xref: /linux/net/ipv6/ip6_input.c (revision 8ca4fc323d2e4ab9dabbdd57633af40b0c7e6af9)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 input
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Ian P. Morris		<I.P.Morris@soton.ac.uk>
9  *
10  *	Based in linux/net/ipv4/ip_input.c
11  */
12 /* Changes
13  *
14  *	Mitsuru KANDA @USAGI and
15  *	YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs().
16  */
17 
18 #include <linux/errno.h>
19 #include <linux/types.h>
20 #include <linux/socket.h>
21 #include <linux/sockios.h>
22 #include <linux/net.h>
23 #include <linux/netdevice.h>
24 #include <linux/in6.h>
25 #include <linux/icmpv6.h>
26 #include <linux/mroute6.h>
27 #include <linux/slab.h>
28 #include <linux/indirect_call_wrapper.h>
29 
30 #include <linux/netfilter.h>
31 #include <linux/netfilter_ipv6.h>
32 
33 #include <net/sock.h>
34 #include <net/snmp.h>
35 #include <net/udp.h>
36 
37 #include <net/ipv6.h>
38 #include <net/protocol.h>
39 #include <net/transp_v6.h>
40 #include <net/rawv6.h>
41 #include <net/ndisc.h>
42 #include <net/ip6_route.h>
43 #include <net/addrconf.h>
44 #include <net/xfrm.h>
45 #include <net/inet_ecn.h>
46 #include <net/dst_metadata.h>
47 
48 static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
49 				struct sk_buff *skb)
50 {
51 	if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) &&
52 	    !skb_dst(skb) && !skb->sk) {
53 		switch (ipv6_hdr(skb)->nexthdr) {
54 		case IPPROTO_TCP:
55 			if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux))
56 				tcp_v6_early_demux(skb);
57 			break;
58 		case IPPROTO_UDP:
59 			if (READ_ONCE(net->ipv4.sysctl_udp_early_demux))
60 				udp_v6_early_demux(skb);
61 			break;
62 		}
63 	}
64 
65 	if (!skb_valid_dst(skb))
66 		ip6_route_input(skb);
67 }
68 
69 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
70 {
71 	/* if ingress device is enslaved to an L3 master device pass the
72 	 * skb to its handler for processing
73 	 */
74 	skb = l3mdev_ip6_rcv(skb);
75 	if (!skb)
76 		return NET_RX_SUCCESS;
77 	ip6_rcv_finish_core(net, sk, skb);
78 
79 	return dst_input(skb);
80 }
81 
82 static void ip6_sublist_rcv_finish(struct list_head *head)
83 {
84 	struct sk_buff *skb, *next;
85 
86 	list_for_each_entry_safe(skb, next, head, list) {
87 		skb_list_del_init(skb);
88 		dst_input(skb);
89 	}
90 }
91 
92 static bool ip6_can_use_hint(const struct sk_buff *skb,
93 			     const struct sk_buff *hint)
94 {
95 	return hint && !skb_dst(skb) &&
96 	       ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
97 }
98 
99 static struct sk_buff *ip6_extract_route_hint(const struct net *net,
100 					      struct sk_buff *skb)
101 {
102 	if (fib6_routes_require_src(net) || fib6_has_custom_rules(net))
103 		return NULL;
104 
105 	return skb;
106 }
107 
108 static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
109 				struct list_head *head)
110 {
111 	struct sk_buff *skb, *next, *hint = NULL;
112 	struct dst_entry *curr_dst = NULL;
113 	struct list_head sublist;
114 
115 	INIT_LIST_HEAD(&sublist);
116 	list_for_each_entry_safe(skb, next, head, list) {
117 		struct dst_entry *dst;
118 
119 		skb_list_del_init(skb);
120 		/* if ingress device is enslaved to an L3 master device pass the
121 		 * skb to its handler for processing
122 		 */
123 		skb = l3mdev_ip6_rcv(skb);
124 		if (!skb)
125 			continue;
126 
127 		if (ip6_can_use_hint(skb, hint))
128 			skb_dst_copy(skb, hint);
129 		else
130 			ip6_rcv_finish_core(net, sk, skb);
131 		dst = skb_dst(skb);
132 		if (curr_dst != dst) {
133 			hint = ip6_extract_route_hint(net, skb);
134 
135 			/* dispatch old sublist */
136 			if (!list_empty(&sublist))
137 				ip6_sublist_rcv_finish(&sublist);
138 			/* start new sublist */
139 			INIT_LIST_HEAD(&sublist);
140 			curr_dst = dst;
141 		}
142 		list_add_tail(&skb->list, &sublist);
143 	}
144 	/* dispatch final sublist */
145 	ip6_sublist_rcv_finish(&sublist);
146 }
147 
148 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
149 				    struct net *net)
150 {
151 	enum skb_drop_reason reason;
152 	const struct ipv6hdr *hdr;
153 	u32 pkt_len;
154 	struct inet6_dev *idev;
155 
156 	if (skb->pkt_type == PACKET_OTHERHOST) {
157 		dev_core_stats_rx_otherhost_dropped_inc(skb->dev);
158 		kfree_skb_reason(skb, SKB_DROP_REASON_OTHERHOST);
159 		return NULL;
160 	}
161 
162 	rcu_read_lock();
163 
164 	idev = __in6_dev_get(skb->dev);
165 
166 	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
167 
168 	SKB_DR_SET(reason, NOT_SPECIFIED);
169 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
170 	    !idev || unlikely(idev->cnf.disable_ipv6)) {
171 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
172 		if (idev && unlikely(idev->cnf.disable_ipv6))
173 			SKB_DR_SET(reason, IPV6DISABLED);
174 		goto drop;
175 	}
176 
177 	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
178 
179 	/*
180 	 * Store incoming device index. When the packet will
181 	 * be queued, we cannot refer to skb->dev anymore.
182 	 *
183 	 * BTW, when we send a packet for our own local address on a
184 	 * non-loopback interface (e.g. ethX), it is being delivered
185 	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
186 	 * It, however, should be considered as if it is being
187 	 * arrived via the sending interface (ethX), because of the
188 	 * nature of scoping architecture. --yoshfuji
189 	 */
190 	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
191 
192 	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
193 		goto err;
194 
195 	hdr = ipv6_hdr(skb);
196 
197 	if (hdr->version != 6) {
198 		SKB_DR_SET(reason, UNHANDLED_PROTO);
199 		goto err;
200 	}
201 
202 	__IP6_ADD_STATS(net, idev,
203 			IPSTATS_MIB_NOECTPKTS +
204 				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
205 			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
206 	/*
207 	 * RFC4291 2.5.3
208 	 * The loopback address must not be used as the source address in IPv6
209 	 * packets that are sent outside of a single node. [..]
210 	 * A packet received on an interface with a destination address
211 	 * of loopback must be dropped.
212 	 */
213 	if ((ipv6_addr_loopback(&hdr->saddr) ||
214 	     ipv6_addr_loopback(&hdr->daddr)) &&
215 	    !(dev->flags & IFF_LOOPBACK) &&
216 	    !netif_is_l3_master(dev))
217 		goto err;
218 
219 	/* RFC4291 Errata ID: 3480
220 	 * Interface-Local scope spans only a single interface on a
221 	 * node and is useful only for loopback transmission of
222 	 * multicast.  Packets with interface-local scope received
223 	 * from another node must be discarded.
224 	 */
225 	if (!(skb->pkt_type == PACKET_LOOPBACK ||
226 	      dev->flags & IFF_LOOPBACK) &&
227 	    ipv6_addr_is_multicast(&hdr->daddr) &&
228 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
229 		goto err;
230 
231 	/* If enabled, drop unicast packets that were encapsulated in link-layer
232 	 * multicast or broadcast to protected against the so-called "hole-196"
233 	 * attack in 802.11 wireless.
234 	 */
235 	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
236 	    (skb->pkt_type == PACKET_BROADCAST ||
237 	     skb->pkt_type == PACKET_MULTICAST) &&
238 	    idev->cnf.drop_unicast_in_l2_multicast) {
239 		SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST);
240 		goto err;
241 	}
242 
243 	/* RFC4291 2.7
244 	 * Nodes must not originate a packet to a multicast address whose scope
245 	 * field contains the reserved value 0; if such a packet is received, it
246 	 * must be silently dropped.
247 	 */
248 	if (ipv6_addr_is_multicast(&hdr->daddr) &&
249 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
250 		goto err;
251 
252 	/*
253 	 * RFC4291 2.7
254 	 * Multicast addresses must not be used as source addresses in IPv6
255 	 * packets or appear in any Routing header.
256 	 */
257 	if (ipv6_addr_is_multicast(&hdr->saddr))
258 		goto err;
259 
260 	skb->transport_header = skb->network_header + sizeof(*hdr);
261 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
262 
263 	pkt_len = ntohs(hdr->payload_len);
264 
265 	/* pkt_len may be zero if Jumbo payload option is present */
266 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
267 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
268 			__IP6_INC_STATS(net,
269 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
270 			SKB_DR_SET(reason, PKT_TOO_SMALL);
271 			goto drop;
272 		}
273 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
274 			goto err;
275 		hdr = ipv6_hdr(skb);
276 	}
277 
278 	if (hdr->nexthdr == NEXTHDR_HOP) {
279 		if (ipv6_parse_hopopts(skb) < 0) {
280 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
281 			rcu_read_unlock();
282 			return NULL;
283 		}
284 	}
285 
286 	rcu_read_unlock();
287 
288 	/* Must drop socket now because of tproxy. */
289 	if (!skb_sk_is_prefetched(skb))
290 		skb_orphan(skb);
291 
292 	return skb;
293 err:
294 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
295 	SKB_DR_OR(reason, IP_INHDR);
296 drop:
297 	rcu_read_unlock();
298 	kfree_skb_reason(skb, reason);
299 	return NULL;
300 }
301 
302 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
303 {
304 	struct net *net = dev_net(skb->dev);
305 
306 	skb = ip6_rcv_core(skb, dev, net);
307 	if (skb == NULL)
308 		return NET_RX_DROP;
309 	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
310 		       net, NULL, skb, dev, NULL,
311 		       ip6_rcv_finish);
312 }
313 
314 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
315 			    struct net *net)
316 {
317 	NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
318 		     head, dev, NULL, ip6_rcv_finish);
319 	ip6_list_rcv_finish(net, NULL, head);
320 }
321 
322 /* Receive a list of IPv6 packets */
323 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
324 		   struct net_device *orig_dev)
325 {
326 	struct net_device *curr_dev = NULL;
327 	struct net *curr_net = NULL;
328 	struct sk_buff *skb, *next;
329 	struct list_head sublist;
330 
331 	INIT_LIST_HEAD(&sublist);
332 	list_for_each_entry_safe(skb, next, head, list) {
333 		struct net_device *dev = skb->dev;
334 		struct net *net = dev_net(dev);
335 
336 		skb_list_del_init(skb);
337 		skb = ip6_rcv_core(skb, dev, net);
338 		if (skb == NULL)
339 			continue;
340 
341 		if (curr_dev != dev || curr_net != net) {
342 			/* dispatch old sublist */
343 			if (!list_empty(&sublist))
344 				ip6_sublist_rcv(&sublist, curr_dev, curr_net);
345 			/* start new sublist */
346 			INIT_LIST_HEAD(&sublist);
347 			curr_dev = dev;
348 			curr_net = net;
349 		}
350 		list_add_tail(&skb->list, &sublist);
351 	}
352 	/* dispatch final sublist */
353 	if (!list_empty(&sublist))
354 		ip6_sublist_rcv(&sublist, curr_dev, curr_net);
355 }
356 
357 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
358 
359 /*
360  *	Deliver the packet to the host
361  */
362 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
363 			      bool have_final)
364 {
365 	const struct inet6_protocol *ipprot;
366 	struct inet6_dev *idev;
367 	unsigned int nhoff;
368 	SKB_DR(reason);
369 	bool raw;
370 
371 	/*
372 	 *	Parse extension headers
373 	 */
374 
375 resubmit:
376 	idev = ip6_dst_idev(skb_dst(skb));
377 	nhoff = IP6CB(skb)->nhoff;
378 	if (!have_final) {
379 		if (!pskb_pull(skb, skb_transport_offset(skb)))
380 			goto discard;
381 		nexthdr = skb_network_header(skb)[nhoff];
382 	}
383 
384 resubmit_final:
385 	raw = raw6_local_deliver(skb, nexthdr);
386 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
387 	if (ipprot) {
388 		int ret;
389 
390 		if (have_final) {
391 			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
392 				/* Once we've seen a final protocol don't
393 				 * allow encapsulation on any non-final
394 				 * ones. This allows foo in UDP encapsulation
395 				 * to work.
396 				 */
397 				goto discard;
398 			}
399 		} else if (ipprot->flags & INET6_PROTO_FINAL) {
400 			const struct ipv6hdr *hdr;
401 			int sdif = inet6_sdif(skb);
402 			struct net_device *dev;
403 
404 			/* Only do this once for first final protocol */
405 			have_final = true;
406 
407 			/* Free reference early: we don't need it any more,
408 			   and it may hold ip_conntrack module loaded
409 			   indefinitely. */
410 			nf_reset_ct(skb);
411 
412 			skb_postpull_rcsum(skb, skb_network_header(skb),
413 					   skb_network_header_len(skb));
414 			hdr = ipv6_hdr(skb);
415 
416 			/* skb->dev passed may be master dev for vrfs. */
417 			if (sdif) {
418 				dev = dev_get_by_index_rcu(net, sdif);
419 				if (!dev)
420 					goto discard;
421 			} else {
422 				dev = skb->dev;
423 			}
424 
425 			if (ipv6_addr_is_multicast(&hdr->daddr) &&
426 			    !ipv6_chk_mcast_addr(dev, &hdr->daddr,
427 						 &hdr->saddr) &&
428 			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) {
429 				SKB_DR_SET(reason, IP_INADDRERRORS);
430 				goto discard;
431 			}
432 		}
433 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
434 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
435 			SKB_DR_SET(reason, XFRM_POLICY);
436 			goto discard;
437 		}
438 
439 		ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
440 				      skb);
441 		if (ret > 0) {
442 			if (ipprot->flags & INET6_PROTO_FINAL) {
443 				/* Not an extension header, most likely UDP
444 				 * encapsulation. Use return value as nexthdr
445 				 * protocol not nhoff (which presumably is
446 				 * not set by handler).
447 				 */
448 				nexthdr = ret;
449 				goto resubmit_final;
450 			} else {
451 				goto resubmit;
452 			}
453 		} else if (ret == 0) {
454 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
455 		}
456 	} else {
457 		if (!raw) {
458 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
459 				__IP6_INC_STATS(net, idev,
460 						IPSTATS_MIB_INUNKNOWNPROTOS);
461 				icmpv6_send(skb, ICMPV6_PARAMPROB,
462 					    ICMPV6_UNK_NEXTHDR, nhoff);
463 				SKB_DR_SET(reason, IP_NOPROTO);
464 			} else {
465 				SKB_DR_SET(reason, XFRM_POLICY);
466 			}
467 			kfree_skb_reason(skb, reason);
468 		} else {
469 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
470 			consume_skb(skb);
471 		}
472 	}
473 	return;
474 
475 discard:
476 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
477 	kfree_skb_reason(skb, reason);
478 }
479 
480 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
481 {
482 	skb_clear_delivery_time(skb);
483 	rcu_read_lock();
484 	ip6_protocol_deliver_rcu(net, skb, 0, false);
485 	rcu_read_unlock();
486 
487 	return 0;
488 }
489 
490 
491 int ip6_input(struct sk_buff *skb)
492 {
493 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
494 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
495 		       ip6_input_finish);
496 }
497 EXPORT_SYMBOL_GPL(ip6_input);
498 
499 int ip6_mc_input(struct sk_buff *skb)
500 {
501 	int sdif = inet6_sdif(skb);
502 	const struct ipv6hdr *hdr;
503 	struct net_device *dev;
504 	bool deliver;
505 
506 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
507 			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
508 			 skb->len);
509 
510 	/* skb->dev passed may be master dev for vrfs. */
511 	if (sdif) {
512 		rcu_read_lock();
513 		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
514 		if (!dev) {
515 			rcu_read_unlock();
516 			kfree_skb(skb);
517 			return -ENODEV;
518 		}
519 	} else {
520 		dev = skb->dev;
521 	}
522 
523 	hdr = ipv6_hdr(skb);
524 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
525 	if (sdif)
526 		rcu_read_unlock();
527 
528 #ifdef CONFIG_IPV6_MROUTE
529 	/*
530 	 *      IPv6 multicast router mode is now supported ;)
531 	 */
532 	if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
533 	    !(ipv6_addr_type(&hdr->daddr) &
534 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
535 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
536 		/*
537 		 * Okay, we try to forward - split and duplicate
538 		 * packets.
539 		 */
540 		struct sk_buff *skb2;
541 		struct inet6_skb_parm *opt = IP6CB(skb);
542 
543 		/* Check for MLD */
544 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
545 			/* Check if this is a mld message */
546 			u8 nexthdr = hdr->nexthdr;
547 			__be16 frag_off;
548 			int offset;
549 
550 			/* Check if the value of Router Alert
551 			 * is for MLD (0x0000).
552 			 */
553 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
554 				deliver = false;
555 
556 				if (!ipv6_ext_hdr(nexthdr)) {
557 					/* BUG */
558 					goto out;
559 				}
560 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
561 							  &nexthdr, &frag_off);
562 				if (offset < 0)
563 					goto out;
564 
565 				if (ipv6_is_mld(skb, nexthdr, offset))
566 					deliver = true;
567 
568 				goto out;
569 			}
570 			/* unknown RA - process it normally */
571 		}
572 
573 		if (deliver)
574 			skb2 = skb_clone(skb, GFP_ATOMIC);
575 		else {
576 			skb2 = skb;
577 			skb = NULL;
578 		}
579 
580 		if (skb2) {
581 			ip6_mr_input(skb2);
582 		}
583 	}
584 out:
585 #endif
586 	if (likely(deliver))
587 		ip6_input(skb);
588 	else {
589 		/* discard */
590 		kfree_skb(skb);
591 	}
592 
593 	return 0;
594 }
595