xref: /openbmc/linux/net/ipv6/ip6_input.c (revision 48ca54e3)
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 INDIRECT_CALLABLE_DECLARE(void tcp_v6_early_demux(struct sk_buff *));
49 static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
50 				struct sk_buff *skb)
51 {
52 	void (*edemux)(struct sk_buff *skb);
53 
54 	if (net->ipv4.sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
55 		const struct inet6_protocol *ipprot;
56 
57 		ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
58 		if (ipprot && (edemux = READ_ONCE(ipprot->early_demux)))
59 			INDIRECT_CALL_2(edemux, tcp_v6_early_demux,
60 					udp_v6_early_demux, skb);
61 	}
62 	if (!skb_valid_dst(skb))
63 		ip6_route_input(skb);
64 }
65 
66 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
67 {
68 	/* if ingress device is enslaved to an L3 master device pass the
69 	 * skb to its handler for processing
70 	 */
71 	skb = l3mdev_ip6_rcv(skb);
72 	if (!skb)
73 		return NET_RX_SUCCESS;
74 	ip6_rcv_finish_core(net, sk, skb);
75 
76 	return dst_input(skb);
77 }
78 
79 static void ip6_sublist_rcv_finish(struct list_head *head)
80 {
81 	struct sk_buff *skb, *next;
82 
83 	list_for_each_entry_safe(skb, next, head, list) {
84 		skb_list_del_init(skb);
85 		dst_input(skb);
86 	}
87 }
88 
89 static bool ip6_can_use_hint(const struct sk_buff *skb,
90 			     const struct sk_buff *hint)
91 {
92 	return hint && !skb_dst(skb) &&
93 	       ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
94 }
95 
96 static struct sk_buff *ip6_extract_route_hint(const struct net *net,
97 					      struct sk_buff *skb)
98 {
99 	if (fib6_routes_require_src(net) || fib6_has_custom_rules(net))
100 		return NULL;
101 
102 	return skb;
103 }
104 
105 static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
106 				struct list_head *head)
107 {
108 	struct sk_buff *skb, *next, *hint = NULL;
109 	struct dst_entry *curr_dst = NULL;
110 	struct list_head sublist;
111 
112 	INIT_LIST_HEAD(&sublist);
113 	list_for_each_entry_safe(skb, next, head, list) {
114 		struct dst_entry *dst;
115 
116 		skb_list_del_init(skb);
117 		/* if ingress device is enslaved to an L3 master device pass the
118 		 * skb to its handler for processing
119 		 */
120 		skb = l3mdev_ip6_rcv(skb);
121 		if (!skb)
122 			continue;
123 
124 		if (ip6_can_use_hint(skb, hint))
125 			skb_dst_copy(skb, hint);
126 		else
127 			ip6_rcv_finish_core(net, sk, skb);
128 		dst = skb_dst(skb);
129 		if (curr_dst != dst) {
130 			hint = ip6_extract_route_hint(net, skb);
131 
132 			/* dispatch old sublist */
133 			if (!list_empty(&sublist))
134 				ip6_sublist_rcv_finish(&sublist);
135 			/* start new sublist */
136 			INIT_LIST_HEAD(&sublist);
137 			curr_dst = dst;
138 		}
139 		list_add_tail(&skb->list, &sublist);
140 	}
141 	/* dispatch final sublist */
142 	ip6_sublist_rcv_finish(&sublist);
143 }
144 
145 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
146 				    struct net *net)
147 {
148 	enum skb_drop_reason reason;
149 	const struct ipv6hdr *hdr;
150 	u32 pkt_len;
151 	struct inet6_dev *idev;
152 
153 	if (skb->pkt_type == PACKET_OTHERHOST) {
154 		dev_core_stats_rx_otherhost_dropped_inc(skb->dev);
155 		kfree_skb_reason(skb, SKB_DROP_REASON_OTHERHOST);
156 		return NULL;
157 	}
158 
159 	rcu_read_lock();
160 
161 	idev = __in6_dev_get(skb->dev);
162 
163 	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
164 
165 	SKB_DR_SET(reason, NOT_SPECIFIED);
166 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
167 	    !idev || unlikely(idev->cnf.disable_ipv6)) {
168 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
169 		if (idev && unlikely(idev->cnf.disable_ipv6))
170 			SKB_DR_SET(reason, IPV6DISABLED);
171 		goto drop;
172 	}
173 
174 	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
175 
176 	/*
177 	 * Store incoming device index. When the packet will
178 	 * be queued, we cannot refer to skb->dev anymore.
179 	 *
180 	 * BTW, when we send a packet for our own local address on a
181 	 * non-loopback interface (e.g. ethX), it is being delivered
182 	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
183 	 * It, however, should be considered as if it is being
184 	 * arrived via the sending interface (ethX), because of the
185 	 * nature of scoping architecture. --yoshfuji
186 	 */
187 	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
188 
189 	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
190 		goto err;
191 
192 	hdr = ipv6_hdr(skb);
193 
194 	if (hdr->version != 6) {
195 		SKB_DR_SET(reason, UNHANDLED_PROTO);
196 		goto err;
197 	}
198 
199 	__IP6_ADD_STATS(net, idev,
200 			IPSTATS_MIB_NOECTPKTS +
201 				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
202 			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
203 	/*
204 	 * RFC4291 2.5.3
205 	 * The loopback address must not be used as the source address in IPv6
206 	 * packets that are sent outside of a single node. [..]
207 	 * A packet received on an interface with a destination address
208 	 * of loopback must be dropped.
209 	 */
210 	if ((ipv6_addr_loopback(&hdr->saddr) ||
211 	     ipv6_addr_loopback(&hdr->daddr)) &&
212 	    !(dev->flags & IFF_LOOPBACK) &&
213 	    !netif_is_l3_master(dev))
214 		goto err;
215 
216 	/* RFC4291 Errata ID: 3480
217 	 * Interface-Local scope spans only a single interface on a
218 	 * node and is useful only for loopback transmission of
219 	 * multicast.  Packets with interface-local scope received
220 	 * from another node must be discarded.
221 	 */
222 	if (!(skb->pkt_type == PACKET_LOOPBACK ||
223 	      dev->flags & IFF_LOOPBACK) &&
224 	    ipv6_addr_is_multicast(&hdr->daddr) &&
225 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
226 		goto err;
227 
228 	/* If enabled, drop unicast packets that were encapsulated in link-layer
229 	 * multicast or broadcast to protected against the so-called "hole-196"
230 	 * attack in 802.11 wireless.
231 	 */
232 	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
233 	    (skb->pkt_type == PACKET_BROADCAST ||
234 	     skb->pkt_type == PACKET_MULTICAST) &&
235 	    idev->cnf.drop_unicast_in_l2_multicast) {
236 		SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST);
237 		goto err;
238 	}
239 
240 	/* RFC4291 2.7
241 	 * Nodes must not originate a packet to a multicast address whose scope
242 	 * field contains the reserved value 0; if such a packet is received, it
243 	 * must be silently dropped.
244 	 */
245 	if (ipv6_addr_is_multicast(&hdr->daddr) &&
246 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
247 		goto err;
248 
249 	/*
250 	 * RFC4291 2.7
251 	 * Multicast addresses must not be used as source addresses in IPv6
252 	 * packets or appear in any Routing header.
253 	 */
254 	if (ipv6_addr_is_multicast(&hdr->saddr))
255 		goto err;
256 
257 	skb->transport_header = skb->network_header + sizeof(*hdr);
258 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
259 
260 	pkt_len = ntohs(hdr->payload_len);
261 
262 	/* pkt_len may be zero if Jumbo payload option is present */
263 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
264 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
265 			__IP6_INC_STATS(net,
266 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
267 			SKB_DR_SET(reason, PKT_TOO_SMALL);
268 			goto drop;
269 		}
270 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
271 			goto err;
272 		hdr = ipv6_hdr(skb);
273 	}
274 
275 	if (hdr->nexthdr == NEXTHDR_HOP) {
276 		if (ipv6_parse_hopopts(skb) < 0) {
277 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
278 			rcu_read_unlock();
279 			return NULL;
280 		}
281 	}
282 
283 	rcu_read_unlock();
284 
285 	/* Must drop socket now because of tproxy. */
286 	if (!skb_sk_is_prefetched(skb))
287 		skb_orphan(skb);
288 
289 	return skb;
290 err:
291 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
292 	SKB_DR_OR(reason, IP_INHDR);
293 drop:
294 	rcu_read_unlock();
295 	kfree_skb_reason(skb, reason);
296 	return NULL;
297 }
298 
299 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
300 {
301 	struct net *net = dev_net(skb->dev);
302 
303 	skb = ip6_rcv_core(skb, dev, net);
304 	if (skb == NULL)
305 		return NET_RX_DROP;
306 	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
307 		       net, NULL, skb, dev, NULL,
308 		       ip6_rcv_finish);
309 }
310 
311 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
312 			    struct net *net)
313 {
314 	NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
315 		     head, dev, NULL, ip6_rcv_finish);
316 	ip6_list_rcv_finish(net, NULL, head);
317 }
318 
319 /* Receive a list of IPv6 packets */
320 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
321 		   struct net_device *orig_dev)
322 {
323 	struct net_device *curr_dev = NULL;
324 	struct net *curr_net = NULL;
325 	struct sk_buff *skb, *next;
326 	struct list_head sublist;
327 
328 	INIT_LIST_HEAD(&sublist);
329 	list_for_each_entry_safe(skb, next, head, list) {
330 		struct net_device *dev = skb->dev;
331 		struct net *net = dev_net(dev);
332 
333 		skb_list_del_init(skb);
334 		skb = ip6_rcv_core(skb, dev, net);
335 		if (skb == NULL)
336 			continue;
337 
338 		if (curr_dev != dev || curr_net != net) {
339 			/* dispatch old sublist */
340 			if (!list_empty(&sublist))
341 				ip6_sublist_rcv(&sublist, curr_dev, curr_net);
342 			/* start new sublist */
343 			INIT_LIST_HEAD(&sublist);
344 			curr_dev = dev;
345 			curr_net = net;
346 		}
347 		list_add_tail(&skb->list, &sublist);
348 	}
349 	/* dispatch final sublist */
350 	if (!list_empty(&sublist))
351 		ip6_sublist_rcv(&sublist, curr_dev, curr_net);
352 }
353 
354 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
355 
356 /*
357  *	Deliver the packet to the host
358  */
359 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
360 			      bool have_final)
361 {
362 	const struct inet6_protocol *ipprot;
363 	struct inet6_dev *idev;
364 	unsigned int nhoff;
365 	SKB_DR(reason);
366 	bool raw;
367 
368 	/*
369 	 *	Parse extension headers
370 	 */
371 
372 resubmit:
373 	idev = ip6_dst_idev(skb_dst(skb));
374 	nhoff = IP6CB(skb)->nhoff;
375 	if (!have_final) {
376 		if (!pskb_pull(skb, skb_transport_offset(skb)))
377 			goto discard;
378 		nexthdr = skb_network_header(skb)[nhoff];
379 	}
380 
381 resubmit_final:
382 	raw = raw6_local_deliver(skb, nexthdr);
383 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
384 	if (ipprot) {
385 		int ret;
386 
387 		if (have_final) {
388 			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
389 				/* Once we've seen a final protocol don't
390 				 * allow encapsulation on any non-final
391 				 * ones. This allows foo in UDP encapsulation
392 				 * to work.
393 				 */
394 				goto discard;
395 			}
396 		} else if (ipprot->flags & INET6_PROTO_FINAL) {
397 			const struct ipv6hdr *hdr;
398 			int sdif = inet6_sdif(skb);
399 			struct net_device *dev;
400 
401 			/* Only do this once for first final protocol */
402 			have_final = true;
403 
404 			/* Free reference early: we don't need it any more,
405 			   and it may hold ip_conntrack module loaded
406 			   indefinitely. */
407 			nf_reset_ct(skb);
408 
409 			skb_postpull_rcsum(skb, skb_network_header(skb),
410 					   skb_network_header_len(skb));
411 			hdr = ipv6_hdr(skb);
412 
413 			/* skb->dev passed may be master dev for vrfs. */
414 			if (sdif) {
415 				dev = dev_get_by_index_rcu(net, sdif);
416 				if (!dev)
417 					goto discard;
418 			} else {
419 				dev = skb->dev;
420 			}
421 
422 			if (ipv6_addr_is_multicast(&hdr->daddr) &&
423 			    !ipv6_chk_mcast_addr(dev, &hdr->daddr,
424 						 &hdr->saddr) &&
425 			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) {
426 				SKB_DR_SET(reason, IP_INADDRERRORS);
427 				goto discard;
428 			}
429 		}
430 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
431 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
432 			SKB_DR_SET(reason, XFRM_POLICY);
433 			goto discard;
434 		}
435 
436 		ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
437 				      skb);
438 		if (ret > 0) {
439 			if (ipprot->flags & INET6_PROTO_FINAL) {
440 				/* Not an extension header, most likely UDP
441 				 * encapsulation. Use return value as nexthdr
442 				 * protocol not nhoff (which presumably is
443 				 * not set by handler).
444 				 */
445 				nexthdr = ret;
446 				goto resubmit_final;
447 			} else {
448 				goto resubmit;
449 			}
450 		} else if (ret == 0) {
451 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
452 		}
453 	} else {
454 		if (!raw) {
455 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
456 				__IP6_INC_STATS(net, idev,
457 						IPSTATS_MIB_INUNKNOWNPROTOS);
458 				icmpv6_send(skb, ICMPV6_PARAMPROB,
459 					    ICMPV6_UNK_NEXTHDR, nhoff);
460 				SKB_DR_SET(reason, IP_NOPROTO);
461 			} else {
462 				SKB_DR_SET(reason, XFRM_POLICY);
463 			}
464 			kfree_skb_reason(skb, reason);
465 		} else {
466 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
467 			consume_skb(skb);
468 		}
469 	}
470 	return;
471 
472 discard:
473 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
474 	kfree_skb_reason(skb, reason);
475 }
476 
477 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
478 {
479 	skb_clear_delivery_time(skb);
480 	rcu_read_lock();
481 	ip6_protocol_deliver_rcu(net, skb, 0, false);
482 	rcu_read_unlock();
483 
484 	return 0;
485 }
486 
487 
488 int ip6_input(struct sk_buff *skb)
489 {
490 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
491 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
492 		       ip6_input_finish);
493 }
494 EXPORT_SYMBOL_GPL(ip6_input);
495 
496 int ip6_mc_input(struct sk_buff *skb)
497 {
498 	int sdif = inet6_sdif(skb);
499 	const struct ipv6hdr *hdr;
500 	struct net_device *dev;
501 	bool deliver;
502 
503 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
504 			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
505 			 skb->len);
506 
507 	/* skb->dev passed may be master dev for vrfs. */
508 	if (sdif) {
509 		rcu_read_lock();
510 		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
511 		if (!dev) {
512 			rcu_read_unlock();
513 			kfree_skb(skb);
514 			return -ENODEV;
515 		}
516 	} else {
517 		dev = skb->dev;
518 	}
519 
520 	hdr = ipv6_hdr(skb);
521 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
522 	if (sdif)
523 		rcu_read_unlock();
524 
525 #ifdef CONFIG_IPV6_MROUTE
526 	/*
527 	 *      IPv6 multicast router mode is now supported ;)
528 	 */
529 	if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
530 	    !(ipv6_addr_type(&hdr->daddr) &
531 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
532 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
533 		/*
534 		 * Okay, we try to forward - split and duplicate
535 		 * packets.
536 		 */
537 		struct sk_buff *skb2;
538 		struct inet6_skb_parm *opt = IP6CB(skb);
539 
540 		/* Check for MLD */
541 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
542 			/* Check if this is a mld message */
543 			u8 nexthdr = hdr->nexthdr;
544 			__be16 frag_off;
545 			int offset;
546 
547 			/* Check if the value of Router Alert
548 			 * is for MLD (0x0000).
549 			 */
550 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
551 				deliver = false;
552 
553 				if (!ipv6_ext_hdr(nexthdr)) {
554 					/* BUG */
555 					goto out;
556 				}
557 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
558 							  &nexthdr, &frag_off);
559 				if (offset < 0)
560 					goto out;
561 
562 				if (ipv6_is_mld(skb, nexthdr, offset))
563 					deliver = true;
564 
565 				goto out;
566 			}
567 			/* unknown RA - process it normally */
568 		}
569 
570 		if (deliver)
571 			skb2 = skb_clone(skb, GFP_ATOMIC);
572 		else {
573 			skb2 = skb;
574 			skb = NULL;
575 		}
576 
577 		if (skb2) {
578 			ip6_mr_input(skb2);
579 		}
580 	}
581 out:
582 #endif
583 	if (likely(deliver))
584 		ip6_input(skb);
585 	else {
586 		/* discard */
587 		kfree_skb(skb);
588 	}
589 
590 	return 0;
591 }
592