xref: /openbmc/linux/net/ipv6/ip6_input.c (revision f125e2d4)
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 
36 #include <net/ipv6.h>
37 #include <net/protocol.h>
38 #include <net/transp_v6.h>
39 #include <net/rawv6.h>
40 #include <net/ndisc.h>
41 #include <net/ip6_route.h>
42 #include <net/addrconf.h>
43 #include <net/xfrm.h>
44 #include <net/inet_ecn.h>
45 #include <net/dst_metadata.h>
46 
47 INDIRECT_CALLABLE_DECLARE(void udp_v6_early_demux(struct sk_buff *));
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 	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 	/* While RFC4291 is not explicit about v4mapped addresses
249 	 * in IPv6 headers, it seems clear linux dual-stack
250 	 * model can not deal properly with these.
251 	 * Security models could be fooled by ::ffff:127.0.0.1 for example.
252 	 *
253 	 * https://tools.ietf.org/html/draft-itojun-v6ops-v4mapped-harmful-02
254 	 */
255 	if (ipv6_addr_v4mapped(&hdr->saddr))
256 		goto err;
257 
258 	skb->transport_header = skb->network_header + sizeof(*hdr);
259 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
260 
261 	pkt_len = ntohs(hdr->payload_len);
262 
263 	/* pkt_len may be zero if Jumbo payload option is present */
264 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
265 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
266 			__IP6_INC_STATS(net,
267 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
268 			goto drop;
269 		}
270 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
271 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
272 			goto drop;
273 		}
274 		hdr = ipv6_hdr(skb);
275 	}
276 
277 	if (hdr->nexthdr == NEXTHDR_HOP) {
278 		if (ipv6_parse_hopopts(skb) < 0) {
279 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
280 			rcu_read_unlock();
281 			return NULL;
282 		}
283 	}
284 
285 	rcu_read_unlock();
286 
287 	/* Must drop socket now because of tproxy. */
288 	skb_orphan(skb);
289 
290 	return skb;
291 err:
292 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
293 drop:
294 	rcu_read_unlock();
295 	kfree_skb(skb);
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 udpv6_rcv(struct sk_buff *));
355 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
356 
357 /*
358  *	Deliver the packet to the host
359  */
360 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
361 			      bool have_final)
362 {
363 	const struct inet6_protocol *ipprot;
364 	struct inet6_dev *idev;
365 	unsigned int nhoff;
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 				goto discard;
427 		}
428 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
429 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
430 			goto discard;
431 
432 		ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
433 				      skb);
434 		if (ret > 0) {
435 			if (ipprot->flags & INET6_PROTO_FINAL) {
436 				/* Not an extension header, most likely UDP
437 				 * encapsulation. Use return value as nexthdr
438 				 * protocol not nhoff (which presumably is
439 				 * not set by handler).
440 				 */
441 				nexthdr = ret;
442 				goto resubmit_final;
443 			} else {
444 				goto resubmit;
445 			}
446 		} else if (ret == 0) {
447 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
448 		}
449 	} else {
450 		if (!raw) {
451 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
452 				__IP6_INC_STATS(net, idev,
453 						IPSTATS_MIB_INUNKNOWNPROTOS);
454 				icmpv6_send(skb, ICMPV6_PARAMPROB,
455 					    ICMPV6_UNK_NEXTHDR, nhoff);
456 			}
457 			kfree_skb(skb);
458 		} else {
459 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
460 			consume_skb(skb);
461 		}
462 	}
463 	return;
464 
465 discard:
466 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
467 	kfree_skb(skb);
468 }
469 
470 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
471 {
472 	rcu_read_lock();
473 	ip6_protocol_deliver_rcu(net, skb, 0, false);
474 	rcu_read_unlock();
475 
476 	return 0;
477 }
478 
479 
480 int ip6_input(struct sk_buff *skb)
481 {
482 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
483 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
484 		       ip6_input_finish);
485 }
486 EXPORT_SYMBOL_GPL(ip6_input);
487 
488 int ip6_mc_input(struct sk_buff *skb)
489 {
490 	int sdif = inet6_sdif(skb);
491 	const struct ipv6hdr *hdr;
492 	struct net_device *dev;
493 	bool deliver;
494 
495 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
496 			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
497 			 skb->len);
498 
499 	/* skb->dev passed may be master dev for vrfs. */
500 	if (sdif) {
501 		rcu_read_lock();
502 		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
503 		if (!dev) {
504 			rcu_read_unlock();
505 			kfree_skb(skb);
506 			return -ENODEV;
507 		}
508 	} else {
509 		dev = skb->dev;
510 	}
511 
512 	hdr = ipv6_hdr(skb);
513 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
514 	if (sdif)
515 		rcu_read_unlock();
516 
517 #ifdef CONFIG_IPV6_MROUTE
518 	/*
519 	 *      IPv6 multicast router mode is now supported ;)
520 	 */
521 	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
522 	    !(ipv6_addr_type(&hdr->daddr) &
523 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
524 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
525 		/*
526 		 * Okay, we try to forward - split and duplicate
527 		 * packets.
528 		 */
529 		struct sk_buff *skb2;
530 		struct inet6_skb_parm *opt = IP6CB(skb);
531 
532 		/* Check for MLD */
533 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
534 			/* Check if this is a mld message */
535 			u8 nexthdr = hdr->nexthdr;
536 			__be16 frag_off;
537 			int offset;
538 
539 			/* Check if the value of Router Alert
540 			 * is for MLD (0x0000).
541 			 */
542 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
543 				deliver = false;
544 
545 				if (!ipv6_ext_hdr(nexthdr)) {
546 					/* BUG */
547 					goto out;
548 				}
549 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
550 							  &nexthdr, &frag_off);
551 				if (offset < 0)
552 					goto out;
553 
554 				if (ipv6_is_mld(skb, nexthdr, offset))
555 					deliver = true;
556 
557 				goto out;
558 			}
559 			/* unknown RA - process it normally */
560 		}
561 
562 		if (deliver)
563 			skb2 = skb_clone(skb, GFP_ATOMIC);
564 		else {
565 			skb2 = skb;
566 			skb = NULL;
567 		}
568 
569 		if (skb2) {
570 			ip6_mr_input(skb2);
571 		}
572 	}
573 out:
574 #endif
575 	if (likely(deliver))
576 		ip6_input(skb);
577 	else {
578 		/* discard */
579 		kfree_skb(skb);
580 	}
581 
582 	return 0;
583 }
584