
1097 	lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1098 	lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1099 }
1100 
1101 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1102 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1103 
1104 void ip_tunnel_need_metadata(void)
1105 {
1106 	static_branch_inc(&ip_tunnel_metadata_cnt);
1107 }
1108 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1109 
1110 void ip_tunnel_unneed_metadata(void)
1111 {
1112 	static_branch_dec(&ip_tunnel_metadata_cnt);
1113 }
1114 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1115 
1116 /* Returns either the correct skb->protocol value, or 0 if invalid. */
1117 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1118 {
1119 	if (skb_network_header(skb) >= skb->head &&
1120 	    (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1121 	    ip_hdr(skb)->version == 4)
1122 		return htons(ETH_P_IP);
1123 	if (skb_network_header(skb) >= skb->head &&
1124 	    (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1125 	    ipv6_hdr(skb)->version == 6)
1126 		return htons(ETH_P_IPV6);
1127 	return 0;
1128 }
1129 EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1130 
1131 const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1132 EXPORT_SYMBOL(ip_tunnel_header_ops);
1133 
1134 /* This function returns true when ENCAP attributes are present in the nl msg */
1135 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
1136 				   struct ip_tunnel_encap *encap)
1137 {
1138 	bool ret = false;
1139 
1140 	memset(encap, 0, sizeof(*encap));
1141 
1142 	if (!data)
1143 		return ret;
1144 
1145 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1146 		ret = true;
1147 		encap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1148 	}
1149 
1150 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1151 		ret = true;
1152 		encap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1153 	}
1154 
1155 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1156 		ret = true;
1157 		encap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1158 	}
1159 
1160 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1161 		ret = true;
1162 		encap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1163 	}
1164 
1165 	return ret;
1166 }
1167 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_encap_parms);
1168 
1169 void ip_tunnel_netlink_parms(struct nlattr *data[],
1170 			     struct ip_tunnel_parm_kern *parms)
1171 {
1172 	if (data[IFLA_IPTUN_LINK])
1173 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1174 
1175 	if (data[IFLA_IPTUN_LOCAL])
1176 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1177 
1178 	if (data[IFLA_IPTUN_REMOTE])
1179 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1180 
1181 	if (data[IFLA_IPTUN_TTL]) {
1182 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1183 		if (parms->iph.ttl)
1184 			parms->iph.frag_off = htons(IP_DF);
1185 	}
1186 
1187 	if (data[IFLA_IPTUN_TOS])
1188 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1189 
1190 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1191 		parms->iph.frag_off = htons(IP_DF);
1192 
1193 	if (data[IFLA_IPTUN_FLAGS]) {
1194 		__be16 flags;
1195 
1196 		flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1197 		ip_tunnel_flags_from_be16(parms->i_flags, flags);
1198 	}
1199 
1200 	if (data[IFLA_IPTUN_PROTO])
1201 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1202 }
1203 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_parms);
1204