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