xref: /openbmc/linux/net/ipv6/seg6_local.c (revision 4fc4dca8) 
1 /*
2  *  SR-IPv6 implementation
3  *
4  *  Authors:
5  *  David Lebrun <david.lebrun@uclouvain.be>
6  *  eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com>
7  *
8  *
9  *  This program is free software; you can redistribute it and/or
10  *        modify it under the terms of the GNU General Public License
11  *        as published by the Free Software Foundation; either version
12  *        2 of the License, or (at your option) any later version.
13  */
14 
15 #include <linux/types.h>
16 #include <linux/skbuff.h>
17 #include <linux/net.h>
18 #include <linux/module.h>
19 #include <net/ip.h>
20 #include <net/lwtunnel.h>
21 #include <net/netevent.h>
22 #include <net/netns/generic.h>
23 #include <net/ip6_fib.h>
24 #include <net/route.h>
25 #include <net/seg6.h>
26 #include <linux/seg6.h>
27 #include <linux/seg6_local.h>
28 #include <net/addrconf.h>
29 #include <net/ip6_route.h>
30 #include <net/dst_cache.h>
31 #ifdef CONFIG_IPV6_SEG6_HMAC
32 #include <net/seg6_hmac.h>
33 #endif
34 #include <net/seg6_local.h>
35 #include <linux/etherdevice.h>
36 #include <linux/bpf.h>
37 
38 struct seg6_local_lwt;
39 
40 struct seg6_action_desc {
41 	int action;
42 	unsigned long attrs;
43 	int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
44 	int static_headroom;
45 };
46 
47 struct bpf_lwt_prog {
48 	struct bpf_prog *prog;
49 	char *name;
50 };
51 
52 struct seg6_local_lwt {
53 	int action;
54 	struct ipv6_sr_hdr *srh;
55 	int table;
56 	struct in_addr nh4;
57 	struct in6_addr nh6;
58 	int iif;
59 	int oif;
60 	struct bpf_lwt_prog bpf;
61 
62 	int headroom;
63 	struct seg6_action_desc *desc;
64 };
65 
66 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
67 {
68 	return (struct seg6_local_lwt *)lwt->data;
69 }
70 
71 static struct ipv6_sr_hdr *get_srh(struct sk_buff *skb)
72 {
73 	struct ipv6_sr_hdr *srh;
74 	int len, srhoff = 0;
75 
76 	if (ipv6_find_hdr(skb, &srhoff, IPPROTO_ROUTING, NULL, NULL) < 0)
77 		return NULL;
78 
79 	if (!pskb_may_pull(skb, srhoff + sizeof(*srh)))
80 		return NULL;
81 
82 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
83 
84 	len = (srh->hdrlen + 1) << 3;
85 
86 	if (!pskb_may_pull(skb, srhoff + len))
87 		return NULL;
88 
89 	if (!seg6_validate_srh(srh, len))
90 		return NULL;
91 
92 	return srh;
93 }
94 
95 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb)
96 {
97 	struct ipv6_sr_hdr *srh;
98 
99 	srh = get_srh(skb);
100 	if (!srh)
101 		return NULL;
102 
103 	if (srh->segments_left == 0)
104 		return NULL;
105 
106 #ifdef CONFIG_IPV6_SEG6_HMAC
107 	if (!seg6_hmac_validate_skb(skb))
108 		return NULL;
109 #endif
110 
111 	return srh;
112 }
113 
114 static bool decap_and_validate(struct sk_buff *skb, int proto)
115 {
116 	struct ipv6_sr_hdr *srh;
117 	unsigned int off = 0;
118 
119 	srh = get_srh(skb);
120 	if (srh && srh->segments_left > 0)
121 		return false;
122 
123 #ifdef CONFIG_IPV6_SEG6_HMAC
124 	if (srh && !seg6_hmac_validate_skb(skb))
125 		return false;
126 #endif
127 
128 	if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0)
129 		return false;
130 
131 	if (!pskb_pull(skb, off))
132 		return false;
133 
134 	skb_postpull_rcsum(skb, skb_network_header(skb), off);
135 
136 	skb_reset_network_header(skb);
137 	skb_reset_transport_header(skb);
138 	skb->encapsulation = 0;
139 
140 	return true;
141 }
142 
143 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr)
144 {
145 	struct in6_addr *addr;
146 
147 	srh->segments_left--;
148 	addr = srh->segments + srh->segments_left;
149 	*daddr = *addr;
150 }
151 
152 int seg6_lookup_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
153 			u32 tbl_id)
154 {
155 	struct net *net = dev_net(skb->dev);
156 	struct ipv6hdr *hdr = ipv6_hdr(skb);
157 	int flags = RT6_LOOKUP_F_HAS_SADDR;
158 	struct dst_entry *dst = NULL;
159 	struct rt6_info *rt;
160 	struct flowi6 fl6;
161 
162 	fl6.flowi6_iif = skb->dev->ifindex;
163 	fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
164 	fl6.saddr = hdr->saddr;
165 	fl6.flowlabel = ip6_flowinfo(hdr);
166 	fl6.flowi6_mark = skb->mark;
167 	fl6.flowi6_proto = hdr->nexthdr;
168 
169 	if (nhaddr)
170 		fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
171 
172 	if (!tbl_id) {
173 		dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
174 	} else {
175 		struct fib6_table *table;
176 
177 		table = fib6_get_table(net, tbl_id);
178 		if (!table)
179 			goto out;
180 
181 		rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
182 		dst = &rt->dst;
183 	}
184 
185 	if (dst && dst->dev->flags & IFF_LOOPBACK && !dst->error) {
186 		dst_release(dst);
187 		dst = NULL;
188 	}
189 
190 out:
191 	if (!dst) {
192 		rt = net->ipv6.ip6_blk_hole_entry;
193 		dst = &rt->dst;
194 		dst_hold(dst);
195 	}
196 
197 	skb_dst_drop(skb);
198 	skb_dst_set(skb, dst);
199 	return dst->error;
200 }
201 
202 /* regular endpoint function */
203 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
204 {
205 	struct ipv6_sr_hdr *srh;
206 
207 	srh = get_and_validate_srh(skb);
208 	if (!srh)
209 		goto drop;
210 
211 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
212 
213 	seg6_lookup_nexthop(skb, NULL, 0);
214 
215 	return dst_input(skb);
216 
217 drop:
218 	kfree_skb(skb);
219 	return -EINVAL;
220 }
221 
222 /* regular endpoint, and forward to specified nexthop */
223 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
224 {
225 	struct ipv6_sr_hdr *srh;
226 
227 	srh = get_and_validate_srh(skb);
228 	if (!srh)
229 		goto drop;
230 
231 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
232 
233 	seg6_lookup_nexthop(skb, &slwt->nh6, 0);
234 
235 	return dst_input(skb);
236 
237 drop:
238 	kfree_skb(skb);
239 	return -EINVAL;
240 }
241 
242 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
243 {
244 	struct ipv6_sr_hdr *srh;
245 
246 	srh = get_and_validate_srh(skb);
247 	if (!srh)
248 		goto drop;
249 
250 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
251 
252 	seg6_lookup_nexthop(skb, NULL, slwt->table);
253 
254 	return dst_input(skb);
255 
256 drop:
257 	kfree_skb(skb);
258 	return -EINVAL;
259 }
260 
261 /* decapsulate and forward inner L2 frame on specified interface */
262 static int input_action_end_dx2(struct sk_buff *skb,
263 				struct seg6_local_lwt *slwt)
264 {
265 	struct net *net = dev_net(skb->dev);
266 	struct net_device *odev;
267 	struct ethhdr *eth;
268 
269 	if (!decap_and_validate(skb, NEXTHDR_NONE))
270 		goto drop;
271 
272 	if (!pskb_may_pull(skb, ETH_HLEN))
273 		goto drop;
274 
275 	skb_reset_mac_header(skb);
276 	eth = (struct ethhdr *)skb->data;
277 
278 	/* To determine the frame's protocol, we assume it is 802.3. This avoids
279 	 * a call to eth_type_trans(), which is not really relevant for our
280 	 * use case.
281 	 */
282 	if (!eth_proto_is_802_3(eth->h_proto))
283 		goto drop;
284 
285 	odev = dev_get_by_index_rcu(net, slwt->oif);
286 	if (!odev)
287 		goto drop;
288 
289 	/* As we accept Ethernet frames, make sure the egress device is of
290 	 * the correct type.
291 	 */
292 	if (odev->type != ARPHRD_ETHER)
293 		goto drop;
294 
295 	if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
296 		goto drop;
297 
298 	skb_orphan(skb);
299 
300 	if (skb_warn_if_lro(skb))
301 		goto drop;
302 
303 	skb_forward_csum(skb);
304 
305 	if (skb->len - ETH_HLEN > odev->mtu)
306 		goto drop;
307 
308 	skb->dev = odev;
309 	skb->protocol = eth->h_proto;
310 
311 	return dev_queue_xmit(skb);
312 
313 drop:
314 	kfree_skb(skb);
315 	return -EINVAL;
316 }
317 
318 /* decapsulate and forward to specified nexthop */
319 static int input_action_end_dx6(struct sk_buff *skb,
320 				struct seg6_local_lwt *slwt)
321 {
322 	struct in6_addr *nhaddr = NULL;
323 
324 	/* this function accepts IPv6 encapsulated packets, with either
325 	 * an SRH with SL=0, or no SRH.
326 	 */
327 
328 	if (!decap_and_validate(skb, IPPROTO_IPV6))
329 		goto drop;
330 
331 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
332 		goto drop;
333 
334 	/* The inner packet is not associated to any local interface,
335 	 * so we do not call netif_rx().
336 	 *
337 	 * If slwt->nh6 is set to ::, then lookup the nexthop for the
338 	 * inner packet's DA. Otherwise, use the specified nexthop.
339 	 */
340 
341 	if (!ipv6_addr_any(&slwt->nh6))
342 		nhaddr = &slwt->nh6;
343 
344 	seg6_lookup_nexthop(skb, nhaddr, 0);
345 
346 	return dst_input(skb);
347 drop:
348 	kfree_skb(skb);
349 	return -EINVAL;
350 }
351 
352 static int input_action_end_dx4(struct sk_buff *skb,
353 				struct seg6_local_lwt *slwt)
354 {
355 	struct iphdr *iph;
356 	__be32 nhaddr;
357 	int err;
358 
359 	if (!decap_and_validate(skb, IPPROTO_IPIP))
360 		goto drop;
361 
362 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
363 		goto drop;
364 
365 	skb->protocol = htons(ETH_P_IP);
366 
367 	iph = ip_hdr(skb);
368 
369 	nhaddr = slwt->nh4.s_addr ?: iph->daddr;
370 
371 	skb_dst_drop(skb);
372 
373 	err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
374 	if (err)
375 		goto drop;
376 
377 	return dst_input(skb);
378 
379 drop:
380 	kfree_skb(skb);
381 	return -EINVAL;
382 }
383 
384 static int input_action_end_dt6(struct sk_buff *skb,
385 				struct seg6_local_lwt *slwt)
386 {
387 	if (!decap_and_validate(skb, IPPROTO_IPV6))
388 		goto drop;
389 
390 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
391 		goto drop;
392 
393 	seg6_lookup_nexthop(skb, NULL, slwt->table);
394 
395 	return dst_input(skb);
396 
397 drop:
398 	kfree_skb(skb);
399 	return -EINVAL;
400 }
401 
402 /* push an SRH on top of the current one */
403 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
404 {
405 	struct ipv6_sr_hdr *srh;
406 	int err = -EINVAL;
407 
408 	srh = get_and_validate_srh(skb);
409 	if (!srh)
410 		goto drop;
411 
412 	err = seg6_do_srh_inline(skb, slwt->srh);
413 	if (err)
414 		goto drop;
415 
416 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
417 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
418 
419 	seg6_lookup_nexthop(skb, NULL, 0);
420 
421 	return dst_input(skb);
422 
423 drop:
424 	kfree_skb(skb);
425 	return err;
426 }
427 
428 /* encapsulate within an outer IPv6 header and a specified SRH */
429 static int input_action_end_b6_encap(struct sk_buff *skb,
430 				     struct seg6_local_lwt *slwt)
431 {
432 	struct ipv6_sr_hdr *srh;
433 	int err = -EINVAL;
434 
435 	srh = get_and_validate_srh(skb);
436 	if (!srh)
437 		goto drop;
438 
439 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
440 
441 	skb_reset_inner_headers(skb);
442 	skb->encapsulation = 1;
443 
444 	err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
445 	if (err)
446 		goto drop;
447 
448 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
449 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
450 
451 	seg6_lookup_nexthop(skb, NULL, 0);
452 
453 	return dst_input(skb);
454 
455 drop:
456 	kfree_skb(skb);
457 	return err;
458 }
459 
460 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
461 
462 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
463 {
464 	struct seg6_bpf_srh_state *srh_state =
465 		this_cpu_ptr(&seg6_bpf_srh_states);
466 	struct ipv6_sr_hdr *srh = srh_state->srh;
467 
468 	if (unlikely(srh == NULL))
469 		return false;
470 
471 	if (unlikely(!srh_state->valid)) {
472 		if ((srh_state->hdrlen & 7) != 0)
473 			return false;
474 
475 		srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
476 		if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3))
477 			return false;
478 
479 		srh_state->valid = true;
480 	}
481 
482 	return true;
483 }
484 
485 static int input_action_end_bpf(struct sk_buff *skb,
486 				struct seg6_local_lwt *slwt)
487 {
488 	struct seg6_bpf_srh_state *srh_state =
489 		this_cpu_ptr(&seg6_bpf_srh_states);
490 	struct ipv6_sr_hdr *srh;
491 	int ret;
492 
493 	srh = get_and_validate_srh(skb);
494 	if (!srh) {
495 		kfree_skb(skb);
496 		return -EINVAL;
497 	}
498 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
499 
500 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
501 	 * which is also accessed by the bpf_lwt_seg6_* helpers
502 	 */
503 	preempt_disable();
504 	srh_state->srh = srh;
505 	srh_state->hdrlen = srh->hdrlen << 3;
506 	srh_state->valid = true;
507 
508 	rcu_read_lock();
509 	bpf_compute_data_pointers(skb);
510 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
511 	rcu_read_unlock();
512 
513 	switch (ret) {
514 	case BPF_OK:
515 	case BPF_REDIRECT:
516 		break;
517 	case BPF_DROP:
518 		goto drop;
519 	default:
520 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
521 		goto drop;
522 	}
523 
524 	if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
525 		goto drop;
526 
527 	preempt_enable();
528 	if (ret != BPF_REDIRECT)
529 		seg6_lookup_nexthop(skb, NULL, 0);
530 
531 	return dst_input(skb);
532 
533 drop:
534 	preempt_enable();
535 	kfree_skb(skb);
536 	return -EINVAL;
537 }
538 
539 static struct seg6_action_desc seg6_action_table[] = {
540 	{
541 		.action		= SEG6_LOCAL_ACTION_END,
542 		.attrs		= 0,
543 		.input		= input_action_end,
544 	},
545 	{
546 		.action		= SEG6_LOCAL_ACTION_END_X,
547 		.attrs		= (1 << SEG6_LOCAL_NH6),
548 		.input		= input_action_end_x,
549 	},
550 	{
551 		.action		= SEG6_LOCAL_ACTION_END_T,
552 		.attrs		= (1 << SEG6_LOCAL_TABLE),
553 		.input		= input_action_end_t,
554 	},
555 	{
556 		.action		= SEG6_LOCAL_ACTION_END_DX2,
557 		.attrs		= (1 << SEG6_LOCAL_OIF),
558 		.input		= input_action_end_dx2,
559 	},
560 	{
561 		.action		= SEG6_LOCAL_ACTION_END_DX6,
562 		.attrs		= (1 << SEG6_LOCAL_NH6),
563 		.input		= input_action_end_dx6,
564 	},
565 	{
566 		.action		= SEG6_LOCAL_ACTION_END_DX4,
567 		.attrs		= (1 << SEG6_LOCAL_NH4),
568 		.input		= input_action_end_dx4,
569 	},
570 	{
571 		.action		= SEG6_LOCAL_ACTION_END_DT6,
572 		.attrs		= (1 << SEG6_LOCAL_TABLE),
573 		.input		= input_action_end_dt6,
574 	},
575 	{
576 		.action		= SEG6_LOCAL_ACTION_END_B6,
577 		.attrs		= (1 << SEG6_LOCAL_SRH),
578 		.input		= input_action_end_b6,
579 	},
580 	{
581 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
582 		.attrs		= (1 << SEG6_LOCAL_SRH),
583 		.input		= input_action_end_b6_encap,
584 		.static_headroom	= sizeof(struct ipv6hdr),
585 	},
586 	{
587 		.action		= SEG6_LOCAL_ACTION_END_BPF,
588 		.attrs		= (1 << SEG6_LOCAL_BPF),
589 		.input		= input_action_end_bpf,
590 	},
591 
592 };
593 
594 static struct seg6_action_desc *__get_action_desc(int action)
595 {
596 	struct seg6_action_desc *desc;
597 	int i, count;
598 
599 	count = ARRAY_SIZE(seg6_action_table);
600 	for (i = 0; i < count; i++) {
601 		desc = &seg6_action_table[i];
602 		if (desc->action == action)
603 			return desc;
604 	}
605 
606 	return NULL;
607 }
608 
609 static int seg6_local_input(struct sk_buff *skb)
610 {
611 	struct dst_entry *orig_dst = skb_dst(skb);
612 	struct seg6_action_desc *desc;
613 	struct seg6_local_lwt *slwt;
614 
615 	if (skb->protocol != htons(ETH_P_IPV6)) {
616 		kfree_skb(skb);
617 		return -EINVAL;
618 	}
619 
620 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
621 	desc = slwt->desc;
622 
623 	return desc->input(skb, slwt);
624 }
625 
626 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
627 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
628 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
629 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
630 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
631 				    .len = sizeof(struct in_addr) },
632 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
633 				    .len = sizeof(struct in6_addr) },
634 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
635 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
636 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
637 };
638 
639 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
640 {
641 	struct ipv6_sr_hdr *srh;
642 	int len;
643 
644 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
645 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
646 
647 	/* SRH must contain at least one segment */
648 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
649 		return -EINVAL;
650 
651 	if (!seg6_validate_srh(srh, len))
652 		return -EINVAL;
653 
654 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
655 	if (!slwt->srh)
656 		return -ENOMEM;
657 
658 	slwt->headroom += len;
659 
660 	return 0;
661 }
662 
663 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
664 {
665 	struct ipv6_sr_hdr *srh;
666 	struct nlattr *nla;
667 	int len;
668 
669 	srh = slwt->srh;
670 	len = (srh->hdrlen + 1) << 3;
671 
672 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
673 	if (!nla)
674 		return -EMSGSIZE;
675 
676 	memcpy(nla_data(nla), srh, len);
677 
678 	return 0;
679 }
680 
681 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
682 {
683 	int len = (a->srh->hdrlen + 1) << 3;
684 
685 	if (len != ((b->srh->hdrlen + 1) << 3))
686 		return 1;
687 
688 	return memcmp(a->srh, b->srh, len);
689 }
690 
691 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
692 {
693 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
694 
695 	return 0;
696 }
697 
698 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
699 {
700 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
701 		return -EMSGSIZE;
702 
703 	return 0;
704 }
705 
706 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
707 {
708 	if (a->table != b->table)
709 		return 1;
710 
711 	return 0;
712 }
713 
714 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
715 {
716 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
717 	       sizeof(struct in_addr));
718 
719 	return 0;
720 }
721 
722 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
723 {
724 	struct nlattr *nla;
725 
726 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
727 	if (!nla)
728 		return -EMSGSIZE;
729 
730 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
731 
732 	return 0;
733 }
734 
735 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
736 {
737 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
738 }
739 
740 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
741 {
742 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
743 	       sizeof(struct in6_addr));
744 
745 	return 0;
746 }
747 
748 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
749 {
750 	struct nlattr *nla;
751 
752 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
753 	if (!nla)
754 		return -EMSGSIZE;
755 
756 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
757 
758 	return 0;
759 }
760 
761 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
762 {
763 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
764 }
765 
766 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
767 {
768 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
769 
770 	return 0;
771 }
772 
773 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
774 {
775 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
776 		return -EMSGSIZE;
777 
778 	return 0;
779 }
780 
781 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
782 {
783 	if (a->iif != b->iif)
784 		return 1;
785 
786 	return 0;
787 }
788 
789 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
790 {
791 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
792 
793 	return 0;
794 }
795 
796 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
797 {
798 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
799 		return -EMSGSIZE;
800 
801 	return 0;
802 }
803 
804 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
805 {
806 	if (a->oif != b->oif)
807 		return 1;
808 
809 	return 0;
810 }
811 
812 #define MAX_PROG_NAME 256
813 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
814 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
815 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
816 				       .len = MAX_PROG_NAME },
817 };
818 
819 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
820 {
821 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
822 	struct bpf_prog *p;
823 	int ret;
824 	u32 fd;
825 
826 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
827 					  attrs[SEG6_LOCAL_BPF],
828 					  bpf_prog_policy, NULL);
829 	if (ret < 0)
830 		return ret;
831 
832 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
833 		return -EINVAL;
834 
835 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
836 	if (!slwt->bpf.name)
837 		return -ENOMEM;
838 
839 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
840 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
841 	if (IS_ERR(p)) {
842 		kfree(slwt->bpf.name);
843 		return PTR_ERR(p);
844 	}
845 
846 	slwt->bpf.prog = p;
847 	return 0;
848 }
849 
850 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
851 {
852 	struct nlattr *nest;
853 
854 	if (!slwt->bpf.prog)
855 		return 0;
856 
857 	nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
858 	if (!nest)
859 		return -EMSGSIZE;
860 
861 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
862 		return -EMSGSIZE;
863 
864 	if (slwt->bpf.name &&
865 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
866 		return -EMSGSIZE;
867 
868 	return nla_nest_end(skb, nest);
869 }
870 
871 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
872 {
873 	if (!a->bpf.name && !b->bpf.name)
874 		return 0;
875 
876 	if (!a->bpf.name || !b->bpf.name)
877 		return 1;
878 
879 	return strcmp(a->bpf.name, b->bpf.name);
880 }
881 
882 struct seg6_action_param {
883 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
884 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
885 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
886 };
887 
888 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
889 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
890 				    .put = put_nla_srh,
891 				    .cmp = cmp_nla_srh },
892 
893 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
894 				    .put = put_nla_table,
895 				    .cmp = cmp_nla_table },
896 
897 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
898 				    .put = put_nla_nh4,
899 				    .cmp = cmp_nla_nh4 },
900 
901 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
902 				    .put = put_nla_nh6,
903 				    .cmp = cmp_nla_nh6 },
904 
905 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
906 				    .put = put_nla_iif,
907 				    .cmp = cmp_nla_iif },
908 
909 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
910 				    .put = put_nla_oif,
911 				    .cmp = cmp_nla_oif },
912 
913 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
914 				    .put = put_nla_bpf,
915 				    .cmp = cmp_nla_bpf },
916 
917 };
918 
919 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
920 {
921 	struct seg6_action_param *param;
922 	struct seg6_action_desc *desc;
923 	int i, err;
924 
925 	desc = __get_action_desc(slwt->action);
926 	if (!desc)
927 		return -EINVAL;
928 
929 	if (!desc->input)
930 		return -EOPNOTSUPP;
931 
932 	slwt->desc = desc;
933 	slwt->headroom += desc->static_headroom;
934 
935 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
936 		if (desc->attrs & (1 << i)) {
937 			if (!attrs[i])
938 				return -EINVAL;
939 
940 			param = &seg6_action_params[i];
941 
942 			err = param->parse(attrs, slwt);
943 			if (err < 0)
944 				return err;
945 		}
946 	}
947 
948 	return 0;
949 }
950 
951 static int seg6_local_build_state(struct nlattr *nla, unsigned int family,
952 				  const void *cfg, struct lwtunnel_state **ts,
953 				  struct netlink_ext_ack *extack)
954 {
955 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
956 	struct lwtunnel_state *newts;
957 	struct seg6_local_lwt *slwt;
958 	int err;
959 
960 	if (family != AF_INET6)
961 		return -EINVAL;
962 
963 	err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
964 					  seg6_local_policy, extack);
965 
966 	if (err < 0)
967 		return err;
968 
969 	if (!tb[SEG6_LOCAL_ACTION])
970 		return -EINVAL;
971 
972 	newts = lwtunnel_state_alloc(sizeof(*slwt));
973 	if (!newts)
974 		return -ENOMEM;
975 
976 	slwt = seg6_local_lwtunnel(newts);
977 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
978 
979 	err = parse_nla_action(tb, slwt);
980 	if (err < 0)
981 		goto out_free;
982 
983 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
984 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
985 	newts->headroom = slwt->headroom;
986 
987 	*ts = newts;
988 
989 	return 0;
990 
991 out_free:
992 	kfree(slwt->srh);
993 	kfree(newts);
994 	return err;
995 }
996 
997 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
998 {
999 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1000 
1001 	kfree(slwt->srh);
1002 
1003 	if (slwt->desc->attrs & (1 << SEG6_LOCAL_BPF)) {
1004 		kfree(slwt->bpf.name);
1005 		bpf_prog_put(slwt->bpf.prog);
1006 	}
1007 
1008 	return;
1009 }
1010 
1011 static int seg6_local_fill_encap(struct sk_buff *skb,
1012 				 struct lwtunnel_state *lwt)
1013 {
1014 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1015 	struct seg6_action_param *param;
1016 	int i, err;
1017 
1018 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1019 		return -EMSGSIZE;
1020 
1021 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1022 		if (slwt->desc->attrs & (1 << i)) {
1023 			param = &seg6_action_params[i];
1024 			err = param->put(skb, slwt);
1025 			if (err < 0)
1026 				return err;
1027 		}
1028 	}
1029 
1030 	return 0;
1031 }
1032 
1033 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1034 {
1035 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1036 	unsigned long attrs;
1037 	int nlsize;
1038 
1039 	nlsize = nla_total_size(4); /* action */
1040 
1041 	attrs = slwt->desc->attrs;
1042 
1043 	if (attrs & (1 << SEG6_LOCAL_SRH))
1044 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1045 
1046 	if (attrs & (1 << SEG6_LOCAL_TABLE))
1047 		nlsize += nla_total_size(4);
1048 
1049 	if (attrs & (1 << SEG6_LOCAL_NH4))
1050 		nlsize += nla_total_size(4);
1051 
1052 	if (attrs & (1 << SEG6_LOCAL_NH6))
1053 		nlsize += nla_total_size(16);
1054 
1055 	if (attrs & (1 << SEG6_LOCAL_IIF))
1056 		nlsize += nla_total_size(4);
1057 
1058 	if (attrs & (1 << SEG6_LOCAL_OIF))
1059 		nlsize += nla_total_size(4);
1060 
1061 	if (attrs & (1 << SEG6_LOCAL_BPF))
1062 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1063 		       nla_total_size(MAX_PROG_NAME) +
1064 		       nla_total_size(4);
1065 
1066 	return nlsize;
1067 }
1068 
1069 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1070 				struct lwtunnel_state *b)
1071 {
1072 	struct seg6_local_lwt *slwt_a, *slwt_b;
1073 	struct seg6_action_param *param;
1074 	int i;
1075 
1076 	slwt_a = seg6_local_lwtunnel(a);
1077 	slwt_b = seg6_local_lwtunnel(b);
1078 
1079 	if (slwt_a->action != slwt_b->action)
1080 		return 1;
1081 
1082 	if (slwt_a->desc->attrs != slwt_b->desc->attrs)
1083 		return 1;
1084 
1085 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1086 		if (slwt_a->desc->attrs & (1 << i)) {
1087 			param = &seg6_action_params[i];
1088 			if (param->cmp(slwt_a, slwt_b))
1089 				return 1;
1090 		}
1091 	}
1092 
1093 	return 0;
1094 }
1095 
1096 static const struct lwtunnel_encap_ops seg6_local_ops = {
1097 	.build_state	= seg6_local_build_state,
1098 	.destroy_state	= seg6_local_destroy_state,
1099 	.input		= seg6_local_input,
1100 	.fill_encap	= seg6_local_fill_encap,
1101 	.get_encap_size	= seg6_local_get_encap_size,
1102 	.cmp_encap	= seg6_local_cmp_encap,
1103 	.owner		= THIS_MODULE,
1104 };
1105 
1106 int __init seg6_local_init(void)
1107 {
1108 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1109 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1110 }
1111 
1112 void seg6_local_exit(void)
1113 {
1114 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1115 }
1116