xref: /openbmc/linux/net/ipv6/ip6_input.c (revision 82b5d164)
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 void ip6_list_rcv_finish(struct net *net, struct sock *sk,
90 				struct list_head *head)
91 {
92 	struct dst_entry *curr_dst = NULL;
93 	struct sk_buff *skb, *next;
94 	struct list_head sublist;
95 
96 	INIT_LIST_HEAD(&sublist);
97 	list_for_each_entry_safe(skb, next, head, list) {
98 		struct dst_entry *dst;
99 
100 		skb_list_del_init(skb);
101 		/* if ingress device is enslaved to an L3 master device pass the
102 		 * skb to its handler for processing
103 		 */
104 		skb = l3mdev_ip6_rcv(skb);
105 		if (!skb)
106 			continue;
107 		ip6_rcv_finish_core(net, sk, skb);
108 		dst = skb_dst(skb);
109 		if (curr_dst != dst) {
110 			/* dispatch old sublist */
111 			if (!list_empty(&sublist))
112 				ip6_sublist_rcv_finish(&sublist);
113 			/* start new sublist */
114 			INIT_LIST_HEAD(&sublist);
115 			curr_dst = dst;
116 		}
117 		list_add_tail(&skb->list, &sublist);
118 	}
119 	/* dispatch final sublist */
120 	ip6_sublist_rcv_finish(&sublist);
121 }
122 
123 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
124 				    struct net *net)
125 {
126 	const struct ipv6hdr *hdr;
127 	u32 pkt_len;
128 	struct inet6_dev *idev;
129 
130 	if (skb->pkt_type == PACKET_OTHERHOST) {
131 		kfree_skb(skb);
132 		return NULL;
133 	}
134 
135 	rcu_read_lock();
136 
137 	idev = __in6_dev_get(skb->dev);
138 
139 	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
140 
141 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
142 	    !idev || unlikely(idev->cnf.disable_ipv6)) {
143 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
144 		goto drop;
145 	}
146 
147 	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
148 
149 	/*
150 	 * Store incoming device index. When the packet will
151 	 * be queued, we cannot refer to skb->dev anymore.
152 	 *
153 	 * BTW, when we send a packet for our own local address on a
154 	 * non-loopback interface (e.g. ethX), it is being delivered
155 	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
156 	 * It, however, should be considered as if it is being
157 	 * arrived via the sending interface (ethX), because of the
158 	 * nature of scoping architecture. --yoshfuji
159 	 */
160 	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
161 
162 	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
163 		goto err;
164 
165 	hdr = ipv6_hdr(skb);
166 
167 	if (hdr->version != 6)
168 		goto err;
169 
170 	__IP6_ADD_STATS(net, idev,
171 			IPSTATS_MIB_NOECTPKTS +
172 				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
173 			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
174 	/*
175 	 * RFC4291 2.5.3
176 	 * The loopback address must not be used as the source address in IPv6
177 	 * packets that are sent outside of a single node. [..]
178 	 * A packet received on an interface with a destination address
179 	 * of loopback must be dropped.
180 	 */
181 	if ((ipv6_addr_loopback(&hdr->saddr) ||
182 	     ipv6_addr_loopback(&hdr->daddr)) &&
183 	    !(dev->flags & IFF_LOOPBACK) &&
184 	    !netif_is_l3_master(dev))
185 		goto err;
186 
187 	/* RFC4291 Errata ID: 3480
188 	 * Interface-Local scope spans only a single interface on a
189 	 * node and is useful only for loopback transmission of
190 	 * multicast.  Packets with interface-local scope received
191 	 * from another node must be discarded.
192 	 */
193 	if (!(skb->pkt_type == PACKET_LOOPBACK ||
194 	      dev->flags & IFF_LOOPBACK) &&
195 	    ipv6_addr_is_multicast(&hdr->daddr) &&
196 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
197 		goto err;
198 
199 	/* If enabled, drop unicast packets that were encapsulated in link-layer
200 	 * multicast or broadcast to protected against the so-called "hole-196"
201 	 * attack in 802.11 wireless.
202 	 */
203 	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
204 	    (skb->pkt_type == PACKET_BROADCAST ||
205 	     skb->pkt_type == PACKET_MULTICAST) &&
206 	    idev->cnf.drop_unicast_in_l2_multicast)
207 		goto err;
208 
209 	/* RFC4291 2.7
210 	 * Nodes must not originate a packet to a multicast address whose scope
211 	 * field contains the reserved value 0; if such a packet is received, it
212 	 * must be silently dropped.
213 	 */
214 	if (ipv6_addr_is_multicast(&hdr->daddr) &&
215 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
216 		goto err;
217 
218 	/*
219 	 * RFC4291 2.7
220 	 * Multicast addresses must not be used as source addresses in IPv6
221 	 * packets or appear in any Routing header.
222 	 */
223 	if (ipv6_addr_is_multicast(&hdr->saddr))
224 		goto err;
225 
226 	skb->transport_header = skb->network_header + sizeof(*hdr);
227 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
228 
229 	pkt_len = ntohs(hdr->payload_len);
230 
231 	/* pkt_len may be zero if Jumbo payload option is present */
232 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
233 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
234 			__IP6_INC_STATS(net,
235 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
236 			goto drop;
237 		}
238 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
239 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
240 			goto drop;
241 		}
242 		hdr = ipv6_hdr(skb);
243 	}
244 
245 	if (hdr->nexthdr == NEXTHDR_HOP) {
246 		if (ipv6_parse_hopopts(skb) < 0) {
247 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
248 			rcu_read_unlock();
249 			return NULL;
250 		}
251 	}
252 
253 	rcu_read_unlock();
254 
255 	/* Must drop socket now because of tproxy. */
256 	skb_orphan(skb);
257 
258 	return skb;
259 err:
260 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
261 drop:
262 	rcu_read_unlock();
263 	kfree_skb(skb);
264 	return NULL;
265 }
266 
267 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
268 {
269 	struct net *net = dev_net(skb->dev);
270 
271 	skb = ip6_rcv_core(skb, dev, net);
272 	if (skb == NULL)
273 		return NET_RX_DROP;
274 	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
275 		       net, NULL, skb, dev, NULL,
276 		       ip6_rcv_finish);
277 }
278 
279 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
280 			    struct net *net)
281 {
282 	NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
283 		     head, dev, NULL, ip6_rcv_finish);
284 	ip6_list_rcv_finish(net, NULL, head);
285 }
286 
287 /* Receive a list of IPv6 packets */
288 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
289 		   struct net_device *orig_dev)
290 {
291 	struct net_device *curr_dev = NULL;
292 	struct net *curr_net = NULL;
293 	struct sk_buff *skb, *next;
294 	struct list_head sublist;
295 
296 	INIT_LIST_HEAD(&sublist);
297 	list_for_each_entry_safe(skb, next, head, list) {
298 		struct net_device *dev = skb->dev;
299 		struct net *net = dev_net(dev);
300 
301 		skb_list_del_init(skb);
302 		skb = ip6_rcv_core(skb, dev, net);
303 		if (skb == NULL)
304 			continue;
305 
306 		if (curr_dev != dev || curr_net != net) {
307 			/* dispatch old sublist */
308 			if (!list_empty(&sublist))
309 				ip6_sublist_rcv(&sublist, curr_dev, curr_net);
310 			/* start new sublist */
311 			INIT_LIST_HEAD(&sublist);
312 			curr_dev = dev;
313 			curr_net = net;
314 		}
315 		list_add_tail(&skb->list, &sublist);
316 	}
317 	/* dispatch final sublist */
318 	ip6_sublist_rcv(&sublist, curr_dev, curr_net);
319 }
320 
321 INDIRECT_CALLABLE_DECLARE(int udpv6_rcv(struct sk_buff *));
322 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
323 
324 /*
325  *	Deliver the packet to the host
326  */
327 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
328 			      bool have_final)
329 {
330 	const struct inet6_protocol *ipprot;
331 	struct inet6_dev *idev;
332 	unsigned int nhoff;
333 	bool raw;
334 
335 	/*
336 	 *	Parse extension headers
337 	 */
338 
339 resubmit:
340 	idev = ip6_dst_idev(skb_dst(skb));
341 	nhoff = IP6CB(skb)->nhoff;
342 	if (!have_final) {
343 		if (!pskb_pull(skb, skb_transport_offset(skb)))
344 			goto discard;
345 		nexthdr = skb_network_header(skb)[nhoff];
346 	}
347 
348 resubmit_final:
349 	raw = raw6_local_deliver(skb, nexthdr);
350 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
351 	if (ipprot) {
352 		int ret;
353 
354 		if (have_final) {
355 			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
356 				/* Once we've seen a final protocol don't
357 				 * allow encapsulation on any non-final
358 				 * ones. This allows foo in UDP encapsulation
359 				 * to work.
360 				 */
361 				goto discard;
362 			}
363 		} else if (ipprot->flags & INET6_PROTO_FINAL) {
364 			const struct ipv6hdr *hdr;
365 			int sdif = inet6_sdif(skb);
366 			struct net_device *dev;
367 
368 			/* Only do this once for first final protocol */
369 			have_final = true;
370 
371 			/* Free reference early: we don't need it any more,
372 			   and it may hold ip_conntrack module loaded
373 			   indefinitely. */
374 			nf_reset(skb);
375 
376 			skb_postpull_rcsum(skb, skb_network_header(skb),
377 					   skb_network_header_len(skb));
378 			hdr = ipv6_hdr(skb);
379 
380 			/* skb->dev passed may be master dev for vrfs. */
381 			if (sdif) {
382 				dev = dev_get_by_index_rcu(net, sdif);
383 				if (!dev)
384 					goto discard;
385 			} else {
386 				dev = skb->dev;
387 			}
388 
389 			if (ipv6_addr_is_multicast(&hdr->daddr) &&
390 			    !ipv6_chk_mcast_addr(dev, &hdr->daddr,
391 						 &hdr->saddr) &&
392 			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
393 				goto discard;
394 		}
395 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
396 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
397 			goto discard;
398 
399 		ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
400 				      skb);
401 		if (ret > 0) {
402 			if (ipprot->flags & INET6_PROTO_FINAL) {
403 				/* Not an extension header, most likely UDP
404 				 * encapsulation. Use return value as nexthdr
405 				 * protocol not nhoff (which presumably is
406 				 * not set by handler).
407 				 */
408 				nexthdr = ret;
409 				goto resubmit_final;
410 			} else {
411 				goto resubmit;
412 			}
413 		} else if (ret == 0) {
414 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
415 		}
416 	} else {
417 		if (!raw) {
418 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
419 				__IP6_INC_STATS(net, idev,
420 						IPSTATS_MIB_INUNKNOWNPROTOS);
421 				icmpv6_send(skb, ICMPV6_PARAMPROB,
422 					    ICMPV6_UNK_NEXTHDR, nhoff);
423 			}
424 			kfree_skb(skb);
425 		} else {
426 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
427 			consume_skb(skb);
428 		}
429 	}
430 	return;
431 
432 discard:
433 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
434 	kfree_skb(skb);
435 }
436 
437 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
438 {
439 	rcu_read_lock();
440 	ip6_protocol_deliver_rcu(net, skb, 0, false);
441 	rcu_read_unlock();
442 
443 	return 0;
444 }
445 
446 
447 int ip6_input(struct sk_buff *skb)
448 {
449 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
450 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
451 		       ip6_input_finish);
452 }
453 EXPORT_SYMBOL_GPL(ip6_input);
454 
455 int ip6_mc_input(struct sk_buff *skb)
456 {
457 	int sdif = inet6_sdif(skb);
458 	const struct ipv6hdr *hdr;
459 	struct net_device *dev;
460 	bool deliver;
461 
462 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
463 			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
464 			 skb->len);
465 
466 	/* skb->dev passed may be master dev for vrfs. */
467 	if (sdif) {
468 		rcu_read_lock();
469 		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
470 		if (!dev) {
471 			rcu_read_unlock();
472 			kfree_skb(skb);
473 			return -ENODEV;
474 		}
475 	} else {
476 		dev = skb->dev;
477 	}
478 
479 	hdr = ipv6_hdr(skb);
480 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
481 	if (sdif)
482 		rcu_read_unlock();
483 
484 #ifdef CONFIG_IPV6_MROUTE
485 	/*
486 	 *      IPv6 multicast router mode is now supported ;)
487 	 */
488 	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
489 	    !(ipv6_addr_type(&hdr->daddr) &
490 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
491 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
492 		/*
493 		 * Okay, we try to forward - split and duplicate
494 		 * packets.
495 		 */
496 		struct sk_buff *skb2;
497 		struct inet6_skb_parm *opt = IP6CB(skb);
498 
499 		/* Check for MLD */
500 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
501 			/* Check if this is a mld message */
502 			u8 nexthdr = hdr->nexthdr;
503 			__be16 frag_off;
504 			int offset;
505 
506 			/* Check if the value of Router Alert
507 			 * is for MLD (0x0000).
508 			 */
509 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
510 				deliver = false;
511 
512 				if (!ipv6_ext_hdr(nexthdr)) {
513 					/* BUG */
514 					goto out;
515 				}
516 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
517 							  &nexthdr, &frag_off);
518 				if (offset < 0)
519 					goto out;
520 
521 				if (ipv6_is_mld(skb, nexthdr, offset))
522 					deliver = true;
523 
524 				goto out;
525 			}
526 			/* unknown RA - process it normally */
527 		}
528 
529 		if (deliver)
530 			skb2 = skb_clone(skb, GFP_ATOMIC);
531 		else {
532 			skb2 = skb;
533 			skb = NULL;
534 		}
535 
536 		if (skb2) {
537 			ip6_mr_input(skb2);
538 		}
539 	}
540 out:
541 #endif
542 	if (likely(deliver))
543 		ip6_input(skb);
544 	else {
545 		/* discard */
546 		kfree_skb(skb);
547 	}
548 
549 	return 0;
550 }
551