xref: /openbmc/linux/net/ipv6/seg6_iptunnel.c (revision f8dd092e)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  SR-IPv6 implementation
4  *
5  *  Author:
6  *  David Lebrun <david.lebrun@uclouvain.be>
7  */
8 
9 #include <linux/types.h>
10 #include <linux/skbuff.h>
11 #include <linux/net.h>
12 #include <linux/module.h>
13 #include <net/ip.h>
14 #include <net/ip_tunnels.h>
15 #include <net/lwtunnel.h>
16 #include <net/netevent.h>
17 #include <net/netns/generic.h>
18 #include <net/ip6_fib.h>
19 #include <net/route.h>
20 #include <net/seg6.h>
21 #include <linux/seg6.h>
22 #include <linux/seg6_iptunnel.h>
23 #include <net/addrconf.h>
24 #include <net/ip6_route.h>
25 #include <net/dst_cache.h>
26 #ifdef CONFIG_IPV6_SEG6_HMAC
27 #include <net/seg6_hmac.h>
28 #endif
29 #include <linux/netfilter.h>
30 
seg6_lwt_headroom(struct seg6_iptunnel_encap * tuninfo)31 static size_t seg6_lwt_headroom(struct seg6_iptunnel_encap *tuninfo)
32 {
33 	int head = 0;
34 
35 	switch (tuninfo->mode) {
36 	case SEG6_IPTUN_MODE_INLINE:
37 		break;
38 	case SEG6_IPTUN_MODE_ENCAP:
39 	case SEG6_IPTUN_MODE_ENCAP_RED:
40 		head = sizeof(struct ipv6hdr);
41 		break;
42 	case SEG6_IPTUN_MODE_L2ENCAP:
43 	case SEG6_IPTUN_MODE_L2ENCAP_RED:
44 		return 0;
45 	}
46 
47 	return ((tuninfo->srh->hdrlen + 1) << 3) + head;
48 }
49 
50 struct seg6_lwt {
51 	struct dst_cache cache;
52 	struct seg6_iptunnel_encap tuninfo[];
53 };
54 
seg6_lwt_lwtunnel(struct lwtunnel_state * lwt)55 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
56 {
57 	return (struct seg6_lwt *)lwt->data;
58 }
59 
60 static inline struct seg6_iptunnel_encap *
seg6_encap_lwtunnel(struct lwtunnel_state * lwt)61 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
62 {
63 	return seg6_lwt_lwtunnel(lwt)->tuninfo;
64 }
65 
66 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
67 	[SEG6_IPTUNNEL_SRH]	= { .type = NLA_BINARY },
68 };
69 
nla_put_srh(struct sk_buff * skb,int attrtype,struct seg6_iptunnel_encap * tuninfo)70 static int nla_put_srh(struct sk_buff *skb, int attrtype,
71 		       struct seg6_iptunnel_encap *tuninfo)
72 {
73 	struct seg6_iptunnel_encap *data;
74 	struct nlattr *nla;
75 	int len;
76 
77 	len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
78 
79 	nla = nla_reserve(skb, attrtype, len);
80 	if (!nla)
81 		return -EMSGSIZE;
82 
83 	data = nla_data(nla);
84 	memcpy(data, tuninfo, len);
85 
86 	return 0;
87 }
88 
set_tun_src(struct net * net,struct net_device * dev,struct in6_addr * daddr,struct in6_addr * saddr)89 static void set_tun_src(struct net *net, struct net_device *dev,
90 			struct in6_addr *daddr, struct in6_addr *saddr)
91 {
92 	struct seg6_pernet_data *sdata = seg6_pernet(net);
93 	struct in6_addr *tun_src;
94 
95 	rcu_read_lock();
96 
97 	tun_src = rcu_dereference(sdata->tun_src);
98 
99 	if (!ipv6_addr_any(tun_src)) {
100 		memcpy(saddr, tun_src, sizeof(struct in6_addr));
101 	} else {
102 		ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
103 				   saddr);
104 	}
105 
106 	rcu_read_unlock();
107 }
108 
109 /* Compute flowlabel for outer IPv6 header */
seg6_make_flowlabel(struct net * net,struct sk_buff * skb,struct ipv6hdr * inner_hdr)110 static __be32 seg6_make_flowlabel(struct net *net, struct sk_buff *skb,
111 				  struct ipv6hdr *inner_hdr)
112 {
113 	int do_flowlabel = net->ipv6.sysctl.seg6_flowlabel;
114 	__be32 flowlabel = 0;
115 	u32 hash;
116 
117 	if (do_flowlabel > 0) {
118 		hash = skb_get_hash(skb);
119 		hash = rol32(hash, 16);
120 		flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
121 	} else if (!do_flowlabel && skb->protocol == htons(ETH_P_IPV6)) {
122 		flowlabel = ip6_flowlabel(inner_hdr);
123 	}
124 	return flowlabel;
125 }
126 
127 /* encapsulate an IPv6 packet within an outer IPv6 header with a given SRH */
seg6_do_srh_encap(struct sk_buff * skb,struct ipv6_sr_hdr * osrh,int proto)128 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
129 {
130 	struct dst_entry *dst = skb_dst(skb);
131 	struct net *net = dev_net(dst->dev);
132 	struct ipv6hdr *hdr, *inner_hdr;
133 	struct ipv6_sr_hdr *isrh;
134 	int hdrlen, tot_len, err;
135 	__be32 flowlabel;
136 
137 	hdrlen = (osrh->hdrlen + 1) << 3;
138 	tot_len = hdrlen + sizeof(*hdr);
139 
140 	err = skb_cow_head(skb, tot_len + skb->mac_len);
141 	if (unlikely(err))
142 		return err;
143 
144 	inner_hdr = ipv6_hdr(skb);
145 	flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
146 
147 	skb_push(skb, tot_len);
148 	skb_reset_network_header(skb);
149 	skb_mac_header_rebuild(skb);
150 	hdr = ipv6_hdr(skb);
151 
152 	/* inherit tc, flowlabel and hlim
153 	 * hlim will be decremented in ip6_forward() afterwards and
154 	 * decapsulation will overwrite inner hlim with outer hlim
155 	 */
156 
157 	if (skb->protocol == htons(ETH_P_IPV6)) {
158 		ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
159 			     flowlabel);
160 		hdr->hop_limit = inner_hdr->hop_limit;
161 	} else {
162 		ip6_flow_hdr(hdr, 0, flowlabel);
163 		hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
164 
165 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
166 
167 		/* the control block has been erased, so we have to set the
168 		 * iif once again.
169 		 * We read the receiving interface index directly from the
170 		 * skb->skb_iif as it is done in the IPv4 receiving path (i.e.:
171 		 * ip_rcv_core(...)).
172 		 */
173 		IP6CB(skb)->iif = skb->skb_iif;
174 	}
175 
176 	hdr->nexthdr = NEXTHDR_ROUTING;
177 
178 	isrh = (void *)hdr + sizeof(*hdr);
179 	memcpy(isrh, osrh, hdrlen);
180 
181 	isrh->nexthdr = proto;
182 
183 	hdr->daddr = isrh->segments[isrh->first_segment];
184 	set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
185 
186 #ifdef CONFIG_IPV6_SEG6_HMAC
187 	if (sr_has_hmac(isrh)) {
188 		err = seg6_push_hmac(net, &hdr->saddr, isrh);
189 		if (unlikely(err))
190 			return err;
191 	}
192 #endif
193 
194 	hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
195 
196 	skb_postpush_rcsum(skb, hdr, tot_len);
197 
198 	return 0;
199 }
200 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
201 
202 /* encapsulate an IPv6 packet within an outer IPv6 header with reduced SRH */
seg6_do_srh_encap_red(struct sk_buff * skb,struct ipv6_sr_hdr * osrh,int proto)203 static int seg6_do_srh_encap_red(struct sk_buff *skb,
204 				 struct ipv6_sr_hdr *osrh, int proto)
205 {
206 	__u8 first_seg = osrh->first_segment;
207 	struct dst_entry *dst = skb_dst(skb);
208 	struct net *net = dev_net(dst->dev);
209 	struct ipv6hdr *hdr, *inner_hdr;
210 	int hdrlen = ipv6_optlen(osrh);
211 	int red_tlv_offset, tlv_offset;
212 	struct ipv6_sr_hdr *isrh;
213 	bool skip_srh = false;
214 	__be32 flowlabel;
215 	int tot_len, err;
216 	int red_hdrlen;
217 	int tlvs_len;
218 
219 	if (first_seg > 0) {
220 		red_hdrlen = hdrlen - sizeof(struct in6_addr);
221 	} else {
222 		/* NOTE: if tag/flags and/or other TLVs are introduced in the
223 		 * seg6_iptunnel infrastructure, they should be considered when
224 		 * deciding to skip the SRH.
225 		 */
226 		skip_srh = !sr_has_hmac(osrh);
227 
228 		red_hdrlen = skip_srh ? 0 : hdrlen;
229 	}
230 
231 	tot_len = red_hdrlen + sizeof(struct ipv6hdr);
232 
233 	err = skb_cow_head(skb, tot_len + skb->mac_len);
234 	if (unlikely(err))
235 		return err;
236 
237 	inner_hdr = ipv6_hdr(skb);
238 	flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
239 
240 	skb_push(skb, tot_len);
241 	skb_reset_network_header(skb);
242 	skb_mac_header_rebuild(skb);
243 	hdr = ipv6_hdr(skb);
244 
245 	/* based on seg6_do_srh_encap() */
246 	if (skb->protocol == htons(ETH_P_IPV6)) {
247 		ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
248 			     flowlabel);
249 		hdr->hop_limit = inner_hdr->hop_limit;
250 	} else {
251 		ip6_flow_hdr(hdr, 0, flowlabel);
252 		hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
253 
254 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
255 		IP6CB(skb)->iif = skb->skb_iif;
256 	}
257 
258 	/* no matter if we have to skip the SRH or not, the first segment
259 	 * always comes in the pushed IPv6 header.
260 	 */
261 	hdr->daddr = osrh->segments[first_seg];
262 
263 	if (skip_srh) {
264 		hdr->nexthdr = proto;
265 
266 		set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
267 		goto out;
268 	}
269 
270 	/* we cannot skip the SRH, slow path */
271 
272 	hdr->nexthdr = NEXTHDR_ROUTING;
273 	isrh = (void *)hdr + sizeof(struct ipv6hdr);
274 
275 	if (unlikely(!first_seg)) {
276 		/* this is a very rare case; we have only one SID but
277 		 * we cannot skip the SRH since we are carrying some
278 		 * other info.
279 		 */
280 		memcpy(isrh, osrh, hdrlen);
281 		goto srcaddr;
282 	}
283 
284 	tlv_offset = sizeof(*osrh) + (first_seg + 1) * sizeof(struct in6_addr);
285 	red_tlv_offset = tlv_offset - sizeof(struct in6_addr);
286 
287 	memcpy(isrh, osrh, red_tlv_offset);
288 
289 	tlvs_len = hdrlen - tlv_offset;
290 	if (unlikely(tlvs_len > 0)) {
291 		const void *s = (const void *)osrh + tlv_offset;
292 		void *d = (void *)isrh + red_tlv_offset;
293 
294 		memcpy(d, s, tlvs_len);
295 	}
296 
297 	--isrh->first_segment;
298 	isrh->hdrlen -= 2;
299 
300 srcaddr:
301 	isrh->nexthdr = proto;
302 	set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
303 
304 #ifdef CONFIG_IPV6_SEG6_HMAC
305 	if (unlikely(!skip_srh && sr_has_hmac(isrh))) {
306 		err = seg6_push_hmac(net, &hdr->saddr, isrh);
307 		if (unlikely(err))
308 			return err;
309 	}
310 #endif
311 
312 out:
313 	hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
314 
315 	skb_postpush_rcsum(skb, hdr, tot_len);
316 
317 	return 0;
318 }
319 
320 /* insert an SRH within an IPv6 packet, just after the IPv6 header */
seg6_do_srh_inline(struct sk_buff * skb,struct ipv6_sr_hdr * osrh)321 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
322 {
323 	struct ipv6hdr *hdr, *oldhdr;
324 	struct ipv6_sr_hdr *isrh;
325 	int hdrlen, err;
326 
327 	hdrlen = (osrh->hdrlen + 1) << 3;
328 
329 	err = skb_cow_head(skb, hdrlen + skb->mac_len);
330 	if (unlikely(err))
331 		return err;
332 
333 	oldhdr = ipv6_hdr(skb);
334 
335 	skb_pull(skb, sizeof(struct ipv6hdr));
336 	skb_postpull_rcsum(skb, skb_network_header(skb),
337 			   sizeof(struct ipv6hdr));
338 
339 	skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
340 	skb_reset_network_header(skb);
341 	skb_mac_header_rebuild(skb);
342 
343 	hdr = ipv6_hdr(skb);
344 
345 	memmove(hdr, oldhdr, sizeof(*hdr));
346 
347 	isrh = (void *)hdr + sizeof(*hdr);
348 	memcpy(isrh, osrh, hdrlen);
349 
350 	isrh->nexthdr = hdr->nexthdr;
351 	hdr->nexthdr = NEXTHDR_ROUTING;
352 
353 	isrh->segments[0] = hdr->daddr;
354 	hdr->daddr = isrh->segments[isrh->first_segment];
355 
356 #ifdef CONFIG_IPV6_SEG6_HMAC
357 	if (sr_has_hmac(isrh)) {
358 		struct net *net = dev_net(skb_dst(skb)->dev);
359 
360 		err = seg6_push_hmac(net, &hdr->saddr, isrh);
361 		if (unlikely(err))
362 			return err;
363 	}
364 #endif
365 
366 	hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
367 
368 	skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
369 
370 	return 0;
371 }
372 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
373 
seg6_do_srh(struct sk_buff * skb)374 static int seg6_do_srh(struct sk_buff *skb)
375 {
376 	struct dst_entry *dst = skb_dst(skb);
377 	struct seg6_iptunnel_encap *tinfo;
378 	int proto, err = 0;
379 
380 	tinfo = seg6_encap_lwtunnel(dst->lwtstate);
381 
382 	switch (tinfo->mode) {
383 	case SEG6_IPTUN_MODE_INLINE:
384 		if (skb->protocol != htons(ETH_P_IPV6))
385 			return -EINVAL;
386 
387 		err = seg6_do_srh_inline(skb, tinfo->srh);
388 		if (err)
389 			return err;
390 		break;
391 	case SEG6_IPTUN_MODE_ENCAP:
392 	case SEG6_IPTUN_MODE_ENCAP_RED:
393 		err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
394 		if (err)
395 			return err;
396 
397 		if (skb->protocol == htons(ETH_P_IPV6))
398 			proto = IPPROTO_IPV6;
399 		else if (skb->protocol == htons(ETH_P_IP))
400 			proto = IPPROTO_IPIP;
401 		else
402 			return -EINVAL;
403 
404 		if (tinfo->mode == SEG6_IPTUN_MODE_ENCAP)
405 			err = seg6_do_srh_encap(skb, tinfo->srh, proto);
406 		else
407 			err = seg6_do_srh_encap_red(skb, tinfo->srh, proto);
408 
409 		if (err)
410 			return err;
411 
412 		skb_set_inner_transport_header(skb, skb_transport_offset(skb));
413 		skb_set_inner_protocol(skb, skb->protocol);
414 		skb->protocol = htons(ETH_P_IPV6);
415 		break;
416 	case SEG6_IPTUN_MODE_L2ENCAP:
417 	case SEG6_IPTUN_MODE_L2ENCAP_RED:
418 		if (!skb_mac_header_was_set(skb))
419 			return -EINVAL;
420 
421 		if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
422 			return -ENOMEM;
423 
424 		skb_mac_header_rebuild(skb);
425 		skb_push(skb, skb->mac_len);
426 
427 		if (tinfo->mode == SEG6_IPTUN_MODE_L2ENCAP)
428 			err = seg6_do_srh_encap(skb, tinfo->srh,
429 						IPPROTO_ETHERNET);
430 		else
431 			err = seg6_do_srh_encap_red(skb, tinfo->srh,
432 						    IPPROTO_ETHERNET);
433 
434 		if (err)
435 			return err;
436 
437 		skb->protocol = htons(ETH_P_IPV6);
438 		break;
439 	}
440 
441 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
442 	nf_reset_ct(skb);
443 
444 	return 0;
445 }
446 
seg6_input_finish(struct net * net,struct sock * sk,struct sk_buff * skb)447 static int seg6_input_finish(struct net *net, struct sock *sk,
448 			     struct sk_buff *skb)
449 {
450 	return dst_input(skb);
451 }
452 
seg6_input_core(struct net * net,struct sock * sk,struct sk_buff * skb)453 static int seg6_input_core(struct net *net, struct sock *sk,
454 			   struct sk_buff *skb)
455 {
456 	struct dst_entry *orig_dst = skb_dst(skb);
457 	struct dst_entry *dst = NULL;
458 	struct seg6_lwt *slwt;
459 	int err;
460 
461 	err = seg6_do_srh(skb);
462 	if (unlikely(err))
463 		goto drop;
464 
465 	slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
466 
467 	local_bh_disable();
468 	dst = dst_cache_get(&slwt->cache);
469 
470 	if (!dst) {
471 		ip6_route_input(skb);
472 		dst = skb_dst(skb);
473 		if (!dst->error) {
474 			dst_cache_set_ip6(&slwt->cache, dst,
475 					  &ipv6_hdr(skb)->saddr);
476 		}
477 	} else {
478 		skb_dst_drop(skb);
479 		skb_dst_set(skb, dst);
480 	}
481 	local_bh_enable();
482 
483 	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
484 	if (unlikely(err))
485 		goto drop;
486 
487 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
488 		return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
489 			       dev_net(skb->dev), NULL, skb, NULL,
490 			       skb_dst(skb)->dev, seg6_input_finish);
491 
492 	return seg6_input_finish(dev_net(skb->dev), NULL, skb);
493 drop:
494 	kfree_skb(skb);
495 	return err;
496 }
497 
seg6_input_nf(struct sk_buff * skb)498 static int seg6_input_nf(struct sk_buff *skb)
499 {
500 	struct net_device *dev = skb_dst(skb)->dev;
501 	struct net *net = dev_net(skb->dev);
502 
503 	switch (skb->protocol) {
504 	case htons(ETH_P_IP):
505 		return NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, net, NULL,
506 			       skb, NULL, dev, seg6_input_core);
507 	case htons(ETH_P_IPV6):
508 		return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, net, NULL,
509 			       skb, NULL, dev, seg6_input_core);
510 	}
511 
512 	return -EINVAL;
513 }
514 
seg6_input(struct sk_buff * skb)515 static int seg6_input(struct sk_buff *skb)
516 {
517 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
518 		return seg6_input_nf(skb);
519 
520 	return seg6_input_core(dev_net(skb->dev), NULL, skb);
521 }
522 
seg6_output_core(struct net * net,struct sock * sk,struct sk_buff * skb)523 static int seg6_output_core(struct net *net, struct sock *sk,
524 			    struct sk_buff *skb)
525 {
526 	struct dst_entry *orig_dst = skb_dst(skb);
527 	struct dst_entry *dst = NULL;
528 	struct seg6_lwt *slwt;
529 	int err;
530 
531 	err = seg6_do_srh(skb);
532 	if (unlikely(err))
533 		goto drop;
534 
535 	slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
536 
537 	local_bh_disable();
538 	dst = dst_cache_get(&slwt->cache);
539 	local_bh_enable();
540 
541 	if (unlikely(!dst)) {
542 		struct ipv6hdr *hdr = ipv6_hdr(skb);
543 		struct flowi6 fl6;
544 
545 		memset(&fl6, 0, sizeof(fl6));
546 		fl6.daddr = hdr->daddr;
547 		fl6.saddr = hdr->saddr;
548 		fl6.flowlabel = ip6_flowinfo(hdr);
549 		fl6.flowi6_mark = skb->mark;
550 		fl6.flowi6_proto = hdr->nexthdr;
551 
552 		dst = ip6_route_output(net, NULL, &fl6);
553 		if (dst->error) {
554 			err = dst->error;
555 			dst_release(dst);
556 			goto drop;
557 		}
558 
559 		local_bh_disable();
560 		dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
561 		local_bh_enable();
562 	}
563 
564 	skb_dst_drop(skb);
565 	skb_dst_set(skb, dst);
566 
567 	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
568 	if (unlikely(err))
569 		goto drop;
570 
571 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
572 		return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
573 			       NULL, skb_dst(skb)->dev, dst_output);
574 
575 	return dst_output(net, sk, skb);
576 drop:
577 	kfree_skb(skb);
578 	return err;
579 }
580 
seg6_output_nf(struct net * net,struct sock * sk,struct sk_buff * skb)581 static int seg6_output_nf(struct net *net, struct sock *sk, struct sk_buff *skb)
582 {
583 	struct net_device *dev = skb_dst(skb)->dev;
584 
585 	switch (skb->protocol) {
586 	case htons(ETH_P_IP):
587 		return NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, net, sk, skb,
588 			       NULL, dev, seg6_output_core);
589 	case htons(ETH_P_IPV6):
590 		return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, net, sk, skb,
591 			       NULL, dev, seg6_output_core);
592 	}
593 
594 	return -EINVAL;
595 }
596 
seg6_output(struct net * net,struct sock * sk,struct sk_buff * skb)597 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
598 {
599 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
600 		return seg6_output_nf(net, sk, skb);
601 
602 	return seg6_output_core(net, sk, skb);
603 }
604 
seg6_build_state(struct net * net,struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)605 static int seg6_build_state(struct net *net, struct nlattr *nla,
606 			    unsigned int family, const void *cfg,
607 			    struct lwtunnel_state **ts,
608 			    struct netlink_ext_ack *extack)
609 {
610 	struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
611 	struct seg6_iptunnel_encap *tuninfo;
612 	struct lwtunnel_state *newts;
613 	int tuninfo_len, min_size;
614 	struct seg6_lwt *slwt;
615 	int err;
616 
617 	if (family != AF_INET && family != AF_INET6)
618 		return -EINVAL;
619 
620 	err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
621 					  seg6_iptunnel_policy, extack);
622 
623 	if (err < 0)
624 		return err;
625 
626 	if (!tb[SEG6_IPTUNNEL_SRH])
627 		return -EINVAL;
628 
629 	tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
630 	tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
631 
632 	/* tuninfo must contain at least the iptunnel encap structure,
633 	 * the SRH and one segment
634 	 */
635 	min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
636 		   sizeof(struct in6_addr);
637 	if (tuninfo_len < min_size)
638 		return -EINVAL;
639 
640 	switch (tuninfo->mode) {
641 	case SEG6_IPTUN_MODE_INLINE:
642 		if (family != AF_INET6)
643 			return -EINVAL;
644 
645 		break;
646 	case SEG6_IPTUN_MODE_ENCAP:
647 		break;
648 	case SEG6_IPTUN_MODE_L2ENCAP:
649 		break;
650 	case SEG6_IPTUN_MODE_ENCAP_RED:
651 		break;
652 	case SEG6_IPTUN_MODE_L2ENCAP_RED:
653 		break;
654 	default:
655 		return -EINVAL;
656 	}
657 
658 	/* verify that SRH is consistent */
659 	if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo), false))
660 		return -EINVAL;
661 
662 	newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
663 	if (!newts)
664 		return -ENOMEM;
665 
666 	slwt = seg6_lwt_lwtunnel(newts);
667 
668 	err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
669 	if (err) {
670 		kfree(newts);
671 		return err;
672 	}
673 
674 	memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
675 
676 	newts->type = LWTUNNEL_ENCAP_SEG6;
677 	newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
678 
679 	if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
680 		newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
681 
682 	newts->headroom = seg6_lwt_headroom(tuninfo);
683 
684 	*ts = newts;
685 
686 	return 0;
687 }
688 
seg6_destroy_state(struct lwtunnel_state * lwt)689 static void seg6_destroy_state(struct lwtunnel_state *lwt)
690 {
691 	dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
692 }
693 
seg6_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)694 static int seg6_fill_encap_info(struct sk_buff *skb,
695 				struct lwtunnel_state *lwtstate)
696 {
697 	struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
698 
699 	if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
700 		return -EMSGSIZE;
701 
702 	return 0;
703 }
704 
seg6_encap_nlsize(struct lwtunnel_state * lwtstate)705 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
706 {
707 	struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
708 
709 	return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
710 }
711 
seg6_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)712 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
713 {
714 	struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
715 	struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
716 	int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
717 
718 	if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
719 		return 1;
720 
721 	return memcmp(a_hdr, b_hdr, len);
722 }
723 
724 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
725 	.build_state = seg6_build_state,
726 	.destroy_state = seg6_destroy_state,
727 	.output = seg6_output,
728 	.input = seg6_input,
729 	.fill_encap = seg6_fill_encap_info,
730 	.get_encap_size = seg6_encap_nlsize,
731 	.cmp_encap = seg6_encap_cmp,
732 	.owner = THIS_MODULE,
733 };
734 
seg6_iptunnel_init(void)735 int __init seg6_iptunnel_init(void)
736 {
737 	return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
738 }
739 
seg6_iptunnel_exit(void)740 void seg6_iptunnel_exit(void)
741 {
742 	lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
743 }
744