xref: /linux/net/ipv4/ip_tunnel_core.c (revision 3e20009988e2470063824c58b19d1c80816cc46d)
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 
iptunnel_xmit(struct sock * sk,struct rtable * rt,struct sk_buff * skb,__be32 src,__be32 dst,__u8 proto,__u8 tos,__u8 ttl,__be16 df,bool xnet,u16 ipcb_flags)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 
__iptunnel_pull_header(struct sk_buff * skb,int hdr_len,__be16 inner_proto,bool raw_proto,bool xnet)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 
iptunnel_metadata_reply(struct metadata_dst * md,gfp_t flags)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 
iptunnel_handle_offloads(struct sk_buff * skb,int gso_type_mask)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  */
iptunnel_pmtud_build_icmp(struct sk_buff * skb,int mtu)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 	return skb->len;
272 }
273 
274 /**
275  * iptunnel_pmtud_check_icmp() - Trigger ICMP reply if needed and allowed
276  * @skb:	Buffer being sent by encapsulation, L2 headers expected
277  * @mtu:	Network MTU for path
278  *
279  * Return: 0 for no ICMP reply, length if built, negative value on error.
280  */
iptunnel_pmtud_check_icmp(struct sk_buff * skb,int mtu)281 static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
282 {
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) {
294 		const struct icmphdr *icmph;
295 
296 		if (!pskb_network_may_pull(skb, iph->ihl * 4 +
297 						offsetofend(struct icmphdr, type)))
298 			return 0;
299 		iph = ip_hdr(skb);
300 		icmph = (void *)iph + iph->ihl * 4;
301 		if (icmp_is_err(icmph->type))
302 			return 0;
303 	}
304 	return iptunnel_pmtud_build_icmp(skb, mtu);
305 }
306 
307 #if IS_ENABLED(CONFIG_IPV6)
308 /**
309  * iptunnel_pmtud_build_icmpv6() - Build ICMPv6 error message for PMTUD
310  * @skb:	Original packet with L2 header
311  * @mtu:	MTU value for ICMPv6 error
312  *
313  * Return: length on success, negative error code if message couldn't be built.
314  */
iptunnel_pmtud_build_icmpv6(struct sk_buff * skb,int mtu)315 static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
316 {
317 	const struct ipv6hdr *ip6h;
318 	struct icmp6hdr *icmp6h;
319 	struct ipv6hdr *nip6h;
320 	struct ethhdr eh;
321 	int len, err;
322 	__wsum csum;
323 
324 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
325 		return -EINVAL;
326 
327 	if (skb_is_gso(skb))
328 		skb_gso_reset(skb);
329 
330 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
331 	pskb_pull(skb, ETH_HLEN);
332 
333 	err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
334 	if (err)
335 		return err;
336 
337 	len = skb->len + sizeof(*icmp6h);
338 	err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
339 	if (err)
340 		return err;
341 
342 	ip6h = ipv6_hdr(skb);
343 	icmp6h = skb_push(skb, sizeof(*icmp6h));
344 	*icmp6h = (struct icmp6hdr) {
345 		.icmp6_type		= ICMPV6_PKT_TOOBIG,
346 		.icmp6_code		= 0,
347 		.icmp6_cksum		= 0,
348 		.icmp6_mtu		= htonl(mtu),
349 	};
350 	skb_reset_transport_header(skb);
351 
352 	nip6h = skb_push(skb, sizeof(*nip6h));
353 	*nip6h = (struct ipv6hdr) {
354 		.priority		= 0,
355 		.version		= 6,
356 		.flow_lbl		= { 0 },
357 		.payload_len		= htons(len),
358 		.nexthdr		= IPPROTO_ICMPV6,
359 		.hop_limit		= ip6h->hop_limit,
360 		.saddr			= ip6h->daddr,
361 		.daddr			= ip6h->saddr,
362 	};
363 	skb_reset_network_header(skb);
364 
365 	csum = skb_checksum(skb, skb_transport_offset(skb), len, 0);
366 	icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
367 					      IPPROTO_ICMPV6, csum);
368 
369 	skb->ip_summed = CHECKSUM_NONE;
370 
371 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
372 	skb_reset_mac_header(skb);
373 
374 	return skb->len;
375 }
376 
377 /**
378  * iptunnel_pmtud_check_icmpv6() - Trigger ICMPv6 reply if needed and allowed
379  * @skb:	Buffer being sent by encapsulation, L2 headers expected
380  * @mtu:	Network MTU for path
381  *
382  * Return: 0 for no ICMPv6 reply, length if built, negative value on error.
383  */
iptunnel_pmtud_check_icmpv6(struct sk_buff * skb,int mtu)384 static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
385 {
386 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
387 	int stype = ipv6_addr_type(&ip6h->saddr);
388 	u8 proto = ip6h->nexthdr;
389 	__be16 frag_off;
390 	int offset;
391 
392 	if (mtu < IPV6_MIN_MTU)
393 		return 0;
394 
395 	if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
396 	    stype == IPV6_ADDR_LOOPBACK)
397 		return 0;
398 
399 	offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
400 				  &frag_off);
401 	if (offset < 0 || (frag_off & htons(~0x7)))
402 		return 0;
403 
404 	if (proto == IPPROTO_ICMPV6) {
405 		struct icmp6hdr *icmp6h;
406 
407 		if (!pskb_may_pull(skb, skb_network_header(skb) +
408 					offset + 1 - skb->data))
409 			return 0;
410 
411 		icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
412 		if (icmpv6_is_err(icmp6h->icmp6_type) ||
413 		    icmp6h->icmp6_type == NDISC_REDIRECT)
414 			return 0;
415 	}
416 
417 	return iptunnel_pmtud_build_icmpv6(skb, mtu);
418 }
419 #endif /* IS_ENABLED(CONFIG_IPV6) */
420 
421 /**
422  * skb_tunnel_check_pmtu() - Check, update PMTU and trigger ICMP reply as needed
423  * @skb:	Buffer being sent by encapsulation, L2 headers expected
424  * @encap_dst:	Destination for tunnel encapsulation (outer IP)
425  * @headroom:	Encapsulation header size, bytes
426  * @reply:	Build matching ICMP or ICMPv6 message as a result
427  *
428  * L2 tunnel implementations that can carry IP and can be directly bridged
429  * (currently UDP tunnels) can't always rely on IP forwarding paths to handle
430  * PMTU discovery. In the bridged case, ICMP or ICMPv6 messages need to be built
431  * based on payload and sent back by the encapsulation itself.
432  *
433  * For routable interfaces, we just need to update the PMTU for the destination.
434  *
435  * Return: 0 if ICMP error not needed, length if built, negative value on error
436  */
skb_tunnel_check_pmtu(struct sk_buff * skb,struct dst_entry * encap_dst,int headroom,bool reply)437 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
438 			  int headroom, bool reply)
439 {
440 	u32 mtu = dst_mtu(encap_dst) - headroom;
441 
442 	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
443 	    (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
444 		return 0;
445 
446 	skb_dst_update_pmtu_no_confirm(skb, mtu);
447 
448 	if (!reply)
449 		return 0;
450 
451 	if (skb->protocol == htons(ETH_P_IP))
452 		return iptunnel_pmtud_check_icmp(skb, mtu);
453 
454 #if IS_ENABLED(CONFIG_IPV6)
455 	if (skb->protocol == htons(ETH_P_IPV6))
456 		return iptunnel_pmtud_check_icmpv6(skb, mtu);
457 #endif
458 	return 0;
459 }
460 EXPORT_SYMBOL(skb_tunnel_check_pmtu);
461 
462 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
463 	[LWTUNNEL_IP_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP_OPTS },
464 	[LWTUNNEL_IP_ID]	= { .type = NLA_U64 },
465 	[LWTUNNEL_IP_DST]	= { .type = NLA_U32 },
466 	[LWTUNNEL_IP_SRC]	= { .type = NLA_U32 },
467 	[LWTUNNEL_IP_TTL]	= { .type = NLA_U8 },
468 	[LWTUNNEL_IP_TOS]	= { .type = NLA_U8 },
469 	[LWTUNNEL_IP_FLAGS]	= { .type = NLA_U16 },
470 	[LWTUNNEL_IP_OPTS]	= { .type = NLA_NESTED },
471 };
472 
473 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
474 	[LWTUNNEL_IP_OPTS_GENEVE]	= { .type = NLA_NESTED },
475 	[LWTUNNEL_IP_OPTS_VXLAN]	= { .type = NLA_NESTED },
476 	[LWTUNNEL_IP_OPTS_ERSPAN]	= { .type = NLA_NESTED },
477 };
478 
479 static const struct nla_policy
480 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
481 	[LWTUNNEL_IP_OPT_GENEVE_CLASS]	= { .type = NLA_U16 },
482 	[LWTUNNEL_IP_OPT_GENEVE_TYPE]	= { .type = NLA_U8 },
483 	[LWTUNNEL_IP_OPT_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 127 },
484 };
485 
486 static const struct nla_policy
487 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
488 	[LWTUNNEL_IP_OPT_VXLAN_GBP]	= { .type = NLA_U32 },
489 };
490 
491 static const struct nla_policy
492 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
493 	[LWTUNNEL_IP_OPT_ERSPAN_VER]	= { .type = NLA_U8 },
494 	[LWTUNNEL_IP_OPT_ERSPAN_INDEX]	= { .type = NLA_U32 },
495 	[LWTUNNEL_IP_OPT_ERSPAN_DIR]	= { .type = NLA_U8 },
496 	[LWTUNNEL_IP_OPT_ERSPAN_HWID]	= { .type = NLA_U8 },
497 };
498 
ip_tun_parse_opts_geneve(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)499 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
500 				    struct ip_tunnel_info *info, int opts_len,
501 				    struct netlink_ext_ack *extack)
502 {
503 	struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
504 	int data_len, err;
505 
506 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
507 			       geneve_opt_policy, extack);
508 	if (err)
509 		return err;
510 
511 	if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
512 	    !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
513 	    !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
514 		return -EINVAL;
515 
516 	attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
517 	data_len = nla_len(attr);
518 	if (data_len % 4)
519 		return -EINVAL;
520 
521 	if (info) {
522 		struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
523 
524 		memcpy(opt->opt_data, nla_data(attr), data_len);
525 		opt->length = data_len / 4;
526 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
527 		opt->opt_class = nla_get_be16(attr);
528 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
529 		opt->type = nla_get_u8(attr);
530 		__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags);
531 	}
532 
533 	return sizeof(struct geneve_opt) + data_len;
534 }
535 
ip_tun_parse_opts_vxlan(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)536 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
537 				   struct ip_tunnel_info *info, int opts_len,
538 				   struct netlink_ext_ack *extack)
539 {
540 	struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
541 	int err;
542 
543 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
544 			       vxlan_opt_policy, extack);
545 	if (err)
546 		return err;
547 
548 	if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
549 		return -EINVAL;
550 
551 	if (info) {
552 		struct vxlan_metadata *md =
553 			ip_tunnel_info_opts(info) + opts_len;
554 
555 		attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
556 		md->gbp = nla_get_u32(attr);
557 		md->gbp &= VXLAN_GBP_MASK;
558 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags);
559 	}
560 
561 	return sizeof(struct vxlan_metadata);
562 }
563 
ip_tun_parse_opts_erspan(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)564 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
565 				    struct ip_tunnel_info *info, int opts_len,
566 				    struct netlink_ext_ack *extack)
567 {
568 	struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
569 	int err;
570 	u8 ver;
571 
572 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
573 			       erspan_opt_policy, extack);
574 	if (err)
575 		return err;
576 
577 	if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
578 		return -EINVAL;
579 
580 	ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
581 	if (ver == 1) {
582 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
583 			return -EINVAL;
584 	} else if (ver == 2) {
585 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
586 		    !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
587 			return -EINVAL;
588 	} else {
589 		return -EINVAL;
590 	}
591 
592 	if (info) {
593 		struct erspan_metadata *md =
594 			ip_tunnel_info_opts(info) + opts_len;
595 
596 		md->version = ver;
597 		if (ver == 1) {
598 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
599 			md->u.index = nla_get_be32(attr);
600 		} else {
601 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
602 			md->u.md2.dir = nla_get_u8(attr);
603 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
604 			set_hwid(&md->u.md2, nla_get_u8(attr));
605 		}
606 
607 		__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags);
608 	}
609 
610 	return sizeof(struct erspan_metadata);
611 }
612 
ip_tun_parse_opts(struct nlattr * attr,struct ip_tunnel_info * info,struct netlink_ext_ack * extack)613 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
614 			     struct netlink_ext_ack *extack)
615 {
616 	int err, rem, opt_len, opts_len = 0;
617 	struct nlattr *nla;
618 	u32 type = 0;
619 
620 	if (!attr)
621 		return 0;
622 
623 	err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
624 			   ip_opts_policy, extack);
625 	if (err)
626 		return err;
627 
628 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
629 		switch (nla_type(nla)) {
630 		case LWTUNNEL_IP_OPTS_GENEVE:
631 			if (type && type != IP_TUNNEL_GENEVE_OPT_BIT)
632 				return -EINVAL;
633 			opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
634 							   extack);
635 			if (opt_len < 0)
636 				return opt_len;
637 			opts_len += opt_len;
638 			if (opts_len > IP_TUNNEL_OPTS_MAX)
639 				return -EINVAL;
640 			type = IP_TUNNEL_GENEVE_OPT_BIT;
641 			break;
642 		case LWTUNNEL_IP_OPTS_VXLAN:
643 			if (type)
644 				return -EINVAL;
645 			opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
646 							  extack);
647 			if (opt_len < 0)
648 				return opt_len;
649 			opts_len += opt_len;
650 			type = IP_TUNNEL_VXLAN_OPT_BIT;
651 			break;
652 		case LWTUNNEL_IP_OPTS_ERSPAN:
653 			if (type)
654 				return -EINVAL;
655 			opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
656 							   extack);
657 			if (opt_len < 0)
658 				return opt_len;
659 			opts_len += opt_len;
660 			type = IP_TUNNEL_ERSPAN_OPT_BIT;
661 			break;
662 		default:
663 			return -EINVAL;
664 		}
665 	}
666 
667 	return opts_len;
668 }
669 
ip_tun_get_optlen(struct nlattr * attr,struct netlink_ext_ack * extack)670 static int ip_tun_get_optlen(struct nlattr *attr,
671 			     struct netlink_ext_ack *extack)
672 {
673 	return ip_tun_parse_opts(attr, NULL, extack);
674 }
675 
ip_tun_set_opts(struct nlattr * attr,struct ip_tunnel_info * info,struct netlink_ext_ack * extack)676 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
677 			   struct netlink_ext_ack *extack)
678 {
679 	return ip_tun_parse_opts(attr, info, extack);
680 }
681 
ip_tun_build_state(struct net * net,struct nlattr * attr,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)682 static int ip_tun_build_state(struct net *net, struct nlattr *attr,
683 			      unsigned int family, const void *cfg,
684 			      struct lwtunnel_state **ts,
685 			      struct netlink_ext_ack *extack)
686 {
687 	struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
688 	struct lwtunnel_state *new_state;
689 	struct ip_tunnel_info *tun_info;
690 	int err, opt_len;
691 
692 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
693 					  ip_tun_policy, extack);
694 	if (err < 0)
695 		return err;
696 
697 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
698 	if (opt_len < 0)
699 		return opt_len;
700 
701 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
702 	if (!new_state)
703 		return -ENOMEM;
704 
705 	new_state->type = LWTUNNEL_ENCAP_IP;
706 
707 	tun_info = lwt_tun_info(new_state);
708 
709 	err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
710 	if (err < 0) {
711 		lwtstate_free(new_state);
712 		return err;
713 	}
714 
715 #ifdef CONFIG_DST_CACHE
716 	err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
717 	if (err) {
718 		lwtstate_free(new_state);
719 		return err;
720 	}
721 #endif
722 
723 	if (tb[LWTUNNEL_IP_ID])
724 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
725 
726 	if (tb[LWTUNNEL_IP_DST])
727 		tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
728 
729 	if (tb[LWTUNNEL_IP_SRC])
730 		tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
731 
732 	if (tb[LWTUNNEL_IP_TTL])
733 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
734 
735 	if (tb[LWTUNNEL_IP_TOS])
736 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
737 
738 	if (tb[LWTUNNEL_IP_FLAGS]) {
739 		IP_TUNNEL_DECLARE_FLAGS(flags);
740 
741 		ip_tunnel_flags_from_be16(flags,
742 					  nla_get_be16(tb[LWTUNNEL_IP_FLAGS]));
743 		ip_tunnel_clear_options_present(flags);
744 
745 		ip_tunnel_flags_or(tun_info->key.tun_flags,
746 				   tun_info->key.tun_flags, flags);
747 	}
748 
749 	tun_info->mode = IP_TUNNEL_INFO_TX;
750 	tun_info->options_len = opt_len;
751 
752 	*ts = new_state;
753 
754 	return 0;
755 }
756 
ip_tun_destroy_state(struct lwtunnel_state * lwtstate)757 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
758 {
759 #ifdef CONFIG_DST_CACHE
760 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
761 
762 	dst_cache_destroy(&tun_info->dst_cache);
763 #endif
764 }
765 
ip_tun_fill_encap_opts_geneve(struct sk_buff * skb,struct ip_tunnel_info * tun_info)766 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
767 					 struct ip_tunnel_info *tun_info)
768 {
769 	struct geneve_opt *opt;
770 	struct nlattr *nest;
771 	int offset = 0;
772 
773 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
774 	if (!nest)
775 		return -ENOMEM;
776 
777 	while (tun_info->options_len > offset) {
778 		opt = ip_tunnel_info_opts(tun_info) + offset;
779 		if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
780 				 opt->opt_class) ||
781 		    nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
782 		    nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
783 			    opt->opt_data)) {
784 			nla_nest_cancel(skb, nest);
785 			return -ENOMEM;
786 		}
787 		offset += sizeof(*opt) + opt->length * 4;
788 	}
789 
790 	nla_nest_end(skb, nest);
791 	return 0;
792 }
793 
ip_tun_fill_encap_opts_vxlan(struct sk_buff * skb,struct ip_tunnel_info * tun_info)794 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
795 					struct ip_tunnel_info *tun_info)
796 {
797 	struct vxlan_metadata *md;
798 	struct nlattr *nest;
799 
800 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
801 	if (!nest)
802 		return -ENOMEM;
803 
804 	md = ip_tunnel_info_opts(tun_info);
805 	if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
806 		nla_nest_cancel(skb, nest);
807 		return -ENOMEM;
808 	}
809 
810 	nla_nest_end(skb, nest);
811 	return 0;
812 }
813 
ip_tun_fill_encap_opts_erspan(struct sk_buff * skb,struct ip_tunnel_info * tun_info)814 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
815 					 struct ip_tunnel_info *tun_info)
816 {
817 	struct erspan_metadata *md;
818 	struct nlattr *nest;
819 
820 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
821 	if (!nest)
822 		return -ENOMEM;
823 
824 	md = ip_tunnel_info_opts(tun_info);
825 	if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
826 		goto err;
827 
828 	if (md->version == 1 &&
829 	    nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
830 		goto err;
831 
832 	if (md->version == 2 &&
833 	    (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
834 	     nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
835 			get_hwid(&md->u.md2))))
836 		goto err;
837 
838 	nla_nest_end(skb, nest);
839 	return 0;
840 err:
841 	nla_nest_cancel(skb, nest);
842 	return -ENOMEM;
843 }
844 
ip_tun_fill_encap_opts(struct sk_buff * skb,int type,struct ip_tunnel_info * tun_info)845 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
846 				  struct ip_tunnel_info *tun_info)
847 {
848 	struct nlattr *nest;
849 	int err = 0;
850 
851 	if (!ip_tunnel_is_options_present(tun_info->key.tun_flags))
852 		return 0;
853 
854 	nest = nla_nest_start_noflag(skb, type);
855 	if (!nest)
856 		return -ENOMEM;
857 
858 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_info->key.tun_flags))
859 		err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
860 	else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_info->key.tun_flags))
861 		err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
862 	else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
863 		err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
864 
865 	if (err) {
866 		nla_nest_cancel(skb, nest);
867 		return err;
868 	}
869 
870 	nla_nest_end(skb, nest);
871 	return 0;
872 }
873 
ip_tun_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)874 static int ip_tun_fill_encap_info(struct sk_buff *skb,
875 				  struct lwtunnel_state *lwtstate)
876 {
877 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
878 
879 	if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
880 			 LWTUNNEL_IP_PAD) ||
881 	    nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
882 	    nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
883 	    nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
884 	    nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
885 	    nla_put_be16(skb, LWTUNNEL_IP_FLAGS,
886 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
887 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
888 		return -ENOMEM;
889 
890 	return 0;
891 }
892 
ip_tun_opts_nlsize(struct ip_tunnel_info * info)893 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
894 {
895 	int opt_len;
896 
897 	if (!ip_tunnel_is_options_present(info->key.tun_flags))
898 		return 0;
899 
900 	opt_len = nla_total_size(0);		/* LWTUNNEL_IP_OPTS */
901 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags)) {
902 		struct geneve_opt *opt;
903 		int offset = 0;
904 
905 		opt_len += nla_total_size(0);	/* LWTUNNEL_IP_OPTS_GENEVE */
906 		while (info->options_len > offset) {
907 			opt = ip_tunnel_info_opts(info) + offset;
908 			opt_len += nla_total_size(2)	/* OPT_GENEVE_CLASS */
909 				   + nla_total_size(1)	/* OPT_GENEVE_TYPE */
910 				   + nla_total_size(opt->length * 4);
911 							/* OPT_GENEVE_DATA */
912 			offset += sizeof(*opt) + opt->length * 4;
913 		}
914 	} else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
915 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_VXLAN */
916 			   + nla_total_size(4);	/* OPT_VXLAN_GBP */
917 	} else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags)) {
918 		struct erspan_metadata *md = ip_tunnel_info_opts(info);
919 
920 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_ERSPAN */
921 			   + nla_total_size(1)	/* OPT_ERSPAN_VER */
922 			   + (md->version == 1 ? nla_total_size(4)
923 						/* OPT_ERSPAN_INDEX (v1) */
924 					       : nla_total_size(1) +
925 						 nla_total_size(1));
926 						/* OPT_ERSPAN_DIR + HWID (v2) */
927 	}
928 
929 	return opt_len;
930 }
931 
ip_tun_encap_nlsize(struct lwtunnel_state * lwtstate)932 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
933 {
934 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP_ID */
935 		+ nla_total_size(4)	/* LWTUNNEL_IP_DST */
936 		+ nla_total_size(4)	/* LWTUNNEL_IP_SRC */
937 		+ nla_total_size(1)	/* LWTUNNEL_IP_TOS */
938 		+ nla_total_size(1)	/* LWTUNNEL_IP_TTL */
939 		+ nla_total_size(2)	/* LWTUNNEL_IP_FLAGS */
940 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
941 					/* LWTUNNEL_IP_OPTS */
942 }
943 
ip_tun_cmp_encap(struct lwtunnel_state * a,struct lwtunnel_state * b)944 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
945 {
946 	struct ip_tunnel_info *info_a = lwt_tun_info(a);
947 	struct ip_tunnel_info *info_b = lwt_tun_info(b);
948 
949 	return memcmp(info_a, info_b, sizeof(info_a->key)) ||
950 	       info_a->mode != info_b->mode ||
951 	       info_a->options_len != info_b->options_len ||
952 	       memcmp(ip_tunnel_info_opts(info_a),
953 		      ip_tunnel_info_opts(info_b), info_a->options_len);
954 }
955 
956 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
957 	.build_state = ip_tun_build_state,
958 	.destroy_state = ip_tun_destroy_state,
959 	.fill_encap = ip_tun_fill_encap_info,
960 	.get_encap_size = ip_tun_encap_nlsize,
961 	.cmp_encap = ip_tun_cmp_encap,
962 	.owner = THIS_MODULE,
963 };
964 
965 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
966 	[LWTUNNEL_IP6_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP6_OPTS },
967 	[LWTUNNEL_IP6_ID]		= { .type = NLA_U64 },
968 	[LWTUNNEL_IP6_DST]		= { .len = sizeof(struct in6_addr) },
969 	[LWTUNNEL_IP6_SRC]		= { .len = sizeof(struct in6_addr) },
970 	[LWTUNNEL_IP6_HOPLIMIT]		= { .type = NLA_U8 },
971 	[LWTUNNEL_IP6_TC]		= { .type = NLA_U8 },
972 	[LWTUNNEL_IP6_FLAGS]		= { .type = NLA_U16 },
973 	[LWTUNNEL_IP6_OPTS]		= { .type = NLA_NESTED },
974 };
975 
ip6_tun_build_state(struct net * net,struct nlattr * attr,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)976 static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
977 			       unsigned int family, const void *cfg,
978 			       struct lwtunnel_state **ts,
979 			       struct netlink_ext_ack *extack)
980 {
981 	struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
982 	struct lwtunnel_state *new_state;
983 	struct ip_tunnel_info *tun_info;
984 	int err, opt_len;
985 
986 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
987 					  ip6_tun_policy, extack);
988 	if (err < 0)
989 		return err;
990 
991 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
992 	if (opt_len < 0)
993 		return opt_len;
994 
995 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
996 	if (!new_state)
997 		return -ENOMEM;
998 
999 	new_state->type = LWTUNNEL_ENCAP_IP6;
1000 
1001 	tun_info = lwt_tun_info(new_state);
1002 
1003 	err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
1004 	if (err < 0) {
1005 		lwtstate_free(new_state);
1006 		return err;
1007 	}
1008 
1009 	if (tb[LWTUNNEL_IP6_ID])
1010 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
1011 
1012 	if (tb[LWTUNNEL_IP6_DST])
1013 		tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
1014 
1015 	if (tb[LWTUNNEL_IP6_SRC])
1016 		tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
1017 
1018 	if (tb[LWTUNNEL_IP6_HOPLIMIT])
1019 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
1020 
1021 	if (tb[LWTUNNEL_IP6_TC])
1022 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
1023 
1024 	if (tb[LWTUNNEL_IP6_FLAGS]) {
1025 		IP_TUNNEL_DECLARE_FLAGS(flags);
1026 		__be16 data;
1027 
1028 		data = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
1029 		ip_tunnel_flags_from_be16(flags, data);
1030 		ip_tunnel_clear_options_present(flags);
1031 
1032 		ip_tunnel_flags_or(tun_info->key.tun_flags,
1033 				   tun_info->key.tun_flags, flags);
1034 	}
1035 
1036 	tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
1037 	tun_info->options_len = opt_len;
1038 
1039 	*ts = new_state;
1040 
1041 	return 0;
1042 }
1043 
ip6_tun_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)1044 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1045 				   struct lwtunnel_state *lwtstate)
1046 {
1047 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1048 
1049 	if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1050 			 LWTUNNEL_IP6_PAD) ||
1051 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1052 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1053 	    nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1054 	    nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1055 	    nla_put_be16(skb, LWTUNNEL_IP6_FLAGS,
1056 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
1057 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1058 		return -ENOMEM;
1059 
1060 	return 0;
1061 }
1062 
ip6_tun_encap_nlsize(struct lwtunnel_state * lwtstate)1063 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1064 {
1065 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP6_ID */
1066 		+ nla_total_size(16)	/* LWTUNNEL_IP6_DST */
1067 		+ nla_total_size(16)	/* LWTUNNEL_IP6_SRC */
1068 		+ nla_total_size(1)	/* LWTUNNEL_IP6_HOPLIMIT */
1069 		+ nla_total_size(1)	/* LWTUNNEL_IP6_TC */
1070 		+ nla_total_size(2)	/* LWTUNNEL_IP6_FLAGS */
1071 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1072 					/* LWTUNNEL_IP6_OPTS */
1073 }
1074 
1075 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1076 	.build_state = ip6_tun_build_state,
1077 	.fill_encap = ip6_tun_fill_encap_info,
1078 	.get_encap_size = ip6_tun_encap_nlsize,
1079 	.cmp_encap = ip_tun_cmp_encap,
1080 	.owner = THIS_MODULE,
1081 };
1082 
ip_tunnel_core_init(void)1083 void __init ip_tunnel_core_init(void)
1084 {
1085 	/* If you land here, make sure whether increasing ip_tunnel_info's
1086 	 * options_len is a reasonable choice with its usage in front ends
1087 	 * (f.e., it's part of flow keys, etc).
1088 	 */
1089 	BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1090 
1091 	lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1092 	lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1093 }
1094 
1095 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1096 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1097 
ip_tunnel_need_metadata(void)1098 void ip_tunnel_need_metadata(void)
1099 {
1100 	static_branch_inc(&ip_tunnel_metadata_cnt);
1101 }
1102 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1103 
ip_tunnel_unneed_metadata(void)1104 void ip_tunnel_unneed_metadata(void)
1105 {
1106 	static_branch_dec(&ip_tunnel_metadata_cnt);
1107 }
1108 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1109 
1110 /* Returns either the correct skb->protocol value, or 0 if invalid. */
ip_tunnel_parse_protocol(const struct sk_buff * skb)1111 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1112 {
1113 	if (skb_network_header(skb) >= skb->head &&
1114 	    (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1115 	    ip_hdr(skb)->version == 4)
1116 		return htons(ETH_P_IP);
1117 	if (skb_network_header(skb) >= skb->head &&
1118 	    (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1119 	    ipv6_hdr(skb)->version == 6)
1120 		return htons(ETH_P_IPV6);
1121 	return 0;
1122 }
1123 EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1124 
1125 const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1126 EXPORT_SYMBOL(ip_tunnel_header_ops);
1127 
1128 /* This function returns true when ENCAP attributes are present in the nl msg */
ip_tunnel_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * encap)1129 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
1130 				   struct ip_tunnel_encap *encap)
1131 {
1132 	bool ret = false;
1133 
1134 	memset(encap, 0, sizeof(*encap));
1135 
1136 	if (!data)
1137 		return ret;
1138 
1139 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1140 		ret = true;
1141 		encap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1142 	}
1143 
1144 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1145 		ret = true;
1146 		encap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1147 	}
1148 
1149 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1150 		ret = true;
1151 		encap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1152 	}
1153 
1154 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1155 		ret = true;
1156 		encap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1157 	}
1158 
1159 	return ret;
1160 }
1161 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_encap_parms);
1162 
ip_tunnel_netlink_parms(struct nlattr * data[],struct ip_tunnel_parm_kern * parms)1163 void ip_tunnel_netlink_parms(struct nlattr *data[],
1164 			     struct ip_tunnel_parm_kern *parms)
1165 {
1166 	if (data[IFLA_IPTUN_LINK])
1167 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1168 
1169 	if (data[IFLA_IPTUN_LOCAL])
1170 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1171 
1172 	if (data[IFLA_IPTUN_REMOTE])
1173 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1174 
1175 	if (data[IFLA_IPTUN_TTL]) {
1176 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1177 		if (parms->iph.ttl)
1178 			parms->iph.frag_off = htons(IP_DF);
1179 	}
1180 
1181 	if (data[IFLA_IPTUN_TOS])
1182 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1183 
1184 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1185 		parms->iph.frag_off = htons(IP_DF);
1186 
1187 	if (data[IFLA_IPTUN_FLAGS]) {
1188 		__be16 flags;
1189 
1190 		flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1191 		ip_tunnel_flags_from_be16(parms->i_flags, flags);
1192 	}
1193 
1194 	if (data[IFLA_IPTUN_PROTO])
1195 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1196 }
1197 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_parms);
1198