xref: /linux/net/ipv6/seg6_local.c (revision d195c39052d1da278a00a6744ce59c383b67b191)
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 #include <net/ip_tunnels.h>
27 #ifdef CONFIG_IPV6_SEG6_HMAC
28 #include <net/seg6_hmac.h>
29 #endif
30 #include <net/seg6_local.h>
31 #include <linux/etherdevice.h>
32 #include <linux/bpf.h>
33 
34 struct seg6_local_lwt;
35 
36 struct seg6_action_desc {
37 	int action;
38 	unsigned long attrs;
39 	int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
40 	int static_headroom;
41 };
42 
43 struct bpf_lwt_prog {
44 	struct bpf_prog *prog;
45 	char *name;
46 };
47 
48 struct seg6_local_lwt {
49 	int action;
50 	struct ipv6_sr_hdr *srh;
51 	int table;
52 	struct in_addr nh4;
53 	struct in6_addr nh6;
54 	int iif;
55 	int oif;
56 	struct bpf_lwt_prog bpf;
57 
58 	int headroom;
59 	struct seg6_action_desc *desc;
60 };
61 
62 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
63 {
64 	return (struct seg6_local_lwt *)lwt->data;
65 }
66 
67 static struct ipv6_sr_hdr *get_srh(struct sk_buff *skb)
68 {
69 	struct ipv6_sr_hdr *srh;
70 	int len, srhoff = 0;
71 
72 	if (ipv6_find_hdr(skb, &srhoff, IPPROTO_ROUTING, NULL, NULL) < 0)
73 		return NULL;
74 
75 	if (!pskb_may_pull(skb, srhoff + sizeof(*srh)))
76 		return NULL;
77 
78 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
79 
80 	len = (srh->hdrlen + 1) << 3;
81 
82 	if (!pskb_may_pull(skb, srhoff + len))
83 		return NULL;
84 
85 	/* note that pskb_may_pull may change pointers in header;
86 	 * for this reason it is necessary to reload them when needed.
87 	 */
88 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
89 
90 	if (!seg6_validate_srh(srh, len))
91 		return NULL;
92 
93 	return srh;
94 }
95 
96 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb)
97 {
98 	struct ipv6_sr_hdr *srh;
99 
100 	srh = get_srh(skb);
101 	if (!srh)
102 		return NULL;
103 
104 	if (srh->segments_left == 0)
105 		return NULL;
106 
107 #ifdef CONFIG_IPV6_SEG6_HMAC
108 	if (!seg6_hmac_validate_skb(skb))
109 		return NULL;
110 #endif
111 
112 	return srh;
113 }
114 
115 static bool decap_and_validate(struct sk_buff *skb, int proto)
116 {
117 	struct ipv6_sr_hdr *srh;
118 	unsigned int off = 0;
119 
120 	srh = get_srh(skb);
121 	if (srh && srh->segments_left > 0)
122 		return false;
123 
124 #ifdef CONFIG_IPV6_SEG6_HMAC
125 	if (srh && !seg6_hmac_validate_skb(skb))
126 		return false;
127 #endif
128 
129 	if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0)
130 		return false;
131 
132 	if (!pskb_pull(skb, off))
133 		return false;
134 
135 	skb_postpull_rcsum(skb, skb_network_header(skb), off);
136 
137 	skb_reset_network_header(skb);
138 	skb_reset_transport_header(skb);
139 	if (iptunnel_pull_offloads(skb))
140 		return false;
141 
142 	return true;
143 }
144 
145 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr)
146 {
147 	struct in6_addr *addr;
148 
149 	srh->segments_left--;
150 	addr = srh->segments + srh->segments_left;
151 	*daddr = *addr;
152 }
153 
154 static int
155 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
156 			u32 tbl_id, bool local_delivery)
157 {
158 	struct net *net = dev_net(skb->dev);
159 	struct ipv6hdr *hdr = ipv6_hdr(skb);
160 	int flags = RT6_LOOKUP_F_HAS_SADDR;
161 	struct dst_entry *dst = NULL;
162 	struct rt6_info *rt;
163 	struct flowi6 fl6;
164 	int dev_flags = 0;
165 
166 	fl6.flowi6_iif = skb->dev->ifindex;
167 	fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
168 	fl6.saddr = hdr->saddr;
169 	fl6.flowlabel = ip6_flowinfo(hdr);
170 	fl6.flowi6_mark = skb->mark;
171 	fl6.flowi6_proto = hdr->nexthdr;
172 
173 	if (nhaddr)
174 		fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
175 
176 	if (!tbl_id) {
177 		dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
178 	} else {
179 		struct fib6_table *table;
180 
181 		table = fib6_get_table(net, tbl_id);
182 		if (!table)
183 			goto out;
184 
185 		rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
186 		dst = &rt->dst;
187 	}
188 
189 	/* we want to discard traffic destined for local packet processing,
190 	 * if @local_delivery is set to false.
191 	 */
192 	if (!local_delivery)
193 		dev_flags |= IFF_LOOPBACK;
194 
195 	if (dst && (dst->dev->flags & dev_flags) && !dst->error) {
196 		dst_release(dst);
197 		dst = NULL;
198 	}
199 
200 out:
201 	if (!dst) {
202 		rt = net->ipv6.ip6_blk_hole_entry;
203 		dst = &rt->dst;
204 		dst_hold(dst);
205 	}
206 
207 	skb_dst_drop(skb);
208 	skb_dst_set(skb, dst);
209 	return dst->error;
210 }
211 
212 int seg6_lookup_nexthop(struct sk_buff *skb,
213 			struct in6_addr *nhaddr, u32 tbl_id)
214 {
215 	return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false);
216 }
217 
218 /* regular endpoint function */
219 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
220 {
221 	struct ipv6_sr_hdr *srh;
222 
223 	srh = get_and_validate_srh(skb);
224 	if (!srh)
225 		goto drop;
226 
227 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
228 
229 	seg6_lookup_nexthop(skb, NULL, 0);
230 
231 	return dst_input(skb);
232 
233 drop:
234 	kfree_skb(skb);
235 	return -EINVAL;
236 }
237 
238 /* regular endpoint, and forward to specified nexthop */
239 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
240 {
241 	struct ipv6_sr_hdr *srh;
242 
243 	srh = get_and_validate_srh(skb);
244 	if (!srh)
245 		goto drop;
246 
247 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
248 
249 	seg6_lookup_nexthop(skb, &slwt->nh6, 0);
250 
251 	return dst_input(skb);
252 
253 drop:
254 	kfree_skb(skb);
255 	return -EINVAL;
256 }
257 
258 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
259 {
260 	struct ipv6_sr_hdr *srh;
261 
262 	srh = get_and_validate_srh(skb);
263 	if (!srh)
264 		goto drop;
265 
266 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
267 
268 	seg6_lookup_nexthop(skb, NULL, slwt->table);
269 
270 	return dst_input(skb);
271 
272 drop:
273 	kfree_skb(skb);
274 	return -EINVAL;
275 }
276 
277 /* decapsulate and forward inner L2 frame on specified interface */
278 static int input_action_end_dx2(struct sk_buff *skb,
279 				struct seg6_local_lwt *slwt)
280 {
281 	struct net *net = dev_net(skb->dev);
282 	struct net_device *odev;
283 	struct ethhdr *eth;
284 
285 	if (!decap_and_validate(skb, IPPROTO_ETHERNET))
286 		goto drop;
287 
288 	if (!pskb_may_pull(skb, ETH_HLEN))
289 		goto drop;
290 
291 	skb_reset_mac_header(skb);
292 	eth = (struct ethhdr *)skb->data;
293 
294 	/* To determine the frame's protocol, we assume it is 802.3. This avoids
295 	 * a call to eth_type_trans(), which is not really relevant for our
296 	 * use case.
297 	 */
298 	if (!eth_proto_is_802_3(eth->h_proto))
299 		goto drop;
300 
301 	odev = dev_get_by_index_rcu(net, slwt->oif);
302 	if (!odev)
303 		goto drop;
304 
305 	/* As we accept Ethernet frames, make sure the egress device is of
306 	 * the correct type.
307 	 */
308 	if (odev->type != ARPHRD_ETHER)
309 		goto drop;
310 
311 	if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
312 		goto drop;
313 
314 	skb_orphan(skb);
315 
316 	if (skb_warn_if_lro(skb))
317 		goto drop;
318 
319 	skb_forward_csum(skb);
320 
321 	if (skb->len - ETH_HLEN > odev->mtu)
322 		goto drop;
323 
324 	skb->dev = odev;
325 	skb->protocol = eth->h_proto;
326 
327 	return dev_queue_xmit(skb);
328 
329 drop:
330 	kfree_skb(skb);
331 	return -EINVAL;
332 }
333 
334 /* decapsulate and forward to specified nexthop */
335 static int input_action_end_dx6(struct sk_buff *skb,
336 				struct seg6_local_lwt *slwt)
337 {
338 	struct in6_addr *nhaddr = NULL;
339 
340 	/* this function accepts IPv6 encapsulated packets, with either
341 	 * an SRH with SL=0, or no SRH.
342 	 */
343 
344 	if (!decap_and_validate(skb, IPPROTO_IPV6))
345 		goto drop;
346 
347 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
348 		goto drop;
349 
350 	/* The inner packet is not associated to any local interface,
351 	 * so we do not call netif_rx().
352 	 *
353 	 * If slwt->nh6 is set to ::, then lookup the nexthop for the
354 	 * inner packet's DA. Otherwise, use the specified nexthop.
355 	 */
356 
357 	if (!ipv6_addr_any(&slwt->nh6))
358 		nhaddr = &slwt->nh6;
359 
360 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
361 
362 	seg6_lookup_nexthop(skb, nhaddr, 0);
363 
364 	return dst_input(skb);
365 drop:
366 	kfree_skb(skb);
367 	return -EINVAL;
368 }
369 
370 static int input_action_end_dx4(struct sk_buff *skb,
371 				struct seg6_local_lwt *slwt)
372 {
373 	struct iphdr *iph;
374 	__be32 nhaddr;
375 	int err;
376 
377 	if (!decap_and_validate(skb, IPPROTO_IPIP))
378 		goto drop;
379 
380 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
381 		goto drop;
382 
383 	skb->protocol = htons(ETH_P_IP);
384 
385 	iph = ip_hdr(skb);
386 
387 	nhaddr = slwt->nh4.s_addr ?: iph->daddr;
388 
389 	skb_dst_drop(skb);
390 
391 	skb_set_transport_header(skb, sizeof(struct iphdr));
392 
393 	err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
394 	if (err)
395 		goto drop;
396 
397 	return dst_input(skb);
398 
399 drop:
400 	kfree_skb(skb);
401 	return -EINVAL;
402 }
403 
404 static int input_action_end_dt6(struct sk_buff *skb,
405 				struct seg6_local_lwt *slwt)
406 {
407 	if (!decap_and_validate(skb, IPPROTO_IPV6))
408 		goto drop;
409 
410 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
411 		goto drop;
412 
413 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
414 
415 	seg6_lookup_any_nexthop(skb, NULL, slwt->table, true);
416 
417 	return dst_input(skb);
418 
419 drop:
420 	kfree_skb(skb);
421 	return -EINVAL;
422 }
423 
424 /* push an SRH on top of the current one */
425 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
426 {
427 	struct ipv6_sr_hdr *srh;
428 	int err = -EINVAL;
429 
430 	srh = get_and_validate_srh(skb);
431 	if (!srh)
432 		goto drop;
433 
434 	err = seg6_do_srh_inline(skb, slwt->srh);
435 	if (err)
436 		goto drop;
437 
438 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
439 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
440 
441 	seg6_lookup_nexthop(skb, NULL, 0);
442 
443 	return dst_input(skb);
444 
445 drop:
446 	kfree_skb(skb);
447 	return err;
448 }
449 
450 /* encapsulate within an outer IPv6 header and a specified SRH */
451 static int input_action_end_b6_encap(struct sk_buff *skb,
452 				     struct seg6_local_lwt *slwt)
453 {
454 	struct ipv6_sr_hdr *srh;
455 	int err = -EINVAL;
456 
457 	srh = get_and_validate_srh(skb);
458 	if (!srh)
459 		goto drop;
460 
461 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
462 
463 	skb_reset_inner_headers(skb);
464 	skb->encapsulation = 1;
465 
466 	err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
467 	if (err)
468 		goto drop;
469 
470 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
471 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
472 
473 	seg6_lookup_nexthop(skb, NULL, 0);
474 
475 	return dst_input(skb);
476 
477 drop:
478 	kfree_skb(skb);
479 	return err;
480 }
481 
482 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
483 
484 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
485 {
486 	struct seg6_bpf_srh_state *srh_state =
487 		this_cpu_ptr(&seg6_bpf_srh_states);
488 	struct ipv6_sr_hdr *srh = srh_state->srh;
489 
490 	if (unlikely(srh == NULL))
491 		return false;
492 
493 	if (unlikely(!srh_state->valid)) {
494 		if ((srh_state->hdrlen & 7) != 0)
495 			return false;
496 
497 		srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
498 		if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3))
499 			return false;
500 
501 		srh_state->valid = true;
502 	}
503 
504 	return true;
505 }
506 
507 static int input_action_end_bpf(struct sk_buff *skb,
508 				struct seg6_local_lwt *slwt)
509 {
510 	struct seg6_bpf_srh_state *srh_state =
511 		this_cpu_ptr(&seg6_bpf_srh_states);
512 	struct ipv6_sr_hdr *srh;
513 	int ret;
514 
515 	srh = get_and_validate_srh(skb);
516 	if (!srh) {
517 		kfree_skb(skb);
518 		return -EINVAL;
519 	}
520 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
521 
522 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
523 	 * which is also accessed by the bpf_lwt_seg6_* helpers
524 	 */
525 	preempt_disable();
526 	srh_state->srh = srh;
527 	srh_state->hdrlen = srh->hdrlen << 3;
528 	srh_state->valid = true;
529 
530 	rcu_read_lock();
531 	bpf_compute_data_pointers(skb);
532 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
533 	rcu_read_unlock();
534 
535 	switch (ret) {
536 	case BPF_OK:
537 	case BPF_REDIRECT:
538 		break;
539 	case BPF_DROP:
540 		goto drop;
541 	default:
542 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
543 		goto drop;
544 	}
545 
546 	if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
547 		goto drop;
548 
549 	preempt_enable();
550 	if (ret != BPF_REDIRECT)
551 		seg6_lookup_nexthop(skb, NULL, 0);
552 
553 	return dst_input(skb);
554 
555 drop:
556 	preempt_enable();
557 	kfree_skb(skb);
558 	return -EINVAL;
559 }
560 
561 static struct seg6_action_desc seg6_action_table[] = {
562 	{
563 		.action		= SEG6_LOCAL_ACTION_END,
564 		.attrs		= 0,
565 		.input		= input_action_end,
566 	},
567 	{
568 		.action		= SEG6_LOCAL_ACTION_END_X,
569 		.attrs		= (1 << SEG6_LOCAL_NH6),
570 		.input		= input_action_end_x,
571 	},
572 	{
573 		.action		= SEG6_LOCAL_ACTION_END_T,
574 		.attrs		= (1 << SEG6_LOCAL_TABLE),
575 		.input		= input_action_end_t,
576 	},
577 	{
578 		.action		= SEG6_LOCAL_ACTION_END_DX2,
579 		.attrs		= (1 << SEG6_LOCAL_OIF),
580 		.input		= input_action_end_dx2,
581 	},
582 	{
583 		.action		= SEG6_LOCAL_ACTION_END_DX6,
584 		.attrs		= (1 << SEG6_LOCAL_NH6),
585 		.input		= input_action_end_dx6,
586 	},
587 	{
588 		.action		= SEG6_LOCAL_ACTION_END_DX4,
589 		.attrs		= (1 << SEG6_LOCAL_NH4),
590 		.input		= input_action_end_dx4,
591 	},
592 	{
593 		.action		= SEG6_LOCAL_ACTION_END_DT6,
594 		.attrs		= (1 << SEG6_LOCAL_TABLE),
595 		.input		= input_action_end_dt6,
596 	},
597 	{
598 		.action		= SEG6_LOCAL_ACTION_END_B6,
599 		.attrs		= (1 << SEG6_LOCAL_SRH),
600 		.input		= input_action_end_b6,
601 	},
602 	{
603 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
604 		.attrs		= (1 << SEG6_LOCAL_SRH),
605 		.input		= input_action_end_b6_encap,
606 		.static_headroom	= sizeof(struct ipv6hdr),
607 	},
608 	{
609 		.action		= SEG6_LOCAL_ACTION_END_BPF,
610 		.attrs		= (1 << SEG6_LOCAL_BPF),
611 		.input		= input_action_end_bpf,
612 	},
613 
614 };
615 
616 static struct seg6_action_desc *__get_action_desc(int action)
617 {
618 	struct seg6_action_desc *desc;
619 	int i, count;
620 
621 	count = ARRAY_SIZE(seg6_action_table);
622 	for (i = 0; i < count; i++) {
623 		desc = &seg6_action_table[i];
624 		if (desc->action == action)
625 			return desc;
626 	}
627 
628 	return NULL;
629 }
630 
631 static int seg6_local_input(struct sk_buff *skb)
632 {
633 	struct dst_entry *orig_dst = skb_dst(skb);
634 	struct seg6_action_desc *desc;
635 	struct seg6_local_lwt *slwt;
636 
637 	if (skb->protocol != htons(ETH_P_IPV6)) {
638 		kfree_skb(skb);
639 		return -EINVAL;
640 	}
641 
642 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
643 	desc = slwt->desc;
644 
645 	return desc->input(skb, slwt);
646 }
647 
648 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
649 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
650 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
651 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
652 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
653 				    .len = sizeof(struct in_addr) },
654 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
655 				    .len = sizeof(struct in6_addr) },
656 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
657 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
658 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
659 };
660 
661 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
662 {
663 	struct ipv6_sr_hdr *srh;
664 	int len;
665 
666 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
667 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
668 
669 	/* SRH must contain at least one segment */
670 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
671 		return -EINVAL;
672 
673 	if (!seg6_validate_srh(srh, len))
674 		return -EINVAL;
675 
676 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
677 	if (!slwt->srh)
678 		return -ENOMEM;
679 
680 	slwt->headroom += len;
681 
682 	return 0;
683 }
684 
685 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
686 {
687 	struct ipv6_sr_hdr *srh;
688 	struct nlattr *nla;
689 	int len;
690 
691 	srh = slwt->srh;
692 	len = (srh->hdrlen + 1) << 3;
693 
694 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
695 	if (!nla)
696 		return -EMSGSIZE;
697 
698 	memcpy(nla_data(nla), srh, len);
699 
700 	return 0;
701 }
702 
703 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
704 {
705 	int len = (a->srh->hdrlen + 1) << 3;
706 
707 	if (len != ((b->srh->hdrlen + 1) << 3))
708 		return 1;
709 
710 	return memcmp(a->srh, b->srh, len);
711 }
712 
713 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
714 {
715 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
716 
717 	return 0;
718 }
719 
720 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
721 {
722 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
723 		return -EMSGSIZE;
724 
725 	return 0;
726 }
727 
728 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
729 {
730 	if (a->table != b->table)
731 		return 1;
732 
733 	return 0;
734 }
735 
736 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
737 {
738 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
739 	       sizeof(struct in_addr));
740 
741 	return 0;
742 }
743 
744 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
745 {
746 	struct nlattr *nla;
747 
748 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
749 	if (!nla)
750 		return -EMSGSIZE;
751 
752 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
753 
754 	return 0;
755 }
756 
757 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
758 {
759 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
760 }
761 
762 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
763 {
764 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
765 	       sizeof(struct in6_addr));
766 
767 	return 0;
768 }
769 
770 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
771 {
772 	struct nlattr *nla;
773 
774 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
775 	if (!nla)
776 		return -EMSGSIZE;
777 
778 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
779 
780 	return 0;
781 }
782 
783 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
784 {
785 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
786 }
787 
788 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
789 {
790 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
791 
792 	return 0;
793 }
794 
795 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
796 {
797 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
798 		return -EMSGSIZE;
799 
800 	return 0;
801 }
802 
803 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
804 {
805 	if (a->iif != b->iif)
806 		return 1;
807 
808 	return 0;
809 }
810 
811 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
812 {
813 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
814 
815 	return 0;
816 }
817 
818 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
819 {
820 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
821 		return -EMSGSIZE;
822 
823 	return 0;
824 }
825 
826 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
827 {
828 	if (a->oif != b->oif)
829 		return 1;
830 
831 	return 0;
832 }
833 
834 #define MAX_PROG_NAME 256
835 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
836 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
837 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
838 				       .len = MAX_PROG_NAME },
839 };
840 
841 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
842 {
843 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
844 	struct bpf_prog *p;
845 	int ret;
846 	u32 fd;
847 
848 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
849 					  attrs[SEG6_LOCAL_BPF],
850 					  bpf_prog_policy, NULL);
851 	if (ret < 0)
852 		return ret;
853 
854 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
855 		return -EINVAL;
856 
857 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
858 	if (!slwt->bpf.name)
859 		return -ENOMEM;
860 
861 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
862 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
863 	if (IS_ERR(p)) {
864 		kfree(slwt->bpf.name);
865 		return PTR_ERR(p);
866 	}
867 
868 	slwt->bpf.prog = p;
869 	return 0;
870 }
871 
872 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
873 {
874 	struct nlattr *nest;
875 
876 	if (!slwt->bpf.prog)
877 		return 0;
878 
879 	nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
880 	if (!nest)
881 		return -EMSGSIZE;
882 
883 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
884 		return -EMSGSIZE;
885 
886 	if (slwt->bpf.name &&
887 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
888 		return -EMSGSIZE;
889 
890 	return nla_nest_end(skb, nest);
891 }
892 
893 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
894 {
895 	if (!a->bpf.name && !b->bpf.name)
896 		return 0;
897 
898 	if (!a->bpf.name || !b->bpf.name)
899 		return 1;
900 
901 	return strcmp(a->bpf.name, b->bpf.name);
902 }
903 
904 struct seg6_action_param {
905 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
906 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
907 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
908 };
909 
910 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
911 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
912 				    .put = put_nla_srh,
913 				    .cmp = cmp_nla_srh },
914 
915 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
916 				    .put = put_nla_table,
917 				    .cmp = cmp_nla_table },
918 
919 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
920 				    .put = put_nla_nh4,
921 				    .cmp = cmp_nla_nh4 },
922 
923 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
924 				    .put = put_nla_nh6,
925 				    .cmp = cmp_nla_nh6 },
926 
927 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
928 				    .put = put_nla_iif,
929 				    .cmp = cmp_nla_iif },
930 
931 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
932 				    .put = put_nla_oif,
933 				    .cmp = cmp_nla_oif },
934 
935 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
936 				    .put = put_nla_bpf,
937 				    .cmp = cmp_nla_bpf },
938 
939 };
940 
941 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
942 {
943 	struct seg6_action_param *param;
944 	struct seg6_action_desc *desc;
945 	int i, err;
946 
947 	desc = __get_action_desc(slwt->action);
948 	if (!desc)
949 		return -EINVAL;
950 
951 	if (!desc->input)
952 		return -EOPNOTSUPP;
953 
954 	slwt->desc = desc;
955 	slwt->headroom += desc->static_headroom;
956 
957 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
958 		if (desc->attrs & (1 << i)) {
959 			if (!attrs[i])
960 				return -EINVAL;
961 
962 			param = &seg6_action_params[i];
963 
964 			err = param->parse(attrs, slwt);
965 			if (err < 0)
966 				return err;
967 		}
968 	}
969 
970 	return 0;
971 }
972 
973 static int seg6_local_build_state(struct net *net, struct nlattr *nla,
974 				  unsigned int family, const void *cfg,
975 				  struct lwtunnel_state **ts,
976 				  struct netlink_ext_ack *extack)
977 {
978 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
979 	struct lwtunnel_state *newts;
980 	struct seg6_local_lwt *slwt;
981 	int err;
982 
983 	if (family != AF_INET6)
984 		return -EINVAL;
985 
986 	err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
987 					  seg6_local_policy, extack);
988 
989 	if (err < 0)
990 		return err;
991 
992 	if (!tb[SEG6_LOCAL_ACTION])
993 		return -EINVAL;
994 
995 	newts = lwtunnel_state_alloc(sizeof(*slwt));
996 	if (!newts)
997 		return -ENOMEM;
998 
999 	slwt = seg6_local_lwtunnel(newts);
1000 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
1001 
1002 	err = parse_nla_action(tb, slwt);
1003 	if (err < 0)
1004 		goto out_free;
1005 
1006 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
1007 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
1008 	newts->headroom = slwt->headroom;
1009 
1010 	*ts = newts;
1011 
1012 	return 0;
1013 
1014 out_free:
1015 	kfree(slwt->srh);
1016 	kfree(newts);
1017 	return err;
1018 }
1019 
1020 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
1021 {
1022 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1023 
1024 	kfree(slwt->srh);
1025 
1026 	if (slwt->desc->attrs & (1 << SEG6_LOCAL_BPF)) {
1027 		kfree(slwt->bpf.name);
1028 		bpf_prog_put(slwt->bpf.prog);
1029 	}
1030 
1031 	return;
1032 }
1033 
1034 static int seg6_local_fill_encap(struct sk_buff *skb,
1035 				 struct lwtunnel_state *lwt)
1036 {
1037 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1038 	struct seg6_action_param *param;
1039 	int i, err;
1040 
1041 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1042 		return -EMSGSIZE;
1043 
1044 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1045 		if (slwt->desc->attrs & (1 << i)) {
1046 			param = &seg6_action_params[i];
1047 			err = param->put(skb, slwt);
1048 			if (err < 0)
1049 				return err;
1050 		}
1051 	}
1052 
1053 	return 0;
1054 }
1055 
1056 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1057 {
1058 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1059 	unsigned long attrs;
1060 	int nlsize;
1061 
1062 	nlsize = nla_total_size(4); /* action */
1063 
1064 	attrs = slwt->desc->attrs;
1065 
1066 	if (attrs & (1 << SEG6_LOCAL_SRH))
1067 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1068 
1069 	if (attrs & (1 << SEG6_LOCAL_TABLE))
1070 		nlsize += nla_total_size(4);
1071 
1072 	if (attrs & (1 << SEG6_LOCAL_NH4))
1073 		nlsize += nla_total_size(4);
1074 
1075 	if (attrs & (1 << SEG6_LOCAL_NH6))
1076 		nlsize += nla_total_size(16);
1077 
1078 	if (attrs & (1 << SEG6_LOCAL_IIF))
1079 		nlsize += nla_total_size(4);
1080 
1081 	if (attrs & (1 << SEG6_LOCAL_OIF))
1082 		nlsize += nla_total_size(4);
1083 
1084 	if (attrs & (1 << SEG6_LOCAL_BPF))
1085 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1086 		       nla_total_size(MAX_PROG_NAME) +
1087 		       nla_total_size(4);
1088 
1089 	return nlsize;
1090 }
1091 
1092 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1093 				struct lwtunnel_state *b)
1094 {
1095 	struct seg6_local_lwt *slwt_a, *slwt_b;
1096 	struct seg6_action_param *param;
1097 	int i;
1098 
1099 	slwt_a = seg6_local_lwtunnel(a);
1100 	slwt_b = seg6_local_lwtunnel(b);
1101 
1102 	if (slwt_a->action != slwt_b->action)
1103 		return 1;
1104 
1105 	if (slwt_a->desc->attrs != slwt_b->desc->attrs)
1106 		return 1;
1107 
1108 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1109 		if (slwt_a->desc->attrs & (1 << i)) {
1110 			param = &seg6_action_params[i];
1111 			if (param->cmp(slwt_a, slwt_b))
1112 				return 1;
1113 		}
1114 	}
1115 
1116 	return 0;
1117 }
1118 
1119 static const struct lwtunnel_encap_ops seg6_local_ops = {
1120 	.build_state	= seg6_local_build_state,
1121 	.destroy_state	= seg6_local_destroy_state,
1122 	.input		= seg6_local_input,
1123 	.fill_encap	= seg6_local_fill_encap,
1124 	.get_encap_size	= seg6_local_get_encap_size,
1125 	.cmp_encap	= seg6_local_cmp_encap,
1126 	.owner		= THIS_MODULE,
1127 };
1128 
1129 int __init seg6_local_init(void)
1130 {
1131 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1132 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1133 }
1134 
1135 void seg6_local_exit(void)
1136 {
1137 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1138 }
1139