xref: /linux/net/ipv4/ip_tunnel_core.c (revision db87bd2ad1f736c2f7ab231f9b40c885934f6b2c)
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 	skb_scrub_packet(skb, xnet);
62 
63 	skb_clear_hash_if_not_l4(skb);
64 	skb_dst_set(skb, &rt->dst);
65 	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
66 	IPCB(skb)->flags = ipcb_flags;
67 
68 	/* Push down and install the IP header. */
69 	skb_push(skb, sizeof(struct iphdr));
70 	skb_reset_network_header(skb);
71 
72 	iph = ip_hdr(skb);
73 
74 	iph->version	=	4;
75 	iph->ihl	=	sizeof(struct iphdr) >> 2;
76 	iph->frag_off	=	ip_mtu_locked(&rt->dst) ? 0 : df;
77 	iph->protocol	=	proto;
78 	iph->tos	=	tos;
79 	iph->daddr	=	dst;
80 	iph->saddr	=	src;
81 	iph->ttl	=	ttl;
82 	__ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
83 
84 	err = ip_local_out(net, sk, skb);
85 
86 	if (dev) {
87 		if (unlikely(net_xmit_eval(err)))
88 			pkt_len = 0;
89 		iptunnel_xmit_stats(dev, pkt_len);
90 	}
91 }
92 EXPORT_SYMBOL_GPL(iptunnel_xmit);
93 
__iptunnel_pull_header(struct sk_buff * skb,int hdr_len,__be16 inner_proto,bool raw_proto,bool xnet)94 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
95 			   __be16 inner_proto, bool raw_proto, bool xnet)
96 {
97 	if (unlikely(!pskb_may_pull(skb, hdr_len)))
98 		return -ENOMEM;
99 
100 	skb_pull_rcsum(skb, hdr_len);
101 
102 	if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
103 		struct ethhdr *eh;
104 
105 		if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
106 			return -ENOMEM;
107 
108 		eh = (struct ethhdr *)skb->data;
109 		if (likely(eth_proto_is_802_3(eh->h_proto)))
110 			skb->protocol = eh->h_proto;
111 		else
112 			skb->protocol = htons(ETH_P_802_2);
113 
114 	} else {
115 		skb->protocol = inner_proto;
116 	}
117 
118 	skb_clear_hash_if_not_l4(skb);
119 	__vlan_hwaccel_clear_tag(skb);
120 	skb_set_queue_mapping(skb, 0);
121 	skb_scrub_packet(skb, xnet);
122 
123 	return iptunnel_pull_offloads(skb);
124 }
125 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
126 
iptunnel_metadata_reply(struct metadata_dst * md,gfp_t flags)127 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
128 					     gfp_t flags)
129 {
130 	IP_TUNNEL_DECLARE_FLAGS(tun_flags) = { };
131 	struct metadata_dst *res;
132 	struct ip_tunnel_info *dst, *src;
133 
134 	if (!md || md->type != METADATA_IP_TUNNEL ||
135 	    md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
136 		return NULL;
137 
138 	src = &md->u.tun_info;
139 	res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
140 	if (!res)
141 		return NULL;
142 
143 	dst = &res->u.tun_info;
144 	dst->key.tun_id = src->key.tun_id;
145 	if (src->mode & IP_TUNNEL_INFO_IPV6)
146 		memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
147 		       sizeof(struct in6_addr));
148 	else
149 		dst->key.u.ipv4.dst = src->key.u.ipv4.src;
150 	ip_tunnel_flags_copy(dst->key.tun_flags, src->key.tun_flags);
151 	dst->mode = src->mode | IP_TUNNEL_INFO_TX;
152 	ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
153 				src->options_len, tun_flags);
154 
155 	return res;
156 }
157 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
158 
iptunnel_handle_offloads(struct sk_buff * skb,int gso_type_mask)159 int iptunnel_handle_offloads(struct sk_buff *skb,
160 			     int gso_type_mask)
161 {
162 	int err;
163 
164 	if (likely(!skb->encapsulation)) {
165 		skb_reset_inner_headers(skb);
166 		skb->encapsulation = 1;
167 	}
168 
169 	if (skb_is_gso(skb)) {
170 		err = skb_header_unclone(skb, GFP_ATOMIC);
171 		if (unlikely(err))
172 			return err;
173 		skb_shinfo(skb)->gso_type |= gso_type_mask;
174 		return 0;
175 	}
176 
177 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
178 		skb->ip_summed = CHECKSUM_NONE;
179 		/* We clear encapsulation here to prevent badly-written
180 		 * drivers potentially deciding to offload an inner checksum
181 		 * if we set CHECKSUM_PARTIAL on the outer header.
182 		 * This should go away when the drivers are all fixed.
183 		 */
184 		skb->encapsulation = 0;
185 	}
186 
187 	return 0;
188 }
189 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
190 
191 /**
192  * iptunnel_pmtud_build_icmp() - Build ICMP error message for PMTUD
193  * @skb:	Original packet with L2 header
194  * @mtu:	MTU value for ICMP error
195  *
196  * Return: length on success, negative error code if message couldn't be built.
197  */
iptunnel_pmtud_build_icmp(struct sk_buff * skb,int mtu)198 static int iptunnel_pmtud_build_icmp(struct sk_buff *skb, int mtu)
199 {
200 	const struct iphdr *iph = ip_hdr(skb);
201 	struct icmphdr *icmph;
202 	struct iphdr *niph;
203 	struct ethhdr eh;
204 	int len, err;
205 
206 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct iphdr)))
207 		return -EINVAL;
208 
209 	if (skb_is_gso(skb))
210 		skb_gso_reset(skb);
211 
212 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
213 	pskb_pull(skb, ETH_HLEN);
214 	skb_reset_network_header(skb);
215 
216 	err = pskb_trim(skb, 576 - sizeof(*niph) - sizeof(*icmph));
217 	if (err)
218 		return err;
219 
220 	len = skb->len + sizeof(*icmph);
221 	err = skb_cow(skb, sizeof(*niph) + sizeof(*icmph) + ETH_HLEN);
222 	if (err)
223 		return err;
224 
225 	icmph = skb_push(skb, sizeof(*icmph));
226 	*icmph = (struct icmphdr) {
227 		.type			= ICMP_DEST_UNREACH,
228 		.code			= ICMP_FRAG_NEEDED,
229 		.checksum		= 0,
230 		.un.frag.__unused	= 0,
231 		.un.frag.mtu		= htons(mtu),
232 	};
233 	icmph->checksum = csum_fold(skb_checksum(skb, 0, len, 0));
234 	skb_reset_transport_header(skb);
235 
236 	niph = skb_push(skb, sizeof(*niph));
237 	*niph = (struct iphdr) {
238 		.ihl			= sizeof(*niph) / 4u,
239 		.version 		= 4,
240 		.tos 			= 0,
241 		.tot_len		= htons(len + sizeof(*niph)),
242 		.id			= 0,
243 		.frag_off		= htons(IP_DF),
244 		.ttl			= iph->ttl,
245 		.protocol		= IPPROTO_ICMP,
246 		.saddr			= iph->daddr,
247 		.daddr			= iph->saddr,
248 	};
249 	ip_send_check(niph);
250 	skb_reset_network_header(skb);
251 
252 	skb->ip_summed = CHECKSUM_NONE;
253 
254 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
255 	skb_reset_mac_header(skb);
256 
257 	return skb->len;
258 }
259 
260 /**
261  * iptunnel_pmtud_check_icmp() - Trigger ICMP reply if needed and allowed
262  * @skb:	Buffer being sent by encapsulation, L2 headers expected
263  * @mtu:	Network MTU for path
264  *
265  * Return: 0 for no ICMP reply, length if built, negative value on error.
266  */
iptunnel_pmtud_check_icmp(struct sk_buff * skb,int mtu)267 static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
268 {
269 	const struct icmphdr *icmph = icmp_hdr(skb);
270 	const struct iphdr *iph = ip_hdr(skb);
271 
272 	if (mtu < 576 || iph->frag_off != htons(IP_DF))
273 		return 0;
274 
275 	if (ipv4_is_lbcast(iph->daddr)  || ipv4_is_multicast(iph->daddr) ||
276 	    ipv4_is_zeronet(iph->saddr) || ipv4_is_loopback(iph->saddr)  ||
277 	    ipv4_is_lbcast(iph->saddr)  || ipv4_is_multicast(iph->saddr))
278 		return 0;
279 
280 	if (iph->protocol == IPPROTO_ICMP && icmp_is_err(icmph->type))
281 		return 0;
282 
283 	return iptunnel_pmtud_build_icmp(skb, mtu);
284 }
285 
286 #if IS_ENABLED(CONFIG_IPV6)
287 /**
288  * iptunnel_pmtud_build_icmpv6() - Build ICMPv6 error message for PMTUD
289  * @skb:	Original packet with L2 header
290  * @mtu:	MTU value for ICMPv6 error
291  *
292  * Return: length on success, negative error code if message couldn't be built.
293  */
iptunnel_pmtud_build_icmpv6(struct sk_buff * skb,int mtu)294 static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
295 {
296 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
297 	struct icmp6hdr *icmp6h;
298 	struct ipv6hdr *nip6h;
299 	struct ethhdr eh;
300 	int len, err;
301 	__wsum csum;
302 
303 	if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
304 		return -EINVAL;
305 
306 	if (skb_is_gso(skb))
307 		skb_gso_reset(skb);
308 
309 	skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
310 	pskb_pull(skb, ETH_HLEN);
311 	skb_reset_network_header(skb);
312 
313 	err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
314 	if (err)
315 		return err;
316 
317 	len = skb->len + sizeof(*icmp6h);
318 	err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
319 	if (err)
320 		return err;
321 
322 	icmp6h = skb_push(skb, sizeof(*icmp6h));
323 	*icmp6h = (struct icmp6hdr) {
324 		.icmp6_type		= ICMPV6_PKT_TOOBIG,
325 		.icmp6_code		= 0,
326 		.icmp6_cksum		= 0,
327 		.icmp6_mtu		= htonl(mtu),
328 	};
329 	skb_reset_transport_header(skb);
330 
331 	nip6h = skb_push(skb, sizeof(*nip6h));
332 	*nip6h = (struct ipv6hdr) {
333 		.priority		= 0,
334 		.version		= 6,
335 		.flow_lbl		= { 0 },
336 		.payload_len		= htons(len),
337 		.nexthdr		= IPPROTO_ICMPV6,
338 		.hop_limit		= ip6h->hop_limit,
339 		.saddr			= ip6h->daddr,
340 		.daddr			= ip6h->saddr,
341 	};
342 	skb_reset_network_header(skb);
343 
344 	csum = skb_checksum(skb, skb_transport_offset(skb), len, 0);
345 	icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
346 					      IPPROTO_ICMPV6, csum);
347 
348 	skb->ip_summed = CHECKSUM_NONE;
349 
350 	eth_header(skb, skb->dev, ntohs(eh.h_proto), eh.h_source, eh.h_dest, 0);
351 	skb_reset_mac_header(skb);
352 
353 	return skb->len;
354 }
355 
356 /**
357  * iptunnel_pmtud_check_icmpv6() - Trigger ICMPv6 reply if needed and allowed
358  * @skb:	Buffer being sent by encapsulation, L2 headers expected
359  * @mtu:	Network MTU for path
360  *
361  * Return: 0 for no ICMPv6 reply, length if built, negative value on error.
362  */
iptunnel_pmtud_check_icmpv6(struct sk_buff * skb,int mtu)363 static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
364 {
365 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
366 	int stype = ipv6_addr_type(&ip6h->saddr);
367 	u8 proto = ip6h->nexthdr;
368 	__be16 frag_off;
369 	int offset;
370 
371 	if (mtu < IPV6_MIN_MTU)
372 		return 0;
373 
374 	if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
375 	    stype == IPV6_ADDR_LOOPBACK)
376 		return 0;
377 
378 	offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
379 				  &frag_off);
380 	if (offset < 0 || (frag_off & htons(~0x7)))
381 		return 0;
382 
383 	if (proto == IPPROTO_ICMPV6) {
384 		struct icmp6hdr *icmp6h;
385 
386 		if (!pskb_may_pull(skb, skb_network_header(skb) +
387 					offset + 1 - skb->data))
388 			return 0;
389 
390 		icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
391 		if (icmpv6_is_err(icmp6h->icmp6_type) ||
392 		    icmp6h->icmp6_type == NDISC_REDIRECT)
393 			return 0;
394 	}
395 
396 	return iptunnel_pmtud_build_icmpv6(skb, mtu);
397 }
398 #endif /* IS_ENABLED(CONFIG_IPV6) */
399 
400 /**
401  * skb_tunnel_check_pmtu() - Check, update PMTU and trigger ICMP reply as needed
402  * @skb:	Buffer being sent by encapsulation, L2 headers expected
403  * @encap_dst:	Destination for tunnel encapsulation (outer IP)
404  * @headroom:	Encapsulation header size, bytes
405  * @reply:	Build matching ICMP or ICMPv6 message as a result
406  *
407  * L2 tunnel implementations that can carry IP and can be directly bridged
408  * (currently UDP tunnels) can't always rely on IP forwarding paths to handle
409  * PMTU discovery. In the bridged case, ICMP or ICMPv6 messages need to be built
410  * based on payload and sent back by the encapsulation itself.
411  *
412  * For routable interfaces, we just need to update the PMTU for the destination.
413  *
414  * Return: 0 if ICMP error not needed, length if built, negative value on error
415  */
skb_tunnel_check_pmtu(struct sk_buff * skb,struct dst_entry * encap_dst,int headroom,bool reply)416 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
417 			  int headroom, bool reply)
418 {
419 	u32 mtu = dst_mtu(encap_dst) - headroom;
420 
421 	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
422 	    (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
423 		return 0;
424 
425 	skb_dst_update_pmtu_no_confirm(skb, mtu);
426 
427 	if (!reply)
428 		return 0;
429 
430 	if (skb->protocol == htons(ETH_P_IP))
431 		return iptunnel_pmtud_check_icmp(skb, mtu);
432 
433 #if IS_ENABLED(CONFIG_IPV6)
434 	if (skb->protocol == htons(ETH_P_IPV6))
435 		return iptunnel_pmtud_check_icmpv6(skb, mtu);
436 #endif
437 	return 0;
438 }
439 EXPORT_SYMBOL(skb_tunnel_check_pmtu);
440 
441 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
442 	[LWTUNNEL_IP_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP_OPTS },
443 	[LWTUNNEL_IP_ID]	= { .type = NLA_U64 },
444 	[LWTUNNEL_IP_DST]	= { .type = NLA_U32 },
445 	[LWTUNNEL_IP_SRC]	= { .type = NLA_U32 },
446 	[LWTUNNEL_IP_TTL]	= { .type = NLA_U8 },
447 	[LWTUNNEL_IP_TOS]	= { .type = NLA_U8 },
448 	[LWTUNNEL_IP_FLAGS]	= { .type = NLA_U16 },
449 	[LWTUNNEL_IP_OPTS]	= { .type = NLA_NESTED },
450 };
451 
452 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
453 	[LWTUNNEL_IP_OPTS_GENEVE]	= { .type = NLA_NESTED },
454 	[LWTUNNEL_IP_OPTS_VXLAN]	= { .type = NLA_NESTED },
455 	[LWTUNNEL_IP_OPTS_ERSPAN]	= { .type = NLA_NESTED },
456 };
457 
458 static const struct nla_policy
459 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
460 	[LWTUNNEL_IP_OPT_GENEVE_CLASS]	= { .type = NLA_U16 },
461 	[LWTUNNEL_IP_OPT_GENEVE_TYPE]	= { .type = NLA_U8 },
462 	[LWTUNNEL_IP_OPT_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 127 },
463 };
464 
465 static const struct nla_policy
466 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
467 	[LWTUNNEL_IP_OPT_VXLAN_GBP]	= { .type = NLA_U32 },
468 };
469 
470 static const struct nla_policy
471 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
472 	[LWTUNNEL_IP_OPT_ERSPAN_VER]	= { .type = NLA_U8 },
473 	[LWTUNNEL_IP_OPT_ERSPAN_INDEX]	= { .type = NLA_U32 },
474 	[LWTUNNEL_IP_OPT_ERSPAN_DIR]	= { .type = NLA_U8 },
475 	[LWTUNNEL_IP_OPT_ERSPAN_HWID]	= { .type = NLA_U8 },
476 };
477 
ip_tun_parse_opts_geneve(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)478 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
479 				    struct ip_tunnel_info *info, int opts_len,
480 				    struct netlink_ext_ack *extack)
481 {
482 	struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
483 	int data_len, err;
484 
485 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
486 			       geneve_opt_policy, extack);
487 	if (err)
488 		return err;
489 
490 	if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
491 	    !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
492 	    !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
493 		return -EINVAL;
494 
495 	attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
496 	data_len = nla_len(attr);
497 	if (data_len % 4)
498 		return -EINVAL;
499 
500 	if (info) {
501 		struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
502 
503 		memcpy(opt->opt_data, nla_data(attr), data_len);
504 		opt->length = data_len / 4;
505 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
506 		opt->opt_class = nla_get_be16(attr);
507 		attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
508 		opt->type = nla_get_u8(attr);
509 		__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags);
510 	}
511 
512 	return sizeof(struct geneve_opt) + data_len;
513 }
514 
ip_tun_parse_opts_vxlan(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)515 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
516 				   struct ip_tunnel_info *info, int opts_len,
517 				   struct netlink_ext_ack *extack)
518 {
519 	struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
520 	int err;
521 
522 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
523 			       vxlan_opt_policy, extack);
524 	if (err)
525 		return err;
526 
527 	if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
528 		return -EINVAL;
529 
530 	if (info) {
531 		struct vxlan_metadata *md =
532 			ip_tunnel_info_opts(info) + opts_len;
533 
534 		attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
535 		md->gbp = nla_get_u32(attr);
536 		md->gbp &= VXLAN_GBP_MASK;
537 		__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags);
538 	}
539 
540 	return sizeof(struct vxlan_metadata);
541 }
542 
ip_tun_parse_opts_erspan(struct nlattr * attr,struct ip_tunnel_info * info,int opts_len,struct netlink_ext_ack * extack)543 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
544 				    struct ip_tunnel_info *info, int opts_len,
545 				    struct netlink_ext_ack *extack)
546 {
547 	struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
548 	int err;
549 	u8 ver;
550 
551 	err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
552 			       erspan_opt_policy, extack);
553 	if (err)
554 		return err;
555 
556 	if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
557 		return -EINVAL;
558 
559 	ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
560 	if (ver == 1) {
561 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
562 			return -EINVAL;
563 	} else if (ver == 2) {
564 		if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
565 		    !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
566 			return -EINVAL;
567 	} else {
568 		return -EINVAL;
569 	}
570 
571 	if (info) {
572 		struct erspan_metadata *md =
573 			ip_tunnel_info_opts(info) + opts_len;
574 
575 		md->version = ver;
576 		if (ver == 1) {
577 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
578 			md->u.index = nla_get_be32(attr);
579 		} else {
580 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
581 			md->u.md2.dir = nla_get_u8(attr);
582 			attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
583 			set_hwid(&md->u.md2, nla_get_u8(attr));
584 		}
585 
586 		__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags);
587 	}
588 
589 	return sizeof(struct erspan_metadata);
590 }
591 
ip_tun_parse_opts(struct nlattr * attr,struct ip_tunnel_info * info,struct netlink_ext_ack * extack)592 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
593 			     struct netlink_ext_ack *extack)
594 {
595 	int err, rem, opt_len, opts_len = 0;
596 	struct nlattr *nla;
597 	u32 type = 0;
598 
599 	if (!attr)
600 		return 0;
601 
602 	err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
603 			   ip_opts_policy, extack);
604 	if (err)
605 		return err;
606 
607 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
608 		switch (nla_type(nla)) {
609 		case LWTUNNEL_IP_OPTS_GENEVE:
610 			if (type && type != IP_TUNNEL_GENEVE_OPT_BIT)
611 				return -EINVAL;
612 			opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
613 							   extack);
614 			if (opt_len < 0)
615 				return opt_len;
616 			opts_len += opt_len;
617 			if (opts_len > IP_TUNNEL_OPTS_MAX)
618 				return -EINVAL;
619 			type = IP_TUNNEL_GENEVE_OPT_BIT;
620 			break;
621 		case LWTUNNEL_IP_OPTS_VXLAN:
622 			if (type)
623 				return -EINVAL;
624 			opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
625 							  extack);
626 			if (opt_len < 0)
627 				return opt_len;
628 			opts_len += opt_len;
629 			type = IP_TUNNEL_VXLAN_OPT_BIT;
630 			break;
631 		case LWTUNNEL_IP_OPTS_ERSPAN:
632 			if (type)
633 				return -EINVAL;
634 			opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
635 							   extack);
636 			if (opt_len < 0)
637 				return opt_len;
638 			opts_len += opt_len;
639 			type = IP_TUNNEL_ERSPAN_OPT_BIT;
640 			break;
641 		default:
642 			return -EINVAL;
643 		}
644 	}
645 
646 	return opts_len;
647 }
648 
ip_tun_get_optlen(struct nlattr * attr,struct netlink_ext_ack * extack)649 static int ip_tun_get_optlen(struct nlattr *attr,
650 			     struct netlink_ext_ack *extack)
651 {
652 	return ip_tun_parse_opts(attr, NULL, extack);
653 }
654 
ip_tun_set_opts(struct nlattr * attr,struct ip_tunnel_info * info,struct netlink_ext_ack * extack)655 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
656 			   struct netlink_ext_ack *extack)
657 {
658 	return ip_tun_parse_opts(attr, info, extack);
659 }
660 
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)661 static int ip_tun_build_state(struct net *net, struct nlattr *attr,
662 			      unsigned int family, const void *cfg,
663 			      struct lwtunnel_state **ts,
664 			      struct netlink_ext_ack *extack)
665 {
666 	struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
667 	struct lwtunnel_state *new_state;
668 	struct ip_tunnel_info *tun_info;
669 	int err, opt_len;
670 
671 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
672 					  ip_tun_policy, extack);
673 	if (err < 0)
674 		return err;
675 
676 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
677 	if (opt_len < 0)
678 		return opt_len;
679 
680 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
681 	if (!new_state)
682 		return -ENOMEM;
683 
684 	new_state->type = LWTUNNEL_ENCAP_IP;
685 
686 	tun_info = lwt_tun_info(new_state);
687 
688 	err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
689 	if (err < 0) {
690 		lwtstate_free(new_state);
691 		return err;
692 	}
693 
694 #ifdef CONFIG_DST_CACHE
695 	err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
696 	if (err) {
697 		lwtstate_free(new_state);
698 		return err;
699 	}
700 #endif
701 
702 	if (tb[LWTUNNEL_IP_ID])
703 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
704 
705 	if (tb[LWTUNNEL_IP_DST])
706 		tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
707 
708 	if (tb[LWTUNNEL_IP_SRC])
709 		tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
710 
711 	if (tb[LWTUNNEL_IP_TTL])
712 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
713 
714 	if (tb[LWTUNNEL_IP_TOS])
715 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
716 
717 	if (tb[LWTUNNEL_IP_FLAGS]) {
718 		IP_TUNNEL_DECLARE_FLAGS(flags);
719 
720 		ip_tunnel_flags_from_be16(flags,
721 					  nla_get_be16(tb[LWTUNNEL_IP_FLAGS]));
722 		ip_tunnel_clear_options_present(flags);
723 
724 		ip_tunnel_flags_or(tun_info->key.tun_flags,
725 				   tun_info->key.tun_flags, flags);
726 	}
727 
728 	tun_info->mode = IP_TUNNEL_INFO_TX;
729 	tun_info->options_len = opt_len;
730 
731 	*ts = new_state;
732 
733 	return 0;
734 }
735 
ip_tun_destroy_state(struct lwtunnel_state * lwtstate)736 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
737 {
738 #ifdef CONFIG_DST_CACHE
739 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
740 
741 	dst_cache_destroy(&tun_info->dst_cache);
742 #endif
743 }
744 
ip_tun_fill_encap_opts_geneve(struct sk_buff * skb,struct ip_tunnel_info * tun_info)745 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
746 					 struct ip_tunnel_info *tun_info)
747 {
748 	struct geneve_opt *opt;
749 	struct nlattr *nest;
750 	int offset = 0;
751 
752 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
753 	if (!nest)
754 		return -ENOMEM;
755 
756 	while (tun_info->options_len > offset) {
757 		opt = ip_tunnel_info_opts(tun_info) + offset;
758 		if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
759 				 opt->opt_class) ||
760 		    nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
761 		    nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
762 			    opt->opt_data)) {
763 			nla_nest_cancel(skb, nest);
764 			return -ENOMEM;
765 		}
766 		offset += sizeof(*opt) + opt->length * 4;
767 	}
768 
769 	nla_nest_end(skb, nest);
770 	return 0;
771 }
772 
ip_tun_fill_encap_opts_vxlan(struct sk_buff * skb,struct ip_tunnel_info * tun_info)773 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
774 					struct ip_tunnel_info *tun_info)
775 {
776 	struct vxlan_metadata *md;
777 	struct nlattr *nest;
778 
779 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
780 	if (!nest)
781 		return -ENOMEM;
782 
783 	md = ip_tunnel_info_opts(tun_info);
784 	if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
785 		nla_nest_cancel(skb, nest);
786 		return -ENOMEM;
787 	}
788 
789 	nla_nest_end(skb, nest);
790 	return 0;
791 }
792 
ip_tun_fill_encap_opts_erspan(struct sk_buff * skb,struct ip_tunnel_info * tun_info)793 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
794 					 struct ip_tunnel_info *tun_info)
795 {
796 	struct erspan_metadata *md;
797 	struct nlattr *nest;
798 
799 	nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
800 	if (!nest)
801 		return -ENOMEM;
802 
803 	md = ip_tunnel_info_opts(tun_info);
804 	if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
805 		goto err;
806 
807 	if (md->version == 1 &&
808 	    nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
809 		goto err;
810 
811 	if (md->version == 2 &&
812 	    (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
813 	     nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
814 			get_hwid(&md->u.md2))))
815 		goto err;
816 
817 	nla_nest_end(skb, nest);
818 	return 0;
819 err:
820 	nla_nest_cancel(skb, nest);
821 	return -ENOMEM;
822 }
823 
ip_tun_fill_encap_opts(struct sk_buff * skb,int type,struct ip_tunnel_info * tun_info)824 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
825 				  struct ip_tunnel_info *tun_info)
826 {
827 	struct nlattr *nest;
828 	int err = 0;
829 
830 	if (!ip_tunnel_is_options_present(tun_info->key.tun_flags))
831 		return 0;
832 
833 	nest = nla_nest_start_noflag(skb, type);
834 	if (!nest)
835 		return -ENOMEM;
836 
837 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_info->key.tun_flags))
838 		err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
839 	else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_info->key.tun_flags))
840 		err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
841 	else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
842 		err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
843 
844 	if (err) {
845 		nla_nest_cancel(skb, nest);
846 		return err;
847 	}
848 
849 	nla_nest_end(skb, nest);
850 	return 0;
851 }
852 
ip_tun_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)853 static int ip_tun_fill_encap_info(struct sk_buff *skb,
854 				  struct lwtunnel_state *lwtstate)
855 {
856 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
857 
858 	if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
859 			 LWTUNNEL_IP_PAD) ||
860 	    nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
861 	    nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
862 	    nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
863 	    nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
864 	    nla_put_be16(skb, LWTUNNEL_IP_FLAGS,
865 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
866 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
867 		return -ENOMEM;
868 
869 	return 0;
870 }
871 
ip_tun_opts_nlsize(struct ip_tunnel_info * info)872 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
873 {
874 	int opt_len;
875 
876 	if (!ip_tunnel_is_options_present(info->key.tun_flags))
877 		return 0;
878 
879 	opt_len = nla_total_size(0);		/* LWTUNNEL_IP_OPTS */
880 	if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, info->key.tun_flags)) {
881 		struct geneve_opt *opt;
882 		int offset = 0;
883 
884 		opt_len += nla_total_size(0);	/* LWTUNNEL_IP_OPTS_GENEVE */
885 		while (info->options_len > offset) {
886 			opt = ip_tunnel_info_opts(info) + offset;
887 			opt_len += nla_total_size(2)	/* OPT_GENEVE_CLASS */
888 				   + nla_total_size(1)	/* OPT_GENEVE_TYPE */
889 				   + nla_total_size(opt->length * 4);
890 							/* OPT_GENEVE_DATA */
891 			offset += sizeof(*opt) + opt->length * 4;
892 		}
893 	} else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
894 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_VXLAN */
895 			   + nla_total_size(4);	/* OPT_VXLAN_GBP */
896 	} else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, info->key.tun_flags)) {
897 		struct erspan_metadata *md = ip_tunnel_info_opts(info);
898 
899 		opt_len += nla_total_size(0)	/* LWTUNNEL_IP_OPTS_ERSPAN */
900 			   + nla_total_size(1)	/* OPT_ERSPAN_VER */
901 			   + (md->version == 1 ? nla_total_size(4)
902 						/* OPT_ERSPAN_INDEX (v1) */
903 					       : nla_total_size(1) +
904 						 nla_total_size(1));
905 						/* OPT_ERSPAN_DIR + HWID (v2) */
906 	}
907 
908 	return opt_len;
909 }
910 
ip_tun_encap_nlsize(struct lwtunnel_state * lwtstate)911 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
912 {
913 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP_ID */
914 		+ nla_total_size(4)	/* LWTUNNEL_IP_DST */
915 		+ nla_total_size(4)	/* LWTUNNEL_IP_SRC */
916 		+ nla_total_size(1)	/* LWTUNNEL_IP_TOS */
917 		+ nla_total_size(1)	/* LWTUNNEL_IP_TTL */
918 		+ nla_total_size(2)	/* LWTUNNEL_IP_FLAGS */
919 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
920 					/* LWTUNNEL_IP_OPTS */
921 }
922 
ip_tun_cmp_encap(struct lwtunnel_state * a,struct lwtunnel_state * b)923 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
924 {
925 	struct ip_tunnel_info *info_a = lwt_tun_info(a);
926 	struct ip_tunnel_info *info_b = lwt_tun_info(b);
927 
928 	return memcmp(info_a, info_b, sizeof(info_a->key)) ||
929 	       info_a->mode != info_b->mode ||
930 	       info_a->options_len != info_b->options_len ||
931 	       memcmp(ip_tunnel_info_opts(info_a),
932 		      ip_tunnel_info_opts(info_b), info_a->options_len);
933 }
934 
935 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
936 	.build_state = ip_tun_build_state,
937 	.destroy_state = ip_tun_destroy_state,
938 	.fill_encap = ip_tun_fill_encap_info,
939 	.get_encap_size = ip_tun_encap_nlsize,
940 	.cmp_encap = ip_tun_cmp_encap,
941 	.owner = THIS_MODULE,
942 };
943 
944 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
945 	[LWTUNNEL_IP6_UNSPEC]	= { .strict_start_type = LWTUNNEL_IP6_OPTS },
946 	[LWTUNNEL_IP6_ID]		= { .type = NLA_U64 },
947 	[LWTUNNEL_IP6_DST]		= { .len = sizeof(struct in6_addr) },
948 	[LWTUNNEL_IP6_SRC]		= { .len = sizeof(struct in6_addr) },
949 	[LWTUNNEL_IP6_HOPLIMIT]		= { .type = NLA_U8 },
950 	[LWTUNNEL_IP6_TC]		= { .type = NLA_U8 },
951 	[LWTUNNEL_IP6_FLAGS]		= { .type = NLA_U16 },
952 	[LWTUNNEL_IP6_OPTS]		= { .type = NLA_NESTED },
953 };
954 
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)955 static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
956 			       unsigned int family, const void *cfg,
957 			       struct lwtunnel_state **ts,
958 			       struct netlink_ext_ack *extack)
959 {
960 	struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
961 	struct lwtunnel_state *new_state;
962 	struct ip_tunnel_info *tun_info;
963 	int err, opt_len;
964 
965 	err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
966 					  ip6_tun_policy, extack);
967 	if (err < 0)
968 		return err;
969 
970 	opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
971 	if (opt_len < 0)
972 		return opt_len;
973 
974 	new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
975 	if (!new_state)
976 		return -ENOMEM;
977 
978 	new_state->type = LWTUNNEL_ENCAP_IP6;
979 
980 	tun_info = lwt_tun_info(new_state);
981 
982 	err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
983 	if (err < 0) {
984 		lwtstate_free(new_state);
985 		return err;
986 	}
987 
988 	if (tb[LWTUNNEL_IP6_ID])
989 		tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
990 
991 	if (tb[LWTUNNEL_IP6_DST])
992 		tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
993 
994 	if (tb[LWTUNNEL_IP6_SRC])
995 		tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
996 
997 	if (tb[LWTUNNEL_IP6_HOPLIMIT])
998 		tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
999 
1000 	if (tb[LWTUNNEL_IP6_TC])
1001 		tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
1002 
1003 	if (tb[LWTUNNEL_IP6_FLAGS]) {
1004 		IP_TUNNEL_DECLARE_FLAGS(flags);
1005 		__be16 data;
1006 
1007 		data = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
1008 		ip_tunnel_flags_from_be16(flags, data);
1009 		ip_tunnel_clear_options_present(flags);
1010 
1011 		ip_tunnel_flags_or(tun_info->key.tun_flags,
1012 				   tun_info->key.tun_flags, flags);
1013 	}
1014 
1015 	tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
1016 	tun_info->options_len = opt_len;
1017 
1018 	*ts = new_state;
1019 
1020 	return 0;
1021 }
1022 
ip6_tun_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)1023 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1024 				   struct lwtunnel_state *lwtstate)
1025 {
1026 	struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1027 
1028 	if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1029 			 LWTUNNEL_IP6_PAD) ||
1030 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1031 	    nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1032 	    nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1033 	    nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1034 	    nla_put_be16(skb, LWTUNNEL_IP6_FLAGS,
1035 			 ip_tunnel_flags_to_be16(tun_info->key.tun_flags)) ||
1036 	    ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1037 		return -ENOMEM;
1038 
1039 	return 0;
1040 }
1041 
ip6_tun_encap_nlsize(struct lwtunnel_state * lwtstate)1042 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1043 {
1044 	return nla_total_size_64bit(8)	/* LWTUNNEL_IP6_ID */
1045 		+ nla_total_size(16)	/* LWTUNNEL_IP6_DST */
1046 		+ nla_total_size(16)	/* LWTUNNEL_IP6_SRC */
1047 		+ nla_total_size(1)	/* LWTUNNEL_IP6_HOPLIMIT */
1048 		+ nla_total_size(1)	/* LWTUNNEL_IP6_TC */
1049 		+ nla_total_size(2)	/* LWTUNNEL_IP6_FLAGS */
1050 		+ ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1051 					/* LWTUNNEL_IP6_OPTS */
1052 }
1053 
1054 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1055 	.build_state = ip6_tun_build_state,
1056 	.fill_encap = ip6_tun_fill_encap_info,
1057 	.get_encap_size = ip6_tun_encap_nlsize,
1058 	.cmp_encap = ip_tun_cmp_encap,
1059 	.owner = THIS_MODULE,
1060 };
1061 
ip_tunnel_core_init(void)1062 void __init ip_tunnel_core_init(void)
1063 {
1064 	/* If you land here, make sure whether increasing ip_tunnel_info's
1065 	 * options_len is a reasonable choice with its usage in front ends
1066 	 * (f.e., it's part of flow keys, etc).
1067 	 */
1068 	BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1069 
1070 	lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1071 	lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1072 }
1073 
1074 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1075 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1076 
ip_tunnel_need_metadata(void)1077 void ip_tunnel_need_metadata(void)
1078 {
1079 	static_branch_inc(&ip_tunnel_metadata_cnt);
1080 }
1081 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1082 
ip_tunnel_unneed_metadata(void)1083 void ip_tunnel_unneed_metadata(void)
1084 {
1085 	static_branch_dec(&ip_tunnel_metadata_cnt);
1086 }
1087 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1088 
1089 /* Returns either the correct skb->protocol value, or 0 if invalid. */
ip_tunnel_parse_protocol(const struct sk_buff * skb)1090 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1091 {
1092 	if (skb_network_header(skb) >= skb->head &&
1093 	    (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1094 	    ip_hdr(skb)->version == 4)
1095 		return htons(ETH_P_IP);
1096 	if (skb_network_header(skb) >= skb->head &&
1097 	    (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1098 	    ipv6_hdr(skb)->version == 6)
1099 		return htons(ETH_P_IPV6);
1100 	return 0;
1101 }
1102 EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1103 
1104 const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1105 EXPORT_SYMBOL(ip_tunnel_header_ops);
1106 
1107 /* 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)1108 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
1109 				   struct ip_tunnel_encap *encap)
1110 {
1111 	bool ret = false;
1112 
1113 	memset(encap, 0, sizeof(*encap));
1114 
1115 	if (!data)
1116 		return ret;
1117 
1118 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1119 		ret = true;
1120 		encap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1121 	}
1122 
1123 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1124 		ret = true;
1125 		encap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1126 	}
1127 
1128 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1129 		ret = true;
1130 		encap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1131 	}
1132 
1133 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1134 		ret = true;
1135 		encap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1136 	}
1137 
1138 	return ret;
1139 }
1140 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_encap_parms);
1141 
ip_tunnel_netlink_parms(struct nlattr * data[],struct ip_tunnel_parm_kern * parms)1142 void ip_tunnel_netlink_parms(struct nlattr *data[],
1143 			     struct ip_tunnel_parm_kern *parms)
1144 {
1145 	if (data[IFLA_IPTUN_LINK])
1146 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1147 
1148 	if (data[IFLA_IPTUN_LOCAL])
1149 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1150 
1151 	if (data[IFLA_IPTUN_REMOTE])
1152 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1153 
1154 	if (data[IFLA_IPTUN_TTL]) {
1155 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1156 		if (parms->iph.ttl)
1157 			parms->iph.frag_off = htons(IP_DF);
1158 	}
1159 
1160 	if (data[IFLA_IPTUN_TOS])
1161 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1162 
1163 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1164 		parms->iph.frag_off = htons(IP_DF);
1165 
1166 	if (data[IFLA_IPTUN_FLAGS]) {
1167 		__be16 flags;
1168 
1169 		flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1170 		ip_tunnel_flags_from_be16(parms->i_flags, flags);
1171 	}
1172 
1173 	if (data[IFLA_IPTUN_PROTO])
1174 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1175 }
1176 EXPORT_SYMBOL_GPL(ip_tunnel_netlink_parms);
1177