xref: /linux/net/ipv4/ip_tunnel_core.c (revision 57885276cc16a2e2b76282c808a4e84cbecb3aae)
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 (dev_recursion_level() > IP_TUNNEL_RECURSION_LIMIT) {
62 		net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
63 				     dev->name);
64 		DEV_STATS_INC(dev, tx_errors);
65 		ip_rt_put(rt);
66 		kfree_skb(skb);
67 		return;
68 	}
69 
70 	dev_xmit_recursion_inc();
71 
72 	skb_scrub_packet(skb, xnet);
73 
74 	skb_clear_hash_if_not_l4(skb);
75 	skb_dst_set(skb, &rt->dst);
76 	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
77 	IPCB(skb)->flags = ipcb_flags;
78 
79 	/* Push down and install the IP header. */
80 	skb_push(skb, sizeof(struct iphdr));
81 	skb_reset_network_header(skb);
82 
83 	iph = ip_hdr(skb);
84 
85 	iph->version	=	4;
86 	iph->ihl	=	sizeof(struct iphdr) >> 2;
87 	iph->frag_off	=	ip_mtu_locked(&rt->dst) ? 0 : df;
88 	iph->protocol	=	proto;
89 	iph->tos	=	tos;
90 	iph->daddr	=	dst;
91 	iph->saddr	=	src;
92 	iph->ttl	=	ttl;
93 	__ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
94 
95 	err = ip_local_out(net, sk, skb);
96 
97 	if (dev) {
98 		if (unlikely(net_xmit_eval(err)))
99 			pkt_len = 0;
100 		iptunnel_xmit_stats(dev, pkt_len);
101 	}
102 
103 	dev_xmit_recursion_dec();
104 }
105 EXPORT_SYMBOL_GPL(iptunnel_xmit);
106 
107 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
108 			   __be16 inner_proto, bool raw_proto, bool xnet)
109 {
110 	if (unlikely(!pskb_may_pull(skb, hdr_len)))
111 		return -ENOMEM;
112 
113 	skb_pull_rcsum(skb, hdr_len);
114 
115 	if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
116 		struct ethhdr *eh;
117 
118 		if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
119 			return -ENOMEM;
120 
121 		eh = (struct ethhdr *)skb->data;
122 		if (likely(eth_proto_is_802_3(eh->h_proto)))
123 			skb->protocol = eh->h_proto;
124 		else
125 			skb->protocol = htons(ETH_P_802_2);
126 
127 	} else {
128 		skb->protocol = inner_proto;
129 	}
130 
131 	skb_clear_hash_if_not_l4(skb);
132 	__vlan_hwaccel_clear_tag(skb);
133 	skb_set_queue_mapping(skb, 0);
134 	skb_scrub_packet(skb, xnet);
135 
136 	return iptunnel_pull_offloads(skb);
137 }
138 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
139 
140 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
141 					     gfp_t flags)
142 {
143 	IP_TUNNEL_DECLARE_FLAGS(tun_flags) = { };
144 	struct metadata_dst *res;
145 	struct ip_tunnel_info *dst, *src;
146 
147 	if (!md || md->type != METADATA_IP_TUNNEL ||
148 	    md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
149 		return NULL;
150 
151 	src = &md->u.tun_info;
152 	res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
153 	if (!res)
154 		return NULL;
155 
156 	dst = &res->u.tun_info;
157 	dst->key.tun_id = src->key.tun_id;
158 	if (src->mode & IP_TUNNEL_INFO_IPV6)
159 		memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
160 		       sizeof(struct in6_addr));
161 	else
162 		dst->key.u.ipv4.dst = src->key.u.ipv4.src;
163 	ip_tunnel_flags_copy(dst->key.tun_flags, src->key.tun_flags);
164 	dst->mode = src->mode | IP_TUNNEL_INFO_TX;
165 	ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
166 				src->options_len, tun_flags);
167 
168 	return res;
169 }
170 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
171 
172 int iptunnel_handle_offloads(struct sk_buff *skb,
173 			     int gso_type_mask)
174 {
175 	int err;
176 
177 	if (likely(!skb->encapsulation)) {
178 		skb_reset_inner_headers(skb);
179 		skb->encapsulation = 1;
180 	}
181 
182 	if (skb_is_gso(skb)) {
183 		err = skb_header_unclone(skb, GFP_ATOMIC);
184 		if (unlikely(err))
185 			return err;
186 		skb_shinfo(skb)->gso_type |= gso_type_mask;
187 		return 0;
188 	}
189 
190 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
191 		skb->ip_summed = CHECKSUM_NONE;
192 		/* We clear encapsulation here to prevent badly-written
193 		 * drivers potentially deciding to offload an inner checksum
194 		 * if we set CHECKSUM_PARTIAL on the outer header.
195 		 * This should go away when the drivers are all fixed.
196 		 */
197 		skb->encapsulation = 0;
198 	}
199 
200 	return 0;
201 }
202 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
203 
204 /**
205  * iptunnel_pmtud_build_icmp() - Build ICMP error message for PMTUD
206  * @skb:	Original packet with L2 header
207  * @mtu:	MTU value for ICMP error
208  *
209  * Return: length on success, negative error code if message couldn't be built.
210  */
211 static int iptunnel_pmtud_build_icmp(struct sk_buff *skb, int mtu)
212 {
213 	const struct iphdr *iph = ip_hdr(skb);
214 	struct icmphdr *icmph;
215 	struct iphdr *niph;
216 	struct ethhdr eh;
217 	int len, err;
218 
219 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct iphdr)))
220 		return -EINVAL;
221 
222 	if (skb_is_gso(skb))
223 		skb_gso_reset(skb);
224 
225 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
226 	pskb_pull(skb, ETH_HLEN);
227 	skb_reset_network_header(skb);
228 
229 	err = pskb_trim(skb, 576 - sizeof(*niph) - sizeof(*icmph));
230 	if (err)
231 		return err;
232 
233 	len = skb->len + sizeof(*icmph);
234 	err = skb_cow(skb, sizeof(*niph) + sizeof(*icmph) + ETH_HLEN);
235 	if (err)
236 		return err;
237 
238 	icmph = skb_push(skb, sizeof(*icmph));
239 	*icmph = (struct icmphdr) {
240 		.type			= ICMP_DEST_UNREACH,
241 		.code			= ICMP_FRAG_NEEDED,
242 		.checksum		= 0,
243 		.un.frag.__unused	= 0,
244 		.un.frag.mtu		= htons(mtu),
245 	};
246 	icmph->checksum = csum_fold(skb_checksum(skb, 0, len, 0));
247 	skb_reset_transport_header(skb);
248 
249 	niph = skb_push(skb, sizeof(*niph));
250 	*niph = (struct iphdr) {
251 		.ihl			= sizeof(*niph) / 4u,
252 		.version 		= 4,
253 		.tos 			= 0,
254 		.tot_len		= htons(len + sizeof(*niph)),
255 		.id			= 0,
256 		.frag_off		= htons(IP_DF),
257 		.ttl			= iph->ttl,
258 		.protocol		= IPPROTO_ICMP,
259 		.saddr			= iph->daddr,
260 		.daddr			= iph->saddr,
261 	};
262 	ip_send_check(niph);
263 	skb_reset_network_header(skb);
264 
265 	skb->ip_summed = CHECKSUM_NONE;
266 
267 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
268 	skb_reset_mac_header(skb);
269 
270 	return skb->len;
271 }
272 
273 /**
274  * iptunnel_pmtud_check_icmp() - Trigger ICMP reply if needed and allowed
275  * @skb:	Buffer being sent by encapsulation, L2 headers expected
276  * @mtu:	Network MTU for path
277  *
278  * Return: 0 for no ICMP reply, length if built, negative value on error.
279  */
280 static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
281 {
282 	const struct icmphdr *icmph = icmp_hdr(skb);
283 	const struct iphdr *iph = ip_hdr(skb);
284 
285 	if (mtu < 576 || iph->frag_off != htons(IP_DF))
286 		return 0;
287 
288 	if (ipv4_is_lbcast(iph->daddr)  || ipv4_is_multicast(iph->daddr) ||
289 	    ipv4_is_zeronet(iph->saddr) || ipv4_is_loopback(iph->saddr)  ||
290 	    ipv4_is_lbcast(iph->saddr)  || ipv4_is_multicast(iph->saddr))
291 		return 0;
292 
293 	if (iph->protocol == IPPROTO_ICMP && icmp_is_err(icmph->type))
294 		return 0;
295 
296 	return iptunnel_pmtud_build_icmp(skb, mtu);
297 }
298 
299 #if IS_ENABLED(CONFIG_IPV6)
300 /**
301  * iptunnel_pmtud_build_icmpv6() - Build ICMPv6 error message for PMTUD
302  * @skb:	Original packet with L2 header
303  * @mtu:	MTU value for ICMPv6 error
304  *
305  * Return: length on success, negative error code if message couldn't be built.
306  */
307 static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
308 {
309 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
310 	struct icmp6hdr *icmp6h;
311 	struct ipv6hdr *nip6h;
312 	struct ethhdr eh;
313 	int len, err;
314 	__wsum csum;
315 
316 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
317 		return -EINVAL;
318 
319 	if (skb_is_gso(skb))
320 		skb_gso_reset(skb);
321 
322 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
323 	pskb_pull(skb, ETH_HLEN);
324 	skb_reset_network_header(skb);
325 
326 	err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
327 	if (err)
328 		return err;
329 
330 	len = skb->len + sizeof(*icmp6h);
331 	err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
332 	if (err)
333 		return err;
334 
335 	icmp6h = skb_push(skb, sizeof(*icmp6h));
336 	*icmp6h = (struct icmp6hdr) {
337 		.icmp6_type		= ICMPV6_PKT_TOOBIG,
338 		.icmp6_code		= 0,
339 		.icmp6_cksum		= 0,
340 		.icmp6_mtu		= htonl(mtu),
341 	};
342 	skb_reset_transport_header(skb);
343 
344 	nip6h = skb_push(skb, sizeof(*nip6h));
345 	*nip6h = (struct ipv6hdr) {
346 		.priority		= 0,
347 		.version		= 6,
348 		.flow_lbl		= { 0 },
349 		.payload_len		= htons(len),
350 		.nexthdr		= IPPROTO_ICMPV6,
351 		.hop_limit		= ip6h->hop_limit,
352 		.saddr			= ip6h->daddr,
353 		.daddr			= ip6h->saddr,
354 	};
355 	skb_reset_network_header(skb);
356 
357 	csum = skb_checksum(skb, skb_transport_offset(skb), len, 0);
358 	icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
359 					      IPPROTO_ICMPV6, csum);
360 
361 	skb->ip_summed = CHECKSUM_NONE;
362 
363 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
364 	skb_reset_mac_header(skb);
365 
366 	return skb->len;
367 }
368 
369 /**
370  * iptunnel_pmtud_check_icmpv6() - Trigger ICMPv6 reply if needed and allowed
371  * @skb:	Buffer being sent by encapsulation, L2 headers expected
372  * @mtu:	Network MTU for path
373  *
374  * Return: 0 for no ICMPv6 reply, length if built, negative value on error.
375  */
376 static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
377 {
378 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
379 	int stype = ipv6_addr_type(&ip6h->saddr);
380 	u8 proto = ip6h->nexthdr;
381 	__be16 frag_off;
382 	int offset;
383 
384 	if (mtu < IPV6_MIN_MTU)
385 		return 0;
386 
387 	if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
388 	    stype == IPV6_ADDR_LOOPBACK)
389 		return 0;
390 
391 	offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
392 				  &frag_off);
393 	if (offset < 0 || (frag_off & htons(~0x7)))
394 		return 0;
395 
396 	if (proto == IPPROTO_ICMPV6) {
397 		struct icmp6hdr *icmp6h;
398 
399 		if (!pskb_may_pull(skb, skb_network_header(skb) +
400 					offset + 1 - skb->data))
401 			return 0;
402 
403 		icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
404 		if (icmpv6_is_err(icmp6h->icmp6_type) ||
405 		    icmp6h->icmp6_type == NDISC_REDIRECT)
406 			return 0;
407 	}
408 
409 	return iptunnel_pmtud_build_icmpv6(skb, mtu);
410 }
411 #endif /* IS_ENABLED(CONFIG_IPV6) */
412 
413 /**
414  * skb_tunnel_check_pmtu() - Check, update PMTU and trigger ICMP reply as needed
415  * @skb:	Buffer being sent by encapsulation, L2 headers expected
416  * @encap_dst:	Destination for tunnel encapsulation (outer IP)
417  * @headroom:	Encapsulation header size, bytes
418  * @reply:	Build matching ICMP or ICMPv6 message as a result
419  *
420  * L2 tunnel implementations that can carry IP and can be directly bridged
421  * (currently UDP tunnels) can't always rely on IP forwarding paths to handle
422  * PMTU discovery. In the bridged case, ICMP or ICMPv6 messages need to be built
423  * based on payload and sent back by the encapsulation itself.
424  *
425  * For routable interfaces, we just need to update the PMTU for the destination.
426  *
427  * Return: 0 if ICMP error not needed, length if built, negative value on error
428  */
429 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
430 			  int headroom, bool reply)
431 {
432 	u32 mtu = dst_mtu(encap_dst) - headroom;
433 
434 	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
435 	    (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
436 		return 0;
437 
438 	skb_dst_update_pmtu_no_confirm(skb, mtu);
439 
440 	if (!reply)
441 		return 0;
442 
443 	if (skb->protocol == htons(ETH_P_IP))
444 		return iptunnel_pmtud_check_icmp(skb, mtu);
445 
446 #if IS_ENABLED(CONFIG_IPV6)
447 	if (skb->protocol == htons(ETH_P_IPV6))
448 		return iptunnel_pmtud_check_icmpv6(skb, mtu);
449 #endif
450 	return 0;
451 }
452 EXPORT_SYMBOL(skb_tunnel_check_pmtu);
453 
454 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
455 	[LWTUNNEL_IP_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP_OPTS },
456 	[LWTUNNEL_IP_ID]	= { .type = NLA_U64 },
457 	[LWTUNNEL_IP_DST]	= { .type = NLA_U32 },
458 	[LWTUNNEL_IP_SRC]	= { .type = NLA_U32 },
459 	[LWTUNNEL_IP_TTL]	= { .type = NLA_U8 },
460 	[LWTUNNEL_IP_TOS]	= { .type = NLA_U8 },
461 	[LWTUNNEL_IP_FLAGS]	= { .type = NLA_U16 },
462 	[LWTUNNEL_IP_OPTS]	= { .type = NLA_NESTED },
463 };
464 
465 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
466 	[LWTUNNEL_IP_OPTS_GENEVE]	= { .type = NLA_NESTED },
467 	[LWTUNNEL_IP_OPTS_VXLAN]	= { .type = NLA_NESTED },
468 	[LWTUNNEL_IP_OPTS_ERSPAN]	= { .type = NLA_NESTED },
469 };
470 
471 static const struct nla_policy
472 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
473 	[LWTUNNEL_IP_OPT_GENEVE_CLASS]	= { .type = NLA_U16 },
474 	[LWTUNNEL_IP_OPT_GENEVE_TYPE]	= { .type = NLA_U8 },
475 	[LWTUNNEL_IP_OPT_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 127 },
476 };
477 
478 static const struct nla_policy
479 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
480 	[LWTUNNEL_IP_OPT_VXLAN_GBP]	= { .type = NLA_U32 },
481 };
482 
483 static const struct nla_policy
484 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
485 	[LWTUNNEL_IP_OPT_ERSPAN_VER]	= { .type = NLA_U8 },
486 	[LWTUNNEL_IP_OPT_ERSPAN_INDEX]	= { .type = NLA_U32 },
487 	[LWTUNNEL_IP_OPT_ERSPAN_DIR]	= { .type = NLA_U8 },
488 	[LWTUNNEL_IP_OPT_ERSPAN_HWID]	= { .type = NLA_U8 },
489 };
490 
491 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
492 				    struct ip_tunnel_info *info, int opts_len,
493 				    struct netlink_ext_ack *extack)
494 {
495 	struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
496 	int data_len, err;
497 
498 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
499 			       geneve_opt_policy, extack);
500 	if (err)
501 		return err;
502 
503 	if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
504 	    !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
505 	    !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
506 		return -EINVAL;
507 
508 	attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
509 	data_len = nla_len(attr);
510 	if (data_len % 4)
511 		return -EINVAL;
512 
513 	if (info) {
514 		struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
515 
516 		memcpy(opt->opt_data, nla_data(attr), data_len);
517 		opt->length = data_len / 4;
518 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
519 		opt->opt_class = nla_get_be16(attr);
520 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
521 		opt->type = nla_get_u8(attr);
522 		__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags);
523 	}
524 
525 	return sizeof(struct geneve_opt) + data_len;
526 }
527 
528 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
529 				   struct ip_tunnel_info *info, int opts_len,
530 				   struct netlink_ext_ack *extack)
531 {
532 	struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
533 	int err;
534 
535 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
536 			       vxlan_opt_policy, extack);
537 	if (err)
538 		return err;
539 
540 	if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
541 		return -EINVAL;
542 
543 	if (info) {
544 		struct vxlan_metadata *md =
545 			ip_tunnel_info_opts(info) + opts_len;
546 
547 		attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
548 		md->gbp = nla_get_u32(attr);
549 		md->gbp &= VXLAN_GBP_MASK;
550 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags);
551 	}
552 
553 	return sizeof(struct vxlan_metadata);
554 }
555 
556 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
557 				    struct ip_tunnel_info *info, int opts_len,
558 				    struct netlink_ext_ack *extack)
559 {
560 	struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
561 	int err;
562 	u8 ver;
563 
564 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
565 			       erspan_opt_policy, extack);
566 	if (err)
567 		return err;
568 
569 	if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
570 		return -EINVAL;
571 
572 	ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
573 	if (ver == 1) {
574 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
575 			return -EINVAL;
576 	} else if (ver == 2) {
577 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
578 		    !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
579 			return -EINVAL;
580 	} else {
581 		return -EINVAL;
582 	}
583 
584 	if (info) {
585 		struct erspan_metadata *md =
586 			ip_tunnel_info_opts(info) + opts_len;
587 
588 		md->version = ver;
589 		if (ver == 1) {
590 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
591 			md->u.index = nla_get_be32(attr);
592 		} else {
593 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
594 			md->u.md2.dir = nla_get_u8(attr);
595 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
596 			set_hwid(&md->u.md2, nla_get_u8(attr));
597 		}
598 
599 		__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags);
600 	}
601 
602 	return sizeof(struct erspan_metadata);
603 }
604 
605 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
606 			     struct netlink_ext_ack *extack)
607 {
608 	int err, rem, opt_len, opts_len = 0;
609 	struct nlattr *nla;
610 	u32 type = 0;
611 
612 	if (!attr)
613 		return 0;
614 
615 	err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
616 			   ip_opts_policy, extack);
617 	if (err)
618 		return err;
619 
620 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
621 		switch (nla_type(nla)) {
622 		case LWTUNNEL_IP_OPTS_GENEVE:
623 			if (type && type != IP_TUNNEL_GENEVE_OPT_BIT)
624 				return -EINVAL;
625 			opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
626 							   extack);
627 			if (opt_len < 0)
628 				return opt_len;
629 			opts_len += opt_len;
630 			if (opts_len > IP_TUNNEL_OPTS_MAX)
631 				return -EINVAL;
632 			type = IP_TUNNEL_GENEVE_OPT_BIT;
633 			break;
634 		case LWTUNNEL_IP_OPTS_VXLAN:
635 			if (type)
636 				return -EINVAL;
637 			opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
638 							  extack);
639 			if (opt_len < 0)
640 				return opt_len;
641 			opts_len += opt_len;
642 			type = IP_TUNNEL_VXLAN_OPT_BIT;
643 			break;
644 		case LWTUNNEL_IP_OPTS_ERSPAN:
645 			if (type)
646 				return -EINVAL;
647 			opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
648 							   extack);
649 			if (opt_len < 0)
650 				return opt_len;
651 			opts_len += opt_len;
652 			type = IP_TUNNEL_ERSPAN_OPT_BIT;
653 			break;
654 		default:
655 			return -EINVAL;
656 		}
657 	}
658 
659 	return opts_len;
660 }
661 
662 static int ip_tun_get_optlen(struct nlattr *attr,
663 			     struct netlink_ext_ack *extack)
664 {
665 	return ip_tun_parse_opts(attr, NULL, extack);
666 }
667 
668 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
669 			   struct netlink_ext_ack *extack)
670 {
671 	return ip_tun_parse_opts(attr, info, extack);
672 }
673 
674 static int ip_tun_build_state(struct net *net, struct nlattr *attr,
675 			      unsigned int family, const void *cfg,
676 			      struct lwtunnel_state **ts,
677 			      struct netlink_ext_ack *extack)
678 {
679 	struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
680 	struct lwtunnel_state *new_state;
681 	struct ip_tunnel_info *tun_info;
682 	int err, opt_len;
683 
684 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
685 					  ip_tun_policy, extack);
686 	if (err < 0)
687 		return err;
688 
689 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
690 	if (opt_len < 0)
691 		return opt_len;
692 
693 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
694 	if (!new_state)
695 		return -ENOMEM;
696 
697 	new_state->type = LWTUNNEL_ENCAP_IP;
698 
699 	tun_info = lwt_tun_info(new_state);
700 
701 	err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
702 	if (err < 0) {
703 		lwtstate_free(new_state);
704 		return err;
705 	}
706 
707 #ifdef CONFIG_DST_CACHE
708 	err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
709 	if (err) {
710 		lwtstate_free(new_state);
711 		return err;
712 	}
713 #endif
714 
715 	if (tb[LWTUNNEL_IP_ID])
716 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
717 
718 	if (tb[LWTUNNEL_IP_DST])
719 		tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
720 
721 	if (tb[LWTUNNEL_IP_SRC])
722 		tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
723 
724 	if (tb[LWTUNNEL_IP_TTL])
725 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
726 
727 	if (tb[LWTUNNEL_IP_TOS])
728 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
729 
730 	if (tb[LWTUNNEL_IP_FLAGS]) {
731 		IP_TUNNEL_DECLARE_FLAGS(flags);
732 
733 		ip_tunnel_flags_from_be16(flags,
734 					  nla_get_be16(tb[LWTUNNEL_IP_FLAGS]));
735 		ip_tunnel_clear_options_present(flags);
736 
737 		ip_tunnel_flags_or(tun_info->key.tun_flags,
738 				   tun_info->key.tun_flags, flags);
739 	}
740 
741 	tun_info->mode = IP_TUNNEL_INFO_TX;
742 	tun_info->options_len = opt_len;
743 
744 	*ts = new_state;
745 
746 	return 0;
747 }
748 
749 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
750 {
751 #ifdef CONFIG_DST_CACHE
752 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
753 
754 	dst_cache_destroy(&tun_info->dst_cache);
755 #endif
756 }
757 
758 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
759 					 struct ip_tunnel_info *tun_info)
760 {
761 	struct geneve_opt *opt;
762 	struct nlattr *nest;
763 	int offset = 0;
764 
765 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
766 	if (!nest)
767 		return -ENOMEM;
768 
769 	while (tun_info->options_len > offset) {
770 		opt = ip_tunnel_info_opts(tun_info) + offset;
771 		if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
772 				 opt->opt_class) ||
773 		    nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
774 		    nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
775 			    opt->opt_data)) {
776 			nla_nest_cancel(skb, nest);
777 			return -ENOMEM;
778 		}
779 		offset += sizeof(*opt) + opt->length * 4;
780 	}
781 
782 	nla_nest_end(skb, nest);
783 	return 0;
784 }
785 
786 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
787 					struct ip_tunnel_info *tun_info)
788 {
789 	struct vxlan_metadata *md;
790 	struct nlattr *nest;
791 
792 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
793 	if (!nest)
794 		return -ENOMEM;
795 
796 	md = ip_tunnel_info_opts(tun_info);
797 	if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
798 		nla_nest_cancel(skb, nest);
799 		return -ENOMEM;
800 	}
801 
802 	nla_nest_end(skb, nest);
803 	return 0;
804 }
805 
806 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
807 					 struct ip_tunnel_info *tun_info)
808 {
809 	struct erspan_metadata *md;
810 	struct nlattr *nest;
811 
812 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
813 	if (!nest)
814 		return -ENOMEM;
815 
816 	md = ip_tunnel_info_opts(tun_info);
817 	if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
818 		goto err;
819 
820 	if (md->version == 1 &&
821 	    nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
822 		goto err;
823 
824 	if (md->version == 2 &&
825 	    (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
826 	     nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
827 			get_hwid(&md->u.md2))))
828 		goto err;
829 
830 	nla_nest_end(skb, nest);
831 	return 0;
832 err:
833 	nla_nest_cancel(skb, nest);
834 	return -ENOMEM;
835 }
836 
837 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
838 				  struct ip_tunnel_info *tun_info)
839 {
840 	struct nlattr *nest;
841 	int err = 0;
842 
843 	if (!ip_tunnel_is_options_present(tun_info->key.tun_flags))
844 		return 0;
845 
846 	nest = nla_nest_start_noflag(skb, type);
847 	if (!nest)
848 		return -ENOMEM;
849 
850 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_info->key.tun_flags))
851 		err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
852 	else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_info->key.tun_flags))
853 		err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
854 	else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
855 		err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
856 
857 	if (err) {
858 		nla_nest_cancel(skb, nest);
859 		return err;
860 	}
861 
862 	nla_nest_end(skb, nest);
863 	return 0;
864 }
865 
866 static int ip_tun_fill_encap_info(struct sk_buff *skb,
867 				  struct lwtunnel_state *lwtstate)
868 {
869 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
870 
871 	if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
872 			 LWTUNNEL_IP_PAD) ||
873 	    nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
874 	    nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
875 	    nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
876 	    nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
877 	    nla_put_be16(skb, LWTUNNEL_IP_FLAGS,
878 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
879 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
880 		return -ENOMEM;
881 
882 	return 0;
883 }
884 
885 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
886 {
887 	int opt_len;
888 
889 	if (!ip_tunnel_is_options_present(info->key.tun_flags))
890 		return 0;
891 
892 	opt_len = nla_total_size(0);		/* LWTUNNEL_IP_OPTS */
893 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags)) {
894 		struct geneve_opt *opt;
895 		int offset = 0;
896 
897 		opt_len += nla_total_size(0);	/* LWTUNNEL_IP_OPTS_GENEVE */
898 		while (info->options_len > offset) {
899 			opt = ip_tunnel_info_opts(info) + offset;
900 			opt_len += nla_total_size(2)	/* OPT_GENEVE_CLASS */
901 				   + nla_total_size(1)	/* OPT_GENEVE_TYPE */
902 				   + nla_total_size(opt->length * 4);
903 							/* OPT_GENEVE_DATA */
904 			offset += sizeof(*opt) + opt->length * 4;
905 		}
906 	} else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
907 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_VXLAN */
908 			   + nla_total_size(4);	/* OPT_VXLAN_GBP */
909 	} else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags)) {
910 		struct erspan_metadata *md = ip_tunnel_info_opts(info);
911 
912 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_ERSPAN */
913 			   + nla_total_size(1)	/* OPT_ERSPAN_VER */
914 			   + (md->version == 1 ? nla_total_size(4)
915 						/* OPT_ERSPAN_INDEX (v1) */
916 					       : nla_total_size(1) +
917 						 nla_total_size(1));
918 						/* OPT_ERSPAN_DIR + HWID (v2) */
919 	}
920 
921 	return opt_len;
922 }
923 
924 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
925 {
926 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP_ID */
927 		+ nla_total_size(4)	/* LWTUNNEL_IP_DST */
928 		+ nla_total_size(4)	/* LWTUNNEL_IP_SRC */
929 		+ nla_total_size(1)	/* LWTUNNEL_IP_TOS */
930 		+ nla_total_size(1)	/* LWTUNNEL_IP_TTL */
931 		+ nla_total_size(2)	/* LWTUNNEL_IP_FLAGS */
932 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
933 					/* LWTUNNEL_IP_OPTS */
934 }
935 
936 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
937 {
938 	struct ip_tunnel_info *info_a = lwt_tun_info(a);
939 	struct ip_tunnel_info *info_b = lwt_tun_info(b);
940 
941 	return memcmp(info_a, info_b, sizeof(info_a->key)) ||
942 	       info_a->mode != info_b->mode ||
943 	       info_a->options_len != info_b->options_len ||
944 	       memcmp(ip_tunnel_info_opts(info_a),
945 		      ip_tunnel_info_opts(info_b), info_a->options_len);
946 }
947 
948 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
949 	.build_state = ip_tun_build_state,
950 	.destroy_state = ip_tun_destroy_state,
951 	.fill_encap = ip_tun_fill_encap_info,
952 	.get_encap_size = ip_tun_encap_nlsize,
953 	.cmp_encap = ip_tun_cmp_encap,
954 	.owner = THIS_MODULE,
955 };
956 
957 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
958 	[LWTUNNEL_IP6_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP6_OPTS },
959 	[LWTUNNEL_IP6_ID]		= { .type = NLA_U64 },
960 	[LWTUNNEL_IP6_DST]		= { .len = sizeof(struct in6_addr) },
961 	[LWTUNNEL_IP6_SRC]		= { .len = sizeof(struct in6_addr) },
962 	[LWTUNNEL_IP6_HOPLIMIT]		= { .type = NLA_U8 },
963 	[LWTUNNEL_IP6_TC]		= { .type = NLA_U8 },
964 	[LWTUNNEL_IP6_FLAGS]		= { .type = NLA_U16 },
965 	[LWTUNNEL_IP6_OPTS]		= { .type = NLA_NESTED },
966 };
967 
968 static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
969 			       unsigned int family, const void *cfg,
970 			       struct lwtunnel_state **ts,
971 			       struct netlink_ext_ack *extack)
972 {
973 	struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
974 	struct lwtunnel_state *new_state;
975 	struct ip_tunnel_info *tun_info;
976 	int err, opt_len;
977 
978 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
979 					  ip6_tun_policy, extack);
980 	if (err < 0)
981 		return err;
982 
983 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
984 	if (opt_len < 0)
985 		return opt_len;
986 
987 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
988 	if (!new_state)
989 		return -ENOMEM;
990 
991 	new_state->type = LWTUNNEL_ENCAP_IP6;
992 
993 	tun_info = lwt_tun_info(new_state);
994 
995 	err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
996 	if (err < 0) {
997 		lwtstate_free(new_state);
998 		return err;
999 	}
1000 
1001 	if (tb[LWTUNNEL_IP6_ID])
1002 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
1003 
1004 	if (tb[LWTUNNEL_IP6_DST])
1005 		tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
1006 
1007 	if (tb[LWTUNNEL_IP6_SRC])
1008 		tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
1009 
1010 	if (tb[LWTUNNEL_IP6_HOPLIMIT])
1011 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
1012 
1013 	if (tb[LWTUNNEL_IP6_TC])
1014 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
1015 
1016 	if (tb[LWTUNNEL_IP6_FLAGS]) {
1017 		IP_TUNNEL_DECLARE_FLAGS(flags);
1018 		__be16 data;
1019 
1020 		data = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
1021 		ip_tunnel_flags_from_be16(flags, data);
1022 		ip_tunnel_clear_options_present(flags);
1023 
1024 		ip_tunnel_flags_or(tun_info->key.tun_flags,
1025 				   tun_info->key.tun_flags, flags);
1026 	}
1027 
1028 	tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
1029 	tun_info->options_len = opt_len;
1030 
1031 	*ts = new_state;
1032 
1033 	return 0;
1034 }
1035 
1036 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1037 				   struct lwtunnel_state *lwtstate)
1038 {
1039 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1040 
1041 	if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1042 			 LWTUNNEL_IP6_PAD) ||
1043 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1044 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1045 	    nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1046 	    nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1047 	    nla_put_be16(skb, LWTUNNEL_IP6_FLAGS,
1048 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
1049 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1050 		return -ENOMEM;
1051 
1052 	return 0;
1053 }
1054 
1055 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1056 {
1057 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP6_ID */
1058 		+ nla_total_size(16)	/* LWTUNNEL_IP6_DST */
1059 		+ nla_total_size(16)	/* LWTUNNEL_IP6_SRC */
1060 		+ nla_total_size(1)	/* LWTUNNEL_IP6_HOPLIMIT */
1061 		+ nla_total_size(1)	/* LWTUNNEL_IP6_TC */
1062 		+ nla_total_size(2)	/* LWTUNNEL_IP6_FLAGS */
1063 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1064 					/* LWTUNNEL_IP6_OPTS */
1065 }
1066 
1067 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1068 	.build_state = ip6_tun_build_state,
1069 	.fill_encap = ip6_tun_fill_encap_info,
1070 	.get_encap_size = ip6_tun_encap_nlsize,
1071 	.cmp_encap = ip_tun_cmp_encap,
1072 	.owner = THIS_MODULE,
1073 };
1074 
1075 void __init ip_tunnel_core_init(void)
1076 {
1077 	/* If you land here, make sure whether increasing ip_tunnel_info's
1078 	 * options_len is a reasonable choice with its usage in front ends
1079 	 * (f.e., it's part of flow keys, etc).
1080 	 */
1081 	BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1082 
1083 	lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1084 	lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1085 }
1086 
1087 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1088 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1089 
1090 void ip_tunnel_need_metadata(void)
1091 {
1092 	static_branch_inc(&ip_tunnel_metadata_cnt);
1093 }
1094 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1095 
1096 void ip_tunnel_unneed_metadata(void)
1097 {
1098 	static_branch_dec(&ip_tunnel_metadata_cnt);
1099 }
1100 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1101 
1102 /* Returns either the correct skb->protocol value, or 0 if invalid. */
1103 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1104 {
1105 	if (skb_network_header(skb) >= skb->head &&
1106 	    (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1107 	    ip_hdr(skb)->version == 4)
1108 		return htons(ETH_P_IP);
1109 	if (skb_network_header(skb) >= skb->head &&
1110 	    (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1111 	    ipv6_hdr(skb)->version == 6)
1112 		return htons(ETH_P_IPV6);
1113 	return 0;
1114 }
1115 EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1116 
1117 const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1118 EXPORT_SYMBOL(ip_tunnel_header_ops);
1119 
1120 /* This function returns true when ENCAP attributes are present in the nl msg */
1121 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
1122 				   struct ip_tunnel_encap *encap)
1123 {
1124 	bool ret = false;
1125 
1126 	memset(encap, 0, sizeof(*encap));
1127 
1128 	if (!data)
1129 		return ret;
1130 
1131 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1132 		ret = true;
1133 		encap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1134 	}
1135 
1136 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1137 		ret = true;
1138 		encap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1139 	}
1140 
1141 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1142 		ret = true;
1143 		encap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1144 	}
1145 
1146 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1147 		ret = true;
1148 		encap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1149 	}
1150 
1151 	return ret;
1152 }
1153 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_encap_parms);
1154 
1155 void ip_tunnel_netlink_parms(struct nlattr *data[],
1156 			     struct ip_tunnel_parm_kern *parms)
1157 {
1158 	if (data[IFLA_IPTUN_LINK])
1159 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1160 
1161 	if (data[IFLA_IPTUN_LOCAL])
1162 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1163 
1164 	if (data[IFLA_IPTUN_REMOTE])
1165 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1166 
1167 	if (data[IFLA_IPTUN_TTL]) {
1168 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1169 		if (parms->iph.ttl)
1170 			parms->iph.frag_off = htons(IP_DF);
1171 	}
1172 
1173 	if (data[IFLA_IPTUN_TOS])
1174 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1175 
1176 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1177 		parms->iph.frag_off = htons(IP_DF);
1178 
1179 	if (data[IFLA_IPTUN_FLAGS]) {
1180 		__be16 flags;
1181 
1182 		flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1183 		ip_tunnel_flags_from_be16(parms->i_flags, flags);
1184 	}
1185 
1186 	if (data[IFLA_IPTUN_PROTO])
1187 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1188 }
1189 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_parms);
1190