xref: /kernel/linux/linux-5.10/net/ipv6/ip6_input.c (revision 8c2ecf20) 
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
47void udp_v6_early_demux(struct sk_buff *);
48void tcp_v6_early_demux(struct sk_buff *);
49static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
50				struct sk_buff *skb)
51{
52	if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) &&
53	    !skb_dst(skb) && !skb->sk) {
54		switch (ipv6_hdr(skb)->nexthdr) {
55		case IPPROTO_TCP:
56			if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux))
57				tcp_v6_early_demux(skb);
58			break;
59		case IPPROTO_UDP:
60			if (READ_ONCE(net->ipv4.sysctl_udp_early_demux))
61				udp_v6_early_demux(skb);
62			break;
63		}
64	}
65
66	if (!skb_valid_dst(skb))
67		ip6_route_input(skb);
68}
69
70int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
71{
72	/* if ingress device is enslaved to an L3 master device pass the
73	 * skb to its handler for processing
74	 */
75	skb = l3mdev_ip6_rcv(skb);
76	if (!skb)
77		return NET_RX_SUCCESS;
78	ip6_rcv_finish_core(net, sk, skb);
79
80	return dst_input(skb);
81}
82
83static void ip6_sublist_rcv_finish(struct list_head *head)
84{
85	struct sk_buff *skb, *next;
86
87	list_for_each_entry_safe(skb, next, head, list) {
88		skb_list_del_init(skb);
89		dst_input(skb);
90	}
91}
92
93static bool ip6_can_use_hint(const struct sk_buff *skb,
94			     const struct sk_buff *hint)
95{
96	return hint && !skb_dst(skb) &&
97	       ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
98}
99
100static struct sk_buff *ip6_extract_route_hint(const struct net *net,
101					      struct sk_buff *skb)
102{
103	if (fib6_routes_require_src(net) || fib6_has_custom_rules(net))
104		return NULL;
105
106	return skb;
107}
108
109static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
110				struct list_head *head)
111{
112	struct sk_buff *skb, *next, *hint = NULL;
113	struct dst_entry *curr_dst = NULL;
114	struct list_head sublist;
115
116	INIT_LIST_HEAD(&sublist);
117	list_for_each_entry_safe(skb, next, head, list) {
118		struct dst_entry *dst;
119
120		skb_list_del_init(skb);
121		/* if ingress device is enslaved to an L3 master device pass the
122		 * skb to its handler for processing
123		 */
124		skb = l3mdev_ip6_rcv(skb);
125		if (!skb)
126			continue;
127
128		if (ip6_can_use_hint(skb, hint))
129			skb_dst_copy(skb, hint);
130		else
131			ip6_rcv_finish_core(net, sk, skb);
132		dst = skb_dst(skb);
133		if (curr_dst != dst) {
134			hint = ip6_extract_route_hint(net, skb);
135
136			/* dispatch old sublist */
137			if (!list_empty(&sublist))
138				ip6_sublist_rcv_finish(&sublist);
139			/* start new sublist */
140			INIT_LIST_HEAD(&sublist);
141			curr_dst = dst;
142		}
143		list_add_tail(&skb->list, &sublist);
144	}
145	/* dispatch final sublist */
146	ip6_sublist_rcv_finish(&sublist);
147}
148
149static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
150				    struct net *net)
151{
152	const struct ipv6hdr *hdr;
153	u32 pkt_len;
154	struct inet6_dev *idev;
155
156	if (skb->pkt_type == PACKET_OTHERHOST) {
157		kfree_skb(skb);
158		return NULL;
159	}
160
161	rcu_read_lock();
162
163	idev = __in6_dev_get(skb->dev);
164
165	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
166
167	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
168	    !idev || unlikely(idev->cnf.disable_ipv6)) {
169		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
170		goto drop;
171	}
172
173	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
174
175	/*
176	 * Store incoming device index. When the packet will
177	 * be queued, we cannot refer to skb->dev anymore.
178	 *
179	 * BTW, when we send a packet for our own local address on a
180	 * non-loopback interface (e.g. ethX), it is being delivered
181	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
182	 * It, however, should be considered as if it is being
183	 * arrived via the sending interface (ethX), because of the
184	 * nature of scoping architecture. --yoshfuji
185	 */
186	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
187
188	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
189		goto err;
190
191	hdr = ipv6_hdr(skb);
192
193	if (hdr->version != 6)
194		goto err;
195
196	__IP6_ADD_STATS(net, idev,
197			IPSTATS_MIB_NOECTPKTS +
198				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
199			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
200	/*
201	 * RFC4291 2.5.3
202	 * The loopback address must not be used as the source address in IPv6
203	 * packets that are sent outside of a single node. [..]
204	 * A packet received on an interface with a destination address
205	 * of loopback must be dropped.
206	 */
207	if ((ipv6_addr_loopback(&hdr->saddr) ||
208	     ipv6_addr_loopback(&hdr->daddr)) &&
209	    !(dev->flags & IFF_LOOPBACK) &&
210	    !netif_is_l3_master(dev))
211		goto err;
212
213	/* RFC4291 Errata ID: 3480
214	 * Interface-Local scope spans only a single interface on a
215	 * node and is useful only for loopback transmission of
216	 * multicast.  Packets with interface-local scope received
217	 * from another node must be discarded.
218	 */
219	if (!(skb->pkt_type == PACKET_LOOPBACK ||
220	      dev->flags & IFF_LOOPBACK) &&
221	    ipv6_addr_is_multicast(&hdr->daddr) &&
222	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
223		goto err;
224
225	/* If enabled, drop unicast packets that were encapsulated in link-layer
226	 * multicast or broadcast to protected against the so-called "hole-196"
227	 * attack in 802.11 wireless.
228	 */
229	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
230	    (skb->pkt_type == PACKET_BROADCAST ||
231	     skb->pkt_type == PACKET_MULTICAST) &&
232	    idev->cnf.drop_unicast_in_l2_multicast)
233		goto err;
234
235	/* RFC4291 2.7
236	 * Nodes must not originate a packet to a multicast address whose scope
237	 * field contains the reserved value 0; if such a packet is received, it
238	 * must be silently dropped.
239	 */
240	if (ipv6_addr_is_multicast(&hdr->daddr) &&
241	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
242		goto err;
243
244	/*
245	 * RFC4291 2.7
246	 * Multicast addresses must not be used as source addresses in IPv6
247	 * packets or appear in any Routing header.
248	 */
249	if (ipv6_addr_is_multicast(&hdr->saddr))
250		goto err;
251
252	skb->transport_header = skb->network_header + sizeof(*hdr);
253	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
254
255	pkt_len = ntohs(hdr->payload_len);
256
257	/* pkt_len may be zero if Jumbo payload option is present */
258	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
259		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
260			__IP6_INC_STATS(net,
261					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
262			goto drop;
263		}
264		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
265			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
266			goto drop;
267		}
268		hdr = ipv6_hdr(skb);
269	}
270
271	if (hdr->nexthdr == NEXTHDR_HOP) {
272		if (ipv6_parse_hopopts(skb) < 0) {
273			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
274			rcu_read_unlock();
275			return NULL;
276		}
277	}
278
279	rcu_read_unlock();
280
281	/* Must drop socket now because of tproxy. */
282	if (!skb_sk_is_prefetched(skb))
283		skb_orphan(skb);
284
285	return skb;
286err:
287	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
288drop:
289	rcu_read_unlock();
290	kfree_skb(skb);
291	return NULL;
292}
293
294int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
295{
296	struct net *net = dev_net(skb->dev);
297
298	skb = ip6_rcv_core(skb, dev, net);
299	if (skb == NULL)
300		return NET_RX_DROP;
301	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
302		       net, NULL, skb, dev, NULL,
303		       ip6_rcv_finish);
304}
305
306static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
307			    struct net *net)
308{
309	NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
310		     head, dev, NULL, ip6_rcv_finish);
311	ip6_list_rcv_finish(net, NULL, head);
312}
313
314/* Receive a list of IPv6 packets */
315void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
316		   struct net_device *orig_dev)
317{
318	struct net_device *curr_dev = NULL;
319	struct net *curr_net = NULL;
320	struct sk_buff *skb, *next;
321	struct list_head sublist;
322
323	INIT_LIST_HEAD(&sublist);
324	list_for_each_entry_safe(skb, next, head, list) {
325		struct net_device *dev = skb->dev;
326		struct net *net = dev_net(dev);
327
328		skb_list_del_init(skb);
329		skb = ip6_rcv_core(skb, dev, net);
330		if (skb == NULL)
331			continue;
332
333		if (curr_dev != dev || curr_net != net) {
334			/* dispatch old sublist */
335			if (!list_empty(&sublist))
336				ip6_sublist_rcv(&sublist, curr_dev, curr_net);
337			/* start new sublist */
338			INIT_LIST_HEAD(&sublist);
339			curr_dev = dev;
340			curr_net = net;
341		}
342		list_add_tail(&skb->list, &sublist);
343	}
344	/* dispatch final sublist */
345	if (!list_empty(&sublist))
346		ip6_sublist_rcv(&sublist, curr_dev, curr_net);
347}
348
349INDIRECT_CALLABLE_DECLARE(int udpv6_rcv(struct sk_buff *));
350INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
351
352/*
353 *	Deliver the packet to the host
354 */
355void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
356			      bool have_final)
357{
358	const struct inet6_protocol *ipprot;
359	struct inet6_dev *idev;
360	unsigned int nhoff;
361	bool raw;
362
363	/*
364	 *	Parse extension headers
365	 */
366
367resubmit:
368	idev = ip6_dst_idev(skb_dst(skb));
369	nhoff = IP6CB(skb)->nhoff;
370	if (!have_final) {
371		if (!pskb_pull(skb, skb_transport_offset(skb)))
372			goto discard;
373		nexthdr = skb_network_header(skb)[nhoff];
374	}
375
376resubmit_final:
377	raw = raw6_local_deliver(skb, nexthdr);
378	ipprot = rcu_dereference(inet6_protos[nexthdr]);
379	if (ipprot) {
380		int ret;
381
382		if (have_final) {
383			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
384				/* Once we've seen a final protocol don't
385				 * allow encapsulation on any non-final
386				 * ones. This allows foo in UDP encapsulation
387				 * to work.
388				 */
389				goto discard;
390			}
391		} else if (ipprot->flags & INET6_PROTO_FINAL) {
392			const struct ipv6hdr *hdr;
393			int sdif = inet6_sdif(skb);
394			struct net_device *dev;
395
396			/* Only do this once for first final protocol */
397			have_final = true;
398
399			/* Free reference early: we don't need it any more,
400			   and it may hold ip_conntrack module loaded
401			   indefinitely. */
402			nf_reset_ct(skb);
403
404			skb_postpull_rcsum(skb, skb_network_header(skb),
405					   skb_network_header_len(skb));
406			hdr = ipv6_hdr(skb);
407
408			/* skb->dev passed may be master dev for vrfs. */
409			if (sdif) {
410				dev = dev_get_by_index_rcu(net, sdif);
411				if (!dev)
412					goto discard;
413			} else {
414				dev = skb->dev;
415			}
416
417			if (ipv6_addr_is_multicast(&hdr->daddr) &&
418			    !ipv6_chk_mcast_addr(dev, &hdr->daddr,
419						 &hdr->saddr) &&
420			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
421				goto discard;
422		}
423		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
424		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
425			goto discard;
426
427		ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
428				      skb);
429		if (ret > 0) {
430			if (ipprot->flags & INET6_PROTO_FINAL) {
431				/* Not an extension header, most likely UDP
432				 * encapsulation. Use return value as nexthdr
433				 * protocol not nhoff (which presumably is
434				 * not set by handler).
435				 */
436				nexthdr = ret;
437				goto resubmit_final;
438			} else {
439				goto resubmit;
440			}
441		} else if (ret == 0) {
442			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
443		}
444	} else {
445		if (!raw) {
446			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
447				__IP6_INC_STATS(net, idev,
448						IPSTATS_MIB_INUNKNOWNPROTOS);
449				icmpv6_send(skb, ICMPV6_PARAMPROB,
450					    ICMPV6_UNK_NEXTHDR, nhoff);
451			}
452			kfree_skb(skb);
453		} else {
454			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
455			consume_skb(skb);
456		}
457	}
458	return;
459
460discard:
461	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
462	kfree_skb(skb);
463}
464
465static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
466{
467	rcu_read_lock();
468	ip6_protocol_deliver_rcu(net, skb, 0, false);
469	rcu_read_unlock();
470
471	return 0;
472}
473
474
475int ip6_input(struct sk_buff *skb)
476{
477	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
478		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
479		       ip6_input_finish);
480}
481EXPORT_SYMBOL_GPL(ip6_input);
482
483int ip6_mc_input(struct sk_buff *skb)
484{
485	int sdif = inet6_sdif(skb);
486	const struct ipv6hdr *hdr;
487	struct net_device *dev;
488	bool deliver;
489
490	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
491			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
492			 skb->len);
493
494	/* skb->dev passed may be master dev for vrfs. */
495	if (sdif) {
496		rcu_read_lock();
497		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
498		if (!dev) {
499			rcu_read_unlock();
500			kfree_skb(skb);
501			return -ENODEV;
502		}
503	} else {
504		dev = skb->dev;
505	}
506
507	hdr = ipv6_hdr(skb);
508	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
509	if (sdif)
510		rcu_read_unlock();
511
512#ifdef CONFIG_IPV6_MROUTE
513	/*
514	 *      IPv6 multicast router mode is now supported ;)
515	 */
516	if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
517	    !(ipv6_addr_type(&hdr->daddr) &
518	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
519	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
520		/*
521		 * Okay, we try to forward - split and duplicate
522		 * packets.
523		 */
524		struct sk_buff *skb2;
525		struct inet6_skb_parm *opt = IP6CB(skb);
526
527		/* Check for MLD */
528		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
529			/* Check if this is a mld message */
530			u8 nexthdr = hdr->nexthdr;
531			__be16 frag_off;
532			int offset;
533
534			/* Check if the value of Router Alert
535			 * is for MLD (0x0000).
536			 */
537			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
538				deliver = false;
539
540				if (!ipv6_ext_hdr(nexthdr)) {
541					/* BUG */
542					goto out;
543				}
544				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
545							  &nexthdr, &frag_off);
546				if (offset < 0)
547					goto out;
548
549				if (ipv6_is_mld(skb, nexthdr, offset))
550					deliver = true;
551
552				goto out;
553			}
554			/* unknown RA - process it normally */
555		}
556
557		if (deliver)
558			skb2 = skb_clone(skb, GFP_ATOMIC);
559		else {
560			skb2 = skb;
561			skb = NULL;
562		}
563
564		if (skb2) {
565			ip6_mr_input(skb2);
566		}
567	}
568out:
569#endif
570	if (likely(deliver))
571		ip6_input(skb);
572	else {
573		/* discard */
574		kfree_skb(skb);
575	}
576
577	return 0;
578}
579