xref: /openbmc/linux/net/ipv6/seg6_local.c (revision a17922de) 
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 static int input_action_end_bpf(struct sk_buff *skb,
463 				struct seg6_local_lwt *slwt)
464 {
465 	struct seg6_bpf_srh_state *srh_state =
466 		this_cpu_ptr(&seg6_bpf_srh_states);
467 	struct seg6_bpf_srh_state local_srh_state;
468 	struct ipv6_sr_hdr *srh;
469 	int srhoff = 0;
470 	int ret;
471 
472 	srh = get_and_validate_srh(skb);
473 	if (!srh)
474 		goto drop;
475 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
476 
477 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
478 	 * which is also accessed by the bpf_lwt_seg6_* helpers
479 	 */
480 	preempt_disable();
481 	srh_state->hdrlen = srh->hdrlen << 3;
482 	srh_state->valid = 1;
483 
484 	rcu_read_lock();
485 	bpf_compute_data_pointers(skb);
486 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
487 	rcu_read_unlock();
488 
489 	local_srh_state = *srh_state;
490 	preempt_enable();
491 
492 	switch (ret) {
493 	case BPF_OK:
494 	case BPF_REDIRECT:
495 		break;
496 	case BPF_DROP:
497 		goto drop;
498 	default:
499 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
500 		goto drop;
501 	}
502 
503 	if (unlikely((local_srh_state.hdrlen & 7) != 0))
504 		goto drop;
505 
506 	if (ipv6_find_hdr(skb, &srhoff, IPPROTO_ROUTING, NULL, NULL) < 0)
507 		goto drop;
508 	srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
509 	srh->hdrlen = (u8)(local_srh_state.hdrlen >> 3);
510 
511 	if (!local_srh_state.valid &&
512 	    unlikely(!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3)))
513 		goto drop;
514 
515 	if (ret != BPF_REDIRECT)
516 		seg6_lookup_nexthop(skb, NULL, 0);
517 
518 	return dst_input(skb);
519 
520 drop:
521 	kfree_skb(skb);
522 	return -EINVAL;
523 }
524 
525 static struct seg6_action_desc seg6_action_table[] = {
526 	{
527 		.action		= SEG6_LOCAL_ACTION_END,
528 		.attrs		= 0,
529 		.input		= input_action_end,
530 	},
531 	{
532 		.action		= SEG6_LOCAL_ACTION_END_X,
533 		.attrs		= (1 << SEG6_LOCAL_NH6),
534 		.input		= input_action_end_x,
535 	},
536 	{
537 		.action		= SEG6_LOCAL_ACTION_END_T,
538 		.attrs		= (1 << SEG6_LOCAL_TABLE),
539 		.input		= input_action_end_t,
540 	},
541 	{
542 		.action		= SEG6_LOCAL_ACTION_END_DX2,
543 		.attrs		= (1 << SEG6_LOCAL_OIF),
544 		.input		= input_action_end_dx2,
545 	},
546 	{
547 		.action		= SEG6_LOCAL_ACTION_END_DX6,
548 		.attrs		= (1 << SEG6_LOCAL_NH6),
549 		.input		= input_action_end_dx6,
550 	},
551 	{
552 		.action		= SEG6_LOCAL_ACTION_END_DX4,
553 		.attrs		= (1 << SEG6_LOCAL_NH4),
554 		.input		= input_action_end_dx4,
555 	},
556 	{
557 		.action		= SEG6_LOCAL_ACTION_END_DT6,
558 		.attrs		= (1 << SEG6_LOCAL_TABLE),
559 		.input		= input_action_end_dt6,
560 	},
561 	{
562 		.action		= SEG6_LOCAL_ACTION_END_B6,
563 		.attrs		= (1 << SEG6_LOCAL_SRH),
564 		.input		= input_action_end_b6,
565 	},
566 	{
567 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
568 		.attrs		= (1 << SEG6_LOCAL_SRH),
569 		.input		= input_action_end_b6_encap,
570 		.static_headroom	= sizeof(struct ipv6hdr),
571 	},
572 	{
573 		.action		= SEG6_LOCAL_ACTION_END_BPF,
574 		.attrs		= (1 << SEG6_LOCAL_BPF),
575 		.input		= input_action_end_bpf,
576 	},
577 
578 };
579 
580 static struct seg6_action_desc *__get_action_desc(int action)
581 {
582 	struct seg6_action_desc *desc;
583 	int i, count;
584 
585 	count = ARRAY_SIZE(seg6_action_table);
586 	for (i = 0; i < count; i++) {
587 		desc = &seg6_action_table[i];
588 		if (desc->action == action)
589 			return desc;
590 	}
591 
592 	return NULL;
593 }
594 
595 static int seg6_local_input(struct sk_buff *skb)
596 {
597 	struct dst_entry *orig_dst = skb_dst(skb);
598 	struct seg6_action_desc *desc;
599 	struct seg6_local_lwt *slwt;
600 
601 	if (skb->protocol != htons(ETH_P_IPV6)) {
602 		kfree_skb(skb);
603 		return -EINVAL;
604 	}
605 
606 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
607 	desc = slwt->desc;
608 
609 	return desc->input(skb, slwt);
610 }
611 
612 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
613 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
614 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
615 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
616 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
617 				    .len = sizeof(struct in_addr) },
618 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
619 				    .len = sizeof(struct in6_addr) },
620 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
621 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
622 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
623 };
624 
625 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
626 {
627 	struct ipv6_sr_hdr *srh;
628 	int len;
629 
630 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
631 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
632 
633 	/* SRH must contain at least one segment */
634 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
635 		return -EINVAL;
636 
637 	if (!seg6_validate_srh(srh, len))
638 		return -EINVAL;
639 
640 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
641 	if (!slwt->srh)
642 		return -ENOMEM;
643 
644 	slwt->headroom += len;
645 
646 	return 0;
647 }
648 
649 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
650 {
651 	struct ipv6_sr_hdr *srh;
652 	struct nlattr *nla;
653 	int len;
654 
655 	srh = slwt->srh;
656 	len = (srh->hdrlen + 1) << 3;
657 
658 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
659 	if (!nla)
660 		return -EMSGSIZE;
661 
662 	memcpy(nla_data(nla), srh, len);
663 
664 	return 0;
665 }
666 
667 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
668 {
669 	int len = (a->srh->hdrlen + 1) << 3;
670 
671 	if (len != ((b->srh->hdrlen + 1) << 3))
672 		return 1;
673 
674 	return memcmp(a->srh, b->srh, len);
675 }
676 
677 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
678 {
679 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
680 
681 	return 0;
682 }
683 
684 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
685 {
686 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
687 		return -EMSGSIZE;
688 
689 	return 0;
690 }
691 
692 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
693 {
694 	if (a->table != b->table)
695 		return 1;
696 
697 	return 0;
698 }
699 
700 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
701 {
702 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
703 	       sizeof(struct in_addr));
704 
705 	return 0;
706 }
707 
708 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
709 {
710 	struct nlattr *nla;
711 
712 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
713 	if (!nla)
714 		return -EMSGSIZE;
715 
716 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
717 
718 	return 0;
719 }
720 
721 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
722 {
723 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
724 }
725 
726 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
727 {
728 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
729 	       sizeof(struct in6_addr));
730 
731 	return 0;
732 }
733 
734 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
735 {
736 	struct nlattr *nla;
737 
738 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
739 	if (!nla)
740 		return -EMSGSIZE;
741 
742 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
743 
744 	return 0;
745 }
746 
747 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
748 {
749 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
750 }
751 
752 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
753 {
754 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
755 
756 	return 0;
757 }
758 
759 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
760 {
761 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
762 		return -EMSGSIZE;
763 
764 	return 0;
765 }
766 
767 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
768 {
769 	if (a->iif != b->iif)
770 		return 1;
771 
772 	return 0;
773 }
774 
775 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
776 {
777 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
778 
779 	return 0;
780 }
781 
782 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
783 {
784 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
785 		return -EMSGSIZE;
786 
787 	return 0;
788 }
789 
790 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
791 {
792 	if (a->oif != b->oif)
793 		return 1;
794 
795 	return 0;
796 }
797 
798 #define MAX_PROG_NAME 256
799 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
800 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
801 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
802 				       .len = MAX_PROG_NAME },
803 };
804 
805 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
806 {
807 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
808 	struct bpf_prog *p;
809 	int ret;
810 	u32 fd;
811 
812 	ret = nla_parse_nested(tb, SEG6_LOCAL_BPF_PROG_MAX,
813 			       attrs[SEG6_LOCAL_BPF], bpf_prog_policy, NULL);
814 	if (ret < 0)
815 		return ret;
816 
817 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
818 		return -EINVAL;
819 
820 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
821 	if (!slwt->bpf.name)
822 		return -ENOMEM;
823 
824 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
825 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
826 	if (IS_ERR(p)) {
827 		kfree(slwt->bpf.name);
828 		return PTR_ERR(p);
829 	}
830 
831 	slwt->bpf.prog = p;
832 	return 0;
833 }
834 
835 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
836 {
837 	struct nlattr *nest;
838 
839 	if (!slwt->bpf.prog)
840 		return 0;
841 
842 	nest = nla_nest_start(skb, SEG6_LOCAL_BPF);
843 	if (!nest)
844 		return -EMSGSIZE;
845 
846 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
847 		return -EMSGSIZE;
848 
849 	if (slwt->bpf.name &&
850 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
851 		return -EMSGSIZE;
852 
853 	return nla_nest_end(skb, nest);
854 }
855 
856 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
857 {
858 	if (!a->bpf.name && !b->bpf.name)
859 		return 0;
860 
861 	if (!a->bpf.name || !b->bpf.name)
862 		return 1;
863 
864 	return strcmp(a->bpf.name, b->bpf.name);
865 }
866 
867 struct seg6_action_param {
868 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
869 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
870 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
871 };
872 
873 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
874 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
875 				    .put = put_nla_srh,
876 				    .cmp = cmp_nla_srh },
877 
878 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
879 				    .put = put_nla_table,
880 				    .cmp = cmp_nla_table },
881 
882 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
883 				    .put = put_nla_nh4,
884 				    .cmp = cmp_nla_nh4 },
885 
886 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
887 				    .put = put_nla_nh6,
888 				    .cmp = cmp_nla_nh6 },
889 
890 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
891 				    .put = put_nla_iif,
892 				    .cmp = cmp_nla_iif },
893 
894 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
895 				    .put = put_nla_oif,
896 				    .cmp = cmp_nla_oif },
897 
898 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
899 				    .put = put_nla_bpf,
900 				    .cmp = cmp_nla_bpf },
901 
902 };
903 
904 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
905 {
906 	struct seg6_action_param *param;
907 	struct seg6_action_desc *desc;
908 	int i, err;
909 
910 	desc = __get_action_desc(slwt->action);
911 	if (!desc)
912 		return -EINVAL;
913 
914 	if (!desc->input)
915 		return -EOPNOTSUPP;
916 
917 	slwt->desc = desc;
918 	slwt->headroom += desc->static_headroom;
919 
920 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
921 		if (desc->attrs & (1 << i)) {
922 			if (!attrs[i])
923 				return -EINVAL;
924 
925 			param = &seg6_action_params[i];
926 
927 			err = param->parse(attrs, slwt);
928 			if (err < 0)
929 				return err;
930 		}
931 	}
932 
933 	return 0;
934 }
935 
936 static int seg6_local_build_state(struct nlattr *nla, unsigned int family,
937 				  const void *cfg, struct lwtunnel_state **ts,
938 				  struct netlink_ext_ack *extack)
939 {
940 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
941 	struct lwtunnel_state *newts;
942 	struct seg6_local_lwt *slwt;
943 	int err;
944 
945 	if (family != AF_INET6)
946 		return -EINVAL;
947 
948 	err = nla_parse_nested(tb, SEG6_LOCAL_MAX, nla, seg6_local_policy,
949 			       extack);
950 
951 	if (err < 0)
952 		return err;
953 
954 	if (!tb[SEG6_LOCAL_ACTION])
955 		return -EINVAL;
956 
957 	newts = lwtunnel_state_alloc(sizeof(*slwt));
958 	if (!newts)
959 		return -ENOMEM;
960 
961 	slwt = seg6_local_lwtunnel(newts);
962 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
963 
964 	err = parse_nla_action(tb, slwt);
965 	if (err < 0)
966 		goto out_free;
967 
968 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
969 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
970 	newts->headroom = slwt->headroom;
971 
972 	*ts = newts;
973 
974 	return 0;
975 
976 out_free:
977 	kfree(slwt->srh);
978 	kfree(newts);
979 	return err;
980 }
981 
982 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
983 {
984 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
985 
986 	kfree(slwt->srh);
987 
988 	if (slwt->desc->attrs & (1 << SEG6_LOCAL_BPF)) {
989 		kfree(slwt->bpf.name);
990 		bpf_prog_put(slwt->bpf.prog);
991 	}
992 
993 	return;
994 }
995 
996 static int seg6_local_fill_encap(struct sk_buff *skb,
997 				 struct lwtunnel_state *lwt)
998 {
999 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1000 	struct seg6_action_param *param;
1001 	int i, err;
1002 
1003 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1004 		return -EMSGSIZE;
1005 
1006 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1007 		if (slwt->desc->attrs & (1 << i)) {
1008 			param = &seg6_action_params[i];
1009 			err = param->put(skb, slwt);
1010 			if (err < 0)
1011 				return err;
1012 		}
1013 	}
1014 
1015 	return 0;
1016 }
1017 
1018 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1019 {
1020 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1021 	unsigned long attrs;
1022 	int nlsize;
1023 
1024 	nlsize = nla_total_size(4); /* action */
1025 
1026 	attrs = slwt->desc->attrs;
1027 
1028 	if (attrs & (1 << SEG6_LOCAL_SRH))
1029 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1030 
1031 	if (attrs & (1 << SEG6_LOCAL_TABLE))
1032 		nlsize += nla_total_size(4);
1033 
1034 	if (attrs & (1 << SEG6_LOCAL_NH4))
1035 		nlsize += nla_total_size(4);
1036 
1037 	if (attrs & (1 << SEG6_LOCAL_NH6))
1038 		nlsize += nla_total_size(16);
1039 
1040 	if (attrs & (1 << SEG6_LOCAL_IIF))
1041 		nlsize += nla_total_size(4);
1042 
1043 	if (attrs & (1 << SEG6_LOCAL_OIF))
1044 		nlsize += nla_total_size(4);
1045 
1046 	if (attrs & (1 << SEG6_LOCAL_BPF))
1047 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1048 		       nla_total_size(MAX_PROG_NAME) +
1049 		       nla_total_size(4);
1050 
1051 	return nlsize;
1052 }
1053 
1054 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1055 				struct lwtunnel_state *b)
1056 {
1057 	struct seg6_local_lwt *slwt_a, *slwt_b;
1058 	struct seg6_action_param *param;
1059 	int i;
1060 
1061 	slwt_a = seg6_local_lwtunnel(a);
1062 	slwt_b = seg6_local_lwtunnel(b);
1063 
1064 	if (slwt_a->action != slwt_b->action)
1065 		return 1;
1066 
1067 	if (slwt_a->desc->attrs != slwt_b->desc->attrs)
1068 		return 1;
1069 
1070 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1071 		if (slwt_a->desc->attrs & (1 << i)) {
1072 			param = &seg6_action_params[i];
1073 			if (param->cmp(slwt_a, slwt_b))
1074 				return 1;
1075 		}
1076 	}
1077 
1078 	return 0;
1079 }
1080 
1081 static const struct lwtunnel_encap_ops seg6_local_ops = {
1082 	.build_state	= seg6_local_build_state,
1083 	.destroy_state	= seg6_local_destroy_state,
1084 	.input		= seg6_local_input,
1085 	.fill_encap	= seg6_local_fill_encap,
1086 	.get_encap_size	= seg6_local_get_encap_size,
1087 	.cmp_encap	= seg6_local_cmp_encap,
1088 	.owner		= THIS_MODULE,
1089 };
1090 
1091 int __init seg6_local_init(void)
1092 {
1093 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1094 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1095 }
1096 
1097 void seg6_local_exit(void)
1098 {
1099 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1100 }
1101