xref: /openbmc/linux/net/ipv6/seg6_local.c (revision c24c57a4) 
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  SR-IPv6 implementation
4  *
5  *  Authors:
6  *  David Lebrun <david.lebrun@uclouvain.be>
7  *  eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com>
8  */
9 
10 #include <linux/types.h>
11 #include <linux/skbuff.h>
12 #include <linux/net.h>
13 #include <linux/module.h>
14 #include <net/ip.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_local.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 <net/seg6_local.h>
30 #include <linux/etherdevice.h>
31 #include <linux/bpf.h>
32 
33 struct seg6_local_lwt;
34 
35 struct seg6_action_desc {
36 	int action;
37 	unsigned long attrs;
38 	int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
39 	int static_headroom;
40 };
41 
42 struct bpf_lwt_prog {
43 	struct bpf_prog *prog;
44 	char *name;
45 };
46 
47 struct seg6_local_lwt {
48 	int action;
49 	struct ipv6_sr_hdr *srh;
50 	int table;
51 	struct in_addr nh4;
52 	struct in6_addr nh6;
53 	int iif;
54 	int oif;
55 	struct bpf_lwt_prog bpf;
56 
57 	int headroom;
58 	struct seg6_action_desc *desc;
59 };
60 
61 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
62 {
63 	return (struct seg6_local_lwt *)lwt->data;
64 }
65 
66 static struct ipv6_sr_hdr *get_srh(struct sk_buff *skb)
67 {
68 	struct ipv6_sr_hdr *srh;
69 	int len, srhoff = 0;
70 
71 	if (ipv6_find_hdr(skb, &srhoff, IPPROTO_ROUTING, NULL, NULL) < 0)
72 		return NULL;
73 
74 	if (!pskb_may_pull(skb, srhoff + sizeof(*srh)))
75 		return NULL;
76 
77 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
78 
79 	len = (srh->hdrlen + 1) << 3;
80 
81 	if (!pskb_may_pull(skb, srhoff + len))
82 		return NULL;
83 
84 	/* note that pskb_may_pull may change pointers in header;
85 	 * for this reason it is necessary to reload them when needed.
86 	 */
87 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
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 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
345 
346 	seg6_lookup_nexthop(skb, nhaddr, 0);
347 
348 	return dst_input(skb);
349 drop:
350 	kfree_skb(skb);
351 	return -EINVAL;
352 }
353 
354 static int input_action_end_dx4(struct sk_buff *skb,
355 				struct seg6_local_lwt *slwt)
356 {
357 	struct iphdr *iph;
358 	__be32 nhaddr;
359 	int err;
360 
361 	if (!decap_and_validate(skb, IPPROTO_IPIP))
362 		goto drop;
363 
364 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
365 		goto drop;
366 
367 	skb->protocol = htons(ETH_P_IP);
368 
369 	iph = ip_hdr(skb);
370 
371 	nhaddr = slwt->nh4.s_addr ?: iph->daddr;
372 
373 	skb_dst_drop(skb);
374 
375 	skb_set_transport_header(skb, sizeof(struct iphdr));
376 
377 	err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
378 	if (err)
379 		goto drop;
380 
381 	return dst_input(skb);
382 
383 drop:
384 	kfree_skb(skb);
385 	return -EINVAL;
386 }
387 
388 static int input_action_end_dt6(struct sk_buff *skb,
389 				struct seg6_local_lwt *slwt)
390 {
391 	if (!decap_and_validate(skb, IPPROTO_IPV6))
392 		goto drop;
393 
394 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
395 		goto drop;
396 
397 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
398 
399 	seg6_lookup_nexthop(skb, NULL, slwt->table);
400 
401 	return dst_input(skb);
402 
403 drop:
404 	kfree_skb(skb);
405 	return -EINVAL;
406 }
407 
408 /* push an SRH on top of the current one */
409 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
410 {
411 	struct ipv6_sr_hdr *srh;
412 	int err = -EINVAL;
413 
414 	srh = get_and_validate_srh(skb);
415 	if (!srh)
416 		goto drop;
417 
418 	err = seg6_do_srh_inline(skb, slwt->srh);
419 	if (err)
420 		goto drop;
421 
422 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
423 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
424 
425 	seg6_lookup_nexthop(skb, NULL, 0);
426 
427 	return dst_input(skb);
428 
429 drop:
430 	kfree_skb(skb);
431 	return err;
432 }
433 
434 /* encapsulate within an outer IPv6 header and a specified SRH */
435 static int input_action_end_b6_encap(struct sk_buff *skb,
436 				     struct seg6_local_lwt *slwt)
437 {
438 	struct ipv6_sr_hdr *srh;
439 	int err = -EINVAL;
440 
441 	srh = get_and_validate_srh(skb);
442 	if (!srh)
443 		goto drop;
444 
445 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
446 
447 	skb_reset_inner_headers(skb);
448 	skb->encapsulation = 1;
449 
450 	err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
451 	if (err)
452 		goto drop;
453 
454 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
455 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
456 
457 	seg6_lookup_nexthop(skb, NULL, 0);
458 
459 	return dst_input(skb);
460 
461 drop:
462 	kfree_skb(skb);
463 	return err;
464 }
465 
466 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
467 
468 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
469 {
470 	struct seg6_bpf_srh_state *srh_state =
471 		this_cpu_ptr(&seg6_bpf_srh_states);
472 	struct ipv6_sr_hdr *srh = srh_state->srh;
473 
474 	if (unlikely(srh == NULL))
475 		return false;
476 
477 	if (unlikely(!srh_state->valid)) {
478 		if ((srh_state->hdrlen & 7) != 0)
479 			return false;
480 
481 		srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
482 		if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3))
483 			return false;
484 
485 		srh_state->valid = true;
486 	}
487 
488 	return true;
489 }
490 
491 static int input_action_end_bpf(struct sk_buff *skb,
492 				struct seg6_local_lwt *slwt)
493 {
494 	struct seg6_bpf_srh_state *srh_state =
495 		this_cpu_ptr(&seg6_bpf_srh_states);
496 	struct ipv6_sr_hdr *srh;
497 	int ret;
498 
499 	srh = get_and_validate_srh(skb);
500 	if (!srh) {
501 		kfree_skb(skb);
502 		return -EINVAL;
503 	}
504 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
505 
506 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
507 	 * which is also accessed by the bpf_lwt_seg6_* helpers
508 	 */
509 	preempt_disable();
510 	srh_state->srh = srh;
511 	srh_state->hdrlen = srh->hdrlen << 3;
512 	srh_state->valid = true;
513 
514 	rcu_read_lock();
515 	bpf_compute_data_pointers(skb);
516 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
517 	rcu_read_unlock();
518 
519 	switch (ret) {
520 	case BPF_OK:
521 	case BPF_REDIRECT:
522 		break;
523 	case BPF_DROP:
524 		goto drop;
525 	default:
526 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
527 		goto drop;
528 	}
529 
530 	if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
531 		goto drop;
532 
533 	preempt_enable();
534 	if (ret != BPF_REDIRECT)
535 		seg6_lookup_nexthop(skb, NULL, 0);
536 
537 	return dst_input(skb);
538 
539 drop:
540 	preempt_enable();
541 	kfree_skb(skb);
542 	return -EINVAL;
543 }
544 
545 static struct seg6_action_desc seg6_action_table[] = {
546 	{
547 		.action		= SEG6_LOCAL_ACTION_END,
548 		.attrs		= 0,
549 		.input		= input_action_end,
550 	},
551 	{
552 		.action		= SEG6_LOCAL_ACTION_END_X,
553 		.attrs		= (1 << SEG6_LOCAL_NH6),
554 		.input		= input_action_end_x,
555 	},
556 	{
557 		.action		= SEG6_LOCAL_ACTION_END_T,
558 		.attrs		= (1 << SEG6_LOCAL_TABLE),
559 		.input		= input_action_end_t,
560 	},
561 	{
562 		.action		= SEG6_LOCAL_ACTION_END_DX2,
563 		.attrs		= (1 << SEG6_LOCAL_OIF),
564 		.input		= input_action_end_dx2,
565 	},
566 	{
567 		.action		= SEG6_LOCAL_ACTION_END_DX6,
568 		.attrs		= (1 << SEG6_LOCAL_NH6),
569 		.input		= input_action_end_dx6,
570 	},
571 	{
572 		.action		= SEG6_LOCAL_ACTION_END_DX4,
573 		.attrs		= (1 << SEG6_LOCAL_NH4),
574 		.input		= input_action_end_dx4,
575 	},
576 	{
577 		.action		= SEG6_LOCAL_ACTION_END_DT6,
578 		.attrs		= (1 << SEG6_LOCAL_TABLE),
579 		.input		= input_action_end_dt6,
580 	},
581 	{
582 		.action		= SEG6_LOCAL_ACTION_END_B6,
583 		.attrs		= (1 << SEG6_LOCAL_SRH),
584 		.input		= input_action_end_b6,
585 	},
586 	{
587 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
588 		.attrs		= (1 << SEG6_LOCAL_SRH),
589 		.input		= input_action_end_b6_encap,
590 		.static_headroom	= sizeof(struct ipv6hdr),
591 	},
592 	{
593 		.action		= SEG6_LOCAL_ACTION_END_BPF,
594 		.attrs		= (1 << SEG6_LOCAL_BPF),
595 		.input		= input_action_end_bpf,
596 	},
597 
598 };
599 
600 static struct seg6_action_desc *__get_action_desc(int action)
601 {
602 	struct seg6_action_desc *desc;
603 	int i, count;
604 
605 	count = ARRAY_SIZE(seg6_action_table);
606 	for (i = 0; i < count; i++) {
607 		desc = &seg6_action_table[i];
608 		if (desc->action == action)
609 			return desc;
610 	}
611 
612 	return NULL;
613 }
614 
615 static int seg6_local_input(struct sk_buff *skb)
616 {
617 	struct dst_entry *orig_dst = skb_dst(skb);
618 	struct seg6_action_desc *desc;
619 	struct seg6_local_lwt *slwt;
620 
621 	if (skb->protocol != htons(ETH_P_IPV6)) {
622 		kfree_skb(skb);
623 		return -EINVAL;
624 	}
625 
626 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
627 	desc = slwt->desc;
628 
629 	return desc->input(skb, slwt);
630 }
631 
632 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
633 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
634 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
635 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
636 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
637 				    .len = sizeof(struct in_addr) },
638 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
639 				    .len = sizeof(struct in6_addr) },
640 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
641 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
642 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
643 };
644 
645 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
646 {
647 	struct ipv6_sr_hdr *srh;
648 	int len;
649 
650 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
651 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
652 
653 	/* SRH must contain at least one segment */
654 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
655 		return -EINVAL;
656 
657 	if (!seg6_validate_srh(srh, len))
658 		return -EINVAL;
659 
660 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
661 	if (!slwt->srh)
662 		return -ENOMEM;
663 
664 	slwt->headroom += len;
665 
666 	return 0;
667 }
668 
669 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
670 {
671 	struct ipv6_sr_hdr *srh;
672 	struct nlattr *nla;
673 	int len;
674 
675 	srh = slwt->srh;
676 	len = (srh->hdrlen + 1) << 3;
677 
678 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
679 	if (!nla)
680 		return -EMSGSIZE;
681 
682 	memcpy(nla_data(nla), srh, len);
683 
684 	return 0;
685 }
686 
687 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
688 {
689 	int len = (a->srh->hdrlen + 1) << 3;
690 
691 	if (len != ((b->srh->hdrlen + 1) << 3))
692 		return 1;
693 
694 	return memcmp(a->srh, b->srh, len);
695 }
696 
697 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
698 {
699 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
700 
701 	return 0;
702 }
703 
704 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
705 {
706 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
707 		return -EMSGSIZE;
708 
709 	return 0;
710 }
711 
712 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
713 {
714 	if (a->table != b->table)
715 		return 1;
716 
717 	return 0;
718 }
719 
720 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
721 {
722 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
723 	       sizeof(struct in_addr));
724 
725 	return 0;
726 }
727 
728 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
729 {
730 	struct nlattr *nla;
731 
732 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
733 	if (!nla)
734 		return -EMSGSIZE;
735 
736 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
737 
738 	return 0;
739 }
740 
741 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
742 {
743 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
744 }
745 
746 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
747 {
748 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
749 	       sizeof(struct in6_addr));
750 
751 	return 0;
752 }
753 
754 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
755 {
756 	struct nlattr *nla;
757 
758 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
759 	if (!nla)
760 		return -EMSGSIZE;
761 
762 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
763 
764 	return 0;
765 }
766 
767 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
768 {
769 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
770 }
771 
772 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
773 {
774 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
775 
776 	return 0;
777 }
778 
779 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
780 {
781 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
782 		return -EMSGSIZE;
783 
784 	return 0;
785 }
786 
787 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
788 {
789 	if (a->iif != b->iif)
790 		return 1;
791 
792 	return 0;
793 }
794 
795 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
796 {
797 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
798 
799 	return 0;
800 }
801 
802 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
803 {
804 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
805 		return -EMSGSIZE;
806 
807 	return 0;
808 }
809 
810 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
811 {
812 	if (a->oif != b->oif)
813 		return 1;
814 
815 	return 0;
816 }
817 
818 #define MAX_PROG_NAME 256
819 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
820 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
821 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
822 				       .len = MAX_PROG_NAME },
823 };
824 
825 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
826 {
827 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
828 	struct bpf_prog *p;
829 	int ret;
830 	u32 fd;
831 
832 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
833 					  attrs[SEG6_LOCAL_BPF],
834 					  bpf_prog_policy, NULL);
835 	if (ret < 0)
836 		return ret;
837 
838 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
839 		return -EINVAL;
840 
841 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
842 	if (!slwt->bpf.name)
843 		return -ENOMEM;
844 
845 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
846 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
847 	if (IS_ERR(p)) {
848 		kfree(slwt->bpf.name);
849 		return PTR_ERR(p);
850 	}
851 
852 	slwt->bpf.prog = p;
853 	return 0;
854 }
855 
856 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
857 {
858 	struct nlattr *nest;
859 
860 	if (!slwt->bpf.prog)
861 		return 0;
862 
863 	nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
864 	if (!nest)
865 		return -EMSGSIZE;
866 
867 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
868 		return -EMSGSIZE;
869 
870 	if (slwt->bpf.name &&
871 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
872 		return -EMSGSIZE;
873 
874 	return nla_nest_end(skb, nest);
875 }
876 
877 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
878 {
879 	if (!a->bpf.name && !b->bpf.name)
880 		return 0;
881 
882 	if (!a->bpf.name || !b->bpf.name)
883 		return 1;
884 
885 	return strcmp(a->bpf.name, b->bpf.name);
886 }
887 
888 struct seg6_action_param {
889 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
890 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
891 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
892 };
893 
894 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
895 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
896 				    .put = put_nla_srh,
897 				    .cmp = cmp_nla_srh },
898 
899 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
900 				    .put = put_nla_table,
901 				    .cmp = cmp_nla_table },
902 
903 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
904 				    .put = put_nla_nh4,
905 				    .cmp = cmp_nla_nh4 },
906 
907 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
908 				    .put = put_nla_nh6,
909 				    .cmp = cmp_nla_nh6 },
910 
911 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
912 				    .put = put_nla_iif,
913 				    .cmp = cmp_nla_iif },
914 
915 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
916 				    .put = put_nla_oif,
917 				    .cmp = cmp_nla_oif },
918 
919 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
920 				    .put = put_nla_bpf,
921 				    .cmp = cmp_nla_bpf },
922 
923 };
924 
925 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
926 {
927 	struct seg6_action_param *param;
928 	struct seg6_action_desc *desc;
929 	int i, err;
930 
931 	desc = __get_action_desc(slwt->action);
932 	if (!desc)
933 		return -EINVAL;
934 
935 	if (!desc->input)
936 		return -EOPNOTSUPP;
937 
938 	slwt->desc = desc;
939 	slwt->headroom += desc->static_headroom;
940 
941 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
942 		if (desc->attrs & (1 << i)) {
943 			if (!attrs[i])
944 				return -EINVAL;
945 
946 			param = &seg6_action_params[i];
947 
948 			err = param->parse(attrs, slwt);
949 			if (err < 0)
950 				return err;
951 		}
952 	}
953 
954 	return 0;
955 }
956 
957 static int seg6_local_build_state(struct nlattr *nla, unsigned int family,
958 				  const void *cfg, struct lwtunnel_state **ts,
959 				  struct netlink_ext_ack *extack)
960 {
961 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
962 	struct lwtunnel_state *newts;
963 	struct seg6_local_lwt *slwt;
964 	int err;
965 
966 	if (family != AF_INET6)
967 		return -EINVAL;
968 
969 	err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
970 					  seg6_local_policy, extack);
971 
972 	if (err < 0)
973 		return err;
974 
975 	if (!tb[SEG6_LOCAL_ACTION])
976 		return -EINVAL;
977 
978 	newts = lwtunnel_state_alloc(sizeof(*slwt));
979 	if (!newts)
980 		return -ENOMEM;
981 
982 	slwt = seg6_local_lwtunnel(newts);
983 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
984 
985 	err = parse_nla_action(tb, slwt);
986 	if (err < 0)
987 		goto out_free;
988 
989 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
990 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
991 	newts->headroom = slwt->headroom;
992 
993 	*ts = newts;
994 
995 	return 0;
996 
997 out_free:
998 	kfree(slwt->srh);
999 	kfree(newts);
1000 	return err;
1001 }
1002 
1003 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
1004 {
1005 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1006 
1007 	kfree(slwt->srh);
1008 
1009 	if (slwt->desc->attrs & (1 << SEG6_LOCAL_BPF)) {
1010 		kfree(slwt->bpf.name);
1011 		bpf_prog_put(slwt->bpf.prog);
1012 	}
1013 
1014 	return;
1015 }
1016 
1017 static int seg6_local_fill_encap(struct sk_buff *skb,
1018 				 struct lwtunnel_state *lwt)
1019 {
1020 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1021 	struct seg6_action_param *param;
1022 	int i, err;
1023 
1024 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1025 		return -EMSGSIZE;
1026 
1027 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1028 		if (slwt->desc->attrs & (1 << i)) {
1029 			param = &seg6_action_params[i];
1030 			err = param->put(skb, slwt);
1031 			if (err < 0)
1032 				return err;
1033 		}
1034 	}
1035 
1036 	return 0;
1037 }
1038 
1039 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1040 {
1041 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1042 	unsigned long attrs;
1043 	int nlsize;
1044 
1045 	nlsize = nla_total_size(4); /* action */
1046 
1047 	attrs = slwt->desc->attrs;
1048 
1049 	if (attrs & (1 << SEG6_LOCAL_SRH))
1050 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1051 
1052 	if (attrs & (1 << SEG6_LOCAL_TABLE))
1053 		nlsize += nla_total_size(4);
1054 
1055 	if (attrs & (1 << SEG6_LOCAL_NH4))
1056 		nlsize += nla_total_size(4);
1057 
1058 	if (attrs & (1 << SEG6_LOCAL_NH6))
1059 		nlsize += nla_total_size(16);
1060 
1061 	if (attrs & (1 << SEG6_LOCAL_IIF))
1062 		nlsize += nla_total_size(4);
1063 
1064 	if (attrs & (1 << SEG6_LOCAL_OIF))
1065 		nlsize += nla_total_size(4);
1066 
1067 	if (attrs & (1 << SEG6_LOCAL_BPF))
1068 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1069 		       nla_total_size(MAX_PROG_NAME) +
1070 		       nla_total_size(4);
1071 
1072 	return nlsize;
1073 }
1074 
1075 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1076 				struct lwtunnel_state *b)
1077 {
1078 	struct seg6_local_lwt *slwt_a, *slwt_b;
1079 	struct seg6_action_param *param;
1080 	int i;
1081 
1082 	slwt_a = seg6_local_lwtunnel(a);
1083 	slwt_b = seg6_local_lwtunnel(b);
1084 
1085 	if (slwt_a->action != slwt_b->action)
1086 		return 1;
1087 
1088 	if (slwt_a->desc->attrs != slwt_b->desc->attrs)
1089 		return 1;
1090 
1091 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1092 		if (slwt_a->desc->attrs & (1 << i)) {
1093 			param = &seg6_action_params[i];
1094 			if (param->cmp(slwt_a, slwt_b))
1095 				return 1;
1096 		}
1097 	}
1098 
1099 	return 0;
1100 }
1101 
1102 static const struct lwtunnel_encap_ops seg6_local_ops = {
1103 	.build_state	= seg6_local_build_state,
1104 	.destroy_state	= seg6_local_destroy_state,
1105 	.input		= seg6_local_input,
1106 	.fill_encap	= seg6_local_fill_encap,
1107 	.get_encap_size	= seg6_local_get_encap_size,
1108 	.cmp_encap	= seg6_local_cmp_encap,
1109 	.owner		= THIS_MODULE,
1110 };
1111 
1112 int __init seg6_local_init(void)
1113 {
1114 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1115 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1116 }
1117 
1118 void seg6_local_exit(void)
1119 {
1120 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1121 }
1122