xref: /openbmc/linux/net/ipv6/seg6_local.c (revision c1d3fb8a) 
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 static int
153 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
154 			u32 tbl_id, bool local_delivery)
155 {
156 	struct net *net = dev_net(skb->dev);
157 	struct ipv6hdr *hdr = ipv6_hdr(skb);
158 	int flags = RT6_LOOKUP_F_HAS_SADDR;
159 	struct dst_entry *dst = NULL;
160 	struct rt6_info *rt;
161 	struct flowi6 fl6;
162 	int dev_flags = 0;
163 
164 	fl6.flowi6_iif = skb->dev->ifindex;
165 	fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
166 	fl6.saddr = hdr->saddr;
167 	fl6.flowlabel = ip6_flowinfo(hdr);
168 	fl6.flowi6_mark = skb->mark;
169 	fl6.flowi6_proto = hdr->nexthdr;
170 
171 	if (nhaddr)
172 		fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
173 
174 	if (!tbl_id) {
175 		dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
176 	} else {
177 		struct fib6_table *table;
178 
179 		table = fib6_get_table(net, tbl_id);
180 		if (!table)
181 			goto out;
182 
183 		rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
184 		dst = &rt->dst;
185 	}
186 
187 	/* we want to discard traffic destined for local packet processing,
188 	 * if @local_delivery is set to false.
189 	 */
190 	if (!local_delivery)
191 		dev_flags |= IFF_LOOPBACK;
192 
193 	if (dst && (dst->dev->flags & dev_flags) && !dst->error) {
194 		dst_release(dst);
195 		dst = NULL;
196 	}
197 
198 out:
199 	if (!dst) {
200 		rt = net->ipv6.ip6_blk_hole_entry;
201 		dst = &rt->dst;
202 		dst_hold(dst);
203 	}
204 
205 	skb_dst_drop(skb);
206 	skb_dst_set(skb, dst);
207 	return dst->error;
208 }
209 
210 int seg6_lookup_nexthop(struct sk_buff *skb,
211 			struct in6_addr *nhaddr, u32 tbl_id)
212 {
213 	return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false);
214 }
215 
216 /* regular endpoint function */
217 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
218 {
219 	struct ipv6_sr_hdr *srh;
220 
221 	srh = get_and_validate_srh(skb);
222 	if (!srh)
223 		goto drop;
224 
225 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
226 
227 	seg6_lookup_nexthop(skb, NULL, 0);
228 
229 	return dst_input(skb);
230 
231 drop:
232 	kfree_skb(skb);
233 	return -EINVAL;
234 }
235 
236 /* regular endpoint, and forward to specified nexthop */
237 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
238 {
239 	struct ipv6_sr_hdr *srh;
240 
241 	srh = get_and_validate_srh(skb);
242 	if (!srh)
243 		goto drop;
244 
245 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
246 
247 	seg6_lookup_nexthop(skb, &slwt->nh6, 0);
248 
249 	return dst_input(skb);
250 
251 drop:
252 	kfree_skb(skb);
253 	return -EINVAL;
254 }
255 
256 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
257 {
258 	struct ipv6_sr_hdr *srh;
259 
260 	srh = get_and_validate_srh(skb);
261 	if (!srh)
262 		goto drop;
263 
264 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
265 
266 	seg6_lookup_nexthop(skb, NULL, slwt->table);
267 
268 	return dst_input(skb);
269 
270 drop:
271 	kfree_skb(skb);
272 	return -EINVAL;
273 }
274 
275 /* decapsulate and forward inner L2 frame on specified interface */
276 static int input_action_end_dx2(struct sk_buff *skb,
277 				struct seg6_local_lwt *slwt)
278 {
279 	struct net *net = dev_net(skb->dev);
280 	struct net_device *odev;
281 	struct ethhdr *eth;
282 
283 	if (!decap_and_validate(skb, NEXTHDR_NONE))
284 		goto drop;
285 
286 	if (!pskb_may_pull(skb, ETH_HLEN))
287 		goto drop;
288 
289 	skb_reset_mac_header(skb);
290 	eth = (struct ethhdr *)skb->data;
291 
292 	/* To determine the frame's protocol, we assume it is 802.3. This avoids
293 	 * a call to eth_type_trans(), which is not really relevant for our
294 	 * use case.
295 	 */
296 	if (!eth_proto_is_802_3(eth->h_proto))
297 		goto drop;
298 
299 	odev = dev_get_by_index_rcu(net, slwt->oif);
300 	if (!odev)
301 		goto drop;
302 
303 	/* As we accept Ethernet frames, make sure the egress device is of
304 	 * the correct type.
305 	 */
306 	if (odev->type != ARPHRD_ETHER)
307 		goto drop;
308 
309 	if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
310 		goto drop;
311 
312 	skb_orphan(skb);
313 
314 	if (skb_warn_if_lro(skb))
315 		goto drop;
316 
317 	skb_forward_csum(skb);
318 
319 	if (skb->len - ETH_HLEN > odev->mtu)
320 		goto drop;
321 
322 	skb->dev = odev;
323 	skb->protocol = eth->h_proto;
324 
325 	return dev_queue_xmit(skb);
326 
327 drop:
328 	kfree_skb(skb);
329 	return -EINVAL;
330 }
331 
332 /* decapsulate and forward to specified nexthop */
333 static int input_action_end_dx6(struct sk_buff *skb,
334 				struct seg6_local_lwt *slwt)
335 {
336 	struct in6_addr *nhaddr = NULL;
337 
338 	/* this function accepts IPv6 encapsulated packets, with either
339 	 * an SRH with SL=0, or no SRH.
340 	 */
341 
342 	if (!decap_and_validate(skb, IPPROTO_IPV6))
343 		goto drop;
344 
345 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
346 		goto drop;
347 
348 	/* The inner packet is not associated to any local interface,
349 	 * so we do not call netif_rx().
350 	 *
351 	 * If slwt->nh6 is set to ::, then lookup the nexthop for the
352 	 * inner packet's DA. Otherwise, use the specified nexthop.
353 	 */
354 
355 	if (!ipv6_addr_any(&slwt->nh6))
356 		nhaddr = &slwt->nh6;
357 
358 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
359 
360 	seg6_lookup_nexthop(skb, nhaddr, 0);
361 
362 	return dst_input(skb);
363 drop:
364 	kfree_skb(skb);
365 	return -EINVAL;
366 }
367 
368 static int input_action_end_dx4(struct sk_buff *skb,
369 				struct seg6_local_lwt *slwt)
370 {
371 	struct iphdr *iph;
372 	__be32 nhaddr;
373 	int err;
374 
375 	if (!decap_and_validate(skb, IPPROTO_IPIP))
376 		goto drop;
377 
378 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
379 		goto drop;
380 
381 	skb->protocol = htons(ETH_P_IP);
382 
383 	iph = ip_hdr(skb);
384 
385 	nhaddr = slwt->nh4.s_addr ?: iph->daddr;
386 
387 	skb_dst_drop(skb);
388 
389 	skb_set_transport_header(skb, sizeof(struct iphdr));
390 
391 	err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
392 	if (err)
393 		goto drop;
394 
395 	return dst_input(skb);
396 
397 drop:
398 	kfree_skb(skb);
399 	return -EINVAL;
400 }
401 
402 static int input_action_end_dt6(struct sk_buff *skb,
403 				struct seg6_local_lwt *slwt)
404 {
405 	if (!decap_and_validate(skb, IPPROTO_IPV6))
406 		goto drop;
407 
408 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
409 		goto drop;
410 
411 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
412 
413 	seg6_lookup_any_nexthop(skb, NULL, slwt->table, true);
414 
415 	return dst_input(skb);
416 
417 drop:
418 	kfree_skb(skb);
419 	return -EINVAL;
420 }
421 
422 /* push an SRH on top of the current one */
423 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
424 {
425 	struct ipv6_sr_hdr *srh;
426 	int err = -EINVAL;
427 
428 	srh = get_and_validate_srh(skb);
429 	if (!srh)
430 		goto drop;
431 
432 	err = seg6_do_srh_inline(skb, slwt->srh);
433 	if (err)
434 		goto drop;
435 
436 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
437 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
438 
439 	seg6_lookup_nexthop(skb, NULL, 0);
440 
441 	return dst_input(skb);
442 
443 drop:
444 	kfree_skb(skb);
445 	return err;
446 }
447 
448 /* encapsulate within an outer IPv6 header and a specified SRH */
449 static int input_action_end_b6_encap(struct sk_buff *skb,
450 				     struct seg6_local_lwt *slwt)
451 {
452 	struct ipv6_sr_hdr *srh;
453 	int err = -EINVAL;
454 
455 	srh = get_and_validate_srh(skb);
456 	if (!srh)
457 		goto drop;
458 
459 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
460 
461 	skb_reset_inner_headers(skb);
462 	skb->encapsulation = 1;
463 
464 	err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
465 	if (err)
466 		goto drop;
467 
468 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
469 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
470 
471 	seg6_lookup_nexthop(skb, NULL, 0);
472 
473 	return dst_input(skb);
474 
475 drop:
476 	kfree_skb(skb);
477 	return err;
478 }
479 
480 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
481 
482 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
483 {
484 	struct seg6_bpf_srh_state *srh_state =
485 		this_cpu_ptr(&seg6_bpf_srh_states);
486 	struct ipv6_sr_hdr *srh = srh_state->srh;
487 
488 	if (unlikely(srh == NULL))
489 		return false;
490 
491 	if (unlikely(!srh_state->valid)) {
492 		if ((srh_state->hdrlen & 7) != 0)
493 			return false;
494 
495 		srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
496 		if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3))
497 			return false;
498 
499 		srh_state->valid = true;
500 	}
501 
502 	return true;
503 }
504 
505 static int input_action_end_bpf(struct sk_buff *skb,
506 				struct seg6_local_lwt *slwt)
507 {
508 	struct seg6_bpf_srh_state *srh_state =
509 		this_cpu_ptr(&seg6_bpf_srh_states);
510 	struct ipv6_sr_hdr *srh;
511 	int ret;
512 
513 	srh = get_and_validate_srh(skb);
514 	if (!srh) {
515 		kfree_skb(skb);
516 		return -EINVAL;
517 	}
518 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
519 
520 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
521 	 * which is also accessed by the bpf_lwt_seg6_* helpers
522 	 */
523 	preempt_disable();
524 	srh_state->srh = srh;
525 	srh_state->hdrlen = srh->hdrlen << 3;
526 	srh_state->valid = true;
527 
528 	rcu_read_lock();
529 	bpf_compute_data_pointers(skb);
530 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
531 	rcu_read_unlock();
532 
533 	switch (ret) {
534 	case BPF_OK:
535 	case BPF_REDIRECT:
536 		break;
537 	case BPF_DROP:
538 		goto drop;
539 	default:
540 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
541 		goto drop;
542 	}
543 
544 	if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
545 		goto drop;
546 
547 	preempt_enable();
548 	if (ret != BPF_REDIRECT)
549 		seg6_lookup_nexthop(skb, NULL, 0);
550 
551 	return dst_input(skb);
552 
553 drop:
554 	preempt_enable();
555 	kfree_skb(skb);
556 	return -EINVAL;
557 }
558 
559 static struct seg6_action_desc seg6_action_table[] = {
560 	{
561 		.action		= SEG6_LOCAL_ACTION_END,
562 		.attrs		= 0,
563 		.input		= input_action_end,
564 	},
565 	{
566 		.action		= SEG6_LOCAL_ACTION_END_X,
567 		.attrs		= (1 << SEG6_LOCAL_NH6),
568 		.input		= input_action_end_x,
569 	},
570 	{
571 		.action		= SEG6_LOCAL_ACTION_END_T,
572 		.attrs		= (1 << SEG6_LOCAL_TABLE),
573 		.input		= input_action_end_t,
574 	},
575 	{
576 		.action		= SEG6_LOCAL_ACTION_END_DX2,
577 		.attrs		= (1 << SEG6_LOCAL_OIF),
578 		.input		= input_action_end_dx2,
579 	},
580 	{
581 		.action		= SEG6_LOCAL_ACTION_END_DX6,
582 		.attrs		= (1 << SEG6_LOCAL_NH6),
583 		.input		= input_action_end_dx6,
584 	},
585 	{
586 		.action		= SEG6_LOCAL_ACTION_END_DX4,
587 		.attrs		= (1 << SEG6_LOCAL_NH4),
588 		.input		= input_action_end_dx4,
589 	},
590 	{
591 		.action		= SEG6_LOCAL_ACTION_END_DT6,
592 		.attrs		= (1 << SEG6_LOCAL_TABLE),
593 		.input		= input_action_end_dt6,
594 	},
595 	{
596 		.action		= SEG6_LOCAL_ACTION_END_B6,
597 		.attrs		= (1 << SEG6_LOCAL_SRH),
598 		.input		= input_action_end_b6,
599 	},
600 	{
601 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
602 		.attrs		= (1 << SEG6_LOCAL_SRH),
603 		.input		= input_action_end_b6_encap,
604 		.static_headroom	= sizeof(struct ipv6hdr),
605 	},
606 	{
607 		.action		= SEG6_LOCAL_ACTION_END_BPF,
608 		.attrs		= (1 << SEG6_LOCAL_BPF),
609 		.input		= input_action_end_bpf,
610 	},
611 
612 };
613 
614 static struct seg6_action_desc *__get_action_desc(int action)
615 {
616 	struct seg6_action_desc *desc;
617 	int i, count;
618 
619 	count = ARRAY_SIZE(seg6_action_table);
620 	for (i = 0; i < count; i++) {
621 		desc = &seg6_action_table[i];
622 		if (desc->action == action)
623 			return desc;
624 	}
625 
626 	return NULL;
627 }
628 
629 static int seg6_local_input(struct sk_buff *skb)
630 {
631 	struct dst_entry *orig_dst = skb_dst(skb);
632 	struct seg6_action_desc *desc;
633 	struct seg6_local_lwt *slwt;
634 
635 	if (skb->protocol != htons(ETH_P_IPV6)) {
636 		kfree_skb(skb);
637 		return -EINVAL;
638 	}
639 
640 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
641 	desc = slwt->desc;
642 
643 	return desc->input(skb, slwt);
644 }
645 
646 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
647 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
648 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
649 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
650 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
651 				    .len = sizeof(struct in_addr) },
652 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
653 				    .len = sizeof(struct in6_addr) },
654 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
655 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
656 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
657 };
658 
659 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
660 {
661 	struct ipv6_sr_hdr *srh;
662 	int len;
663 
664 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
665 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
666 
667 	/* SRH must contain at least one segment */
668 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
669 		return -EINVAL;
670 
671 	if (!seg6_validate_srh(srh, len))
672 		return -EINVAL;
673 
674 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
675 	if (!slwt->srh)
676 		return -ENOMEM;
677 
678 	slwt->headroom += len;
679 
680 	return 0;
681 }
682 
683 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
684 {
685 	struct ipv6_sr_hdr *srh;
686 	struct nlattr *nla;
687 	int len;
688 
689 	srh = slwt->srh;
690 	len = (srh->hdrlen + 1) << 3;
691 
692 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
693 	if (!nla)
694 		return -EMSGSIZE;
695 
696 	memcpy(nla_data(nla), srh, len);
697 
698 	return 0;
699 }
700 
701 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
702 {
703 	int len = (a->srh->hdrlen + 1) << 3;
704 
705 	if (len != ((b->srh->hdrlen + 1) << 3))
706 		return 1;
707 
708 	return memcmp(a->srh, b->srh, len);
709 }
710 
711 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
712 {
713 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
714 
715 	return 0;
716 }
717 
718 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
719 {
720 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
721 		return -EMSGSIZE;
722 
723 	return 0;
724 }
725 
726 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
727 {
728 	if (a->table != b->table)
729 		return 1;
730 
731 	return 0;
732 }
733 
734 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
735 {
736 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
737 	       sizeof(struct in_addr));
738 
739 	return 0;
740 }
741 
742 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
743 {
744 	struct nlattr *nla;
745 
746 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
747 	if (!nla)
748 		return -EMSGSIZE;
749 
750 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
751 
752 	return 0;
753 }
754 
755 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
756 {
757 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
758 }
759 
760 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
761 {
762 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
763 	       sizeof(struct in6_addr));
764 
765 	return 0;
766 }
767 
768 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
769 {
770 	struct nlattr *nla;
771 
772 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
773 	if (!nla)
774 		return -EMSGSIZE;
775 
776 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
777 
778 	return 0;
779 }
780 
781 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
782 {
783 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
784 }
785 
786 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
787 {
788 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
789 
790 	return 0;
791 }
792 
793 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
794 {
795 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
796 		return -EMSGSIZE;
797 
798 	return 0;
799 }
800 
801 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
802 {
803 	if (a->iif != b->iif)
804 		return 1;
805 
806 	return 0;
807 }
808 
809 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
810 {
811 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
812 
813 	return 0;
814 }
815 
816 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
817 {
818 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
819 		return -EMSGSIZE;
820 
821 	return 0;
822 }
823 
824 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
825 {
826 	if (a->oif != b->oif)
827 		return 1;
828 
829 	return 0;
830 }
831 
832 #define MAX_PROG_NAME 256
833 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
834 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
835 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
836 				       .len = MAX_PROG_NAME },
837 };
838 
839 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
840 {
841 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
842 	struct bpf_prog *p;
843 	int ret;
844 	u32 fd;
845 
846 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
847 					  attrs[SEG6_LOCAL_BPF],
848 					  bpf_prog_policy, NULL);
849 	if (ret < 0)
850 		return ret;
851 
852 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
853 		return -EINVAL;
854 
855 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
856 	if (!slwt->bpf.name)
857 		return -ENOMEM;
858 
859 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
860 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
861 	if (IS_ERR(p)) {
862 		kfree(slwt->bpf.name);
863 		return PTR_ERR(p);
864 	}
865 
866 	slwt->bpf.prog = p;
867 	return 0;
868 }
869 
870 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
871 {
872 	struct nlattr *nest;
873 
874 	if (!slwt->bpf.prog)
875 		return 0;
876 
877 	nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
878 	if (!nest)
879 		return -EMSGSIZE;
880 
881 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
882 		return -EMSGSIZE;
883 
884 	if (slwt->bpf.name &&
885 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
886 		return -EMSGSIZE;
887 
888 	return nla_nest_end(skb, nest);
889 }
890 
891 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
892 {
893 	if (!a->bpf.name && !b->bpf.name)
894 		return 0;
895 
896 	if (!a->bpf.name || !b->bpf.name)
897 		return 1;
898 
899 	return strcmp(a->bpf.name, b->bpf.name);
900 }
901 
902 struct seg6_action_param {
903 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
904 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
905 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
906 };
907 
908 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
909 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
910 				    .put = put_nla_srh,
911 				    .cmp = cmp_nla_srh },
912 
913 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
914 				    .put = put_nla_table,
915 				    .cmp = cmp_nla_table },
916 
917 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
918 				    .put = put_nla_nh4,
919 				    .cmp = cmp_nla_nh4 },
920 
921 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
922 				    .put = put_nla_nh6,
923 				    .cmp = cmp_nla_nh6 },
924 
925 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
926 				    .put = put_nla_iif,
927 				    .cmp = cmp_nla_iif },
928 
929 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
930 				    .put = put_nla_oif,
931 				    .cmp = cmp_nla_oif },
932 
933 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
934 				    .put = put_nla_bpf,
935 				    .cmp = cmp_nla_bpf },
936 
937 };
938 
939 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
940 {
941 	struct seg6_action_param *param;
942 	struct seg6_action_desc *desc;
943 	int i, err;
944 
945 	desc = __get_action_desc(slwt->action);
946 	if (!desc)
947 		return -EINVAL;
948 
949 	if (!desc->input)
950 		return -EOPNOTSUPP;
951 
952 	slwt->desc = desc;
953 	slwt->headroom += desc->static_headroom;
954 
955 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
956 		if (desc->attrs & (1 << i)) {
957 			if (!attrs[i])
958 				return -EINVAL;
959 
960 			param = &seg6_action_params[i];
961 
962 			err = param->parse(attrs, slwt);
963 			if (err < 0)
964 				return err;
965 		}
966 	}
967 
968 	return 0;
969 }
970 
971 static int seg6_local_build_state(struct nlattr *nla, unsigned int family,
972 				  const void *cfg, struct lwtunnel_state **ts,
973 				  struct netlink_ext_ack *extack)
974 {
975 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
976 	struct lwtunnel_state *newts;
977 	struct seg6_local_lwt *slwt;
978 	int err;
979 
980 	if (family != AF_INET6)
981 		return -EINVAL;
982 
983 	err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
984 					  seg6_local_policy, extack);
985 
986 	if (err < 0)
987 		return err;
988 
989 	if (!tb[SEG6_LOCAL_ACTION])
990 		return -EINVAL;
991 
992 	newts = lwtunnel_state_alloc(sizeof(*slwt));
993 	if (!newts)
994 		return -ENOMEM;
995 
996 	slwt = seg6_local_lwtunnel(newts);
997 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
998 
999 	err = parse_nla_action(tb, slwt);
1000 	if (err < 0)
1001 		goto out_free;
1002 
1003 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
1004 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
1005 	newts->headroom = slwt->headroom;
1006 
1007 	*ts = newts;
1008 
1009 	return 0;
1010 
1011 out_free:
1012 	kfree(slwt->srh);
1013 	kfree(newts);
1014 	return err;
1015 }
1016 
1017 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
1018 {
1019 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1020 
1021 	kfree(slwt->srh);
1022 
1023 	if (slwt->desc->attrs & (1 << SEG6_LOCAL_BPF)) {
1024 		kfree(slwt->bpf.name);
1025 		bpf_prog_put(slwt->bpf.prog);
1026 	}
1027 
1028 	return;
1029 }
1030 
1031 static int seg6_local_fill_encap(struct sk_buff *skb,
1032 				 struct lwtunnel_state *lwt)
1033 {
1034 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1035 	struct seg6_action_param *param;
1036 	int i, err;
1037 
1038 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1039 		return -EMSGSIZE;
1040 
1041 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1042 		if (slwt->desc->attrs & (1 << i)) {
1043 			param = &seg6_action_params[i];
1044 			err = param->put(skb, slwt);
1045 			if (err < 0)
1046 				return err;
1047 		}
1048 	}
1049 
1050 	return 0;
1051 }
1052 
1053 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1054 {
1055 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1056 	unsigned long attrs;
1057 	int nlsize;
1058 
1059 	nlsize = nla_total_size(4); /* action */
1060 
1061 	attrs = slwt->desc->attrs;
1062 
1063 	if (attrs & (1 << SEG6_LOCAL_SRH))
1064 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1065 
1066 	if (attrs & (1 << SEG6_LOCAL_TABLE))
1067 		nlsize += nla_total_size(4);
1068 
1069 	if (attrs & (1 << SEG6_LOCAL_NH4))
1070 		nlsize += nla_total_size(4);
1071 
1072 	if (attrs & (1 << SEG6_LOCAL_NH6))
1073 		nlsize += nla_total_size(16);
1074 
1075 	if (attrs & (1 << SEG6_LOCAL_IIF))
1076 		nlsize += nla_total_size(4);
1077 
1078 	if (attrs & (1 << SEG6_LOCAL_OIF))
1079 		nlsize += nla_total_size(4);
1080 
1081 	if (attrs & (1 << SEG6_LOCAL_BPF))
1082 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1083 		       nla_total_size(MAX_PROG_NAME) +
1084 		       nla_total_size(4);
1085 
1086 	return nlsize;
1087 }
1088 
1089 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1090 				struct lwtunnel_state *b)
1091 {
1092 	struct seg6_local_lwt *slwt_a, *slwt_b;
1093 	struct seg6_action_param *param;
1094 	int i;
1095 
1096 	slwt_a = seg6_local_lwtunnel(a);
1097 	slwt_b = seg6_local_lwtunnel(b);
1098 
1099 	if (slwt_a->action != slwt_b->action)
1100 		return 1;
1101 
1102 	if (slwt_a->desc->attrs != slwt_b->desc->attrs)
1103 		return 1;
1104 
1105 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1106 		if (slwt_a->desc->attrs & (1 << i)) {
1107 			param = &seg6_action_params[i];
1108 			if (param->cmp(slwt_a, slwt_b))
1109 				return 1;
1110 		}
1111 	}
1112 
1113 	return 0;
1114 }
1115 
1116 static const struct lwtunnel_encap_ops seg6_local_ops = {
1117 	.build_state	= seg6_local_build_state,
1118 	.destroy_state	= seg6_local_destroy_state,
1119 	.input		= seg6_local_input,
1120 	.fill_encap	= seg6_local_fill_encap,
1121 	.get_encap_size	= seg6_local_get_encap_size,
1122 	.cmp_encap	= seg6_local_cmp_encap,
1123 	.owner		= THIS_MODULE,
1124 };
1125 
1126 int __init seg6_local_init(void)
1127 {
1128 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1129 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1130 }
1131 
1132 void seg6_local_exit(void)
1133 {
1134 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1135 }
1136