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