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
48static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
49 struct sk_buff *skb)
50{
51 if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) &&
52 !skb_dst(skb) && !skb->sk) {
53 switch (ipv6_hdr(skb)->nexthdr) {
54 case IPPROTO_TCP:
55 if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux))
56 tcp_v6_early_demux(skb);
57 break;
58 case IPPROTO_UDP:
59 if (READ_ONCE(net->ipv4.sysctl_udp_early_demux))
60 udp_v6_early_demux(skb);
61 break;
62 }
63 }
64
65 if (!skb_valid_dst(skb))
66 ip6_route_input(skb);
67}
68
69int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
70{
71 /* if ingress device is enslaved to an L3 master device pass the
72 * skb to its handler for processing
73 */
74 skb = l3mdev_ip6_rcv(skb);
75 if (!skb)
76 return NET_RX_SUCCESS;
77 ip6_rcv_finish_core(net, sk, skb);
78
79 return dst_input(skb);
80}
81
82static void ip6_sublist_rcv_finish(struct list_head *head)
83{
84 struct sk_buff *skb, *next;
85
86 list_for_each_entry_safe(skb, next, head, list) {
87 skb_list_del_init(skb);
88 dst_input(skb);
89 }
90}
91
92static bool ip6_can_use_hint(const struct sk_buff *skb,
93 const struct sk_buff *hint)
94{
95 return hint && !skb_dst(skb) &&
96 ipv6_addr_equal(a1: &ipv6_hdr(skb: hint)->daddr, a2: &ipv6_hdr(skb)->daddr);
97}
98
99static struct sk_buff *ip6_extract_route_hint(const struct net *net,
100 struct sk_buff *skb)
101{
102 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net) ||
103 IP6CB(skb)->flags & IP6SKB_MULTIPATH)
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(list: &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(nskb: skb, oskb: 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(head: &sublist))
138 ip6_sublist_rcv_finish(head: &sublist);
139 /* start new sublist */
140 INIT_LIST_HEAD(list: &sublist);
141 curr_dst = dst;
142 }
143 list_add_tail(new: &skb->list, head: &sublist);
144 }
145 /* dispatch final sublist */
146 ip6_sublist_rcv_finish(head: &sublist);
147}
148
149static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
150 struct net *net)
151{
152 enum skb_drop_reason reason;
153 const struct ipv6hdr *hdr;
154 u32 pkt_len;
155 struct inet6_dev *idev;
156
157 if (skb->pkt_type == PACKET_OTHERHOST) {
158 dev_core_stats_rx_otherhost_dropped_inc(dev: skb->dev);
159 kfree_skb_reason(skb, reason: SKB_DROP_REASON_OTHERHOST);
160 return NULL;
161 }
162
163 rcu_read_lock();
164
165 idev = __in6_dev_get(dev: skb->dev);
166
167 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
168
169 SKB_DR_SET(reason, NOT_SPECIFIED);
170 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
171 !idev || unlikely(idev->cnf.disable_ipv6)) {
172 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
173 if (idev && unlikely(idev->cnf.disable_ipv6))
174 SKB_DR_SET(reason, IPV6DISABLED);
175 goto drop;
176 }
177
178 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
179
180 /*
181 * Store incoming device index. When the packet will
182 * be queued, we cannot refer to skb->dev anymore.
183 *
184 * BTW, when we send a packet for our own local address on a
185 * non-loopback interface (e.g. ethX), it is being delivered
186 * via the loopback interface (lo) here; skb->dev = loopback_dev.
187 * It, however, should be considered as if it is being
188 * arrived via the sending interface (ethX), because of the
189 * nature of scoping architecture. --yoshfuji
190 */
191 IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(dst: skb_dst(skb))->dev->ifindex : dev->ifindex;
192
193 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
194 goto err;
195
196 hdr = ipv6_hdr(skb);
197
198 if (hdr->version != 6) {
199 SKB_DR_SET(reason, UNHANDLED_PROTO);
200 goto err;
201 }
202
203 __IP6_ADD_STATS(net, idev,
204 IPSTATS_MIB_NOECTPKTS +
205 (ipv6_get_dsfield(hdr) & INET_ECN_MASK),
206 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
207 /*
208 * RFC4291 2.5.3
209 * The loopback address must not be used as the source address in IPv6
210 * packets that are sent outside of a single node. [..]
211 * A packet received on an interface with a destination address
212 * of loopback must be dropped.
213 */
214 if ((ipv6_addr_loopback(a: &hdr->saddr) ||
215 ipv6_addr_loopback(a: &hdr->daddr)) &&
216 !(dev->flags & IFF_LOOPBACK) &&
217 !netif_is_l3_master(dev))
218 goto err;
219
220 /* RFC4291 Errata ID: 3480
221 * Interface-Local scope spans only a single interface on a
222 * node and is useful only for loopback transmission of
223 * multicast. Packets with interface-local scope received
224 * from another node must be discarded.
225 */
226 if (!(skb->pkt_type == PACKET_LOOPBACK ||
227 dev->flags & IFF_LOOPBACK) &&
228 ipv6_addr_is_multicast(addr: &hdr->daddr) &&
229 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
230 goto err;
231
232 /* If enabled, drop unicast packets that were encapsulated in link-layer
233 * multicast or broadcast to protected against the so-called "hole-196"
234 * attack in 802.11 wireless.
235 */
236 if (!ipv6_addr_is_multicast(addr: &hdr->daddr) &&
237 (skb->pkt_type == PACKET_BROADCAST ||
238 skb->pkt_type == PACKET_MULTICAST) &&
239 idev->cnf.drop_unicast_in_l2_multicast) {
240 SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST);
241 goto err;
242 }
243
244 /* RFC4291 2.7
245 * Nodes must not originate a packet to a multicast address whose scope
246 * field contains the reserved value 0; if such a packet is received, it
247 * must be silently dropped.
248 */
249 if (ipv6_addr_is_multicast(addr: &hdr->daddr) &&
250 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
251 goto err;
252
253 /*
254 * RFC4291 2.7
255 * Multicast addresses must not be used as source addresses in IPv6
256 * packets or appear in any Routing header.
257 */
258 if (ipv6_addr_is_multicast(addr: &hdr->saddr))
259 goto err;
260
261 skb->transport_header = skb->network_header + sizeof(*hdr);
262 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
263
264 pkt_len = ntohs(hdr->payload_len);
265
266 /* pkt_len may be zero if Jumbo payload option is present */
267 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
268 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
269 __IP6_INC_STATS(net,
270 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
271 SKB_DR_SET(reason, PKT_TOO_SMALL);
272 goto drop;
273 }
274 if (pskb_trim_rcsum(skb, len: pkt_len + sizeof(struct ipv6hdr)))
275 goto err;
276 hdr = ipv6_hdr(skb);
277 }
278
279 if (hdr->nexthdr == NEXTHDR_HOP) {
280 if (ipv6_parse_hopopts(skb) < 0) {
281 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
282 rcu_read_unlock();
283 return NULL;
284 }
285 }
286
287 rcu_read_unlock();
288
289 /* Must drop socket now because of tproxy. */
290 if (!skb_sk_is_prefetched(skb))
291 skb_orphan(skb);
292
293 return skb;
294err:
295 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
296 SKB_DR_OR(reason, IP_INHDR);
297drop:
298 rcu_read_unlock();
299 kfree_skb_reason(skb, reason);
300 return NULL;
301}
302
303int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
304{
305 struct net *net = dev_net(dev: skb->dev);
306
307 skb = ip6_rcv_core(skb, dev, net);
308 if (skb == NULL)
309 return NET_RX_DROP;
310 return NF_HOOK(pf: NFPROTO_IPV6, hook: NF_INET_PRE_ROUTING,
311 net, NULL, skb, in: dev, NULL,
312 okfn: ip6_rcv_finish);
313}
314
315static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
316 struct net *net)
317{
318 NF_HOOK_LIST(pf: NFPROTO_IPV6, hook: NF_INET_PRE_ROUTING, net, NULL,
319 head, in: dev, NULL, okfn: ip6_rcv_finish);
320 ip6_list_rcv_finish(net, NULL, head);
321}
322
323/* Receive a list of IPv6 packets */
324void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
325 struct net_device *orig_dev)
326{
327 struct net_device *curr_dev = NULL;
328 struct net *curr_net = NULL;
329 struct sk_buff *skb, *next;
330 struct list_head sublist;
331
332 INIT_LIST_HEAD(list: &sublist);
333 list_for_each_entry_safe(skb, next, head, list) {
334 struct net_device *dev = skb->dev;
335 struct net *net = dev_net(dev);
336
337 skb_list_del_init(skb);
338 skb = ip6_rcv_core(skb, dev, net);
339 if (skb == NULL)
340 continue;
341
342 if (curr_dev != dev || curr_net != net) {
343 /* dispatch old sublist */
344 if (!list_empty(head: &sublist))
345 ip6_sublist_rcv(head: &sublist, dev: curr_dev, net: curr_net);
346 /* start new sublist */
347 INIT_LIST_HEAD(list: &sublist);
348 curr_dev = dev;
349 curr_net = net;
350 }
351 list_add_tail(new: &skb->list, head: &sublist);
352 }
353 /* dispatch final sublist */
354 if (!list_empty(head: &sublist))
355 ip6_sublist_rcv(head: &sublist, dev: curr_dev, net: curr_net);
356}
357
358INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
359
360/*
361 * Deliver the packet to the host
362 */
363void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
364 bool have_final)
365{
366 const struct inet6_protocol *ipprot;
367 struct inet6_dev *idev;
368 unsigned int nhoff;
369 SKB_DR(reason);
370 bool raw;
371
372 /*
373 * Parse extension headers
374 */
375
376resubmit:
377 idev = ip6_dst_idev(dst: skb_dst(skb));
378 nhoff = IP6CB(skb)->nhoff;
379 if (!have_final) {
380 if (!pskb_pull(skb, len: skb_transport_offset(skb)))
381 goto discard;
382 nexthdr = skb_network_header(skb)[nhoff];
383 }
384
385resubmit_final:
386 raw = raw6_local_deliver(skb, nexthdr);
387 ipprot = rcu_dereference(inet6_protos[nexthdr]);
388 if (ipprot) {
389 int ret;
390
391 if (have_final) {
392 if (!(ipprot->flags & INET6_PROTO_FINAL)) {
393 /* Once we've seen a final protocol don't
394 * allow encapsulation on any non-final
395 * ones. This allows foo in UDP encapsulation
396 * to work.
397 */
398 goto discard;
399 }
400 } else if (ipprot->flags & INET6_PROTO_FINAL) {
401 const struct ipv6hdr *hdr;
402 int sdif = inet6_sdif(skb);
403 struct net_device *dev;
404
405 /* Only do this once for first final protocol */
406 have_final = true;
407
408
409 skb_postpull_rcsum(skb, start: skb_network_header(skb),
410 len: 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, ifindex: sdif);
416 if (!dev)
417 goto discard;
418 } else {
419 dev = skb->dev;
420 }
421
422 if (ipv6_addr_is_multicast(addr: &hdr->daddr) &&
423 !ipv6_chk_mcast_addr(dev, group: &hdr->daddr,
424 src_addr: &hdr->saddr) &&
425 !ipv6_is_mld(skb, nexthdr, offset: 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 if (!xfrm6_policy_check(NULL, dir: XFRM_POLICY_IN, skb)) {
432 SKB_DR_SET(reason, XFRM_POLICY);
433 goto discard;
434 }
435 nf_reset_ct(skb);
436 }
437
438 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
439 skb);
440 if (ret > 0) {
441 if (ipprot->flags & INET6_PROTO_FINAL) {
442 /* Not an extension header, most likely UDP
443 * encapsulation. Use return value as nexthdr
444 * protocol not nhoff (which presumably is
445 * not set by handler).
446 */
447 nexthdr = ret;
448 goto resubmit_final;
449 } else {
450 goto resubmit;
451 }
452 } else if (ret == 0) {
453 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
454 }
455 } else {
456 if (!raw) {
457 if (xfrm6_policy_check(NULL, dir: XFRM_POLICY_IN, skb)) {
458 __IP6_INC_STATS(net, idev,
459 IPSTATS_MIB_INUNKNOWNPROTOS);
460 icmpv6_send(skb, ICMPV6_PARAMPROB,
461 ICMPV6_UNK_NEXTHDR, info: nhoff);
462 SKB_DR_SET(reason, IP_NOPROTO);
463 } else {
464 SKB_DR_SET(reason, XFRM_POLICY);
465 }
466 kfree_skb_reason(skb, reason);
467 } else {
468 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
469 consume_skb(skb);
470 }
471 }
472 return;
473
474discard:
475 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
476 kfree_skb_reason(skb, reason);
477}
478
479static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
480{
481 skb_clear_delivery_time(skb);
482 rcu_read_lock();
483 ip6_protocol_deliver_rcu(net, skb, nexthdr: 0, have_final: false);
484 rcu_read_unlock();
485
486 return 0;
487}
488
489
490int ip6_input(struct sk_buff *skb)
491{
492 return NF_HOOK(pf: NFPROTO_IPV6, hook: NF_INET_LOCAL_IN,
493 net: dev_net(dev: skb->dev), NULL, skb, in: skb->dev, NULL,
494 okfn: ip6_input_finish);
495}
496EXPORT_SYMBOL_GPL(ip6_input);
497
498int ip6_mc_input(struct sk_buff *skb)
499{
500 int sdif = inet6_sdif(skb);
501 const struct ipv6hdr *hdr;
502 struct net_device *dev;
503 bool deliver;
504
505 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
506 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
507 skb->len);
508
509 /* skb->dev passed may be master dev for vrfs. */
510 if (sdif) {
511 rcu_read_lock();
512 dev = dev_get_by_index_rcu(net: dev_net(dev: skb->dev), ifindex: sdif);
513 if (!dev) {
514 rcu_read_unlock();
515 kfree_skb(skb);
516 return -ENODEV;
517 }
518 } else {
519 dev = skb->dev;
520 }
521
522 hdr = ipv6_hdr(skb);
523 deliver = ipv6_chk_mcast_addr(dev, group: &hdr->daddr, NULL);
524 if (sdif)
525 rcu_read_unlock();
526
527#ifdef CONFIG_IPV6_MROUTE
528 /*
529 * IPv6 multicast router mode is now supported ;)
530 */
531 if (atomic_read(v: &dev_net(dev: skb->dev)->ipv6.devconf_all->mc_forwarding) &&
532 !(ipv6_addr_type(addr: &hdr->daddr) &
533 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
534 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
535 /*
536 * Okay, we try to forward - split and duplicate
537 * packets.
538 */
539 struct sk_buff *skb2;
540 struct inet6_skb_parm *opt = IP6CB(skb);
541
542 /* Check for MLD */
543 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
544 /* Check if this is a mld message */
545 u8 nexthdr = hdr->nexthdr;
546 __be16 frag_off;
547 int offset;
548
549 /* Check if the value of Router Alert
550 * is for MLD (0x0000).
551 */
552 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
553 deliver = false;
554
555 if (!ipv6_ext_hdr(nexthdr)) {
556 /* BUG */
557 goto out;
558 }
559 offset = ipv6_skip_exthdr(skb, start: sizeof(*hdr),
560 nexthdrp: &nexthdr, frag_offp: &frag_off);
561 if (offset < 0)
562 goto out;
563
564 if (ipv6_is_mld(skb, nexthdr, offset))
565 deliver = true;
566
567 goto out;
568 }
569 /* unknown RA - process it normally */
570 }
571
572 if (deliver)
573 skb2 = skb_clone(skb, GFP_ATOMIC);
574 else {
575 skb2 = skb;
576 skb = NULL;
577 }
578
579 if (skb2) {
580 ip6_mr_input(skb: skb2);
581 }
582 }
583out:
584#endif
585 if (likely(deliver))
586 ip6_input(skb);
587 else {
588 /* discard */
589 kfree_skb(skb);
590 }
591
592 return 0;
593}
594

source code of linux/net/ipv6/ip6_input.c